Optical Performance Monitoring: Advanced Techniques for Next-Generation Photonic Networks

Optical Performance Monitoring

This page intentionally left blank

Optical Performance Monitoring Advanced Techniques for Next-Generation Photonic Networks

Calvin C. K. Chan, Ph.D. The Chinese University of Hong Kong

AMSTERDAM • BOSTON • HEIDELBERG • LONDON NEW YORK • OXFORD • PARIS • SAN DIEGO SAN FRANCISCO • SINGAPORE • SYDNEY • TOKYO Academic Press is an imprint of Elsevier

Academic Press is an imprint of Elsevier 30 Corporate Drive, Suite 400, Burlington, MA 01803, USA 525 B Street, Suite 1900, San Diego, California 92101-4495, USA The Boulevard, Langford Lane, Kidlington, Oxford, OX5 1 GB, UK # 2010 ELSEVIER Inc. All rights reserved. No part of this publication may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Details on how to seek permission, further information about the Publisher’s permissions policies and our arrangements with organizations such as the Copyright Clearance Center and the Copyright Licensing Agency, can be found at our website: www.elsevier.com/permissions. This book and the individual contributions contained in it are protected under copyright by the Publisher (other than as may be noted herein). Notices Knowledge and best practice in this field are constantly changing. As new research and experience broaden our understanding, changes in research methods, professional practices, or medical treatment may become necessary. Practitioners and researchers must always rely on their own experience and knowledge in evaluating and using any information, methods, compounds, or experiments described herein. In using such information or methods they should be mindful of their own safety and the safety of others, including parties for whom they have a professional responsibility. To the fullest extent of the law, neither the Publisher nor the authors, contributors, or editors, assume any liability for any injury and/or damage to persons or property as a matter of products liability, negligence or otherwise, or from any use or operation of any methods, products, instructions, or ideas contained in the material herein. Library of Congress Cataloging-in-Publication Data Chan, Calvin C. K. Optical performance monitoring : advanced techniques for next-generation photonic networks / Calvin C. K. Chan. p. cm. Includes bibliographical references and index. ISBN 978-0-12-374950-5 (alk. paper) 1. Optical fiber communication. 2. Network performance (Telecommunication) I. Title. TK5103.592.F52C48 2010 621.3820 75—dc22 2009046134 British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library. For information on all Academic Press publications visit our Web site at www.elsevierdirect.com Printed in The United States of America. 10 11 12 9 8 7 6 5 4 3 2

1

To my beloved wife Esther and our lovely kids, Tsz-ching and Lang-ho.

This page intentionally left blank

Contents List of Acronyms................................................................................................................................. xv List of Figures .................................................................................................................................... xxi List of Tables....................................................................................................................................... xli Preface ............................................................................................................................................... xliii Acknowledgments .............................................................................................................................. xlv List of Contributors .......................................................................................................................... xlvii About the Editor ................................................................................................................................ xlix

CHAPTER 1 OPTICAL PERFORMANCE MONITORING: PERSPECTIVES AND CHALLENGES ......................................................................................... 1 1.1 Introduction................................................................................................................. 1 1.1.1 Overarching vision .............................................................................................2 1.1.2 Challenges ..........................................................................................................3 1.2 Physical-Layer Measurements and Routing Decisions in Today’s Optical Networks ........................................................................................................ 4 1.3 Signal Parameters Requiring Monitoring and OPM Techniques ............................. 5 1.3.1 Optical impairments ...........................................................................................5 1.3.2 OPM techniques .................................................................................................8 1.4 Laudable OPM-Enabled Functionalities in Next-Generation Optical Networks ..... 9 1.4.1 Robust and stable operation...............................................................................9 1.4.2 Accommodate transparency .............................................................................10 1.4.3 Impairment-aware routing................................................................................10 1.4.4 Secure links ......................................................................................................12 1.4.5 Optical supervisory channel.............................................................................13 1.5 Smart Network Operation and Security................................................................... 13 1.5.1 Smart network operation..................................................................................14 1.5.2 Security.............................................................................................................14 1.6 Summary ................................................................................................................... 16

CHAPTER 2 OPTICAL SIGNAL-TO-NOISE RATIO MONITORING ............................... 21

2.1 Introduction............................................................................................................... 21 2.2 Linear Interpolation Techniques .............................................................................. 22 2.2.1 Optical spectrum analysis ................................................................................22 2.2.2 Out-of-band noise measurement ......................................................................23 2.2.3 Potential problems............................................................................................24 2.3 Polarization-Based Techniques ................................................................................ 26 2.3.1 Operating principles .........................................................................................26 2.3.2 Potential problems and limitations ..................................................................30 2.3.3 Methods to overcome limitations ....................................................................36

vii

viii

Contents

2.4 Interferometer-Based Technique .............................................................................. 41 2.4.1 Operating principle...........................................................................................41 2.4.2 Potential problems and limitations ..................................................................42 2.4.3 Method to overcome limitations......................................................................42 2.5 Beat Noise Analysis Techniques ............................................................................. 43 2.5.1 Operating principle...........................................................................................43 2.5.2 Potential problems and limitations ..................................................................45 2.5.3 Methods to overcome limitations ....................................................................47 2.6 OSNR Estimation Technique Based on the Operating Condition of Optical Amplifiers .............................................................................. 56 2.6.1 Operating principle...........................................................................................56 2.6.2 Link-based OSNR monitoring technique ........................................................57 2.6.3 Potential problems and limitations ..................................................................59 2.7 Summary ................................................................................................................... 61

CHAPTER 3 CHROMATIC DISPERSION MONITORING............................................. 67 3.1 Introduction............................................................................................................... 67 3.2 Chromatic Dispersion and Its Effects on Optical Fiber Systems ........................... 68 3.2.1 Fiber chromatic dispersion...............................................................................68 3.2.2 Systems limitations due to chromatic dispersion............................................70 3.2.3 Dispersion effects in the presence of fiber nonlinearities ..............................73 3.2.4 The need for chromatic dispersion monitoring...............................................75 3.3 Chromatic Dispersion Monitoring Techniques........................................................ 77 3.3.1 Measurement of RF spectrum .........................................................................77 3.3.2 Measurement of relative group delay between VSB signals .........................80 3.3.3 Histogram monitoring techniques....................................................................81 3.3.4 All-optical spectral analysis using nonlinear optics .......................................85 3.3.5 Electronic monitoring techniques ....................................................................88 3.3.6 Other chromatic dispersion monitoring techniques ........................................89 3.3.7 Differentiate chromatic dispersion from polarization mode dispersion.........90 3.4 Summary ................................................................................................................... 92 Acknowledgments ........................................................................................................... 92

CHAPTER 4 POLARIZATION MODE DISPERSION MONITORING ............................ 101

4.1 Introduction............................................................................................................. 101 4.2 PMD Monitoring Based on Measurement of RF Tone ........................................ 106 4.3 PMD Monitoring Based on Measurement of Degree of Polarization .................. 111 4.4 Electronic PMD Monitoring Techniques............................................................... 117 4.5 Other PMD Monitoring Techniques ...................................................................... 121 4.6 Summary ................................................................................................................. 121 Acknowledgments ......................................................................................................... 122

Contents

ix

CHAPTER 5 TIMING MISALIGNMENT MONITORING............................................. 127 5.1 Introduction............................................................................................................. 127 5.2 Monitoring of Timing Alignment .......................................................................... 128 5.2.1 Synchronization of pulse carver and data modulation .................................128 5.2.2 Synchronization for phase remodulation.......................................................134 5.2.3 Synchronization for I/Q data and data/pulse carver .....................................134 5.2.4 OTDM clock recovery using timing misalignment of data pulses ..............136 5.3 Investigation of the Effects of Timing Misalignment........................................... 137 5.3.1 Clock/data synchronization in CPFSK systems ............................................137 5.3.2 Misalignment between pulse carver/data modulator in RZ-DPSK systems............................................................................................................137 5.3.3 Misalignment between ASK and DQPSK modulation in ASK/DQPSK orthogonal modulation systems..............................................138 5.4 Mitigation of Timing Misalignment ...................................................................... 140 5.4.1 Hybrid OTDM scheme for demultiplexing with better timing misalignment tolerance ..................................................................................140 5.4.2 Novel remodulation scheme for colorless high-speed WDM-PON without remodulation synchronization .....................................140 5.4.3 Misalignment mitigation using MLSE equalizers ........................................141 5.5 Summary ................................................................................................................. 141

CHAPTER 6 OPTICAL PERFORMANCE MONITORING BASED ON ASYNCHRONOUS AMPLITUDE HISTOGRAMS ................................... 145 6.1 Introduction............................................................................................................. 145 6.2 Monitoring Techniques Based on Analysis of Asynchronous Histograms .......... 149 6.2.1 Q-factor monitoring .......................................................................................149 6.2.2 OSNR monitoring using asynchronous histograms ......................................157 6.3 General Concepts on the Acquisition and Processing of Amplitude Histograms .............................................................................................................. 163 6.3.1 Sampling noise ...............................................................................................163 6.3.2 Averaging effects ...........................................................................................168 6.4 Summary ................................................................................................................. 172

CHAPTER 7 OPTICAL PERFORMANCE MONITORING BASED ON ASYNCHRONOUS DELAY-TAP SAMPLING......................................... 175 7.1 Introduction............................................................................................................. 175 7.2 Technique................................................................................................................ 176 7.2.1 Phase portrait..................................................................................................177 7.2.2 Pattern recognition .........................................................................................180 7.3 Experiment .............................................................................................................. 182 7.3.1 Network emulator...........................................................................................182

x

Contents

7.3.2 Multi-impairment monitor .............................................................................183 7.3.3 First-order PMD .............................................................................................183 7.3.4 Results for 10-G NRZ and 40-G NRZ-DPSK ..............................................183 7.3.5 System testing ................................................................................................185 7.4 Discussion ............................................................................................................... 187 7.4.1 Extension to new impairments ......................................................................188 7.4.2 Application to higher-order formats..............................................................188 7.5 Summary ................................................................................................................. 191

CHAPTER 8 OPTICAL PERFORMANCE MONITORING BASED ON OPTICAL SAMPLING .......................................................................... 193

8.1 Introduction............................................................................................................. 193 8.1.1 Data encoding in the electric field of optical waves....................................193 8.1.2 Temporal characterization of optical signals ................................................194 8.1.3 Linear optical sampling .................................................................................196 8.2 LOS Principle and Properties................................................................................. 196 8.2.1 Coherent detection .........................................................................................196 8.2.2 Various implementations of coherent detection for optical performance monitoring ..........................................................................................199 8.2.3 Polarization and wavelength sensitivity ........................................................201 8.2.4 Phase sensitivity .............................................................................................202 8.2.5 Digital phase tracking ....................................................................................204 8.3 Implementations of LOS ........................................................................................ 205 8.3.1 Balanced photodetection ................................................................................206 8.3.2 Direct photodetection .....................................................................................208 8.3.3 LOS with four-wave mixing..........................................................................208 8.3.4 Correction of imperfections ...........................................................................209 8.4 Optical Performance Monitoring with LOS .......................................................... 211 8.4.1 Characterization of amplified spontaneous emission....................................211 8.4.2 Phase and amplitude noise measurements ....................................................211 8.4.3 Nonlinear phase noise ....................................................................................213 8.4.4 Nonlinear phase shift measurement...............................................................216 8.4.5 Digital processing of sampled electric field .................................................216 8.4.6 Characterization of the electric field of periodic sources ............................217 8.5 Recent Results and Related Techniques................................................................ 218 8.6 Summary ................................................................................................................. 219 Acknowledgments ......................................................................................................... 219

CHAPTER 9 OPTICAL PERFORMANCE MONITORING BASED ON RF PILOT TONES.......................................................................... 223

9.1 Introduction............................................................................................................. 223

Contents

xi

9.2 Performance Monitoring Techniques Using AM Pilot Tones .............................. 224 9.2.1 Operating principle.........................................................................................224 9.2.2 Potential problems..........................................................................................225 9.2.3 Scalability .......................................................................................................232 9.2.4 Typical applications .......................................................................................234 9.3 Performance Monitoring Techniques Using PM and FM Pilot Tones................. 238 9.3.1 Using PM pilot tones .....................................................................................238 9.3.2 Using FM pilot tones .....................................................................................243 9.4 Dispersion Monitoring Techniques for Adaptive Compensators.......................... 245 9.4.1 CD monitoring techniques using AM and PM pilot tones...........................246 9.4.2 CD monitoring technique using chirped pilot tone ......................................249 9.4.3 CD monitoring technique using pilot tone carried by broadband light source...................................................................................252 9.4.4 PMD monitoring technique using SSB pilot tone ........................................256 9.5 Summary ................................................................................................................. 258

CHAPTER 10 OPTICAL PERFORMANCE MONITORING BASED ON ELECTRONIC DIGITAL SIGNAL PROCESSING .................................. 261 10.1 Introduction........................................................................................................... 261 10.2 OPM in Digital Direct Detection Systems .......................................................... 263 10.2.1 The channel model for direct detection systems ......................................264 10.2.2 State-based equalization based on MLSE .................................................265 10.2.3 State-based OSNR estimation....................................................................268 10.2.4 Referenced parameter estimation ..............................................................271 10.2.5 Conclusion ..................................................................................................274 10.3 OPM in Digital Coherent Receivers.................................................................... 274 10.3.1 Theory.........................................................................................................275 10.3.2 Joint estimation of linear channel parameters ..........................................280 10.3.3 Conclusion ..................................................................................................295 10.4 Summary............................................................................................................... 296

CHAPTER 11 OPTICAL PERFORMANCE MONITORING BASED ON NONLINEAR OPTICAL TECHNIQUES .................................................................... 301 11.1 11.2 11.3 11.4

Introduction........................................................................................................... 301 Nonlinear Optics................................................................................................... 303 OPM Techniques Using Nonlinear Optics .......................................................... 304 Key Challenges..................................................................................................... 309 11.4.1 Sensitivity ...................................................................................................309 11.4.2 Cost, size, and complexity.........................................................................312 11.4.3 Impairment isolation ..................................................................................312 11.5 Summary............................................................................................................... 314

xii

Contents

CHAPTER 12 OPTICAL PERFORMANCE MONITORING OF OPTICAL PHASE–MODULATED SIGNALS ....................................................... 319

12.1 Introduction........................................................................................................... 319 12.2 Performance of Phase-Modulated Signals........................................................... 320 12.2.1 Signal impairments.....................................................................................321 12.2.2 Generation and detection of N-PSK signals .............................................325 12.3 Optical Performance Monitoring ......................................................................... 327 12.3.1 Monitoring techniques ...............................................................................327 12.3.2 Comparison of monitoring techniques ......................................................343 12.4 Summary............................................................................................................... 345 Acknowledgments ......................................................................................................... 346

CHAPTER 13 OPTICAL PERFORMANCE MONITORING FOR COHERENT OPTICAL SYSTEMS ........................................................................................ 351

13.1 Historical Aspect of Coherent Optical Systems.................................................. 351 13.2 Single-Carrier and Multicarrier Coherent Optical Systems................................ 352 13.2.1 Principle of coherent detection..................................................................353 13.2.2 Single-carrier coherent optical system ......................................................354 13.2.3 Coherent optical OFDM system ................................................................356 13.2.4 Comparison of single-carrier and multicarrier coherent optical system.............................................................................................360 13.3 OPM Using Coherent Detection .......................................................................... 361 13.3.1 OPM without receiver electrical equalization...........................................361 13.3.2 OPM with receiver electrical equalization................................................366 13.4 OPM in CO-OFDM Systems ............................................................................... 368 13.4.1 Optical channel model ...............................................................................369 13.4.2 Principle of OPM through optical channel estimation .............................371 13.5 Progress in OPM for CO-OFDM Systems .......................................................... 373 13.5.1 Simulation model and results ....................................................................373 13.5.2 Optical performance monitoring in CO-OFDM systems with 4-QAM ...375 13.5.3 OPM in CO-OFDM systems with 16-QAM modulation .........................378 13.6 OPM Experiment Results..................................................................................... 379 13.7 Summary............................................................................................................... 381

CHAPTER 14 OPTICAL PERFORMANCE MONITORING IN OPTICAL TRANSPORT NETWORKS ................................................................ 385

14.1 Introduction........................................................................................................... 385 14.2 Overview............................................................................................................... 385 14.2.1 Business interface model ...........................................................................385 14.2.2 Generic OTN service requirements ...........................................................386 14.2.3 OTN: A network of networks....................................................................387

Contents

xiii

14.3 Generic Modeling Principles for Transport Networks........................................ 388 14.3.1 Top-level functional architecture...............................................................388 14.3.2 Control plane functions..............................................................................388 14.3.3 Management functions ...............................................................................388 14.3.4 Transport functions ....................................................................................390 14.4 Modeling of Multilayer Networks ....................................................................... 390 14.4.1 Application of partitioning concept...........................................................392 14.4.2 Application of the layering concept ..........................................................392 14.4.3 Transport entities: trails and connections..................................................392 14.4.4 Characteristic information..........................................................................393 14.5 Optical Transport Network–Layered Structure ................................................... 393 14.5.1 OTN-layer networks...................................................................................393 14.5.2 Layer management .....................................................................................394 14.5.3 OTN information structure ........................................................................394 14.6 OTN Services ....................................................................................................... 395 14.6.1 All-Optical Networks .................................................................................395 14.7 Test and Measurement Tasks in Optical Networking......................................... 396 14.7.1 Lightpath provisioning ...............................................................................396 14.7.2 Service assurance .......................................................................................398 14.8 Optical Performance Monitoring ......................................................................... 400 14.8.1 Optical-layer signal quality supervision requirements..............................401 14.8.2 Optical power .............................................................................................403 14.8.3 Channel wavelength ...................................................................................403 14.8.4 OSNR..........................................................................................................403 14.8.5 Q-factor measurement................................................................................407 14.8.6 OTUk, ODUkT, and ODUkP signal quality supervision .........................409 14.8.7 What is missing? ........................................................................................410 14.9 Implementation Issues .......................................................................................... 411 14.9.1 Accuracy requirements...............................................................................411 14.9.2 External versus embedded monitoring ......................................................412 14.9.3 Monitoring points.......................................................................................413 14.9.4 Recommended measurement interval........................................................414 14.9.5 Risk management aspects ..........................................................................415 14.9.6 Improved fault diagnostics by event correlation ......................................417 14.10 Future Challenges............................................................................................... 419 14.11 Summary............................................................................................................. 419

CHAPTER 15 OPTICAL PERFORMANCE MONITORING IN OPTICAL LONG-HAUL TRANSMISSION SYSTEMS ............................................................. 423 15.1 Introduction........................................................................................................... 423

xiv

Contents

15.2 Elements of a Long-Haul Transmission System................................................. 424 15.3 System Performance Measures ............................................................................ 430 15.4 OPM in a Long-Haul Transmission System ....................................................... 432 15.4.1 OPM functions and applications ...............................................................432 15.4.2 Optical device monitoring .........................................................................433 15.4.3 OSNR monitoring along transmission line ...............................................435 15.4.4 Transmission performance testing and analysis........................................438 15.4.5 Service-oriented system design .................................................................442 15.5 Summary............................................................................................................... 445 Index................................................................................................................................... 447

List of Acronyms 2R 3R AAH ACRZ ADC AF AIS AM AMI ANN AON APS ASE ASIC ASK ASTN AWG AWGN BER BIP BLS BPF BPSK CCI CD CI CM CMA CMOS CNR CO-OFDM CP CPE CPFSK CSRZ CW DB DBPSK DBR DC

Regeneration, reshaping Regeneration, reshaping, retiming Asynchronous amplitude histogram Alternate chirped return to zero Analog-to-digital converter Amplitude filter Alarm indication signal Amplitude modulation Alternate mark inversion Artificial neural network All-optical network Automatic protection switching Amplified spontaneous emission Application-specific integrated circuit Amplitude shift keying Automated switched-transport network Array waveguide grating Additive white Gaussian noise Bit error rate Bit-interleaved parity Broadband light source Bandpass filter Binary phase-shift keying Connection controller interface Chromatic dispersion Characteristic information Connection monitoring Constant modulus algorithm Complementary metal-oxide semiconductor Carrier-to-noise ratio Coherent optical OFDM Cyclic prefix Common phase error Continuous-phase frequency-shift keying Carrier-suppressed return to zero Continuous wave Duobinary Differential binary phase-shift keying Distributed Bragg reflector Direct current

xv

xvi

List of Acronyms

DCF DCM DD-LMS DFB DFT DGD DGEF DI DML DOP DPASK DPSK DQPSK DSB DSP DTS DWDM EAM EBPF ECC EDC EDFA EO ESNR FBG FDE FDI FEC FFT FIR FM FOM FROG FSR FWM GVD HNLF IC ICI IEEE IM/DD IP IrDI ISI

Dispersion-compensating fiber Dispersion-compensating module Decision-directed least mean square Distributed feedback Discrete Fourier transform Differential group delay Dynamic gain equalizing filter Delay interferometer Directly modulated laser Degree of polarization Differential phase-amplitude-shift keying Differential phase-shift keying Differential quadrature phase-shift keying Double sideband Digital signal processing Delay-tap sampling Dense wavelength-division multiplexing Electro-absorption modulator Electrical bandpass filter Embedded control channel Electronic dispersion compensation Erbium-doped fiber amplifier Electrical to optical Electrical signal-to-noise ratio Fiber Bragg grating Frequency-domain equalization Forward defect indication Forward error correction Fast Fourier transform Finite impulse response Frequency modulation Figure of merit Frequency-resolved optical gating Free spectral range Four-wave mixing Group velocity dispersion Highly nonlinear fiber Integrated circuit Intercarrier interference Institute of Electrical and Electronic Engineers Intensity modulation/direct detection Internet Protocol Interdomain interface Intersymbol interference

List of Acronyms

ITU LD LEAF LO LOS LP LPF LS LSB MCM MEMS MI MIMO ML MLSE MMSE MPDR MPI MPLS MSPE MTBF MTTR MZ MZI MZM NE NLSE NMS NPN NRZ NZ-DSF OA OADM OAM OBPF OCA OCC OCE OCG OCh OChr OCM ODUk ODUkP

International Telecommunication Union Laser diode Large effective area fiber Local oscillator Linear optical sampling Low pass Low-pass filter Least square Lower sideband Multicarrier modulation Micro-electro-mechanical systems Modulator index Multiple-input and multiple-output Maximum likelihood Maximum likelihood sequence equalizer Minimum mean-square estimation Monitoring power dynamic range Multiple path interference Multiprotocol label switching Multisymbols phase estimation Mean time before failure Mean time to repair Mach-Zehnder Mach-Zehnder interferometer Mach-Zehnder modulator Network element Nonlinear Schro¨dinger equation Network management system Nonlinear phase noise Non-return to zero Nonzero dispersion-shifted fiber Optical amplifier Optical add/drop multiplexer Operations, administration, and maintenance Optical bandpass filter Optical channel analyzer Optical channel carrier Optical channel estimation Optical channel group Optical channel Optical channel with reduced functionality Optical channel monitoring Optical data unit of level k Optical data unit of level k, path

xvii

xviii

ODUkT OE OEO OFDM OH OMS ONE ONU OOK OOS OPM OPPL OPS OPSn OPUk OS OSA OSC OSNR OTDR OTM OTN OTR OTS OTU OTUk OTUkV OXC PA PBS PD PDF PDG PDL PDM PHB PM PM PMD PMDC PMF PolMux PON PRBS

List of Acronyms

Optical data unit of level k, tandem connection sublayer Optical-to-electrical Optical-electrical-optical Orthogonal frequency-division multiplexing Overhead Optical multiplex section Optical network element Optical network unit On-off keying Optical transport multiplex overhead signal Optical performance monitoring Optical phase-locked loop Optical physical section Optical physical section of level n Optical channel payload unit of level k Optical section Optical spectrum analyzer Optical supervisory channel Optical signal-to-noise ratio Optical time-domain reflectometer Optical transport module Optical transport network Optical-to-RF Optical transmission section Optical channel transport unit Optical channel transport unit of level k Optical channel transport unit of level k, functionally standardized Optical crossconnect Preamplifier Polarization beam splitter Photodetector Probability density function Polarization-dependent gain Polarization-dependent loss Polarization-division multiplexing Polarization hole burning Polarization maintaining Phase modulator Polarization mode dispersion Polarization mode dispersion compensator Polarization-maintaining fiber Polarization multiplexing Passive optical network Pseudo-random bit sequence

List of Acronyms

PSD PSP PXC QAM QoS QPM QPSK RAH RF RFSA ROADM RSOA RTO RWA Rx RZ S&H SBS SCM SDH SE SES SHG SISO SITO SLA SMF SNR SOA SONET SOP SPIDER SPM SRB SRS SSB SSMF TC TCM TDE TDM TIA TITO TMN

Power spectral density Principal states of polarization Photonic crossconnect Quadrature amplitude modulation Quality of service Quasi phase matched Quadrature phase-shift keying Reference asynchronous histogram Radio frequency RF spectrum analyzer Reconfigurable optical add-drop multiplexer Reflective semiconductor optical amplifier RF-to-optical Routing wavelength assignment Receive Return zero Sample and hold Stimulated Brillouin scattering Subcarrier multiplexing Synchronous digital hierarchy Spectral efficiency Severely errored second Second harmonic generation Single-input and single-output Single-input and two-output Service-level agreement Single-mode fiber Signal-to-noise ratio Semiconductor optical amplifier Synchronous optical network State of polarization Spectral phase interferometry for direct electric-field reconstruction Self-phase modulation Stimulated Rayleigh back-scattering Stimulated Raman scattering Single sideband Standard single-mode fiber Tandem connection Tandem connection monitoring Time-domain equalization Time-division multiplexing Transimpedance amplifier Two-input and two-output Telecommunication(s) management network

xix

xx

TPA TTI Tx UNI VOA VPN VSB WB WDM XGM XPM xPSK xQAM ZF

List of Acronyms

Two-photon absorption Trail trace identifier Transmit User (to) network interface Variable optical attenuator Virtual private network Vestigial sideband Wavelength blocker Wavelength-division multiplexing Crossgain modulation Crossphase modulation Multilevel phase-shift keying Multilevel quadrature amplitude modulation Zero forcing

List of Figures FIGURE FIGURE FIGURE FIGURE FIGURE

1.1 1.2 1.3 1.4 1.5

FIGURE 1.6 FIGURE 1.7 FIGURE 1.8 FIGURE 2.1 FIGURE 2.2 FIGURE 2.3 FIGURE 2.4 FIGURE 2.5

FIGURE 2.6 FIGURE 2.7 FIGURE 2.8 FIGURE 2.9 FIGURE 2.10 FIGURE 2.11 FIGURE 2.12 FIGURE 2.13

Window of operability is shrinking as networks become more complex. A self-managed network with optical performance monitoring. Features of ubiquitous monitoring for robust and self-managed networks. Overview of various optical impairments within the network. Future heterogeneous networks should accommodate various types of traffic and use optimal channel characteristics for each application/user. The required hardware should be reconfigurable and transparent. Multivariable routing. (a) Denial of service, and (b) eavesdropping of an intrusive wavelength using the nonlinear effects in the fiber link. Identification and localization of network impairments allow network resources to be adapted for compensation, data re-routing, and resource reallocation. Graphical description of OSNR measurement based on optical spectrum. (a) Conceptual diagram of optical spectrum analyzer. (b) Relationship between level of optical noise and resolution bandwidth (BW) of OSA. Out-of-band noise measurement using AWG. Out-of-band noise measurement using OSAs, when there are DGEs within the optical link. The OSAs measure the power of out-of-band ASE noise for the OSNR monitoring, as well as the channel powers for the gain equalization. (a) Example of a dynamically reconfigurable transparent optical network configured with ROADMs. Optical spectrum measured at point A, (b) when unmodulated CW signals are transmitted, or (c) when 10-Gb/s NRZ signals are transmitted. (Resolution bandwidth: 0.05 nm.) (a) Optical spectrum of 43-Gb/s RZ-DQPSK signals. (b) Optical spectrum measured after turning off modulators. (Resolution: 0.1 nm; div, division.) Principle of OSNR monitoring based on polarization-nulling technique. Copyright # 2006 IEEE. Configuration of polarization-nulling technique based on (a) adaptive polarization control with feedback, or (b) rotating quarter-wave plate and polarizer. (a) Correlation between DOP and OSNR. (b) OSNR monitoring sensitivity to DOP error. Measured DOP and normalized power of WDM signals. Copyright # 2006 IEEE. Illustration of the error mechanism caused by PMD in the execution of the polarization-based OSNR monitoring technique in (a) the time domain and (b) the frequency domain. OSNR monitoring error caused by PMD (mean PMD = 3.22 ps). Copyright # 2001 IEEE. Illustration of error mechanism caused by nonlinear birefringence in execution of polarization-based OSNR monitoring technique.

xxi

xxii

List of Figures

FIGURE 2.14

FIGURE 2.15 FIGURE 2.16

FIGURE 2.17 FIGURE 2.18 FIGURE 2.19 FIGURE 2.20 FIGURE 2.21 FIGURE 2.22 FIGURE 2.23 FIGURE 2.24 FIGURE 2.25 FIGURE 2.26 FIGURE 2.27 FIGURE 2.28 FIGURE 2.29

Maximum OSNR errors caused by nonlinear birefringence (a) measured in a twochannel experiment (200-GHz spacing, and 0- and 7-dBm input power for the probe and pump, respectively) (Copyright # 2006 IEEE), or (b) measured in various 640-km-long fiber links with six channels (200-GHz spacing, 0-dBm/ channel input power, 80 km
8 spans) (Copyright # 2001 IEEE). Illustration of the error mechanism caused by PDL in execution of polarizationbased OSNR monitoring technique. (a) Cumulative probability of errors in the measured OSNRs by using the polarization-nulling technique due to partially polarized ASE noise in a transmission link consisting of 15 spans (average PDL/span = 0.57 dB). (b) Probability that the error in the measured OSNRs by using the polarization-nulling technique becomes >1 dB (due to the partially polarized ASE noise caused by PDL). Copyright # 2006 OSA. (a) Fourier components of the Stokes parameters of an optical signal measured in a 120-km-long aerial fiber link. (b) OSNR measured by using the polarizationnulling technique in a 120-km-long aerial fiber link. Copyright # 2004 OSA. Schematic diagram of the polarization-nulling technique improved by using additional optical filter. PBS, polarization beam splitter; BPF, bandpass filter; PD, photo detector. Copyright # 2006 IEEE. (a) Illustration of off-center filtering technique for OSNR monitoring. (b) Effects of filter detuning for 39.81-Gb/s, 2.5-ps, full-width at half-maximum (FWHM) RZ signal. Copyright # 2004 IEEE. Schematic diagram of the polarization-nulling technique, improved by multiplefrequency measurement and PMD compensation. Copyright # 2006 IEEE. Principle of spectral SOP measurement technique. (a) Schematic diagram of OSNR monitoring technique based on MZDI interferometer. (b) Total power measurement with constructive interference. (c) Noise power measurement with destructive interference. Schematic diagram of modified optical interferometer for OSNR monitoring. (a) Power spectral density of receiver noises caused by signal-ASE and ASE-ASE beating. (b) Measured receiver noise spectrum from 40 to 50 kHz when OSNR is 20 dB or 30 dB. Principle of OSNR monitoring technique based on low-frequency beat noise analysis. ADC, analog-to-digital converter; B, bit rate; FFT, fast Fourier transform; PD, photodetector. RF spectrum of 10-Gb/s NRZ signal with PRBS pattern (pattern length: 223 – 1). (a) Principle of OSNR monitoring technique based on high-frequency beat noise analysis. (b) RF spectrum of 2.5-Gb/s NRZ signal with 20-dB or 30-dB OSNR. OSNR of 2.5-Gb/s NRZ signal measured by monitoring beat noise at (a) 2.5-GHz null point or (b) 10 GHz. Schematic diagram of OSNR monitoring technique based on polarization diversity.

List of Figures

FIGURE 2.30 FIGURE 2.31 FIGURE 2.32 FIGURE 2.33 FIGURE 2.34 FIGURE 2.35 FIGURE 2.36 FIGURE 2.37 FIGURE 2.38 FIGURE 2.39 FIGURE 2.40 FIGURE 2.41 FIGURE 2.42

FIGURE 2.43

FIGURE 2.44 FIGURE 3.1 FIGURE 3.2

xxiii

Waveforms of 1.25-Gb/s signal. The upper graph shows the outputs of two arms of the polarization-diversity receiver; the lower graph shows the sum and difference of outputs. Copyright # 2005 IEEE. Conversion from polarization variation to intensity variation by polarizationdiversity receiver with subtraction circuit. Schematic diagram of orthogonal polarization delayed-homodyne technique for monitoring OSNR. (a) Principle of nullifying data spectrum using the orthogonal polarization delayedhomodyne technique. (b) Measured RF spectra of 10-Gb/s NRZ signal with and without data spectrum nullifying. Schematic diagram of OSNR monitoring technique based on frequency diversity. RF spectra of the signal after (a) one photodiode and (b) balanced subtraction. Copyright # 2005 IEEE. Effect of chromatic dispersion on OSNR monitoring technique based on a frequency-diversity receiver. Copyright # 2008 IEEE. Schematic diagram of orthogonal polarization self-heterodyne technique for monitoring OSNR. Operating principle of orthogonal polarization self-heterodyne technique for OSNR monitoring. (a) Relation of polarization states of the signals in branches 1 and 2. (b) Effect of PMD on electrical spectrum. Copyright # 2007 IEEE. (a) Schematic diagram of OSNR monitoring technique based on synchronously gated signal. RF spectra of signal (b) without gating pulses and (c) with gating pulses. Copyright # 2006 OSA. Schematic diagram of OSNR monitoring technique based on operating condition of optical amplifiers. OCM, optical channel monitor; PM, power monitor. Copyright # 2008 IEEE. (a) Schematic diagram of OSNRlink monitoring. (b) Schematic diagram of OPM manager GUI, graphic user interface; LSP, label-switched path; CLI, command line interface; GMPLS, generalized multiprotocol label switching; SNMP, Simple Network Management Protocol. Copyright # 2009 IEEE. Monitoring error of link OSNR caused by total power monitoring (a) without calibration and (b) with 50% of optical noise power calibrated (contour plots). Span loss = gain of EDFA = 20 dB; noise figure of EDFA = 8 dB; optical noise bandwidth = 30 nm; input power/channel to fiber = 0 dBm. Copyright # 2009 IEEE. Maximum OSNR monitoring error in the worst-case scenario for optical link with AGC-EDFAs (a) under normal operating conditions, or (b) with an increase of 10 dB in span loss at worst position (contour plots). Copyright # 2009 IEEE. Dispersion coefficient, D, as a function of wavelength in conventional silica single-mode fiber. Copyright # 2003 IEEE. Dispersion coefficient in a dispersion-shifted fiber.

xxiv

List of Figures

FIGURE 3.3 FIGURE 3.4

FIGURE 3.5 FIGURE 3.6

FIGURE 3.7

FIGURE 3.8 FIGURE 3.9

FIGURE 3.10 FIGURE 3.11 FIGURE 3.12 FIGURE 3.13 FIGURE 3.14 FIGURE 3.15

CD values for several commercially available types of transmission fiber. Origin of CD in data transmission. (a) CD is caused by the frequency-dependent refractive index in fiber. (b) Non zero spectral width due to data modulation. (c) Dispersion leads to pulse broadening, proportional to the transmission distance and data rate. f, frequency; v, velocity. Transmission distance limitations due to uncompensated dispersion in SMF as a function of data rate for intensity-modulated optical signals. Copyright # 2001 IEEE. (a) The glass that a photon in the l3 pulse “sees” changes as other channels (with potentially varying power) move to coincide with the l3 pulse. (b) System performance (SNR) versus fiber dispersion. Higher dispersion is preferred to reduce XPM effects. Copyright # 1994 IEEE. (a, b) FWM induces new spectral components via nonlinear mixing of two wavelength signals. (c) The signal degradation due to FWM products falling on a third data channel can be reduced by even small amounts of dispersion. Copyright # 1995 IEEE. Dispersion map of basic dispersion-managed system. Positive dispersion transmission fiber alternates with negative dispersion compensation elements such that total dispersion is nearly zero end to end. (a) Zero-dispersion wavelength shifts due to temperature change; thus, dispersion itself changes at a fixed wavelength (b) For a 40-Gb/s, 1000-km fiber link, 30 C temperature change causes dispersion beyond system limit. Copyright # 2000 IEEE. Principle of RF fading used for dispersion monitoring: RF tone within data band fades due to CD. Copyright # 2002 IEEE. (a) RF fading due to dispersion for 7- and 9-GHz tones. Solid lines represent theoretical results and points are experimental results. (b) Theoretical curve for measurable range of CD. Copyright # 2002 IEEE. (a) Clock regenerating effect for NRZ data. (b) Clock fading effect for RZ data. Solid lines represent without SPM; dashed lines represent with SPM; dotted lines represent experimental. Copyright # 2001 IEEE. (a) Experimental setup. RF power at half of data rate (5 GHz) is measured after MZI with path difference of 100 ps. (b) Received RF power as function of dispersion. Copyright # 2005 IEEE. Conceptual diagram for monitoring CD using optical VSB filtering. Recovered bits from either part of spectrum arrive at slightly different times depending on CD. Copyright # 2002 IEEE. Phase shift between the two VSB signals versus the normalized. The three lines are simulation results for NRZ data, Gaussian filter (dashed dotted line); RZ data, Gaussian filter (solid line); and RZ data, fiber Fabry-Perot filter (dashed line). Scatter points are experimental for 10-Gb/s RZ data using a fiber Fabry-Perot filter. Copyright # 2002 IEEE.

List of Figures

FIGURE 3.16 FIGURE 3.17 FIGURE 3.18

FIGURE 3.19 FIGURE 3.20 FIGURE 3.21

FIGURE 3.22 FIGURE 3.23 FIGURE 3.24 FIGURE 3.25 FIGURE 3.26 FIGURE 3.27 FIGURE 4.1 FIGURE 4.2 FIGURE 4.3 FIGURE 4.4 FIGURE 4.5 FIGURE 4.6

xxv

(a) Eye diagram and (b) histogram with synchronous sampling. (c) Eye diagram and (d) histogram with asynchronous sampling. Copyright # 2004 IEEE. Portraits processing of delay-tap sample pairs to create phase portraits. Labels on phase portrait represent the sampled bit sequences. Copyright # 2007 IEEE. Eye diagrams and phase portraits for NRZ: (a) OSNR = 35 dB and no impairment; (b) OSNR = 25 dB; (c) OSNR = 35 dB and CD = 800 ps/nm; (d) OSNR = 35 dB and PMD = 40 ps; (e) OSNR = 35 dB and crosstalk = 25 dB; and (f) OSNR = 25 dB, CD = 800 ps/nm, PMD = 40 ps, and crosstalk = 25dB. Copyright # 2007 IEEE. Monitor versus actual values of various impairments and signal quality measures for simultaneous mixtures of OSNR, CD, PMD, and filter offset from 10-Gb/s NRZ simulations. Copyright # 2007 IEEE. Principle of residual dispersion monitoring via SPM and filtering. Copyright # 2002 IEEE. Comparison of long-pass, spectral monitoring signal (solid squares) with typical 40-Gb/s RZ receiver BER penalty (open circles), both plotted versus residual dispersion. Eye diagrams are indicated for three residual dispersion values. Lines are a guide to the eye. Copyright # 2002 IEEE. Experimental setup to vary noise and accumulated dispersion on a data signal. The wavelength-converted monitoring signal is generated by mixing Pdata with CW signal PCW in SOA and selected by the optical filter. Copyright # 2005 IEEE. Monitor signal for 40-Gb/s data as function of accumulated dispersion. Copyright # 2005 IEEE. (a) Typical measured data for logarithm of BER versus decision threshold (Copyright # 1993 IEEE). (b) BER as a function of the received optical SNR (Copyright # 1988 Holt, Rinehart, and Winston). RF tone fading due to CD and PMD. System setup of the CD monitoring scheme suppressing PMD and chirp effects. Copyright # 2006 IEEE. CD monitoring error, (a) versus DGD without and with PMD cancellation, and (b) versus a parameter without and with chirp suppression. Copyright # 2006 IEEE. Origin of PMD. Illustration of input optical pulse with power transmitted on the two PSPs, each arriving at a different time. (a) Probability distribution of DGD in typical fiber. (b) System performance (BER) fluctuations due to changes in temperature caused by PMD. Copyright # 1991 IEEE. Graphical representation of all-order PMD effect on an optical pulse. Transmission distance limitations for a 40-Gb/s NRZ system due to combination of fiber PMD and PMD of cascaded in-line optical components found in amplifier sites. Copyright # 2004 IEEE/OSA. Explanation of PMD-induced RF power fading in an SSB SCM system in optical domain. Copyright # 2004 IEEE.

xxvi

List of Figures

FIGURE 4.7 FIGURE 4.8 FIGURE 4.9 FIGURE 4.10 FIGURE 4.11 FIGURE 4.12 FIGURE 4.13 FIGURE 4.14

FIGURE 4.15

FIGURE 4.16 FIGURE 4.17 FIGURE 4.18

FIGURE 4.19

FIGURE 4.20

Received RF power variation versus DGD for eighth, quarter, half, and bit rate frequency components. Copyright # 2004 IEEE/OSA. Concept of CD-insensitive RF power fading using optical bandpass filtering. Copyright # 2004 IEEE. CD-induced RF clock power fading under various DGD values (a) without bandpass filtering and (b) with bandpass filtering. Insets are RF clocks when DGD is 23 ps and CD is 0 and 640 ps/nm, respectively. Copyright # 2004 IEEE. Experimental setup of simultaneous PMD and OSNR monitoring through enhanced RF spectrum analysis by adding large DGD element. FMLL, fiber mode–locked laser. Copyright # 2005 IEEE. PMD monitoring results for 10-Gb/s, 2.5% RZ data by adding large DGD element. OSNR varies from 15 to 35 dB. Copyright # 2005 IEEE. (a) Conceptual diagram of PMD monitoring technique for DPSK/DQPSK. (b) RF power increases with decreasing FSR of polarization-based interferometer filter (i.e., with increasing DGD values). Copyright # 2008 IEEE. Experimental results of (a) RF power measured at 170 MHz for PMD monitoring of NRZ-DQPSK and NRZ-DPK, and (b) CD dependence for PMD monitoring with DGD 23 ps and 40 ps, respectively. Copyright # 2008 IEEE. Schematic illustration of DOP signal degradation by PMD. (a) Optical waveform and SOP of signal without PMD. (b) Optical waveform and SOP of signal with PMD. The x and y axes correspond to two PSPs of the transmission media. Copyright # 2001 IEEE/OSA. DOP as function of DGD for 10-Gb/s NRZ data modulated by MZ modulator. Plots, experiment; dashed line, rectangular waveform approximation; thin lines, numerical simulation. All simulated DOP curves are relatively unaffected by chirp parameter a and by fiber dispersion of 350 ps/nm. Copyright # 2001 IEEE/OSA. Theoretical results of minimum DOP versus DGD (relative to bit time, Tb) as pulse width of RZ signal varies. Copyright # 2004 IEEE/OSA. Sensitivity of DOP reduction as a function of DGD (first-order PMD). Copyright # 2001 IEEE/OSA. Measured DOP reduction with scrambled input polarization of 40-Gb/s RZ signal. (a) First-order PMD of 1.25 ps. (b) Second-order PMD, concatenation of two unaligned birefringent secitons (6-ps and 4-ps DGD). Copyright # 2001 IEEE/ OSA. Prior to optical filtering, an RZ signal that undergoes DGD equal to the pulse width is completely deplorized, thus limiting the DGD monitoring range of DOPbased DGD monitors. After filtering, the signal is partially polarized, allowing DOP-based monitoring of the DGD. Copyright # 2001 IEEE/OSA. Frequency-domain illustration of reducing depolarization via symmetric narrowband optical filtering. Short optical pulses have a wide optical spectrum, enhancing the effects of DGD-induced depolarization. A narrowband filter shrinks the optical spectrum, thus reducing these depolarization effects and increasing the DGD monitoring range. Copyright # 2001 IEEE/OSA.

List of Figures

FIGURE 4.21

FIGURE 4.22 FIGURE 4.23 FIGURE 4.24 FIGURE 4.25 FIGURE 4.26 FIGURE 4.27 FIGURE 5.1 FIGURE 5.2 FIGURE 5.3 FIGURE 5.4 FIGURE 5.5

FIGURE 5.6 FIGURE 5.7 FIGURE 5.8 FIGURE 5.9 FIGURE 5.10 FIGURE 5.11 FIGURE 5.12

xxvii

(a) Experimental results for minimum DOP versus DGD for 40-Gb/s 50% RZ signals. (b) Simulation results for minimum DOP versus DGD for 40-Gb/s NRZ signals before and after asymmetric partial optical filtering. Copyright # 2001 IEEE/OSA. (a) Design of error monitor with analog integrator, and (b) typical characteristics of integrator voltage Uint versus monitor threshold U1 for first-order PMD signals with variable DGD and Y = 0.5. Copyright # 2001 IEEE/OSA. (a) BER versus eye opening for all orders PMD statistics. (b) BER after compensation versus eye opening using eye monitoring. Copyright # 2001 IEEE/OSA. (a) Eye diagram and (b) histogram with synchronous sampling. (c) Eye diagram and (d) histogram with asynchronous sampling. Copyright # 2004 IEEE/OSA. Portrait processing of delay-tap sample pairs to create phase. Labels on phase portrait represent sampled bit sequences. Copyright # 2007 IEEE/OSA. Eye diagrams and phase portraits for NRZ with OSNR = 35 dB at (a) no impairment and (b) PMD = 40 ps. Copyright # 2007 IEEE/OSA. Measurement of effective DGD (root mean square error of 3.1 ps) in presence of OSNR levels ranging from 13.5 to 25 dB. Copyright # 2009 IEEE/OSA. The effect of timing misalignment between pulse carver and data modulator. Copyright # 2003 IEEE. Measured spectrum asymmetry due to timing misalignment. Copyright # 2002 OSA. Measured spectra for (a) aligned and (b) misaligned cases. The first spectral null occurs at around 6.4 GHz. Copyright # 2003 IEEE. (a) Simple alignment-detection scheme. (b) Measurements: microwave monitoring of misalignment. Copyright # 2003 IEEE. (a) Illustration of timing alignment between pulse carver and data modulator. (b) Calculated signal spectra with timing alignment between pulse carver and data modulator of (i) 0, (ii) 0.3, and (iii) 0.5 T in a 10-Gb/s RZ-DPSK system with 0.28-T pulsewidth. Copyright # 2005 IEEE. (a) Proposed setup for monitoring clock misalignment. (b) Frequency-to-intensity conversion characteristic of delay-and-add discriminator. Copyright # 2006 IEEE. (a) Two-tap plot for various modulation timing misalignment. (b) Measured d and t parameters for various timing misaligments. Copyright # 2006 IEEE. Misalignment monitoring of an RZ-DQPSK transmitter. Copyright # 2008 IEEE. Monitoring signal power for (a) I/Q data misalignment and (b) carver/data misalignment. Copyright # 2008 IEEE. (a) Basic structure of CPFSK modulator with synchronous control. Numerically calculated optical modulation spectra for (b) CPFSK and (c) BPSK (DPSK). Copyright # 2006 IEEE. Modulation characteristics of PM and MZM. Copyright # 2006 IEEE. Receiver power penalty versus t0 for various ASK formats and DQPSK modulation methods. (a–c) sampling phases are optimized. Copyright # 2006 IEEE.

xxviii

List of Figures

FIGURE 6.1 FIGURE 6.2

FIGURE 6.3

FIGURE 6.4 FIGURE 6.5 FIGURE 6.6 FIGURE 6.7 FIGURE 6.8 FIGURE 6.9

FIGURE 6.10 FIGURE 6.11

FIGURE FIGURE FIGURE FIGURE

6.12 6.13 6.14 6.15

(a) Synchronous and (b) asynchronous eye diagrams and corresponding histograms of an NRZ signal. Examples of asynchronous histograms acquired from an NRZ signal (a) in presence of ASE noise for different values of signal-to-noise ratio; (b) in presence of intraband crosstalk; and (c) when the signal is impaired by fiber dispersion. Copyright # 2009, Institution of Engineering and Technology. Examples of histogram acquisition systems. (a) Based on electrical sampling using an unsynchronized digital oscilloscope (# 2009, IEEE). (b) Based on optical sampling using a sum-frequency generation crystal. Copyright # 2009, Institution of Engineering and Technology. Schematic of asynchronous histogram for an average Q-factor analysis technique. Copyright # 2009, Institution of Engineering and Technology. Relationship between Q-factor and average Q-factor using analysis of an NRZ signal. ○, a = 0.1; x, a = 0.2;
, a = 0.3; □, a = 0.4; þ, a = 0.49. Copyright # 2009, Institution of Engineering and Technology. Relationship between Q-factor and average Q-factor using analysis of an NRZ signal for different values of dispersion-impairing 10-Gb/s signal under analysis. ○, 0 ps/nm; x, 1190 ps/nm;
, 1530 ps/nm. Copyright # 2009, IEEE. Schematic of histogram crosspoint elimination method. Copyright # 2009, Institution of Engineering and Technology. Illustration of BER estimation method using multi-Gaussian fitting of asynchronous histogram after process to eliminate crosspoint data. Copyright # 2009, Institution of Engineering and Technology. Histogram of mark symbol of signal affected by intraband crosstalk (signal-tocrosstalk ratio of 19 dB) and degraded by Gaussian noise (a) without deconvolution and (b) after deconvolution and filtering. In the case of (b), the crosstalk floor around the mark symbol becomes clearly visible and may be evaluated. Copyright # 2009, Institution of Engineering and Technology. Diagram of method to estimate an asynchronous histogram of the signal impaired with the ASE noise proposed. Asynchronous histogram acquired from simulated signals with □, raised-cosine pulse shape; •, rectangular pulse shape filtered by Bessel filter with bandwidth of 70% of signal’s bit rate;
, 40-Gb/s raised-cosine signal degraded by accumulated chromatic dispersion of 34 ps/nm; continuous lines, estimated histograms using the raised-cosine approximation. Copyright # 2009, IEEE. Diagram of histogram estimation method A. Diagram of histogram estimation method B. Schematic diagram of OMS for OSNR evaluation using asynchronous histograms. Asynchronous histograms of reference signal considering an OSNR of 28 dB (dashed line) and signal under analysis with OSNR values of 14, 20, and 30 dB (continuous lines) for (a) power at OMS input of 0 dBm and (b) power at OMS input of 5 dBm. Optical preamplification in OMS is not considered in this case. Copyright # 2009, IEEE.

List of Figures

FIGURE 6.16

FIGURE 6.17

FIGURE 6.18 FIGURE 6.19 FIGURE 6.20

FIGURE 6.21 FIGURE 6.22 FIGURE 6.23 FIGURE 6.24

FIGURE 7.1 FIGURE 7.2 FIGURE 7.3 FIGURE 7.4

FIGURE 7.5

FIGURE 7.6

xxix

Dependence of estimated OSNR on OSNRSIG for OSNRREF = 28 dB (□), OSNRREF = 22 dB (
), and OSNRREF = 16 dB (○). Optical amplification or filtering in OMS is not considered in this case. The insets present eye diagrams of signal under analysis for OSNR values of 16 and 30 dB. (a) Power at OMS input of 0 dBm and (b) power at OMS input of 5 dBm. Copyright # 2009, IEEE. Dependence of estimated OSNR on OSNRSIG for OSNRREF = 28 dB (□), OSNRREF = 22 dB (
), and OSNRREF = 16 dB (○). An EDFA is used within the OMS for pre-amplification. The insets present the eye diagram of the signal under analysis for OSNR values of 16 and 30 dB. (a) Power at OMS input of 20 dBm and (b) power at the OMS input of 25 dBm. Copyright # 2009, IEEE. Simplified block diagram of S&H circuit. General block diagram of S&H system. Copyright # 2009, IEEE. Signal-to-sampling-noise ratio as function of aperture time for different values of transition time. (a) Using rectangular sampling impulse. (b) Using triangular sampling impulse. Numerical simulation results, symbols; analytical results, continuous line. General block diagram of equivalent S&H system. Copyright # 2009, IEEE. Simulated nonideal histogram acquisition system for optical monitoring. qeq/q and qcor/q as function of aperture time for 40-Gb/s signal. •, qeq/q for q = 8; ○, qeq/q for q = 10; n, qcor/q for q = 8; □, qcor/q for q = 10. Copyright # 2009, IEEE. Asynchronous histogram of a 40-Gb/s signal obtained with a nonideal sampling system from simulated signals with □, aperture time of 28 ps; •, aperture time of 44 ps. Continuous lines represent estimated histogram. Arrows indicate new relative maximums that result from averaging effect induced by nonideal sampling. Copyright # 2009, IEEE. Schematic of asynchronous delay-tap sampling technique. Sample pairs are separated by a fixed delay,
t. NRZ phase portraits for (a) 1-bit delay, and (b) ¼-bit delay. Phase portraits of an NRZ signal showing the effects of small changes in tap delay. Eye diagrams and phase portraits (1-bit delay, ¼-bit delay) for 10-Gb/s NRZ: (a) OSNR = 35 dB and no impairments; (b) OSNR = 25 dB; (c) CD = 800 ps/nm; (d) DGD = 40 ps; (e) crosstalk = 25 dB; and (f) OSNR = 25 dB, CD = 800 ps/nm, PMD = 40 ps, and crosstalk = 25 dB. Setup for generation of training sets. The impairment emulator adds known combinations of OSNR, CD, and first-order PMD to clean transponder signals. A polarization controller ensures a random distribution of power splits between principal states. AS, asynchronous sampler and outboard processing; PC, polarization controller; TDCM, tunable dispersion compensation module; TOF, tunable optical filter; Tx, transponder under test; VOA, variable optical attenuator. Experimental phase portraits for (a–c) 10-Gb/s NRZ and (d–f) 40-Gb/s DPSK showing the effects of CD and DGD; the OSNR is 14 dB for all cases. The tap

xxx

List of Figures

FIGURE 7.7

FIGURE 7.8

FIGURE 7.9

FIGURE 7.10 FIGURE 7.11

FIGURE 8.1

FIGURE 8.2

FIGURE 8.3

delay was set at 25 ps for both bit rates. The impairment levels were DGDeff (ps), absCD (ps/nm): (a) 0,0, (b) 39,0, (c) 0,1200, (d) 0,0, (e) 14,0, and (f) 2,400. Experimental measurements (gray) of simultaneous OSNR, CD, and DGD impairments for (a–c) 10-Gb/s NRZ and (d–f) 40-Gb/s NRZ-DPSK. Results for 250 test cases are ordered along the x axis by true values (shown in black). The test errors, stdTe, are quoted at the 2s level. Setup for 10-G NRZ WDM system test. OSNR and CD were monitored at the three tap points. An additional 400 ps/nm dispersion was added at each of the tap points as a further check of CD accuracy. The input power to the monitor was 18 dBm. Simulation results for 10-Gb/s NRZ demonstrating monitoring of simultaneous OSNR, CD, DGD, in-band crosstalk, and optical filter offset. The training ranges were for OSNR, 11–25 dB; CD, 1400 to 1400 ps/nm; and DGD, 0–50 ps (random g), crosstalk 15–24 dB, and filter offset 0–12 GHz. The training set consisted of 2000 random combinations of these impairments. Predictions for 1000 test cases (gray) are shown; true values are shown in black. The RMS error at the 2s level are (a) OSNR 0.3 dB, (b) absCD 15 ps/nm, (c) DGD 1.6 ps, (d) crosstalk 1.4 dB, and (e) filter offset 0.7 GHz. Simulation results for simultaneous measurements of absCD and DGDeff for 40Gb/s RZ-DQPSK. The training set contained 2000 cases with OSNR ranging from 14 to 28 dB; CD, 800 to 800 ps/nm; and DGD, 0 to 25 ps. Simulated phase portraits for (a–c) 40-Gb/s RZ-DQPSK and (d–f) 80-Gb/s polarization-multiplexed DPSK showing the effects of CD and DGD. The OSNR is 14 dB for all cases. The tap delay was set at 25 ps for both bit rates. The impairment levels were DGDeff (ps), absCD (ps/nm): (a) 0,0, (b) 25,0, (c) 0,800, (d) 0,0, (e) 25,0, and (f) 0,800. Data-encoded optical signals represented by intensity and phase as a function of time (left column) and complex electric field at center of time slot (right column). The signals correspond to (a) on-off keying, (b) binary phase-shift keying, (c) quaternary phase-shift keying, and (d) 16-state quadrature amplitude modulation. Layout for the direct measurement of the complex interference between data source and local oscillator. The two sources are split at the splitters SLO and SDATA, and recombined pairwise at the combiners CA and CB. Balanced photodetection of the two outputs of each combiner yields the in-phase and quadrature components of the interference if a relative p/2 phase shift is introduced in the optical path of one of the two sources between splitters and combiners. Copyright # 2006 OSA. Spectral representation of the sources involved in various implementations of linear optical sampling. (a) the Monochromatic local oscillator is spectrally overlapped with the signal. (b) the Broadband pulsed local oscillator is overlapped with the signal. (c) the Nonlinear interaction between the short pump pulse and the signal leads to an idler pulse, and the monochromatic signal is overlapped with the generated idler.

List of Figures

FIGURE 8.4 FIGURE 8.5

FIGURE 8.6

FIGURE 8.7

FIGURE 8.8

FIGURE 8.9

FIGURE 8.10

FIGURE 8.11

FIGURE 8.12

xxxi

Train of sampling pulses in the time domain showing the carrier-phase evolution under the field envelope. Phase samples measured on a 10-Gb/s BPSK signal. The phase is shown (a) without any processing, (b) after removal of a linear term leading to rotation in complex plane, and (c) after removal of both a linear term and a slowly-varying phase. Schematic of a polarization interferometer. The orthogonally polarized data and sampling sources are combined with a nonpolarizing 3-dB coupler. The two outputs of the coupler are sent to identical setups leading to a pair of balanced photodetectors, excepting that a p/2 phase shift is introduced in one arm, so that the real and imaginary parts of the interference are measured. (a) Picture of a 90-degree optical hybrid made with silicon-on-silica. (b) Setup for coherent photodetection of a data source with a copolarized pulsed local oscillator using the waveguide 90-degree optical hybrid. The sampling source is spectrally filtered to match the optical spectrum of the data source. The relative phase between the two measured signals is controlled by applying a low continuous voltage to a thermo-optic coupler. Copyright # 2005 OSA. Setup for the coherent photodetection of a data source by gating with four-wave mixing and detection with a monochromatic local oscillator. The sampling and data source are combined and propagate in a highly nonlinear fiber. The idler resulting from four-wave mixing of the sampling source acting as a pump on the data source is filtered and detected by homodyne detection with a local oscillator. Courtesy of Mathias Westlund and Peter A. Andrekson; copyright # 2009 OSA. Experimental results obtained with four-wave mixing and coherent detection. (a) Constellation diagram of 625,000 samples measured on a 10-GBaud QPSK signal. Electric-field samples located at the center of the bit slots in a time interval corresponding to 20% of the bit period are plotted in black. Other samples appear in gray, while lines correspond to interbit transitions averaged over a large number of similar transitions in the bit sequence. (b) Measured phase as function of position in the bit sequence. (c) Close-up of (b) in a 10-bit intervals. Courtesy of Mathias Westlund and Peter A. Andrekson; copyright # 2009 OSA. Constellation diagrams measured (a) on an ASE source, (b) on a monochromatic source with an OSNR of 15.8 dB, (c) on a binary PSK signal with an OSNR of 18 dB, and (d) on a binary PSK signal with an OSNR of 13 dB. (e) Measured variance of one of the symbols along the real and imaginary axes s2x and s2y versus the measured OSNR. Line of dots represents the theoretical relation between variance and OSNR. Copyright # 2006 OSA. (a, b) Constellation diagrams of phase-modulated signals generated with a phase modulator for differential phase of p/2 and p. (c) Standard deviation of the amplitude s
and phase sc of one of the symbols as a function of the differential phase (respectively round markers and square markers). Copyright # 2006 OSA. (a, b) Constellation diagrams of BPSK signal generated with Mach-Zehnder modulator for two different amplitudes of the drive voltage. (c) Standard deviation

xxxii

List of Figures

FIGURE 8.13

FIGURE 8.14

FIGURE 8.15

FIGURE 8.16

FIGURE 8.17

FIGURE 8.18

FIGURE 9.1 FIGURE 9.2

of the amplitude and phase of one of the symbols as a function of the drive voltage (respectively round markers and square markers). Copyright # 2006 OSA. Constellation diagrams measured at input and output of wavelength converter set to reduce the phase noise of a BPSK signal. (a) and (c) are measured before the wavelength converter, while (b) and (d) are the corresponding converted signals. Copyright # 2008 IEEE. Constellation diagrams measured after propagation of a noisy signal in a highly nonlinear fiber at (a) low power and (b) high power. The coupling between the intensity and phase of the samples is a sign of Gordon-Mollenauer phase noise. This coupling is quantified in (c) as a function of the average power of the source. Copyright # 2006 OSA. (a) Constellation diagram of a QPSK signal measured after two roundtrips in a recirculating loop (the propagation distance in transmission fiber is 800 km). Gordon-Mollenauer phase noise can be seen. (b) Coupling between intensity and phase in constellation diagrams measured for various propagation distances and/or launch powers. Measured relative instantaneous power and phase of an amplitude-modulated optical source after propagation in a nonlinear fiber. The coupling between power and phase is due to self-phase modulation, and the nonlinear coefficient of the fiber can be quantified with these data. Copyright # 2005, Institution of Engineering and Technology. Examples of measured constellation diagrams of a 10.7-Gb/s BPSK signal (a) after the transmitter, and (b) after propagation in 90 km of standard single-mode fiber. The trajectories of the electric field in the complex plane are plotted with continuous gray lines and the values of the field at the center of the bit slot are plotted with a black round marker. Courtesy of Michael G. Taylor. (a) Temporal transmission and phase of a semiconductor optical amplifier depleted by a short optical pulse. (b) Temporal intensity and phase of an optical pulse carved by an electro-absorption modulator. In (a), the period of the depleting pulse is 100 ps, while in (b), the period of the drive voltage is 25 ps. The lines correspond to the quantities measured with linear optical sampling and the markers correspond to the same quantities measured with the spectrogram technique. Copyright # 2005 OSA. Pilot-tone-based optical performance monitoring technique. Pilot-tone generation and detection methods. (a) Adding a small sinusoidal current to the laser’s bias current. (b) Dithering bias voltage of external modulator. (c) PM tone generation by using phase modulator. (d) Pilot-tone detection using FFT. (e) Using tunable electrical bandpass filter. (f) Using tunable local oscillator for the down-conversion of tone frequency. LD, laser diode; AM, amplitude modulator; PM, phase modulator; PD, photodetector; A/D, analog-to-digital converter; FFT, fast Fourier transform; BPF, tunable bandpass filter; RFD, radio frequency power detector; LOSC, tunable local oscillator.

List of Figures

FIGURE 9.3

FIGURE 9.4 FIGURE 9.5 FIGURE 9.6 FIGURE 9.7 FIGURE 9.8 FIGURE 9.9 FIGURE 9.10

FIGURE 9.11

FIGURE 9.12 FIGURE 9.13 FIGURE 9.14 FIGURE 9.15 FIGURE 9.16 FIGURE 9.17

FIGURE FIGURE FIGURE FIGURE

9.18 9.19 9.20 9.21

xxxiii

Effects of modulation index and frequency of pilot tone on 10-Gb/s NRZ signal (pattern length = 231  1). (a) Pilot-tone-induced power penalty measured at low tone frequencies. (b) Maximum allowable modulation indices of high-frequency pilot tones for 0.5-dB penalty. Effect of highpass filter on data signal. Eye closure penalty calculated while varying low cut-off frequency in comparison with simulation results (bit rate = 2.5 Gb/s, pattern length = 27  1). Experimental setup. Power penalty measured while varying the bit rate (low cut-off frequency = 1 MHz). Power penalty measured while varying tone frequency (bit rate = 2.5 Gb/s, pattern length = 27  1, MI = 12%). Mechanisms of performance degradation caused by XGM and SRS. Measured optical and electrical spectra. (a) Optical spectrum measured after 640km transmission. (b) Electrical spectrum measured after 640-km transmission (without using control channel). (c) Electrical spectrum measured after first EDFA (with using control channel). (d) Electrical spectrum measured after 640-km transmission (with using control channel). Effects of ghost tones on pilot-tone-based monitoring technique. (a) WDM signals with corresponding pilot tones. (b) Ghost tones generated by XGM and SRS. (c) Pilot tones measured without using a demultiplexing filter. (d) Pilot tones measured after demultiplexing WDM channels. Scalability of pilot-tone-based monitoring technique. Solid line, limits imposed by SRS; dashed line, limits imposed by XGM. Principle of tone-based channel power and wavelength monitoring technique. (a) Experimental setup. (b) Imaginary part of FFT-converted data. (c) Real part of FFT-converted data. Monitoring errors of 16 WDM channels: (a) channel power and (b) wavelength. Experimental setup used for demonstration of pilot-tone-based monitoring technique for optical path and crosstalk. Electrical spectra of pilot tones measured at the output of the switch for l1 in OXC 4 (a) under normal operation, and (b) under switch failure condition. (a) Schematic representation of OADM implemented with the pilot-tone-based monitoring technique. (b) Schematic representation of OXC implemented with the pilot-tone-based monitoring technique. OSW, optical switch; FPF, Fabry-Parot filter; PD, photodetector; Tx, transmitter; Rx, receiver; ADC, analog-to-digital converter; FFT, fast Fourier transform. Operating principle of monitoring technique based on PM pilot tones. Experimental setup of PM pilot-tone-based monitoring technique. Amplitudes and ratio of AM components measured at two adjacent AWG ports. Frequency monitoring error measured while varying tone frequency (accumulated dispersion = 1000 ps/nm) when the WDM channel was operated at 192.8 THz.

xxxiv

List of Figures

FIGURE 9.22 FIGURE 9.23 FIGURE 9.24 FIGURE 9.25 FIGURE 9.26 FIGURE 9.27 FIGURE 9.28 FIGURE 9.29 FIGURE 9.30 FIGURE 9.31 FIGURE 9.32 FIGURE 9.33 FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE

9.34 9.35 9.36 9.37 9.38 9.39 9.40

FIGURE 10.1

FIGURE 10.2 FIGURE 10.3 FIGURE 10.4

FIGURE 10.5

Frequency monitoring errors measured after transmission of 10-Gb/s signals over 320 km of SMF. The inset shows BER curve measured when PM pilot tone was added to optical signal. Experimental setup. Amplitudes and ratio of pilot tones (channel 3) measured using PD 3 and PD 4. (a) Frequency and (b) power errors of seven WDM signals measured after 640-km transmission over SMF. (a) Experimental setup to measure the CD. (b) MI of pilot tones measured at the receiver while varying the CD. Effects of SPM and PMD on pilot-tone-based CD monitoring technique. Monitoring errors caused by (a) SPM and (b) PMD. Operating principle of CD monitoring technique using chirped pilot tones. Experimental setup used for demonstrating CD monitoring technique based on chirped pilot tone. Normalized tone power measured while varying CD. Maximum monitoring error due to PMD measured while varying DGD. Monitoring error due to SPM measured while varying signal power incident on SMF. Experimental setup used for demonstrating CD monitoring technique based on pilot tone carried by BLS. Measured power penalty due to pilot tone. Measured power variation in received pilot tone as a function of CD. Effect of PMD on received power of pilot tones. Experimental setup to demonstrate in WDM system. Measured dispersion and monitoring error of each WDM signal. Magnitude of pilot tone versus tone frequency and DGD. Magnitudes of pilot tones measured while varying both DGD and CD (a) when DSB pilot tone is used, and (b) when SSB pilot tone is used. Electrical noise-free eye prior to ADC at 1700-ps/nm residual CD with symbol ðmÞ patterns di ðm ¼ 3Þ aligned to each interference pattern (a) and according to lookup table obtained at sampling instant t0 = 0 (1 sample/symbol, 4-bit quantization) at OSNR = 14 dB (b). State-based OSNR estimation at a residual CD of 1700 ps/nm and variations in the sampling phase t0. State-based OSNR estimation for NRZ and DB modulation with analog and quantized samples, and for transmission with significant SPM and XPM. Two examples for reference-based estimation of launch power–induced SPM and residual CD for (a) Plp = 0 dBm, CD = 1250 ps/nm, and (b) Plp = 15 dBm, CD = 3050 ps/nm. The black circles indicate the given parameter set, and the white arrow indicates the estimated parameter set. Note: The diagram is rotated in (b). Digital coherent receiver with one optical 90 hybrid for each polarization, ADC, and subsequent digital postprocessing by timing recovery, FIR filter bank, and carrier phase recovery.

List of Figures

FIGURE 10.6 FIGURE 10.7 FIGURE 10.8 FIGURE 10.9 FIGURE 10.10

FIGURE 10.11 FIGURE 10.12 FIGURE 10.13 FIGURE 10.14 FIGURE 10.15 FIGURE 10.16 FIGURE 10.17 FIGURE 10.18 FIGURE 10.19 FIGURE 10.20 FIGURE 10.21 FIGURE 10.22 FIGURE 11.1

xxxv

Implementation of FIR filter structure with outer and inner butterfly arrangement. Exemplary coefficients of 11 tap FIR filters for CD = 1000 ps/nm and mean DGD = 30 ps. Effect of PDL depending on input polarization. Copyright # 2009 IEEE. Recovery of the quadratic phase of H^CD(f), for example, of CD = 1000 ps/nm. Copyright # 2009 IEEE. Joint estimation for given CD = 1000 ps/nm and DGD varying from 0, 30 and 60–90 ps over a range of OSNR values between 12 and 24 dB (a) and estimation of pure CD over a wide range (b). The bars indicate the standard deviation of the estimation. Copyright # 2009 IEEE. Deviation mCD and biased error sCD in ps/nm of the mean value of estimated CD compared to given CD for combinations of CD and PDL. Copyright # 2009 IEEE. Estimation example for a CD value of 120,000 ps/nm in a 112-Gb/s PDM-QPSK uncompensated transmission link. The initial scan is performed in steps of dCD = 200 ps/nm. Copyright # 2009 IEEE. Estimation performance versus mean DGD (full-order PMD) for RZ and NRZ pulse shaping of 112-Gb/s PDM-QPSK (OSNR = 14 dB) and PDM-16-QAM (OSNR = 17 dB). Copyright # 2009 IEEE. Exemplary state of polarization evolution in Stokes space. SOP evolution in presence of PDL for 43-Gb/s PDM-QPSK. Copyright # 2009 IEEE. Estimation example for a PMD of 10 ps of the instantaneous mean DGD ht(f)i for CD = 1000 ps/nm and 9-dB worst-case PDL. Note that the distribution of ht(j)i proves to be nearly Maxwellian. Copyright # 2009 IEEE. Statistics of exemplary DGD estimation for CD = 1000 ps/nm and 9-dB worst-case PDL with respect to deviation of the estimation error sDGD, mean value mDGD, and standard deviation within 5-ps DGD windows. Copyright # 2009 IEEE. Deviation mDGD and biased error sDGD in ps of the mean value of estimated DGD compared to given DGD for combinations of CD and PDL. Copyright # 2009 IEEE. Simulative PDL estimation for 43-Gb/s PDM-QPSK. Copyright # 2009 IEEE. Averaged PDL power spectrum adapted to various levels according to the influence of noise power in the system. Copyright # 2009 IEEE. Unbiased estimation error sPDL over individual channel realizations with individual DGD for combinations of CD and PDL at an OSNR of (a) 18 dB and (b) 22 dB. Copyright # 2009 IEEE. PDL estimation performance at OSNR = 22 dB versus the number of concatenated PDL elements with a mean PDL of 5 dB. Copyright # 2009 IEEE. (a) Common nonlinear-threshold-based monitor integrated with 2-R optical regenerator. Monitoring signal is ratio of average power measured at P2 normalized by input power at tap P1. The variable optical attenuator can be used to

xxxvi

List of Figures

FIGURE 11.2

FIGURE 11.3

FIGURE 11.4 FIGURE 11.5 FIGURE 11.6

FIGURE 12.1 FIGURE 12.2 FIGURE 12.3 FIGURE 12.4 FIGURE 12.5 FIGURE 12.6 FIGURE 12.7 FIGURE 12.8

maintain constant input power. (b) Optical spectrum of 40-Gb/s NRZ-OOK signal generated by XPM in SOA and OSNR monitoring sensitivity obtained by moving output filter to different wavelengths. Copyright # 2005 IEEE. Chromatic dispersion monitoring on 40-Gb/s RZ-DPSK signals using XPM in highly nonlinear fiber. (a) Setup and concept: only clock tones from the RZ pulse train are observed in XPM-induced spectrum of CW probe, but chromatic dispersion distorts the signal-creating power at other wavelengths for nearly background-free monitoring. (b) Monitoring signal variation at 25-GHz offset from probe wavelength for various signal input powers. Copyright # 2008 IEEE. (a) 640-GHz-intensity power spectrum generated using crossphase modulation in fiber. (b) Chromatic dispersion and (c) polarization mode dispersion monitoring signals generated by detecting clock tone in intensity power spectrum of a 320Gb/s signal. Permission from Macmillan Publishers Ltd., Nature Photonics # 2009. Common monitoring configurations on an amplified long-haul link. Comparison of optical spectra of signals using (a) linear photodetector and (b) nonlinear two-photon absorption detector. Nonlinear detection reveals differences between 40-Gb/s NRZ-DPSK, 10-Gb/s RZ-OOK signals, and filtered ASE noise. Experimental dependence of the zero-delay peak value of autocorrelation trace as function of OSNR for accumulated dispersions of 0 and 470 ps/nm. Solid lines are curves to guide the eye. Total average power (ASE noise plus signal) is constant upon filtering with a 0.2-nm optical bandpass filter. (a) Comparison of required OSNR between DPSK and OOK formats. (b) Comparison of nonlinear tolerance between RZ-DPSK and RZ-OOK formats. (a) Power penalty as function of accumulated chromatic dispersion for 42.7-Gb/s RZ-DQPSK signal. (b) Power penalty as function of DGD for 42.7-Gb/s RZDQPSK signal. (a) Comparison of robustness to optical filtering for OOK and RZ-DPSK formats. (b) Sensitivity penalty as function of frequency offset for DPSK and DQPSK format. Phase modulation of the optical signal: (a) generation of DPSK signal using MZM, (b) generation of RZ-DQPSK in parallel configuration, and (c) generation of RZDQPSK in serial configuration. Demodulator and balanced receiver for (a) DPSK and (b) DQPSK. (a) Phase distribution and decision regions of received signals with NPN for nonlinear phase shift of 1 rad (left) and 2 rad (right). (b) Corrected signal distribution for nonlinear phase shift of 1 rad (left) and 2 rad (right). (a) Diagram of CD-independent PMD monitoring technique. (b) Alignment of polarizer with respect to signal in case without and with DGD. (c) Variation of RF spectrum power depending on level of DGD. Partial-bit DI for NRZ signal monitoring. Fraction of bit interferes with following bit, which leads to pulse carving in destructive port. Pulses produce a strong clock tone.

List of Figures xxxvii

FIGURE 12.9 Modification to transmitter for pilot tone monitoring techniques. FIGURE 12.10 Improvement of carrier-to-noise ratio (CNR) in DPSK signal over OOK signal for pilot tone monitoring technique. FIGURE 12.11 Diagram of optical circuit for mixing of optical signal with local oscillator light (90 optical hybrid). FIGURE 12.12 Signal constellation obtained using linear optical sampling: (a) PSK signal modulated with MZM, (b) signal modulated with PM at driving voltage of 2/3vP, and (c) PSK signal with OSNR = 13 dB. FIGURE 12.13 (a) Evaluation of OSNR using linear optical sampling. (b) Evaluation of NPN using linear optical sampling. FIGURE 12.14 (a) Diagram of self-homodyne phasor monitor. (b) Symbol values obtained from monitor. (c) Constellation diagram after rotation through phasor-estimated phase. FIGURE 12.15 Constellation diagram obtained with differential phasor. (a) Phase modulation with driving voltage of VP and 1/2VP. (b) PSK modulation without and with phase mismatch between arms of MZM. FIGURE 12.16 Waveform of RZ-DPSK signal and corresponding asynchronous amplitude histogram. Figure also shows parameters used for monitoring signal impairments. FIGURE 12.17 Waveform and the corresponding AAH for the NRZ-DPSK signals. (a) Parameter for monitoring OSNR. (b) Parameters for monitoring of CD. FIGURE 12.18 (a) Diagram of the monitor based on delay-tap sampling. (b) Acquisition of two samples with relative delay
t. (c) Construction of delay-tap plot. FIGURE 12.19 Two-tap plot constructed by sampling the constructive and destructive ports of the demodulator DI. FIGURE 12.20 Diagram of phase-offset monitoring technique based on optical receiver and limiting amplifier. FIGURE 12.21 Analysis of OPM techniques by type of monitored impairment. FIGURE 13.1 Configuration of coherent receiver with balanced detector. FIGURE 13.2 Configuration of phase and polarization-diversity receiver. FIGURE 13.3 (a) Basic single-carrier coherent optical system. (b) Block diagram of digital signal processing. FIGURE 13.4 Block diagram of RF OFDM transmitter/receiver. FIGURE 13.5 Time-domain OFDM signal for one complete OFDM symbol with cyclic prefix. FIGURE 13.6 CO-OFDM system with direct up-/down-conversion. FIGURE 13.7 Simplified block diagram of coherent optical spectrum analyzer. LO, local oscillator; PD, photodetector; TZ, transimpedance amplifier. Copyright # 2002 IEEE. FIGURE 13.8 Comparison of measurement of DFB-LD linewidth by COSA and by diffraction grating-based OSA. Copyright # 2002 IEEE. FIGURE 13.9 Schematic for in-band OSNR and spectrum monitoring based on swept coherent detection. FBC, fiber Bragg grating; PC, polarization controller; VOA, variable optical attenuator. Copyright # 2006 IEEE. FIGURE 13.10 Operating principle of in-band, high-resolution swept coherent detection scheme. Copyright # 2006 IEEE.

xxxviii List of Figures

FIGURE 13.11 Block diagram of CD and PMD monitoring using coherent detection. ADC, analog-to-digital converter; BPF, bandpass filter; ESA, electrical spectrum analyzer; OSA, optical spectrum analyzer. Copyright # 2005 IEEE. FIGURE 13.12 Schematic of coherent frequency–selective polarimeter. BPF, bandpass filter; LPF, low-pass filter; PA, power amplifier; PH, photodetector; PT, polarization transformer. Copyright # 2004 IEEE. FIGURE 13.13 Coherent transmission system with butterfly-structured equalizer filter. Copyright # 2008 IEEE. FIGURE 13.14 Block diagram of real-time monitoring transmission experiment. Tunable laser and polarimeter were used to measure PMD of channel independently of monitor. Copyright # 2008 IEEE. FIGURE 13.15 Schematic of investigated OFDM system. OFDM parameters: data rate 100 Gb/s, 256 subcarriers, 16-QAM, 12.5% CP. Twelve spans: 80 km-SSMF per span, D = 17 ps/nm/km, mean PMD = 10 ps. Copyright # 2009 IEEE. FIGURE 13.16 (a) Estimated DGDs at OSNR = 20 dB using Savitzky-Golay filtering. (b) Estimated inverse SNR versus optical input power. Copyright # 2009 IEEE. FIGURE 13.17 Conceptual diagram of TITO coherent optical MIMO-OFDM model. FIGURE 13.18 Performance of CD monitoring through channel estimation. FIGURE 13.19 Monitored system Q and OSNR as function of input OSNR. FIGURE 13.20 Experimental setup for optical performance monitoring with SITO-MIMO CO-OFDM system. (There is a polarization controller before DGD emulator.) AWG, arbitrary waveform generator; DMZ, dual MZ modulator; EDFA, erbium-doped fiber amplifier; TDS, time-domain sampling scope. FIGURE 13.21 (a) OSNR monitoring result. (b) Q-factor monitoring result. Both are measured in back-to-back transmission. FIGURE 13.22 Estimated channel responses for (a) x and (b) y polarization components X-axes are the frequencies normalized to OFDM subcarrier spacing. FIGURE 13.23 (a) CD monitoring versus transmission distance with and without DGD. (b) CD monitoring error versus DGD after 1008-km transmission. FIGURE 13.24 Monitoring results in CO-OFDM system with 16-QAM for (a) OSNR and (b) Q-factor. FIGURE 13.25 CD monitoring result in CO-OFDM system with 16-QAM. FIGURE 14.1 Business interface model. FIGURE 14.2 Data networking and transport networking. FIGURE 14.3 ASTN architecture. FIGURE 14.4 Modeling of layer networks. FIGURE 14.5 OTN information structure. FIGURE 14.6 Transfer parameters of reconfigurable OADM without amplifiers. FIGURE 14.7 TMN-integrated maintenance process. FIGURE 14.8 OTN performance metrics. FIGURE 14.9 OSNR measurement method. FIGURE 14.10 Noise shaping. (a) Various noise floors caused by OADM. (b) OSNR errors caused by various noise floors.

List of Figures

FIGURE FIGURE FIGURE FIGURE FIGURE

14.11 14.12 14.13 14.14 14.15

FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE FIGURE

14.16 14.17 14.18 14.19 15.1 15.2 15.3 15.4 15.5

FIGURE FIGURE FIGURE FIGURE

15.6 15.7 15.8 15.9

FIGURE 15.10 FIGURE 15.11 FIGURE 15.12 FIGURE 15.13

xxxix

Power spectral density. In-band OSNR measurement with polarization extinction method. Eye diagram and amplitude histograms. Error probabilities. Influence of dispersion on 10-Gb/s binary signal. Without dispersion: (a), (c), and (e); with dispersion: (b), (d), and (f). BIP8 processing. Transmitter (a), Receiver (b). Optical performance monitoring overview. Embedded monitoring and external test equipment. Fault coverage versus cost for monitoring equipment. Typical bidirectional long-haul optical transmission link. Multiple-channel DWDM transmitter/receiver. Amplified link. Raman amplifier. Accumulated dispersion along long-haul transmission system. (a) Dispersion management system. (b) Schematic of evolution of cumulated dispersion along transmission distance. Fiber dispersion. Example of OSA/OCA measurement in DWDM transmitter. OSA/OCA monitoring at receiver node. Required OSNR and receiver sensitivity: achieving error-free operation (BER < 10–12) after 1000-km SMF transmission with no CD compensation using special SCM transmitter/receiver. Intermediate repeater node with and without add-drop channels: (a) add-drop node, and (b) intermediate node without add-drop. Fiber PMD. Example of 40-G transceiver using PMD compensator. Spectrum of long-haul transmission.

This page intentionally left blank

List of Tables Table 3.1

Overview of Dispersion Tolerance at 42.7 Gb/s

Table 3.2 Table 4.1

Summary of Selected Advanced CD Monitoring Techniques Summary of Major PMD Monitoring Techniques

Table 5.1 Table 7.1

Comparisons of Monitoring Schemes for Timing Misalignment Summary of Optical Performance Monitoring Techniques Covered in This Chapter

Table 7.2

Independent Validation of Impairment Emulator

Table 7.3

Test Errors for Transponder T2 Showing Effects of Including Combinations of Transponders in Training

Table 7.4

Simultaneous OSNR and CD Measurements on 800-km WDM Test Bed

Demonstrating Improvement in DGD and CD Accuracy with Increasing Size of Training Set Table 10.1 Comparison of Measured and Estimated OSNR from Measured Data for Various Combinations of Residual CD and Launch Power Table 7.5

Table 10.2 Normalization Constants Table 11.1 Nonlinear Effects and Their Applications to OPM Table 12.1 Comparison of Measurement Ranges between OPM Techniques for Phase-Modulated Signals Table 13.1 OFDM Parameters for 4-QAM Transmission Table 13.2 OFDM Parameters for 16-QAM Transmission Table 14.1 G.805 Architectural Components Table 14.2 OTN Services Table 14.3 OTN Supervision Table 14.4 Comparison of OTN Monitoring Methods Table 15.1 Optical Signal Performance under Normal Conditions Table 15.2 Optical Signal Performance for EDFA under Normal Conditions Table 15.3 Typical OSNR Requirements for Long-Haul System Table 15.4 Example of 10G System Budget Allocation (NRZ, 2000 km, SMF) Table 15.5 Chromatic Dispersion and Polarization Mode Dispersion Requirements for Typical Direct Detection System

xli

This page intentionally left blank

Preface These days, optical networks worldwide have been widely deployed in various network scales, including long-haul backbone and metropolitan areas, as well as regional broadband access. In order to assure a certain quality of service and a service-level agreement of the data delivery as requested by the network service subscribers, network management plays a very important role in the operation and administration of practical optical networks. Performance management is among the key aspects of network management that assures the signal quality during signal transmission, switching, and routing. It provides several important network functions including (1) providing feedback in the adaptive signal compensators and equalizers; (2) control of network elements; (3) link setup, control, and optimization; and (4) fault forecasting, detection, diagnosis, and localization, as well as resilience mechanism activation. Conventionally, performance management in optical systems and networks involves simple measurements of system parameters such as optical power, wavelength values, and the system noise levels, followed by the correlation analysis using the network management software. With the ever-increasing network bandwidth demand, as well as the advances in optical technologies, the signal transmission data rate on each wavelength channel has been increasing beyond 40 Gb/s or even more. At the same time, the wavelength-division multiplexing technique has also been widely adopted to combine multiple wavelength channels on the same fiber link, thus it has greatly enhanced the system transmission capacity. However, such high-speed optical signals are more vulnerable to some optical system impairments such as fiber chromatic dispersion, polarization mode dispersion, fiber nonlinearity, etc., and the induced performance degradations are getting more prominent. Therefore, more advanced optical signal processing techniques are required to characterize the signal quality and assure the transmission performance of such high-speed optical signals in various kinds of modulation formats. Over the past decade there have been numerous research efforts in devising many advanced techniques for optical performance monitoring (OPM) for optical systems and networks. Most of the reported OPM techniques have been scattered throughout journals, conference proceedings, and several book chapters. Yet, there is no such book available in the market solely for the discussion of OPM techniques. This book aims to provide comprehensive and in-depth discussions of these advanced OPM techniques, which would be expected to play an important role to facilitate the network management of the next-generation photonic networks. All chapters in this book have been prepared and written by the leading experts and researchers in the OPM field. The first chapter provides a comprehensive overview of all the relevant aspects of OPM, as well as its perspectives and challenges. The following fourteen chapters are organized in three different sections. The first section comprises Chapters 2–5, which discuss the OPM techniques for several common performance metrics, including optical signal-to-noise ratio, chromatic dispersion, polarization mode dispersion, and timing alignment, respectively. Each chapter provides an extensive overview and discussion of various feasible approaches to characterize the respective performance metric. These approaches will be further compared in terms of their technical merits and limitations, as well as their robustness in the presence of other system impairments. The second section comprises Chapters 6–11, which discuss several selected advanced OPM techniques, based on asynchronous amplitude histograms, asynchronous delay-tap sampling, linear

xliii

xliv

Preface

optical sampling, radio-frequency pilot tones, electronic digital signal processing techniques, and nonlinear optical effects, respectively. The goal of these chapters is to provide a detailed and indepth discussion of the principles of the advanced optical signal processing techniques that could be applied for OPM. Although some of these OPM techniques may have also been included as the feasible approaches in the chapters in the first section, the focus and treatment of this discussion will be much different. In the first section, the principles of the feasible approaches for monitoring the individual performance metric may not be discussed in detail, as the variety of techniques is the focus. In contrast, in the second section, the principles and theories of those OPM techniques are discussed in detail, and OPM is just their application. I believe this treatment of discussion in both breadth and depth will give the reader a more thorough understanding of both the principles and the applications of the advanced OPM techniques discussed in this book. The third section comprises Chapters 12–15, which discuss the applications and roles of OPM in various optical systems and networks, including phase-modulated optical systems, coherent optical systems, optical transport networks, and optical long-haul transmission systems, respectively. In view of the different system requirements and unique characteristics of these optical systems and networks, the considerations of OPM and the respective techniques will be quite different. On the whole, this book is intended as a general reference for researchers in both academia and the industry who work in the field of optical networks. It covers most of the recent advances of the optical signal processing techniques for OPM in various optical systems and networks. The comprehensive discussions have been carefully organized to illustrate the principles and applications of OPM in both breadth and depth. Calvin C. K. Chan Hong Kong, 2010

Acknowledgments First, I would like to express my gratitude to all chapter contributors who have kindly spent their precious time and made tremendous efforts to contribute to this book. All of the chapter contributors are really the leading experts in this area of optical performance monitoring. I believe their collective efforts will definitely make this book a great success. I would like to thank Prof. Lian-kuan Chen, Prof. Kwok-wai Cheung, Prof. Frank Tong, and Prof. Chinlon Lin at the Chinese University of Hong Kong for their enlightenment, encouragement, and collaboration in this area of optical networks and system technologies. I would also like to thank Elsevier for giving me the opportunity to prepare this book and the valuable support and excellent project management from its staff, especially Tim Pitts, Melanie Benson, Naomi Robertson, and Sarah Binns. Their full support has made a smooth preparation of this book. Last, but not the least, I am most indebted to my wife Esther, daughter Ariel, and son Leo, for their consistent love, encouragement, and patience.

xlv

This page intentionally left blank

List of Contributors Trevor B. Anderson Monitoring Division Inc., Melbourne, Australia; and National ICT Australia Ltd. (NICTA), Victoria Research Laboratory, Australia Paulo Andre´ Instituto de Telecomunicac¸o˜es and Departamento de Fı´sica, Campus de Santiago, Aveiro, Portugal Lian-Kuan Chen The Chinese University of Hong Kong, Hong Kong SAR, The People’s Republic of China Yun C. Chung Department of Electrical Engineering and Computer Science, Korea Advanced Institute of Science and Technology, Korea Ken Clarke Monitoring Division Inc., Melbourne, Australia Liliana Costa Instituto de Telecomunicac¸o˜es, Campus de Santiago, Aveiro, Portugal Sarah D. Dods Monitoring Division Inc., Melbourne, Australia Christophe Dorrer Laboratory for Laser Energetics, University of Rochester, Rochester, New York Wolfgang Grupp JDSU Deutschland GmbH, Eningen u.A., Germany Fabian N. Hauske Huawei Technologies, European Research Center, Munich, Germany Don Hewitt National ICT Australia Ltd. (NICTA), Victoria Research Laboratory, Australia Xin Jiang Department of Engineering Science and Physics, College of Staten Island; and The City University of New York, Staten Island Daniel C. Kilper Bell Laboratories, Alcatel-Lucent, New Jersey Adam Kowalczyk Monitoring Division Inc., Melbourne, Australia; and National ICT Australia Ltd. (NICTA), Victoria Research Laboratory, Australia Bartłomiej Kozicki NTT Network Innovation Laboratories, NTT Corporation, Japan

xlvii

xlviii

List of Contributors

Maxim Kuschnerov University of the Federal Armed Forces, Munich, Germany; and Institute for Communications Engineering, Neubiberg, Germany Jun Haeng Lee Samsung Advanced Institute of Technology, Samsung Electronics, Korea Jonathan C. Li Monitoring Division Inc., Melbourne, Australia Ruben S. Luis Instituto de Telecomunicac¸o˜es, Campus de Santiago, Aveiro, Portugal; and Center of Volcanology and Geological Risk Assessment, University of Azores, Ponta Delgada, Portugal Zhongqi Pan Department of Electrical and Computer Engineering, University of Louisiana at Lafayette Paul K.J. Park Samsung Electronics, Korea William Shieh Department of Electrical and Electronic Engineering, The University of Melbourne, Victoria, Australia Yan Tang Department of Electrical and Electronic Engineering, The University of Melbourne, Victoria, Australia Anto´nio Teixeira Instituto de Telecomunicac¸o˜es and Departamento de Electro´nica, Telecomunicac¸o˜es e Informatica, Campus de Santiago, Aveiro, Portugal Alan Willner Department of Electrical Engineering, University of Southern California, Los Angeles Xiaoxia Wu Department of Electrical Engineering, University of Southern California, Los Angeles Jeng-Yuan Yang Department of Electrical Engineering, University of Southern California, Los Angeles Xingwen Yi Department of Electrical and Computer Engineering, University of California, Davis Changyuan Yu National University of Singapore; and A*STAR Institute for Infocomm Research, Singapore

About the Editor Calvin C. K. Chan received his B.Eng., M.Phil., and Ph.D. degrees from the Chinese University of Hong Kong, all in Information Engineering. In September 1997, he joined the Department of Electronic Engineering at the City University of Hong Kong as a Research Assistant Professor. At both universities he worked on design and experimental demonstration of high-speed all-optical timedivision-multiplexed and wavelength-division-multiplexed tunable channel multi-access networks. He also worked on surveillance techniques for fault identification in various kinds of optical network elements. In June 1999, he joined Bell Laboratories, Lucent Technologies in Holmdel, NJ, United States, as a member of the technical staff, where he worked on control of widely tunable semiconductor lasers and realization of an optical packet-switch fabric with terabit-per-second capacity. In 2001, he served as Senior Optical System Engineer at Jedai Broadband Networks, Inc. in New Jersey, where he worked on the design of optical access networks and optical wireless systems. In August 2001, he joined the Department of Information Engineering at the Chinese University of Hong Kong, and now serves as an Associate Professor there. He has served as a member of the Technical Program Committees on many international conferences, including the prestigious OFC/NFOEC, OECC, ICCCAS, Photonic in Switching, APOC, ICOCN, COIN, ChinaCom, ICAIT, etc. Currently, he serves as an Associate Editor for IEEE/OSA Journal of Optical Communications and Networking. Dr. Chan has published more than 180 technical papers in refereed international journals and conferences, as well as two book chapters. He holds one issued U.S. patent. His main research interests include optical metro/access network architectures and their enabling technologies, high-speed optical packet-switching techniques, and optical network management.

xlix

This page intentionally left blank

CHAPTER

Optical performance monitoring: Perspectives and challenges

1

Alan Willner, Xiaoxia Wu, Jeng-Yuan Yang Department of Electrical Engineering, University of Southern California, Los Angeles, California, USA

1.1 INTRODUCTION Today’s telecommunication networks function in a fairly static fashion and are built to operate within well-defined specifications. A single parameter out of specification can bring down the entire optical network. Since the parameters can change over time, operating and managing an existing network requires a fair amount of labor resources. As the demand for increasing bandwidth and higher data rates creates increasingly complex networks, more variables are introduced that decrease the mean-time-to-failure and increase the mean-time-to-repair if a failure occurs. The forthcoming IEEE 100 Gb/s standard, 802.3ba, provides greater bandwidth, higher data rates, and a mixture of advanced modulation formats, promising to tighten the window of network operability and further, as illustrated in Figure 1.1, emphasizing the need for a “smart” network capable of managing its operation as parameters vary.1,2 Optical performance monitoring (OPM) is one potential means of either widening this window or helping to maintain channel operation within a small window, such that rapid growth of high-performance optical networks can be enabled.

Format Number of channels

Bit rate

Power

Nonlinearities Polarization effects

Dispersion

FIGURE 1.1 Window of operability is shrinking as networks become more complex. © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00001-8

1

2

CHAPTER 1 Optical performance monitoring

Highly desirable features of a future “smart” network include higher robustness, reconfigurability, flexibility, and security. To enable robust and smart operation, the network should be able to: (1) measure its physical state and the quality of the propagating data signals; (2) automatically diagnose and repair the failures; (3) take actions before data loss and failure occur; (4) allocate resources, including signal wavelength/power, tunable compensation/equalization, data coding, path determination, and channel bandwidth; (5) change routing tables and redirect traffic based on physicallayer conditions; and (6) detect accidental and malicious security risks. All of these desirable features require the ability to first measure the network elements, paths, and data in an accurate and repeatable fashion. This deficiency in the measurement science and metrology of telecommunication networks leads to the overbuilding and inefficient use of current networks. The ability to measure the network infrastructure and resources in a real-time fashion is typically known as network monitoring. In order to enable robust, self-managed, and smart operation, OPM should be able to accurately measure and rapidly reflect the physical states of network elements, paths, and the quality of propagating data signals. The ability to measure both the data quality and the network path’s physical state in a real-time fashion has traditionally proven difficult due to the large number of impairments, the variety of data formats, the variety of data rates, and the continuous growth and change of the network structure. As a result, multiple impairments that cause the alarms must be isolated, localized, and compensated, requiring real-time monitoring and dynamic feedback control. This chapter will first provide an overarching vision and critical challenges of OPM, and then describe the specific parameters that a network might want to monitor, laudable OPM-enabled functionalities for next-generation optical networks, and desirable features of OPM.

1.1.1 Overarching vision The growth of Internet-driven wired traffic has been increasing exponentially, well beyond the simple addition of new users. Fueled by the ability to traverse appreciable distances of fiber optics, optical communication networks will undoubtedly continue to grow, and this growth will be symbiotic in that growth in one area depends on the available capacity in others. Traditionally, data are passed from the wireless or radio-frequency (RF) domain to the fiber optic backbone for long-distance transmission. “Nortel estimates that by 2010 there will be 10 devices connected to the network for every person using them, resulting in five billion connection points.”3 This puts an enormous strain on network resources as users contend for increasing bandwidth and better quality of service. Today, networks suffer from an inability to assess both the physical state of the network and the quality of the propagating data, leading to the over design and inefficient use of network resources. In order to enable a robust, scalable, self-managed, and smart operation, the network should be able to accurately monitor the physical state of network elements, paths, and the quality of propagating data signals. It is becoming clear that OPM will be playing an increasingly important role in managing the emerging networks. One of the major operational areas that carriers probably need to deal with is very quickly identifying a fault that has occurred, where it occurred, and what caused it. As networks move to all-optical technologies, identifying these faults will become more difficult. This is because many of the faults are what we call “soft faults,” caused by things like optical impairments, which do not bring the whole system down, but rather result in noise like conditions that impair communications. OPM will be an essential tool to identify these soft faults and will probably play an important role in reducing the operation costs for the future optical networks.4 Figure 1.2 illustrates a future self-managed network with OPMs integrated.

1.1 Introduction

• • • •

Network controller

Network elements

O P M

Re-route traffic Network control Network management Provisioning

Network elements

Network elements OPM

3

O P M

• Heterogeneous traffic • Variable rates and formats • Various OSNR/CD/PMD

FIGURE 1.2 A self-managed network with optical performance monitoring.

To operate at higher capacity, performance monitors can keep the system above the “red line” and operate under tougher conditions. Network carriers will not need to overbuild (i.e., waste money) the network. Some basic issues are already being tracked in many networks. For example, optical networks look for sudden power loss that would follow a fiber break or a laser/amplifier outage. Moreover, given the widespread adoption of wavelength-division multiplexing (WDM) for transmission and routing, many networks monitor any deleterious wavelength drifts of the channels. Furthermore, it is helpful when existing networks measure the optical signal-to-noise ratio (OSNR) so that the optical amplifier–generated gain and noise can be controlled; note that many receivers are indeed limited by the accumulated amplified spontaneous emission (ASE) noise. All the above parameters can be monitored, to the first order, using fairly straightforward low-speed equipment, such as reflectometers, power meters, spectrometers, and filters. Some companies have been formed to provide this binary level of monitoring, yet none of these methods can provide a holistic system approach capable of simultaneously measuring impairments, locating their source, and providing accurate information that network controllers can use to compensate and re-route.

1.1.2 Challenges As data rates increase and network architectures become more complex, it becomes more difficult to predict and manage data impairments due to the fact that the degradations can change with time. In order to enable a robust and cost-effective “self-managed” operation, the optical network should be able to agilely monitor the physical state of the network and the quality of propagating data signals, automatically diagnose and repair the network, redirect the traffic, and dynamically allocate resources, as illustrated in Figure 1.3. Unfortunately, high-data-rate optical networks are quite susceptible to “noncatastrophic” signaldegrading problems due to channel propagation, in which there is sufficient signal power, but the data bits themselves are unrecoverable due to various fiber dispersive and nonlinear effects. This scenario is a result of the more complex nature of present and future transmission systems, especially when parallel wavelength channels are propagating simultaneously along the fiber. For example,

4

CHAPTER 1 Optical performance monitoring

Locate faults

Ubiquitous monitoring Detect attacks

Diagnose and assess Repair damage

Reroute and balance traffic

Malicious behavior

Telcos: Human error (~1/3 of outages)

FIGURE 1.3 Features of ubiquitous monitoring for robust and self-managed networks.

the in-band OSNR may vary among channels, and when coupled with repeated filtering and amplification, direct measurement can be very difficult. Additionally, impairments can interact with each other to degrade the data signal. The effort of research work is to develop the efficient OPM techniques and to overcome the following significant technical challenges.5–10 1. Monitor and isolate all degrading effects: Provide a real-time monitoring scheme that efficiently predicts all degrading effects on the high-speed data with a minimum of optical hardware. 2. Localize any degradation: Localize any degrading effects for network repair and traffic re-routing without requiring monitors at infinitely small distances. 3. Fast response time: The output of the optical monitor must be interpreted to provide meaningful information on the physical state of the network. The output is a simple number processed from the summary data, which can be used by the network controller to efficiently monitor the specific link or data channel. 4. Take action to dynamically allocate resources: Interpret the output of the network controller in order to change the channel wavelength, optical and/or electrical dispersion compensator, data modulation format, bit rate, gain of an amplifier, and the allocated channel bandwidth. 5. Prevent denial-of-service and eavesdropping: Keep track of any and all wavelengths and determine whether any nonlinear products are being generated. Protect the system in the secure condition by notching out any unwanted wavelengths.

1.2 PHYSICAL-LAYER MEASUREMENTS AND ROUTING DECISIONS IN TODAY’S OPTICAL NETWORKS The development of optical amplifiers and WDM has allowed systems to carry more traffic over longer distances, but this makes performance monitoring more challenging. At present, monitoring of performance in the physical layer primarily involves a combination of individual component alarms,

1.3 Signal parameters requiring monitoring and OPM techniques

5

aggregate power, and in some cases, optical channel monitoring (OCM).11 Component alarms include monitoring of parameters such as amplifier pump laser power or temperature controller limits. In this case, OPM is indirectly realized through the assumption that if all of the components are working correctly, then the signal must be good. This is a very powerful technique that largely comes for free, but unfortunately there are signal failures that violate this assumption. OCM includes measurements of channel power, presence, and wavelength. From an OPM perspective, channel monitoring as well as aggregate power monitoring are extensions of component alarms in that they indirectly measure signal quality. The term “OPM” is often applied to OCM devices with the additional capability of monitoring the OSNR, as well as other signal quality measures, such as chromatic dispersion (CD) and polarization mode dispersion (PMD), or jitter. Band-channel monitors that are sensitive to the per-channel signal-to-noise ratio (SNR) are referred to as “signal quality monitors.” Techniques such as Q-factor monitoring are perhaps the closest optical analog of the electronic performance monitor. As optical communication systems have become more sophisticated, the need has arisen for more sophisticated performance monitoring. Fortunately, many technologies that enable improved system reach and capacity can simultaneously be used to enhance system performance monitoring, which will be discussed in subsequent chapters of this book. Note that in these cases, the cost of the performance monitors can be fully justified on the basis of the system’s improved performance. Not all monitors are so easily justified, but service providers have a history of deploying additional equipment when necessary. There is no level of performance monitoring that is superfluous; there are only levels that are too expensive.8

1.3 SIGNAL PARAMETERS REQUIRING MONITORING AND OPM TECHNIQUES 1.3.1 Optical impairments

The specific parameters that a network might want to monitor range from the most basic to the highly complex. Some issues are already being monitored in many networks. For example, optical networks monitor the abrupt power loss that would follow a fiber break or a laser/amplifier outage. Moreover, given the widespread adoption of WDM for transmission and routing, many networks monitor any deleterious wavelength drifts of the channels. When the existing networks monitor the OSNR, simultaneously controlling the optical amplifier–generated gain and noise is helpful; note that many receivers are indeed limited by the accumulated ASE noise. High-data-rate optical networks are quite susceptible to various “noncatastrophic” signal degrading problems due to fiber transport, in which there is sufficient optical signal power but the data bits themselves are unrecoverable. When parallel wavelength channels are propagating simultaneously along the fiber, this scenario becomes more complex. Although the signal quality and bit error rate can be determined after a high-speed photodetector, such signal recovery cannot readily provide the location and cause of a problem; the signal’s phase information will be lost using common direct detection and the origin of the problem might remain a mystery. The root causes cannot be specified in advance because they include all of the various modes of component failures, but it might be helpful to further divide impairments into either component fault effects or optical transmission impairments.

6

CHAPTER 1 Optical performance monitoring

Component faults include individual or multiple component malfunctions, improperly installed or configured equipment, and damage or intrusion to the network. Impairments due to such faults are as diverse as the components and network designs deployed in the field and cannot be comprehensively cataloged. Kartalopoulos cataloged the more common component failure modes in DWDM systems.12 To appreciate the range of possibilities, we consider for example optical amplifier failure modes. Failing pump lasers will result in reduced power at particular points in the transmission path or within internal sections of the amplifier. Subsequent amplification can convert these low power levels into excess noise that will show up in the optical spectrum. Unstable pump lasers can also result in excess noise on the signal, but in this case it may not show up in the optical spectrum. If the power control loop on a pump laser fails and it runs high, then the signal levels will be high, which can enhance a host of nonlinear effects in transmission, such as self-phase modulation (SPM) or cross-phase modulation. Optical transmission impairments, which can also be referred to as fiber degradations, will cause various signal distortions and thus result in diverse information (e.g., eye diagrams, optical and electrical power variations, and polarization fluctuations) for monitoring. For example, PMD results from the birefringence nature of the fiber core, which is different from the CD effect. However, it is difficult to monitor and isolate PMD from CD since both of them cause pulse distortions and change the monitored information. They might result in a similar variation (e.g., RF clock-tone power fading), making the monitored information ambiguous. Moreover, each of the optical impairments gives rise to different degrading effects for different data modulation formats. Note that the future challenge is to develop a simple and unique monitoring technique that can monitor each of the optical impairments simultaneously for any input data formats. Essential fiber-based degrading effects due to transmission include the following: •

•

Chromatic dispersion. Because the refractive index of fiber is slightly dependent on frequency, different frequencies, and thus different channels, propagate at different speeds. Data modulation causes monochromatic laser light to spread out in frequency roughly on the same order as the data modulation. This causes each bit to spread in time. The unit of chromatic dispersion is (ps/nm)/km; thus, shorter time pulses, wider frequency spread due to data modulation, and longer fiber lengths will each contribute to temporal dispersion. The CD tolerance is proportional to the inverse of the square of the bit rate and thus 40-Gb/s signals are 16 times more susceptible to CD than 10-Gb/s signals. Due to the reduced tolerance to CD for the high-data-rate signals (i.e., >40 Gb/s), the accuracy of monitoring becomes increasingly crucial. CD also changes slightly with temperature, and there is some uncertainty in any deployed fiber plant. Polarization mode dispersion and polarization-dependent loss. The core of an optical fiber is not perfectly circular, and the resultant ellipse has two orthogonal axes. The “effective” refractive index of a waveguide, which determines the speed of light, depends on the glass material itself as well as the shape of the waveguide relative to the traveling wave. Therefore, light polarized along one axis travels at a different speed than does the light polarized along the orthogonal axis. PMD has emerged as a key issue for deploying high-speed optical communication systems. Differential group delay (DGD), which is known as the first-order PMD, causes walk-off in time between the two orthogonal polarization states. Moreover, the PMD effects on the data signal are stochastic, time varying, and a random process that acts on each WDM channel differently. A system can operate flawlessly for 364 days and then break down temporarily on the 365th day. Any adaptive PMD compensator must require accurate PMD monitoring in order to

1.3 Signal parameters requiring monitoring and OPM techniques

•

•

•

7

dynamically track the degrading effects due to time-varying conditions. Furthermore, the two polarization axes exhibit a different loss, the so-called polarization-dependent loss (PDL). PMD and PDL can interact to produce higher-order distortions.13 PMD and PDL are two of the key limitations in systems that operate at greater than 10 Gb/s, especially in high-speed polarization multiplexed systems. OSNR degradation. One of the major parameters to monitor is OSNR degradation since it provides direct and important information about the channel quality inside a dynamic network. Typically, OSNR can be measured by the linear interpolation between the channels using a tunable narrowband filter and an optical spectrum analyzer (OSA). However, in reconfigurable optical networks, WDM channels are added, dropped, or cross-connected in the optical layer. One channel may travel through different paths, different numbers of optical amplifiers, and cascaded filtering effects, resulting in a nonuniform noise spectrum compared with adjacent channels. The accumulated noise level may be quite different from channel to channel. As a result, the accurate OSNR of each channel must be monitored in-band. Fiber nonlinearities. The refractive index of fiber is slightly dependent on optical power, with higher intensities experiencing a higher index. Moreover, the electric fields of several channels will mix with each other and produce sum and difference frequency products. Nonlinearities can be controlled by carefully introducing and balancing chromatic dispersion, perhaps with fixed and/or tunable compensation. Some typical nonlinear effects, which tend to degrade significantly the signal integrity, include SPM (the intensity profile of an optical pulse on a single channel causes an index-of-refraction profile and, thus, the higher intensity center of a pulse travels slower than the lower-intensity pulse wings); XPM (when considering many WDM channels, power from channel 2 through channel N can distort the index profile that is experienced by channel 1, which can be translated into a lightwave speed distortion); and FWM (the optical intensity propagating through the fiber is the square of the electric field and when one squares the sum of different channels’ fields, products emerge at various sum and difference frequencies). If a WDM channel exists at one of the FWM beat-term frequencies, then the beat term will interfere with this other channel and potentially distort the data. Other nonlinear effects include stimulated Rayleigh scattering (SRS), stimulated Brillouin scattering (SBS), and so on. Frequency chirp. Frequency variation across the time period of a bit is defined as the frequency chirp, which results in a phase-modulation, parasitic-causing frequency broadening of the data. For high-speed and long-haul transmission systems, chirp interacts with CD to further distort the data. Besides directly modulated lasers and optical modulators, even passive components such as an optical filter may also induce chirp. Each component normally results in a different type of chirp, making it very challenging to be monitored, distinguished, and compensated.

Figure 1.4 gives an overview of various optical impairments within the network, which also includes some other notable transmission impairments, such as: • • • • • •

Amplifier noise Amplifier distortion and transients Timing jitter Interference effects (MPI) Optical filter distortion Linear crosstalk

8

CHAPTER 1 Optical performance monitoring

• Power loss • Wavelength drift • Chirp

• Loss, CD, PMD • SPM, XPM, FWM • MPI Transmitter

Fiber

Amplifier

Receiver • Noise • Bandwidth • Sensitivity

Network element

• ASE noise • Gain fluctuation • Power saturation

• Wavelength alignment • Power condition • Linear crosstalk

FIGURE 1.4 Overview of various optical impairments within the network.

1.3.2 OPM techniques Advances in measurement science and metrology enable a more accurate and flexible characterization of optical signals. Increasing the transmission capacity can be attained by utilizing the amplitude, phase, and polarization of optical waves to send information at ever-increasing data rates. At the higher transmission rates, even a weak impairment in the link could severely distort some of the optical wave properties that impact information capacity. Comprehensive OPM will be required to manage the communication network with enhanced security and full utilization. A comprehensive approach also means developing a set of measurements and the corresponding methods and tools through metrology designed to meet the demanding requirements of highly utilized networks. For example, in data-intensive networks, we require a technology to monitor transmission degrading effects simultaneously. On the other hand, in local clients, cost-effective and repeatable measurements are required without the need for a short response time. OPM can be performed by measuring changes to the data and determining “real-time” changes resulting from various impairments, such that a change in a particular effect will change a measured parameter. This can employ (1) electrical postprocessing techniques in the cases of either non-coherent detection or coherent detection,14–16 which are referred to as time-domain OPM; or (2) optical techniques to monitor changes in a RF tone power or in the spectral channel power distribution,17 which are referred to as frequency-domain OPM. •

Time-domain OPM. Besides the specific case of using coherent detection for OPM,14–16 several techniques have been proposed for OPM using offline digital signal processing of received electrical data signals.18–31 Five of these methods18–22 utilize amplitude histograms, power distributions, or asynchronous sampling to monitor the optical performance in various forms, such as bit error rate (BER); four23–26 employ delay-tap plots to distinguish among impairments; three27– 29 use pattern recognition techniques to identify multiple impairments; and the rest30,31 use parameters derived from eye diagrams, histograms, and delay-tap plots for the same purpose.

1.4 Laudable OPM-enabled functionalities in next-generation optical networks

•

9

In addition, linear optical sampling (LOS) has been applied to recover the phase and amplitude of a high-speed advanced data modulation format simultaneously.32–38 LOS requires a low-speed pulsed-laser limited only by the pulsewidth, and thus LOS is very applicable to the monitoring of high-speed advanced modulation formats. These techniques will be reviewed in great detail in Chapters 2–11. Frequency-domain OPM. The frequency-domain measurement approaches are proposed for OPM, which are cost-effective and simply implemented. Although in some cases the frequency domain approaches tend to monitor a single impairment at one time, several frequency-domain techniques can potentially form an OPM toolbox, and in particular meet the technical requirements imposed in achieving a smart assignment and management of network resources.39–45 Similar frequency-domain techniques have also been applied to monitor the time misalignments in the transmitter with advance phase-modulated formats.46,47 Repeatability and stability enabled by the frequency domain measurement approach are highly desirable in future “smart” networks. Detailed discussions about the frequency-domain OPM techniques will be given in subsequent chapters.

Driven by the need for spectral efficiency, the high-speed, phase-modulated data channel is becoming increasingly important for 100-Gb/s Ethernet and higher-capacity optical networks.48 In order to further advance the metrology and accuracy of measurements for robust and reconfigurable high-capacity optical networks, developing more cost-effective, efficient, and stable OPM techniques becomes increasingly crucial.2,49,50 Chapters 12–15 will review the use of OPM in various transmission systems, including high-speed, optical-phase modulation formats, synchronous digital hierarchy (SDH), and synchronous optical network (SONET).

1.4 LAUDABLE OPM-ENABLED FUNCTIONALITIES IN NEXT-GENERATION OPTICAL NETWORKS OPM plays an important role for maintenance and management of high-speed, intelligent, and reconfigurable optical networks. In this section, we give a brief introduction to the network functionalities that can be enabled by using OPM.

1.4.1 Robust and stable operation Multiple impairments on optical signals are time varying due to a changing environment, drift of components, and rapid reconfiguration of network paths. Moreover, the fiber impairments depend on complex interactions of linear and nonlinear fiber effects, which in turn are a function of the signal power, data rate, and data modulation format. These properties imply that the OPM should provide real-time information about the quality of these transmitted signals and also isolate the specific cause and location of the problem. To enable a robust and stable operation, the monitoring information corresponding to the accumulated impairment due to each specific degrading effect should probably be deployed ubiquitously around the network. Furthermore, it can be quite advantageous to determine when a data signal is beginning to degrade, so that the network can take action to correct the problem (i.e., change a laser wavelength, tune a compensator) or to route the traffic around the degraded area.51

10

CHAPTER 1 Optical performance monitoring

1.4.2 Accommodate transparency High-speed (i.e., >40 Gb/s) differential phase-shift-keying (DPSK) data format has generated much excitement due to its high receiver sensitivity and tolerance to nonlinearities as compared to the conventional on-off-keying (OOK) format.48 For high-capacity transmission systems, spectrally efficient data modulation formats are highly desirable due to their ability to pack more information into narrow wavelength channels, enhance the tolerance to fiber dispersion, and remove the necessity for high-speed components. The use of multilevel PSK formats, such as differential quadrature phaseshift-keying (DQPSK), offers high spectral efficiency and tolerance to fiber-dispersive effects and leverages the limitations on the available operation bandwidth of the RF/optical modules. Furthermore, fitting more bits/s/Hz on any channel will require the utilization of polarization multiplexing two independent data streams on the two orthogonal polarization states to reduce the received power, required hardware, and network management cost.52 As a consequence, the high-capacity transmission systems tend to polarization multiplex two DPSK or DQPSK channels for better receiver sensitivity and robustness to dispersion and nonlinearities.53 It would be highly desirable for the performance monitor to be independent of the input data format and to accommodate a convergence of different traffic types, including amplitude modulation, phase modulation, and polarization multiplexing for binary and multilevel coding (i.e., OOK, PSK, and QPSK). Particularly for an optical high-capacity network, it seems inefficient to build a separate optical network/monitor to accommodate each one. Therefore, the performance monitor should be able to effectively handle a wide variety of high-rate traffic. Since each application might have a different set of optimal requirements, one can envision that one network might be required to transmit different modulation formats, a wide range in data rates, commercial and military traffic, and variable quality of service (QoS), as shown in Figure 1.5. Therefore, OPM should accommodate transparency.14,51

1.4.3 Impairment-aware routing In the existing networks, traffic routing is typically based either on a shortest-path calculation or on paths that satisfy certain network QoS constraints (e.g., data rate, delay, jitter, or packet loss). However,

Different modulation formats

Variable QoS

Variable bit rate

Future heterogeneous network

Multiple wavelength ranges

Subcarrier multiplexing (D + A)? Circuit + packet switching?

FIGURE 1.5 Future heterogeneous networks should accommodate various types of traffic and use optimal channel characteristics for each application/user. The required hardware should be reconfigurable and transparent.

1.4 Laudable OPM-enabled functionalities in next-generation optical networks

11

it might be quite advantageous for the network controller to also take into account the variable physical state of the network, especially given that transparency domains are growing, data rates are increasing, and channel spacing is denser. It might be valuable for the network controller and routing algorithms to consider the optical-layer impairments that degrade the BER of the signal.54 Performance monitoring provides valuable information such that routing tables themselves can dynamically reflect the state of the physical links to the network controller. For example, each link has a set of parameters (a, b, c) and each node has a set of parameters (a, b), as illustrated in Figure 1.6. The ranges of these parameters will be determined. Impairment-aware routing enables routing tables to be modified based on the physical changes of the links so that the network controller can agilely control and manage the heterogeneous networks. The network must interpret the “cost function” for the routing tables and determine ranges of these parameters for inclusion into the network model. Therefore, OPM might be required to provide valuable information such that routing tables themselves can dynamically reflect the state of the physical network links, the fidelity of the channels, and the addition/deletion of nodes. This “impairment-aware” routing would enable routing tables to be modified based on physical changes in the links, instead of routing decisions being based simply on of the fewest hops and shortest links between source and destination. This results in severe challenges in terms of algorithm complexity and interpreting monitor output for any protocol. Network links today are generally assigned either a “1” or “0” depending on whether the link is considered working or malfunctioning. However, a link might function well for some traffic but not sufficiently well for all traffic. Therefore, we emphasize that the network could assign “weights” to each link, such that a partially functioning link can still be used for some traffic, while re-routing other traffic. This is analogous to closing down only two lanes of a four-lane highway, which is significantly more preferable for traffic management than shutting down the entire highway. The backbone network will likely be a transparent WDM system. The key benefits of all-optical transport networks are the transparency to bit rate, protocol, and modulation format of all the various wavelength channels propagating in the system. However, key challenges exist when determining an

a. Fiber length b. Signal degradation c. Amplification and transients , c i>
α. Component nonidealities β. Signal degradation

FIGURE 1.6 Multivariable routing.

12

CHAPTER 1 Optical performance monitoring

optimum path through the network, since an optical wavelength might accumulate different physical impairments as it is switched through the network. These non idealities will be imposed by both the transmission links and the optical switching nodes. The performance penalty due to the different optical impairments will depend on many factors, including the bit rate, modulation format, and electronic processing within the receiver.55–58 Without physical impairment awareness, a network-layer routing and wavelength assignment (RWA) algorithm might rapidly provision a light path that cannot meet the signal quality requirement. Therefore, the control plane of an optical transparent network should incorporate the characteristics of the physical layer in establishing a light path for a new connection. To greatly simplify the network management, some margin in BER fluctuation could be considered to mitigate the effects of traffic distribution on a light path’s quality. An impairment-aware RWA algorithm requires some form of OPM.

1.4.4 Secure links As optical communication systems have become more sophisticated, they have also become vulnerable to denial of service and eavesdropping in new and unique ways. Measurement and monitoring could provide enhanced security to identify suspicious activity and to initiate preventive measures against denial of service and eavesdropping resulting from human error or malicious attack.59,60 An effective scheme is based on an understanding of the physics behind data transmission in an optical WDM network. Being specific to erbium-doped fiber amplifiers (EDFAs), a pump light is used to excite the erbium atoms to higher orbits, and the input signal stimulates them to release excess energy as photons in phase and at the same wavelength. EDFAs boost wavelengths in the 1550-nm range, and the pump light is typically 1480 nm or 980 nm. As a simple example, as shown in Figure 1.7, when an unwelcome high-power wavelength is added, severely degrading effects could be caused as a measure of achieving a denial-of-service attack. Furthermore, an intense unwanted

Denial of service l1

l2

lp

l3

Eavesdropping ldata(t)

l4

[2lp - ldata](t) λ

λ Sleeper wavelength

lp

• Overwhelms EDFAs • Incoherent/coherent crosstalk • Fiber nonlinearities (a)

lp

lp

Sleeper wavelength

[2lp - ldata](t) lp

lp

λ

• Eavesdrop on channel via undetected FWM products (b)

FIGURE 1.7 (a) Denial of service, and (b) eavesdropping of an intrusive wavelength using the nonlinear effects in the fiber link.

1.5 Smart network operation and security

13

wavelength can saturate and reduce the gain of an optical amplifier for all existing channels. Additionally, simple crosstalk can also destroy the data on the channels. If this wavelength disappears, the network returns to normal without any lasting trace of the “culprit.” Moreover, this new wavelength could generate optical nonlinear FWM products that would contain the data information of the existing channels, such that only mixing products that contain the original data bits need be recovered by an eavesdropper. In general, the network should be able to find the spatial and spectral locations of all wanted and unwanted wavelength channels and detect the generation of any nonlinear mixing products.

1.4.5 Optical supervisory channel In some optical networks, a separate wavelength may be allocated to carry management and control information. Unlike the service-bearing data channels that might travel transparently through a node, an optical supervisory channel (OSC) might be electro-optically terminated, processed, and regenerated at each adjacent node.61–63 Since the OSC wavelength may not fall within the amplifier gain bandwidth, OSC signals might be demultiplexed and detected without propagation through an optical amplifier. After processing of the incoming management data, the outgoing OSC signal is generated and added to the appropriate output port. Some bit-oriented data, such as the alarm indication signal and remote defect indication signal, are specific to a particular service-bearing data wavelength and are forwarded to the correct output port after processing. Some commercial optical systems include OSC, such as CIENA’s CN 4200 OSC module.64 Built around a dedicated 1510-nm laser and receiver, the OSC provides an out-of-band, full-duplex communications channel for remote node management, monitoring, and control. The OSC optically segregates network management and control from user data, so even if the OSC is lost, data forwarding continues uninterrupted. Equipped with a 1510-nm MUX/DEMUX filter, the OSC travels the same fiber as the dense WDM stream and always terminates on neighboring nodes. Because it has a dedicated wavelength, the OSC data are entirely independent of the user data. The OSC carries a 100-Mb/s, Ethernet-over-IP management channel used for internodal management and control. Additionally, the module’s built-in controller CPU allows it to assume the added role of system controller. The OSC can help determine major changes in the system, but it tends not to accurately determine subtle yet destructive fiber-based impairments on the data channels (i.e., CD PMD, nonlinearities, and OSNR).

1.5 SMART NETWORK OPERATION AND SECURITY The problem of impairments is further complicated by the wide range for modulation formats that are being considered for current and future networks. The forthcoming IEEE 100-Gb/s Ethernet Standard, 802.3ba, has provisions for many potential formats, both noncoherent and coherent. PSK can provide higher receiver sensitivity while multilevel PSK formats (xPSK) and multilevel quadrature-amplitudemodulation (xQAM) provide multiple bits per symbol, allowing lower symbol rate operation. Polarization multiplexing allows two channels to share the same wavelength by propagating on orthogonal polarization axis. Any potential OPM method must be capable of handling the wide array of data formats and data rates that networks may employ. It is also worth noting that future networks may

14

CHAPTER 1 Optical performance monitoring

use a combination of data rates and formats while Robert Metcalfe (Plenary speaker of OFC 2009) has estimated that “we may start seeing the first commercial use of Terabit Ethernets by 2015.” This ability to provide accurate and repeatable network performance monitoring is the foundation for future “smart” operation. Optical networks suffer from the aforementioned impairments caused by the variation of components and the physical medium, such as fiber, copper wires, and even air. These impairments are also time varying due to changes in nature, network use, temperature, vibration, and component age. Without the ability to measure these changes at the rate in which they occur, the impairment rate, today’s networks are forced to operate far from the limits imposed by these impairments. This has created an overbuilding of networks and an inefficient use of network resources. By advancing the OPM techniques of optical networks, the time-varying nature of the network can be tracked and problem components, links, and nodes can be identified and located. This ability to assess the state of the network serves two extremely important functions.

1.5.1 Smart network operation The real-time monitoring of network elements, paths, and channels will enable: (1) continuous measurement of network status and quality; (2) automatic network diagnostics and repair; (3) allocation of network resources, including signal wavelength and power, tunable compensation and equalization, data coding, path determination, and channel bandwidth; (4) and changing of routing tables and redirection of traffic based on physical-layer conditions. These “smart” network behaviors will allow for network resources to be utilized to their full potential. This network metrology will deter the overbuilding of networks at high additional cost. High-data-rate WDM communication systems are highly susceptible to deleterious channel-based impairments. Even relatively straightforward properties, such as the OSNR, can be quite challenging to maintain in a wavelength-routed network that contains many spectral filters. Hundreds of optical channels with various data rates and modulation formats are added, dropped, and routed to different network links through the optical network elements (ONEs). As a consequence, each channel will experience different degradations in heterogeneous optical networks. These impairments are not static and can vary over time due to changing network paths, temperature, and routine maintenance. Moreover, today’s optical networks function in a fairly static fashion. Particularly as the bit rates increase and the modulation formats become more complex, it becomes quite challenging to determine data degradation mechanisms and provide stable and accurate physical states to the network controller for data dropping, correction, compensation, and routing. In order to enhance the functionality of network control and management, OPM becomes an increasingly practical tool to overcome such challenges and to agilely reflect the physical state of the network and the quality of all data channels for “smart” network operation, as illustrated in Figure 1.8.

1.5.2 Security Over the past two decades, it is has become clear that fiber optics plays a critical role as a key part of any high-capacity communications system, and this is true for friend and foe alike. Fiber optics is unrivaled in its ability to transport high-speed data, and it is relatively unaffected by electromagnetic interference. Optical communications have become ubiquitous, meaning that any vulnerability to

1.5 Smart network operation and security

15

B

A

Fib

er

Network element

link

C

High-speed channels

E “Adaptive” resources Diagnose and repair BW allocation Gain / loss Dispersion Compensation l routing Look-up tables

Impairments can interact

D

CD PMD + PDL OSNR NL Chirp Equipment failure • … • • • • • •

FIGURE 1.8 Identification and localization of network impairments allow network resources to be adapted for compensation, data re-routing, and resource reallocation.

data exfiltration and infiltration would be devastating. It is estimated that “25% of network outages are caused by human errors.”7 The ability to detect and prevent accidental and malicious network outages is vital for efficient, stable, and secure network operation. For smart and secure operation, it is critical to: 1. Monitor and isolate all impairments: Efficient real-time prediction of all degrading effects on each of the high-speed channels with a minimum of optical and electrical hardware and signal processing algorithms. 2. Diagnose and localize any degradation: Accurately identify and localize any degrading effect with a minimum number of required monitors. 3. Be transparent to changing network data rates and modulation formats: As networks are continually being updated, the type of data on any given channel can be different from all neighboring channels while continuously fluctuating to meet network demands. 4. Localize human errors and intent to specific points in the network: It is critical to measure the spatial and spectral locations of all wanted and unwanted wavelength channels and to detect the generation of any nonlinear mixing products. Continuous performance monitoring will provide the information needed by network controllers to provide the highest-level quality and efficiency. Networks can be operated much closer to their operational limits, no longer requiring expensive and inefficient overbuilding. Simultaneously, network performance can be continuously updated where individual channels can be optimized to provide the optimal quality of service while making use of the network in the most efficient fashion, saving power and energy.

16

CHAPTER 1 Optical performance monitoring

1.6 SUMMARY Today’s network infrastructure is required to adapt for anticipated future growth and be capable of detecting impairments and identify threats, intentional or otherwise. This deficiency in the measurement science and metrology of networks leads to the overbuilding and inefficient use of current networks. The ability to monitor both the data quality and the network paths’ physical state in a realtime fashion has traditionally proven difficult due to the large number of impairments, variety of data formats, variety of data rates, and the continuous growth and change of the network structure. Specific technical challenges include: 1. Localize any security breaches to specific points in the network. 2. Isolate the specific cause and location of a degrading effect, rather than simply sound an alarm. 3. The monitor should accommodate various advanced modulation formats. It would be desirable to enable “smart” and secure network operation, such that the network can provide accurate and repeatable network performance monitoring. The desirable functionalities of OPM highly depend on the system’s intelligence, complexity, and bit rates. The right balance should be kept among monitoring coverage, sensitivity, and cost. These key features of the desired OPM provide the following advantages: (1) simplicity in implementation; (2) cost effectiveness; (3) fast response time and minimal latency; (4) applicability to other formats without any modification on the monitor; (5) fine resolution and wide monitoring range; (6) ability to isolate each impairment from others without any ambiguity; and (7) the potential to be integrated onto photonic chips.

ACKNOWLEDGMENTS The authors thank the support of the DARPA CORONET PARAGON Program, the U.S. Department of Commerce, and Cisco Systems. We also thank Dr. Anjali Agarwal, Dr. Janet Jackel, Dr. Ronald A. Skoog from Telcordia Technologies, Dr. Tasshi Dennis, Dr. Paul Hale, Dr. Jeffrey A. Jargon from National Institute of Standards and Technology, and Dr. Loukas Paraschis from Cisco Systems for valuable discussions.

REFERENCES 1. Saleh AAM, Simmons JM. Evolution toward the next-generation core optical network. IEEE/OSA J Lightwave Technol 2006;24(9):3303–21. 2. Chen LK, Chan CCK, Lu GW, Ku YC, Ho ST, Lin C. Optical performance monitoring and network diagnosis in reconfigurable optical network. Proc Soc Photo Opt Instrum Eng 2007;6784:67841I. 3. Herman W. 40G to 100G–What’s All the Fuss about Optical Gigabits? Nortel; March 2008; http://www 2.nortel.com/go/news_detail.jsp?cat_id¼-9742&oid¼100238328. 4. Private communication. 5. Meflah L, Thomsen B, Mitchell J, Bayvel P, Lehmann G, Santoni S, et al. Advanced optical performance monitoring for dynamically reconfigurable networks. In: 10th European Conference on Networks and Optical Communications (NOC); 2005, p. 554–61. 6. Pinart C, Amrani A, Junyent G. Monitoring service “Health” in intelligent, transparent optical networks. In: Optical networks and technologies. Boston: Springer; 2005. p. 179–86.

References

17

7. Shapiro RJ. The Internet’s capacity to handle fast-rising demand for bandwidth. US Industry Association (USIIA); 2007. 8. Woodward SL. Monitors to ensure the performance of photonic networks. In: Conference on optical fiber communications (OFC) 0 07, paper OMM1, Anaheim, CA; March 2007. 9. Willner AE, Pan Z, Yu C. Optical performance monitoring. In: Kaminow IP, Li T, Willner AE, editors. Optical fiber telecommunications VB. San Diego: Elsevier and Academic Press; 2008 [Chapter 7]. 10. Chung YC. Optical performance monitoring techniques: current status and future challenges. In: European conference on optical communication (ECOC) 0 08, paper We.1.D.1. Brussels; September 2008. 11. Kilper DC, Bach R, Blumenthal DJ, Einstein D, Landolsi T, Ostar L, et al. Optical performance monitoring. IEEE/OSA J Lightwave Technol 2004;22(1):294–304. 12. Kartalopoulos SV. Fault detectability in DWDM. New York: IEEE; 2001. 13. Yan LS, Yu Q, Xie Y, Willner AE. Experimental demonstration of the system performance degradation due to the combined effect of polarization-dependent loss with polarization-mode dispersion. IEEE Photon Technol Lett 2002;14(2):224–6. 14. Shieh W, Tucker RS, Chen W, Yi X, Pendock G. Optical performance monitoring in coherent optical OFDM systems. Opt Express 2007;15(2):350–6. 15. Sun H, Wu K, Roberts K. Real-time measurements of a 40 Gb/s coherent system. Opt Express 2008;16 (2):873–9. 16. Hauske FN, Geyer JC, Kuschnerov M, Piyawanno K, Duthel T, Fludger CRS, et al. Optical performance monitoring from FIR filter coefficients in coherent receivers. In: Conference on optical fiber communications (OFC) 0 08, paper OThW2. San Diego, CA; February 2008. 17. Luo T, Pan Z, Motaghian Nezam SMR, Yan LS, Sahin A, Willner AE. Chromatic-dispersion-insensitive PMD monitoring using optical off-center bandpass filtering. In: Conference on optical fiber communications (OFC) 0 03, paper ThY3. Atlanta, GA; March 2003. 18. Shake I, Takara H, Kawanishi S, Yamabayashi Y. Optical signal quality monitoring method based on optical sampling. Electron Lett 1998;34(22):2152–4. 19. Hanik N, Gladisch A, Caspar C, Strebel B. Application of amplitude histograms to monitor performance of optical channels. Electron Lett 1999;35(5):403–4. 20. Ohteru S, Takachio N. Optical signal quality monitor using direct Q-factor measurement. IEEE Photon Technol Lett 1999;11(10):1307–9. 21. Andre´ P, Teixeira A, Lima M, Monteiro P, Luis R, Fonseca D. Asynchronous sampled amplitude histogram models for optical performance monitoring in high-speed networks. J Opt Netw 2004;3(8):636–42. 22. Shake I, Takara H, Kawanishi S. Simple measurement of eye diagram and BER using high-speed asynchronous sampling. IEEE/OSA J Lightwave Technol 2004;22(5):1296–302. 23. Dods SD, Anderson TB. Optical performance monitoring technique using delay tap asynchronous waveform sampling. In: Conference on optical fiber communications (OFC) 0 06, paper OThP5. Anaheim, CA; March 2006. 24. Dods SD, Anderson TB, Clarke K, Bakaul M, Kowalczyk A. Asynchronous sampling for optical performance monitoring. In: Conference on optical fiber communications (OFC) 0 07, paper OMM5. Anaheim, CA; March 2007. 25. Kozicki B, Maruta A, Kitayama K. Experimental demonstration of optical performance monitoring for RZ-DPSK signals using delay-tap sampling method. Opt Express 2008;16(6):3566–76. 26. Kozicki B, Maruta A, Kitayama K. Transparent performance monitoring of RZ-DQPSK systems employing delay-tap sampling. J Opt Netw 2007;6(11):1257–69. 27. Skoog RA, Banwell TC, Gannett JW, Habiby SF, Pang M, Rauch ME, et al. Automatic identification of impairments using support vector machine pattern classification on eye diagrams. IEEE Photon Technol Lett 2006;18(22):2398–400.

18

CHAPTER 1 Optical performance monitoring

28. Anderson TB, Dods SD, Clarke K, Bedo J, Kowalczyk A. Multi-impairment monitoring for photonic networks. In: European conference on optical communication (ECOC) 0 07, paper 3.5.1. Berlin; September 2007. 29. Anderson TB, Clarke K, Beaman D, Ferra H, Birk M, Zhang G, et al. Experimental demonstration of multiimpairment monitoring on a commercial 10 Gbit/s NRZ WDM channel. In: Conference on optical fiber communications (OFC) 0 09, paper OThH7. San Diego; March 2009. 30. Wu X, Jargon J, Skoog RA, Paraschis L, Willner AE. Applications of artificial neural networks in optical performance monitoring. IEEE/OSA J Lightwave Technol 2009;27(16):3580–9. 31. Jargon JA, Wu X, Willner AE. Optical performance monitoring using artificial neural networks trained with parameters derived from delay-tap asynchronous sampling. In: Conference on optical fiber communications (OFC) 0 09, paper OThH1. San Diego, CA; March 2009. 32. Dorrer C, Doerr CR, Kang I, Ryf R, Leuthold J, Winzer PJ. Measurement of eye diagrams and constellation diagrams of optical sources using linear optics and waveguide technology. IEEE/OSA J Lightwave Technol 2005;23(1):178–86. 33. Dorrer C, Kilper DC, Stuart HR, Raybon G, Raymer MG. Linear optical sampling. IEEE Photon Technol Lett 2003;15(12):1746–8. 34. Dorrer C, Leuthold J, Doerr CR. Direct measurement of constellation diagrams of optical sources. In: Conference on optical fiber communications (OFC) 0 04, paper PDP33. Los Angeles, CA; February 2004. 35. Dorrer C. Monitoring of optical signals from constellation diagrams measured with linear optical sampling. IEEE/OSA J Lightwave Technol 2006;20(1):313–21. 36. Dennis T, Williams PA, Coddington I, Newbury NR. Word-synchronous linear optical sampling of 40 Gb/s QPSK signals. In: Conference on optical fiber communications (OFC) 0 09, paper OThH3. San Diego, CA; March 2009. 37. Williams PA, Dennis T, Coddington I, Swann WC, Newbury NR. Vector signal characterization of highspeed optical components by use of linear optical sampling with milliradian resolution. IEEE Photon Technol Lett 2008;20(23):2007–9. 38. Williams PA, Dennis T, Coddington I, Newbury NR. Polarization-sensitive linear optical sampling for characterization of NRZ polarization-multiplexed QPSK. In: Conference on optical fiber communications (OFC) 0 09, paper OThH2. San Diego, CA; March 2009. 39. Motaghian Nezam SMR, McGeehan JE, Willner AE. Theoretical and experimental analysis of the dependence of a signal’s degree of polarization on the optical data spectrum. IEEE/OSA J Lightwave Technol 2004;22(3):763–72. 40. Lee JH, Choi HY, Shin SK, Chung YC. A review of the polarization-nulling technique for monitoring optical-signal-to-noise ratio in dynamic WDM networks. IEEE/OSA J Lightwave Technol 2006;24 (11):4162–71. 41. Yu Q, Pan Z, Yan L-S, Willner AE. Chromatic dispersion monitoring technique using sideband optical filtering and clock phase-shift detection. IEEE/OSA J Lightwave Technol 2002;20(12):2267–71. 42. Ji HC, Park KJ, Lee JH, Chung HS, Son ES, Han KH, et al. Optical performance monitoring techniques based on pilot tones for WDM network applications. J Opt Netw 2004;3(7):510–33. 43. Pan Z, Yu Q, Xie Y, Havstad SA, Willner AE, Starodubov DS, et al. Real-time group-velocity dispersion monitoring and automated compensation without modifications of the transmitter. Opt Commun 2004;230:145–9. 44. Luo T, Yu C, Pan Z, Wang Y, Arieli Y, Willner AE. Dispersive effects monitoring for RZ data by adding a frequency-shifted carrier along the orthogonal polarization state. IEEE/OSA J Lightwave Technol 2005;23 (10):3295–301. 45. Yang J-Y, Zhang L, Wu X, Yilmaz O, Zhang B, Willner AE. All-optical chromatic dispersion monitoring for phase-modulated signals utilizing cross-phase modulation in a highly-nonlinear fiber. IEEE Photon Technol Lett 2008;20(19):1642–4.

References

19

46. Tsai KT, Lu G-W, Chen L-K, Way WI. Alignment monitoring technique for pulse carver and data modulator in RZ-DPSK systems using an optical frequency discriminator. IEEE Photon Technol Lett 2006;18 (10):1119–21. 47. Wu X, Christen L, Zhang B, Peng W-R, Yang J-Y, Zhang L, et al. Synchronization monitoring of I/Q data and pulse carving misalignment for a parallel-type RZ-DQPSK transmitter by measuring RF clock tone/low frequency power. IEEE Photon Technol Lett 2008;20(24):2138–40. 48. Winzer PJ, Essiambre R-J. Advanced optical modulation formats. Proc IEEE 2006;94(5):952–85. 49. Grupp W. Monitoring requirements for optical transparent networks. In: Conference on optical fiber communications (OFC) 0 06, paper OWN5. Anaheim, CA; March 2006. 50. Jiang X, Feng KM, Cardakli M, Cai JX, Willner AE, Grubsky V, et al. Control monitoring of routing bits and data packets in WDM networks using wavelength-to-time mapping. IEEE Photon Technol Lett 1999;11(9):1186–8. 51. Willner AE. The optical network of the future: can optical performance monitoring enable automated, intelligent and robust systems? Opt Photon News 2006;17:30–5. 52. van den Borne D, Jansen SL, Gottwald E, Krummrich PM, Khoe GD, de Waardt H. 1.6-b/s/Hz spectrally efficient transmission over 1700 km of SSMF using 40  85.6-Gb/s POLMUX-RZ-DQPSK. IEEE/OSA J Lightwave Technol 2007;25(1):222–32. 53. Chandrasekhar S, Liu X. Experimental investigation of system impairments in polarization multiplexed 107-Gb/s RZ-DQPSK. In: Conference on optical fiber communications (OFC) 0 08, paper OThU7. San Diego, CA; February 2008. 54. Strand J, Chiu A, Tkach R. Issues for routing in the optical layer. IEEE Commun Mag 2001;39:81–7. 55. Guild K. Impairment-aware routing for OBS and OPS networks. In: Proc. IEEE international conference on transparent optical networks (ICTON 2006), vol. 3. Nottingham, UK; 2006. p. 61. 56. Huang Y, Heritage JP, Mukherjee B. Connection provisioning with transmission impairment consideration in optical WDM networks with high-speed channels. IEEE/OSA J Lightwave Technol 2005;23(3):982–93. 57. Carpenter T, Shallcross D, Gannett J, Jackel J, Von Lehmen A. Maximizing the transparency advantage in optical networks. In: Conference on optical fiber communications (OFC) 0 03, paper FA2. Atlanta, GA; March 2003. 58. Martinez R, Pinart C, Cugini F, Andriolli N, Vakarenghi L, Castoldi P, et al. Challenges and requirements for introducing impairment-awareness into the management and control planes of ASON/GMPLS WDM networks. IEEE Commun Mag 2006;44(12):76–85. 59. Rejeb R, Pavlosoglou I, Leeson MS, Green RJ. Management issues in transparent optical networks. In: Proc. 6th international conference on transparent optical networks (ICTON), vol. 1. Wroclaw, Poland; July 2004. p. 248–54. 60. Shaneman K, Gray S. Optical network security: technical analysis of fiber tapping mechanisms and methods for detection and prevention. In: Proc MILCOM, vol. 2. Monterey, CA;. p. 711–6. 61. Maeda MW. Management and control of transparent optical networks. IEEE J Sel Areas Commun 1998; 16(7):1018–23. 62. Lee HJ, Bang JH, Ko J-S. Dynamic gain control of booster amplifier (EDFA) using an optical supervisory channel power adjustment in WDM transmission systems. In: Proc. joint 4th IEEEICATM. Seoul, Korea; April 2001. p. 278–81. 63. Shin J, Kwon Y, Ko J. Optical supervisory channel subsystem for 1.6T WDM transmission system. In: Proc ICALT 2004, vol. 1. Phoenix Park, Korea. p. 402–4. 64. Ciena. Optical Supervisory Channel Module. Linthicum, MD; http://www.ciena.com/files/Optical_Supervisory_Channel_Module_DS.pdf.

This page intentionally left blank

CHAPTER

Optical signal-to-noise ratio monitoring *

2 Jun Haeng Lee*, Yun C. Chung

{

Samsung Advanced Institute of Technology, Samsung Electronics, Yongin-si Gyeonggi-do, Korea Department of Electrical Engineering and Computer Science, Korea Advanced Institute of Science and Technology, Daejeon, Korea

{

2.1 INTRODUCTION The optical signal-to-noise ratio (OSNR) is one of the most useful parameters for estimating the quality of a signal directly in the optical layer. This is mainly because the OSNR can be correlated to the end-terminal bit error rate (BER) of the transmitted optical signal through an optically amplified link.1 In addition, since the OSNR is transparent to both the bit rate and the modulation format of the optical signal, it is ideally suited for use in dynamically reconfigurable optical networks. In fact, the OSNR can be used in these networks for link setup and optimization, root-cause analysis of system problems, the setup of an early signal degradation alarm, resilience mechanism activation, service-level agreement (SLA) verification, and so on. Thus, for the efficient operation and maintenance of such dynamic optical networks, it is highly desirable to have the capability to monitor the OSNR of each wavelength-division multiplexed (WDM) channel. Previously, the OSNR has been measured by using the linear interpolation technique, in which the power of amplified spontaneous emission (ASE) noise is measured at wavelengths between the WDM channels and then interpolated into the signal’s wavelength. This technique could accurately estimate the OSNR in a conventional point-to-point transmission system (where the noise spectrum is more or less uniform). However, in a modern dynamically reconfigurable network, WDM signals are added/ dropped or crossconnected directly in the optical layer.2 Thus, each signal could traverse different routes and pass through a different number of optical amplifiers. In addition, the noise spectrum in these networks may not be uniform due to the optical filtering that occurs in various network elements, such as a reconfigurable optical add/drop multiplexer (ROADM) and optical crossconnect (OXC). As a result, the accumulated noise levels in these networks could be quite different from channel to channel. Thus, ASE noise residing within a signal’s bandwidth (and, consequently, the true value of the OSNR) cannot be measured by using the conventional linear interpolation technique. During the past decade, many researchers have attempted to develop a technique capable of monitoring the true value of OSNR (i.e., in-band OSNR), even in a dynamic networking environment.3–5 For this purpose, it is necessary to differentiate and detect small noise components hidden behind the large signal. For example, such differentiation (and the detection of in-band noises) could be achieved by utilizing the different optical characteristics of the signal and noises (i.e., ASE noises © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00002-X

21

22

CHAPTER 2 Optical signal-to-noise ratio monitoring

are unpolarized and incoherent, while the optical signal is polarized and coherent). In this chapter we review various OSNR monitoring techniques that are based on these principles.

2.2 LINEAR INTERPOLATION TECHNIQUES 2.2.1 Optical spectrum analysis

In this technique, the spectrum of the ASE noise is assumed to be almost uniform. Thus, as long as no narrow optical-filtering component is used in the network, the levels of optical signals and background ASE noises can be easily estimated from the measured optical spectrum. Figure 2.1 shows a typical optical spectrum of WDM signals (and ASE noises) measured by using an optical spectrum analyzer (OSA). Note that the ASE noise levels in the background can be easily identified between WDM channels. Thus, the noise level at the operating wavelength of each WDM channel can be estimated by interpolating from the noise levels measured at the wavelengths in the immediate neighborhood of the channel of interest. The concept of this OSNR monitoring technique is further outlined in the inset of Figure 2.1. The OSNR can be easily determined by the difference (in dB) between the levels of the optical signal and the ASE noises. For example, the OSNR of the channel with the shortest wavelength in Figure 2.1 is found to be about 23 dB. A simple pattern analysis algorithm would be necessary to automatically determine the peaks (i.e., signal levels) and valleys (i.e., ASE levels) needed for OSNR estimation. Figure 2.2(a) depicts the operating principle of the OSA. The OSA measures the power of each spectral component conceptually by sweeping the center wavelength of a narrowband optical filter. Various implementations of the OSA using diffraction gratings or tunable filters have been explained elsewhere,6 and several cost-effective solutions for obtaining the optical spectrum can be found in References 7 and 8. The bandwidth of the ASE noises sliced by the filter is usually proportional to the filter bandwidth, which is often called the resolution bandwidth of the OSA. Thus, the power level of the measured ASE noises strongly correlates with the resolution bandwidth of the OSA, as illustrated in Figure 2.2(b). For example, the level of the ASE noises increases by 10 dB as the

Signal level

REF

23 dB

Optical power (dBm), 5 dB/div

Normal (A)

OSNR

Noise level WLOFST

Wavelength (nm), 3 nm/div

FIGURE 2.1 Graphical description of OSNR measurement based on optical spectrum.

2.2 Linear interpolation techniques

23

Optical power detector Optical signal Wavelength tunable narrowband filter

Display

(a) Optical power (dBm)

Resolution BW: 10R Resolution BW: R

10 dB

Wavelength (b)

FIGURE 2.2 (a) Conceptual diagram of optical spectrum analyzer. (b) Relationship between level of optical noise and resolution bandwidth (BW) of OSA.

resolution bandwidth of the OSA increases ten-fold. On the other hand, the levels of the WDM signals are not very sensitive to the resolution bandwidth of the OSA, unless the bandwidths of the optical signals are much wider than the resolution bandwidth. Thus, the measured OSNR is usually inversely proportional to the resolution bandwidth of the OSA. In optical communication systems, a 0.1-nm resolution bandwidth is assumed to be used if there is no note about it; otherwise, the resolution bandwidth must be provided with the OSNR value.

2.2.2 Out-of-band noise measurement If the ASE noises have a flat and wide optical spectrum, the noise power does not need to be measured at all wavelengths. It is possible to estimate the noise levels of other wavelengths by measuring the noise power at an out-of-band wavelength and extrapolating it to the in-band wavelengths, as illustrated in Figure 2.3.9 This technique demultiplexes the WDM signals by using a demultiplexer such as an arrayed waveguide grating (AWG), and it measures the power of each port of the demultiplexer by using a photo detector (PD). While the channel powers are measured using the ports of the channel wavelengths, the noise power is measured using the port of the out-of-band wavelength. If the out-of-band ASE noises are filtered by wavelength blockers (WBs), such as a dynamic gain equalizer (DGE) within the optical link, it is necessary to measure the noise power at every input of WB and add all of the measured values to obtain the total noise power, as shown in Figure 2.4.10 Subsequently, the OSNR of each channel is estimated from the measured powers by assuming that

24

CHAPTER 2 Optical signal-to-noise ratio monitoring

Signal l1 ~ ln

Transmission characteristics of AWG Optical tap Signals

ln

ASE noise

AWG

OSNR calculation

l1

lm

ASE noise OSNR monitor

PD

1

2

n

m Port # of AWG

FIGURE 2.3 Out-of-band noise measurement using AWG.

Tx

Rx DGE

DGE

Tx

Rx OSA

OSA

OSA OSNR Calculation

FIGURE 2.4 Out-of-band noise measurement using OSAs, when there are DGEs within the optical link. The OSAs measure the power of out-of-band ASE noise for the OSNR monitoring, as well as the channel powers for the gain equalization.

the ASE noises have a uniform spectrum or at least a deterministic spectral shape. However, this assumption is valid only for a static optical link. The spectrum of the ASE noises tends to be sensitive to channel-loading conditions such as the number of channels and the wavelengths occupied by those channels. Thus, if the channel-loading conditions are changed in the optical link, the correlation between in-band and out-of-band noise levels could be significantly reduced, and the accuracy of this technique is consequently degraded.

2.2.3 Potential problems Linear interpolation techniques are very useful for measuring the OSNR of WDM signals in point-topoint optical links. However, in dynamically reconfigurable networks, the accuracy of these techniques could be severely degraded because the spectrum of the ASE noises could have irregular levels that vary with the wavelength.11 Figure 2.5(a) shows an example of such a case in an optical network consisting of ROADMs. Eight wavelength channels are multiplexed and transmitted over seven

2.2 Linear interpolation techniques

Point A

l1

l8

ADM

ADM

ADM

ADM

ADM

ADM

ADM

l8

l7

l6

l3

l5

l1

l2

25

l1

l8

MUX

DMUX (a) 10

10 Optical power (dBm)

Optical power (dBm)

0 −10 −20 −30 −40 −50 1545

True OSNR

10 Gb/s

CW

λ1 1548

λ8 1551

1554

1557

1560

1563

0 −10 10 −20 −30 −40 −50 1545

False OSNR 1548

1551

1554

1557

Wavelength (nm)

Wavelength (nm)

(b)

(c)

1560

1563

FIGURE 2.5 (a) Example of a dynamically reconfigurable transparent optical network configured with ROADMs. Optical spectrum measured at point A, (b) when unmodulated CW signals are transmitted, or (c) when 10-Gb/s NRZ signals are transmitted. (Resolution bandwidth: 0.05 nm.)

add/drop multiplexing (ADM) nodes. At each ADM node, a certain wavelength channel is dropped and a new channel having an identical wavelength is added. The other wavelength channels pass through the node transparently in the optical layer. In this scenario, each wavelength channel should traverse a different number of ADM nodes. Thus, the wavelength channels have different OSNR levels when they are observed at point A of the link. Figure 2.5(b) shows the optical spectrum obtained at point A when unmodulated continuous-wave (CW) signals are transmitted. The spectrum of the ASE noises is significantly reshaped due to optical filtering at each ADM node. Since the ASE noises between channels are significantly filtered out at every node, the levels of the ASE noises under the channel wavelengths have no correlation with those between the channels. In addition, it is not easy to identify the levels of the filtered ASE noises if the channels are modulated at a high bit rate, as shown in Figure 2.5(c). Thus, in such a case, linear interpolation techniques could induce a large error in OSNR measurement. The situation is exacerbated when high-speed phase-modulated signals are transmitted with dense channel spacing. The spectral power of the phase-modulated signals tends to be well distributed like ASE noises since they often have no carrier component carrying strong optical power. Figure 2.6 compares the spectrum of 43-Gb/s return-to-zero (RZ) differential quadrature phase-shift keying

Optical power (dBm), 5 dB/div

CHAPTER 2 Optical signal-to-noise ratio monitoring

Optical power (dBm), 5 dB/div

26

Wavelength (nm), 0.65 nm/div

Wavelength (nm), 0.65 nm/div

(a)

(b)

FIGURE 2.6 (a) Optical spectrum of 43-Gb/s RZ-DQPSK signals. (b) Optical spectrum measured after turning off modulators. (Resolution: 0.1 nm; div, division.)

(DQPSK) signals with that of CW signals; the spectrum was measured after turning off the modulators. One group of channels (i.e., the four channels located on the left side) was transmitted over a longer route than the other. Thus, the channels in this group have worse OSNR values than those in the other, as clearly seen in Figure 2.6(b), which was obtained when the CW signals were transmitted. However, it is impossible to identify this difference from the spectrum in Figure 2.6(a), which was obtained with the 43-Gb/s RZ-DQPSK signals. Based on this optical spectrum, the linear interpolation techniques may determine that all channels have the same OSNR. Therefore, to accurately monitor the OSNR in reconfigurable transparent optical networks, it is necessary to measure the in-band ASE noises under the channel’s wavelength.

2.3 POLARIZATION-BASED TECHNIQUES The ASE noises have an even power distribution for every polarization direction, while an optical signal typically contains the entire power in a single polarization state. Therefore, many studies have attempted to monitor in-band OSNR by utilizing these properties of the optical signal and ASE noises.12–24 This section introduces the operating principle of polarization-based OSNR monitoring techniques; it also investigates how various polarization effects influence the monitoring performance and discusses the several techniques that have been developed to mitigate these effects.

2.3.1 Operating principles 2.3.1.1 Polarization-nulling technique Figure 2.7 shows the operating principle of the polarization-nulling technique.20 An arbitrarily polarized (linear, circular, or elliptical) optical signal can be linearly polarized by using a polarization controller. However, this polarization controller does not confer any changes upon the polarization state of

2.3 Polarization-based techniques

27

Signal + Polarized ASE noise

Linear polarizer Linear polarizer

Arbitrarily polarized signal + ASE noise

Polarized ASE noise

Polarization controller

Linearly polarized signal + ASE noise

FIGURE 2.7 Principle of OSNR monitoring based on polarization-nulling technique.20

Copyright © 2006 IEEE.

the unpolarized ASE noises since the ASE noises have an even power distribution along every polarization direction (i.e., completely depolarized). The mixture of the linearly polarized signal with the unpolarized ASE noises can be split into two orthogonal polarization components by using two orthogonally aligned linear polarizers, as shown in Figure 2.7. The power of the optical signal together with the polarized ASE noises can be measured with the first linear polarizer, which is aligned with the signal’s polarization direction. On the other hand, only the polarized ASE noises can be measured by using the second linear polarizer, which is aligned with the orthogonal state of the signal’s polarization direction. The polarized ASE noises have a power corresponding to one-half the total ASE noise. Thus, the optical powers measured after the first and second linear polarizers can be expressed as Pp ¼ Ps þ 0:5 Pn

(2.1)

Po ¼ 0:5 Pn ;

(2.2)

and where Ps and Pn represent the powers of the optical signal and unpolarized ASE noises, respectively. Using these measured powers (i.e., Pp and Po), we can obtain the OSNR as

 Ps Bn P p  Po B n ¼ ; (2.3) OSNR ¼ Pn B r 2 Po Br where Bn is the noise-equivalent bandwidth—which is usually determined by the passband of the channel section filter—and Br is the resolution bandwidth that defines the OSNR (usually 0.1 nm).20 Thus, the OSNR can be estimated at the resolution of Br simply by measuring Pp and Po. Figure 2.8 shows two typical configurations of the polarization-nulling technique. The state of polarization (SOP) of an optical signal can be adaptively controlled by referencing the optical power

28

CHAPTER 2 Optical signal-to-noise ratio monitoring

Feedback Power Linear polarizer detector

Channel selection filter

Controller Polarization controller (a) WDM signals Channel selection filter

l/4 plate

Linear polarizer

Power detector Oscilloscope

(b)

FIGURE 2.8 Configuration of polarization-nulling technique based on (a) adaptive polarization control with feedback, or (b) rotating quarter-wave plate and polarizer.

measured at the output of a linear polarizer (Figure 2.8(a)).12 For example, the power of the polarized ASE noises can be obtained by setting the polarization controller to the minimum optical power. It is also possible to align the SOP of the optical signal to the transmission axis of the linear polarizer without using a feedback loop; this can be achieved simply by rotating a quarter-wave plate and a linear polarizer at different speeds (Figure 2.8(b)).13,14 In this case, the maximum and minimum powers after the polarizer can be obtained periodically.

2.3.1.2 Stokes parameter analysis Measuring the Stokes parameters of an optical signal is another polarization-based technique for monitoring the OSNR.21–24 Degree of polarization (DOP), one of the Stokes parameters, serves as an indicator of the power of ASE noises and is given by DOP ¼

Power of polzarized part Ps ; ¼ P s þ Pn Total power

(2.4)

where Ps and Pn represent the optical powers of the signal and the unpolarized ASE noises, respectively. Thus, the OSNR can be directly estimated from DOP as
 DOP Bn ; (2.5)  OSNR ¼ 1  DOP Br where Bn is the noise-equivalent bandwidth and Br is the resolution bandwidth. The underlying principle of this technique is basically identical to that of the polarization-nulling technique described in Section 2.3.1.1. However, Stokes parameters can also provide the polarization states of the signal, which can in turn be utilized to reveal the deleterious effect of polarization mode dispersion

2.3 Polarization-based techniques

30 Sensitivity to DOP error (dB/%)

50

25 OSNR (dB)

29

20

15

10 90 91 92 93 94 95 96 97 98 99 100 DOP (%)

40 30 20 10 0 90 91 92 93 94 95 96 97 98 99 100 DOP (%) (b)

(a)

FIGURE 2.9 (a) Correlation between DOP and OSNR. (b) OSNR monitoring sensitivity to DOP error.

DOP / Normalized power

(PMD) (see Section 2.3.2.1). The detailed procedure for measuring Stokes parameters can be found in Reference 25. Figure 2.9(a) shows the correlation between DOP and OSNR, obtained by using Equation (2.5) with Bn/Br ¼ 1. It should be noted that the accuracy of the OSNR monitoring process becomes very sensitive to the DOP error as the OSNR increases, as shown in Figure 2.9(b). For example, when the OSNR is around 25 dB, a mere 0.1% error in DOP measurement could cause a 1.4-dB error in OSNR estimation. Thus, very high accuracy in DOP measurement is required for monitoring the OSNR, especially when the OSNR of an optical signal is greater than 20 dB. Figure 2.10 shows the DOP and normalized optical spectrum of WDM signals measured by sweeping the wavelength.21 The positions of the DOP peaks are well correlated with those of the spectral peaks. The spectral Stokes parameter measurement is also beneficial for the simultaneous monitoring of OSNR and PMD.

1 0.8

DOP Power

0.6 0.4

ASE noise

0.2 0

1546

1548

1550

1552

1554 1556

1558

Wavelength (nm)

FIGURE 2.10 Measured DOP and normalized power of WDM signals.21

Copyright © 2006 IEEE.

1560

30

CHAPTER 2 Optical signal-to-noise ratio monitoring

2.3.2 Potential problems and limitations The performance of polarization-based OSNR monitoring techniques may be degraded if an optical signal experiences various polarization effects within an optical transmission link. For example, an error could occur if the signal is depolarized due to PMD or nonlinear birefringence. The accuracy could also be compromised if the ASE noises are partially polarized due to polarization-dependent loss (PDL). In addition, failures in tracking rapid fluctuations in the SOP of an optical signal could cause a large error.

2.3.2.1 Polarization mode dispersion PMD induces a time shift between two orthogonal polarization components of an optical signal. Thus, some portion of the signal power could leak into the orthogonal polarization state where the two polarization components are not overlapping, as illustrated in Figure 2.11(a). This effect can be described in the frequency domain, as shown in Figure 2.11(b). When the optical signal (broadened by modulation) propagates through a fiber link having non-negligible PMD, each spectral component of the modulated signal could have a different polarization state at the end of the fiber due to PMD.26 Thus, it is impossible to nullify all the spectral components of the signal simultaneously by using the polarizer aligned to the orthogonal state from the signal’s polarization; this, in turn, results in monitoring errors.14,20 The effect of PMD can be included in Equations (2.1) and (2.2) as Pp ¼ Ps ð1  ePMD Þ þ 0:5 Pn

(2.6)

Po ¼ Ps ePMD þ 0:5 Pn ;

(2.7)

and

w2 w1 w3 w0 w4

Linear polarizer

w w0 1

w2

w1 w0

Birefringent fiber

w2 w1 w 3 w0 w4

w3

w w4 3

w4 Linear polarizer

Birefringent fiber

foptical (a)

(b)

FIGURE 2.11 Illustration of the error mechanism caused by PMD in the execution of the polarization-based OSNR monitoring technique in (a) the time domain and (b) the frequency domain.

2.3 Polarization-based techniques

31

where ePMD represents the fraction of the optical signal leaked into the orthogonal polarization state from the signal’s polarization due to PMD. Assuming that only a first-order PMD exists, we can estimate this term as sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð 2 1 1 2 2 (2.8) ePMD ¼  cos y þ sin y cosðtDoÞPðDoÞdDo ; 2 2 where y represents the polarization angle between the input SOP of the optical signal and the PMD vector of the transmission link, t is the magnitude of the PMD vector, Do is the optical angular frequency offset from the center frequency, ð and P(Do) is the power spectral density function of the modulated optical signal defined as PðDoÞdDo ¼ 1.20,27 Thus, by substituting Equations (2.6) and (2.7) in Equation (2.3), the OSNR error caused by PMD can be described as
   OSNRr 1 þ 2ePMD OSNRr ðBr =Bn Þ ¼ 10 log ; ErrorðdBÞ ¼ 10 log OSNRm 1  2ePMD

(2.9)

where OSNRr is the real OSNR, and OSNRm is the measured OSNR. As expected, the OSNR error caused by PMD increases with ePMD. Since ePMD is dependent on the magnitude of the PMD vector and the spectral width of the optical signal, the OSNR error increases with PMD and the transmission speed. This equation shows that the OSNR error is also dependent on OSNRr; if the OSNR is high, even a small ePMD could cause a large monitoring error. Figure 2.12 shows the OSNR error caused by PMD when the OSNR is measured after singlechannel transmission over an optical link with a mean PMD of 3.22 ps.14 The error bars represent the fluctuations of the measured OSNR errors caused by the random nature of PMD. The monitoring error caused by PMD increases with the bit rate because the spectral bandwidth of the optical signal is broadened. In general, PMD causes a non-negligible amount of error in the execution of polarizationbased OSNR monitoring techniques if the system is operating at high speed (i.e., 10 Gb/s). 3

OSNR error (dB)

Experiment 2

Calculation

1

0 –1 0.4

1.0

2.5 Bit rate (Gb/s)

5.0

10.0

FIGURE 2.12 OSNR monitoring error caused by PMD (mean PMD ¼ 3.22 ps).14

Copyright © 2001 IEEE.

32

CHAPTER 2 Optical signal-to-noise ratio monitoring

2.3.2.2 Nonlinear birefringence When WDM signals are transmitted through an optical fiber, the polarization state of one wavelength channel (probe) can be affected by the intensity of other channels (pumps); this is called nonlinear birefringence.28,29 Thus, the polarization state of the probe signal could be scattered by the intensitymodulated pump signal. The speed of polarization scattering (i.e., order of data rate) is typically much faster than the polarization-adjusting time of the polarization-nulling technique. Thus, it would be difficult to ensure complete linear polarization of this polarization-scattered signal during the OSNR measurement; this results in overestimation of the noise power and causes monitoring errors, as shown in Figure 2.13.14,20,30 The depth of nonlinear polarization scattering depends on the relative polarization angle between channels. The worst-case scenario is when the SOP of the probe signal is 90 apart from that of the pump signal on the Poincare sphere.20,28 The effects of nonlinear birefringence in an M-channel WDM system can be analyzed as follows: Because of nonlinear birefringence, a small portion of the signal power could be included in the noise power, which is measured by using the linear polarizer set in the SOP orthogonal to the signal. Neglecting the pulse distortion caused by chromatic dispersion (CD), we can estimate this portion of the signal power as follows for the ith channel in an M-channel WDM system (for the worst-case analysis)14 as ð

1 h1  s^i ^tiPðoÞdo 2 2 0 8 ! ð M < n o 2 X 16 B2 ¼  Fij ðoÞcos fij ðoÞ þ 41  J 0 @ 2 3 : j¼1; j6¼i ij

eNL;i ¼

M X

n

ij Fij ðoÞsin fij ðoÞ

j¼1; j6¼i

o

!2 91=2 = ;

13 C7 A5PðoÞdo;

(2.10)

Linear polarizer

l1

Demultiplexer l1 l2

Linear polarizer

Nonlinear fiber

q

FIGURE 2.13 Illustration of error mechanism caused by nonlinear birefringence in execution of polarization-based OSNR monitoring technique.

2.3 Polarization-based techniques

33

where s^i and ^t are the normalized Stokes vectors of the ith channel and the transmission axis of a polarizer, respectively; hi denotes the time average; o is the angular frequency; J0 is the Bessel function of the first kind of order 0; N is the number of spans; Zij is the link enhancement factor (ij ¼ N for the dispersion-compensated link and ij ¼ jsin(No dijL/2)/sin(o dijL/2)j for the link without dispersion compensation)31; L is the span length; dij is the group velocity mismatch between channels i and j; Fij is the AC portion of the nonlinear phase shift of the ith channel caused by the jth channel; and ’ij is the phase retardation factor.28 From this result, it can be inferred that the effect of nonlinear birefringence is sensitive to the dispersion coefficient of the transmission fiber and the dispersion map of the link. The monitoring error caused by nonlinear birefringence could be obtained by replacing ePMD with eNL in Equation (2.9) as   1 þ 2eNL OSNRr ðBr =Bn Þ : (2.11) ErrorðdBÞ ¼ 10 log 1  2eNL Figure 2.14(a) shows the effect of nonlinear birefringence, as measured in a simple two-channel experiment in which a CW signal (probe) and a sinusoidally modulated signal (pump) were transmitted together over a 40-km-long, standard single-mode fiber (SMF) link.20 In contrast with the effect of PMD, the monitoring error is maximized at low frequencies. The error substantially decreases with the modulation frequency. This is because the nonlinear polarization scattering should be averaged out due to the group velocity mismatch between channels caused by chromatic dispersion (CD) as the modulation frequency of the pump signal increases. This effect is clearly observed in Figure 2.14 20

1.5

Exp

Cal

OSNR error (dB)

OSNR error (dB)

Calculation Experiment 1.0

0.5

SMF NZDSF SMF+DCF

15

10

5

0.0

0 0.1

Modulation frequency (Hz)

1.0 Bit rate (Gb/s)

(a)

(b)

1E8

1E9

10

FIGURE 2.14 Maximum OSNR errors caused by nonlinear birefringence (a) measured in a two-channel experiment (200-GHz spacing, and 0- and 7-dBm input power for the probe and pump, respectively)20 (Copyright © 2006 IEEE ), or (b) measured in various 640-km-long fiber links with six channels (200-GHz spacing, 0-dBm/channel input power, 80 km  8 spans) (Copyright © 2001 IEEE ).14

34

CHAPTER 2 Optical signal-to-noise ratio monitoring

(b), which shows the effect of nonlinear birefringence measured in various fiber links.14 The OSNR monitoring error measured in the nonzero dispersion-shifted fiber (NZDSF) link is larger than that in the SMF link due to its smaller dispersion coefficient and effective area. However, the largest monitoring error is observed in the SMF þ DCF (dispersion-compensating fiber) link since the DCF has the smallest effective area and the effect of nonlinear birefringence generated in each span linearly accumulates in the dispersion-compensated fiber link. Although this result includes the effect of PMD, the effect was not observed clearly except in the 10-Gb/s case in the SMF þ DCF link, which had the largest PMD (3.22 ps). From this result, it can be concluded that the effect of nonlinear birefringence predominates over the effect of PMD with regard to low-speed signals (<10 Gb/s). However, it should be noted that the effect of nonlinear birefringence can be significantly reduced for purely phase-modulated signals that have no intensity variation.

2.3.2.3 Polarization-dependent loss The polarization-based, OSNR monitoring technique assumes that ASE noises are fully unpolarized. Based on this assumption, it estimates the power of the ASE noises within the signal’s bandwidth by measuring only the noise power in the polarization state orthogonal to the signal’s SOP. However, if the ASE noises are partially polarized due to PDL, the noise power in the orthogonal polarization state may no longer be identical to the noise power in the state parallel to the signal’s SOP. Thus, it could cause an error in the OSNR measured by using the polarization-based OSNR monitoring technique, as shown in Figure 2.15.32,33 If the OSNR is much higher than 10 dB, this monitoring error (caused by partially polarized ASE noise) can be estimated as s  n^Þ; ErrorðdBÞ ¼ 10 log½1  DOPASE ð^ Signal + Polarized ASE noise

Linear polarizer

Linearly polarized signal + Unpolarized ASE noise

(2.12)

Different amount of ASE noise

Linear polarizer

Max Polarized ASE noise Min

PDL element

Signal + Partially polarized ASE noise

FIGURE 2.15 Illustration of the error mechanism caused by PDL in execution of polarization-based OSNR monitoring technique.

2.3 Polarization-based techniques

35

where DOPASE is the DOP of the ASE noises; s^ and n^ represent the normalized Stokes vectors of the signal and the partially polarized ASE noises, respectively; and s^  n^ represents their inner product.32 Thus, the error is maximized when the SOPs of the signal and the ASE noises are located in parallel (i.e., s^ n^ ¼ 1) or anti-parallel positions (i.e., s^  n^ ¼ 1) on a Poincare sphere. The probability that the error caused by PDL in the OSNR measured using the polarization-based OSNR monitoring technique will be greater than x dB is given by 

 1 2ð10x=10  1Þ 2ð1  10x=10 Þ erfc pffiffiffi þ erfc pffiffiffi : (2.13) Probabilityfjerrorj  xðdBÞg  2 phDOPN i phDOPN i For hDOPN i 1 and x < 3, where hDOPNi is the average DOP of the ASE noise after the transmission of N amplifier spans.33 Figure 2.16(a) shows the cumulative probability of the errors in the measured OSNRs due to partially polarized ASE noise in a transmission link consisting of 15 spans (average PDL/span ¼ 0.57 dB).33 Figure 2.16(b) shows the plot of the probability of the error in the measured OSNR being >1 dB versus the PDL/span and the number of spans. If the PDL/span is <0.2 dB, the effect of PDL on the measured OSNR using the polarization-nulling technique is very small, even for an ultra-long-distance transmission system. For example, the probability that the error in the measured OSNR becomes >1 dB is <104 in a 50-span transmission system as long as the PDL/span is <0.2 dB. The PDL/span could be raised to >0.2 dB in a network utilizing ROADMs. In such cases, an increased PDL/span would certainly increase the error probability.

50 10 –3

40 Number of spans

Cumulative probability

10 0

10 –1

10 –2

Experiment

0.2 0.4 0.6 0.8 1.0 OSNR monitoring error (dB) (a)

10 –4 30 10 –5 20 10

Calculation 10 –3 0.0

10 –2

1.2

0 0.0

0.1

0.2 0.3 PDL/span (dB)

0.4

(b)

FIGURE 2.16 (a) Cumulative probability of errors in the measured OSNRs by using the polarization-nulling technique due to partially polarized ASE noise in a transmission link consisting of 15 spans (average PDL/span ¼ 0.57 dB). (b) Probability that the error in the measured OSNRs by using the polarization-nulling technique becomes >1 dB (due to the partially polarized ASE noise caused by PDL).33 Copyright © 2006 OSA.

36

CHAPTER 2 Optical signal-to-noise ratio monitoring

2.3.2.4 Polarization fluctuation When an optical signal traverses an installed fiber link—especially an aerial fiber link—its SOP could rapidly fluctuate in response to environmental conditions.34–36 Figure 2.17(a) shows the Fourier transform of an SOP fluctuation measured in a 120-km-long aerial fiber link installed in a field.35 The fluctuations have two strong frequency components at 60 Hz and 0.3 Hz. The 60-Hz peak is attributed to Faraday rotation caused by the current in the electrical power transmission line since it is identical to the current frequency used in electrical power lines. The other dominant peak at 0.3 Hz is caused by wind and the pendulum motion of the optical ground wire. However, current technologies can track polarization at a speed that is fast enough not to cause a meaningful error in the polarization-based OSNR monitoring technique, as shown in Figure 2.17(b).35

2.3.3 Methods to overcome limitations 2.3.3.1 Additional optical filtering

1.6

30

1.4

28

1.2 1.0

S1 S2

0.8

S3

Measured OSNR (dB)

Intensity of S-parameters (×10−2)

Figure 2.18 shows the schematic diagram of the improved polarization-nulling technique implemented using an additional optical filter.15,20 The optical signal, which is demultiplexed by using the first optical bandpass filter (BPF), is sent to a polarization beam splitter (PBS) via a polarization controller. This controller is used to maximize the signal power P1 in one arm. Thus, the signal and ASE noises are split into two polarization components after the PBS. If the effects of PMD and nonlinear birefringence could be neglected, one polarization component would contain only the polarized ASE noises, while the other would contain both the signal and polarized ASE noises. However, if the signal is depolarized by PMD and/or nonlinear birefringence, a small amount of signal power could leak into the arm in which we intend to measure only the noise power. To compensate for this, the polarized ASE noise component is split into two parts (P2 and P3) by using a 3-dB coupler; one part is then filtered by using the second optical BPF. This filter is used to reduce the bandwidth of

0.6 0.4 0.2 0.0

0

40

80 120 160 Frequency (Hz) (a)

200

26 24 22 20 18 16

0

5

10 15 20 25 30 35 40 45 Time (min) (b)

FIGURE 2.17 (a) Fourier components of the Stokes parameters of an optical signal measured in a 120-km-long aerial fiber link. (b) OSNR measured by using the polarization-nulling technique in a 120-km-long aerial fiber link.35 Copyright © 2004 OSA.

2.3 Polarization-based techniques

37

Signal ASE PBS WDM signal Polarization controller

BPF1

Small-portion signal due to PMD or nonlinear birefringence

PD

P1

PD

P2

PD

P3

1:1

BPF2

FIGURE 2.18 Schematic diagram of the polarization-nulling technique improved by using additional optical filter. PBS, polarization beam splitter; BPF, bandpass filter; PD, photo detector.20 Copyright © 2006 IEEE.

the ASE noise. Neglecting the loss of each path (which can be calculated and compensated for, if necessary), we can express the optical powers detected at photodiodes as P1 ¼ Ps ð1  eÞ þ 0:5 Pn ;

(2.14)

P2 ¼ 0:5 Ps e þ 0:25 Pn ;

(2.15)

P3 ¼ 0:5 Ps e þ 0:25aPn ;

(2.16)

and where e is the ratio of the signal power excluded from P1 (thus, included in P2 and P3) due to PMD, nonlinear birefringence, and/or incomplete polarization control; and a is the bandwidth reduction factor determined by the transmission characteristics of the first and second BPFs. By using these equations, the OSNR can be obtained as
  
 Ps B n ðP1 þ 2P2 Þ Bn ¼ : (2.17) 1 OSNR ¼ Pn Br Br 4ðP2  P3 Þ=ð1  aÞ However, in practice, it is difficult to use this technique when the ASE noise has a bandwidth comparable to that of the high-speed optical signal. In such cases, e in Equation (2.16) can differ from that in Equation (2.15) because the signal spectrum also narrows on account of tight optical filtering. In addition, this technique requires the precise alignment of two tunable filters when selecting a target channel from WDM signals.

2.3.3.2 Off-center narrow filtering OSNR monitoring errors caused by PMD or nonlinear birefringence can be significantly reduced by measuring the noise power at a frequency where the signal power is low.16,17 Figure 2.19(a) shows the principle of the off-center narrow filtering technique proposed for this purpose. The spectral power of a modulated optical signal is sensitive to frequency; it dramatically decreases as the frequency deviates from the center frequency of the signal. On the other hand, the ASE noises have

CHAPTER 2 Optical signal-to-noise ratio monitoring

Filtered spectrum Signal spectrum

Monitoring errors (dB)

38

4 2 0 −2 −4 Without PMD With 10-ps PMD

–6 −8 –10

ASE

1.0 l1

l2

(a)

1.5 2.0 2.5 3.0 3.5 Offset position of narrowband filter from center carrier (nm) (b)

FIGURE 2.19 (a) Illustration of off-center filtering technique for OSNR monitoring. (b) Effects of filter detuning for 39.81-Gb/s, 2.5-ps, full-width at half-maximum (FWHM) RZ signal.17 Copyright © 2004 IEEE.

an almost flat spectrum over a much wider frequency range than the optical signal. Thus, if the noise power is measured by the polarization-based technique at a frequency different from the center frequency, the fraction of the optical signal leaked into the orthogonally polarized ASE noises (due to PMD or nonlinear birefringence) should be significantly reduced. Figure 2.19(b) shows that the monitoring error caused by PMD decreases as the offset from the center frequency increases. In addition, narrow filtering further reduces the effect of PMD. If the filter bandwidth is sufficiently narrow, the term cos(tDo) in Equation (2.8) can be approximated to 1  (tDo)2. Thus, by assuming that y ¼ p/4 and P(Do) ¼ 1/Bf, where Bf is the filter bandwidth, Equation (2.8) can be simplified as vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi "ð #2ffi u 2 Bf =2 u ðtBf Þ2 1 1 1 1 1  ðtDoÞ : (2.18) dDo  ePMD   t þ Bf 48 2 2 2 2 Bf =2 Equation (2.18) indicates that the fraction of the optical signal leaked into orthogonally polarized noises due to PMD is proportional to the square of the filter bandwidth (i.e., Bf). On the other hand, the power of the filtered ASE noises (Pn) is linearly proportional to the filter bandwidth and is given by Pn ¼ NASE Bf ;

(2.19)

where NASE is the power spectral density of the ASE noises. Thus, narrow filtering decreases ePMD much more rapidly than the power of the ASE noises, which mitigates the effect of PMD.

2.3.3.3 Multiple-frequency measurement Although the off-center narrow filtering technique could reduce the monitoring error caused by PMD or nonlinear birefringence, it cannot completely remove it. Its performance can be limited when there

2.3 Polarization-based techniques

39

Filtered spectrum

Signal spectrum

ASE l1 PMD compensator

Tunable BPF

l2 Polarization controller

l1, l2

PD

Pl1,null Pl2,null

Polarizer PD

Pl1, Pl2

FIGURE 2.20 Schematic diagram of the polarization-nulling technique, improved by multiple-frequency measurement and PMD compensation.20 Copyright © 2006 IEEE.

is significant depolarization of an optical signal due to large PMD or nonlinear birefringence. Figure 2.20 shows a schematic diagram of another technique proposed to completely negate the effect of PMD and nonlinear birefringence.20,37 This technique measures the noise power at two frequencies (for instance, both at the center and at the slope of the signal’s spectrum). The optical signal is first sent to a PMD compensator (PMDC) to reduce the effect of PMD on the OSNR measurement. After the PMDC, the signal is sent to a tunable BPF having a bandwidth much narrower than the signal bandwidth. The narrow bandwidth filtering process further reduces the effect of PMD, including higher-order PMD that is not removed by using the PMDC. The filter is first adjusted to the center of a specific WDM channel. Then, the power of the filtered signal (Pl1) is measured before the linear polarizer. At the same time, the ASE noise power polarized orthogonal to the signal (Pl1,null) is measured by using the polarization-nulling technique. The tunable BPF is then adjusted again to measure the powers of the signal (Pl2) and the polarized ASE noise (Pl2,null) on the slope of the signal’s spectrum (i.e., at a frequency a few gigahertz away from the center frequency). If the optical signal is slightly depolarized after the transmission due to nonlinear birefringence, a small portion of the optical signal may be present in addition to the polarized ASE noise in Pl1,null or Pl2,null. Thus, Pl1,null and Pl2,null can be expressed as 1 Pl1;null ¼ ðPl1  Pn ÞeNL þ Pn 2

(2.20)

1 Pl2;null ¼ ðPl2  Pn ÞeNL þ Pn 2

(2.21)

and

40

CHAPTER 2 Optical signal-to-noise ratio monitoring

where Pn is the power of the filtered ASE noise and eNL is the portion of the optical signal that leaks into the orthogonally polarized ASE noise due to nonlinear birefringence. Thus, the power of the ASE noise and the OSNR can be accurately estimated without considering the effect of eNL as Pn ¼

2ðPl1 Pl2;null  Pl1;null Pl2 Þ Pl1  Pl2  2Pl1;null þ 2Pl2;null

(2.22)

and OSNR ¼

Pt  Pn Bt =Bf ; Pn Br =Bf

(2.23)

where Pt is the total power of the optical signal and ASE noise within the bandwidth Bt (measured by scanning the tunable filter over the signal’s entire spectrum), and Bf is the bandwidth of the tunable BPF. This technique assumes that the effect of nonlinear birefringence (i.e., eNL) is even for all spectral components. However, this assumption may not hold if nonlinear birefringence significantly interacts with large PMD. Since PMD causes divergence of the SOPs of the spectral components, a different amount of nonlinear polarization scattering could be induced on each spectral component.

2.3.3.4 Spectral SOP measurement The OSNR monitoring technique based on Stokes parameter analysis can also separate the effect of PMD by measuring the signal’s SOP at multiple wavelengths with very high spectral resolution, as shown in Figure 2.21.23 The PMD spreads the SOP of an optical signal on a Poincare sphere, which causes a reduction in the signal’s DOP. It can be shown that the measured DOP in a system suffering from PMD will be the product of the two DOP factors from the ASE noises and PMD as given by23 ð ð ð Ps ðlÞdl sðlÞPs ðlÞdl sðlÞPs ðlÞdl ¼ ð ð ¼ DOPPMD  DOPOSNR ; (2.24) DOPMEAS ¼ ð Ps ðlÞ þ Pn ðlÞdl Ps ðlÞdl Ps ðlÞ þ Pn ðlÞdl

Filter spectrum

Modulated signal spectrum l5 l1

l4 l2

l1

l2

l3

l4

l5

FIGURE 2.21 Principle of spectral SOP measurement technique.

l3

2.4 Interferometer-based technique

41

where s(l) is the normalized Stokes vector as a function of frequency, and Ps,n(l) is the spectral power density of the signal and noise. Thus, it is possible to monitor the OSNR of the optical signal without considering the effect of PMD if the DOPPMD is estimated from the powers and SOPs of the spectral components of the optical signal measured at multiple wavelengths. However, this technique is still vulnerable to the effect of nonlinear birefringence.

2.4 INTERFEROMETER-BASED TECHNIQUE 2.4.1 Operating principle

Several techniques have been proposed for using the coherency property of an optical signal when monitoring the OSNR.38–41 The basic assumption is that the optical signal is highly coherent, whereas ASE noises are incoherent. These techniques are conceptually configured with a tunable filter to select a channel to be monitored, a Mach-Zehnder delay interferometer (MZDI), and a power meter, as shown in Figure 2.22(a). The MZDI consists of two 3-dB couplers that provide separate paths, with an optical delay and a phase adjuster in one path. The phase adjuster tunes the phase of an optical signal in one path so that the two paths are combined constructively or destructively at the output of the interferometer. The optical delay is used to control the free spectral range (FSR) of the MZDI—the spectral spacing between two adjacent transmission peaks. The total

Optical Phase + delay adjuster Tunable filter

3-dB coupler

3-dB coupler

Power meter

(a) MZDI transmission characteristic Power (dBm)

Power (dBm)

Signal

Wavelength (b)

ASE noise

Wavelength

(c)

FIGURE 2.22 (a) Schematic diagram of OSNR monitoring technique based on MZDI interferometer. (b) Total power measurement with constructive interference. (c) Noise power measurement with destructive interference.

42

CHAPTER 2 Optical signal-to-noise ratio monitoring

power—that is, signal plus ASE noise power—can be measured with the constructive interference, as shown in Figure 2.22(b). On the other hand, the noise power can be measured by eliminating the optical signal with the destructive interference, as shown in Figure 2.22(c). The OSNR of the optical signal can be estimated from the noise and the total optical power. This technique was found to be insensitive to CD or polarization effects.41

2.4.2 Potential problems and limitations The interferometer-based technique seems capable of monitoring in-band noise by eliminating the optical signal based on the assumption that the optical signal is highly coherent whereas the ASE noise is incoherent; however, this is not the case. The destructive MZDI eliminates some portion of the ASE noises, as well as the spectral components of the optical signal since its transmission characteristic is common to the signal and the ASE noises, regardless of their coherency properties. The destructive MZDI is nothing but a notch filter that is centered at the signal’s wavelength; it rejects not only the optical signal but also the ASE noises. As a result, it is difficult to use this technique for monitoring the OSNR of WDM signals with high spectral density. In this case, the ASE noises filtered by the channel selection filter should have a narrow bandwidth comparable to that of the modulated signal. The MZDI with large FSR removes the excess noise power, whereas the MZDI with small FSR leaves some portion of the high-speed optical signal unsuppressed. Both cases induce an error in OSNR monitoring techniques.

2.4.3 Method to overcome limitations One possible solution to overcome the limitation described in Section 2.4.2 is to use the polarization property of the ASE noises, as shown in Figure 2.23. By using a PBS, the optical signal is split into two orthogonal polarization components having equal power. Two outputs of the PBS are combined in the same polarization state by inserting a half-wave plate in one path. Thus, the optical signal can be canceled out by adjusting the phase of one path so that the two paths interfere destructively, as is the case in an MZDI. On the other hand, because the ASE noises from the two paths have no correlation with each other, they should be added in the random phase statistically (i.e., as the sum of optical powers), regardless of phase adjustment. As a result, the power of the in-band ASE noise can be measured accurately. Having an infinite FSR is the best means to completely remove the optical signal with this technique since the noise power is not affected by the interferometer. Thus, unlike PBS Tunable filter

Polarization controller

Half-wave plate

Phase adjuster

3-dB coupler

FIGURE 2.23 Schematic diagram of modified optical interferometer for OSNR monitoring.

Power meter

2.5 Beat noise analysis techniques

43

the MZDI-based technique, an optical delay in adjusting the FSR is not necessary. However, this technique would be sensitive to various polarization effects such as PMD and nonlinear birefringence, as described in Section 2.3.2.

2.5 BEAT NOISE ANALYSIS TECHNIQUES Amplification of an optical signal brings about considerable improvement in the sensitivity of an optical receiver. Since the power of the optical signal can be sufficiently amplified at the receiver through optical preamplification, the thermal and circuit noises that occur inside the receiver no longer limit the performance of the receiver. Instead, the noises caused by beating between the optical signal and the ASE noises become a dominant source of receiver noise; consequently, many researchers have attempted to estimate the OSNR by analyzing the electrical noises at the receiver. This section introduces several key techniques based on beat noise analysis that are proposed to estimate the in-band OSNR of the optical signal.

2.5.1 Operating principle The electrical noises at the optical receiver consist of various components such as shot noises (Nshot), thermal noises (Nthermal), circuit noises (Ncircuit), and beat noises (Nbeat)42: Ntotal ¼ Nbeat þ Nshot þ Nthermal þ Ncircuit :

(2.25)

Beat noises occur when an optical signal is detected together with the ASE noises by a square-law receiver (i.e., direct detection receiver). From the electrical fields of the optical signal (Es) and the ASE noises (Ease), the current (Ircv) at the square-law receiver can be obtained as 

Ircv ðtÞ ¼ RjEs ðtÞ þ Ease ðtÞj2 ¼ RjEs ðtÞj2 þ 2R  RefEs ðtÞEase ðtÞg þ RjEase ðtÞj2 ;

(2.26)

where R is the responsivity of the photodiode of the receiver. The second term on the right side of Equation (2.26) represents the signal-spontaneous beat noises, which are caused by beating between the optical signal and the ASE noises. The third term represents the spontaneous-spontaneous beat noises caused by beating among ASE noises. The spectral density functions of the signal-spontaneous (Ss-ase) and spontaneous-spontaneous (Sase-ase) beat noises can be expressed as42 ssase ð f Þ /

2Ps Pn Bo

(2.27)

and sasease ð f Þ /

P2n ðBo  f Þ; B2o

(2.28)

where Ps and Pn denote the average power of the optical signal and the ASE noises, respectively, and Bo is the optical bandwidth (i.e., the bandwidth of the ASE noises). Figure 2.24(a) shows the power spectral density of the receiver noises caused by these beat noises. Thus, the electrical power of the beat noises (Pbeat) at low frequencies (i.e., f Bo ) can be expressed as

CHAPTER 2 Optical signal-to-noise ratio monitoring

Power spectral density

44

2IsIase Bo OSNR = 20 dB

Iase2

OSNR = 30 dB

Bo (a)

Be

Bo Frequency

40.0 kHz

50.0 kHz

(b)

FIGURE 2.24 (a) Power spectral density of receiver noises caused by signal-ASE and ASE-ASE beating. (b) Measured receiver noise spectrum from 40 to 50 kHz when OSNR is 20 dB or 30 dB.

Pbeat  A



 2Ps Pn P2n 2AP2s 1 Bo ¼ ; þ þ Bo Bo Br OSNR 2Br  OSNR2

(2.29)

where A is a constant and OSNR is the OSNR of the optical signal measured with the resolution bandwidth of Br.42 The power of the beat noises strongly depends on the OSNR of the signal, as shown in Figure 2.24(b). On the other hand, the thermal and circuit noises in Equation (2.25) are independent of the optical signal, and the shot noises are only decided by the total received optical power; this means that these noises can be premeasured and calibrated, if necessary, regardless of the OSNR level. This noise-level calibration is useful for improving monitoring accuracy, especially for high OSNR values. However, in most cases, the beat noise power can simply be approximated to the total noise power, even without noise-level calibration. In any case, the power of the beat noises (Pbeat) can be estimated from the total power of the receiver noises. The total optical power (Ptotal), consisting of the signal power and the ASE noise power, is given by42
 1 Bo : (2.30) Ptotal ¼ Ps þ Pase ¼ Ps 1 þ OSNR Br From Equations (2.29) and (2.30), we have only two unknowns (i.e., Ps and OSNR), since A, Bo, and Br are constants and Ptotal and Pbeat are measurable. Thus, the in-band OSNR can be estimated by measuring the total optical power and the electrical power of the receiver noises. Figure 2.25 shows the principle of the OSNR monitoring technique based on low-frequency beat noise analysis.42 If the signal comprises repetitive bit patterns, such as pseudo-random bit sequence (PRBS) patterns, the signal spectrum should consist of discrete spectral components, as shown at the bottom of Figure 2.25. In this case, the electrical power of the beat noises can be measured at low frequencies (i.e., between the DC and the first spectral component) by using low-speed electronics consisting of an analog-to-digital converter (ADC) and a fast Fourier transform (FFT). This technique splits the optical signal into two branches and measures the total optical power in one branch and the electrical power of the beat noises in the other.

2.5 Beat noise analysis techniques

45

Total power measurement Tunable BPF

OSNR calculation

PD

PD

ADC

FFT

Power spectral density

Noise power measurement Signal Receiver noise OSNR change

Δν

B

Noise measuring point

2B

Frequency

Δν = B/pattern length

FIGURE 2.25 Principle of OSNR monitoring technique based on low-frequency beat noise analysis. ADC, analog-to-digital converter; B, bit rate; FFT, fast Fourier transform; PD, photodetector.

2.5.2 Potential problems and limitations The OSNR monitoring technique based on low-frequency noise analysis can be applied to only a very limited number of cases. For example, this technique can be utilized for monitoring the OSNR if the phase of an optical carrier carries user traffic like PSK signals.43,44 In this case, the signal spectrum does not interfere with the low-frequency beat noises since the photodiode detects only the optical intensity. However, if the intensity of the optical carrier carries information, the signal spectrum interferes with the low-frequency beat noises. Thus, this technique can be applied only to optical signals that carry traffic with a pattern length no longer than 215  1.42 As shown in Figure 2.26, an intensity-modulated signal with a long pattern has densely spaced spectral components, which makes it difficult to measure the noise power at low frequencies, regardless of the signal components involved. The most critical problem arises from the fact that no frequency region is free from the signal spectrum at low frequencies if the signal carries real random patterns. Real user traffic has random patterns rather than repetitive ones such as PRBS patterns. Thus, in this case, the lowfrequency noise analysis technique cannot be utilized for monitoring the OSNR because of the signal spectrum. In addition, it should be noted that low-frequency noises are also induced by multipath interference (MPI) noises, which can cause an error in OSNR estimation.45 It is possible to estimate the OSNR by measuring the power of the beat noises at high frequencies, as shown in Figure 2.27(a).46 For example, a non-return-to-zero (NRZ) signal has spectral null points at frequencies corresponding to the harmonics of the clock frequency. In addition, the power

46

CHAPTER 2 Optical signal-to-noise ratio monitoring

hp REF 0.0 dBm 10 dB/

ATTEN 10 dB 10 Gb/s 23 2 −1 PRBS

20 dB

30 dB

START

10.0 kHz RES BW 100 Hz

VBW 10 Hz

STOP 20.0 kHz SWP 30.0

FIGURE 2.26

Signal 10 dB/div

Power spectral density

RF spectrum of 10-Gb/s NRZ signal with PRBS pattern (pattern length: 223 – 1).

OSNR 20 dB 30 dB 2.5 GHz

Frequency Noise measuring point (a)

9.8 GHz 1 GHz

Frequency

11 GHz

(b)

FIGURE 2.27 (a) Principle of OSNR monitoring technique based on high-frequency beat noise analysis. (b) RF spectrum of 2.5-Gb/s NRZ signal with 20-dB or 30-dB OSNR.

spectral density of the received signal decreases to dramatically low levels as the frequency exceeds the second null point. Thus, the power of the beat noises can be measured at these frequencies for estimating the OSNR. Figure 2.27(b) shows the RF spectra of a 2.5-Gb/s NRZ signal when it has an OSNR of 20 or 30 dB.46 The power density is sensitive to the OSNR at high frequencies where there are negligible signal spectral components. Figure 2.28 shows the performance of the OSNR monitoring technique based on high-frequency noise analysis for the 2.5-Gb/s NRZ signal at various transmission distances. When the OSNR was estimated from the power level of the 2.5-GHz null point, the accuracy fluctuated significantly, as shown in Figure 2.28(a). This mainly occurred when the residual CD was large and, as a result, the shape of the RF spectrum of the received signal was significantly transformed. The 2.5-GHz null point disappeared due to interference caused by nonzero

2.5 Beat noise analysis techniques

OSNROSA

28 27 26 25 24 23 22 21 20

OSNRNoise

OSNR (dB)

OSNR (dB)

30 29

0

100

200 300 400 500 600 Transmission distance (km)

700

(a)

30 29 28 27 26 25 24 23 22 21 20

47

OSNROSA OSNRNoise

0

80 160 240 320 400 480 560 640 720 Transmission distance (km) (b)

FIGURE 2.28 OSNR of 2.5-Gb/s NRZ signal measured by monitoring beat noise at (a) 2.5-GHz null point or (b) 10 GHz.

residual CD. This problem was resolved by measuring the noise power in the 10-GHz region where the signal spectrum was negligible, as shown in Figure 2.28(b). However, this technique requires RF components that operate at much higher frequencies than the bit rate, which can be a critical issue for high-speed signals with a data rate of >40 Gb/s. The high-frequency noise analysis technique is robust to MPI noises since the beat noises caused by the MPI noises occur at low frequencies.43

2.5.3 Methods to overcome limitations There are several ways to overcome the limitations imposed by the interference between the beat noises and the signal’s spectral components on the OSNR monitoring technique based on the beat noise analysis. For example, the beat noises can be accurately measured if the signal spectrum is eliminated—at least at a certain frequency—without changing the noise density. Polarization- or frequency-diversity detection is a common technique used to achieve this goal.

2.5.3.1 Polarization-diversity technique When a highly polarized optical signal is split into two orthogonally polarized components (by aligning the optical signal into a PBS at a 45-degree angle), the waveform of two outputs is highly correlated. On the other hand, two orthogonally polarized components of the unpolarized ASE noises are completely uncorrelated in nature. Thus, if the signal and the ASE noises are detected by using a polarization-diversity receiver with a subtraction circuit, the signal spectrum can be diminished without affecting the density of the beat noises.47,48 Figure 2.29 shows the schematic diagram of the OSNR monitoring technique based on this principle. A polarization controller and a PBS can split an optical signal into two orthogonally polarized parts with well-balanced power. Subsequently, each component is detected separately and a subtraction circuit cancels out the waveform of two outputs, to obtain the beat noises only. Since the beat noises from the PDs are completely uncorrelated to each other, they are not subtracted but added in terms of power. As a result, it is possible to

48

CHAPTER 2 Optical signal-to-noise ratio monitoring

Total power measurement OSNR calculation

PD PBS PD Polarization controller

Tunable filter

–

Noise measurement

DSP

PD

FIGURE 2.29 Schematic diagram of OSNR monitoring technique based on polarization diversity.

Linear units

Arm 1, Arm 2

0

0.2

0.4

0.6

0.8

0.4 Time (ns)

0.6

0.8

Linear units

Arm 1 + Arm 2, Arm 1 – Arm 2

0

0.2

FIGURE 2.30 Waveforms of 1.25-Gb/s signal. The upper graph shows the outputs of two arms of the polarization-diversity receiver; the lower graph shows the sum and difference of outputs.43 Copyright © 2005 IEEE.

estimate the OSNR of the signal by measuring the power of beat noises without the signal spectrum. Figure 2.30 shows the waveform of a 1.25-Gb/s NRZ signal at the outputs of two arms of the polarization-diversity receiver, together with their sum and difference47; this figure clearly shows the operating principle of this technique. If the low-frequency noises are measured for monitoring the OSNR, the subtraction circuit and noise measurement part can be simply implemented by using a low-speed digital signal processing (DSP) circuit.48 Since this technique can eliminate the signal spectrum ideally over all frequencies, the noise power can be measured at any frequency region. However, it is preferable to measure low-frequency noises, not only because doing so is cost effective, but also because it allows us to negate the influence of PMD.48 If the optical signal is depolarized due to PMD, the signal spectrum in high frequencies can become high at the output of the subtraction circuit, due to the waveform mismatch of the two outputs. This technique is insensitive to CD since the outputs of the polarization-diversity receiver have an identical waveform. However, the effect of nonlinear birefringence (described in

2.5 Beat noise analysis techniques

Polarization-modulated PBS optical signal

49

Intensity-modulated electrical signal PD – PD

FIGURE 2.31 Conversion from polarization variation to intensity variation by polarization-diversity receiver with subtraction circuit.

Section 2.3.2.2) can pose a problem since the polarization-diversity receiver with the subtraction circuit converts polarization variation into intensity variation, as shown in Figure 2.31. The effect of nonlinear birefringence is maximized at low frequencies; thus, the accuracy of noise measurement at low frequencies can be significantly degraded if substantial polarization scattering occurs due to nonlinear birefringence in WDM systems.

2.5.3.2 Orthogonal polarization delayed-homodyne technique The orthogonal polarization delayed-homodyne technique nullifies the signal spectrum for a single frequency, rather than for the entire spectrum.49 For this purpose, as shown in Figure 2.32, an optical signal is split into two polarization components by using a polarization controller and a PBS. Subsequently, two polarizations are recombined via another PBS after the signal on one branch passes through an optical delay. The part consisting of the differential group delay (DGD) and the PD can be equivalently replaced by a polarization-diversity receiver with an electrical delay and an add-circuit. If the polarization controller aligns the optical signal with the first PBS so that two branches have equal optical power, the signal current at the receiver can be expressed as IðtÞ ¼ 0:5FðtÞ þ 0:5Fðt þ DtÞ;

(2.31)

where Dt is the DGD between two polarization components introduced by the optical delay and F(t) is the waveform of the received signal when Dt ¼ 0. Then, by using a Fourier transform, the power spectral density of the received signal can be obtained as 2 ð 2 ð (2.32) Pð f Þ ¼ IðtÞejot dt ¼ ½1  sin2 ðp f D tÞ FðtÞejot dt ; where o is an angular frequency (¼2pf).49,50 Equation (2.32) indicates that the signal spectrum is nullified at frequencies satisfying the condition of sin(pfDt) ¼ 1. Therefore, it is possible to measure the power of the beat noises at these frequencies, as shown in Figure 2.33(a). Figure 2.33(b) shows the electrical power spectra of a 10-Gb/s NRZ signal received after 0- and 62.5-ps DGD.49 The signal spectrum is diminished at 8 GHz, when the signal experiences a 62.5-ps DGD, as expected from Equation (2.32).

50

CHAPTER 2 Optical signal-to-noise ratio monitoring

Total power measurement OSNR calculation

PD PBS

PBS Optical delay

Polarization controller

Tunable filter

PD

BPF fc =

Noise measurement

1 2Δt

PBS PD

Delay +

PD

FIGURE 2.32 Schematic diagram of orthogonal polarization delayed-homodyne technique for monitoring OSNR.

10 dB/div

Power spectral density

PBS

Original signal

Δt

PBS

S(f ) = [1 - sin2(πfΔt)] |F(f )|2

1 2Δt

Frequency (a)

2 GHz

12 GHz (b)

FIGURE 2.33 (a) Principle of nullifying data spectrum using the orthogonal polarization delayed-homodyne technique. (b) Measured RF spectra of 10-Gb/s NRZ signal with and without data spectrum nullifying.

2.5 Beat noise analysis techniques

51

Since the monitoring frequency can be controlled by changing the optical delay, any frequency can be utilized for monitoring the OSNR. However, a high-frequency region is preferred since nullifying a low frequency requires a large amount of DGD. A high-frequency region also has the benefit of a low nonlinear birefringence effect, as described in Section 2.3.2.2. Even if the noise power is measured at high frequencies, this technique is also robust to PMD because it nullifies the signal spectrum for only a single frequency. The orthogonal polarization delayed-homodyne technique is insensitive to CD.

2.5.3.3 Frequency-diversity technique The signal spectrum can be also eliminated by using a frequency-diversity receiver without affecting the profile of the beat noises, as shown in Figure 2.34.51,52 Two optical filters—namely, BPF1 and BPF2, which are centered at different frequencies—filter out the upper or lower sidebands of an optical signal, respectively. If BPF1 and BPF2 are well aligned at the same distance from the center of the signal spectrum and the optical signal has a symmetrical spectrum, the waveform of the outputs of the two BPFs is highly correlated. Thus, it is possible to cancel out the signal spectrum by using a balanced subtraction receiver. On the other hand, the beat noises should not be canceled out if the overlapping of the passband between BPF1 and BPF2 is negligible since the ASE noises from BPF1 and BPF2 are not correlated with each other. Thus, the beat noise power can be measured to estimate the OSNR, regardless of the signal spectrum. Figure 2.35 shows the RF spectra of the signal received with one photodiode or the balanced subtraction receiver.51 By using the balanced subtraction receiver, the signal spectrum is well removed over the whole frequency range. However, high-frequency components can be regenerated if the optical signal experiences CD as shown in Figure 2.36.52 Since the CD induces different amounts of phase delay for the upper and lower sidebands of the optical signal, the high-frequency components from two outputs are not completely in-phase at the receiver. As a result, they cannot be canceled out by using the subtraction circuit; this leads to noise measurements being made at low frequencies, which is preferable. This technique should be free from any polarization effects such as PMD (although a higher-order PMD could create a problem since it induces frequency-sensitive distortion) and nonlinear birefringence since it does not utilize polarization properties. However, Total power measurement OSNR calculation

PD

Modulated optical signal BPF1

Tunable filter

BPF1

– BPF2

BPF2

PD Noise measurement

PD Optical frequency

FIGURE 2.34 Schematic diagram of OSNR monitoring technique based on frequency diversity.

CHAPTER 2 Optical signal-to-noise ratio monitoring

−40

−40

MKR 10.00 GHz 48.83 dBm

−50 −60

Electrical spectrum (dBm/1MHz)

Electrical spectrum (dBm/1MHz)

52

Modulated signal

−70 −80 −90

0

4

8 12 Frequency (GHz)

16

20

−50 MKR 10.00 GHz −60 –71.75 dBm −70 −80 −90

0

4

8 12 Frequency (GHz)

16

20

(b)

(a)

FIGURE 2.35

Uncorrelated noise (dBm)

RF spectra of the signal after (a) one photodiode and (b) balanced subtraction.47

–71

RF spectra

Copyright © 2005 IEEE.

100 km 50 km 25 km

–75

0 km

–79

Receiver bandwidth 0

1

2

3

4 5 6 Frequency (GHz)

7

8

9

10

FIGURE 2.36 Effect of chromatic dispersion on OSNR monitoring technique based on a frequency-diversity receiver.48 Copyright © 2008 IEEE.

it is difficult to use this technique for signals with a nonsymmetrical optical spectrum, such as vestigial sideband (VSB)–filtered signals or single sideband (SSB) orthogonal frequency-division multiplexed (OFDM) signals. In addition, accurate tuning of two optical filters can be very difficult from a practical viewpoint; this is especially true for WDM signals since the filters should be also tuned to select the wavelength channel to be monitored.

2.5.3.4 Orthogonal polarization self-heterodyne technique This technique generates beat noises at a certain RF frequency by making, from an optical signal, two spectral components that are orthogonally polarized to each other and mixing them.53 The optical signal is split into two branches and a spectral component is selected by a narrow BPF in each branch, as shown in Figure 2.37. BPF1 and BPF2 are centered at different frequencies so that uncorrelated ASE noise components can be selected. A polarization controller is used in one branch to

2.5 Beat noise analysis techniques

53

Total power measurement OSNR calculation

PD f1 Narrow BPF 1

Tunable filter

PD

Elec BPF

RFPD

fcenter = Δf Narrow BPF 2

Polarization controller

f2

FIGURE 2.37 Schematic diagram of orthogonal polarization self-heterodyne technique for monitoring OSNR.

ensure orthogonal polarization of the spectral components of the two branches. Dynamic polarization control is required since the polarization state of the signal spectrum can be diverged due to PMD, which is a time-varying random process. Subsequently, the outputs of two branches are combined and received by a PD. Since the signal components are in orthogonal polarization and the ASE noises are completely unpolarized, mixing the two spectral components generates beat noises at around the frequency corresponding to jf1  f2j without the signal spectrum, as shown in Figure 2.38, where f1 and f2 are the center frequencies of BPF1 and BPF2, respectively. After the PD, there is an electrical BPF (EBPF) centered at jf1  f2j and an RF power detector (RFPD) to detect the power of the beat noises. To ensure that the signal spectrum does not get into the frequencies where beat noises occur, the two optical filters should satisfy the following requirement53: Df ¼ jf1  f2 j > ðB2 þ B1 Þ=2;

Filtered signal @ f1

Signal spectrum Δf

Filtered signal @ f2

(2.33)

y-pol beating Signal spectrum y-pol Beat noise

y-pol Pol. Cont. x-pol x-pol

ASE

Optical frequency

PD

x-pol beating

Δf

FIGURE 2.38 Operating principle of orthogonal polarization self-heterodyne technique for OSNR monitoring.

Electrical frequency

54

CHAPTER 2 Optical signal-to-noise ratio monitoring

where B1 and B2 are the bandwidths of BPF1 and BPF2, respectively. It is known that for filters with a 30-dB bandwidth of 2Df, the maximum monitoring error for a 30-dB OSNR is about 3 dB when the signal has an equal power level at the center frequencies of the two filters.53 The optical powers of branches 1 and 2 after narrow filtering can be expressed as P1 ¼ Ps1 þ Ppn1 þ Pon1

(2.34)

P2 ¼ Ps2 þ Ppn2 þ Pon2 ;

(2.35)

and Pp;o n

where Ps is the signal power and is the power of the ASE noise component in the parallel or orthogonal polarization state to the signal component. Then, the electrical power detected after EBPF (i.e., the power of beat noises) is given by PDf ¼ AðPs1 Pon2 þ Ps2 Pon1 þ Ppn1 Pon2 þ Pon1 Ppn2 Þ;

(2.36)

where A is a constant related to the design specifications of the receiver. Assuming that the ASE noises are completely unpolarized (i.e., Ppn1 ¼ Pon1 ¼ Pn1 =2 and Ppn2 ¼ Pon2 ¼ Pn2 =2), we can obtain the power of the ASE noises after BPF1 as qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ðaP1 þ P2 Þ  ðaP1 þ P2 Þ2  8aPDf =A Pn1 ¼ ; (2.37) 2a where a is the ratio of the noise power after BPF2 and BPF1 (Pn2 ¼ aPn1).53 The OSNR can be obtained from Equation (2.23) together with (2.37). This technique is immune to first-order PMD. Figure 2.39(a) shows the relationship between the SOPs of the filtered signals from branches 1 and 2 when the optical signal is depolarized due to the first-order PMD.53 The SOPs of the spectral components distribute along a circle in a Poincare sphere due to PMD. Positions A and B are the initial SOPs of the signals after BPF1 and BPF2, respectively. By rotating the SOP of the signal from BPF2 using the polarization controller, its position can be moved from B to D through C. Thus, after the combining coupler, each pair of spectral components between branches 1 and 2 with frequency difference Df is orthogonally polarized in any combination, which means that the signal-signal beating term can be completely eliminated at Df, even when considerable depolarization of the optical signal occurs due to first-order PMD. However, the signal-signal beating term can increase due to PMD as the frequency diverges from Df, as shown in Figure 2.39(b). Thus, to be insensitive to PMD effects, the bandwidth of the EBPF should be sufficiently narrow so as not to include any signal-signal beating term. Despite the inherent insensitivity of this system to first-order PMD, a higher-order PMD can pose a problem by inducing a different amount of depolarization on the spectral components from two branches. If there is no strong interaction between PMD and nonlinear birefringence within the link, all the spectral components experience the same amount of nonlinear polarization scattering. Thus, by accurately balancing the delay of the two branches, the orthogonality of the polarizations between the spectral components from the two branches can be preserved, regardless of polarization scattering; this feature makes this technique insensitive to nonlinear birefringence.

2.5 Beat noise analysis techniques

55

S3 f2 + B2/2

A

f2 – B2/2

B

f1 – B1/2

f1 + B1/2 f2 – B2/2 f2 + B2/2 θ θ f2 + B2/2 D f2 – B2/2

S2

Signal-signal beating term

Electrical power density

C

Δf -

Signal-ASE beating term

B1 + B2 2

Δf

Δf +

B1 + B2 2

Electrical frequency

(a)

(b)

FIGURE 2.39 (a) Relation of polarization states of the signals in branches 1 and 249. (b) Effect of PMD on electrical spectrum. Copyright © 2007 IEEE.

2.5.3.5 Technique based on synchronously gated signal Figure 2.40(a) shows the operating principle of the OSNR monitoring technique using RF spectral analysis of a gated framing signal.54 A gating device such as an intensity modulator allows only the repetitive patterns in each header to pass through, and it suppresses the other parts of frames, including payload data, to reveal the noise spectrum for monitoring the OSNR. For this purpose, the gating device should be modulated by the gating signal synchronized to the frame headers. A high extinction ratio operation of the gating device is required for achieving a wide dynamic range when monitoring the OSNR. Figure 2.40(b) shows the RF spectrum of a received signal carrying synchronous traffic, which consists of repetitive short patterns in the header and PRBS with a pattern length of 223  1 in the payload. The signal spectrum prevents accurate measurement of the beat noise density at low frequencies, as described in Section 2.5.2. On the other hand, the noise spectrum is revealed by gating synchronous traffic so that only the repetitive patterns in the headers are allowed through, as shown in Figure 2.40(c). This technique can be applied to any optical signal carrying synchronous traffic with a repetitive header structure, such as synchronous optical network or synchronous digital hierarchy (SONET/SDH) frames. However, it seems difficult to acquire a gating signal if the OSNR needs to be monitored at a transparent optical node where there is no optical-electrical-optical (OEO) conversion. In addition, during the monitoring process, a large optical power is required since only a small portion of the optical signal and ASE noises are utilized for monitoring after the gating device. This technique is insensitive to polarization effects since it does not utilize any polarization property.

56

CHAPTER 2 Optical signal-to-noise ratio monitoring

Framing synchronized pulses

Payload

ON

OFF

Tunable filter

PD

Gating

AD converter

PD

Spectral tones of payloads Df = 1/(223–1)Tb

–200

Spectral tones of frame headers

Noise

–100 0 100 Frequency (kHz)

FFT

OSNR calculation

(a)

Amplitude (a.u.)

Amplitude (a.u.)

Frame-synchronized gating pulses

200

–200

Frame spectral spacing Df = 1/Tframe Noise

–100 0 100 Frequency (kHz)

(b)

200

(c)

FIGURE 2.40 (a) Schematic diagram of OSNR monitoring technique based on synchronously gated signal. RF spectra of signal (b) without gating pulses and (c) with gating pulses.54 Copyright © 2006 OSA.

2.6 OSNR ESTIMATION TECHNIQUE BASED ON THE OPERATING CONDITION OF OPTICAL AMPLIFIERS Instead of measuring the OSNR directly from an optical signal itself, it is also possible to estimate it based on the noise characteristics of erbium-doped fiber amplifiers (EDFAs) and their operating conditions like input power and gain.55 This technique is based on a well-established OSNR calculation method that has played a key role in the design of optical transmission systems for many years.56 In this section, we review the operating principle of this technique and the link-based OSNR monitoring technique based on it.

2.6.1 Operating principle If the OSNR of the optical signal is infinite at the EDFA input, its OSNR (OSNREDFA) at the EDFA output is given by OSNREDFA ¼

Pin G ; 2nsp hvðG  1ÞBo

(2.38)

2.6 OSNR estimation technique

57

where Pin is the signal power at the EDFA input, G is the gain of the EDFA, nsp is the population inversion parameter, h is Plank’s constant, n is the optical frequency, and Bo is the optical bandwidth that measures the OSNR (usually 0.1 nm).56 Taking logarithms of both sides of Equation (2.38) gives (and assuming G 1) OSNREDFA ðdBÞ ¼ Pin ðdBmÞ þ OSNR0 ;

(2.39)

where OSNR0 is a constant related to the noise characteristics of the EDFA. Thus, if OSNR0 is measured and is not too sensitive to the operating conditions of the EDFA, OSNREDFA can be estimated simply by monitoring the input signal power (Pin) to the EDFA. In other words, for WDM signals, the channel power monitoring at the input of every EDFA makes it possible to estimate the OSNRs of all channels simultaneously. The constant OSNR0 usually has a wavelength-dependent value. Thus, a set of OSNR0 values measured at multiple wavelengths would be required to accurately determine the wavelength dependency of OSNREDFA.55 For an optical link consisting of multiple EDFA spans, the OSNRs (OSNRlink) of the WDM signals at the output of the link can be obtained as follows based on the OSNREDFA of each channel of all EDFAs: , N

X  1=OSNREDFA;i ðlÞ ; (2.40) OSNRlink ðlÞ ¼ 1 i¼1

where l is the channel wavelength and N is the number of EDFAs comprising the optical link. The monitored OSNRlink(l)s of all links comprising the light path of a certain wavelength channel can be used in exactly the same way to estimate the end-to-end OSNR of the channel as , ! M X 1=OSNRinit þ ð1=OSNRlink; j Þ ; (2.41) OSNRend ¼ 1 j¼1

where OSNRinit is the initial OSNR of the channel at the ingress node, OSNRlink,j is the OSNRlink of the channel at the jth link, and M is the number of optical links comprising the light path.55

2.6.2 Link-based OSNR monitoring technique The estimation of OSNRlink(l) using Equations (2.39) and (2.40) requires the optical power of each channel to be monitored at the input of every EDFA, which increases the monitoring cost. This problem can be overcome by restricting the points of channel power monitoring to only the input and output of the link,55,57 as shown in Figure 2.41. Optical channel monitors (OCMs) can be used for channel power monitoring at the input and output of the link. Only total input power is monitored by using simple power detectors for the in-line EDFAs. Subsequently, the channel powers at the input of the in-line EDFAs are estimated based on the measured total input power and the channel power information obtained at the input and output of the link. In this case, a large gain-tilt variation among the EDFAs can give rise to errors in channel power estimation and, in turn, lead to errors in OSNR estimation.55 This technique is insensitive to CD and any polarization effects since it monitors only optical powers. This technique calculates the OSNR of an optical signal based on the channel powers measured at every EDFA through which the signal passes. Thus, it is necessary to aggregate all the power

58

CHAPTER 2 Optical signal-to-noise ratio monitoring

Link 1

Link 2 Optical node

Link 3 Optical node

OCM

PM

PM

OCM

FIGURE 2.41 Schematic diagram of OSNR monitoring technique based on operating condition of optical amplifiers. OCM, optical channel monitor; PM, power monitor.55 Copyright © 2008 IEEE.

information via a communication channel such as the management channel. This can be achieved on a link basis, as shown in Figure 2.42(a).57 The optical performance monitor (OPM) located at the output of the link periodically retrieves sets of power information from OCMs and EDFAs via the management channel and calculates the OSNRlink of each wavelength channel. Subsequently, the OPM provides the performance parameter of each channel (i.e., OSNRlink), which is utilized for Optical node Optical OCM node

OCM OPM

Management plane

(a) OPM manager Result viewer (GUI)

Timer

Management of Update SNMP link and LSP manager performance LSP Restoration information trigger CLI to GMPLS controller

SNMP trap receiver

SNMP GET SNMP M -plane trap

Remote OPM Remote OPM

GMPLS controller

Operator C-plane

GMPLS controller

(b)

FIGURE 2.42 (a) Schematic diagram of OSNRlink monitoring. (b) Schematic diagram of OPM manager57 GUI, graphic user interface; LSP, label-switched path; CLI, command line interface; GMPLS, generalized multiprotocol label switching; SNMP, Simple Network Management Protocol. Copyright © 2009 IEEE.

2.6 OSNR estimation technique

59

end-to-end OSNR estimation. Figure 2.42(b) shows a block diagram of the OPM manager, which can be a part of a network management system (NMS), to manage distributed OPMs and light paths. The OPM manager retrieves the performance information of each link from remote OPMs and updates a database that contains the monitored performance of optical links and established light paths. The OPM manager can periodically retrieve data according to an internal timer or occasionally by event traps sent from remote OPMs. The OPM manager calculates the end-to-end OSNR of every label-switched path (LSP) based on the link constituents of LSPs and the OSNRlink values of corresponding links.

2.6.3 Potential problems and limitations Since this technique estimates the OSNR of optical signals based on signal powers, it needs to aggregate power-monitoring information from the entire the network through the management channel. Thus, failure of the management channel or a few power monitors could limit the operation of the OSNR monitoring. In addition, it cannot estimate the OSNR of an alien signal from other networks if the originating network does not provide the final OSNR of the signal. This technique also has accuracy limitations that are caused by total power monitoring and the gain-tilt variation among the EDFAs, which will be discussed in this section in greater detail.

2.6.3.1 Limitations due to total power monitoring The proposed technique estimates the channel powers of WDM signals based on the total power, instead of monitoring each channel at the input of in-line EDFAs. Although this scheme could be used to substantially reduce monitoring costs, the channel powers could be overestimated due to accumulated broadband optical noise if the number of wavelength channels is very small. Consequently, the link OSNR could be overestimated. Figure 2.43(a) shows the calculated monitoring error caused by this problem during the estimation of the link OSNR.55 The monitoring error should increase as the number decreases since the sum of channel powers becomes comparable to the power of optical noise. The monitoring error also increases with the number of spans per link since optical noise is invariably accumulated. On the other hand, the monitoring of total power in the subsequent link is not seriously affected by optical noise that accumulated in previous links since the majority of the optical noise, except the in-band noise under the channels, should be removed by optical filtering (via demultiplexing and multiplexing) at each node. Thus, it is possible to calibrate out the noise power from the total power by using monitored sets of OSNREDFA within the link. Figure 2.43(b) shows that the monitoring error could be substantially reduced when 50% of the noise power is calibrated out. For example, the monitoring error would be <1 dB even if three channels are transmitted over ten spans. More accurate calibration could further reduce the error.

2.6.3.2 Limitations due to gain-tilt variation of EDFAs The proposed technique assumes that all EDFAs within a link have identical gain-tilt. Thus, variation of gain-tilt among EDFAs could result in errors in OSNR estimation. Figure 2.44 shows the limitation imposed by gain-tilt variation.55 The maximum OSNR estimation error for the worst-case scenario was plotted as a function of the gain-tilt variation and the number of EDFAs comprising the link. As the number of EDFAs or the gain-tilt variation increased, the monitoring error was exacerbated. Under normal operating conditions, as shown in Figure 2.44(a), the product of the gain-tilt

CHAPTER 2 Optical signal-to-noise ratio monitoring

Link OSNR (dB) 24 22

Number of channels

28 26 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

20

Number of channels

60

0.50 dB

1.0 dB 1.5 dB 2.0 dB 1

2

3

2.5 dB

4 5 6 7 8 Number of spans / link (a)

9

15 14 13 12 11 10 9 8 7 6 5 4 3 2 1

20

0.50 dB 1.0 dB 1.5 dB 1

10

Link OSNR (dB) 24 22

28 26

2

3

4 5 6 7 8 Number of spans / link (b)

9

10

FIGURE 2.43 Monitoring error of link OSNR caused by total power monitoring (a) without calibration and (b) with 50% of optical noise power calibrated (contour plots). Span loss ¼ gain of EDFA ¼ 20 dB; noise figure of EDFA ¼ 8 dB; optical noise bandwidth ¼ 30 nm; input power/channel to fiber ¼ 0 dBm.55 Copyright © 2009 IEEE.

4.0

4.0 >2.0 dB

3.0

3.5 Gain-tilt variation (dB)

Gain-tilt variation (dB)

3.5

1.5 dB

2.5

1.0 dB

2.0 1.5

0.5 dB

1.0 0.5

>2.0 dB

3.0 2.5

1.5 dB

2.0

1.0 dB

1.5 1.0

0.5 dB

0.5

0.0 1

2

3

4 5 6 7 Number of EDFAs (a)

8

9

10

0.0 1

2

3

4 5 6 7 8 Number of EDFAs/link (b)

9

FIGURE 2.44 Maximum OSNR monitoring error in the worst-case scenario for optical link with AGC-EDFAs (a) under normal operating conditions, or (b) with an increase of 10 dB in span loss at worst position (contour plots).55 Copyright © 2009 IEEE.

10

2.7 Summary

61

variation and the number of EDFAs (i.e., gain-tilt variation  the number of EDFAs) was estimated to be 14 dB for a monitoring error of 1 dB. For example, when the optical link consisted of five EDFAs, the tolerable gain variation was about 2.8 dB (i.e., 1.4 dB). On the other hand, the tolerable gain-tilt variation was reduced to 1.4 dB when the number of EDFAs was increased to 10. Figure 2.44(b) shows the maximum OSNR estimation error calculated when a 10-dB increase in span loss was considered in the span where the worst monitoring error occurs (severe increases of >10 dB in the loss were not considered since such a large loss should initiate a power-monitor alarm). The monitoring error was found to increase slightly. This was because OSNR degradation could be dominantly devoted by several EDFAs, which had a significant difference between the actual channel powers and the estimated channel powers at the input of the EDFAs. For the optical link with five EDFAs, the tolerable gain variation was reduced to 2.2 dB for a monitoring error of 1 dB. However, the problem caused by the significant gain-tilt variation can be solved easily by adding only one or two OCMs at the intermediate EDFAs.

2.7 SUMMARY For the proper operation and maintenance of dynamic optical networks, it is essential to monitor the OSNR of each channel accurately. This is because the OSNR can be used to assess the signal’s quality directly in the optical layer without requiring OEO interface. However, the conventional linear interpolation technique based on optical spectrum analysis cannot measure the true value of OSNR (i.e., in-band OSNR) in such dynamic optical networks. To solve this problem, a number of techniques have been proposed to date; most of them exploit the fact that ASE noises are unpolarized and incoherent. This chapter reviewed some of these techniques and offered detailed descriptions of their operating principles, performance limitations caused by various physical effects, and potential means to overcome the limitations and thereby improve performance. The polarization-nulling technique and the DOP analysis technique are based on the assumption that the signals are highly polarized, while the ASE noises are completely unpolarized. However, in practice, it is quite possible that this assumption may not hold because of various polarization effects that occur within the transmission links. For example, the signal could be depolarized by PMD and nonlinear birefringence, and the ASE noises could be partially polarized due to PDL. In addition, for use in aerial fiber links, these techniques should be able to track rapid fluctuations in polarization that are caused by wind and electric currents in neighboring power lines. Several techniques have been developed to overcome some of these problems. For example, improved polarization-nulling techniques either calibrated out the small amount of signal power that leaks into the noise in the orthogonal polarization state (due to PMD or nonlinear birefringence) by using an additional optical filter or measure the noise power at the slope of the signal’s spectrum after narrow filtering. The DOP analysis technique measured the spectral SOPs of the optical signal to separate the contributions of PMD and ASE noises to the decreased DOP. The polarization-based techniques were found to be robust against PDL (as long as the PDL/span was <0.2 dB) and rapid polarization fluctuations (which occurred in the aerial fiber link). The interferometer-based technique exploits the fact that ASE noises are incoherent whereas optical signals are coherent. Thus, this technique measured the noise power after eliminating the signal component using the MZDI with destructive interference. However, this technique could not accurately

62

CHAPTER 2 Optical signal-to-noise ratio monitoring

monitor the in-band OSNR since the ASE noises could also be seriously affected by destructive interference (especially when we needed to monitor the OSNR of the WDM signals with very high spectral efficiency). It is possible to avoid this problem by interfering two orthogonal polarization components destructively after combining them into the same polarization state. In this case, the ASE noises could be separated from the optical signal without any change despite of the destructive interference of the optical signal. The techniques based on beat noise analysis estimate the in-band OSNR by analyzing the beat noises caused by beating between the optical signal and ASE noises at the square-row detector. Thus, the noise density in the low-frequency region can be monitored by using a low-speed PD and electronics as long as the optical signal is not modulated (or modulated with a signal having a pattern length of 215  1). However, in cases where the optical signal is modulated with a long repetitive (or random) pattern, this technique cannot accurately measure noise density in the low-frequency region, due to the interference between the beat noises and the signal component. This problem could be solved by measuring the beat noises in the frequency region much higher than the signal’s bit rate, but doing so would require the use of expensive components such as high-speed PD and electronics. Several research groups have attempted to mitigate this problem by eliminating the signal’s spectral components in the low- or medium-frequency regions (without affecting the noise density at the receiver). For example, polarization-diversity and spectral-diversity receiver techniques could be used to achieve this goal as they cancel out the signal components from two branches without changing the beat noises. The orthogonal polarization delayed-homodyne technique could also make the signal spectrum disappear at a certain frequency for in-band monitoring of the OSNR. It would also be possible to remove the signal’s spectral components in the low-frequency region by suppressing the random patterns while leaving the repetitive ones by using the synchronously gated signal. On the other hand, the beat noises could be generated in the frequency region where there is no signal component by using the orthogonal polarization self-heterodyne technique. The OSNR estimation technique deduces the OSNR by measuring only the signal powers (assuming that the noise characteristics of the optical amplifiers are known). The link-based OSNR monitoring technique estimates the OSNR on a per-link basis, instead of the conventional per-channel basis. Thus, the amount of ASE noises that accumulates in each link is measured separately. Subsequently, a centralized OPM manager retrieves the noise information of each link and updates the end-to-end OSNR of the optical signal of each established light path. Although this technique merely calculates the OSNR instead of measuring it, it offers the possibility of accurately monitoring the in-band OSNRs of WDM signals, even in dynamic optical networks.

REFERENCES 1. Agrawal GP. Fiber-optic communication systems. 3rd ed. New York: John Wiley & Sons; 2002. p. 226–78. 2. Hill GR, Chidgey PJ, Kaufhold F, Lynch T, Sahlen O, Gustavsson M, et al. A transport network layer based on optical network elements. J Lightwave Technol 1993;11(5):667–79. 3. Chung YC. Optical performance monitoring techniques. In: European conference on optical communications, paper We.1.D.1. Brussels; September 2008. 4. Kilper DC, Bach R, Blumenthal DJ, Einstein D, Landolsi T, Ostar L, et al. Optical performance monitoring. J Lightwave Technol 2004;22(1):294–304.

References

63

5. Optical Monitoring for DWDM Systems. ITU-T recommendation G.697; June 2004. 6. Derickson D. Optical signal analysis. In: Fiber optic test and measurement. Upper Saddle River, NJ: Prentice Hall; 1997. p. 87–130. 7. Otsuka K, Maki T, Sampei Y, Tachikawa Y, Fukushima N, Chikama T. A high-performance optical spectrum monitor with high-speed measuring time for WDM optical networks. In: European conference on optical communications. Edinburgh; September 1997. p. 147–9. 8. Asahi K, Yamashita M, Hosoi T, Nakaya K, Konoshi C. Optical performance monitor built into EDFA repeaters for WDM networks. In: Optical fiber communication conference, paper THO2. San Jose, CA; 1998. 9. Suzuki H, Takachio N. Optical signal quality monitor built into WDM linear repeaters using semiconductor arrayed waveguide grating filter monolithically integrated with eight photodiodes. Electron Lett 1999;35(10):836–7. 10. Eiselt MH, Garrett LD. Simple method to measure OSNR in transmission systems with channel filtering. Electron Lett 2006;42(7):409–10. 11. Kilper DC, Chandrasekhar S, Buhl L, Agarwal A, Maywar D. Spectral monitoring of OSNR in high speed networks. In: European conference on optical communications, paper 7.4.4. Copenhagen; 2002. 12. Rasztovits-Wiech M, Danner M, Leeb WR. Optical signal-to-noise ratio measurement in WDM networks using polarization extinction. In: European conference on optical communications. Madrid; 1998. p. 549–50. 13. Jung DK, Kim CH, Chung YC. OSNR monitoring technique using polarization-nulling method. In: Optical fiber communication conference, paper WK4. Baltimore; 2000. 14. Lee JH, Jung DK, Kim CH, Chung YC. OSNR monitoring technique using polarization-nulling method. IEEE Photon Technol Lett 2001;13(1):88–90. 15. Lee JH, Chung YC. Improved OSNR monitoring technique based on polarization-nulling method. Electron Lett 2001;37(15):972–3. 16. Chung YC, Shin SK, Kim CH. OSNR monitoring method and apparatus using tunable optical bandpass filter and polarization-nulling method. U.S. patent 20040114923; June 17, 2004. 17. Cheung M-H, Chen L-K, Chan C-K. PMD-insensitive OSNR monitoring based on polarization-nulling with off-center narrow-band filtering. IEEE Photon Technol Lett 2004;16(11):2562–4. 18. Yan L-S, Yao XS, Shi Y, Willner AE. Simultaneous monitoring of both optical signal-to-noise ratio and polarization-mode dispersion using polarization scrambling and polarization beam splitting. J Lightwave Technol 2005;23(10):3290–4. 19. Lu G-W, Cheung M-H, Chen L-K, Chan C-K. Simple PMD-insensitive OSNR monitoring scheme assisted by transmitter side polarization scrambling. Opt Expr 2006;14(1):58–62. 20. Lee JH, Choi HY, Shin SK, Chung YC. A review of the polarization-nulling technique for monitoring optical signal-to-noise ratio in dynamic WDM network. J Lightwave Technol 2006;24(11):4162–71. 21. Petersson M, Sunnerud H, Karlsson M, Olsson B-E. Performance monitoring in optical networks using Stokes parameters. IEEE Photon Technol Lett 2004;16(2):686–8. 22. Petersson M, Sunnerud H, Olsson B-E, Karlsson M. Optical performance monitoring using degree of polarization in presence of polarization-mode dispersion. In: European conference on optical communications, paper Tu3.6.2. Stockholm; 2004. 23. Skold M, Olsson B-E, Sunnerud H, Karlsson M. PMD-insensitive DOP-based OSNR monitoring by spectral SOP measurements. In: Optical fiber communication conference, paper OThH3. Anaheim, CA; 2005. 24. Skold M, Olsson B, Karlsson M, Andrekson PA, Swahn T, Martensson J, et al. Field trial of polarizationassisted optical performance monitoring operating in an 820 km WDM system. In: European conference on optical communications, paper Th.2.5.1. Cannes; 2006.

64

CHAPTER 2 Optical signal-to-noise ratio monitoring

25. Derickson D. Polarization measurement In: Fiber optic test and measurement. Upper Saddle River, NJ: Prentice Hall; 1997. p. 220–45. 26. Eickhoff W, Yen Y, Ulrich R. Wavelength dependence of birefringence in single-mode fiber. Appl Opt 1981;20(19):3428–35. 27. Motaghian Nezam SMR, McGeehan JE, Willner AE. Theoretical and experimental analysis of the dependence of a signal’s degree of polarization on the optical data spectrum. J Lightwave Technol 2004; 22(3):763–72. 28. Phillips MR, Ott DM. Crosstalk due to optical fiber nonlinearities in WDM CATV lightwave systems. J Lightwave Technol 1999;17(10):1782–92. 29. Phillips MR, Woodward SL. Cross-polarization modulation: theory and measurement of a two-channel WDM system. IEEE Photon Technol Lett 2005;17(10):2086–8. 30. Xie C, Mo¨ller L, Kilper DC, Mollenauer LF. Impact of crossphase modulation induced polarization scattering on optical PMD compensation in WDM systems. Opt Lett 2003;28(23):2303–5. 31. Chiang TK, Kagi N, Marhic ME. Cross-phase modulation in fiber links with multiple optical amplifiers and dispersion compensators. J Lightwave Technol 1996;14(3):249–60. 32. Feuer MD. Measurement of OSNR in the presence of partially polarized ASE. IEEE Photon Technol Lett 2005;17(2):435–7. 33. Lee JH, Chung YC. Effect of polarization-dependent loss on optical signal-to-noise ratio monitoring technique based on polarization-nulling method. Opt Expr 2006;14(12):5045–9. 34. Wuttke J, Krummrich PM, Rosch J. Polarization oscillations in aerial fiber caused by wind and power-line current. IEEE Photon Technol Lett 2003;15(6):882–4. 35. Kim CH, Lee YB, Ji SK, Choi SD, Park MW, Kim MS, et al. Performance of OSNR monitor based on polarization-nulling technique in aerial fiber link. In: Optical fiber communication conference, paper FF3. Los Angeles; 2004. 36. Kim CH, Lee YB, Ji SK, Choi SD, Hamn SP, Kim MS, et al. Performance of an OSNR monitor based on the polarization-nulling technique. J Opt Network 2004;3(6):388–95. 37. Choi HY, Lee JH, Jun SB, Chung YC, Shin SK, Ji SK. Improved polarization-nulling technique for monitoring OSNR in WDM network. In: Optical fiber communication conference, paper OThP2. Anaheim, CA; 2006. 38. Tao Z, Chen Z, Fu L, Wu D, Xu A. A novel method to monitor OSNR using a Mach-Zehnder interferometer. In: Conference on lasers and electro optics, paper ThB2-4. Baltimore; 2001. 39. Lin X, Li Yan. Multiple-channel OSNR monitoring with a single detector. IEEE Photon Technol Lett 2004;16(12):2637–9. 40. Ku Y-C, Chan C-K, Chen L-K. A novel robust OSNR monitoring technique with 40-dB dynamic range using phase modulator embedded fiber loop mirror. In: Optical fiber communication conference, paper OWN6. Anaheim, CA; 2006. 41. Liu X, Kao Y-H, Chandrasekhar S, Kang I, Cabot S, Buhl LL. OSNR monitoring method for OOK and DPSK based on optical delay interferometer. IEEE Photon Technol Lett 2007;19(15):1172–4. 42. Shin SK, Park KJ, Chung YC. A novel optical signal-to-noise ratio monitoring technique for WDIM networks. In: Optical fiber communication conference, paper WK6. Baltimore; 2000. 43. Dorrer C, Liu X. Noise monitoring of optical signals using RF spectrum analysis and its application to phase-shift-keyed signals. IEEE Photon Technol Lett 2004;16(7):1781–3. 44. Choi HY, Takushima Y, Chung YC. OSNR monitoring technique for DPSK/DQPSK signals based on selfheterodyne detection. IEEE Photon Technol Lett 2008;20(13):1124–6. 45. Lee JH, Yoshikane N, Tsuritani T, Otani T. Dynamic monitoring of physical link performance for path computation in transparent optical networks. In: Opto-electronics and communications conference, paper WeK4. Yokohama; 2008.

References

65

46. Youn CJ, Shin SK, Park KJ, Chung YC. OSNR monitoring technique using high-frequency. In: Asia-pacific optical and wireless communications, paper 4584-03. Beijing; 2001. 47. Dods SD, Anderson TB. In-band optical signal-to-noise monitoring technique using polarization diversity. In: The 18th annual meeting of the IEEE lasers and electro-optics society, paper WEE2. Sydney; 2005. 48. Anderson TB, Clarke K, Dods SD, Bakaul M. Robust, low cost, in-band optical signal to noise monitoring using polarization diversity. In: Optical fiber communication conference, paper OMM3. Anaheim, CA; 2007. 49. Youn CJ, Park KJ, Lee JH, Chung YC. OSNR monitoring technique based on orthogonal delayedhomodyne method. IEEE Photon Technol Lett 2002;14(10):1469–71. 50. Ishikawa G, Ooi H. Polarization-mode dispersion sensitivity and monitoring in 40-Gbit/s OTDM and 10-Gbits/s NRZ transmission experiments. In: Optical fiber communication conference, paper WC5. San Jose, CA; 1998. 51. Chen W, Tucker RS, Yi X, Shieh W, Evans JS. Optical signal-to-noise ratio monitoring using uncorrelated beat noise. IEEE Photon Technol Lett 2005;17(11):2484–6. 52. Bakaul M. Low-cost PMD-insensitive and dispersion tolerant in-band OSNR monitor based on uncorrelated beat noise measurement. IEEE Photon Technol Lett 2008;20(11):906–8. 53. Xie C, Kilper DC, Mo¨ller L, Ryf R. Orthogonal-polarization heterodyne OSNR monitoring insensitive to polarization-mode dispersion and nonlinear polarization scattering. J Lightwave Technol 2007; 25(1):177–83. 54. Tian X, Su Y, Hu W, Leng L, Hu P, Yi L, et al. Nonlinearity-tolerant in-band OSNR monitoring for synchronous traffic using gated-signal RF spectral analysis. In: Optical fiber communication conference, paper OThP1. Anaheim, CA; 2006. 55. Lee JH, Yoshikane N, Tsuritani T, Otani T. In-band OSNR monitoring technique based on link-by-link estimation for dynamic transparent optical networks. J Lightwave Technol 2008;26(10):1217–25. 56. Zyskind JL, Barry R, Pendock G, Cahill M, Ranka J. High capacity, ultra-long-haul networks. In: Kaminow T, Li T, editors. Optical fiber communications IVB. San Diego, CA: Academic; 2002. 57. Lee JH, Guo H, Tsuritani T, Yoshikane N, Otani T. Field trial of all-optical networking controlled by intelligent control plane with assistance of optical performance monitors. J Lightwave Technol 2009; 27(2):94–100.

This page intentionally left blank

CHAPTER

Chromatic dispersion monitoring

3 Zhongqi Pan

Department of Electrical and Computer Engineering, University of Louisiana at Lafayette

3.1 INTRODUCTION In general, optical impairments can be classified into catastrophic and noncatastrophic problems. Catastrophic problems tend to decrease the optical power and include individual or multiple component malfunctions, improperly installed or configured equipment, fiber breaks, and damage or intrusion to the network. Impairments due to such faults are as diverse as the components and network designs deployed in the field. In addition to faulty outages, there are many other well-known effects that are always present and must be minimized or controlled.1 These might be classified as “noncatastrophic” problems, in which there is sufficient optical signal power but that the data bits themselves are unrecoverable due to various linear and nonlinear effects. Chromatic dispersion (CD) is one of the most basic characteristics of fiber and can cause noncatastrophic problems. More importantly, its effect increases as the square of the bit rate increases for intensity modulation with direct detection (IM/DD) systems. This makes CD significantly more important for
40-Gb/s systems than for lower-rate systems.2,3 Any slight inaccuracies in dispersion management will result in severe power penalties. Several scenarios may require monitoring of accumulated dispersion on a fiber channel, including (1) due to repair and maintenance, the link length or fiber type itself may change; (2) CD has a temperature dependence that causes a change in net link dispersion; (3) future intelligent networks may perform restoration and protection in the optical domain, and then CD should be monitored and compensated after each reconfiguration; (4) any wavelength drifts of the transmitters could result in chirp of a signal through an optical filter, such that the chirp will interact with the fiber dispersion; and (5) since tunable electrical or optical modules might be used for managing accumulated CD for
40-Gb/s telecommunication networks, it might be valuable to implement some method of dynamic monitoring to drive an optical or electrical compensator.3–5 The monitoring techniques can be analog or digital.1,4 Digital techniques use high-speed electronic logic to process digital information encoded on the optical waveform. Measurements on the digital signal are used to infer characteristics of the optical signal. Digital methods have the strongest correlation with the bit error rate (BER), but are usually less effective at isolating the effects of individual impairment. Analog measurement techniques treat the optical signal as an analog waveform and attempt to measure specific characteristics of this waveform. These measurements are typically protocol-independent and can be subdivided further into time-domain methods or spectral methods. © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00003-1

67

68

CHAPTER 3 Chromatic dispersion monitoring

Time-domain monitoring includes eye diagram measurements and auto- or cross-correlation measurements. Spectral methods must be broken down into optical spectrum and amplitude power spectrum (radio frequency, RF) measurements. The optical spectrum is conveniently measured using highly sensitive optical techniques and can provide optical noise information. The amplitude power spectrum is a better measure of signal quality because it measures the spectrum of the signal that is encoded on the optical carrier. Noise and distortion on the amplitude power spectrum will usually directly translate to impairments on the signal. This chapter will address the issues surrounding the monitoring of chromatic dispersion in highperformance optical systems. In Section 3.2, we will describe the basics of fiber’s CD and its effects on communication systems. Section 3.3 includes discussions on monitoring of accumulated CD, including various techniques in both optical and electrical domains. The chapter ends with a brief discussion and comparison of various monitoring approaches. It is worth mentioning selected recent trends on ultra-high-speed (over 100-Gb/s) and highspectral-efficiency (>2-bit/s/Hz) applications in fiber communications. One emerging technique is using coherent detection associated with digital signal processing that allows compensating for most fiber linear impairments such as CD and polarization mode dispersion (PMD).6,7 Optical orthogonal frequency-division multiplexing (OFDM) has also been proposed and demonstrated to ease the transmission limits induced by CD and PMD.8,9 While this chapter mostly will focus on binary modulation schemes and direct detection systems, coherent and OFDM systems are covered in other chapters in this book. The reader should review these chapters for more in-depth information about these areas.

3.2 CHROMATIC DISPERSION AND ITS EFFECTS ON OPTICAL FIBER SYSTEMS This section will first explain the fundamentals of fiber’s CD. Second, issues are addressed surrounding how CD affects system design in the presence of fiber’s nonlinearity for WDM systems. Finally, the need for CD monitoring to enable robust optical systems will be discussed.

3.2.1 Fiber chromatic dispersion In any medium other than vacuum and in any waveguide structure (other than ideal infinite free space), the speed at which different electromagnetic frequencies travel varies. This is the essence of fiber CD. In fact, there are many kinds of dispersion among optical fibers and each of them may work in different ways. This section focuses on the fiber systems using single-mode fibers (SMFs) that have only one fundament mode (LP01). CD (or intramodal dispersion), a combination of material and waveguide dispersion, is considered as one of the most important characteristics of a SMF.10 In a SMF, the refractive index at different wavelengths is slightly different, and therefore each wavelength will travel at a slightly different speed. This is essence of material dispersion. Material dispersion can be changed by varying the composition of the glass in the fiber core and cladding. In contrast, waveguide dispersion is caused by the shape or waveguide structure and refractive index profile of the fiber core and cladding. Therefore, it can be controlled by careful design of the index profile and can be used to counteract material dispersion. In conventional SMF, material dispersion

3.2 Chromatic dispersion and its effects on optical fiber systems

69

30 Material dispersion

Dispersion [ps/(km·nm)]

20 10

Total dispersion

0 Waveguide dispersion

–10 –20 –30 1.1

1.2

1.3 1.4 1.5 Wavelength (mm)

1.6

1.7

FIGURE 3.1 Dispersion coefficient, D, as a function of wavelength in conventional silica single-mode fiber.11 Copyright © 2003 IEEE.

and waveguide dispersion tend to cancel each other in wavelengths close to 1310 nm. CD is negative for short wavelengths and positive for longer wavelengths. The units of CD are picoseconds per nanometer per kilometer, meaning that wider frequency spread and longer fiber lengths will each contribute linearly to temporal dispersion. Figure 3.1 shows the dispersion coefficient, D (ps/(nm  km)), of a conventional SMF with the material and waveguide contributions plotted separately.11,12 Considering that CD in optical fibers is due to the frequency-dependent nature of the propagation characteristics for both the material and the waveguide structure, the speed of light of a particular wavelength l can be expressed as follows using a Taylor series expansion of the value of the refractive index as a function of the wavelength: co co ¼ : (3.1) vðlÞ ¼ n ð lÞ @n @2n 2 no ðlo Þ þ dl þ 2 ðdlÞ @l @l Here, co is the speed of light in vacuum, lo is a reference wavelength, and the terms in @n=@l and @ 2 n=@l2 are associated with CD and the dispersion slope (i.e., the variation of CD with wavelength), respectively. As mentioned earlier, since material dispersion depends on the composition of the glass, and waveguide dispersion depends on the geometry of the waveguide, we can then change the fiber doping and core (which effectively changes the waveguiding geometry) to change both material and waveguide dispersion and thus move the dispersion zero point to other wavelengths, such as 1550 nm. Figure 3.2 shows the dispersion curve of one of these fibers, known as dispersion-shifted fiber (DSF).13 DSF, which has both the dispersion zero and the loss minimum located at 1550 nm, can be used to attempt the best design for single-channel transmissions. In fact, CD is not always bad for fiber transmission. It is necessary for the deployment of wavelength-division multiplexed (WDM) systems. When the fiber dispersion is near zero in a

70

CHAPTER 3 Chromatic dispersion monitoring

Dispersion (ps/nm/km)

20 10

1300-nm optimized 1400

0

1100

1200

1300

Wavelength (nm)

1500 1600

1700

–10 –20

Dispersion shifted

FIGURE 3.2 Dispersion coefficient in a dispersion-shifted fiber.13

WDM system, different channels travel at almost the same speed. Any nonlinear mixing effects that require phase matching between different wavelength channels will accumulate at a higher rate than if wavelengths travel at widely different speeds (the case of higher dispersion fiber). The deleterious nonlinear effects that tend to destroy the signal integrity are self-phase modulation (SPM), crossphase modulation (XPM), and four-wave mixing (FWM).14,15 Both FWM and XPM are strengthened by interactions between wavelengths over long propagation distances. A dispersion value as small as a few ps/(nm  km) is sufficient to make XPM and FWM negligible since the different wavelength channels are not phase matched and “walk off” from each other quickly, thus ensuring that they interact with each other only over relatively short distances. To mitigate the effects of nonlinearities, the fibers may need to introduce relatively modest amounts of CD. The intent was to avoid distorting the signal with too much dispersion, but still introduce enough dispersion to counteract the nonlinear effects. Two of the best-known, nonzero dispersion-shifted fibers (NZDSFs) introduced in the mid-1990s are Corning’s large effective area fiber (LEAF)16 and Lucent’s TrueWave fiber. The dispersion of a NZDSF is roughly 4–6 ps/(nm  km), low enough to allow transmission over longer distances than SMF but with a dispersion value large enough to reduce FWM and XPM that occurred in DSF. Figure 3.3 shows the dispersion curve for several of these commercialized NZDSFs. Note that there are continuing efforts and studies from both the academy and industry to improve fiber’s dispersion characteristics for next-generation fiber systems.

3.2.2 Systems limitations due to chromatic dispersion The velocity of a single monochromatic wavelength in fiber is constant. However, data modulation causes a broadening of the spectrum of even the most monochromatic laser pulse. Thus, all modulated data has a nonzero spectral width that spans several wavelengths, and the different spectral components of the modulated data travel at different speeds. In particular, for digital data intensity modulated on an optical carrier, CD leads to pulse broadening, which in turn leads to CD limiting the maximum data rate that can be transmitted through optical fiber. The units of CD are (ps/nm)/km; thus, shorter time pulses, wider frequency spread due to data modulation, and longer fiber lengths will each contribute to temporal dispersion, as shown in Figure 3.4.

Dispersion (ps/nm)

3.2 Chromatic dispersion and its effects on optical fiber systems

71

20 16 12 8 4 0 –4

1510 1530 1550 1570 1590 1610 Wavelength (nm)

SMF Tera Light Pure Guide E-LEAF

True Wave RS True Wave Classic DSF

FIGURE 3.3 CD values for several commercially available types of transmission fiber.3 Photon velocity (f ) =

Speed of light in vacuum Index of refraction (f ) (a)

vj

Information bandwidth of data

0

11

0

vi

vk

Fourier transform

1 0

Time

fcarrier

(b)

Temporal pulse spreading

F [distance, (bit rate)2 ]

ps/nm km

Time (c)

FIGURE 3.4 Origin of CD in data transmission. (a) CD is caused by the frequency-dependent refractive index in fiber. (b) Non zero spectral width due to data modulation. (c) Dispersion leads to pulse broadening, proportional to the transmission distance and data rate. f, frequency; v, velocity.3

The data rate and the data modulation format can significantly affect the sensitivity of a system to CD. For a given system, a pulse will disperse more in time for a wider frequency distribution of the light and for a longer length of fiber. Higher data rates inherently have both shorter pulses and wider frequency spreads. Therefore, as network speed increases, the impact of CD rises precipitously as the square of the increase in data rate. The quadratic increase with the data rate is a result of two effects,

72

CHAPTER 3 Chromatic dispersion monitoring

each with a linear contribution. First, a doubling of the data rate makes the spectrum twice as wide, doubling the effect of dispersion. Second, the same doubling of the data rate makes the data pulses only half as long, thus making it twice as sensitive to dispersion. The combination of a wider signal spectrum and a shorter pulse width is what leads to the overall quadratic impact—when the bit rate increases by a factor of 4, the effects of CD increase by a factor of 16.3 A rule for the maximum distance over which data can be transmitted is to consider a broadening of the pulse equal to the bit period. For bit period B, dispersion value D, and spectral width Dl, the dispersion-limited distance is given by LD ¼

1 1 1 ¼ / 2: D  B  Dl D  B  ðcBÞ B

(3.2)

For example, for single-mode fiber, D ¼ 17 ps/nm/km; thus, for 10-Gb/s data the distance is LD ¼ 52 km, as shown in Figure 3.5. In fact, a more exact calculation shows that for 60 km, the dispersion-induced power penalty is less than 1 dB.17 The power penalty for uncompensated dispersion rises exponentially with transmission distance. Signals at 40-Gb/s are 16 times more sensitive to CD than 10-Gb/s signals. Therefore, CD is one of the main impairments that limit the performance of optical fiber systems. For robust high–bit rate systems, it is essential that dispersion be compensated to within tight tolerances. It is important to note that data modulation format types have substantial impacts on the fiberdispersive and nonlinear effects.18 Some of these formats carry information through on-off keying (OOK), but also modulate the optical phase in a non-information-bearing way in order to enhance the signals’ robustness to CD, optical filtering, and/or nonlinearities. This group includes formats such as optical duobinary (DB), chirped return-to-zero (CRZ), and alternating-phase OOK formats, such as carrier-suppressed return-to-zero (CSRZ). Single-sideband (SSB) modulations have narrower spectra

10000 4000 Dispersion limit (km)

NZDSF 4 ps/nm/km 1000 1000

250

100

60 60 15

10

15 SMF = 17 ps/nm/km 3.5

1

1

2.5

10 20 Data rate (Gb/s)

40

100 160

FIGURE 3.5 Transmission distance limitations due to uncompensated dispersion in SMF as a function of data rate for intensity-modulated optical signals.17 Copyright © 2001 IEEE.

3.2 Chromatic dispersion and its effects on optical fiber systems

73

Table 3.1 Overview of Dispersion Tolerance at 42.7 Gb/s19 Modulation Format

CD [ps/nm] (2-dB Penalty)

NRZ-OOK 50% RZ-OOK 67% CSRZ-OOK Duobinary (DB) 33% RZ-AMI VSB-NRZ-OOK VSB-CSRZ NRZ-DPSK 50% RZ-DPSK NRZ-DQPSK 50% RZ-DQPSK

54 48 42 211 (152) 49 63 (155) 51 (154) 74 (161) 50 (161) 168 (176) 161 (186)

width that can also reduce the dispersion effect. For example, Table 3.1 quantifies the accumulated CD that yields a 2-dB penalty at 42.7 Gb/s19 for various modulation formats with direct detection. Most modulation formats exhibit dispersion tolerances on the order of 50 ps/nm, except for some spectrally narrowed formats, such as DB and quadrature phase-shift keying (QPSK). Note that the numbers in brackets refer to the dispersion tolerance in a system with five optical add-drop multiplexers (OADMs).

3.2.3 Dispersion effects in the presence of fiber nonlinearities The refractive index of fiber is not only dependent on the frequency of light (CD), but also on the optical intensity (optical power) inside the fiber core: nð f ; PÞ ¼ nð f Þ þ n2

P Aeff

(3.3)

where n( f ) is the linear part of the refractive index, P is the optical power inside the fiber, and n2 is the nonlinear index coefficient for silica fibers. The typical value of n2 is 2.6  1020 m2/W. This number takes into account the averaging of the polarization states of the light as it travels in the fiber. The intensity dependence of the refractive index gives rise to three major nonlinear effects, such as SPM, XPM, and FWM. Although the nonlinear coefficient is small, the long transmission lengths and high optical powers that have been made possible by the use of optical amplifiers can cause a large enough nonlinear phase change to play a significant role in state-of-the-art light wave systems. For instance, XPM could cause severe degradation of signal to noise (SNR) if the fiber’s dispersion is too small,14 as shown in Figure 3.6. Another experiment has shown that if a WDM channel exists at one of the four wave-mixing beat-term frequencies, the beat term will interfere coherently with other WDM channels and potentially destroy the data, as shown in Figure 3.7.15 On the other hand, fiber nonlinearity in fact is not always “bad.” One kind of optical pulse, known as solitons,20,21 actually takes advantage of fiber nonlinear effects, specifically SPM, to overcome the pulse-broadening effects of CD.22 This is due to a unique combination of factors for which the dispersion exactly cancels the SPM (i.e., these two effects have the same magnitude but are opposite in sign), making for a nondispersive pulse.

74

CHAPTER 3 Chromatic dispersion monitoring

v1 l1

l1 n1⬘

v2

l2 n1

l2

Fiber l3

l2

Fiber

v3

l3

l3

Time

SNR (dB)

l1 n1⬘⬘

Time

Time (a) Performance vs. dispersion

20 18 16 14 12 10 8 6 4 2 0

3 channels 1-nm spacing 360-km span Power = 15-mW/channel

0

2

4

6

8

10

12

14

16

18

20

Dispersion (ps/nm/km) (b)

FIGURE 3.6 (a) The glass that a photon in the l3 pulse “sees” changes as other channels (with potentially varying power) move to coincide with the l3 pulse. (b) System performance (SNR) versus fiber dispersion. Higher dispersion is preferred to reduce XPM effects.14 Copyright © 1994 IEEE.

While zero-dispersion transmission is not a good idea for WDM systems, a large value of the accumulated dispersion at the end of a fiber link is also undesirable for intersymbol interference (ISI) caused by pulse spreading. One solution is to have a “dispersion map,” alternating sections of positive and negative dispersion, as seen in Figure 3.8. This is a very powerful concept: at each point along the fiber the dispersion has some non zero value, eliminating FWM and XPM, but the total dispersion at the end of the fiber link is nearly zero, so that no pulse broadening is induced. The most advanced systems require periodic dispersion compensation, as well as pre- and postcompensation (before and after the transmission fiber). In order to optimally manage the accumulation of nonlinear effects, each fiber system should have an optimum dispersion map, requiring its own specific arrangement of positive and negative dispersion elements. In conclusion, whenever WDM is used, CD must be present to minimize fiber’s nonlinear effects, even though it is possible to manufacture dispersion-shifted fiber that induces zero CD. CD is a necessary “evil” for the deployment of WDM systems.23 CD should always be managed within an optimum range rather than minimized to zero. This dispersion range depends on data rate, formats, power level, and many other factors.

3.2 Chromatic dispersion and its effects on optical fiber systems

Out of fiber

Into fiber f1

f

f2

2f1 - f2 f1 f2 2f2 - f1

DSF D ~ 0 (a)

Degradation on optical spectrum (unequal channel spacing) Power (dBm)

75

f

Degradation on bit stream (equal channel spacing)

–10

D = –0.2 (ps/nm/km)

–20

D = –1 (ps/nm/km) D = –2 (ps/nm/km) 26 ns

–30 –40 1542 1544 1546 1548 Wavelength (nm)

16 ns

21 ns

(b)

(c)

FIGURE 3.7 (a, b) FWM induces new spectral components via nonlinear mixing of two wavelength signals. (c) The signal degradation due to FWM products falling on a third data channel can be reduced by even small amounts of dispersion.15 Copyright © 1995 IEEE.

Accumulated dispersion (ps/nm)

Positive dispersion transmission fiber

–D

Negative dispersion element

–D

–D Dtotal ≈0

+D

–D +D

–D

+D

–D

Distance (km)

FIGURE 3.8 Dispersion map of basic dispersion-managed system. Positive dispersion transmission fiber alternates with negative dispersion compensation elements such that total dispersion is nearly zero end to end.3

3.2.4 The need for chromatic dispersion monitoring As mentioned above, unlike other fiber impairments (such as noise, attenuation, and PMD), CD needs to be optimized for different systems, rather than be minimized to zero. In a perfect world, all fiber links would have a known, discrete, and unchanging value of CD. Network operators would then deploy fixed dispersion compensators periodically along every fiber link to match the fiber

76

CHAPTER 3 Chromatic dispersion monitoring

Dispersion (ps/nm)

30 20

dl o /dT~ 0.03 nm /⬚C

Δl o(T )

10

e

op

. sl isp

ΔD

D

0 –10

1250

1350 1450 1550 Wavelength (nm)

(a)

1650

Accumulated dispersion change ΔD (ps/nm)

dispersion. Unfortunately, several vexing issues may necessitate that dispersion compensators are tunable—that is, the ability to adjust the amount of dispersion to match system requirements. First is the most basic business issue of inventory management. Network operators typically do not know the exact length of a deployed fiber link nor its CD value. Moreover, fiber plants periodically undergo upgrades and maintenance, leaving new and non exact lengths of fiber behind. Therefore, operators would need to keep in stock a large number of compensator models, and even then the compensation would only be approximate. Second, we must consider the sheer difficulty of
40-Gb/s signals. The tolerable threshold for accumulated dispersion for a
40-Gb/s data channel is 16 times smaller than at 10 Gb/s. If the compensation value does not exactly match the fiber to within a few percent points of the required dispersion value, then the communication link will not work. Tunability is considered a key enabler for this bit rate and beyond. Third, the accumulated dispersion changes slightly with temperature, which begins to be an issue for
40-Gb/s systems and 10Gb/s ultra-long-haul systems. In fiber, the zero-dispersion wavelength changes with temperature at a typical rate of 0.03 nm/ C.24 It can be shown that a not-uncommon 50 C variation along a 1000-km 40-Gb/s link can produce significant degradation (see Figure 3.9). Fourth, we are experiencing the dawn of reconfigurable optical networking. In such systems, the network path, and therefore the accumulated fiber dispersion, can change. It is important to note that even if the fiber spans are compensated span-by-span, the pervasive use of compensation at the transmitter and receiver suggests that optimization and tunability based on path will still be needed. Other issues that increase the need for tunability include: (1) laser and (de)mux wavelength drifts for which a data channel no longer resides on the flat-top portion of a filter, thereby producing a chirp on the signal that interacts with the fiber’s CD; (2) changes in signal power that change both the link’s nonlinearity and the optimal system dispersion map; and (3) small differences that exist in transmitter-induced signal chirp. Based on the above-mentioned reasons, in almost all
40-Gb/s IM/DD systems, highly accurate dispersion management must be implemented, potentially requiring tunable dispersion compensators that should be accompanied by dynamic monitoring of the accumulated CD.25 100

50

Dispersion slope∼ 0.08 ps/nm2/km NRZ 40 Gb/s limit

0

L = 200 km

–50

L = 500 km NRZ 40 Gb/s limit L = 1000 km

–100 – 40 –30 –20 –10 0 10 20 30 40 Temperature change, ΔT (°C)

(b)

FIGURE 3.9 (a) Zero-dispersion wavelength shifts due to temperature change; thus, dispersion itself changes at a fixed wavelength24 (b) For a 40-Gb/s, 1000-km fiber link, 30 C temperature change causes dispersion beyond system limit.3 Copyright © 2000 IEEE.

3.3 Chromatic dispersion monitoring techniques

77

Moreover, the ability to reconfigure a network to provide dynamic services causes many traditionally static network properties to become dynamic quantities. In order to enable robust and cost-effective automated operation, the future network should probably be able to: (1) intelligently monitor the physical state of the network as well as the quality of propagating data signals, (2) automatically diagnose and repair the network, (3) allocate resources, and (4) redirect traffic. To achieve this, optical performance monitoring should isolate the specific impairments and location of the cause rather than simply sound an alarm. CD monitoring is one of the most important problems among these impairments. In summary, CD-induced impairments are time varying due to a changing environment, drift of components, and rapid reconfiguration of network paths. Moreover, CD impairments depend on complex interactions of nonlinear fiber effects, which in turn are a function of the signal power, data rate, and data modulation format. These properties imply that CD monitoring should provide realtime information about the quality of these transmitted signals.

3.3 CHROMATIC DISPERSION MONITORING TECHNIQUES Basically CD monitoring techniques measure one of the signal’s parameters or property that could be affected by the fiber’s CD. These parameters include the signal’s spectral component in optical and/ or electrical domain, signal waveform distortion, and/or the channel’s BER. Some techniques may add extra information at the transmitter and then detect it at the receiver. In this section, we will describe the various promising techniques in details.

3.3.1 Measurement of RF spectrum One method that has been demonstrated for real-time CD monitoring and may be applied more generally as optical performance monitoring is to detect the conversion of a phase-modulated (PM) signal into an amplitude-modulated (AM) signal due to CD.26 A second method is inserting subcarriers (RF tones) at the transmitter. The subcarrier approach measures the resulting delay of the subcarrier sidebands relative to the baseband and can be used to measure the accumulated dispersion with fine and medium accuracy without knowledge of the signal transport history.27–37 Figure 3.10 shows the basic principle of RF fading of a subcarrier tone within the data band. The tone carries no data and has a narrow spectral line width that is less than a few MHz. Since the tone is within the frequency band of the data, it experiences the same dispersion as the data and can be used as a dispersion sensor. The upper and lower sidebands of the subcarrier will initially be in phase with each other at the transmitter. However, a phase difference will accumulate during transmission between the two sidebands as a result of CD. The degree of phase mismatch is related to the amount of accumulated dispersion when the subcarrier is opto-electronically detected, and the subcarrier power fades according to the phase mismatch in a predictable manner, as in  2  pfsc lDL ; (3.4) I ð fsc Þ ¼ I0 m cos c where I0 is the peak photocurrent, m is the modulation depth, fsc is the subcarrier tone frequency, l is carrier wavelength, DL is the total accumulated dispersion, and c is the speed of light in vacuum. Figure 3.11(a) shows the measured and theoretical fading curves for tone frequencies of 7 and 9 GHz,

78

CHAPTER 3 Chromatic dispersion monitoring

Transmitter

Receiver

Optical spectrum

Optical spectrum Dispersive fiber

l0 l 0 – fsc l 0 + fsc Electrical signal

In phase

l0 l 0 – fsc l 0 + fsc Electrical signal

t

Out of phase

t

FIGURE 3.10 Principle of RF fading used for dispersion monitoring: RF tone within data band fades due to CD.28 Copyright © 2002 IEEE.

4000 7-GHz tone

3500

9-GHz tone –44

Dispersion (ps/nm)

RF tone power (dBm)

–42

–46 –48 –50

3000 2500 2000 1500 1000 500

–52 0

200

400 600 800 1000 1200 Dispersion (ps/nm) (a)

0

4

6 8 10 Tone frequency (GHz)

12

(b)

FIGURE 3.11 (a) RF fading due to dispersion for 7- and 9-GHz tones. Solid lines represent theoretical results and points are experimental results. (b) Theoretical curve for measurable range of CD.28 Copyright © 2002 IEEE.

3.3 Chromatic dispersion monitoring techniques

79

in which the fading follows a sin2 function with increasing dispersion. Based on this experiment, an important characteristic of this technique is that the sensitivity increases for higher dispersion values since higher values reside on a steeper portion of the curve. It is possible to reconfigure the measurement range and sensitivity of the monitor by simply tuning the RF tone frequency. Figure 3.11(b) shows the relationship between the tone frequency and the measurable dispersion range, in which a smaller range results in a higher sensitivity. Using the high subcarrier frequency will provide higher sensitivity. On the other hand, using the low subcarrier frequency will provide a wider measurement range, but the sensitivity may not be enough. One solution is to add two RF tones to improve the monitoring range and sensitivity without increasing the system complexity.38,39 Another obstacle of AM pilot-tone-based monitoring techniques is the difficulty in discriminating the sign of accumulated dispersion. One proposed method is adding an extra FM pilot tone to detect differential phase delay, thus determining positive or negative dispersion.40 A third method is optical frequency modulation (FM) at the transmitter. Since the group-delay time of the transmitted signal varies with the carrier frequency, the phase of the clock signal recovered from the transmitted signal at the receiver is also modulated at the same frequency. Therefore, the CD of the transmission system can be obtained from the phase deviation of the clock signal.41,42 All above-mentioned methods are simple and applicable to WDM systems, but require modification of the transmitter. Based on the dispersion-induced RF power fading effect, an alternative technique is to extract the bit rate frequency component (clock) from photodetected data and monitor its RF power.43–45 This technique does not require modification of the transmitter, but it is bit rate and modulation format dependent. For example, the effect of dispersion for a non-return-to-zero (NRZ) signal is to induce an increase in the power in the spectral region of the clock. Within a range of values of the signal dispersion, the power is proportional to the amount of accumulated dispersion. Beyond this distance, the clock power fades again, and then continues to change periodically with the distance. As opposed to NRZ systems, RZ data contain a clock component, and within a range of values of the signal dispersion, clock power gradually fades away. Figure 3.12(a) and (b) show the results of the clock regenerating effect for a NRZ signal and the clock fading effect for a RZ signal in a 10-Gb/s system. Since the clock signal represents the degradation, or distortion, of both NRZ and RZ data, this approach actually monitors system performance. Note that fiber nonlinearities may change the measurement range and clock power, requiring initial calibrations for different transmitted optical power levels. Although this approach cannot isolate CD, like other tone-fading techniques, it is sensitive to a variety of distortion effects including PMD and pulse carver misalignment, which is advantageous for fault localization. Except for measuring the clock signal at the receiver, other methods suggest measuring RF power or phase at other frequencies, such as the half–bit rate frequency.46 The proposed optical filters at the receiver include delay-and-add filter, partial optical filter, imbalanced Mach-Zehnder interferometer (MZI), and fiber loop.47–50 This is due to dispersion-induced phase modulation to the intensitymodulated signal. By measuring the phase modulation present at a specific frequency component, the dispersion of the link can be determined. This method does not require the addition of RF tones at the transmitter as well, and potentially can provide a wide mentoring range and high resolution. Figure 3.13(a) shows the experiment setup and Figure 3.13(b) shows the measure of RF power as a function of the fiber dispersion.47

80

CHAPTER 3 Chromatic dispersion monitoring

0

–10 –20

40-km SMF

–30

20-km SMF

–40 Back to back

–50 –1000

0 500 –500 Accumulated dispersion (ps/nm)

Normalized clock power (dB)

Normalized clock power (dB)

0

–10 Back to back

–20 20-km SMF

–30 40-km SMF

–40 –1000

1000

(a)

–500 0 500 Accumulated dispersion (ps/nm)

1000

(b)

FIGURE 3.12 (a) Clock regenerating effect for NRZ data. (b) Clock fading effect for RZ data. Solid lines represent without SPM; dashed lines represent with SPM; dotted lines represent experimental.44 Copyright © 2001 IEEE.

Transmitter

Dispersion measurement system RF spectrum analyzer

Dispersive link

τ

Power at 5 GHz (dBm)

–50

Pulse pattern generator

–55 –60 –65 –70 –75 –2500 –2000 –1500 –1000 –500

Receiver

0

500

1000 1500

Dispersion (ps/nm) (a)

(b)

FIGURE 3.13 (a) Experimental setup. RF power at half of data rate (5 GHz) is measured after MZI with path difference of 100 ps. (b) Received RF power as function of dispersion.47 Copyright © 2005 IEEE.

3.3.2 Measurement of relative group delay between VSB signals One powerful technique for CD monitoring is detecting the relative group delay between the upper and lower vestigial sideband (VSB) signals in transmitted data51–53: the lower and upper vestigial sidebands are obtained by tuning an optical filter away from the optical spectrum center of the

3.3 Chromatic dispersion monitoring techniques

1.5 Intensity

Entire channel

VSB-L Filtered spectrum 40-Gb/s RZ data

1.0 0.5 0.0

f Dispersion

0

O/E

50 100 Time (ps)

Filtered spectrum

Intensity

150 Δt

1.5

VSB-U

f

81

1.0 0.5 0.0

0

50 100 Time (ps)

150

FIGURE 3.14 Conceptual diagram for monitoring CD using optical VSB filtering. Recovered bits from either part of spectrum arrive at slightly different times depending on CD.51 Copyright © 2002 IEEE.

double-sideband data, as shown in Figure 3.14. Typically, in high-speed optical transmission systems, the optical signal with either an NRZ or RZ format has two optical sidebands that carry the same data information. Each sideband can be selected using a tunable optical filter with a frequency bandwidth approximately equal to the data bit rate. In general, a VSB (or single sideband) filtering is implemented by tuning the filter apart from the spectral center of the optical signal. Since the two optical sidebands occupy different wavelength ranges, fiber CD induces a relative group delay between the lower and upper VSB signals (see Figure 3.14). This group delay can be measured in general through clock recovery and phase-sensitive detection. Figure 3.15 shows the simulation and experimental results of phase shift between the upper and lower VSB signals as a function of CD. This technique requires no modification at the transmitter; is highly sensitive; is unaffected by PMD, fiber nonlinearity, and transmitter chirp; and can be applied to WDM signals by sweeping the optical filter. In this approach, monitoring occurs through phase detection of the recovered clock signal; thus, the range of CD that can be monitored is limited by the range of phase shift from 180 to þ180 . There is also a trade-off between the monitoring accuracy and speed. One proposed solution is a dithered-filter method: a received optical signal is filtered by a dithering optical bandpass filter and then measure the CD from the delay time caused by the filtering. Although the delay-induced phase deviation might be small, the accuracy could be kept high due to the synchronous detection at the dithering frequency.52

3.3.3 Histogram monitoring techniques Sampling a received signal can obtain the histograms that allow for signal quality assessment (such as BER). This method is sensitive to signal distortion and noise, and thus could be used to monitor

82

CHAPTER 3 Chromatic dispersion monitoring

Phase shift (degree)

360 240 120 RZ

0 –120 –240

NRZ

–360 –20 –15 –10 –5 0 5 10 –100 × GVD × (bit rate)2

15

20

FIGURE 3.15 Phase shift between the two VSB signals versus the normalized. The three lines are simulation results for NRZ data, Gaussian filter (dashed dotted line); RZ data, Gaussian filter (solid line); and RZ data, fiber Fabry-Perot filter (dashed line). Scatter points are experimental for 10-Gb/s RZ data using a fiber Fabry-Perot filter.51 Copyright © 2002 IEEE.

CD-induced waveform distortions. There are generally two sampling approaches: synchronous and asynchronous sampling. An eye diagram is actually a synchronous amplitude distribution within an entire bit period. It is a commonly used tool to analyze the quality of signals, which can be used for performance monitoring. The CD could be assessed by synchronously sampling the eye diagram on both sides.54 The key drawback of eye monitoring is the required clock recovery. When the clock information is absent, the signal can only be asynchronously sampled. The amplitude histogram is obtained by random sampling that spans the entire bit period. With a sufficient number of random samples, an asynchronous histogram can evenly represent the pulse amplitude distribution in a bit period.55 The asynchronous amplitude histogram technique may be a promising method for low-cost, bit rate–transparent channel performance monitoring due to unnecessary clock recovery. A comparison of synchronous and asynchronous diagrams has been shown in Figure 3.16.55 Figure 3.16(a) is the synchronous temporal window (solid-line box) at the widest eye opening. Figure 3.16(b) is the corresponding histogram. Based on the histogram, the factor and the BER can be estimated.56 Figures 3.16(c) and (d) are the asynchronously sampled eye diagram and histogram obtained within the entire bit period. The main difference between the asynchronous and the synchronous histograms lies at the middle counts, the “cross” in the eye diagram between the mark and space levels. Within the cross point region, the asynchronous histogram (Figure 3.16(d)) contains counts, while the synchronous histogram (Figure 3.16(b)) contains essentially no counts. The cross point count is related to the pulse rise time and fall time that may be affected by dispersionrelated impairment. In fact, asynchronous histogram can evaluate the signal quality. It is sensitive to both OSNR and other fiber impairment such as CD and PMD. Monitoring the Q-factor or OSNR for both OOK and DPSK signals using this method has been suggested.57–64 Monitoring of both CD and PMD has also been investigated and demonstrated theoretically and experimentally.65,66 However, since this is the electrical-domain measurement, the challenge is how to measure the absolute value of the Q-factor or OSNR, CD, and PMD simultaneously.

Amplitude (a.u.)

Amplitude (a.u.)

3.3 Chromatic dispersion monitoring techniques

s1

m1 s0

Time (ps) (a)

83

m0

Amplitude (a.u.)

Amplitude (a.u.)

Count (b)

Time (ps) (c)

s1

m1

s0

m0

Count (d)

FIGURE 3.16 (a) Eye diagram and (b) histogram with synchronous sampling. (c) Eye diagram and (d) histogram with asynchronous sampling.55 Copyright © 2004 IEEE.

Recently, an asynchronous sampling technique is demonstrated that can measure multiple simultaneous impairments.67–70 The optical waveform is sampled in pairs separated by a known physical delay Dt, as shown in Figure 3.17. The sample time between the pairs, Ts, is not related to the bit rate of the monitored signal, and can be many orders of magnitude slower. Plotting the pairs produces information-rich patterns called phase portraits, which are of similar complexity to eye diagrams, but do not require clock recovery. A fundamental difference between the two techniques is that

Δt = T Ts

(0.1) Ts

x2

1

10

110

0 01

y x3

001

y3

x1 y1

(1.1) 011

y2 Time

(0.0)

100

(1.0)

x

FIGURE 3.17 Portraits processing of delay-tap sample pairs to create phase portraits. Labels on phase portrait represent the sampled bit sequences.67 Copyright © 2007 IEEE.

84

CHAPTER 3 Chromatic dispersion monitoring

Clean

ASE

CD

PMD

Crosstalk

All

(a)

(b)

(c)

(d)

(e)

(f)

FIGURE 3.18 Eye diagrams and phase portraits for NRZ: (a) OSNR ¼ 35 dB and no impairment; (b) OSNR ¼ 25 dB; (c) OSNR ¼ 35 dB and CD ¼ 800 ps/nm; (d) OSNR ¼ 35 dB and PMD ¼ 40 ps; (e) OSNR ¼ 35 dB and crosstalk ¼ 25 dB; and (f) OSNR ¼ 25 dB, CD ¼ 800 ps/nm, PMD ¼ 40 ps, and crosstalk ¼ 25dB.67 Copyright © 2007 IEEE.

the phase portrait contains information about the probability distributions of closely spaced samples, or equivalently, distributions of waveform slopes. This information is absent in eye diagrams, which are constructed from samples separated by large periods. For a one-bit delay, Dt ¼ T, the technique geometrically separates the various 3-bit sequences, and in particular separates out the 010 and 101 sequences, which are generally most susceptible to signal distortion. The effects of OSNR, CD, PMD, and interferometric crosstalk on a 10-Gb/s NRZ signal have been simulated. The resulting phase portraits and eye diagrams are shown in Figure 3.18. In each case the tap delay for the phase portrait was chosen to be a 1-bit period. Figure 3.18(a) shows the results for no optical impairment (OSNR ¼ 35 dB), with a clean eye and a well-defined geometric shape in the phase portrait. Figure 3.18(b) shows the effects of reducing the OSNR to 25 dB. The predominant effect is to broaden the high-power regions of both the eye and the two tap plots, but the underlying geometric shapes are not affected. Figure 3.18(c) shows OSNR of 35 dB with 800 ps/nm of CD, corresponding to 50 km of SMF. The eye diagram shows the characteristic narrowing of the peaks. In the phase portrait, the dispersion causes the diagonal to curve in toward the origin. This curvature increases with dispersion, offering a potential dispersion monitoring tool. Figure 3.18(d) shows an OSNR of 35 dB with 30 ps of first-order PMD, with the power split equally between the polarization axes. The eye shows the characteristic “triangularization.” And a closer inspection shows significant differences in the distribution of points along the 3-bit transitions. Figure 3.18(e) shows an OSNR of 35 dB with a single source of interferometric crosstalk at 25 dB. Both the eye and two-tap plots show similar broadening to the OSNR degradation, but different noise statistics. Finally, Figure 3.18(f) shows an OSNR of 25 dB with the combined degradations of CD, PMD, and crosstalk. These results show that the phase portraits contained impairment signatures that could be exploited to separate and measure impairments even in cases where they occur simultaneously. Figure 3.19 shows the simulation results of extracting the signal quality and the underlying cause of signal degradation by analyzing the phase portraits, including OSNR, CD, PMD, filter dispersion, or some combination.70

Multi-impairment monitor

3.3 Chromatic dispersion monitoring techniques

OSNR (dB)

Dispersion (ps/nm)

25

85

First-order PMD (ps) 50

1500

40

20

1000 30

15

500

10

10

0 10

20

20

15 20 Actual

0

25

Filter offset (GHz)

500 1000 1500 Actual

0

Jitter (%)

50 40

10

0

20 40 Actual Q (linear)

10

30 0 20 –10 –20 –20

5

10 0 Actual

20

0

0

20 40 Actual

0

0

5 10 Actual

FIGURE 3.19 Monitor versus actual values of various impairments and signal quality measures for simultaneous mixtures of OSNR, CD, PMD, and filter offset from 10-Gb/s NRZ simulations.70 Copyright © 2007 IEEE.

3.3.4 All-optical spectral analysis using nonlinear optics Nonlinear optical signal processing could also be used to detect time-domain pulse distortion induced by CD. The operating principle basically is that a dispersed optical pulse has reduced peak intensity, thus resulting in a weaker nonlinear effect either in fiber or semiconductors. The nonlinear process could be SPM, XPM, FWM, optical parametric amplify in highly nonlinear fibers or cross gain/cross phase modulation in optical semiconductor amplifiers (OSAs), power-dependent optical spectral shift in OSAs, or two-photon absorption in semiconductors.71–82 One of the main advantages of these approaches is one can detect the average optical power in a certain wavelength range using a low-speed detector, and thus avoid the high-speed electrical-domain signal processing. As an example, Figure 3.20 shows the principle of dispersion monitoring via SPM in a Kerr medium followed by optical filtering.71 An amplitude-modulated signal will be spectrally broadened when propagating through a Kerr medium, and a measure of this spectral broadening can be used to determine the degree to which pulses were temporally broadened at the input of the fiber by residual dispersion. Figure 3.21 shows the measured optical power versus the residual dispersion by recording the optical spectrum of a 40-Gb/s RZ data stream at the output of a nonlinear fiber while varying the dispersion-induced pulse broadening at the input. It shows clear correlation between spectral broadening at the output and residual dispersion at the input of the nonlinear fiber. This experiment

86

CHAPTER 3 Chromatic dispersion monitoring

TDC Filter

Tap Kerr nonlinearity

SPM

Time

Frequency

FIGURE 3.20 Principle of residual dispersion monitoring via SPM and filtering.71

Copyright © 2002 IEEE.

3

0.08 2

0.06 0.04

1

0.02 0.00 –75

BER penalty (dB)

Filtered power/total power

0.10

0 –50

–25 0 25 50 75 Residual dispersion (ps/nm)

100

FIGURE 3.21 Comparison of long-pass, spectral monitoring signal (solid squares) with typical 40-Gb/s RZ receiver BER penalty (open circles), both plotted versus residual dispersion. Eye diagrams are indicated for three residual dispersion values. Lines are a guide to the eye.71 Copyright © 2002 IEEE.

3.3 Chromatic dispersion monitoring techniques

Optical noise and dispersion monitor

CW

PCW

SOA

PM OF

Pdata

DCMs / SSMF

87

PCNV

Sampling scope

Tx

Rx

ASE

OSA

BERT

FIGURE 3.22 Experimental setup to vary noise and accumulated dispersion on a data signal. The wavelength-converted monitoring signal is generated by mixing Pdata with CW signal PCW in SOA and selected by the optical filter.76 Copyright © 2005 IEEE.

demonstrated that spectral broadening is a good measure of residual dispersion of a RZ data stream, and is directly correlated to BER impairments. Figure 3.22 shows one scheme of a SOA-based CD monitoring technique.76 The data signal is combined with a continuous-wave (CW) field at the desired conversion wavelength and coupled into the SOA. The CW field is influenced by the data signal through both cross gain and cross phase modulation in the SOA. A 2R regenerated and wavelength-converted signal is recovered by filtering the red-shifted portion of the CW field. The monitoring signal (at a different filtering wavelength) as a function of accumulated dispersion is shown in Figure 3.23. The converted signal power is 1.2 Monitor signal (a.u.)

Red filter (+0.2 nm)

2R filter

1.0 0.8 0.6 0.4 0.2 0.0 50 100 –150 –100 –50 0 Accumulated dispersion [ps/nm]

150

FIGURE 3.23 Monitor signal for 40-Gb/s data as function of accumulated dispersion.76

Copyright © 2005 IEEE.

88

CHAPTER 3 Chromatic dispersion monitoring

normalized to the zero dispersion point. Thus, the reduction in the average level of the marks translates directly into a reduction in the red shift. The 2R filter provides a dispersion monitoring signal with coarse sensitivity, whereas when using the red-filter location, a highly sensitive signal is obtained. Secondary peaks at high dispersion values are due to intersymbol interference.

3.3.5 Electronic monitoring techniques Electronic monitoring techniques are based on the analysis of the electrical signal after the O/E conversion at the receiver, such as Q/BER measurement and eye diagram analysis. The error correlation information from FEC and the coefficients obtained from an electrical equalizer at the receiver can also be used for performance monitoring to identify and quantify optical distortions and perform network monitoring without additional expensive optical equipment. By comparing the coefficients of an adaptive electrical equalizer to precomputed coefficient vectors within the digital-signal processor in the receiver, it is possible to identify and quantify the most common distortions of an optical link.83–85 BER is the ultimate measurement of system performance, and is the preferred parameter to use for fault management. In fact, this is precisely the parameter used in electronic networks. However, one difficulty for BER monitoring in optical networks is that the signal is typically error free within the network. For monitoring at an amplifier site, the signal is amplified only and not regenerated. Therefore, noise will pass through and continue to accumulate. Measurement of the BER at the location of the fault would result in error-free measurement. However, when the signal reaches the end terminal, it is not error free due to accumulated noise and performance degradation on the BER is observed. In order to detect the degradation within the network, one solution is to use noise loading. In this case, noise is intentionally added to the signal in order to bring the BER to a measurable level and then the additional noise caused by the impairment can be detected. The common method to compensate for the low sensitivity of BER monitoring is to use Q-factor monitoring.56,86 The Q-factor is obtained by adjusting the decision threshold voltage of the monitor receiver away from the optimum level so that errors are recorded. Figure 3.24(a) shows typical measured data for the logarithm of the BER versus the threshold in the decision circuit. Once an error rate is generated, changes to that rate can be monitored and small degradations become visible. Several such techniques have been developed for measuring the Q-factor.87,88 Note that the Q-factor is essentially the SNR. Figure 3.24(b) shows the BER as a function of the received optical power.89 If Q-factor is measured using a receiver, then it is precisely the electronic SNR. If measured by other means such as optical sampling, then it is the in-band optical SNR. It is defined as the difference between the average value of the marks (ones) and of the spaces (zeros) divided by the sum of the standard deviations of the noise distributions around each. Due to the strong correlation between Q-factor and BER, the Q-factor measurement is highly effective for fault management. Q-factor is sensitive to the same impairments that impact the end terminal receiver with the appropriate sensitivity. Although the cost of this approach may be high for many embedded network monitoring applications, a portable unit can be a valuable tool in troubleshooting faults, particularly to target the rare complication that is not identified by the embedded optical channel monitor. For the system with a coherent receiver, CD can be monitored and compensated using the minimum mean-square-error (MMSE) equalizer solution, by estimating the channel from a training

3.3 Chromatic dispersion monitoring techniques

100

–2 Optimum decision point

Q = 8.5

–6

–8

m1

10–2 Probability of error

–4 Log (BER)

89

m0

–10 –0.4

–0.2 0.0 0.2 Decision levels (volts)

0.4

10–4 10–6 10–8 10–10 10–12 –5

0

(a)

5

10 S , (dB) N (b)

15

20

FIGURE 3.24 (a) Typical measured data for logarithm of BER versus decision threshold56 (Copyright © 1993 IEEE ). (b) BER as a function of the received optical SNR89 (Copyright © 1988 Holt, Rinehart, and Winston).

sequence.90,91 A training sequence may increase the receiver complexity if a correlator is used, thus prohibiting from large training sequences that can estimate large values of dispersion. A blind dispersion estimation algorithm was also proposed for mentoring and compensating impairments.85 This method is suitable for uncompensated transmission and is able to precisely estimate large arbitrary dispersion at a fixed adaptation length and low complexity. Coherent demodulation in general allows complete compensation for linear impairments in the electrical domain, instead of using optical compensators. (See Chapter 13 for detailed information on monitoring techniques for coherent systems.)

3.3.6 Other chromatic dispersion monitoring techniques Other techniques of CD monitoring include: 1. Injecting broadband amplified spontaneous emission that has been modulated with an RF tone into the start of the link to be monitored, and then determining the dispersion-induced group delay by measuring the phase shifts in this RF tone for wavelengths across the band.92 2. Using coherent detection down-converts the spectrum of the optical signal into the RF domain; the carrier and the two sidebands then are selected separately by bandpass filters. The CD can be evaluated from the relative time delay between two clock components.93 3. When using second-harmonic generation (SHG) with chirped quasi-phase-matched (QPM) LiNbO3 waveguides, the pulse broadening can be evaluated through SHG power.94

90

CHAPTER 3 Chromatic dispersion monitoring

4. Using data stream intensity autocorrelations usually depend on instantaneous optical intensity. Thus, this technique can yield information on signal characteristics such as peak power, pulsewidth, or signal duty cycle.95 5. All-optical sampling system that utilizes FWM in a highly nonlinear fiber.96

3.3.7 Differentiate chromatic dispersion from polarization mode dispersion One of the biggest challenges for CD monitoring is how to differentiate the CD-induced impairment from other effects, such as noise, PMD, and nonlinearities.97 Among these effects, PMD is the most difficult because both CD and PMD are linear impairments and cause similar distortion to optical signals. Several proposed CD monitoring techniques are inherently immune to the effect of PMD, such as VSB filtering and phase-delay detection,51 dithering transmitter frequency or receiver filter and measuring time delay,41,42 coherent detection with RF signal processing,93 and the histogram evaluation method.59,67 Many publications focus on simultaneously monitoring CD and PMD,98–100 and the studies focused on monitoring the power of the RF components, histogram evaluations, and coherent detection. This section will focus on the RF measurement, as another two methods can easily differentiate CD and PMD effects. The CD monitoring techniques based on RF tone measurement are relatively simple, but PMD and the chirp of the external modulator may influence CD monitoring. It is well known that when an RF-modulated light travels along an optical path, both CD and PMD will result in the RF power fading at the photodetector. As shown in Figure 3.25, CD causes a phase difference between the two sidebands, and PMD induces the DGD between the two PSPs. Both of these effects will introduce RF power fading in the electrical domain after photodetection. (See Chapter 4 of this book for more indepth information.) Considering the phase difference between the two RF sidebands and the PMD and chirp effects, the detected RF power with double sideband (DSB) at the photodetector is given by35,101,102:  
PSSB ¼ P0 1 þ a2 H ð f0 Þ 2 H ð f0 þ fRF Þ 2 =4  1  4gð1  gÞ sin2 ðpfRF DtÞ ; (3.5)

Δτ

Two sidebands Carrier Lower sideband

Upper sideband

Upper

t

CD (freq t delay)

Lower l

Upper Lower

In phase

Out of phase

Power

Upper sidebands Axis 1 f Axis 2 In phase

FIGURE 3.25 RF tone fading due to CD and PMD.4

Δτ

t PMD Axis 1 (axis t delay) Axis 2 Out of phase

3.3 Chromatic dispersion monitoring techniques

91

where P0 is the RF power without CD and PMD effects, which depends on the loss or gain that the signal experiences; g and Dt are PMD-related factors, including the power-splitting ratio and DGD between the two PSPs; a is a parameter that relates the instantaneous intensity-induced phase variation of the modulated light, also known as chirp parameter102; Dtotal is the total CD induced by fibers and other optical components; fRF is the RF; l is the carrier wavelength; and c is the speed of light in a vacuum. This equation indicates that the faded RF power could be employed in CD monitoring, but the DGD (first-order PMD) may considerably influence the RF power, which leads to monitoring errors for the RF-tone-based CD monitoring techniques. An interesting technique is proposed to suppress the effects of PMD and chirp.102 By inserting a filter to remove one of the transmitted sideband tones just before the detector, the detected RF power of the single-sideband (SSB) signal is given by (assuming the lower-sideband [LSB] tone is removed):  
(3.6) PSSB ¼ P0 1 þ a2 H ðf0 Þ 2 H ðf0 þ fRF Þ 2 =4  1  4gð1  gÞ sin2 ðpfRF DtÞ ; where H is the electrical field transfer function of the optical filter and f0 is the optical carrier frequency. By taking the ratio of the RF power with DSB to that with SSB, the power variation-related term P0 and PMD-related term 1  4g(1  g)sin2(pfDt) are canceled. The RF power ratio is then given by
2 4 cos2 pDtotal l2 fRF =c þ arctan a (3.7) R¼ jH ð f0 Þj2 jH ð f0 þ fRF Þj2 ; where |H| is fixed and can be easily measured. Therefore, CD can be monitored by this RF power ratio without the influence of PMD. Furthermore, the monitoring error induced by the small chirp fluctuation can be suppressed using two RF tones and a CD offset. Figure 3.26 shows the experimental setup. Experimental results in Figure 3.27 show that this technique could accurately monitor the accumulated CD without being affected by the PMD and small chirp fluctuation. Another experiment has been demonstrated that simultaneously monitors and isolates CD and PMD or NRZ OOK and DPSK signals.103,104 The RF clock tone power is monitored at the output ports of an unbalanced Mach-Zehnder delay line interferometer (DLI) with a quarter-bit delay in one arm. It is observed that the clock power from the constructive port of the DLI grows with an

EDFA Laser source

Data

MZM

Fiber

EDFA

TFPF

PMD emulator

90:10 coupler Rx

RF power detector

+

CD offset

PD DSB

Ratio RF power detector

fRF 1, fRF 2 (time multiplexed)

PD SSB FBG filter

50:50 coupler

CD monitoring

FIGURE 3.26 System setup of the CD monitoring scheme suppressing PMD and chirp effects.102

Copyright © 2006 IEEE.

92

CHAPTER 3 Chromatic dispersion monitoring

300 CD monitoring error (ps/nm)

CD monitoring error (ps/nm)

Without PMD cancellation With PMD cancellation

200 150 100 50 0

0

10

200

Without chirp suppression (7 GHz) Without chirp suppression (9.9 GHz) With chirp suppression (7 GHz and 9.9 GHz)

100 0 –100 –200 –300 –0.4

–0.2

DGD (ps)

0 a parameter

(a)

(b)

20

30

40

0.2

0.4

FIGURE 3.27 CD monitoring error, (a) versus DGD without and with PMD cancellation, and (b) versus a parameter without and with chirp suppression.102 Copyright © 2006 IEEE.

increase in CD and with a decrease in PMD, whereas the clock power from the destructive port grows with a decrease in both CD and PMD. By appropriately adding and subtracting the constructive and destructive clock powers, the individual contributions of CD and PMD can be derived simultaneously while the sensitivity is also increased. The same setup can be used for simultaneous OSNR monitoring.105

3.4 SUMMARY In summary, CD is a major degrading effect and limits system performance such as transmission distance and data rate in fiber communication networks. Its effect increases as the square of the bit rate increases for IM/DD systems. For high-speed and intelligent optical fiber networks, CD needs to be monitored and controlled precisely. The value of monitoring increases with increasing intelligence and bit rates. Keeping an appropriate balance among monitoring coverage, sensitivity, and cost is recommended. In this chapter, we have covered a number of techniques on CD monitoring. Some of these techniques are summarized in Table 3.2.

ACKNOWLEDGMENTS The author wishes to thank the kind and critical help from Junyi Wang and Phillip Regan, and all the members of UL Lafayette’s Optics Laboratory and USC’s OC Laboratory family.

Table 3.2 Summary of Selected Advanced CD Monitoring Techniques Techniques

Monitor

Comments

Clock Tone Measurement

CD PMD OSNR Fast response time (sub ms); requires transmitter modification; requires additional consideration to isolate CD and PMD effects

CD PMD OSNR Fast response time (sub ms); no requirement to modify transmitter; only for single channel operation; requires additional consideration to isolate CD and PMD effects

Phase Measurement (with VSB filtering)

Nonlinear Process

Histogram Analysis

Electrical Techniques

CD

CD OSNR

No modification at the transmitter; unaffected by PMD; can be applied to WDM signals by sweeping the optical filter; requires a highperformance phase detector

Low-speed optical power detection; no requirement for high-speed circuit; potentially higher power sensitivity

CD PMD OSNR In-band and cost-effective monitoring; response time may be relatively slow; isolating various impairments are still under investigation

CD PMD OSNR Need high-speed circuits; directly related to system performance; may differentiate CD and PMD if using coherent receiver and DSP

3.4 Summary

RF Tone Measurement

93

94

CHAPTER 3 Chromatic dispersion monitoring

REFERENCES 1. Kilper DC, Bach R, Blumenthal DJ, Einstein D, Landolsi T, Ostar L, et al. Optical performance monitoring. IEEE/OSA J Lightwave Technol 2004;22(1):294–304. 2. Agrawal GP. Lightwave technology, telecommunication systems. New York: John Wiley and Sons, Inc.; 2005. 3. Willner AE, Hoanca B. Fixed and tunable management of fiber chromatic dispersion. In: Kaminow IP, Li T, editors. Optical fiber telecommunications IV. New York: Academic Press; 2002. 4. Willner AE, Pan Z, Yu C. Optical performance monitoring. In: Kaminow IP, Li T, Willner AE, editors. Optical fiber telecommunications VB. New York: Academic Press; 2008 [chapter 7]. 5. Win MZ, Winters JH, Vitetta GM. Equalization techniques for mitigation transmission impairments. In: Kaminow Tingye, Li Tingye, editors. Optical fiber telecommunications IV. New York: Academic Press; 2002. 6. Kikuchi K. Coherent optical communication systems. In: Kaminow IP, Li T, Willner AE, editors. Optical fiber telecommunications V B. New York: Academic Press; 2008 [chapter 4]. 7. Ip E, Lau APT, Barros DJF, Kahn JM. Coherent detection in optical fiber systems. Opt Expr 2008; 16(2):753–91. 8. Shieh W, Bao H, Tang Y. Coherent optical OFDM: theory and design. Opt Expr 2008;16(2):841–59. 9. Jansen SL, Morita I, Forozesh K, Randel S, van den Borne D. Optical OFDM: a hype or is it for real? In: 34th European conference and exhibition on optical communication (ECOC)’08, paper Mo.3.E.3. Brussels; September 2008. 10. Keiser G. Optical fiber communications. 3rd ed. New York: McGraw-Hill; 2000. 11. Andre PS, Pinto AN, Pinto JL. Effect of temperature on the single mode fibers chromatic dispersion. SBMO/IEEE MTT-S Int Microwave Optoelectron Conf 2003;1:231–4. 12. Agrawal GP. Fiber-optic communication systems. New York: Wiley; 2002. 13. Kaiser P, Keck DB. Fiber types and their status. In: Miller SE, Kaminow IP, editors. Optical fiber telecommunications II. New York: Academic Press; 1988 [chapter 2]. 14. Marcuse D, Chraplyvy AR, Tkach RW. Dependence of cross-phase modulation on channel number in fiber WDM systems. IEEE/OSA J Lightwave Technol 1994;12(5):885–90. 15. Tkach RW, Chraplyvy AR, Forghieri F, Gnauck AH, Derosier RM. Four-photo mixing and high-speed WDM system. IEEE/OSA J Lightwave Technol 1995;13(5):841–9. 16. Knudsen SN, Veng T. Large effective area dispersion compensating fiber for cabled compensation of standard single mode fiber. In: Optical fiber communication conference 2000, vol.1; 2000. p. 98–100. 17. Garrett LD. All About chromatic dispersion in dense WDM optical fiber transmission. In: Invited Short Course at Optical Fiber Communication Conference. Anaheim, CA; 2001. 18. Pecci P, Lanne S, Frignac Y, Antona J-C, Charlet G, Bigo S. Tolerance to dispersion compensation parameters of six modulation formats in systems operating at 43 Gbit/s. Electron Lett 2003;39(25):1844–6. 19. Winzer PJ, Essiambre R-J. Advanced modulation formats for high-capacity optical transport networks. IEEE/OSA J Lightwave Technol 2006;24(12):4711–28. 20. Hasegawa A, Tappert F. Transmission of stationary nonlinear optical pulses in dispersive dielectric fibers. Appl Phys Lett 1973;23:142–4. 21. Mollenauer LF, Stolen RH, Gordon JP. Experimental observation of picosecond pulse narrowing and solitons in optical fibers. Phys Rev Lett 1980;45:1095–7. 22. Gordon JP, Haus HA. Random walk of coherently amplified solitons in optical fiber transmission. Optics Lett 1986;11:665–7. 23. Willner AE. Chromatic dispersion and polarization-mode dispersion: managing key limitations in optical communication systems. Opt Photon News OPN Trends 2002;16–21.

References

95

24. Kato T, Koyano Y, Nishimura M. Temperature dependence of chromatic dispersion in various types of optical fibers. In: Optical Fiber Communication Conference (OFC’00), vol. 1; 2000. p. 104–6. 25. Willner AE. The optical network of the future: can optical performance monitoring enable automated, intelligent and robust systems? Opt Photon News 2006;17(3):30–5. 26. Chraplyvy A, Tkach R, Buhl L, Alferness R. Phase modulation to amplitude modulation conversion of CW laser light in optical fibers. IEE Electron Lett 1986;22:409–11. 27. Dimmick TE, Rossi G, Blumenthal DJ. Optical dispersion monitoring technique using double sideband subcarriers. IEEE Photon Technol Lett 2000;12(7):900–2. 28. Petersen MN, Pan Z, Lee S, Havstad SA, Willner AE. Online chromatic dispersion monitoring and compensation using a single inband subcarrier tone. IEEE Photon Technol Lett 2002;14:570–2. 29. Liu N, Zhong W-D, Wen YJ, Li Z. New transmitter configuration for subcarrier multiplexed DPSK systems and its applications to CD monitoring. Opt Expr 2007;15(3):839–944. 30. Murakami M, Imai T, Aoyama M. A remote supervisory system based on subcarrier overmodulation for submarine optical amplifier systems. IEEE/OSA J Lightwave Technol 1996;14:671–7. 31. Dimmick TE, Rossi G, Blumenthal DJ. Optical dispersion monitoring technique using double sideband subcarriers. IEEE Photon Technol Lett 2000;12:900–2. 32. Liu A, Pendock GJ, Tucker RS. Chromatic dispersion monitoring using time-multiplexed in-band RF tones. In: Proc. optical fiber communication conference, paper OThH6. Anaheim, CA; 2005. 33. Tsai KT, Way WI. Chromatic-dispersion monitoring using an optical delay-and-add filter. IEEE/OSA J Lightwave Technol 2005;23(11):3737–47. 34. Wang Y, Pan Z, Sahin AB, Yan L-S, Yu C, Willner AE. In-line chromatic dispersion monitoring using optically-added phase-modulated in-band tones for 10 Gb/s system. In: Proc. optical fiber communication conference, paper WP3. Atlanta, GA; March 2003. 35. Park KJ, Youn CJ, Lee JH, Chung YC. Performance comparisons of chromatic dispersion-monitoring techniques using pilot tones. IEEE Photon Technol Lett 2003;15:873–5. 36. Jun SB, Kim H, Park PKJ, Lee JH, Chung YC. Pilot-tone-based WDM monitoring technique for DPSK systems. IEEE Photon Technol Lett 2006;18(20):2171–3. 37. Yoshimi H, Takushima Y, Kikuchi K. Novel optical fiber dispersion monitor using optical single sideband modulation. In: Conference on lasers and electro-optics (CLEO)’02, paper CThU1, vol. 1. Baltimore, MD; 2002. p. 576–8. 38. Liu N, Zhong WD, Yi X, Wang Y, Lu C. Chromatic dispersion monitoring using the power ratio of two RF tones with a dispersion offset. In: Proc. optical fiber communication conference (OFC)’04, paper MF81. Los Angeles, CA; 2004. 39. Liu A, Pendock GJ, Tucker RS. Chromatic dispersion monitoring using time-multiplexed in-band RF tones, In: Proc. optical fiber communication conference (OFC)’05, paper OThH6. Anaheim, CA; 2005. 40. Park PKJ, Jun SB, Chung YC. A novel chromatic dispersion monitoring technique using frequency-modulated and amplitude-modulated pilot tones. In: Conference on optical fiber communication (OFC)’06, paper JThB37. Anaheim, CA; 2006. 41. Takushima Y, Yoshimi H, Ozeki Y, Kikuchi K, Yamauchi H, Taga H. In-service dispersion monitoring in 32  10.7 Gbps WDM transmission system over transatlantic distance using optical frequency-modulation method. IEEE/OSA J Lightwave Technol 2004;22(1):257–65. 42. Sandel D, Mirvoda V, Wu¨st F, Noe´ R, Weiske C-J. Signed online chromatic dispersion detection at 40 Gb/s based on arrival time detection with a 60-attosecond dynamic accuracy. Electron Lett 2002;38(17):984–6. 43. Ishikawa G, Ooi H. Demonstration of automatic dispersion equalization in 40 Gb/s OTDM transmission. In: Proc. European conference on optical communication. Madrid, Spain; 1998. p. 519–20. 44. Pan Z, Yu Q, Xie Y, Havstad SA, Willner AE, Starodubov DS, et al. Real-time group-velocity dispersion monitoring and automated compensation without modifications of the transmitter. Opt Commun 2004; 230(1–3):145–9.

96

CHAPTER 3 Chromatic dispersion monitoring

45. Kim S-M, Lee C-H. The efficient clock-extraction methods of NRZ signal for chromatic dispersion monitoring. IEEE Photon Technol Lett 2005;17(5):1100–2. 46. Zhao J, Lu C, Li Z, Tam HY, Wai PKA. Optical signal monitoring of DPSK signals using RF power detection. In: Joint conference of opto-electronics and communications conference and Australian conference on optical fibre technology (OECC/ACOFT)’08. Sydney, Australia; 2008. 47. Campillo AL. Chromatic dispersion-monitoring technique based on phase-sensitive detection. IEEE Photon Technol Lett 2005;17(6):1241–3. 48. Tsai KT, Way WI. Chromatic-dispersion monitoring using an optical delay-and-add filter. IEEE/OSA J Lightwave Technol 2005;23(11):3737–47. 49. Nezam SMRM, McGeehan JE, Willner AE. Chromatic dispersion monitoring using partial optical filtering and phase-shift detection of bit rate and doubled half bit rate frequency components. In: Optical fiber communication conference, (OFC)’04, paper ThU2, vol. 2. Los Angeles, CA; 2004. 50. Ku Y-C, Chan C-K, Chen L-K. Chromatic dispersion monitoring technique using birefringent fiber loop. In: Conference on optical fiber communication (OFC)’06, paper OFN2. Anaheim, CA; 2006. 51. Yu Q, Pan Z, Yan L-S, Willner AE. Chromatic dispersion monitoring technique using sideband optical filtering and clock phase-shift detection. IEEE/OSA J Lightwave Technol 2002;20(12):2267–71. 52. Hirota Y, Takushima Y, Matsumoto S, Yoshiara K. Real-time monitoring and fast adaptive compensation of chromatic dispersion in a 40-Gbps optical cross-connect network. In: 31st European conference on optical communication (ECOC)’05, paper We1.2.3, vol. 3. Glasgow, UK; 2005; p. 325–6. 53. Meflah L, Thomsen B, Mitchell J, Bayvel P. Chromatic dispersion monitoring of a multi-channel 40 Gbit/s system for dynamically reconfigurable networks. In: 31st European conference on optical communication (ECOC)’05, paper Th3.2.5, vol. 4. Glasgow, UK; 2005; p. 935–6. 54. Benlachtar Y, Killey RI, Bayvel P. Chromatic dispersion monitoring using synchronous sampling. In: Optical fiber communication/national fiber optic engineers conference (OFC/NFOEC)’06, paper OWK4. Anaheim, CA; 2006. 55. Chen H, Poon AW, Cao X-R. Transparent monitoring of rise time using asynchronous amplitude histograms in optical transmission systems. IEEE/OSA J Lightwave Technol 2004;22:1661–7. 56. Bergano NS, Kerfoot FW, Davidson CR. Margin measurements in optical amplifier systems. IEEE Photon Technol Lett 1993;5:304–6. 57. Weinert CM, Schmidt C, Weber HG. Application of asynchronous amplitude histograms for performance monitoring of RZ signals. In: Optical fiber communication conference (OFC)’01, paper WDD41. Anaheim, CA; 2001. 58. Kikuchi N, Hayase S, Sekine K, Sasaki S. Performance of chromatic dispersion monitoring using statistical moments of asynchronously sampled waveform histograms. IEEE Photon Technol Lett 2005; 17(5):1103–5. 59. Li Z, Li G. Chromatic dispersion and polarization-mode dispersion monitoring for RZ-DPSK signals based on asynchronous amplitude-histogram evaluation. IEEE/OSA J Lightwave Technol 2006;24(7):2859–66. 60. Kawakami H, Yoshida E, Kubota H, Miyamoto Y. Novel signed chromatic dispersion monitoring technique based on asymmetric waveform distortion in DQPSK receiver. In: Joint conference of opto-electronics and communications and australian conference on optical fibre technology (OECC/ACOFT)’08; July 2008. 61. Hanik N, Gladisch A, Caspar C, Strebel B. Application of amplitude histograms to monitor the performance of optical channels. IEE Electron Lett 1999;35(5):403–4. 62. Li Z, Li G. In-line performance monitoring for RZ-DPSK signals using asynchronous amplitude histogram evaluation. IEEE Photon Technol Lett 2006;18(3):472–4. 63. Shake I, Takara H. Averaged Q-factor method using amplitude histogram evaluation for transparent monitoring of optical signal-to-noise ratio degradation in optical transmission system. IEEE/OSA J Lightwave Technol 2002;20(8):1367–73.

References

97

64. Inui T, Komukai T, Mori K, Morioka T. 160-gb/s adaptive dispersion equalization using an asynchronous dispersion-induced chirp monitor. IEEE/OSA J Lightwave Technol 2005;23(6):2039–45. 65. Li Z, Li G. Chromatic dispersion and polarization-mode dispersion monitoring for RZ-DPSK signals based on asynchronous amplitude-histogram evaluation. IEEE/OSA J Lightwave Technol 2006;24(7): 2859–66. 66. Shake I, Takara H. Chromatic dispersion dependence of asynchronous amplitude histogram evaluation of NRZ signal. IEEE/OSA J Lightwave Technol 2003;21(10):2154–61. 67. Anderson TB, Dods SD, Clarke K, Bedo J, Kowalczyk A. Multi-impairment monitoring for photonic networks. In: Proc. Europe conference on optical communication, paper 3.5.1. Berlin; 2007. 68. Dods SD, Anderson TB. Optical performance monitoring technique using delay tap asynchronous waveform sampling, In: Proc. optical fiber communication conference, paper OThP5. Anaheim, CA; 2006. 69. Anderson TB, Dods SD. Asynchronous measurement of chromatic dispersion from waveform distortion, In: Proc. optical fiber communication conference, paper OWN4. Anaheim, CA; 2006. 70. Dods SD, Anderson TB, Clarke K, Bakaul M, Kowalczyk A. Simultaneous and independent monitoring of OSNR, chromatic and polarization mode dispersion for NRZ-OOK, DPSK and Duobinary. In: Proc. optical fiber communication conference, paper OMM5. Anaheim, CA; 2007. 71. Westbrook PS, Eggleton BJ, Raybon G, Hunsche S, Her T. Measurement of residual chromatic dispersion of a 40-Gb/s RZ signal via spectral broadening. IEEE Photon Technol Lett 2002;14(3):346–8. 72. Li S, Kuksenkov DV. A novel dispersion monitoring technique based on four-wave mixing in optical fiber. IEEE Photon Technol Lett 2004;16(3):942–4. 73. Ng TT, Blows JL, Mok JT, McKerracher RW, Eggleton BJ. Cascaded four-wave mixing in fiber optical parametric amplifiers: application to residual dispersion monitoring. IEEE/OSA J Lightwave Technol 2005;23(2):818–26. 74. Yang J-Y, Zhang L, Wu X, Yilmaz OF, Zhang B, Willner AE. All-optical chromatic dispersion monitoring for phase-modulated signals utilizing cross-phase modulation in a highly nonlinear fiber. IEEE Photon Technol Lett 2008;20(19):1642–4. 75. Luo T, Yu C, Pan; Z, Wang Y, McGeehan JE, Adler M, et al. All-optical chromatic dispersion monitoring of a 40-Gb/s RZ signal by measuring the XPM-generated optical tone power in a highly nonlinear fiber. IEEE Photon Technol Lett 2006;18(2):430–2. 76. Vorreau P, Kilper DC, Leuthold J. Optical noise and dispersion monitoring with SOA-based optical 2R regenerator. IEEE Photon Technol Lett 2005;17(1):244–6. 77. Shi Y, Chen M, Xie S. A novel low-power chromatic dispersion monitoring technique employing SOA spectral shift. Opt Commun 2004;230:297–300. 78. Wielandy S, Fishteyn M, Her T, Kudelko D, Zhang C. Dispersion monitoring and automatic compensation based on a differential nonlinear detection scheme. In: Optical fiber communications conference (OFC)’03, paper TuD5, vol. 1. Atlanta, GA; 2003. p. 166–8. 79. Bondarczuk K, Maguire PJ, Barry LP, O’Dowd J, Guo WH, Lynch M, et al. Chromatic dispersion monitoring of 80-Gb/s OTDM data signal via two-photon absorption in a semiconductor microcavity. IEEE Photon Technol Lett 2007;19(1):21–3. 80. Guo WH, Donegan JF, Barry LP. Expanding the range of chromatic dispersion monitoring with two-photon absorption in semiconductors. In: European conference on lasers and electro-optics and the international quantum electronics conference (CLEOE-IQEC)’07. Baltimore, MD; 2007. 81. Bondarczuk K, Maguire PJ, Barry LP, O’Dowd J, Guo WH, Lynch M, et al. High-speed chromatic dispersion monitoring of a two-channel WDM system using a single TPA microcavity. In: Conference on lasers and electro-optics and quantum electronics and laser science (CLEO/QELS)’08, paper CThAA7. San Jose, CA; 2008.

98

CHAPTER 3 Chromatic dispersion monitoring

82. Maguire PJ, Bondarczuk K, Barry LP, O’Dowd J, Guo WH, Lynch M, et al. Dispersion monitoring for high-speed WDM networks via two-photon absorption in a semiconductor microcavity. In: Proc. international conference on transparent optical networks, vol. 4. Nottingham, UK; June 2006. p. 138–41. 83. Meflah L, Thomsen B, Mitchell JE, Bayvel P, Lehmann G, Santori S, et al. Advanced optical performance monitoring for dynamically reconfigurable networks. In: Proc. 10th European conference on networks and optical communications. London; 2005. p. 554–61. 84. Woodward S, Nelson LE, Feuer MD, Zhou X, Magill PD, Foo S, et al. Characterization of real-time PMD and chromatic dispersion monitoring in a high-PMD 46-Gbps transmission system. IEEE Photon Technol Lett 2008;20(24):2048–50. 85. Kuschnerov M, Hauske FN, Piyawanno K, Spinnler B, Napoli A, Lankl B. Adaptive chromatic dispersion equalization for non-dispersion managed coherent systems. In: Optical fiber communication conference and exposition and the national fiber optic engineers conference (OFC/NFOEC)’09, paper OMT1. San Diego, CA; 2009. 86. ITU-T Recommendation O.201. 87. Wiesmann R, Bleck O, Heppner H. Cost effective performance monitoring in WDM systems. In: Proc. optical fiber communication conference. Baltimore, MD; 2000. p. 171–3. 88. Downie JD, Tebben DJ. Performance monitoring of optical networks with synchronous and asynchronous sampling, In: Proc. optical fiber communication conference, paper WDD50. Anaheim, CA; 2001. 89. Jones WB Jr. Introduction to optical fiber communication systems. New York: Holt, Rinehart, and Winston; 1988. 90. Ishihara K, Kobayashi T, Kudo R, Takatori Y, Sano A, Yamada E, et al. Coherent optical transmission with frequency-domain equalizers, In: European conference on optical communication (ECOC)’08, paper We.2.E.3. Brussels, Belgium; 2008. 91. Geyer JC, Fludger CRS, Duthel T, Schulien C, Schmauss B. Performance monitoring using coherent receivers. In: Optical fiber communication conference and exposition and the national fiber optic engineers conference (OFC/NFOEC)’09, paper OThH5. San Diego, CA; 2009. 92. Pendock GJ, Yi X, Yu C, Shieh W. Dispersion-monitoring in WDM systems by injecting modulated ASE. IEEE Photon Technol Lett 2008;20(10):821–3. 93. Fu B, Hui R. Fiber chromatic dispersion and polarization-mode dispersion monitoring using coherent detection. IEEE Photon Technol Lett 2005;7:1561–3. 94. Yang S-D, Jiang Z, Weiner AM, Parameswaran KR, Fejer MM. Ultrasensitive chromatic dispersion monitoring for 10 GHz pulse train by quasi-phase-matched LiNbO3 waveguides. IEE Electron Lett 2005;41:554–6. 95. Dinu M, Kilper DC, Stuart HR. Optical performance monitoring using data stream intensity autocorrelation. IEEE J Lightwave Technol 2006;24:1194–202. 96. Westlund M, Andrekson PA, Sunnerud H, Hansryd J, Li J. High-performance optical-fiber-nonlinearitybased optical waveform monitoring. IEEE/OSA J Lightwave Technol 2005;23(6):2012–22. 97. Youn C. Effects of SPM and PMD on chromatic dispersion monitoring techniques using pilot tone. In: Optical fiber communications conference (OFC)’03, paper WP2, vol. 4. Atlanta, GA; 2003. p. 403–4. 98. Shi Y, Chen M, Xie S. Simultaneous polarization mode dispersion and chromatic dispersion monitoring method in 40 Gbit/s system by polarization modulation. In: Optical fiber communication conference (OFC)’06, paper JThB39. Anaheim, CA; March 2006. 99. Liu N, Zhong W-D, Wen YJ, Wang Y. PMD-insensitive chromatic dispersion monitoring. In: 31st European conference on optical communication (ECOC)’05, paper We4.P.001, vol. 3. Glasgow, UK; 2005. p. 503–4. 100. Park KJ, Lee JH, Youn CJ, Chung YC. A simultaneous monitoring technique for polarization-mode dispersion and group-velocity dispersion. In: Optical fiber communication conference and exhibit (OFC)’02, paper WE4. Anaheim, CA; 2002. p. 199–200.

References

99

101. Devaux F, Sorel Y, Kerdiles JF. Simple measurement of fiber dispersion and of chirp parameter of intensity modulated light emitter. IEEE/OSA J Lightwave Technol 1993;11:1937–40. 102. Liu N, Zhong W-D, Wen YJ, Lu C, Cheng L, Wang Y. PMD and chirp effects suppression in RF tonebased chromatic dispersion monitoring. IEEE Photon Technol Lett 2006;18:673–5. 103. Lize´ YK, Christen L, Yang J-Y, Saghari P, Nuccio S, Willner AE, et al. Simultaneous monitoring of chromatic dispersion and PMD for OOK and DPSK using partial-bit-delay-assisted clock tone detection. In: Proc. European conference on optical communication (ECOC), paper Mo4.4.7. Cannes; September 2006. 104. Lize YK, Christen L, Yang J-Y, Saghari P, Nuccio S, Willner AE, et al. Independent and simultaneous monitoring of chromatic and polarization-mode dispersion in OOK and DPSK transmission. IEEE Photon Technol Lett 2007;19(1):3–5. 105. Lize´ YK, Yang J-Y, Christen LC, Wu X-X, Nuccio S, Wu T, et al. Simultaneous and independent monitoring of OSNR, chromatic and polarization mode dispersion for NRZ-OOK, DPSK and duobinary. In: Proc. optical fiber communications conference, paper OWB4. Anaheim, CA; March 2007.

This page intentionally left blank

CHAPTER

Polarization mode dispersion monitoring

4 Changyuan Yu

National University of Singapore, Singapore A*STAR Institute for Infocomm Research, Singapore

4.1 INTRODUCTION Polarization mode dispersion (PMD) is an important impairment in high-speed reconfigurable optical networks. PMD is based on the fact that a given signal spectral component of the optical data propagates along two identifiable principal states of polarization (PSPs) within a non circular, symmetric fiber core, and these two spectral copies travel down the fiber at slightly different speeds. PMD accumulates due to fiber and any slight birefringence of many in-line components. By its inherent nature, deleterious PMD effects are stochastic, time varying, vibration dependent, and temperature dependent. Moreover, these effects include first- and higher-order components. Therefore, PMD monitoring is required in order to either dynamically tune a compensator or simply to determine the network location that must be diagnosed and repaired.1 This chapter provides a review of PMD monitoring techniques. Single-mode fibers actually support two perpendicular polarizations of the original transmitted signal (fundamental mode). In an ideal fiber (perfect), these two modes are indistinguishable, and have the same propagation constants owing to the cylindrical symmetry of the waveguide. However, practical fibers are not perfect and, as a result, the two perpendicular polarization components (along the two PSPs) of the signal light may travel at different speeds and consequently arrive at the end of the fiber at different times. This phenomenon is called PMD. As shown in Figure 4.1, the major cause of PMD is the asymmetry of the fiber optic strand. Asymmetry is a description of the fact that the fiber core is slightly out-of-round, or oval. Fiber asymmetry may be inherent in the fiber from the manufacturing process, or it may be a result of mechanical stress on the deployed fiber. The inherent asymmetries of the fiber are fairly constant over time, while the mechanical stress due to movement of the fiber can vary, resulting in a dynamic aspect to PMD. The mechanical stress on the optical fiber can originate from a variety of sources. One source that is very difficult to control is the day/night and seasonal heating and cooling of the optical fiber. Although much fiber is deployed in the ground and often within conduits, it is still subject to temperature variations and corresponding mechanical stress. Another source of mechanical stress can originate from nearby sources of vibration. For example, much fiber is deployed alongside railroad tracks because of the ease of right-of-way and construction. © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00004-3

101

102

CHAPTER 4 Polarization mode dispersion monitoring

Geometric

Internal stress

External stress

Bend Twist

FIGURE 4.1 Origin of PMD.

However, vibration from passing trains can contribute to stress on the optical fiber. In addition, wind can cause swaying of the fiber cable and can contribute to PMD of fiber deployed aerially. The difference in propagation constants of the two PSPs is responsible for PMD in the fiber, and can be related to the difference in refractive indices between the two orthogonal polarization axes as bo
be ¼ ono =c
one =c ¼ oDneff =c;

(4.1)

where no and ne are the effective refractive indices of two orthogonal states, and △neff is the differential index of refraction. The differential index of refraction is a measure of birefringence in the fiber, and is usually between 10
7 and 10
5. As the first-order approximation, this birefringence causes the power in each optical pulse to split between the two PSPs of the fiber and travel at different speeds, creating a differential group delay (DGD) between the two modes that results in pulse spreading and intersymbol interference (ISI), as shown in Figure 4.2. Since PMD is caused by the different transmission speeds of the signal’s two states of polarization (SOPs) as they propagate along a fiber having a small birefringence, and the birefringence of a fiber changes randomly along a fiber link, PMD is a statistically random quantity. PMD can be characterized by DGD between the two PSPs after a given length of fiber. Because of random variations in the perturbations along a fiber span, PMD in long fiber spans accumulates in a random-walk-like process that leads to a square root of transmission-length dependence.2 Therefore, PMD does not

Fast PSP 1

PSP 2 Slow

Side view

DGD = Time delay

FIGURE 4.2 Illustration of input optical pulse with power transmitted on the two PSPs, each arriving at a different time.

4.1 Introduction

103

have a single value for a given span of fiber. Rather, it is described in terms of average DGD, and a fiber has a distribution of DGD values over time. The probability of the DGD of a fiber section being a certain value at any particular time follows a Maxwellian distribution (see Figure 4.3). The probability of DGD ¼ Dt is given by " # 32Dt2 4Dt2 exp ; (4.2) PDFðDtÞ ¼ p2 hDti3 phDti2

Probability of DGD

Ambient temperature (°C) Bit error rate

with mean value hDti. PMD is usually expressed in ps/km1/2 in long fiber spans, and the typical value of hDti is 0.1–10 ps/km1/2.3,4 In addition to the time variance of DGD, PMD also varies over wavelengths, known as higherorder PMD. This variance results in an optical dispersion that is a function of both the channel bandwidth and the value of DGD over that bandwidth. In the presence of higher-order PMD, an input pulse can spread and become depolarized, as the different frequencies making up the pulse spectrum will experience a slightly different polarization evolution as they propagate down the fiber, due to the frequency-dependent nature of DGD and PSPs.5–9 Figure 4.4 is a graphical representation of

Average (mean)

DGD

10–5

Signal through fiber

10–6 10–7 Signal through reference

20

Sunset Sunrise

10 00

(a)

12 4 8 Elapsed time (hours) (b)

16

FIGURE 4.3 (a) Probability distribution of DGD in typical fiber. (b) System performance (BER) fluctuations due to changes in temperature caused by PMD. Copyright © 1991 IEEE.

At distance = 0

At distance = L Higher-order PMD

First-order PMD

FIGURE 4.4 Graphical representation of all-order PMD effect on an optical pulse.

104

CHAPTER 4 Polarization mode dispersion monitoring

the effect of PMD (both first and higher order) on an optical pulse. The optical pulse and its constituent photons travel from the source, or transmitter, at distance ¼ 0, along the single-mode optical fiber. After some distance L PMD has affected the pulse, the two orthogonal polarization states of the light are separated by some time (i.e., DGD) and also broadened by high-order PMD. If PMD is severe, the receiver at some distance L cannot accurately decode the optical pulse, and bit errors can result. If the bit errors caused by PMD are too numerous, then the transmitted information is too corrupt to recover and the transmission link should be considered out of service. In a cascaded fiber link, besides the PMD of fiber, there may be many discrete components (i.e., isolators, couplers, wavelength multiplexers), which are also polarization dependent due to molecular asymmetry (anisotropy) of the waveguide material. Although PMD caused by polarization dependence of a single component may be negligible, cascaded components may add significant PMD in a long link. The combined PMD-induced broadening in a long link may be up to a few tens of ps, which can degrade systems operating at 10 Gb/s. In systems operating at 40 Gb/s, PMD has been proved to be detrimental.10 Figure 4.5 shows a plot of fiber PMD versus maximum transmission distance for a 40-Gb/s non-return-to-zero (NRZ) system. In order to enable ultrafast communications over long distances of optical fiber, the critical issue of PMD must be considered. PMD is also an issue in subcarrier multiplexed (SCM) transmission systems, especially in millimeter wave systems where the modulation frequency is high enough to be strongly affected by the severe radio frequency (RF) power fading induced by PMD.11,12 Many reports show that the transmission of analog and digital SCM signals over fiber can be severely affected by PMD and polarization-dependent loss (PDL).12–14 In general, double-sideband (DSB) transmission will be affected by chromatic dispersion (CD) because of the relative time (phase) shift that develops between the 1

1 ps PMD in amplifier sites No PMD in amplifier sites Fiber PMD (ps/km1/2)

0.8

No PMD in components

1 ps PMD in components

Old fiber PMD = 0.5 ps/km1/2

26 km

26 km

New fiber PMD = 0.1 ps/km1/2

625 km

320 km

Future fiber PMD = 0.05 ps/km1/2

2500 km

480 km

Fiber type 0.6

0.4

Distance

0.2

0

0

100

200 300 400 Maximum transmission distance (km)

500

600

FIGURE 4.5 Transmission distance limitations for a 40-Gb/s NRZ system due to combination of fiber PMD and PMD of cascaded in-line optical components found in amplifier sites. Copyright © 2004 IEEE /OSA.

4.1 Introduction

105

upper and lower sidebands due to frequency-dependent velocities in the fiber. To avoid this, many SCM systems employ single-sideband (SSB) subcarriers and are therefore relatively immune to CD.15 However, PMD-induced RF power fading remains a problem even for SSB signals, since the relative polarization state of the carrier to the sideband changes through the transmission fiber.16,17 For example, in a 40-GHz optical SCM system, the RF power after detection is completely faded over a fiber link with an average DGD of only 5 ps when the instantaneous DGD is only 12.5 ps, near the tail of the Maxwellian DGD distribution. The deleterious PMD-induced power-fading effect in SCM can be described in the time domain as follows. The light can be decomposed along two orthogonal PSPs, with one axis traveling faster than the other. The time delay between the faster and slower axes causes a phase difference in the corresponding received subcarrier signals. This phase difference induces destructive interference and may lead to serious RF power fading that is a function of subcarrier frequency and accumulated DGD. Another explanation involves the polarization state in the optical frequency domain. PMD-induced RF power fading occurs when the polarization state of the optical carrier wave is not aligned with the polarization state of the subcarrier, since PMD will cause the state of polarization (SOP) of the carrier and subcarrier to wander at different rates, as shown in Figure 4.6. To ensure robust transmission of SCM signals through long fiber links, PMD monitoring and mitigation are necessary.17–19 Within optical fiber communication systems, another polarization effect, PDL, expressed as the ratio of the maximum to minimum transmission on a log scale as the launch polarization is rotated through all possible states, has also been recognized as a critical polarization-related impairment. Due to non-negligible PDL in various in-line optical components, such as switches, isolators, couplers, filters, and circulators, when the optical pulse passes through these optical components, it splits between two orthogonal polarization modes that attenuate each optical pulse replica differently. PDL can cause deleterious effects in a fiber transmission link, such as optical power fluctuations resulting in random optical signal-to-noise ratio (OSNR) variations due to polarization-state wandering during propagation, and limited PMD compensator performance. On the other hand, there is polarizationdependent gain (PDG) in optical communication systems, which is due to anisotropic gain saturation in fiber amplifiers and appears in amplified fiber links. The source of PDG in erbium-doped fiber amplifiers (EDFAs) has been identified as polarization hole burning (PHB): signals with orthogonal Fiber output SSB subcarrier transmitter

SOP changes while propagating through the fiber

l

SOPcarrier No fading

l

SOPcarrier SOPSSB Partial fading

Carrier SSB

Fiber link with PMD

l

SOPSSB

Optical spectrum Change of SOPcarrier

l SOPcarrier ^SOPSSB

Total fading!

Change of SOPSSB

FIGURE 4.6 Explanation of PMD-induced RF power fading in an SSB SCM system in optical domain.

Copyright © 2004 IEEE.

106

CHAPTER 4 Polarization mode dispersion monitoring

states of polarization can utilize different subsets of gain-producing ions.20 PDG can randomly degrade the OSNR, inducing significant fluctuations in the BER over time. Although the PDG from a single amplifier is quite small and negligible, the PDG effects from cascaded amplifiers in the overall optical link can result in a several-dB fluctuation in the received Q-factor. Note that the interaction between PMD and PDL/PDG may lead to significant overall performance degradation, which dramatically surpasses the result of adding the degradations induced by the two impairments independently.21–25 When PDL exists in a fiber link also impaired by PMD, the PSPs of the fiber may be no longer orthogonal to each other, the probability distribution of the DGD degenerates from its Maxwellian shape, and the PDL seen by the wavelength-division multiplexing (WDM) channels may become uncorrelated. PMD- and PDL/PDG-induced problems can be reduced simply by regeneration—that is, shortening the optical transmission distance. However, from a network point of view, a regeneration site is an inefficient and costly optical-electronic conversion site. Adding to the expense and inefficiency of a regeneration site is the fact that most long-haul transmission systems are now multiwavelength, dense wavelength-division multiplexing (DWDM) systems. In this application, the transmission link must first be demultiplexed, then regenerated, and then multiplexed again. This is a very costly operation compared to the preferred alternative of a multiwavelength amplifier. From a network and cost perspective, a more efficient method of addressing the PMD problem is to fix the effects of PMD while the transmission is in an optical state, before a receiver tries to decode the bits. A PMD compensator (PMDC) deployed at the destination of the transmission system can reduce the effects of the PMD in the fiber, and ensure that the optical bits are correctly decoded by the receiver before they are to be routed and switched. The most reliable and efficient PMDC technology is the use of adaptive optics to realign and correct the pulses of dispersed optical bits, which need PMD monitoring to provide control signals. In summary, deleterious PMD effects are stochastic, time varying, temperature dependent, and worsen as the bit rate rises. Moreover, the instantaneous first-order PMD (i.e., DGD) follows a Maxwellian probability distribution, always with some finite possibility of a network outage, and it can interact with other polarization-related impairments such as PDL and PDG. In addition, reconfigurable optical networking will change the signal path and thus the accumulation of fiber-based effects including PMD, and periodic repair and maintenance of the fiber plant will alter the fiber itself. Since the accumulated PMD can vary on a millisecond time scale,26 it requires real-time PMD monitoring in order to either dynamically tune a compensator or for network control and management. A PMD monitor that is simple, fast (response time less than a millisecond), sensitive, and robust would facilitate a proper system performance.

4.2 PMD MONITORING BASED ON MEASUREMENT OF RF TONE Based on spectral analysis such as RF tones, a number of monitoring techniques have been demonstrated to provide appropriate control signals for PMD mitigation.27–32 A given optical frequency component splits on two orthogonal PSPs and each replica travels down the fiber with a different speed that dephases these two replicas. This effect reduces the corresponding spectral component in the detected RF power spectrum through destructive interference. Therefore, the RF power is a function of PMD, which can be used for PMD monitoring.

4.2 PMD monitoring based on measurement of RF tone

107

Considering the phase difference between the two RF sidebands and the PMD and chirp effects, the detected RF power with DSB at the photodetector is given by27,33–35 2 PDSB ¼ P0 ½1
4gð1
gÞ sin2 ðpfRF DtÞ  ð1 þ a2 Þ cos2 ðpDtotal l2 fRF =c þ arctan aÞ;

(4.3)

where P0 is the RF power without CD and PMD effects, which depends on the loss or gain that the signal experiences; g and △t are PMD-related factors: the power splitting ratio and DGD between the two PSPs; a is a parameter that relates the instantaneous intensity-induced phase variation of the modulated light, also known as chirp parameter; Dtotal is the total CD induced by fibers and other optical components; fRF is the RF frequency; l is the carrier wavelength; and c is the speed of light in a vacuum. This equation indicates that the faded RF power could be employed in PMD monitoring, but the CD may considerably influence the RF power, which leads to monitoring errors for the RF tone-based PMD monitoring techniques. Figure 4.7 shows the RF power variation as a function of DGD for different frequency components, when CD is a constant. As an example, for return-to-zero (RZ) data, its strong RF clock tone can be used as a monitoring signal. Unfortunately, as shown in Equation (4.3), CD will also affect the power in the RF tone, since dispersion causes a relative time delay between the upper- and lower-frequency optical clocks. After power detection, the RF clock will fade when these two optical clocks are out of phase due to CD. Therefore, CD will cause ambiguity in PMD monitoring using the recovered RF tone. A potential solution to CD-insensitive PMD monitoring is to use a narrowband optical filter centered at either the upper or the lower optical clock tones before a photodetector,36 which is shown in Figure 4.8. Since only one optical clock tone frequency is detected, any CD effects are negated. The RF clock power is then due solely to the beating of one optical clock tone and the carrier. Because the first-order PMD effect still causes power fading for single-sideband signals, this technique can be used to monitor PMD. For a 10-Gb/s system, a narrow filter is centered at the upper 10-GHz optical clock tone and the power of the 10-GHz RF clock is measured to monitor the PMD. As shown in Figure 4.9, the RF tone power gives accurate DGD values and is insensitive to CD up to 640 ps/nm.

Relative frequency power

1.0 0.125Rb 0.8

(g = 0.5) 0.5Rb

Rb

0.6 0.4

0.25Rb

0.2 0.0

0

1

2 DGD/Tb

3

4

FIGURE 4.7 Received RF power variation versus DGD for eighth, quarter, half, and bit rate frequency components. Copyright © 2004 IEEE /OSA.

108

CHAPTER 4 Polarization mode dispersion monitoring

Electrical domain

~DSB

Optical spectrum

Without filter

Detection

SMF l

f Clock fades with PMD, CD

l

~SSB Detection

Electrical spectrum

f

f Clock fades with PMD only

With partial filtering l

FIGURE 4.8 Concept of CD-insensitive RF power fading using optical bandpass filtering.

10 DGD = 0 ps 23 ps

–10 –20

43 ps

–30

PMD = 23 ps CD = 0 ps/nm

–40

–49.3 dBm

PMD = 23 ps CD = 640 ps/nm

faded

Relative clock power (dB)

Relative clock power (dB)

10 0

Copyright © 2004 IEEE.

0

DGD = 0 ps 23 ps

–10 –20

43 ps

–30

PMD = 23 ps CD = 0 ps/nm

PMD = 23 ps CD = 640 ps/nm

–40

–43.8 dBm

–43.9 dBm

–50

–50 0 160 320 480 640 Chromatic dispersion (ps/nm)

0 160 320 480 640 Chromatic dispersion (ps/nm)

(a)

(b)

FIGURE 4.9 CD-induced RF clock power fading under various DGD values (a) without bandpass filtering and (b) with bandpass filtering. Insets are RF clocks when DGD is 23 ps and CD is 0 and 640 ps/nm, respectively. Copyright © 2004 IEEE.

On the other hand, the clock tone does not appear at the receiver for NRZ data. A dispersive, fiber Bragg grating (FBG) is used to generate the required clock tone for PMD monitoring,37 although this technique is still CD sensitive. Another technique is using a narrowband FBG notch filter to filter off one of the optical clock sidebands; the RF clock tone can be recovered from the beating between the carrier and the remaining optical clock sideband. Again, the recovered RF clock power depends on the relative polarization state of the carrier to the optical clock sideband, which is determined by PMD of the transmission link. CD only affects the phase of the recovered RF clock tone but not the amplitude. Therefore, the recovered RF clock power can be used as a PMD monitoring signal, and is insensitive to CD.38 Excluding the effects on the clock tone, the RF power fading due to PMD also generates a notch in the electrical spectrum after detection. A number of these notch components (also called dip frequencies) have been proposed as PMD monitors in a link.27,28 From Equation (4.3), we can see that if the DGD of the system is large enough, the dip frequency can be measured more easily and

4.2 PMD monitoring based on measurement of RF tone

109

PRBS

Attenuator Receiver MOD

FMLL

PC DGD emulator

Attenuator ASE

OSNR and PMD monitoring module

RF amplifier RF spectrum analyzer Note: Dip position shift Dip minimum power

90:10

EDFA

Receiver

Scrambling 99:1 PM fiber

Power meter

Incoming signal

Large DGD Polarization Optical element controller filter

FIGURE 4.10 Experimental setup of simultaneous PMD and OSNR monitoring through enhanced RF spectrum analysis by adding large DGD element. FMLL, fiber mode–locked laser. Copyright © 2005 IEEE.

accurately, thus facilitating derivation of the DGD as the dip frequency is pushed into the more readily accessible low-frequency regime of the signal’s RF spectrum. In order to improve the monitoring sensitivity, a large-DGD element can be intentionally introduced at the monitoring module to increase the total DGD value.39 At the same time, OSNR can be monitored based on the orthogonal delayed-homodyne principle in this scheme.40 Figure 4.10 shows the demonstrated module for simultaneously monitoring PMD and OSNR. The monitoring signal is sent to an optical filter, a polarization controller (PC), and then a piece of polarization-maintaining (PM) fiber with a large DGD value. To monitor PMD, the PC in front of the PM fiber is scrambled, and the position shift of the RF spectral dip (i.e., the minimum dip frequency fmin and the maximum dip frequency fmax) are measured by an RF spectrum analyzer. By considering the transmission fiber as one span of fiber (with DGD value of b1) and the PM fiber in the monitoring module as another (with DGD value of b2), the maximum and the minimum overall DGD (DGDmax and DGDmin) can be expressed as DGDmax ¼ b1 þ b2 DGDmin ¼ jb1
b2 j:

(4.4)

As DGDmax is proportional to fmax while DGDmin is proportional to fmin, the DGD values of b1 and b2 can be calculated from Equation (4.4). The cascade of a large DGD element can help move the dip position to the low-frequency regime of the spectrum even when the transmission link DGD is small. Monitoring at low-frequency components with an improved sensitivity is highly desirable because it not only eliminates the use of high-speed electronics, but also desensitizes the monitoring to higherorder PMD. In addition, since the position shift, instead of the absolute RF power, of the dip is measured, the PMD monitoring scheme is strictly dependent on the transmission link DGD and is

110

CHAPTER 4 Polarization mode dispersion monitoring

60

5

0

40

–5

20 0

Monitoring errors (ps)

DGD by proposed method (ps)

10 OSNR = 35 dB/0.1 nm OSNR = 30 dB/0.1 nm OSNR = 25 dB/0.1 nm OSNR = 20 dB/0.1 nm OSNR = 15 dB/0.1 nm

80

–10

0

20 40 60 DGD by PMD emulator (ps)

80

FIGURE 4.11 PMD monitoring results for 10-Gb/s, 2.5% RZ data by adding large DGD element. OSNR varies from 15 to 35 dB. Copyright © 2005 IEEE.

independent of the signal bit rate. Therefore, no prior knowledge between the RF power level and the DGD values is needed. The PMD monitoring is thus quite robust for different OSNR levels of the signal. Figure 4.11 shows the experimental results of PMD monitoring by this technique. Another interesting CD-insensitive PMD monitor has been reported recently for differential phaseshift keying (DPSK) and differential quadrature phase-shift keying (DQPSK).41,42 As shown in Figure 4.12, DGD causes walk-off in time between the two PSPs. After passing through a polarization beam splitter (PBS), the two polarizations are beating together, resulting in a periodic filter response. Polarization beam splitter

Fiber

RF spectrum analyzer

(a) Optical spectrum

Δt

t

RF spectrum High DGD Low DGD

Power

DGD

lcarrier

f 170 MHz

f

(b)

FIGURE 4.12 (a) Conceptual diagram of PMD monitoring technique for DPSK/DQPSK. (b) RF power increases with decreasing FSR of polarization-based interferometer filter (i.e., with increasing DGD values). Copyright © 2008 IEEE.

4.3 PMD monitoring based on measurement of degree of polarization

–40

–40

– 45

–45

10-Gb/s NRZ-DPSK 20-Gb/s NRZ-DQPSK

10-Gb/s NRZ-DPSK

DGD = 40 ps

–50 –55 20-Gb/s NRZ-DQPSK –60

RF power (dBm)

RF power (dBm)

111

–50 –55 –60 DGD = 23 ps

–65

–65 –70

–70 0

20

40

60

80

100

0

100 200 300 400 500 600 700

DGD (ps)

Chromatic dispersion (ps/nm)

(a)

(b)

FIGURE 4.13 Experimental results of (a) RF power measured at 170 MHz for PMD monitoring of NRZ-DQPSK and NRZ-DPK, and (b) CD dependence for PMD monitoring with DGD 23 ps and 40 ps, respectively. Copyright © 2008 IEEE.

The resulting free spectral range (FSR) is equal to 1/Dt, where Dt is the DGD experienced throughout the link. As a result of this filtering effect, the spectral content of the transmitted RF spectrum changes with DGD. As the DGD of the link increases, the RF power also increases. Therefore, it is possible to measure variations in power level of the RF spectrum and correlate these variations to the first-order PMD of the link. Because the RF content is not impacted much by CD at lower RF frequencies, this monitoring technique is relatively insensitive to CD. Besides the dispersion insensitivity, this monitoring scheme operates on very low frequencies, such as 170 MHz in the experiment, and therefore does not require any high-speed components, such as a high-speed photodiode. The proposed PMD monitoring technique was demonstrated in both 20-Gb/s NRZ-DQPSK and 10-Gb/s NRZ-DPSK systems. Figure 4.13(a) illustrates that the RF power at 170 MHz increases by 20 dB in the presence of 0–100 ps of DGD, for both NRZ-DQPSK and DPSK. Figure 4.13(b) illustrates the sensitivity of the proposed scheme to CD. The measured RF powers at 170 MHz with 23 and 40 ps DGD vary within 1 dB in the presence of 0–650 ps/nm CD.

4.3 PMD MONITORING BASED ON MEASUREMENT OF DEGREE OF POLARIZATION The degree of polarization (DOP) is given by the ratio of the power of the polarized part of the light to the total power of the light. By measuring the DOP of the received signal light, one can evaluate the signal affected by PMD.43,44 The use of the DOP to monitor the effects of PMD has a number of advantages over other techniques, including the following:

112

CHAPTER 4 Polarization mode dispersion monitoring

y

1

0

1

y

1

SOP

A

time x x (a) y

1

0

1

y

1

B(t = TS) time

SOP

A(t = Tm)

x x DGD

Ts

Tm (b)

Te

C(t = Te)

FIGURE 4.14 Schematic illustration of DOP signal degradation by PMD. (a) Optical waveform and SOP of signal without PMD. (b) Optical waveform and SOP of signal with PMD. The x and y axes correspond to two PSPs of the transmission media. Copyright © 2001 IEEE /OSA.

1. 2. 3. 4.

No need for high-speed devices. Simplicity. Bit rate independent. Unaffected by other degrading effects such as CD or chirp.45

The mechanism of DOP degradation of the signal via PMD is illustrated in Figure 4.14. Without PMD, the optical signal is pure polarized light with a single SOP, as in Figure 4.14(a). With PMD, the SOP at the edge (SOP B and SOP C) and at the mid point (SOP A) of “1”s of NRZ signal become different due to the PMD-induced pulse shift, as in Figure 4.14(b). The decreased amount of DOP of the signal corresponds to the amount of signal pulse distortion caused by PMD. Figure 4.15 shows the DOP as a function of DGD for 10-Gb/s NRZ data modulated by LiNbO3 Mach-Zehnder (MZ) modulator, which is insensitive to chirp and CD in the fiber. Note that the DOP is pulse-width dependent.46,47 As an example, for an ideal 50% RZ data pulse, as the signal (launched at 45 with respect to the PSPs of the DGD element or fiber link) passes through the DGD element, the “fast” and “slow” halves of a given pulse travel at different speeds and thus no longer overlap. As this overlap decreases, the signal replicas on the two PSPs become more and more depolarized, and the measured DOP is reduced. If the signal experiences DGD equal to or greater than half of the bit time, but less than the full bit duration (for 10 Gb/s, this would be 50 ps), the two signal replicas no longer coincide in any way, and the measured DOP is zero. Thus, for RZ signals there is a limit to the monitoring range of a DOP-based DGD monitor—once the DGD equals or exceeds the pulse width (regardless of the bit rate), the measured DOP will reach its first minimum point. On the other hand, an NRZ signal can be considered a “100%-duty-cycle” RZ signal, and can never be completely depolarized, even after experiencing >1-bit duration’s worth of DGD. Due to the pseudo-random nature of the data, the minimum DOP is 0.5. Figure 4.16 shows theoretical results of minimum DOP versus DGD (relative to the bit time, Tb) as the pulse width of an RZ signal varies.

4.3 PMD monitoring based on measurement of degree of polarization

1

g = 0.5

0.8

DOP

113

MZ(a=0)

0.6

Rectangular(α=0) 0.4

a > 0; 0 ps/nm a > 0; +350 ps/nm a > 0; –350 ps/nm a < 0; 0 ps/nm a = 0; 0 ps/nm

0.2

MZ(a>0,a<0)

0 0

50

100 DGD (ps)

150

200

FIGURE 4.15 DOP as function of DGD for 10-Gb/s NRZ data modulated by MZ modulator. Plots, experiment; dashed line, rectangular waveform approximation; thin lines, numerical simulation. All simulated DOP curves are relatively unaffected by chirp parameter a and by fiber dispersion of 350 ps/nm. Copyright © 2001 IEEE /OSA.

W = Tb W = 0.5 Tb W = 0.3 Tb W = 0.1 Tb

1.0

DOP

0.8 0.6 0.4 0.2 0.0 0

0.2

0.4 0.6 0.8 DGD / Tb (ps)

1

FIGURE 4.16 Theoretical results of minimum DOP versus DGD (relative to bit time, Tb) as pulse width of RZ signal varies. Copyright © 2004 IEEE /OSA.

Therefore, DOP-based techniques may suffer from the following disadvantages: a small DGD monitoring range for short-pulse RZ signals, and a lack of sensitivity for NRZ signals, as shown in Figure 4.17. In addition, they are affected by higher-order PMD. As shown in Figure 4.18, the maximum DOP equals 1 when the SOP of the input signal aligns with the fiber’s PSP (first-order only). The minimum DOP depends on the data rate, modulation format, and so on. While higherorder PMD is included, it decreases the signal’s maximum DOP at the receiver less than unity,48 which also limits the sensitivity of monitoring.

114

CHAPTER 4 Polarization mode dispersion monitoring

1.0 10 Gb/s RZ 40 Gb/s RZ 160 Gb/s RZ

RZ

0.5

0.0

10 Gb/s NRZ

DOP

DOP

1.0

0

5

10 DGD (ps)

15

20

NRZ

0.5

0.0

0

20

40 60 DGD (ps)

80

100

FIGURE 4.17 Sensitivity of DOP reduction as a function of DGD (first-order PMD).

Maximum DOP = 1

Copyright © 2001 IEEE /OSA.

Maximum DOP = 0.61 Minimum DOP = 0.21

Minimum DOP = 0.21

First-order PMD only (a)

Higher-order PMD included (b)

FIGURE 4.18 Measured DOP reduction with scrambled input polarization of 40-Gb/s RZ signal. (a) First-order PMD of 1.25 ps. (b) Second-order PMD, concatenation of two unaligned birefringent secitons (6-ps and 4-ps DGD). Copyright © 2001 IEEE /OSA.

It would be highly desirable to have a PMD monitor for which the DOP can be measured to obtain a wide DGD monitoring window and wide dynamic range for both RZ and NRZ signals. Partial optical filtering has been proposed to dramatically enhance the DGD monitoring range and/or DOP dynamic range/sensitivity to DGD in DOP-based DGD monitors.46,49,50 A single optical filter is added in the monitoring tap-off line, centered either on the optical spectrum (“symmetrically”) or offset from the center of the spectrum by the bit rate frequency—that is, for 10-Gb/s data, offset by 10 GHz (“asymmetrically”). PMD monitoring is based on the DOP measurement of the portion of the optical spectrum that is passed by the filter—a “partial signal spectrum” DOP measurement. This technique is applied to 10-Gb/s carrier-suppressed RZ (CSRZ) and alternate-chirped RZ (ACRZ) signals for which the pulse widths are half of the bit time and the results show that the monitoring range can be extended from 50 to 70 ps and from 70 to 100 ps, respectively. Moreover, the monitoring range of the 25-ps pulse in 20-Gb/s RZ signals is extended from 26 to 45 ps. Using asymmetric optical filtering in NRZ systems can double the DOP dynamic range without any reduction in DGD monitoring range.

4.3 PMD monitoring based on measurement of degree of polarization

115

We solve these problems using a single added optical element, an optical filter placed in the monitoring tapline, that is centered either on the optical spectrum (symmetrically) or offset from the center of the spectrum by the bit rate frequency, which, for 10-Gb/s data, is offset by 10 GHz (asymmetrically). We measure the DOP of the portion of the optical spectrum that is passed by the filter, thereby performing a partial-signal-spectrum DOP measurement. As multiplication in the frequency domain corresponds to convolution in the time domain, a narrowband optical filter (with a broad time-domain response) can be used to broaden a pulse in time via convolution. The effect of this filter on a 50% RZ signal in the worst case (g ¼ 0.5) is shown in Figure 4.19. In a standard monitoring configuration, when the DGD is greater than or equal to the pulse width and less than 1-bit duration (e.g., 12.5 ps for a 40-Gb/s 50% RZ signal), the optical pulse replicas on the two orthogonal PSPs no longer coincide with each other, and as such the signal is completely depolarized (the minimum DOP is zero). The point at which the signal reaches this complete depolarization is the maximum DGD monitoring range, and is typically equal to the pulse width of the signal regardless of the bit rate. The narrow transfer function of the filter in the frequency domain means that it has a wide response time, and this response convolved with the short pulses in the time domain results in a broadening of the signal replica pulses. These broadened pulses may overlap where they did not before, thus shifting upward the minimum DOP versus DGD curve, extending the DGD monitoring window. The frequency-domain explanation for this repolarization is shown in Figure 4.20. The link DGD causes the SOP of frequency components within the optical spectrum to rotate with respect to the central optical frequency by an amount 2p △f DGD (in Stokes space), where △f is the frequency offset of a given component from the center of the optical spectrum. This is known as the SOP walk-off effect. As short optical pulses in time have a wide optical spectrum, depolarization effects are much more pronounced for short-pulse-width signals. A narrowband optical filter shrinks the optical spectrum, and thus reduces the depolarization effects. Figure 4.21(a) shows minimum DOP versus DGD curves for 12.5-ps pulse width (50%), 40-Gb/s RZ signals. Prior to asymmetric partial optical filtering with a 36-GHz first-order Gaussian filter, 50% RZ DOP ↑

DOP = 0

DOP = 1 Tb /2 45⬚

Slow axis DGD

Optical filter

~ ~ ~ Fast axis Constant polarization

Δt = DGD = Tb / 2 Total depolarization

DGD Partial depolarization

FIGURE 4.19 Prior to optical filtering, an RZ signal that undergoes DGD equal to the pulse width is completely deplorized, thus limiting the DGD monitoring range of DOP-based DGD monitors. After filtering, the signal is partially polarized, allowing DOP-based monitoring of the DGD. Copyright © 2001 IEEE /OSA.

116

CHAPTER 4 Polarization mode dispersion monitoring

Optical spectrum

–

Slow-axis DGD λ

+

Optical filter

λ Frequency-dependent SOP

Constant SOP Fast axis

~ ~ ~

λ Canceling depolarization effects

FIGURE 4.20 Frequency-domain illustration of reducing depolarization via symmetric narrowband optical filtering. Short optical pulses have a wide optical spectrum, enhancing the effects of DGD-induced depolarization. A narrowband filter shrinks the optical spectrum, thus reducing these depolarization effects and increasing the DGD monitoring range. Copyright © 2001 IEEE /OSA.

1.0

1.0 40-Gb/s RZ (50%) With filtering Without filtering

0.8

0.8

DOP

DOP

0.6

0.6

0.4

0.4

0.2

0.2

0

0

5

10

15

DGD (ps) (a)

20

Asymmetric filtering Broadband symmetric filtering

25

0

40 Gb/s NRZ

0

20

40

60

80

100

DGD(ps) (b)

FIGURE 4.21 (a) Experimental results for minimum DOP versus DGD for 40-Gb/s 50% RZ signals. (b) Simulation results for minimum DOP versus DGD for 40-Gb/s NRZ signals before and after asymmetric partial optical filtering. Copyright © 2001 IEEE /OSA.

the monitoring range is limited to 13 ps. However, after asymmetric partial optical filtering, the DGD monitoring range is extended to a full bit time, 25 ps. Figure 4.21(b) shows the simulation results for 40-Gb/s NRZ systems after broadband symmetric (conventionally used to remove the effects of amplified spontaneous emission [ASE] noises on the DOP measurement) and narrowband asymmetric filtering. Without asymmetric filtering, the DOP dynamic range varies from 1 to 0.5. However, after asymmetric filtering using the optical filter, the monitoring range remains the same, but the DOP dynamic range is almost doubled.

4.4 Electronic PMD monitoring techniques

117

This limiting of DGD monitoring range is present in short-pulse-width CSRZ and ACRZ systems as well. However, as the optical spectrum for ACRZ and CSRZ signals differs from that of RZ signals, and the DOP is dependent on the optical spectrum, while this technique of partial optical filtering can be applied to these signals, the shape, order, and bandwidth of the filter must be optimized for the individual data modulation format to ensure the maximum DGD monitoring range and DOP dynamic range. By optimizing the parameters of the partial optical filter, for 10-Gb/s 50% CSRZ and ACRZ signals, the monitoring range can be extended from 50 to 70 ps and from 70 to 100 ps, respectively.50

4.4 ELECTRONIC PMD MONITORING TECHNIQUES Electronic monitoring techniques are based on the analysis of the electrical signal after the O/E conversion at the receiver, such as Q/BER measurement, eye diagram analysis, and histogram analysis. The error correlation information from forward error correction (FEC) and the coefficients obtained from an electrical equalizer at the receiver can also be used for optical performance monitoring (OPM) to identify and quantify optical distortions and perform network monitoring without additional expensive optical equipment. By comparing the coefficients of an adaptive electrical equalizer to pre computed coefficient vectors, it is possible to identify and quantify the most common distortions of an optical link.51,52 Eye diagram is a commonly used tool to analyze the quality of signals, which can be used for PMD monitoring.53–56 Figure 4.22 shows an eye monitor, which is a SiGe integrated circuit (IC) consisting of two decision circuits in parallel.56 The upper decision gate acts as an active data channel, whereas the lower gate is the monitor gate working with a variable threshold to characterize the edges of the eye at the variable phase. By dithering the monitor threshold, pseudo-errors are generated. Using an EXOR gate, these pseudo-errors are detected and added within the integration time by an analog integrator. The eye opening is determined at a target integrator voltage that corresponds to a certain pseudo-error rate in the monitor decision gate. Figure 4.23(a) exhibits excellent correlation between the bit error rate (BER) and eye opening down to 10
10 of BER. Using an eye-opening

10

Recovered data U0

C0 EXOR

U1

C1 (a)

dt

Ueye

Uint [V ]

Decision circuits

U1

70 ps 60 ps 50 ps 30 ps 0 ps

1

Ueye Utarget

0.1 0.05

0.1

0.15 0.2 Ut [V ]

0.25

0.3

(b)

FIGURE 4.22 (a) Design of error monitor with analog integrator, and (b) typical characteristics of integrator voltage Uint versus monitor threshold U1 for first-order PMD signals with variable DGD and Y ¼ 0.5. Copyright © 2001 IEEE /OSA.

118

CHAPTER 4 Polarization mode dispersion monitoring

10–6

10–3

10–5 10–7 10–9 10–11 0.05 0.07 0.09 0.11 0.13 0.15 Eye opening [V ] (a)

BER

BER

10–7 10–8 10–9 10–10 10–11 0.12

0.13 0.14 Eye monitor [V ] (b)

0.15

FIGURE 4.23 (a) BER versus eye opening for all orders PMD statistics. (b) BER after compensation versus eye opening using eye monitoring. Copyright © 2001 IEEE /OSA.

monitor, a PMD compensator was tested with 1045 independent PMD conditions at 10 Gb/s. The correlation between the eye opening and BER after optical compensation is good down to 3 10
10, as shown in Figure 4.23(b). The key drawbacks of eye monitoring are the clock recovery requirement and the difficulties of isolating various impairments. An eye diagram is actually a synchronous amplitude distribution within an entire bit period. When the clock information is absent, the signal can only be asynchronously sampled. The amplitude histogram is obtained by random sampling that spans the entire bit period. With a sufficient number of random samples, an asynchronous histogram can evenly represent the pulse amplitude distribution in a bit period.56 The asynchronous amplitude histogram technique may be a promising method for low-cost, bit rate–transparent channel performance monitoring due to unnecessary clock recovery. A comparison of synchronous and asynchronous diagrams has been shown in Figure 4.24.57 Figure 4.24(a) is the synchronous temporal window (solid-line box) at the widest eye opening. Figure 4.24(b) is the corresponding histogram. Based on the histogram, the Q-factor and the BER can be estimated.58 Figures 4.24(c) and (d) are the asynchronously sampled eye diagram and histogram obtained within the entire bit period. The main difference between the asynchronous and the synchronous histograms lies at the middle counts, the “cross” in the eye diagram, between the mark and space levels. Within the cross-point region, the asynchronous histogram (Figure 4.24(d)) contains counts, while the synchronous histogram (Figure 4.24(b)) contains essentially no counts. The cross point count is related to the pulse rise time and fall time that may be affected by dispersion-related impairments. In fact, an asynchronous histogram can evaluate the signal quality. It is sensitive to both OSNR and other fiber impairment such as CD and PMD. This method has been suggested to monitor the Qfactor or OSNR for both on-off keying (OOK) and DPSK signals.51–61 Monitoring of both CD and PMD has also been investigated and demonstrated theoretically and experimentally.62–64 However, since this is an electrical-domain measurement, the challenge is how to measure the absolute value of PMD and other degrading effects separately.

Amplitude (a.u.)

Amplitude (a.u.)

4.4 Electronic PMD monitoring techniques

s1

m1 s0

m0

Time (ps)

Count

(a)

(b)

Amplitude (a.u.)

Amplitude (a.u.)

119

s1

m1 s0

m0

Count (d)

Time (ps) (c)

FIGURE 4.24 (a) Eye diagram and (b) histogram with synchronous sampling. (c) Eye diagram and (d) histogram with asynchronous sampling. Copyright © 2004 IEEE /OSA.

Other than the traditional histogram, an asynchronous sampling technique called asynchronous delay-tap sampling is proposed that can measure multiple simultaneous impairments.65–68 The optical waveform is sampled in pairs separated by a known physical delay Dt, as shown in Figure 4.25. The sample time between the pairs, Ts, is not related to the monitored signal bit Δt = T TS

TS

x2

y3

x1

x3 y1

y2 Time (1.1)

0 1

10

110

011 01

y

001

(0.1)

(0.0)

100 x

(1.0)

FIGURE 4.25 Portrait processing of delay-tap sample pairs to create phase. Labels on phase portrait represent sampled bit sequences. Copyright © 2007 IEEE /OSA.

120

CHAPTER 4 Polarization mode dispersion monitoring

rate, and can be many orders of magnitude slower. Plotting the pairs produces information-rich patterns called phase portraits, which are of similar complexity to eye diagrams, but do not require clock recovery. A fundamental difference between the two techniques is that the phase portrait contains information about the probability distributions of closely spaced samples, or equivalently, distributions of waveform slopes. This information is absent in eye diagrams, which are constructed from samples separated by large periods. For a 1-bit delay, Dt ¼ T, the technique geometrically separates the various 3-bit sequences, and in particular separates out the 010 and 101 sequences, which are generally most susceptible to signal distortion. The effects of PMD on a 10-Gb/s NRZ signal have been simulated. The resulting phase portraits and eye diagrams are shown in Figure 4.26. In each case, the tap delay for the phase portrait was chosen to be a 1-bit period. Figure 4.26(a) shows the results for no optical impairment (OSNR ¼ 35 dB), with a clean eye and a well-defined geometric shape in the phase portrait. Figure 4.26(b) shows the same OSNR of 35 dB with 40 ps of first-order PMD, with the power split equally between the polarization axes. The eye shows the characteristic “triangularization.” And a closer inspection shows significant differences in the distribution of points along the phase portraits. Other impairments such as OSNR, CD, and interferometric crosstalk can be monitored at the same time by analyzing the phase portraits.65 Recently, this monitoring technique was demonstrated for a 10-Gb/s NRZ channel from a commercial transponder with randomly aligned first-order PMD in a long-haul WDM test bed with 50-GHz channel spacing.69 The results showed that, independent of the OSNR, the effective DGD and CD can be measured. The DGD measurement is shown in Figure 4.27. Moreover, the artificial neural network (ANN) technique has been introduced for PMD and other impairments monitoring based on analyzing eye diagrams and delay-tap sampling, which can be a powerful tool for monitoring the performance of optical channels.70,71

Clean

(a)

With PMD

(b)

FIGURE 4.26 Eye diagrams and phase portraits for NRZ with OSNR ¼ 35 dB at (a) no impairment and (b) PMD ¼ 40 ps. Copyright © 2007 IEEE /OSA.

4.5 Other pmd monitoring techniques

121

30

Predicted DGDeff (ps)

25 20 15 10 5 0 –5

0

5

10 15 True DGDeff (ps)

20

25

FIGURE 4.27 Measurement of effective DGD (root mean square error of 3.1 ps) in presence of OSNR levels ranging from 13.5 to 25 dB. Copyright © 2009 IEEE/OSA.

4.5 OTHER PMD MONITORING TECHNIQUES Another straightforward technique involves measuring the phase difference between two optical frequency components for the two orthogonal PSPs.72 This technique requires polarization tracking at the receiver to be able to find the PSPs so the phase can be measured. The SOP “string” length73 has been introduced and demonstrated to have strong correlation with PMD-induced penalty in a channel. The SOP string represents the length of the wavelengthdependent SOP trace on the Poincare sphere over the modulation bandwidth of the channel. The DGD and the angle between the PSP and the launch SOP were first measured directly, and then the string length was estimated. While such an extensive approach is suitable for laboratory studies, these DGD and PSP measurements are less appropriate for real systems. Following that, Wang et al.74 demonstrated nonintrusive, near-real-time estimation of the PMD-induced system penalty from direct measurements of the output SOP string lengths in a field trial, although a special custom-built, high-speed, high-spectral-resolution polarimeter is required in the system.75 Other techniques of PMD monitoring include using nonlinear effects,76–80 real-time optical Fourier transformation,81 finite-impulse-response (FIR) filters and spectrum monitoring,82 and coherent detection.83–85 Most of these techniques are relatively complicated to be applied to real online systems currently.

4.6 SUMMARY In summary, PMD is an important impairment in high-speed reconfigurable optical networks, affects system performance, and might be monitored. By its inherent nature, deleterious PMD effects are stochastic, time varying, vibration dependent, and temperature dependent. In addition, reconfigurable optical networking will change the signal path and thus the accumulated PMD. Therefore, real-time

122

CHAPTER 4 Polarization mode dispersion monitoring

Table 4.1 Summary of Major PMD Monitoring Techniques Techniques

RF Tone Measurement

DOP Measurement

Electronic Monitoring

Comments

Needs high-speed devices

No need for high-speed devices

Needs high-speed devices

Insensitive to other distortion sources

Needs synchronization for eye analysis

Requires additional techniques to increase sensitivity and DGD range

Response time may be relatively slow for histogram analysis

Affected by other distortion sources Sensitivity and DGD range depends on monitored frequency

Isolation of various impairments still under investigation

PMD monitoring is required in order to dynamically tune a compensator or for network control and management. A variety of techniques on PMD monitoring were covered in this chapter. Selected major PMD monitoring techniques are summarized in Table 4.1. In order to enable robust and cost-effective automated operation of high-speed reconfigurable optical networks, it is desirable to intelligently monitor the state of the network as well as the quality of propagating data signals, isolate specific impairments including PMD, and automatically diagnose and repair the network. The value of monitoring increases with increasing intelligence and bit rates. And we should seek to maintain the right balance among monitoring coverage, sensitivity, and cost.

ACKNOWLEDGMENTS The author would like to thank Shaoliang Zhang and Jing Yang for their valuable help.

REFERENCES 1. Willner AE. The optical network of the future: can optical performance monitoring enable automated, intelligent and robust systems? Opt Photon News 2006;17(3):30–5. 2. Poole CD. Statistical treatment of polarization dispersion in single-mode fiber. Opt Lett 1988;13(8):687–9. 3. Poole CD. Measurement of polarization-mode dispersion in single-mode fibers with random mode coupling. Opt Lett 1989;14(10):523–5. 4. Namihira Y, Wakabayashi H. Fiber length dependence of polarization mode dispersion measurement in long-length optical fibers and installed optical submarine cables. J Opt Commun 1991;12(1):1–8. 5. Poole CD, Tkach RW, Chraplyvy AR, Fishman DA. Fading in lightwave systems due to polarization-mode dispersion. IEEE Photon Technol Lett 1991;3(1):68–70. 6. Poole CD, Winters JH, Nagel JA. Dynamic equation for polarization dispersion. Opt Lett 1991;16(6):372–4. 7. Foschini GJ, Poole CD. Statistical theory of polarization dispersion in single mode fibers. IEEE/OSA J Lightwave Technol 1991;9(11):1439–56.

References

123

8. Jopson RM, Nelson LE, Kogelnik H, Foschini GJ. Statistics of second-order PMD depolarization. IEEE/ OSA J Lightwave Technol 2001;19(12):1882–6. 9. Ciprut P, Gisin B, Gisin N, Passy R, Von der Weid JP, Prieto F, Zimmer CW. Second-order polarization mode dispersion: impact on analog and digital transmissions. IEEE/OSA J Lightwave Technol 1998;16(5):757–71. 10. Shieh W. On the second-order approximation of PMD. IEEE Photon Technol Lett 2000;12(3):290–2. 11. Willner AE, Motaghian SMR, Yan LS, Pan Z, Hauer M. Monitoring and control of polarization-related impairments in optical fiber systems. IEEE/OSA J Lightwave Technol 2004;22(1):106–25. 12. Schmuck H. Effect of polarization-mode-dispersion in fiber-optic millimeter-wave systems. IEE Electron Lett 1994;30(18):1503–4. 13. Adamczyk OH, Sahin AB, Yu Q, Lee S, Willner AE. Statistics of PMD-induced power fading for intensitymodulated double-sideband and single-sideband microwave and millimeter-wave signals. IEEE Trans Microw Theory Tech 2001;49(10):1962–7. 14. Poole CD, Darcie TE. Distortion related to polarization-mode dispersion in analog lightwave systems. IEEE/OSA J Lightwave Technol 1993;11(11):1749–59. 15. Smith GH, Novak D, Ahmed Z. Technique for optical SSB generation to overcome dispersion penalties in fibre-radio systems. IEE Electron Lett 1997;33(1):74–5. 16. Hofstetter R, Schmuck H, Heidemann R. Dispersion effects in optical millimeter-wave systems using selfheterodyne method for transport and generation. IEEE Trans Microw Theory Tech 1995;43(9):2263–9. 17. Yu C, Yu Q, Pan Z, Sahin AB, Willner AE. Optically compensating the PMD-induced RF power fading for single-sideband subcarrier-multiplexed systems. IEEE Photon Technol Lett 2004;16(1):341–3. 18. Nezam SMRM, McGeehan JE, Willner AE. DOP-based PMD monitoring in optical subcarrier multiplexed systems by carrier/sideband equalization. In: Proc. conference on optical fiber communication, paper ThY4. Atlanta, GA; 2003. p. 593–5. 19. Hui R, Allen C, Demarest K. PMD insensitive SCM optical receiver using polarization diversity. IEEE Photon Technol Lett 2002;14(11):1632–4. 20. Mazurczyk VJ, Zyskind JL. Polarization hole burning in erbium doped fiber amplifiers. In: Proc. conference on lasers and Electro-Optics/Quantum Electronics and Laser Science, paper CPD26. Baltimore, MD; 1993. 21. Haunstein HF, Kallert HM. Influence of PMD on the performance of optical transmission systems in the presence of PDL. In: Proc. optical fiber communication conference, vol. 3, paper WT4. Anaheim, CA; 2001. p. 17–22. 22. Gisin N, Huttner B. Combined effects of polarization mode dispersion and polarization dependent loss. Opt Commun 1997;142(1-3):119–25. 23. Huttner B, Geiser C, Gisin N. Polarization-induced distortions in optical fiber networks with polarizationmode dispersion and polarization-dependent losses. IEEE J Sel Top Quantum Electron 2000;6(2):317–29. 24. Yan LS, Yu Q, Xie Y, Willner AE. Experimental demonstration of the system performance degradation due to the combined effect of polarization dependent loss with polarization mode dispersion. IEEE Photon Technol Lett 2002;14(2):224–6. 25. Yan L-S, Yu Q, Luo T, McGeehan JE, Willner AE. Deleterious system effects due to low-frequency polarization scrambling in the presence of nonnegligible polarization-dependent loss. IEEE Photon Technol Lett 2003;15:464–6. 26. Bulow H, Baumert W, Schmuck H, Mohr F, Schulz T, Kuppers F, et al. Measurement of the maximum speed of PMD fluctuation in installed field fiber. In: Proc. conference on optical fiber communication, San Diego, CA; 1999. p. 83–5. 27. Ishikawa G, Ooi H. Polarization-mode dispersion sensitivity and monitoring in 40-Gbit/s OTDM and 10Gbit/s NRZ transmission experiments. In: Proc. optical fiber communication conference. Washington, DC; 1998. p. 117–9.

124

CHAPTER 4 Polarization mode dispersion monitoring

28. Heismann F, Fishman DA, Wilson DL. Automatic compensation of first-order polarization mode dispersion in a 10 Gb/s transmission system. In: Proc. European conference on optical communication, vol. 1. Washington, DC; 1998. p. 529–30. 29. Takahashi T, Imai T, Aiki M. Automatic compensation technique for timewise fluctuating polarization mode dispersion in in-line amplifier systems. Electron Lett 1994;30(4):348–9. 30. Pua HY, Peddanarappagari K, Zhu B, Allen C, Demarest K, Hui R. An adaptive first-order polarizationmode dispersion compensation system aided by polarization scrambling: theory and demonstration. IEEE/ OSA J Lightwave Technol 2000;18(6):832–41. 31. Noe R, Sandel D, Yoshida-Dierolf M, Hinz S, Mirvoda V, Schopflin A, et al. Polarization mode dispersion compensation at 10, 20, and 40 Gb/s with various optical equalizers. IEEE/OSA J Lightwave Technol 1999;17(9):1602–16. 32. Luo T, Yu C, Pan Z, Wang Y, Arieli Y, Willner AE. Dispersive effects monitoring for RZ data by adding a frequency-shifted carrier along the orthogonal polarization state. IEEE/OSA J Lightwave Technol 2005;23(10):3295–301. 33. Park KJ, Youn CJ, Lee JH, Chung YC. Performance comparisons of chromatic dispersion-monitoring techniques using pilot tones. IEEE Photon Technol Lett 2003;15(6):873–5. 34. Devaux F, Sorel Y, Kerdiles JF. Simple measurement of fiber dispersion and of chirp parameter of intensity modulated light emitter. IEEE/OSA J Lightwave Technol 1993;11(12):1937–40. 35. Liu N, Zhong W-D, Wen YJ, Lu C, Cheng L, Wang Y. PMD and chirp effects suppression in RF tone-based chromatic dispersion monitoring. IEEE Photon Technol Lett 2006;18(5):673–5. 36. Luo T, Pan Z, Nezam SMRM, Yan LS, Sahin AB, Willner AE. PMD monitoring by tracking the chromaticdispersion-insensitive RF power of the vestigial sideband. IEEE Photon Technol Lett 2004;16(9):2177–9. 37. Motaghian Nezam SMR, Song YW, Yu C, McGeehan J, Sahin AB, Willner AE. First-order PMD monitoring for NRZ data using optical clock regeneration techniques. IEEE/OSA J Lightwave Technol 2004; 22(4):1086–93. 38. Yu C, Wang Y, Luo T, Pan Z, Motaghian Nezam SMR, Sahin AB, et al. Chromatic-dispersion-insensitive PMD monitoring for NRZ data based on clock power measurement using a narrowband FBG notch filter. In: Proc. Europe conference on optical communication, paper Tu4.2.3. Rimini, Italy; 2003. 39. Lu G-W, Cheung M-H, Chen L-K, Chan C-K. Simultaneous PMD and OSNR monitoring by enhanced RF spectral dip analysis assisted with a local large-DGD element. IEEE Photon Technol Lett 2005; 17(12):2790–2. 40. Joun CY, Park KJ, Lee JH, Chung YC. OSNR monitoring technique based on orthogonal delayedhomodyne method. IEEE Photon Technol Lett 2002;14(10):1469–71. 41. Yang J-Y, Zhang L, Christen LC, Zhang B, Nuccio S, Wu X, et al. Chromatic-dispersion-insensitive PMD monitoring of 20-Gb/s DQPSK and 10-Gb/s DPSK using DGD-generated polarization-based interferometer filter. In: Proc. Europe conference on optical communication, paper 3.5.3. Berlin; 2007. 42. Yang J-Y, Zhang L, Christen LC, Zhang B, Nuccio S, Wu X, et al. Polarization-mode-dispersion monitoring for phase-modulated signals using DGD-generated interferometric filter. IEEE Photon Technol Lett 2008; 20(2):150–2. 43. Roy F, Francia C, Bruyere F, Penninckx D. Simple dynamic polarization mode dispersion compensator. In: Proc. optical fiber communication conference, vol. 1. San Diego, CA; 1999. p. 275–8. 44. Lanne S, Idler W, Thiery J-P, Hamaide J-P. Fully automatic PMD compensation at 40 Gbit/s. IEE Electron Lett 2002;38(1):40–1. 45. Kikuchi N. Analysis of signal degree of polarization degradation used as control signal for optical polarization mode dispersion compensation. IEEE/OSA J Lightwave Technol 2001;19(4):480–6. 46. Motaghian Nezam SMR, Yan L-S, McGeehan JE, Shi YQ, Willner AE, Yao S. Wide-dynamic-range DGD monitoring by partial optical signal spectrum DOP measurement. In: Proc. optical fiber communication conference. Anaheim, CA; 2002. p. 924–6.

References

125

47. Reza Motaghian Nezam SM, McGeehan JE, Willner AE. Theoretical and experimental analysis of the dependence of a signal’s degree of polarization on the optical data spectrum. IEEE/OSA J Lightwave Technol 2004;22(3):763–72. 48. Rosenfeldt H, Knothe C, Ulrich R, Brinkmeyer E, Feiste U, Schubert C, et al. Automatic PMD compensation at 40 Gbit/s and 80 Gbit/s using a 3-dimensional DOP evaluation for feedback. In: Proc. optical fiber communication conference, vol. 4, paper PD27. Anaheim, CA; 2001. 49. Motaghian Nezam SMR, McGeehan JE, Willner AE. Dependence of DOP-based DGD monitors on the optical power spectrum of equal pulse-width data formats. In: Proc. conference on optical fiber communication, paper MF91. Atlanta, GA; 2003. 50. Motaghian Nezam SMR, Yan L-S, McGeehan JE, Shi YQ, Willner AE, Yao XS. Enhancing the dynamic range and DGD monitoring windows in DOP-based DGD monitors using symmetric and asymmetric partial optical filtering. IEEE/OSA J Lightwave Technol 2004;22(4):1094–102 [Special issue on polarization-mode-dispersion]. 51. Meflah L, Thomsen B, Mitchell JE, Bayvel P, Lehmann G, Santori S, et al. Advanced optical performance monitoring for dynamically reconfigurable networks. In: Proc. 10th European conference on networks and optical communications. London, UK; 2005. p. 554–61. 52. Woodward SL, Nelson LE, Feuer MD, Zhou X, Magill PD, Foo S, et al. Characterization of real-time PMD and chromatic dispersion monitoring in a high-PMD 46-Gb/s transmission system. IEEE Photon Technol Lett 2008;24(24):2048–50. 53. Buchali F, Baumert W, Bulow H, Poirrier J. A 40 Gb/s eye monitor and its application to adaptive PMD compensation. In: Proc. conference on optical fiber communication. Anaheim, CA; 2002. p. 202–3. 54. Buchali F, Baumert W, Bulow H, Feiste U, Ludwig R, Weber HG. Eye monitoring in a 160 Gbit/s RZ field transmission system. In: Proc. European conference on optical communication. Amsterdam, The Netherlands; 2001. p. 288–9. 55. Bulow H. Electronic equalization of transmission impairments. In: Proc. conference on optical fiber communication, paper TuE4. Anaheim, CA; 2002. p. 24–5. 56. Buchali F, Lanne S, Thiery J-P, Baumert W, Bulow H. Fast eye monitor for 10 Gbit/s and its application for optical PMD compensation. In: Proc. optical fiber communication conference, paper TuP5. Anaheim, CA; 2001. 57. Chen H, Poon AW, Cao X-R. Transparent monitoring of rise time using asynchronous amplitude histograms in optical transmission systems. IEEE/OSA J Lightwave Technol 2004;22(7):1661–7. 58. Bergano NS, Kerfoot FW, Davidson CR. Margin measurements in optical amplifier systems. IEEE Photon Technol Lett 1993;5(3):304–6. 59. Hanik N, Gladisch A, Caspar C, Strebel B. Application of amplitude histograms to monitor the performance of optical channels. IEE Electron Lett 1999;35(5):403–4. 60. Li Z, Li G. In-line performance monitoring for RZ-DPSK signals using asynchronous amplitude histogram evaluation. IEEE Photon Technol Lett 2006;18(3):472–4. 61. Shake I, Takara H. Averaged Q-factor method using amplitude histogram evaluation for transparent monitoring of optical signal-to-noise ratio degradation in optical transmission system. IEEE/OSA J Lightwave Technol 2002;20(8):1367–73. 62. Li Z, Li G. Chromatic dispersion and polarization-mode dispersion monitoring for RZ-DPSK signals based on asynchronous amplitude-histogram evaluation. IEEE/OSA J Lightwave Technol 2006;24(7):2859–66. 63. Shake I, Takara H. Chromatic dispersion dependence of asynchronous amplitude histogram evaluation of NRZ signal. IEEE/OSA J Lightwave Technol 2003;21(10):2154–61. 64. Kozicki B, Takuya O, Hidehiko T. Optical performance monitoring of phase-modulated signals using asynchronous amplitude histogram analysis. IEEE/OSA J Lightwave Technol 2008;26(10):1353–61. 65. Anderson TrevorB, Dods SarahD, Clarke Ken, Bedo Justin, Kowalczyk Adam. Multi-impairment monitoring for photonic networks. In: Proc. Europe conference on optical communication, paper 3.5.1. Berlin; 2007.

126

CHAPTER 4 Polarization mode dispersion monitoring

66. Dods SD, Anderson TB. Optical performance monitoring technique using delay tap asynchronous waveform sampling. In: Proc. optical fiber communication conference, paper OThP5. Anaheim, CA; 2006. 67. Anderson TB, Dods SD. Asynchronous measurement of chromatic dispersion from waveform distortion. In: Proc. optical fiber communication conference, paper OWN4. Anaheim, CA; 2006. 68. Dods SD, Anderson TB, Clarke K, Bakaul M, Kowalczyk A. Simultaneous and independent monitoring of OSNR, chromatic and polarization mode dispersion for NRZ-OOK, DPSK and Duobinary. In: Proc. optical fiber communication conference, paper OMM5. Anaheim, CA; 2007. 69. Anderson T, Clarke K, Beaman D, Ferra H, Birk M, Zhang G, et al. Experimental demonstration of multiimpairment monitoring on a commercial 10 Gbit/s NRZ WDM channel. In: Proc. optical fiber communication conference, paper OThH7. San Diego, CA; 2009. 70. Jargon JA, Wu X, Willner AE. Optical performance monitoring using artificial neural networks trained with eye-diagram parameters. IEEE Photon Technol Lett 2009;21(1):54–6. 71. Jargon JA, Wu X, Willner AE. Optical performance monitoring by use of artificial neural networks trained with parameters derived from delay-tap asynchronous sampling. In: Proc. optical fiber communication conference, paper OThH1. San Diego, CA; 2009. 72. Hakki BW. Polarization mode dispersion compensation by phase diversity detection. IEEE Photon Technol Lett 1997;9(1):121–3. 73. Boroditsky M, Cornick K, Antonelli C, Brodsky M, Dods SD, Frigo NJ, et al. Comparison of system penalties from first- and multiorder polarization-mode dispersion. IEEE Photon Technol Lett 2005;17(8):1650–2. 74. Wang SX, Weiner AM, Boroditsky M, Brodsky M. Monitoring PMD-induced penalty and other system performance metrics via a high-speed spectral polarimeter. IEEE Photon Technol Lett 2006;18(16):1753–5. 75. Wang SX, Weiner AM. Fast wavelength-parallel polarimeter for broadband optical networks. Opt Lett 2004;29(9):923–5. 76. Wielandy S, Fishteyn M, Zhu B. Optical performance monitoring using nonlinear detection. IEEE/OSA J Lightwave Technol 2004;22(3):784–93. 77. Lu G-W, Xie C, Ku Y-C, Chen L-K, Chan C-K. Enhanced PMD monitoring with frequency-resolved SOP rotation by self-phase modulation. IEEE Photon Technol Lett 2004;16(9):2180–2. 78. Ito T, Sato R, Kasahara R, Ogawa I, Kawaguchi Y, Inoue Y, et al. Bit rate and format independent PMD compensation with differential group delay monitoring using SOA/PLC hybrid integrated wavelength converter. IEE Electron Lett 2003;39(1):84–6. 79. Huh JY, Chung YC. Simultaneous monitoring technique for OSNR and PMD based on four-wave mixing in SOA. In: Proc. optical fiber communication conference, paper OThW1. San Diego, CA; 2008. 80. Pelusi MD, Luan F, Vo TD, Lamont MRE, Madden SJ, Choi D-Y, et al. Ultra-broadband (THz) RF spectrum monitoring of high-speed optical signals using a Chalcogenide waveguide chip. In: Proc. optical fiber communication conference, paper OThH6. San Diego, CA; 2009. 81. Llorente R, Clavero R, Marti J. PMD monitoring by spectral SOP rotation using a real-time optical fourier transformer. In: Proc. optical fiber communication conference, paper OME26. Anaheim, CA; 2005. 82. Bohn M, Rosenkranz W, Krummrich P, Horst F, Offrein BJ, Bona GL. Experimental verification of combined adaptive PMD and GVD compensation in a 40Gb/s transmission using integrated FIR-filters and spectrum monitoring. In: Proc. optical fiber communications conference, paper TuG3. Los Angeles, CA; 2004. 83. Fu B, Hui R. Fiber chromatic dispersion and polarization-mode dispersion monitoring using coherent detection. IEEE Photon Technol Lett 2005;7(7):1561–3. 84. Roudas I, Piech GA, Mlejnek M, Mauro Y, Chowdhury DQ, Vasilyev M. Coherent frequency-selective polarimeter for polarization-mode dispersion monitoring. IEEE/OSA J Lightwave Technol 2004; 22(4):953–67. 85. Geyer JC, Fludger CRS, Duthel T, Schulien C, Schmauss B. Performance Monitoring Using Coherent Receivers. In: Proc. optical fiber communication conference, paper OThH5. San Diego, CA; 2009.

CHAPTER

Timing misalignment monitoring

5 Lian-Kuan Chen

The Chinese University of Hong Kong, Hong Kong SAR, The People’s Republic of China

5.1 INTRODUCTION With the surge of bandwidth demand for the Internet and high-speed computer interconnects, data modulation is required at much higher speed. To circumvent the distortion brought about by the fiber dispersion and nonlinear effects, advanced modulation formats are employed. For high-speed transmission systems with data rates
40 Gb/s, the return-to-zero (RZ) pulse is generally used together with the data modulation such as on-off keying (OOK) or differential phase-shift keying (DPSK).1,2 For a given bit error rate (BER), DPSK offers the advantage of requiring a 3-dB lower optical signalto-noise ratio (SNR) than OOK by using balanced detection. For these types of modulations, two modulators—the first for RZ-pulse carving and the second for data modulation—are used in series.3–5 The two modulators have to be synchronous such that the data encoding part will fall on the carved pulses. Figure 5.1 depicts results of the data modulation when the two modulators are aligned or misaligned. The pulse distortion is evident when there is timing misalignment between the pulse carver and data modulator. Thus, it is essential to have a monitoring mechanism to detect any timing misalignment between the two parts. Even if the two signals are synchronized properly initially, timing misalignment may still happen due to environment changes. It is reported that timing drift of 2–5 ps was observed when the temperature of a 1-m-long optical fiber between a pulse caver and a data modulator was changed by 50 C.6 This may cause a power penalty of around 1 dB for an optically pre amplified receiver in 40-Gb/s systems. Thus, an automatic alignment is essential to compensate the timing drift between pulse carving and data modulation. Fortunately, such a timing drift process is rather slow; thus, it is possible to realize automatic alignment using cost-effective, low-speed components.

FIGURE 5.1 The effect of timing misalignment between pulse carver and data modulator.4 © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00005-5

Copyright © 2003 IEEE.

127

128

CHAPTER 5 Timing misalignment monitoring

Other than the synchronization between the pulse carver and data modulator for RZ-OOK and RZ-DPSK signals, timing alignment is also needed in other applications such as the multiplexing and demultiplexing of OTDM channels, synchronization for data remodulation, and synchronization for I/Q data and data/pulse carver. These applications will also be discussed in this chapter. Similar to other optical performance monitoring discussed throughout this book, the following performance metrics should be considered when assessing different monitoring schemes: n

n n n

Monitoring power dynamic range (MPDR)—The ratio of the maximum and minimum detected monitoring signal power (unit: dB). Monitoring sensitivity—The change of monitoring signal per unit timing misalignment (unit: dB/ps). Monitoring range—The range of timing misalignment that can be monitored (unit: picosecond, ps). Complexity—For example, the number of components needed to construct the monitoring device, the difficulty of extracting the misalignment information, or whether a high-speed detector is required.

This chapter is organized as follows. In Section 5.2, various monitoring schemes for timing alignment will be discussed. A comparison table for different monitoring schemes for pulse carving and data modulation will be given. In Section 5.3, the detrimental effects of timing misalignment will be further investigated for selected modulation formats. Various schemes to mitigate the distortion due to the timing misalignment are given in Section 5.4. Finally, Section 5.5 summarizes the chapter.

5.2 MONITORING OF TIMING ALIGNMENT

5.2.1 Synchronization of pulse carving and data modulation RZ pulse is commonly used for high-speed transmission systems as it provides better tolerance to chromatic dispersion and polarization mode dispersion. RZ pulses are generated by a pulse carver and then are data modulated by using a proper modulation format, such as OOK and DPSK. The pulse carver used is a high-speed-intensity modulator, such as the Mach-Zehnder modulator (MZM) or electroabsorption modulator (EAM). The pulse carver can also be placed after the data modulator. In the following, the monitoring schemes for timing misalignment will be discussed for the cases of RZ-DPSK and RZ-OOK modulations, respectively. For high-speed data modulation (>40 Gb/s), EAM is very attractive for data modulation. In Reference 3, a device to detect and correct the timing misalignment between the pulse carver and the EAM for data modulation in a 10-Gb/s ultra-long-haul soliton transmission system is demonstrated. The scheme utilizes the chirping characteristics of the EAM for the timing misalignment monitoring. When timing misalignment occurs, the RF power of the upper sideband and lower sideband of the signal spectrum will increase or decrease. As shown in the inset of Figure 5.2, the solid (dashed) curve is the optical spectrum when pulse carver leads (lags) the data modulator by 20 ps. By taking the power difference between the 10-GHz and 10-GHz sideband, the timing offset can be inferred, as shown in Figure 5.2. The two sidebands are extracted using a wedge etalon. The timing misalignment and the power difference depict a linear relation within 20 ps of misalignment, showing that it is a feasible scheme for misalignment monitoring using the sidebands’ power difference. The device can also serve as a wavelength locker.

(Power (10GHz)-Power (–10GHz))/Power (0GHz)

5.2 Monitoring of timing alignment

129

0.2 Simulation Measurement 0.1

0.0

–0.1

–0.2

–20 –10 0 10 20 Frequency (GHz)

–30

–20

10 –10 0 Timing offset (ps)

20

30

FIGURE 5.2 Measured spectrum asymmetry due to timing misalignment.3

Copyright © 2002 OSA.

As shown in Figure 5.1, the signal is distorted due to timing misalignment. This will lead to the change in spectrum. Therefore, by measuring the signal spectrum, the information of timing alignment can be obtained. In Reference 4, a novel approach for aligning the pulse carver and data modulator in a 9.9532-Gb/s, OC-192 optical transmitter using dual MZMs is presented. A mathematical explanation for the spectral phenomenon is presented and shown to agree quite well with measurement and simulation results. When there is misalignment, it is shown that there is a spectral null at 0.64  bit rate, when the MZM is biased at the point of maximum optical transmission (Figure 5.3). A simple alignment detection is shown in Figure 5.4(a) to measure the radio frequency (RF) power of the first spectral null with a 2-GHz spectral width around the null frequency. The relationship between the measured RF power and the relative time delay between the data modulator and pulse carver is given in Figure 5.4(b). Around 7.5 dB excursion in the RF power is obtained when the relative delay between the carver and data modulator is changed from the perfect alignment case to the maximum misalignment case. This technique relies solely on the photodiode-detected microwave power spectrum of the transmitted optical signal. As the timing misalignment is a slow process, low-speed analog-to-digital (A/D) and digital-to-analog (D/A) converters can be used. However, the proposed simple scheme is affected by the signal power variation in the transmission. To circumvent the problem, a more robust scheme that resembles a low-cost RF spectrum analyzer is also presented in Reference 4. A monitoring technique for the timing-misalignment for RZ-DPSK systems is proposed and demonstrated in Reference 5. The proposed alignment-monitoring module is composed of a polarizationmaintaining fiber (PMF), polarizer, and power meter. Due to the limited response of the phase modulator, the falling edge of a pulse and the rising edge of the subsequent pulse exhibit phase transition when the two pulses are encoded with two different phases. After propagating through a PMF, the phase transition period will induce a change of the degree of polarization (DOP), and thus a power change, when the signal passes through a polarizer. To achieve better resolution, the polarizer can be orthogonally aligned to the state of polarization (SOP) of the incoming signal; thus, reduction

CHAPTER 5 Timing misalignment monitoring

–20.0

–20.0

–30.0

–30.0

–40.0

–40.0 Power (dBm)

Power (dBm)

130

–50.0 –60.0 –70.0

–50.0 –60.0 –70.0

–80.0

–80.0

–90.0

–90.0

–100 0.00

5.00

10.0

15.0

–100 0.00

20.0

5.00

10.0

(a)

15.0

20.0

(b)

FIGURE 5.3 Measured spectra for (a) aligned and (b) misaligned cases. The first spectral null occurs at around 6.4 GHz.4

BPF

Low-Speed A/D

LPF

Photo- Schottky diode detector

CPU Low-Speed D/A

fdata /2

VoltageControlled Phase Shifter To Carver

(a)

Average power in a 2-GHz bandwidth (dBm)

Copyright © 2003 IEEE.

–80 –81 –82 –83 –84 –85 –86 –87 –88 –89 200 220 240 260 280 300 320 340 360 380 Time delay (ps)

(b)

FIGURE 5.4 (a) Simple alignment-detection scheme. (b) Measurements: microwave monitoring of misalignment.4 Copyright © 2003 IEEE.

of the signal’s DOP will result in the increase of monitoring signal power. When there is no timing misalignment, the phase transition period resides in the space between two adjacent pulses. When timing misalignment occurs, the phase transition period will move toward the center of the leading pulse or trailing pulse, resulting in the increase of monitoring signal power due to the decrease in DOP. Both the simulation and measurement results show that the proposed technique can monitor the BER degradation caused by timing misalignment between the pulse carver and data modulator. Only a low-speed power meter is needed. The scheme, however, has limited MPDR (0.2 dB for 50-ps

5.2 Monitoring of timing alignment

131

misalignment), and is susceptible to noise and environmental changes. To avoid the polarization fluctuation at the output of the PMF, further control of interferometric stability is needed. To enhance the monitoring sensitivity, a monitoring scheme that employs a simple off-center optical filter is proposed to extract optical power for alignment monitoring in RZ-DPSK systems.7,8 The operation principle is based on the property that frequency chirp is induced at the phase transition edge after data modulation by a phase modulator, as shown in Figure 5.5(a). With the increased misalignment amount, the signal spectrum is broadened, and power nulls will appear in the signal spectrum with the nulls’ spacing of 1/T, the bit rate (Figure 5.5(b)). Thus, by filtering out a narrow slice from the edge of the signal spectrum, any misalignment-induced spectrum broadening will be translated into an increase in the output power from the optical filter. By monitoring the optical power at the optical filter output, the alignment status can be obtained. Compared with the previous monitoring scheme using degree of polarization,6 in which the MPDR is 0.2 dB, a larger MPDR of 3.35 dB is achieved for the same timing alignment range of half-bit period, thus achieving a higher monitoring sensitivity. The monitoring power depends on the filter bandwidth and frequency offset from the signal’s center wavelength. A maximum MPDR of 3.35 dB is obtained when the optical filter with a bandwidth of 0.22 nm is placed at a 0.5-nm offset from the signal center wavelength. This simple alignment monitoring scheme has the desirable features of high-speed operation, polarization independence, and possible integration with transmitter for synchronization feedback control. In Reference 9, another technique for monitoring the timing alignment between a pulse carver and a phase modulator in RZ-DPSK systems is proposed. To monitor the spectrum broadening caused by timing misalignment, an optical frequency discriminator and a microwave detector centered at one-half of the phase modulation data rate are used, as shown in Figure 5.6. The frequency

Phase transient region

40

(i)

20

...

Worst case

...

Best case

...

...

Magnitude (a.u.)

Data pattern

0

40

–15

–10

–5

0

5

10

15

–15

–10

–5

0

5

10

15

–15

–10

–5

0

5

10

15

(ii)

20 0

40

(iii)

20 0

Misalignment tolerance range (a)

f/f0 (b)

FIGURE 5.5 (a) Illustration of timing alignment between pulse carver and data modulator. (b) Calculated signal spectra with timing alignment between pulse carver and data modulator of (i) 0, (ii) 0.3, and (iii) 0.5 T in a 10-Gb/s RZ-DPSK system with 0.28-T pulsewidth.8 Copyright © 2005 IEEE.

132

CHAPTER 5 Timing misalignment monitoring

Clock section

Data modulation

CW laser

Pulse carver

Phase modulator

Wavelength controller

Voltagecontrolled phase shifter

Controller

Tap coupler

NRZ data Clock

Optical RZ-DPSK receiver

|H( f )| 0.5 FSR

Optical frequency discriminator H(f)

f

RF power monitoring

[dV(t)]pk-to-pk

BPF@fc /2

(a)

(b)

FIGURE 5.6 (a) Proposed setup for monitoring clock misalignment. (b) Frequency-to-intensity conversion characteristic of delay-and-add discriminator.9 Copyright © 2006 IEEE.

discriminator, which is a thin-film filter with a linear frequency transition range of 0.3 nm, converts the frequency chirp into power variation. The RF power at 5.3 GHz of the frequency discriminator output is measured. This timing-misalignment detection method can be used when the phase modulation is implemented by a dual-arm MZM or a phase modulator. Compared with aforementioned monitoring schemes, this polarization-independent method can achieve a much larger MPDR of 17.5 dB within a range of 40 ps of misalignment, and therefore better monitoring sensitivity. In addition to RZ-DPSK, this method can also be applied to other modulation techniques such as RZ- and carrier-suppressed RZ-OOK modulation, which use two optical modulators for data and clock. In Reference 10, the delay-tap synchronous waveform sampling technique is proposed and experimentally demonstrated to monitor the alignment between the pulse carver and the data modulator in RZ-OOK systems. The proposed alignment monitoring scheme is based on a delay-tap, asynchronous waveform sampling technique.11 The two-tap scatter plot obtained for timing alignment within 50 to 50 ps is shown in Figure 5.7(a). The features of the plot are analyzed and two parameters, d and t, are defined to capture the information of timing misalignment. The parameter d represents the average distance of every sample point from the diagonal, while t represents the average angle of the sample points from the origin. Only the top-right part of the plot is considered, so as to reduce the computation time and the influence from the vertical and the horizontal edges. Figure 5.7(b) depicts the value of d and t for various misalignments. By using both d and t values derived from the measurement, the timing misalignment can be uniquely determined. This proposed scheme is able to determine both the misalignment magnitude and the sign of misalignment. The scheme is successfully demonstrated for monitoring the timing misalignment of a 10-Gb/s RZ-OOK transmitter. As it is not based on the frequency chirp of the modulator, the scheme is also applicable to the dual-drive Mach-Zehnder (MZ) modulator that has a very small chirp.

5.2 Monitoring of timing alignment

–30 ps

–40 ps

–50 ps

+30 ps

+40 ps

+50 ps

133

0 ps

1.8e–3

0.86

1.6e–3

0.84 0.82

1.4e–3

0.80 1.2e–3 0.78 1.0e–3

0.76

8.0e–4 6.0e–4 –50 –40 –30 –20 –10 0 10 20 30 40 Timing misalignment (ps) (b)

0.74 0.72 50

t, average angle of sample points (rad)

d, deviation from diagonal (a.u.)

(a)

FIGURE 5.7 (a) Two-tap plot for various modulation timing misalignment. (b) Measured d and t parameters for various timing misaligments.10 Copyright © 2006 IEEE.

Two-photon absorption (TPA) is a nonlinear process that can be used to monitor waveform distortion.12 At the same average input power, the efficiency of TPA processes in Si-APD depends on the optical waveform. Thus, TPA can be employed for monitoring the waveform distortion generated by the timing misalignment of the pulse carver and data modulation signals. The misalignment monitoring in optical transmitters with 20-Gb/s RZ-OOK and CSRZ-OOK modulations is demonstrated.13 The TPA efficiency decreases when the misalignment exceeds  10 ps. For RZ-OOK, the monitoring signal is around 3-2.7 (r.u.) for misalignment of 10 to 28 ps, whereas for CSRZ-OOK, it is 235-222 (r.u.) for misalignment from 10 to 28 ps. Within  10 ps, the TPA efficiency is less sensitive to misalignment variation. Since this monitoring scheme is not

134

CHAPTER 5 Timing misalignment monitoring

wavelength dependent, packaging it in MZMs has been proposed. No additional optical loss is incurred. The S-APD is quite sensitive and the destructive port of the modulator has sufficient power to be used for monitoring. Low-cost and low-speed electronics can be used to monitor high-speed data signals, as the scheme is data-rate independent. A few schemes to monitor the timing alignment have been described in this section. To summarize, in Table 5.1, the comparison of different schemes is provided based on the assessment of monitoring schemes as discussed in Section 5.1.

5.2.2 Synchronization for phase remodulation Synchronized phase remodulation to erase the original phase information and write the new phase information for phase-encoded modulation formats is a versatile signal processing technique for various applications in optical networks. The applications include optical label swapping in optical networks, provisioning of the optical virtual private network in a passive optical network (PON), the remodulation of the downstream signal for upstream transmission in wavelength-division multiplexing PON (WDM-PON), and the generation of spectral-efficient orthogonal modulation format.14–16 The second-phase modulator for phase remodulation needs to be in synch with the firstphase modulator; otherwise, waveform distortion is induced, leading to the change in the signal’s optical spectrum. In Reference 17, a simple and polarization-insensitive monitoring scheme for synchronized phase remodulation is demonstrated by using offset filtering with a narrowband optical bandpass filter (OBPF). The monitoring for the phase remodulation of an NRZ- and RZ-DPSK signal with a data rate of 10.61 Gb/s is investigated. To maximize the MPDR, the central wavelength of the OBPF needs to be optimized. Experimental results show that MPDR of 1.2 and 3.8 dB are obtained for NRZ-DPSK, and RZ-DPSK formats at an OBPF wavelength offset of 0.45 and 0.4 nm, respectively.

5.2.3 Synchronization for I/Q data and data/pulse carver Multilevel formats, such as differential quadrature phase-shift keying (DQPSK), are highly desirable for high-speed, long-haul transmission systems due to their tolerance of fiber nonlinearity and better spectral efficiency.2,18 In DQPSK, the in-phase (I) and quadrature-phase (Q) data are simultaneously transmitted in a single symbol time. When a pulse carver is used for generating RZ pulses, timing alignment monitoring for the synchronization of pulse carver and data modulator is needed. In addition, synchronization for the two orthogonal data channels is also necessary. In Reference 19, the authors experimentally demonstrate a technique for monitoring the time misalignment of in-phase/ quadrature-phase (I/Q) data streams and pulse carver/data in a 20-Gb/s RZ-DQPSK transmitter, as shown in Figure 5.8. The scheme is based on the measurement of the RF clock-tone power at 10 GHz (for I/Q data misalignment) and the low-frequency power at 600 MHz (for carver/data misalignment). As shown in Figure 5.9(a), an MPDR of 18 dB can be achieved for the monitoring of I/Q data misalignment by measuring the 10-GHz RF power. The monitoring scheme is more effective when the I/Q misalignment is more than 20 ps. For the carver/data misalignment, an MPDR of 6 dB is achieved when I/Q data misalignment is 0%, as shown in Figure 5.9(b). When there is 20% of I/Q data misalignment, the MPDR for the carver/data misalignment is degraded.

Table 5.1 Comparisons of Monitoring Schemes for Timing Misalignment Kim6

Sinsky4

Ku7 and Lu8

Tsai9

Ku10

Tian13

Yes

No

No

No

No

Yes

No

0.4 dB

0.2 dB

7.5 dB

3.35 dB

t: 0.73 to 0.84

0.46 dB (RZ) and 0.27 dB (CSRZ) for misalignment from 10 to 28 ps 28 to 28

30 to 30

50 to 50

50 to 50

50 to 50

17.5 dB (theoretical: 1 limited by phase detector noise) 40 to 40

10-Gb/s RZ-OOK No

10-Gb/s RZ-DPSK No

10 Gb/s RZ-OOK Yes

10.61-Gb/s RZ-DPSK No

10.61-Gb/s RZ-DPSK Yes

10-Gb/s OOK

No

Yes

No

No

No

No

d: (8 to 13.5)  104 50 to 50

Yes

20-Gb/s RZ-OOK and CSRZ-OOK No; data-rate independent No

5.2 Monitoring of timing alignment

Detection of the sign of misalignment Monitoring power dynamic range Misalignment range (ps) Data modulation High-speed receiver needed? Polarization dependent

Kang3

135

136

CHAPTER 5 Timing misalignment monitoring

Data I

Clock DQPSK

RZ-DQPSK

CW π/2

MZM

Data Q

RF power

RZ-DQPSK RF spectrum (direct detection)

RF power

DQPSK RF spectrum (direct detection)

Rs I/Q aligned

Rs freq misaligned

Rs carver aligned

Rs freq misaligned

FIGURE 5.8 Misalignment monitoring of an RZ-DQPSK transmitter.19

Copyright © 2008 IEEE.

0%

Normalized RF Power at 600 MHz (dB)

Normalized RF power at 10 GHz (dB)

0 –4 –8 50 ps/div –12 –16 50 ps/div –20 –50 –40 –30 –20 –10 0 10 20 30 40 I/Q data misalignment (% symbol period)

(a)

50

20%

–20% I/Q data misalignment

7 50 ps/div 6 5 4 3 2 50 ps/div 1 0 –1 –2 –50 –40 –30 –20 –10 0 10 20 30 40 50 Carver/data misalignment (% symbol period)

(b)

FIGURE 5.9 Monitoring signal power for (a) I/Q data misalignment and (b) carver/data misalignment.19 Copyright © 2008 IEEE.

5.2.4 OTDM clock recovery using timing misalignment of data pulses Albeit timing misalignment between the pulse carver and data modulator incurs power penalty and generally is an issue that needs to be resolved, timing misalignment can also be employed for useful applications. One example is to use the timing misalignment for the base-rate clock recovery in optical time-division multiplexed (OTDM) systems, where the line-rate clock is n times of the base rate with n being the number of TDM channels. In Reference 20, a base-rate clock recovery

5.3 Investigation of the effects of timing misalignment

137

technique is analyzed and demonstrated. In the proposed scheme, an electrical clock is extracted from an ultra-high-speed OTDM RZ signal with a detector and a bandpass filter. The basic idea is that when there is a small misalignment between adjacent pulses in the transmitted data, a subharmonic tone at the base-rate frequency will be generated. The base-rate tone is recovered as a clock signal at the receiver. A clock recovery circuit is experimentally demonstrated for the demultiplexing of 40- and 80-Gb/s optical TDM systems to 10 Gb/s. The effect of filter bandwidth, word length, and strength of timing shift on the received timing jitter are investigated. It is shown numerically that correlated TDM channels will decrease the jitter of the recovered clock considerably.

5.3 INVESTIGATION OF THE EFFECTS OF TIMING MISALIGNMENT In the previous section, various monitoring schemes for timing misalignment are discussed. In this section, the investigations of the effects of timing misalignment for some modulation formats are presented.

5.3.1 Clock/data synchronization in CPFSK systems Continuous-phase frequency-shift keying (CPFSK) is a promising modulation format because it has a compact spectrum and its receiver sensitivity can be improved with differential detection. A CPFSK modulator is implemented by embedding two MZMs in each arm of a main modulator.21 In the CPFSK modulator, synchronization between the clock and data signals is required, similar to the case of pulse carver and data modulation in RZ-DPSK transmitters. In Reference 22, a synchronous control technique for external optical modulation for CPFSK at 10 Gb/s is demonstrated. The modulator is shown in Figure 5.10(a). The simulated modulation spectral for CPFSK and BPSK is given in Figure 5.10(b) for comparison. In the proposed CPFSK scheme, the FSK signal in the upperor lower-sideband state synchronously shifts to the other state at the time when their phases are the same. Timing control between the sinusoidal clocks and the data signal is required to enable a synchronous control that gives rise to continuous-phase modulation. The delay between the clock and baseband signal is adjusted to achieve the optimal condition for the CPFSK modulation by monitoring its optical spectrum. The accuracy of the timing control required for the synchronization control is investigated. The timing misalignment causes phase discontinuity at the frequency shift and results in eye closure of the received signal. Experimental results show that the allowable timing misalignment for less than 1-dB power penalty of the receiver sensitivity is more than 25 ps, which is 25% of each bit period.

5.3.2 Misalignment between pulse carver/data modulator in RZ-DPSK systems The power penalty induced by the timing misalignment between pulse carver and data modulator for RZ-DPSK systems has been shown in the previous sections in various misalignment monitoring schemes. In Reference 23, the effect of the pulse shape is investigated. A 10-Gb/s Gaussian-shaped pulse train with a 33% duty cycle and raised-cosine NRZ data with the pulse shape sharpness factor a ¼ 0, 0.4, and 0.8 are assumed. It is shown that for an RZ-DPSK system, the timing misalignment– induced distortion (TMID) can be generalized as intersymbol interference (ISI). For a raised-cosineshaped NRZ pulse, the misalignment tolerance range would reduce from 25 to 15 ps in 10-Gb/s RZDPSK systems when a increases from 0 to 0.8. As expected, a narrower pulse is more tolerant of

138

CHAPTER 5 Timing misalignment monitoring

MZ-a MZ-c Laser

MZ-b

RF-a

RF-b π/2

Intensity, 20 dB/div

Modulation spectra (simulated) CPFSK

DATA

Wavelength offset, 0.5 nm/div

Δφb

(b)

Phase locked Sinusoidal clock

Baseband data

f0 = B/2 [Hz]

B [bit/s]

(a)

Intensity, 20 dB/div

cf BPSK (DPSK)

Wavelength offset, 0.5 nm/div (c)

FIGURE 5.10 (a) Basic structure of CPFSK modulator with synchronous control. Numerically calculated optical modulation spectra for (b) CPFSK and (c) BPSK (DPSK).22 Copyright © 2006 IEEE.

timing misalignment. The reason that the power penalty depends on parameter a is twofold. First, the temporally less-confined NRZ pulses with larger a values would experience more ISI. Second, when timing is misaligned, much energy leaks into the phase transition region, which enhances performance sensitivity to a. As TMID can be generalized as ISI, it suggests that TMID can be compensated by employing electronic equalization.

5.3.3 Misalignment between ASK and DQPSK modulation in ASK/DQPSK orthogonal modulation systems Recently, multilevel optical orthogonal modulations combining amplitude-shift keying and differential quadrature phase-shift keying (ASK/DQPSK) have drawn much attention as they exhibit better spectral efficiency, which helps alleviate some of the transmission distortions. In Reference 24, the tolerance for timing misalignment between 10-Gb/s ASK data and 20-Gb/s DQPSK data in ASK/DQPSK orthogonal modulation systems is investigated. Three different ASK formats are used, namely NRZ, RZ, and inverse-RZ. For DQPSK, the modulation can be implemented by a phase modulator (PM) or MZM, which leads to different phase and intensity characteristics, as shown in

5.3 Investigation of the effects of timing misalignment

Intensity

139

Phase PM

PM

MZM

MZM

FIGURE 5.11 Modulation characteristics of PM and MZM.24

Copyright © 2006 IEEE.

Figure 5.11. It is shown that impairment from timing misalignment strongly depends on ASK formats and DQPSK phase-modulation methods (PM or MZM). Monte-Carlo simulations were performed to investigate the dependence. In Figure 5.12, it is shown that DQPSK modulation methods influence the timing misalignment sensitivity of ASK 3

Receiver power penalty (dB)

3 2.5 2 1.5 1

NRZ–ASK/DQPSK

RZ–ASK/DQPSK

2

MZM, DQPSK MZM, ASK PM, DQPSK PM, ASK

MZM, DQPSK

1

MZM, ASK 0

0.5

PM, DQPSK PM, ASK

–1

0 –2

–0.5 –1

–50 –40 –30 –20 –10 0 10 20 30 40 50 Misaligned time t0 (ps)

–3

–50 –40 –30 –20 –10 0 10 20 30 40 50 Misaligned time t0 (ps)

(a)

(b)

Receiver Power Penalty (dB)

3 2

IRZ–ASK/DQPSK

1 0 –1 –2

MZM, DQPSK

MZM, ASK PM, ASK

PM, DQPSK –3

–50 –40 –30 –20 –10 0 10 20 30 40 50 Misaligned time t0 (ps)

(c)

FIGURE 5.12 Receiver power penalty versus t0 for various ASK formats and DQPSK modulation methods. (a–c) sampling phases are optimized.24 Copyright © 2006 IEEE.

140

CHAPTER 5 Timing misalignment monitoring

signals, but not of DQPSK signals. For instance, as shown in Figure 5.12(a), ASK data experienced large power penalty variation when DQPSK data were using MZM for modulation, but not when using PM. This can be attributed to the waveform distortion from the intensity characteristic of the DQPSK, as shown in Figure 5.11. In contrast, ASK formats influence the DQPSK signal’s timing misalignment sensitivity, but not the ASK signal’s. DQPSK is more robust to timing misalignment when ASK is using IRZ format. The results also show that the performance of IRZ-ASK/DQPSK can be improved by using certain timing misalignments, as depicted in Figure 5.12(c).

5.4 MITIGATION OF TIMING MISALIGNMENT Monitoring of timing misalignment and effects of timing misalignment for different modulation formats were discussed in the previous sections. In this section, a few schemes that alleviate the impairment induced by timing misalignment for different applications are presented.

5.4.1 Hybrid OTDM scheme for demultiplexing with better timing misalignment tolerance In an OTDM system, as the data rate increases, demultiplexing of optical tributary channels becomes more and more challenging. It would be desirable to allow channel demultiplexing with a larger timing misalignment tolerance or a relaxed switching window. In Reference 25, a novel OTDM scheme with hybrid RZ-ASK/RZ-DPSK formats that eases the demultiplexing difficulty is proposed and investigated. The RZ-ASK channels are interleaved with the RZ-DPSK channels in time domain. For an OOK channel in the proposed hybrid OTDM signal, the crosstalk from adjacent DPSK channels is a constant power, which mainly alters the detection threshold. Similarly, for a DPSK channel in a hybrid OTDM signal, the demultiplexing with a relaxed switching window will gate part of the adjacent OOK bits. After DPSK demodulation with delayed interferometer (DI), the gated part of the OOK bit will destructively interfere with that of its previous bit; thus, it will vanish or become onequarter of its original power, depending on whether these two consecutive bits are the same or not. For 40-Gb/s OTDM systems, the tolerances to demultiplexing timing misalignment are improved by 180% and 70%, for RZ-ASK and RZ-DPSK channels, respectively. In Reference 26, the proposed scheme is applied to an 84.88-Gb/s OTDM system with a channel data rate of 10.61 Gb/s. Compared with the conventional OTDM system with homogenous modulation format for different OTDM channels, the demultiplexed channel in a hybrid OTDM signal suffers from less degradation caused by the possible crosstalk from the adjacent channels. 84.88- to 10.61-Gb/s hybrid OTDM demultiplexing is achieved with a relatively wide switching window, which cannot be realized by using the conventional OTDM.

5.4.2 Novel remodulation scheme for colorless high-speed WDM-PON without remodulation synchronization Wavelength-division multiplexing passive optical network (WDM-PON) has aroused much attention for next-generation broadband access architecture, due to its large dedicated bandwidth per user and upgrade flexibility. A centralized light source is desirable in WDM-PON because such source in the central office (CO) eliminates the need of the wavelength-specific transmitters at the optical

5.5 Summary

141

network units (ONUs) and features cost reductions of wavelength management on the customer side. For previous schemes using downstream DPSK and upstream infinite-extinction ratio (ER) OOK, it is shown that despite its better back-to-back sensitivity, closed eye is exhibited for the upstream signal when there is no chromatic dispersion compensation and remodulation synchronization. At ONU, the downstream DPSK signal exhibits intensity fluctuation at the phase transition period, thus the upstream data at ONU needs to be synchronous with the downstream data during remodulation. In Reference 27, a remodulation scheme using downstream 10-Gb/s OOK and upstream 10-Gb/s DPSK is proposed for a 30-km-range colorless WDM-PON without dispersion compensation and remodulation synchronization. The downstream OOK signal has an ER of 4.9 dB, resulting in a 3.5-dB, back-to-back penalty. The upstream DPSK data are encoded using a phase modulator. After transmission, only 1-dB penalty for both 10-Gb/s downstream and 10-Gb/s upstream signals is demonstrated. It is shown that the proposed scheme is robust for remodulation misalignment, with the receiver sensitivity variation less than 0.6 dB.

5.4.3 Misalignment mitigation using MLSE equalizers As discussed in Reference 23, the distortion induced by timing misalignment between the pulse carver and data modulator can be generalized as intersymbol interference; thus, it can be mitigated by electronic equalization. A novel method using an electronic equalizer is proposed to mitigate the impairment from timing misalignment in RZ and carrier-suppressed RZ systems.28 Simulations results show that a maximum-likelihood sequence equalizer (MLSE) significantly reduces timing misalignment-induced power penalty and sampling phase sensitivity. Simultaneous mitigation of both timing misalignment-induced distortion and polarization-mode dispersion (PMD) using a single MLSE is also demonstrated. In Reference 29, the performance of maximum-likelihood sequence estimation (MLSE) receiver in the presence of the impairments from both the pulse carver/data modulator timing misalignment and PMD in optically amplified 10-Gb/s RZ systems are investigated. The results show that by using the MLSE receiver, the power penalty for the worst timing misalignment, where the eye is completely closed, is limited to 6 dB in the absence of PMD and 9 dB in the presence of the worst PMD. The investigation validates the effectiveness of an MLSE receiver for combating both timing misalignment and PMD with shared electrical devices, which therefore reduces the required number of compensation components.

5.5 SUMMARY Various timing alignment monitoring schemes have been presented in this chapter. For future optical networks, as the line rate increases, the timing alignment requirement becomes more and more stringent. Thus, considering the scalability of the monitoring schemes is essential. Most monitoring schemes exhibit a small monitoring power change, and thus limited resolution, when the timing misalignment is small. Although the power penalty is small when the timing misalignment is small, enhancing the monitoring sensitivity is desirable while simultaneously avoiding monitor range reduction. Another issue is the response time of monitoring. Some monitoring schemes take advantage of the slow change of timing misalignment due to environment variation, and therefore employ only low-speed detection circuits to simplify the design. A much faster monitoring scheme may be

142

CHAPTER 5 Timing misalignment monitoring

necessary for future optical packet-switching networks. Considering WDM applications, it would be useful if multichannel timing misalignment monitoring schemes can be devised. Finally, it is always desirable to have a monitoring scheme that can facilitate the multi-impairment monitoring.30 It would be interesting to implement an integrated monitoring scheme that is capable of monitoring timing misalignment and other impairments simultaneously, such as wavelength detuning and power variation at the transmitter. In conclusion, various timing misalignment monitoring schemes were discussed and compared in this chapter. Performance metrics, including MPDR, monitoring range, data rate, complexity, and so on, were discussed to facilitate comparison. Although this chapter is mainly focused on the synchronization between pulse carver and data modulation, various investigations into the effect of timing misalignment and means to mitigate effects of timing misalignment are also presented. It is hoped that the information provided in this chapter will be useful for more general applications that require timing synchronization.

REFERENCES 1. Xu C, Liu X, Wei X. Differential phase-shift keying for high spectral efficiency optical transmissions. IEEE J Sel Top Quantum Electron 2004;10:281–93. 2. Gnauck AH, Winzer PJ. Optical phase-shift-keyed transmission. J Lightwave Technol 2005;23:115–30. 3. Kang I, Mollenauer L, Greene B, Grant A. A novel method for synchronizing the pulse carver and electroabsorption data modulator in RZ optical transmitters. In: Proc. OFC, paper ThCC4. Anaheim, CA; 2002. 4. Sinsky JH. High-speed data and pulse-carver alignment in dual Mach-Zehnder modulator optical transmitters using microwave signal processing. J Lightwave Technol 2003;21:412–23. 5. Monitoring alignment between pulse carvers and phase modulators in optical systems. United States Patent 6972842. 6. Kim H, Doerr CR, Pafchek R, Stulz LW, Bernasconi P. Alignment monitoring of the pulse carver and data modulator for RZ-DPSK systems. IEEE Photon Technol Lett 2003;15(11):1594–6. 7. Ku YC, Lu GW, Chan CK, Chen LK. Novel technique for modulation alignment monitoring in RZ-DPSK systems using off-center optical filtering. In: Proc. OFC, paper OWJ4. Anaheim, CA; 2005. 8. Lu GW, Ku YC, Chen LK, Chan CK. A novel technique for pulse-carver and data alignment monitoring in RZ-DPSK systems using off-center optical filtering. IEEE Photon Technol Lett 2005;17(3):711–3. 9. Tsai KT, Lu GW, Chen LK, Winston I, Way WI. Alignment monitoring technique for pulse carver and data modulator in RZ-DPSK systems using an optical frequency discriminator. IEEE Photon Technol Lett 2006;18(10):1119–21. 10. Ku YC, Chan CK. High-speed data and pulse-carver alignment in RZ-OOK systems using delay tap asynchronous waveform sampling. In: Proc. ECOC, paper Tu4.2.2. Cannes, France; 2006. 11. Dods SD, and Anderson TB. Optical Performance Monitoring Technique Using Delay Tap Asynchronous Waveform Sampling. In: Proc. OFC, paper OThP5. Anaheim, USA; 2006. 12. Tian C, Kinoshita S. Waveform monitoring with polarization-dependence-eliminated two-photon absorption in Si-APD in high-speed WDM systems. In: Proc. ECOC, paper We4.P.70. Glasgow, UK; 2005. 13. Tian C, Naito T. Data and pulse-carver alignment in high-speed optical transmitters monitored with two-photon absorption in Si-APD. In: Proc. SPIE, paper 63532Y. Soc Photo Opt Instrum Eng 2006;6353. 14. Hung W, Chan CK, Chen LK, Tong F. A bit-serial optical packet label swapping scheme using DPSK encoded labels. IEEE Photon Technol Lett 2003;15(11):1630–2.

References

143

15. Tian Y, Su Y, Yi L, Leng L, Tian X, He H, et al. Optical VPN in PON based on DPSK erasing/rewriting and DPSK/IM formatting using a single Mach-Zehnder modulator. In: Proc. ECOC, paper Tu4.5.6. Cannes, France; 2006. 16. Pun S-S, Chan C-K, Chen L-K. Demonstration of a novel optical transmitter for high-speed differential phase shift keying/inverse return-to-zero orthogonally modulated signals. IEEE Photon Technol Lett 2005;17(12):2763–5. 17. Zhao J, Ku YC, Chen LK. Simple monitoring technique for synchronised phase remodulation using narrowband optical filtering. IEE Electron Lett 2007;43(6):63–4. 18. Kim H, Essiambre R-J. Transmission of 820 Gb/s DQPSK signals over 310-km SMF with 0.8-b/s/Hz spectral efficiency. IEEE Photon Technol Lett 2003;15:769–71. 19. Wu X, Christen L, Zhang B, Peng W-R, Yang J-Y, Zhang L, et al. Synchronization monitoring of I/Q data and pulse carving misalignment for a parallel-type RZ-DQPSK transmitter by measuring RF clock tone/low frequency power. IEEE Photon Technol Lett 2008;20(24):2138–40. 20. Hansryd J, Andrekson PA, Bakhshi B. Prescaled clock recovery based on small timing misalignment of data pulses. J Lightwave Technol 2001;19(1):105–13. 21. Kawanishi T, Sakamoto T, Shinada S, Izutsu M, Higuma K, Fujita T, et al. High-speed optical FSK modulator for optical packet labeling. In: Proc. OFC, paper PDP16. Los Angeles, CA; 2004. 22. Sakamoto T, Kawanishi T, Miyazaki T, Izutsu M. 10-Gb/s external modulation in optical CPFSK format. IEEE Photon Technol Lett 2006;18(8):968–70. 23. Zhao J, Chen LK, Chen CK. Performance degradation induced by pulse carver/data modulator misalignment in RZ-DPSK systems. In: Proc. IEEE LEOS, paper JThE62. Baltimore, MD; May 2005. 24. Zhao J, Chen LK, Chan CK. Tolerance of misalignment between ASK and DQPSK modulation in ASK/ DQPSK orthogonal modulation systems. In: Proc. OECC, paper 5F3-3. Kaohsiung, Taiwan; 2006. 25. Deng N, Chan CK. Enhanced tolerance to demultiplexing misalignment in an OTDM system with hybrid RZ-ASK/DPSK formats. In: Proc. OFC/NFOEC, paper JWA51. Anaheim, CA; March 2007. 26. Deng N, Chan CK, Chen Lk. A hybrid OTDM scheme with enhanced demultiplexing performance. IEEE Photon Technol Lett 2007;19(19):1454–6. 27. Zhao J, Chen LK, Chan CK. Novel re-modulation scheme to achieve colorless high-speed WDM-PON with enhanced tolerance to chromatic dispersion and re-modulation misalignment. In: Proc. OFC/NFOEC, paper OWD2. Anaheim, CA; March 2007. 28. Zhao J, Chen LK, Chan CK. Mitigation of timing misalignment-induced distortion using electronic equalizer in RZ/CSRZ systems. IEEE Photon Technol Lett 2005;17(5):1106–8. 29. Zhao J, Chen LK, Chan CK. Maximum likelihood sequence estimation in the presence of timing misalignment and polarization mode dispersion in optically amplified return-to-zero systems. In: Proc. IEEE ICC, paper CT20-7. Istanbul, Turkey; June 2006. 30. Chen LK, Chan CK, Lu GW, Ku YC, Lin CL. Optical performance monitoring and network diagnosis in reconfigurable optical networks. In: Proc. APO 2007, paper 6784-54. Wuhan, PRC; November 2007.

This page intentionally left blank

CHAPTER

6

Optical performance monitoring based on asynchronous amplitude histograms ,{

,{

Ruben S. Luis* , Liliana Costa*, Anto´nio Teixeira* , Paulo Andre´*

,}

*

Instituto de Telecomunicac¸o˜es, Campus de Santiago, Aveiro, Portugal Center of Volcanology and Geological Risk Assessment, University of Azores, Ponta Delgada, Portugal { Departamento de Electro´nica, Telecomunicac¸o˜es e Informa´tica, University of Aveiro, Campus de Santiago, Aveiro, Portugal } Departamento de Fı´sica, University of Aveiro, Campus de Santiago, Aveiro, Portugal

{

6.1 INTRODUCTION Optical signals traveling along a network are fundamentally degraded by amplified spontaneous emission (ASE) noise due to optical amplification, crosstalk, and signal distortion as a result of transmission through optical components and optical fibers.1 With the rapid increase of transmission bit rates and the channel density in wavelength-division multiplex (WDM) systems, the impact of signal distortion or crosstalk on the quality of the optical signals may equal or even surpass the currently dominant impact of ASE noise.1 As such, there is an urgent need for the development of optical monitoring systems (OMSs), which effectively distinguish and quantify the impact of these forms of degradation. Generally, the development of OMS is performed subsequent to prior advancements in transmission technologies or components. Furthermore, the cost of advanced OMS often imposes development on a requirement basis—that is, OMSs are usually developed as support subsystems for specific applications, when necessary. For this reason, OMSs are usually one step behind the development of mainstream optical transmission systems. It is not uncommon that a given optical monitoring technology only reaches maturity after the transmission technology that it was initially developed for is dropped in favor of alternative advancements. Although this is not a rule, one may consider that the deployment of commercial optical transmission systems, which requires an established maturity of all subsystems, has been performed with limited or simply inexistent optical monitoring technology. Given the rapid increase in recent years of requirements for optical transmission systems in terms of bit rate, distance, and overall flexibility, one may expect that this cycle will be broken in the near future as the OMS becomes an enabler of optical transmission systems. Hence, the development of OMSs since the implementation of WDM systems has brought a large number of solutions, which is steadily increasing.1 In this context, it may be assumed that the most interesting optical monitoring technologies are those that can be easily adapted or upgraded to support the rapid advances of other technologies used

© 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00006-7

145

146

CHAPTER 6 OPM based on asynchronous amplitude histograms

in optical transmission systems. Following this premise, OMS based on the analysis of asynchronous histograms have attracted significant attention due to their simplicity and remarkable flexibility.2 Asynchronous histograms may be acquired by sampling the signal under analysis at an arbitrary rate, much lower than the symbol rate, which avoids the need for clock recovery systems. This potentially reduces the overall cost of the monitor and gives it flexibility by allowing the application of the same monitor to signals with arbitrary bit rates. Besides the simplicity of the histogram acquisition system, the analysis of asynchronous histograms can be performed using relatively simple algorithms, which may be implemented via software. This allows improving and upgrading the OMS indefinitely, simply by replacing its firmware rather than its hardware. To illustrate the usage of asynchronous histograms, consider Figures 6.1(a) and (b), which show a comparison between the simulated synchronous and asynchronous eye diagrams of a non-return-tozero (NRZ) signal and the corresponding histograms. In the case of the histogram acquired synchronously, the considered samples are located within an interval of 10% of the bit period around the center of the eye diagram. As shown in Figure 6.1(a), the synchronous histogram of the signal presents two main peaks corresponding to the samples acquired during the mark and space symbols. The statistical distribution of the samples around the peaks reflects the noise distributions associated with the corresponding symbols. In the case of the asynchronous histogram, one may also distinguish two main peaks corresponding to the mark and space symbols, similar to the case of the synchronous histogram. However, it is not possible to entirely dissociate the samples corresponding to the mark and space Synchronous histogram Electrical current (a.u.)

Electrical current (a.u.)

Synchronous eye diagram

Time (a.u.)

Occurrences (a) Asynchronous histogram

Electrical current (a.u.)

Electrical current (a.u.)

Asynchronous eye diagram

Time (a.u.)

Occurrences (b)

FIGURE 6.1 (a) Synchronous and (b) asynchronous eye diagrams and corresponding histograms of an NRZ signal.

6.1 Introduction

147

symbols. This is resulted from the large number of samples located between the two peaks, which have been acquired during the transitions between symbols, also referred to as crosspoint data. A study of the changes in the shape of asynchronous histograms as a result of optical impairments was first presented by Hanik et al.2 To illustrate this impact, consider Figures 6.2(a–c), which present examples of asynchronous histograms degraded by different levels of ASE noise, intraband crosstalk, and group velocity dispersion (GVD), respectively. In the first case, it is shown that decreasing the optical signal-to-noise ratio (OSNR) broadens and reduces the magnitude of the peaks corresponding to the mark and space symbols. This is particularly evident for the mark symbol, where the beating between signal and ASE noise is stronger. Note that the magnitude of the crosspoint data is not significantly affected when varying the OSNR. In the case of Figure 6.2(b), the OSNR is maintained and the amount of intraband crosstalk is increased by combining the original signal under analysis with a delayed replica. The impact of intraband crosstalk manifests as an emerging floor around the mark symbol. 0.125

0.16

S/N = 31dB (BER<10–12) 19 dB (BER = 10–10)

0.100

S/N = 20 dB (BER<10–12) 15 dB (10–11)

0.12

13 dB (10–6)

P,A

P,A

0.075 16 dB (BER = 10–8)

0.050

0.08 0.04

0.025 0 30

0

90

30 60 Amplitude A (a.u.) (a)

0 –20

0

20 40 60 Amplitude A (a.u.) (b)

80

0.125 noise (BER<10–12)

0.100

P,A

0.075 0.050

160 km SMF (BER<10–12) 320 km (10–11)

0.025

480 km (10–9)

0 –20

0

20 40 60 Amplitude A (a.u.) (c)

80

100

FIGURE 6.2 Examples of asynchronous histograms acquired from an NRZ signal (a) in presence of ASE noise for different values of signal-to-noise ratio; (b) in presence of intraband crosstalk; and (c) when the signal is impaired by fiber dispersion.2 Copyright © 2009, Institution of Engineering and Technology.

148

CHAPTER 6 OPM based on asynchronous amplitude histograms

In the case of Figure 6.2(b), the OSNR is maintained and the amount of accumulated dispersion is increased by placing different standard fiber lengths prior to the optical receiver. This experiment has been conducted using a 5-Gb/s signal.2 The main characteristic of the impact of dispersion on asynchronous histograms is the increase of the amplitude of the crosspoint data. This results from the GVD-induced increase of the symbol transition time. It is also shown that the GVD-induced distortion of the signal takes the form of multiple rails arising near the nominal level of the space symbol. Note also that the mark symbol is broadened, similar to what occurs when the OSNR is decreased. Hanik et al. have shown that these modifications can be easily analyzed to infer information regarding the status of the signal under analysis. This information may regard the signal quality (Q-factor or bit error rate [BER] estimation) or the optical impairments that affect the signal (OSNR, GVD, etc.). Since the initial works by Hanik et al. and also by Shake et al.,3 a large number of asynchronous histogram acquisition and analysis techniques have been proposed, adapting the original principle to multiple modulation formats and evaluating different optical transmission impairments. For brevity reasons, this work will only approach a few of the most significant. Section 6.2 describes three distinct

Tunable LD (1553 nm)

LiNbO3 modulator

~

PPG

Dispersive medium

10 Gb/s

Error detector

Dispersion D ps/nm

EDFA Variable attenuator

O/E Sampling oscilloscope

BW = 6 GHz

(a)

Optical laser source

Optical signal Modulate

Coupler

ATT

O/E

Error detector

Q-factor monitor system

Pulse repetition rate f0 Optical frequency wsig

Transmission line

Signal processing circuit Sampling optical SF light pulse f0 / n – Δf,wsam f0 / n – Δf,wsam + wsig Sampling optical pulse source SFG crystal

APD

(b)

FIGURE 6.3 Examples of histogram acquisition systems. (a) Based on electrical sampling using an unsynchronized digital oscilloscope4 (# 2009, IEEE). (b) Based on optical sampling using a sum-frequency generation crystal.3 Copyright © 2009, Institution of Engineering and Technology.

6.2 Monitoring techniques based on analysis of asynchronous histograms

149

techniques to analyze asynchronous histograms in order to obtain information regarding the quality of the signal. Furthermore, Section 6.2 also includes a technique to address the OSNR of the signal under analysis based on the comparison of the acquired histograms with reference histograms. Asynchronous histograms may be acquired by sampling the optical signal at an arbitrary rate in the electrical domain4 or optical domain,3 as illustrated in Figures 6.3(a) and (b), respectively. In the case of electrical sampling, the signal under analysis is detected using a fast photodetector and electrically sampled at an arbitrary bit rate using an unsynchronized oscilloscope. Note that the presented scheme is used for experimental purposes only, and the oscilloscope would be replaced by a significantly less-costly standalone electrical sampling system. The samples then follow to a histogram counter, which is responsible for computing the asynchronous histogram. The fast photodetector is required to detect the signal with minimal distortion. In the case of optical sampling, the samples are produced by an optical sampling gate and follow to a slow photodetector followed by the sample counter. As such, the bandwidth requirements of the photodetector are simply to exceed the sampling frequency, avoiding distortion of the acquired samples. In both cases, the main bandwidth requirements are with the sampling gate (optical or electrical), which should be as high as possible to avoid averaging effects. The limitations of the acquisition system are a significant aspect of the design of monitoring systems based on asynchronous histograms. These problems are addressed in Section 6.3, where several methods are proposed to evaluate the requirements of the acquisition system.

6.2 MONITORING TECHNIQUES BASED ON ANALYSIS OF ASYNCHRONOUS HISTOGRAMS 6.2.1 Q-factor monitoring

Q-factor or BER monitoring is the preferred method for signal quality evaluation in optical transmission systems.1 The information provided by multiple Q-factor (or BER) monitors placed within an optical network allows the detection and localization of signal-degrading faults, which is fundamental for maintenance and operation tasks. To achieve this, OMSs based on digital processing of information contained in the monitored signal (e.g., error block counting or forward-error correction codes) are often used in the terminal points of optical networks, or in intermediate points, where electrical regeneration of the optical signals is performed. These systems are effective for obtaining in-service estimates of the BER, and are usually integrated with the digital receivers of the transmission systems to minimize cost. However, as the evolution of transparent optical networks tends to reduce the number of electrical regenerators, digital performance monitoring systems may become too costly to implement in intermediate points of the network as standalone applications. What is greatly desired is a method for signal quality monitoring that can provide a good measure of signal quality without the complexity of termination.5 In this context, the analysis of asynchronous histograms has been extensively proposed as a simple, low-cost, and effective means to perform Q-factor monitoring in optical networks. Most of the available techniques to estimate Q-factor through asynchronous histograms available in the literature are bit rate independent and insensitive to frame format or any other signal structure, requiring only that a continuous bit stream is present for analysis. Shake et al. have contributed significantly to the analysis of asynchronous histograms by developing methods to estimate the Q-factor and corresponding BER, based on processing the

150

CHAPTER 6 OPM based on asynchronous amplitude histograms

Amplitude

Amplitude s1,avg

m1,avg

am m

m0,avg

Mark Thresholds

am s0,avg Number of times

Space Time

FIGURE 6.4 Schematic of asynchronous histogram for an average Q-factor analysis technique.3 Copyright © 2009, Institution of Engineering and Technology.

asynchronous histograms to remove the crosspoint data.3–8 Figure 6.4 illustrates the basis of Shake’s method to analyze asynchronous histograms. One begins by taking the values of electrical current corresponding to the maximums of the peaks associated with the mark and space symbols, m1,avg and m0,avg, respectively. Note that although these values do not necessarily correspond to the actual nominal values of electrical current for the mark and space symbols m1 and m0, respectively, they are a fair approximation if the signal under analysis is undistorted. The next step involves defining two thresholds below and above m1,avg and m0,avg, respectively, to eliminate the crosspoint data. These thresholds should be located at the electrical currents ath
Dm below and above m1,avg and m0,avg, respectively.3 The terms ath and Dm ¼ m1,avg  m0,avg are an adjustment constant and the difference between the electrical current levels associated with the mark and space symbols, respectively. After eliminating the crosspoint data it becomes possible to estimate the standard deviations of the electrical current associated with the mark and space symbols as s1,avg and s0,avg, respectively. This allows the definition of an average Q-factor as m1;avg  m0;avg : (6.1) Qavg ¼ s1;avg þ s0;avg It has been shown in Reference 3 that the average Q-factor, Qavg, is proportional to the Q-factor of the signal under analysis, Q. This is illustrated in Figure 6.5, extracted from Reference 3. Furthermore, the proportionality constant that relates these quantities is maintained even if the signal under analysis is degraded by GVD.4 A variation of this technique was developed to simultaneously monitor the impact of GVD on the Q-factor. The GVD monitoring technique takes into account the fact that the GVD-induced distortion alters the difference between the average power levels associated with the mark and space symbols and the standard deviation of the mark symbol.8,9 This effect is illustrated in Figure 6.6, extracted from Reference 4, where the average Q-factors computed from a signal affected by different amounts of GVD are compared. Shake’s method quantifies the GVD-induced distortion through the waveform distortion parameter Xs1 þ Xm. The terms Xs1 and Xm correspond to the variation of s1,avg and Dm

6.2 Monitoring techniques based on analysis of asynchronous histograms

151

12

Averaged Q-factor

Averaged Q-factor (Qavg), dB

14

10 8 6 4 2 12

14 Q(t0), dB

16

FIGURE 6.5 Relationship between Q-factor and average Q-factor using analysis of an NRZ signal. ○, a ¼ 0.1; x, a ¼ 0.2; D, a ¼ 0.3; □, a ¼ 0.4; þ, a ¼ 0.49.3 Copyright © 2009, Institution of Engineering and Technology.

11

Qavg, dB

10

9

8

7

6

6

8

10

12 Q, dB

14

16

18

FIGURE 6.6 Relationship between Q-factor and average Q-factor using analysis of an NRZ signal for different values of dispersion-impairing 10-Gb/s signal under analysis. ○, 0 ps/nm; x, 1190 ps/nm; D, 1530 ps/nm.4 Copyright © 2009, IEEE.

between two consecutive measurements, respectively. A variation of the average Q-factor accompanied by a strong variation of the waveform distortion parameter indicates that the degradation (or restoration) of the signal is a result of a change of the accumulated dispersion.8 Otherwise, the monitor assumes that the quality variation results from fluctuations of the ASE noise power within the signal bandwidth.

152

CHAPTER 6 OPM based on asynchronous amplitude histograms

The main limitation of Shake’s method is that it assumes that the noise distributions around the mark and space symbols are Gaussian, leading to an incorrect estimate. To improve this aspect, Rasztovits-Wiech et al. proposed an alternative approach, where a more accurate distribution of noise is considered while taking into account distortion of the signal due to intersymbol interference.10 The method proposed by Rasztovits-Wiech et al. begins by eliminating the crosspoint data by mirroring the outer areas of the asynchronous histogram around the power levels corresponding to the mark and space symbols.10 This approach was also used in Reference 11 to analyze the Q-factor of RZ signals, where the distinction of the mark symbols in asynchronous histograms is less evident. As illustrated in Figure 6.7, the region of the histogram between m1,avg and m0,avg is eliminated and then reconstructed by flipping the outer regions of the histogram about the axis defined by the mark and space levels. The next step in the Rasztovits approach is to approximate the reconstructed histogram with an estimated histogram, given by " # Nb X hSIG;i ðan  ai Þ2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
exp  2 ; (6.2) hEST;n ¼ 2sEST ðai Þ 2p
s2EST ðai Þ i¼1 where s2EST ðai Þ is the estimated variance of the Gaussian process associated with the current level ai. It is also assumed that the histograms have Nb bins and the ith bin is centered in the electrical current amplitude level ai. The term hSIG,i is the histogram of the signal under analysis in the absence of noise. Note that Equation (6.2) assumes that the estimated histogram is a sum of strategically centered Gaussian distributions weighted by the probability density function of the signal in the absence of noise, as illustrated in Figure 6.8.10 This is a significant improvement from the Gaussian approach by Shake et al. and allows handling histograms from signals degraded by intersymbol interference (ISI). By matching the estimated histogram with the measured histogram, it becomes possible to have a set of estimated values for s2EST ðai Þ and hSIG,i, which allow the computation of the BER as 2 3 ! !7 6 X X 7 16 D  a a  D i i 7; þ (6.3) Pe  6 hSIG; i
erfc pffiffiffi hSIG; i
erfc pffiffiffi 6 24 2sEST ðai Þ 7 2sEST ðai Þ 5 i i ai < D

ai  D

Cut

Amplitude

Amplitude

FIGURE 6.7 Schematic of histogram crosspoint elimination method.10

Copyright © 2009, Institution of Engineering and Technology.

6.2 Monitoring techniques based on analysis of asynchronous histograms

Probability density

0

Decision threshold D

153

1

Amplitude l0, N

l0, 1

l1, 1

l1, M

FIGURE 6.8 Illustration of BER estimation method using multi-Gaussian fitting of asynchronous histogram after process to eliminate crosspoint data.10 Copyright © 2009, Institution of Engineering and Technology.

where D is an optimizable decision threshold. This method is applicable to asynchronous histograms acquired from binary signals with arbitrary forms of distortion. It is more effective than the previously described Shake’s method in the sense that the latter assumes that the mark and space symbols are well defined, which is untrue for signals with ISI. As a disadvantage, Rasztovits’ method requires a considerable amount of computation in order to estimate the values of sSIG (ai) using nonlinear minimization algorithms. Rasztovits’ approach was further developed in References 12–17, which present simplified techniques to reduce the computation required to numerically fit the estimated histograms to the histograms acquired from the signal under analysis. The approaches proposed by Rasztovits and Shake required altering the shape of the asynchronous histogram to allow a distinction between the mark and space symbols, and from that the signal performance was extracted. Alternatively, other authors have proposed techniques to extract information on the optical impairments degrading the signal under analysis without changing the shape of the asynchronous histogram. As an example, Chen et al. have proposed to use the crosspoint data of asynchronous histograms to determine the transition time of the signal impulses, which was shown previously to relate directly with the amount of GVD.18,19 The latter reference also extended this study to include the impact of polarization mode dispersion. Another example is the estimation of in-band crosstalk with synchronous histograms proposed by Weinert et al. in References 20 and 21 and further developed in Reference 22. This technique uses a deconvolution method to remove the impact of ASE noise from the synchronous histograms leaving only the impact of in-band crosstalk, which may be easily analyzed. The principle of this technique is illustrated in Figure 6.9. Further developments of this technique have allowed an evaluation of intersymbol interference and jitter.23 However, its application to asynchronous histograms is not straightforward due to the nonuniform conditional distribution of the Gaussian and arc-sin probability density functions of ASE noise and in-band crosstalk, respectively, which limit the application of deconvolution methods.24 This limitation may be handled by converting the asynchronously acquired samples in synchronous samples using their intrinsic correlation, as proposed in References 7, 25, and 26.

154

CHAPTER 6 OPM based on asynchronous amplitude histograms

PDF

WP

70

90

110 130 U, mV (a)

70

150

90

110 130 U, mV (b)

150

FIGURE 6.9 Histogram of mark symbol of signal affected by intraband crosstalk (signal-to-crosstalk ratio of 19 dB) and degraded by Gaussian noise (a) without deconvolution and (b) after deconvolution and filtering. In the case of (b), the crosstalk floor around the mark symbol becomes clearly visible and may be evaluated.21 Copyright © 2009, Institution of Engineering and Technology.

An interesting method to analyze the Q-factor without eliminating the crosspoint data has been proposed by Andre´ et al. in Reference 27 and further developed in Reference 17. The basis of Andre´’s approach is to numerically compute an estimated histogram, which may be compared with the measured histogram. The shape of the estimated histogram is modified by adjusting the parameters that condition it (mark and space levels, transition time, noise variances, or others) until a match between estimated and measured histograms is found. Figure 6.10 presents a schematic of the method proposed in Reference 27 to compute the estimated histogram.

m1

* m0

m1 tr

Theoretical histogram of signal in absence of noise, hSIG,i

Component of Include impact of mark symbol, noise associated with mark hSIG1,i symbol m0

m1

sEST,1

sEST,0

Estimated histogram associated with mark symbol m0

m0

m1

Estimated histogram, hEST,i

* Component of space symbol, hSIG0,i

Include impact of noise associated with space symbol

Estimated histogram associated with space symbol

FIGURE 6.10 Diagram of method to estimate an asynchronous histogram of the signal impaired with the ASE noise proposed.27

6.2 Monitoring techniques based on analysis of asynchronous histograms

155

In more detail, Andre´’s method consists in calculating the theoretical histogram of the signal under analysis, in the absence of noise, hSIG,i, and numerically includes the impact of optical and electrical noise. To illustrate this method, we begin by calculating hSIG,i by considering that the received signal, in the absence of noise and neglecting the influence of the sampling system, is a linear superposition of NRZ binary impulses with equal probability for the mark and space symbols. The transitions from space to mark and from mark to space have amplitude shapes given by s01(t) and s10(t), respectively. The transitions from mark to mark and from space to space have constant levels m1 and m0, respectively. We also consider that the histograms have Nb bins. The ith bin is centered in the electrical current amplitude level ai and presents width D. Under these conditions, the value of the ith bin of the theoretical histogram of the signal in the absence of noise is given by27 8 0; for ai > m1 þ D=2 or ai < m0  D=2 > > > 1 > 2T þ s1 > 01 ðai þ D=2Þ  s10 ðai þ D=2Þ > ; for m0  D=2  ai  m0 þ D=2 > > > 4T > > < 1 1 1 1 hSIG;i ¼ s01 ðai þD=2Þs01 ðai þD=2Þþs10 ðai þD=2Þs10 ðai þD=2Þ ; for m þD=2 < a < m D=2 i 0 1 > > 4T > > > > 1 > 2T  s1 > 01 ðai þ D=2Þ þ s10 ðai þ D=2Þ > > ; for m1  D=2  ai  m1 þ D=2; > : 4T (6.4) 1 where T, s1 01 ðaÞ and s01 ðaÞ are the bit period of the signal and the inverse functions of s01(t) and s10(t), respectively. To include the influence of electrical noise in the histogram it is simply necessary to convolve hSIG,i with the histogram of the electrical noise, which in most cases one may assume to be Gaussian, to obtain an estimate of the histogram of the received signal. In the case of optical and electrical noise the method becomes more complex due to the nonlinear characteristics of the photodetector. In these cases, the histogram of the signal can be separated in two components, ( ( hSIG;i for ai < I
s hSIG;i for ai  I
s hSIG1;i ¼ ; (6.5) hSIG0;i ¼ 0 otherwise 0 otherwise

where I
s is the average current level. hSIG0,i and hSIG1,i correspond to the areas of the histogram related to the space and mark symbols, respectively, as represented in Figure 6.10. After the separation, each histogram component is convolved with Gaussian noise with the corresponding standard deviation, sEST,0 and sEST,1. The final histogram is obtained by adding the histogram components. To illustrate the application of this technique, consider a simulated signal under analysis with NRZ format and space to mark raised-cosine transitions given by s01 ðtÞ ¼ ðm1 þ m0 Þ=2  ðm1  m0 Þ=2
cosðpt=tr Þ;

0 < t < tr ;

(6.6)

where tr is the transition time. The mark to space transition is given by s10(t) ¼ m1 þ m0  s01(t). Figure 6.11 presents a comparison between the histograms of the signal under analysis and the corresponding estimated histogram in a logarithmic scale. The OSNR of the signal under analysis has been adjusted to provide a Q-factor of 8, in linear units. The parameters of the estimated

156

CHAPTER 6 OPM based on asynchronous amplitude histograms

Occurrences

104

103

102

101 –0.2

0.0

0.2

0.4

0.6 0.8 1.0 1.2 Current level (a.u.)

1.4

1.6

1.8

FIGURE 6.11 Asynchronous histogram acquired from simulated signals with □, raised-cosine pulse shape; •, rectangular pulse shape filtered by Bessel filter with bandwidth of 70% of signal’s bit rate; D, 40-Gb/s raised-cosine signal degraded by accumulated chromatic dispersion of 34 ps/nm; continuous lines, estimated histograms using the raised-cosine approximation.17 Copyright © 2009, IEEE.

histogram (m0, m1, s0, s1, and tr) have been adjusted to match the histogram of the signal under analysis using the Nelder-simplex nonlinear minimization algorithm. Figure 6.11 shows a reasonable match between the estimated histogram and the histogram of the signal under analysis. The most evident discrepancy occurs in the central area and corresponds to the point of separation between the two components of the estimated histogram. However, it does not contribute significantly to the error of the match. The resulting set of parameters for the estimated histogram can be used to evaluate the Q-factor using Equation (6.31). In this example, the estimated Q-factor takes a value of 8.2, which is fairly close to the original value of 8. Furthermore, the method allows an estimation of the transition time of 8 ps, close to the actual value of 6.25 ps. The main limitation of the presented technique is that prior knowledge of the received pulse shape is required to obtain its theoretical histogram. Nevertheless, approximations, like the raised-cosine transition presented previously, can be used to model signals with different shapes, as long as some similarity between the actual shape of the pulse and the theoretical model is maintained. To illustrate this, Figure 6.11 presents a comparison between the histogram of a simulated rectangular signal filtered by a Bessel filter with a bandwidth of 70% of the bit rate and the corresponding estimated histogram, obtained with the raised-cosine approximation. As can be noted, the approximation is quite acceptable and an estimated Q-factor of 8.2 has been obtained. Additionally, the estimated value of tr is 23 ps, which reflects the effect of the filter on the pulse transition time. Figure 6.11 also presents a comparison between the simulated histogram of a raised-cosine signal, under the same conditions as above, but impaired by an accumulated GVD of 34 ps/nm. This reduced the real Q-factor of the signal under analysis to 4.4. Even though the signal and noise powers at the receiver are the same as in the previous case, the GVD degradation was enough to prevent the fitting algorithm from finding an appropriate set of parameters. Hence, an incorrect Q-factor estimate of 2.3 was achieved in these conditions.

6.2 Monitoring techniques based on analysis of asynchronous histograms

157

6.2.2 OSNR monitoring using asynchronous histograms OSNR monitoring has long been considered an essential method to address the signal quality in optical transmission systems. The most widespread method to infer the OSNR of optical signals is based on analyzing the corresponding optical spectrum.28,29 Such techniques are based on evaluating the power spectrum of the signal under analysis, where the ASE noise may be recognized by the characteristic flat noise floor whereas the WDM channel components present peaks at the corresponding wavelengths. The OSNR is proportional to the ratio (or difference in logarithmic units) between the power of the peaks and the noise floor. However, the deployment of dense WDM systems with reduced channel spacing and high bit rates renders this approach unfeasible as the noise floor in the optical spectrum is undistinguishable from the signal. In these cases, the edges of the spectra of the WDM channels superimpose, which restricts the measurement of OSNR unless one may find a region of the spectrum where there are no channels, to identify the level of the noise floor. A significantly more complex case arises from the deployment of meshed optical network architectures using optical crossconnectors. This allows for each channel within the WDM signal to present its particular OSNR. In these cases, each channel has a given in-band noise level, which is not clearly distinguishable from the channel component. As such, the measurement of the noise floor using a region of the spectrum without WDM channels is useless to determine the OSNR of each channel. For these reasons, meshed optical networks using dense WDM systems favor the application of in-band OSNR monitoring methods, where the OSNR monitoring is carried out for each channel individually.28 In this context, the analysis of asynchronous histograms provides an effective method to evaluate the OSNR (in-band or otherwise) of a signal under analysis, regardless of other optical impairments affecting it. This method has been proposed in Reference 30 and is based on the previously discussed approaches by Rasztovits-Wiech and Andre´, and is based on comparing histograms measured from the signal under analysis with reference histograms. To understand this technique, consider again Equation (6.2) to model an asynchronous histogram. However, instead of using the theoretical histogram of the signal in the absence of noise, hSIG,i, the proposed approach uses a reference histogram, hREF,i. The reference histogram must acquire from the signal under analysis at a calibration stage, where the OSNR can be measured using appropriate measurement instrumentation. The OSNR of the signal during the calibration station will be referred to here as osnrREF in linear units or OSNRREF in logarithmic units. Equation (6.2) then becomes " # Nb X hREF;i ðan  ai Þ2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
exp  2 hEST;n ¼ : (6.7) 2sEST ðai Þ 2p
s2EST ðai Þ i¼1 Note that if osnrREF is very high, we may consider that the reference histogram is a reasonable approximation of the theoretical signal histogram in the absence of noise and includes the impact of any detrimental influence of the monitoring system, such as nonideal sampling. To further simplify this approach, one may consider that the most significant noise component of the signal under analysis is the signal-ASE beat noise. Hence, the estimated noise variance s2EST ðai Þ may be approximated by ai ; (6.8) s2EST ðai Þ  bsp
osnrEST where bsp is a constant depending on the characteristics of the monitoring system and osnrEST is the estimated OSNR. By combining Equations (6.7) and (6.8), one only needs to adjust the value of

158

CHAPTER 6 OPM based on asynchronous amplitude histograms

osnrEST until a reasonable match between the estimated and measured histograms is found. Under such conditions, one may assume that osnrEST is a fair estimation of the real OSNR of the signal under analysis, osnrSIG. Note that the validity of this approach requires that the impact of ASE noise on the signal under analysis is larger than the impact of noise components that are signal independent, particularly noise generated by the monitoring system itself quantified by its variance, s2OMS . This condition may be expressed as s2SIG ðai Þ > s2OMS :

(6.9)

When taking into consideration Equation (6.8), the condition of Equation (6.9) becomes ai > s2OMS
osnrSIG =bsp ;

(6.10)

which implies that the proposed approximation is only valid above a given power level or below a maximum measurable osnrSIG. These limits are conditioned by the variance of the electrical noise generated by the OMS, in agreement with the general conditions for OSNR measurements predicted in Reference 29. The main drawback of this method is the requirement of a noiseless reference signal. The corresponding reference asynchronous histogram (RAH) may be easily acquired in laboratory conditions but would be difficult to obtain in a field installation. Consider now the case of a reference histogram acquired from a signal with finite OSNR. The noise affecting the reference signal then becomes characterized by a signal-ASE noise beat variance given by s2REF ðai Þ ¼ bsp
ai =osnrREF . The initial value of OSNRREF may be measured at a calibration stage, using external equipments when installing the OMS in the field. Alternatively, the output of the network planning may be used to estimate this value. Subsequent histogram acquisitions will reflect the changes of OSNR of the signal under analysis with respect to OSNRREF. Assume the case where osnrSIG < osnrREF —that is, the signal under analysis presents a higher degradation than the reference signal. If this is the case, the method proposed by Equation (6.7) may be used to produce the estimated histogram by including the impact of additional noise on the reference histogram to approximate the measured histogram, as represented in Figure 6.12. In this case, the match between the measured and the estimated histograms will be found when the estimated variance of the signal-ASE beat noise reflects the difference between the corresponding noise variances of the signal under analysis and the reference signal—that is, 1 1  osnrREF : s2EST ðai Þ  s2SIG ðai Þ  s2REF ðai Þ ¼ bsp
ai
½osnrSIG

(6.11)

Finally, assuming that when a match between the measured and the estimated histogram is found, we have osnrSIG  osnrEST, yielding 1 1 s2EST ðai Þ ¼ bsp
ai
½osnrEST  osnrREF ;

(6.12)

which replaces Equation (6.8) for the implementation of the novel method. Note that if osnrREF is known from the calibration stage, the only unknown variable is osnrEST. The use of Equations (6.7) and (6.12) to produce the estimated histogram using a reference histogram with finite OSNR higher than the OSNR of the signal under analysis will be referred to here as method A. Similar to the case of a reference histogram with very high OSNR, the validity of the approximation used in method A requires that the variance of the signal-ASE noise beat affecting the signal under analysis is higher than the variance of the noise affecting the reference

6.2 Monitoring techniques based on analysis of asynchronous histograms

159

RAH presents finite OSNRREF < OSNRSIG

Reference asynchronous histogram, hREF,i Estimated histogram, hEST,i, presenting estimated OSNR, osnrEST Linear superposition of Gaussian distributions centered in ai with variance s 2EST (ai)

Fitting using nonlinear minimization method

Measured histogram, hMEA,i, presenting unknown OSNR, osnrSIG

FIGURE 6.12 Diagram of histogram estimation method A.

signal added to the variance of the OMS noise, assuming that the latter may be approximately described by a Gaussian distribution. This is represented by the condition s2SIG ðai Þ > s2REF ðai Þ þ s2OMS :

(6.13)

Note that for very high osnrREF, the contribution of the ASE noise affecting the reference signal at the OMS input may be neglected in comparison with the contribution of the OMS noise. As such, the validity condition of Equation (6.13) may be written as osnrSIG < bsp
ai =s2OMS . This implies that the range of method A depends directly on the power at the OMS input. However, when reducing osnrREF the impact of the OMS noise becomes negligible and the validity condition may be written simply as osnrSIG < osnrREF. Note that the latter is independent of the power at the OMS input, suggesting that reducing osnrREF allows an increase of the sensitivity of the OMS at the cost of reducing the maximum measurable OSNR. Consider now the case when the signal under analysis is less degraded than the reference signal—that is, osnrSIG < osnrREF. When using method A, the nonlinear minimization algorithm will diverge to an arbitrarily high value of osnrEST, corresponding to the lowest possible value of s2EST ðai Þ ¼ s2REF ðai Þ. As such, the condition of failure of method A may be identifiedPwhen the distance between the b 2 ¼ Nn¼1 jhREF;n  hEST;n j2 , converges estimated and reference histograms, defined in this work as dRE to zero. Although it is not possible to calculate an estimate of the measured histogram by including the impact of ASE noise in the reference histogram, as proposed for method A, it is possible to do just the opposite. If the signal under analysis is less degraded by ASE noise than the reference signal,

160

CHAPTER 6 OPM based on asynchronous amplitude histograms

MAH presents OSNRSIG < OSNRREF

Measured asynchronous histogram, hMEA,i, presenting unknown OSNR, osnrSIG Estimated histogram, hEST,i

Linear superposition of Gaussian distributions centered in ai with variance s 2EST (ai)

Fitting using nonlinear minimization method

Reference asynchronous histogram, hREF,i, presenting known OSNR, osnrREF

FIGURE 6.13 Diagram of histogram estimation method B.

we may calculate an estimate of the reference histogram by including the impact of ASE noise in the MAH, as represented in Figure 6.13. The amount of ASE noise that has to be included to find a match will allow an estimation of the osnrSIG. This method will be referred here as method B. In this case, Equation (6.7) becomes " # Nb X hMEA;i ðan  ai Þ2 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
exp  2 ; (6.14) hEST;n ¼ 2sEST ðai Þ 2p
s2EST ðai Þ i¼1 where hMEA,i represents the measured histogram. In addition, considering that the roles of the reference signal and the signal under analysis are swapped, Equation (6.12) will become 1 1  osnrEST : s2EST ðai Þ ¼ bsp
ai
½osnrREF

(6.15)

Equations (6.14) and (6.15) allow estimating the OSNR of the signal under analysis when it is lower than the OSNR of the reference signal. Similar to Equation (6.13), the validity of the approximation used to estimate the reference histogram is given by s2REF > s2SIG þ s2OMS :

(6.16)

Note that for high osnrSIG, the impact of the ASE noise degrading the signal under analysis is negligible in condition of Equation (6.16), and the validity condition becomes osnrREF < bsp
ai =s2OMS . This implies that the maximum osnrREF is conditioned by the power at the OMS input. Furthermore,

6.2 Monitoring techniques based on analysis of asynchronous histograms

161

it indicates that the reachable OSNR range may be increased by reducing osnrREF, similar to method A. For low osnrREF, the OMS noise is negligible and the validity condition simply becomes osnrSIG > osnrREF. The remaining problem with the proposed methods is the decision between using method A or B. Since osnrSIG is usually unknown, it may not be compared with osnrREF in order to select the appropriate method. As such, the solution is to initially use method A and test its condition of failure, defined previously. If the value of dRE converges to zero, one may assume that osnrSIG < osnrREF. As such, a new calculation will be required using method B. To validate the previously described method, consider the OMS presented in Figure 6.14. A pseudo-random binary sequence of 214  1 NRZ symbols at 40 Gb/s is used as the signal under analysis and fed into the OMS. The monitoring system will consist of an unsynchronized digital oscilloscope equipped with an optical front end with a bandwidth of 65 GHz to obtain the asynchronous amplitude histograms. In addition, the signal at the OMS input is preamplified using an EDFA with a noise figure of approximately 6 dB. An optical filter with a bandwidth of 0.6 nm is used after the preamplifier to limit the optical noise. Each histogram will be computed by acquiring 106 asynchronous samples of the signal. We begin by validating the OSNR estimation method using NRZ signals without preamplification in the OMS. The reference signal will be obtained in nearly noiseless conditions, with an OSNR of 28 dB. Figures 6.15(a) and (b) present examples of the measured histograms for OSNRSIG ¼ 14, 20, and 30 dB, as well as the reference histogram with OSNRREF ¼ 28 dB. Average powers at the input of the OMS of 0 dBm and 5 dBm have been considered. In the first case, the impact of ASE noise is clearly distinguishable on the shape of all presented histograms, particularly in the region of the histograms corresponding to the mark level. When the power at the OMS input is reduced to 5 dBm [see Figure 6.15(b)], the impact of the electrical noise, generated at the oscilloscope receiver, and the quantization noise, resulting from the limited resolution of the analog-to-digital converter (ADC) of the oscilloscope, becomes evident from the broadening of the peaks of the mark and space symbols, and the spikes of the histograms, respectively. In these conditions, the histogram obtained for OSNRSIG ¼ 30 dB is indistinguishable from the reference histogram with OSNRREF ¼ 28 dB. Figures 6.16(a) and (b) present the dependence of OSNREST on OSNRSIG, considering average powers of 0 dBm and 5 dBm at the input of the OMS, respectively, for OSNRREF ¼ 16 dB, 22 dB, and 28 dB. For a qualitative analysis, the insets of Figures 6.16(a) and (b) present eye diagrams of the signal at the oscilloscope input.

Optical monitoring system Signal under analysis 40-Gb/s NRZ

EDFA Optical filter

Digital oscilloscope

FIGURE 6.14 Schematic diagram of OMS for OSNR evaluation using asynchronous histograms.

CHAPTER 6 OPM based on asynchronous amplitude histograms

OSNRSIG = 30 dB

OSNRSIG = 14 dB OSNRSIG = 20 dB

OSNRSIG = 30 dB

Occurrences (a.u.)

OSNRREF = 28 dB

Occurrences (a.u.)

162

OSNRREF = 28 dB

OSNRSIG = 14 dB OSNRSIG = 20 dB

Electrical current (a.u.) (a)

Electrical current (a.u.) (b)

FIGURE 6.15 Asynchronous histograms of reference signal considering an OSNR of 28 dB (dashed line) and signal under analysis with OSNR values of 14, 20, and 30 dB (continuous lines) for (a) power at OMS input of 0 dBm and (b) power at OMS input of 5 dBm. Optical preamplification in OMS is not considered in this case.30

36

36

32

32

OSNREST, dB

OSNREST, dB

Copyright © 2009, IEEE.

28 24 20 16 12

28 24 20 16 12

12

16

20 24 OSNRSIG, dB (a)

28

32

12

16

20 24 OSNRSIG, dB

28

32

(b)

FIGURE 6.16 Dependence of estimated OSNR on OSNRSIG for OSNRREF ¼ 28 dB (□), OSNRREF ¼ 22 dB (D), and OSNRREF ¼ 16 dB (○). Optical amplification or filtering in OMS is not considered in this case. The insets present eye diagrams of signal under analysis for OSNR values of 16 and 30 dB. (a) Power at OMS input of 0 dBm and (b) power at OMS input of 5 dBm.30 Copyright © 2009, IEEE.

The usage of a preamplified OMS allows the acquisition of valid asynchronous histograms with an average power at the OMS input as low as 25 dBm. Figures 6.17(a) and (b) present the dependence of the estimated OSNR, OSNREST, on the actual signal OSNR, OSNRSIG, assuming that the signal used to acquire the reference histogram presents an OSNR of OSNRREF ¼ 16, 22, and 28 dB, and considering average powers of 20 and 25 dBm at the input of the OMS, respectively. In the case of OSNRREF ¼ 28 dB, the OSNR estimates follow a behavior similar to the one observed for higher powers in Figures 6.16(a) and (b). Note that in the case of an average power at the OMS

36

36

32

32

OSNREST, dB

OSNREST, dB

6.3 Acquisition and processing of amplitude histograms

28 24 20 16 12 12

16

20 24 OSNRSIG, dB

(a)

28

32

163

28 24 20 16 12 12

16

20 24 OSNRSIG, dB

28

32

(b)

FIGURE 6.17 Dependence of estimated OSNR on OSNRSIG for OSNRREF ¼ 28 dB (□), OSNRREF ¼ 22 dB (D), and OSNRREF ¼ 16 dB (○). An EDFA is used within the OMS for pre-amplification. The insets present the eye diagram of the signal under analysis for OSNR values of 16 and 30 dB. (a) Power at OMS input of 20 dBm and (b) power at the OMS input of 25 dBm.30 Copyright © 2009, IEEE.

input of 25 dBm, OSNREST is independent of OSNRSIG, for values above 24 dB. This limitation is due to the noise generated by the digital oscilloscope and also due to the noise generated by the preamplifier, which masks the impact of ASE noise on the signal under analysis. However, it occurs for an average power at the OMS input 20 dB below the corresponding case without an optical preamplifier, demonstrating the advantage of the preamplified approach. When OSNRREF is reduced to 22 and 16 dB, we verify that the maximum measurable OSNR is limited to 24 and 26 dB, respectively, for errors below 1 dB and for both considered values of power at the OMS input. Note that in all considered cases, this technique required the adjustment of a single variable, the OSNR, to find a match between the estimated and measured histograms. It has been shown that the maximum measurable OSNR decreases with the power of the signal at the input of the OMS due to the influence of the noise generated in the OMS. However, this limitation may be overcome by lowering the value of the reference OSNR. The technique based on reference histograms may be applied, in principle, to arbitrary modulation formats.30 In the same reference, its application is demonstrated for 40-Gb/s NRZ signals with and without GVD-induced degradation, and also for 40-Gb/s RZ signals. In the latter case, the RZ signal is converted to a NRZ signal within the OMS by narrowband optical filtering.

6.3 GENERAL CONCEPTS ON THE ACQUISITION AND PROCESSING OF AMPLITUDE HISTOGRAMS 6.3.1 Sampling noise

Optical or electrical, most real-life signals have a continuous and analog nature. In order to process these signals using digital circuitry, analog-to-digital conversion is required. However, a digital signal is discrete in amplitude and time. To obtain an amplitude- and time-discrete signal from an

164

CHAPTER 6 OPM based on asynchronous amplitude histograms

amplitude- and time-continuous signal a sampling process is required. The fundamental principal of sampling is the repeated quasi-instantaneous capture of a time-varying waveform by a sampling gate.31 A train of narrow impulses controls the sampling gate. After a sample of the continuoussensor signal is obtained, an ADC circuit can be used to obtain a corresponding digital value. The existing ADC circuits require a constant input for a certain amount of time in order to obtain a correct digital output. Therefore, a sample obtained with a narrow impulse is usually not enough for the ADC to function properly. In order to retain the impulse amplitude an analog memory is used between the sampling circuit and the ADC. The simplest analog memory is a capacitor, known as a hold capacitor. The charge accumulated when the gate is closed is proportional to the amplitude of the sample and can be used for analog-to-digital conversion. The combination of the sampling circuit with the hold capacitor is known as a sample-and-hold (S&H) circuit. An example of a simplified S&H block diagram is presented in Figure 6.18. When the sampling gate is closed, the input sensor signal current, iin(t), charges the hold capacitor. The time required for the sampling gate to open is known as the aperture time. After the gate is open, the capacitor discharges with a given fall-off rate, depending on the input impedance of the following circuitry. In most cases, a high input impedance amplifier is used to decrease the falloff rate. The time required for the hold capacitor to discharge should be sufficient for the ADC to produce a valid digital value for the acquired sample. Some of the main limitations of S&H systems are as follows: 1. Finite aperture time. The S&H takes a period of time to capture a sample of the sensor signal. Since the signal will vary during this time, the sampled signal can be slightly off. 2. Sampling jitter. Temporal variations of the sampling frequency lead to uncertainty of the sampling instant. This problem is particularly hazardous in synchronous sampling systems. 3. Signal feedthrough. Due to construction limitations of the sampling gate, some current “bleeds” through the switch to the capacitor causing the voltage to change slightly. 4. Signal droop. The voltage being held on the capacitor starts to slowly decrease over time due to the limited input resistance of the following circuitry. 5. Aperture uncertainty. Variations from sample to sample of the aperture time lead to additional sampling noise. For the characterization of sampling systems for the acquisition of asynchronous histograms, it may be considered that the input impedance of the ADC is high enough to assume a constant voltage at

Sampling gate

iin(t)

R

C Impulse generator

FIGURE 6.18 Simplified block diagram of S&H circuit.

6.3 Acquisition and processing of amplitude histograms

165

the hold capacitor terminals. Additionally, currently existing fast ADCs allow the reduction of feedthrough. As so, limitations 3 and 4 will be neglected. Temporal variations of the aperture time (limitation 5) will also not be considered because the sampler characteristics are assumed perfect. The sampling jitter (limitation 2) leads to significant error in systems where the sampling frequency is related with the frequency of the sensor signal. These systems include all digital receivers, oscilloscopes, and time-domain reflectometry devices. However, when considering asynchronous sampling, the sampling frequency is not related to the frequency of the signal under analysis. Therefore, sampling jitter does not affect significantly the overall accuracy of the system. The aperture time (limitation 1) is the time required for the sampling gate to close after a sample is acquired. In an ideal situation, the aperture time is zero and the output signal of the S&H is proportional to the signal value immediately before the sampling instant. However, in a nonideal situation, the sampling gate requires a finite time to switch. During this time, the input signal of the S&H is changed, leading to sampling error. In order to evaluate the effects of finite aperture time in asynchronous sampling, one can assume the general model for an S&H system presented in Figure 6.19. The pulse generator creates a pulse train at the sampling frequency Fa with a pulse shape given by c(t). It is considered that c(t) is norR malized in a way that c(t) ¼ 1. The pulse train is multiplied by the signal under analysis, x(t), and the result is inserted in an integrator. This integrator is such that its output is a constant signal in each sampling period. The output value of the integrator for the kth sampling interval, xk, is given by þ1 ð

xk ¼

xðtÞ
cðt  k
Ta Þ
dt;

(6.17)

1

where Ta ¼ 1/Fa is the sampling period. In an ideal sampling system, the time required by the sampling gate to open is infinitesimal, with the sampling impulse given by c(t) ¼ d(t), where d(t) is the Dirac function. In such cases, we obtain xk ¼ x(k
Ta) and the sampling error is null. In a realistic case the sampling pulse is not infinitesimal and sampling error will occur. To address this error without loosing generality, we may consider the Taylor series expansion of x(t) around the sampling instant xðtÞ ¼

þ1 X ðt  k
Ta Þm m¼0

x(t )

m!


xðmÞ ðk
Ta Þ;

∫ . dt +∞

Σ y (t − k . Ta)

k=−∞

FIGURE 6.19 General block diagram of S&H system.17

Copyright © 2009, IEEE.

xk

(6.18)

166

CHAPTER 6 OPM based on asynchronous amplitude histograms

where x(m)(t) is the mth order derivate of x(t). By replacing Equation (6.18) in Equation (6.17) we have xk ¼ xðk
Ta Þ þ ek ;

(6.19)

where ek is the sampling error given by ek ¼

1 ðmÞ X x ðk
Ta Þ m¼1

m!


cm ;

(6.20)

and the term cm is given by þ1 ð

cm ¼

tm
cðtÞ
dt:

(6.21)

1

Equation (6.19) shows that the sampling error depends mainly on the derivates of the sensor signal and the shape of the sampling function. Furthermore, if c(t) is an even function, the contribution of the even-order derivates of x(t) to the error is null.32 The overall sampling error can be considered as a form of noise and measured by its variance, s2SN ; " #2 Ns 1 X 1 X cm ðmÞ 2 sSN ¼

x ðk
Ta Þ ; (6.22) Ns k¼1 m¼1 m! where Ns is the number of acquired samples. Assuming that the main contribution for the sampling error results from the first-order derivate and neglecting the contributions of the higher-order derivates, s2SN can be approximated by s2SN  c21
s2x;1 ;

(6.23)

where s2x;m is the variance of the samples acquired from the mth order derivate of x(t), defined as s2x;m ¼

Ns 1 X
½xðmÞ ðk
Ta Þ2 : Ns k¼1

(6.24)

In the particular case that even sampling impulses are used, the odd-order derivates of the sensor signal do not contribute to the error signal. As an approximation, only the contribution of the second-order derivate will be considered. The contribution of the higher-order derivates will be neglected. As so, s2SN can be approximated by s2SN 

c22 2
sx;2 : 4

(6.25)

Equations (6.23) and (6.25) show that the sampling error depends essentially on the sampling impulse shape, and the variance of the first- or second-order derivates of the signal under analysis in the sampling instant.

6.3 Acquisition and processing of amplitude histograms

167

The impact of the sampling noise may be characterized by a signal-to-sampling-noise ratio (SSNR), defined by   Ps (6.26) SSNR ¼ 10
log10 2 ; sSN where Ps is the signal power. As an application example, Figures 6.20(a) and (b) present the SSNR calculated for an S&H with rectangular and triangular sampling pulses, respectively. In both cases, an aperture time ap is considered. It has been assumed that the signal under analysis is a binary NRZ signal with a raised-cosine transition function with transition time tr and symbol period T. Under these conditions, it may be shown that the sampling noise variance is approximated by s2SN 

a4p p4
3 9216
T tr

(6.27)

when a rectangular sampling pulse is used. Similarly, when using a triangular sampling impulse, the sampling noise variance may be approximated by s2SN 

a2p p2
: 144
T tr

(6.28)

Figure 6.20(a) or (b) shows that an increase of the transition time leads to a significant decrease of the SSNR. This decrease is more evident for low values of the aperture time. Comparing Figure 6.20(a) with Figure 6.20(b), the advantage of using a rectangular sampling impulse to a triangular sampling impulse becomes evident. The latter case presents a significant degradation of the SSNR due to the contribution of the sensor signal first-order derivate.

90

90 Sampling pulse ap

80

70

tr/T = 0.4

60

SSNR (dB)

SSNR (dB)

70

50 40 tr/T = 0.1

30

60 50

20 10 0.1 ap /T

(a)

1.0

tr/T = 0.4

30

10 0

tr/T = 0.1

40

20

0.0

Sampling pulse ap

80

0 0.0

0.1 ap/T

1.0

(b)

FIGURE 6.20 Signal-to-sampling-noise ratio as function of aperture time for different values of transition time. (a) Using rectangular sampling impulse. (b) Using triangular sampling impulse. Numerical simulation results, symbols; analytical results, continuous line.

168

CHAPTER 6 OPM based on asynchronous amplitude histograms

Figure 6.20 also presents numerical simulation results to validate Equations (6.27) and (6.28). Note that the validity of this approach is limited to the validity of the Taylor series expansion of the signal under analysis around the sampling point. Since this signal is generally not an analytic function in the case of the considered optical systems, it may not agree with its Taylor expansion in all points. For digital signals, the critical points are those where the derivates of the signal are not continuous in the neighborhood of the symbol transitions. This justifies the decay of this approach for high sampling aperture times, as shown in Figure 6.20.

6.3.2 Averaging effects The computation of the sampling noise as a means to characterize an S&H system is effective when the signal under analysis is noiseless and one needs to address its shape. This is particularly significant for measurement instrumentation, such as digital oscilloscopes. However, in monitoring systems based on sampling the signal under analysis, such as the case of asynchronous histograms, it becomes important to address not only the signal but also the noise degrading this signal. As such, more effective methods to characterize the sampling system can be devised, in particular, methods focusing on analyzing the impact of the monitoring system on the signal under analysis.17 Consider the simplified block diagram for the equivalent S&H system, presented in Figure 6.21. In this model, the nonideal sampling system presented in Figure 6.19 is replaced by an ideal sampling system preceded by an equivalent sampling filter with an impulse response heq(t). In this system, the kth sample is given by ð þ1 xðtÞ
heq ðk
Ta  tÞ
dt: (6.29) xk ¼ xeq ðk
Ta Þ ¼ 1

By comparing Equations (6.29) and (6.17), it can be shown that the two systems become equivalent by having heq(t) ¼ c(t). However, the system presented in Figure 6.21 clearly indicates that the performance quality measurements performed with the acquired samples evaluate xeq(t) instead of x(t). As such, this form of modeling allows treating the impact of nonideal sampling as an additional filter applied to the signal under analysis. Note that in reality such a filter does not exist, and for this reason one usually refers this phenomenon as averaging effect. To assess the impact of the sampling process on the evaluation of signals degraded by noise, consider the signal under analysis given by x(t) ¼ s(t) þ n(t), where s(t) and n(t) are signal and noise

x(t)

heq (t)

xeq (t )

ò . dt

xk

+⬁

Σ d (t − k . Ta ) k=1 FIGURE 6.21 General block diagram of equivalent S&H system.17

Copyright © 2009, IEEE.

6.3 Acquisition and processing of amplitude histograms

169

components, respectively. The ratio between the variance of the noise after the equivalent sampling filter, s2n;eq , and the variance of the monitored noise, s2n , is given by Ð þ1 2 s2n;eq 1 Sn ð f Þ
jHeq ð f Þj
df ; (6.30) ¼ Ð þ1 s2n 1 Sn ð f Þ
df where Sn(f) is the power spectral density of the noise affecting the signal under analysis and Heq ð f Þ ¼ F½heq ðtÞ is the transfer function of the equivalent sampling filter, with F½
 as the Fourier transform operator. Applying the Schwarz’s inequality to Equation (6.30) it can be shown that s2n;eq  s2n . Therefore, if the distortion of the signal component is considered negligible, the sampling process may influence quality estimates by reducing the evaluated noise power. To illustrate the impact of averaging effects on optical monitoring systems based on the analysis of asynchronous histograms, consider the block diagram presented in Figure 6.22. The signal under analysis is initially filtered through an optical filter with a low-pass equivalent transfer function Ho( f ). After that, the signal is detected by a p-i-n photodetector with unitary responsivity. An electrical filter, with transfer function given by Hr( f ), models the frequency limitations of the receiver electronics. Finally, the filtered electrical current is sampled for histogram counting. The sampling system is assumed to have a structure equivalent to the system presented in Figure 6.21. The optical and electrical filter bandwidths are assumed to be large enough to neglect signal distortion due to filtering. In these conditions, the Q-factor, in linear units, can be approximated by33 m  m0 q¼ 1 ; (6.31) s1 þ s0 where mm and sm are the values of the mean and standard deviation of the electrical current for the symbol m, with m ¼ 0 and 1 for the space and mark symbols, respectively. If the signal-spontaneous emission beat noise is considered dominant, the standard deviation of the electrical current for the symbol m can be approximated by34 ð þ1 s2m ¼ ksp ðPm Þ jHr ð f Þj2
jHo ð f Þj2
df ; (6.32) 1

where ksp(Pm) is a term that depends on the noise characteristics and the optical power at the receiver input for the symbol m, Pm. Due to the nonideal sampling process, the variance of the measured noise is given by ð þ1 2 sm;eq ¼ ksp ðPm Þ jHeq ð f Þj2
jHr ð f Þj2
jHo ð f Þj2
df : (6.33) 1

Signal under analysis 40-Gb/s NRZ

Nonideal S&H

ADC

FIGURE 6.22 Simulated nonideal histogram acquisition system for optical monitoring.

Histogram counter

170

CHAPTER 6 OPM based on asynchronous amplitude histograms

Therefore, the measured Q-factor for the equivalent system becomes qeq ¼

m1  m0 q ¼ ; s1;eq þ s0;eq k

where the correction factor, k, is given by Ð þ1 jHeq ð f Þj2
jHr ð f Þj2
jHo ð f Þj2
df 2 : k ¼ 1 Ð þ1 2 2 1 jHr ð f Þj
jHo ð f Þj
df

(6.34)

(6.35)

Equation (6.34) shows that if the aperture time is short enough to neglect distortion of the signal under analysis (i.e., when the SSNR is high), one may consider that the impact of the nonideal sampling on quality measurements may be reduced to a proportional adjustment.17,35 The correction factor allows a simple and direct adjustment of the measured Q-factor in conditions where noise averaging due to the nonideal sampling process is dominant. This factor is only dependent on the physical characteristics of the measurement system and can be obtained by calibration of the device. To illustrate the application of the correction factor, consider the optical monitoring system based on the analysis of asynchronous histograms presented in Figure 6.22. In this system, the signal under analysis is simulated as a 40-Gb/s NRZ deBruijn sequence of 220 symbols. Finite sampling aperture time was considered by simulating rectangular sampling impulses. The OSNR of the signal under analysis is adjusted to provide simulated signals with Q-factors of q ¼ 8 and 10, in linear units. The OMS computes the Q-factor, qeq, without correction due to nonideal sampling using the asynchronous histogram analysis technique developed by Andre´ et al.,27 which is described in detail in Section 6.2.1. In addition, the corrected values of the Q-factor, qcor, have been calculated using the correction factor for this OMS. The dependence of qeq/q and qcor/q on the aperture time is presented in Figure 6.23, where each value was obtained by averaging the results from 16 simulations. In order to have a better understanding of Figure 6.23, it may be divided in three different regions. For sampling apertures below 20 ps, the averaging effect of nonideal sampling influences mainly the noise component. Therefore, qeq increases steadily with the aperture time due to the noise averaging. In this area, the correction factor can be applied, providing accurate values of the Q-factor. From sampling apertures of 2035 ps near the symbol period, the signal component also becomes significantly affected by the nonideal sampling, and the model used to obtain the estimated histogram becomes inaccurate. The distortion induced by the nonideal sampling in the signal component leads to new relative maximums in the simulated histogram, which are interpreted as noise by the fitting algorithm. To illustrate this effect, Figure 6.24 presents the asynchronous histograms obtained with an aperture time of 28 ps for a Q-factor of 8. The new relative maximums lead to an overestimation of the noise component, reducing the growth of the estimated value of qeq. As the correction factor continues to increase steadily with the aperture time, the values of qcor start to decrease. However, a margin of 25% of error is maintained. For aperture times above 35 ps the averaging of the signal and noise components due to nonideal sampling has degraded the shape of the histogram to a point where the histogram analysis technique fails completely. To illustrate this effect, Figure 6.24 also presents the asynchronous histograms obtained with an aperture time of 44 ps for a Q-factor of 8. In these conditions, the Q-factor estimates performed over the simulated histograms lose physical meaning.

6.3 Acquisition and processing of amplitude histograms

171

3.5 3.0 qeq /q

2.5 2.0

qcor /q

1.5 1.0 0.5 0.0 0

5

10

15

20

25

30

35

40

Sampling aperture time (ps)

FIGURE 6.23 qeq /q and qcor /q as function of aperture time for 40-Gb/s signal. •, qeq /q for q ¼ 8; ○, qeq /q for q ¼ 10; n, qcor / q for q ¼ 8; □, qcor /q for q ¼ 10.17 Copyright © 2009, IEEE.

Normalized histogram

104

ap = 28 ps

103

ap = 44 ps 102 –0.2

0.0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

Current level (mA)

FIGURE 6.24 Asynchronous histogram of a 40-Gb/s signal obtained with a nonideal sampling system from simulated signals with □, aperture time of 28 ps; •, aperture time of 44 ps. Continuous lines represent estimated histogram. Arrows indicate new relative maximums that result from averaging effect induced by nonideal sampling.17 Copyright © 2009, IEEE.

172

CHAPTER 6 OPM based on asynchronous amplitude histograms

6.4 SUMMARY This work has shown that by using simple analysis techniques, such as the average Q-factor, one may employ asynchronous histograms to derive information that would otherwise require a complete optical receiver as well as a BER counter to estimate. More complex techniques, such as the decomposition of asynchronous histograms, allow handling signals that suffer from distortion to the point where the original signal format is unrecognizable. In addition, it has been shown that histogram comparison techniques either using theoretical histograms or reference histograms can be used to extract information on the signal under analysis. Some practical aspects on the design of asynchronous sampling systems for the acquisition of asynchronous histograms have been considered. In particular, the most demanding aspect has been shown to be the sampling aperture time. It has been shown that the aperture time of the sampling system conditions the performed measurement by restricting the noise or distorting the signal under analysis. As a consequence, compensation methods have to be used either by adjusting the performed measurement or by using reference histograms. In summary, the analysis of asynchronous histograms has survived the evolution of optical communications systems as an effective proposal for advanced optical performance monitoring thanks to its remarkable flexibility. Asynchronous histogram acquisition systems can be easily designed for very wide ranges of bit rates and arbitrary signal formats. An extensive set of simple and robust methods to analyze asynchronous histograms has been developed over the years, and application to novel modulation formats can be carried out with relatively little effort.

REFERENCES 1. Kilper D, Bach R, Blumenthal D, Einstein D, Landolsi T, Ostar L, et al. Optical performance monitoring. IEEE/OSA J Lightwave Technol 2004;22(1):294–304. 2. Hanik N, Gladisch A, Caspar C, Strebel B. Application of amplitude histograms to monitor performance of optical channels. IEE Electron Lett 1999;35(5):403–4. 3. Shake I, Takara H, Kawanishi S, Yamabayashi Y. Optical signal quality monitoring method based on optical sampling. IEE Electron Lett 1998;32(22):2152–4. 4. Shake I, Takara H, Uchiyama K, Yamabayashi Y. Quality monitoring of optical signals influenced by chromatic dispersion in a transmission fiber using averaged Q-factor evaluation. IEEE Photon Technol Lett 2001;13(4):385–7. 5. Shake I, Takara H. Averaged Q-factor method using amplitude histogram evaluation for transparent monitoring of optical signal-to-noise ratio degradation in optical transmission system. IEEE/OSA J Lightwave Technol 2002;20(8):1367–73. 6. Shake I, Otani E, Takara H, Uchiyama K, Yamabayashi Y, Morioka T. Bit rate flexible quality monitoring of 10 to 160 Gbit/s optical signals based on optical sampling technique. IEE Electron Lett 2000;36(25): 2087–8. 7. Shake I, Takara H, Kawanishi S. Simple Q factor monitoring for BER estimation using opened eye diagrams captured by high-speed asynchronous electroopical sampling. IEEE Photon Technol Lett 2003;15(4):620–2. 8. Shake I, Takara H. Chromatic dispersion dependence of asynchronous amplitude histogram evaluation of NRZ signal. IEEE/OSA J Lightwave Technol 2003;21(10):2154–61.

References

173

9. Shake I, Takara H, Morioka T. Determination of the origin of BER degradation utilizing asynchronous amplitude histograms. In: Proc. Pacific Rim conference on lasers and electro-optics, 2001 (CLEO/Pacific 2001), Chiba, Japan, vol. 2; p. II-560–II-561. 10. Rasztovits-Wiech M, Studer K, Leeb W. Bit error probability estimation algorithm for signal supervision in all-optical networks. IEE Electron Lett 1999;35(20):1754–5. 11. Weinert C, Schmidt C, Weber H. Application of asynchronous amplitude histograms for performance monitoring of RZ signals. In: Proc. optical fiber communication conference 2001 (OFC 2001), Anaheim, California, USA, vol. 3; 2001. p. WDD41-1–WDD41-3. 12. Andre´ P, Pinto J, Teixeira A, Rocha J, Almeida T, Pousa M. Optical-signal-quality monitor for bit-errorratio assessment in transparent DWDM networks based on asynchronously sampled amplitude histograms. OSA J Opt Network 2002;1(3):118–28. 13. Andre´ P, Pinto J. Bit error rate assessment in DWDM transparent networks using optical performance monitor based in asynchronous sampling. In: Proc. optical fiber communication conference 2002 (OFC 2002); Anaheim, California, USA, 2002. p. 749–50. 14. Teixeira A, Andre´ P, Lima M, da Rocha J, Pinto J. Characterization of high bit rate optical signals by low rate asynchronous sampling. In: Proc. lasers and electro-optics society 2002 (LEOS 2002), Glasgow, Scotland, UK, vol. 2; 2002. p. 625–6. 15. Ye D, Zhong W. Improved BER monitoring based on amplitude histogram and multi-Gaussian curve fitting. OSA J Opt Network 2007;6(6):584–98. 16. Ding L, Zhong W, Lu C, Wang Y. New bit-error-rate monitoring technique based on histograms and curve fitting. OSA Opt Expr 2004;12(11):2507–11. 17. Luı´s R, Andre´ P, Teixeira A, Monteiro P. Performance monitoring in optical networks using asynchronously acquired samples with nonideal sampling systems and intersymbol interference. IEEE/OSA J Lightwave Technol 2004;22(11):2452–9. 18. Chen H, Poon A, Cao X. Amplitude histogram-based method for inline pulse rise time monitoring without eye-diagram. In: Proc. conference on lasers and electro-optics 2003 (CLEO 2003), Washington, D.C., USA, 2003. p. 1511–3. 19. Chen H, Poon A, Cao X. Transparent monitoring of rise time using asynchronous amplitude histograms in optical transmission systems. IEEE/OSA J Lightwave Technol 2004;22(7):1661–7. 20. Weinert C, Caspar C, Konitzer M, Rohde M. Histogram method for identification and evaluation of crosstalk. In: Proc. optical fiber communication conference 2000 (OFC 2000), Baltimore, Maryland, USA, vol. 3; 2000. p. 56–8. 21. Weinert C, Caspar C, Konitzer M, Rohde M. Histogram method for identification and evaluation of crosstalk. IEE Electron Lett 2000;36(6):558–9. 22. Benlachtar Y, Killey R, Bayvel P. Identification of sources of degradation in optical channels using deconvolution technique. In: Proc. optical fiber communications conference 2003 (OFC 2003), Atlanta, Georgia, USA, vol. 1; March 2003. p. 109–10. 23. Weinert C. Gaussian deconvolution method for identification of impairments in optical signal transmission. OSA J Opt Network 2004;3(6):361–87. 24. Luı´s R, Teixeira A, Andre´ P, Monteiro P. Evaluation of intra-band crosstalk using asynchronous histograms. In: Proc. European conference on networks and optical communications 2004 (NOC 2004), Eindhoven, The Netherlands, 2004. p. 374–81. 25. Noirie L, Ce´rou F, Moustakides G, Audouin O, Peloso P. New transparent optical monitoring of the eye and BER using asynchronous under-sampling of the signal. In: Proc. European conference on optical communication 2002 (ECOC 2002), Copenhagen, Denmark, vol. 5; 2002. p. 1–2.

174

CHAPTER 6 OPM based on asynchronous amplitude histograms

26. Mobilon E, Barros M, Lopes A. Experimental verification of an eye diagram reconstruction technique based on asynchronous undersampling. In: Proc. international conference on microwave and optoelectronics 2005; Brası´lia, Brazil, 2005. p. 603–6. 27. Andre´ P. Optoelectronic components for high bitrate photonic networks. Ph.D. Thesis, Department of Electronics and Telecommunications Engineering, University of Aveiro, Aveiro, Portugal, 2002. 28. Kilper D, Weingartner W. Monitoring optical network performance degradation due to amplifier noise. IEEE/OSA J Lightwave Technol 2003;21(5):1171–8. 29. Yang W. Sensitivity issues of optical performance monitoring. IEEE Photon Technol Lett 2002;14(1): 107–9. 30. Luı´s R, Teixeira A, Monteiro P. Optical signal-to-noise ratio estimation using reference asynchronous histograms. IEEE/OSA J Lightwave Technol 2009;27(6):731–43. 31. Kahrs M. 50 Years of RF and microwave sampling. IEEE Trans Microw Theory Tech 2003;51(6): 1787–805. 32. Sauerwald M. Effects of aperture time and jitter in a sampled data system. In: National semiconductor application note AD-03. Available at: http://www.national.com. 1994. 33. Bergano N, Kerfoot F, Davidson C. Margin measurements in optical amplifier systems. IEEE Photon Technol Lett 1993;5(3):304–6. 34. Rebola J, Cartaxo A. Power penalty assessment in optically preamplified receivers with arbitrary optical filtering and signal-dependent noise dominance. IEEE/OSA J Lightwave Technol 2002;20(3):401–8. 35. Li Z, Li G. Chromatic dispersion and polarization-mode dispersion monitoring for RZ-DPSK signals based on asynchronous amplitude-histogram evaluation. IEEE/OSA J Lightwave Technol 2006;24(7):2859–66.

CHAPTER

Optical performance monitoring based on asynchronous delay-tap sampling

7

Trevor B. Anderson*,{, Sarah D. Dods*, Adam Kowalczyk*,{, Ken Clarke*, Don Hewitt{, and Jonathan C. Li* *

Monitoring Division Inc., Melbourne, Australia National ICT Australia Ltd. (NICTA), Victoria Research Laboratory, Australia

{

7.1 INTRODUCTION Reconfigurable and self-managed optical networks offer the potential for significant operational savings through automated fault diagnosis and management, path provisioning, and optimization.1,2 However, realizing this potential requires real-time monitoring of optical impairments and their distribution within the network. In addition to OSNR, these impairments include PMD, four-wave mixing, chromatic dispersion, reflections, and both baseband and in-band crosstalk. Future networks will also potentially carry a wide variety of formats and data rates. To accommodate these requirements, optical performance monitors will need to be ubiquitously distributed throughout the network, compatible with multiple formats and capable of monitoring multiple simultaneous impairments. We have recently introduced a new monitoring technique known as asynchronous delay-tap sampling that promises to satisfy the above requirements. With this technique, multiple impairment measurements as well as signal quality are extracted from a two-dimensional histogram of the signal.3,4 This histogram, known as a phase portrait, provides the information richness of an eye diagram without the requirement of clock extraction. The phase portraits contain unique impairment signatures that can be discovered using statistical pattern recognition techniques. With this approach we are able to not just classify, but to quantitatively monitor simultaneous combinations of impairments. A key advantage of the technique is that a simple direct detection receiver can be used to monitor a variety of bit rates and modulation formats without the need for demodulation of the signal or modification of the receiver bandwidth. The monitoring of a new impairment or format is enabled by downloading the relevant pattern recognition algorithm to the monitor. Many alternative techniques have been proposed to measure different subsets of impairments. These techniques can be broadly classed as either spectral5 or sampling based. The former include the use of RF tones and measurement of RF clock power. These techniques are, however, format dependent. The laboratory eye diagram is the most familiar sampling technique for measuring signal quality and estimating the underlying causes,6 but requires tunable clock recovery and can be difficult to extract for strongly distorted signals. The asynchronous histogram technique has been proposed as an alternative sampling technique that does not require clock recovery,7,8 but while sensitive to © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00007-9

175

176

CHAPTER 7 OPM based on asynchronous delay-tap sampling

multiple impairments, unique identification of a particular impairment in the presence of other impairments remains a challenge. Monitoring of digital coherent systems using the information available in the digital equalizer at the receiver has recently been demonstrated for CD, PMD and OSNR, and PDL.9,10 While these techniques provide a cost-effective solution at the receiver, the real-time sampling requirements mean that they cannot be readily used at intermediate points in the network. In contrast, the potential for a distributed solution based on a simple direct detection receiver is a key attribute of the asynchronous delay-tap technique. Early simulations of the asynchronous delay-tap technique for 10-Gb/s NRZ showed the potential for simultaneous monitoring of OSNR, CD, DGD, and crosstalk, as well the ability to measure signal quality parameters Q and timing jitter.3 The first experimental results using pattern recognition (with a training set of 300 cases) were demonstrated for simultaneous combinations of OSNR, CD, and DGD.4 An important limitation of these earlier works was that the signal polarization was such that first-order PMD had a fixed and equal power split between principal states. In recent work,11 the polarization restriction has been removed, with the first demonstration of CD and first-order PMD with random power splits on a commercial 10-Gb/s NRZ WDM test channel. This work used an automated network emulator that enabled larger training sets (900 cases) to be created. More recently the technique has been demonstrated in simulations and experiments for a variety of formats including 40-Gb/s NRZ-DPSK, 10-Gb/s RZ-DPSK, and 20-Gb/s RZ-DQPSK.12,13 We note that while the ability to monitor phase-modulated formats with a simple direct detection receiver is a key benefit of the technique, the method is also compatible with coherent-based receivers.14 In this chapter, we review the technique beginning with a discussion and interpretation of the phase portrait in Section 7.2, and a brief overview of pattern recognition approaches. Experimental demonstrations of OSNR, CD, and DGD monitoring for NRZ and DPSK formats are described in Section 7.3, where we also introduce a measure of DGD that takes into account arbitrary power splits between principal polarization states. Early proof of principle demonstrations, based on single transponder and laboratory-generated impairments, is further validated at 10-Gb/s with WDM system measurements and training over combined transponders. To make the technique more robust, the training set can be extended to include other effects that may potentially induce errors, such as alternative sources of noise, filter drift, or delay variations. In Section 7.4, the ability of the technique to accommodate optical filter-induced distortions and crosstalk is demonstrated for 10-Gb/s NRZ. Finally, the format independence of the technique is demonstrated with simulations of CD and DGD monitoring for 40-Gb/s RZ-DQPSK using the same monitoring hardware employed in 10-Gb/s NRZ. A summary of the techniques covered in this chapter is shown in Table 7.1.

7.2 TECHNIQUE The two key components to the technique that we describe in this section are the construction of the phase portraits followed by the application of statistical pattern recognition techniques to extract impairment signatures.

7.2 Technique

177

Table 7.1 Summary of Asynchronous delay-tap sampling Techniques Technique

Formats Demonstrated

Impairments Demonstrated (range)

Asynchronous sampling AI—support vector machine3,4,11,23,24

10G NRZ 40G NRZ-DPSK 40G RZ-DQPSK* 80G PMDQPSK*

OSNR (10, 30) dB CD (1400, 1400) ps/nm PMD (0, 60) ps In-band crosstalk* (15–25) dB Filter offset* (12,12) GHz OSNR (8.7, 35) dB CD (600, 600) ps/nm PMD OSNR (10, 30) dB CD (0, 800) ps/nm CD* (0,400) ps/nm

Asynchronous sampling—Hough transformation8,12,13 Asynchronous sampling—Homodyne detection14 Asynchronous sampling—Hausdorff measure17 Asynchronous sampling AI—artificial neural networks18

10G 20G 40G 10G

RZ-DPSK RZ-DQPSK* NRZ-DPSK NRZ-DPSK

40G NRZ-DPSK* 10G NRZ*

OSNR* (15, 30) dB CD* (0, 55) ps/nm PMD* (0, 10) ps

*Indicates simulation results.

7.2.1 Phase portrait Asynchronous delay-tap sampling is an alternative to the eye diagram that uses the joint probability density function (pdf) of a signal x(t), and its delayed version x(t þ Dt) to characterize the signal.1 This pdf, known as a phase portrait, is sensitive to waveform distortion and noise and contains unique signatures of impairments. To generate the phase portrait, the waveform is sampled in pairs separated by a known delay Dt, as shown for the NRZ signal in Figure 7.1. The phase portrait is then created by binning the sample pairs into a two-dimensional histogram, as shown in Figure 7.2(a) and (b) for 1-bit and ¼-bit delays, respectively. We emphasize that the sampling is asynchronous, in that the time between the pairs, Ts, is not related to the monitored signal bit rate, and can be many orders of magnitude longer with thousands of bits between sample pairs. A fundamental difference between the eye diagram and phase portrait is that the latter contains information about the

1

Unbeknown to the authors in the early development of this monitoring technique, delay-tap sampling is used in the field of nonlinear time series analysis16 where it is referred to as the method of “time-delay embedding.” It is also where the term phase portrait originates and in this context the word “phase” has no relation to optical phase. (The space of delayed variables can be shown through Taken’s theorem to be isomorphically related to the phase space of the underlying system parameters). The technique has been applied to problems as diverse as the analysis of electrocardiograms. The optimal choice of delay is the subject of numerous papers; a common heuristic is to use a delay equal to the correlation time of the signal.

178

CHAPTER 7 OPM based on asynchronous delay-tap sampling

Asynchronous clock From monitor port

ADC x Δt

ADC y

FIGURE 7.1 Schematic of asynchronous delay-tap sampling technique. Sample pairs are separated by a fixed delay, Dt.

(1,1)

x

(1,0)

Δt = ¼T

y

Δt = T

y

(0,1)

101 and 010 transitions (0,0)

(a)

x

(b)

FIGURE 7.2 NRZ phase portraits for (a) 1-bit delay, and (b) ¼-bit delay.

correlation between samples on the time scale of a bit period. This information is absent in eye diagrams that are constructed from samples that are separated by long periods. To help interpret the 1-bit delay-phase portrait, we begin by noting that the corners of the phase portrait represent sample pairs of well-defined marks and spaces. The lines joining these points originate from samples taken during the bit transitions between these states. For example, the diagonal line represents the 010 and 101 transition. Because the sampling is asynchronous, the waveform is sampled uniformly in time and therefore the density of points in these lines is inversely proportional to the slope of the waveform. For an idealized rectangular waveform, there would only be the four corner points; in contrast, a triangular waveform would give a uniform distribution of points along the lines. As we will show in the following, waveform distortions in the time-domain manifest as variations in both the curvature and density of these lines. In Figure 7.3 we show the effects of changing the tap delay. For small offsets from the 1-bit delay, we see that the diagonal line splits into separate curves for the 101 and 010 transitions. The splitting is a useful means of comparing the alignment of the bit rate and delay. In another application, the increased splitting or distortion in the phase portrait has been proposed as a method for aligning the pulse carver and data modulator in RZ-OOK modulators.15 We note that while the choice of 1-bit delay Dt is a useful tool for visualization, it is not critical for pattern recognition. (This enables a fixed delay of 25 ps, for example, to be used for both 10- and

7.2 Technique

Δt = T(1 − 1/16)

Δt = T

179

Δt = T(1 − 1/16)

FIGURE 7.3 Phase portraits of an NRZ signal showing the effects of small changes in tap delay.

40-Gb/s signals.) For short time delays, such as ¼ T, we can interpret the phase portrait as approximating the joint probability density of the amplitude and slope of the waveform with respect to time. The width of this phase portrait, as measured along its minor axis, can be interpreted as an approximate measure of the maximum slope of the waveform. The curves above and below the line y ¼ x represent the positive and negative slopes of the waveform, respectively. An asymmetry about this line therefore represents an asymmetry in rising and falling transients. In addition to impairment monitoring, the ¼-bit phase portraits can also be used, in an analogous manner to the eye diagram, to measure signal quality. For example, the standard signal quality measure Q can be approximated from the distribution of points along the lines y ¼ x corresponding to sample pairs of zero slope. In a more sophisticated approach, we could apply pattern recognition techniques to compensate for chromatic dispersion and monitor Q midspan as it would be measured at the receiver postcompensation. To help understand the effects of different impairments on the phase portrait, we have simulated OSNR, CD, PMD, and in-band crosstalk on a 10-Gb/s NRZ signal. The resulting phase portraits and eye diagrams are shown in Figure 7.4. Figure 7.4(a) shows the results for no optical impairment ASE noise

Dispersion

PMD

Crosstalk

All

(a)

(b)

(c)

(d)

(e)

(f)

T/4

Delay = T

Eye

Clean

FIGURE 7.4 Eye diagrams and phase portraits (1-bit delay, ¼-bit delay) for 10-Gb/s NRZ: (a) OSNR ¼ 35 dB and no impairments; (b) OSNR ¼ 25 dB; (c) CD ¼ 800 ps/nm; (d) DGD ¼ 40 ps; (e) crosstalk ¼ 25 dB; and (f) OSNR ¼ 25 dB, CD ¼ 800 ps/nm, PMD ¼ 40 ps, and crosstalk ¼ 25 dB.

180

CHAPTER 7 OPM based on asynchronous delay-tap sampling

(OSNR 35 dB), with a clean eye, and with a well-defined geometric shape in the phase portrait. Figure 7.4(b) shows the effects of reducing the OSNR to 25 dB. The predominant effect is to broaden the high-power regions of both the eye and phase portrait, but the underlying geometric shapes are not affected. Figure 7.4(c) shows OSNR of 35 dB with 800 ps/nm of CD. The eye diagram shows the characteristic narrowing of the peaks and reduction in the amplitude of the 010 transition as measured at a 1-bit period in width. In the phase portrait, this manifests as the diagonal curving in toward the origin (the intersection of the diagonal and the line y ¼ x represents the height of the 010 pulse as measured at 1-bit period width). Figure 7.4(d) shows the effects of 30 ps of first-order PMD, with the power split equally between the principal polarization states. The eye shows the characteristic “triangularization,” but still looks very clean. Interestingly, at first glance, the phase portrait does not show the effects of PMD. However, a closer inspection shows significant differences in the distribution of points along the lines. Figure 7.4(e) shows the phase portrait with a single source of in-band crosstalk at 25 dB. Both the eye and two-tap plots show similar broadening to the OSNR degradation, but different noise statistics. Finally, Figure 7.4(f) shows OSNR of 25 dB with the combined degradations from Figures 7.4(a)–(e). These initial results suggested that the phase portraits contained impairment signatures that could be exploited to enable us to distinguish between varieties of optical impairments. As we will show in the next section, this is indeed the case; in particular, there exist features that enable us to identify and measure individual impairments even in the cases where they occur simultaneously, as in Figure 7.4(f). However, such a determination of individual impairment values for mixed impairments requires more sophisticated approaches than handcrafted measurement of image features. For instance, the curvature of the diagonal line in the 1-bit phase portrait has been proposed as a measure of CD; however, it is obscured by ASE noise and reduced by PMD. To address this issue in a principled manner, we used statistical learning techniques to automate feature selection. As emphasized in the introduction, the technique can monitor both amplitude and phasemodulated formats without the need for expensive demodulation. For example, Figures 7.6(d)–(f) and Figures 7.11(a)–(c) show sample phase portraits (25-ps delay) for 40-Gb/s NRZ-DPSK and 40-Gbs/s RZ-DQPSK signals, respectively. In the former case, the NRZ-like structure is due to the partial demodulation of the signal resulting from the tuneable channel selection filter. In general, the amplitude variation and structure of the phase portraits for these formats derives from a combination of the residual amplitude variation in the format, phase-to-amplitude conversion from impairments (i.e., CD), and finally, phase-to-amplitude conversion due to the channel selection filter.

7.2.2 Pattern recognition Phase portraits can be treated as images and lend themselves well to a variety of image analysis and pattern recognition techniques. These techniques, which we briefly describe next, include Hough transforms,12 Hausdorff measures,17 artificial neural networks,18 Kernel methods, and support vector machines.19,20 In the first example, a ¼-bit phase portrait was used to monitor a 10-Gb/s RZ-DPSK signal.12 For this case, a Q estimate (using the distribution of points along the major axis of the phase portrait) was used to monitor OSNR, while the width of the phase portrait was used to monitor CD. In practice, the width is obtained by using an image processing technique known as a Hough transformation.

7.2 Technique

181

Although this technique was demonstrated for simultaneous OSNR and CD monitoring, the presence of cross-coupling between the impairments limited the sensitivity of the technique. A simulation of CD monitoring in the presence of OSNR and PMD was demonstrated for 40-Gb/s NRZ-DPSK using a shape comparison method known as Hausdorff distance in Reference 17. The demonstration was, however, limited to CD monitoring with OSNR robustness above 22 dB. An alternative model that overcomes the coupling between impairments18 uses artificial neural networks (ANNs) to map phase portrait features onto the three impairments—OSNR, CD, and DGD. The features used to describe the phase portrait, denoted by ( r 1 , s1, r3 , s3, x2 , y2 , Q31), are obtained by dividing the portrait into quadrants, as shown on the left in Figure 7.5. Here r and s represent the expected value and standard deviation of the radial coordinate in the first and third quadrants, respectively, with x2 and y2 representing the mean values of the first and second samples in the second quadrant. The fourth quadrant was not used, as it was assumed to be symmetric with respect to the second. The final parameter is similar to the approximate Q-factor described above; it is defined as Q31 ¼ ðr3  r1 Þ=ðs1 þ s3 Þ. The ANN learns the relationships between the input features and the output impairments. The ANN architecture used had a three-layer structure, consisting of an input, hidden, and output layers. Using this approach, a correlation coefficient between input and output predictions of 0.97 was obtained for a simulated 10-Gb/s NRZ system with relatively few (140) training cases. Although strongly correlated, it is noted that the CD predictions show significant offsets at low CDs for OSNRs at the lower end of the training range (18 dB). An important limitation of this result was the restriction to worst-case DGD with equal power splits in the principle states. (For a general PMD power split ratio, g, the phase portrait is not symmetrical about the y ¼ x axis.) While the features used in the above approaches have the advantages of being interpretable, they are applicable for only a restricted range of formats and impairments. For this reason, the approach used by the authors is to automate the feature selection. The first step in the training process is to bin the sample pairs into a two-dimensional histogram. For the work presented here, a 3030 histogram was used and all 900 features (representing the number of hits in each bin) are available for training.

PC

DGD

TDCM

RF splitter

γ

Tx_1

OSNR Tx_i

TOF

AS

20 GHz

VOA

VOA Network emulator

TOF 30 GHz

Variable delay

Asynchronous multi-impairment monitor

FIGURE 7.5 Setup for generation of training sets. The impairment emulator adds known combinations of OSNR, CD, and first-order PMD to clean transponder signals. A polarization controller ensures a random distribution of power splits between principal states. AS, asynchronous sampler and outboard processing; PC, polarization controller; TDCM, tunable dispersion compensation module; TOF, tunable optical filter; Tx, transponder under test; VOA, variable optical attenuator.

182

CHAPTER 7 OPM based on asynchronous delay-tap sampling

A key advantage of this approach is that it can be used for any format or tap delay without the need for fine-tuning of the algorithm. In our approach, we use kernel-based techniques to model the impairments. The predictor for each impairment can, in general, be assumed to be a weighted nonlinear combination of features. For example, the predictor for absCD can be written as jCDj ¼

n X

  a j k xj ; x ;

j¼1

  P where k() is in general a nonlinear kernel function jCDji ¼ nj¼1 aj k xj ; xi of the phase portrait feature vector x to be measured and the n phase portraits in the training set. The training process determines the weights aj, which jointly minimizes the error function, and a regularization term, the latter of which is used to avoid over-fitting.20 In practice, the training sets are created using a “network emulator” to add known quantities of the impairments to a clean signal. Training is done independently for each of the impairments by minimizing the impairment prediction error (e.g., CD) in the presence of the other background impairments (e.g., OSNR and DGD). This enables the CD predictor to be valid over the DGD and OSNR range included in training. For impairment levels outside this range, the CD will be expected to suffer from OSNR and DGD dependence. We point out that the CD prediction does not require simultaneous DGD or OSNR predictions to be made, as the predictors are independently trained. Finally, we note that the pattern recognition techniques described here can also be applied to the analysis of eye diagrams. In particular, we note the use of support vector machines6 and ANNs. An experimental comparison of monitoring with eye diagrams and phase portraits was shown in Reference 4, in which it was found that the phase portrait outperformed eye diagrams.

7.3 EXPERIMENT We demonstrated the technique on 10-Gb/s NRZ and 40-Gb/s DPSK signals with simultaneous OSNR, CD, and DGD with random polarization. Figure 7.5 shows a schematic of the experimental setup used for training and testing. Signals for training can be sourced directly from transponders, as in this demonstration, or alternatively from a tap point at the start of the link. For this demonstration, the 10-Gb/s signals were generated with commercial transponders, while the 40-Gb/s signal was generated with a tunable DFB modulated with a dual-drive Mach Zehnder. In both cases, 223  1 PRBS was used. The clean signals were passed to a custom network emulation unit with known combinations of impairments for the training phase. The signal was then input to the delay-tap multiimpairment monitor to create and analyze the phase portraits. Control of the network emulator and the data processing of the multi-impairment monitor were done by an external laptop and an on-board processor, respectively.

7.3.1 Network emulator In the network emulator, the signal was amplified with an EDFA to overcome emulator loss, and filtered with a tunable 30-GHz (@ 3 dB) filter. A polarization controller was used to randomly select a polarization state and ensure a random distribution of g. Various known combinations of DGD and

7.3 Experiment

183

CD were then added to the monitored signal. The DGD emulator had a range of 0 to 100 ps, and g ¼ Ifast/(Ifast þ Islow) was derived from DC optical power measurements (averaged over the sampling time) using taps in the fast and slow axes of the DGD emulator. The polarization controller was programmed to step through random polarization states and was held constant during the sampling of each phase portrait. The OSNR was controlled by coupling in a variable amount of ASE, generated by filtering and amplifying the output of an EDFA. Varying levels of OSNR were used to ensure that the CD and PMD predictions were valid across normal operational ranges. The ASE source was placed after the PMD emulator to ensure that the measurement of g was not compromised by the added ASE noise. CD was controlled using a tunable dispersion compensation module with a range of 1400 to 1400 ps/nm. The impaired signal was then passed to the MIM for sampling and analysis.

7.3.2 Multi-impairment monitor In the monitor, the signal was amplified to deliver a constant power of 0 dBm to the photodiode, after then filtering with a tunable optical filter with a 3-dB bandwidth of 30 GHz. The signal was then fed to a 20-GHz receiver followed by a 50:50 splitter with tunable electrical delay in one arm. The signal and delayed ports were asynchronously sampled at 40 kS/s using a customized dual-channel A/D. For every network emulator setting, a phase portrait was generated from 40,000 sample pairs. Training and test sets were created by programming the emulator to step through combinations of CD from 1400 to 1400 ps/nm (200-ps/nm steps) and DGD from 0 to 22.5 ps (2.5 ps steps), while varying the OSNR between 15 and 25 dB (eight levels), giving a total of approximately 2000 phase portraits. We note that the polarization state, and hence g, varied randomly across this phase portrait data set. In practice, allowing for emulator tuning setup times, the training time is approximately 4 h.

7.3.3 First-order PMD In our first demonstration of the technique,3,4 the impairments included were OSNR, CD, and DGD, where the latter was restricted to worst-case polarization alignment of the signal and the principal states. More recently, we have demonstrated the technique at 10-Gb/s taking into account both the differential group delay (DGD) and a random power split g between principal states. To do this, we define an effective differential group delay, DGDeff ¼ 4g(1-g)DGD, that is proportional to first-order string length and is a measure of first-order PMD system penalty.21,22 When the signal is aligned with either principal state of polarization (PSP) (g ¼ 0 or 1), DGDeff ¼ 0 and the signal remains undistorted regardless of the level of first-order PMD. The factor of 4 in DGDeff is chosen so that when g ¼ ½, DGDeff ¼ DGD. The advantages of using DGDeff are that it is directly related to the induced signal distortion, and thus the first-order PMD-induced system penalty, and it provides a dynamic measure of feedback for PMD compensation.

7.3.4 Results for 10-G NRZ and 40-G NRZ-DPSK Figure 7.6 shows a selection of phase portraits taken from the training sets. For the 10-Gb/s signals, we noted characteristic narrowing of the phase portrait and more uniform width due to DGD. A closer inspection shows that the phase portrait is not perfectly symmetric; this can be partly

CHAPTER 7 OPM based on asynchronous delay-tap sampling

DGDeff: 0.4 ps, absCD: 0

DGDeff: 39.3 ps, absCD: 0

DGDeff: 1.4 ps, absCD: 1200

(a)

(b)

(c)

DGDeff: 0.4 ps, absCD: 0

DGDeff: 14.0 ps, absCD: 0

DGDeff: 1.9 ps, absCD: 400

(d)

(e)

(f)

40-Gb/s DPSK

10-Gb/s NRZ

184

FIGURE 7.6 Experimental phase portraits for (a–c) 10-Gb/s NRZ and (d–f) 40-Gb/s DPSK showing the effects of CD and DGD; the OSNR is 14 dB for all cases. The tap delay was set at 25 ps for both bit rates. The impairment levels were DGDeff (ps), absCD (ps/nm): (a) 0,0, (b) 39,0, (c) 0,1200, (d) 0,0, (e) 14,0, and (f) 2,400.

attributed to g not equal to 1. The optical filter bandwidth of 30 GHz partially demodulates the 40-G DPSK signal resulting in the characteristic NRZ portrait. The effects of CD are clearly seen in the loop structure of Figure 7.6(f) and, as with 10-Gb/s NRZ, there is a reduction in the width of the portrait with increasing DGD. For each format, a set of 1500 phase portraits generated by the emulator was used to generate independent prediction models for OSNR, CD, and DGDeff. Tests were carried out on a further 250 randomly selected impairment combinations that were not included in the training set. Test results are shown, using an average of five measurements for each impairment combination, in the ordered plots of Figure 7.7. For 10 G, we find that RMS error at the 2s level for OSNR is 0.2 dB, and jCDj is 55 ps/nm over a range of 1400 ps/nm to 1400 ps/nm and 3.6 ps for DGDeff over 0–60 ps. For 40-Gb/s DPSK, the result for OSNR is 0.7 dB, and jCDj is 9 ps/nm over a range of 400 ps/nm to 400 ps/nm and 1.5 ps for DGDeff over 022 ps. A breakdown of the 10-G results shows that the accuracy of the DGDeff degrades, as expected, with poor OSNR and high CD. For 40-G DPSK, however, it was found that the DGDeff accuracy degrades with poor OSNR, but surprisingly improves with increasing CD. A possible interpretation of this is that the CD-induced, phase-to-amplitude conversion provides a larger waveform (Figure 7.6(f)) for DGDeff to distort.

7.3 Experiment

OSNR (dB), stdTe = 0.22 17

absCD (ps/nm), stdTe = 55.0 1500

10-Gb/s NRZ

DGDeff (ps), stdTe = 3.6 60 50

16 1000

15

40 30

14

500

20

13 12

10 0

100

200

300

0

0

100

(a) OSNR (dB), stdTe = 0.69

25

20

15

0

100

200

300

0

0

100

(b)

30

40-Gb/s DPSK

185

200 (d)

300

absCD (ps/nm), stdTe = 9.1 500

25

400

20

300

15

200

10

100

5

0

0

100

200 (e)

200

300

(c)

300

0

DGDeff (ps), stdTe = 1.5

0

100

200

300

(f)

FIGURE 7.7 Experimental measurements (gray) of simultaneous OSNR, CD, and DGD impairments for (a–c) 10-Gb/s NRZ and (d–f) 40-Gb/s NRZ-DPSK. Results for 250 test cases are ordered along the x axis by true values (shown in black). The test errors, stdTe, are quoted at the 2s level.

The previous results can only be regarded as proof of principle since they have used training and test sets generated from the same transponder and impairment emulator. Tests for transponder dependency were performed on three commercial 10-G transponders. The results for different combinations of training sets are shown in Table 7.2. We found that training and testing on separate transponders can give rise to significant errors. Training sets from combined transponders, however, add robustness to the measurements. Verification of the impairment emulator is shown in test results on independent impairments in Table 7.3. In this case, the CD for test cases is generated from varying lengths of SMF fiber and the PMD from birefringent fiber. For further verification, the technique has been recently demonstrated on a 10-Gb/s NRZ WDM system test bed that includes ROADMs.23

7.3.5 System testing Measurements of CD and OSNR were performed on an 800-km, 10-Gb/s NRZ test bed (50-GHz channel spacing) using three monitoring points, as illustrated in Figure 7.8. At each point, the CD

186

CHAPTER 7 OPM based on asynchronous delay-tap sampling

Table 7.2 Test Errors for Transponder T2 Showing Effects of Including Combinations of Transponders in Training Training Set

OSNR (dB)

TX3 TX1TX3 TX1TX2TX3 TX2

absCD (ps/nm)

0.6 0.7 0.5 0.4

DGDeff (ps)

51 45 40 26

2.8 2.6 2.5 2.1

Table 7.3 Independent Validation of Impairment Emulator OSNR (dB)

absCD (ps/nm)

DGDeff (ps)

True Measured

13.0 13.3

0 55

9.0 9.3

True Measured

13.0 13.4

0 8

14.0 15.2

True Measured

13.0 13.7

576 572

0.0 1.6

True Measured

13.0 13.2

576 547

9.0 7.1

True Measured

13.0 13.2

576 557

14.0 13.0

Monitoring point 1 8x 10-G

100 km

Monitoring point 2

100 km

100 km

100 km

ROADM

Tx

DCM 60

DCM 100 DCM 40

DCM 120

DCM 80 20 dB

DCM 110 DCM 120 8x 10-G Rx

100 km

100 km

DCM 60 ROADM

DCM 60

100 km

DCM 60

100 km

Monitoring point 3

FIGURE 7.8 Setup for 10-G NRZ WDM system test. OSNR and CD were monitored at the three tap points. An additional 400 ps/nm dispersion was added at each of the tap points as a further check of CD accuracy. The input power to the monitor was 18 dBm.

7.4 Discussion

187

Table 7.4 Simultaneous OSNR and CD Measurements on 800-km WDM Test Bed Tap Point


CD (ps/nm)

True OSNR (dB)

OSNR Mean Offset (dB)

CD Measured (ps/nm)

1

400 0 400 400 0 400 400 0 400

19.3 19.3 19.3 19.3 19.3 19.3 16.8 16.8 16.8

0.3 0.1 0.4 0.0 0.3 0.4 0.2 0.2 1.0

420 0 þ430 430 0 þ430 172 0 520

2

3

(18.4) (20.5) (20.5) (21.3) (21.3) (21.3) (18.5) (18.6) (18.6)

(0.3) (0.2) (0.1) (0.5) (0.2) (0.1) (0.3) (2.0) (6.0)

(520) (þ400) (520) (þ400)

Note: Values in parentheses show measurements made at 3-dBm power levels.

was varied over three levels using an additional þ/400 ps/nm of dispersion-compensating fiber. A total of 30 measurements of CD and OSNR were made for each level. Measurements were then repeated at different power levels (0 and 3 dBm) to test for the effects of SPM. The true OSNR was measured with an OSA. An independent CD measurement at point 1 was 120 ps/nm and at point 3 was 1000 ps/nm. The latter was outside the training range. The results are summarized in Table 7.4. We see that the mean (systematic) error in the OSNR prediction is less than 0.5 dB for tap points before and after the ROADM and show minimal sensitivity to CD. The degradation in accuracy at monitoring point 3 with 0 and þ400 ps/nm is due to the residual CD falling outside of the training range. The standard deviation in the 30 OSNR measurements for each setting was 0.2 dB. A repeat of the measurements at 3-dBm launch powers again showed excellent agreement with independent OSNR measurements. The measured residual CD at points 1 and 2 was within 10 ps/nm of true CD. The measured residual value at point 3 was in error, because it lay outside the training range. The standard deviation in CD measurements was 25 ps/nm. The CD range increases with 3-dB launch power from 860 to 920 ps/nm and may indicate the onset of SPM effects on CD predictions.

7.4 DISCUSSION In work to date, it has been assumed that the monitor has access to knowledge of the formats and bit rates being monitored from the network management system and can select appropriate predictors from an on-board library. For the monitoring of systems with alien wavelengths, this information may not be available and automatic identification of format would be required. Automatic format and bit rate identification is a feature that the technique is well suited to. As an example, we have demonstrated that a single predictor can predict impairments at both 10.3- and 10.7-Gb/s FEC rates.

188

CHAPTER 7 OPM based on asynchronous delay-tap sampling

Table 7.5 Demonstrating Improvement in DGD and CD Accuracy with Increasing Size of Training Set Number of Training Cases

DGDeff Error (ps)

500 750 1000 1250 1500

3.6 2.7 2.5 2.0 1.9

jCDj Error (ps/nm) 17.3 14.1 13.0 11.1 10.4

Table 7.5 shows the improvement in accuracy for 40-Gb/s DPSK obtained with increasing training set size. There is a relatively rapid improvement in performance for the first 1000 training cases, which then plateaus for greater than 2000 cases. In general, however, the required number of training cases will vary depending on the number and range of impairments. Allowing for emulator tuning times, the training time for 2000 cases is approximately 4 h. For the experimental setup described in this paper the measurement time was 1.2 s. This is dominated by the sampling time; the processing time required for prediction takes less than 20 ms. In future implementations, cost-effective solutions at sampling rates of 50 MS/s will be achievable with currently available samples and old technology. This will enable measurement times of better than 50 ms and provide an increased ability to exploit the improvement in accuracy that is achievable through averaging of multiple measurements.

7.4.1 Extension to new impairments In principle, in order to make the technique more robust, one can extend the training set to include other effects that can potentially induce errors, such as optical nonlinearities, filter drift, and delay variations. To illustrate this ability, we have extended the number of simultaneous impairments to include optical filter offset and in-band crosstalk, in addition to OSNR, CD, and DGD (with random g). The results for a simulated 10-Gb/s NRZ signal are shown in Figure 7.9. The filter offset result shows that the technique is able to distinguish between filter-induced dispersion (e.g., from ROADMs) and fiber-induced dispersion. We noted that the mean optical channel power is not used as an input for predictions, and hence the optical filter loss was not being used to measure the offset. The crosstalk introduced into the model was from a single interferer with the same wavelength as the signal. The non-Gaussian statistics of this interference noise enable the machine learning to distinguish the crosstalk from OSNR. Although the crosstalk is determined with only limited accuracy, its presence causes little degradation to the OSNR accuracy.

7.4.2 Application to higher-order formats A key advantage of our approach is that it can be applied without the need for fine tuning of features or modification of hardware. For example, using the same algorithms and hardware as used

7.4 Discussion

OSNR (dB), stdTe = 0.33

absCD (ps/nm), stdTe = 14.66

25

189

DGDeff (ps), stdTe = 1.56 50

1200

40 20

800

30 20

15

400 10

10

0

500 (a)

0

1000

0

500 (b)

In-band crosstalk (db), stdTe = 1.35 30

0

1000

0

500 (c)

1000

Filter offset (GHz), stdTe = 0.64 15

25 10 20 5 15 10

0

500 (d)

1000

0

0

500 (e)

1000

FIGURE 7.9 Simulation results for 10-Gb/s NRZ demonstrating monitoring of simultaneous OSNR, CD, DGD, in-band crosstalk, and optical filter offset. The training ranges were for OSNR, 11–25 dB; CD, 1400 to 1400 ps/nm; and DGD, 0–50 ps (random g), crosstalk 15–24 dB, and filter offset 0–12 GHz. The training set consisted of 2000 random combinations of these impairments. Predictions for 1000 test cases (gray) are shown; true values are shown in black. The RMS error at the 2s level are (a) OSNR 0.3 dB, (b) absCD 15 ps/nm, (c) DGD 1.6 ps, (d) crosstalk 1.4 dB, and (e) filter offset 0.7 GHz.

for 10-Gb/s NRZ and 40-Gb/s DPSK, we find that simulations of 40-Gb/s RZ-DQPSK (see Figures 7.10 and 7.11(a)–(c)) give results that are consistent with the NRZ and DPSK results for simultaneous CD and DGDeff presented in this chapter. In addition, we note that ability of machine learning techniques to discern patterns in very “unstructured” phase portraits holds promise for applying the technique to monitoring polarization-multiplexed systems. By way of example, in Figure 7.11(d, e) we show the effects of CD and DGD on the phase portrait for a simulated 80-Gb/s polarization-multiplexed RZ-QPSK. We emphasize that these portraits have been obtained assuming the same direct detection–based monitor (25-ps delay and 20-GHz receiver) used for 10-Gb/s NRZ signals.

190

CHAPTER 7 OPM based on asynchronous delay-tap sampling

absCD (ps/nm), stdTe = 15.1

DGDeff (ps), stdTe = 1.45 25

800 40-Gb/s DQPSK

20 600 15 400

10

200 0

5 0

100

200

300

0

0

(a)

100

200

300

(b)

FIGURE 7.10 Simulation results for simultaneous measurements of absCD and DGDeff for 40-Gb/s RZ-DQPSK. The training set contained 2000 cases with OSNR ranging from 14 to 28 dB; CD, 800 to 800 ps/nm; and DGD, 0 to 25 ps.

DGDeff: 25.0 ps, absCD: 0

(a)

(b)

DGDeff: 0.0 ps, absCD: 0

DGDeff: 25.0 ps, absCD: 0

(d)

(e)

DGDeff: 0.0 ps, absCD: 800

40-Gb/s DQPSK

DGDeff: 0.0 ps, absCD: 0

(c)

80-Gb/s PM-DQPSK

DGDeff: 0.0 ps, absCD: 800

(f)

FIGURE 7.11 Simulated phase portraits for (a–c) 40-Gb/s RZ-DQPSK and (d–f) 80-Gb/s polarization-multiplexed DPSK showing the effects of CD and DGD. The OSNR is 14 dB for all cases. The tap delay was set at 25 ps for both bit rates. The impairment levels were DGDeff (ps), absCD (ps/nm): (a) 0,0, (b) 25,0, (c) 0,800, (d) 0,0, (e) 25,0, and (f) 0,800.

References

191

7.5 SUMMARY The asynchronous delay-tap sampling technique is a promising in-service monitoring technique capable of measuring simultaneous impairments for multiple bit rates and formats. The two core components of the technique are the characterization of the signal with an asynchronously sampled phase portrait and the extraction of impairment features using statistical machine learning techniques. A major advantage of the technique is that a direct detection–based receiver with a set bandwidths can be used for multiple formats and bit rates. In this chapter we have reviewed the technique beginning with a discussion and interpretation of the phase portrait and a brief overview of some pattern recognition approaches demonstrating its application to NRZ, DPSK, and QPSK formats for simultaneous measurements of OSNR, CD, and DGDeff. Early proofs of principle demonstrations, based on single transponder- and laboratorygenerated impairments, have been further validated at 10-Gb/s with WDM system measurements. Key challenges for practical realization of the technique include a demonstration of robustness with respect to transponder variations and compatibility with ROADMs. We find that robustness to transponder variations can be improved by training over multiple transponders. Similarly, the ability to distinguish dispersion in the presence of ROADM filter effects may be enhanced by including filter variations in the training set. Finally, our experience to date suggests that the remarkable ability of machine learning techniques to discern patterns in extremely “unstructured” phase portraits holds promise for application of the technique to the monitoring of advanced 100-Gb/s formats and the “next generation” of impairments.

REFERENCES 1. Kilper DC, Bach R, Blumenthal DJ, Einstein D, Landolsi T, Ostar L, et al. Optical performance monitoring. J Lightwave Technol 2004;22:294–304. 2. Kaminow IP, Li T, Willner AE. Optical performance monitoring. In: Optical fiber telecommunications V B. New York, New York: Academic Press; 2008. 3. Dods SD, Anderson TB. Optical performance monitoring technique using delay tap asynchronous waveform sampling. In: Proc. OFC/NFOEC, paper OThP5. Anaheim, California; 2006. 4. Anderson T, Dods SD, Clarke K, Bedo J, Kowalczyk A. Multi-impairment monitoring in photonic networks. In: Proc. ECOC, paper 3.5.1. Berlin, Germany; 2007. 5. Ji H, Park K, Lee J, Chung H, Son E, Han K, et al. Optical performance monitoring techniques based on pilot tones for WDM network applications. J Opt Network 2004;3:510–33. 6. Skoog RA, Banwell TC, Gannett JW, Habiby SF, Pang M, Rauch ME, et al. Automatic identification of impairments using support vector machine pattern classification on eye diagrams. IEEE Photon Technol Lett 2006;18:2398–400. 7. Hanik N, Gladisch A, Caspar C, Strebel B. Application of amplitude histograms to monitor performance of optical channels. Electron Lett 1999;35:403–4. 8. Kozicki B, Takuya O, Hidehiko T. Optical performance monitoring of phase-modulated signals using asynchronous amplitude histogram analysis. J Lightwave Technol 2008;26:1353–61. 9. Geyer JC, Hauske FN, Fludger CRS, Duthel T, Schulien C, Kuschnerov M, et al. Channel parameter estimation for polarization diverse coherent receivers. IEEE Photon Technol Lett 2008;20(10):776–8.

192

CHAPTER 7 OPM based on asynchronous delay-tap sampling

10. Mayrock M, Haunstein H. Performance monitoring in optical OFDM systems. In: Proc. OFC/NFOEC, paper OWM3. San Diego, California; 2009. 11. Anderson T, Clarke K, Beaman D, Ferra H, Birk M, Guodong Z, et al. Experimental demonstration of multi-impairment monitoring on a commercial 10 Gb/s NRZ channel. In: Proc. OFC/NFOEC, paper OThH7. San Diego, California; 2009. 12. Kozicki B, Maruta A, Kitayama K. Experimental demonstration of optical performance monitoring for RZ-DPSK signals using delay-tap sampling method. Opt Expr 2008;16:3566–76. 13. Kozicki B, Maruta A, Kitayama K. Asynchronous optical performance monitoring of RZ-DQPSK signals using delay-tap sampling. In: Proc. ECOC, paper P060. Berlin, Germany; 2007. 14. Choi HY, Takushima Y, Chung YC. Multiple impairment monitoring technique using optical field detection and asynchronous delay-tap sampling method. In: Proc. OFC/NFOEC, paper 0ThJ5. San Diego, California; 2009. 15. Ku YC, Chan CK. High-speed data and pulse-carver alignment in RZ-OOK systems using delay tap asynchronous waveform sampling. In: Proc. ECOC, paper Tu4.2.2. Cannes, France; 2006. 16. Kantz H, Schreiber T. Nonlinear time series analysis. Cambridge and New York: Cambridge University Press; 2004. 17. Zhao J, Lu C, Lam K-M, Li Z, Tam HY, Wai PKA. A novel optical signal monitoring method of DPSK signal based on delay tap sampling and Hausdorff distance measure. In: Proc. CLEO/QELS, paper JWA108. San Jose, California; 2008. 18. Jargon JA, Wu X, Willner AE. Optical performance monitoring using artificial neural networks with features derived from asynchronous delay tap sampling. In: Proc. CLEO/QELS, paper OThH1. San Jose, California; 2009. 19. Vapnik V. Statistical learning theory. New York, New York: Wiley; 1998. 20. Duda RO, Hart PE, Stork DG. Pattern classification. New York, New York: Wiley; 2000. 21. Kogelink H, Jopson RM, Nelson LE. Polarization-mode dispersion. In: Kaminow I, Li T, editors. Optical fiber telecommunications, IVB. New York: Academic Press; 2002. 22. Cornick KE, Boroditsky M, Finch S, Dods SD, Farrell PM. Experimental comparison of PMD-induced system penalty models. IEEE Photon Technol Lett 2006;18:1149–51. 23. Anderson T, Beaman D, Li JC, Jerphagnon O, Le Rouzic E, Neddam F, et al. Demonstration of simultaneous OSNR and CD monitoring using asynchronous delay tap sampling on an 800 km WDM test bed. In: Proc., paper 9.3.4 ECOC. Vienna, Austria; 2009. 24. Clarke K, Anderson T, Dods SD. Monitoring of multiple modulation formats using asynchronous delay-tap sampling. In: Proc. ACOFT, paper MoA1–2. Melbourne, Australia; 2007.

CHAPTER

Optical performance monitoring based on linear optical sampling

8 Christophe Dorrer

Laboratory for Laser Energetics, University of Rochester, Rochester, New York

8.1 INTRODUCTION

8.1.1 Data encoding in the electric field of optical waves Modulating and measuring the phase of optical waves are enablers in a wide variety of applications in optics. In optical telecommunications, optical waves with time-varying instantaneous power I(t) and phase c(t) are used to transmit information between a transmitter and receiver. Although these quantities are usually thought p offfiffiffiffiffiffiffi as independent, it is better to think in terms of the electric field of the modulated waves EðtÞ ¼ IðtÞ exp½icðtÞ since this is really the physical quantity of interest. Encoding information in the electric field of optical waves is attractive for several reasons.1,2 Phase-modulated signals operate with constant or periodic intensity, and the deleterious effects of nonlinear propagation can be reduced relatively to the effects experienced by signals based on amplitude modulation. Various discrete phase levels can be encoded and decoded unambiguously. Phasemodulated formats can be spectrally efficient, in the sense that good transmission properties are obtained while using less bandwidth than an on-off keying (OOK) system. This is particularly important for WDM applications. Phase modulation can be combined with amplitude modulation to convey multiple information bits per time slot. Figure 8.1 displays examples of data-encoded optical waveforms: on the left side, the temporal intensity and phase are plotted independently, while on the right side, a constellation diagram represents the distribution of the electric field in the complex plane at the center of the time slot. The OOK signal in Figure 8.1(a) uses the absence or presence of energy in the time slot to encode binary information, and does not rely on phase modulation to carry information. The binary phase-shift keying (BPSK) and quaternary phase-shift keying (QPSK) signals in Figures 8.1(b) and (c) only use phase encoding to encode 1 bit or 2 bits in a given time slot, respectively. The quadrature amplitude modulation (QAM) signal in Figure 8.1(d) relies on amplitude and phase modulation of the electric field to encode information. Since square-law photodetectors are only sensitive to the intensity of the electric field, one cannot directly measure the phase of an optical wave and directly decode a phase-modulated signal by photodetection. However, interference of two optical waves leads to an intensity that depends on the relative phase of the two waves. For example, spectral interferometry uses the spectrally resolved interference between two optical waves to recover their spectral phase difference as a function of the optical frequency, while spatial interferometry can measure the phase difference between two optical © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00008-0

193

194

CHAPTER 8 Optical performance monitoring based on linear optical sampling

(a) t (b)

π t π/2

(c) t (d) t

FIGURE 8.1 Data-encoded optical signals represented by intensity and phase as a function of time (left column) and complex electric field at center of time slot (right column). The signals correspond to (a) on-off keying, (b) binary phase-shift keying, (c) quaternary phase-shift keying, and (d) 16-state quadrature amplitude modulation.

waves as a function of the spatial variables. The information encoded in an optical telecommunication signal can be recovered by measuring the interferometric beating between the signal and the local oscillator (LO)—that is, a reference wave provided by the receiver. One can also generate a delayed replica of the signal, and interfere the signal and its delayed replica to recover the phase difference between successive bits of the data without a local oscillator. This is generally known as differential phase-shift keying, and popular corresponding telecommunication formats are DPSK (two phase levels) and DQPSK (four phase levels).

8.1.2 Temporal characterization of optical signals The characterization of time-varying optical signals has been a challenging problem since the invention of lasers. This is due to the lack of phase sensitivity of square-law photodetectors and the relatively slow response times of electronic phenomena when compared to optical phenomena. The second-order intensity autocorrelation is a well-known approach to obtain temporal information about optical signals. Its measurement is simple: two time-delayed replicas of the test signal are generated and sent to a nonlinear medium properly phase matched for up-conversion. The energy of the up-converted signal is measured as a function of the interreplica relative delay. For isolated optical pulses, a functional form is often assumed for the temporal intensity (e.g., Gaussian or secanthyperbolic), in which case the intensity duration can be estimated from the duration of the intensity autocorrelation. The intensity autocorrelation only gives indirect information about the instantaneous power of the test signal and does not contain any phase information. It can, however, be used in

8.1 Introduction

195

advanced optical performance monitoring, as intensity correlation between different times in a data stream can be used to quantify impairments.3 Indirect approaches have been devised for the characterization of the electric field of optical signals.4,5 The optical signal is modified in various controlled ways; for example, it is gated by a modulator or by a nonlinear interaction with a short optical pulse, filtered in the spectral domain, or sent in a dispersive element. The energy of the modified signal is then measured with a timeintegrating photodetector. The experimental trace is composed of the measured energies as a function of the parameters of the different filters that have been used (e.g., the delay between the pulse and the temporal modulation, or the center frequency of the passband of a spectral filter). Under some conditions, the electric field of the signal (intensity and phase) can be retrieved from the measured experimental trace. Many techniques based on different principles have been demonstrated in the field of ultrafast optics. For example, frequency-resolved optical gating (FROG) uses nonlinear optics to measure a spectrogram of the test source, from which the electric field can be reconstructed using iterative blind-deconvolution phase-retrieval algorithms.6 The spectrogram is measured by nonlinear interaction of several replicas of the test source, such as sum-harmonic generation or four-wave mixing; the optical spectrum of the nonlinear signal is measured as a function of the delay between the replicas and the optical frequency. Another popular technique, spectral phase interferometry for direct electric-field reconstruction (SPIDER), uses the interference of two spectrally sheared test pulses to directly reconstruct the spectral phase of the test source.7 The spectral shear is obtained by nonlinear conversion with a chirped optical pulse, and the spectral phase is reconstructed by Fourier processing of a spectral interferogram measured for example by a spectrometer. These techniques rely on nonlinear optics, and their application to signals in the telecommunication environment can be difficult. Linear techniques based on similar principles but implemented with telecommunication components have also been demonstrated for use with lowaverage power sources.8,9 While these techniques have been successfully applied to repetitive optical signals, including pulses in the telecommunication environment, they would be difficult to use for single-shot measurements of nonrepetitive events such as data-encoded telecommunication signals. There are few reports of the measurement of such signals. One approach for single-shot characterization of the intensity of an optical signal is to use time magnification, a general process by which a temporally scaled copy of the test waveform is generated.10,11 This scaling is generally obtained using a concept similar to spatial imaging with magnification—that is, by using quadratic temporal and spectral phase modulations to mimic the effect of lenses and free-space propagation. A quadratic temporal phase modulation can be induced by a temporal phase modulator driven by a sinusoidal drive provided that the optical pulse is synchronized with one of the extrema of the modulation. It can also be obtained by nonlinear interaction with a chirped optical pulse. The quadratic spectral phase modulation can be obtained simply by propagation in a dispersive fiber. Temporal magnification was extended to the single-shot characterization of the electric field using heterodyning with a monochromatic source.12 Characterizing fast optical phenomena remains difficult for nonrepetitive optical signals. For these signals, it is mandatory to measure a meaningful experimental trace in a single shot. The measurement of temporal samples of the time-varying instantaneous power of a data-encoded optical signal is often used in the telecommunication environment. It is typically performed with a gate having a bandwidth higher than the bandwidth of the signal in order to temporally resolve the intensity of the optical wave. The measured samples are then plotted in a statistical representation

196

CHAPTER 8 Optical performance monitoring based on linear optical sampling

(e.g., an eye diagram), and performance can be assessed using various metrics. Photodetection and analog-to-digital conversion typically have a bandwidth lower than 100 GHz. High-bandwidth optical sampling systems have been implemented where the test optical signal is gated by a high-bandwidth, low-repetition-rate gate, and the energy of the resulting pulse is measured using a slower analog-to-digital converter.4,13 The corresponding sample is representative of the instantaneous power of the test signal at the time defined by the gate. Higher bandwidths can be obtained using nonlinear interactions with short optical pulses, including sum-frequency generation and four-wave mixing. By itself, this approach can at best provide temporal samples of the intensity that can be grouped into an eye diagram. No phase information is generally obtained by sampling techniques, whether they are based on photodetection, temporal modulators, or nonlinear interactions.

8.1.3 Linear optical sampling Wave interference is a well-known physical phenomenon to perform phase-to-intensity conversion. By the interference of two optical waves, the relative phase difference between the waves becomes apparent in the resulting intensity—that is, the phase of a test optical source can be obtained by beating with a reference optical source, a process known as coherent detection, homodyning, or heterodyning. Coherent photodetection is the basis of linear optical sampling (LOS): characterization of the temporal electric field of an optical source is obtained by interference with a local oscillator and photodetection. The local oscillator can be a quasimonochromatic source, in which case the interference must be measured with fast photodetectors. The local oscillator can also be a short optical pulse providing the necessary temporal resolution, in which case the interference can be resolved with slower photodetectors. Diagnostics based on this principle are advantageous since they can in principle provide a high-resolution temporal representation of the electric field of a test source. This information is valuable for optical performance monitoring since the properties of the signal can be assessed in a somewhat general fashion. Coherent detection is used extensively for optical telecommunications.2 This aspect is not described in this chapter, which focuses on the application of coherent detection, particularly with a short optical pulse, as a diagnostic for optical performance monitoring. The principle of LOS is presented in Section 8.2. Section 8.3 describes experimental implementations of LOS. Section 8.4 presents various experimental results demonstrating the application of LOS to optical performance monitoring of telecommunication signals and characterization of devices and fibers. Finally, Section 8.5 is a nonexhaustive description of related techniques and recent results.

8.2 LOS PRINCIPLE AND PROPERTIES 8.2.1 Coherent detection

The real electric field e of an optical wave is a solution of Maxwell’s equations. It is convenient to work with the analytic signal E, which verifies eðtÞ ¼ EðtÞ þ E  ðtÞ:

(8.1)

8.2 LOS principle and properties

197

The analytic signal is complex, and therefore can be expressed uniquely in terms of an amplitude and phase, EðtÞ ¼jEðtÞjexp½icðtÞ þ ic0  io0 t;

(8.2)

where jE(t)j is the time-dependent field amplitude (i.e., the envelope of the field oscillations), o0 is the carrier frequency (usually chosen near the center of the pulse spectrum), c(t) is the time-dependent phase, and c0 is a constant, known as the carrier-envelope offset (CEO) phase. The square of the envelope, I(t) ¼jE(t)j2, is the time-dependent instantaneous power of the pulse that can be measured if a detector of sufficient bandwidth is available (note that absolute measurement of the instantaneous power is usually not required). The temporal phase takes into account the changes in the frequency of the field as a function of time via the instantaneous frequency O ¼ @c/@t. The CEO phase describes the relative location of the field envelope and carrier, which is usually different from pulse to pulse, even in a pulse train from a mode-locked laser unless active stabilization is used.14 The frequency representation of the analytic signal is defined by the Fourier transform, ð þ1 ~ ~ dtEðtÞ expðiotÞ: (8.3) EðoÞ ¼j EðoÞjexp½i’ðoÞ ¼ 1

~ Here j EðoÞ j is the spectral amplitude and ’(o) is the spectral phase. The square of the spectral 2 ~ ~ , is the spectral density, which can be obtained with an optical spectrum amplitude, IðoÞ ¼j EðoÞj analyzer. The spectral phase describes the relative phase of the optical frequencies composing the pulse, and its derivative @’/@o is the group delay T(o) at the corresponding frequency—that is, the time of arrival of a subset of optical frequencies of the pulse around o. A pulse with a constant group delay—that is, a linear spectral phase—is said to be Fourier-transform limited because it is the shortest pulse that can be obtained for a given optical spectrum.15 Expressions for the LOS signal with various local oscillators are obtained using the notations of Figure 8.2. The interference of two optical waves with fields E1(t) and E2(t) leads to the intensity I12 ðtÞ ¼ j E1 ðtÞ þ E2 ðtÞj2  ¼ I1 ðtÞ þ I2 ðtÞ þ 2p Re½E 1 ðtÞE2ffi ðtÞ ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ I1 ðtÞ þ I2 ðtÞ þ 2 I1 ðtÞI2 ðtÞ cos½c1 ðtÞ  c2 ðtÞ:

(8.4)

The phase difference c1 – c2 between the two waves is therefore revealed as an intensity modulation after interference. If E1 is a phase-encoded telecommunication signal, E2 can be a local oscillator generated at the receiver, or E2 can be a delayed version of E1. For LOS, the local oscillator is an independent optical wave provided by the diagnostic, and the fields are written E2 ¼ ELO and E1 ¼ EDATA. When using a 90-degree optical hybrid, as pictured in Figure 8.2, the two fields are split at the splitters SLO and SDATA and recombined at the combiners CA and CB. The interference measured with a square-law photodetector with impulse response R on one output of the combiner CA is Ð þ1 SA;1 ðtÞ ¼ 1 Rðt  tÞ j EDATA ðtÞ þ ELO ðtÞj2 dt Ð þ1 Ð þ1 Ð þ1 ¼ 1 Rðt  tÞIDATA ðtÞdt þ 1 Rðt  tÞILO ðtÞdt þ 2 1 Rðt  tÞRe½EDATA ðtÞELO ðtÞdt; (8.5)

198

CHAPTER 8 Optical performance monitoring based on linear optical sampling

Data source

CA

SLO

Local oscillator

– SA

SDATA

CB

– SB

p/2 phase shift

FIGURE 8.2 Layout for the direct measurement of the complex interference between data source and local oscillator. The two sources are split at the splitters SLO and SDATA, and recombined pairwise at the combiners CA and CB. Balanced photodetection of the two outputs of each combiner yields the in-phase and quadrature components of the interference if a relative p/2 phase shift is introduced in the optical path of one of the two sources between splitters and combiners.19 Copyright © 2006 OSA.

where Re(z) represents the real part of the complex argument z. The signal measured on the other port of the combiner CA is Ð þ1 SA;2 ðtÞ ¼ 1 Rðt  tÞ j EDATA ðtÞ  ELO ðtÞj2 dt Ð þ1 Ð þ1 Ð þ1 ¼ 1 Rðt  tÞIDATA ðtÞdt þ 1 Rðt  tÞILO ðtÞdt  2 1 Rðt  tÞ Re½EDATA ðtÞELO ðtÞdt: (8.6) Balanced photodetection isolates the interferometric term from the phase-independent terms, leading to the signal SA ðtÞ ¼ SA;1 ðtÞ  SA;2 ðtÞ Ð þ1 ¼ 4 1 Rðt  tÞ Re½EDATA ðtÞELO ðtÞdt:

(8.7)

Since the recovery of the complex interference from its real part is in general ambiguous, one usually arranges for the simultaneous measurement of its imaginary part. This can be achieved by adding a relative p/2 phase shift on one of the fields before interference. Balanced photodetection at the two outputs of combiner CB leads to SB ðtÞ ¼ SB;1 ðtÞ  SB;2 ðtÞ Ð þ1 ¼ 4 1 Rðt  tÞ RefEDATA ðtÞ½iELO ðtÞ gdt Ð þ1 ¼ 4 1 Rðt  tÞ Im½EDATA ðtÞELO ðtÞdt:

(8.8)

8.2 LOS principle and properties

199

It is customary to call the quantities SA and SB the in-phase and quadrature component of the interference between the two sources, which are also known as the I and Q components. SA and SB can be combined to obtain the complex interference of the data and local oscillator: SðtÞ ¼

ð þ1

1 Rðt  tÞEDATA ðtÞELO ðtÞdt ¼ ½SA ðtÞ þ iSB ðtÞ: 4 1

(8.9)

Balanced photodetection benefits from the intrinsic cancellation of the intensity of each source. For practical reasons, particularly if high-speed photodetectors are required, a single photodetector might be used instead of a balanced photodetector. In this case, it is not possible to extract one of the quadratures from a single sample described by Equation (8.5). However, it is possible to extract the interference between the two sources if a large collection of samples is obtained, and the average value of these samples is removed. Since the average value of the interference term is zero for uncorrelated sources, one obtains SðtÞ ¼

ð þ1 1

1 Rðt  tÞEDATA ðtÞELO ðtÞdt ¼ f½SA;1 ðtÞ  hSA;1 ðtÞi þ i½SB;1 ðtÞ  hSB;1 ðtÞig; 2

(8.10)

where the average value of the two measured signals (denoted by brackets) is calculated on a large collection of measured samples. In practice, the removal of the constant terms of each collection of samples can be performed by AC-coupled photodetectors, since they naturally remove any timeindependent offset on the measured signals.

8.2.2 Various implementations of coherent detection for optical performance monitoring For optical performance monitoring, various implementations of coherent detection are now described in a schematic way to emphasize their general properties (Figure 8.3).

Signal

CW LO

Pulsed LO

Signal

Frequency (a)

Signal

Pump pulse

Frequency (b)

Idler

CW LO

Frequency (c)

FIGURE 8.3 Spectral representation of the sources involved in various implementations of linear optical sampling. (a) the Monochromatic local oscillator is spectrally overlapped with the signal. (b) the Broadband pulsed local oscillator is overlapped with the signal. (c) the Nonlinear interaction between the short pump pulse and the signal leads to an idler pulse, and the monochromatic signal is overlapped with the generated idler.

200

CHAPTER 8 Optical performance monitoring based on linear optical sampling

8.2.2.1 Coherent detection with a monochromatic local oscillator The local oscillator can be a monochromatic source at the optical frequency oDATA (Figure 8.3(a)), with ELO(t) ¼ exp(ioDATAt) (cases where the local oscillator frequency is not exactly oDATA or is time varying are treated later).16,17 With photodetectors having an impulse response faster than the temporal features of the data electric-field envelope, the measured signal is ð þ1 SðtÞ ¼jEDATA ðtÞjexp½icDATA ðtÞ RðtÞdt: (8.11) 1

The signal measured at time t is therefore a representation of the data electric field at that time. A continuous representation of the electric field can be obtained by continuously measuring the signals SA and SB and combining them. This approach is identical to regular homodyne coherent detection in telecommunication systems, except that in practice, no phase-locking is required provided that the phase can be tracked numerically. The bandwidth of a diagnostic based on coherent detection with a monochromatic local oscillator is fundamentally limited by the bandwidth of the photodetectors since the role of the monochromatic LO is only to provide phase sensitivity.

8.2.2.2 Coherent detection with a short optical pulse If the local oscillator is a short optical pulse (Figure 8.3(b)), it can be used to measure the electric field of the signal with high temporal resolution.18,19 If the impulse response of the photodetectors is much longer than the duration of the local oscillator pulse, ð þ1 EDATA ðtÞELO ðtÞdt: (8.12) SðtÞ ¼ RðtÞ 1

The signal therefore has the shape of the photodetection impulse response and a complex amplitude proportional to the overlap of the electric fields of the data source and LO pulse. The two quadrature signals can be sampled with low-speed, analog-to-digital converters to obtain a highresolution sample of the electric field. The diagnostic’s bandwidth is set by the duration of the local oscillator, and not by the photodetectors. Sources with rates as high as 640 Gb/s have been characterized with a pulsed local oscillator. The data electric field is gated by the interference with the LO, and LOS is therefore fundamentally different from other sampling techniques for which gating of the intensity is performed. Samples are only obtained when a local oscillator pulse is present in the interferometer. This situation is common when sampling rapidly varying signals: high-bandwidth gates are usually obtained at a somewhat low repetition rate. Multiple samples are obtained when a train of sampling pulses is used: each pulse samples the data at a different time and a statistical representation is obtained by combining the measured samples. Photodetectors with response slower than the duration of the sampling pulses but faster than the period of the sampling train must be used to ensure independent measurements of successive samples of the data field. To understand some fundamental properties of LOS with a short local oscillator, it is particularly Ð þ1 instructive to consider the integral 1 EDATA ðtÞELO ðtÞdt in the frequency domain: ð þ1 ð þ1 do  : (8.13) EDATA ðtÞELO ðtÞdt ¼ E~DATA ðoÞE~LO ðoÞ 2p 1 1

8.2 LOS principle and properties

201

Equation (8.13) shows that a LOS signal is obtained only if the local oscillator spectrum overlaps with the data spectrum. This property arises from the low-bandwidth homodyne photodetection: Since only slow signals can be measured by time-integrating photodetectors, only the interference of one optical frequency of the test source with the same optical frequency of the local oscillator participates in the signal. Consequently, when the local oscillator has a constant spectral density over the spectral support of the data source, Equation (8.13) simplifies to ð þ1 ð þ1 do ¼jEDATA ð0Þj exp½icDATA ð0Þ; EDATA ðtÞELO ðtÞdt / E~DATA ðoÞ 2p 1 1 which means that the sample measured by a short local oscillator arriving at t ¼ 0 is directly proportional to the data temporal electric field at t ¼ 0. The influence of the LO field can be calculated directly from Equation (8.13); this is done in Reference 20 for a Gaussian LO with a second-order dispersion.

8.2.2.3 Phase-preserving temporal gating and monochromatic local oscillator Sampled coherent detection has also been implemented with distinct gating and homodyning sources (Figure 8.3(c)). In this case, the gating source must provide the temporal resolution while preserving phase information. An appropriate process for this operation is four-wave mixing, for example, implemented in a highly nonlinear fiber.21 The data signal EDATA(t) is mixed with a high-energy pump pulse EPUMP(t) and generates an idler signal described by the electric field EIDLER ðtÞ ¼ E2PUMP ðtÞEDATA ðtÞ ¼ IPUMP ðtÞj EDATA ðtÞj exp½2icPUMP ðtÞ  icDATA ðtÞ  2ioPUMP t þ ioDATA t:

(8.14)

The idler is therefore generated around the optical frequency oIDLER ¼ 2oPUMP – oDATA, and its phase verifies cIDLER(t) ¼ 2cPUMP(t) – cDATA(t), therefore ensuring that the data phase is related to the measured idler phase. When the pump pulse is significantly shorter than the data signal, the idler field simplifies to EIDLER ðtÞ ¼ IPUMP ðtÞj EDATA ð0Þj exp½2icPUMP ðtÞ  icDATA ð0Þ  ioIDLER t:

(8.15)

Coherent detection can be performed using a local oscillator around the frequency oIDLER and a photodetector with bandwidth higher than the repetition rate of the sampling source but lower than the bandwidth of the idler signal to obtain the signal: ð þ1 IPUMP ðtÞ exp½2icPUMP ðtÞdt: (8.16) SðtÞ ¼ RðtÞjEDATA ð0Þj exp½icDATA ð0Þ 1

The signal of Equation (8.16) can be sampled with a slow analog-to-digital converter to obtain a sample of the complex conjugate of the data electric field. In this approach, temporal resolution is provided by the pump pulse, while phase sensitivity is provided by the monochromatic LO. This can be used to decouple the requirements imposed on these two sources.

8.2.3 Polarization and wavelength sensitivity Because sampled coherent detection is based on the linear interference of the signal with the local oscillator, the two sources must be copolarized when reaching the photodetectors for optimal operation, although they might have different polarization states at the input or inside the diagnostic, such

202

CHAPTER 8 Optical performance monitoring based on linear optical sampling

as when different polarization states are used in a polarization interferometer. This point is of particular practical importance since polarization mode dispersion in the non-polarization-maintaining optical fibers used in deployed optical networks leads to time-varying polarization states. Laboratory experiments usually rely on polarization controllers to adjust the polarization state of the test source, but for complete characterization of a source in the general case, polarization diversity—that is, a setup characterizing the source along two orthogonal states—must be implemented. In LOS, two identical setups operating on two polarization states of the data field can be used. Polarization sensitivity is common in other high-speed optical sampling techniques, since most nonlinear effects are polarization dependent. Coherent detection is also sensitive to the spectral overlap of the test source and local oscillator. With a monochromatic local oscillator, if the local oscillator frequency departs significantly from the center of the spectrum of the data source, high-frequency oscillations occur in the signal to be measured, which increases the bandwidth requirement for the photodetectors. With the low-speed photodetectors used in sampled coherent photodetection with a pulsed local oscillator, the local oscillator must have a spectrum encompassing the spectral support of the test source. In the absence of overlap, no signal is observed. This can be used to selectively characterize wavelength-multiplexed channels: Following Equation (8.13), the spectral support of the LO can be chosen to overlap with the spectral support of a given channel to characterize this channel. With the absence of spectral overlap with other channels present at the input of the diagnostic, these channels are not sampled.

8.2.4 Phase sensitivity 8.2.4.1 Monochromatic local oscillator Unless the same optical source is used to generate the data signal and the local oscillator (which constrains the transmitter and diagnostic to be collocated) or active phase-locking is used, the monochromatic local oscillator beating with the test source does not exactly cancel the carrier frequency of the test source. For a monochromatic frequency oLO and frequency noise doLO(t), the electric field of the local oscillator is described by ðt doLO ðt0 Þdt0 : (8.17) ELO ðtÞ ¼ exp½ioLO t  i 1

Ðt Identically, the data have their own phase noise represented by the integral  1 doDATA ðt0 Þdt0 . Equation (8.11) can be written as ðt ½doLO ðt0 Þ  doDATA ðt0 Þdt0 g; (8.18) SðtÞ ¼j EDATA ðtÞ j expficDATA ðtÞ þ iðoLO  oDATA Þt þ i 1

where a proportionality constant has been omitted. Equations (8.18) and (8.11) differ by a linear temporal phase due toÐ the difference in the average optical carrier of the two sources (oLO – oDATA)t t and by a noise term 1 ½doLO ðt0 Þ  doDATA ðt0 Þdt0 . The carrier-difference term is a systematic rotation at a constant rate in the complex plane and the noise term is a slowly-varying function, so that the induced phase terms can in practice be tracked and removed, therefore allowing the recovery of the data phase.

8.2 LOS principle and properties

203

8.2.4.2 Pulsed local oscillator For a pulsed local oscillator, one considers a train of sampling pulses described by the electric field 1 X n¼1

ðt j ELO ðt  nTÞj exp½icLO ðt  nTÞ  ioLO t  i

doLO ðt0 Þdt0 :

(8.19)

1

This train of pulses is represented by a periodic modulation with period T of a high-frequency carrier with mean frequency oLO. Because oLOT is not necessarily a multiple of 2p, the electric field of Equation (8.19) is not periodic even in the absence of noise. Only the envelope of the field is periodic, as pictured in Figure 8.4. Precise phase stabilization is possible,14 but digital processing allows implementing coherent detection without active stabilization. In the absence of phase noise (i.e., doLO being identically zero), in the referential of each pulse centeredÐ at t ¼ nT, the t phase difference between consecutive pulses is a constant. With the phase noise 1 doLO ðt0 Þdt0 , there is a slow drift of the phase when considering the time interval corresponding to a large number of sampling pulses. For all practically useful sources, there is no significant random change of the instantaneous frequency over the temporal support of each sampling pulse and across a sequence of successive pulses. When the photodetectors have bandwidth higher than the repetition rate of the train of pulses but much slower than the duration of the sampling pulses, the output is a collection of discrete samples corresponding to the sampling of the electric field of the test source at times nT. Using Equations (8.12) and (8.19), the nth sample is proportional to ð nT ½doLO ðt0 Þ  doDATA ðt0 Þdt0 g: Sn ¼j EDATA ðnTÞ j expficDATA ðnTÞ þ inTðoLO  oDATA Þ þ i 1

(8.20) The discrete samples measured with a pulsed local oscillator at a low repetition rate T are therefore expressed similarly to the samples measured with a monochromatic local oscillator and continuous acquisition. With a pulsed local oscillator, LOS samples are therefore samples of the electric field of the data source measured at a rate given by the repetition rate of the sampling source. It is obvious that j Sn j2 ¼ IDATA ðnTÞ;

(8.21)

which means that the intensity of the data source is perfectly sampled if the local oscillator spectrum encompasses the spectrum of the data signal.22,23 However, T

T

Time

FIGURE 8.4 Train of sampling pulses in the time domain showing the carrier-phase evolution under the field envelope.

204

CHAPTER 8 Optical performance monitoring based on linear optical sampling

argðSn Þ ¼ cDATA ðnTÞ þ ðoLO  oDATA ÞnT þ

ð nT 1

½doLO ðt0 Þ  doDATA ðt0 Þdt0 :

(8.22)

The phase (oLO – oDATA)nT induces a rotation of the samples in the complex plane occurring at a constant rate. One should note that unless special care is taken in setting oLO ¼ oDATA, these quantities are different, and rotation is therefore boundÐ to occur. The finite linewidth of the two sources nT impacts the recovered phase through the integral 1 ½doLO ðt0 Þ  doDATA ðt0 Þdt0 . This integral modifies the rotation rate since the induced rotation rate (angle per sample) is approximately equal to T [doLO(nT) – doDATA(nT)]. Phase tracking can be used to remove these effects. However, it should be noted that since samples are measured at a much lower rate than with continuous sampling, phase tracking is more difficult for identical linewidth of the considered sources. With continuous sampling, a sampling rate of the order of 10 GSamples/s is possible, while mode-locked sources appropriate for pulsed coherent detection usually have repetition rates around 100 MHz.

8.2.4.3 Pulsed gate and monochromatic local oscillator For a pulsed sampling source described by Equation (8.19) and a local oscillator described by Equation (8.17), the measured signal is Ð nT Sn ¼ jEDATA ðnTÞj expficDATA ðnTÞ  inTð2oPUMP  oDATA  oLO Þ  i 1 : (8.23) ½2doPUMP ðt0 Þ  doDATA ðt0 Þ  doLO ðt0 Þdt0 g: Identical considerations relate to the phase terms of Equation (8.23): There is a constant-rate rotation Ð nT induced by the term nT(2oPUMP – oDATA – oLO) and phase drifts due to 1 ½2doPUMP ðt0 Þ doDATA ðt0 Þ  doLO ðt0 Þdt0 . Each optical source contributes to the noise, and there are therefore three sources of phase noise in this case, but phase tracking techniques identical to other implementations of coherent detection can be used.

8.2.5 Digital phase tracking The collection of complex samples obtained with coherent sampled detection can be expressed as Sðtn Þ ¼jEDATA ðtn Þj exp½icDATA ðtn Þ þ iOtn þ iyðtn Þ;

(8.24)

where O is a rotation rate assumed constant over the collection of samples, y is a slowly-varying function, and tn is a temporal variable set by the sampling rate of the acquisition system. Since one is interested in the electric field of the data signal, methods for the removal of the phase terms Otn þ y(tn) must be used. Figure 8.5 illustrates the effect of the various terms of Equation (8.24) with a set of experimental data measured on a BPSK signal. In Figure 8.5(a), the measured phase before processing is shown. In Figure 8.5(b), the linear term has been removed, leading to a slowly-varying phase evolution on which the binary data modulation is superimposed. In Figure 8.5(c), the slowlyvarying phase has been removed, and the two phase levels separated by p can be seen. Digital phase tracking can be performed in various ways. One can observe, for example, that arg½Sðtnþ1 Þ  arg½Sðtn Þ ¼ ½cDATA ðtnþ1 Þ  cDATA ðtn Þ þ Oðtnþ1  tn Þ þ ½yðtnþ1 Þ  yðtn Þ:

(8.25)

For phase-encoded signals with discrete phase levels (e.g., BPSK, QPSK, . . .), the first term in the rightside of Equation (8.25) corresponds to discrete differential phase levels. The second term is a

8.3 Implementations of LOS

205

Phase (rad)

3 2 1 0 −1 −2 −3

0

100

200 300 Sample

400 0

(a)

100

200 300 Sample

400 0

100

(b)

200 300 Sample

400

(c)

FIGURE 8.5 Phase samples measured on a 10-Gb/s BPSK signal. The phase is shown (a) without any processing, (b) after removal of a linear term leading to rotation in complex plane, and (c) after removal of both a linear term and a slowly-varying phase.

constant. The third term is a slowly-varying function of the sample number. Since the differential phase levels are discrete, the samples of Equation (8.25) corresponding to a single level chosen arbitrarily can be isolated. The evolution of the phase of the isolated subset can be tracked using a slowly-varying function, which leads to the function O(tnþ1 – tn) þ y(tnþ1) – y(tn). This determined phase function can be subtracted from the differential phase before concatenation, which leads to the estimated phase cDATA(tn). If the modulation format is known and consists of M levels equally spaced in the interval [0, 2p] (e.g., M ¼ 2 for BPSK and M ¼ 4 for QPSK, with each level corresponding to an Mth root of 1), the rotation and slowly-varying phase drifts can be removed by considering the quantity M arg[S(tn)], which is the phase of [S(tn)]M, written as M arg½Sðtn Þ ¼ McDATA ðtn Þ þ MOtn þ Myðtn Þ:

(8.26)

By definition, the phase modulation for ideal noiseless data verifies McDATA(tn) ¼ 0, so that Equation (8.26) can be lowpass filtered to obtain the phase drift between sources. (Note that noise on the measured data signal is assumed to be uncorrelated between samples and should not be filtered.) The phase drift can be subtracted from the data to obtain the phase of the sampled complex interference cDATA(nT). Techniques for phase tracking are described extensively in Reference 24. Phase tracking is made easier if the phase variations of the two sources are small during the acquisition time of a representative sample set. Because samples are acquired at the rate of the analog-to-digital converters in the case of a monochromatic LO and at the repetition rate of the sampling source when using a pulsed LO, phase tracking should generally be easier with monochromatic sources. As a general rule, it is preferable to operate with sources having a long coherence time (i.e., narrow linewidth for a monochromatic LO, or narrow spectral linewidth of each optical frequency of the spectral comb for a periodic sampling source) and sampling sources with a high repetition rate, all other things being kept equal.

8.3 IMPLEMENTATIONS OF LOS This section describes various experimental implementations and practical aspects of LOS.

206

CHAPTER 8 Optical performance monitoring based on linear optical sampling

8.3.1 Balanced photodetection If balanced photodetectors with adequate bandwidth are available, a 90-degree optical hybrid can be used. One practical advantage of balanced photodetection is that the noninterferometric terms (i.e., the intensities of each source) do not need to be sampled—that is, the entire dynamic range of the analog-to-digital conversion is used to sample the phase-sensitive signal. The 90-degree optical hybrid generally uses distinct optical paths as represented in Figure 8.2. These optical paths can be implemented with free-space propagation, optical fibers, or waveguide. Stability of the phase difference between different electric fields must be ensured during the acquisition of a representative set of samples. This can usually be ensured without active stabilization, even with optical fibers.16

8.3.1.1 Polarization 90-degree optical hybrid Polarization interferometers are commonly used in interferometric measurements and have been used for coherent detection.23,25 They use two orthogonal polarization modes to implement an environmentally stable interferometer. A polarizer with axes rotated by 45 degrees relative to the input states projects the two states on the two output ports of the polarizer, therefore acting like a beam splitter. A relative phase shift between two orthogonally polarized optical waves can be introduced by waveplates properly aligned. Advantageously, since two orthogonally polarized sources can propagate along the same optical path, phase fluctuations due to turbulence or thermal drift have little impact on the interference that occurs after projection along a common axis. Polarization states can be used in an optical hybrid, for example, with the following implementation. Let’s consider two orthogonal axes x and y and the combination of the local oscillator and data source with linear orthogonal polarization states: ˘

˘

˘

˘

˘

E ¼ ELO x þ EDATA y:

(8.27)

˘

˘

˘

This field is split into two identical fields E A ¼pEffiffiBffi ¼ p1ffiffi2 E with p a nonpolarizing beam splitter. Sendffiffiffi ing EA to a polarizer with axes along ð x þ yÞ= 2 and ð x  yÞ= 2 leads to the fields 1 E A;1 ¼ ½ELO þ EDATA ð x þ yÞ; (8.28) 2 and 1 E A;2 ¼ ½ELO  EDATA ð x  yÞ: (8.29) 2 ˘

˘

˘

˘

˘

˘

˘

˘

˘

˘ ˘

Balanced square-law photodetection of these two fields gives SA ¼ Re½EDATA ELO :

(8.30)

˘

EB propagates into a quarter waveplate aligned to induce a relative p/2 phase shift between the two orthogonally polarized fields. After the waveplate, one has 1 E B ¼ pffiffiffi ½iELO x þ EDATA y: (8.31) 2 pffiffiffi pffiffiffi Balanced photodetection after a polarizer oriented along ð x þ yÞ= 2 and ð x  yÞ= 2 gives ˘

˘

˘

˘

˘

˘

SB ¼ Im½EDATA ELO :

(8.32)

˘

8.3 Implementations of LOS

3-dB coupler

207

p /2 phase l /4

l /2

Data source Sampling source

l /2

Polarizer Balanced photodetector

FIGURE 8.6 Schematic of a polarization interferometer. The orthogonally polarized data and sampling sources are combined with a nonpolarizing 3-dB coupler. The two outputs of the coupler are sent to identical setups leading to a pair of balanced photodetectors, excepting that a p/2 phase shift is introduced in one arm, so that the real and imaginary parts of the interference are measured.

A schematic of a polarization interferometer achieving these operations is presented in Figure 8.6, based on an implementation of LOS previously published.23 The two input sources have orthogonal polarization states and are split and coupled by a nonpolarizing 3-dB coupler. The two outputs are independently processed to yield the two balanced photodetection signals. In the lower optical path, a 45-degree polarization rotation is performed so that a polarizer combines the two sources along its horizontal and vertical axes. In the upper path, a differential p/2 phase shift is added on one of the sources by propagation through a quarter-wave plate with axes horizontal and vertical. A 45-degree polarization rotation is then performed by a half-wave plate, and balanced photodetection is performed on the two outputs of a polarizer. Since operations on polarization states can be performed easily by combinations of birefringent elements, polarization interferometers suitable for coherent sampling can be integrated.

8.3.1.2 Waveguide 90-degree optical hybrid Waveguide structures are attractive for compact stable implementations of optical hybrids. In the 1.5- and 1.3-micron telecommunication regions, waveguides with low linear losses can be fabricated using, for example, Silicon-on-Insulator. A 90-degree waveguide optical hybrid (Figure 8.7(a)) exactly replicates the splitting and recombining shown in Figure 8.2.18 The waveguide optical hybrid has been used in a LOS setup to sample a data source with a copolarized free-running mode-locked laser (Figure 8.7(b)). Each source is independently split and recombined, and the sampling signals are measured by two pairs of balanced photodetectors. The repetition rate of the sampling laser was 10 MHz, and the bandwidth of the photodetectors was 800 MHz. The broadband mode–locked laser (40 nm) was spectrally filtered to match the spectrum of the data source, typically with a 3-nm filter. The analog-to-digital conversion of the photodetector signals is synchronized with the sampling events by sending part of the input sampling pulse onto a photodiode. Since the optical path length differences between splitting and recombining of the different sources are smaller than the temporal duration/resolution, it is necessary and sufficient to control the optical phase of one of the split sources to get the real and imaginary part of the same complex interference. This can

208

CHAPTER 8 Optical performance monitoring based on linear optical sampling

Sampling source Data source

Pulse generator

Hybrid Filter

SA

BDA

PC

A/D

SB

BDB V

(a)

(b)

FIGURE 8.7 (a) Picture of a 90-degree optical hybrid made with silicon-on-silica. (b) Setup for coherent photodetection of a data source with a copolarized pulsed local oscillator using the waveguide 90-degree optical hybrid. The sampling source is spectrally filtered to match the optical spectrum of the data source. The relative phase between the two measured signals is controlled by applying a low continuous voltage to a thermo-optic coupler.18 Copyright © 2005 OSA.

be achieved using the electro-optic effect (change of index induced by a voltage) or by a thermooptic effect (change of index controlled by the temperature). In the latter case, a thermo-optic coupler is driven by a control voltage, so that a change in voltage effectively controls the relative phase. This implementation was used to characterize various phase-encoded sources at 10 Gb/s and 40 Gb/s, as well as OOK optical sources at rates as high as 640 Gb/s.

8.3.2 Direct photodetection If balanced photodetectors are not available, only two ports of a 90-degree optical hybrid need to be used. In this case, a simpler structure where the data source is, for example, circularly polarized while the sampffiffiffi 21 pling source is linearly pffiffiffi polarized can be used. With the input field EðtÞ ¼ ELO ðtÞð x þ i yÞ= 2 þEDATA ðtÞð x þ yÞ= 2, the fields at the two outputs of a polarizing beam splitter with axes x and y are ˘ ˘

˘

˘

˘

˘

SA;1 ¼

1 1 1 j ELO þ EDATA j2 ¼ j ELO j2 þ j EDATA j2 þ Re½EDATA ELO ; 2 2 2

(8.33)

SB;1 ¼

1 1 1 j iELO þ EDATA j2 ¼ j ELO j2 þ j EDATA j2 þ Im½EDATA ELO : 2 2 2

(8.34)

˘

and

Therefore, the real and imaginary part of the interference can be obtained by considering a large collection of samples and removing the average value of the two measured signals, following Equation (8.10). In practice, this removal can be performed by AC-coupled photodetectors.

8.3.3 LOS with four-wave mixing Coherent detection of a data-encoded signal can be performed with an independent optical gate and local oscillator, provided that the optical gating mechanism preserves the phase information—that is, it is possible to retrieve the data phase from the measured phase of the gated signal. This was implemented using four-wave mixing in a highly nonlinear fiber, a process that can be phase matched to provide high temporal resolution and large spectral coverage.21,26 As described previously, the phase

8.3 Implementations of LOS

209

lpump = 1547 nm 1.5625 GHz TFWHM = 4.5 ps Sampling pulse source

LO PC

HNLF

PC PBS

2 nm

Data source L = 50 m = 10 W–1km–1 idler = 1530 nm

PC

ADC 8 bits ADC 8 bits

DSP (offline)

FIGURE 8.8 Setup for the coherent photodetection of a data source by gating with four-wave mixing and detection with a monochromatic local oscillator. The sampling and data source are combined and propagate in a highly nonlinear fiber. The idler resulting from four-wave mixing of the sampling source acting as a pump on the data source is filtered and detected by homodyne detection with a local oscillator.21 Courtesy of Mathias Westlund and Peter A. Andrekson; copyright © 2009 OSA.

of the generated idler is related to the phase of the data by cIDLER(t) ¼ 2cPUMP(t) – cDATA(t). This phase can be measured by coherent photodetection with a monochromatic local oscillator having an optical frequency close to oIDLER ¼ 2oPUMP – oDATA. A suitable setup is presented in Figure 8.8. The copolarized data and sampling source are combined and propagate in a highly nonlinear fiber at a power sufficient to generate a measurable idler. The sampling rate is set by the repetition rate of the sampling source (1.5 GSamples/s). The idler is spectrally filtered and is detected by coherent photodetection with a monochromatic laser and two photodetectors. The linearly polarized idler and circularly polarized local oscillator are combined and sent to a polarizing beam splitter, and direct photodetection is performed on each output port of the splitter to obtain the real and imaginary part of the interference signal. Software synchronization was performed to yield samples of the electric field of the data source over the complete bit sequence of a PRBS signal. Examples of experimental results are presented in Figure 8.9 for a 10-GBaud QPSK signal. The constellation diagram clearly shows the four symbols separated by p/2. These four levels are also clearly visible on the plots of the measured phase as a function of the location in the bit sequence, which were obtained by software synchronization.

8.3.4 Correction of imperfections 8.3.4.1 Quadrature amplitude correction When the data and sampling sources are not mutually coherent, the phase variations between the two sources allow correction of experimental imperfections in the optical setup or photodetectors. For the signals SA and SB, one can show that hSA i ¼ hSB i ¼ 0;

(8.35)

hS2A i ¼ hS2B i:

(8.36)

and

210

CHAPTER 8 Optical performance monitoring based on linear optical sampling

Optical phase

p 100 ps p /2 0 –p/2 Time

Optical phase

(a)

(c)

p p /2 0 –p/2 0

Time

27 − 1 bits

(b)

FIGURE 8.9 Experimental results obtained with four-wave mixing and coherent detection. (a) Constellation diagram of 625,000 samples measured on a 10-GBaud QPSK signal. Electric-field samples located at the center of the bit slots in a time interval corresponding to 20% of the bit period are plotted in black. Other samples appear in gray, while lines correspond to interbit transitions averaged over a large number of similar transitions in the bit sequence. (b) Measured phase as function of position in the bit sequence. (c) Close-up of (b) in a 10-bit intervals.21 Courtesy of Mathias Westlund and Peter A. Andrekson; copyright © 2009 OSA.

A large collection of samples {SA} and {SB} measured over a time interval much longer than the mutual coherence time of the two sources can therefore be corrected by removal of the DC part of each collection, therefore ensuring hSAi ¼ hSBi ¼ 0, and normalization by its standard deviation, therefore ensuring hS2A i ¼ hS2B i.

8.3.4.2 Quadrature phase correction For two mutually incoherent sources, the relative phase shift D’AB between the two signals SA and SB verifies cosðD’AB Þ ¼

2hSA  SB i : S2A þ S2B

(8.37)

This can be used experimentally to tune the differential phase between the measured signals SA and SB and set it to its optimal value p/2, or numerically correct imperfections.18,27 The sampled complex

8.4 Optical performance monitoring with LOS

211

signal can always be written as the sum of the two measured signals as long as their relative angle is known and not a multiple of p. If the adjustment of the relative phase is not experimentally possible, the complex signal z can be calculated using the measured signals, which are SA ¼ Re(z) and SB ¼ Re[z exp(iD’AB)], to obtain z ¼ ReðzÞ þ i ImðzÞ SB  cosðD’AB ÞSA ¼ SA þ i : sinðD’AB Þ

(8.38)

This should be restricted to large values of sin(D’AB) (i.e., D’AB close to (p/2), since noise on the measured signals strongly impacts the noise on the measured complex interference for small values of this quantity.

8.4 OPTICAL PERFORMANCE MONITORING WITH LOS This section presents various examples of optical performance monitoring performed with coherent detection.

8.4.1 Characterization of amplified spontaneous emission Amplified spontaneous emission (ASE) is common in optical telecommunication systems since all amplifiers have a finite signal-to-noise ratio. ASE is not correlated to the (noise-free) signal, and the electric field of the noisy signal is ESIGNAL ðtÞ ¼ EDATA ðtÞ þ EASE ðtÞ;

(8.39)

where the phase of EASE is uniformly distributed in the interval [0,2p] and the modulus of EASE has a normal distribution. The symbols of the noisy signal are therefore isotropically spread in the complex plane. In Figure 8.10(a), the constellation diagram of an ASE source is plotted, in which case EDATA(t) ¼ 0. In Figure 8.10(b), the constellation diagram of a continuous-wave (CW) source coupled with an ASE source is plotted. In Figures 8.10(c) and (d), the constellation diagrams of a BPSK signal with two different OSNRs are plotted. As the OSNR is decreased, the spread of the symbols increases in the complex plane. Noting that both the variance of the noise on each symbol and the OSNR are proportional to a bilinear function of the electric field of the ASE, a linear relation between these two quantities is expected, and can indeed be derived. Figure 8.10(e) demonstrates the link between variance of the sampled electric field of a given symbol along the real and imaginary axis and a linear function of the OSNR. A good agreement is obtained down to an OSNR of the order of 25 dB in this particular implementation.

8.4.2 Phase and amplitude noise measurements From LOS data, the amplitude and phase noise on a symbol can be obtained simply by considering the amplitude and phase spread of the symbols in the complex plane. Exemplary constellation diagrams of phase-encoded sources are shown in Figure 8.11. These phase-encoded signals were generated by directly driving a LiNbO3 phase modulator with a two-level PRBS drive. Tuning the

212

CHAPTER 8 Optical performance monitoring based on linear optical sampling

(a)

(b)

(c)

(d)

1

sx 2 sy 2

Variance

10–1

Theory 10–2 10–3 10–4

5

10

15

20

25

30

35

40

45

OSNR (dB) (e)

FIGURE 8.10 Constellation diagrams measured (a) on an ASE source, (b) on a monochromatic source with an OSNR of 15.8 dB, (c) on a binary PSK signal with an OSNR of 18 dB, and (d) on a binary PSK signal with an OSNR of 13 dB. (e) Measured variance of one of the symbols along the real and imaginary axes s2x and s2y versus the measured OSNR. Line of dots represents the theoretical relation between variance and OSNR.19 Copyright © 2006 OSA.

voltage difference between the two levels changes the relative optical phase of the two levels and increases the noise on each level, as can be observed in Figures 8.11(a) and (b). However, since only phase modulation is performed, the amplitude noise does not depend on the amount of phase modulation. Figure 8.11(c) presents a quantification of the phase and amplitude noise as a function of the phase difference between the two encoded levels: the phase noise increases linearly while the amplitude noise is constant, as expected. In practice, this approach to the generation of phase-encoded signals is avoided because of the inherent phase noise induced by the noise on the drive voltage. It is preferred to use a Mach-Zehnder modulator and benefit from the sign change of the transfer function when going through extinction. This provides a phase shift exactly equal to p regardless of the modulation amplitude, and amplitude noise can be minimized by proper tuning of the amplitude of the modulation. These properties are demonstrated in Figure 8.12. The two constellation diagrams plotted in Figures 8.12(a) and (b) were measured for AC drive voltages with different amplitudes, the bias of the Mach-Zehnder modulator being set for extinction. A p phase shift between levels is obtained regardless of the drive amplitude, and the phase noise does not depend significantly on the drive amplitude. However, the amplitude noise decreases as the drive amplitude is increased,

8.4 Optical performance monitoring with LOS

(b)

0.3

0.3

0.2

0.2

0.1

0.1

sy

sr

(a)

213

0

0 0

1

2 Phase (rad) (c)

3

4

FIGURE 8.11 (a, b) Constellation diagrams of phase-modulated signals generated with a phase modulator for differential phase of p/2 and p. (c) Standard deviation of the amplitude sr and phase sc of one of the symbols as a function of the differential phase (respectively round markers and square markers).19 Copyright © 2006 OSA.

as the sinusoidal transfer function of the modulator clamps the amplitude modulation when the voltage amplitude is sufficient (Figure 8.12(c)). The direct characterization of the amplitude and phase noise is important in many aspects of optical telecommunications. Taking as an example the BPSK regeneration experiment described in Reference 28, the wavelength-conversion process modifies the amplitude and phase noise properties of the signal. A performance assessment of the channel in terms of bit error rate only leads to an indirect understanding of the properties of the converter. Figure 8.13 demonstrates the phaseregeneration capability of the proposed setup when operating on signals with phase noise. The constellation diagrams measured on the input signal (Figures 8.13(a) and (c)) and on the converted signal (Figures 8.13(b) and (d)) demonstrate a reduction of the phase noise, which could lead to an appreciable transmission performance improvement. The standard deviation of the phase is reduced from 0.3 rad to 0.14 rad after conversion (Figures 8.13(c) and (d)).

8.4.3 Nonlinear phase noise Because of the intensity dependence of the index of materials, optical waves propagating in long distances of optical fibers accumulate a nonlinear phase shift. The effect of this phase shift depends on the modulation format, power into the fiber spans, chromatic dispersion of the fiber spans, and more. In general, the propagation of a source with instantaneous power p(t) generates an optical phase

214

CHAPTER 8 Optical performance monitoring based on linear optical sampling

(b)

0.15

0.15

0.1

0.1

0.05

0.05

sy

sr

(a)

0 0

2

4

6

8

0 10

Voltage (V) (c)

FIGURE 8.12 (a, b) Constellation diagrams of BPSK signal generated with Mach-Zehnder modulator for two different amplitudes of the drive voltage. (c) Standard deviation of the amplitude and phase of one of the symbols as a function of the drive voltage (respectively round markers and square markers).19 Copyright © 2006 OSA.

(a)

(b)

(c)

(d)

FIGURE 8.13 Constellation diagrams measured at input and output of wavelength converter set to reduce the phase noise of a BPSK signal. (a) and (c) are measured before the wavelength converter, while (b) and (d) are the corresponding converted signals.28 Copyright © 2008 IEEE.

equal to gLp(t) by self-phase modulation in a medium of length L and nonlinear coefficient g. This process therefore couples the temporal phase and amplitude of the electric field. The ASE-signal beat noise of a source with finite OSNR leads to random amplitude fluctuations of the field, which induce nonlinear phase noise during nonlinear propagation (Gordon-Mollenauer phase noise).29 An example of the BPSK constellation diagram with such noise is displayed in Figure 8.14. Figure 8.14(a)

8.4 Optical performance monitoring with LOS

(a)

215

(b)

0.4

C (rad)

0.3 0.2 0.1 0 0

10

20 30 Power (mW)

40

50

(c)

FIGURE 8.14 Constellation diagrams measured after propagation of a noisy signal in a highly nonlinear fiber at (a) low power and (b) high power. The coupling between the intensity and phase of the samples is a sign of GordonMollenauer phase noise. This coupling is quantified in (c) as a function of the average power of the source.19 Copyright © 2006 OSA.

represents the constellation diagram measured after propagation in a nonlinear fiber at low power, and the reduced OSNR is indicated by the isotropic spread of the symbols. Figure 8.14(b) represents the constellation diagram measured after propagation in the same fiber at a power sufficient to induce significant nonlinearities. A clear coupling between phase and amplitude of the samples measured for each symbol is observed. The coupling between intensity and phase of the samples can be quantified with the measured constellation diagrams. For example, one can define C as the slope of the fit of the measured phase versus the measured intensity for a collection of samples (the intensity being normalized to an average value equal to one). As can be seen in Figure 8.14(c), the coupling increases as the average input power into the fiber increases. This is expected from the linear relation between induced nonlinear phase and power in the nonlinear medium. Figure 8.15 demonstrates the use of LOS with a short local oscillator on a 20-Gb/s QPSK signal propagating in a recirculating loop. The constellation diagrams of the transmitted signal were measured for different input average power and number of roundtrips in the loop (i.e., propagation distance in the transmission fiber). A typical constellation diagram after propagation is displayed in Figure 8.15(a), showing a clear coupling between intensity and phase of the four symbols of the QPSK signal. The amplitude of the coupling is expected to increase with the injected power and the transmission distance. This is confirmed by the plot presented in Figure 8.15(b), which

216

CHAPTER 8 Optical performance monitoring based on linear optical sampling

0.5

C (rad)

0.4 0.3 0.2

800 km 1200 km

0.1 0 0

1000 2000 3000 Power × distance (mW.km)

(a)

(b)

FIGURE 8.15 (a) Constellation diagram of a QPSK signal measured after two roundtrips in a recirculating loop (the propagation distance in transmission fiber is 800 km). Gordon-Mollenauer phase noise can be seen. (b) Coupling between intensity and phase in constellation diagrams measured for various propagation distances and/or launch powers.

shows the measured phase-intensity coupling as a function of the product of the input power and transmission distance. The data in this plot were obtained with seven constellation diagrams, measured for either two or three roundtrips in the recirculating loop (propagation distance of 800 km and 1200 km in transmission fiber, respectively). Direct detection and quantification of nonlinear phase noise is made possible by phase-sensitive sampling.

8.4.4 Nonlinear phase-shift measurement Because of its ability to measure both intensity and phase of an optical wave, LOS can directly measure the nonlinearities introduced by optical fibers and components. One way to accomplish this is by propagation of a modulated wave into a test fiber at different optical powers and comparison of the measured output temporal phase to the measured output temporal intensity.30 Figure 8.16 shows a comparison of the measured power and phase of a modulated signal after propagation in a nonlinear fiber. Self-phase modulation leads to proportionality of these two quantities, and the nonlinear coefficient of the fiber can be recovered after proper scaling.

8.4.5 Digital processing of sampled electric field When operated at a sufficiently high sampling rate, coherent photodetection leads to continuous representations of the electric field. Since a continuous set of samples of the complex temporal field is obtained, various distortions can be directly measured on data-encoded signals, and digital postdetection compensation can be performed.16,17,31 Figure 8.17 shows 10-Gb/s constellation diagrams obtained with LOS implemented with a monochromatic laser,32 demonstrating the effect of chromatic dispersion. The two symbols are visible in the complex plane, and the constellation diagrams before (Figure 8.17(a)) and after chromatic dispersion (Figure 8.17(b)) are clearly different. In particular, chromatic dispersion leads to separation of the bit centers along a circle in the complex plane

8.4 Optical performance monitoring with LOS

217

0.4

–0.4

1

Δj (rad)

Δp

0

0 –1 0

10

20

30 40 Sample

50

60

FIGURE 8.16

3

3

2

2

1

1 Im(E)

Im(E)

Measured relative instantaneous power and phase of an amplitude-modulated optical source after propagation in a nonlinear fiber. The coupling between power and phase is due to self-phase modulation, and the nonlinear coefficient of the fiber can be quantified with these data.30 Copyright © 2005, Institution of Engineering and Technology.

0 –1

–1 –2

–2 –3 –3 –2 –1

0

0 1 Re(E) (a)

2

3

–3 –3 –2 –1

0 1 Re(E)

2

3

(b)

FIGURE 8.17 Examples of measured constellation diagrams of a 10.7-Gb/s BPSK signal (a) after the transmitter, and (b) after propagation in 90 km of standard single-mode fiber. The trajectories of the electric field in the complex plane are plotted with continuous gray lines and the values of the field at the center of the bit slot are plotted with a black round marker. Courtesy of Michael G. Taylor.

with different associated trajectories. Postdetection digital processing of the sampled electric field can be used to compensate impairments; for example, chromatic dispersion can be compensated by convolution with the impulse response corresponding to exp(i’(2) o2/2), where ’(2) is the accumulated second-order dispersion.16

8.4.6 Characterization of the electric field of periodic sources LOS can also characterize periodic sources, in the context of optical telecommunications or more generally in ultrafast optics.33 Assuming a period TTEST for the test source and TSAMPLING for the sampling source, samples of the intensity and phase of the source can be measured at the repetition

218

CHAPTER 8 Optical performance monitoring based on linear optical sampling

1

4 1

1

Phase (rad)

2

Intensity (a.u.)

3

0

Phase (rad)

Transmission (a.u.)

1

0 0

0

20

40

60

–1 80 100

0 0

5

10

15

Time (ps)

Time (ps)

(a)

(b)

20

–1 25

FIGURE 8.18 (a) Temporal transmission and phase of a semiconductor optical amplifier depleted by a short optical pulse. (b) Temporal intensity and phase of an optical pulse carved by an electro-absorption modulator. In (a), the period of the depleting pulse is 100 ps, while in (b), the period of the drive voltage is 25 ps. The lines correspond to the quantities measured with linear optical sampling and the markers correspond to the same quantities measured with the spectrogram technique.33 Copyright © 2005 OSA.

rate of the sampling laser—that is, every TSAMPLING. The samples can be assembled into a representation of the electric field of the test source in one single period by numerical resynchronization. Defining dt ¼ TSAMPLING – nTTEST, the collection of measured samples can be plotted with an axis having a sampling rate dt. Measurement of the amplitude of the correlation between two periodic laser sources has also been performed with two photodiodes.34 Figure 8.18 displays examples of electric field measurement in the context of pulse carving and device characterization. In Figure 8.18 (a), the transmission and phase of a semiconductor optical amplifier depleted by a short optical pulse are shown. The carrier depletion leads to a significant decrease of the transmission and a large induced phase. Recovery occurs with the transmission and phase reaching their predepletion levels. Figure 8.18(b) shows the intensity and phase of an optical pulse generated by carving a monochromatic laser with an electro-absorption modulator driven by a 40-GHz sinusoidal voltage. Phase modulation occurs at the same time as amplitude modulation, and it is of practical importance to understand the coupling between these two quantities. In these two cases, the results are in very good agreement with independent measurements performed with the linear spectrogram technique.8

8.5 RECENT RESULTS AND RELATED TECHNIQUES Various techniques based on coherent detection have been recently demonstrated. Two optical hybrids have been used to characterize differentially encoded signals by sampling two successive bits of the data stream. This is achieved by adding a relative delay equal to the bit period on the sampling source going into one of the hybrids, so that the difference between the phases measured by the first and second sampling system is essentially the phase difference between two successive bits.35,36 Because two phase samples separated by a small delay are compared, the requirement on the coherence time of the interfering sources is decreased, but only differential measurements

Acknowledgments

219

of the phase are obtained. An impairment such as nonlinear phase noise is clearly visible on samples of the electric field of the data source, but might not be as clearly identifiable with the differential phase. If the delay between the two sampling systems is decreased, the measured phase difference scales like the instantaneous frequency around the sampled times.27 A configuration with two optical hybrids has also been used to characterize optical devices.37 In that demonstration, one of the hybrids is used to provide a phase reference signal on a monochromatic laser, while the other hybrid measures the same laser after phase modulation. The two signals are essentially sampled by the same sampling laser at the same time, so that the fluctuating phases of the monochromatic laser and sampling laser exactly cancel out in the differential phase. This provides a measurement of the actual phase introduced, for example, by a modulator without a strong requirement on the coherence of the sampling source and monochromatic laser. High-resolution, high-precision measurements of optical components have also been obtained using coherent detection of two phase-locked femtosecond oscillators.38 Polarization diversity implementations of linear sampling have been demonstrated.39–41 In one of these implementations (Reference 41), sampling of two combined channels was performed, and the intensity of the independent channels was obtained by digital processing. A precision time base has also been used for word-synchronous sampling.42 This allowed measurement and averaging of various bit patterns. Software synchronization has also been used with some implementations of coherent sampling.21 Finally, direct coherent detection has seen impressive developments thanks partly to signal processing techniques that allow direct impairment compensation using digital processing of the electric field.31 The associated phase-tracking algorithms can be used for LOS. Optical components such as 90-degree optical hybrids are now widely commercially available.

8.6 SUMMARY Coherent homodyne detection can be used as a high-resolution temporal diagnostic for optical sources. This capability is unique and advantageous for optical performance monitoring. Valuable assessment of the performance of telecommunication systems and devices is obtained with statistical measurements of the electric field of the optical signals. Because the technology of optical telecommunications is constantly evolving, diagnostics are evolving accordingly. Since coherent detection has seen increased interest over the past few years for optical transmission, diagnostics providing relevant information on the intensity and phase of optical signals naturally fit into this evolution.

ACKNOWLEDGMENTS This work was partially supported by the U.S. Department of Energy Office of Inertial Confinement Fusion under Cooperative Agreement No. DE-FC52-08NA28302, the University of Rochester, and the New York State Energy Research and Development Authority. The support of DOE does not constitute an endorsement by DOE of the views expressed in this article. Some of the concepts and experiments presented in this chapter were developed by the author while working at Bell Laboratories (Lucent Technologies). The author is thankful to many of his former colleagues for fruitful interactions, in particular Christopher Doerr, Inuk Kang, Daniel Kilper, Howard Stuart, and

220

CHAPTER 8 Optical performance monitoring based on linear optical sampling

Peter Winzer (all from Bell Laboratories), as well as Michael Raymer (Oregon Center for Optics, University of Oregon). Fruitful discussions with Mathias Westlund (Chalmers University of Technology), Peter Andrekson (Chalmers University of Technology), and Michael Taylor (Atlantic Sciences) during the preparation of this chapter are also acknowledged.

REFERENCES 1. Ip E, Lau AP, Barros DJF, Kahn JM. Coherent detection in optical fiber systems. Opt Express 2008;16:753–91. 2. Ho KP. Phase-modulated optical communication systems. Berlin: Springer; 2005. 3. Dinu M, Kilper DC, Stuart HR. Optical performance monitoring using data stream intensity autocorrelation. J Lightwave Technol 2006;24:1194–202. 4. Dorrer C. High-speed measurements for optical telecommunication systems. IEEE J Sel Top Quantum Electron 2006;12:843–58. 5. Walmsley IA, Dorrer C. Characterization of ultrashort electromagnetic pulses. Adv Opt Photon 2009;1:308–437. 6. Trebino R. Frequency resolved optical gating: the measurement of ultrashort optical pulses. New York: Kluwer Academic Publishers; 2002. 7. Iaconis C, Walmsley IA. Spectral phase interferometry for direct electric-field reconstruction of ultrashort optical pulses. Opt Lett 1998;23:792–4. 8. Dorrer C, Kang I. Simultaneous temporal characterization of telecommunication optical pulses and modulators using spectrograms. Opt Lett 2002;27:1315–7. 9. Dorrer C, Kang I. Highly sensitive direct femtosecond pulse characterization using electro-optic spectral shearing interferometry. Opt Lett 2003;28:477–9. 10. Bennett CV, Scott RP, Kolner BH. Temporal magnification and reversal of 100 Gb/s optical data with an up-conversion time microscope. Appl Phys Lett 1994;65:2513–5. 11. Salem R, Foster MA, Turner-Foster AC, Geraghty DF, Lipson M, Gaeta AL. High-speed optical sampling using a silicon-chip temporal magnifier. Opt Express 2009;17:4324–9. 12. Dorrer C. Single-shot measurement of the electric field of optical waveforms by use of time magnification and heterodyning. Opt Lett 2006;30:3237–9. 13. Schmidt-Langhorst C, Weber H-G. Optical sampling techniques. Ultra-high-speed optical transmission technology. Berlin: Springer; 2007. 14. Cundiff ST, Ye J. Phase stabilization of mode-locked lasers. J Mod Opt 2005;52:201–19. 15. Walmsley IA, Waxer L, Dorrer C. The role of dispersion in ultrafast optics. Rev Sci Instrum 2001;72:1–29. 16. Taylor MG. Coherent detection method using DSP for demodulation of signal and subsequent equalization of propagation impairments. IEEE Photon Technol Lett 2004;16:674–6. 17. Ly-Gagnon D-S, Tsukamoto S, Katoh K, Kikuchi K. Coherent detection of optical quadrature phase-shift keying signals with carrier phase estimation. J Lightwave Technol 2006;24:12–21. 18. Dorrer C, Doerr CR, Kang I, Ryf R, Leuthold J, Winzer P. Measurement of eye diagrams and constellation diagrams of optical sources using linear optics and waveguide technology. J Lightwave Technol 2005;23:178–86. 19. Dorrer C. Monitoring of optical signals from constellation diagrams measured with linear optical sampling. J Lightwave Technol 2006;24:313–21. 20. Kim I, Kim C, Li G. Requirements for the sampling source in coherent linear sampling. Opt Express 2004;12:2723–30.

References

221

21. Sko¨ld M, Westlund M, Sunnerud H, Andrekson PA. All-optical waveform sampling in high-speed optical communication systems using advanced modulation formats. J Lightwave Technol 2009;27:3662–71. 22. Ito F. Demultiplexed detection of ultrafast optical signal using interferometric cross-correlation technique. J Lightwave Technol 1997;15:930–7. 23. Dorrer C, Kilper DC, Stuart HR, Raybon G, Raymer MG. Linear optical sampling. IEEE Photon Technol Lett 2003;15:1746–8. 24. Taylor MG. Phase estimation methods for optical coherent detection using digital signal processing. J Lightwave Technol 2009;27:901–14. 25. Leeb WR. Optical 90 hybrid for Costas-type receivers. Electron Lett 1990;26:1431–2. 26. Westlund M, Andrekson PA, Sunnerud H, Hansryd J, Li J. High-performance optical-fiber-nonlinearitybased optical waveform monitoring. J Lightwave Technol 2005;23:2012–22. 27. Okamoto K, Ito F. Dual-channel linear optical sampling for simultaneously monitoring ultrafast intensity and phase modulation. J Lightwave Technol 2009;27:2169–75. 28. Kang I, Dorrer C, Zhang L, Dinu M, Rasras M, Buhl LL, et al. Characterization of the dynamical processes in all-optical signal processing using semiconductor optical amplifiers. IEEE J Sel Top Quantum Electron 2008;14:758–69. 29. Gordon JP, Mollenauer LF. Phase noise in photonic communications systems using linear amplifiers. Opt Lett 1990;15:1351–3. 30. Dorrer C. Direct measurement of nonlinear coefficient of optical fiber using linear optical sampling. Electron Lett 2005;41:8–10. 31. Li G. Recent advances in coherent optical communication. Adv Opt Photon 2009;1:279–307. 32. Taylor MG. Measurement of phase diagrams of optical communication signals using sampled coherent detection. In: Proc. symposium on optical fiber measurements. Boulder, CO; 2004. p. 163–4. 33. Dorrer C. Complete characterization of periodic optical sources by use of sampled test-plus-reference interferometry. Opt Lett 2005;30:2022–4. 34. Benkler E, Telle HR, Weingarten K, Krainer L, Spu¨hler G, Keller U. Characterization of ultrashort optical pulse properties by amplitude-modulation-balanced heterodyne gating. Opt Lett 2005;30:2016–8. 35. Okamoto K, Ito F. Ultrafast measurement of optical DPSK signals using 1-symbol delayed dual-channel linear optical sampling. IEEE Photon Technol Lett 2008;20:948–50. 36. Okamoto K, Ito F. Nearly shot-noise-limited performance of dual-channel linear optical sampling for ultrafast DPSK signals. IEEE J Quantum Electron 2009;45:711–9. 37. Williams PA, Dennis T, Coddington I, Swann WC, Newbury N. Vector signal characterization of high-speed optical components by use of linear optical sampling with milliradian resolution. IEEE Photon Technol Lett 2008;20:2007–9. 38. Coddington I, Swann WC, Newbury N. Coherent linear optical sampling at 15 bits of resolution. Opt Lett 2009;34:2153–5. 39. Okamoto K, Fan X, Ito F. Ultrafast sampling of complex polarization components for characterizing polarization mode dispersion. In: Proc. optical fiber communication conference, paper OTuN6. Anaheim, California; 2007. 40. Williams PA, Dennis T, Coddington I, Newbury N. Polarization-sensitive linear optical sampling for characterization of NRZ polarization-multiplexed QPSK. In: Proc. optical fiber communication conference, paper OThH2. San Diego, California; 2009. 41. Okamoto K, Ito F. Simultaneous WDM signal detection realized by ultrafast field sampling. Opt Express 2009;17:6696–702. 42. Dennis T, Williams PA, Coddington I, Newbury N. Word-synchronous linear optical sampling of 40 Gb/s QPSK signals. In: Proc. optical fiber communication conference, paper OThH3. San Diego, California; 2009.

This page intentionally left blank

CHAPTER

Optical performance monitoring based on RF pilot tones

9

Paul K.J. Park*, Yun C. Chung{ *Samsung Advanced Institute of Technology, Samsung Electronics, Korea Department of Electrical Engineering and Computer Science, Korea Advanced Institute of Science and Technology, Korea

{

9.1 INTRODUCTION For the proper operation and maintenance of modern dynamic wavelength-division-multiplexed (WDM) networks, it is essential to directly monitor the signal’s quality in the optical layer.1 For this purpose, there have been many efforts to utilize pilot tones, which represent small sinusoidal components added to WDM signals.2–11 For example, it has been reported that pilot tones can be used for monitoring various optical parameters of WDM signals such as optical channel power, wavelength, crosstalk, and so on. Unlike other optical performance monitoring techniques, the pilot-tone-based techniques can monitor these parameters without using the expensive demultiplexing filters such as tunable optical filters and diffraction gratings. Thus, this technique can be extremely cost effective. In addition, this technique is well suited for use in a dynamic WDM network, since the pilot tones are bound to follow their corresponding optical signals anywhere in the network. Therefore, the optical path of each WDM signal can be monitored simply by tracking its tone frequency. Although the pilot-tone-based monitoring technique has many advantages, it also has some limitations (particularly when the pilot tone is obtained simply by dithering the bias current of the transmitter laser). First, the pilot tone can impose unwanted amplitude modulation on the data signal and degrade receiver sensitivity.12 Second, the performance of the pilot-tone-based monitoring technique can be deteriorated by the “ghost tones” caused by the crossgain modulation (XGM) and stimulated Raman scattering (SRS).13 These problems can be mitigated by utilizing proper amplitudes and frequencies of pilot tones.2 However, for the use in the long-haul network with a large number of channels, it may still be necessary to restrict the number of WDM channels to be monitored simultaneously (e.g., by using an optical bandpass filter). Pilot tone can also be used for monitoring the chromatic dispersion (CD) and polarization mode dispersion (PMD) for adaptive compensators.8–11 These techniques typically utilize high-frequency (>1 GHz) pilot tones. However, the main drawback of these techniques is the difficulty in separating their effects, since the magnitude of such high-frequency pilot tone is dependent on both types of dispersions.2 In order to solve this problem, the phase-modulated (PM) pilot tones are used for monitoring CD instead of the conventional amplitude-modulated (AM) pilot tones.10 In contrast, the use of singlesideband (SSB) pilot tones has been proposed for monitoring the PMD without the effect of CD.2 © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00009-2

223

224

CHAPTER 9 Optical performance monitoring based on RF pilot tones

In this chapter, we review various pilot-tone-based monitoring techniques proposed for the proper operation and maintenance of high-capacity WDM networks. In Section 9.2, we describe the operating principle of the pilot-tone-based monitoring technique and estimate its scalability. We also introduce the technique’s typical applications and implementation examples in this section. In Section 9.3, we present the recently proposed monitoring techniques based on the PM and frequencymodulated (FM) pilot tones. In Section 9.4, we review the pilot-tone-based monitoring techniques for CD and PMD used in adaptive compensators. We first review the monitoring techniques for CD and compare their performances by considering the effects of self-phase modulation (SPM) and PMD. In addition, we described the recent advancements in CD monitoring techniques obtained by utilizing the chirped pilot tone and the AM pilot tone carried by a broadband light source. We also review the PMD monitoring technique that is based on the SSB pilot tones and discuss its performance in the presence of CD. Finally, the chapter is summarized in Section 9.5.

9.2 PERFORMANCE MONITORING TECHNIQUES USING AM PILOT TONES In this section we first describe the operating principle of the pilot-tone-based monitoring technique and discuss its potential problems and possible solutions. Then, we estimate the scalability of the pilot-tone-based monitoring techniques and determine the proper range of tone frequencies for the targeted network. In addition, we review the techniques for monitoring various optical parameters of WDM signals such as optical power, wavelength, optical path, and crosstalk. We close this section with some implementation examples.

9.2.1 Operating principle Figure 9.1 shows that an optical signal is generated from node A and then transmitted to node C via node B. A pilot tone (i.e., a small sinusoidal component) is added to the optical signal at node A. At node B, this pilot tone can be extracted by using a low-speed photodiode and used for optical performance monitoring. It should be noted that in a dynamically configurable WDM network, pilot tones can also be used to monitor the optical paths of a WDM signal. This is because the pilot tone is bound to follow the corresponding WDM signal to wherever in the network once the tone is attached. Thus, we can monitor the optical path of each WDM signal only by tracking the tone frequency.

Pilot tone

Node A

Pilot tone Node B

Signal

Node C

FIGURE 9.1 Pilot-tone-based optical performance monitoring technique.2

9.2 Performance monitoring techniques using AM pilot tones

AM

LD

AM

LD

225

PM

LD

Bias

Bias

Data

Bias

Bias Data

(a)

(b)

(c)

PD

PD

A/D

BPF

FFT

RFD

RFD

(d)

(e)

(f)

PD LOSC

FIGURE 9.2 Pilot-tone generation and detection methods. (a) Adding a small sinusoidal current to the laser’s bias current. (b) Dithering bias voltage of external modulator. (c) PM tone generation by using phase modulator. (d) Pilottone detection using FFT. (e) Using tunable electrical bandpass filter. (f) Using tunable local oscillator for the down-conversion of tone frequency. LD, laser diode; AM, amplitude modulator; PM, phase modulator; PD, photodetector; A/D, analog-to-digital converter; FFT, fast Fourier transform; BPF, tunable bandpass filter; RFD, radio frequency power detector; LOSC, tunable local oscillator.2

For using the pilot tone in practical systems, it is necessary that pilot tones should be added into and extracted from WDM signals anywhere in the network. Figure 9.2 shows typical techniques used for the generation and detection of pilot tones.2 A pilot tone could be generated by dithering the laser’s bias current (Figure 9.2(a)),5 the bias voltage of the amplitude modulator (Figure 9.2(b)),9 or the phase modulator (Figure 9.2(c)).10 These techniques would require a slight modification of the existing transmitter and help in suppressing the stimulated Brillouin scattering.14 For the detection of pilot tones, a technique based on the fast Fourier transform (FFT) can be utilized as shown in Figure 9.2(d).5 This technique is attractive since every pilot tone added to WDM signals can be detected simultaneously without any scanning mechanism. In case of
GHz tone frequency, a tunable electrical bandpass filter or a tunable local oscillator can be used for the tone detection, as shown in Figures 9.2(e) and (f), respectively.

9.2.2 Potential problems 9.2.2.1 Receiver sensitivity degradation When the AM pilot tone is added to an optical signal, it could interfere with data and cause deleterious effects.2 In earlier studies, it has been reported that these effects are dependent only on the modulation indices (MIs) and not on the frequencies of pilot tones. (MI is defined as the ratio between the peak amplitude of the pilot tone and the average signal amplitude.12) Thus, it would be required to maintain the MI within the acceptable level. However, because of the frequency response characteristics of the optical receiver, the maximum allowable MI is also dependent on

CHAPTER 9 Optical performance monitoring based on RF pilot tones

2.5

Receiver response

14 MI = 12 %

2.0

0.8 0.6

1.5

0.4 0.2

B.W. = 150 kHz

0.0

1.0

1

10

100

1000

MI = 6 % 0.5 0.0

12

1.0

MI of RF tone (%)

Power penalty (dB)

1.2

10 100 1000 Tone frequency (kHz) (a)

10

Data spectrum Fequency

8 6 4

Due to LPF characteristics of receiver

2

MI = 3 % 1

RF tone

Power

226

0

10000

1

2 3 4 5 6 7 8 9 10 Tone frequency (GHz) (b)

FIGURE 9.3 Effects of modulation index and frequency of pilot tone on 10-Gb/s NRZ signal (pattern length ¼ 231  1). (a) Pilot-tone-induced power penalty measured at low tone frequencies. (b) Maximum allowable modulation indices of high-frequency pilot tones for 0.5-dB penalty.2

the tone frequency. For example, Figure 9.3 shows the power penalties of a 10-Gb/s non-return-tozero (NRZ) signal (pattern length ¼ 2311) measured while varying the MI and the frequency of the pilot tone. The results show that the power penalty could be substantially reduced by using a tone frequency lower than
100 kHz. This was because the receiver used in the experiment had a lowfrequency cut-off at approximately 150 kHz. Thus, the power penalty could be maintained within 0.5 dB even when the MI was as high as 12%. However, the MI should be smaller than 6% when we used the tone frequency in the range of 1 MHz to 4.5 GHz. When the tone frequency was higher than 4.5 GHz, the pilot-tone-induced power penalty decreased because of the roll-off characteristics of the 10-Gb/s receiver. In order to reduce the tone-induced power penalty at a low tone frequency, it is possible to decrease the amplitude of a pilot tone by inserting a highpass filter into the optical receiver.15 However, if the highpass filter is used in the optical receiver, the waveform (i.e., “1” and “0” bits) of the data signal would be distorted as shown in Figure 9.4. In principle, this waveform can be distorted further as the number of consecutive 1- or 0-bits increases. When n consecutive 1-bits pass through the highpass filter in the optical receiver, the normalized amplitude of the nth bit at decision time x can be described as x ¼ e2pf0 ðn2Þ=B : 1

(9.1)

where f0 and B are the low cut-off frequency of the optical receiver and the bit rate, respectively. Similarly, in the case of n consecutive 0-bits, the normalized amplitude of the nth bit at the decision time, y, can be described as y ¼ 1  e2pf0 ðn2Þ=B ; 1

(9.2)

9.2 Performance monitoring techniques using AM pilot tones

227

Highpass filter “1”

“1”

“0”

“0”

“0”

Input data

“0”

Decision time

Decision time

1-bit

Output data

1-bit x

y 0-bit 0-bit

1/B

1/B

FIGURE 9.4 Effect of highpass filter on data signal.15

Equations (9.1) and (9.2) indicate that the highpass filter induces an eye closure of the data signal. Thus, the eye closure penalty, Pc, can be calculated as Pc ¼ 10 logðx  yÞ:

(9.3)

Using these equations, we calculated the eye closure penalty while varying the low cut-off frequency in comparison with simulation results (using OptSim 3.0), as shown in Figure 9.5. In the calculation, we assumed that the bit rate and pattern length were 2.5 Gb/s and 271, respectively. The results 3.5

Eye closure penalty (dB)

3

Calculation Simulation

2.5 2 1.5 1 0.5 0 10

100 1000 10,000 Cut-off frequency (kHz)

100,000

FIGURE 9.5 Eye closure penalty calculated while varying low cut-off frequency in comparison with simulation results (bit rate ¼ 2.5 Gb/s, pattern length ¼ 27  1).15

228

CHAPTER 9 Optical performance monitoring based on RF pilot tones

show that the eye closure penalty increases with an increase in the low cut-off frequency of the optical receiver. However, the eye closure penalty is negligible (<0.1 dB) when the low cut-off frequency is less than 1 MHz. In practical transmission systems, the bit error rate (BER) is typically used for evaluating the system performance. By using Olsson’s model16 and Equations (9.1) and (9.2), we can calculate the effect of the highpass filter on the BER of the data signal. In particular, when we utilize a pseudo-random-bit-sequence (PRBS) signal, the overall BER, BERt, can be expressed as X  BERt ¼ (9.4) Pprob BERb ; where Pprob and BERb represent the probability and BER of each PRBS pattern, respectively, and Pprob can be described as17 Pprob ¼

2nk2 ; ð2n  1Þ

(9.5)

where 2n  1 represents the total number of PRBS patterns and k is the number of consecutive 1-bits. These equations show that the BER is dependent on the type of PRBS pattern in the presence of a highpass filter. In order to confirm this, we measured the BER of the PRBS signal as shown in Figure 9.6. The output power and the operating wavelength of the distributed feedback (DFB) laser were 3 dBm and 1555 nm, respectively. We modulated the DFB laser using a LiNbO3 modulator. The modulated signal was sent to a photodetector via an optical attenuator. We then measured the BER of the detected signal while varying the bit rate from 155 Mb/s to 2.5 Gb/s (pattern length ¼ 27  1 and 231  1). The low cut-off frequency of our receiver chain was 50 kHz. In the experiment, we intentionally changed the low cut-off frequency of the optical receiver by using various highpass filters. Figure 9.7 shows the power penalty (@BER ¼ 109) measured while varying the bit rate when the low cut-off frequency of the optical receiver was 1 MHz, in comparison with the theoretically calculated lines. The power penalty of the PRBS signal increases with an increase in the pattern length increases. The results show that the power penalty was measured to be less than 0.3 dB at the bit rate of 2.5 Gb/s even when the pattern length was increased to 231  1. Thus, this technique can mitigate the tone-induced degradation in the receiver sensitivity while maintaining the performance of the data signal.

LiNbO3 modulator

LD (1555 nm)

Optical attenuator

10 90

Pattern generator

Signal generator

Data

Modulator driver

Clock BERT

FIGURE 9.6 Experimental setup.15

Highpass filter

RF amplifier

Bessel filter

RF amplifier

PD

Power meter

9.2 Performance monitoring techniques using AM pilot tones

229

Power penalty (dB)

2 27 − 1 PRBS 231 − 1 PRBS Calculation

1.6 1.2 0.8 0.4 0 0.1

1 Bit rate (Gb/s)

10

FIGURE 9.7 Power penalty measured while varying the bit rate (low cut-off frequency ¼ 1 MHz).15

Figure 9.8 shows the power penalty measured while varying the tone frequency when the bit rate and pattern length were 2.5 Gb/s and 27  1, respectively (MI ¼ 12%). The results show that we can reduce the tone-induced power penalty by
0.8 dB (tone frequency ¼ 1 MHz) by inserting the highpass filter (1-MHz cut-off) into the optical receiver. In particular, the low cut-off frequency of the commercial 10-Gb/s optical receiver is lower than 100 kHz. In this case, our calculation shows that the power penalty can be negligible (<0.1 dB) even when the low cut-off frequency of the 10-Gb/s optical receiver is increased up to
10 MHz (@27  1 PRBS).

Power penalty (dB)

2

1.5

Without filter With filter (1-MHz cut-off)

1

0.5

0 100

1000 Tone frequency (kHz)

FIGURE 9.8 Power penalty measured while varying tone frequency (bit rate ¼ 2.5 Gb/s, pattern length ¼ 27  1, MI ¼ 12%).15

230

CHAPTER 9 Optical performance monitoring based on RF pilot tones

9.2.2.2 Effect of ghost tones When the AM pilot-tone-based monitoring technique is used in an amplified WDM network, the efficiency of the technique could be deteriorated by the XGM of an erbium-doped fiber amplifier (EDFA) and/or SRS.2,18,19 Figure 9.9 illustrates the mechanisms of these performance degradations. When a pilot tone is injected into the EDFA, the amplifier gain becomes modulated because of its slow gain dynamic property. This XGM could generate unwanted crosstalk components at the WDM channels.19 We designated these crosstalk components as “ghost tones.” Such ghost tones could not only cause measurement errors of the monitoring technique but also mislead the network operators (i.e., the dropped channels would be interpreted as though they existed). This problem could be simply solved by using high-frequency (>1 MHz) pilot tones19 and/or automatic gaincontrolled EDFAs.13 However, in this case, the performance could be impaired by the interactions between the optical signals and pilot tones via SRS.13 In order to exemplify these problems, we performed a long-distance WDM transmission experiment using eight DFB lasers operating at 1547.72 to 1558.98 nm (channel spacing: 200 GHz). Each laser was slightly dithered by modulating its bias current with a small sinusoidal component ranging from 101 to 115 kHz (separation, 2 kHz; MI, 10%). The transmission link consisted of 640 km of single-mode fiber (SMF) and eight EDFAs. In order to evaluate the effect of XGM, we dropped six out of eight WDM channels after transmission over 640 km of SMF and measured the optical spectrum of WDM signals and the electrical spectrum of their corresponding pilot tones. Figure 9.10 shows the results. Although the optical spectrum in Figure 9.10(a) shows that six channels were completely dropped, the electrical spectrum of pilot tones in Figure 9.10(b) shows the ghost tones generated by XGM at the frequencies corresponding to the dropped channels. In order to suppress these ghost tones, we applied a dynamic gain control unit (using an additional laser) in the first EDFA of the transmission link.20 However, in this case, we turned off the fifth WDM channel operating at the same wavelength (1554.13 nm) with the control channel. This was done to maintain Fiber EDFA

XGM

Optical carrier 1 f1

Optical carrier 2

Pilot tone f1 f2

SRS

Optical carrier 1

f2

Freq

f 2 f1 f 1 f2

Optical carrier 2

Optical carrier 1

f2 f1 f1 f2

Ghost tones

Freq

FIGURE 9.9 Mechanisms of performance degradation caused by XGM and SRS.2

f2 f1

f1

SRS

Optical carrier 2 f2

f2

f1

Ghost tones

f1 f2

Freq

9.2 Performance monitoring techniques using AM pilot tones

5.0 dB/D

RES: 0.2 nm SENS:NORM HLD AVG: 5

231

SMPL:1001

–20.0

10 dB/div 10 –30.0 REF dBm

Ghosttones tones ghost

–40.0

–50.0

–60.0 1546.10 nm

MON:SGL

1554.05 nm

1.59 nm/D

1562.00 nm

118 kHz

98 kHz

(a)

(b)

10 dB/div

10 dB/div

Ghosttones tones ghost

Noghost ghosttones tones no

Control channel 98 kHz

118 kHz (c)

98 kHz

118 kHz (d)

FIGURE 9.10 Measured optical and electrical spectra. (a) Optical spectrum measured after 640-km transmission. (b) Electrical spectrum measured after 640-km transmission (without using control channel). (c) Electrical spectrum measured after first EDFA (with using control channel). (d) Electrical spectrum measured after 640-km transmission (with using control channel).2

the total input power of the EDFA to be constant. Figure 9.10(c) shows the electrical spectrum of pilot tones measured at the output of the first EDFA. The ghost tones were removed completely as we compensated the effect of XGM using the control channel. However, when we measured the electrical spectrum after transmission over 640 km of SMF, the ghost tones reappeared, as shown in Figure 9.10(d). The amplitudes of these ghost tones were measured to be larger for the channels separated further from the control channel and increased in proportion to the square of the signal power. Thus, we concluded that these ghost tones were caused by the SRS-induced crosstalk.13 Figure 9.11(a) shows that, by using pilot tones, each WDM channel can be represented by a distinct pilot tone in the electrical spectrum. However, as WDM signals traverse the optical network, these pilot tones could generate ghost tones on the other WDM channels via XGM and SRS, as

232

CHAPTER 9 Optical performance monitoring based on RF pilot tones

Fiber WDM signals with tones

After EDFA & fiber

EDFA

BPF PD

Pilot tone nc y ef req ton Pil

ot-

f1

f

ue

f3 f2

Pil

ot-

ton

ef req

ue

nc y

f

f3

f2

f1 Ghost tone (c)

l1 l2 l3 l Optical wavelength (a)

PD

(d)

l1 l2 l3 l Optical wavelength (b)

FIGURE 9.11 Effects of ghost tones on pilot-tone-based monitoring technique. (a) WDM signals with corresponding pilot tones. (b) Ghost tones generated by XGM and SRS. (c) Pilot tones measured without using a demultiplexing filter. (d) Pilot tones measured after demultiplexing WDM channels.2

shown in Figure 9.11(b). These ghost tones could cause measurement errors if we tap a small portion of light from the transmission fiber and detect the pilot tones of all WDM channels using a single photodiode, as shown in Figure 9.11(c). However, this problem could be avoided by demultiplexing the WDM channels before the detection of the pilot tones, as shown in Figure 9.11(d). This is because the demultiplexed WDM signal does not have any ghost tone (caused by other WDM channels) at its corresponding tone frequency.

9.2.3 Scalability A pilot-tone-based monitoring technique is extremely cost effective since we can tap a small portion of light from the transmission fiber and use it for the performance monitoring of every WDM signal without using the demultiplexing filter. However, in this case, the accuracy of the pilot-tone-based monitoring technique could be seriously affected by the ghost tones caused by XGM and SRS. Thus, we investigated the maximum size of the optical network that the pilot-tone-based monitoring technique could support. For this estimation of scalability, we calculated the maximum number of EDFA spans as a function of the number of WDM channels.19 We assumed that the RF power of a ghost tone should be at least 10-dB smaller than that of the original pilot tone. This was to avoid any misinterpretation of the dropped and passed channels by the network operators. Under this condition, the

9.2 Performance monitoring techniques using AM pilot tones

233

Number of EDFA spans (80 km)

32 28 400 MHz 24

1 MHz

20

500 kHz

16

100 MHz

12 300 kHz 8

10 MHz Tone freq = 100 kHz

4 0

0

16

<1 MHz 32 48 Number of WDM channels

64

80

FIGURE 9.12 Scalability of pilot-tone-based monitoring technique. Solid line, limits imposed by SRS; dashed line, limits imposed by XGM.2

ghost-tone-induced measurement error for optical power would be less than 0.5 dB. We also assumed that the channel spacing was 100 GHz and the optical power of each channel was 3 dBm. The results are shown in Figure 9.12. The solid and dashed lines represent the limitations imposed by the SRS and XGM, respectively.13 The effect of XGM appears to be dominant when the tone frequency is low (<100 kHz) and the number of WDM channels is small (<32). However, the effect of XGM is rapidly reduced as the tone frequency increases. When the tone frequency is higher than 300 kHz, the effect of SRS becomes dominant. In contrast, the effect of SRS is not sensitive to the frequencies of pilot tones unless the frequencies are considerably higher than 10 MHz. Low-frequency pilot tones (<10 MHz) can be detected easily by using an off-the-shelf A/D converter and FFT processor. However, it limits the maximum network size that the pilot-tone-based monitoring technique can support. For example, when we use the pilot tones in the range of 300 kHz, the maximum network size is limited to five EDFA spans for the WDM system with 32 channels due to SRS. Thus, unless an expensive tunable optical filter is used, the application of the low-frequency pilot tones will be confined to metropolitan area networks. For the use of low-frequency pilot tones in a longhaul WDM network, it is necessary to restrict the number of channels to be monitored at a time. For example, the maximum network size can be increased to 20 EDFA spans (1600 km) by monitoring a group of 16 WDM channels simultaneously. This could be achieved by using a coarse WDM demultiplexer and an optical switch or a broadband tunable filter. Another technique could be the use of highfrequency tones in the order of several hundreds of MHz. Since the ghost tone is well suppressed in this frequency range, it will not be necessary to use optical bandpass filters. However, in this case, a highspeed A/D converter (
1 GS/s) is required for the detection of pilot tones. In addition, as the tone frequency is increased further (>1 GHz), the amplitude of the pilot tone becomes sensitive to CD and PMD. Thus, the high-frequency pilot tones are mostly used for the PMD compensators and tunable dispersion compensators rather than for monitoring the performance of optical signals.

234

CHAPTER 9 Optical performance monitoring based on RF pilot tones

9.2.4 Typical applications For the efficient operation and maintenance of a dynamic WDM network, it is essential to monitor various optical parameters such as channel power, wavelength, optical path, and crosstalk. In this section, we review several techniques capable of monitoring these parameters by using pilot tones.

9.2.4.1 Optical channel power and wavelength monitoring The magnitude of a pilot tone is proportional to the optical power of its corresponding WDM channel. Thus, the optical powers of WDM channels can be monitored easily by using pilot tones.3,5 The wavelengths of WDM channels can also be monitored by using pilot tones and a solid etalon filter.5 Figure 9.13(a) shows a typical experimental setup of the channel power and wavelength monitoring technique using pilot tones. In this experiment, each WDM laser was slightly dithered at a unique tone frequency by adding a small sinusoidal current to the laser’s bias current. After transmission, the WDM signals were sent to the monitoring module and then split into two parts. One part was directly sent to a photodiode for the monitoring of channel power. The channel power could be monitored simply by multiplying the detected amplitude of the pilot tone by its MI. For wavelength monitoring, the other part was sent a photodetector via the solid etalon filter, which had resonance peaks

Bias and pilot tone

LD 1 Power

Fiber

PD 1

Modulator

A/D FP etalon filter

Bias and pilot tone

LD 16

Wavelength

FFT

PD 2 Monitoring module

0

0.08

–0.05

0.06

–0.1

0.04

Amplitude (a.u.)

Amplitude (a.u.)

(a)

–0.15 –0.2 –0.25 –0.3

0 –0.02 –0.04 –0.06

–0.35 –0.4 –30

0.02

–20 –10 0 10 20 Normalized frequency (GHz) (b)

30

–0.08 –30

–20 –10 0 10 20 Normalized frequency (GHz) (c)

FIGURE 9.13 Principle of tone-based channel power and wavelength monitoring technique. (a) Experimental setup. (b) Imaginary part of FFT-converted data. (c) Real part of FFT-converted data.2

30

9.2 Performance monitoring techniques using AM pilot tones

235

matched to International Telecommunication Union (ITU) grids. Thus, the free-spectral range of this etalon filter was identical to the channel spacing. The detected signal was then digitized by using an A/D converter and processed with FFT. We tuned the operating frequency of a WDM laser by adjusting its temperature and measured the amplitudes of the real and imaginary parts of the FFTtransformed signal. The results are shown in Figures 9.13(b) and (c). In these figures, the horizontal axis represents the difference between the laser frequency and its corresponding resonance peak of the etalon filter. Thus, these figures could be used for determining the operating frequency of a WDM laser. For example, Figure 9.13(b) shows that we can determine two possible laser frequencies (one is lower and the other is higher than the resonance peak) by using the measured amplitude of the imaginary part. The correct operating frequency (between these two frequencies) can then be selected by using the sign of the real part, as shown in Figure 9.13(c). Figure 9.14 shows the power and wavelength errors of 16 WDM channels measured by using the monitoring technique that is based on pilot tones. The maximum power error was measured to be less than 0.4 dB even when the channel power incident on the monitoring module was varied from 24 to 45 dBm. The wavelength error was less than 3 GHz as long as the WDM channels operated within 40 GHz from their corresponding ITU grids.5 These results demonstrate that the technique based on pilot tones is capable of monitoring the channel powers and wavelengths of WDM signals. However, the accuracy of this technique could be degraded by ghost tones (caused by XGM and SRS) as we increase the transmission distance and/or the number of WDM channels. In order to avoid this problem, it has been proposed to demultiplex the WDM channels by using arrayed-waveguide grating (AWG) before the detection of pilot tones.6

9.2.4.2 Optical path and channel identification In the future all-optical transport network, it is envisioned that the optical paths of WDM signals can be frequently changed by the optical crossconnects (OXCs). Thus, for the proper operation and management of such a dynamically reconfigurable network, it will be necessary to monitor the optical paths of every WDM channel. Pilot tones are well suited for this purpose since they are bound to 4 CH 1 CH 2 CH 3 CH 4 CH 5 CH 6 CH 7 CH 8 CH 9 CH 10 CH 11 CH 12 CH 13 CH 14 CH 15 CH 16

1

0

–1

–2 –40 –37 –34 –31 –28 –25 –22 –19 –16 –13 –10 –7 Input power (dBm)

Frequency error (GHz)

Power errror (dB)

2

CH 1 CH 2 CH 3 CH 4 CH 5 CH 6 CH 7 CH 8 CH 9 CH 10 CH 11 CH 12 CH 13 CH 14 CH 15 CH 16

2

0

–2

–4 –60

–40

–20 0 20 40 Normalized frequency (GHz)

(a)

FIGURE 9.14 Monitoring errors of 16 WDM channels: (a) channel power and (b) wavelength.2

(b)

60

80

236

CHAPTER 9 Optical performance monitoring based on RF pilot tones

follow their corresponding optical signals in the network. Thus, the optical paths of WDM signals can be monitored by simply tracking their corresponding tone frequencies since every channel is assigned to a unique tone frequency. In earlier studies, it has been proposed to monitor the optical paths of WDM signals within the OXC by using low-frequency pilot tones (0.1–5 KHz).21–23 These techniques added pilot tones to WDM signals at the input of OXC and then monitored them at the OXC’s output in order to verify that the optical path was reconfigured as desired. However, these pilot tones could propagate to the next nodes along with WDM signals and cause confusions to network operators (if every OXC utilized an identical set of tone frequencies). In addition, these low-frequency tones could generate ghost tones while traversing the amplified transmission link. In order to avoid these problems, it has been proposed to erase the pilot tones at the outputs of OXC by fiber Bragg gratings (FBGs)22 and voltage-controlled attenuators.23 There have been some efforts to extend these monitoring techniques for use in the entire network.3,4,7 For this purpose, every channel in the network should be assigned to a unique tone frequency. Thus, the channel power and the optical path of a WDM signal can be monitored from the originating node to the destination node by tracking its corresponding pilot tone. However, the performances of these techniques can be seriously impaired by XGM and SRS, particularly in a long-haul network with a large number of WDM channels. In order to overcome this problem, we measured the pilot tones after the WDM signals were demultiplexed in the OXC.7 Thus, unlike the previous techniques, this technique is not sensitive to ghost tones. Figure 9.15 shows the experimental setup used for demonstrating this technique. In this demonstration, four 44 OXCs, made of waveguide-grating routers and 640-km SMF PRBS

110 kHz

DFB l4

82 kHz

DFB l1

l2 OXC 1

•••

DFB l 3

l3 l4

OXC 2

•••

100 kHz

l1

•••

DFB l 2

•••

•••

90 kHz

•••

80 kHz

MOD

•••

10 Gb/s

DFB l 1

•••

PD array 84 kHz

DFB l1 Attenuator

OXC CONTROLLER (FFT)

A/D

OXC 3

86 kHz

DFB l1

OXC 4

FIGURE 9.15 Experimental setup used for demonstration of pilot-tone-based monitoring technique for optical path and crosstalk.2

9.2 Performance monitoring techniques using AM pilot tones

237

thermo-optic polymer switches, were used. We transmitted four WDM signals operating at 1550.92 to 1555.75 nm from OXC 1 to OXC 4 via 640 km of SMF and 8 EDFAs. These WDM signals had unique tone frequencies in the range of 80 to 110 kHz. We also sent one channel from each OXC, all operating at 1550.92 nm, to OXC 4. In OXC 4, we monitored the pilot tones at the outputs of optical switches. Figure 9.16(a) shows the electrical spectrum measured at one of the output fibers of the optical switch for l1 in OXC 4. From this figure, we can easily identify the optical path of channel 1, originating from OXC 1 by measuring the frequency component at 80 kHz. The optical power of this channel can also be monitored by measuring the magnitude of the pilot tone. Figure 9.16(a) also shows the ghost tones (at 90, 100, and 110 kHz) generated by XGM during the transmission over 640 km of SMF. However, it should be noted that there is no ghost tone at 80 kHz since we demultiplexed the WDM channels before the detection of pilot tones. When the optical switches in OXC fail to operate properly, intraband crosstalk—which is a difficult parameter to measure by using only optics—may be generated. This technique could also be used for the monitoring of such intraband crosstalk. In order to simulate the failure of an optical switch, we replaced the optical switch for l1 in OXC 4 with a 44 coupler and measured the pilot tones. The crosstalk level was adjusted by using the variable attenuators placed at the input fibers to OXC 4. The result in Figure 9.16(b) clearly shows the crosstalk components at 82, 84, and 86 kHz (caused by the WDM signals operating at the same wavelength with channel 1 but originating from different OXCs). This technique may require considerable tone frequencies since every channel in the network has to utilize a unique tone frequency. In order to mitigate this problem, the use of dual tones for each channel has been proposed.4 In this technique, the first tone represents the originating node, while the second tone indicates the channel number. Thus, we can monitor the optical paths of every channel in the network with a small number of tone frequencies. However, as the number of WDM channels increases,

10 dB/div

10 dB/div ~ 20 dB Ghost tones

Crosstalks

75

Ghost tones

RF power (dB)

RF power (dB)

No crosstalks

80

85

90 95 100 Frequency (kHz) (a)

105

110

115

~ 50 dB

75

80

85

90 95 100 Frequency (kHz) (b)

105

110

115

FIGURE 9.16 Electrical spectra of pilot tones measured at the output of the switch for l1 in OXC 4 (a) under normal operation, and (b) under switch failure condition.2

238

CHAPTER 9 Optical performance monitoring based on RF pilot tones

the performance of this technique can be seriously impaired by ghost tones (since all the WDM channels originating from the same node have an identical first tone).

9.2.4.3 Implementation examples In this section, we describe some examples of various network elements utilizing the pilottone-based monitoring technique. Figure 9.17 shows the schematic representations of the optical add/drop multiplexer (OADM) and OXC used in an all-optical network testbed.24 In order to utilize the pilot-tone-based monitoring technique, the bias current of each laser was slightly dithered at a unique tone frequency in the range of
100 kHz. The modulation index of the pilot tone was set to be less than 6% in order to avoid any excessive power penalty. The frequencies and modulation indices of these pilot tones were verified at the outputs of transmitters. Using these pilot tones, we monitored the optical powers and wavelengths of every added, dropped, and through channel. We also verified the optical paths of every channel by monitoring their tone frequencies. The monitored information was then sent to the neighboring nodes through the out-of-band supervisory channel. However, these OADM and OXC could not be used in an ultra-long-haul network with a large number of WDM channels because of ghost tones. Thus, for use in such a large-scale network, it would be necessary to add some tunable bandpass filters, as indicated in Figure 9.11 (see also Section 9.2.2.2).

9.3 PERFORMANCE MONITORING TECHNIQUES USING PM AND FM PILOT TONES The AM pilot tone causes the degradation in receiver sensitivity and ghost tone as described in Section 9.2.2. In order to eliminate these detrimental effects of an AM pilot tone, it is desirable not to use the amplitude modulation in the generation of the pilot tone. Thus, we introduce recently proposed monitoring techniques utilizing PM25 and FM26 pilot tones in this section. These techniques can overcome problems inherent in the monitoring technique based on the AM pilot tones while maintaining all the advantages of using pilot tones.

9.3.1 Using PM pilot tones Figure 9.18 shows the operating principle of the optical channel monitoring technique using a PM pilot tone.25 The optical phase of an optical carrier including the PM pilot tone f(t) can be described as fðtÞ ¼ f0 þ Df sinðotÞ;

(9.6)

where f0 and
f are the initial phase and the peak phase deviations of the optical carrier, respectively, and o is the angular modulation frequency of the pilot tone. When this PM pilot tone passes through the frequency discriminator, the amplitude of the optical signal becomes modulated because of the PM-AM conversion.27 In this case, the magnitude of the AM component
P can be expressed as DP / Tðn0 ÞT 0 ðn0 ÞDfo=2p;

(9.7)

where T(n) and T0 (n) represent the transmission function of the frequency discriminator and its first derivative, respectively. Equation (9.7) indicates that the magnitude of the AM component differs

9.3 Performance monitoring techniques using PM and FM pilot tones • Channel power • Channel ID • Wavelength • Channel power • Channel ID

• Channel power • Channel ID Total power

Tunable Tunable BPF BPF

Tx

PD PD

Rx

Tx

Tx

fM 2

fM1

Rx fMN

PD

PD Tunable Tunable BPF BPF

OSW OSW • Tx, Rx power • Switch X-talk OADM OADM Controller Controller (FFT) (FFT)

PD

PD PD Total power

Total power To other nodes

PD Total power

FPF ADC ADC

PD

PD

Rx

Tunable Tunable BPF

Total power PD

ADC ADC

Total power

239

Supervisory channel

To other nodes

(a) l λ11

l λ22

lλN N

Tunable Tunable BPF BPF PD PDarray array PD

fM1

• Ch power • Ch ID

• Tx, Rx power • Switch X-talk

OXC OXC Controller Controller (FFT) (FFT)

Tunable Tunable BPF BPF

PD PD

ADC ADC

OSW OSW

OSW OSW

Rx

OpticalSW SW Optical

ADC ADC

Tx

• Ch power • Ch ID • Wavelength

FPF FPF PD PD PD

• Total power

To other nodes

Supervisory channel

To other nodes

(b)

FIGURE 9.17 (a) Schematic representation of OADM implemented with the pilot-tone-based monitoring technique. (b) Schematic representation of OXC implemented with the pilot-tone-based monitoring technique. OSW, optical switch; FPF, Fabry-Parot filter; PD, photodetector; Tx, transmitter; Rx, receiver; ADC, analog-to-digital converter; FFT, fast Fourier transform.2

CHAPTER 9 Optical performance monitoring based on RF pilot tones

Amplitude

240

Frequency discriminator

PM-AM conversion

PM pilot tones

Optical frequency

Operating frequency of WDM signal

FIGURE 9.18 Operating principle of monitoring technique based on PM pilot tones.25

according to the operating frequency of the WDM signal. Thus, the operating frequency of the WDM signal can be easily monitored by measuring the magnitude of the AM component. Figure 9.19 shows the experimental setup used for demonstrating the principle of the monitoring technique based on PM pilot tones. The DFB lasers were operated at 192.4–193.2 THz (channel spacing ¼ 200 GHz). The optical phase of one DFB laser (channel 3) was modulated by using a LiNbO3 phase modulator. All WDM signals were combined and modulated at 10 Gb/s (pattern length: 231  1) using a high-speed LiNbO3 modulator and then sent to the AWG via an EDFA chain for the optical channel monitoring (AWG input power ¼ 3 dBm). Figure 9.20 shows the amplitudes of AM components (tones) of channel 3 and the ratio between these amplitudes measured at two adjacent AWG ports while varying the optical frequency (tone frequency ¼ 504 MHz, peak phase deviation ¼ 0.5 p). In this figure, A represents the amplitude of the AM component measured at the third input port of the AWG, while B represents the amplitude obtained at the fourth input port. These AM components are proportional to the transmission function of AWG as expressed in Equation (9.7). The result shows that the ratio curve is nearly linear (
0.4 dB/GHz) with the operating frequency of the optical signal. In particular, the crossover points of the AWG were anchored at the ITU-standardized frequencies by using a temperature controller.6 Thus, the operating frequencies of WDM signals can be estimated by measuring the ratios of tone amplitudes at the corresponding tone frequencies and AWG ports. In addition, the optical power of each WDM channel can also be easily monitored by measuring their corresponding tone amplitudes. However, Equation (9.7) indicates that, if the tone frequencies were too low, the magnitude of the AM components would not be measured. The monitoring sensitivity of this technique could be improved by increasing the tone frequency, but it would deteriorate the accuracy because of the PM-AM conversion caused by the CD.28 Thus, the tone frequency should be optimized to obtain the best performance. Figure 9.21 shows the monitoring error measured while varying the tone frequency (accumulated dispersion ¼ 1000 ps/nm) when the WDM channel was operating at 192.8 THz. The results show that the monitoring error increases when the

9.3 Performance monitoring techniques using PM and FM pilot tones

DCF

CH 1 CH 2 CH 3

Phase modulator

Intensity modulator

80 EDFA

EDFA

10 Gb/s CH 4

320-km SMF

CH 5

Multiwavelength meter

RF spectrum analyzer

PD PD

Attenuator

AWG

FIGURE 9.19 Experimental setup of PM pilot-tone-based monitoring technique.25

40

–50

Tone amplitudes (dB)

20

0

–70 B/A –80

–20

–90

–40 192.76 192.78 192.8 192.82 192.84 192.86 Optical frequency (THz)

FIGURE 9.20 Amplitudes and ratio of AM components measured at two adjacent AWG ports.25

Ratio of tone amplitudes (dB)

B

A –60

241

242

CHAPTER 9 Optical performance monitoring based on RF pilot tones

Monitoring error (GHz)

2.5 2 1.5 1 0.5 0 0.1

1 Tone frequency (GHz)

10

FIGURE 9.21 Frequency monitoring error measured while varying tone frequency (accumulated dispersion ¼ 1000 ps/nm) when the WDM channel was operated at 192.8 THz.25

tone frequency is lower than 200 MHz. This is mainly because the magnitude of the AM components is too small to be measured accurately (RF power <73 dBm, carrier-to-noise ratio (CNR) <10 dB, and resolution bandwidth ¼ 50 kHz). In contrast, when the tone frequency is higher than 1 GHz, the monitoring error is also increased by the CD. Thus, we found that the frequency of PM pilot tone should be approximately 500 MHz for accurate optical channel monitoring. Figure 9.22 shows the frequency monitoring errors measured after the transmission of 10-Gb/s signals over 320 km of dispersion-managed SMF using PM pilot tones (SMF launch power ¼ 3 dBm and dispersion-compensating fiber (DCF) launch power ¼ 3 dBm). In the experiment, the optical 20 –3

BER(bg)

Monitoring errors (GHz)

–4

10

–5

CH1 CH2 CH3 CH4 CH5

Without tone With PM tone

–6 –7 –8 –9 –10 –11 –29.0 –28.0 –27.0 –26.0 –25.0 –24.0 Measured received power (dBm)

0

–10

–20 –60

–40

–20 0 20 Frequency deviations (GHz)

40

60

FIGURE 9.22 Frequency monitoring errors measured after transmission of 10-Gb/s signals over 320 km of SMF. The inset shows BER curve measured when PM pilot tone was added to optical signal.25

9.3 Performance monitoring techniques using PM and FM pilot tones

243

phases of the lasers were modulated at 500–508 MHz (peak deviation ¼ 0.5p) with a channel separation of 2 MHz. The horizontal axis represents the optical frequency deviation from the ITU-standardized frequency of each optical channel. In this measurement, we intentionally changed the optical frequency of each laser up to 50 GHz using a temperature controller. The optical frequencies of five WDM channels could be measured with an accuracy of better than 3 GHz when the WDM lasers were operating within 40 GHz from the ITU-standardized frequencies. The monitoring errors were mainly attributed to the uneven transmission characteristics of AWG. When phase-modulated optical signals propagate through an optical amplifier and an optical filter, undesirable AM components can be generated because of the PM-AM conversion. However, the performance of this technique cannot be deteriorated because the transmission characteristics of the conventional optical amplifier and optical filter have nearly flat responses at the operating frequencies (e.g., the typical transmission slopes of our optical amplifiers and optical filters were less than 0.001 and 0.01 dB/GHz, respectively). Thus, we did not observe any significant AM components because of the PM-AM conversion after the transmission over a 320-km link (including five two-stage EDFAs and one optical filter). Moreover, the performance of the monitoring technique based on PM pilot tones could not be deteriorated because of the PM-AM conversion caused by the CD after the transmission over a 320-km SMF (residual dispersion ¼ 200 ps/nm) by using the optimum tone frequencies. In addition, the inset shows the BER curve measured when a PM pilot tone was added to the optical signal. These results confirm that the PM pilot tone does not degrade the receiver sensitivity.

9.3.2 Using FM pilot tones Figure 9.23 shows the experimental setup used for demonstrating the principle of the optical channel monitoring technique based on FM pilot tones.26 We used seven DFB lasers operating at 192.4–193.6 THz (channel spacing ¼ 200 GHz). The optical frequency of each laser was slightly dithered with a small sinusoidal current (3 mA) ranging from 10 to 16 kHz with a separation of 1 kHz. The peak deviation of the laser frequency induced by the sinusoidal current was measured to be 0.3–0.56 GHz. However, this dithering also generated a small AM component. For the demonstration of this principle, we canceled out this AM component by using an additional intensity modulator. However, for the practical use of this technique, it would be necessary to employ either distributed Bragg reflector (DBR) lasers (i.e., wavelength-tunable sources) or temperature dithering of DFB lasers for the generation of pure FM pilot tone. The outputs of seven lasers were combined using a star coupler and then modulated at 10 Gb/s (pattern length ¼ 231  1) using a LiNbO3 modulator. The modulated signals were first sent to a 13-km-long SMF for decorrelation. The transmission link was composed of eight 80-km-long SMFs. Thus, the total transmission distance was 640 km. The EDFA module consisted of a two-stage EDFA and DCF as an interstage component. The gain of each EDFA module was identical to the span loss of an 80-km-long SMF (18 dB). A small portion of the WDM signal was then tapped and sent to the monitoring module, which consisted of a 18 AWG (channel spacing: 200 GHz, crosstalk: 32 dB), eight photodetectors, an A/D converter, and a microprocessor. When the frequency-modulated WDM signal passed through an AWG, the optical power became modulated because of the transmission characteristics of the AWG. Although an AWG was used in this experiment, it should also be possible to use either a Fabry-Perot etalon filter or FBG. Figure 9.24 shows the amplitudes of AM components (tones) of channel 3 operating at 192.8 THz (tone frequency ¼ 12 kHz) measured by using PD 3 and PD 4 and the ratio between these

244

CHAPTER 9 Optical performance monitoring based on RF pilot tones

DCF

10 kHz

Laser Laser 11

11 kHz

Laser 22 Laser

12 kHz

Laser Laser 33

13 kHz

Laser Laser 44

14 kHz

Laser 55 Laser

15 kHz

Laser Laser 66

13

80

80

Modulator 640 km SMF

10 Gb/s

Multiwavelength meter

Laser Laser77

Modulator Modulator Phase Phase Shifter shifter

16 kHz

Processor Processor

FFT

A/D A/D

PD1 PD2 PD3 PD4 PD5 PD6 PD7 PD7 PD8

Attenuator

AWG

FIGURE 9.23

0

25

–5

20

–10

A

B

15 10

–15

5

–20

0

–25 –30 –35

–5 A/B

–10 –15 –20

–40 –25 192.74 192.76 192.78 192.8 192.82 192.84 192.86 Optical frequency (THz)

FIGURE 9.24 Amplitudes and ratio of pilot tones (channel 3) measured using PD 3 and PD 4.26

Ratio of tone amplitude (dB)

Tone amplitude (dB)

Experimental setup.26

20

5

16

4

12

3

8

2

Power error (dB)

Frequency error (GHz)

9.4 Dispersion monitoring techniques for adaptive compensators

4 0 –4 –8

1 0 –1 –2

–12

–3

–16

–4

–20 –80

–60

–40 –20 0 20 40 Frequency deviation (GHz)

(a)

60

80

245

–5 –80

–60

–40 –20 0 20 40 Frequency deviation (GHz)

60

80

(b)

FIGURE 9.25 (a) Frequency and (b) power errors of seven WDM signals measured after 640-km transmission over SMF.26

amplitudes. In this figure, A represents the amplitude of the AM component measured by PD 3, while B represents the amplitude obtained by PD 4. Thus, when the WDM channel is operating precisely at 192.8 THz, the ratio, A/B, should be 0 dB. This ratio curve was nearly linear with the optical frequency deviation of channel 3. Thus, the optical frequencies of all WDM channels could be estimated by measuring the ratios of tone amplitudes at the corresponding tone frequencies and AWG ports. The optical power of each WDM channel could also be monitored easily by measuring their corresponding tone amplitudes. Figure 9.25 shows the frequency and power errors of seven WDM signals measured after a 640-km transmission over SMF using this technique. The horizontal axis represents the optical frequency deviation from the ITU-standardized frequency of each WDM channel. In this measurement, we intentionally changed the optical frequency of each laser up to 60 GHz using temperature controllers. The optical frequencies and powers of seven WDM channels could be measured with accuracies better than 4 GHz and 1 dB, respectively, when the lasers were operating within 30 GHz from the ITU-standardized frequencies. When frequency-modulated optical signals propagate through an optical fiber, undesirable AM components could be generated by FM-AM conversion because of the CD. However, in our experiment, these AM components were not observed since we used sufficiently low tone frequencies.28

9.4 DISPERSION MONITORING TECHNIQUES FOR ADAPTIVE COMPENSATORS Recently, the capacity of the WDM network has been increased significantly by using a large number of high-speed (>10 Gb/s) channels. In such a high-speed network, the most important limiting factors would be CD and PMD. In order to overcome these limitations, various dispersion compensators have been proposed and demonstrated. However, in a dynamic network, the WDM channel

246

CHAPTER 9 Optical performance monitoring based on RF pilot tones

could experience different CD and PMD whenever the network is reconfigured. In addition, both CD and PMD are sensitive to ambient temperature. Thus, for the efficient compensation of CD and PMD, we should be able to monitor them accurately. In earlier studies, it has been reported that CD and PMD could be monitored by using high-frequency pilot tones.8–11,29 However, the main obstacle in these techniques is the difficulty in differentiating the effects of CD and PMD. In this section, we review the CD and PMD monitoring techniques based on pilot tones and discuss their limitations and possible solutions.

9.4.1 CD monitoring techniques using AM and PM pilot tones The CD of each WDM channel can be monitored by measuring either the magnitude of the AM pilot tone8,29 or the magnitude of the AM component converted from the PM pilot tone (i.e., PM-AM conversion).10 However, these pilot-tone-based monitoring techniques can be affected by both SPM and PMD.28 Thus, in this section, we compare the performances of these CD monitoring techniques based on the high-frequency pilot tones by considering SPM and PMD.

9.4.1.1 Operating principles When an optical signal with AM or PM pilot tone is propagated along an optical fiber, the phase difference between the lower and upper sidebands could be changed by CD.8,10 As a result, the magnitude of the transmitted pilot tone would be changed. Thus, we could monitor the CD of a WDM channel by measuring the magnitude of the received AM or PM pilot tone. The electrical powers of the received AM (PAM) and PM pilot tones (PPM) could be described as  2 2 2 2 pDLl f ðAM pilot toneÞ (9.8) PAM / m cos c and  PPM / J02 ðAÞJ12 ðAÞ sin2

pDLl2 f 2 c

 ðPM pilot toneÞ;

(9.9)

where c is the speed of light, l is the wavelength, m and f are MI and tone frequency, D and L are the dispersion parameter and fiber length, A is the depth of phase modulation, and J0 and J1 are the firstkind Bessel functions of the order 0 and 1, respectively. Equation (9.8) shows that the magnitude of the AM pilot tone decreases with an increase in the fiber length. However, if the tone frequency f were too low, the magnitude of the AM pilot tone would remain nearly constant. The resolution of this technique (i.e.,
PAM/
(DL)) could be improved by increasing the tone frequency, but it would reduce the measurement range. Thus, the tone frequency should be optimized to obtain both reasonable resolution and measurement range. The dispersion of 1000 ps/nm would cause a 1-dB power penalty for 10-Gb/s NRZ signals. Our calculation shows that, when the tone frequency is 8 GHz, the measurement range is given by 0–1000 ps/nm. Thus, in most of the previous reports, a high-frequency tone (
8 GHz) was used for the monitoring of CD.8,29 In contrast, Equation (9.9) shows that the magnitude of the AM component generated by the PM-AM conversion would increase with an increase in CD even when the tone operates in the low-frequency region. Thus, unlike the techniques using AM pilot tones, a technique using a relatively low-frequency (
2 GHz) PM pilot tone would make it possible to monitor the value of CD.

9.4 Dispersion monitoring techniques for adaptive compensators

LD

IM or or IM PM PM

IM

247

Power Power meter meter

BPF BPF SMF PD

PRBS PRBS

10 Gb/s 231–1

Signal generator RFSA RFSA

(a) 12 PM pilot tone (2 GHz) AM pilot tone (8 GHz) Calculation

Modulation index (%)

10 8 6 4 2 0

0

400

800 1200 1600 Chromatic dispersion (ps/nm)

2000

(b)

FIGURE 9.26 (a) Experimental setup to measure the CD. (b) MI of pilot tones measured at the receiver while varying the CD.2

Figure 9.26(a) shows the experimental setup used for comparing the performances of the dispersion monitoring techniques using AM and PM pilot tones. The output power and operating wavelength of the DFB laser were set to be 3 dBm and 1550 nm, respectively. The output of the DFB laser was modulated at 10 Gb/s (pattern length ¼ 231  1) by using a LiNbO3 intensity modulator. We added an AM pilot tone to the 10-Gb/s NRZ signal by using an additional LiNbO3 intensity modulator (or a PM pilot tone by using an additional LiNbO3 phase modulator). The output signal of the transmitter was sent to the SMF via an EDFA and a variable attenuator. After transmission, the electrical power of the received pilot tone was measured by using an electrical spectrum analyzer. The signal power incident on the photodetector was adjusted to be 10 dBm by using another variable attenuator. Figure 9.26(b) shows the MIs of AM and PM pilot tones measured at the receiver while varying CD (i.e., using different lengths of the fiber) in comparison with the theoretically calculated lines. The frequencies of the AM and PM pilot tones were set to be 8 and 2 GHz, respectively. The depth of phase modulation was set to be 0.09p. The results show that both techniques provide a resolution of approximately 0.006%  ps/nm. However, unlike in the case of the technique using high-frequency (8-GHz) AM pilot tones, in the case of the technique using the relatively low-frequency (2-GHz) PM pilot tone, the measurement range could be extended to >15,000 ps/nm.

248

CHAPTER 9 Optical performance monitoring based on RF pilot tones

9.4.1.2 Effects of SPM and PMD When the high-frequency pilot tone is used, the performance of the CD monitoring technique could be deteriorated by SPM.28 This is because the AM component of the pilot tone could modulate the refractive index of optical fiber and cause phase modulation because of the SPM. The SPM-induced phase modulation would then generate an additional AM component because of the PM-AM conversion that occurs in the dispersive fiber, which, in turn, causes monitoring errors. Figure 9.27(a) shows the SPM-induced monitoring errors measured while varying the signal power incident on the 60-km-long SMF (dispersion ¼ 1000 ps/nm), in comparison with theoretically calculated curves.30 When the signal power incident on the SMF was 10 dBm, the monitoring errors of the techniques using AM and PM pilot tones were measured to be 266 ps/nm and 18 ps/nm, respectively. Thus, the dispersion monitoring technique using the AM pilot tone was considerably more sensitive to SPM than the technique using the PM pilot tone. This was mainly because we set the frequency of the AM pilot tone (8 GHz) to be significantly higher than that of the PM pilot tone (2 GHz) in order to obtain similar resolutions. In general, the effect of SPM increases with an increase in the square of the tone frequency.30 Both CD monitoring techniques based on AM and PM pilot tones could be affected by PMD since the magnitude of the pilot tone decreases with an increase in the PMD. In general, the effect of PMD on the power of the received AM pilot tone can be described as28 P / 1  4gð1  gÞ sin2 ðpf DtÞ;

(9.10)

where g and
t are the power ratio between the slow and fast axes and the differential group delay (DGD), respectively. Thus, we measured the effect of PMD on these techniques after replacing the SMF with a PMD emulator in Figure 9.26(a) (CD ¼ 0 ps/nm). Figure 9.27(b) shows the maximum PMD-induced errors measured while varying the DGD of the PMD emulator. When the DGD was 30 ps, the maximum monitoring error was measured to be 420 ps/nm for the technique using AM pilot tones, while it was less than 30 ps/nm for the technique using PM pilot tones. Thus, 800 Monitoring error (ps / nm)

Monitoring error (ps/nm)

100 0 −100 −200

PM pilot tone (2 GHz) AM pilot tone (8 GHz) Calculation

−300 −400 −6

−4

−2

0 2 4 6 Input power (dBm) (a)

8

10

12

PM pilot tone (2 GHz) AM pilot tone (8 GHz) Calculation

600 400 200 0 −200

0

10

20 30 DGD (ps)

40

50

(b)

FIGURE 9.27 Effects of SPM and PMD on pilot-tone-based CD monitoring technique. Monitoring errors caused by (a) SPM and (b) PMD.2

9.4 Dispersion monitoring techniques for adaptive compensators

249

we concluded that the monitoring technique using PM pilot tones exhibits a superior performance to the technique using AM pilot tones by suppressing the effects of both SPM and PMD.

9.4.2 CD monitoring technique using chirped pilot tone The CD monitoring techniques based on AM and PM pilot tones cannot determine the sign of accumulated dispersion since the magnitude of a pilot tone is dependent on only the amount of dispersion. Thus, in this section, we introduce the monitoring technique based on the direct detection of high-speed chirped pilot tones that can monitor both the sign and the amount of CD simultaneously. Figure 9.28 shows the operating principle of the CD monitoring technique using chirped pilot tones.31 The chirped (i.e., FM and AM are generated simultaneously) pilot tones can be generated by using the direct modulation of DFB LD. When an optical signal with an FM pilot tone is propagated along the optical fiber, the phase difference between the lower and upper sidebands could be changed by CD.10,25 As a result, the magnitude of the transmitted pilot tone would be changed (i.e., FM-AM conversion). When the total dispersion is in the normal dispersion regime (3000 < dispersion < 0 ps/nm) and the tone frequency is lower than 2 GHz, the AM component generated by the FM-AM conversion constructively interferes with the original AM pilot tone. In contrast, in the anomalous dispersion regime (0 < dispersion < 3000 ps/nm), as shown in Figure 9.28, the overall magnitude of the pilot tone decreases with an increase in the CD because of the destructive interference.28,32 In this case, the interference depends on the tone frequency and the relative phase difference between the FM-AM component and the original AM tone. Using these phenomena, we could monitor both the sign and the amount of CD simultaneously. Figure 9.29 shows the experimental setup. The output power and operating wavelength of the DFB laser were set to be 3 dBm and 1550 nm, respectively. We added a chirped pilot tone to the output of the DFB laser by modulating the injection current (peak modulation current ¼ 1 mA). When the injection current of the DFB laser is modulated, the lasing frequency of the DFB laser becomes modulated in addition to the amplitude modulation. This is mainly because the ambient temperature of the active region is modulated when the modulation frequency is lower than a few MHz.33 In this case, the peak deviation of the lasing frequency (
n) caused by thermal modulation decreases with an increase in the modulation frequency since the thermal response of a DFB laser is inversely

Amplitude

Amplitude

Time

Time FM-AM conversion

Amplitude

AM pilot tone

AM pilot tone

Time FM pilot tone

Optical frequency

Optical fiber (anomalous region)

Operating frequency of WDM signal

FIGURE 9.28 Operating principle of CD monitoring technique using chirped pilot tones.31

Time

250

CHAPTER 9 Optical performance monitoring based on RF pilot tones

Power meter

DFB laser

Amplitude modulator EDFA Attenuator

RF signal

Filter

Fiber

Pattern generator (10 Gb/s, 231–1) Polarization controller

Attenuator

PD

Spectrum analyzer

EDFA

PMD emulator

FIGURE 9.29 Experimental setup used for demonstrating CD monitoring technique based on chirped pilot tone.31

proportional to the frequency. In contrast, when the modulation frequency is higher than a few MHz, the frequency deviation increases with an increase in the frequency. This is because the lasing frequency is modulated because of the carrier-density modulation (i.e., frequency chirp). The frequency-deviation rate of the DFB laser was measured to be 0.2 GHz/mA when the modulation frequency was 2 GHz. The output of the DFB laser was modulated at 10 Gb/s (pattern length: 231  1) by using a LiNbO3 intensity modulator. The output signal of the transmitter was sent to various optical fibers via an EDFA and a variable attenuator. The transmitted signal was filtered by using an optical filter and then sent to a high-speed PD. The input signal of PD was controlled to maintain a constant power (3 dBm) by adjusting an optical attenuator. The spectrum analyzer was used for measuring the electrical power of the pilot tone. Figure 9.30 shows the electrical tone power measured while varying the CD (fiber input power ¼ 5 dBm, and tone frequency ¼ 0.4, 0.8, 1, 2, 4, 6 GHz). The measured data are normalized to the

Normalized tone power (dB)

10 5 0 0.4 GHz 0.8 GHz 1 GHz 2 GHz 4 GHz 6 GHz

–5 –10 –15 –1500

–1000

–500 0 500 Dispersion (ps/nm)

FIGURE 9.30 Normalized tone power measured while varying CD.31

1000

1500

9.4 Dispersion monitoring techniques for adaptive compensators

251

tone power at 0 ps/nm. The total link dispersion value for which the tone produces a maximum signal in direct detection is shifted away from 0 ps/nm. For directly modulated laser diodes, this maximum is shifted from 0 ps/nm to a negative dispersion (normal dispersion regime), the exact value of which depends on the modulation frequency. The results show that the tone frequency should be lower than 4 GHz in order to increase the measurement range (>700 ps/nm). However, it is necessary to increase the tone frequency to improve the monitoring resolution. Thus, we found out that the optimum tone frequencies were in the range of 2–3 GHz (monitoring resolution: <50 ps/nm @ 2 GHz). When the tone frequency is 2 GHz, the measured curve was nearly linear with the dispersion (3000 < dispersion < 3000 ps/nm). Thus, the CD could be estimated by measuring the tone amplitudes. However, the performance of the pilot-tone-based monitoring techniques could be affected by PMD and SPM.28 Thus, using a PMD emulator, we measured the effects of PMD on this technique, as shown in Figure 9.29. Figure 9.31 shows the maximum PMD-induced monitoring error measured while varying the DGD (dispersion ¼ 0 ps/nm). The monitoring error could be estimated simply by measuring the change in the received tone power caused by the PMD. In this case, the monitoring error could not be changed by the dispersion because the measured tone power had a linear relationship with the dispersion, as shown in Figure 9.30. The results show that the maximum monitoring error was measured to be less than 100 ps/nm even when the DGD was set to be 50 ps. The 100-ps/nm monitoring error would be acceptable for a practical 10-Gb/s transmission system because it is approximately one-tenth of the total amount of dispersion (i.e., 1000 ps/nm), which causes a 1-dB system penalty. Figure 9.32 shows the SPM-induced monitoring error measured while varying the signal power incident on the 60-km-long SMF. Even when the signal power was increased to be 6 dBm, the monitoring error was measured to be less than 100 ps/nm. In a practical 10-Gb/s WDM transmission system, the DGD and fiber launch power are usually designed to be less than 50 ps and 6 dBm, respectively. However, even when the DGD and the fiber launch power are higher than 50 ps/nm and 6 dBm in a particular system, Figures 9.31 and 9.32 show that the performance of this monitoring technique could not be seriously deteriorated by PMD and SPM. 160 Monitoring error (ps/nm)

140 120 100 80 60 40 20 0

0

10

20 30 40 Differential group delay (ps)

FIGURE 9.31 Maximum monitoring error due to PMD measured while varying DGD.31

50

60

252

CHAPTER 9 Optical performance monitoring based on RF pilot tones

Monitoring error (ps/nm)

0 –50 –100 –150 –200 –250 –300 –350

0

2

4 6 8 Fiber input power (dBm)

10

12

FIGURE 9.32 Monitoring error due to SPM measured while varying signal power incident on SMF.31

9.4.3 CD monitoring technique using pilot tone carried by broadband light source In this section, we describe the monitoring technique using the AM pilot tone carried by a broadband light source (BLS). The performance of this technique is almost insensitive to the PMD unlike that of the conventional AM pilot-tone-based monitoring technique. Figure 9.33 shows the experimental setup used for demonstrating the CD monitoring technique based on the AM pilot tone carried by the BLS. Thus, the AM pilot tone can be obtained by directly modulating the reflective semiconductor optical amplifier (RSOA) (instead of the transmitter lasers) as in Reference 34. In order to monitor the CD experienced by each WDM channel, the output of the RSOA is combined with the optical signal by using a coupler. Thus, the AM pilot tone traverses the same optical path with the signal. Hence, the pilot tone and signal are exposed to the same amount of the CD. However, because of the broad spectral width of the RSOA, the pilot tone is more sensitive to the CD than the optical signal. Thus, by using this technique, we can utilize a lower-frequency pilot tone compared to the case of using the conventional technique based on the AM pilot tone.

10-Gb/s DPSK PD

VOA

Tx

SMF 0.6 nm

EDFA

EDFA

RFSA

1 nm

RSOA f = 2 GHz

FIGURE 9.33 Experimental setup used for demonstrating CD monitoring technique based on pilot tone carried by BLS.34

9.4 Dispersion monitoring techniques for adaptive compensators

253

When the pilot tone is generated by using the BLS, the power variation in the received AM pilot tone, caused by the CD, can be expressed as   2  p (9.11) P / exp  DLsl f 2 where sl is the spectral width of the BLS, D is the dispersion parameter, L is the fiber length, and f is the tone frequency.35 This expression shows that the frequency of the pilot tone required for monitoring the CD of up to 1000 ps/nm is only approximately 2 GHz (assuming that the spectral width of the BLS is 0.8 nm), while it is
8 GHz when the conventional monitoring technique based on the AM pilot tone is used.8 Since this technique can utilize the low-frequency pilot tone, the performance of this technique is less sensitive to PMD. Figure 9.33 shows the experimental setup used for evaluating the performance of this technique. We modulated the output of the DFB laser operating at 1549.32 nm with a 10-Gb/s differential phase-shift keying (DPSK) signal by using a LiNbO3 modulator. The pilot tone was generated by directly modulating an RSOA at 2 GHz. The MI of the pilot tone was 100%. We combined the output of the RSOA (carrying a 2-GHz AM pilot tone) with the 10-Gb/s DPSK signal and sent it to an optical bandpass filter. This filter (3-dB bandwidth ¼ 0.6 nm) was used for simulating the AWGs used in WDM networks. Thus, the 10-Gb/s DPSK signal traversed together with the 2-GHz AM pilot tone (which was carried by 0.6-nm amplified spontaneous emission [ASE] light centered around the signal). The combined signals passed through the SMF in various lengths. After passing through the SMF, the power of the received AM pilot tone was measured by using a photodiode (PD) and an RF spectrum analyzer (RFSA). When this technique is used, the BER performance of the 10-Gb/s DPSK signal can degrade because of the ASE light carrying the AM pilot tone. Thus, in order to minimize this degradation, it is necessary to limit the output power of the RSOA. In order to evaluate this limit, we measured the power penalty for the 10-Gb/s DPSK signal caused by the 0.6-nm ASE light. Figure 9.34 shows 4.0

Power penalty (dB)

3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0 –40 –38 –36 –34 –32 –30 –28 –26 –24 –22 –20 Power difference (dB)

FIGURE 9.34 Measured power penalty due to pilot tone.34

254

CHAPTER 9 Optical performance monitoring based on RF pilot tones

the power penalty measured as a function of the power difference between the signal and the ASE light. In this figure, we defined the power difference as the ratio between the output power of the RSOA (within the optical bandwidth of 0.1 nm) and the total power of the DPSK signal. The power penalty increased with an increase in this power difference. The power penalty of 0.5 dB was observed at the power difference of 28 dB. Thus, throughout this experiment, we set the power difference to be 28 dB. We measured the power of the AM pilot tone while varying the CD from 0 to 1000 ps/nm. These reference values of the CD of the SMF were measured by using the modulation phase-shift method.36 Figure 9.35 shows the measured power of the AM pilot tone (normalized with the value measured at the CD of 0 ps/nm) in comparison with the theoretically calculated curve using Equation (9.11). The measured data agreed well with the calculated curve. The power of the AM pilot tone (carried by the sliced ASE light) decreased continuously as a function of the CD (caused by the use of the BLS). This is in marked contrast with the results obtained by other pilot-tone-based dispersion monitoring techniques (where the power of the pilot tone is changed periodically).28 Thus, by using this technique, it is possible to increase the monitoring range to be wider than that of the previous pilottone-based techniques. It should be noted that the performance of the CD monitoring technique based on the AM pilot tone is sensitive to PMD.28 However, this technique should be less sensitive to PMD than the conventional AM-tone-based techniques because of its use of a relatively low tone frequency. In order to evaluate the effect of PMD, we compared the performance of this technique and that of the conventional techniques based on the AM pilot tone while varying DGD from 0 to 50 ps by using a commercial PMD emulator. For this comparison, we set the frequency and MI of the AM pilot tone to be 8 GHz and 12%, respectively, for the conventional technique in order to maintain the pilottone-induced power penalty within 0.5 dB.35 The CD was set to be 360 ps/nm. Figure 9.36 shows the maximum PMD-induced variations (i.e., @ g ¼ 0.5) in the powers of the received AM pilot tones measured as a function of DGD. In this figure, the calculated curves were obtained by using

Relative power (dB)

0

Calculation Measurement

–4 –8 –12 –16 –20 –24

0

200

400 600 800 1000 Chromatic dispersion (ps/nm)

FIGURE 9.35 Measured power variation in received pilot tone as a function of CD.34

1200

9.4 Dispersion monitoring techniques for adaptive compensators

255

Tone power variation (dB)

0 –2 –4 –6 –8 Calculation Proposed technique (2-GHz AM tone) Conventional technique (8-GHz AM tone)

–10 –12

0

10

20

30 DGD (ps)

40

50

FIGURE 9.36 Effect of PMD on received power of pilot tones.34

Equation (9.10). The result shows that, when the conventional technique is used, the power of the received pilot tone can be reduced by an additional 3 dB (because of PMD) at the DGD of 30 ps. This result indicates that the monitoring error can be as large as 420 ps/nm when the DGD is 30 ps.28 In contrast, when this technique was used (i.e., the tone frequency was 2 GHz), the monitoring error was measured to be less than 20 ps/nm under the same conditions. Since the RSOA has a broad spectral width (
30 nm), this technique can monitor the CD of multiple WDM signals by using single RSOA (instead of dithering every WDM transmitter). Figure 9.37 shows the experimental setup used for evaluating the performance of this technique in a WDM system. We multiplexed eight DFB lasers (channel spacing ¼ 100 GHz) and modulated these WDM signals at 10 Gb/s in the DPSK format. The WDM signals were then combined with the ASE light (generated by RSOA) carrying the 2-GHz AM pilot tone and traversed through the SMF link. The CD of this link was set to be either 400 or 620 ps/nm. After demultiplexing each signal by using an optical bandpass filter, we measured the power of the received AM pilot tones by using this monitoring technique. Figure 9.38 shows the measured CDs and the monitoring errors of the eight WDM

LD1

10-Gb/s DPSK

193.0 THz PD

VOA SMF LD8 193.7 THz

EDFA RSOA

FIGURE 9.37 Experimental setup to demonstrate in WDM system.34

EDFA

0.6 nm

RFSA

256

CHAPTER 9 Optical performance monitoring based on RF pilot tones

700

150

Measured CD (ps/nm)

50 500

0 –50

400 –100 300

1

2

3

4 5 6 Channel number

7

8

Measured CD error (ps/nm)

100 600

–150

FIGURE 9.38 Measured dispersion and monitoring error of each WDM signal.34

channels. From this result, we confirmed that this technique could measure the CD of every channel within the monitoring error of 20 ps/nm.

9.4.4 PMD monitoring technique using SSB pilot tone Because of the random birefringence of the optical fiber, PMD has statistical and time-varying characteristics. Thus, an accurate PMD monitoring technique is vital for the efficient compensation of PMD. Recently, it has been proposed that PMD be monitored by using pilot tones.11 However, as discussed in Section 9.4.1.2, this technique is sensitive not only to PMD but also to CD. In this section, we describe a monitoring technique using SSB pilot tones to mitigate this problem.2 Equation (9.10) indicates that the detected magnitude of the pilot tone, P, decreases with an increase in PMD. Thus, we could monitor the PMD of the WDM signal by measuring the magnitude of the pilot tone. For the use of pilot tones in the PMD monitoring technique, we should first determine the proper tone frequency. Figure 9.39 shows the magnitudes of pilot tones measured while varying the tone frequency and DGD. For this measurement, we used the same experimental setup shown in Figure 9.26(a) after replacing the SMF with a PMD emulator. The state of polarization (SOP) of the signal incident on the PMD emulator was adjusted to minimize the magnitude of the detected pilot tone (i.e., g ¼ 0.5). The result shows that the resolution of this technique (i.e.,
P/
(DGD)) could be improved by increasing the tone frequency, although it would reduce the measurement range. In contrast, when DGD was larger than 50 ps, the power penalty (@BER ¼ 109) was measured to be greater than 5 dB for the 10-Gb/s NRZ signal (pattern length ¼ 231  1). Thus, we concluded that the proper tone frequency would be 10 GHz if the maximum DGD were smaller than 50 ps. Under this condition, the MI of pilot tone should be set to a value less than 14% in order to maintain the tone-induced power penalty within 0.5 dB (refer to Figure 9.3(b)).

9.4 Dispersion monitoring techniques for adaptive compensators

257

Measured pilot tone (a.u.)

1.2 1.0

DGD = 0 ps

0.8

DGD = 20 ps

0.6 DGD = 30 ps 0.4 Calculation Measurement

0.2

DGD = 40 ps

DGD = 50 ps 0.0

0

2

4 6 8 10 Tone frequency (GHz)

12

14

FIGURE 9.39 Magnitude of pilot tone versus tone frequency and DGD.2

The accuracy of the PMD monitoring technique based on the high-frequency pilot tone could be deteriorated by CD. This is because the phase difference between the lower and upper sidebands of a pilot tone could be changed by CD, which, in turn, decreases the magnitude of the pilot tone as CD increases. This problem could be solved by using a SSB pilot tone for the monitoring of PMD. The SSB pilot tone is inherently insensitive to CD since it has only one sideband. Figure 9.40 shows the magnitudes of SSB pilot tones (tone frequency ¼ 10 GHz) measured while varying both CD and 5

0

Measured pilot tone (dB)

Measured pilot tone (dB)

5

–5 –10

CD = 0 ps/nm CD = 170 ps/nm CD = 340 ps/nm CD = 680 ps/nm

–15 –20 –25 –30 –35

0

10

20 30 DGD (ps) (a)

40

50

0 –5 –10 –15

CD = 0 ps/nm CD = 170 ps/nm CD = 340 ps/nm CD = 680 ps/nm

–20 –25 –30 –35

0

10

20 30 DGD (ps)

40

(b)

FIGURE 9.40 Magnitudes of pilot tones measured while varying both DGD and CD (a) when DSB pilot tone is used, and (b) when SSB pilot tone is used.2

50

258

CHAPTER 9 Optical performance monitoring based on RF pilot tones

DGD, in comparison with the conventional double-sideband (DSB) pilot tones. As expected, the magnitude of the DSB pilot tone decreased as CD increased. However, by using the SSB pilot tone, we could substantially reduce this sensitivity to CD and monitor PMD with a reasonable accuracy even when CD was as high as 680 ps/nm.

9.5 SUMMARY In this chapter, we reviewed various optical performance monitoring techniques based on the pilot tone. These techniques utilize a small sinusoidal signal, operating at a unique frequency, imposed on each WDM channel. As a result, various optical parameters, such as optical channel powers, wavelengths, and optical paths of multiple WDM channels can be monitored simultaneously without the need to demultiplex them. However, in the case of using the AM pilot tone, the receiver sensitivity in on-off keying (OOK) system can be degraded as the pilot tone interferes with the data signal. If necessary, we can avoid this AM-tone-induced sensitivity degradation simply by utilizing a highpass filter within the optical receiver. On the other hand, when the frequency of the AM pilot tone is lower than 10 MHz, the performance of this monitoring technique can be seriously deteriorated by the ghost tones originating from XGM and SRS. In fact, these effects limit the maximum network size that this monitoring technique can support. Thus, it appears that the monitoring technique based on the low-frequency AM pilot tone is most useful for metro network applications. For the use of this monitoring technique in the long-haul network, we need to either restrict the number of channels that can be monitored at a time (e.g., by using an optical bandpass filter) or increase the tone frequency to be higher than 300 MHz. It is also possible to mitigate this limitation by utilizing the PM and FM pilot tones. The pilot-tone-based monitoring technique can also be used for the adaptive compensation of CD and PMD. However, one of the key issues here is the difficulty of separating the effects of CD and PMD. There have been many attempts to solve this problem. For example, to monitor the CD without the effects of SPM and PMD, it is advantageous to utilize the PM pilot tone rather than the AM pilot tone. The same objective can also be achieved by utilizing either the chirped pilot tone or the AM pilot tone carried by the BLS. For the accurate monitoring of PMD, we can utilize the SSB pilot tone and improve the tolerance to CD. Recently, it has been reported that the pilot-tone based optical performance monitoring technique is well suited for the use in DPSK systems.35–36 For example, the pilot-tone based technique can simultaneously monitor 100 times more WDM channels (without using optical demultiplexers) in DPSK systems than in conventional OOK systems.35 In addition, by using the PM pilot tone, we can accurately monitor the frequency offset between the optical signal and the delayed interferometer (DI) in a direct detection DPSK system.36 In the near future, the optical network is expected to utilize various types of advanced modulation formats such as quadrature phase-shift keying (QPSK), quadrature amplitude modulation (QAM), and orthogonal frequency-division multiplexing (OFDM). Thus, it is needed to further improve the pilot-tone based optical performance monitoring technique for the use in these next-generation optical networks.

References

259

REFERENCES 1. Bonenfant PA. Tutorial on protection and restoration in optical networks. In: Proc. optical fiber communication conference, paper FF1. San Diego, CA; 1999. 2. Ji HC, Park KJ, Lee JH, Chung HS, Son ES, Han KH, et al. Optical performance monitoring techniques based on pilot tones for WDM network applications. J Opt Network 2004;3(7):510–33. 3. Hill GR, Chidgey PJ, Kaufhold F, Lynch T, Sahlen O, Gustavsson M, et al. A transport network layer based on optical network elements. IEEE J Lightwave Technol 1993;11(5/6):667–79. 4. Ji HC, Park KJ, Kim JK, Shin SK, Jang MJ, Chung YC. Demonstration of optical path monitoring technique using dual tones in all-optical WDM transport network. In: Proc. optoelectronics and communication conference, paper 14A4-4. Chiba, Japan; 2000. 5. Park KJ, Shin SK, Chung YC. Simple monitoring technique for WDM networks. IEE Electron Lett 1999;35(5):415–7. 6. Youn CJ, Shin SK, Park KJ, Chung YC. Optical frequency monitoring technique using arrayed-waveguide grating and pilot tones. IEE Electron Lett 2001;37(16):1032–3. 7. Ji HC, Kim JK, Chung YC. Optical path and crosstalk monitoring technique using pilot tones in all-optical WDM transport network. In: Proc. Asia-Pacific optical and wireless communications conference, paper 4584-2. Beijing, China; 2001. 8. Petersen MN, Pan Z, Lee S, Havstad SA, Willner AE. Online chromatic dispersion monitoring and compensation using a single inband subcarrier tone. IEEE Photon Technol Lett 2002;14(4):570–2. 9. Dimmick TE, Rossi G, Blumenthal DJ. Optical dispersion monitoring technique using double sideband subcarriers. IEEE Photon Technol Lett 2000;12(7):900–2. 10. Park KJ, Youn CJ, Lee JH, Chung YC. Chromatic dispersion monitoring technique in WDM network. In: Proc. optical fiber communication conference, paper ThGG88. Anaheim, CA; 2002. 11. Motaghian Nezam SMR, Wang Y, Hauer M, Lee S, Willner AE. Simultaneous PMD monitoring of several WDM channels using subcarrier tones. In: Proc. conference of lasers and electro-optics, paper CFE1. Baltimore, MD; 2001. 12. Murakami M, Imai T, Aoyama M. A remote supervisory system based on subcarrier overmodulation for submarine optical amplifier systems. IEEE J Lightwave Technol 1996;14(5):671–7. 13. Chung HS, Shin SK, Park KJ, Woo HG, Chung YC. Effects of stimulated Raman scattering on pilot-tone based supervisory technique. IEEE Photon Technol Lett 2000;12(6):731–3. 14. ITU-T Study Group 15. Tone modulation for suppressing stimulated Brillouin scattering and for channel identification on systems using in-line OFAs and WDM. Contrib Quest 1994;25/15:. 15. Park PKJ, Kim H, Chung YC. Effects of low cut-off frequency of optical receiver on the performance of lightwave systems using pilot tones. Opt Commun 2006;261(2):245–8. 16. Olsson NA. Lightwave systems with optical amplifiers. IEEE J Lightwave Technol 1989;7(7):1071–82. 17. Lam AW, Tantaratana S. Theory and applications of spread-spectrum systems. Piscataway, NJ: IEEE Press; 1994 [chapter 6]. 18. Sun Y, Saleh AAM, Zyskind JL, Wilson DL, Srivastava AK, Sulhoff JW. Time-dependent perturbation theory and tones in cascaded erbium-doped fiber amplifier systems. IEEE J Lightwave Technol 1997;15(7):1083–7. 19. Park PKJ, Kim CH, Chung YC. Performance analysis of low-frequency pilot-tone-based monitoring techniques in amplified wavelength-division-multiplexed networks. Opt Eng 2008;47(2):025009. 20. Srivastava AK, Zyskind JL, Sun Y, Ellson J, Newsome G, Tkach RW, et al. Fast-link control protection of surviving channels in multiwavelength optical networks. IEEE Photon Technol Lett 1997;9(12):1667–9. 21. Kong E, Tong F, Ho K-P, Chen L-K, Chan C-K. An optical-path supervisory cross-connects using pilottones, In: Proc conference of lasers and electro-optics, paper FT4. Baltimore, MD; 1999.

260

CHAPTER 9 Optical performance monitoring based on RF pilot tones

22. Chan C-K, Kong E, Tong F, Chen L-L. A novel optical-path supervisory scheme for optical cross connects in all-optical transport network. IEEE Photon Technol Lett 1998;10(6):899–901. 23. Chu KU, Lee CH, Shin SY. Optical path monitoring based on the identification of optical cross-connect input ports. In: Proc. optical fiber communication conference, paper FJ5. Baltimore, MD; 2000. 24. Kim JK, Ji HC, Chung HS, Kim CH, Shin SK, Hyun DH, et al. Demonstration of fast restorable all-optical WDM network. IEICE Trans Commun 2001;E84-B(5):1119–26. 25. Park KJ, Ji HC, Chung YC. Optical channel monitoring technique using phase-modulated pilot tones. IEEE Photon Technol Lett 2005;17(11):2481–3. 26. Park KJ, Youn CJ, Lee JH, Chung YC. Optical path, wavelength, and power monitoring technique using frequency-modulated pilot tones, In: Proc optical fiber communication conference, paper FF1. Los Angeles, CA; 2004. 27. Phillips MR, Ott DM. Crosstalk caused by nonideal output filters in WDM lightwave systems. IEEE Photon Technol Lett 2000;12(8):1094–6. 28. Park KJ, Youn CJ, Lee JH, Chung YC. Performance comparisons of chromatic dispersion-monitoring techniques using pilot tones. IEEE Photon Technol Lett 2003;15(6):873–5. 29. Rossi G, Dimmick TE, Blumenthal DJ. Optical performance monitoring in reconfigurable WDM optical networks using subcarrier multiplexing. IEEE J Lightwave Technol 2000;18(12):1639–48. 30. Park KJ, Youn CJ, Lee JH, Chung YC. Effect of self-phase modulation on group-velocity dispersion measurement technique using PM-AM conversion. IEE Electron Lett 2002;38(21):1247–8. 31. Park PKJ, Jun SB, Chung YC. Chromatic dispersion monitoring technique based on chirped pilot tones. Opt Commun 2006;266(1):280–3. 32. Yoshida S, Iwashita K. Influence of amplitude modulation induced by LD direct modulation on FM signal transmission. IEEE Photon Technol Lett 1990;2(12):929–31. 33. Welford D, Alexander SB. Magnitude and phase characteristics of frequency modulation in directly modulated GaAlAs semiconductor diode lasers. IEEE J Lightwave Technol 1985;3(5):1092–9. 34. Choi HY, Park PKJ, Chung YC. Chromatic dispersion monitoring technique using pilot tone carried by broadband light source. IEEE Photon Technol Lett 2009;21(9):578–80. 35. Jun SB, Kim H, Park PKJ, Lee JH, Chung YC. Pilot-tone-based WDM monitoring technique for DPSK systems. IEEE Photon Technol Lett 2006;18(20):2171–3. 36. Ji HC, Park PKJ, Kim H, Lee JH, Chung YC. A novel frequency-offset monitoring technique for directdetection DPSK systems. IEEE Photon Technol Lett 2006;18(8):950–2.

CHAPTER

Optical performance monitoring based on electronic digital signal processing

10

Fabian N. Hauske*, Maxim Kuschnerov{ *Huawei Technologies, European Research Center, Munich, Germany { University of the Federal Armed Forces, Munich, Germany

10.1 INTRODUCTION Future dynamically reconfigurable all-optical networks will add more flexibility to legacy transport networks, replacing static point-to-point links by routed optical paths with switched wavelengths.1 By adaptive routing, the channel characteristics are altered depending on the selected path, which requires an adaptive equalizer to compensate for the residual deterministic distortions. Even today’s existing systems equipped with different modulation formats at various data rates due to the ongoing upgrade from 10 Gb/s to 40 Gb/s and to 100 Gb/s,2 and several reconfigurable optical add/drop modules (ROADMs), induce time-varying distortions that have to be compensated in order to increase required system margins and to guarantee a reliable quality of service (QoS).3,4 At the same time, the increasing demand for higher transmission capacities boosted the development of high-speed transmission, increasing the bit rate per channel from 10 Gb/s to 100 Gb/s and beyond.5,6 Due to the limited bandwidth within the optical transmission band, spectrally efficient higher-order modulation formats have been introduced in combination with coherent detection.6,7 As these modulation formats employ all physical dimensions to encode data, polarization in particular, a fasttracking and compensation of time-varying polarization effects is required. It is critically important for network providers to prevent network failures to avoid large data loss. In order to react quickly on any detrimental influence in the network, it is not sufficient to rely on the bit error rate (BER) prior to the forward error correction (FEC) or on the bit-interleaved parity (BIP) bytes of the synchronous optical network (SONET) layer or the synchronous digital hierarchy (SDH) layer to extend fault management and to guarantee the QoS. Besides, these parameters do not allow identification of the failure’s origin. Thus, a parameter monitoring directly from the optical layer is required.8,9 Various techniques to measure optical signal parameters are known that typically tap the optical signal and provide an analysis based on the properties of the optical spectrum.8–11 Alternatively, techniques like asynchronous sampling12 or electrical eye monitoring13 provide analysis of the transmitted signal in the electrical domain. Another possibility for performance monitoring directly evaluates the channel estimate from known sequence training, as it is done in orthogonal frequency-division multiplexing (OFDM) transmission.14 Several other techniques are presented throughout this book. All previously mentioned technologies have one thing in common: They require external devices, © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00010-9

261

262

CHAPTER 10 OPM based on electronic digital signal processing

additional resources, or access to the transmitter. Despite the extra cost of such devices, they either do not rely on the information carrying signal itself (e.g., pilot tones), or they do not include the properties of the optical front end and the electrical receiver with all possible influences by filters, amplifiers, or analog-to-digital converters (ADCs). However, those influences can significantly alter the signal quality and should be included in the channel parameter estimation. In this chapter we focus on single-carrier systems without polarization multiplexing in case of direct detection systems and with polarization multiplexing in case of coherent detection systems. Blind adaptation is assumed throughout the paper,6 although similar results are obtained with known sequence training. Since digital signal processing (DSP) has met the requirements of high-speed processing, digital equalization of optical transmission links has opened a cost-efficient solution to mitigate optical (and electrical) distortions. Furthermore, it relaxes system requirements, increasing the robustness and flexibility in optical networks.15,17 Analyzing equalizer filter coefficients allows one to conclude the kind and the strength of various optical impairments, opening the way for optical performance monitoring (OPM) based on electronic DSP. A prerequisite to such equalizer-based parameter estimations is the exact knowledge of the channel model and signal statistics. Optical transmission systems can be classified according to the down-conversion and demodulation of the optical signal from the transmission band into the baseband at the receiver. As an essential part of the total channel transfer function, each demodulation scheme calls for a certain family of equalizers. The two main types are direct detection and coherent detection. Direct detection systems apply simple arrangements of square-law devices, such as photodiodes that transfer the modulation of the optical intensity into the electrical domain. With a delay interferometer (DI) in the optical receiver front end, which converts differential phase information into amplitude information, the subdivision of the differential detection systems also allows to transfer the phase information of the optical signal into the electrical domain. Binary modulation formats such as on-off keying (OOK) and differential (binary) phase-shift keying (D(B)PSK), as well as differential quaternary phase-shift keying (DQPSK), are the main types in direct detection systems that have been deployed and investigated in recent years.18 By the nature of the square-law device, the transmission channel between the modulator at the transmitter and the detection at the receiver becomes a nonlinear channel. Efficient receiver-side electrical equalization requires algorithms based on the maximum-likelihood sequence estimation (MLSE) principle.19 Direct detection systems with MLSE are typically dispersion limited, with residual chromatic dispersion (CD) and optical signal-to-noise ratio (OSNR) being the main impairments to be monitored. The influence of fiber nonlinearities and polarization mode dispersion (PMD) is secondary. The square-law detection in direct detection systems only transfers the intensity of the optical field into the electrical domain. Thus, OPM based on the analysis of the equalizer can retrieve limited information about the optical channel. In the following, the possibilities of OPM in direct detection systems with equalization by means of MLSE are presented in Section 10.2. After a short introduction to the channel model (Section 10.2.1), the noise statistics prior to the ADC and the state model of the equalizer based on MLSE are briefly described (Section 10.2.2). The state model, which usually serves the equalizer to generate the metrics, is then employed for OPM. A systematic approach to separate the contributions from deterministic signal components and noise leads to an estimation of the OSNR (Section 10.2.3), followed by a method to identify individual deterministic signal distortions such as CD or self-phase modulation (SPM) based on

10.2 OPM in digital direct-detection systems

263

built-in references (Section 10.2.4). In a short conclusion, we sum up and discuss the estimation results (Section 10.2.5). In contrast, coherent detection systems employing down-conversion with the aid of a local oscillator (LO) and two optical 90
hybrids arranged for a polarization-diverse demodulation, transfer the full information with respect to amplitude, phase, and polarization from the optical domain into the electrical domain. This allows for polarization-division multiplexing (PDM) in combination with rich signal constellations employing both amplitude and phase. Common higher-order modulation formats are quaternary phase-shift keying (QPSK), also called 4-quadrature amplitude modulation (QAM), 16-QAM,7 or higher orders of QAM.20 Due to the linear transfer function of the coherent receiver, the transmission channel can be described as a merely linear channel, where only the degree of optical nonlinearities weakens this assumption. Due to the unitary nature of CD and PMD, electrical equalization can be obtained by linear finite impulse response (FIR) filters. In coherent detection systems, complete information of the optical signal is available in the electrical domain, which allows for extensive OPM based on equalizer properties. After describing all elements contributing to the linear channel model (Section 10.3.1.1), we discuss the methods to identify the channel parameters from the properties of the equalizer (Section 10.3.1.2). Joint estimation of all parameters is demonstrated based on simulations and on measured data from off-line experiments, considering the mutual influence on the estimation in particular (Section 10.3.2). Two methods to estimate the value of residual CD are demonstrated (Section 10.3.2.1). Each method refers to a respective equalization structure, one for compensated links (see “Residual CD” section) and one for uncompensated links (see “Dispersion Estimation in Uncompensated Link” section). Within the polarization effects, estimations are presented of the instantaneous spectrum of differential group delay (DGD), mean DGD, and state of polarization (SOP) (Section 10.3.2.2). Methods to obtain an accurate estimation of polarization-dependent loss (PDL) are shown (Section 10.3.2.3), and finally, a conclusion rounds up the results (Section 10.3.3). In Section 10.4, we sum up the results of the presented OPM methods and discuss their possibilities and limitations. Also, a prospect for future developments in this field is given.

10.2 OPM IN DIGITAL DIRECT-DETECTION SYSTEMS Understanding parameter extraction from the analysis of equalizer properties requires the knowledge of equalization principles and techniques. Direct detection systems are by nature nonlinear systems due to the square-law characteristics of the photodiode. Transferring only the intensity of the optical signal into the electrical domain, it is impossible to recover the absolute information of the optical phase. In addition, differential detection systems that allow transferring differential phase information into the electrical domain are nonlinear systems due to the self-homodyne down-conversion within the DI of the optical front end. Only with special DSP techniques, such as multisymbol phase estimation (MSPE)21,22 or optical field reconstruction,23 can the information of the optical field be recovered. However, MSPE is a decision-directed method that cannot be combined with equalization by MLSE, and the technique of optical field reconstruction seems to be out of scope due to the high implementation complexity. However, the required implementation complexity only allows for equalization of binary modulation formats with state-of-the-art technology.15,16 Concepts for

264

CHAPTER 10 OPM based on electronic digital signal processing

higher-order modulation formats have been solely presented in offline experiments or with simulated data.24,26 The optimum equalizer in direct detection systems employs MLSE, which delivers decisions based on the signal statistics of a sequence of received samples. It does not provide an improved decision variable, nor a compensated signal or closed-form compensation function. Thus, the estimation of single deterministic channel parameters is difficult. A range of publications discuss the signal statistics that are crucial for equalizer performance,27,28 but very little has been published on performance monitoring.29,30 In the following, we present a systematic separation of deterministic signal components and statistical noisy distortions based on the analysis of MLSE equalizer properties. This allows the estimation of the noise power, which relates to the given OSNR value. Furthermore, we provide a reference-based method to distinguish between single deterministic signal distortions.

10.2.1 The channel model for direct-detection systems In the following, we provide a simplified channel model for direct detection with a single photo diode that covers the typical case of widely deployed 10-Gb/s binary and pseudo-multilevel systems.18 Existing ADC technology in combination with digital equalizers allows an effective mitigation of impairments in such systems. A similar model can be derived for differential detection systems. All deterministic signal distortions like CD, PMD, SPM, and filtering, as well as noise, are combined in the signal rðtÞ ¼ F frð f Þg

(10.1)

with rðtÞ ¼ ½rX ðtÞ; rY ðtÞT defined by its components rX ðtÞ and rY ðtÞ according to the two polarization states. Each component can be described by amplitude and phase or in a complex valued notation giving rðtÞ four dimensions of freedom. For the linear or weakly nonlinear optical channel where Kerr nonlinearities can be neglected, the signal rð f Þ results from rð f Þ ¼ Hð f Þsð f Þ þ nð f Þ

(10.2)

with the transmitted signal sð f Þ ¼ ½sX ð f Þ; sY ð f ÞT affected by the linear channel transfer function Hð f Þ and additive white Gaussian noise (AWGN) nð f Þ ¼ ½nX ð f Þ; nY ð f ÞT . It should be noted that the components sX ð f Þ and sY ð f Þ do not carry different information as would be the case with polarization multiplexing. Furthermore, we define gð f Þ ¼ Hð f Þsð f Þ with gðtÞ ¼ ½gX ðtÞ; gY ðtÞT as the component containing only the deterministic channel distortions. For simplicity, we refer to a model with a single concentrated element of noise loading prior to the receiver instead of several noise elements distributed along the transmission link, as would be the case for several amplification stages along the link. A detailed description of the linear channel transfer function including the effects of CD, PMD, and PDL is given in Section 10.3.1.1. If Kerr nonlinearities cannot be neglected, the received signal rðtÞ can be obtained with the aid of the nonlinear Schro¨dinger equation (NLSE), including the effects of SPM and crossphase modulation (XPM).31 The well-known, split-step algorithm approximates the solution of the NLSE by a finite elements approach.30 Strictly speaking, optical and electrical filtering in the receiver jointly affect signal and noise. We will comment on this effect in Section 10.2.3.

10.2 OPM in digital direct-detection systems

265

The photodiode detects the optical intensity by 2 2 2 2 r~ðtÞ ¼ jrX ðtÞj2 þ jrY ðtÞj2 ¼ rX;Re ðtÞ þ rX;Im ðtÞ þ rY;Re ðtÞ þ rY;Im ðtÞ;

(10.3)

which is fed into the ADC stage, sampled at t ¼ t0 þ nTs ; and quantized to yield r~ðnÞ. The parameter t0 defines the sampling phase with t0 ¼ 0 centered in the middle of each symbol, and Ts defines the duration between two sampling instants. Given the symbol duration T, the ratio T=Ts ¼ 1 refers to single-fold sampling with one sample per symbol, and T=Ts ¼ 2 refers to two-fold over-sampling with two samples per symbol. Higher sampling rates are not advisable for equalization. It becomes clear from Equation (10.3) that separation between the contributions from X and Y polarization is not possible, and that all linear and nonlinear distortions are superimposed. Thus, systematic identification of single channel impairments from r~ðnÞ is out of scope. However, the deterministic signal components and noise contributions can be identified from the signal statistics in a state-based model, which is typically applied in digital equalizers employing the MLSE principle.

10.2.2 State-based equalization based on MLSE

^ of length N, which is the argument of the maximization The MLSE delivers an estimated sequence d of the maximum likelihood (ML) criterion by ^ ¼ arg maxfpð~rjdðNÞ Þg; d i

(10.4)

dðmÞ

ðNÞ

where ~r defines the vector of the received sequence, di defines one possible transmitted message ðNÞ out of i 2 f1; 2; . . . ; 2N g, and pð~rjdi Þ defines the joint probability density function (PDF) that a ðNÞ sequence ~r has been received under the condition of a sent message di . Assuming statistical independent samples r~ðnÞ, we can simplify the ML criterion into ðNÞ

pð~rjdi Þ ¼

N Y n¼1

ðNÞ

pð~ r ðnÞjdi Þ;

(10.5)

ðNÞ

the product of individual PDFs pð~r ðnÞjdi Þ.32 Given the typical filter bandwidths, single-fold sampling and two-fold over-sampling satisfy the statistical independence of adjacent samples.33 The Viterbi algorithm implements the MLSE principle with a final-state machine reducing the length of dðNÞ to subsets dðmÞ covering only the channel memory length m. It applies iterative steps of adding a branch metric ldðmÞ ðnÞ on top of a path metric Ldðm1Þ ðnÞ ¼ Ldðm1Þ ðn  1Þ þ ldðmÞ ðnÞ;

(10.6) ðm1Þ

comparing two path metrics Ldðm1Þ ðnÞ that lead to the same sequence di i

and selecting

ðmÞ

the one with the larger probability. Each state transition d is composed of the state of its origin ðm1Þ ðn  1Þ ¼ dðm1Þ and a possible decision d. For example, for a memory length of m ¼ 3, the state di ðmÞ

½dn2 ; dn1  ¼ ½0; 1 evolves into the state transitions di ðnÞ ¼ ½dn2 ; dn1 ; dn  ¼ ½0; 1; 0 and ðmÞ d ðnÞ ¼ ½dn2 ; dn1 ; dn  ¼ ½0; 1; 1. In the next step, the “oldest” decision dn2 is omitted, giving the j

ðm1Þ

two states di

ðm1Þ

ðnÞ ¼ ½dn1 ; dn  ¼ ½1; 0 and dj

ðnÞ ¼ ½dn1 ; dn  ¼ ½1; 1. It is clear also that

266

3

CHAPTER 10 OPM based on electronic digital signal processing

⫻10–3

2.5

Amplitude (a.u.)

111 2 0.06

1.5

011

110

0.04

010 1

0.02 0

101

0.5

001

0 0

1/4

000 1/2 Time (t/T )

(a)

2

100 3/4

4

6 Quan

8 tizatio 10 12 n bin

14

111 110 101 n 100 sitio 011 ran 010 te t 001 Sta 16 000

1

(b)

FIGURE 10.1 ðmÞ

Electrical noise-free eye prior to ADC at 1700-ps/nm residual CD with symbol patterns di ðm ¼ 3Þ aligned to each interference pattern (a) and according to lookup table obtained at sampling instant t0 ¼ 0 (1 sample/symbol, 4-bit quantization) at OSNR ¼ 14 dB (b). ðmÞ ðm1Þ dl ðnÞ ¼ ½dn2 ; dn1 ; dn  ¼ ½1; 1; 1 competes for the same state dj ðnÞ ¼ ½dn1 ; dn  ¼ ½1; 1. Only 34 the one with the higher aligned path metric will survive. ðmÞ In reality, we are only interested in the PDFs pð~ r jdi Þ. If m equals or exceeds the channel memory, each state transition will be governed by one deterministic signal component superimposed by the ðmÞ same stationary noise process. In other words, each decision pattern di leads to a distinct interference pattern and vice versa. The interference patterns can be observed easily in a noise-free eye diagram, as depicted in Figure 10.1(a). It can be seen that the channel memory slightly exceeds the memory ðmÞ length m of the state model. Each decision pattern di actually covers a small bundle of traces. Given a residual CD of 1700 ps/nm, the memory length should be sufficient, but filtering induces additional intersymbol interference extending the channel memory length. Within the accuracy of the ADC resolution, this effect should be minor. For the OSNR estimation described in the next section, a memory length of m ¼ 5 is chosen. Because the variable r~ is quantized, we can represent the complete set of all possible PDFs in a matrix, also called a lookup table, with coordinates defined by the quantization bin and the state transition, respectively. An example of such a lookup table is given in Figure 10.1(b). The lookup tables of the equalizer can be based on histograms15 or on parametric descriptions of the channel properties.27 The latter imply previous parameter estimation itself. They are obtained by blind channel acquisition, starting with default values and performing a continuous update based on tupels ^ ^ ð~ r ðnÞ; dðnÞÞ of received samples r~ðnÞ and aligned equalizer decisions dðnÞ. In the following, parameter estimation based on histograms is presented under the assumption of a sufficient optical and electrical filter bandwidth, such that the signal statistics can be described by higher-order chi-square PDFs. The statistically independent Gaussian noise processes before square-law detection can be described by their mean values gX;Re ; gX;Im ; gY;Re ; and gY;Im and the variance s2 in each dimension.

10.2 OPM in digital direct-detection systems

267

The total noise power from ASE refers to 4s2 . As shown in Equation (10.3), the four dimensions of the Gaussian noise process are transformed into a fourth-order, noncentral chi-square distribution by square-law detection. The noncentrality parameter is defined by the contributions of all four mean values32: g2 ¼ g2X;Re þ g2X;Im þ g2Y;Re þ g2Y;Im :

(10.7) ðmÞ

It is now clear that the PDFs stored in the lookup table are chi-square distributed. Each PDF pð~ r jdi Þ is ðmÞ defined by the noncentrality parameter g2i aligned to the state transition di that generates the according interference pattern. Unlike the Gaussian distribution, the first-order moment and the second-order moment of the chi-square statistics are governed by both the noncentrality parameter and the noise power 4s2 . Thus, we cannot read the parameters from a single PDF. Instead, we compare the statistical ðmÞ ðmÞ properties of two distinct PDFs aligned to two distinct transitions di and dj ði 6¼ jÞ. For each transition, the first- and second-order moments are given, respectively, by27   ðmÞ E r~jdi (10.8) ¼ 4s2 þ g2i ;     2  2 ðmÞ ðmÞ : E r~2 jdi ¼ 8s4 þ 4s2 g2i þ 4s2 þ g2i ¼ 8s4 þ 4s2 g2i þ E r~di

(10.9)

^ to After equalization, we can sort the received samples r~ðnÞ with the aid of the detected symbols dðnÞ calculate the first- and second-order moments as N   ðmÞ  1 X  ^ðmÞ ; ^ ~ r j d E r~jd ¼ j i i N j¼1

(10.10)

N    2 X ^ðmÞ ¼ 1 ^ðmÞ : E r~2 jd r~j jd (10.11) i i N j¼1  ðmÞ     ðmÞ    ^ ^ðmÞ ; E r~jd ^ðmÞ ; we obtain27 ^ By use of E r~jd ; E r~2 jd ; and E r~2 jd i i j j    ðmÞ     ðmÞ  ^ ^ ^ðmÞ  E2 r~jd ^ðmÞ þ E2 r~jd E r~2 jd  E r~2 jd i i j j  ðmÞ   ðmÞ  4s2 ¼ ; (10.12) ^ ^ ~ E r~jd  E r j d i j  ðmÞ  ^ , we can furthermore calculate As the total noise power 4s2 is identical to each E r~jd i  ðmÞ  ^ g2i ¼ E r~jd (10.13)  4s2 ; i

which delivers a noise-free representation of the deterministic channel distortions at the sampling instant. Thus, through this state-based approach, we can separate the deterministic channel distortions from the statistical channel distortions. In our model, we neglect the impact of the optical bandpass filter and the electrical low-pass (LP) filter necessary for channel separation and noise suppression, respectively. Both filters alter the noise process. The optical filter reduces the total noise power and modifies the white noise

268

CHAPTER 10 OPM based on electronic digital signal processing

process into a colored noise process. Additional intersymbol interference might be induced to the deterministic signal components. However, we can still assume white noise at the sampling instant given typical optical filter bandwidths that only induce a negligible correlation between adjacent samples.33 After the square-law device in Equation (10.3), mixed terms with signal and noise do not allow a separate treatment of both effects. Electrical LP filtering again reduces the noise power. It can be shown with the aid of the characteristic function that LP filtering modifies the fourth-order chi-square distribution into the convolution of weighted higher-order chi-square distributions. For tight narrowband filtering, the resulting distribution tends toward a Gaussian bell curve. However, the mean value and the variance of this distribution still jointly depend on the noise power as well as the deterministic signal components. Based on our experience, narrowband filtering with cut-off frequencies down to 0.4/T (T: symbol duration) do not affect parameter estimation. A similar case applies for differential detection with balanced photodiodes. The resulting PDF can be obtained by the convolution of the fourth-order chi-square distributions after each photodiode ðNÞ ðNÞ r PD2 jdi Þ. Note: The minus sign in the second term is due to subtraction in the pð~ r PD1 jdi Þpð~ balanced detector. Obviously, this does not lead to a higher-order chi-square distribution, as the balanced signal is defined on the positive and negative axes. The resulting PDF still remains slightly asymmetric, but turns into a Gaussian shape. The interested reader is referred to Reference 35. It has been shown that equalization based on metrics with a Gaussian assumption delivers excellent results.25 As mentioned before, the Gaussian assumption is not useful for parameter estimation because it does not allow separation between signal and noise. Applying an estimation based on Equations (10.10)–(10.13) regardless of this fact still led to reasonable parameter estimation with a slightly decreased precision.

10.2.3 State-based OSNR estimation The value of the OSNR is one of the core parameters to assess the signal quality of an optical transmission line. It requires knowledge of the total signal power g2, the noise bandwidth Bn, and the total noise power 4s2 . Methods to monitor the signal from histograms, which lack a systematic separation between signal and noise, have been previously published.36,37 Also, a state-based approach to estimate statistical parameters for parametric metrics is proposed in Reference 27, but no evaluation with respect to OSNR or other signal parameters is given. However, this approach yields a reliable OSNR estimation with only a little more effort.30 The total signal power can be easily calculated by 2 1 X g2 ; m 2 i¼1 i m

g2 ¼

(10.14)

which allows us to define a signal-to-noise ratio (SNR) r¼

g2 : 4s2

(10.15)

If we take into account that the OSNR is defined with a noise bandwidth of Dl ¼ 0:1 nm around the carrier wavelength l0 (carrier frequency f0 ), we can scale the SNR to OSNR ¼ r

Bn f0 Dl l0

k ¼ kscale r:

(10.16)

10.2 OPM in digital direct-detection systems

269

The factor k accounts for the receiver sensitivity, noise factor, and other detailed properties of the receiver. The noise bandwidth Bn is defined by the bandwidths of the filters in the optical and electrical domains. If a narrow LP filter in the electrical domain is applied, specifying the noise bandwidth exactly is difficult, because after square-law detection, the signal and the noise terms are partly mixing. The SNR will still be proportional to the OSNR, where the relation can be simply adjusted by a scaling factor kscale . The scaling factor can be obtained from detailed simulations including all receiver properties, or it can be calibrated with the aid of a reference measurement at a known OSNR. In principle, the noise term 4s2 also contains noisy distortions added by the electrical circuitry, quantization noise, and statistical distortions induced by XPM from neighboring channels. Typically, those distortions are low compared to the influence of the ASE noise. However, whether those influences should be referred to as “OSNR” is a question of definition. If the proposed algorithm is applied, the estimation provides the in-band OSNR. In the following, we present estimations of the OSNR based on Monte Carlo simulations that include ASE noise, quantization noise, and XPM contributing to the total noise power. A transmission link with 10-Gb/s non-return-to-zero (NRZ)-OOK modulation over 100 km of standard single-mode fiber (SSMF), with dispersion-compensating fiber (DCF) to adjust the residual dispersion, was simulated. At the receiver, AWGN was added before the optical bandpass filter (Gaussian, 2nd order, Bopt ¼ 25 GHz) with a subsequent electrical LP filter (Bessel, 20th order, Bel ¼ 7 GHz). Single-fold sampling with one sample per symbol is applied, and a Viterbi equalizer with 16 states and 32 state transitions (m ¼ 5) provides the digital decisions based on histograms. Each estimation is based on a block of 106 samples, which refers to an estimation of the first- and second-order moments based on N ¼ 106/32 ¼ 31,250 samples for each transition in average (see Equations (10.10) and (10.11)). All OSNR estimations are normalized with a calibration at OSNR ¼ 10 dB. Figure 10.2 shows the estimated OSNR at a residual CD of 1700 ps/nm and variations of the sampling instant. A high quantization resolution for the estimation of the first- and second-order moments was employed, which refers to analog samples. In this case, the noise power consists only of contributions from ASE noise. Within a range of 2–15 dB the estimation shows excellent linear ðmÞ ðmÞ behavior without significant deviation. As long as the two chosen transitions di and dj belong to distinct traces in the eye diagram (compare Figure 10.1(a)), the estimation is independent from the sampling instant. Furthermore, it is independent from deterministic distortions given that the memory length of the channel does not exceed the equalizer memory. Figure 10.3 shows that there is also no difference between different pulse shapes and OOK-modulation formats with NRZ and duobinary (DB) modulation. Further evaluations provide equal results for return-to-zero (RZ) formats. However, if the pulse returns to zero, no information can be retrieved in this region, which results in a clear sampling phase dependency with the estimation preferably obtained from the center of the symbol. The influence of various sources of noise is shown in Figure 10.3. The given OSNR only refers to the contribution of ASE noise. It can be observed that for high OSNR values, the contribution of quantization noise becomes visible, which decreases the estimated OSNR compared to the given OSNR. This effect becomes even stronger for a 3-bit quantization. However, for a 4-bit quantization, the effect is rather minor with a deviation of less than 1 dB at a given OSNR of 15 dB. This relates to the low BER penalty of the MLSE for such a quantization resolution where even a 3-bit ADC delivers a respectable performance.38

CHAPTER 10 OPM based on electronic digital signal processing

Estimated OSNR (0.1 nm) (dB)

270

15 13 11 9 7 5 3 15 ve 13 n O 11 SN 9 R (0. 7 5 1n m) 3 (dB )

Gi

–t/2

+t/2

+t/4 0.0 –t/4 phase Sampling

FIGURE 10.2 State-based OSNR estimation at a residual CD of 1700 ps/nm and variations in the sampling phase t0.

Estimated OSNR (0.1 nm) (dB)

15

NRZ analog samples NRZ 4-bit quantization DB analog samples NRZ analog samples + XPM NRZ analog samples + SPM

13 11 9 7 5 3 1

3

5

7 9 11 Given OSNR (0.1 nm) (dB)

13

15

FIGURE 10.3 State-based OSNR estimation for NRZ and DB modulation with analog and quantized samples, and for transmission with significant SPM and XPM.

Another noise source is XPM, which was simulated with three channels to each side of the channel under observation (total of seven channels) at high launch powers. In this case, the contribution of XPM strongly affects the OSNR estimation with an offset of nearly 5 dB. Furthermore, the influence of SPM with a launch power of 15 dBm is shown, where the system is strongly degraded and the channel

10.2 OPM in digital direct-detection systems

271

memory already exceeds the filter memory. For low OSNR values, the noise dominates the system, which still results in an accurate estimation. As the OSNR increases, the deterministic signal components dominate and the influence of SPM starts to degrade the estimation where deterministic signal components are interpreted as noise. This results in an underestimation for high OSNR values. All evaluations prove a systematic overestimation for low OSNR values, which slightly depends on the degree of the signal distortions. The electrical LP filter affects both deterministic signal components and noise. The influence of the first is met by the state model, where the influence of the latter should be met by the PDF model. If noise dominates the system, electrical filtering alters the PDF from the chi-square distribution toward a Gaussian distribution. In this case, Equations (10.8) and (10.9) are no longer valid. To prove the concept, measured data with offline processing was applied.29 A detailed description of the experimental setup is given in the following section. Various configurations of combined launch power Plp and residual CD have been evaluated. The given OSNR was measured in the optical domain with an optical spectrum analyzer (OSA). The estimated OSNR is based on a state model with 25 state transitions. Upon over-sampling with two samples per symbol, the estimated OSNR is the average of the estimation on each sample. Table 10.1 compares measured and estimated OSNR for selected combinations of strong signal distortions. The estimated OSNR shows accurate results independent from the deterministic signal distortions, which is in excellent agreement with simulations carried out in Reference 30. As described in the previous section, it is difficult to obtain an accurate description of the PDF for differential detection systems. Nonetheless, assume a chi-square distribution led to reasonable results that permitted estimation of the OSNR in case of DBPSK and DQPSK modulation. In addition, the application of over-sampling with two samples per symbol is possible, where only one sample is employed for the estimation or an average over the results from both samples could be applied. In summary, OSNR estimation proves to be accurate as long as the equalizer memory is sufficient. It is robust for any sampling instant, as well as any direct detection modulation format and intrachannel deterministic distortion. As noisy distortions induced by quantization or XPM also degrade system performance, we believe that their influence is not detrimental to the estimation but rather reflects a reliable picture of the actual system condition.

10.2.4 Referenced parameter estimation As mentioned previously, Equation (10.13) provides a noise-free representation of the signal at the sampling instant. As the photodiode irreversibly combines amplitude and phase distortions, there is no systematic approach to separate the contributions from each deterministic channel distortion. Table 10.1 Comparison of Measured and Estimated OSNR from Measured Data for Various Combinations of Residual CD and Launch Power Condition Plp Plp Plp Plp

¼ ¼ ¼ ¼

0 dBm, 0 dBm, 0 dBm, 15 dBm,

Measured OSNR (dB) CD CD CD CD

¼ ¼ ¼ ¼

1250 1250 3050 3050

ps/nm ps/nm ps/nm ps/nm

10.0 13.0 14.3 15.0

Estimated OSNR (dB) 10.3 13.1 14.0 14.8

272

CHAPTER 10 OPM based on electronic digital signal processing

Still, every combination of impairments will lead to a specific eye diagram, and thus we propose an estimation of combined channel distortions based on references. Therefore, we compare the probe set g2i of the received signal to a pool of references g2i;ref with known parameters. The references are either obtained from simulations or from a reference measurement. In both cases, the properties of the receiver should be identical to the actual system, as all filtering effects and receiver characteristics will affect the signal. The best matching reference indicates the estimated parameters by Qref ;min

2m  2 X ¼ min g2i  g2i;ref ; ref

(10.17)

i¼1

which refers to the minimum Euclidean distance between the set g2i and g2i;ref . In the following, we show an example parameter estimation of combined residual CD and launch power–induced SPM, as demonstrated in Reference 29. We verify our method by both Monte Carlo simulations and measured data. Measurements were carried out for a 10-Gb/s NRZ-OOK (pseudo-random binary sequence, PRBS 2151), single-channel transmission with variations in launch power Plp, residual CD (adjusted by according transmission line), and OSNR. Up to CD ¼ 1700 ps/nm, the transmission is only single span; beyond this point, two spans with reamplification are applied. After a 35 GHz optical bandpass filtering, and a photodiode, a sampling oscilloscope (2 samples per symbol, 8-bit ADC, 8-GHz bandwidth) saved 2  106 samples of the probe signal. Offline processing with timing recovery implemented by a resampling routine and a 16-state Viterbi equalizer (32 state transitions, m ¼ 5) was applied. The references for given parameters of Plp ¼ 0, 2, . . . , 18 dBm and CD ¼ 0, 200, . . . , 3600 ps/nm were built from simulations under the same channel conditions. First, the noise power and OSNR are estimated (for results, see previous section). After noise-free components g2i have been obtained, they are compared to the references. Two results are presented in Figure 10.4, which show all Euclidean distances Qref . The solid curves show Qref when the channel was simulated, and the wireframe curves show Qref when the channel was measured. It can be clearly seen that in both cases only one global minimum establishes where the gradient with respect to residual dispersion is steeper than the gradient toward Plp. This typically results in a more accurate estimation for CD compared to Plp. The global minimum indicates Qref;min , which is aligned to the estimated parameters for CD and Plp. The shapes of Qref for simulated data represent the ideal case because the references were built by simulated data under the same conditions. The shape for measured data is almost identical despite a small offset, which proves the accuracy of the model for reference simulations. In case of a given combination of Plp ¼ 0 dBm with CD ¼ 1250 ps/nm, the estimated parameters were Plp,estim ¼ 0 dBm with CDestim ¼ 1000 ps/nm (Figure 10.4(a)). In the case of a given combination of Plp ¼ 15 dBm with CD ¼ 3050 ps/nm, estimated parameters were Plp,estim ¼ 16 dBm with CDestim ¼ 2800 ps/nm (Figure 10.4(b)). The results comprise a quite accurate estimate for the given single channel with a maximum of two spans. It is clear that over-sampling provides a better representation of the noise-free eye, which leads to greater estimation accuracy. Basically, all direct detection modulation formats can be applied if the respective reference is provided. This is also the weak point of the estimation. The reference should model the channel as accurately as possible. Covering all possible combinations of parameters to provide the references is

10.2 OPM in digital direct-detection systems

Measured

6

Qref (normalized)

273

4

2

18

lp

6

3000

2400

1800 1200 CD (ps/nm)

0 600

P

0 3600

(d Bm )

12 Simulated

0

(a)

Qref (normalized)

6

4 Measured

2

6

Bm P lp (d

)

0 0

Simulated

12 18 0

600

1200 1800 2400 CD (ps/nm)

3000

3600

(b)

FIGURE 10.4 Two examples for reference-based estimation of launch power–induced SPM and residual CD for (a) Plp ¼ 0 dBm, CD ¼ 1250 ps/nm, and (b) Plp ¼ 15 dBm, CD ¼ 3050 ps/nm. The black circles indicate the given parameter set, and the white arrow indicates the estimated parameter set. Note: The diagram is rotated in (b).

almost impossible, especially in switched networks with influence from narrowband filtering and in long fiber links with nonlinearities distributed over several spans. Unfortunately, flexible switched optical networks are predestined for monitoring those parameters. Furthermore, the model relies on the choice of the sampling instant, which is controlled by the timing recovery circuit of the

274

CHAPTER 10 OPM based on electronic digital signal processing

receiver. Although over-sampling makes the equalizer performance robust to slight variations of the sampling phase,38 this reference-based parameter estimation strongly depends on the sampling instant, as the reference is bound to a certain sampling phase. In conclusion, the estimation of combined deterministic channel parameters in direct detection systems is more difficult than and not as reliable as estimation of the OSNR. Unfortunately, there are no other references in this field that provide OPM cost effectively as a by-product of the electrical equalizer.

10.2.5 Conclusion In conclusion, OPM from the state model of MLSE equalizers in direct detection systems is elaborate and only allows for limited access to individual channel parameters. Although there is a systematic separation of deterministic signal components and noisy distortions, the identification of individual deterministic distortions like CD or PMD is not possible, nor is the identification of different noise sources. As the parameter space can be reduced to typical operation points of the system, a reference-based method comparing the actual system state to a set of known system states delivers respectable results. However, as soon as the system exceeds the reference parameter space or if different system conditions refer to similar state models, the estimation becomes unreliable. Similarly, the estimation of pure OSNR is altered by the influence of XPM. With significant SPM, the channel memory length is likely to exceed the equalizer memory length, which also degrades the estimation (compare discussion in Section 10.3.3). Although the MLSE principle might be an interesting candidate to solve problems with nonlinearity in coherent detection systems, the importance of direct detection systems with equalizers based on MLSE is likely to decline in the future. Thus, we see no urgent need to further investigate the reliability and robustness of proposed methods or to develop alternative concepts based on DSP in direct detection systems.

10.3 OPM IN DIGITAL COHERENT RECEIVERS In contrast to direct detection, the polarization-diverse coherent receiver preserves all signal properties and linearly transfers the optical field into the electrical domain. The coherent optical front end in combination with DSP for data recovery is also referred to as a “digital coherent receiver.” The digital coherent receiver not only compensates for all deterministic linear channel impairments, namely CD and all-order PMD, but also enables a systematic parameter estimation and a comprehensive OPM,39,40 which allows signal quality measurement, fault management extension, and judging the QoS.8,10 Given optimum equalizer filter acquisition by continuous adaptation algorithms,41 the filter impulse response relates to the inverse channel impulse response. Due to the properties of the optical channel operated at moderate launch powers in the weakly nonlinear regime, the contributions of CD and polarization effects such as PMD and PDL to the total channel impulse response can be separated. In principle, all linear channel distortions that can be compensated for by the filter can be estimated nearly instantaneously and directly from the optical channel. No BIP bytes from higher layers have to be analyzed,8 and no pilot tones and no training sequence are required.14

10.3 OPM in digital coherent receivers

275

Unlike recent approaches that typically tap the optical signal and evaluate optical signal properties by external devices,12,13 and unlike classical methods mainly based on OSA, OPM based on the analysis of the filter impulse response does not decrease signal power and does not require devices other than the inevitable equalizer. This leads to cost-efficient, robust, and reliable in-service estimation of residual CD, the mean DGD, and evaluation of the instantaneous PMD spectrum including the polarization rotation, which has been demonstrated on data obtained from transmission experiments with offline processing42,43 and simulations.39,43 In Reference 40, the filter analysis of a commercially available equalizer showed excellent long-term estimation results of PMD and CD. Recently, we extended previous results by PDL estimation, and proved the practicality of our OPM algorithms by joint estimation of all linear channel impairments.44 In the following, we present a comprehensive overview of reliable and robust methods for OPM based on FIR filter analysis. We focus on practical solutions, including a discussion of the influence of various filter adaptation principles. Parts of this work have been presented in Reference 44 with copyright © 2009 IEEE.

10.3.1 Theory Cost-efficient systems with digital coherent receivers are typically operated in the linear or weakly nonlinear regime. A higher power increases the OSNR but also increases the influence of fiber nonlinearities,45 which are interpreted as noise by the linear equalizer.47 The weakly nonlinear regime defines a trade-off between the requirements for a high OSNR and a relatively low impact of nonlinearities leading to a BER around 103. Under this assumption, the channel transfer function can be described by a linear concatenation of all deterministic linear channel impairments. Using polarizationdiverse coherent demodulation and ADC, all properties of the optical field are transferred from the optical domain into the electrical domain. With digital processing, the data can be recovered by mitigating noise and compensating for channel impairments. The core element of this data recovery consists of the equalizer with several blocks of FIR filters. In the following, we provide the theory for the linear optical channel model, the impulse response of the digital FIR filter, and the filter analysis for parameter extraction.

10.3.1.1 The optical channel

In general, the transmitter sends complex valued signals s( f ) ¼ [sX( f ), sY( f )]T on both X and Y polarizations. The optical fiber channel cross-couples information between both polarizations, which can be described by the following matrix vector multiplication:      sX ð f Þ nX ð f Þ HXX ð f Þ HYX ð f Þ þ : (10.18) rð f Þ ¼ Hð f Þsð f Þ þ nð f Þ ¼ HXY ð f Þ HXX ð f Þ sY ð f Þ nY ð f Þ The linear channel transfer matrix is composed of the transfer function D( f ) and concatenated elements Ei and Ui( f ), accounting for PDL and higher-order PMD, as in Hð f Þ ¼ Dð f Þ

N Y i¼1

Ei Ui ð f Þ:

(10.19)

276

CHAPTER 10 OPM based on electronic digital signal processing

The PMD matrices  ui ð f Þ Ui ð f Þ ¼ vi ð f Þ

vi ð f Þ ui ð f Þ



 ¼

ejðfi þ2pf ti Þ=2 0



0

ejðfi þ2pf ti Þ=2

cosai sinai

sinai cosai

 (10.20)

induce a phase shift fi and a DGD ti between the two orthogonal fast and slow axes of the birefringent element. The angle between the slow axis and the polarization state of sX is ai. The instantaneous DGD ht( f)i defines the average over the DGD spectrum p t(ffiffiffiffiffiffiffiffiffiffiffiffiffiffi f ). Individual pffiffiffiffi realizations t( f, t) follow a Maxwellian distribution with mean value htð f ; tÞi ¼ 8=ð3pÞ D p N , given a PMD parameter pffiffiffiffiffiffiffi Dp in ½ps= km. The Hermitian PDL elements   1 0 (10.21) Ei ¼ 0 ki attenuate one polarization relative to the other with the attenuation factor 0 < ki  1 and PDLi ½dB ¼ 20logðki Þ. Analogous to PMD, the instantaneous PDL thatp depends ffiffiffiffiffiffiffiffiffiffiffiffiffiffi on rotation pffiffiffiffi angles N for PDL around a mean value of hPDLðtÞ½dBi ¼ 8=ð3pÞ D ai is Maxwellian distributed PDL pffiffiffiffiffiffiffi parameter DPDL ½dB= km.48 The linear transfer function Dð f Þ ¼ HCD ð f ÞHAF ð f Þ

(10.22)

HCD ð f Þ ¼ expðjf 2 cÞ

(10.23)

contains the all-pass transfer function

with c ¼ CDl p=c responsible for residual CD and an amplitude filter (AF) function HAF( f ) with a linear phase accounting for any bandpass filtering during transmission or LP filtering after demodulation in the receiver. 2

10.3.1.2 The equalizer

Typically, coherent intradyne demodulation with a LO and two optical 90
hybrids, one for each polarization, is applied in the receiver. After the ADC stage, a complex-valued, FIR filter bank arranged in a butterfly structure is located between the timing recovery and the carrier phase recovery (Figure 10.5).6,40

LO

ADC

ADC

Carrier recovery

Optical 90⬚ hybrid

ADC

FIR butterfly equalizer

X pol.

Timing recovery

PBS

ADC Dispersion compensation

Y pol.

FIGURE 10.5 Digital coherent receiver with one optical 90
hybrid for each polarization, ADC, and subsequent digital postprocessing by timing recovery, FIR filter bank, and carrier phase recovery.

10.3 OPM in digital coherent receivers

Meeting channel requirements, the filtering is described by    rX ð f Þ WXX ð f Þ WYX ð f Þ zð f Þ ¼ Wð f Þrð f Þ ¼ WXY ð f Þ WYY ð f Þ rY ð f Þ

277

(10.24)

with the filtering function W( f ) and the filtered signal zð f Þ ¼ ½zX ð f Þ; zY ð f ÞT . In the most general case, we receive the estimate ^sð f Þ of the transmitted signal s( f ) after the carrier recovery stage. FIR filter coefficients can be adapted by various means. Commonly used in OFDM systems, known training symbols can be periodically transmitted that allow computing the inverse channel transfer function directly.14 Provided that the channel memory does not exceed the equalizer memory length, the filter adapts to the zero-forcing (ZF) solution Wð f Þ ¼ H1 ð f Þ;

(10.25)

which is an optimum solution regarding parameter extraction for OPM. However, inverting the AF 1 ð f Þ strongly enhances noise in the regions component that typically follows a LP characteristic, HAF of the stop band of HAF( f ). The minimum mean-square error (MMSE) equalizer solution jointly minimizes the influence of noise and deterministic channel distortions. In contrast to the ZF solution, it reduces noise enhancement with a trade-off between ideally inverting HAF( f ) and imposing a LP filter characteristic. This noisepower-dependent filter component overlaps with the ideal inverse channel transfer function H1( f ) and affects the high-frequency components toward the band edges. Thus, for the components around f ¼ 0 we can assume W( f ) H1( f ), which restricts our parameter estimation to the center taps of W( f ) only. In the following, we assume an ideal ZF filter for the derivation of channel parameters. For blind convergence and continuous tracking of time-varying effects like polarization rotations, a feedback error criterion based on z( f ) or on the decision of ^sð f Þ can be employed. In the first case, we speak of the constant modulus algorithm (CMA); the second case refers to the decision-directed, least mean square (DD-LMS) algorithm.41,45 While the LMS algorithm is derived from the MMSE equalization criterion, the CMA was also shown to approximately converge to the MMSE solution.46 The FIR filter bank comprises an outer butterfly structure to equalize all polarization effects and an inner butterfly structure to realize complex multiplications of a real valued implementation. In the most general case, imperfections at the transmitter or receiver, such as I/Q imbalance, can be mitigated by a filter bank consisting of 16 real-valued FIR filters (Figure 10.6). In this case, the components wXX,I1(t) and wXX,I2(t) adapt to slightly different FIR coefficients. Thus, we obtain rx(t )

w⬘x

wxx

zx(t ) Re{.}

wxy

Q1 –Q2

wyx ry (t )

w⬘y

wyy

I1

zy (t )

Im{.}

I2

FIGURE 10.6 Implementation of FIR filter structure with outer and inner butterfly arrangement.

278

CHAPTER 10 OPM based on electronic digital signal processing

YX

XX

0.4 0.2

0.2

0

0

–0.2

–0.2

–0.4 –10

–5

0 XY

5

10

–0.4 –10

0.4

0.4

0.2

0.2

0

0

–0.2

–0.2

–0.4 –10

–5

0

I Q

0.4

5

10

–0.4 –10

–5

0 YY

5

10

–5

0

5

10

FIGURE 10.7 Exemplary coefficients of 11 tap FIR filters for CD ¼ 1000 ps/nm and mean DGD ¼ 30 ps.

 wXX ðtÞ ¼ 0:5 ðwXX;I1 ðtÞ þ wXX;I2 ðtÞÞ þ jðwXX;Q1 ðtÞ þ wXX;Q2 ðtÞÞ

(10.26)

by averaging over the components. The same applies to the other 12 components to yield wXY(t), wYX(t), and wYY(t), respectively, which reduces the filter bank to the outer butterfly structure. An implementation with only eight real-valued filters and a separate I/Q-imbalance compensation directly leads to the four complex-valued components wXX(t), wXY(t), wYX(t), and wYY(t) without averaging. An example of equalizer coefficients is shown in Figure 10.7. Actually, the filter impulse response is given in a discrete-time representation with w(n) sampled at time instants nTs (Ts is sampling period). For purposes of simplicity, we carry on with a continuous time notation. We obtain the filter transfer matrix by the discrete Fourier transform of the impulse response  

wXX ðtÞ wYX ðtÞ : (10.27) Wð f Þ ¼ F wXY ðtÞ wYY ðtÞ In the most general case, an over-sampling factor of S samples per symbol (Ts ¼ T/S-spaced samples, where T is symbol duration) leads to a bandwidth of S/T in the digital domain. The number of taps defines the resolution within the bandwidth. It is clear that a higher sampling rate and a larger number of taps are beneficial to represent the channel transfer function. With the ZF assumption, Equation (10.25) evolves into Wð f Þ ¼ H1 ð f Þ ¼ D1 ð f Þ

1 Y i¼N;1

1 U1 i ð f ÞEi

(10.28)

10.3 OPM in digital coherent receivers

279

by inverting Equation (10.19). With the aid of individual properties described in Equations (10.20)– (10.22), we can now separate the components for extracting the desired parameters. Although the inverse of each element D( f ), Ei( f ), and Ui( f ) can be calculated easily, a general description with a large number of PDL and PMD elements cannot be obtained. However, due to the unitary nature of the PMD matrices Ui( f ), the square root of the determinant of W( f ) becomes !1=2 1 Y pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 1 detðU1 detðWð f ÞÞ ¼ D2 ð f Þ i ð f ÞÞdetðEi Þ i¼N;1 (10.29) N Y 1=2; 1 ¼ D ð f Þ ðki Þ i¼1 1 ð f Þ has a linear phase in which eliminates the PMD contributions. Given that the amplitude filter HAF the transmission band, the phase  1  pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (10.30) detðWð f ÞÞ ¼ f 2 c arg H^CD ¼ arg

only depends on the quadratic phase transfer function imposed by CD. In contrast, the amplitude transfer function N pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Y   H^PDL ð f Þ ¼  detðWð f ÞÞ ¼ jHAF ð f Þj1 ðki Þ1=2

(10.31)

i¼1

clearly becomes a function of the PDL attenuation. In particular, this is valid for the center taps 1 ð f Þ has unity gain. It should be noted that for multiple concatenated PDL elements as where HAF described in Equation (10.19), the resulting PDL is not simply given by the product Pki, as the rotations ai can vary the impact of each PDL element. For example, one PDL element could impose an attenuation on the X polarization only, and the next element to the Y polarization only. The effect of both elements together would result in pure fiber attenuation identical to each polarization. Still, the estimation proves to be accurate if we extract the PDL attenuation from Equation (10.31). In the following, we refer to H^PDL ð f Þ as the equivalent PDL spectrum. Normalizing W( f ) by the square root of its determinant, we obtain   1   Y ui vi ki 0 1 k vi ui 0 1 i Wð f Þ i¼N;1 WUE ð f Þ ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ¼ 1 Y detðWð f ÞÞ 1=2 (10.32) ki i¼N;1 !    1=2 1   Y ui vi u I ð f Þ vI ð f Þ ki 0 : ¼ ¼ 1=2 vi ui vII ð f Þ uII ð f Þ 0 ki i¼N;1 Reference 49 proposes to further separate the hermitian elements from the unitary elements to estimate the resulting PMD matrix and the PDL. In the “All-order PMD” section (p. 288), we will present an averaging method for the DGD estimation that tolerates the remaining PDL contributions. In our evaluations, the estimation of PDL directly obtained from Equation (10.31) also proved to be

280

CHAPTER 10 OPM based on electronic digital signal processing

more accurate than separating the PMD and PDL elements from Equation (10.32). This is an outcome of the fact that concatenations of estimations accumulate large errors. With Equations (10.30), (10.31), and (10.32), we can now access the desired parameters for CD, PMD, and PDL as demonstrated in the following section.

10.3.2 Joint estimation of linear channel parameters The parameter estimation is demonstrated on a 111-Gb/s (27.75-GBaud) polarization-divisionmultiplexed (PDM) QPSK transmission at an OSNR of 22 dB within a 0.1-nm bandwidth if not specified otherwise. In Reference 42, the validity of a joint estimation of CD and DGD has been demonstrated on measured data with offline processing on a PC. Because measured data with all combined effects including higher-order PMD and PDL were not available, we are restricted to simulations only. Simulations of the linear optical channel with concatenated elements for PMD, CD, and PDL provide the data of the linear optical channel data. If not stated otherwise, only a single element Ei of lumped worst-case PDL is applied to keep the parameter space manageable. Worstcase PDL applies if the axes of the PDL element are aligned with the polarizations of the signal such that only one polarization isQ affected by PDL,50 as shown in Figure 10.8. N¼20 Ui ð f Þ is generated by 20 DGD segments with randomly rotating All-order PMD Uð f Þ ¼ i¼1 39 Jones matrices. The instantaneous mean DGD ht( f )i is obtained from the resulting DGD spectrum t( f ) of U( f ). At the receiver, AWGN is loaded onto the signal. An optical Gaussian bandpass filter (second order, 35 GHz) and an electrical Bessel LP filter (fifth order, 19.425 GHz) provide the AF. In between the AF filters, the optical front end provides coherent demodulation. The ADC stage digitizes the received signal with over-sampling at two samples per symbol. The same blind, digital data-recovery algorithms, as applied in the offline experiments with measured data,6,42 recover the clock phase, adapt the filter coefficients with an impulse response of 21 taps, and recover the carrier phase. The chosen data sets lead to a worst-case BER of 0.026 for a worst-case PDL of 12 dB and 1500-ps/nm residual CD. Once the filter is converged, a fixed routine jointly estimates all the desired parameters from the coefficients of the 16 individual filters in the complex FIR butterfly structure.

Before PDL

q = 0°

q = 45°

After PDL

90°

90° q

FIGURE 10.8 Effect of PDL depending on input polarization.50

Copyright © 2009 IEEE.

90°

>90°

10.3 OPM in digital coherent receivers

10.3.2.1 Estimation of CD Residual CD

281

1

Once we have obtained the phase transfer function of H^CD from Equation (10.30), which is represented by a complex valued vector, we can extract the phase ’( f ) to estimate c and the residual CD by 1

’ð f Þ ¼ cf 2 ¼ arctan

ImðH^CD ð f ÞÞ : 1 ReðH^ ð f ÞÞ

(10.33)

CD

After unwrapping the p-ambiguity of the phase, the weighting factor c of the parabolic phase can be estimated with a quadratic fit (Figure 10.9). 1 As we do not fully isolate H^CD from all the other components in W( f ), the CD estimation can only rely on the center taps around the DC component, where the AF transfer functions of the optical and the electrical filters typically have a linear-phase response. At the same time, we avoid the highfrequency components degraded by the MMSE solution. Thus, we restrict the quadratic fit to about three-quarters of the inner samples around the minimum of ’( f ). The residual CD was estimated for the combination of a given CD ranging from 0 to 1500 ps/nm in steps of 250 ps/nm and given PDL ranging from 0 to 12 dB in steps of 1 dB. For a mean PMD of 10 ps and each combination of CD and PDL, we generated N ¼ 1000 realizations of individual DGD t( f) with a mean DGD ht( f )i ranging from 2 to 35 ps. The statistics of the CD estimation are evaluated with respect to the deviation of the mean value ! N 1X CDestim;i ; (10.34) mCD ¼ CDgiven  N i¼1 which indicates an offset of the estimation. Furthermore, we define the deviation sCD, which is given by vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u N u1 X sCD ¼ t (10.35) ðCDgiven  CDestim;i Þ2 ; N i¼1

arg(HCD(f )) (rad)

0

–5 Phase ϕ Unwrapped phase Quadratic fit

–10

–15 –10

–8

–6

2 4 –4 –2 0 Normalized frequency (taps)

6

8

10

FIGURE 10.9

Recovery of the quadratic phase of H^CD(f ), for example, of CD ¼ 1000 ps/nm.42

Copyright © 2009 IEEE.

282

CHAPTER 10 OPM based on electronic digital signal processing

as the square root of the average error power. In contrast to the common definition of the standard deviation, the biased error sCD is not centralized around the mean value, and indicates the quality of the estimation including the offset of the mean value. In Figure 10.4 we can see that the deviation of the mean value mCD does not significantly degrade for increased PDL. As long as the channel memory length is well below the filter memory length, there is a slight tendency toward an overestimation that systematically increases for larger values of a given CD. As soon as the channel memory starts to exceed the filter memory, which is the case for 21 T/2-spaced taps at 1500 ps/nm residual CD,6 an underestimation is observed. This is in excellent accordance with results obtained from measured data42 (Figure 10.10). Phase unwrapping starts to become critical in this region and ambiguities of multiple p phase offsets between adjacent taps can no longer be resolved. Values above 1500 ps/nm deliver estimation offsets larger than 100 ps/nm. In contrast, the linear-phase transfer function at zero CD is distinguished from the quadratic phase transfer function at nonzero CD values with high accuracy. The estimation error represented by the noncentralized deviation sCD in Figure 10.11 is similar to the deviation of the mean value. It does not exceed 6% for any combination of CD and PDL, which delivers high precision. The precision could be further increased by compensating for the systematic deviation of the mean value, which has not yet been applied to the data set. In addition, a long-term average could reduce the deviation sCD as the value of residual CD is typically constant over time and noise-induced variations of the filter taps could be mitigated.40 The same investigation for a mean PMD of 20 ps with the mean DGD ht( f )i ranging from 2 to 55 ps did not present significant differences. Furthermore, no discrepancy of the estimation for bestcase and worst-case PDL has been observed. Evaluations with several distributed PDL elements along the link led to similar results. Negative values of CD can be estimated analogously as shown in Reference 39. Recalling results from measured data with variations of the OSNR from Reference 39, we can conclude that the estimation of the residual CD is independent from PDL, PMD, and OSNR.

Dispersion estimation in uncompensated links In most installed links, the effect of CD is reversed by arranging DCFs according to the dispersion map. The limited residual dispersion can be compensated by the FIR butterfly filter. In uncompensated links, the DCFs are omitted and the bulk of the dispersion has to be compensated in the receiver using signal processing.51,52 Here, the equalization of CD, PMD, and PDL can be separated into two subcomponents. Whereas PMD and PDL can change in microseconds,53 CD only changes slowly over time as long as the optical link is not altered by path switching. The CD compensation 0 1 ð f Þ ¼ HCD ð f Þ can therefore be separated from polarization-dependent effects as shown in filter WX;Y Figure 10.6. CD compensation can be performed either as time-domain equalization (TDE) or frequencydomain equalization (FDE). Typically, FDE reaches the breakeven point over TDE in terms of complexity at relatively low values of CD.54 With higher channel distortion, the adaptation using common blind algorithms like CMA and LMS increases disproportionately and convergence cannot be guaranteed. In the following, a different approach for estimating CD will be presented. Once the 0 can be derived from the analytiapproximate CD value is known, the equalizer coefficients for WX;Y cal function in Equation (10.6).

10.3 OPM in digital coherent receivers

283

1050 1030

90

990 970

60

950 930 910

30 Mean of estimated DGD Mean of estimated CD

30

890 870 850 90

0 0

Estimated CD (ps/nm)

Estimated DGD (ps)

1010

60 DGD (ps) (a)

2000

Estimated CD (ps/nm)

1500 1000 500 0 –2000

–1000

0

1000

2000

–500 –1000 –1500 –2000 CD (ps/nm) (b)

FIGURE 10.10 Joint estimation for given CD ¼ 1000 ps/nm and DGD varying from 0, 30 and 60–90 ps over a range of OSNR values between 12 and 24 dB (a) and estimation of pure CD over a wide range (b). The bars indicate the standard deviation of the estimation.42 Copyright © 2009 IEEE.

284

CHAPTER 10 OPM based on electronic digital signal processing

80 60

mCD (ps/nm)

40

CD = 0 ps/nm CD = 500 ps/nm CD = 1000 ps/nm CD = 1250 ps/nm CD = 1500 ps/nm

20

0

–20 –40

0

2

4

6 PDL (dB)

8

10

12

90 80 70 CD = 0 ps/nm CD = 500 ps/nm

sCD (ps/nm)

60

CD = 1000 ps/nm CD = 1250 ps/nm CD = 1500 ps/nm

50 40 30 20 10 0 0

2

4

6 PDL (dB)

8

10

12

FIGURE 10.11 Deviation mCD and biased error sCD in ps/nm of the mean value of estimated CD compared to given CD for combinations of CD and PDL.44 Copyright © 2009 IEEE.

The search algorithm presented in Reference 55 for an FDE example evaluates a cost function derived from the CMA based on preloaded equalizer coefficients. It is given by 2 X 2 X J½k ¼ qi;m ½kj2  Ri;m ½bj; (10.36) i¼1 m¼1

where qi,m[k] is the CD compensator output signal for symbol k, i ¼ x,y for the two polarizations, and m ¼ 1,2 for the two samples of each symbol. Ri,m[b] is the normalization constant of the odd and

10.3 OPM in digital coherent receivers

285

even samples for fast Fourier transform (FFT) block b. In order to account for the changing timing phase due to the timing frequency offset between the transmitter and receiver oscillators, the normalization constant Ri,m[b] is determined separately for every FFT block. It is computed from the mean power for the odd and even samples after equalization given by qi;m ½b ¼

1

NX FFT 1

NFFT

d¼0

jqi;m ½k þ dj2 ;

(10.37)

where NFFT is the number of symbols processed per FFT. Here, a simple case differentiation of q i;1 ½b is sufficient to give optimum results regardless of RZ or NRZ pulse the power ratio qi;2 ½b= shaping (see Table 10.2). In the following, the parameters have been optimized using RZ pulse shaping with the constraint of a minimum estimation error. They are set to  ¼ 1.25, Ra ¼ 0.6, Rb ¼ 1.5, and Rc ¼ 2, with the signal power in each polarization normalized to half the total signal power. For every value of CD that is loaded in the filter, the total cost function is given by 2 NX max 1 1 X  Ji ½k þ d; 8W½CD 2 ½CDmax : dCD : CDmin : J½CD ¼ Nmax i¼1 d¼0

(10.38)

0  The minimum J½CD indicates the correct value of CD, giving W½CD ¼ WX;Y ð f Þ. After a rough scan with a large dCD, the dispersion values around the minimum can be scanned at a higher resolution, improving the estimation precision. An example of estimating the optimum filter value with a rough scan is given in Figure 10.12. The main limiting parameter for the algorithm is PMD, since its effect is not modeled in the transfer function W[CD], resulting in estimate deterioration. Figure 10.13 shows the standard deviation of the estimator versus the mean DGD for 112-Gb/s PDM-QPSK and 16-QAM. Here, a maximum dispersion of 48,000 ps/nm was computed from the 3-dB penalty limit of the required OSNR at BER ¼ 1e3 for an FFT size of 1024 symbols, where half of the samples were used for overlap. In general, the overlap size should be larger than the length of the channel impulse response. An average over eight FFT blocks was used for the cost function computation at each CD value. Although the computation of Ri,m[b] was optimized for RZ-QPSK, NRZ-QPSK as well as 16-QAM also perform similarly. With varying FFT sizes, estimation performance remains virtually  and Noverlap the same, if the identical number of symbols Nmax is used for the computation of J½CD, and CDmax are appropriately scaled with the FFT size. A joint CD estimation in combination with the FIR butterfly filter allows further refinement of the compensator CD value.

Table 10.2 Normalization Constants Power Ratio q i;1 ½b >  qi;2 ½b= 1= < qi;2 ½b= q i;1 ½b   q i;1 ½b < 1= qi;2 ½b=

Ri,1[b]

Ri,2[b]

Ra Rb Rc

Rc Rb Ra

286

CHAPTER 10 OPM based on electronic digital signal processing

0.50 0.48

Cost function

0.46 0.44 0.42 0.40 0.38 0.36

0

2.5

5

7.5

10

12.5

15

Chromatic dispersion / (ps/nm) ´ 104

FIGURE 10.12 Estimation example for a CD value of 120,000 ps/nm in a 112-Gb/s PDM-QPSK uncompensated transmission link. The initial scan is performed in steps of dCD ¼ 200 ps/nm.55 Copyright © 2009 IEEE.

Standard deviation of estimation (ps/nm)

150 NRZ-QPSK RZ-QPSK

125

NRZ-16QAM RZ-16QAM

100 75 50 25 0

0

5

10

15

20 25 30 35 Mean DGD (ps)

40

45

50

FIGURE 10.13 Estimation performance versus mean DGD (full-order PMD) for RZ and NRZ pulse shaping of 112-Gb/s PDMQPSK (OSNR ¼ 14 dB) and PDM-16-QAM (OSNR ¼ 17 dB).56 Copyright © 2009 IEEE.

10.3 OPM in digital coherent receivers

287

10.3.2.2 Estimation of PMD Within the polarization effects, we are interested in the value of DGD and the polarization rotation. Both effects vary over time; thus, we are interested in the instantaneous values to monitor the transients. Depending on the signal bandwidth and the mean DGD in the link, several orders of PMD can be present. However, only two parameters are of crucial interest. Rotation of the absolute signal polarization can be viewed as zero-order PMD and determines the fastest gradients that have to be tracked by the equalizer.53 The DGD is the magnitude of the first-order derivative of the PMD vector and fully describes all statistical properties of the all-order PMD.

Polarization rotation Polarization is usually visualized in the three-dimensional Stokes space, with the vector S ¼ ½S1 ; S2 ; S3  given by S1 ¼ cosð2’Þcosð2cÞ S2 ¼ sinð2’Þcosð2cÞ S3 ¼ sinð2cÞ with

(10.39)

0 n h ð f Þ o1 XY  Im hXX ð fÞ jhXY ð f Þj @ n o A: ; cX ¼ arctan ’X ¼ arctan hXY ð f Þ jhXX ð f Þj Re 

(10.40)

hXX ð f Þ

The same applies for ’Y and cY . Here, the central tap of the equalizer coefficients at f ¼ 0 in the frequency domain is used for computation. An example state of polarization evolution is given in Figure 10.14.

FIGURE 10.14 Exemplary state of polarization evolution in Stokes space.

288

CHAPTER 10 OPM based on electronic digital signal processing

TX

82 kHz

PDL 6 dB

41 kHz

DGD 60 ps

CD

RX

1000 ps/nm

Rate (rotations/ms)

200 180

X tributary Y tributary Average

160 140 120 100 80 0

5

10

15

20

25

Time (μs)

FIGURE 10.15 SOP evolution in presence of PDL for 43-Gb/s PDM-QPSK.49

Copyright © 2009 IEEE.

Given the Stokes vector, the SOP rotation frequency is given by49   1 1 arcsin jSðt2 Þ  Sðt1 Þj : fpo1 ¼ ðt2  t1 Þp 2

(10.41)

In Reference 49, the authors demonstrated the evolution of SOP in the presence of PDL. Given the depolarization induced by PDL (Figure 10.8), periodic SOP rotation leads to a different SOP rotation frequency in the two polarizations, with the mean SOP corresponding to the input SOP rotation in the polarization scrambler (Figure 10.15).

All-order PMD Without loss of generality, we provide a reliable DGD estimation with the aid of appropriate averaging for only one PMD and one PDL element. In a first step, we derive the matrix elements uI ð f Þ; uII ð f Þ; vI ð f Þ; and vII ð f Þ of WUE( f ) in Equation (10.32) with respect to the angular frequency (indexed by o). Now we simply multiply the appropriate elements to yield pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi testim ð f Þ ¼ 2 uI;o ð f ÞuII;o ð f Þ  vI;o ð f ÞvII;o ð f Þ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ; (10.42) ¼ 2 juo ð f Þj2 þ jvo ð f Þj2 which cancels the PDL contributions. No further matrix separation operations are required. This operation refers to calculating the determinant of the derivative of WUE( f ) with eigenvalues uo ð f Þ and vo ð f Þ.57 The same applies for channels with multiple concatenated elements of PMD or PDL, or both. For the same reason as in AF and the MMSE solution (compare CD estimation), we limit the estimation of the mean value over the DGD spectrum htestim( f )i to a centered bandwidth of 3/(2T). One should be aware that theoretically the estimates u^ ¼ ðuI uII Þ1=2 ; v^ ¼ ðvI vII Þ1=2 , and testim( f ) provide further polarization information, including the polarization rotation gradient over time and the evolution of the DGD spectrum over time for a sufficient update rate of the equalizer.

10.3 OPM in digital coherent receivers

289

Estimated DGD 〈τestim(f )〉 (ps)

25

20

15

10

5 Deviation of the mean value mDGD 0

0

5

10 15 Given DGD 〈τ(f )〉 (ps)

20

25

FIGURE 10.16 Estimation example for a PMD of 10 ps of the instantaneous mean DGD ht(f )i for CD ¼ 1000 ps/nm and 9-dB worst-case PDL. Note that the distribution of ht(j )i proves to be nearly Maxwellian.44 Copyright © 2009 IEEE.

An example estimation of the DGD htestim( f )i for given realizations ht( f )i combined with 1000ps/nm CD and 9-dB PDL appears in Figure 10.16. The Maxwellian-distributed instantaneous DGD values range from 2 to 25 ps. On average, the estimation tends toward a slight overestimation, which becomes clear in Equation (10.42). As we add up the absolute values of uo and vo, any estimation error increases the result. For this reason, we expect the estimation of very low DGD values below 5 ps to be offset. Despite the offset induced by PDL (compare Figure 10.16), the result is in agreement with Reference 40, where no PDL was present. Because the given DGD ht( f )i varies for each individual case, we modify the deviation of the mean value from (10.34) to yield mDGD ¼

N 1X DGDgiven;i  DGDestim;i N i¼1

(10.43)

and the biased error sDGD becomes sDGD

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi u N u1 X  2 ¼t DGDgiven  DGDestim;i : N i¼1

(10.44)

Figure 10.17 evaluates the data of the given example with respect to mDGD and sDGD for DGD windows of 5 ps. Increasing the given DGD, the deviation of the mean value slightly decreases and the standard deviation centered around the mean value increases. This indicates that the estimation error sDGD is governed by an offset of the estimation for low DGD values, and adopts

290

CHAPTER 10 OPM based on electronic digital signal processing

4

Statistics of estimation (ps)

3 2 1 0 –1 –2 σDGD

–3

mDGD

–4 –5

Standard deviation

0-5

5-10

10-15 15-20 Given DGD 〈τ(f )〉 (ps)

20-25

0-5

FIGURE 10.17 Statistics of exemplary DGD estimation for CD ¼ 1000 ps/nm and 9-dB worst-case PDL with respect to deviation of the estimation error sDGD, mean value mDGD, and standard deviation within 5-ps DGD windows.44 Copyright © 2009 IEEE.

a larger variance toward high DGD values. This typical scheme has been observed in nearly all data sets. Low DGD values up to 5 ps suffer from the estimation offset induced by the absolute values in Equation (10.42), which becomes negligible as the DGD increases. However, for large DGD values, the standard deviation of the estimation increases, which dominates the estimation error above 20 ps. In conclusion, the estimation regression curve is not purely linear. In particular, the estimates for low DGD values could be improved with an appropriate calibration. It should be noted that the statistics above 15 ps are based on a weak basis of less than 50 incidents per DGD window. In the following, no correction or calibration in the DGD estimation is applied. We now apply the DGD estimation for constant values of CD and PDL over a wide range of given DGD. This reflects the situation of a transmission over installed fiber with time-variant DGD around a known mean PMD (compare Reference 40). Results of mDGD and sDGD for combinations of CD and PDL with a mean PMD of 10 ps are given in Figure 10.18. For a sufficient filter memory length, the deviation of the mean value mDGD and the estimation error sDGD are only slightly affected by residual CD but increase for high values of PDL to about 3 and 3 ps, respectively. We believe that for large PDL, the perfect symmetry required for an ideal cancellation in Equations (10.32) and (10.42) is degraded by the MMSE solution. Similarly, the estimation error sDGD increases for high values of CD where the filter memory is insufficient. In that case, the separation of effects by Equation (10.32) is no longer fully valid. In conclusion, despite the deterministic influence of large PDL and the statistical influence of large CD, the instantaneous mean DGD ht( f )i can be robustly estimated over a wide range. In addition, earlier estimations based on measured data proved to be independent from the OSNR.42 In contrast to the estimation of CD, which is tailored to a parabolic phase transfer function tolerating slight

10.3 OPM in digital coherent receivers

3.5

2 1.5

CD = 0 ps/nm CD = 500 ps/nm CD = 1000 ps/nm CD = 1250 ps/nm CD = 1500 ps/nm

1 0.5 0 –0.5 –1 –1.5

CD = 0 ps/nm CD = 500 ps/nm CD = 1000 ps/nm CD = 1250 ps/nm CD = 1500 ps/nm

3

σDGD (ps)

mDGD (ps)

291

2.5

2

–2 1.5

–2.5 –3 –3.5

0

2

4

8 6 PDL (dB)

10

12

1

0

2

4

6 8 PDL (dB)

10

12

FIGURE 10.18 Deviation mDGD and biased error sDGD in ps of the mean value of estimated DGD compared to given DGD for combinations of CD and PDL.44 Copyright © 2009 IEEE.

deviations, each imperfection in the filter convergence and each error during the separation process degrade the estimation. Still, accuracy of less than 3 ps for reasonable values of CD and PDL is a respectable result.

10.3.2.3 Estimation of PDL The distortion of PDL cannot be compensated, as it attenuates the signal and thus degrades the OSNR. Still, it imposes a very critical condition for blind equalizer convergence, which is likely to converge to only one polarization state and suppress the other polarization state. In the following, we present two methods to estimate the PDL from the FIR coefficients. The first method refers to the ideal theoretic solution; however, the second method proves to be more reliable and robust in practical systems.

Eigenvalue estimation In Reference 49, PDL estimation from the eigenvalues ðl1 ; l2 Þ of the square root of the equalizer autocorrelation matrix Að f Þ was given by qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Wð f Þ : (10.45) Að f Þ ¼ WN ð f ÞWN ð f ÞH ; WN ð f Þ ¼ detðWð f ÞÞ The PDL value can then be computed as PDLð f ÞdB ¼ 20log10

 ! l   1   :  l2 

(10.46)

The authors in Reference 49 presented a continuous estimation example, as shown in Figure 10.19. For a preset PDL of 6 dB, the average observed time showed an underestimation of about 0.5 dB.

292

CHAPTER 10 OPM based on electronic digital signal processing

TX

PDL 6 dB

82 kHz

41 kHz

DGD 60 ps

CD

RX

1000 ps/nm

Estimated PDL (dB)

6 Instantaneous Moving average

5.8 5.6 5.4 5.2 5

5

0

10

15

20

25

Time (μs)

FIGURE 10.19 Simulative PDL estimation for 43-Gb/s PDM-QPSK.49

Copyright © 2009 IEEE.

Determinant estimation Basically, the equivalent PDL spectrum from Equation (10.14) provides a function of the parameters ki necessary for the PDL estimation. Until now, the equalizer model has been treated as an ideal ZF filter. Due to the property of the MMSE equalizer to suppress noise enhancement, it will impose individual solutions for different OSNR values as well as for different PDL values (Figure 10.20). 3

|det(HFIR)|(a.u.)

2.5 2 1.5 1 OSNR 18 dB, PDL 0 dB OSNR 18 dB, PDL 10 dB OSNR 22 dB, PDL 0 dB OSNR 22 dB, PDL 10 dB

0.5 0 –10

–8

–6

–4 –2 0 2 4 6 Normalized frequency (taps)

8

10

FIGURE 10.20 Averaged PDL power spectrum adapted to various levels according to the influence of noise power in the system.44 Copyright © 2009 IEEE.

10.3 OPM in digital coherent receivers

293

Thus, an exact estimation of the PDL necessitates considering the influence of OSNR on the PDL estimation. The equivalent PDL power spectrum H^PDL ð f Þ from Equation (10.31) is depicted for two different OSNR levels and two PDL values in Figure 10.10. The various power levels around the center of the PDL spectrum are clearly governed by the k1/2 factor of the PDL. The ratio of the center taps for 0-dB PDL and 10-dB PDL is about 1.7, which relates to the ratios of k1/2 for each chosen PDL. On top of those power levels, the equalizer imposes an LP characteristic, which leads to a lower maximum power level in the center of the transmission band for a lower OSNR. In return, the power levels toward the edges of the transmission band are increased, where the AF strongly decreases the influence of noise. Thus, the equalizer boosts the deterministic components in this area to obtain the optimum trade-off. In this example, it can be clearly seen that without calibration the PDL estimation would be strongly degraded by the MMSE solution reacting on the OSNR. Furthermore, we noted that after the filter has been converged, the continuous update in the tracking mode of the equalizer slightly varies the filter coefficients, which leads to significant variations in the PDL power spectrum, similar to Figure 10.19. Thus, we average on several consecutive filter constellations of the same channel constellation with updates in between. Only the noise, update parameters, and filter coefficients vary during this period. To avoid a complex scaling and correction term in Equation (10.31), we simply compare the five centered taps to references of the equivalent PDL spectrum that have been obtained from extensive simulations for OSNR variations with a single lumped PDL element. The best matching reference leads us to the estimated PDL. The same references are used for evaluations with a single PDL element or with multiple concatenated PDL elements. To our knowledge, there is no solution for reliably estimating the value of OSNR from any point in the DSP circuitry. Any in-band SNR will be affected by the PDL itself. Furthermore, the signal variance prior to equalization is governed by intersymbol interference, and even after equalization it will contain noise from nonlinearities.47 Thus, the value of the OSNR needs to be obtained from an external OSNR monitor. Similar to the evaluations of CD, we simulated combinations of CD and PDL varying from 0, 500, 1000, 1250, and 1500 ps/nm and from 0 to 10 dB, respectively, each with 100 individual realizations of DGD for a mean PMD of 10 ps. In a first simulation setup with only one lumped PDL element, the PDL varies randomly between worst-case and best-case PDL by varying the angle of the signal polarization prior to the PDL element. The data are evaluated for an OSNR of 18 and 22 dB. Analogous to the unbiased estimation error defined in Equation (10.35), we define sPDL for the estimation of the PDL, which is the biased deviation of the estimated PDL given in dB units. The estimation error sPDL for an OSNR of 18 and 22 dB is given in Figure 10.21. The estimation indicates a large deviation for low PDL values close to 0 dB. In this region, the impact of PDL can hardly be detected by the equalizer. This effect closely relates to the negligible OSNR penalty induced by low PDL.50 The optimum estimation is achieved for a PDL above 2 dB. As system performance strongly degrades for an OSNR of 18 dB combined with a PDL of 10 dB, the estimation also degrades. For a higher OSNR of 22 dB, this point is not reached until a 10-dB PDL. Finally, clear influence of residual dispersion on the estimation can be seen increasing the estimation error for higher values of CD. The inverse of HCD ð f Þ is not modeled by the equalizer as an ideal all-pass function in this case. In a second simulation setup, the PDL has been accumulated to yield a mean of hPDLðtÞ½dBi ¼ 5 dB by distributed PDL elements according to Equation (10.2), where each element imposed the same attenuation ki. As mentioned, the instantaneous PDL depends on polarization states

294

CHAPTER 10 OPM based on electronic digital signal processing

1 CD 0 ps/nm CD 500 ps/nm CD 1000 ps/nm CD 1250 ps/nm CD 1500 ps/nm

Estimation error σPDL (dB)

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0

2

4 6 Given PDL (dB) (a)

8

10

1 CD 0 ps/nm CD 500 ps/nm CD 1000 ps/nm CD 1250 ps/nm CD 1500 ps/nm

Estimation error σPDL (dB)

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0

0

2

4 6 Given PDL (dB) (b)

8

10

FIGURE 10.21 Unbiased estimation error sPDL over individual channel realizations with individual DGD for combinations of CD and PDL at an OSNR of (a) 18 dB and (b) 22 dB.44 Copyright © 2009 IEEE.

and varies within a Maxwellian distribution.48 The given value of the instantaneous PDL was obtained using the Mueller method described in Reference 58. Identical references were used as in the case of single-element PDL. In Figure 10.22, it can be seen that estimation performance does not greatly differ between a single PDL element and multiply concatenated elements, which verifies previous results for a single lumped PDL element and verifies the use of Equation (10.31) for PDL estimation.

10.3 OPM in digital coherent receivers

295

0.4 0 ps/nm 1000 ps/nm 1250 ps/nm 1500 ps/nm

0.35

σPDL (dB)

0.3 0.25 0.2 0.15 0.1 0.05

1

2

4

8 16 32 64 128 256 512 Number of PDL elements

FIGURE 10.22 PDL estimation performance at OSNR ¼ 22 dB versus the number of concatenated PDL elements with a mean PDL of 5 dB.44 Copyright © 2009 IEEE.

The largest deviation of the estimation at 0-dB PDL combined with a CD of 1500 ps/nm shows precision of 1 dB with a typical deviation around 0.2 dB for given PDL values larger than 2 dB. PMD showed no significant impact on the PDL estimation. In addition, there was no difference between worst-case PDL and best-case PDL using one lumped element, and multiple concatenated PDL elements nearly led to the same results. However, to achieve this robust and precise PDL estimation, a priori knowledge of the OSNR is required, which has to be obtained from an external source. To the best of our knowledge, reliable and precise OSNR estimation from the equalizer for combined linear channel distortions has not been presented so far.

10.3.3 Conclusion The concept of evaluating the filter impulse response of the digital equalizer to estimate all deterministic linear channel distortions such as residual CD, PMD, and PDL has been proved in theory and by extensive simulation. The practicality of robust and precise estimation has been demonstrated over a wide range of combined distortions. Preceding investigations based on measured and simulated data are in agreement.40,43 Although the theory of the linear optical channel model provides comprehensive access to all desired parameters, many details of the practical implementation of estimation algorithms must be considered to obtain a precise and robust estimation. Estimates of statistical noise distortions have not been demonstrated. On the one hand, evaluating the mean value and variance of the signal zðtÞ to obtain an estimation of the SNR is simple42; this is a valuable parameter for monitoring signal quality prior to the decision. However, as the MMSE solution jointly mitigates noise enhancement and intersymbol interference, the variance of zðtÞ is also

296

CHAPTER 10 OPM based on electronic digital signal processing

influenced by “unequalized” deterministic signal components. In particular, this is valid for deterministic nonlinear effects such as SPM that are interpreted as noise by the linear equalizer. In contrast to the ZF solution applicable in OFDM systems with training, where identification of OSNR and SPM- and XPM-induced nonlinearities is possible,47 blind estimation based on MMSE does not allow for a reliable identification of noise sources. It has also been demonstrated in Section 10.3.2.3 that the OSNR affects the estimation of PDL and that PDL degrades the OSNR. To our knowledge, no reliable estimation of the in-band OSNR from DSP exists. By nature, the FIR filter is a linear equalizer compensating for linear channel distortions only. The parameter estimation presented in Section 10.3.2 assumes a linear or weakly nonlinear channel with negligible influence of SPM and XPM on the filter convergence. In presence of significant nonlinearities the filter might converge to suboptimum solutions or might not converge at all, although the BER in this case would be well above the FEC limit. Some DSP techniques partly increase the nonlinear tolerance of the receiver,59,60 but until now no effective compensation of nonlinearities has been demonstrated. Monitoring nonlinearities from DSP is also not possible.

10.4 SUMMARY In this chapter we have presented methods to monitor properties of the optical channel via analysis of the digital equalizer at the receiver. We assume the equalizer to blindly adapt to the channel transfer function without a training sequence or a training signal. Prerequisite to the estimation are a sufficient equalizer memory length and a reliable update algorithm to converge the equalizer, to match the channel conditions and track time-varying changes. In direct detection systems where equalization is based on the MLSE principle, a systematic separation of deterministic signal components and noise has been demonstrated, which allows for an estimation of the noise power, and OSNR in particular. Furthermore, with a reference-based method, single deterministic distortions could be identified. Compared to the extensive capabilities in coherent detection systems, OPM in direct detection systems is elaborate and limited. The main reason is the nonlinearity of the photodiode that irreversibly superimposes deterministic channel impairments and transforms the optical Gaussian noise into a complicated higher-order, chi-square-distributed process. Given the negligible influence of SPM and XPM, the optical channel, the optical front end, and the receiver impose a linear channel transfer function that can be fully compensated by linear FIR filters. Due to the individual characteristics of linear optical distortions like CD, PMD, or PDL, their contributions can be systematically identified. This results in precise estimation of all linear parameters over a wide range. The distinction between noisy distortions and deterministic signal components is automatically computed by the update algorithm of the equalizer, which suppresses the influence of the zero-mean Gaussian noise by appropriate averaging. A systematic separation of noise terms is not possible via analyzing equalizer properties, as the MMSE solution of the blind update algorithms jointly minimizes the influence of deterministic channel distortions and noise. Inevitably, the remaining deterministic components in the equalized signal mix with noise terms. Only a ZF solution as applied in OFDM systems is capable of such a noise term identification.47 The great advantage of equalizer-based OPM is its cost-efficient and low-complexity implementation as a by-product of existing DSP circuits. Basically, all distortions that can be compensated can be monitored.

References

297

At the same time, its great disadvantage is the necessity of a digital equalizer including the optical front end and all stages of sophisticated signal processing for equalizer convergence and data recovery. With state-of-the-art technology, this restricts cost-efficient application to transmission link terminals. Future technologies with a high degree of integration might enable portable OPM devices based on the same principle as tapping optical signals at any interface in the network.

REFERENCES 1. Shukla V, Brown D, Hunt CJ, Mueller T, Varma E. Next-generation optical network—enabling dynamic bandwidth services. In: Proc. optical fiber communication conference (OFC), paper NWB3. Anaheim, CA; 2007. 2. Xia TJ, Wellbrock G, Peterson D, Lee W, Pollock M, Basch B, et al. Multi-rate (111-Gb/s, 2  43-Gb/s, and 8  10.7-Gb/s) transmission at 50-GHz channel spacing over 1040-km field-deployed fiber. In: Proc. European conference optical communication (ECOC), paper Th.2.E.2. Brussels; 2008. 3. Pachnicke S, Obholz M, Voges E, Krummrich PM, Gottwald E. Electronic EDFA gain control for the suppression of transient gain dynamics in long-haul transmission systems. In: Proc. optical fiber communication conference (OFC), paper JWA15. Anaheim, CA; 2007. 4. Raybon G, Winzer PJ. 100 Gb/s challenges and solutions. In: Proc. optical fiber communication conference (OFC), paper OTuG1. San Diego, CA; 2008. 5. Lach E, Schuh K, Junginger B, Veith G, Lutz J, Mo¨ller M. Challenges for 100 Gbit/s ETDM transmission and implementation. In: Proc. optical fiber communication conference (OFC), paper OWE1. Anaheim, CA; 2007. 6. Fludger CRS, Duthel T, van den Borne D, Schulien C, Schmidt ED, Wuth T, et al. Coherent equalisation and POLMUX-RZ-DQPSK for robust 100-GE transmission. J Lightwave Technol 2008;26(1):64–72. 7. Winzer PJ, Gnauck AH. 112-Gb/s polarization-multiplexed 16-QAM on a 25-GHz WDMGrid. In: Proc. European conference optical communication (ECOC), post-deadline paper Th.3.E.5. Brussels, September; 2008. 8. Chung YC. Optical performance monitoring techniques: current status and future challenges. In: Proc. European conference optical communication (ECOC), paper OThW2. Brussels; 2008. 9. Woodward SL. Monitors to ensure the performance of photonic networks. In: Proc. optical fiber communication conference (OFC), paper OMM1. Anaheim, CA; 2007. 10. Kilper DC, Bach R, Blumenthal DJ, Einstein D, Landolsi T, Ostar L, et al. Optical performance monitoring. J Lightwave Technol 2004;22(1):294–304. 11. Lu GW, Cheung MH, Chen LK, Chan CK. Simultaneous PMD and OSNR monitoring by enhanced RF spectral dip analysis assisted with a local large DGD element. IEEE Photon Technol Lett 2005;17(12): 2790–2. 12. Dods SD, Anderson TB, Clarke K, Bakaul M, Kowalczyk A. Asynchronous sampling for optical performance monitoring. In: Proc. optical fiber communication. conference (OFC), paper OMM5. Anaheim, CA; 2007. 13. Yoshida N, Nogushi H, Abe J, Uchida H, Ozaki M, Kanemitsu S, et al. High-speed and simple estimation algorithm of waveform distortions using a 43-Gb/s SiGe-based eye-monitoring LSI. In: Proc. European conference optical communication (ECOC), paper We1D3. Brussels; 2008. 14. Yi X, Shieh W, Ma Y, Tang Y, Pendock GJ. Experimental demonstration of optical performance monitoring in coherent optical OFDM systems. In: Proc. optical fiber communication conference (OFC), paper OThW3. San Diego, CA; 2008.

298

CHAPTER 10 OPM based on electronic digital signal processing

15. Fa¨rbert A, Langenbach S, Stojanovic N, Dorschky C, Kupfer T, Shulien C, et al. Performance of a 10.7 Gb/s receiver with digital equalizer using maximum likelihood sequence estimation. In: Proc. European conference optical communication (ECOC), post-deadline paper Th4.1.5. Stockholm; 2004. 16. Griffin RA, Swenson N, Crivelli D, Carrer H, Hueda M, Voois P, et al. Combination of InP MZM transmitter and monolithic CMOS 8-state MLSE receiver for dispersion tolerant 10 Gb/s transmission. In: Proc. optical fiber communication conference (OFC), paper OThO2. San Diego, CA; 2008. 17. Sun H, Wu KT, Roberts K. Real-time measurements of a 40 Gb/s coherent system. Opt Exp 2008;16(2): 873–9. 18. Winzer PJ, Essiambre RJ. Advanced optical modulation formats. Proc. of the IEEE 2006;94(5):952–85. 19. Haunstein HF, Sauer-Greff W, Dittrich A, Sticht K, Urbansky R. Principles for electronic equalization of polarization-mode dispersion. J Lightwave Technol 2004;22(4):1169–82. 20. Yoshida M, Goto H, Omiya T, Kasai K, Nakazawa M. Frequency division multiplexed 1 Gsymbol/s, 64 QAM coherent optical transmission with a spectral efficiency of 8.6 bit/s/Hz. In: Proc. European conference optical communication (ECOC), paper Mo4D5. Brussels; 2008. 21. van den Borne D, Calabro` S, Jansen SL, Gottwald E, Khoe GD, de Waardt H. Differential quadrature phase shift keying with close to homodyne performance based on multi-symbol phase estimation. In: Proc. IEE seminar on electrical signal processing in optical communications, paper 12. London; 2006. 22. Liu X. Receiver sensitivity improvement in optical DQPSK and DQPSK/ASK through data-aided multisymbol phase estimation. In: Proc. European conference optical communication (ECOC), paper We.2.5.6. Cannes; 2006. 23. Kikuchi N, Sekine K, Sasaki S. Modulation/demodulation schemes for optical multi-level transmission. In: Proc. optical fiber communication conference (OFC), paper OThL3. Anaheim, CA; 2006. 24. Kikuchi N, Mandai K, Sasaki S. Experimental demonstration of incoherent optical multilevel staggeredAPSK (amplitude- and phase-shift keying) signaling. In: Proc. optical fiber communication conference (OFC), paper OMI3. San Diego, CA; 2008. 25. Alfiad MS, van den Borne D, Hauske FN, Napoli A, Lankl B, Koonen AMJ, et al. Dispersion tolerant 21.4-Gb/s DQPSK using simplified Gaussian joint-symbol MLSE. In: Proc. optical fiber communication conference (OFC), paper OThO3. San Diego, CA; 2008. 26. Hauske FN, Lankl B, Xie C, Schmidt ED. Iterative electronic equalization utilizing low complexity MLSEs for 40 Gbit/s DQPSK modulation. In: Proc. optical fiber communication conference (OFC), paper OMG2. Anaheim, CA; 2007. 27. Agazzi OE, Hueda MR, Carrer HS, Crivelli DE. Maximum-likelihood sequence estimation in dispersive optical channels. J Lightwave Technol 2005;23(2):749–63. 28. Alic´ N, Papen GC, Saperstein RE, Milstein LB, Fainman Y. Signal statistics and maximum likelihood sequence estimation in intensity modulated fiber optic links containing a single optical preamplifier. Opt Exp 2005;13(12):4568–79. 29. Hauske F, Kuschnerov M, Piyawanno K, Lankl B, Schmidt ED. Optical performance monitoring in amplitude sampling receivers. In: Proc. optoelectronics communication conference (OECC), paper 13B4-3. Yokohama; 2007. 30. Hauske FN, Lankl B, Xie C, Schmidt ED. State-based OSNR estimation in amplitude sampling receivers. In: Proc. European conference optical communication (ECOC), paper Mo4.4.4. Cannes; 2006. 31. Agrawal GP. Nonlinear fiber optics. 3rd ed. San Diego: Academic Press; 2001. 32. Proakis JG. Digital communications. 4th ed. New York: McGraw-Hill; 2001. 33. Foggi T, Forestieri E, Colavolpe G, Prati G. Maximum-likelihood sequence detection with closed-form metrics in OOK optical systems impaired by GVD and PMD. J Lightwave Technol 2006;24(8):3073–87. 34. Proakis JG. Digital communications. 4th ed. New York: McGraw-Hill; 2001. 35. Ho KP. Phase-modulated optical communication systems. Berlin: Springer; 2005.

References

299

36. Fa¨rbert A, Elbers JP, Bock H, Neuhauser R, Glingener C. Practical method to measure optical SNR in multi-terabit systems. In: Annual meeting of the IEEE lasers & electro-optics society (LEOS), paper MD2. San Diego, CA; 2001. 37. Hanik N, Gladisch A, Caspar C, Strebel B. Application of amplitude histograms to monitor performance of optical channels. Electron Lett 1999;35(5):403–4. 38. Kupfer T, Langenbach S, Stojanovic N, Gehrke S, Whiteaway J. Performance of MLSE in optical communication systems. In: Proc. European conference optical communication (ECOC), paper Th9.1.1. Berlin; 2007. 39. Hauske FN, Kuschnerov M, Piyawanno K, Spinnler B, Schmidt ED, Lankl B. DGD estimation from FIR filter taps in presence of higher order PMD. In: Proc. European conference optical communication (ECOC), paper Tu1D4. Brussels; 2008. 40. Woodward SL, Nelson LE, Feuer MD, Zhou X, Magill PD, Foo S, et al. Characterization of real-time PMD and chromatic dispersion monitoring in a high-PMD 46-Gbs transmission system. IEEE Photon Technol Lett 2008;20(24):2048–50. 41. Benvenuto N, Cherubini G. Algorithms for communications systems and their applications. Chichester: John Wiley & Sons; 2005. 42. Hauske FN, Geyer J, Kuschnerov M, Piyawanno K, Lankl B, Duthel T, et al. Optical performance monitoring from FIR filter coefficients in coherent receivers. In: Proc. optical fiber communication conference (OFC), paper OThW2. San Diego, CA; 2008. 43. Geyer JC, Hauske FN, Fludger CRS, Duthel T, Schulien C, Kuschnerov M, et al. Channel parameter estimation for polarization diverse coherent receivers. IEEE Photon Technol Lett 2008;20(10):776–8. 44. Hauske FN, Kuschnerov M, Spinnler B, Lankl B. Optical performance monitoring in digital coherent receivers. J Lightwave Technol 2009;27(16):3623–31. 45. Savory SJ, Gavioli G, Killey RI, Bayvel P. Transmission of 42.8 Gbit/s polarization multiplexed NRZ-QPSK over 6400 km of standard fiber with no optical dispersion compensation. In: Proc. optical fiber communication conference (OFC), paper OTuA1. Anaheim, CA; 2006. 46. Haykin S. “Unsupervised Adaptive Filtering, Volume II: Blind Deconvolution”, 2000, John Wiley & Sons. 47. Mayrock M, Haunstein H. Optical monitoring for non-linearity identification in CO-OFDM transmission systems. In: Proc. optical fiber communication conference (OFC), paper JThA58. San Diego, CA; 2008. 48. Steinkamp A, Vorbeck S, Voges E. Polarization mode dispersion and polarization dependent loss in optical fiber systems. In: Proc. Optics East. Philadelphia, PA; 2004. p. 243–54. 49. Geyer JC, Fludger CRS, Duthel T, Schulien C, Schmauss B. Performance monitoring using coherent receivers. In: Proc. optical fiber communication conference (OFC), paper OThH5. San Diego, CA; 2009. 50. Duthel T, Fludger CRS, Geyer J, Schulien C. Impact of polarisation-dependent loss on coherent POLMUXNRZ-DQPSK. In: Proc. optical fiber communication conference (OFC), paper OThU5. San Diego, CA; 2008. 51. Savory SJ, Gavioli G, Killey RI, Bayvel P. Electronic compensation of chromatic dispersion using a digital coherent receiver. Opt Expr 2007;15(5):2120–6. 52. Curri V, Poggiolini P, Carena A, Forghieri F. Dispersion compensation and mitigation of nonlinear effects in 111-Gb/s WDM coherent PM-QPSK systems. IEEE Photon Technol Lett 2008;20(17):1473–5. 53. Krummrich PM, Kotten K. Extremely fast (microsecond timescale) polarization changes in high-speed long-haul WDM transmission system. In: Proc. optical fiber communication conference (OFC), paper FI3. Los Angeles, CA; 2004. 54. Spinnler B, Hauske FN, Kuschnerov M. Adaptive equalizer complexity in coherent optical receivers. In: Proc. European conference optical communication (ECOC), paper We.2.E.4. Brussels; 2008. 55. Kuschnerov M, Hauske FN, Piyawanno K, Spinnler B, Napoli A, Lankl B. Adaptive chromatic dispersion equalization for non-dispersion managed coherent systems. In: Proc. optical fiber communication conference (OFC), paper OMT1. San Diego, CA; 2009.

300

CHAPTER 10 OPM based on electronic digital signal processing

56. Kuschnerov M, Hauske FN, Piyawanno K, Spinnler B, Alfiad MS, Napoli A, et al. DSP for Single-Carrier Receivers, J Lightwave Technol 2009;27(16):3614–22. 57. Gordon JP, Kogelnik H. PMD fundamentals: polarization mode dispersion in optical fibers. Proc. Nat Acad Sci USA 2005;97(9):4541–50. 58. Mecozzi A, Shtaif M. The statistics of polarization-dependent loss in optical communication systems. IEEE Photon Technol Lett 2002;14(3):313–5. 59. Kuschnerov M, van den Borne D, Piyawanno K, Hauske FN, Fludger CRS, Duthel T, et al. Joint-polarization carrier phase estimation for XPM-limited coherent polarization-multiplexed QPSK transmission with OOK-neighbors. In: Proc. European conference on optical communication (ECOC), paper Mo.4.D.2. Brussels; 2008. 60. Kahn JM, Ip E. Principles of digital coherent receivers for optical communications. In: Proc. optical fiber communication conference (OFC), paper OTuG5. San Diego, CA; 2009.

CHAPTER

Optical performance monitoring based on nonlinear optical techniques

11 Daniel C. Kilper

Bell Laboratories, Alcatel-Lucent, Holmdel, New Jersey

11.1 INTRODUCTION Optical performance monitoring (OPM) is concerned with obtaining performance information about an optical communication signal. Given that the target is an optical signal, it is natural to consider that optical processing techniques might be used to efficiently extract this information. In some cases, the motivation for using OPM is due to the difficulty or cost associated with terminating the signal at a receiver and obtaining performance using electronics. Furthermore, a wide array of optical techniques is available.1,2 Among these are methods that take advantage of nonlinear optical phenomena. Nonlinear optics deal with processes in which optical fields modify the properties of a material with which they are interacting. From such effects, new optical or electronic signals can be generated that are nonlinear functions of the originating optical signal properties and can be used for monitoring. Switching and logic devices depend on input-output (transfer) functions that are nonlinear and thus nonlinear optics are a method to perform such operations directly on an optical signal. Several factors motivate the use of nonlinear optics for OPM applications. Most optical techniques, whether linear or nonlinear, have the advantage of wide bandwidth. Spectroscopy is an example of a linear optical technique commonly used for monitoring optical signals with the key attribute being a bandwidth covering hundreds of terahertz (THz). Many nonlinear optical phenomena, such as two-photon absorption (TPA) and four-wave mixing, also have bandwidths measured in THz. For nonlinear methods, however, both the frequency or wavelength range of the material and the speed of the process must be taken into account. Most nonlinear processes involve atomic phenomena that are very fast—in the pico- to femto-second range. In some devices, however, such as semiconductor optical amplifiers (SOAs), many of the stronger nonlinear effects are dependent on slower processes related to carrier dynamics and may evolve in nanoseconds.3 In typical monitoring applications using nonlinear optics, wide bandwidth information is converted to a low-speed signal that can then be processed easily and at low cost using conventional electronics. TPA, for example, can measure the statistical variance of the intensity of a high-speed signal (THz bandwidths) and provide the result as a low-speed photocurrent.4 Another advantage of nonlinear optics is that complex computations can often be implemented using simple device configurations. Calculating the variance of a signal using two-photon absorption has already been mentioned. Cross- and auto-correlation, level detection, spectrum analysis, and digitalto-analog conversion are other examples. Again, this is also an advantage for linear optics, © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00011-0

301

302

CHAPTER 11 OPM based on nonlinear optical techniques

although nonlinear optics provide the added dimension of including signal amplitude-dependent quantities and thus provide a broader range of processing tools. For digital signals, the primary performance indicator is usually the bit error rate (BER). The BER can be determined using a threshold detector and comparing the resulting bits with a reference. Self-referencing techniques involve comparing the bits recovered using two or more different thresholds or by using variable thresholds. Through optical level detection and gating, nonlinear optics provide the capability of obtaining digital bit-related performance indicators directly from the optical bits. Although monitoring using linear optical processes such as diffraction have found widespread use, using nonlinear optics in transmission system monitoring applications has met with limited success. Unlike linear methods, nonlinear optics require intense optical fields to generate a large effect—by the very nature of the phenomena. Of course, there are methods that compensate for this such as using specialized materials with high nonlinear coefficients, using small area-guided modes or focused beams, or mixing with intense control or local oscillator fields. Nevertheless, the added complication of these methods or materials detracts from the overall benefit of a nonlinear approach. Optical methods also suffer in monitoring applications because their wideband advantage is often diminished due to band limiting on the optical signals in communication systems.5 Filtering in optical switches such as reconfigurable optical add-drop multiplexers (ROADMs), especially in high-spectral-efficiency, dense wavelength-division multiplexing (DWDM), results in signals that have little spectral content beyond the symbol rate. Therefore, the additional bandwidth capability provided by the optical techniques is not needed in many in-service monitoring applications. Since the optical signals are generated and received using opto-electronics operating at or near the symbol rate, opto-electronic components are readily available covering the bandwidths of interest for these signals. For monitoring signal quality in a 40-Gb/s system with 50-GHz channel spacing, 40-GHz bandwidth is usually adequate. There is little benefit to using a device with 100-GHz bandwidth because there are no signals or noise at 100 GHz—everything is filtered optically to below 50 GHz. In practice there is some benefit to measuring a signal at a higher bandwidth than the signal rate, but this level of detail is more suited for specialized test and measurement analysis than inservice monitoring. In fact, some success in the use of nonlinear optics in test and measurement applications has been noted due to the higher premium placed on performance and the relaxed requirements with respect to size and cost.6,7 Several circumstances lend themselves to the use of nonlinear optics for OPM applications. The first is for performance verification and monitoring on an all-optical 2- or 3-R regenerator. Such a regenerator must itself be built using nonlinear optics and have a clear advantage over electronics if it is deployed. Thus, the potential to realize an advantage using optics is already present, and in order for the all-optical regenerator to provide the full functionality of an opto-electronic (OEO) regenerator, it would need an all-optical OPM capability. A second circumstance in which nonlinear optics may be advantageous is in cases in which the optical signal is not readily terminated at the monitoring location. For example, some dispersion maps will cause the optical signal to be highly dispersed during transmission only to be reconstituted at the receive location, such as pseudo-linear transmission.8 Monitoring such a signal at an intermediate location along the transmission path might require additional hardware such as a tunable dispersion compensator. In such cases, the extra optics required for the nonlinear method can be traded against the additional signal conditioning hardware required to recover the signal for reception, improving the cost-benefit equation. However, the recent trend toward flexible ROADM-based transparent networks,5 which support adding and dropping

11.2 Nonlinear optics

303

traffic at any node along a path, have moved systems away from this situation. Trends toward higher bit rates using advanced modulation formats and coherent reception9 may provide a new opportunity for nonlinear optics. As the transceivers become more sophisticated and integrate more optical components, there might arise opportunities for simple nonlinear optics–based methods to retrieve OPM data without requiring a full receiver. One final circumstance in which nonlinear optics can find applications is in situations in which the nonlinear signal is readily available due to other required processes. Recently photonic integration has received increased application in communication systems10 and often involves complex signal manipulation in optical waveguides. Two-photon absorption can be an impediment in these devices. However, by collecting the carriers generated by TPA, a monitoring signal can be provided. Examples have been demonstrated for use in tuning the optical pulse carver in a waveguide-based, return-to-zero (RZ) optical source.11 In this case, the nonlinear monitoring comes largely for free. Due to the high field densities and long interaction lengths in photonic integrated chips, it’s likely that nonlinear optics will find greater use in such devices, some of which might involve OPM capabilities.

11.2 NONLINEAR OPTICS Nonlinear optical effects are usually described using the material susceptibility.12 An intense optical field, E, will apply forces to the electrons in a material and create a polarization, P. This polarization can be expressed as a power series in the electric field through the susceptibility, w: Pðz; tÞ ¼ e0 wð1Þ Eðz; tÞ þ e0 wð2Þ E2 ðz; tÞ þ e0 wð3Þ E3 ðz; tÞ þ    ¼ Pð1Þ þ Pð2Þ þ Pð3Þ þ    ;

(11.1)

where e0 is the permittivity of free space. The various orders in susceptibility and ∂ polarization correspond to the different orders of nonlinear phenomena. For simplicity, only the scalar form is shown here. In general, the electric field and the polarization are vectors, and the jth order susceptibility w(j) is a tensor of rank j þ 1, relating all vector products of E to a given component of the polarization. The resulting electric field is found using Maxwell’s wave equation in which the polarization acts as a source term: ⇀

⇀

1 ∂2 E ∂2 P r E 2 2 ¼m 2 : ∂t c ∂t ⇀2 !

(11.2)

Since the polarization is related to the position of the charges in the host system—that is, the dipole moment—the second derivative of the polarization is proportional to the acceleration of the charges due to the applied electrical field. These accelerating charges generate a new electric field that interacts with the applied field to create the nonlinear response. Often the electric field is a superposition of multiple components and in general can be written in vector form as X⇀ ⇀ ⇀ ⇀ E ðr ; tÞ: (11.3) E ðr ; tÞ ¼ i

i

For the case in which all fields are aligned in the x direction, Equation (11.3) becomes ⇀

∗ E ¼ 12 x^E~1 ðx; tÞ ejfk1;x xn1 t’1 ðx;tÞg þ 12 x^E~1 ðx; tÞ ejfk1;x xn1 t’1 ðx;tÞg þ 12 x^E~2 ðx; tÞ ejfk2;x xn2 t’2 ðx;tÞg þ    ;

(11.4)

304

CHAPTER 11 OPM based on nonlinear optical techniques

where ki,x is the wave vector, ni is the optical frequency, and fi is the optical phase. The quantities E~i ðx; tÞ are the slowly varying field amplitudes. Many nonlinear effects can be well described using the slowly varying form of Maxwell’s equation in one dimension, which is a first-order partial differential equation.13 These equations should form the starting point for any treatment of nonlinear phenomena used for monitoring applications. There are examples in the literature in which attempts were made to overly simplify the action of the nonlinear response and apply an ad hoc description. These inevitably lead to incorrect results or hide important dimensions. A common example is the case in which noise due to amplified spontaneous emission (ASE) from optical amplifiers is considered. Pure amplifier ASE generates an electric field that is randomly polarized chaotic light. Chaotic light is characterized by a Gaussian distributed electric field amplitude, E~i ðx; tÞ.14 This corresponds to a continuous-wave field that has been disturbed by random phase jumps (or equivalently the superposition of many fields with random phases). For simplicity, ASE noise is often approximated as Gaussian-distributed intensity noise. While this approximation is incorrect, it coincidentally yields BER estimates for photodetection that are reasonably accurate.15 This coincidence should not be expected to occur for an arbitrary optical phenomenon, so in general one must resort to the chaotic light model and solve Equation (11.2) or the slowly varying equivalent. Although many nonlinear optical phenomena such as four-wave mixing and the Kerr nonlinearity in fibers have been shown to lead to nonclassical effects, semiclassical models are adequate for monitoring techniques used in systems today. The high losses and classical sources used in conventional optical communication systems remove the intrinsically quantum features. Complex methods can be used to create and preserve quantum communication channels for very specialized applications,16 and in these cases a fully quantum description might be needed.14 Depending on the number of electric fields and the relative strength of each term in the susceptibility, many different optical phenomena are described. Optical absorption and refraction are determined from the first-order linear term. The second-order susceptibility, w(2), describes processes such as second harmonic generation. The second-order response requires that the material be noncentrosymmetric. Most optical fiber is centrosymmetric and therefore the dominant nonlinear response comes from w(3). However, numerous crystals and some specialized optical fibers are noncentrosymmetric and have a strong w(2) response. The third-order susceptibility, w(3), describes many effects including four-wave mixing, Kerr nonlinearity, TPA, and Raman scattering.

11.3 OPM TECHNIQUES USING NONLINEAR OPTICS Table 11.1 lists common nonlinear optical phenomena used in OPM techniques associated with various orders of nonlinear susceptibility. References are provided for many papers that describe OPM techniques related to these effects. The reader is also directed to recent tutorials1,6,7 and book chapters.2,17 There are several different aspects of nonlinear optical phenomena that can be exploited for monitoring purposes. Perhaps the most common use is optical level detection or bit amplitude detection. The idea is to set a power threshold so that a monitor output is generated when the instantaneous optical signal power is above the threshold and no output if the signal power is below the threshold. This method is a form of optical demodulation and it is often used in optical regeneration techniques. For monitoring purposes, however, a low-speed, signal quality–dependent output is usually the goal.

11.3 OPM techniques using nonlinear optics

305

Table 11.1 Nonlinear Effects and Their Applications to OPM Order

Phenomenon

Measurement

References

w

Second harmonic generation

Optical sampling Frequency-resolved optical gating Bit amplitude detection Autocorrelation Bit amplitude detection Bit amplitude detection Intensity power spectrum Bit amplitude detection Phase noise extraction Bit amplitude detection

18,19,20 21,22 23–29 23, 30 31, 32 33–37 38–40 41–44 45 46

(2)

w(3)

Two-photon absorption (TPA) Self-phase modulation (SPM) Crossphase modulation (XPM) Four-wave mixing (FWM)

*

Laser injection locking

*Note: Nonlinearity is created by device operation and not directly from the material response.

For example, the average power generated by the threshold detection output can be related to the number of mark bits in the signal. If the signal is distorted, fewer bits will cross the threshold and the monitor output will decrease. In practice it is difficult to create a sharp threshold. Nevertheless, an output that is nonlinear in the input peak power is usually enough to provide a quality indication: larger output from larger 1-bits. Such monitors can be highly effective for detecting the influence of various forms of dispersion or distortion, including group velocity dispersion, polarization mode dispersion, and self-phase modulation. These methods tend to be only weakly responsive to changes in noise. Often, however, a change in optical signal-to-noise ratio (OSNR) can be detected if the total signal power is held constant as the OSNR degrades. The increase in noise power results in a decrease in signal power (for the same total optical power in the channel). These thresholding monitors can be very sensitive to such signal power changes, particularly if referenced to the average power measured in a standard linear detector (e.g., a photodiode). In fact, since most of these techniques are sensitive to changes in the average power, such referencing is usually required. Since optical thresholding is an essential element in a signal regenerator, thresholding monitors have been investigated for use with optical regenerator technologies. One promising approach for optical regeneration is to use crossphase modulation (XPM) to transfer the data on an impaired signal to a new continuous-wave (CW) probe field. Due to the intensity dependence of the XPM, the probe field will undergo a strong frequency shift only when a well-formed mark or 1-bit is present. By optically filtering this signal, these frequency shifts are converted to an amplitude modulation. Noise on the spaces or 0-bits is rejected by the filter. Using a bandpass filter, large frequency shifts can also be rejected resulting in reduction of noise on the marks as well. This type of regenerator is shown in Figure 11.1(a). A similar configuration is used for self-phase modulation, except that the probe signal is removed. Since noise is being rejected, the difference in power between the regenerated signal and the input signal is related to the signal quality. In practice, the sensitivity to changes in signal quality may not be large, since the signal itself is a large background on this measurement. By moving the filter to other portions of the spectrum, however, greater sensitivity can be obtained. Figure 11.1(b) shows an example in which the filter position is optimized for OSNR sensitivity

306

CHAPTER 11 OPM based on nonlinear optical techniques

CW probe

Regenerator Nonlinear device

Data in

P1

P2 Filter

(a) –34

–16 13.9-dB OSNR

21.1-dB OSNR

–39

–26

–44

–36 13.9-dB OSNR

21.1-dB OSNR

–46 1549 1550

1548.5

1550.5

4.0

0 Power (dBm/0.02 nm)

1551

–10 3.0 –20 –30

2.0

–40 1.0 –50 –60 1547.5

Monitor sensitivity (dB)

–49 1548

0.0 1548.5

1549.5

1550.5

Wavelength (nm)

(b)

FIGURE 11.1 (a) Common nonlinear-threshold-based monitor integrated with 2-R optical regenerator. Monitoring signal is ratio of average power measured at P2 normalized by input power at tap P1. The variable optical attenuator can be used to maintain constant input power. (b) Optical spectrum of 40-Gb/s NRZ-OOK signal generated by XPM in SOA and OSNR monitoring sensitivity obtained by moving output filter to different wavelengths.33 Copyright © 2005 IEEE.

by probing different locations. In this case, XPM in an SOA is used. The response is complicated, however, by the presence of gain saturation and four-wave mixing. Even for optimal placement of the filter, the ASE noise generated by the SOA limits the range of the monitoring signal to roughly a factor of 2. In order to take advantage of the optimal filter placement, the tap in Figure 11.1(a) would need to be moved to before the filter and an additional monitoring filter would need to be added on the tap P2. XPM tends to be more strongly influenced by signal distortions such as that caused by fiber dispersion and self-phase modulation. The dependence of the monitoring signal on dispersion in Figure 11.1 varies by roughly an order of magnitude. In all cases, the input signal power to the SOA and the SOA pump power were held constant. Similar methods for amplitude-level detection have been implemented using highly nonlinear fiber. In the case of RZ phase-modulated signals such as RZ-DPSK, the RZ pulse train is not modulated, and therefore the data signal does not appear in the spectrum of the probe signal after XPM in

11.3 OPM techniques using nonlinear optics

307

the nonlinear medium. This effect was shown to provide nearly background-free monitoring using highly nonlinear fiber.37 Figure 11.2(a) shows the experimental setup and the different XPM-induced probe spectra depending on different amounts of chromatic dispersion on the data signal. By monitoring the spectral minimum between the probe carrier and the induced RZ clock tone, a large monitoring signal variation is generated (up to 16 dB), as shown in Figure 11.2(b).

Distorted pulses due to CD XPM effect

RZ-DPSK signal

Optical spectrum of inserted CW probe

Optical spectrum of the CW probe (after HNLF)

HNLF

CD ≠ 0 ΔP

CW probe CD = 0

lprobe lprobe

(a)

Input power: 12 dBm

Relative optical power (dB)

16

10 dBm

12

8

5 dBm 4

0

0

20

40 60 80 Chromatic dispersion (ps/nm)

100

120

(b)

FIGURE 11.2 Chromatic dispersion monitoring on 40-Gb/s RZ-DPSK signals using XPM in highly nonlinear fiber. (a) Setup and concept: only clock tones from the RZ pulse train are observed in XPM-induced spectrum of CW probe, but chromatic dispersion distorts the signal-creating power at other wavelengths for nearly background-free monitoring. (b) Monitoring signal variation at 25-GHz offset from probe wavelength for various signal input powers.37 Copyright © 2008 IEEE.

308

CHAPTER 11 OPM based on nonlinear optical techniques

numeric –40

–80 0

640 320 Frequency (GHz) (a)

–40

–60

–80 –1.5

0 GVD (ps/nm) (b)

1.5

Optical tone power (dBm)

Power (dBm)

0

Optical tone power (dBm)

Another approach to optical level detection is to use variable-threshold optical logic gates to probe the bit amplitudes. A common method in opto-electronic-based OPM is to measure the Q-factor of the signal using a variable threshold receiver. The Q-factor for an on-off keying (OOK) signal is the eye opening divided by the sum of the standard deviations of the noise on the marks and spaces.1,2 One of the most reliable methods for obtaining the Q-factor is to scan the eye opening by varying the decision level of the receiver. This method can be implemented using optical logic gates. Often, however, the threshold for the nonlinear process is not sharp enough. Recently, logic based on laser injection locking has been used to implement a dual-threshold bit error detector.46 Injection locking can provide a sharp transition as the output signal of a laser jumps to a new wavelength depending on the amplitude of a small injected control signal. In this case, a high and low threshold is applied to gates on the incoming optical data and an XOR operation detects when bits fall within the transition region result in errors. Unfortunately, the speed of the injection-locking process is limited by cavity dynamics so that operations beyond 40 Gb/s may be difficult to achieve. Other methods for signal monitoring include transforming the signal in ways that highlight signal quality–dependent features. For example, XPM has been used to generate a broadband-intensity power spectrum.38 Using this approach, radio-frequency (RF) spectral monitoring techniques39,40 can be applied to very high-frequency signals. Recently power spectra exceeding 640 GHz were generated using a photonic chip device.40 Figure 11.3(a) shows two monitoring signals obtained by detecting the level of a clock tone in the intensity power spectrum of a 320-Gb/s OOK signal. The intensity autocorrelation can also be obtained using nonlinear detection, and then a monitoring signal can be extracted either directly from the autocorrelation30 or from the power spectrum obtained by Fourier transformation of the autocorrelation. Optical parametric amplification through four-wave mixing in fibers is well known to provide phase-dependent gain. In another novel technique, this process has been used to amplify the phase noise on a phase-modulated signal and to cancel the modulated signal in order to detect the noise.45

–45

–50

–55 0

1.5625 DGD (ps)

3.125

(c)

FIGURE 11.3 (a) 640-GHz-intensity power spectrum generated using crossphase modulation in fiber. (b) Chromatic dispersion and (c) polarization mode dispersion monitoring signals generated by detecting clock tone in intensity power spectrum of a 320-Gb/s signal.40 Permission from Macmillan Publishers Ltd., Nature Photonics © 2009.

11.4 Key challenges

309

Optical sampling can also be implemented using nonlinear detection in order to generate a lowspeed representation of the signal.6,7 These methods can be highly effective for obtaining modulation waveforms of very high-speed signals that are not accessible using electronic sampling techniques. Although recently optical sampling was implemented using a linear optical technique, nonlinear methods are typically used to obtain a gate that is determined by a short optical pulse. The short pulses can have a low-speed repetition rate so that the samples can be detected with conventional (low-speed, low-cost) opto-electronics. The optical waveform can be reconstructed from the resulting samples. Because a short-pulse optical sampling source is required, these techniques are usually too costly for use in OPM applications. As mentioned above, the wide bandwidth provided by optical sampling is often not required for monitoring communication signals. However, such methods can be very effective in test and measurement applications.

11.4 KEY CHALLENGES 11.4.1 Sensitivity

Monitoring sensitivity to changes in signal characteristics is particularly important for nonlinear techniques because of the high intensity needed to create the nonlinear response. Before addressing the device sensitivity, it is important to understand the requirements for using devices in optical systems. In general, this evaluation must be carried out for each specific transmission system for which the monitoring will be used. Here, requirements that are common in many long-haul DWDM transmission systems are described. Figure 11.4 shows the different configurations that can be used to probe a link at a repeater site. A typical line amplifier will have two stages with various devices inserted midstage such as a dispersion-compensating module (DCM) or dynamic gain-equalizing

Input span

Amplifier stage 1 P11

Tap 1

DCM P12

Tap 2

Output span

Amplifier stage 2 P13

Tap 3

P14 Tap 4

Booster amplifier From taps

OPM

A Monitor path configurations

Tunable filter From taps

OPM

FIGURE 11.4 Common monitoring configurations on an amplified long-haul link.

B

310

CHAPTER 11 OPM based on nonlinear optical techniques

filter (DGEF). Taps can be located at each of the input and output ports of the amplifier stages. The power available for monitoring is a simple product of the losses from the line tap location to the monitor: Pmi ¼ Pliifc, where i, f, and c are the losses due to the ith (i ¼ 1–4) tap, loss due to the filter, and coupling losses due to fiber connections, respectively, and Pli is the optical power on the line at the ith tap. Note that the optical power can be expressed as the peak or average power for the signal or channel to be monitored. The line powers at tap 4, which correspond to the span launch powers, are typically in the range 1 to þ5 dBm per channel for standard single-mode fiber (SSMF), erbium-doped amplifier (EDFA)–based DWDM systems. The use of coherent modulation formats may shift this range down by roughly 5 dB, and the use of Raman amplification can shift the power down by as much as 5–15 dB. Other fiber types that tend to generate stronger nonlinear impairments such as non-zero-dispersionshifted fiber may require launch powers shifted down by another 2–5 dB. The input power to a DCM is 5–10 dB below the span input power, and the DCM loss is in the range of 4–10 dB. Thus, Pl2 ¼ RpPl4 and Pl3 ¼ DPl2, where Rp is the ratio of the DCM input power to the span input power and D is the DCM loss. The tap 1 location tends to be very problematic for monitoring because it can vary widely in power due to different span losses. Span losses may vary between 10 and 35 dB, depending on the length and the different splice or coupling losses involved. Although low-loss SSMF fiber can have a loss coefficient of 0.16 dB/km, transmission spans in a deployed system may have losses as high as 0.3–0.4 dB/km (averaged over the entire span), due to numerous splices or intra-office patch panels. Span lengths can vary from a few kilometers to over 100 km. Taps on the transmission line are usually 17–20 dB, although taps as high as 5 dB might still be tolerated at specific locations in a repeater design, for example, at the midstage access. Although tap 4 has the highest power, power is at a premium at that location because any additional tap losses will require more pump power in the amplifiers. A tunable filter may add 3–10 dB of loss and another 1 dB for coupling losses. Thus, the total loss from the line to the monitor might be in the range of 21–31 dB for a typical monitoring point. The power on the line can range from 5 dBm to 36 dBm per channel, considering all tap locations. Most nonlinear techniques will be limited to monitoring from tap 2 or 4 because of the available power. At these points the channel power might vary between þ5 and 11 dBm. Wavelength-dependent ripple in the optical power spectrum must also be considered in this estimate; thus, for a 0-dBm target-channel launch power, the channel powers might vary 3 dBm due to ripple. Two monitor path configurations are shown in Figures 11.4(a) and (b). A booster amplifier is used in configuration A. It is interesting to note that the line-to-monitor loss is within the range of span losses, with 22 dB being roughly the loss of a long-haul transmission span. Thus, inserting a booster amplifier before the OPM device is similar to monitoring the signal after one additional span, at least as far as the impact of the amplifier noise figure is concerned. With this in mind, the power into the OPM device for configuration A might be as high as 5 dBm for an EDFA system. Of course, such an amplifier adds significantly to the cost of monitoring. However, some savings can be derived from the fact that the amplifier is only single channel and might not need gain flattening, assuming that power control is available. Thus, two power targets are of interest for line-monitoring OPM techniques in long-haul DWDM systems: 0 dBm for configuration A and  25 dBm for configuration B, where the cost of the booster is the price for using configuration A. The strength of the nonlinear optical response will depend on the peak signal power. For an intensitymodulated OOK signal, the peak power is the power in a mark or 1-bit. If the signal is non-return-to-zero

11.4 Key challenges

311

(NRZ) with equal numbers of marks and spaces, then the peak power is twice the mean power, and for RZ the peak power can be expressed as P1 ¼ Pavg/dRm, where Rm is the ratio of the number of marks to the total number of symbols, and d is the duty factor (d ¼ 0.5 for a pulse width that is half the symbol width). Thus, for a RZ-OOK with equal 1s and 0s and a 33% duty factor, the power in the 1-bit is six times the average power. Note that this relationship is independent of the bit rate. Although shorter pulses yield a higher peak power for the same average power, this is only true if the pulse rate is constant. In communication signals, the pulse width or symbol width usually scales with the bit rate and therefore short-pulse, high–bit rate signals do not provide a significant advantage in terms of sensitivity for nonlinear methods. Increasing the peak power in transmission also has the disadvantage that it increases impairments due to fiber nonlinearities. The complexities of nonlinear transmission tend to keep the peak signal amplitudes low. The number of photons in a mark for a 10-Gb/s NRZ-OOK signal with 0-dBm average power is given by n ¼ 2Pavgt/hn ¼ 2  103  100  1012/(6.63  1034  3  108)/1550  109 ¼ 1.6  106 photons, where t is the bit period, h is Planck’s constant, and n is the frequency of light. With another 20-dB loss, this becomes 104 photons per bit, and scaling to 100-Gb/s signals, only 1000 photons per bit may be available for monitoring. The prospects for monitoring using such small signals are further diminished when one considers the resulting counting statistics. For classical detection, Poisson statistics will yield a signal-to-noise ratio (SNR) of n/n1/2 ¼ n1/2, where n is the number of photons. Assuming perfect efficiency, at 104 photons, the measurement SNR is 20 dB. This puts a boundary on the available accuracy for a given measurement. Nevertheless, nonlinear techniques have been demonstrated with sufficient sensitivity for monitoring as small as 400 photons per bit.19 Depending on the type of nonlinearity that is being exploited, a figure of merit can be derived for the material system or device design. This figure of merit enables one to evaluate the potential of a given material for use in the associated nonlinear effect or for selectingbetween different device  3 Re wð3Þ is perhaps the most configurations. The nonlinear refractive index coming from w(3), n ¼ 8n commonly used effect. The nonlinear response of the material is a function of the material susceptibility and the fields involved. Thus, the interaction length and the achievable field intensities are also important. Often materials with a large nonlinear index (real part of w(3)) exhibit strong TPA (imaginary part of w(3)) or large linear losses. Silica fiber has a relatively small n2, but can yield large effects from relatively small optical fields due to the fact that it has very low loss, small core size, and small TPA. Usually a nonlinear process will take place in a region of material in which a high optical intensity is achieved. For a given physical configuration, the product of the length of this region and the intensity is a common figure of merit: FOM1 ¼ ILeff. For optical fiber, the effective length of the interaction region is determined by the fiber losses, such that ILeff  P/Aeff a, where P is the power, Aeff is the effective area, and a is the absorption coefficient. In waveguides, the interaction length is often limited by TPA. In this case, a good figure of merit  is the ratio of the real to the imag3 Im wð3Þ is the TPA coefficient and l is inary part of w(3), which is FOM2 ¼ n2/bl, where b ¼ 4nl the wavelength. A key challenge is to find a material with a large figure of merit that is also robust for use in telecom applications. Optical fibers are an obvious choice and have been used in many techniques. Chalcogenide glasses have an exceptionally high FOM and can be used in waveguides or fibers, but have not been qualified for telecom applications. Another important quantity is the nonlinearity coefficient for fiber propagation: g ¼ Aneff2 l, which is a measure of the strength of n2-related nonlinear phenomena in fibers. Highly nonlinear fibers such as chalcogenides yield nonlinearity coefficients roughly 103 greater than silica fiber, which is 2.

312

CHAPTER 11 OPM based on nonlinear optical techniques

11.4.2 Cost, size, and complexity Any optical technique, whether linear or nonlinear, is complicated by the difficulties associated with packaging optical components. In many cases, the packaging costs are the dominant costs for an optical device. A key benefit of photonic integration is that several optical devices can be combined in a single chip, and therefore packaging is not required for every device. Even simple functions, such as an optical splitter, require bulky packaging. A 2  2 optical fiber coupler can be 2–4 cm in length and 0.5 cm in diameter and require fiber bend radii of several centimeters. OPM techniques can be divided into two categories: (1) techniques that provide unique functionality not available to opto-electronic methods or (2) techniques that compete with opto-electronic techniques. Pulse distortion, for example, can be measured using nonlinear optics to obtain the RF power spectrum38–40 or by receiving the signal and measuring the RF spectral power at specific frequencies.47 A photodetector matched to commercial bit rates and RF electronics to accomplish the opto-electronic measurement are compact and low-cost commodity items. A nonlinear optical technique that provides similar functionality must have a clear advantage over the opto-electronic approach. Just a few simple optical components will quickly put an optical solution at a disadvantage. Two approaches have been proposed to deal with this: (1) use photonic integration, as photonic chips solve the size problem and reduce cost through less packaging; or (2) use simple devices. A TPA detector is an example of a simple device—it’s essentially just a photodetector. However, the TPA device may need additional optics to provide polarization diversity and a strong lens to reduce the spot size and enhance the response.

11.4.3 Impairment isolation As with most OPM techniques, impairment isolation is often difficult using nonlinear optics. Many nonlinear techniques show strong sensitivity to the signal distortion, however, there is often little opportunity to distinguish what the root cause of that distortion might be. The same can be said for techniques that measure optical noise. Of course, many OPM applications do not require root cause analysis or the measurement of a specific quantity. For cases in which impairment isolation is important, nonlinear optics in principle have an advantage since they provide access to the native optical signal with very wide bandwidth. TPA benefits from being a simple measurement, but only has one degree of freedom, which can depend on many signal features. Figure 11.5 shows the different TPA spectra obtained for two different signals and filtered ASE noise by scanning a 0.2-nmwide filter in front of a TPA detector.48 The corresponding linear spectra obtained with a photodetector show little difference between the signals, although the TPA spectra show a difference of as much as a factor of 2. In principle, these differences can be exploited to identify signals with diverse modulation formats; the difficulty is that the TPA amplitudes also depend on the amount of distortion and noise present. On the other hand, by adding an interferometer in front of the TPA device, the full autocorrelation can be measured, which provides significantly more detail to potentially isolate the impact of different impairments.30 The interferometer does bring in significantly greater complication; however, fixed and tunable delay-line interferometers have become common for use in receivers for differential phase-shift-keyed signals, and therefore might find use in this application. One of the more challenging phenomena that complicate impairment isolation is periodicity of the signal amplitude with large variations in dispersion. Optical pulses that undergo fiber dispersion

11.4 Key challenges

313

40-Gb/s Filtered 10-Gb/s NRZ-DPSK ASE RZ-OOK Power (mW)

8.0 6.0 4.0 2.0 0.0 (a)

TPA (a.u.)

2.0 1.5 1.0 0.5 0.0 1564.0

1564.5

1565.0 1565.5 Wavelength (nm)

1566.0

1566.5

(b)

FIGURE 11.5 Comparison of optical spectra of signals using (a) linear photodetector and (b) nonlinear two-photon absorption detector. Nonlinear detection reveals differences between 40-Gb/s NRZ-DPSK, 10-Gb/s RZ-OOK signals, and filtered ASE noise.

will broaden and decrease in amplitude until neighboring pulses begin to overlap. Since neighboring pulses are generated from the same source, they will be coherent and can overlap and constructively interfere. As a result, pulses will periodically reform as pulses in different bit slots interfere. This effect is an analog of the optical Talbot effect. Amplitude-dependent nonlinearities, therefore, will tend to give a periodic response with increasing dispersion. As a result, dispersion measurements often will be ambiguous with respect to large differences in dispersion. One approach to monitoring optical noise using nonlinear optics is to rely on the different response that is obtained from a pure-noise signal compared to a noise-free signal. The TPA rate, for example, is higher for a 33% RZ-OOK-modulated signal than it is for ASE light. Therefore, the TPA will be reduced as the relative contribution of ASE increases for a fixed total average power. The TPA rate of NRZ-OOK-modulated light, however, is identical to the TPA rate for ASE light, and therefore the signal must be modified in order to detect an ASE contribution. This also complicates ASE monitoring for RZ-modulated signals, because fiber dispersion will broaden RZ pulses causing them to approach an NRZ waveform, and thereby remove the sensitivity to ASE. Figure 11.6 shows how the zero-delay peak in an autocorrelation measurement (corresponding to the direct TPA response) of a 10-Gb/s RZ-ASK signal, varies with OSNR for two different values of dispersion. At 470 ps/nm of accumulated dispersion, the autocorrelation peak no longer changes with OSNR: dispersion has transformed the signal so that it is similar to an NRZ pattern. Naturally, such effects can be used to isolate the contributions of different impairments, assuming sufficient degrees of freedom are available to access the various sensitivity conditions for an arbitrary input.

CHAPTER 11 OPM based on nonlinear optical techniques

AC peak signal (a.u.)

314

DL = 0 ps/nm DL = –470 ps/nm

10

15

20 OSNR (dB)

25

30

FIGURE 11.6 Experimental dependence of the zero-delay peak value of autocorrelation trace as function of OSNR for accumulated dispersions of 0 and 470 ps/nm. Solid lines are curves to guide the eye. Total average power (ASE noise plus signal) is constant upon filtering with a 0.2-nm optical bandpass filter.30

In general, not only do multiple impairments influence the monitoring signal strengths, but they also influence the sensitivity that the monitoring signal has to the other impairments.

11.5 SUMMARY Nonlinear optics provides a toolbox of diagnostic measurements with potential for use in OPM applications. Advanced functionality combined with wide bandwidth and almost unlimited bit rate scalability is the primary advantage of nonlinear approaches. In some cases, the ability to directly manipulate the optical signal provides capabilities or performance not achievable using electronic processing following photodetection. Targeting OPM opportunities in which electronic processing falls short is a promising route to commercialization. Often, however, OPM applications require simple, low-cost solutions and the higher performance is not required. On the other hand, test and measurement applications tend to place more of a premium on performance and need to handle lab signals not found in deployed systems. As a result, nonlinear optics can be expected to continue to find value in the laboratory, particularly as bit rates and the complexity of optical transmission systems increase. Optical monitoring devices have historically found utility in the test lab before being embraced in systems. For example, optical spectrometers were widely used in labs and for system maintenance, but were only recently deployed in systems. Greater use of photonic integration likely will lead to new applications for nonlinear optics. In electronics, monitoring is frequently implemented with little additional processing or cost. A similar argument can be made in the case of using nonlinear optics with other photonic integrated chip devices because additional packaging may not be required and many of the photonic components may be readily available on chips. While the role of nonlinear optics in OPM applications is still unclear, the promise due to the capabilities that are offered and the increased potential due to advances in transmission functionality and capacity remain strong.

References

315

REFERENCES 1. Kilper DC, Bach R, Blumenthal DJ, Einstein D, Landolsi T, Ostar L, et al. Optical performance monitoring. J Lightwave Technol 2004;22:294. 2. Willner AE, Pan Z, Yu C. Optical performance monitoring. In: Kaminow IP, Li T, Willner AE, editors. Optical fiber communications V B. San Diego, CA: Academic Press; 2008. 3. Agrawal GP. Nonlinear fiber optics. 2nd ed. San Diego, CA: Academic Press; 1995. 4. Guo WH, O’Dowd J, Lynch M, Bradley AL, Donegan JF, Barry LP, et al. Two-photon absorption generated by optically amplified signals. Electron Lett 2008;44:1087–8. 5. Feuer MD, Kilper DC, Woodward S. ROADMs and their system applications. In: Kaminow IP, Li T, Willner AE, editors. Optical fiber communications V B. San Diego, CA: Academic Press; 2008. p. 293–344. 6. Westlund M, Andrekson PA, Sunnerud H, Hansryd J, Li J. High-performance optical-fiber-nonlinearitybased optical waveform monitoring. J Lightwave Technol 2005;23:2012–22. 7. Dorrer C. High-speed measurements for optical telecommunication systems. IEEE J Sel Top Quant Electron 2006;12:843–58. 8. Raybon G, Mikkelsen B, Essiambre R-J. Transmission of short pulse OTDM signals over nonzerodispersion fiber. Proc LEOS Annual Meet 2000;1:106–7. 9. Liu X, Gill DM, Chandrasekhar S. Optical technologies and techniques for high bit rate fiber transmission. Bell Labs Tech J 2006;11:83–104. 10. Doerr CR, Okamoto K. Planar lightwave circuits in fiber-optic communications. In: Kaminow IP, Li T, Willner AE, editors. Optical fiber communications V B. San Diego, CA: Academic Press; 2008. 11. Griffin RA, Walker RG, Buck BJ, Powell R, Langley LN, Hall J, et al. 40 Gb/s RZ GaAs transmitter with integrated waveform monitoring, paper 10.5. Proc. ECOC 2002. Copenhagen; September 8–12, 2002. 12. Boyd RW. Nonlinear optics. San Diego, CA: Academic Press; 1992. 13. Meystre P, Sargent M. Elements of quantum optics. Berlin: Springer-Verlag; 1990. 14. Loudon R. The quantum theory of light. 2nd ed. Oxford: Oxford University Press; 1983. 15. Humblet PA, Azizoglu M. On the bit error rate of lightwave systems with optical amplifiers. J Lightwave Technol 1991;9:1576–82. 16. Kumar P, Chen J, Voss PL, Li X, Lee KF, Sharping JE. Fiber-optic quantum information technologies. In: Kaminow T, Li T, Willner AE, editors. Optical fiber communications V B. San Diego, CA: Academic Press; 2008. 17. Radic S, Moss DJ, Eggleton BJ. Nonlinear optics in communications: from crippling impairment to ultrafast tools. In: Kaminow IP, Li T, Willner AE, editors. Optical fiber communications V B. San Diego, CA: Academic Press; 2008. 18. Langrock C, Kumar S, McGeehan JE, Willner AE, Fejer MM. All-optical signal processing using w(2) nonlinearities in guided wave devices. J. Lightwave Technol 2006;24:2579–92. 19. Yang S-D, Weiner AM, Parameswaran KR, Fejer MM. 400-photon-per-pulse ultrashort pulse autocorrelation measurement with aperiodically poled lithium niobate waveguides at 1.55 mm. Opt Lett 2004;29:2070–2. 20. Shake I, Otani E, Takara H, Uchiyama K, Yamabayashi Y, Morioka T. Bit rate flexible quality monitoring of 10 to 160 Gbit/s optical signals based on optical sampling technique. Electron Lett 2000;36:2087–8. 21. Barry LP, Dudley JM, Bollond PG, Harvey JD, Leonhardt R. Simultaneous measurement of optical fibre nonlinearity and dispersion using frequency resolved optical gating. Electron Lett 1997;33:707–8. 22. Dudley JM, Barry LP, Bollond PG, Harvey JD, Leonhardt R. Characterizing pulse propagation in optical fibers around 1550 nm using frequency-resolved optical gating. Opt Fiber Tech 1998;4:237–65. 23. Barry LP, Thomsen BC, Dudley JM, Harvey JD. Autocorrelation and ultrafast optical thresholding at 1.5 um using a commercial InGaAsP 1.3 um laser diode. Electron Lett 1998;34:358–60.

316

CHAPTER 11 OPM based on nonlinear optical techniques

24. Inui T, Tamura KR, Mori K, Morioka T. Bit rate flexible chirp measurement technique using two-photon absorption. Electron Lett 2002;38:1459–60. 25. Wielandy S, Fishteyn M, Zhu B. Optical performance monitoring using nonlinear detection. J Lightwave Technol 2004;22:784–93. 26. Inui T, Komukai T, Mori K, Morioka T. 160-Gb/s adaptive dispersion equalization using an asynchronous dispersion-induced chirp monitor. J Lightw Technol 2005;23:2039–45. 27. Tian C, Kinoshita S. Polarization-independent waveform monitoring with two-photon absorption in Si-APD in high-speed transmission systems. In: Proc. of ECOC, paper We4.P.070 Glasgow, UK; 2005. 28. Bondarczuk K, Maguire PJ, Barry LP, O’Dowd J, Guo WH, Lynch M, et al. Chromatic dispersion monitoring of 80-Gb/s OTDM data signal via two-photon absorption in a semiconductor microcavity. IEEE Photon Technol Lett 2007;19:21–3. 29. Bondarczuk K, Maguire PJ, Reid D, Barry LP, O’Dowd J, Guo WH, et al. Chromatic dispersion monitoring for high-speed WDM systems using two-photon absorption in a semiconductor microcavity. IEEE J Quantum Electron 2009;45:223–32. 30. Dinu M, Kilper DC, Stuart HR. Optical performance monitoring using data stream intensity autocorrelation. J Lightwave Technol 2006;24:1194–202. 31. Westbrook PS, Her TH, Eggleton BJ, Hunsche S, Raybon G. Measurement of pulse degradation using all-optical 2R regenerator. Electron Lett 2002;38:1193–4. 32. Westbrook PS, Eggleton BJ, Raybon G, Hunsche S, Her TH. Measurement of residual chromatic dispersion of a 40-Gb/s RZ signal via spectral broadening. IEEE Photonics Technol Lett 2002;14:346–8. 33. Vorreau P, Kilper DC, Leuthold J. Optical noise and dispersion monitoring with SOA-based optical 2R regenerator. IEEE Photonics Technol Lett 2005;17:244. 34. Adams R, Rochette M, Ng TT, Eggleton BJ. All-optical in-band OSNR monitoring at 40 Gb/s using a nonlinear optical loop mirror. IEEE Photon Technol Lett 2006;18:469–71. 35. Luo T, Yu C, Pan Z, Wang Y, McGeehan JE, Adler M, et al. Optical chromatic dispersion monitoring of a 40-Gb/s RZ signal by measuring the XPM-generated optical tone power in a highly nonlinear fiber. IEEE Photon Technol Lett 2006;18:430–2. 36. Yang J-Y, Zhang L, Wu T, Wu X, Christen LC, Nuccio S, et al. Chromatic dispersion monitoring of 40-Gb/s RZ-DPSK and 80-Gb/s RZ-DQPSK data using cross-phase modulation in highly-nonlinear fiber and a simple power monitor. In: Proc. OFC/NFOEC. paper OTuG5. San Diego, CA; 2008. 37. Yang J-Y, Zhang L, Wu X, Yilmaz OF, Zhang B, Willner AE. All-optical chromatic dispersion monitoring for phase-modulated signals utilizing cross-phase modulation in a highly nonlinear fiber. IEEE Photon Technol Lett 2008;20:1642–4. 38. Dorrer C, Maywar D. RF spectrum analysis of optical signals using nonlinear optics. J Lightwave Technol 2004;22:266–74. 39. Blows JL, Hu PF, Eggleton BJ. Differential group delay monitoring using an all-optical signal spectrum analyzer. Opt Commun 2006;260:288–91. 40. Pelusi M, et al. Photonic-chip-based radio-frequency spectrum analyzer with terahertz bandwidth. Nat Photonics 2009;3:139–43. 41. Li S, Kuksenkov DV. A novel dispersion monitoring technique based on four-wave mixing in optical fiber. IEEE Photon Technol Lett 2004;16:942–4. 42. Ng TT, Blows JL, Mok JT, et al. Cascaded four wave mixing in fiber optical parametric amplifiers: application to residual dispersion monitoring. J Lightwave Technol 2005;23:818–26. 43. Ng TT, Blows JL, Eggleton BJ. In-band OSNR monitoring using fiber optical parametric amplifier. Electron Lett 2005;41:352–3. 44. Foster MA, Salem R, Geraghty DF, Turner-Foster AC, Lipson M, Gaeta AL. Silicon-chip-based ultrafast optical oscilloscope. Nature 2008;456:81–4.

References

317

45. Lu G-W, Miyazaki T. Optical phase noise extraction and amplification technique through four-wave mixing and its application to optical phase noise monitoring in (D)PSK systems. IEEE Photon Technol Lett 2009;21:881–3. 46. Chan LY, Qureshi KK, Wai PKA, Moses B, Lui LFK, Tam HY, et al. All-optical bit-error monitoring system using cascaded inverted wavelength converter and optical NOR gate. IEEE Photon Technol Lett 2003;15:593–5. 47. Petersen MN, Pan Z, Lee S, Havstad SA, Willner AE. Online chromatic dispersion monitoring and compensation using a single inband subcarrier tone. IEEE Photon Technol Lett 2002;14:570–2. 48. O’Dowd J, Kilper DC, Guo WH, Donegan JF, Chandrasekhar S. Optical channel monitoring using two-photon absorption. In: Proc CLEO Europe. Munich, Germany, 2007.

This page intentionally left blank

CHAPTER

Optical performance monitoring of optical phase–modulated signals

12 Bartłomiej Kozicki

NTT Network Innovation Laboratories, NTT Corporation, Japan

12.1 INTRODUCTION The continuing growth of the traffic volume generated by broadband Internet connections, distributed storage services, or cloud/grid computing is placing a rising demand on network operators to provide increasing amounts of bandwidth. This is expected to occur without significant changes in pricing of the services, which forces network operators to look for novel solutions in the field of optical transport networks. In order to take advantage of the available resources efficiently, the envisaged transparent optical network will accommodate dense wavelength-division-multiplexed (DWDM), high–bit rate channels.1 Consequently, the currently dominating on-off keying (OOK) modulation formats will pose a challenge due to limited tolerance toward chromatic dispersion (CD) and nonlinear distortions, as well as relatively low receiver sensitivity, which constrict the network design to a small number of transparent nodes.2 The solution lies in the adaptation of advanced modulation formats that, in the context of optical modulation, relate to all formats beyond binary intensity modulation.3 Advances in high-speed transponder electronics techniques such as controlled signal predistortion, maximum-likelihood sequence estimation (MLSE), and forward error correction (FEC) made increases in bit rates of individual channels possible.3–5 More importantly, the introduction of binary and multilevel phase modulation gave the opportunity to develop formats with high spectral efficiency (SE) and receiver sensitivity.6 As a result, a number of modulation formats have been introduced to the optical communication systems. In particular, phase-modulation formats have gained renewed interest in optical domain in the context of differential encoding and balanced detection.7,8 The return-to-zero differential phaseshift keying (RZ-DPSK) modulation provides an approximately 3-dB receiver sensitivity advantage when compared to RZ-OOK, as shown in Figure 12.1(a). It has been reported that the DPSK modulation formats can tolerate larger amounts of polarization mode dispersion (PMD) than OOK systems in both compensated and uncompensated systems.9 The RZ-DPSK format provides resilience to narrow optical filtering8 and crossphase modulation (XPM) both in fiber10,11 and in semiconductor optical amplifiers.12 Spectral efficiency of 0.4 bit/s/Hz in the RZ-DPSK format can be further enhanced by alternate phase modulation of intense RZ pulses, thereby forming the carrier-suppressed RZ-DPSK (CSRZ-DPSK) format.13 The CSRZ-DPSK format has been shown to allow transmission with SE of up to 0.8 bit/s/Hz.14,15 Similarly, high SE has been achieved with multilevel phase modulation. The return-to-zero, differential quadrature phase-shift keying (RZ-DQPSK) signal occupies © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00012-2

319

320

CHAPTER 12 OPM of optical phase–modulated signals

–4 –5

2.7 dB 21

OOK

–7 –8

Q-factor (dB)

log10 (BER)

–6

DPSK

–9 2.8 dB

–10 –11 –12 10

40 Gb/s

RZ-DPSK 18

Nonlinear tolerance improvement Sensitivity improvement

15 12

12

14 16 OSNR (dB / 0.1nm)

18

20

0

3 6 Channel power (dBm)

(a)

9

(b)

FIGURE 12.1 (a) Comparison of required OSNR between DPSK and OOK formats. (b) Comparison of nonlinear tolerance between RZ-DPSK and RZ-OOK formats.6,7

only 50% of the bandwidth of an RZ-DPSK signal at the same bit rate.16 The DQPSK format has also been shown to provide better linear performance than other quaternary formats, such as differential phase-amplitude-shift keying (DPASK).17,18 This is achieved through encoding 2 bits in four phase states of one symbol. The format relaxes the requirement for high-speed modulator electronics, the compressed spectrum ensures resilience to CD, and the longer symbol duration compared to the binary modulation makes DQPSK more robust to PMD. In conjunction with polarization multiplexing (PolMux), the RZ-DQPSK format can achieve SE of up to 1.6 bit/s/Hz.6 The multilevel encoding of RZ-DQPSK results in an approximately 1-dB receiver sensitivity penalty when compared to the RZ-DPSK format.19 Nevertheless, the high SE of the RZ-DQPSK and the receiver sensitivity advantage of the RZ-DPSK make these formats an attractive solution for transparent transmission. Multilevel modulation formats exceeding four levels have also been explored both in simulation and experiment. Formats like ASK-DPSK, 8DPSK, 16-quadrature amplitude modulation (16-QAM), and 64-QAM promise even higher SE. However, the increase in the number of bits per symbol results in a reduction of the distance between the signal points in the constellation map, which in turn causes degradation of receiver sensitivity, as well as increases the impact of CD and PMD.20–22

12.2 PERFORMANCE OF PHASE-MODULATED SIGNALS The use of phase modulation in optical communication systems is not a new concept, as the phasemodulated systems had been studied in the context of single-span transmission with coherent detection before development of the erbium-doped fiber amplifiers (EDFAs).23 The deployment of DPSK in optical transmission systems is advantageous from several viewpoints. In a linear channel dominated by the amplified spontaneous emission (ASE) noise, the DPSK signal demonstrates the same level of bit error ratio (BER) at an OSNR 3-dB lower compared to the OOK when balanced detection is used.

12.2 Performance of phase-modulated signals

321

The decrease of required OSNR is not the only advantage of the DPSK system. The RZ-DPSK format also shows improved performance during transmission through a nonlinear channel especially in the context of reduced XPM influence, as can be observed in Figure 12.1(b). In RZ-DPSK transmission, all bits have the same pulse power. In addition, in the case of pulse-overlapped 40-Gb/s transmission, the DPSK format is found to undergo less overall nonlinear penalty than OOK.6 The influence of intrachannel four-wave mixing (iFWM) on the phase-modulated format is lower with the same average power due to the lower peak power and the correlation of nonlinear phase shifts between any two neighboring bits. The RZ-DPSK format has been shown to be more robust than OOK to narrowband filtering.24 Further improvements in filtering performance are achieved in the multilevel RZ-DQPSK signal, which occupies 50% of the spectrum of the RZ-DPSK signal generated at the same bit rate. The compressed spectrum allows achieving high SE in DWDM systems and high tolerance to CD. Longer bit duration, compared to the binary modulation, also makes the DQPSK more robust against PMD.19

12.2.1 Signal impairments Transmission of signals in a transparent domain results in the accumulation of optical impairments. The impairments degrade the quality of the signal at the receiver, leading to reception errors. This section describes the most crucial optical phenomena that influence the performance of phasemodulated channels.

12.2.1.1 ASE noise/OSNR Optical amplification, most commonly realized using EDFAs, is accompanied by ASE noise. The variance of signal-spontaneous beat noise can be observed in the electrical waveform of the signal as a fluctuation around the signal level value. In OOK systems, the noise level can be related to a performance parameter called the Q-factor, which can be used to estimate system BER. However, in phase-modulated systems, the output of the balanced receiver is based on the phase component and noise distribution is non-Gaussian. Therefore, the Q-parameter cannot be used as a reliable method to determine the BER in phase-modulated systems.25 The performance of a beat noise–limited receiver can be fully characterized by the level of optical signal-to-noise ratio (OSNR) necessary to obtain a target BER.19 For this reason, a reduction in OSNR reduces the margin for other system impairments. The OSNR is customarily defined independently of the data rate in an optical resolution bandwidth Bo of 0.1 nm. The reported OSNR required to achieve BER of 1
109 in an RZ-DQPSK signal at 20 Gb/s is approximately 17.5 dB.26 The RZDPSK signal at the same bit rate requires 2.3 dB less OSNR to achieve the same BER.27 It has to be stressed that the required OSNR measured in a fixed-resolution bandwidth is increased by 3 dB for every doubling in signal bit rate.

12.2.1.2 Chromatic dispersion While the level of ASE-related OSNR reduction scales with the number of traversed amplifiers, the influence of CD scales with the length of transmission fiber. Moreover, it scales with the square of the bit rate, as the spectrum of a higher bit rate signal is wider and the bit period is shorter. The uncompensated CD results in receiver penalty. In the case of phase-modulated signals, a 2-dB penalty occurs for accumulated CD of 32 ps/nm in a 42.7-Gb/s RZ-DPSK signal28; this

322

CHAPTER 12 OPM of optical phase–modulated signals

2.5

4

2.0

p 0 0

Time (ps)

50

2

Penalty (dB)

Penalty (dB)

6

Differential phase (rad)

Intensity (a.u.)

8

1.5

1.0

0.5

0 –200 –100 0 100 200 Chromatic dispersion (ps/nm) (a)

0

0 5 10 15 20 25 30 Differential group delay (ps) (b)

FIGURE 12.2 (a) Power penalty as function of accumulated chromatic dispersion for 42.7-Gb/s RZ-DQPSK signal. (b) Power penalty as function of DGD for 42.7-Gb/s RZ-DQPSK signal.27

corresponds to 512 ps/nm in a 10-Gb/s signal. In the case of an RZ-DQPSK signal, the same penalty of 2 dB was reported for accumulated CD of 130 ps/nm in a 42.7-Gb/s signal, as seen in Figure 12.2(a). This corresponds to 520 ps/nm in a 20-Gb/s RZ-DQPSK signal.27 The agreement between these results stems from the fact that the 10-Gb/s RZ-DPSK and 20-Gb/s RZ-DQPSK signals occupy the same bandwidth and have the same symbol rate.

12.2.1.3 Polarization mode dispersion Apart from CD, PMD is another dispersive process affecting the quality of optical transmission. It occurs as a result of random deviations from the circularity of the fiber causing birefringence between the polarization axes.29 The optical signal, ideally consisting of two degenerate modes, experiences a polarizationdependent differential group delay (DGD) during transmission. Random coupling between the states of polarization due to fiber stress or nonlinear birefringence lead to fluctuations of DGD in time. This makes PMD more difficult to control and compensate than CD.30 Unlike in the case of CD, where the amount of temporal spreading is related to the optical bandwidth of the signal, the (first-order) PMD is independent of the frequency of signal components. The penalty induced by PMD on the optical channel depends predominantly on the waveform of the signal. For that reason, the influence scales linearly with signal bit rate.19 The DPSK signals generally perform better than the OOK signals,9 and in the case of a 42.7-Gb/s DPSK, a penalty of 2 dB was reported at a PMD-induced DGD of 14 ps.28 The RZ-DQPSK signal has been shown to perform significantly better than both DPSK and OOK due to the reduced symbol rate for a given bit rate.19,31 The 42.7-Gb/s RZ-DQPSK signal suffers a 2-dB penalty at DGD of 28 ps, as seen in Figure 12.2(b). In both cases, the 2-dB penalty occurs at a DGD of approximately 50% of symbol duration.

12.2 Performance of phase-modulated signals

323

12.2.1.4 Filtering and frequency offset An important penalty source in optical networks stems from signal spectrum filtering in reconfigurable optical add-drop multiplexer (ROADM) filters of multiple optical nodes. The impact of spectrum narrowing on the performance of DPSK signals has been studied in Reference 24. It has been shown that the DPSK modulation with balanced detection increases the robustness to optical filtering due to the better tolerance to intersymbol interference (ISI), as seen in Figure 12.3(a). In the case of DQPSK modulation, the reduced spectrum of the signal ensures that the penalty due to filtering is significantly lower than in the case of DPSK of the same bit rate.19 For a 40-Gb/s signal it reaches 1 dB after transmission over 10 filter stages with a bandwidth of 44 GHz.27 This has also been confirmed for PolMux DQPSK signals.32 The stated values of a filtering penalty are obtained under the assumption that the frequency of the optical carrier matches the predetermined grid and is aligned with all-optical components in the signal path. Without the proper supervision, the center frequencies of optical components will vary with time and fluctuate due to environmental factors. The frequency drift of the lasers or the optical components leads to relative frequency misalignment in the network and may cause optical power loss, crosstalk between channels, or penalty in the delay interferometer used for the demodulation of a DPSK or DQPSK channel. For a 40-Gb/s DPSK signal, a 1.2-GHz frequency offset in the demodulator leads to a power penalty of 1 dB.33 For the same bit rate, a 1-dB penalty in the DQPSK is incurred at only a 0.2-GHz frequency offset. Both cases are illustrated in Figure 12.3(b). Finally, it should be noted that the laser linewidth also has an influence on the performance of phasemodulated channels. In the case of DQPSK signals, the laser phase noise stemming from the nonzero

3 DQPSK

RZ-OOK RZ-DPSK, balanced RZ-DPSK, constructive port RZ-DPSK, destructive port

30 26

DPSK 2 1.5 X6

1 0.5

22 18

Sensitivity penalty [dB]

Required OSNR (dB)

2.5

NRZ 33%RZ

1

1.5 2 2.5 3 Optical filter bandwidth / data rate (a)

3.5

0 –10

–5 0 5 Frequenty offset (% bit rate)

10

(b)

FIGURE 12.3 (a) Comparison of robustness to optical filtering for OOK and RZ-DPSK formats.24 (b) Sensitivity penalty as function of frequency offset for DPSK and DQPSK format.33

324

CHAPTER 12 OPM of optical phase–modulated signals

linewidth of the laser was shown to impair the tolerance to accumulated CD.34 The influence of laser linewidth is even more pronounced when the number of phase levels is increased.35

12.2.1.5 Nonlinear phenomena Linear impairments described in the preceding sections affect the performance of optical signals in proportion with the traversed distance, number of optical amplifiers, and signal bit rate. Another group of transmission effects influencing the quality of the signal is dependent on optical power. The origin of nonlinear phenomena stems from the confinement of transmitted light to the small effective area of the fiber core, which results in high optical densities influencing the refractive index of silica through the optical Kerr effect. As the intensity of the optical field changes due to modulation of the signal amplitude, signal-induced refractive index fluctuations affect the phase of the optical signal. During transmission in dispersive fiber, the fluctuations in phase are converted into intensity variations, which distort the waveform shape.3 The influence of nonlinear phenomena on the signal depends on the employed modulation format as well as characteristics of the optical transmission platform. For systems with low SE, phenomena such as four-wave mixing (FWM), XPM, and self-phase modulation (SPM) have to be taken into account. When a low-dispersion fiber, such as a non-zero dispersion-shifted fiber (NZ-DSF), is employed, the FWM is the dominant effect.3 The phase-modulated formats are affected by the nonlinear phenomena to a lesser degree.36 The higher receiver sensitivity of the RZ-DPSK as well as the RZ-DQPSK format allows transmitting the signal with lower power compared to the OOK case, while achieving the same BER.7 This allows reducing the effect of nonlinear phenomena. Moreover, the RZ pulse carving was shown to reduce the distortion due to XPM for the RZ-DPSK11 and RZDQPSK signals.37 Because the waveform pattern is not dependent on data and does not change from bit to bit, the effect of XPM averages as the neighboring WDM channels pass each other due to CD. Therefore, no timing jitter is introduced to the signal.6 At symbol rates of 40 Gb/s and channel spacing of 100 GHz, both RZ-DPSK and RZ-DQPSK improve resilience to nonlinear phenomena by 3 dB, compared to OOK signals.27,37 Despite improved tolerance to XPM, the phase-modulated transmission is susceptible to another form of impairment stemming from nonlinear signal-noise interactions. The nonlinear phase noise (NPN), also called Gordon-Mollenauer noise, results from the conversion of amplitude fluctuations into phase fluctuations.38 The influence of NPN is especially pronounced in long-haul transmission systems39 and in the case where OSNR is poor during the transmission (regardless of the receiver OSNR).19 The most important types of signal-noise interactions are the SPM-induced NPN and the XPM-induced NPN.40 The random intrachannel (SPM) or interchannel (XPM) amplitude fluctuations are translated into phase modulation, which results in a penalty in the DPSK and DQPSK signal during reception. The penalty is stochastic and depends on the signal power and propagation distance, as well as the level of noise in the signal.34,41 Low-CD fibers increase the effect of NPN, as is also the case for other nonlinearities. The effect is most pronounced for signals in low symbol-rate systems where long coherence time is required. Moreover, the phase-modulated channels with differential receivers (DPSK) show higher sensitivity to NPN than the formats not employing selfhomodyne reception (PSK).42,43 It has been reported that phase noise cancels the 3-dB sensitivity advantage of the differential receiver.41 In the case of multilevel modulation formats, the penalty due to the NPN is expected to be larger than that of DPSK due to smaller spacing between the points in the constellation diagram.

12.2 Performance of phase-modulated signals

325

12.2.2 Generation and detection of N-PSK signals

In-phase Optical power

MZM transmission

The advantageous properties of phase-modulated formats come at a price of complexity in transmitter and receiver structures, compared to the OOK transmission. DPSK modulation can be performed either by a phase modulator (PM) or by a Mach-Zehnder modulator (MZM) biased at zero transmission and driven at twice the switching voltage.8 If a phase modulator is used, a constant-envelope signal is generated; however, the phase transitions between bits are not instantaneous and introduce chirp. Employing the MZM allows switching the phase between consecutive symbols precisely by p. The latter approach introduces residual amplitude modulation and is illustrated in Figure 12.4(a). The two neighboring transmission maximum points have opposite phase, which produces a phase shift of p regardless of drive voltage imperfections. Conversely, in the case of PM, the optical phase directly follows the electrical signal. Therefore, the phase transitions are limited by driver bandwidth and the imperfections of drive signal are reflected in phase distortions. The chirp or fluctuations in intensity can be suppressed in RZ-DPSK format, in which case an additional pulse carver is employed. DQPSK signal generation requires a more complicated modulator structure, imprinting the in-phase and quadrature data components. The modulators can be arranged either in parallel or serial fashion. Respective arrangements are presented in Figures 12.4(b) and (c). In the parallel arrangement, continuous-wave (CW) light is split between two nested MZMs operating as phase modulators. The phase of the optical carrier is modulated at the symbol rate producing dual DPSK signals. The phase of one of the DPSK tributaries is rotated by p/2, and subsequently the signals are combined. The

p

0

p p

CW 0

p 2 Quadrature

Time DPSK drive signal

Pulse carver

Data

Precoder

Time

(b)

In-phase

Quadrature

CW

Pulse carver

PM

(a) Data

Precoder (c)

FIGURE 12.4 Phase modulation of the optical signal: (a) generation of DPSK signal using MZM, (b) generation of RZ-DQPSK in parallel configuration, and (c) generation of RZ-DQPSK in serial configuration.

326

CHAPTER 12 OPM of optical phase–modulated signals

resulting field is a DQPSK signal with the bit rate equal to the sum of symbol rates in the tributaries. A pulse carver adds the RZ shaping to create the RZ-DQPSK signal. This arrangement provides modulator stability and precise carrier-phase modulation.3 The serial arrangement, presented in Figure 12.4(c), uses an MZM to generate the 0 or p in-phase component and a phase modulator to add the p/2 quadrature component. As in the previous case, a pulse carver is necessary to generate the RZ-DQPSK signal. The latter modulation scheme can be realized using discrete components.44 Other modulator structures have also been proposed in the literature. Details can be found in Reference 45 for generation of RZ/CSRZDPSK signals and in Reference 46 for DQPSK signals. The generation of a higher number of signal levels has been shown using the adaptation of the described modulators.47 The noncoherent phase-modulated systems use direct detection at the receiver. However, the square-law detection of a photodiode (PD) is not sensitive to the phase of the optical carrier. For that reason, a delay interferometer (DI) is inserted in the path of the signal in order to convert the phase modulation into amplitude modulation. Typically, a Mach-Zehnder interferometer (MZI) is employed. An MZI followed by a balanced receiver is shown in Figure 12.5(a). The multilevel phase encoding in the DQPSK signal requires a more complicated receiver structure. Typically, it consists of a pair of MZIs, as shown in Figure 12.5(b). The delay introduced by the interferometers is equivalent to the symbol rate. Contrary to the case of DPSK, in DQPSK, the MZIs introduce a differential optical phase shift between the branches. The differential phase is equal to p/4 and p/4 for the upper and lower branch, respectively. The output voltage signals after balanced detection are proportional to cos(△’i) þ sin(△’i) and cos(△’i) sin(△’i), where △’i is the differential phase shift between the neighboring symbols. This produces the in-phase (I) and quadrature (Q) data components, respectively.48 The complex structure of the DQPSK receiver causes an approximate 1-dB power penalty. Proper operation of the optical receiver has a major impact on the performance of both DPSK and DQPSK signals.8 For balanced DPSK detection, the detector amplitude imbalance of 40% will result in a 1-dB OSNR penalty. Similarly, a 1-dB OSNR penalty may be expected when the delay imbalance between the detector arms reaches 0.4 of the symbol period. Finally, a receiver detuning by frequency equal to 2% of signal bit rate will cause an OSNR penalty of 1 dB. In the case of DQPSK signals, the amplitude imbalance causing an OSNR penalty of 1 dB is also approximately 40%.49 The penalty due to delay imbalance varies depending on signal shaping. In the case of RZ-DQPSK modulation, the OSNR penalty reaches 1 dB at a delay imbalance of 0.4 of the symbol period.50 The sensitivity to frequency detuning is increased and results in an OSNR penalty of 1 dB for a detuning of 1.7% of the symbol rate.

I

π 4

–

π 4

–

a –

DI

–

Q

b (a)

FIGURE 12.5 Demodulator and balanced receiver for (a) DPSK and (b) DQPSK.

(b)

12.3 Optical performance monitoring

327

12.3 OPTICAL PERFORMANCE MONITORING Optical performance monitoring (OPM) techniques provide the means to analyze signal quality directly in the optical domain. The quality is estimated in terms of a single performance parameter. Currently, monitoring performance in the physical domain involves individual component alarms or optical power, inferring that the performance is satisfactory if optical components operate within the assumed limits and the optical channels are present.51 However, the target for OPM is the development of more sophisticated parameters, allowing the analysis of signal quality. Apart from signal power and OSNR, which are the most commonly considered impairments,52,53 accumulated CD and PMD are the parameters that can provide insight into optical channel performance.54 When considering phase-modulated transmission systems, additional factors need to be taken into consideration. The first is the laser frequency offset, which incurs a power penalty at the receiver. The second is the phase noise, which may stem from the nonzero laser linewidth or from nonlinear conversion of ASE into phase fluctuations. In the following sections, the techniques for monitoring the dominant impairments affecting the DPSK and DQPSK signals are described. First, the methods for estimating and suppression of NPN are presented in Section 12.3.1.1. Although these techniques do not provide feedback upon which signal quality can be estimated, they have been shown effective in reducing the penalty by monitoring optical waveform parameters. Next, the group of techniques monitoring the optical spectrum for estimation of the OSNR is discussed in Section 12.3.1.2. These techniques are shared with other modulation formats, and this section only discusses the merits and demerits of existing approaches. Subsequently, Section 12.3.1.3 focuses on monitoring techniques based on polarization nulling and DOP monitoring. Section 12.3.1.4 explains the group of techniques based on measurements of electrical radio frequency (RF) spectrum and clock tone. In Section 12.3.1.5, the techniques exploiting pilot tones are presented. While Sections 12.3.1.2 through 12.3.1.5 discuss the monitoring techniques observing the signal spectrum, the methods presented in Sections 12.3.1.6 through 12.3.1.8 focus on methods analyzing waveform samples. Section 12.3.1.6 discusses the application of linear optical sampling and phasor-based monitoring to phase-modulated signals. Sections 12.3.1.7 and 12.3.1.8 explain the methods of asynchronous sample analysis with histograms and two-tap plots, respectively. Finally, in Section 12.3.1.9, other monitoring techniques employing a receiver and signal processing are presented. Section 12.3.2 summarizes and compares the presented OPM techniques.

12.3.1 Monitoring techniques 12.3.1.1 Nonlinear phase noise compensation As mentioned in Section 12.2.1.5, the NPN stems from the interaction of ASE noise generated in the amplifiers and the signal through the fiber’s nonlinear refractive index. As a result, the fluctuations in intensity change signal propagation speed, influencing the signal phase. The relationship between the intensity fluctuation and phase noise has been employed as a feedback mechanism to counteract the adverse effects of NPN. Although there have been no reports on developing the parameter for quantifying the level of NPN for monitoring purposes, several techniques suppressing phase noise using the feedback signal have been presented in the literature. In one approach, a phase noise compensator consisting of a power monitor and phase modulator was proposed.55,56 A broadband photodetector placed before the receiver analyzes the incoming signal power to determine the amount of phase rotation necessary to counter that incurred through

328

CHAPTER 12 OPM of optical phase–modulated signals

the Kerr effect. The phase rotation is subsequently applied in the phase modulator. The proper adjustment of the amount of phase rotation and delay enables increasing the OSNR margin by 2.5 dB under the nonlinear transmission regime. This scheme has been shown to further improve DPSK signal performance in the nonlinear regime for a multispan transmission when polarization diversity is added to the compensation scheme.57 The principle of opposite-phase rotation correlated to instantaneous power fluctuation has also been demonstrated using nonlinear elements. In Reference 58, the periodically poled LiNbO3 waveguide provides a negative nonlinear phase shift, reducing the Q-penalty by approximately 6 dB for a single channel. The efficiency of this technique is reduced for a DWDM system due XPM-NPN, in which case a Q-penalty reduction of approximately 3 dB is observed. A different approach to compensating the NPN using nonlinear elements was shown in Reference 59. The DQPSK transmission is precompensated by a phase-preserving amplitude limiter. Employing FWM saturation in a highly nonlinear fiber (HNLF), the fluctuation of amplitude of the pulses is suppressed at the transmitter. This allows increasing the transmission power in a densely dispersion-managed system by 5 dB without incurring the penalty due to NPN. The compensation of nonlinear phase rotation was also demonstrated using numerical methods after signal reception. In Reference 60, the boundary decision threshold of the receiver was modified to follow the after-transmission spiral rotation of the phase, as shown in Figure 12.6(a). The constellation of a DPSK signal is restored by using a lookup table or phase subtraction correlated to the signal power, thereby reducing the standard deviation of NPN and suppressing the nonlinear phase rotation (Figure 12.6(b)). A similar phase-compensation approach in a digital coherent receiver is presented in Reference 61. The level of NPN affecting performance of the phase-modulated channels is dependent not only on the power and distance of transmission but also on the level of launched OSNR. The following section briefly discusses demonstrated techniques for spectral monitoring of OSNR in optical links.

12.3.1.2 OSNR/power monitoring The monitoring of optical power and OSNR through optical spectrum analysis is the most basic OPM method and the only one that has been standardized.62 It is also largely modulation-format independent. It compares the optical power contained in the signal to the power of ASE, typically within a bandwidth of 0.1 nm. It has been shown that the monitor placed within an optical network needs to provide information about the level of OSNR for levels of up to 26 dB for a 10-Gb/s signal.53 However, the analysis of spectrum of a DWDM signal transmitted over OADMs is likely to provide incorrect measurements as the densely spaced channels will obscure the level of OSNR. Moreover, channels with different OSNR levels are added and dropped and the ASE is shaped by the optical filters.63 A spectrum-based, OSNR monitoring technique solving the above issue has been presented for OOK and DPSK signals.64 It is based on cascaded, long-period fiber gratings equipped with an optically tunable phase shifter. The first grating splits the signal into a core mode and cladding mode that are recombined in the second grating. The optically tunable phase shifter, consisting of an Ytterbium-doped optical fiber, controls the phase difference between the core and cladding modes. Depending on the relative phase between the modes, either the power of incoherent ASE noise or the combined power of coherent signal and ASE noise can be measured, allowing the calculation

90

90 Region for “0”

Region for “0”

80

80

70

70

60

60 (0,0)

(0,0)

50

(0,0)

(0,0)

40

40 30 Region for “1”

= 2.0

(a)

Region for “1”

Region for “0”

Region for “1”

20 10

= 1.0

30 Region for “0”

20 10

= 1.0

= 2.0

(b)

FIGURE 12.6 (a) Phase distribution and decision regions of received signals with NPN for nonlinear phase shift of 1 rad (left) and 2 rad (right). (b) Corrected signal distribution for nonlinear phase shift of 1 rad (left) and 2 rad (right).60

12.3 Optical performance monitoring

Region for “1”

50

329

330

CHAPTER 12 OPM of optical phase–modulated signals

of OSNR within the channel. The method was shown to provide an OSNR monitoring range from 0 to 28 dB without being influenced by the accumulated CD or PMD. A similar interferometerbased, OSNR monitoring technique using a partial-bit-delay interferometer was demonstrated for a 40-Gb/s DPSK signal.65 The monitoring range of 5–25 dB was shown to be independent of the influence of both CD and PMD.

12.3.1.3 Polarization nulling and DOP monitoring The OSNR monitoring based on polarization nulling is an alternative to the spectrum-based OSNR monitoring and has been first presented in Reference 66. Basing on the assumption that the signal is contained in a single polarization, while the ASE noise is uniformly distributed between the polarizations, the monitoring of OSNR using a slowly rotating quarter-wave plate and a linear polarizer was demonstrated. By measuring the maximum and minimum optical power passing these two elements, the ratio between the signal and noise power can be extracted. The method was further refined through additional parameters and signal filtering in order to address the problem of signal depolarization due to PMD, or partial polarization of ASE due to polarization-dependent loss (PDL).67 Polarization nulling has been shown effective in monitoring the OSNR of RZ-DPSK signals from 22 to 33 dB.68 A different approach toward monitoring impairments using signal polarization is based on analysis of the degree of polarization (DOP). It has been demonstrated both without69 and with70 polarization scrambling at the transmitter. The OSNR can be measured as a function of signal depolarization due to the presence of unpolarized ASE noise. Apart from OSNR monitoring, the DOPbased technique has been shown effective in monitoring the influence of PMD in phase-modulated signals.71 PMD causes temporal misalignment between signal components traveling in the orthogonal states of polarization. The difference in arrival time between the two polarizations at the receiver reduces the signal’s DOP. For a DPSK signal, the DOP has been shown to provide unambiguous measurement of PMD for DGD values of up to 1.0 symbol period for the NRZ-DPSK format and 0.5 symbol period for RZ-DPSK. Similar results have been presented for the DQPSK modulation.72

12.3.1.4 RF spectrum/tone monitoring In contrast to the OPM techniques presented in Section 12.3.1.2, which utilized the optical spectrum, this section focuses on the electrical RF spectrum of the received optical channel. The RF spectrum corresponds to the power spectrum of the intensity, and is affected by optical impairments such as reduction in OSNR, CD, or PMD. This makes analysis of the RF spectrum a powerful tool for OPM. In the case of OOK modulation, the RF spectrum contains a strong component stemming from data modulation. In comparison, the RF component originating from signal beating and ASE noise in the receiver is relatively small. This makes the RF spectrum–based monitoring of OSNR in intensitymodulated systems impractical. However, in the case of phase-modulated signals, the power spectrum does not bear the data information, and is therefore lower than in OOK systems. The low-frequency components of the phase-modulated signal do not change even in the presence of accumulated CD. The small background (signal) RF component allows monitoring the power of the beat component between the ASE noise and the signal at low frequencies. This property has been employed in evaluating the level of OSNR.73 A slow optical detector with an electrical bandpass filter (BPF) is used to monitor the OSNR from 15 to 35 dB by observing the power around the 6-MHz frequency. Similarly, a low-frequency RF component was used to monitor the OSNR level in a singlechannel 80-Gb/s PolMux RZ-DPSK signal.74 A narrowband optical filter is used to extract a portion of the optical signal spectrum and ASE noise after an EDFA. Assuming a constant power level after

12.3 Optical performance monitoring

331

the filter, the ratio of signal power to the power of noise is reduced with the OSNR reduction. Observation of the RF spectrum component in the vicinity of 250 MHz enables OSNR monitoring from 11 to 29 dB. The use of low-frequency components ensures that the method is not influenced by accumulated CD or PMD. A modification of this technique also allows tracking of the PMD level. Observation of lowfrequency components in an 80-Gb/s PolMux RZ-DPSK signal was shown effective in PMD monitoring from 0 to 12 ps.75 This method takes advantage of the interaction between two signals in orthogonal polarizations after the polarizer and is illustrated in Figure 12.7. The polarizer is aligned with the polarization of one of the data streams. When no PMD is affecting the signal, there are no observable intensity components in the RF spectrum. However, in the presence of PMD, the orthogonal signals become depolarized and the crosstalk component between them can be observed after the polarizer. The observation of frequency tone in the vicinity of 250 MHz provides a CDindependent evaluation of PMD. It must, however, be noted that the sensitivity of the monitor depends on alignment precision between the polarizer and the polarization-multiplexed channel. Higher-frequency components of the DPSK signal are more susceptible to CD influence. The phase components of the signal are converted to intensity fluctuation with the accumulation of CD, which makes them appear in the RF spectrum of the signal. In contrast to OSNR monitoring with a low-frequency BPF, a high-frequency BPF can be used to observe the level of phase-toamplitude conversion, and therefore to monitor the level of accumulated CD. This approach was assumed in Reference 76, where a broadband BPF is used to observe the CD. The monitoring range is dependent on filter bandwidth, and for a 10-GHz-wide filter, at which the method showed the Slow axis DGD emulator

Pol-muxed RZ-DPSK PC H

H PC

RF spectrum analyzer

V 45° relative to DGD

Low-speed Polarizer detector

Fast axis

Rotating

V

(a) RF spectrum

No DGD With DGD

High DGD

Polarizer RF power

H

V

No DGD ΔPRF

f <250 Mhz (CD insensitive)

(b)

(c)

FIGURE 12.7 (a) Diagram of CD-independent PMD monitoring technique. (b) Alignment of polarizer with respect to signal in case without and with DGD. (c) Variation of RF spectrum power depending on level of DGD.75

332

CHAPTER 12 OPM of optical phase–modulated signals

highest sensitivity, the monitoring range reached from 0 to 800 ps/nm. The method was shown to limit the influence of OSNR on the monitoring result by analyzing the RF power within two independent frequency bands. The phase-to-amplitude conversion of the RZ-DPSK signal was also used for monitoring the level of accumulated CD in conjunction with CW probe and fiber nonlinearity.77 The CD-induced fluctuation of intensity in the phase-modulated signal is transferred to the probe through XPM in a highly nonlinear fiber. By measuring the degree to which the XPM modulates the CW light, the level of accumulated CD can be estimated up to 120 ps/nm for a 40-Gb/s RZ-DPSK signal. This method can also potentially be used for an 80-Gb/s RZ-DQPSK signal with a similar monitoring sensitivity. A particular case of RF spectrum monitoring is observation of the clock tone level. Accumulated CD causes a relative delay between the upper and lower sidebands of the signal. In the case of RZshaped, phase-modulated channels, the clock component in the RF spectrum will fade and rise periodically with the increase in CD due to the interference of components in the two sidebands. For the NRZ-shaped signals, these clock components are out of phase when the CD equals zero and produce no clock component in the RF spectrum. As the amount of accumulated CD is increased, the clock components interfere constructively. This phenomenon has been used to estimate the level of accumulated CD in NRZ-DPSK and RZ-DPSK signals.71 For 10-Gb/s signals, the measurement ranges from 0 to 600 ps/nm and 0 to 900 ps/nm for NRZ-DPSK and RZ-DPSK, respectively. In a system employing 10-Gb/s DQPSK modulation, the clock tone measurement showed sensitivity to the level of accumulated CD up to 720 ps/nm.72 In the abovementioned technique, the clock level is related to the amount of accumulated CD. However, the clock tone level is also influenced by PMD, resulting in ambiguity and possible measurement errors. A technique that allows monitoring of both CD and PMD in OOK and NRZ-DPSK signals was presented in Reference 78. It analyzes the RF clock power at the output of an unbalanced DI. The principle is illustrated in Figure 12.8. The DI has a quarter-bit delay in one arm. Inside the DI, the original signal interferes with itself for 75% of the bit period and with the phase of the following bit for 25% of the bit period. At the output, the constructive port introduces almost no change to the signal due to the large free spectral range (FSR) of the interferometer. The destructive port filters signal sidebands, leading to an RZ pulse train at its output. The 10-GHz clock at the output of both arms of DI is used to independently analyze CD and PMD. The clock power from the constructive port grows with an increase in CD and a decrease in PMD. Conversely, the clock from the destructive port grows with the decrease in both CD and PMD. Using these observations, CD can be evaluated from 0 to 600 ps/nm and PMD from 0 to 50 ps for a 10-Gb/s NRZ-DPSK signal.

12.3.1.5 Pilot tone monitoring The monitoring techniques described in the previous sections are based on the inherent properties of the signal’s optical or electrical spectrum. Optical impairments that affect signal quality also influence certain features of the spectrum. This concept is extended by introducing a distinct frequency tone that can be observed regardless of the information-carrying portion of the signal. The tone format, frequency, and intensity are designed to reflect the optical impairments affecting the signal during transmission. This principle is employed in OPM techniques based on pilot tones.79 The pilot tone monitoring techniques require a modification to the transmitter, which is illustrated in Figure 12.9. A low-frequency RF-modulated tone is added to the optical channel. The tone may modulate the amplitude, frequency, or phase of the carrier. It may be added through

12.3 Optical performance monitoring

333

Pulse-carving Spectral domain time domain 1/4-bit time delay Input NRZ signal

4x bit rate FSR

Destructive port

Constructive port

FIGURE 12.8 Partial-bit DI for NRZ signal monitoring. Fraction of bit interferes with following bit, which leads to pulse carving in destructive port. Pulses produce a strong clock tone.78

Bias

LD

LD

AM

Bias

LD Data

Bias

AM

PM

Bias Data

(a)

(b)

(c)

FIGURE 12.9 Modification to transmitter for pilot tone monitoring techniques.79

modulation of the laser diode bias (Figure 12.9(a)) or through an external-intensity or phase modulator (Figures 12.9(b) and (c)).80 At the receiver, the tone can be filtered from the signal using a BPF after the PD and analyzed using an RF power detector. The drawbacks of this method are the necessity to modify the transmitter and the degradation in receiver sensitivity of the monitored signal due to reduced extinction ratio, crossgain modulation in optical amplifiers, or stimulated Raman scattering.79 The realization of an intensity-modulated, pilot tone monitoring technique in phase-modulated systems has been shown to have a relatively small impact on the signal. Since the intensity spectrum of the phase-modulated channels does not contain low-frequency components, the power of pilot tone can be reduced while maintaining the tone-to-signal ratio, as shown in Figure 12.10. Using this method, a 300-MHz tone was employed to monitor the optical signal power, wavelength, and path in the optical DPSK signal.81 However, when compared to the pilot tone monitoring of the intensitymodulated channels, monitoring of transmission impairments has not been realized due to the tone’s low frequency. The pilot tone monitoring technique is not limited to intensity-modulated tones. The phasemodulated pilot tone has been shown effective in monitoring the frequency offset between the signal and the demodulator DI.82 A low-modulation index-phase dither tone is added to the DPSK signal. At the receiver, the dithered phase is converted to intensity in the DI when the center frequency of the channel and the interferometer does not match. The power of the recovered tone is proportional to the frequency offset and allows monitoring the detuning of the demodulator from 2 to 2 GHz with a resolution of 1 MHz. Similarly, in Reference 83 an intensity tone with a frequency equal

334

CHAPTER 12 OPM of optical phase–modulated signals

70

Res: 10 kHz

OOK (NRZ) DPSK

60

CNR (dB)

50

10 dB/div

40 30 20 10 10 dB/div

0

0

5

10 15 Modulation index (%)

20

25

FIGURE 12.10 Improvement of carrier-to-noise ratio (CNR) in DPSK signal over OOK signal for pilot tone monitoring technique.81

to half of the signal bit rate was used to monitor the frequency offset between the optical source and the demodulator. In this case, however, the presence of tone resulted in a power penalty of 0.4 dB.

12.3.1.6 Linear optical sampling and phasor monitoring The OPM techniques described thus far in this chapter have been focused on analyzing the properties of optical or electrical signal spectra that are influenced by transmission impairments. These techniques are popular because, in most cases, they do not require high-speed components. However, transmission impairments primarily influence the signal waveform. Accordingly, a number of monitoring techniques have been developed for direct signal analysis. Traditionally, the waveforms are characterized by measures such as rise and fall time, extinction ratio, or jitter. These values are calculated from the intensity eye diagrams of the sampled optical waveform.84 The introduction of phase–modulation formats opened a new field for sampling-based OPM. On the one hand, a new set of parameters had to be developed for analyzing intensity samples, as the waveforms are different from OOK systems. On the other hand, sampling techniques enabling the analysis of signal constellation provided new insight into both intensity and phase evolution of the channel. This section describes the OPM techniques that analyze the optical field based on samples acquired using linear optical sampling and self-homodyne phasor monitoring. Linear optical sampling is a technique for capturing optical signal intensity and phase with high temporal resolution. The signal is mixed with optical pulses generated by a local source. A 90 optical hybrid splits and recombines the two fields while introducing 0 and p/2 relative phase shifts. The interference patterns are received by balanced photodiodes, as shown in Figure 12.11. Interference between the two fields is used to recover the incoming signal’s intensity and phase. The use of short, repetitive optical pulses for the local oscillator forms a gate for the sampling process, thereby reducing the requirement for high-speed electronic sampling components.85 Linear optical sampling allows analysis of signal constellations generated by optical phase– modulated sources. It is capable of revealing modulation distortion and OSNR degradation, as

12.3 Optical performance monitoring

Source under test

SA

CA

SS

335

SD

Sampling source

SB

CB

p/2 phase shift

FIGURE 12.11 Diagram of optical circuit for mixing of optical signal with local oscillator light (90 optical hybrid).81

(a)

(b)

(c)

FIGURE 12.12 Signal constellation obtained using linear optical sampling: (a) PSK signal modulated with MZM, (b) signal modulated with PM at driving voltage of 2/3VP, and (c) PSK signal with OSNR ¼ 13 dB.85

illustrated in Figure 12.12. Figure 12.12(a) presents a constellation diagram of a PSK signal modulated by MZM. The phase noise is relatively small, while the noise stemming from imperfections in the drive signal introduces a minor variation in samples along the real axis. Alternatively, a constellation diagram of a PSK signal generated by a phase modulator driven by voltage of 2/3VP is shown in Figure 12.12(b). The incomplete inversion of phase and the phase noise stemming from the drive signal noise can be clearly observed in the constellation diagram. Finally, Figure 12.12(c) shows the effect of ASE noise on the constellation diagram when the OSNR of the signal is reduced to 13 dB. Analysis of the obtained constellation diagrams allows quantifying the amount of signal distortion.86 Taking Figure 12.12(c) as an example, by calculating the variance of samples in phase or amplitude around the mean value of the symbol, it is possible to estimate the level of ASE noise. This value is directly related to OSNR when normalized to the signal’s average power. An example of OSNR measurement using this technique is shown in Figure 12.13(a). Moreover, the correlation between the intensity and phase of the sample in the constellation diagram can be used for estimation of the NPN affecting the phase-modulated signal. The correlation has been used as the parameter in estimating the NPN level, showing a linear relationship to the signal power before transmission, as demonstrated in Figure 12.13(b). The linear optical sampling technique also allows monitoring of

CHAPTER 12 OPM of optical phase–modulated signals

Variance

1

sx2 sy2 Theory

10–1

0.4 C (rad)

336

10–2

0.2 0.1

10–3 10–4

0.3

0 5

0

10 15 20 25 30 35 40 45 OSNR (dB) (a)

10

20 30 40 Power (mW) (b)

50

FIGURE 12.13 (a) Evaluation of OSNR using linear optical sampling. (b) Evaluation of NPN using linear optical sampling.86

multilevel phase-modulated signals, as was demonstrated in Reference 87 for a 43-Gb/s RZ-QPSK signal. Similar to the case of PSK signal monitoring, the statistical distribution of samples allows monitoring the OSNR level, while the correlation between the phase and amplitude values provides a parameter for evaluating the level of NPN. The ability to measure the performance of phase-modulated channels provided by the constellation diagrams has attracted research toward the development of techniques allowing simple acquisition of both intensity and phase information of the signal. One such example is the self-homodyne phasor monitor presented in Reference 88. The monitor consists of a delay interferometer and a 120 optical hybrid followed by three optical detectors, as shown in Figure 12.14(a). The output of the detectors is sampled and based on the three values, and the phase and intensity of the samples are calculated in the constellation diagram, as shown in Figure 12.14(b). The relative phase between the arms of the delay interferometer is not matched exactly to the operating wavelength; therefore, the obtained constellation is rotated by an angle d. The rotation is compensated numerically by calculating the differential phasor indicating the real axis of the constellation, as shown in Figure 12.14(c).88 The calculated values of phase and intensity can be represented, not only in a form of constellation diagram, but can also indicate the phasor trajectory, phase eye diagram, and demodulated output. This opens the possibility of measuring the imperfections in the signal driving the phase modulator,

P1(t)

Input 1 T 2×2 coupler

Phase delay

(a)

Output I1(t)

1

2

2 P2(t)

3

3

I2(t)

P3(t) I3(t) 3×3 coupler Detector

Output signal I1(t) I2(t) I3(t)

æ è

æ è

Y¢

Estimation of δ δ + rotation

Y

X¢ Intermediate phasor (X ¢, Y⬘)

X Differential phasor (X , Y )

(b)

FIGURE 12.14 (a) Diagram of self-homodyne phasor monitor. (b) Symbol values obtained from monitor. (c) Constellation diagram after rotation through phasor-estimated phase.88

12.3 Optical performance monitoring

337

including the insufficient modulation depth or temporal skew between the arms of the modulator. These impairments are illustrated in Figures 12.15(a) and (b), respectively. In the case of multilevel phase-modulated signals, the phasor diagram enables precise adjustment of bias voltages in stacked MZM. Finally, the impairment due to ASE noise can be observed in the phasor diagram by measuring the ratio of the amplitude noise to the signal level. This enabled the monitoring of OSNR in range from 13 to 25 dB. It ought to be mentioned that other examples of constellation diagram–based monitoring using self-homodyne detection have been presented in the literature. Monitoring of signal evolution and analysis of modulator adjustment have been demonstrated for both RZ-DQPSK89 and RZ-8-APSK90 modulations. These techniques use the modulated or unmodulated signal as a reference source and employ digital signal processing (DSP) for analysis and presentation of the acquired samples. Readers are referred to the respective articles for details of these techniques.

Dqmax = ±p

1.5 Quadrature, Y (a.u)

Quadrature, Y (a.u)

1.5 1 0.5 0 –0.5 –1

Δqmax = ± 0.5p

1 0.5 0 –0.5 –1 –1.5

–1.5 –1.5 –1–0.5 0 0.5 1 1.5

–1.5 –1–0.5 0 0.5 1 1.5

In phase, X (a.u.)

In phase, X (a.u.)

(a)

1.5 Quadrature, Y (a.u)

Quadrature, Y (a.u)

1.5 1 0.5 0 –0.5 –1 –1.5

1 0.5 0 –0.5 –1 –1.5

–1.5 –1–0.5 0 0.5 1 1.5 In phase, X (a.u.)

–1.5 –1–0.5 0 0.5 1 1.5 In phase, X (a.u.)

(b)

FIGURE 12.15 Constellation diagram obtained with differential phasor. (a) Phase modulation with driving voltage of VP and 1/2VP. (b) PSK modulation without and with phase mismatch between arms of MZM.88

338

CHAPTER 12 OPM of optical phase–modulated signals

12.3.1.7 Asynchronous amplitude histogram

m1 mm

Intensity

The monitoring techniques employing constellation diagrams provide ample information about the modulator characteristics and OSNR level affecting the signal. However, no methods employing the constellation diagrams to evaluate the influence of transmission impairments in phase-modulated channels have been presented to date. In order to obtain information about the level of CD or PMD affecting the signal, the samples need to be digitally processed. One of the methods to analyze the samples is to build a histogram of signal amplitude values. Such an approach is used to obtain a synchronous eye diagram, which can be used to extract information about signal noise, jitter, or waveform distortion. However, it requires a costly and format-dependent clock recovery circuit. A powerful OPM technique using asynchronous amplitude samples has been demonstrated for OOK signals.91 The asynchronous amplitude histogram (AAH) OPM technique takes advantage of the statistical properties of the amplitude histogram to extract information about signal OSNR, accumulated CD, and PMD also in phase-modulated signals.92–94 The AAH technique satisfies many of the requirements put on the OPM system. It is asynchronous and allows monitoring of multiple signal impairments. The histogram is created by first sorting the samples by their value. The values are mapped onto histogram bins that uniformly divide the dynamic range of sample values into n levels. The histogram is formed by counting the number of amplitude samples falling into each of the bins and plotting the count as a function of the bin value. An example of an amplitude histogram for a 10-Gb/s RZ-DPSK signal is shown in Figure 12.16. The horizontal axis of the histogram represents the number of sample points in respective bins, while the vertical axis corresponds to the sample value. The peaks of the histogram correspond to the valley and the peak of signal waveform, respectively. The samples in between the peaks correspond to the crossover points of the rising and falling edge. In order to analyze the shape of the histogram, a distribution is fitted to each of the peaks. Histogram parameters reflect the properties of signal waveform. Therefore, by tracking the statistical properties of the histogram, it is possible to evaluate the level of impairments affecting the optical signal.

20 ps / div N1

Nm

NOISE DISPERSION

m0 Time

FIGURE 12.16 Waveform of RZ-DPSK signal and corresponding asynchronous amplitude histogram. Figure also shows parameters used for monitoring signal impairments.92

339

Intensity

Intensity

12.3 Optical performance monitoring

m90 mavg

mavg savg 20 ps/div

20 ps/div Time (a)

Time (b)

FIGURE 12.17 Waveform and the corresponding AAH for the NRZ-DPSK signals. (a) Parameter for monitoring OSNR. (b) Parameters for monitoring of CD.92

OSNR monitoring using AAH relies on dependence of the variance of signal-spontaneous beat noise on the power of ASE noise. In the case of an NRZ-DPSK signal generated by a phase modulator, the waveform is a constant-amplitude signal, as illustrated in Figure 12.17(a). The amplitude histogram acquired from the waveform consists of a single peak with mean value mavg. The fluctuation of waveform amplitude due to signal-spontaneous beat noise can be observed through the standard deviation savg of histogram distribution, which is inversely proportional to the level of OSNR. With this technique, the OSNR was evaluated from 22 to 38 dB in a 10-Gb/s NRZ-DPSK signal.92 OSNR monitoring has also been demonstrated for RZ-DPSK and RZ-DQPSK formats. In this case, the histogram differs from that of the NRZ-DPSK waveform. It is comprised of two distribution peaks corresponding to the RZ pulse peak and the level of waveform valley between the pulses. OSNR reduction due to the change in ASE noise level causes a waveform fluctuation due to signal beating with ASE noise. Waveform fluctuation results in a reduced ratio between the histogram peak height and histogram height at median level, as shown in Figure 12.16. This phenomenon enables evaluation of the OSNR from 17 to 27 dB for the RZ-DPSK signal and from 17 to 30 dB for the RZ-DQPSK signal.92 In this range the parameter scales linearly with the OSNR. The AAH also serves to evaluate the level of residual CD affecting the signal. In the case of the NRZ-DPSK signal, the change of phase between bits introduces a frequency chirp proportional to the phase-modulator bandwidth. CD accumulation results in phase-to-amplitude conversion. This can be observed in the histogram that develops multiple peaks proportional to the conversion level. As CD does not influence the mean or standard deviation of the central histogram peak, it is possible to evaluate the accumulated CD level independently from the OSNR level. For a 10-Gb/s NRZ-DPSK, evaluation of CD is possible from 600 to þ600 ps/nm.92 The effect of CD accumulation in the RZ-DPSK or RZ-DQPSK signal can be observed in the waveform as spreading of pulses in time. This is reflected in the amplitude histogram, as shown in Figure 12.16. By measuring the distance between the peaks of the histogram, a parameter for evaluation of the accumulated CD can be established. CD monitoring for a 10-Gb/s RZ-DPSK signal and 20-Gb/s RZ-DQPSK signal was demonstrated from 600 to þ600 ps/nm.92 As the AAH reflects the statistical properties of waveform

340

CHAPTER 12 OPM of optical phase–modulated signals

amplitude, it is also sensitive to other transmission impairments, such as PMD. PMD monitoring has been demonstrated using the AAH for DGD up to 50 ps in the 10-Gb/s RZ-DPSK94 and 20-Gb/ s RZ-DQPSK92 signals. An alternative approach to monitoring the performance of phase-modulated signals using histogram analysis was presented in Reference 95. A self-clocked sampling module based on sumfrequency generation recovers a synchronized waveform of a demodulated channel. The samples forming the center of the obtained eye diagram are used for construction of the histogram. A Q-parameter is calculated from the statistical properties of the histogram and related to transmission impairments in 10-Gb/s RZ-DPSK and 40-Gb/s RZ-DQPSK signals.

12.3.1.8 Two-tap sampling

OBPF

PD

Power splitter

E(X) Sampling scope E(Y) Δt

Intensity

The asynchronous monitoring of transmission impairments in phase-modulated signals using amplitude histograms presented in the previous section reconstructs signal properties from individual intensity samples. This concept may be extended to the acquisition of sample pairs with a known relationship providing a reference in time or phase. A number of OPM techniques using sample pairs have been developed. The principle of the delay-tap sampling technique, first demonstrated in Reference 96, is shown schematically in Figure 12.18(a). The OPM device input is the optical signal tapped from the transmission fiber. An optical BPF (OBPF) selects a single WDM channel for processing. The optical signal is split into two branches. One branch contains an optical delay line that introduces a time shift Dt between the optical signals. Subsequently, the light in both branches is converted into the electrical domain by a pair of photodiodes. Finally, the signals are fed into a sampling oscilloscope operating in X-Y mode. The voltage sampled in respective branches serves as X (E(X)) and Y (E(Y)) parameters in further analysis. The oscilloscope is operated by an external clock, independent of the monitored signal. Simultaneous sampling of the electrical signal in two branches produces sample pairs. Effectively, this can be considered as sampling of the original waveform at two instants separated by the delay Dt. This is illustrated in Figure 12.18(b). The value of the delay shown in the figure is on the order of one-tenth of a symbol period. The exact choice of delay depends on which property of the signal is analyzed, as explained in the following paragraphs. The electrical signal samples at E(Y)

tX tY E(X )

E(Y) Clock Time

(a)

(b)

E(X )

(c)

FIGURE 12.18 (a) Diagram of the monitor based on delay-tap sampling. (b) Acquisition of two samples with relative delay Dt. (c) Construction of delay-tap plot.97

12.3 Optical performance monitoring

341

times tX and tY produce sample pairs with values E(X) and E(Y). The sample pairs are used to construct the delay-tap plot, as illustrated in Figure 12.18(c). Sample values serve as coordinates in the Cartesian system to plot individual points. The overlapping points from consecutive sampling instants form the delay-tap plot. This allows graphical decomposition of the waveform without synchronizing with the signal. The shape of the plot depends on the delay Dt and is influenced by changes in the signal waveform due to transmission impairments. Subsequent analysis of the delay-tap plot allows extraction of the information on the level of signal degradation. In order to monitor OSNR degradation in the RZ-DPSK or RZ-DQPSK signal, a short delay Dt of one-tenth of a symbol period is used. The short delay between samples E(X) and E(Y) forms a narrow delay-tap plot that allows separation of the distributions of waveform peaks and valleys. The frequency of occurrence of sample pairs along the plot diagonal forms a histogram from which a noise parameter is calculated.97 This parameter is similar but not equivalent to the synchronous Q-parameter due to the difference in statistical distribution of samples at the maximum eye opening captured by the synchronous sampling. Using the delay-tap plot analysis, the OSNR can be evaluated from 9 to 32 dB for the RZ-DPSK signal97 and from 7 to 31 dB for the RZ-DQPSK signal.98 Monitoring the influence of residual CD on RZ-DPSK and RZ-DQPSK signals using the delaytap plots relies on the waveform pulse carving and the relative phase of neighboring bits. The differential phase modulation conveys information in phase shifts between two consecutive symbols. For the RZ-DPSK signal, the possible phase shifts are 0 or p; for RZ-DQPSK, the possible phase shifts are 0, p/2, or p. When an optical pulse is distorted by residual CD, it spreads beyond the allocated bit slot. The peak power of the pulse is reduced. Moreover, the shape of the waveform in between pulses is the result of interference between two neighboring bits and depends on the phase shift between the bits, as the constructive or destructive interference occurs. In order to track this phenomenon with the DTS technique, a 1-bit delay Dt, is employed. This allows comparing the waveforms of two neighboring bits. An algorithm based on the Hough transform is used to estimate accumulated CD from 600 to þ600 ps/nm for both the 10-Gb/s RZ-DPSK97 and 20-Gb/s RZ-DQPSK signals.98 The delay-tap plot monitoring technique in conjunction with the Hough transform was also used to evaluate PMD for DGD of up to one-half a symbol period in the RZ-DQPSK signal. Alternatively, the analysis of the delay-tap plot using the Hausdorff distance for CD monitoring in NRZ-DPSK signals was presented in Reference 99. A different approach to delay-tap plot analysis was presented in Reference 100. The plots are formed using a fixed delay and the resulting images are analyzed using pattern recognition. This technique has been shown to be capable of evaluating impairment due to ASE noise, accumulated CD, and PMD. The two-dimensional plots for monitoring optical impairments can also be constructed by sampling the constructive and destructive ports of the DI used for demodulation of the DPSK signal.101 The two-tap plot is constructed by accumulating the samples with x values equal to the level of the signal in the destructive port and y values equal to the level of the signal in the constructive port, as shown in Figure 12.19. The analysis of the plot yields a Q-parameter that linearly increases with the conventional Q-parameter. The concept of two-tap sampling has been extended to monitoring both the amplitude and phase of the NRZ-DPSK signal.102 The field of the monitored channel is detected using a homodyne receiver. The signal is mixed with a local oscillator light in an optical hybrid and detected by two photodiodes. Two individual delay-tap plots are created for the phase and noise components, each with a delay of one-fifth of a symbol period. Analysis of the frequency histograms acquired along

342

CHAPTER 12 OPM of optical phase–modulated signals

Two-tap histogram

Count

100

50

0 phase change

p phase change

0 Co

ns

4 cti ve 3 po rt 2 vo lta 1 ge ,V 0 (⫻ –1 10 – 4 )

tru

–1

10

–4 )

r t vo

o ve p

cti estru

4

, V(⫻ ltage

2

1 0

3

D

FIGURE 12.19 Two-tap plot constructed by sampling the constructive and destructive ports of the demodulator DI.101

the diagonal of each delay-tap plot enables estimation of transmission impairments. The delay-tap plot constructed from the amplitude component allows estimation of the OSNR from 10 to 30 dB, while the plot constructed from the phase component allows evaluation of accumulated CD from 0 to 750 ps/nm. OSNR and CD monitoring in a single channel is performed without a tunable optical filter, as the channel is separated by the frequency of the local oscillator.

12.3.1.9 Receiver-based monitoring This section explains a group of OPM techniques that employ the optical receiver in order to analyze impairments in the phase-modulated signals. A technique for monitoring the phase offset of the DQPSK demodulator from the center frequency of the signal is demonstrated in Reference 103. The technique employs a demodulator based on a 1-bit delay interferometer followed by a 90 hybrid and a pair of balanced photodiodes, as shown in Figure 12.20. The signal after the PD is limited to a value equal to the level obtained at the optimum phase of the demodulator. When a phase mismatch occurs, leading to the rotation of the constellation diagram with respect to the demodulator axes, the quadrature component that exceeds the limit is cropped, while the other component’s power is reduced. Consequently, the average power of the monitor signal after the limiter drops while the phase mismatch between the modulator and the center frequency of the signal increases. Dithering of the phase offset in one of the arms of the demodulator allows reaching the optimum phase. Operation of this technique was proven experimentally for a 43-Gb/s RZ-DQPSK signal in the presence of accumulated CD of up to 250 ps/nm.

12.3 Optical performance monitoring

Demodulator 1-bit delay Dt Signal

Control

343

I

Phase shifter Df

Q

+ Limiter

RF power meter

Lock-in amplifier

Monitor signal

Monitor

FIGURE 12.20 Diagram of phase-offset monitoring technique based on optical receiver and limiting amplifier.103

Finally, analysis of the synchronous eye diagram for monitoring a 40-Gb/s signal was demonstrated in Reference 104. The eye diagram received after the demodulator was analyzed using a neural network. The algorithm was trained with a set of signals affected by simulated impairments that enabled monitoring of OSNR, CD, and PMD. The evaluation range for the OSNR impairment was 18–30 dB, for the accumulated CD impairment from 7.5 to 52.5 ps/nm, and for PMD from 1.25 to 8.75 ps. This technique was also verified experimentally for monitoring impairments due to OSNR and CD.

12.3.2 Comparison of monitoring techniques In order to realize OPM functionality, the presented techniques must fulfill a number of requirements. The desired characteristics of an OPM system are as follows: • • • • • • •

Transparency Asynchronous operation Multi-impairment monitoring Sensitivity Wide dynamic range Repeatability Affordable design

The concept of OPM transparency is in agreement with that of network transparency. This means flexibility in terms of signal bit rate and modulation format. On the other hand, transparency means that monitoring itself does not degrade the monitored signal. The requirement for asynchronous operation is one of the conditions for realizing transparent monitoring. The monitoring of multiple impairments is another important factor that should be satisfied by the monitoring equipment. A single technique should provide information about the degradations that are most likely to influence signal quality. The measurement or evaluation done through the OPM technique should be sensitive, have a wide dynamic range, and be repeatable. The dynamic range of the measurement defines the range of impairment values that the monitor is able to track, and it depends on the impairment type, signal bit rate, modulation format, and position of the device in the network.

344

CHAPTER 12 OPM of optical phase–modulated signals

Finally, repeatability or reliability of the measurement determines whether the parameter is uniformly evaluated when the impairment level remains constant. The last requirement of a successful OPM technique is affordable design. Although it is not directly related to the performance of the monitor, it plays an important role as the capital expenditures related to the deployment of the OPM equipment should be outweighed by the savings in operational expenditures. Therefore, affordable OPM techniques that provide a suitable range of measurement parameters are more likely to find their way into the realm of transparent networks. This section focuses on the comparison of OPM techniques for the phase-modulated signals from the perspective of general requirements for OPM equipment. The techniques presented in this chapter may be arranged by the type of impairment that they are capable of monitoring. Such a comparison is presented in Figure 12.21. The center part of the figure lists the signal impairments described in Section 12.2.1. Impairments are linked to the techniques presented in Sections 12.3.1.2–12.3.1.9 that are capable of monitoring the respective impairments. The left side of the figure lists the spectrum-based monitoring techniques, while the right side presents the sampling-based techniques. It can be observed that most of the focus is concentrated on monitoring OSNR, CD, and PMD. Monitoring of NPN, power, Q-factor, and frequency offset receives less attention in the literature. Moreover, it may be observed that the sampling-based techniques are more versatile than the spectrum-based techniques. It must, however, be remembered that these techniques require high-speed components, such as broadband detectors and high-speed sampling. The versatility of the monitoring techniques in terms of modulation format and impairment (including the evaluation range) are shown in Table 12.1. Measurement range values are recalculated for the 10-Gsymbol/s signals. Spectrum-based monitoring techniques provide wider ranges for monitoring impairments. Therefore, it may be concluded that the sampling-based techniques are more versatile, while requiring high-speed components, whereas the spectrum-based techniques should be chosen when the monitoring range is the priority. Spectrum-based techniques

Signal impairments OSNR

Spectral monitor Polarization nulling/DOP RF spectrum/clock tone Pilot tone

CD PMD NPN Power Q-factor Frequency offset

FIGURE 12.21 Analysis of OPM techniques by type of monitored impairment.

Sampling-based techniques Linear optical sampling/ optical field detection Asynchronous histogram Delay tap Receiver/neural network

12.4 Summary

345

Table 12.1 Comparison of Measurement Ranges between OPM Techniques for Phase-Modulated Signals Technique (references) Optical spectrum DOP70 DOP71 DOP72

64

RF spectrum76 RF spectrum73 RF spectrum74,75

XPM77 Clock tone71,78

Clock tone72 Pilot tone79,83 AAH92,93,94

DTS99 DTS97,98,99

Format

Impairment

Range (for 10-Gsymbol/s signal)

NRZ-DPSK

OSNR OSNR PMD PMD CD OSNR OSNR

0–28 (dB) 12–25 (dB) 0–100 (ps) 0–100 (ps) 0–50 (ps) 0–800 (ps/nm) 15–35 (dB) 11–29 (dB)

PMD

0–48 (ps)

CD CD PMD CD CD Offset CD PMD OSNR

0–1920 (ps/nm) 0–600 (ps/nm) 0–50 (ps) 0–900 (ps/nm) 0–720 (ps/nm) 02 (0–883) (GHz) 0–600 (ps/nm) 14–38 (dB) 11–32 (dB) 17–35 (dB) 0–400 (ps/nm) 0–600 (ps/nm) 0–50 (ps) 9–32 (dB) 0–850 (ps/nm) 12–19 (dB) 10–30 (dB) 0–800 (ps/nm) 0–1.35 (GHz) 0–640 (ps/nm) 0–36 (ps) 18–30 (dB)

NRZ/RZ-DPSK NRZ-DQPSK RZ-DQPSK NRZ-DPSK RZ-DPSK Pol-Mux RZ-DPSK Pol-Mux RZ-DPSK RZ-D(Q)PSK NRZ-DPSK RZ-DPSK NRZ/RZ-DQPSK NRZ-DPSK NRZ-DPSK RZ-DPSK RZ-DQPSK NRZ-DPSK

RZ-D(Q)PSK DTS100 DTS101 Delay-tap optical field102

NRZ-DPSK NRZ-DPSK

Receiver103 Neural network104

RZ-DQPSK NRZ-DPSK

CD CD PMD OSNR CD Q-factor OSNR CD Offset CD PMD OSNR

12.4 SUMMARY This chapter discussed the most recent research in the OPM field for phase-modulated channels. Both the spectrum- and sampling-based techniques developed for the intensity-modulated channels have been adapted to accommodate the phase-modulated signals. In addition, new techniques

346

CHAPTER 12 OPM of optical phase–modulated signals

focusing on signal phase analysis have been introduced. By allowing the observation of transmission impairments directly in the optical layer, these techniques bring optical communication systems one step closer to transparent operation. It must, however, be remembered that the constant drive toward higher capacity leads to the introduction of more sophisticated modulation formats, covering more dimensions of modulation space, more levels of modulation, and higher symbol rates. These formats pose new challenges for OPM techniques as system tolerances become tighter and more sensitive parameters have to be employed. Therefore, it may be expected that OPM for phase-modulated signals will keep developing together with advances in transmission techniques.

ACKNOWLEDGMENTS The author would like to thank Hidehiko Takara of NTT Network Innovation Laboratories, as well as Ken-ichi Kitayama and Akihiro Maruta of Osaka University for valuable discussions.

REFERENCES 1. O’Mahony MJ, Politi C, Klonidis D, Nejabati R, Simeonidou D. Future optical networks. J Lightwave Technol 2006;24:4684–96. 2. Gladisch A, Braun RP, Breuer D, Ehrhardt A, Foisel HM, Jaeger M, et al. Evolution of terrestrial optical system and core network architecture. Proc. IEEE 2006;94:869–91. 3. Winzer P, Essiambre RJ. Advanced optical modulation formats. Proc. IEEE 2006;94:952–85. 4. Miyamoto Y. Advanced modulation formats for high-capacity optical transport network. In: Proc. 31st European conference on optical communication (ECOC) 2005, Glasgow, Scotland, vol. 4, p. 777–80; 2005. 5. Mizuochi T. Recent progress in forward error correction and its interplay with transmission impairments. IEEE J Sel Top Quantum Electron 2006;12:544–54. 6. Xu C, Liu X, Wei X. Differential phase-shift keying for high spectral efficiency optical transmissions. IEEE J Sel Top Quantum Electron 2004;10:281–93. 7. Charlet G. Progress in optical modulation formats for high-bit rate WDM transmissions. IEEE J Sel Top Quantum Electron 2006;12:469–83. 8. Gnauck A, Winzer P. Optical phase-shift-keyed transmission. J Lightwave Technol 2005;23:115–30. 9. Xie C, Moller L, Haunstein H, Hunsche S. Comparison of system tolerance to polarization-mode dispersion between different modulation formats. IEEE Photon Technol Lett 2003;15:1168–70. 10. Vassilieva O, Hoshida T, Choudhary S, Castanon G, Kuwahara H, Terahara T, et al. Numerical comparison of NRZ, CS-RZ and IM-DPSK formats in 43 Gbit/s WDM transmission. In: Proc. LEOS 2001, San Diego, CA, vol. 2, p. 673–74; 2001. 11. Miyano T, Fukutoku M, Hattori K, Ono H. Suppression of degradation induced by SPM/XPM þ GVD in WDM transmission using a bit-synchronous intensity modulated DPSK signal. In: Proc. OECC 2000, Chiba, Japan, p. 580–81; 2000. 12. Cho PS, Khurgin JB. Suppression of cross-gain modulation in SOA using RZ-DPSK modulation format. IEEE Photon Technol Lett 2003;15:162–4. 13. Hirano A, Miyamoto Y. Novel modulation formats in ultra-high-speed transmission systems, and their applications. In: Proc. OFC 2004, Los Angeles, CA, paper ThM1. 2004. 14. Zhu B, Nelson LE, Stulz S, Gnauck AH, Doerr C, Leuthold J, et al. High spectral density long-haul 40-Gb/s transmission using CSRZ-DPSK format. J Lightwave Technol 2004;22:208–14.

References

347

15. Zhu B, Nelson LE, Stulz S, Gnauck AH, Doerr C, Leuthold J, et al. 6.4 Tb/s (160  42.7 Gb/s) transmission with 0.8 bit/s/Hz spectral efficiency over 32  100 km of fiber using CSRZ-DPSK format. In: Proc. OFC 2003, Atlanta, GA, paper PD19. 2003. 16. Wree C, Leibrich J, Eick J, Rosenkranz W, Mohr D. Experimental investigation of receiver sensitivity of RZ-DQPSK modulation using balanced detection. In: Proc. OFC 2003, Atlanta, GA, paper ThE5. 2003. 17. Ohm M, Speidel J. Quaternary optical ASK-DPSK and receivers with direct detection. IEEE Photon Technol Lett 2003;15:159–61. 18. Yadin Y, Orenstein M, Shtaif M. Optical DPASK and DQPSK: a comparative analysis for linear and nonlinear transmission. IEEE J Sel Top Quantum Electron 2006;12:581–8. 19. Winzer PJ, Essiambre R-J. Advanced optical modulation formats. In: Kaminow IP, Li T, Willner AE, editors. Optical fiber telecommunications V B: Systems and Networks. New York: Academic Press; 2008. 20. Sekine K, Kikuchi N, Mandai K, Sasaki S. Advanced multi-level transmission systems. In: Proc. OFC 2008, San Diego, CA, paper OMI4. 2008. 21. Serbay M, Wree C, Rosenkranz W. Experimental investigation of RZ-8DPSK at 310.7 Gb/s. In: Proc. LEOS 2005, Sydney, Australia, p. 483–84; 2005. 22. Nazarathy M, Simony E. Performance limits of multilevel DPSK. IEEE Photon Technol Lett 2005;17:2310–2. 23. Agrawal GP. Fiber-optic communication systems. Wiley series in microwave and optical engineering, 3rd ed. New York: John Wiley & Sons, Inc.; 2002. 24. Winzer P, Chandrasekhar S, Kim H. Impact of filtering on RZ-DPSK reception. IEEE Photon Technol Lett 2003;15:840–2. 25. Bosco G, Poggiolini P. On the Q factor inaccuracy in the performance analysis of optical direct-detection DPSK systems. IEEE Photon Technol Lett 2004;16:665–7. 26. Boffi P, Marazzi L, Paradiso L, Parolari P, Righetti A, Siano R, et al. Experimental comparison of RZ-IMDD and RZ-DQPSK performance in a standard 2000-km DWDM system. In: Proc. CLEO 2004, San Francisco, CA, paper CThBB5. 2004. 27. Gnauck A, Winzer P, Dorrer C, Chandrasekhar S. Linear and nonlinear performance of 42.7-Gb/s singlepolarization RZ-DQPSK format. IEEE Photon Technol Lett 2006;18:883–5. 28. Sinsky J, Adamiecki A, Gnauck A, Burrus C, Leuthold J, Wohlgemuth O, et al. RZ-DPSK transmission using a 42.7-Gb/s integrated balanced optical front end with record sensitivity. J Lightwave Technol 2004;22:180–5. 29. Willner A, Nezam S, Yan L, Pan Z, Hauer M. Monitoring and control of polarization-related impairments in optical fiber systems. J Lightwave Technol 2004;22:106–25. 30. Kim H, Doerr C, Pafchek R, Stulz L, Bernasconi P. Polarisation-mode dispersion impairments in directdetection differential phase-shift-keying systems. Electron Lett 2002;38:1047–8. 31. Wang J, Kahn JM. Impact of chromatic and polarization-mode dispersions on DPSK systems using interferometric demodulation and direct detection. J Lightwave Technol 2004;22:362–71. 32. Chandrasekhar S, Liu X. Experimental investigation of system impairments in polarization multiplexed 107-Gb/s RZ-DQPSK. In: Proc. OFC 2008, San Diego, CA, paper OThU7. 2008. 33. Kim H, Winzer P. Robustness to laser frequency offset in direct-detection DPSK and DQPSK systems. J Lightwave Technol 2003;21:1887–91. 34. Gene J, Soler M, Killey R, Prat J. Investigation of 10-Gb/s optical DQPSK systems in presence of chromatic dispersion, fiber nonlinearities, and phase noise. IEEE Photon Technol Lett 2004;16:924–6. 35. Freund R, Seimetz M. Higher order modulation formats using coherent detection and electronic distortion equalisation for application in future backbone networks. In: Proc. IEEE/LEOS 2008, Newport Beach, CA, paper WD3.3. 2008. 36. Gnauck A. 40-Gb/s RZ-differential phase shift keyed transmission. In: Proc. OFC 2003, Atlanta, GA, p. 450–51; 2003.

348

CHAPTER 12 OPM of optical phase–modulated signals

37. Wree C, Leibrich J, Rosenkranz W. RZ-DQPSK format with high spectral efficiency and high robustness towards fiber nonlinearities. In: Proc. ECOC 2002, Copenhagen, Denmark, paper 9.6.6. 2002. 38. Gordon P, Mollenauer LF. Phase noise in photonic communications systems using linear amplifiers. Optics Lett 1990;15:1351–3. 39. Xu C, Liu X, Mollenauer LF, Wei X. Comparison of return-to-zero differential phase-shift keying and ONOFF keying in long-haul dispersion managed transmission. IEEE Photon Technol Lett 2003;15:617–9. 40. Kim H. Cross-phase-modulation-induced nonlinear phase noise in WDM direct-detection DPSK systems. J Lightwave Technol 2003;21:1770–4. 41. Kim H, Gnauck A. Experimental investigation of the performance limitation of DPSK systems due to nonlinear phase noise. IEEE Photon Technol Lett 2003;15:320–2. 42. Ho KP. Performance degradation of phase-modulated systems due to nonlinear phase noise. IEEE Photon Technol Lett 2003;15:1213–5. 43. Liu X. Nonlinear effects in phase shift keyed transmission. In: Proc. OFC 2004, Los Angeles, CA, paper ThM4. 2004. 44. Boffi P, Marazzi L, Paradiso L, Parolari P, Righetti A, Setti D, et al. 20 Gb/s differential quadrature phaseshift keying transmission over 2000 km in a 64-channel WDM system. Opt Commun 2004;237:319–23. 45. Wen YJ, Nirmalathas A, Lee DS. RZ/CSRZ-DPSK and chirped NRZ signal generation using a single-stage dual-electrode Mach-Zehnder modulator. IEEE Photon Technol Lett 2004;16:1041–135. 46. Ohm M, Freckmann T. Comparison of different DQPSK transmitters with NRZ and RZ impulse shaping. In: Proc. IEEE/LEOS workshop on advanced modulation formats, Rio Grande, Puerto Rico, paper ThB2. ; 2004. 47. Ho KP, Cuei HW. Generation of arbitrary quadrature signals using one dual-drive modulator. J Lightwave Technol 2005;23:764–70. 48. Griffin RA, Carter AC. Optical differential quadrature phase-shift key (oDQPSK) for high capacity optical transmission. In: Proc. OFC 2002, Anaheim, CA, paper WX6. 2002. 49. Bosco G, Poggiolini P. Analysis of impact of receiver imperfections on performance of optical DQPSK systems. Electron Lett 2004;40:1147–9. 50. Bosco G, Poggiolini P. On the joint effect of receiver impairments on direct-detection DQPSK systems. J Lightwave Technol 2006;24:1323–33. 51. Kilper DC, Bach R, Blumenthal D, Einstein D, Landolsi T, Ostar L, et al. Optical performance monitoring. J Lightwave Technol 2004;22:294–304. 52. Weingartner W, Kilper DC. OSNR monitoring for fault management in high speed networks. In: Proc. ECOC 2002, Copenhagen, Denmark, paper 7.4.5. 2002. 53. Kilper DC, Weingartner W. Monitoring optical network performance degradation due to amplifier noise. J Lightwave Technol 2003;21:1171–8. 54. Willner AE, Pan Z, Yu C. Optical performance monitoring. In: Kaminow IP, Li T, Willner AE, editors. Optical fiber telecommunications V B: systems and networks. New York: Academic Press; 2008. 55. Xu C, Liu X. Postnonlinearity compensation with data-driven phase modulators in phase-shift keying transmission. Electron Lett 2002;27:1619–21. 56. Hansryd J, van Howe J, Xu C. Experimental demonstration of nonlinear phase jitter compensation in DPSK modulated fiber links. IEEE Photon Technol Lett 2005;17:232–4. 57. Zhu G, Mollenauer L, Xu C. Experimental demonstration of postnonlinearity compensation in a multispan DPSK transmission. IEEE Photon Technol Lett 2006;18:1007–9. 58. Liu X, Wei X, Slusher RE, McKinstrie CJ. Improving transmission performance in differential phaseshift-keyed systems by use of lumped nonlinear phase-shift compensation. Optics Lett 2002;27:1616–8. 59. Matsumoto M, Kamio T. Nonlinear phase noise reduction of DQPSK signals by a phase-preserving amplitude limiter using four-wave mixing in fiber. IEEE J Sel Top Quantum Electron 2008;14:610–5. 60. Ho KP, Kahn JM. Electronic compensation technique to mitigate nonlinear phase noise. J Lightwave Technol 2004;22:779–83.

References

349

61. Kikuchi K. Electronic post-compensation for nonlinear phase fluctuations in a 1000-km 20-Gbit/s optical quadrature phase-shift keying transmission system using the digital coherent receiver. Opt Express 2008;16:889–96. 62. ITU-T. Optical monitoring for DWDM systems. ITU-T G.697. 2004. 63. Kilper DC, Chandrasekhar S, Buhl L, Agarwal A, Maywar D. Spectral monitoring of OSNR in high-speed networks. In: Proc. ECOC 2002, Copenhagen, Denmark, paper 7.4.4. 2002. 64. Jeon SW, Kim YH, Lee BH, Jung MA, Park CS. OSNR monitoring technique based on cascaded long-period fiber grating with optically tunable phase shifter. Opt Express 2008;16:20603–9. 65. Liu X, Kao YH, Chandrasekhar S, Kang I, Cabot S, Buhl LL. OSNR monitoring method for OOK and DPSK based on optical delay interferometer. IEEE Photon Technol Lett 2007;19:1172–4. 66. Lee J, Jung D, Kim C, Chung Y. OSNR monitoring technique using polarization-nulling method. IEEE Photon Technol Lett 2004;13:88–90. 67. Lee JH, Choi HY, Shin SK, Chung YC. A review of the polarization-nulling technique for monitoring optical-signal-to-noise ratio in dynamic WDM networks. J Lightwave Technol 2006;24:4162–71. 68. Moench W, Larikova J. Measuring the optical signal-to-noise ratio in agile optical networks. In: Proc. OFC 2007, Anaheim, CA, paper NWC1. 2007. 69. Skold M, Olsson B-E, Sunnerud H, Karlsson M. PMD-insensitive DOP-based OSNR monitoring by spectral SOP measurements. In: Proc. OFC 2005, Anaheim, CA, paper NWC1. 2007. 70. Lu GW, Cheung MH, Chen LK, Chan CK. Simple PMD-insensitive OSNR monitoring scheme assisted by transmitter-side polarization scrambling. Opt Express 2006;14:58–62. 71. Pan Z, Wang Y, Song Y, Motaghian R, Havstad S, Willner A. Monitoring chromatic dispersion and PMD impairments in optical differential phase-shift-keyed (DPSK) systems. In: Proc. OFC 2003, Atlanta GA, p. 402–3; 2003. 72. Wang Y, Hu S, Yan L, Yang JY, Willner AE. Chromatic dispersion and polarization mode dispersion monitoring for multi-level intensity and phase modulation systems. Opt Express 2007;15:14038–43. 73. Dorrer C, Liu X. Noise monitoring of optical signals using RF spectrum analysis and its application to phase-shift-keyed signals. IEEE Photon Technol Lett 2004;16:1781–3. 74. Yang JY, Zhang L, Yue Y, Arbab VR, Agarwal A, Paraschis L, et al. Optical signal-to-noise ratio monitoring of an 80 Gbits/s polarization-multiplexed return-to-zero differential phase-shift keying channel. Optics Lett 2009;34:1006–8. 75. Yang J-Y, Zhang L, Yue Y, Christen LC, Zhang B, Jackel J, et al. CD-insensitive PMD monitoring of an 80-Gb/s polarization-multiplexed RZ-DPSK channel using a polarizer and a low-speed detector. In: Proc. OFC 2009, San Diego, CA, paper OThJ2. 2009. 76. Zhao J, Qureshi KK, Li ZH, Lu C, Tam HY, Wai PKA. Chromatic dispersion monitoring of DPSK signals using RF power detection. In: Proc. SPIE 7136, paper 71361H. 2008. 77. Yang J-Y, Zhang L, Wu T, Wu X, Christen LC, Nuccio S, et al. Chromatic dispersion monitoring of 40-Gb/s RZ-DPSK and 80-Gb/s RZ-DQPSK data using cross-phase modulation in highly-nonlinear fiber and a simple power monitor. In: Proc. OFC 2008, San Diego, CA, paper OTuG5. 2008. 78. Lize YK, Christen L, Yang JY, Saghari P, Nuccio S, Willner AE, et al. Independent and simultaneous monitoring of chromatic and polarization-mode dispersion in OOK and DPSK transmission. IEEE Photon Technol Lett 2007;19:3–5. 79. Ji HC, Park KJ, Lee JH, Chung HS, Son ES, Han KH, et al. Optical performance monitoring techniques based on pilot tones for WDM network applications. J Opt Netw 2004;3:510–33. 80. Park KJ, Ji HC, Chung YC. Optical channel monitoring technique using phase-modulated pilot tones. IEEE Photon Technol Lett 2005;17:2481–3. 81. Jun S, Kim H, Park P, Lee J, Chung Y. Pilot-tone-based WDM monitoring technique for DPSK systems. IEEE Photon Technol Lett 2006;18:2171–3. 82. Ji HC, Park PKJ, Kim H, Lee JH, Chung YC. A novel frequency-offset monitoring technique for directdetection DPSK systems. IEEE Photon Technol Lett 2006;18:950–2.

350

CHAPTER 12 OPM of optical phase–modulated signals

83. Christen L, Nuccio S, Lize YK, Jayachandran N, Willner AE, Paraschis L. Stabilization of a 40 Gb/s DPSK delay-line interferometer using half bit-rate AM pilot tone monitoring. In: Proc. CLEO 2007, Baltimore, MD, paper CMJJ2. 2007. 84. Andrekson P. High resolution optical waveform and eye diagram monitoring. In: Proc. OFC 2007, Anaheim, CA, paper OThN3. 2007. 85. Dorrer C, Doerr CR, Kang I, Ryf R, Leuthold J, Winzer PJ. Measurement of eye diagrams and constellation diagrams of optical sources using linear optics and waveguide technology. J Lightwave Technol 2005;23:178–86. 86. Dorrer C. Monitoring of optical signals from constellation diagrams measured with linear optical sampling. J Lightwave Technol 2006;24:313–21. 87. Dorrer C, Gnauck AH, Winzer PJ, Chandrasekhar S. Investigation of 42.7-Gb/s quadrature phase-shift keying (QPSK) signals using linear optical sampling. In: Proc. CLEO 2005, Baltimore, MD, paper CMHH4. 2005. 88. Takushima Y, Choi HY, Chung YC. Monitoring techniques for phase and OSNR of DPSK/DQPSK signals. In: Proc. SPIE 2008, paper 71360U. 2008. 89. Tanimura K, Ohta H. Monitoring of DPSK/DQPSK signals using 1-bit delayed self-homodyne detection with optical phase diversity. In: Proc. ECOC 2007, Berlin, Germany, paper P065. 2007. 90. Kikuchi N, Sekine K, Sasaki S. Time-resolved waveform measurement of high-speed phase-modulated optical signals using self-homodyne interferometry. In: Proc. ECOC 2005, Glasgow, Scotland, p. 391–92; 2005. 91. Shake I, Takara H. Averaged Q-factor method using amplitude histogram evaluation for transparent monitoring of optical signal-to-noise ratio degradation in optical transmission system. J Lightwave Technol 2002;20:1367–73. 92. Kozicki B, Ohara T, Takara H. Optical performance monitoring of phase-modulated signals using asynchronous amplitude histogram analysis. J Lightwave Technol 2008;26:1353–61. 93. Li Z, Lu C, Wang Y, Li G. In-service signal quality monitoring and multi-impairment discrimination based on asynchronous amplitude histogram evaluation for NRZ-DPSK systems. IEEE Photon Technol Lett 2005;17:1998–2000. 94. Li Z, Li G. Chromatic dispersion and polarization-mode dispersion monitoring for RZ-DPSK signals based on asynchronous amplitude-histogram evaluation. J Lightwave Technol 2006;24:2859–66. 95. Yoshikane N, Tsuritani T, Lee JH, Otani T. All-optical performance monitoring of multi-bitrate and multiformat signals based on optical sampling for transparent optical networks. In: Proc. SPIE 2008, paper 713637. 2008. 96. Anderson TB, Dods SD, Wong E, Farrell PM. Asynchronous measurement of chromatic dispersion from waveform distortion. In: Proc. OFC 2006, Anaheim, CA, paper OWN4. 2006. 97. Kozicki B, Maruta A, Kitayama K. Experimental demonstration of optical performance monitoring for RZ-DPSK signals using delay-tap sampling method. Opt Express 2008;16:3566–76. 98. Kozicki B, Maruta A, Kitayama KI. Experimental investigation of delay-tap sampling technique for online monitoring of RZ-DQPSK signals. IEEE Photon Technol Lett 2009;21:179–81. 99. Zhao J, Lu C, Lam KM, Li ZH, Tam HY, Wai PK. A novel optical signal monitoring method of DPSK signal based on delay tap sampling and Hausdorff distance measure. In: Proc. CLEO/QELS 2008, San Jose, CA, paper JWA108. 2008. 100. Clarke K, Anderson T, Dods S. Monitoring of multiple modulation formats using asynchronous delay-tap sampling. In: Proc. COIN-ACOFT 2007, Melbourne, Australia, paper MoA1–2. 2007. 101. Jong KC, Tsao HW, Lee SL. Q-factor monitoring of optical signal-to-noise ratio degradation in optical DPSK transmission. Electron Lett 2008;44:761–3. 102. Choi HY, Takushima Y, Chung YC. Multiple-impairment monitoring technique using optical field detection and asynchronous delay-tap sampling method. In: Proc. OFC 2009, San Diego, CA, paper OThJ5. 2009. 103. Kawakami H, Yoshida E, Miyamoto Y, Oguma M, Itoh T. Simple phase offset monitoring technique for 43 Gbit/s optical DQPSK receiver. Electron Lett 2008;44:437–8. 104. Wu X, Jargon J, Willner AE. Off-line monitoring of OSNR/CD/PMD degradation effects using neuralnetwork-based training sequences. In: Proc. ECOC 2008, Brussels, Belgium, paper We 3.D.6. ; 2008.

CHAPTER

Optical performance monitoring for coherent optical systems

13

Yan Tang*, Xingwen Yi{, William Shieh* *Department of Electrical and Electronic Engineering, The University of Melbourne, Victoria, Australia { University of Electronic Science and Technology of China, Chengdu, China

13.1 HISTORICAL ASPECT OF COHERENT OPTICAL SYSTEMS Most commercial optical communication systems use intensity modulation/direct detection (IM/DD) schemes in which the intensity of semiconductor lasers is modulated to carry the information and the optical signal is detected directly by a photodiode. In contrast to the IM/DD scheme, a coherent optical communication system detects the transmitted signal using homodyne or heterodyne detection schemes. It not only transmits information by modulating the intensity of the optical carrier, but also the phase or the polarization. Coherent optical communication systems were extensively studied in 1980s.1–3 Compared with direct detection, coherent detection offers the following advantages: 1. Improved receiver sensitivity. With sufficient local oscillator (LO) power, the shot-noise-limited receiver sensitivity can be achieved using coherent detection. 2. Improved spectral efficiency enables high-capacity transmissions, which is particularly attractive for high-speed transmission systems. However, with the invention of erbium-doped fiber amplifiers (EDFAs), direct detection systems could achieve receiver sensitivity within a few decibels of coherent receivers, which made the shot-noise-limited receiver sensitivity of the coherent receiver less attractive. In addition, the technical difficulties of coherent detection make it less practical. Coherent detection requires sophisticated manipulation and processing of phase and polarization. Since the state of polarization (SOP) of the incoming optical signal is scrambled in the fiber, a dynamic polarization controller is needed to match the SOP of the signal and LO. The dynamic polarization controller is usually a bulky and expensive device.4 The difficulty in stable locking of the carrier phase drift also prevents practical application of the coherent detection. Consequently, further research of coherent optical communications had been almost abandoned for nearly 10 years. Coherent detection has resurged to attract great interest in recent years, which is highlighted by remarkable theoretical and experimental demonstrations from various groups around the world.5–9 The drive behind using coherent communication techniques nowadays is twofold. First, current coherent detection systems are heavily entrenched in silicon-based digital signal processing (DSP). By taking advantage of high-speed DSP, both polarization and phase management can be easily © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00013-4

351

352

CHAPTER 13 OPM for coherent optical systems

realized, and thus a free running laser can be used as a local oscillator. Optical coherent detection in conjunction with high-speed DSP has the potential to increase spectral efficiency and the ability to compensate linear transmission impairments such as chromatic dispersion (CD) and polarization mode dispersion (PMD) in the electrical domain.10–12 Second, in contrast to the optical network system that was dominated by a low-speed, point-to-point, and single-channel system a decade ago, modern optical communication systems have advanced to massive wavelength-division multiplexing (WDM) and reconfigurable optical networks with a transmission speed per channel approaching 100 Gb/s. The primary aim of coherent communications has shifted toward supporting these high-speed dynamic networks by simplifying network installation, monitoring, and maintenance. Besides, orthogonal frequency-division multiplexing (OFDM), which has emerged and thrived in the radio frequency (RF) domain during the past decade, has gradually encroached into the optical domain. Compared to the single-carrier coherent detection system, the combination of OFDM and coherent detection not only realizes a robust dispersion transmission, but also brings the benefits of computation efficiency and easiness of channel and phase estimation. As optical fiber communication systems evolve toward more advanced all-optical fiber networks, optical performance monitoring (OPM) has become increasingly important. Coherent communication brings OPM the following advantages: 1. With rapid advances in high-speed DSP and receiver electronic equalization, the optical channel response can be accurately monitored with coherent detection. 2. High-frequency resolution enables the precise identification of optical power, optical signal-tonoise ratio (OSNR), data rate, and modulation format of the signal in the optical domain, which cannot be achieved in conventional IM/DD systems. 3. Coherent detection linearly converts the optical field into the RF field. Since RF technology is more mature and accessible than optical technology, extending signal processing from the optical domain into the RF domain greatly enhances the system’s functionality and flexibility.22 In this chapter we review recent developments on OPM techniques based on coherent detection, especially in multicarrier optical coherent detection systems. We first introduce the basic theory of optical coherent detection for both single-carrier and multicarrier systems in Section 13.2. Section 13.3 reviews various coherent detection–based OPM techniques that are categorized according to whether OPM is accomplished with or without receiver electrical equalization. In Section 13.4 we introduce the concept and theory of optical channel estimation (OCE), namely, OPM is achieved via extraction of various optical parameters from the estimated OCE. Section 13.5 reports the recent progress in OPM for coherent optical OFDM (CO-OFDM) systems in terms of simulation and experimental demonstration.

13.2 SINGLE-CARRIER AND MULTICARRIER COHERENT OPTICAL SYSTEMS Conventional single-carrier systems use one frequency to carry all the data. Multicarrier systems use multicarrier modulation (MCM) schemes by which the transmitted data stream is divided into several parallel lower–bit rate subcarriers. OFDM is a special form of MCM in the sense that the subcarriers are partially overlapped in the frequency domain, but yet orthogonal to each other. In the context of optical communications, CO-OFDM is a multicarrier optical system that combines OFDM

13.2 Single-carrier and multicarrier coherent optical systems

353

techniques and optical coherent detection. We present the principle of coherent detection in Section 13.2.1, and then give a brief introduction to single-carrier systems in Section 13.2.2. Section 13.2.3 focuses on CO-OFDM systems in terms of architecture and signal processing. Finally, Section 13.2.4 compares single-carrier and multicarrier CO-OFDM systems.

13.2.1 Principle of coherent detection The basic idea behind coherent detection is to mix the received signal with another continuous lightwave (CW) emitted from the LO before feeding into the photodetector. The optical field of the received optical signal ES (only single polarization is considered) can be expressed as Es ðtÞ ¼ As  expð jos t þ jfs Þ;

(13.1)

where As is the complex amplitude, and os and fs are, respectively, the angular frequency and the phase of the input optical signal. The optical field of the LO can be expressed as ELO ¼ ALO  expð joLO t þ jfLO Þ;

(13.2)

where ALO, oLO, and fLO are the complex amplitude, angular frequency, and phase of the LO, respectively. For coherent detection, a balanced receiver is generally used to reject the common mode component, such as suppressing the DC component and minimizing the laser relative intensity noise (RIN). Figure 13.1 shows the block diagram of coherent detection with a balanced receiver. The optical signal and the LO are mixed with a 3-dB coupler that adds a 180 phase shift to either the signal or the LO field and splits into two equal parts that are detected by two photodetectors. When the signal and LO share the same polarization, the output of the balanced receiver can be given as pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi (13.3) IðtÞ ¼ Iþ ðtÞ  I ðtÞ ¼ 2R Ps ðtÞPLO cosðoIF t þ fs ðtÞ  fLO ðtÞÞ; where oIF ¼ os  oLO; Ps and PLO are, respectively, the power of the signal and LO; fLO(t) is the phase of LO; and R is the responsivity of the photodetector. Dependent on whether the LO frequency, oLO, is set close to the signal frequency, the coherent receiver can be divided into homodyne receiver and heterodyne receiver. The OSNR sensitivity of Balanced receiver

ES

3-dB coupler

E1

I+(t)

E2

I–(t)

I(t) ELO

Polarization controller

FIGURE 13.1 Configuration of coherent receiver with balanced detector.

Laser

Photodetector

354

CHAPTER 13 OPM for coherent optical systems

the homodyne receiver is the same as the heterodyne counterpart. The homodyne receiver requires an optical hybrid to recover in-phase (I) and quadrature (Q) components of the signal, and thus needs twice as many optical components. But the heterodyne receiver needs an image rejection filter and twice as much photodetector bandwidth. Both homodyne and heterodyne receivers need a polarization controller to align the SOPs of LO and signal, or preferably, a polarization diversity detector,8 which will be discussed below. Without a polarization controller, because of the random changes resulted from the birefringence of the fiber, the polarization of the incoming signal is usually misaligned to the SOP of the LO. Thus, the receiver sensitivity of coherent detection is highly dependent on the SOP of the incoming signal in the absence of a polarization controller. A coherent receiver with polarization-diverse architecture can be employed to solve this problem. The schematic of a phase- and polarization-diverse receiver is shown in Figure 13.2. The incoming signal with arbitrary SOP is divided into two orthogonal linear polarization components with a polarization beam splitter (PBS) and fed into two homodyne I/Q receivers. A DSP module performs the functionalities such as analog-to-digital conversion, polarization alignment, and so on. Thus, coherent detection allows us to obtain full information on the optical signal—that is, signal amplitude, signal phase, and signal SOP.

13.2.2 Single-carrier coherent optical systems Figure 13.3(a) shows the communication system architecture of a single-carrier coherent optical system. At the transmitter end, the single-carrier system employs a relatively “conventional” and simpler architecture, where discrete digital modulation is fed into the two arms of the QPSK (optical I/Q) I/Q receiver I IIx

Esx PBS 90⬚ optical hybrid

Es ELOx

IQx

D S P IIy PBS

Esy 90⬚ optical hybrid

ELO ELOy

I/Q receiver II

FIGURE 13.2 Configuration of phase and polarization-diversity receiver.

IQy

Signals from phase and polarization diverse receiver

I

Clock extract and retiming

p/2 Phase and polarization diversity receiver

Laser

Normalization and orthogonalization DSP

Q QPSK modulator

ADC/ DSP

Digital filtering (a) Frequency estimation and carrier recovery

Symbol estimation and FEC

(b)

FIGURE 13.3 (a) Basic single-carrier coherent optical system. (b) Block diagram of digital signal processing.

13.2 Single-carrier and multicarrier coherent optical systems

Analog to digital conversion

355

356

CHAPTER 13 OPM for coherent optical systems

modulator. At the receiver, the incoming signal is first detected with the phase- and polarizationdiversity coherent receiver to obtain the full information of the optical signal. Then the downconverted electrical signal is processed by the DSP module with the process functions shown in Figure 13.3(b).13 The analog-to-digital converter (ADC) and clock extraction sub-blocks are used to digitize the received signal and synchronize the four channels with an integer number of samples per symbol. The normalization and orthogonalization block is used to compensate for imperfections in the 90 hybrid and the variation in responsivity of the four photodiodes.13 Digital filtering is introduced to compensate for transmission impairments. Generally speaking, the functionality of digital filtering can be categorized13 as compensation for (1) polarization-independent impairments such as chromatic dispersion and (2) polarization-dependent effects such polarization rotations and PMD dispersion. Additional nonlinear filtering can be implemented if nonlinear impairments such as self-phase modulation or nonlinear phase noise are to be compensated. After the digital filtering block, the phase and frequency mismatch between the incoming signal and the local oscillator is then compensated. One possible phase estimation algorithm for the M-ary PSKmodulated signal is to take the Mth power of the digital signal to remove the phase modulation. Finally, the data can be obtained after the symbol estimation and forward error correction (FEC).

13.2.3 Coherent optical OFDM systems OFDM is a special form of a broader class of MCM. An important property is that the subcarriers are partly overlapped since each of their subcarrier frequencies is orthogonal with every other subcarrier frequency. OFDM has been widely studied in mobile communications to combat hostile frequencyselective fading and has been incorporated into various wireless network standards. CO-OFDM is an optical equivalent of RF-OFDM that combines the technique of “optical coherent detection” and “OFDM.”14 The coherent detection brings OFDM a much-needed linearity in RF-to-optical (RTO) up-conversion and optical-to-RF (OTR) down-conversion. OFDM brings coherent system computation efficiency and ease of channel and phase estimation. Currently, many CO-OFDM experiments using offline signal processing8,9,15,16 have been demonstrated. The complementary metal-oxide semiconductor (CMOS), application-specific integrated circuit (ASIC) chips recently demonstrated for single-carrier coherent systems17,18 signify that current silicon speeds can support 40-Gb/s OFDM transmission systems. A real-time CO-OFDM system with a 3-Gb/s data rate19 has been demonstrated recently. Because of its superior scalability with the bit rate of transmission systems, CO-OFDM is well-positioned to be an attractive modulation format choice for the next generation of ultra-high-speed optical networks.

13.2.3.1 Principle of OFDM Each subcarrier of an MCM signal can be represented as a complex wave, t

sk ¼ Ck ðtÞej2pfk ;

(13.4)

where k ¼ 1, 2,. . ., Nsc, Ck(t) represents complex data at the kth subcarrier. For the sake of simplicity, the pulse shape term is dropped. The MCM-transmitted signal s(t) is the combination of all subcarriers, which can be written as sðtÞ ¼

Nsc 1 X Ck e2pfk t : Nsc k¼1

(13.5)

13.2 Single-carrier and multicarrier coherent optical systems

357

The OFDM signal employs an overlapped yet orthogonal signal set to reduce bandwidth consumption. It can be seen that if the condition fk  ft ¼ m

1 Ts

(13.6)

is satisfied, then the two subcarriers are orthogonal to each other over one OFDM symbol period Ts. In reality, a number of OFDM symbols are grouped together as an OFDM frame, which might also include additional so-called pilot symbols for synchronization and channel estimation. For an OFDM frame with multiple OFDM symbols, the output is the summation of multiple Equation (13.5), which can be expressed as sðtÞ ¼

Nsc 1 X X

Cki Sk ðt  iTs Þ;

(13.7)

i¼1 k¼1

Sk ðtÞ ¼ PðtÞe j2pfk t ; and

 PðtÞ ¼

1 0

ð0 < t < T s Þ ; ðt  0; t > Ts Þ

(13.8)

(13.9)

where cki is the ith information symbol at the kth subcarrier, Sk is the waveform for the kth subcarrier, Nsc is the number of subcarriers, fk is the frequency of the subcarrier, and Ts is the symbol period. The detected information symbol c0 ki is given by ð Ts ð Ts 0  r ðt  iTs Þsk dt ¼¼ r ðt  iTs Þexpðj2pfk tÞdt: (13.10) cki ¼ 0

0

The modulation/demodulation of the OFDM signal can be performed by inverse discrete Fourier transform/discrete Fourier transform (IDFT/DFT) of the input/output information. The corresponding architecture of the RF-OFDM transmitter/receiver using IDFT/DFT and digital-to-analog/analog-todigital converter (DAC/ADC) is shown in Figure 13.4.

13.2.3.2 Cyclic prefix for OFDM A cyclic prefix is created to prevent intersymbol interference (ISI) when an OFDM signal is transmitted in a dispersive channel. As shown in Figure 13.5, the cyclic prefix (CP) is essentially an identical copy of the last portion of the OFDM symbol appended before the OFDM symbol. This CP preserves the orthogonality of the subcarriers and prevents ISI between successive OFDM symbols. The condition for ISI-free OFDM transmission is given by tG < DG ;

(13.11)

where tG is the time delay of the OFDM symbol introduced by CD and PMD.

13.2.3.3 CO-OFDM system architecture A linear transformation in modulation, transmission, and demodulation is the most critical requirement for implementing OFDM in the optical domain. A generic CO-OFDM system uses direct up-/down-conversion (DC) architecture, shown in Figure 13.6, which can be divided into five

358

CHAPTER 13 OPM for coherent optical systems

RF OFDM transmitter

Data S/P

Subcarrier symbol mapper

IDFT

D/A

LPF

I

D/A

LPF

Q

GI

RF OFDM receiver LPF

A/D DFT

LPF

A/D

Data symbol decision

P/S

Data

• DFT window synchronization • Frequency window offset compensation • Subcarrier recovery

FIGURE 13.4 Block diagram of RF OFDM transmitter/receiver. Ts, OFDM symbol period ΔG guard interval

ts, observation period

Identical copy

FIGURE 13.5 Time-domain OFDM signal for one complete OFDM symbol with cyclic prefix.

functional blocks, including (1) the RF-OFDM transmitter, (2) the RTO up-converter, (3) the optical channel, (4) the OTR down-converter, and (5) the RF-OFDM receiver.14 The architecture of RF-OFDM transmitter/receiver is shown in Figure 13.4, and the functionalities of the RF-OFDM transmitter include: 1. Converting the digital data from serial to parallel into a “block” of bits. 2. Mapping the information symbol onto the two-dimensional complex signal.

13.2 Single-carrier and multicarrier coherent optical systems

359

RF-to-optical up-converter I MZM

LD1 Data

RF-OFDM transmitter

90°

MZM Q

Optical link

PD1 I Data

–

PD2 LD2

RF-OFDM receiver

PD3 90°

Q –

PD4

Optical-to-RF down-converter

FIGURE 13.6 CO-OFDM system with direct up-/down-conversion.

3. Performing IDFT of the signal to obtain the time-domain OFDM signal. 4. Inserting pilot symbols and a guard interval for receiver processing, and performing DAC to generate real-time signal SB(t), which can be shown in Equations (13.7)–(13.9). The RTO up-converter uses an optical I/Q modulator that comprises two MZMs with a 90 phase offset to up-convert the real/imaginary parts of the real-time analog-baseband OFDM signal from the RF domain directly to the optical domain. The directly up-converted optical OFDM signal can be expressed as EðtÞ ¼ eðjoLD1 tþjfLD1 Þ  SB ðtÞ;

(13.12)

where oLD1 and fLD1 are, respectively, the angular frequency and phase of the transmitter laser. The up-converted signal E(t) transmits through the optical medium with impulse response h(t), and the received optical signal becomes EðtÞ ¼ eðjoLD1 tþjfLD1 Þ  SB ðtÞhðtÞ;

(13.13)

where  stands for convolution. The optical OFDM signal is then fed into the OTR down-converter, and uses two pairs of balanced receivers to perform I/Q detection optically. The output signal of the OTR down-converter is rðtÞ ¼ eðjoo f f tþjDfÞ r0 ðtÞ; r0 ðtÞ ¼ SB ðtÞhðtÞ:

(13.14)

In the RF-OFDM receiver, the down-converted near-DC-OFDM signal is first sampled with an ADC. Then the signal needs to go through three sophisticated levels of synchronizations before

360

CHAPTER 13 OPM for coherent optical systems

the symbol decision can be made. A detailed description of receiver signal processing can be found in the next section. The three levels of synchronizations are: 1. DFT window synchronization 2. Frequency synchronization 3. Subcarrier recovery Assuming successful completion of DFT window synchronization and frequency synchronization, the RF-OFDM signal after DFT of the sampled value of Equation (13.14) becomes rki ¼ efi hki cki þ nki ;

(13.15)

where rki is the received information symbol, fi is the OFDM symbol phase or common phase error (CPE), hki is the frequency domain channel transfer function, and nki is the random noise. After the phase and channel estimation, an estimated value of cki, c^ki , is given by the zero-forcing method as c^ki ¼

hki jhki j2

eifi gki :

(13.16)

c^ki is used for the symbol decision to recover the transmitter value cki, which is subsequently mapped back to the original transmitted digital bits.

13.2.4 Comparison of single-carrier and multicarrier coherent optical systems For the single-carrier and multicarrier system, there are two conspicuous differences. First, the singlecarrier system usually employs a relatively “conventional” and simpler architecture. In contrast, for the CO-OFDM architecture, the electronic DSP module and DAC are required for complex OFDM signal generation at the transmit end. The OFDM transmitter strictly enforces linearity in each component associated with the CO-OFDM transmitter. Second, in the single-carrier systems, the information is coded in the time domain, whereas in CO-OFDM, the information is encoded in the frequency domain. Given these two basic differences, CO-OFDM provides the following advantages: 1. Ease of signal processing. In CO-OFDM-based multicarrier systems, with the use of pilot symbols and pilot subcarriers, channel estimation and phase estimation can be relatively simple. However, in single-carrier coherent systems, channel estimation has to rely on blind equalization, such as the constant modulus algorithm (CMA) or decision feedback, all of which are prone to error propagation. The phase estimation usually adopts the Viterbi algorithm, which is mostly effective for the pure phase modulation and less effective for other constellation modulation. Furthermore, differential phase coding needs to be employed to resolve the intrinsic phase ambiguity for the Mth-power law algorithm, resulting in approximately a factor of 2 for the bit error rate (BER) increase.6 2. Higher-order modulation. With increasing spectral efficiency in transmission, optical transmission systems should be designed to support the higher-order modulation format. For CO-OFDM systems, the scale-to-higher-order modulation format can be simply achieved via the software to reconfigure DSP and DAC. In contrast, the higher-order, single-carrier optical system requires more complicated optical modulator configurations, which inevitably increase system complexity and cost.

13.3 OPM using coherent detection

361

3. Tight bounding of spectral components. The OFDM spectral shape is tightly bounded and is more tolerant to the filter narrowing effect. Even if the edge subcarriers are attenuated by the narrowing filtering, certain bit- and power-loading schemes in the similar spirit to water-filling algorithm20 can be employed to mitigate the effect. In contrast, the filtering narrowing effect not only causes pulse distortion, but also makes it susceptible to timing jitter in single-carrier systems.

13.3 OPM USING COHERENT DETECTION Various monitoring techniques using coherent detection for both single-carrier and multicarrier systems have been proposed in the literature to monitor one or multiple parameters. Generally, OPMs with coherent detection can be grouped into two categories: OPM without receiver electrical equalization, and OPM with receiver electrical equalization. OPM without receiver electrical equalization usually relies on external devices such as an optical spectrum analyzer (OSA),21 RF devices,22 frequency-selective polarimeter,23 and so on. Recently, many OPMs have been proposed based on the receiver electrical equalization technique,24 which takes advantage of powerful and costeffective silicon signal-processing capabilities. They do not require expensive external devices to evaluate optical properties or to tap the optical signal, which eventually reduces the effective received optical power. In addition, DSP-based OPM techniques are adaptable to varying data rates and modulation formats, and are capable of realizing jointly monitoring OSNR, BER, Q-factor, CD, and PMD.

13.3.1 OPM without receiver electrical equalization 13.3.1.1 OSNR monitoring With the development of laser technology, tunable lasers with wide wavelength coverage, continuouswavelength tunability, and narrow spectral linewidth are commercially available. Coherent receivers incorporated with such a tunable local laser can be used as a high-spectral-resolution OSA. OSAs with coherent detection can be used for precise identification of optical power, OSNR,21 data rate, and modulation format of the signal in the optical domain, which cannot be achieved by conventional grating-based OSAs. Baney et al.25 demonstrated a coherent optical spectrum analysis method based on a swept-tuned optical LO and a coherent receiver that provides fine resolution and high dynamic range. Figure 13.7 is a simplified block diagram of the proposed coherent optical spectrum analyzer (COSA). The incoming signal is combined with a tunable LO via an optical coupler, and a balanced receiver is used to perform OE conversion. The LO frequency is swept across the measurement wavelength range to display the optical spectrum. Figure 13.8 shows the DFB-LD spectral measurement by a COSA and a grating-based OSA that had been set to the narrowest resolution of 80 pm. The LO emitted 2-mW power and scanned at a nominal rate of 62 GHz/s.25 It was reported that COSA was able to clearly resolve the DFB lineshape, including the relaxation sidebands and the central laser peak. Furthermore, the dynamic range achieved by COSA was as high as 70 dB. Tian et al.21 proposed an in-band polarization-assisted OSNR and spectrum monitoring technique based on the swept coherent detection. As shown in Figure 13.9, the swept coherent detector consists

362

CHAPTER 13 OPM for coherent optical systems

In

Es(t)

TZ

EA(t) PD

ELO(t)

Coupler

Δ

out

TZ

EB(t)

LO PD

n(t) = no + g t

FIGURE 13.7 Simplified block diagram of coherent optical spectrum analyzer. LO, local oscillator; PD, photodetector; TZ, transimpedance amplifier.25 Copyright © 2002 IEEE.

Optical power (dBm)

0

–20 COSA

OSA

–40

–60 30 GHz –80 1560.90

1561.00 Wavelength (nm)

1561.15

FIGURE 13.8 Comparison of measurement of DFB-LD linewidth by COSA and by diffraction grating-based OSA.25

Copyright ©

2002 IEEE.

Incoming signal pS RF power meter

50:50

Swept coherent detector

PC2

pLO

PC3

LO

Wavelength tunable

FIGURE 13.9 Schematic for in-band OSNR and spectrum monitoring based on swept coherent detection. FBC, fiber Bragg grating; PC, polarization controller; VOA, variable optical attenuator.21 Copyright © 2006 IEEE.

13.3 OPM using coherent detection

363

of a tunable LO laser, two polarization controllers to adjust the polarization states of the signal and LO, a balanced detector, and a RF power meter. By adjusting the polarization of the signal and LO laser, the proposed method was capable of measuring the signal and the in-band ASE noise spectra separately; the OSNR then can be obtained by integrating the measured spectra. The average RF power at the output of the balanced detector is !2 * !2 + X X ! ! ! ! 2 jAS;k jP LO  P S þ ALO Ai P LO  P ASE;i hPE ðtÞi / jALO j k k (13.17) ! ! 2 2 / jALO j Ps P LO  P S þ 12jALO PASE; where ALO is the LO signal amplitude, and PASE and PS are, respectively, the ASE noise power and !

!

!

signal power. P S and P LO are the signal and LO polarization states. Ai and P ASE are the amplitude !

!

and polarization states of the ASE at frequency fi. hi denotes the time average. If P S ⊥P LO , then

! ! P S P LO

¼ 0, the minimal RF power is detected, and the output is a direct indicator of the in-band !

!

! !

ASE power at a specific frequency. If P S jjP LO , then P S P LO ¼ 1, the RF power meter gives the maximum power, which contains the signal power and the half of the in-band ASE power. The monitored optical spectrum can be obtained by scanning the wavelength of the LO over the optical signal, as shown Figure 13.10. Then the OSNR can be obtained by integrating over the measured spectra for signal and noise power. However, the accuracy of the polarization-assisted OSNR monitoring scheme could be affected by PMD and polarization scattering induced by interchannel cross phase modulation (XPM) in WDM systems26,27; therefore, additional measurement must be used to reduce errors caused by polarization scattering.

13.3.1.2 CD/PMD monitoring Hui et al.22 proposed a combined CD and PMD monitoring technique based on RF signal processing methods that depends on the signal processing on the RF signal after heterodyne detection using external devices such as RF amplifiers, RF filters, RF mixers, and so on. The principle and experiment setup fLO tunable Signal spectrum Max power, pS // pLO Min power, pS ^ pLO

In-band ASE fi 2B

f

Optical frequency

FIGURE 13.10 Operating principle of in-band, high-resolution swept coherent detection scheme.21

Copyright © 2006 IEEE.

364

CHAPTER 13 OPM for coherent optical systems

10-Gbit/s RZ

SMF PMD emulator

EDFA OSA

Tunable laser

BPF

BPF

25 GHz

10 GHz

Oscilloscope CD and PMD calculation GPIB

3-dB coupler Polarization controller/ scrambler

ESA

RF amp Mixer

BPF PD

15 GHz

BPF

BPF

5 GHz

10 GHz

ADC

Power meter

FIGURE 13.11 Block diagram of CD and PMD monitoring using coherent detection. ADC, analog-to-digital converter; BPF, bandpass filter; ESA, electrical spectrum analyzer; OSA, optical spectrum analyzer.22 Copyright © 2005 IEEE.

are shown in Figure 13.11. Since RF technology is more mature than the optical alternative, extending signal processing from the optical domain into the RF domain greatly enhances the system’s functionality and flexibility. CD monitoring is based on the fact that the optical spectrum with RZ or NRZ digital modulation typically has two redundant clock-frequency components. Due to the CD, these two clock components propagate at different speeds. Since the coherent heterodyne detection linearly shifts the optical spectrum into the RF domain, the relative phase-delay information of the optical signal is preserved. The fiber CD can be evaluated from the relative time delay between two recovered RF clocks. In the RF domain, the carrier and the two clock frequencies can be selected separately by three bandpass filters (BPFs). As shown in Figure 13.11, because of the heterodyne detection, the frequency of two RF clocks are both shifted with a frequency equal to the IF carrier frequency. Thus, the carrier component is further split into two components, each used to mix with the upper and lower sidebands independently to generate two baseband clocks. The CD can be evaluated from the relative time delay Dt between these two recovered clocks by Dt ¼ Dl2 Rb =c;

(13.18)

where D is the fiber accumulated CD, Rb is the data rate, l is the signal wavelength, and C is the speed of light. The basic idea of PMD monitoring is to measure the differential polarization walk-off between any two different frequency components within the optical spectrum. When only considering the first-order DGD (Dt), a relatively angular walk-off between two frequencies with Df distance can be used to represent such polarization walk-off: D’ ¼ p  Df  Dt:

(13.19)

Since the RF signal is a linear representation of the optical signal, to perform such measurement, the power of two RF frequencies, f1 and f2, is measured. In Reference 22, the down-converted optical

13.3 OPM using coherent detection

365

carrier ( f1) and lower sideband ( f2) are measured. The RF power of two selected frequencies can be expressed as22 P1 ¼ 1 Plo Psig ðf1 Þ cos2 ð’Þ ; P2 ¼ 2 Plo Psig ðf2 Þ cos2 ð’ þ D’Þ

(13.20)

where 1/2 is the combined effect of two detectors’ responsivities and the relative amplitude of RF frequency f1/2. Plo/sig is the optical power of LO/signal. ’ represents the polarization mismatch between one of the selected frequencies and the LO. D’ is the SOP angle between frequencies f1 and f2. Normalized powers of P1 and P2 are measured to eliminate the uncertainty of the power spectral densities of two frequencies, detector responsivity and RF amplifier gain. The first-order DGD can be obtained through the normalized power cos2’(t) and cos2(’(t) þ D’). However, these methods are not bit rate or modulation-format transparent because several RF BPFs with particular center frequencies are needed in the implementation of these approaches. Roudas et al.23 proposed a PMD monitoring technique based on a frequency-selective polarimeter using coherent heterodyne detection. Due to the inherent high-frequency resolution and power sensitivity of coherent detection, the frequency-selective polarimeter with coherent detection offers superior accuracy compared to its direct detection counterparts. The proposed method is capable of measuring variation in the Stokes parameters as a function of frequency. Figure 13.12 shows the block diagram of such a coherent detection–based, frequency-selective polarimeter, which consists of a 3-dB coupler, balanced receiver, polarization transformer, and electronic preamplifier. A BPF with a center frequency of fc is used to cut a very thin slice of the modulated signal. A square-law detector and a low-pass filter are used to measure the power of this spectral slice, which can be used to estimate the SOP of the received signal. When the LO and the received signal are both planar monochromic waves, the photocurrent at the output of the low-pass filter is iLPF ¼ R2 Ps Plo ð1 þ e^s e^lo Þ;

(13.21)

PH Er1

Er

EIo

3-dB coupler

i1 itot

BPF

(.)2

LPF

i2 Er2

PT

PA

PH

Square-law detector

LO

FIGURE 13.12 Schematic of coherent frequency–selective polarimeter. BPF, bandpass filter; LPF, low-pass filter; PA, power amplifier; PH, photodetector; PT, polarization transformer.23 Copyright © 2004 IEEE.

366

CHAPTER 13 OPM for coherent optical systems

where R is the responsivity of the photodiode, and ps and plo are the average power of the received signal and LO, respectively. e^s ; e^lo are the normalized Stokes vectors corresponding to the SOP of the received signal and LO, respectively. The Stokes components [Sx, Sy, Sz] of the signal around the frequency fc can be estimated with the following equation: 2 32 3 3 2 ð1Þ iLPF jð1Þ Sðx1Þ Sy Sðz1Þ Sx 6 7 1 7 6i j 7 6 Sð2Þ Sðy2Þ Sð2Þ 76 (13.22) S ¼ 2 4 LPF ð2Þ 5  1: 4 x z 54 y 5 R Ps Plo ð3Þ iLPF jð3Þ Sz Sðx3Þ Sy Sðz3Þ R2PsPlo can be estimated from two measurements of the photocurrent corresponding to two antiparallel (in Stokes space) LO-SOPs. iLPF|(k), k ¼ 1, 2, 3 represent three different measurements of the photocurrent corresponding to three noncoplanar LO-SOPs with known Stokes components ½Skx ; Sky ; Skz ; k ¼ 1; 2; 3. For example, LO-SOPs can be set to 0 , linear 45 , and right- or left-circular polarization. Estimation of the Stokes parameter variance as a function of frequency can be performed by tuning the LO frequency at closed-space intervals and repeating Equations (13.21) and (13.22) for each frequency.

13.3.2 OPM with receiver electrical equalization For single-carrier coherent optical systems, Hauske28 demonstrated the CD, DGD, and OSNR monitoring techniques by analyzing finite-impulse response (FIR) filter coefficients. This technique uses a polarization-diversified coherent receiver, as shown in Figure 13.13, to linearly map the optical signal into electrical domains. A bank of FIR filters is applied to the digital signal after analog-to-digital conversion. The filtering is induced by the blind adaptive algorithm to minimize ISI, and thus BER is minimized. The equalizer filter consists of four complex-valued FIR filters arranged in a butterfly structure. The filter’s transfer function is described with a single Jones matrix, " # 1 1 ð f Þ HYX ð fÞ HXX 1 ; (13.23) H ð fÞ ¼ 1 1 HXY ð f Þ HYY ð fÞ

HYX HXY HYY

+

X-pol

Y-pol

Decider

+

Carrier recovery

Y⬘-pol

HXX

Equalizer H

Y-pol

X⬘-pol

Coherent reception ADC Clock recovery

Transmitter

X-pol

Fiber channel F

where the matrix elements are the transfer functions of complex-valued filters. The filter’s transfer function can be assumed as the inverse of the fiber link once the tap algorithm for blind adaptation is converged. Then channel parameters such as CD and DGD can be obtained from filter coefficients. Experiment results based on the 111-Gb/s PolMUX-RZ-DQPSK data showed that the OSNR

FIGURE 13.13 Coherent transmission system with butterfly-structured equalizer filter.28

Copyright © 2008 IEEE.

13.3 OPM using coherent detection

367

could be estimated with a precision of 1 dB. At zero DGD, the highest standard deviation of CD monitoring is 63 ps/nm. Real-time in-service CD and PMD monitoring based on the coefficient extraction of an FIR filter were demonstrated.29 Figure 13.14 shows the experimental setup. The system under test consists of an 80-channel, dual-polarization QPSK system and 800-km fiber with distributed high PMD. Nortel eDC40G circuit packs are used as the dual-polarization QPSK transceiver. The system uses the OC192 signal as the payload. Compared with independent measurements of the PMD, the DGD monitor exhibited tolerance of 12 ps (95% confidence level) over a range of 10–123 ps. Experimental results showed that CD monitoring is independent of instantaneous DGD. The average CD monitor reading was within 32 ps/nm of the independently measured CD. For multicarrier coherent optical systems, with the emergence of CO-OFDM systems, the combination of coherent detection and multicarrier transmission brings about an advanced OPM concept, the so-called OCE.30,31 As proposed by Shieh et al.,30 OCE is equivalent to channel estimation in wireless OFDM communication systems. Because all optical parameters, including OSNR, CD,

MUX

79

QPSK D1

D3

Delay

11.5Gb/s 231–1 PRBS

80

D2

…

2 4

MUX

D4

Tunable transceiver #1

Polarization controller 1

Tunable laser

Booster EDFA WSS

80 loading channels …

1 3

30-km SSMF PBC

QPSK

x3

100-km TWRS

100-km TWRS

100-km TWRS

DGE IV or V or VI x3 Noise loading

OSA WSS

Span 8 100-km TW-RS

I or II or III

Polarimeter Tunable transceiver #2

FIGURE 13.14 Block diagram of real-time monitoring transmission experiment. Tunable laser and polarimeter were used to measure PMD of channel independently of monitor.29 Copyright © 2008 IEEE.

368

X1

CHAPTER 13 OPM for coherent optical systems

TX proc DAC

s1(t)

2⫻4 90⬚ hybrid

I/Q mod

X2

TX proc DAC

ADC, RX proc

Y1

ADC, RX proc

Y2

r2(t)

I/Q mod s2(t)

r1(t)

n times

2⫻4 90− hybrid

FIGURE 13.15 Schematic of investigated OFDM system. OFDM parameters: data rate 100 Gb/s, 256 subcarriers, 16-QAM, 12.5% CP. Twelve spans: 80 km-SSMF per span, D ¼ 17 ps/nm/km, mean PMD ¼ 10 ps.32 Copyright © 2009 IEEE.

and PMD, are embedded in the optical channel response, most of them can be extracted and accurately monitored at the same time after OCE. Therefore, in CO-OFDM systems, the principle of OCE is to estimate the optical channel response by using training symbols. Once channel response is known, OPM is the extraction of various optical parameters from the estimated optical channel response. More importantly, performance monitoring by OCE is basically free because it is embedded as a part of intrinsic receiver signal processing. Such a monitoring device could be also placed anywhere in the network without concern about the large residual CD of the monitored signal. A similar concept of channel estimation based on optical OFDM with coherent detection is proposed by Mayrock et al.32 A simulated optical OFDM system with polarization-diverse coherent receiver and simulation parameters are shown, respectively, in Figure 13.15. Two orthogonally polarized, optical OFDM signals are generated independently. At the receiver, with the DSP, the optical channel matrix is obtained through the pilot symbols. Then the accumulated CD and the differential group delay (DGD) on the subcarrier basis can be extracted from the channel matrix. To minimize the ASE noise impact, Savitzky-Golay filters are used to improve monitoring accuracy. Figure 13.16(a) depicts estimated DGDs that were obtained at an OSNR of 20 dB with polynomial degree p ¼ 1. W is the number of data samples processed by the filter. Figure 13.16(b) summarizes the estimated inverse SNR versus optical input power. For low optical powers, N/S is dominated by ASE noise, and thus an estimate for the OSNR can be deduced directly.31 At higher signal power levels, self-phase modulation is the dominant additive noise. Thus, nonlinear signal degradation can be identified with some additional effort by intentionally inserted signal power variations.33 A detailed discussion of the signal processing of OPM with OCE in CO-OFDM systems is the focus of Section 13.4.

13.4 OPM IN CO-OFDM SYSTEMS As shown in Section 13.2.3, the principle of OFDM is to use a large number of low-speed orthogonal subcarriers to transmit a high-speed data stream. Therefore, each subcarrier only occupies a narrow frequency band, and the channel response for each subcarrier is approximately flat, even though the

13.4 OPM in CO-OFDM systems

260

Branch 1

–12 10 log|N/S|

Δt(d) (ps)

220 200 180 160

Branch 2

–14 –16 –18 –20

140 120

–10

Reference W=5 W = 11 W = 17

240

369

0

50

100

150

Carrier Index (a)

200

250

–22 –16 –14 –12 –10 –8 –6 –4 Popt (dBm)

–2

0

2

(b)

FIGURE 13.16 (a) Estimated DGDs at OSNR ¼ 20 dB using Savitzky-Golay filtering. (b) Estimated inverse SNR versus optical input power.32 Copyright © 2009 IEEE.

global channel response is not. The global channel response is simply the combination of channel responses of all the subcarriers. Furthermore, by using coherent detection and polarization-diversity detection, the optical field can be linearly down-converted to the electrical domain. Consequently, the channel information can be obtained through receiver signal processing. It should be noted that OPM by OCE in a CO-OFDM system can achieve a very fast response, such as less than a microsecond if tens of OFDM symbols are used as the preamble for the OCE. This monitoring speed could be sufficient to accommodate the CD and OSNR change from the environment disturbance. OCE can also be conducted in single-carrier systems but with increased computation complexity.11 This section mainly discusses OPM in CO-OFDM systems.

13.4.1 Optical channel model It is well-known that single-mode fiber supports two polarization modes. Thus, instead of being represented as a single element, the CO-OFDM signal model requires the mathematical description of the polarization effects as well as the fiber CD. Therefore, the optical channel model for a COOFDM signal can be treated as a two-input-two-output (TITO), multiple-input multiple-output (MIMO) OFDM model, which is intrinsically represented by a two-element Jones vector familiar to the optical communication community. Figure 13.17 shows a complete TITO-MIMO CO-OFDM system that consists of two CO-OFDM transmitters (one for each polarization), an optical link, and two CO-OFDM receivers. The dashed line on the devices indicates variations of the MIMO architecture with the option to remove the device from the configuration. The other MIMO architecture includes single-input-single-output (SISO), single-input-two-output (SITO), and two-input-singleoutput (TISO). As discussed in Section 13.2.3, the CO-OFDM transmitter comprises an RF-OFDM

370

CHAPTER 13 OPM for coherent optical systems

Span 1

Span m

Optical OFDM transmitter I

Optical OFDM receiver I PBC

Fiber

PBS

Fiber

Optical OFDM transmitter II

EDFA

EDFA

Optical OFDM receiver II

Optical link with PMD/PDL

FIGURE 13.17 Conceptual diagram of TITO coherent optical MIMO-OFDM model.

transmitter and RTO up-converter, whereas the CO-OFDM receiver comprises an OTR downconverter and RF-OFDM receiver. Each fiber span includes the effects of CD and PMD/PDL. The optical noise is added from the optical amplifiers (OAs) at the end of each span. The fiber nonlinear effect is not considered in this channel model, but we will investigate the influence of fiber nonlinearities on OPM through OCE. Similar to the single-polarization OFDM signal model represented by Equations (13.7)–(13.9), the transmitted OFDM time-domain signal s(t) of the MIMO-OFDM model is described using the Jones vector given by sðtÞ ¼

Nsc 1 X X

cik Pðt  iTs Þej2pfk ðtiTs Þ ;

(13.24)

i¼1 k¼1

sðtÞ ¼ ½ sx and

 PðtÞ ¼

1 0

sy T ; cik ¼ ½ cxik

cyik T ;

ð0 < t < Ts Þ : ðt  0; t > Ts Þ

(13.25)

(13.26)

We use i and k as the indices for the OFDM symbol and OFDM subcarrier, respectively. sx and sy are the two polarization components for s(t), and cik is the transmitted OFDM information symbol in the form of Jones vector for the kth subcarrier in the ith OFDM symbol. The Jones vector cik is employed to describe the generic OFDM information symbol regardless of any polarization configuration for the OFDM transmitter. The received information symbol after the proper DFT window and frequency offset synchronization is given by c0ik ¼ ejfi ejFDðfk Þ Tk cik þ nik ; Tk ¼

N Y l¼1

exp ð12 jbl fk  12al Þs;

(13.27) (13.28)

and 2 ; FD ð fk Þ ¼ f0 þ 2pt0 fk þ pcDt fk2 =fLD1

(13.29)

13.4 OPM in CO-OFDM systems

371

where c0ik ¼ ½ cikx ciky T is the received information symbol in the form of the Jones vector for the kth subcarrier in the ith OFDM symbol, nik ¼ ½ cxik cyik T is the noise including two polarization components, Tk is the Jones matrix for the fiber link, N is the number of PMD/PDL cascading elements represented by their birefringence vector bl and PDL vector al,34 s is the Pauli matrix vector,34 fk is the frequency of kth subcarrier, FD( fk) is the phase owing to the fiber-accumulated CD Di, and fi is the OFDM symbol phase noise owing to the phase noises from the lasers and RF-LO at both the transmitter and receiver.2 fi is usually dominated by laser phase noise. In the channel model of Equation (13.27), CD and DGD are independent of the OFDM symbol index because they are treated as slowly varying compared within the time duration of OFDM symbols. The laser phase-noise term becomes CPE after FFT, common to one OFDM symbol. Although the transmission model of Equation (13.27) includes the frequency responses of transmitter and receiver components, their effect can be considered as stationary and calibrated out in the initial stage, and will not be discussed. 0

0

13.4.2 Principle of OPM through optical channel estimation In order to perform the channel estimation, the phase noise fi for each OFDM symbol has to be obtained through pilot subcarriers introduced in Section 13.2.3. Removing the phase noise fi from Equation (13.27), we obtain 0

c pik ¼ Hð fk Þcik þ npik ;

(13.30)

Hð fk Þ ¼ ejFD ðfk Þ Tk ;

(13.31)

and

 Hð fk Þ ¼

Hxx ð fk Þ

Hxy ð fk Þ

Hyx ð fk Þ

Hyy ð fk Þ

 ;

(13.32)

0

where C pik ¼ C 0ik ejfi , and npik ¼ nik ejfi are, respectively, the received symbol and noise after phase noise compensation. H( fk) is a 2  2 channel response matrix that includes CD and DGD/ PDL. The received signal after channel compensation and phase noise compensation is 0

p 1 cR ik ¼ H ð fk Þc ik :

(13.33)

The goal of OCE is to estimate the four elements of H( fk) through signal processing. Once H( fk) is known, OPM is the extraction of various optical parameters from the estimated optical channel response. One possible method to estimate H( fk) for the TITO-MIMO-OFDM system is by using training symbols in thepreamble using alternate polarization launch35—that is, successive transmis 0 c . Thus, the four elements of H( fk) can be expressed as sion of 1 and 0 c2 Hxx Hyx Hxy Hyy

¼< c01 =c1 ¼< c02 =c1 ¼< c01 =c2 ¼< c02 =c2

> > > >;

(13.34)

where c10 and c20 are the received training symbols, and hi denotes the average overall training symbols.

372

CHAPTER 13 OPM for coherent optical systems

13.4.2.1 CD monitoring From Equation (13.27), it is apparent that phase change of the channel response is mainly induced by the CD. Once the channel transfer function H( fk) is known, the subcarrier phase is given by Fð fk Þ ¼ argðHð fk ÞÞ;

(13.35)

where arg() stands for the phase for a complex signal. According to Equation (13.29), the accumulated chromatic dispersion Dt can be estimated by a simple second curve fitting of F( fk) as a function of the subcarrier frequency. Note that CD monitoring is only based on the phase curves of the channel response, and therefore immune to the DGD-induced amplitude change.

13.4.2.2 DGD monitoring It can be shown that the amplitude response of Uxx( fk) and Uvx( fk) can be expressed as28 ( jUxx ð fk Þj2 ¼ a þ b cosð2pfk tÞ aþb1 ; jUyx ð fk Þj2 ¼ 1  a  b cosð2pfk tÞ a  b  1

(13.36)

where a and b are constants. |Uxx( fk)|2 and |Uyx( fk)|2 essentially represent the multipath interference due to DGD t. By analyzing Equation (13.36), such interference/fading is periodic, which follows a cosine function with a period determined by t. Thus, the amplitude fading of the OFDM subcarrier is caused by DGD, which can be estimated by the inverse of the fading period.

13.4.2.3 System Q-factor monitoring Another important parameter to monitor is the system Q-factor. A live system could run error-free even without FEC for an extended period, making it hard to detect the system margin by measuring BER directly. From Equation (13.27), we can see that each subcarrier channel is essentially a linear channel with additive white Gaussian noise. Subsequently, the BER of a QPSK system is given by36  pffiffiffi BER ¼ 0:5  er f c ESNR= 2 (13.37) and ESNR ¼ hhc0ki i2i =d2k ik;

(13.38)

where ESNR is the (electrical) SNR ratio per bit, hik stands for the averaging over the subcarriers or the index k, hciik ii is the expectation value of the received symbol for subcarrier k, and qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi dk ¼ hjc0ik j2 ii  hciik i2i is the standard deviation of the received symbol for subcarrier k. Equation (13.38) shows that ESNR can be obtained by first constructing the constellation of the received symbol and then performing the computation of ESNR for each constellation point. We further convert the BER in Equation (13.37) into the Q-value,37 which is commonly used in the optical community. From Equation (13.37), the system Q is thus given by Q ¼ 10 log10 ð2ESNRÞ:

(13.39)

From Equations (13.38) and (13.39), the system Q can be effectively monitored by computing the subcarrier symbol spread in the constellation diagram. Since the CD-induced delay spread and the ISI can be completely removed in a CO-OFDM system, the electrical noise characterized by d is predominately from the accumulation of ASE noise from optical amplifiers, and it can be shown that

13.5 Progress in OPM for CO-OFDM systems

1 1 ¼A þ B; ESNR OSNR

373

(13.40)

where A is a proportional constant between ESNR and OSNR, and B is attributed to the background noise not accounted for by ASE noise, which is mainly from the phase noise of the transmit/receive lasers. From Equations (13.37)–(13.40), we can see that by acquiring hc0 iki and dk through receiver signal processing, the ESNR for the OFDM signal can be computed, and subsequently both the system Q and the OSNR can be monitored. The coefficients A and B in Equation (13.40) can be obtained empirically with a calibration procedure by measuring ESNRs against a series of known OSNRs and performing a linear fit between 1/ESNR and 1/OSNR. It is quite instructive to explicitly write out the ideal coherent detection performance for a QPSKmodulated CO-OFDM system where the linewidths of the transmit/receive lasers are assumed to be zero. From Equation (13.37), the corresponding BER, Q, and ESNR in this ideal condition can be given by  B0 ; (13.41) BER ¼ 0:5  er f c OSNR R  B0 Q ¼ 10 log10 2OSNR ; (13.42) R and ESNR ¼ 2OSNR

B0 ; R

(13.43)

where B0 is the optical ASE noise bandwidth used for OSNR measurement (12.5 GHz for 0.1-nm bandwidth), R Nsc Df is the total system symbol transmission rate, and Nsc and Df are the number of subcarriers and channel spacing of the subcarriers, respectively.

13.5 PROGRESS IN OPM FOR CO-OFDM SYSTEMS In this section we report the recent progress in OPM for CO-OFDM systems in terms of simulation and experimental demonstration. We show that in polarization, multiplexed CO-OFDM systems, critical optical system parameters including fiber CD, PMD, Q-value, and OSNR can be accurately monitored without resorting to separate monitoring devices.

13.5.1 Simulation model and results A Monte Carlo simulation is carried out to demonstrate the CD, Q, and OSNR monitoring for a SISOMIMO-CO-OFDM system. The OFDM parameters are a symbol period of 25.6 ns, a guard time of 3.2 ns, and 256 subcarriers. BPSK encoding is used for each subcarrier, resulting in a total bit rate of 10 Gb/s. The linewidth of the transmitter and receiver lasers are assumed to be 100 kHz each, which is close to the value achieved with commercially available semiconductor lasers.38,39 The linked ASE noise from the optical amplifiers is assumed to be additive white Gaussian noise, and the phase noise of the lasers is modeled as white frequency noise characterized by its linewidth. The CD is assumed

CHAPTER 13 OPM for coherent optical systems

Monitored CD (ps/nm)

60000

100 Monitored CD (ps/nm) Monitored CD error (ps/nm) 50

40000 0 20000 –50 0

0

Monitored CD error (ps/nm)

374

–100 60000

20000 40000 Link CD (ps/nm)

FIGURE 13.18 Performance of CD monitoring through channel estimation.

to be constant within the OFDM spectrum. Eight-block OFDM symbols, each consisting of 100 OFDM symbols, are used for extracting various parameters, including CD, system Q, and OSNR. In the following text, we use “calculate” to mean the BER results obtained by Monte Carlo simulation, and “monitor” to mean the interpolation results obtained by Equations (13.41)–(13.43). Figure 13.18 shows the monitored CD from the receiver signal processing. The input OSNR is set at 3.8 dB, which gives a BER of 103 for a CD below 34,000 ps/nm. We can see that CD up to 50,000 ps/nm can be monitored with an accuracy of 50 ps/nm. The simultaneous large dynamic range and good accuracy of CD monitoring are the unique features of the OFDM modulation format, namely, a large number of subcarriers spread across a wide spectrum of 10 GHz, resulting in good accuracy, and narrow subcarrier channel spacing of 44.6 MHz, resulting in wide dynamic range. This wide dynamic range is an improvement of more than one order of magnitude over a prior report using single or a few auxiliary subcarriers.40 Figure 13.19 shows the monitored system Q and OSNR though OCE. The Q is calculated from 7 to 12 dB by Monte Carlo simulation—that is, direct BER simulation with signal duration of 2

20

Calculated Q Monitored Q

Q Margin

15

OSNR error

10

5

–2

–6

–10

0

5

10

15 OSNR (dB)

FIGURE 13.19 Monitored system Q and OSNR as function of input OSNR.

20

25

–14 30

Monitored OSNR error (dB)

System Q (dB)

25

13.5 Progress in OPM for CO-OFDM systems

375

20.5 ms, represented by the solid squares in Figure 13.19. This demonstrates good agreement with the monitored Q by Equation (13.39). Beyond that, we rely on Equation (13.39) for system Q estimation. To appreciate the advantage of this approach, for instance, at an input OSNR of 20 dB, the system Q for this OSNR is monitored to be 21.3 dB, which gives a Q-margin of 11.5 dB over a BER of 103. Such a method of Q-margin prediction at high OSNRs is similar to that in direct detected systems.35 Thus, the margin monitoring is achieved nonintrusively. Note that this level of system margin cannot be measured directly. Additionally, the OSNR is monitored by computing ESNR and estimating OSNR using Equation (13.40). The curve with solid triangles in Figure 13.19 shows that the OSNR can be monitored with errors within 0.5 dB for an input OSNR dynamic range of 1–20 dB. The maximum OSNR that can be monitored is limited by laser phase noise.

13.5.2 Optical performance monitoring in CO-OFDM systems with 4-QAM In this section we focus on the experimental demonstration of OSNR, Q-factor, fiber CD, and DGD monitoring through OCE in CO-OFDM systems.31

13.5.2.1 Experiment setup The experimental setup in this work comprises a generic SITO-MIMO CO-OFDM transmission system, as shown in Figure 13.20. The transmitter and receiver laser sources in this work both have a specified linewidth of less than 100 kHz. The optical carrier wavelength is about 1555 nm. The data rate is 10.7 Gb/s. The OFDM parameters used in the experiment are listed in Table 13.1. In the experiment, 11 pilot subcarriers are used for phase estimation. The spacing of OFDM subcarriers

OBF Transmitter EDFA

LD1 DMZ Laser

DGD Optical attenuator

Q

I AWG

Balanced receiver I

Optical hybrid

Balanced receiver II

Optical hybrid

TDS

LD2 PBS

Polarization diversity receiver

FIGURE 13.20 Experimental setup for optical performance monitoring with SITO-MIMO CO-OFDM system. (There is a polarization controller before DGD emulator.) AWG, arbitrary waveform generator; DMZ, dual MZ modulator; EDFA, erbium-doped fiber amplifier; TDS, time-domain sampling scope.

376

CHAPTER 13 OPM for coherent optical systems

Table 13.1 OFDM Parameters for 4-QAM Transmission FFT Length 128

Subcarrier Number 88

Symbol Period 12.8 ns

Guard Interval 1.6 ns

Pilot Subcarrier Number 11

Effective Data Rate 10.7 Gb/s

is 1/(12.8  109) ¼ 78.125 MHz. The RF-OFDM sequence is generated by an arbitrary waveform generator at 10 GSa/s to emulate the DAC, and the electrical-to-optical (EO) direct up-conversion is realized by a dual MZ modulator configured as an optical I/Q modulator. A home-built PMD emulator follows the transmitter to generate different DGD. A recirculating loop, including 100.8-km standard SMF fiber and an EDFA to compensate the loss, is used to emulate the long-haul transmission. As the PMD emulator is outside the recirculating loop, only the first-order PMD is emulated. The optical signal coupled out from the recirculating loop transmits through an optical attenuator and another EDFA to evaluate the system performance through ASE noise loading. The output signal after fiber transmission is detected using a polarization-diversity coherent receiver with intermediatefrequency (IF) down-conversion detection. After the optical-to-electrical (OE) down-conversion, the RF-OFDM signal is sampled by a real-time scope (TDS) at 20 GSa/s to emulate the ADC, and the resultant digital sequences are uploaded to a computer for DSP. The OCE is conducted by sending preambles in the OFDM frame. In this experiment, 40 training symbols (equivalent to an OPM response time in the order of microseconds) are used in the preamble to keep a balance between OCE accuracy and OPM response speed.

13.5.2.2 OSNR and Q-factor monitoring In this section we monitor the OSNR and Q-factor by loading different levels of ASE noise in a back-to-back transmission. The OSNR measured by an OSA is used as the reference. Both OSNR and Q-factor monitoring are derived from the electrical SNR of the received RF-OFDM signal according to Equations (13.42) and (13.43). SNR (per bit) is calculated from the noise spreading of received signal constellations based on Equation (13.38). The best SNR in our system is about 17 dB, which is used to determine and subtract the background noise (mainly from the RF components). The OSNR monitoring result is shown in Figure 13.21(a). Due to background noise, it is difficult to monitor the high OSNR in the system. However, the monitored OSNR error is within 0.5 dB over a range from 6 to 18 dB, which covers the main OSNR dynamic range of interest. Figure 13.21(b) shows the monitored Q-factor result. The Q-factor calculated from the BER, the calculated Q, is included for comparison. It can be seen that the monitored Q agrees with the calculated Q within 0.6 dB, which implies that our system is dominated by Gaussian noise. When the BER is low, it is difficult to obtain the meaningful calculated Q, which signifies the importance of Q-factor monitoring. The calculated Q is limited to 12.7 dB due to the maximum number of OFDM symbols processed.

13.5.2.3 CD and DGD monitoring The CO-OFDM channel response is mainly determined by CD and PMD in the optical link. In this section we first estimate the channel response and then extract the parameters for CD and DGD in the long-haul transmission.

20

2

16

1

12

0

8

–1

4

5

10 15 OSNR by OSA (dB)

377

Monitoring error (dB)

Monitored OSNR (dB)

13.5 Progress in OPM for CO-OFDM systems

–2 20

(a)

Q-factor (dB)

1 15 0 10

5

Monitored Q Calculated Q Q-factor error 5

10

15

–1

Q-factor error (dB)

2

20

–2 20

OSNR by OSA (dB) (b)

FIGURE 13.21 (a) OSNR monitoring result. (b) Q-factor monitoring result. Both are measured in back-to-back transmission.

Figure 13.22 shows an example of channel response of the 1008-km and 900-ps DGD transmission. Since we use a SITO-MIMO architecture, we can estimate the channel responses for both polarization components corresponding to Hxx( fk) and Hyx( fk) in Equation (13.32). Figure 13.22 clearly shows that due to the CD the phase response is parabolic and the magnitude has an apparent DGD-induced, frequency-selective fading. As discussed in Section 13.4.2, CD is estimated by second curve fitting of the phase response, and the DGD is estimated by the inverse of the period of the fading as shown in Equation (13.36); for example, the period in Figure 13.22(a) is 1.1 GHz, which corresponds to a DGD of 909 ps. Figure 13.23(a) shows the CD monitoring result versus transmission distance. Without DGD, the monitored CD increases linearly with the transmission distance. The CD parameter is around 16.4 ps/ nm/km, which corresponds to the dispersion of the SMF used in the experiment. The CD monitoring result with fixed 900-ps DGD is also shown in Figure 13.23(a). Compared with the non-DGD monitoring result, the DGD causes about 5% of the monitoring variation. The main reason for such monitoring variation is the phase ripples, as illustrated in Figure 13.22. The launch power into the fiber is about 0 dBm. It has been shown that the fiber nonlinearity at this power level is very strong.41

CHAPTER 13 OPM for coherent optical systems

5 0 –5

–10

–10 –15

–15 –20

–20

0

10

20 30 40 50 60 70 Frequency (⫻78.125 MHz) (a)

80

–25 90

5

–5 Magnitude (dB, relative)

Phase (rad)

Magnitude (dB, relative)

–5

0 –5

–10

–10 –15

–15

Phase (rad)

378

–20 –20

0

10

20 30 40 50 60 70 Frequency (⫻78.125 MHz) (b)

80

–25 90

FIGURE 13.22 Estimated channel responses for (a) x and (b) y polarization components X-axes are the frequencies normalized to OFDM subcarrier spacing.

Therefore, Figure 13.23(a) also in essence indicates that the CD monitoring is robust against fiber nonlinearity. Figure 13.23(b) shows the CD monitoring error versus varying DGD at two different OSNR conditions after 1008-km transmission with about 7-dBm launch power. The high OSNR (13–14 dB) condition has a BER smaller than 105, whereas the low OSNR (6.5–7.5 dB) has a BER greater than 103. The CD monitoring errors of both OSNR conditions are within 5%, which shows that the CD monitoring can be immune to DGD and ASE noise.

13.5.3 OPM in CO-OFDM systems with 16-QAM modulation The OPM in CO-OFDM systems with 16-QAM is conducted in this section. The experiment setup is similar to the one used for 4-QAM, excepting that the PMD effect is excluded. The data rate is 10 Gb/s and the OFDM parameters are shown in Table 13.2.

13.6 OPM experiment results

10%

16000

5%

12000 0% 8000 Without DGD With DGD DGD-induced CD variation

4000 0

0

200

400 600 800 Transmission distance (km) (a)

–5%

DGD-induced CD variation (dB)

Chromatic dispersion

20000

379

–10% 1000

Relative error

4% 2% 0% –2% –4%

High OSNR Low OSNR 0

400

800

1200

DGD (ps) (b)

FIGURE 13.23 (a) CD monitoring versus transmission distance with and without DGD. (b) CD monitoring error versus DGD after 1008-km transmission.

Table 13.2 OFDM Parameters for 16-QAM Transmission FFT Length 128

Number of Subcarrier 44

Symbol Period 13.6 ns

Guard Interval 0.8 ns

Number of Pilot Subcarrier 8

Data Rate 10 Gb/s

13.6 OPM EXPERIMENT RESULTS Figure 13.24 shows the monitoring results of OSNR and Q-factor in the back-to-back transmission. When the OSNR is below 18 dB, the OSNR monitoring error is below 0.5 dB, as shown in Figure 13.24(a). The bigger monitoring error beyond 18 dB is due to the finite SNR of the RF components in the transmitter and receiver. Figure 13.25 shows the CD monitoring results. The monitored CD linearly increases with the transmission distance and the CD coefficient of the transmission fiber is calculated as 16.22 ps/nm/km.

CHAPTER 13 OPM for coherent optical systems

Monitored OSNR (dB)

20 1

18 16

0 14 12

8

–1

OSNR-mon OSNR-mon offset

10 8

10

12 14 16 OSNR by OSA (dB) (a)

18

–2 20

2

16

Q-factor (dB)

14

1

12 0 10

4 Monitored Q Calculated Q Q-factor error

8 6

Monitoring error (dB)

2

22

8

10

12

14

16

18

–1

Q-factor error (dB)

380

–2 20

OSNR by OSA (dB) (b)

FIGURE 13.24 Monitoring results in CO-OFDM system with 16-QAM for (a) OSNR and (b) Q-factor.

Chromatic dispersion (ps/nm)

20000 16000 12000 8000 4000 0

0

200

400 600 800 1000 Transmission distance (km)

FIGURE 13.25 CD monitoring result in CO-OFDM system with 16-QAM.

1200

References

381

13.7 SUMMARY We have reviewed the principle and progress of the OPM techniques in coherent optical systems with a focus on MCM. We have presented the theoretical background as well as experimental demonstration of OPM in CO-OFDM systems. In particular, we have taken advantage of OCE using advanced DSP to realize a fast and joint estimation of channel parameters as OSNR, Q-factor, CD, and PMD.

REFERENCES 1. Giles RC, Reichman KC. Optical self-homodyne DPSK transmission at 1-Gbit/s and 2-Gbit/s over 86 km of fiber. IEE Electron Lett 1987;23(22):1180–1. 2. Kazovsky LG, Benedetto S, Willner AE. Optical fiber communication systems. Norwood, CT: Artech House; 1996. 3. Okoshi T, Kikuchi K. Coherent optical fiber communications. Berlin: Springer; 1988. 4. Noe R, Sandel D, Yoshida-Dierolf M, Hinz S, Mirvoda V, Schopflin A, et al. Polarization mode dispersion compensation at 10, 20, and 40 Gb/s with various optical equalizers. IEEE/OSA J Lightwave Technol 1999;17(9):1602–16. 5. Savory SJ, Gavioli G, Killey RI, Bayvel P. Electronic compensation of chromatic dispersion using a digital coherent receiver. OSA Opt Express 2007;15(5):2120–6. 6. Ly-Gagnon DS, Tsukarnoto S, Katoh K, Kikuchi K. Coherent detection of optical quadrature please-shift keying signals with carrier phase estimation. IEEE/OSA J Lightwave Technol. 2006;24(1):12–21. 7. Charlet G, Renaudier J, Salsi M, Mardoyan H, Tran P, Bigo S. Efficient mitigation of fiber impairments in an ultra-long haul transmission of 40 Gbit/s polarizationmultiplexed data, by digital processing in a coherent receiver. In: Technical digest of optical fiber communication conference and exposition and the national fiber optic engineers conference (OFC/NFOEC), paper PDP17. Anaheim, CA; 2007. 8. Shieh W, Yi X, Ma Y, Tang Y. Theoretical and experimental study on PMD-supported transmission using polarization diversity in coherent optical OFDM systems. OSA Opt Express 2007;15(16):9936–47. 9. Jansen SL, Morita I, Tanaka H. 16  52.5-Gb/s, 50-GHz spaced, POLMUX-COOFDM transmission over 4,160 km of SSMF enabled by MIMO processing KDDI R&D laboratories. In: Proc. of European conference on optical communications (ECOC), paper PD1.3. Berlin; 2007. 10. Taylor MG. Coherent detection method using DSP for demodulation of signal and subsequent equalization of propagation impairments. IEEE Photon Technol Lett 2004;16(2):674–6. 11. Ip E, Kahn JM. Digital equalization of chromatic dispersion and polarization mode dispersion. IEEE/OSA J Lightwave Technol 2007;25(8):2033–43. 12. Kikuchi K. Phase-diversity homodyne receiver for coherent optical communications. In: Technical digest of optical amplifiers and their applications/coherent optical technologies and applications, paper CThB3. Whistler, BC, Canada; 2006. 13. Seb J Savory. Digital filters for coherent optical receivers. OSA Opt Express 2008;16(2):804–17. 14. Shieh W, Athaudage C. Coherent optical orthogonal frequency division multiplexing. IEE Electron Lett 2006;42(10):587–9. 15. Shieh W, Yi X, Tang Y. Transmission experiment of multi-gigabit coherent optical OFDM systems over 1000 km SSMF fiber. IEE Electron Lett 2007;43(3):183–5. 16. Jansen SL, Morita I, Takeda N, Tanaka H. 20-Gb/s OFDM transmission over 4,160-km SSMF enabled by RFpilot tone phase noise compensation. In: Technical digest of optical fiber communication conference and exposition and the national fiber optic engineers conference (OFC/NFOEC), paper PDP15. Anaheim, CA; 2007.

382

CHAPTER 13 OPM for coherent optical systems

17. Roberts K. Electronic dispersion compensation beyond 10 Gb/s. In: Technical digest of IEEE LEOS summer topical meeting, paper MA2.3. Portland, OR; 2007. 18. Sitch J. Implementation aspects of high-speed DSP for transmitter and receiver signal processing. In: Technical digest of IEEE LEOS summer topical meeting, paper MA4.3. Portland, OR; 2007. 19. Yang Q, Chen S, Ma Y, Shieh W. Real-time reception of multi-gigabit coherent optical OFDM signals. OSA Opt Express 2009;17(10):7985–92. 20. Proakis J, Salehi M. Digital communications. 5th ed. Singapore: McGraw-Hill International Edition; 2008. 21. Tian X, Su Y, Hu W, Leng L, Hu P, He H, et al. Precise in-band OSNR and spectrum monitoring using high-resolution swept coherent detection. IEEE Photon Technol Lett 2006;18(1):145–7. 22. Fu B, Hui R. Fiber chromatic dispersion and polarization-mode dispersion monitoring using coherent detection. IEEE Photon Technol Lett 2005;17(7):1561–3. 23. Roudas I, Piech G, Mlejnek M, Mauro Y, Chowdhury D, Vasilyev M. Coherent frequency-selective polarimeter for polarization-mode dispersion monitoring. IEEE/OSA J Lightwave Technol 2004;22(4):953–67. 24. Buchali F. Electronic dispersion compensation for enhanced optical transmission. In: Technical digest of optical fiber communication conference and exposition and the national fiber optic engineers conference (OFC/NFOEC), paper OWR5, Anaheim, CA; 2006. 25. Baney DM, Szafraniec B, Motamedi A. Coherent optical spectrum analyzer. IEEE Photon Technol Lett 2002;14(3):355–7. 26. Lee JH, lung DK, Kim CH, Chung YC. OSNR monitoring technique using polarization-nulling method. IEEE Photon Technol Lett 2001;13(1):88–90. 27. Xie C, Mtiller L, Kilper DC, Mollenauer LF. Impact of cross-phase modulation induced polarization scattering on optical PMD compensation in WDM systems. OSA Opt Lett 2003;28(23):2303–5. 28. Hauske FN, Geyer J, Kuschnerov M, Piyawanno K, Duthel T, Fludger CRS, et al. Optical performance monitoring from FIR filter coefficients in coherent receivers. In: Technical digest of optical fiber communication conference and exposition and the national fiber optic engineers conference (OFC/NFOEC), paper OThW2. San Diego, CA; 2008. 29. Woodward SL, Nelson LE, Feuer MD, Zhou X, Magill PD, Foo S, et al. Characterization of real-time PMD and chromatic dispersion monitoring in a high-PMD 46-Gb/s transmission system. IEEE Photon Technol Lett 2008;20(24):2048–50. 30. Shieh W, Tucker R, Chen W, Yi X, Pendock G. Optical performance monitoring in coherent optical OFDM systems. OSA Opt Express 2007;15(2):350–6. 31. Yi X, Shieh W, Ma Y, Tang Y, Pendock GJ. Experimental demonstration of optical performance monitoring in coherent optical OFDM systems. In: Technical digest of optical fiber communication conference and exposition and the national fiber optic engineers conference (OFC/NFOEC), paper OThW3. San Diego, CA; 2008. 32. Mayrock M, Haunstein H. Performance monitoring in optical OFDM systems. In: Proc. conference on optical fiber communication (OFC), paper OWM3. San Diego, CA 2009. 33. Mayrock M, Haunstein H. Optical monitoring for non-linearity identification in CO-OFDM transmission systems. In: Proc. conference on optical fiber communication (OFC), paper JThA58. San Diego, CA 2008. 34. Gisin N, Huttner B. Combined effects of polarization mode dispersion and polarization dependent losses in optical fibers. Opt Commun 1997;142(1–3):119–25. 35. Shieh W, Yi X, Ma Y, Tang Y. Theoretical and experimental study on PMD-supported transmission using polarization diversity in coherent optical OFDM systems. Opt. Express 2007;15:9936–47. 36. Proakis J. Digital communications. 3rd ed. New York: WCB/McGraw-Hill; 1995 [chapter 5]. 37. Bergano NS, Kerfoot FW, Davidsion CR. Margin measurements in optical amplifier system. IEEE Photon Technol Lett 1993;5(3):304–6. 38. Berger JD, Anthon D. Tunable MEMS devices for optical networks. Opt Photon News 2003;14:43–9.

References

383

39. Ip E, Kahn J, Anthon D, Hutchins J. Linewidth measurements of MEMS-based tunable lasers for phaselocking applications. IEEE Photon Technol Lett 2005;17(10):2029–31. 40. Liu A, Pendock GJ, Tucker RS. Improved chromatic dispersion monitoring using single RF monitoring tone. OSA Opt Express 2006;14(11):4611–6. 41. Ma Y, Shieh W, Yi X. Characterization of nonlinearity performance for coherent optical OFDM signals under influence of PMD. IEE Electron Lett 2007;43(17):943–5.

This page intentionally left blank

CHAPTER

Optical performance monitoring in optical transport networks

14 Wolfgang Grupp

JDSU Deutschland GmbH, Eningen u.A., Germany

14.1 INTRODUCTION This chapter introduces the functional elements required for the implementation of optical transport services, and presents the role of optical performance monitoring functions in the operation and maintenance process of optical layer networks and their integration into a telecommunications management network (TMN). The applicability of optical power monitoring, spectral analysis, and Q-factor measurements for optical layer supervision will be discussed, followed by a cost-benefit assessment.

14.2 OVERVIEW Driven by the tremendous growth in IP-based data traffic, telecommunications providers are facing the urgent need to manage ultra-high-capacity networks and provide the capability for evolving to a multiservice transport platform that supports packet-based data transport as well as legacy traffic. The underlying optical transport network (OTN) infrastructure must be inherently scalable to handle this traffic and flexible enough to respond to rapidly changing traffic patterns. At the same time, the exponential growth in data traffic has not translated into a proportional increase in revenue, forcing network operators to minimize the cost per bit not only by reducing equipment cost but also by optimizing their business processes. The following sections introduce the business environment and the high-level network architecture required for advanced optical transport services.

14.2.1 Business interface model Deregulation and global competition in the telecommunications industry have created a situation where multiple service providers and transport providers are cooperating to deliver an increasing range of services to the end user. Figure 14.1 illustrates the business interfaces that can exist between service and transport providers and end users. An end user may interface directly with a transport provider or via a service provider. Service providers are interconnected with transport providers and service and transport providers usually have interconnections agreements with their peers. © 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00014-6

385

386

CHAPTER 14 Optical performance monitoring in optical transport networks

User

Service provider

Transport provider

FIGURE 14.1 Business interface model.

Business relations are formalized in terms of service-level agreements (SLAs) that define traffic patterns, quality-of-service (QoS) parameters, monitoring procedures, reaction patterns in case of a service degradation—and penalties! Test equipment for SLA validation plays an important role as an objective arbiter when conflicts between service providers and users have to be resolved.

14.2.2 Generic OTN service requirements The emergence of intelligent optical switches and networking protocols will offer enhanced functionality to digital clients of the optical layer network, including the following: •

•

•

Fast and cost-effective lightpath provisioning to set up and release OTN connections with customer-defined attributes (e.g., bit rate, framing, protection schemes), including the configuration when the two endpoints lie in different operators’ network management domains. End-to-end network visibility. Each service provider needs to maintain its own view of all end-to-end connections, including connection segments that are managed by interconnection partners. A transport service provider is therefore responsible for disseminating all relevant information about shared resources to all its peers. Enhanced network survivability on the basis of optical channel protection and restoration. This includes dissemination of fault information from the transport provider owning the faulty resource to other transport providers using that resource. Operators can offer flexible restoration and recovery policies depending on service classes, for example, by offering 50-ms optical protection switching to “gold customers,” whereas “best-effort customers” would have to live with the much slower restoration procedures offered by the IP layer.

From the viewpoint of an OTN service user, an optical link is just a point-to-point, “big-dumb-pipe” interconnect between routers and switches. The intelligence required for provisioning and maintaining this “dumb pipe” is completely delegated to the optical layer, which takes care of the intricacies of network planning, link engineering, and service assurance.

14.2 Overview

387

14.2.3 OTN: A network of networks A large OTN can be viewed as a “network of networks,” made up of separately managed subnetworks with 3R (regeneration, reshaping, retiming) regeneration at the user-network interface (UNI) and at the interdomain interface (IrDI) between two subnetworks. Each of the subnetworks may have its proprietary fault, performance, and configuration management processes for the network layer and the network element layer, as illustrated in Figure 14.2. The IrDI for the OTN hierarchy has been standardized in ITU-Rec. G.7091, and is gaining acceptance in the provider community. The so-called “digital wrapper” (Figure 14.5) defined in G.709 is a common encapsulation format for all kinds of data signals and offers comprehensive monitoring and management capabilities (see Sections 14.8 and 14.9). End-to-end integration across different administrative domains takes place at the network and service levels by exchanging configuration and performance information about the optical transport

Data client

Data client

End-to-end application

End-to-end OTN connection UNI

UNI

3R

3R Administrative domain A

Administrative domain B IrDI NNI 3R

3R

ONE ONE

ONE ONE

IaDI ONE NNI

ONE

ONE NMI

NMI Xcoop

FIGURE 14.2 Data networking and transport networking.

388

CHAPTER 14 Optical performance monitoring in optical transport networks

entities (optical transmission section [OTS], optical multiplex section [OMS], and optical channel [OCh] trails and connections as defined in ITU-T Rec. G.872) via Xcoop interfaces, or in the future via signaling at the UNI or the network node interface (NNI), or simply by FAX or telephone.2

14.3 GENERIC MODELING PRINCIPLES FOR TRANSPORT NETWORKS The following section gives a summary of the principles and architectural components defined in ITU-T Rec. G.805 that form the basis for modeling of multilayer and multiprovider transport networks.3

14.3.1 Top-level functional architecture Provisioning of optical transport services requires integration of transport functions, management functions, and control functions in a multivendor/multiprovider environment. A management concept based on layering and partitioning and abstraction was developed to master the complexity of optical networking in a global multivendor/multiprovider environment. Those concepts are the enablers for flexible and cost-effective network integration, rapid provisioning of broadband services, and controllable service quality on optical networks. The separation into data networking and transport networking is a first step toward layered functional architecture. However, this simple approach is not sufficient for modeling the full range of networking and control processes required for implementation of intelligent transport services. The various functions that constitute a telecommunications network can be classified into two broad functional groups. One is the control and management functional group, which realizes various ancillary services and operations and maintenance functions. The other is the transport functional group, which transfers any telecommunications information from one point to another point(s). Automated provisioning of optical transport services requires the integration of transport functions, management functions and control functions. Figure 14.3 shows the top-level architecture of an automated switched-transport network (ASTN) and the main functional areas allocated to the control plane, transport plane, and management plane.

14.3.2 Control plane functions Control plane functions include neighbor discovery, signaling, routing, and link management to support automated provisioning of connections in an optical layer network. The message exchange between distributed control plane processes is done either via a separate overlay network, via a separate signaling channel on the same fiber, or via the embedded communication channels in the digital wrapper overhead.

14.3.3 Management functions Management functions comprise the five classical network management functional areas, namely fault, configuration, accounting, performance, and security management (FCAPS). A telecommunications management network (TMN) provides the means to transport and process information related to the management of telecommunications networks.4 A TMN can vary in complexity from a very

14.3 Generic modeling principles for transport networks

389

Control plane OCC

NNI signaling

OCC

OCC

UNI signaling

CCI

• Resource discovery • Routing • Signaling • Link management • Call detail records

Transport plane • Transmission • Switching • Multiplexing

Client User data

Management plane UNI user network I/F NNI network node I/F OCC optical connection controller CCI connection controller I/F

TMN Operations system

• Fault management • Configuration management • Accounting • Performance management • Security management

FIGURE 14.3 ASTN architecture.

simple connection between an operations system (OS) and a single piece of telecommunications equipment to a complex network interconnecting many different types of OSs and telecommunications equipment. Testing and monitoring are primarily needed for fault and configuration management and for providing data for performance management applications. Therefore, those three functional areas will be described in more detail in the following sections.

14.3.3.1 Configuration management Configuration management is responsible for system and channel activation, addition of new channels, setting parameters that control the routine operation of the transmission system, changing the configuration of the network resources (e.g., switching from normal mode into test mode), and keeping track about the current condition of the system (network view).

14.3.3.2 Fault management Fault management is a set of functions and procedures that enable the detection, isolation, and correction of abnormal operation of the telecommunication network and its environment. The six basic fault management applications according to ITU-T M.3400 are:5 •

RAS quality assurance. RAS stands for reliability, availability, and survivability. RAS quality assurance establishes the reliability criteria that guide the design policy for redundant equipment (a responsibility of configuration management), and the policies of other function groups in this area.

390

•

•

• •

•

CHAPTER 14 Optical performance monitoring in optical transport networks

Alarm surveillance. A TMN provides the capability to monitor failures of network elements (NE) in near-real time. When such a failure occurs, an indication is made available by the NE. Based on this, a TMN determines the nature and severity of the fault. For example, it may determine the effect of the fault on the services supported by the faulty equipment. Fault localization. Where the initial failure information is insufficient for fault localization it has to be augmented with information obtained by additional failure localization routines. The routines can employ internal or external test systems and can be controlled by a TMN (see ITU-T M.20).6 Fault correction. Fault correction transfers data concerning the repair of a fault and for the control of procedures that use redundant resources to replace equipment or facilities that have failed. Testing. Testing can be carried out in one of two ways. In one case, a TMN directs a given NE to carry out analysis of circuit or equipment characteristics. Processing is executed entirely within the NE and the results are automatically reported to the TMN, either immediately or on a delayed basis. Another method is where the analysis is carried out within the TMN. In this case, the TMN merely requests that the NE provide access to the circuit or equipment of interest and no other messages are exchanged with the NE (see ITU-T Rec. G.7710/Y.1701).7 Trouble administration. Trouble administration transfers trouble reports originated by customers and trouble tickets originated by proactive failure detection checks. It supports action to investigate and clear the trouble and provides access to the status of services and the progress in clearing each trouble.

14.3.3.3 Performance management Performance management provides functions to evaluate and report upon the behavior of telecommunication equipment and the effectiveness of the network, or NE. Its role is to gather and analyze statistical data for the purpose of monitoring and correcting the behavior and effectiveness of the network, NEs, or other equipment, and to aid in planning, provisioning, maintenance, and the measurement of quality. As such, it is carrying out the performance measurement phase of ITU-T M.20.

14.3.4 Transport functions The transport plane in optical networks is organized as a stack of layer networks with a client-server relation between adjacent layers. Each layer network offers services to its client layers and in turn uses services from the underlying server layer. A layer network provides its clients with a transport service using a layer-specific information structure known as the characteristic information (CI).

14.4 MODELING OF MULTILAYER NETWORKS The following section introduces the generic principles and the terminology defined in ITU-Rec. G.805 for modeling the functionality of multilayer transport networks. Table 14.1 is a summary of the different types of architectural components used in G.805, and Figure 14.4 is an example how these components are used recursively for modeling the layering and partitioning of a multilayer network.

14.4 Modeling of multilayer networks

391

Table 14.1 G.805 Architectural Components Architectural Components Topological components Access group Layer network Subnetwork Link

Reference points

Transport entities

Access point (AP) Connection point (CP) Termination connection point (TCP)

Trail Connections:

Transport processing functions Trail termination function Adaptation function

– Network connection (NC) – Subnetwork connection (SNC) – Link connection (LC)

Adaptation functions are used for mapping client-layer characteristic information into a common server layer format so that it can be transported over a trail in the server-layer network. From a transport network functional viewpoint, therefore, the adaptation function falls between the layer networks. Each layer network can be separately partitioned to reflect its topology or management structure (Figure 14.4).

Trail

AP

AP Trail termination

Network connection TCP Client layer

CP

Link connection

TCP

SNC Client-to-server adaptation

Server layer

AP

AP

Trail

CP

CP

TCP LC

LC

TCP SNC

Server-layer trails = client-layer links

FIGURE 14.4 Modeling of layer networks.

Trail termination

392

CHAPTER 14 Optical performance monitoring in optical transport networks

14.4.1 Application of partitioning concept The partitioning concept is important as a framework for defining: • • • • •

The network structure within a layer network. Administrative boundaries between network operators jointly providing connections within a single-layer network. Domain boundaries within a layer network of a single operator to allow the apportioning of performance objectives to the architectural components. Routing domain boundaries within the layer network of a single operator. The part of a layer network or subnetwork that is under the control of a third party for routing purposes (e.g., customer network management).

14.4.2 Application of the layering concept A layer network provides its clients with a transport service, and the operation of the layer network is independent of whatever client happens to use the layer network. Information that passes between any client to the layer network is common to all clients and is the minimum needed to be consistent with the definition of the transport service offered. The operations within a layer network that are independent of its clients include the control of forwarding, resource reservation, traffic demerging, and the OAM and recovery of the transport service. All of these operations are internal to a layer network. By definition, a layer network does not rely on any client information to perform these operations, and therefore all information required to perform these operations is independent of whatever client is using the layer network. The layering concept of the transport network allows: • • • • •

Each layer network to be described using similar functions. Independent design and operation of each layer network. Each layer network to have its own operations, and diagnostic and automatic failure recovery capability. The possibility of adding or modifying a layer network without affecting other layer networks from the architectural viewpoint. Simple modeling of networks that contain multiple transport technologies.

14.4.3 Transport entities: trails and connections Transport entity is a generic term used for describing transport network functionality. Transport entities provide transparent information transfer between reference points within a layer network. Examples of reference points are connection points (CPs) or termination connection points (TCPs). There is no change in information between input and output other than that resulting from degradation in the transfer process. Two basic entities are distinguished according to whether or not the information transferred is monitored for integrity. Entities without monitoring are termed connections, and monitored entities are termed trails. Trails are formed by combining trail termination functions and a network connection. Connections are further categorized into network, subnetwork, and link connections.

14.5 Optical transport network–layered structure

393

14.4.4 Characteristic information A signal with a layer-specific format is transferred on “network connections.” The CI consists of the adapted client CI plus overhead.

14.5 OPTICAL TRANSPORT NETWORK–LAYERED STRUCTURE In this chapter the term OTN system is used to describe devices that are compliant with the requirements specified in the ITU-T Recommendations G.872, G.709, G.798, G.874, and G.874.1.8,9,10

14.5.1 OTN-layer networks The optical domain of the OTN is based on DWDM multiplexing technology, but provides standardized interfaces and methods to supervise and manage the network. The layered structure of OTN is shown in Figure 14.5. It comprises three layer networks: optical channel, optical multiplex section, and optical transmission section.

14.5.1.1 Optical transmission section–layer network This layer network provides functionality for transmission of optical signals on optical media of various types (e.g., G.652, G.653, and G.655 fiber).

Client

Client

Client layer OTN OH G.709 Digital wrapper

OH OH

Client

Optical payload unit Optical data unit

Optical channel payload

FEC

OTU optical transport unit

OSC Optical channels

l1 l2 OCh OH

…

lN Optical multiplex section

OMS OH OTS OH

FIGURE 14.5 OTN information structure.

Optical transmission section

Optical transport module

394

CHAPTER 14 Optical performance monitoring in optical transport networks

14.5.1.2 Optical multiplex section–layer network This layer network provides functionality for networking of a multiwavelength optical signal. Note that a “multiwavelength” signal includes the case of just one optical channel.

14.5.1.3 Optical channel–layer network The optical channel (OCh) is an information structure consisting of the information payload (OCh_PLD) with a certain bandwidth and nonassociated overhead (OCh_OH) for management of the optical channel. Nonassociated overhead means the overhead information is transported on a physically separate channel. The OCh-layer network provides end-to-end transmission of optical channels for transparently conveying client information of varying format (e.g., SDH STM-N, gigabit Ethernet, cell-based ATM, etc.), including optical channel connection rearrangement for flexible network routing.

14.5.2 Layer management Each layer network has its own management functions, including: • •

•

Monitoring processes for ensuring integrity of the adapted client information. Processing of the layer-specific overhead information, which is transported either together with the adapted information (associated overhead) or on a physically separate link (nonassociated overhead). Operations, administrations, and maintenance processes for enabling network-level functions, such as connection provisioning, QoS parameter exchange, and network survivability.

14.5.3 OTN information structure The OTN hierarchy is based on the OCh. The OCh payload is modulated onto an OCh carrier (OCC) that has a defined wavelength li, which is chosen from the ITU-T DWDM grid.

14.5.3.1 The G.709 digital wrapper Current optical technology does not provide all of the functionality required for the management and maintenance of optical channels in a manner that is independent of the digital client layer. This is overcome by the use of the optical transport unit (OTU) defined in ITU-T Rec. G.709 as a common encapsulation technique for digital clients (digital wrapper) and as a mechanism for supporting aspects of management and maintenance information that cannot currently be supported by optical channel trail termination functions. The optical channel transport unit (OTU) is the highest electrical multiplexing level, and includes a forward error correction (FEC) function. It is applied between points in the network where the signals are converted from the optical to the electrical domain. The optical channel data unit (ODU) carries an optical channel payload unit (OPU). The client signals are mapped into the OPU, which is not processed at intermediate sites of a transport network. The ODU signals are defined as the end-to-end networking entities, comparable to the former SDH virtual containers. The OTUk bit rates defined today are 2.7, 10.7, 43, and 112 Gbps. The frame structure contains 4
4080 bytes regardless of the bit rate. This implies that the frame duration is not constant like in SDH, but varies with the bit rate of the actual signal.

14.6 OTN services

395

The G.709 digital wrapper specifies the interfaces for interconnection between service providers and network operators, and facilitates midspan meeting between equipment from different vendors. As such, the OTU layer network is the recommended client-layer network of the optical channel, enabling a managed network at a wavelength granularity. The benefits are threefold: • •

•

First, the unified encapsulation scheme is protocol-agnostic, allowing carriers to manage native data services. Second, the ODUk tandem connection monitoring enables performance management and fault isolation across multiple domains. This allows each service providers to maintain their own SLAs and fault isolation procedures over several carrier networks. Third, the FEC allows increasing the span length between 3R regeneration points.

In summary, a network architecture built on the G.709 hierarchy has many advantages regarding operation and maintenance. However, these benefits can only be used if the OTU-OH is terminated, which means OEO conversion, which in turn conflicts with the vision of a true all-optical network!

14.5.3.2 Optical transport module (OTM-n[r].m): G.872 The OTM is the information structure that is transported across an ONNI. The indexes n and m define the number of supported wavelengths and bit rates at the interface. Two OTM structures are defined: • •

OTM with full functionality (OTM-n.m) OTM with reduced functionality (OTM-0.m and OTM-nr.m)

The OTM-n.m consists of up to n multiplexed optical channels and an OTM overhead signal to support the nonassociated overhead. The OTM-0.m consists of a single optical channel without a specific color assigned. The OTM-nr.m consists of up to n multiplexed optical channels. Nonassociated overhead is not supported.

14.5.3.3 Optical supervisory channel A common optical supervisory channel (OSC) is used to transport the nonassociated overhead information of each optical channel as well as the OMS- and OTS-layer network overhead, but also provides a transport medium for maintenance signals and other management data.

14.6 OTN SERVICES Different types of OTN services can be easily mapped onto the OTN-layer model, since each layer offers well-defined transport capabilities to its clients. Table 14.2 is an overview on the different service types and the layers that provide these services.

14.6.1 All-Optical Networks The original approach in OTN was based on the G.709 digital wrapper concept with 3R regeneration at each connection point, but with the introduction of optical amplifiers and reconfigurable optical add/drop multiplexers (OADMs), there is an increasing interest in all-optical networking applications over long transparent lightpaths with no OEO conversion at the intermediate connection points.

396

CHAPTER 14 Optical performance monitoring in optical transport networks

Table 14.2 OTN Services Service Type/Options

Characteristic Information

Client Access

Dark fiber

Optical transport module:

OTS/OMS layer

• Protected • Unprotected Leased wavelength

• Protected • Unprotected • On demand via control plane Optical transport service

• Protection • Tandem connection monitoring • FEC • On demand

• Optical multiplex of order n • Optical supervisory channel • Optical signal of defined maximum

OCh layer

bandwidth and signal-to-noise ratio

• Data stream that constitutes nonassociated-channel overhead

• ODU payload area for transport of

ODUk layer

digital client signals

• ODU overhead for transport of associated overhead

Photonic crossconnects and OADMs are available today, with capacity, space requirements, power consumption, reliability, and cost suitable for use in telecommunication networks. With this evolution, it is possible to build large all-optical networks (AONs) with long, transparent spans between 3R regenerators. This transition from the opaque networks used today toward AON brings many new challenges with regards to network planning, lightpath provisioning, and service assurance, which also have significant impact on testing.

14.7 TEST AND MEASUREMENT TASKS IN OPTICAL NETWORKING The purpose of optical monitoring is to provide configuration, fault, and performance management processes with reliable and timely status information about the transport entities and network resources at each layer in an OTN. Lightpath provisioning and service assurance are the two broad application classes that rely heavily on testing. A third application is the test of the physical medium—that is, the optical fiber itself.

14.7.1 Lightpath provisioning The following steps are necessary in order to fulfill the request for a new network connection, regardless of whether the provisioning process is done manually, via a centralized TMN, or by means of a distributed control plane: • • • • •

Route selection Degradation estimate Connection setup Test Connect user

14.7 Test and measurement tasks in optical networking

397

Depending on link length, channel spacing, supported bit rates, and so on, the effort for the degradation estimate increases. For instance, in a small network it is likely that any path selected will be within the allowable range of impairments—that is, for purposes of path computation, the impairments may be neglected. For larger networks, there are two options: •

•

Perform a detailed calculation of the impairments and optimize the power levels in the network. The computations involved in this option require detailed knowledge of the characteristics of the optical network elements in the path and sophisticated simulation programs, which are quite time consuming. Perform a “simple” calculation of impairments and apply conservative rules to ensure a low probability of unacceptable impairments due to estimation errors. This may result in too many rejected connection requests, which means lost revenue for the transport provider. Less conservative rules increase the probability that a connection is accepted, but also increase the probability that the impairments are actually out of limits.

Recommendation G.680 deals with calculating the effect of cascading multiple optical network elements (ONEs) on the degradation of the optical signal quality in order to decide whether the overall degradation is compatible with the error performance objectives at the trail termination point. The calculation of the overall degradation of an optical channel route is based on the degradation function of each ONE involved. The degradation functions in ITU-T G.680 are defined for every class of ONEs, such as PXCs and OADMs.11 For instance, the table in Figure 14.6 lists the parameters that should be considered when modeling the transfer characteristic of an OADM with amplifiers: For cascaded ONEs, estimation rules follow: 1. Degradation of each parameter in isolation from the others. 2. Combined effect of all of the impairments on the overall performance objectives that must be met at the trail termination point. From the above, it is clear that any degradation estimate is afflicted with uncertainty due to the approximations made in the underlying simulation models and the uncertainty in the parameter values of the ONEs. This creates a need for two types of test applications: connection verification and characterization of network elements.

14.7.1.1 Connection verification In many cases, verifying the connection quality before connecting a user is mandatory. At minimum, the parameters that are guaranteed in the SLA must be validated. ITU-T Rec. M.2401 defines the exact error performance objectives, limits, and procedures for bringing into service (BIS) and maintenance of multioperator international ODUk paths and OTUk sections in OTN following the G.709 standard.12 M.2401 is based on error performance parameters that are derived from BIP and FEC calculations, but this is not applicable for transparent OTNs, which do not use the digital wrapper. In these cases, only optical signal quality parameters like the optical signal-to-noise ratio (OSNR) or Q-factor can be used. As a side effect of this connection verification, the measurement results can also be used to improve the estimation process.

398

CHAPTER 14 Optical performance monitoring in optical transport networks

Parameter Channel frequency range

Wavelengths not dropped

Channel insertion loss Input to output Input to drop Add to output

dB dB dB

Channel insertion loss deviation

dB

Ripple

Input

Output

Drop

Add

Unit GHz

Channel chromatic dispersion

dB ps/nm

Channel differential group delay (DGD) Input to output Input to drop Add to output

ps ps ps

Channel polarization-dependent loss (PDL) Input to output Input to drop Add to output

dB dB dB

Reflectance

dB

Adjacent channel isolation input to drop

dB

Nonadjacent channel isolation input to drop

dB

Channel extinction input to output

dB

Reconfigure time

ms

Channel uniformity

dB

FIGURE 14.6 Transfer parameters of reconfigurable OADM without amplifiers.

14.7.1.2 Characterization of network elements A second test application in this context is the characterization of individual network elements to feed the database for the link simulation with accurate parameter values. These measurements should be repeated at regular intervals to address aging and drift. This type of test application requires measurement of a broad range of optical parameters. This is usually done by setting the NE or link remotely into a special test mode and by connecting high-end, portable test equipment that is operated locally by qualified technicians. This combination of TMN control and local operation is called “operator-assisted testing.” This is an economic approach in cases where changes are infrequent or slow, thus allowing expensive test equipment to be shared among several locations.

14.7.2 Service assurance Service assurance is the process responsible for executing maintenance activities for customer satisfaction. Network status and performance are constantly monitored to detect possible failures. Performance data are collected and analyzed to identify potential problems and fix them without impacting on the customer side. This process manages SLAs and reports service performance levels to the customer. It involves receiving problem reports from customers, informing customers of problem status, and ensuring restoration and repair. It ensures that customers are being adequately serviced as per committed QoS.

14.7 Test and measurement tasks in optical networking

399

14.7.2.1 The “five nines” Large transparent spans that are operated close to physical limits to meet the demand for higher channel bit rates, narrow channel spacing in DWDM systems, and dynamically changing network configuration in ROADM-based networks make it increasingly difficult for network operators to guarantee the famous “five nines”—that is, 99.999% availability to clients of optical transport services. Such availability means downtime of only 5 minutes in 1 year. Two different maintenance approaches exist to reach the “five nines”: • •

Reactive maintenance relies on failure detection and automatic protection switching in near-real time. Proactive maintenance aims at detecting faults before the client layer is affected.

14.7.2.2 Reactive maintenance Figure 14.7 illustrates a typical maintenance cycle based on the reactive model. The process is based on monitoring devices that are embedded in the network elements as part of the trail termination or adaptation functions. If a critical defect is detected, the layer management in the NE is informed

Notifications

Test management Queries OSC Mgt I/F

Events

Digital transmission analyzer

Real-time actions

Monitor

Layer mgt

Embedded monitoring

FIGURE 14.7 TMN-integrated maintenance process.

Q-factor

Optical spectral analyzer

400

CHAPTER 14 Optical performance monitoring in optical transport networks

immediately, which then responds in near-real time with corrective actions, such as protection switching or shutdown of the defective link, following a preprogrammed reaction pattern. After completion of these near-real-time actions, notifications are sent to the network management system with an update of the system status. As a next step, the network management system (NMS) retrieves the current network configuration from the configuration database and starts the TMN-assisted fault isolation process by querying the MIBs of the NEs in the defective link for any abnormal events, such as loss of signal, excessive laser temperature, and so on. In a second step, test management functions are used to initiate equipment self-tests or loop-back tests from a central location. If this remote procedure is not successful, a service technician is dispatched who continues the search using high-end portable test equipment. Often it is necessary to put the network equipment into a specific test mode, which is done either remotely by the NMS or by the technician using a local craft terminal. Once the fault is identified, the repair process is initiated, and after completion, the link is set back to its normal mode.

14.7.2.3 Proactive maintenance Proactive maintenance relies on trend analysis of characteristic performance parameters that are observed at regular intervals over a long period of time. Repair is initiated as soon as negative trend is visible, normally long before the client layer is affected.

14.8 OPTICAL PERFORMANCE MONITORING Transport entities are fundamental “managed objects” in network management. The purpose of optical monitoring is to provide configuration, fault, and performance management applications with reliable and timely status information about the transport entities at each layer in an OTN. For that purpose, the termination points of each network layer are equipped with simple but efficient supervision functions that continuously measure the signal attributes that characterize the service offered by that layer. Two basic supervision tasks are defined in ITU-T Rec. G.872, namely connection supervision (including connectivity and continuity supervision) and signal quality supervision.2 Connectivity supervision monitors the integrity of the routing of the trail between sink and source by analyzing a unique trail trace identifier (TTI), which is transported in the frame overhead of the OTUk or in the OSC. Continuity supervision checks the continuity of a trail by monitoring the presence/absence of the CI using simple monitoring devices that are usually implemented in the trail termination functions. Signal quality supervision in general monitors the performance of a trail. If the performance falls below a certain threshold, this might activate a defect indication. The monitoring of an optical network can be performed following different approaches. In SDH networks and OTNs, specific bytes are dedicated to carry and provide information on the quality of the transport layer. In these cases, accessing and retrieving such information would require conversion of the optical signal back to the electrical domain and processing of the frame at that level. In contrast, optical monitoring provides the function of monitoring the optical signal directly without processing the electrical frame.

14.8 Optical performance monitoring

401

14.8.1 Optical-layer signal quality supervision requirements Either embedded or external monitoring equipment may be used to measure the optical signal quality parameters defined in ITU-T Rec. G.69713 at the OTS, OMS, and OCh layers.13,3 These parameters are defined from a client perspective—that is, they may represent the combined effect of several server-layer impairments in a single comprehensive metric (e.g., Q-factor or OSNR) that allows a simple estimate of the expected client-layer performance. An ideal optical monitoring method should meet the following criteria: • • • • • • • •

The primary goal is to provide meaningful and precise performance data for each sublayer. In-service measurement is mandatory. No overhead in form of additional payload bits, additional monitoring channels on a separate wavelength, or pilot signals modulated onto the OCh payload should be required. The measurement method should be independent of the client’s bit rate or frame format. The measurement method should have no impact on the payload signal—that is, low power drain and no additional signal delay. Short measurement time is essential, in particular when fast reactions to changes in the network status are required. Integration into a network management system is essential. Preferably the monitoring devices should be part of the network elements and controllable via the element manager. Last, and most important, the cost of monitoring equipment must be extremely low.

The optical signal quality of a fiber link is determined by a large number of physical effects. Some of these have been identified/defined by Q.6 in Table 14.1 of G.697 as follows: • • • • • • • • • • • • • • • •

Attenuation Optical channel power changes due to gain variations Frequency (or wavelength) deviation from nominal Polarization mode dispersion (PMD) (first and higher orders) Four-wave mixing (FWM) Amplified spontaneous emission (ASE) noise in OA Chromatic dispersion Chromatic dispersion slope Reflections Laser noise Interchannel crosstalk Interferometric crosstalk Crossphase modulation (XPM) Self-phase modulation (SPM) Stimulated Brillouin scattering (SBS) Stimulated Raman scattering (SRS)

These effects have to be carefully balanced when dimensioning an optical connection. As mentioned in Section 14.7.1, measurement of all these parameters requires a large range of specialized and expensive test equipment, which collides with the cost barrier for optical-layer supervision equipment.

402

CHAPTER 14 Optical performance monitoring in optical transport networks

OSNR Q-factor Attenuation

BIP, FEC

ASE Noise FWM

Chromatic dispersion

XPM PMD

SPM Brillouin scattering

BER

Raman scattering

FIGURE 14.8 OTN performance metrics.

On the other side, a client of an optical transport service is mainly interested in a concise set of optical performance metrics that represent the cumulative effect of these physical impairments and that allows a reliable estimate of the bit error rate (BER) that can be expected for the data application (Figure 14.8). Based on the status of the current technology and associated cost, G.697 has identified a subset of five optical parameters: total power, channel power, channel wavelength, OSNR, and Q-factor. These parameters complement the BIP8 and FEC mechanisms available at the OTU layer and improve the OTN supervision capabilities at the OCh, OMS, and OTS layers (Table 14.3). Table 14.3 OTN Supervision Connection Supervision Layer Optical channel

G.709 wrapper

Connectivity

Continuity

TTI in OTUk OH

Loss of signal Loss of frame Channel power

TTI transported in OCh overhead in the OSC

OMS

Not required (one-to-one relation with OTS)

Total power

OTS

TTI transported in OTS overhead in the OSC

Total power OSC alarms

Monitoring type

Embedded

Signal Quality Supervision BIP8 FEC Channel power Channel wavelength OSNR Q-factor Spectral distribution gain tilt

Embedded or external

14.8 Optical performance monitoring

403

The termination points of each network layer in an OTN are equipped with simple but efficient performance monitoring functions that continuously measure the quality parameters that characterize the service offered by that layer.

14.8.2 Optical power Optical power is the fundamental quantity in OTN surveillance. Optical power monitors can be found at every port of optical amplifiers, optical add-drop multiplexers, and transceivers, and are accessible through the element managers controlling these network elements.

14.8.3 Channel wavelength Monitoring of the channel carrier wavelength is mainly required in older WDM systems where no wavelength lockers are installed. Channel wavelength monitoring can be done either by an optical spectral analyzer, which is connected to the fiber with a tap coupler, or by channel monitors, which are embedded in optical network elements.

14.8.4 OSNR OSNR measurement currently uses the principle of measuring the noise between channels in order to estimate the noise at the channel wavelength (see Figure 14.9). This method works well for simple point-to-point systems with nothing but fiber and amplifiers in the optical path. For more complex DWDM systems, however, the introduction of any element that causes shaping of the noise between channels renders this method inaccurate.14 Noise shaping can occur when optical channels with different individual OSNR are multiplexed onto one common fiber without prior 3R regeneration (Figure 14.10). In that case, the noise power measured between two adjacent channels is not representative for the true OSNR in the passband of these two channels, and does not allow a reliable prediction of the BER in each channel.

Power

Pi + Ni

Ni N(li –Δl)

N(li + Δl) Wavelength

FIGURE 14.9 OSNR measurement method.

404

CHAPTER 14 Optical performance monitoring in optical transport networks

Add OADM Drop

(a)

Signal

“Noise” (b)

FIGURE 14.10 Noise shaping. (a) Various noise floors caused by OADM. (b) OSNR errors caused by various noise floors.

For a realistic OSNR measurement in the presence of noise shaping, it is essential to measure the filtered noise value in the passband of the optical filters in a system (often called “in-band” OSNR measurement). The power spectral density (PSD) in the passband of a filter is the superposition of the noise PSD N(l) and the signal PSD S(l) (Figure 14.11). Measurement of the OSNR requires, therefore, some method to separate the signal components from the noise. The methods used today exploit differences between signal and noise with regard to spectral distribution or state of polarization (SOP).

14.8 Optical performance monitoring

405

Pt = ∫ [S (l) + N (l)]d l B1

S(l) N2 = ∫ N (l)d l B2 N(l)

B1

B2

FIGURE 14.11 Power spectral density.

14.8.4.1 OSNR measurement with a narrowband spectral analyzer Measurement of in-band OSNR by spectral analysis is only possible if a frequency band B2 exists where only the noise is present. The OSNR in the frequency interval B1 is given by OSNR½dB ¼ 10  log P1 =N1 ; where P1 is the signal power and N1 is the noise power in this interval. Assuming that N(l) is flat in the passband of the optical filter, the noise power N1 can be estimated from the noise power N2 measured in a frequency interval B2 where no signal components are present. N2 is given by ð ð NðlÞdl Nw dl ¼ Nw  B2 ; N2 ¼ B2

B2

where Nw is the power spectral density of the noise. With Nw ¼ N2 =B2 ; the noise power N1 is given by N1 ¼ N2 

B1 : B2

406

CHAPTER 14 Optical performance monitoring in optical transport networks

The signal power P1 can be calculated from the total power Pt measured in the frequency interval B1 as P1 ¼ Pt  N1 ¼ Pt  N2 

B1 : B2

The OSNR in the frequency interval B1 is then given by OSNR½dB ¼ 10  log P1 =N1 ¼ 10  log P1 =N2 þ 10  log B2 =B1 : The total power Pt in the frequency interval B1 and the noise power N2 in the frequency interval B2 can be measured, for example, with an optical power meter combined with appropriate weighting filters with noise-equivalent bandwidths B1 and B2 or by numerical integration of the power spectral density measured with a high-resolution optical spectral analyzer. The interval B1 must be chosen to be large enough to capture the full signal spectrum. The International Electrotechnical Commission (IEC)1 recommends a reference bandwidth of 0.1 nm, which is sufficient for narrowband signals but is too low for signals with bit rates of 10 Gb/s or higher. At those higher bit rates, the signal spectrum occupies the full channel bandwidth and completely overlaps with the noise floor, which requires a different measurement principle, as illustrated in Figure 14.12.

14.8.4.2 OSNR measurement by polarization extinction The polarization extinction method15 makes use of the fact that—in a first approximation—the optical transmission signal is polarized, whereas the ASE noise is unpolarized. In its simplest form, a combination of a variable polarization controller and a polarization splitter/filter is used to separate the polarized signal from the unpolarized noise. By variation of the polarization controller in front of the PBS, it is possible to suppress the polarized signal and get access to the nonpolarized in-band noise at one branch, where the other branch shows the signal plus noise.

½ ASE

In-band OSNR DSP

Signal + ASE Polarization controller

Polarizing beam splitter Signal + ½ ASE

FIGURE 14.12 In-band OSNR measurement with polarization extinction method.

14.8 Optical performance monitoring

407

One technical problem of the polarization extinction measurements is to make sure that the SOP is stable during the measurements. Moreover, the SOP of the optical signal can change rapidly over frequency due to PMD and to nonlinear birefringence. Therefore, a high suppression is not possible over a larger optical bandwidth. State-of-the-art OSNR measurement instruments use short measurement intervals and wavelength-selective measurements with narrow bandwidth to overcome these problems.16–18 Polarization crosstalk caused by imperfections in the optical front end of the instrument can be reduced by measuring the OSNR not at the center wavelength, but rather, “off-center” in a frequency band where the signal energy is lower. Again, this requires a tunable narrowband selective filter, preferably an optical spectral analyzer. In a polarization multiplex signal, there is a signal on each of the two orthogonal polarizations. Regardless of the polarization controller setting, there is always a linear combination of both signals at each output of PBS, making it impossible to isolate the noise signal.

14.8.5 Q-factor measurement The Q-factor measurement forms a bridge between the classical optical parameters (power, OSNR, and wavelength) and the digital end-to-end performance parameters. The Q-factor is measured in the time domain by analyzing the pulse shape of the digital OCh payload signal, which is not possible with simple optical parameters alone. Q-factor is a comprehensive measure for the signal quality of an optical channel that takes into account the effects of noise, crosstalk, filtering, and linear/nonlinear distortions on the pulse shape, which is not possible with spectral analysis alone. The Q-factor measurement is nonintrusive, independent of bit rate and data format, and can be applied to any kind of digital client signal, regardless of whether a “digital wrapper” is used or not.19

14.8.5.1 Q-factor theory Each digital receiver has a decision circuit that decides whether an incoming binary signal is at the logical 0 level or at the logical 1 level by sampling the received signal and comparing the sample value to a threshold Vth. In a noise-free system, the received signal would have only two states, namely 2 or 0. In a real system, these two signal levels can vary considerably due to additive noise and the linear and nonlinear distortions caused by the transmission medium or line equipment. This means that the sampled receive signal must be regarded as a random variable with probability distributions px(xj0) and px (xj1), respectively, depending on whether a 0 or a 1 was transmitted. Each of the two random variables has a mean value m and a variance s. Even with optimal threshold settings, there is a certain probability P(1j0) for detecting a 1, although a 0 was transmitted and a certain probability P(0j1) for detecting a 0 although a 1, was transmitted (Figure 14.13). The total BER is given as BER ¼ Pð0 j 1Þ  Pð0Þ þ Pð1 j 0Þ  Pð1Þ; where Pð0 j 1Þ ¼

ð Vth 1

px ðx j 1Þ dx

408

CHAPTER 14 Optical performance monitoring in optical transport networks

P(1I0)

1

m1,s1

Decision threshold 0

m0,s0

Sampling point P(0I1)

FIGURE 14.13 Eye diagram and amplitude histograms.

and

1 ð

px ðx j 0Þ dx:

Pð1 j 0Þ ¼ Vth

P(1) and P(0) are the a priori probabilities of a 1 and a 0 respectively.

14.8.5.2 The Q-factor definition The Q-factor is a concise performance metric for fast characterization of optical digital transmission systems. The Q-factor is defined as m  m0 Q¼ 1 ; s1 þ s0 where m0 and m1 are the mean values and s0 and s1 are the variances of the probability density functions px(xj0) and px(xj1). Q can interpreted as a signal-to-noise ratio (SNR) where the signal energy increases with the distance between m0 and m1 and the noise energy increases when s0 and/or s1 are increased.

14.8.5.3 Relation between Q and BER As shown in Figure 14.14, the BER is a function of the SNR (and therefore of Q), since the overlap between the two probability density functions (PDFs) decreases when signal energy is increased (i.e., the curves are moved farther apart) and the overlap region increases when the noise level is increased (i.e., when the curves become broader). However, there is no straightforward relation between BER and Q, since BER depends on the actual shape of the PDFs px(xj0) and px(xj1), whereas Q only uses the first and second moment of the PDF (m and s), which in general are not sufficient to define the PDFs completely.   1 jm1  m0 j 1 ¼ erfcðQÞ PError ðVopt Þ ¼ erf 2 s0 þ s1 2 Unfortunately, there is no such straightforward relation when the dominant impairments are caused by other effects, such as intersymbol interference, chromatic dispersion, or crosstalk. In those cases, there may be significant deviations from the Gaussian PDF.

14.8 Optical performance monitoring

m0,s0

409

px(x|0) px(x|1)

P(0I1)

m1,s1 P(1I0)

x 0

V th

1

FIGURE 14.14 Error probabilities.

Under ideal conditions (only additive Gaussian noise, no linear or nonlinear distortions, etc.), the BER of a binary optical channel should be the same as that indicated by a Q-factor measurement. However, these idealized conditions are rarely present in real systems and the correlation between the Q-factor of an optical signal and the BER measured after regeneration is influenced by the different receiver characteristics (noise bandwidth, impulse response, etc.) in the regenerator compared to that of the Q-factor meter. An additional factor that has a serious effect on the validity of a Q-factor measurement at any point in an optical path is the residual dispersion present at that point. However, the Q-factor is used to characterize OChs; therefore, a Q-factor measurement is most relevant at termination points of OChs (points X and Y), which are normally dispersion compensated. Q-factor measurements at intermediate points of an optical path are only possible with proper dispersion compensation at these points. Figure 14.15 shows the effect of chromatic dispersion on a 10-Gb/s binary signal. Dispersion may cause significant intersymbol interference, which can significantly increase the BER of the binary client signal. This degradation of the binary signal cannot be detected with an OSA alone (Figures 14.15(c) and (d)). A digital oscilloscope can give at least a qualitative assessment, while a Q-factor analyzer gives quantitative measure for the signal quality of a binary signal in the form of a Q-factor value and a BER estimate (Figures 14.15(e) and (f)).

14.8.6 OTUk, ODUkT, and ODUkP signal quality supervision A BIP8 is used for each of these layers as defined in G.709. At the source side, the BIP8 shall be computed over the OPUk frame (columns 15–3824). The computed BIP8 is inserted in the BIP8 byte position of the relevant overhead field of the second following frame, as shown in Figure 14.15(a). At the sink side, the BIP8 is computed over the OPUk (columns 15–3824 of the frame). The BIP8 value generated at the source shall be extracted from the BIP8 byte position of the relevant overhead field. The computed BIP8 value of the second preceding frame is compared with the BIP8 value extracted from the current frame, as shown in Figure 14.15(b). If there is a mismatch between the two values, a near-end errored block (nN_B) is detected and the number of BIP violations (nBIPV) is forwarded to the trail termination function at the source side (Figure 14.16). A further indication of the signal quality is provided by the number of bits corrected by the FEC at the OTUk layer.

410

CHAPTER 14 Optical performance monitoring in optical transport networks

Eye pattern without dispersion (a)

FIGURE 14.15 Influence of dispersion on 10-Gb/s binary signal. Without dispersion: (a), (c), and (e); with dispersion: (b), (d), and (f). Continued

14.8.7 What is missing? The digital monitoring techniques based on parity checks (BIP8) or on the evaluation of the FEC and the simple optical power monitors applied at the OTS, OMS, and OCh levels cover a large part of the standard supervision tasks, but they leave a gap at the OMS and the OCh layer when it comes to characterizing the power spectrum of a WDM signal and the signal quality of individual optical channels. High-resolution optical spectral analysis and Q-factor measurements can close this gap to a certain extent, but at a high cost compared to embedded monitoring devices, which forces network operators to find the best compromise between fault coverage and cost.

14.9 Implementation issues

411

Eye pattern with dispersion (b)

FIGURE 14.15—CONT’D

14.9 IMPLEMENTATION ISSUES 14.9.1 Accuracy requirements

The accuracy of embedded monitoring devices must be sufficiently high to provide meaningful input for automated management decisions. This can often be achieved with relatively low effort compared to a general-purpose test instrument, since in many cases only the deviations from a known nominal value are of interest, and the normal operating range of network elements is narrow. External monitoring equipment must have higher accuracy and a wider measurement range than embedded monitoring devices, since it must provide reliable absolute measurements over the full operating range of an optical transmission system. External monitoring equipment serves a different purpose than embedded monitoring equipment. It is used for measuring additional, more sophisticated performance parameters, or when the absolute value of certain performance parameters is required.

412

CHAPTER 14 Optical performance monitoring in optical transport networks

Optical spectrum without dispersion (c)

FIGURE 14.15—CONT’D

Main applications are the monitoring of SLAs, which may include application-specific performance parameters. Other applications are the location of hard-to-find failures that cannot be isolated by the embedded monitoring devices, as well as function tests and accurate parameter measurements during installation, line-up, or repair.

14.9.2 External versus embedded monitoring Monitoring equipment for rapidly changing optical parameters is normally embedded in each network element in the form of simple, low-cost level monitors at each optical interface. Slowly varying parameters can be monitored by external high-end equipment that may be switched via remote control between several optical monitoring points. Measurement intervals are determined empirically, based on long-term statistics (Figure 14.17).

14.9 Implementation issues

413

Optical spectrum with dispersion (d)

FIGURE 14.15—CONT’D

14.9.3 Monitoring points In contrast to embedded monitoring equipment, external monitoring instruments are usually not permanently installed, but rather connected on demand to critical network segments. They are mostly used in an interactive mode, and are often remote controlled from a network operation center. For cost reasons, expensive test equipment can be shared between several monitoring points, such as by using tap couplers and optical switches, as shown in Figure 14.18. Figure 14.18 shows an example of how embedded monitoring functions and remotely operated external test equipment can be integrated under a common fault and performance management system. The embedded monitoring devices are controlled via the element managers of the network elements, while the external equipment is normally controlled by a separate test manager that executes the test suites requested by the management system. The element managers are also used to

414

CHAPTER 14 Optical performance monitoring in optical transport networks

Q-factor without dispersion (e)

FIGURE 14.15—CONT’D

configure the NEs into a test mode for out-of-service tests. Element managers and the test manager are coordinated by the fault and performance management system, which is responsible for configuration of the test environment and for processing of the raw data provided by the embedded sensors and the external instruments. Monitoring points for external test equipment must be carefully planned, with regard to the penalties introduced by optical splitters and with regard to the dispersion compensation equipment (DCE) required for Q-factor measurements.

14.9.4 Recommended measurement interval The rate of change of the optical impairments that are observed determines the interval between two measurements. As an example, loss of signal due to a fiber break must be detected in a few milliseconds in order to accomplish protection switching within 50 ms. This means that the optical power level must be measured with a sampling period well below 50 ms.

14.9 Implementation issues

415

Q-factor with dispersion (f)

FIGURE 14.15—CONT’D

Slow degradations, for example, due to drift or aging, are normally detected by a trend analysis, which is updated in larger intervals (typically 15 min to 24 h). An example is BIP8 errors. If a digital wrapper (G.709) with FEC is used, the error performance may also be monitored by evaluating the number of bits corrected by the FEC.

14.9.5 Risk management aspects The decision on whether a certain type of monitoring procedure should be installed or not depends on the cost of the required equipment and the risk associated with the event that could be detected by that equipment. The risk is defined as follows: R ¼ total cost of failure event  likelihood of the event High-risk areas are generally those where many clients are affected by a failure. One famous— and frequent—example is the fiber optic cable being cut by an excavator. (Fortunately, the fault-finding procedure is extremely simple in this case: Just follow the fiber duct and look for a construction site.)

416

CHAPTER 14 Optical performance monitoring in optical transport networks

1

3284

14 15

1

Frame i

2

OPUk

3 4

BIP8

1

Frame i+1

2

OPUk

3 4 1

Frame i+2

2

OPUk

3 4 (a) 1

14 15

3284

1 2

Frame i−2 OPUk

3 4 BIP8

1 2

Frame i−1 OPUk

3

XOR

4 1 2

Frame i

# of BIP violations

OPUk

3

≥ 1?

4

(b)

Near-end errorred block

FIGURE 14.16 BIP8 processing. Transmitter (a), Receiver (b).

Open or dirty connectors are other examples of frequent man made errors. Fortunately, all these failures are easily detected by the optical power monitors, which are meanwhile standard in optical NEs. Failures in the physical medium such as bad splices or excessive fiber bending are normally found already by an acceptance test after installation, using an optical time-domain reflectometer (OTDR). Critical fibers can also be monitored in service with OTDRs working in the 1650-nm range outside the normal OTN wavelength range.

14.9 Implementation issues

Out of service

Performance monitors

In service

NE characterization Link characterization Repair

Connection monitoring Rapid changes

Sampling time << 50 ms

Slow changes

Permanent monitoring

Embedded, low-cost sensors

Total power

417

Channel power

Level monitors

Measurement on demand

External, high-end equipment

Channel wavelength

OSNR

High-end OSA

Signal quality

Q-factor meter sampling scope

FIGURE 14.17 Optical performance monitoring overview.

Other frequent causes of failure are misconfigurations of network elements or permuted connectors, which can be identified through the TTIs in the OTUk overhead or in the OSC (if implemented). Those easy-to-find defects probably account for >90% of the failure events. Failures after software upgrades or failures after inappropriate “repair” attempts are also common, but are more difficult to resolve. Often, an extensive, TMN-assisted search strategy is necessary.

14.9.6 Improved fault diagnostics by event correlation As stated previously, a large fraction of failures at the optical network layers can be detected with simple monitoring devices such as power monitors or DWDM channel monitors. Fault identification and fault location can be further improved by combining error symptoms reported from different sensors. This event correlation is often implemented as part of the adaptation or trail termination processes in the network element. Event correlation is an efficient method yielding comprehensive fault coverage with inexpensive monitoring devices (Figure 14.19).

CHAPTER 14 Optical performance monitoring in optical transport networks

Test manager

OQM lj

lk Frame analyzer

OSA with drop

Fault and performance management

418

Q-factor meter

Switch

...

OSC l1 ... ln DCE 1

E/O

M U X

TX

D E M U X

Tap

TX

O/E O/E RX RX

Embedded monitoring

FIGURE 14.18 Embedded monitoring and external test equipment. Fault coverage 100% OSNR, Q-factor… 80%

OCH power monitor + Digital wrapper or client-side monitoring + Intelligent event correlation 20%

FIGURE 14.19 Fault coverage versus cost for monitoring equipment.

Cost

Element manager

E/O

14.11 Summary

419

For instance, permanently installed high-end test equipment, such as OSAs or Q-factor testers, is a rare exception. This type of equipment is usually shared among several monitoring points as shown in Figure 14.16, and used only on demand.

14.10 FUTURE CHALLENGES Power monitors at the OTS, OMS, and OCh layers and the digital monitoring techniques based on parity checks (BIP8) and on the evaluation of the FEC applied at the ODUkP, ODUkT, and OTUk layers will certainly be the essential surveillance mechanisms in future optical transmission systems with bit rates in the 100-Gb/s range and advanced modulation formats such as dual-polarization QPSK or OFDM. However, Q-factor and in-band OSNR measurements in their current form are not applicable for these modulation formats. A modulation analyzer that could display the signal constellation would be of great interest, but due to the fact that there will be no standardized 100-G long-haul interface, it is difficult to find an economic solution for a 100-G signal analyzer that can support all the modulation formats currently under investigation. A coherent receiver with a flexibly programmable digital signal processor appears to be a suitable approach, but the feasibility with currently available components is uncertain. There is also a foreseeable lack of standardized test procedures for optical multilane interfaces, such as the 4
25–Gb/s WDM defined in G.709 Amendment 3.

14.11 SUMMARY Optical monitoring is an important complement to the monitoring techniques applied at the digital client layers of the optical-layer network. Optical monitoring equipment supports two main network management processes: •

Service assurance ○ By providing timely and reliable signal quality data for all transport entities in the optical-layer network • Lightpath provisioning ○ By providing network planning tools and configuration management applications with accurate parameter values of the optical network elements. ○ By validation of the error performance objectives of a link after provisioning (bringinginto-service test). Monitoring equipment for service assurance is normally embedded in each network element in the form of simple, low-cost power monitors at each optical interface. Together with the digital monitoring techniques at the OTU layer and the self-test functions in the network elements, this is often sufficient for signal quality supervision at the optical layer with high-fault coverage and a very low probability of false alarms. Table 14.4 compares the monitoring methods presented in the previous sections with regard to: • • • •

Suitability for signal quality measurements required by G.697 Generic requirements for optical monitoring equipment listed in Section 14.10.1 Implementation complexity Cost

420

Measurement Method

G.697 Parameters

Features

Total power Channel power Channel wavelength Q-factor OSNR BER estimate In-service measurement Measurement speed Optical power drain Overhead Client independence O/E conversion Clock recovery Frame decoding TMN integration Cost

Power Meter

Channel Monitor

OSA

Polarization Extinction

Q-Factor

Digital Wrapper

Yes Yes No No No No Yes High Low 0 Yes No No No Good Low

Yes Yes No No No No Yes High Low 0 Yes No No No Good Medium

Yes Yes Yes No (Yes) No Yes Medium Low 0 Yes No No No Possible High

Yes Yes No No Yes No Yes Medium Low 0 Yes No No No Possible High

Yes Yes No Yes Yes Possible Yes Medium Medium 0 Yes Yes (Yes) No Possible High

No No No No No Good Yes High 0 <10% Yes Yes Yes Yes Good Low

CHAPTER 14 Optical performance monitoring in optical transport networks

Table 14.4 Comparison of OTN Monitoring Methods

References

421

Permanently installed high-end monitoring equipment, like OSA’s Q-factor meters or remote fiber test systems, is usually found only on high-risk links such as undersea cables or high-capacity trunk lines. In contrast to service assurance applications, network equipment characterization and bringinginto-service tests are performed out of service using high-end test equipment that is not permanently installed but rather shared among various locations.

REFERENCES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19.

ITU-T Rec. G.709: Interfaces for the Optical Transport Network (OTN); March 2003. ITU-T Rec. G.872: Architecture of optical transport networks; Nov. 2001. ITU-T Rec. G.805: Generic functional architecture of transport networks; March 2000. ITU-T Rec. M.3010: Principles for a telecommunications management network; Feb. 2000. ITU-T Rec. M.3400: TMN management functions; Feb. 2000. ITU-T Rec. M.20: Maintenance philosophy for telecommunication networks; Oct. 1992. ITU-T Rec. G.7710/Y.1701: Common equipment management function requirements; July 2007. ITU-T Rec. G.798: Characteristics of optical transport network hierarchy equipment functional blocks Dec. 2006. ITU-T Rec. G.874: Management aspects of optical transport network elements; March 2008. ITU-T Rec. G.874.1: Optical transport network (OTN): protocol-neutral management information model for the network element view; Jan. 2002. ITU-T Rec. G.680: Physical transfer functions of optical network elements; July 2007. ITU-T Rec. M.2401: Error performance limits and procedures for bringing-into-service and maintenance of multi-operator international paths and sections within an optical transport network; Dec. 2003. ITU-T Rec. G.697: Optical monitoring for DWDM systems; Nov. 2009. IEC 61280-2- Ed. 2.0: Fibre optic communication subsystem test procedures—part 2-9: digital systems— optical signal-to noise ratio measurement for dense wavelength division multiplexed systems. Rasztovits-Wiech M, Danner M, Leeb WR. Optical signal-to-noise measurement in WDM networks using polarization extinction. In: ECOC 1998, Madrid Spain, Proc. 549D; 1998. Moench W, Larikova J. Measuring the optical signal-to-noise ratio in agile optical networks, In: OFC 2007, San Diego, CA., NWCl; 2007. http://www.teralinkcom.com/. http://www.jdsu.com/product-literature/mts-tb_osas_ds_fop_tm_ae.pdf. Kilper C, Bach R, Blumenthal DJ, Einstein D, Landolsi T, Ostar L, Preiss M, Willner AE. Optical performance monitoring. J. Lightwave Technol 2004;22:294.

This page intentionally left blank

CHAPTER

Optical performance monitoring in optical long-haul transmission systems

15 Xin Jiang

Department of Engineering Science and Physics, College of Staten Island, The City University of New York, Staten Island

15.1 INTRODUCTION In the past several decades, optical long-haul transmission systems have gained significant progress in terms of speed and capacity. Dense wavelength-division-multiplexed (WDM) systems with 10 Gb/s per channel have been widely deployed all over the world. 40-Gb/s systems are moving from lab demos and field trials to commercial deployments. The next generation of higher-speed optical systems operating at 100 Gb/s is under development and being standardized. Effectively monitoring these high-speed, long-haul transmissions has been challenging to system engineers and operators. Optical long-haul transmission systems refer to those that can transmit signals over fiber at long distance (>1000 km) without expensive optical-electrical-optical (OEO) regeneration in the middle of the link. The transmission distance can extend to several thousand kilometers. There is no explicit definition for the boundary of the long-haul transmission. But a long-haul system usually has to address system degradations due to long fiber transmission link. Many elements can impact optical signal transmission as well as system performance. The dominant impairments for long-haul transmission are broadband noise from fiber amplifiers, e.g., accumulated amplified spontaneous emission (ASE) and fiber dispersions, including chromatic dispersion (CD) and polarization mode dispersion (PMD). These elements will limit the transmission capacity and distance, and have to be optimized in system design. Fiber nonlinear effects, on the contrary, are generally detrimental to system performance and preferably should be minimized in design. Usually, these elements would worsen the signal deterioration when added up, but on occasion two elements can cancel each other and improve the system instead. A good example is to use local dispersion of fiber to combat transmission nonlinearity via dispersion map management.3 Thus, performance monitoring on these optical effects are essential to long-haul system operation. On top of that, all these physical impacts increase dramatically as the data rate increases. System operators have to depend on certain optical performance monitoring (OPM) tools to ensure that the system delivers the required performance. In addition, long-haul systems usually carry high-speed, high-bandwidth, and large-capacity data that demand very high reliability. As system speed and capacity increase to accommodate fast traffic growth, it is more important to incorporate effective OPM into the system in order to simplify system

© 2010 Elsevier Inc. All rights reserved. Doi: 10.1016/B978-0-12-374950-5.00015-8

423

424

CHAPTER 15 OPM in optical long-haul transmission systems

design, facilitate operations, enhance system performance, and reduce overall operational cost. OPMs as core physical-layer performance monitoring tools have also been used in system installation, provision, tune-up, fault management, and upgrade management. This chapter reviews the needs and technical requirements of OPM technologies from the long-haul system transmission design and operation point of view.

15.2 ELEMENTS OF A LONG-HAUL TRANSMISSION SYSTEM Today’s long-haul optical transmission system is built on several core technologies: advanced lowloss fibers, high-speed opto-electronic devices, fiber amplifiers, and WDM technologies. Figure 15.1 shows a typical point-to-point long-haul transmission link. It consists of a pair of unidirectional fiber links and the terminal equipments on two ends. The fiber links are made of a series of fiber spans and fiber amplifiers. The eastbound and westbound fiber amplifiers are colocated in the intermediate nodes. Local channels can be added or dropped at the intermediate nodes. For simplicity, the intermediate add-drop function is not shown in the figure. The terminal equipment is made of optical transmitters and receivers, as well as multiplexers and demultiplexers.

Fiber plant

Terminal node

Terminal node Intermediate nodes

Tx/Rx Tx/Rx

Tx/Rx Preamplifier

Power amplifier

Tx/Rx

Tx/Rx

Tx/Rx

Preamplifier

Inline amplifier

FIGURE 15.1 Typical bidirectional long-haul optical transmission link.

Inline amplifier

Power amplifier

15.2 Elements of a long-haul transmission system

425

Optical Transmitter. The long-haul optical transmitter uses a directly modulated laser or a laser with an external modulator to generate an optical modulated signal. Low-cost direct modulated lasers (DMLs) are widely used in 2.5-Gb/s systems, but few are used in 10-Gb/s systems due to high chirp. The most common optical transmitter design for long-haul systems is to use a narrowband DFB laser with an external chirp-free Mach-Zehnder modulator (MZM). Electro-absorption-modulated lasers are also widely used in 10-G systems. The separation of wavelength control from modulation control enables better transmitter optimization for high-speed systems beyond 10 Gb/s. The laser wavelength must be well controlled so that it can fit into the narrow frequency requirements of the dense WDM system. Modulation optimization is the key technology for long-haul transmission. Modulation waveform and frequency characteristics directly impact the transmission distance for high-speed optical signals. The most widely deployed formats are binary-intensity modulations due to their simplicity. While non-return-to-zero (NRZ) dominates in 2.5-G long-haul systems, other formats, such as returnto-zero (RZ) and duo-binary (DB), are widely used in 10-G systems since they can provide better performance in reducing dispersion distortion and intersymbol interference (ISI). Other modulation formats, such as DPSK and DQPSK, are adapted as the data rate increases to 40 G. Advanced modulation formats are seen as a promising direction to enable long-distance transmission for high-speed data of 40 Gb/s and 100 Gb/s. Optical Receiver. The long-haul optical receiver typically uses low-noise p-i-n photodiode with electronic transimpedance amplifier (TIA). The photodiode is a broadband optical power detector that converts optical signals into electronic signals. The TIA amplifies the weak signal to improve the sensitivity of the receiver. Both the optical filter bandwidth before the photodiode and the electronic filter bandwidth after the photodiode could impact receiver performance. On top of that, forward error correction (FEC) is broadly used to improve receiver performance. In fact, the FEC coding has been adapted as a standard requirement in the 10-Gb/s optical transmission system. As the speed goes to 40 G and higher, enhanced FEC and other signal processing techniques, such as electronic dispersion compensation (EDC) and polarization mode dispersion compensation (PMDC), are adapted to continuously extend the regeneration-free reach of the optical line. Wavelength-Division Multiplexing. Besides transmitters and receivers, optical amplifiers, WDM multiplexers, demultiplexers, and other optical devices are employed in terminal equipment. Figure 15.2 shows a typical 80-channel DWDM terminal configuration for a long-haul system. At the left is the transmitter configuration. Eighty optical signals are operated on the International Telecommunication Union (ITU) wavelength grid with 50-GHz separation. They are grouped into two banks of odd and even channels. The 40 channels in each bank are combined by an arrayedwaveguide grating router (AWG) and then combined by a 50-G/100-G interleaver. This type of configuration can increase channel isolation and minimize the crosstalk from adjacent channels. The erbium-doped fiber amplifier (EDFA) is used to compensate the loss introduced by the AWG and interleaver and boost the optical signal to the desired level. At the right is the optical receiver configuration. The incoming signal is first preamplified to normal levels by the EDFA, and then separated through the demultiplexer, which is the reverse copy of the transmitter side. Before the signal is fed into the individual optical receiver, it may need to pass through a single-channel postcompensator to compensate for residual chromatic dispersion or other types of dispersion.

426

CHAPTER 15 OPM in optical long-haul transmission systems

Post-compensator Receiver AWG1 l1

Transmitter AWG1 l1 l3

Interleaver

EDFA

l3

Interleaver

EDFA

l79

l79 AWG2

AWG2

l2

l2

l4

l4

l80

l80

FIGURE 15.2 Multiple-channel DWDM transmitter/receiver.

Optical Amplifier. EDFAs brought revolutionary change to optical fiber communication systems. Since the invention of EDFAs, power degradation is no longer the main restriction for long-distance transmission. EDFAs not only provide high gain, low noise, and very broad bandwidth, they are also fiber compatible, and independent of data rate and modulation format. Figure 15.3 shows the schematic of a simple long-haul system with cascaded fiber amplifiers. These amplifiers are categorized based on their functions. The first is a power amplifier, also called booster or postamplifier, which is used after the transmitter to increase output power. The second, used in front of a receiver to improve its sensitivity, is called a preamplifier. The third category is found on inline amplifiers, which are cascaded in the middle of the link to compensate for link loss. Typically, variable optical attenuators (VOAs) are placed after those amplifiers so that the output power can be adjusted based on the system design. To keep the system simple, the gain of each amplifier is set to be equal to its previous

Tx

Rx

Power amplifier

FIGURE 15.3 Amplified link.

Inline amplifier

Inline amplifier

Preamplifier

15.2 Elements of a long-haul transmission system

427

span loss. The simple design is to set every span at the same length. But in reality, most terrestrial long-haul systems cannot have uniform span length due to various restrictions. So the tunable optical attenuators would be used to adjust span loss to keep the system simple. Each EDFA can operate in two modes: constant gain mode and cost power mode. In most cases, the EDFA is working in constant gain mode to accommodate WDM signals. For traditional point-to-point long-haul systems with no dynamic add-drop, each EDFA is actually slightly saturated to compensate for fluctuations in optical power due to various factors. For new systems that employ reconfigurable optical add-drop multiplexers (ROADMs), linear gains are preferred to avoid power fluctuation due to wavelength add-drop. For an EDFA, pump laser power, input and output signal power, gain, and noise are parameters to be monitored. Raman Amplifier. Raman amplification is an alternative amplification technology and has been increasingly implemented in long-haul system. The Raman amplifier is different from the EDFA in that it is a distributed amplification system. Figure 15.4 shows a schematic of the Raman amplifier and power profile along the transmission line. The Raman amplifier makes use of stimulated Raman scattering (SRS) within the fiber, which transfers the energy of higher-frequency pump signals to lower-frequency signals. The amplification occurs along the transmission fiber for the distributed Raman amplifier. The typical configuration is a backward pump scheme, as indicated in the Figure 15.4, which would introduce less noise.1 The low-noise feature and large gain bandwidth make Raman amplification very attractive to long-haul systems. In some applications, such as when a large span or extra-wide bandwidth is required, the Raman amplifier is the only one that can be used. This amplifier requires much higher power than the EDFA. In practice, a Raman amplifier uses multiple pump lasers to realize high gain and flatness. Using a polarization multiplexer, two pump lasers with the same center frequency can be used to double pump power and reduce the polarization dependency of Raman gain. It is also common to use a WDM coupler to add up two to three pump lasers with different center frequency. When using a different wavelength, pump power can be increased, and bandwidth is enlarged as well. By adjusting the ratio of these pump powers, Raman amplifier can achieve flat gain. To obtain optimum performance, the pump power of each laser has to be set according to the signal spectrum received.

WDM/PBS Pump LD2 Pump LD1

FIGURE 15.4 Raman amplifier.

Launch power

Filter Signal power (dBm)

WDM

(1) Raman amplification

(2) No Raman amplification

Distance (km)

dP

428

CHAPTER 15 OPM in optical long-haul transmission systems

Transmission Fiber. Transmission fiber used for long-haul systems is single mode fiber with low loss, low PMD parameter, and high nonlinear threshold. It turns out that standard single mode fiber, e.g., SMF-28, is one of the best fibers for long-haul WDM system transmissions due to its high chromatic dispersion (CD) coefficient and low nonlinear coefficient. Fiber dispersion and nonlinear parameters play critical roles in long-haul transmission. Chromatic Dispersion. CD is one of the main limiting factors for long-haul transmission systems with data rates of 10 Gb/s and higher. The maximum transmission distance over standard singlemode fiber (SMF) for a 10-Gb/s NRZ system is limited to 150 km.1 Typically, using a nonchirped, external modulated transmitter, 10-G binary signals can transmit over SMF about 60–80 km with a 1-dB penalty. CD causes various frequency components in the signal to travel at different speeds. This introduces ISI via optical pulse spreading. The pulse spreading increases linearly with transmission distance. Once the spreading becomes comparable to the pulse width of the optical signal, it causes errors in signal detection. Considering that the pulse width of optical signals declines as the data rate increases, the CD impact to the system will increase four times as the data rate doubles. Many technologies have been developed to combat the CD impairment. Subcarrier modulation, duo-binary modulation, optical dispersion compensation, and EDC have been used in 10-G long-haul transmission systems. The most commonly used method is adding dispersion compensation fiber (DCF) or dispersion compensation module (DCM) in the middle stage of the optical amplifier so that it periodically reduces the dispersion accumulation. Figure 15.5 shows a typical dispersion management system. There is no need to add DCF or DCM in each inline amplifier. Since CD would xM SMF (L)

SMF (L)

SMF (L)

TX(i )

DCM(i)

RX(i )

DCM

Cumulated dispersion

(a)

DCM(i )

Distance

0 DCM(j )

(b)

FIGURE 15.5 Accumulated dispersion along long-haul transmission system. (a) Dispersion management system. (b) Schematic of evolution of cumulated dispersion along transmission distance.

15.2 Elements of a long-haul transmission system

429

SMF fiber dispersion 25

Dispersion (ps/nm/km)

20 15 10 5 0 –5 –10 1250

1300

1350

1400 1450 1500 Wavelength (nm)

1550

1600

1650

FIGURE 15.6 Fiber dispersion.

actually help to reduce the fiber nonlinear effect, the desired scenario is to maintain nonzero dispersion along the transmission link and reduce low overall dispersion at the end of the transmission line. Figure 15.5 shows that the DCM is added every three spans. Due to the dispersion slope, dispersion in the various channels of the WDM system differs, as shown in Figure 15.6. An individual DCM has to be added at the end of the transmission link for long-haul transmission. Slope compensations are also available and common these days. For ultra-long-haul systems, additional DCMs must be added at the transmitter side for precompensation in order to reduce the effects of nonlinearity. In theory, the CD effect is linear and can be 100% compensated by DCF or DCM. In practice, this is costly and not necessary. Small amount of residual CD can be used to offset the chirp distortion introduced by the transmitter. There is usually a window of residual dispersion in which the transceiver will work well. Polarization Mode Dispersion. PMD is another impairment in the long-haul system. Due to fiber imperfections, the transmission speed of the two-polarization mode behaves a bit differently, which is PMD. Although the PMD effect is much smaller compared to that of CD, accumulated PMD is a major impairment for high-speed, long-haul transmission. The key characteristic of PMD is its random variation with environment parameters; thus, it shows a stochastic behavior, which increases the complexity of PMD compensation. Typically, PMD is not an issue for regular optical systems at the date rate of 10 G and lower, except on some old high-PMD fibers. As the data rate increases, the PMD tolerance will decrease by the same factor that the symbol rate increases. For a 40- or 100-Gb/s system, the PMD tolerance is reduced by a factor of 4 and 10, respectively. Therefore, the PMD impairment on transmission performance has to be mitigated. PMD compensation for a traditional intensity modulation followed by direct detection (IMDD) system is a major challenge for optical systems operating at 40-G and higher speed, but it is no longer an issue for new coherent systems using electronic signal processing.5,10

430

CHAPTER 15 OPM in optical long-haul transmission systems

15.3 SYSTEM PERFORMANCE MEASURES As a digital communication system, the ultimate measure of the long-haul system performance is the bit error rate (BER). However, there are many other parameters that can be used to evaluate complex optical transmission system performance. Receiver Sensitivity. Receiver sensitivity, the traditional measure of receiver performance, is defined as the minimum received optical signal power at a specific BER (e.g., 10
9)4 in the backto-back configuration. This parameter shows the quality of receiver design. The better the receiver sensitivity, the better the system performance in terms of longer transmission distance and the greater the tolerance to fiber impairments. However, receiver sensitivity is not the critical measure for long-haul systems that consist of many optical amplifiers. In the optically amplified long-haul system, receiver sensitivity is replaced by optical signal-to-noise ratio (OSNR) as the base for performance comparison. Nevertheless, backto-back receiver sensitivity itself is still a good measure of performance for high-speed receiver design. OSNR and OSNR Penalties. Each EDFA adds extra noise to the signal during its amplification. The noise is called amplified spontaneous emission (ASE). After many cascaded fiber amplifiers, the accumulated ASE noise will dominate the channel and causes signal detection errors. The signal strength over the ASE noise is the OSNR. It is directly correlated to the signal-to-noise ratio (SNR) after the O/E conversion. The OSNR is frequently quoted as a major measure of optical amplifier system performance since ASE is the principal source of noise in long-haul transmission systems. OSNR is defined1 as OSNR ¼

P 2 Bref Nase

;

(15.1)

where P is the average signal power (includes two orthogonal polarizations), Bref is the optical reference bandwidth (typically chosen as 0.1 nm or 12.5 GHz at 1550 nm), and Nase is the power spectral density of ASE in each polarization. The Nase is determined by amplifier line parameters, such as span loss and the amplifier noise figure. By definition, the OSNR is independent of data rate, modulation format, and other transmitter characteristics, as well as transmission impairments. Since the OSNR measurement does not involve a transmitter nor a receiver, it has been broadly adapted as a basic parameter to measure optical signal transmission performance. In fact, most optical domain impairments in optical system performance are evaluated based on the OSNR penalties that they introduce. Pre-FEC BER. Because long-haul transmission systems need to carry large amounts of backbone traffic, they should be fault tolerant. In fact, the BER requirement for an operating optical telecommunication link is very low, typically <10
15.1 In order to meet this requirement, optical long-haul transmission systems have developed FEC techniques. For most long-haul transmission systems, the BER of transmitted data after the FEC is always zero and not a good indicator for system performance analysis. Alternatively, the pre-FEC BER—the BER before error correction—is directly related to optical signal integrity, and is referred to as a primary indicator of the optical system performance. Although the coding scheme requires a different precorrection BER, a 2  10
3 BER is common, which corresponds to the threshold of the enhanced Reed-Solomon code, as the criterion to evaluate the system performance.

15.3 System performance measures

431

Q-Factor. To guarantee transmission quality, most systems do not operate near the boundary. System design typically allocates an adequate margin for signal-degrading effects, and commonly reserves 3–5 dB for unknown degrading effects and aging. So, the real system is usually operating in the very small pre-FEC BER area, such as <10
12. Such small pre-FEC BER would take time to measure and is thus not effective for performance monitoring. The Q measurement, a practical alternative in such situations, is based on the eye diagram of the receiving signal before the decision circuit. The relationship between the Q and BER for binary on-off keying (OOK) system is given by1   1 Q expð
Q2 =2Þ pffiffiffiffiffiffi BER ¼ erfc pffiffiffi  ; (15.2) 2 2 Q 2p where Q is given by Q¼

I1
I0 : s1 þ s0

(15.3)

The approximate form of BER is obtained by using the asymptotic expansion of the efrc function, and is valid for Q > 3. The I1 and I0 are mean values of 1 and 0, respectively. The s1 and s0 are variances of 1 and 0, respectively. Eye Diagram. The eye diagram is another important alternative BER measurement for binary OOK systems. There are two types of noises that can impact system performance: amplitude noise and timing jitter. The simple BER measurement cannot separate the two effects. The eye diagram provides more information on both, especially time jitter, and also measures the extinction ratio. Traditional eye diagram measurement usually involves using a digital oscilloscope and requires collecting a lot of data, which takes time. Thus, it is not suitable for real-time processing. This will be improved for future systems using digital signal processing (DSP). Signal Constellation Diagram. For new systems using multilevel signal formats and coherent detection technologies, the signal constellation diagram is very useful for providing both amplitude and phase information on signal and noise. With the help of the constellation diagram, electrical SNR can be estimated and optimized to achieve better performance. Dispersion Tolerance. Many factors in the long-haul optical transmission link can cause significant optical signal degradation. The main factors include CD, PMD, and fiber nonlinearity. These impairments introduce ISI distortion at the receiver, and therefore cause performance degradation. Many techniques have been investigated for effective measurement of CD and PMD and performance analysis.8,9 However, most of these measurements are complicated and not accurate enough. A practical method to deal with the impacts of CD and PMD is to allocate appropriate system penalties to CD and PMD impairments. The CD-induced OSNR penalty is defined as the additional OSNR required to get the same BER with and without the CD impairment. Accordingly, the PMDinduced OSNR penalty is defined as the additional OSNR required to compensate for the PMD effect. These are effective performance measures when both CD impairment and PMD impairment are small and can be treated independent of each other. The CD tolerance or PMD tolerance refers to the range of residual CD or PMD that induces OSNR penalty within 1 dB. For 40-Gb/s and higher-speed systems, this criterion can be relaxed to 2 dB.1 In practice, the system dispersion map is calculated based on the premeasured parameters of CD for each span. The overall accumulated CD can be reduced to a small range within the CD tolerance by

432

CHAPTER 15 OPM in optical long-haul transmission systems

DCF and DCM. Such method is effective since the CD effect is deterministic and linear with the fiber length. The CD tolerance can be used to evaluate the system robustness against the CD impairment. PMD is not an issue for regular systems at the data rate of 10 Gb/s and below. Most fibers used in long-haul system transmission are low-PMD fibers. Only a few systems need to deal with high-PMD fiber. However, as the data rates move up to 40 and 100 Gb/s, both CD and PMD become the real challenge for long-haul transmission. Both impairments of CD and PMD need to be monitored and corrected. Nonlinear Penalty. The nonlinear penalty is introduced by signal distortions from nonlinear interactions of optical signal with optic-fiber. The main nonlinear effects that can impact the long-haul fiber transmissions are self-phase modulation (SPM), cross-phase modulations (XPM), four-wave mixing (FWM), stimulated Raman scattering (SRS), and stimulated Brillouin scattering (SBS). These nonlinear interactions in fiber communication systems depend on a large numbers of parameters, such as signal power, fiber nonlinear coefficient, fiber dispersion, WDM channel number, channel spacing, transmission distance, and amplification scheme. It is difficult to accurately measure the nonlinear transmission interactions and penalties. Fiber nonlinearities occur when optical signals are strong, which limits the upper boundary of signal power in fiber transmissions. In most cases, nonlinearities are detrimental to the system performance and should be minimized. In general, nonlinear penalty should be kept within a few dB for any given impairment since the system performance will deteriorate dramatically with further increase in the impairment. Nonlinearity is becoming a major issue for coherent detection systems with DSP. All linear impairments such as CD and PMD can be easily corrected in electronic domain in the absence of fiber nonlinearity. Wavelength Accuracy and Stability. In the past decades, optical transmission systems have moved to more dense WDM architecture, with the channel spacing reduced from 200 GHz, to 100 GHz, 50 GHz, and even 25 GHz. Large numbers of narrowband wavelength multiplexers and demultiplexers are used in long-haul transmission systems as capacity and distance increase. Accurate wavelength control and alignment become critical to maintain satisfactory system performance. Active WDM components, such as lasers, AWG, and tunable filters, require active wavelength control. Many wavelength-related devices are passive, and their parameters are fixed and predetermined in the system design. Bandwidth mismatch and misalignment of these WDM devices can cause signal degradation. Therefore, system-level wavelength measurements and drifting monitoring become necessary for system performance evaluation.

15.4 OPM IN A LONG-HAUL TRANSMISSION SYSTEM 15.4.1 OPM functions and applications

The overall performance of a system is determined by the transmission quality of its signals, as well as its abilities to discover signal degradation, locate failures, analyze root causes, and quickly recover the system. OPM plays an important role in signal integrity tests and system performance controls. The following are basic functions that OPM could provide for long-haul systems: 1. Monitor the status of optical devices. 2. Provide feedback to optimize optical devices.

15.4 OPM in a long-haul transmission system

3. 4. 5. 6.

433

Monitor signal integrity along the transmission line. Detect optical link failures and report to system management. Provide feedback for transmission link optimization. Estimate fiber impairments and assist in physical-layer debugging.

15.4.2 Optical device monitoring Optical device monitoring is one of the basic functions of OPM. The optical components in longhaul systems that require monitoring are the transmitter, receiver, optical amplifier, and active WDM components. The major monitoring parameters are power and wavelength.

15.4.2.1 Transmitter and receiver monitoring Fixed-wavelength lasers are widely used in long-haul systems. Tunable lasers are increasingly deployed in the field. These lasers must be operated on the ITU grid and require accurate wavelength locking and power stabilization. Table 15.1 lists basic parameters of optical signal performance defined by GR-253/GR-1377, ITU-T G.957, and IEEE 802.3z. For a long-haul DWDM system, power variation and wavelength drifting can lead to system performance degradation. Tight power control and accurate wavelength locking are required to prevent degradation. The typical accuracy requirement for power control is 0.5 dB. For extra-long-distance transmission, the accuracy requirement for power control would be 0.1 dB. The accuracy of wavelength locking depends on multiple parameters, such as the data rate (which will impact the bandwidth of modulated signals), channel spacing, and filter bandwidth. For 100-G and 50-G spacing systems (which can support 40 and 80 channels in the C-band, respectively), the wavelength accuracy of 0.1 nm is sufficient. But when the channel spacing reduces to 25 G, tighter wavelength control is needed. Typical wavelength accuracy would be determined by temperature-controlled feedback circuitry. Reliability is another vital element of OPM. The failure rate of optical components is several orders higher than electronic components. Optical systems are required to continuously monitor power and wavelength variations in order to determine the sign of symmetric degradation. Typically, data are saved for some periods. Short-term data are stored in local devices and limited by the memory size of these devices. Long-term data are saved in remote servers that can store records for days and months. Many systems provide tools for data retrieval and performance analysis. Table 15.1 Optical Signal Performance under Normal Conditions Element

Parameters

Transmitter

Maximum output power/minimum output power Wavelength (ITU-grid) Extinction ratio Eye diagram Minimum input power/overload input power Filtering performance

Receiver

434

CHAPTER 15 OPM in optical long-haul transmission systems

15.4.2.2 EDFA monitoring Performance of long-haul optical transmission systems is determined by amplifier performance. The most commonly used amplifiers in long-haul transmission are EDFAs. The EDFA is a low-cost, broadband, high-gain, high-power, and low-noise amplifier. Performance measurements for an EDFA include the maximum and minimum gain, gain bandwidth, gain flatness, gain stability, gain tilt, and maximum ASE noise level, as listed in Table 15.2. In practice, pump laser power, input signal, output signal, signal spectrum, gain spectrum, and noise figures are widely monitored when EDFAs are in operation.

15.4.2.3 Raman amplifier monitoring The Raman amplifier is another widely used fiber amplifier in long-haul systems. Raman amplification is a distributed process where signal amplification takes place inside the transmission fiber. Measuring Raman gain or noise directly is difficult. Typically, in a Raman-amplified system, the power and spectrum of amplified output signals are monitored. Raman amplifications rely on the SRS effect of transmission fiber, which provides gain over a limited wavelength region. Using two to three pump lasers with slightly different wavelengths in the 1480-nm region comprises a broadband amplifier that covers both the C-band and L-band (65 nm). The powers and wavelengths of Raman pumps need to be optimized to obtain a flat spectrum. The characteristics of transmission fiber impact Raman gain and must be taken into account in amplifier design. Commonly used fibers include standard SMF, dispersion-shift fiber (DCF), LEAF fiber, and TureWave fiber. The Raman gain coefficients of these fibers are significantly different. Different pump powers are required to achieve the same gain. Optimizing a Raman amplifier is more complicated than for an EDFA. Raman amplification provides more flexibility than EDFA. Both gain and gain shape can be adjusted by pump powers. Raman amplifier adjustments require use of an optical spectrum analyzer (OSA).

15.4.2.4 WDM device monitoring Most WDM devices are passive devices, such as mux/demux, interleaver, WDM coupler, OADM, and bandpass optical filters. Active devices are tunable filters and ROADMs. Each device has limited passband and adjacent channel rejection. These devices will cascade in long-haul systems and narrow the overall effective passband. The impacts of narrow passband and neighboring channel crosstalk grow linearly with the number of cascaded devices. The central wavelength of these devices must be aligned and controlled.

Table 15.2 Optical Signal Performance for EDFA under Normal Conditions Element

Parameters

Optical Amplifiers

Maximum gain/minimum gain Gain bandwidth Gain flatness (per input power level and channel number) Gain stability Gain tilt Maximum ASE

15.4 OPM in a long-haul transmission system

435

15.4.3 OSNR monitoring along transmission line OSNR monitoring, a major OPM function for long-haul transmissions, is typically performed by an optical channel analyzer (OCA) or OSA. Cost-effective standalone OSA/OCAs have been developed to provide accurate power and wavelength measurement, as well as rather good OSNR measurements using a narrowband tunable filter.

15.4.3.1 OSA/OCA in transmitter and receiver Figure 15.7 shows how an OSA/OCA is used in an 80-channel terminal transmitter. This configuration, which provides excellent channel isolation, uses two-array waveguide gratings (AWGs) and one interleaver to combine 80 channels. A 1  4 optical switch and an OSA are used to measure all 80 channels. The OSA can provide power, wavelength, and OSNR measurements, as well as channel balancing, wavelength fine-tuning, and wavelength-drift monitoring. Figure 15.8 shows the corresponding configuration of the 80-channel receiver. One OSA/OCA is placed after the preamplifier to measure the receiving signals spectrum. The OSA/OCA of this receiver can be the same one used for the transmitter, resulting in OPM cost savings. Unlike testing equipment, the OCA/OSA for performance monitoring does not require very high resolution and accuracy. These requirements trade off for fast sweep speed, small size, light weight, and low cost. Appropriate resolution is selected to match with channel spacing. For a regular 100-G/50-G spacing system, 0.1-nm (or 12.5-GHz) resolution is sufficient. For some ultradense WDM systems, especially certain new systems with high spectral efficiency, better resolution is needed.

Transmitter

AWG1

l1 l3 Interleaver EDFA

l79 AWG2 l2 l4

OSA l80

FIGURE 15.7 Example of OSA/OCA measurement in DWDM transmitter.

Optical switch

436

CHAPTER 15 OPM in optical long-haul transmission systems

AWG1

Postcompensator

Receiver l1 l3

Interleaver EDFA

l79 AWG2 l2 OSA

l4

l80

FIGURE 15.8 OSA/OCA monitoring at receiver node.

15.4.3.2 OSNR requirements One key design requirement of a long-haul optical transmission system is to ensure that the accumulated noise is always well below the optical signals—that is, maintenance of sufficiently high OSNR through the link. The OSNR is a function of launched power and other line parameters. For example, the OSNR at a transmission line composed of N identical amplification spans can be expressed by the following equation1:   1000 þ Pin
NF
LSP
10 log10 N  58 þ Pin
NF
LSP
10 log10 N: OSNRN ðdBÞ ¼ 10 log10 hnBref (15.4) The Pin, NF, and LSP are the launch power (in dBm), effective noise figure of a span (in dB), and the span loss (in dB), respectively. To maintain high performance, the OSNRN must be larger than the required OSNR for a receiver. The required OSNR is defined as the minimum OSNR at the receiver sensitivity to achieve a predefined BER. Typically, the OSNR is a function of receiving power, as shown in Figure 15.9. When the OSNR is less than 16 dB, a small decrease in OSNR results in a significant increase of receiving power to achieve the same BER. When OSNR is larger than 16 dB, the curve becomes flat. Therefore, the OSNR of 16 dB is determined by experiment as the required OSNR for this system. Table 15.3 lists the typical OSNR requirements for various systems at the receiver end in backto-back configuration. Target BERs vary by system since diverse FEC schemes are adapted. In theory, to achieve the same BER, if the data rate quadruples, the required OSNR also needs to increase by 4 (6 dB).

Rx power (dBm/port)

15.4 OPM in a long-haul transmission system

437

–9 –11 –13 –15 –17 –19 13

14

15

16 17 18 19 OSNR (dB/0.1 nm)

20

21

22

FIGURE 15.9 Required OSNR and receiver sensitivity: achieving error-free operation (BER < 10–12) after 1000-km SMF transmission with no CD compensation using special SCM transmitter/receiver.

Table 15.3 Typical OSNR Requirements for Long-Haul System1,3,6 Data Rate

OSNR Req (dB)

Notes

2.5-Gb/s NRZ-OOK 10-Gb/s NRZ-OOK 10-Gb/s RZ-OOK 40-Gb/s NRZ-OOK 40-Gb/s RZ-OOK

10 14 12 16 14

Target Target Target Target Target

BER BER BER BER BER

¼ ¼ ¼ ¼ ¼

10
9, 10
5, 10
5, 10
3, 10
3,

no FEC using RS FEC using RS FEC using enhanced FEC using enhanced FEC

Increasing the launch power into fiber can improve the OSNR until fiber nonlinearities kick in. The fiber nonlinear effects, such as SPM, XPM, or FWM, are proportional to signal intensity and will cause irreversible signal degradation. To maintain the best performance, the launch power must be distant from the nonlinear upper boundary, which depends on fiber type, data rate, and total transmission distance, among other factors. The lower boundary for the signal power is determined by noise level. Shortening amplifier spans helps to raise input power into amplifiers and lower individual amplifiers’ noise. But a short span means that more amplifiers are needed to achieve the same transmission distance. More amplifiers generate more noise. Balancing these nonlinear effects with accumulated ASE noise to achieve top performance is an optimization target for long-haul systems.

15.4.3.2 OSNR monitoring in intermediate amplifier node OSA/OCAs can be used to optimize amplifier link. There is no need to add an OSA/OCA at each inline amplifier. For a long-haul system, adding OSA/OCA every three to five amplifiers along the transmission link can provide sufficient information for system optimization. Figure 15.10 shows configurations of intermediate nodes with and without add-drop channels. For the add-drop node (on the left), there are four amplifiers, two for eastbound and two for westbound. This configuration can compensate extra loss introduced by the add-drop coupler and dispersion-compensating modules

438

CHAPTER 15 OPM in optical long-haul transmission systems

Add1

Drop1

Mon1

Mon2

Mon1

Mon2

Attn1 In1

Amp1

Out1

Amp2 WDM1

WDM2 DCM1

In1

Attn1

Attn2 DCM1 DCM2

DCM2 WDM4

WDM3 Amp3

Amp4

Mon4

In2

Out2

Attn3

Attn4

Amp2

In2

Attn2 Drop2

Add2

Out1

Amp1

Mon3

(a)

Mon4

Mon3 (b)

FIGURE 15.10 Intermediate repeater node with and without add-drop channels: (a) add-drop node, and (b) intermediate node without add-drop.

(DCMs). VOAs are used for power adjustment and gain-tilt control. For regular nodes without adddrop channels (on the right), two amplifiers are used, one for each direction, as shown in Figure 15.10(b). The DCMs are optional. The OSA/OCA is connected to any monitor port (Mon1 to Mon4) or connected to all monitor ports via an optical switch. These inline OSA/OCAs provide abundant information on optical signal quality along the transmission link, including signal power, noise level, OSNR, and so on. The key component for OSA/OCAs is narrowband tunable filters. Many commonly used tunable devices, such as Fabry-Perot filters, Bragg grating filters, free-space or MEMS diffractive optics, and dielectric thin-film filters, meet the requirements for bandwidth and free spectrum range (FSR) of OSA/OCAs. The scan time over the whole spectrum cannot be too long. The maximum scan time should be around 30 s to 1 min—the shorter, the better. The accuracy of power measurement for OSA is around 0.1–0.5 dB. The wavelength resolution is determined by a system channel plan. Typically, 0.1 nm is sufficient for a typical DWDM system with 100-G/50-G spacing. Higher resolution is needed for ultradense WDM systems, such as the 25-G spacing system.

15.4.4 Transmission performance testing and analysis 15.4.4.1 System penalties and system margins Signal distortions due to fiber impairments are major concerns for long-distance-system transmissions. These impairments may be limiting factors to system performance. Some of major impairments include: 1. 2. 3. 4.

Chromatic dispersion Polarization effects and polarization mode dispersion Fiber nonlinearity-induced distortion and crosstalk WDM device–induced distortion and crosstalk

15.4 OPM in a long-haul transmission system

439

Table 15.4 Example of 10-G System Budget Allocation (NRZ, 2000 km, SMF) System Element Tx/Rx OAS Fiber Fiber WDM Fiber Margin Margin

Parameters Back-to-back required OSNR for BER ¼ 10 Penalty due to channel power ripple and tilt Dispersion penalty PMD penalty Interchannel WDM crosstalk Nonlinear transmission penalty Aging and repairs System margin Required EOL minimum-received OSNR

Value or Penalty (dB)
3

10.0 0.5 1.0 1.0 0.5 1.0 1.5 3.0 18.5

The impacts of these impairments vary by system parameter, such as data rate, modulation format, fiber type, fiber span, transmission distance, and signal power, among others. A systematic way to handle these impairments is to separately evaluate them and allocate appropriate system penalties for each one. Table 15.4 shows a budget allocation example. This is a standard 10-G IMDD system over SMF fiber. The transmission distance is 2000 km. The OSNR for transmitting and receiving optical signals over a 2000-km fiber is 10 dB with pre-FEC BER of 10
3. The system penalties allocated for dispersion, PMD, interchannel crosstalk, and nonlinear effect are 1, 1, 0.5, and 1 dB, respectively. These values are very conservative compared to actual degradation. The total OSNR budget for this system is 18.5 dB with a 3-dB margin. Use of the budget table effectively simplifies the system design and guarantees the required operation. The system margin and penalty for individual impairment can be determined by simulations and experiments.

15.4.4.2 CD and dispersion-compensation management Combating CD impairment is the principal challenge for 10-G long-haul transmission system design. On the one hand, CD-induced signal-pulse broadening will cause ISI distortions and errors when it interferes with the signal detection process at the receiver. On the other hand, high CD is desired to combat disruptive nonlinear impairments. Since the CD effect is a linear and deterministic process along the transmission link, it can be compensated using DCF or DCM, which have a negative dispersion coefficient as transmission fiber. In practice, the DCM technique or dispersion map is widely used to minimize both CD and nonlinear impacts on system performance. A commonly used parameter for dispersion performance evaluation is CD tolerance, which is defined as the amount of residual CD that will introduce a 1-dB OSNR penalty (some 40-G or 100-G systems allocate a 2-dB penalty for CD effects). Table 15.5 lists the CD tolerance (third column) for some commonly used modulation formats. The numbers given in the table are typical values for each category. They are rough estimations as references for OPM. For the sake of simplicity, the accumulated dispersion is converted to the SMF length (D ¼ 17 ps/nm-km), and put into brackets. This length refers to the dispersion-tolerance distance that a system can tolerate without optical dispersion compensation. The dispersion-tolerance distance helps to explain how much a system is vulnerable to CD and dispersion map variations. For example, the CD tolerance for a 2.5-G NRZ system is 16,000 ps/nm. If converted to SMF length, the

440

CHAPTER 15 OPM in optical long-haul transmission systems

Table 15.5 Chromatic Dispersion and Polarization Mode Dispersion Requirements for Direct Detection System1–3,7,10

Data Rate 2.5 Gb/s 10 Gb/s

40 Gb/s

Modulation Format

CD Tolerance (ps/nm) (1-dB penalty)

NRZ-OOK NRZ-OOK RZ-OOK CSRZ-OOK DB NRZ-OOK RZ-OOK CSRZ-OOK DB NRZ-DPSK RZ-DPSK NRZ-DQPSK RZ-DQPSK

16000 (940 km) 1000 (59 km) 800 (47 km) 768 (45 km) 4000 (235 km) 62 (3.6 km) 50 (3 km) 48 (2.8 km) 250 (15 km) 56 (3.3 km) 45 (2.6 km) 120 (7 km) 150 (9 km)

PMD Tolerance (ps) (1-dB penalty) – 32 40 44 24 8 10 11 6 10 10 20 21

dispersion-tolerance distance is 940 km, which means that the 2.5-Gb/s signal can transmit over 940 km of SMF fiber with no need for dispersion compensation. The dispersion-tolerance distances for 10- and 40-Gb/s NRZ systems are reduced to 59 km and 3.6 km, respectively. This indicates that a 10-G system needs dispersion compensation if its transmission distance is larger than 59 km. For a 40-G system, dispersion compensation is required after a 4-km transmission. Furthermore, dispersion map accuracy for 10-G and 40-G systems has to be well controlled within 59 km and 3.6 km. For a long-haul system over 1000 km, such requirements are tight. In theory, the CD tolerance is reduced by a factor of 16 as the data rate is increased by a factor of 4, assuming the same modulation format. For long-haul transmission systems, dispersion slope has to be taken into account in dispersion management. As shown in Figure 15.6, fiber dispersion is a function of wavelength. The slope of the dispersion curve for SMF-28 is about 0.058 ps/nm2-km. Consider a C-band WDM system with a 32-nm usable bandwidth: after a 1000-km transmission, the side channels will accumulate extra 930 ps/nm compared to the central channel. These extra CDs have to be compensated separately by adding channel-based pre- and postcompensation at the transmitters and receivers, respectively. Various tunable CD compensators have been developed to simplify dispersion management and enable dynamic optimization. Although CD can be 100% compensated through appropriate dispersion management, the requirements on accurate dispersion management for 40-G and higher speeds are very tight and challenging in terms of actual operations. Advanced modulation formats, especially DPSK and DQPSK, have shown better performance in terms of required OSNR and tolerance to CD and PMD effects, among other parameters, and have become prevalent technologies for 40 G. Recently, new EDCs using DSP technology have been developed and demonstrated for 40-G and 100-G systems. EDC is a cost-effective replacement of optical tunable CD compensators. With coherent detection technology, digital dispersion compensation can compensate very large amounts

15.4 OPM in a long-haul transmission system

441

of dispersion and may eliminate most optical CD compensation. CD will not be a transmissionlimiting factor for long-haul transmission.

15.4.4.3 PMD monitoring and compensation PMD is another type of fiber impairment that impacts long-haul transmissions when a system is upgraded to 40 G or 100 G. PMD impairment is usually much smaller than that of CD at any given data rate. However, for a traditional IMDD system, CD can be 100% compensated using DCF and DCM. There is no simple and effective PMD compensation method. For traditional IMDD systems at 40 and 100 Gb/s, PMD becomes a significant transmission impairment that limits system performance. Unlike CD, PMD is more vulnerable to ambient conditions, such as stress, vibration, and temperature changes. The challenge of PMD impairments is that both the principal state of polarization (PSP) and the value of the dispersion between the two PSPs—that is, DGD—vary with time. The variation is stochastic,13 seen in the snapshot shown in Figure 15.11. Three widely used solutions to PMD issues are (1) use of low-PMD fiber and low-PDL optical components; (2) use of the advanced modulation format with high PMD tolerance; and (3) adaptive PMD compensation by adaptive electrical filters or in optical domain by using a birefringent DGD equalizer. The low-PMD fiber solution works for green-field deployment but is not applicable to existing systems. Many approaches14–17 for PMD measurement and PMD compensation have been discussed in the literature. Figure 15.12 shows an example of a PMD compensator used in a 40-G receiver. The basic principle is using two polarization controllers (PC1 and PC2) with two

4.000 3.500

DGD (ps)

3.000 2.500 2.000 1.500 1.000 0.500 0.000 1520.000

1540.000

1560.000

1580.000 Wavelength (nm)

FIGURE 15.11 Fiber PMD.

1600.000

1620.000

442

CHAPTER 15 OPM in optical long-haul transmission systems

PC1

PD

PC2 Delay 1

Delay 2

Δτ1

Δτ2

Polarimeter

FEC

DOP error

Controller

FIGURE 15.12 Example of 40-G transceiver using PMD compensator.

fixed-DGD delay lines to generate a PMD that can cancel the PMD generated through the transmission link. A polarimeter or other device that can detect the degree of polarization (DOP) is used to provide feedback to the two polarization controllers. The control signal would also be provided by FEC error detection from the receiver. Commercial PMDC techniques can compensate 8–10 ps mean DGDs, and are independent of signal format and wavelength. But optical PMDCs are applied on a per-channel basis, and are complex and expensive. Studies have demonstrated that many advanced modulation formats can tolerate more PMD than OOK. A widely used advanced modulation format, DQPSK, can effectively increase tolerance to PMD by increasing its pulse width. Table 15.5 provides PMD tolerance (the fourth column) for different modulation formats. Recent studies have shown that upon using coherent detection and a digital linear equalizer, PMD can be fully compensated and will not be a limiting factor for 40-G and 100-G systems.19 Successful transmission of a 40-G signal over 3000 km of installed fiber has been demonstrated using adaptive DPSK without PMD compensation. The measured mean PMD of that experiment is 4.3 ps.18

15.4.5 Service-oriented system design Previous sections focus on physical transmission performance for long-haul systems. This section discusses other system performance considerations. A service-oriented system focuses on superior service performance. The service-oriented long-haul system design must consider the following characteristics: 1. High reliability and availability: Most long-haul links are parts of backbone networks and carry large amounts of data. The reliability requirements on long-haul systems are much higher than those of local and access networks. Besides error-free transmission, very low system outage and failure protections are usually required. 2. Scalability: A long-haul system should be designed for full capacity but need not operate at full capacity at inception. A high-performance system is able to start with low initial cost that can

15.4 OPM in a long-haul transmission system

3. 4. 5. 6. 7.

443

meet near-term expectations and is capable of scaling economically to meet growing demands and new services. Cost effective: The system needs to trade off between equipment and installation costs to realize overall savings. Easy operation and maintenance. Balance between flexibility and simplicity. Decent upgrades and downgrades. Easy management.

15.4.5.1 System availability and system performance System availability is critical to the business success of service providers. Revenue growth largely depends on deploying value-added services such as real-time triple play (video, voice, and data), VPN, and other mission-critical applications. These services have stringent system-reliability requirements. Various levels of performance testing must be done to ensure that a deployed system can deliver acceptable quality of service (QoS) and can comply with the performance metrics specified in service-level agreements (SLAs). System availability is a measure of users’ abilities to access services and is defined by11 Network Availability ¼

MTBF : MTBF þ MTTR

(15.5)

Mean time before failure (MTBF) is a measure of network reliability. Mean time to repair (MTTR) measures the average time it takes for a network to recover and return to full operation after a failure. Both average failure times and average repair times contribute to the high availability of a network and need to be addressed in a high-performance system. The ultimate goal is to minimize service outage. From the system provider point of view, there are two types of outages: unexpected and planned. An unexpected outage is caused by fiber cut, serious signal deterioration, hardware failure, software failure, power outage, and others system faults when the system is in operation. The planned outage is the result of system maintenance or upgrade. The goal of system performance monitoring is to prevent or reduce system failure by monitoring the health of system links and various components. Fiber link performance metrics require continuous updating. Other nonphysical elements, such as redundant system architecture, fast fault detection, and system protection mechanisms have to be taken into consideration. The ultimate requirement is to keep end-user communication reliable.

15.4.5.2 Fault detection and isolation System performance measures involve the ability to detect signal degradations, locate and isolate failures and faults, report alarms, and restore traffic. A high-performance system usually requires monitoring of performance parameters for all connections supported in the transmission system. It is important to have a symmetrical way to handle performance metrics and communicate status and alarms to system management. An OPM can provide multiple levels of information regarding the monitored parameters. For example, an OSA can measure the optical spectrum. In addition, the OSA can provide WDM channel information—that is, identification of missing and occupied channels. It can also monitor channel wavelength drifting. OPM can be designed to play an important role in fault detection and fault isolation.

444

CHAPTER 15 OPM in optical long-haul transmission systems

15.4.5.3 Real-time performance monitoring A long-haul system must quickly detect failure and response. The carrier-grade telecommunication system requires failure restoration within 50 ms. Failure detection ranges between a few milliseconds to dozens of milliseconds. To respond quickly, the system must be able to tell which part of the system has problems—that is, to tell if the problem is caused by physical failure of hardware, software bugs, or fiber impairment issues. Time requirements for OPM are highly dependent on the type of parameters being monitored. Hardware failure and link failure are serious failures and require immediate action and reporting. The carrier-grade criteria for signal failure detection and reporting are under 50 ms, based on the SONET/SDH standard. For high-speed, long-haul systems where the system budget and tolerance for fiber impairment are tight, OPM is required to provide feedback for locking, stabilization, optimization, and compensation. These applications require that the OPM runs at speeds ranging from microseconds to seconds.

15.4.5.4 Remote performance monitoring In a long-haul system, much equipment operates in intermediate nodes located at remote sites. These remote sites are hundreds of miles away from the operation center and may be located in isolated places. Debugging and maintenance at such remote sites are time consuming, difficult, and costly. This requires that OPM on those nodes be remotely accessed and controlled. Remote performance monitoring helps to save time and reduce cost for operations and debugging. OPM is a powerful tool for performance management.

15.4.5.5 System installation and upgrade As a performance management tool, OPM plays an important role in accelerating system installation and upgrade.

Automatic power balancing Power balancing of a long-haul transmission system is a major task during a system deployment and upgrade. After long-haul transmission, a small variation along the amplifier link can be enlarged multiple times. Both EDFA and Raman amplifiers have unflatten gain spectrum and gain tilt. Combined with wavelength-dependent loss of transmission fiber and other wavelength-dependent defects, these result in significant power imbalances for WDM channels after long-distance transmission. Figure 15.13 shows an example of the signal spectrum after a 1200-km transmission. The power variation can be as large as 10 dB. In order to maintain constant performance, a consistent level of OSNR for all channels is necessary. The gain tilt must be corrected by adjusting a series of amplifiers along the link. The optical power of each channel must be adjusted accordingly. Manual power balancing is very inefficient considering the large number of channel counts and amplifier spans involved. Many systems use OPMs to balance power automatically, which dramatically reduces installation costs and saves time.

“Future-proof” system upgrade It is predicted that future optical transmission networks will be hybrid networks that consist of hybrid channels running at 10 G, 40 G, and 100 G. Network operators prefer to upgrade their networks on

15.5 Summary

445

FIGURE 15.13 Spectrum of long-haul transmission.

demand with minimal network redesign and expense. Many system vendors are developing products that overlay a 40-Gb/s system on existing 10-Gb/s infrastructure. An IEEE standard is in development to set the specifications that allow future 40-G and 100-G signal transmission over existing 10-G DWDM systems. It is expected that long-haul systems will be gradually upgraded from 10 G to 40 G and 100 G in the near future. The OPM would play an important role in these “future-proof” system upgrades. There are two levels of system upgrades. The first level adds more channels to full capacity; the second level gradually upgrades channels from a low to high data rate. A hitless upgrade requires that existing traffic remain intact through the upgrading process. A non-hitless upgrade requires that system turn-down time is minimized. In any case, OPM is required. OPM can closely monitor the optical physical parameters during the upgrade process to ensure that new channels and systems can be optimized and turn up as quickly as possible while interruptions of existing traffic remain low.

15.5 SUMMARY Optical performance monitoring plays a critical role in long-haul transmission system operations. It would be very costly and difficult to process all channels via electronic monitoring. Optical performance monitoring has shown competitive advantages to simplify system design, optimize system performance, shorten system installation, and lower operations cost. OPM introduces multilevel performance testing for a long-haul transmission system. As nextgeneration optical transmission systems move up to 40 Gb/s and 100 G, new OPM functions are needed to accurately monitor the performance of devices and transmission links. The development of cost-effective OPM technology is crucial to the continued growth of the optical communications industry.

446

CHAPTER 15 OPM in optical long-haul transmission systems

REFERENCES 1. Agrawal GP. Fiber-optic communication systems. 2nd ed. New York: John Wiley and Sons, Inc. 1997. 2. Kaminow I, Li T, editors. Optical fiber telecommunications IIIA. New York: Academic Press; 1997. 3. Kaminow I, Li T, Willner A, editors. Optical fiber telecommunications VB. New York: Academic Press; 2008. 4. Kazovsky L, Benedetto S, Willner A. Optical fiber communication systems. Norwood, CT: Artech House; 1996. 5. van den Borne D, Duthel T, Fludger CRS, et al. Electrical PMD compensation in 43-Gb/s POLMUX-NRZDQPSK enabled by coherent detection and equalization. In: Proc. ECOC. paper 8.3.1. Berlin; 2007. 6. Farbert A, Langenbach S, Stojanovic N, et al. Performance of a 10.7-Gb/s receiver with digital equalizer using Maximum Likelihood Sequence Estimation. In: Proc. ECOC. paper Th4.1.5. 2004. 7. Pecci P, Lanne S, Frignac Y, et al. Tolerance to dispersion compensation parameters of six modulation formats in systems operating at 43-Gb/s. In: Proc. ECOC. paper WE3.5.5. 2003. 8. Lize YK, Christen L, Yang J-Y, et al. Independent and simultaneous monitoring of chromatic and polarization-mode dispersion in OOK and DPSK transmission. IEEE Photon Technol Lett 2007;19:3–5. 9. Cornick KE, Hinton K, Dods SD, et al. A framework for evaluating the system penalty from polarizationmode dispersion using different performance monitoring techniques. J Lightwave Technol 2008;26(13): 1790–7. 10. Daikoku M, Yoshikane N, Morita I. Performance comparison of modulation formats for 40-Gb/s DWDM transmission systems. In: Proc. OFC/NFOEC, paper OFN2. 2005. 11. Ciciora W, Farmer J, Large D, Adams M. Modern Cable Television Technology - Video, voice, and Data Communications. 2nd ed. San Francisco: Morgan Kaufmann Publishers; 2004. 12. Derickson D. Fiber optic test and measurement. Upper Saddle River, NJ: Prentice-Hall; 1998. 13. Brodsky M, Frigo NJ, Boroditsky M, et al. Polarizaiton mode dispersion of installed fibers. J Lightwave Technol 2006;24:4584–99. 14. Luo T, Pan Z, Nezam SMRM, et al. PMD monitoring by tracking the chromatic-dispersion-insensitive RF power of the vestigial sideband. IEEE Photon Technol Lett 2004;16:2177–9. 15. Nezam SMRM, McGeehan JE, Willner AE. Theoretical and experimental analysis of the dependence of a signal’s degree of polarization on the optical data spectrum. J Lightwave Technol 2004;22:763–72. 16. Wang SX, Weiner AM. A complete spectral polarimeter design for lightwave communication systems. J Lightwave Technol 2006;24:3982–91. 17. Yan L-S, Steve Yao X, Shi Y, et al. Simultaneous monitoring of both optical signal-to-noise ratio and polarization-mode dispersion using polarization scrambling and polarization beam splitting. J Lightwave Technol 2005;23(10):3290–4. 18. Chen DZ, Xia TJ, Wellbrock G, et al. New field trial distance record of 3040 km on wide reach WDM with 10 G and 40 Gb/s transmission including OC-768 traffic without regeneration. J Lightwave Technol 2007;25:28–37. 19. Sun H, Kuang-Tsan W, Roberts K. Real-time measurements of a 40 Gb/s coherent system. Opt Express 2008;16(2):873.

Index Page numbers followed by “f ” indicate figures and “t” indicate tables.

A 4-QAM, see Quaternary phase-shift keying AAH, see Asynchronous amplitude histogram Add/drop multiplexing (ADM), 24–25 ADM, see Add/drop multiplexing AM, see Amplitude modulation AM pilot tones, see Amplitude modulation pilot tones Amplified spontaneous emission, (ASE), 21, 22–23, 25–26, 145, 147–148, 153 linear optical sampling characterization, 211, 212f noise, 34–35, 36–37, 38, 42–43, 158, 160–161 Amplifier distortion and transients, 7 noise, 7 Amplitude modulation (AM), radiofrequency spectrum measurement for chromatic dispersion monitoring, 77–79, 78f, 80f Amplitude modulation (AM) pilot tones chromatic dispersion monitoring chirped pilot tone monitoring, 249–251, 249f, 250f, 251f, 252f effect of self-phase modulation and polarization mode dispersion, 248–249, 248f operating principles, 246–247, 247f pilot tones carried by broadband light source, 252–256, 252f, 253f, 254f, 255f, 256f experimental setup, 240, 241f implementation examples, 238, 239f optical channel power and wavelength monitoring, 234–235, 234f, 235f optical path and channel identification, 235–238, 236f, 237f overview, 223–224 principle of performance monitoring, 224–225, 224f, 225f, 238–243, 240f

problems with performance monitoringreceiver sensitivity degradation, 225–229, 226f, 227f, 228f, 229f ghost tones, 230–232, 230f, 231f, 232f scalability, 232–233, 237f Amplitude-shift keying (ASK), 138–140, 139f ANN, see Artificial neural network Arrayed waveguide grating (AWG) out-of-band noise measurement, 23–24, 24f wavelength-division multiplexed system demultiplexing, 235 Artificial neural network (ANN), asynchronous delay-tap sampling, 181, 181f ASE, see Amplified spontaneous emission ASK, see Amplitude-shift keying Asynchronous amplitude histogram (AAH) acquisition and processing averaging effects, 168–171, 169f, 171f sampling noise, 163–168, 164f, 165f, 167f, 168f systems, 148f, 149 eye diagrams, 146f optical signal-to-noise ratio effects, 147–148, 147f optical signal-to-noise ratio monitoring, 157–163, 159f, 160f, 161f, 162f, 163f overview, 145–149 phase-modulated signal optical performance monitoring techniques, 338–340, 338f, 339f Q-factor monitoring, 149–156, 150f, 151f, 152f, 153f, 154f, 156f reference asynchronous histogram, 159 Asynchronous delay-tap sampling application to higher-order formats, 188–190, 190f, 191 comparison of monitoring techniques, 177t

447

448

Index

Asynchronous delay-tap sampling (Continued) experimental measurements differential phase-shift keying systems, 183–185, 184f, 185f, 186t first-order polarization mode dispersion, 183 multi-impairment monitor, 183 network emulator, 182–183 system testing, 185–187, 186f, 187t, 188t extension to new impairments, 188, 189f overview, 175–176 pattern recognition, 180–182, 181f phase portrait, 177–180, 178f, 179f AWG, see Arrayed waveguide grating

B Bandpass filter (BPF), 36–37, 51, 52–54, 340 Beat noise analysis, optical signal-to-noise ratio monitoring operating principle, 43–44, 44f, 45f problems and limitations, 45–47, 46f, 47f solutions to problems frequency diversity technique, 51–52, 51f, 52f orthogonal polarization delayed homodyne technique, 49–51, 50f orthogonal polarization self-heterodyne technique, 52–54, 53f, 55f polarization diversity technique, 47–49, 48f synchronously gated signal, 55, 56f BER, see Bit error rate Binary phase-shift keying (BPSK), 193, 213–215, 214f BIP, see Bit-interleaved parity Bit error rate (BER), 8, 10–11, 21, 67–68, 82 asynchronous amplitude histogram for Qfactor monitoring, 149–156, 150f, 151f, 152f, 153f, 154f, 156f chromatic dispersion monitoring, 88, 89f long-haul transmission system, 430 polarization mode dispersion monitoring, 117–118, 118f Bit-interleaved parity (BIP), 261 BPF, see Bandpass filter BPSK, see Binary phase-shift keying

C CD, see Chromatic dispersion Chirped pilot tone, see Chromatic dispersion Chromatic dispersion (CD) asynchronous delay-tap sampling, 181, 181f, 182, 183–185, 188t differentiation from polarization mode dispersion, 90–92, 90f, 91f, 92f dispersion tolerance, 73t effects in presence of fiber nonlinearities, 73–74, 74f, 75f electronic digital signal processing monitoring in digital coherent receivers, 281–282, 281f, 283f, 284f, 285t, 286f fiber chromatic dispersion, 68–70, 69f, 70f, 71f long-haul transmission system, 428–429, 428f, 429f overview, 6, 67–68 systems limitations, 70–73, 71f, 72f zero-dispersion wavelength shifts and temperature, 76f Chromatic dispersion monitoring all-optical spectral analysis using nonlinear optics, 85–88, 86f, 87f coherent detection, 363–366, 364f, 365f coherent optical orthogonal frequencydivision-multiplexing system optical performance monitoring, 372, 376–378, 378f, 379f comparison of techniques, 93t electronic monitoring, 88–89, 89f group delay measurement between vestigialsideband signals, 80–81, 81f, 82f histogram monitoring, 81–84, 83f, 84f, 85f long-haul transmission system performance monitoring and compensation, 439–441, 440t miscellaneous techniques, 89–90 need, 75–77, 76f nonlinear optical performance monitoring, 307f, 308f phase-modulated signal optical performance monitoring techniques, 330–332, 331f, 341

Index

pilot tone monitoring chirped pilot tone monitoring, 249–251, 249f, 250f, 251f, 252f effect of self-phase modulation and polarization mode dispersion, 248–249, 248f operating principles, 246–247, 247f pilot tones carried by broadband light source, 252–256, 252f, 253f, 254f, 255f, 256f radiofrequency spectrum measurement, 77–79, 78f, 80f scenarios, 67 CMOS, see Complementary metal-oxide semiconductor Coherent optical systems coherent detection principles, 353–354, 353f, 354f coherent optical orthogonal frequencydivision-multiplexing system architecture, 357–360, 359f cyclic prefix, 357, 358f optical performance monitoring chromatic dispersion monitoring, 372, 376–378, 378f, 379f experimental results, 379–381, 380f optical channel estimation, 371–373 optical channel model, 369–371, 370f optical signal-to-noise ratio monitoring, 376, 377f polarization mode dispersion monitoring, 372, 376–378, 378f, 379f 16-QAM systems, 378, 379t Q-factor monitoring, 372–373, 376, 377f quaternary phase-shift keying, 375–378, 376t simulations, 373–375, 374f overview, 356–360 principles, 356–357, 358f single-carrier versus multicarrier systems, 360–361 historical perspective, 351–352 multicarrier modulation, 352–353

449

optical performance monitoring with coherent detection advantages, 352 chromatic dispersion monitoring, 363–366, 364f, 365f optical signal-to-noise ratio monitoring, 361–363, 362f, 363f polarization mode dispersion monitoring, 363–366, 364f, 365f receiver electrical equalization, 366–368, 366f, 367f, 368f, 369f single-carrier systemoverview, 354–356, 355f Coherent optical orthogonal frequency-divisionmultiplexing system (CO-OFDM) architecture, 357–360, 359f cyclic prefix, 357, 358f optical performance monitoring chromatic dispersion monitoring, 372, 376–378, 378f, 379f experimental results, 379–381, 380f optical channel estimation, 371–373 optical channel model, 369–371, 370f optical signal-to-noise ratio monitoring, 376, 377f polarization mode dispersion monitoring, 372, 376–378, 378f, 379f 16-QAM systems, 378, 379t Q-factor monitoring, 372–373, 376, 377f quaternary phase-shift keying, 375–378, 376t simulations, 373–375, 374f overview, 356–360 principles, 356–357, 358f single-carrier versus multicarrier systems, 360–361 Complementary metal-oxide semiconductor (CMOS), 356 Continuous-phase frequency-shift keying (CPFSK), clock/data synchronization in systems, 137, 138f CO-OFDM, see Coherent optical orthogonal frequency-division-multiplexing system CPFSK, see Continuous-phase frequency-shift keying

450

Index

Cross-phase modulation (XPM), 6, 7, 264, 305–306, 324 Cyclic prefix, orthogonal frequency-divisionmultiplexing coherent optical system, 357, 358f

D DCF, see Dispersion-compensating fiber DCM, see Dispersion-compensating module Degree of polarization (DOP) optical signal-to-noise ratio monitoring, 28–29, 29f phase-modulated signal optical performance monitoring techniques, 330 polarization mode dispersion monitoring, 111–117, 112f, 113f, 114f, 115f, 116f timing alignment monitoring, 129–131 Delay-tap sampling, see Asynchronous delay-tap sampling; Two-tap sampling Denial of service, 12f Dense wavelength division multiplexing (DWDM), 106, 302, 309–310 DGE, see Dynamic gain equalizer DGEF, see Dynamic gain-equalizing filter Differential group delay, see Polarization mode dispersion Differential phase-shift keying (DPSK), 10, 110–111, 110f, 137–138, 183–185, 184f, 185f, 306–307, 325f, 326f, 337f, 339f Differential quadrature phase-shift keying (DQPSK), 10, 25–26, 110–111, 110f, 111f, 134, 136f, 138–140, 139f, 325f, 326f Digital signal processing, see Electronic digital signal processing Dispersion coefficient, 69f, 70f Dispersion-compensating fiber (DCF), 33–34, 428–429 Dispersion-compensating module (DCM), 309–310, 428–429 Dispersion-shifted fiber (DSF), 69, 70f DOP, see Degree of polarization DPSK, see Differential phase-shift keying

DQPSK, see Differential quadrature phase-shift keying DSF, see Dispersion-shifted fiber DWDM, see Dense wavelength division multiplexing Dynamic gain equalizer (DGE), 23–24 Dynamic gain-equalizing filter (DGEF), 309–310

E EAM, see Electroabsorption modulator Eavesdropping, 12f EDFA, see Erbium-doped fiber amplifier Electroabsorption modulator (EAM), timing alignment monitoring, 128 Electronic digital signal processing monitoring in digital coherent receivers chromatic dispersion estimation, 281–282, 281f, 283f, 284f, 285t, 286f joint estimation of linear channel parameters, 280–295, 280f overview, 274–296 polarization mode dispersion estimation all-order polarization mode dispersion, 288–291, 289f, 290f, 291f polarization dependent loss estimation, 291–295, 292f, 294f, 295f polarization rotation, 287–288, 287f, 288f theory equalizer, 276–280, 276f, 277f, 278f optical channel, 275–276 monitoring in digital direct detection systems channel model for direct detection systems, 264–265 optical signal-to-noise ratio monitoring, 268–271, 270f, 271t reference parameter estimation, 271–274, 273f state-based equalization based on maximum likelihood sequence estimation, 265–268 overview, 261–263 prospects, 295–296

Index

Erbium-doped fiber amplifier (EDFA), 12–13 long-haul transmission system performance monitoring, 434, 434t operating conditions for optical signal-tonoise ratio estimation link-based monitoring, 22f, 57–59, 58f operating principle, 56–57 problems and limitations gain tilt variation, 59–61, 60f total power monitoring, 59, 60f Eye diagram long-haul transmission system, 431 polarization mode dispersion monitoring, 117f, 119f, 120f synchronous versus asynchronous, 146f

F FBG, see Fiber Bragg grating FEC, see Forward error correction Fiber Bragg grating (FBG), polarization mode dispersion monitoring, 108 FM, see Frequency modulation FM pilot tones, see Frequency modulation pilot tones Forward error correction (FEC), 261, 319 Four-wave mixing linear optical sampling, 208–209, 209f, 210f nonlinear bandwidths, 301 phase-modulated signal nonlinear phenomena, 324 Frequency diversity technique, optical signal-tonoise ratio monitoring, 51–52, 51f, 52f Frequency modulation (FM), radiofrequency spectrum measurement for chromatic dispersion monitoring, 77–79, 78f, 80f Frequency modulation (FM) pilot tones experimental setup, 243, 244f performance monitoring, 243–245, 244f principles, 243–245 Frequency-resolved optical gating (FROG), 195 FROG, see Frequency-resolved optical gating

451

G Ghost tones, amplitude modulation pilot tone monitoring, 230–232, 230f, 231f, 232f Gordon-Mollenauer noise, see Nonlinear phase noise Group velocity dispersion (GVD), 147–148, 150–151, 153, 156, 163 GVD, see Group velocity dispersion

H Histogram, see Asynchronous amplitude histogram

I Interferometer, see Mach-Zehnder delay interferometer Intersymbol interference (ISI), 152–153 ISI, see Intersymbol interference

L Linear crosstalk, 7 Linear optical sampling (LOS) balanced photodetection ninety-degree optical hybrid of polarization, 206–207, 207f waveguide ninety-degree optical hybrid, 207–208, 208f coherent detection implementations monochromatic local oscillator, 200 overview, 199–201, 199f phase-preserving temporal gating and monochromatic local oscillator, 201 short optical pulse, 200–201 principles, 196–205, 198f corrections quadrature amplitude correction, 209–210 quadrature phase correction, 210–211 data encoding in electric field of optical waves, 193–194, 194f digital phase tracking, 204–205, 205f

452

Index

Linear optical sampling (LOS) (Continued) direct photodetection, 208 four-wave mixing, 208–209, 209f, 210f optical performance monitoring amplified spontaneous emission characterization, 211, 212f digital processing of sampled electric field, 216–217, 217f electric field characterization from periodic sources, 217–218, 218f nonlinear phase noise, 213–216 nonlinear phase shift measurement, 216, 217f phase and amplitude noise measurements, 211–213, 213f, 214f, 215f overview, 9, 196 phase sensitivity monochromatic local oscillator, 202 pulsed gate and monochromatic local oscillator, 204 pulsed local oscillator, 203–204 phase-modulated signal optical performance monitoring techniques, 334–337, 335f, 337f polarization and wavelength sensitivity, 201–202 prospects, 218–219 temporal characterization of optical signals, 194–196 LO, see Local oscillator Local oscillator (LO), 193–194, 200, 202, 203–204, 263, 351 Long-haul transmission system bidirectional link components, 424f chromatic dispersion, 428–429, 428f, 429f optical amplifier, 426–427, 426f optical performance monitoring chromatic dispersion monitoring and compensation, 439–441, 440t erbium-doped fiber amplifier, 434, 434t fault detection and isolation, 443 functions and applications, 432–433 network availability and performance, 443 optical signal-to-noise ratio monitoring intermediate amplifier node, 437–438, 438f

optical channel analyzer, 435, 435f, 436f requirements, 436–437, 437f, 437t polarization mode dispersion monitoring and compensation, 440t, 441–442, 441f, 442f Raman amplifier, 434 real-time monitoring, 444 receiver, 433, 433t remote monitoring, 444 service-oriented system design, 442–445 system installation and upgrade, 444–445, 445f system performance penalties or margins, 438–442, 439t transmitter, 433, 433t wavelength-division multiplexed device monitoring, 434 optical transmitter, 425 performance measures bit error rate, 430 dispersion tolerance, 431–432 eye diagram, 431 optical signal-to-noise ratio, 430 Q-factor, 431 receiver sensitivity, 430 signal constellation diagram, 431 wavength accuracy and stability, 432 polarization mode dispersion, 429 Raman amplifer, 427, 427f wavelength division multiplexing, 425 LOS, see Linear optical sampling

M Mach-Zehnder delay interferometer (MZDI) chromatic dispersion monitoring, 79, 80f optical signal-to-noise ratio monitoring operating principle, 41–42, 41f problems and limitations, 42 solutions to limitations, 42, 42f Mach-Zehnder modulator (MZM) phase-shift keying signal modulation, 334–335, 335f, 337f timing alignment monitoring, 128, 129, 131–132

Index

Maximum likelihood sequence estimation (MLSE) state-based equalization for electronic digital signal processing monitoring, 265–268 timing misalignment mitigation, 141 Maxwell wave equation, polarization as source term, 303 MLSE, see Maximum likelihood sequence estimation MPDR, 129–132, 134 MPI, see Multipath interference MSPE, see Multisymbol phase estimation Multilevel phase-shift keying (xPSK), 13–14 Multilevel quadrature-amplitude modulation (xQAM), 13–14 Multipath interference (MPI), noise, 45–47 Multiple frequency measurement, optical signalto-noise ratio monitoring, 38–40, 39f Multisymbol phase estimation (MSPE), 263–264 MZDI, see Mach-Zehnder delay interferometer MZM, see Mach-Zehnder modulator

N NLSE, see Nonlinear Schro¨dinger equation Nonlinear birefringence, optical signal-to-noise ratio monitoring problems, 32–34, 32f, 33f Nonlinear optics optical performance monitoring techniques challenges cost, size, and complexity, 309–314 impairment isolation, 312–314, 313f, 314f sensitivity, 309–311, 309f chromatic dispersion monitoring, 307f, 308f comparison of types, 305t optical signal-to-noise ratio monitoring, 305–306 overview, 301–303 prospects, 314 Q-factor measurement, 308 overview, 303–304 phase-modulated signal nonlinear phenomena, 324

453

Nonlinear phase noise (NPN) compensation, 327–328, 329f linear optical sampling, 336f overview, 324 Nonlinear Schro¨dinger equation (NLSE), 264 Non-zero dispersion-shifted fiber (NZ-DSF), 69–70, 324 NPN, see Nonlinear phase noise NZ-DSF, see Non-zero dispersion-shifted fiber

O OADM, see Optical add/drop multiplexer OEO conversion, see Optical-electro-optical conversion OFDM, see Orthogonal frequency-divisionmultiplexing Off-center narrow filtering, optical signal-tonoise ratio monitoring, 37–38, 38f On-off keying (OOK), 10, 118, 132, 262 OOK, see On-off keying OPM, see Optical performance monitoring Optical add/drop multiplexer (OADM), 238 Optical cross-connect (OXC), 21, 235, 237–238, 237f Optical-electro-optical (OEO) conversion, 55 Optical long-haul transmission system, see Long-haul transmission system Optical performance monitoring (OPM) challenges, 3–4, 4f frequency domain monitoring, 9 network functionality impairment-aware routing, 10–12, 11f optical supervisory channel, 13 robustness and stability, 9 secure links, 12f, 14–15 transparency, 10, 10f optical channel monitoring, 4–5 optical impairments requiring monitoring, 5–9, 8f overarching vision, 2–3, 3f overview, 1–4, 1f security, 13–15 smart network operation, 13–15, 15f time domain monitoring, 8

454

Index

Optical semiconductor amplifier (OSA), 85 Optical signal-to-noise ratio (OSNR) asynchronous delay-tap sampling, 181, 181f, 182, 183–185 asynchronous histograms, 147–148, 147f degradation, 7 long-haul transmission system, 430 Optical signal-to-noise ratio monitoring asynchronous amplitude histograms, 157–163, 159f, 160f, 161f, 162f, 163f beat noise analysis operating principle, 43–44, 44f, 45f problems and limitations, 45–47, 46f, 47f solutions to problems frequency diversity technique, 51–52, 51f, 52f orthogonal polarization delayed homodyne technique, 49–51, 50f orthogonal polarization self-heterodyne technique, 52–54, 53f, 55f polarization diversity technique, 47–49, 48f synchronously gated signal, 55, 56f coherent detection, 361–363, 362f, 363f coherent optical orthogonal frequencydivision-multiplexing system optical performance monitoring, 376, 377f electronic digital signal processing, 268–271, 270f, 271t erbium-doped fiber amplifier operating conditions for estimation link-based monitoring, 22f, 57–59, 58f operating principle, 56–57 problems and limitations gain tilt variation, 59–61, 60f total power monitoring, 59, 60f interferometer-based monitoring operating principle, 41–42, 41f problems and limitations, 42 solutions to limitations, 42, 42f linear interpolation limitations, 24–26, 25f, 26f optical spectrum analyzer, 22–23, 22f, 23f out-of-band noise measurement, 23–24, 24f

long-haul transmission system intermediate amplifier node, 437–438, 438f optical channel analyzer, 435, 435f, 436f requirements, 436–437, 437f, 437t nonlinear optical performance monitoring, 305–306 optical transport networks narrowband spectral analyzer, 405–406, 406f polarization extinction, 406–407, 406f principles, 403–407, 403f, 404f, 405f phase-modulated signal optical performance monitoring techniques, 328–330 polarization-based monitoring polarization-nulling technique, 26–29, 27f, 28f problems and limitations nonlinear birefringence, 32–34, 32f, 33f polarization dependent loss, 34–35, 34f, 35f polarization fluctuation, 36, 36f polarization mode dispersion, 30f, 31f, 36–41 solutions to problems additional optical filtering, 36–41, 37f multiple frequency measurement, 38–40, 39f off-center narrow filtering, 37–38, 38f spectral state of polarization measurement, 40–41, 40f Stokes parameter analysis, 28–29, 29f simultaneous polarization mode dispersion monitoring, 109f Optical spectrum analyzer (OSA), optical signalto-noise ratio monitoring, 22–23, 22f, 23f, 24f Optical supervisory channel (OSC), 13, 395 Optical time domain reflectometer (OTDR), 416 Optical time-division multiplexed (OTDM) system clock recovery using timing misalignment of data pulses, 136–137 hybrid optical time-division multiplexed scheme for demultiplexing, 140 Optical transport network (OTN) business interface model, 385–386, 386f

Index

data networking and transport networking, 387–388, 387f generic service requirements, 386 information structure G.709, 394–395 G.872, 395 optical supervisory channel, 395 layer networks management functions, 394 optical channel layer network, 393 optical multiplex section layer network, 394 optical transmission section layer network, 393 lightpath provisioning connection verification, 397 network element characterization, 398 overview, 396–398, 398f modeling of multilayer networks characteristic information, 393 G.805 architectural components, 391t layering concept, 392 overview, 390–393, 391f partitioning concept, 392 trails and connections, 392 modeling principles control plane functions, 388 management functions configuration management, 389 fault management, 389–390 overview, 388–390 performance management, 390 top-level functional architecture, 388, 389f transport functions, 390 optical performance monitoring channel wavelength, 403 comparison of techniques, 420t gaps, 413–414 implementation issues accuracy, 411–412 event correlation and fault diagnostics, 417–419, 418f external versus embedded monitoring, 412, 418f intervals of measurement, 414–415

455

monitoring points, 413–414 risk management, 415–417 optical power, 403 optical signal-to-noise ratio monitoring narrowband spectral analyzer, 405–406, 406f polarization extinction, 406–407, 406f principles, 403–407, 403f, 404f, 405f overview, 400–410, 402f, 402t prospects, 419 Q-factor measurement bit error rate relationship, 408–409, 409f definition of Q-factor, 408 theory, 407–408, 408f signal quality supervision OTUk, ODUkT, and ODUkP, 409, 410f, 416f requirements, 401–403 service assurance five nines, 399 overview, 398–400 proactive maintenance, 400 reactive maintenance, 399–400, 399f services, 395–396, 396t Orthogonal frequency-division-multiplexing (OFDM), 68, 351–352, see also Coherent optical orthogonal frequency-divisionmultiplexing system Orthogonal polarization delayed homodyne technique, optical signal-to-noise ratio monitoring, 49–51, 50f Orthogonal polarization self-heterodyne technique, optical signal-to-noise ratio monitoring, 52–54, 53f, 55f OSA, see Optical semiconductor amplifier OSA, see Optical spectrum analyzer OSC, see Optical supervisory channel OSNR, see Optical signal-to-noise ratio OTDM system, see Optical time-division multiplexed system OTDR, see Optical time domain reflectometer OTN, see Optical transport network Out-of-band noise, measurement, 23–24, 24f OXC, see Optical cross-connect

456

Index

P PBS, see Polarization beam splitter PC, see Polarization controller PDF, see Probability density function PDG, see Polarization dependent gain PDL, see Polarization dependent loss Phase-modulated (PM) pilot tones chromatic dispersion monitoring chirped pilot tone monitoring, 249–251, 249f, 250f, 251f, 252f effect of self-phase modulation and polarization mode dispersion, 248–249, 248f operating principles, 246–247, 247f pilot tones carried by broadband light source, 252–256, 252f, 253f, 254f, 255f, 256f experimental setup, 240, 241f performance monitoring, 240, 241f, 242–243, 242f principles, 238–243, 240f Phase-modulated signals optical performance monitoring techniques asynchronous amplitude histogram, 338–340, 338f, 339f comparison of techniques, 343–344, 344f, 345t degree of polarization monitoring, 330 linear optical sampling, 334–337, 335f, 337f nonlinear phase noise compensation, 327–328, 329f optical signal-to-noise ratio monitoring, 328–330 overview, 327–344 pilot tone monitoring, 332–334, 333f, 334f polarization nulling, 330 power monitoring, 328–330 radiofrequency spectrum/tone monitoring, 327–343, 331f, 333f receiver-based monitoring, 342–343, 343f two-tap sampling, 340–342, 340f, 342f

performance amplified spontaneous emission noise, 321 chromatic dispersion, 321–322, 322f filtering and frequency offset, 323–324, 323f N-PSK signal generation and detection, 325–326, 325f, 326f nonlinear phenomena, 324 optical signal-to-noise ratio, 321 overview, 320–326, 320f polarization mode dispersion, 322 Pilot tones, see Amplitude modulation pilot tones; Frequency modulation pilot tones; Phase-modulated pilot tones; Phasemodulated signals; Single sideband pilot tone PM pilot tones, see Phase-modulated pilot tones PMD, see Polarization mode dispersion PMF, see Polarization-maintaining fiber Polarization beam splitter (PBS), 36–37, 47–48, 353–354 Polarization controller (PC), 108–109 Polarization dependent gain (PDG), 105–106 Polarization dependent loss (PDL) electronic digital signal processing monitoring in digital coherent receivers, 291–295, 292f, 294f, 295f optical signal-to-noise ratio monitoring problems, 34–35, 34f, 35f overview, 6, 105–106 Polarization diversity technique, optical signalto-noise ratio monitoring, 47–49, 48f Polarization fluctuation, optical signal-to-noise ratio monitoring problems, 36, 36f Polarization-maintaining fiber (PMF), 108–109, 129–131 Polarization mode dispersion (PMD) compensator, 38–39, 39f, 106 differential group delay, 6, 49–51, 102–103, 103f, 251f differentiation from chromatic dispersion, 90–92, 90f, 91f, 92f

Index

long-haul transmission system, 429 optical signal-to-noise ratio monitoring problems, 30–31, 30f, 31f origins, 102f overview, 6, 101–106, 102f wavelength dependence, 104 Polarization mode dispersion monitoring asynchronous delay-tap sampling, 181, 181f, 182, 183–185, 188t coherent detection, 363–366, 364f, 365f coherent optical orthogonal frequencydivision-multiplexing system optical performance monitoring, 372, 376–378, 378f, 379f comparison of techniques, 122t degree of polarization measurement, 111–117, 112f, 113f, 114f, 115f, 116f electronic digital signal processing monitoring in digital coherent receivers all-order polarization mode dispersion, 288–291, 289f, 290f, 291f polarization dependent loss estimation, 291–295, 292f, 294f, 295f polarization rotation, 287–288, 287f, 288f electronic monitoring, 117–120, 117f, 119f, 120f, 121f long-haul transmission system performance monitoring and compensation, 440t, 441–442, 441f, 442f miscellaneous techniques, 121 phase-modulated signal optical performance monitoring techniques, 331, 331f, 332, 341 radiofrequency tone measurement, 106–111, 107f, 108f, 109f, 110f, 111f simultaneous optical signal-to-noise ratio monitoring, 109f Polarization-nulling technique optical signal-to-noise ratio monitoring, 26–29, 27f, 28f, 39f phase-modulated signal optical performance monitoring techniques, 330 PRBS, see Pseudo-random bit sequence

457

Probability density function (PDF), state-based equalization based on maximum likelihood sequence estimation, 265–268 Pseudo-random bit sequence (PRBS), 44, 46f Pulse carver misalignment between pulse carver data modulator in return-to-zero differential phase-shift keyingsystems, 137–138 timing alignment monitoring, 128–137, 129f, 130f, 131f, 132f, 133f

Q QAM, see Quadrature amplitude modulation Q-factor, see also Bit error rate asynchronous amplitude histograms for monitoring, 149–156, 150f, 151f, 152f, 153f, 154f, 156f average value, 150, 151f chromatic dispersion monitoring, 88 coherent optical orthogonal frequencydivision-multiplexing system optical performance monitoring, 372–373, 376, 377f long-haul transmission system, 431 optical transport network measurement bit error rate relationship, 408–409, 409f definition of Q-factor, 408 theory, 407–408, 408f QPSK, see Quaternary phase-shift keying Quadrature amplitude modulation (QAM), 193, 375–378 Quaternary phase-shift keying (QPSK), 193, 215–216, 216f coherent optical orthogonal frequencydivision-multiplexing system optical performancre monitoring, 375–378, 376t

R Radiofrequency pilot tones, see Amplitude modulation pilot tones; Frequency modulation pilot tones; Phase-modulated pilot tones; Single sideband pilot tone

458

Index

Radiofrequency spectrum/tone monitoring chromatic dispersion monitoring, 77–79, 78f, 80f phase-modulated signal optical performance monitoring techniques, 327–343, 331f, 333f polarization mode dispersion monitoring, 106–111, 107f, 108f, 109f, 110f, 111f RAH, see Reference asynchronous histogram Raman amplifer, long-haul transmission system overview, 427, 427f performance monitoring, 434 Reconfigurable add/drop multiplexer (ROADM), 21, 35, 261, 302, 323 Reference asynchronous histogram (RAH), 159 Reflective semiconductor optical amplifier (RSOA), 252, 253, 255–256 Refractive index, fiber, 73 ROADM, see Reconfigurable add/drop multiplexer Routing and wavelength assignment (RWA), 11–12 RSOA, see Reflective semiconductor optical amplifier RWA, see Routing and wavelength assignment

S Sample-and-hold circuit, 164, 165f, 168f SBS, see Stimulated Brillouin scattering SCM transmission, see Subcarrier-multiplexed transmission SDH, see Synchronous digital hierarchy Self-phase modulation (SPM), 6, 7, 73, 85–87, 86f, 248–249, 252f, 324 Semoiconductor optical amplifier (SOA) chromatic dispersion monitoring, 87–88, 87f nonlinear effects, 301, 305–306 Signal constellation diagram, long-haul transmission system, 431 Signal-to-noise ratio (SNR), versus fiber dispersion, 74f Signal-to-sampling noise ratio (SSNR), 167, 167f

Single sideband (SSB) pilot tone, polarization mode dispersion monitoring, 256–258, 257f Single sideband orthogonal frequency division multiplexed signal, 51–52 Single-mode fiber (SMF), fiber chromatic dispersion, 68–70, 69f, 70f, 71f SMF, see Single-mode fiber SNR, see Signal-to-noise ratio SOA, see Semoiconductor optical amplifier SONET, see Synchronous optical network SOP, see State of polarization Spectral phase interferometry for direct electricfield reconstruction (SPIDER), 195 Speed of light, wavelength dependence in fibers, 69 SPIDER, see Spectral phase interferometry for direct electric-field reconstruction SPM, see Self-phase modulation SRS, see Stimulated Rayleigh scattering SSB pilot tone, see Single sideband pilot tone SSNR, see Signal-to-sampling noise ratio State of polarization (SOP) overview, 27–28 polarization mode dispersion origins, 102–103 spectral measurement, 40–41, 40f string length and polarization mode dispersion monitoring, 121 timing alignment monitoring, 129–131 Stimulated Brillouin scattering (SBS), 7 Stimulated Rayleigh scattering (SRS), 7 Subcarrier-multiplexed (SCM) transmission, polarization mode dispersion effects, 104–105, 105f Synchronous digital hierarchy (SDH), 55, 261 Synchronous optical network (SONET), 55, 261 Synchronously gated signal, optical signal-tonoise ratio monitoring, 55, 56f

T Telecommunications management network, see Optical transport network

Index

Timing alignment misalignment effects investigation clock/data synchronization in continuous-phase frequency-shift keying systems, 137, 138f misalignment between amplitude-shift keying and differential quadrature phase shift keying modulation, 138–140, 139f misalignment between pulse carver data modulator in return-to-zero differential phase-shift keying systems, 137–138 overview, 127–128, 127f misalignment mitigation hybrid optical time-division multiplexed scheme for demultiplexing, 140 maximum likelihood sequence equalizer, 141 remodulation scheme for wavelength division multiplexing passive optical network, 140–141 monitoring comparison of techniques, 135t optical time-division multiplexed clock recovery using timing misalignment of data pulses, 136–137 performance metrics, 128 synchronization for in-phase/quadraturephase data and data/pulse carver, 134–135, 136f synchronization for phase remodulation, 134 synchronization of pulse carver and data modulation, 128–137, 129f, 130f, 131f, 132f, 133f Timing jitter, 7 TPA, see Two-photon absorption

459

Transmission distance chromatic dispersion effects, 72f polarization mode dispersion effects, 104f Two-photon absorption (TPA) nonlinear bandwidths, 301 timing alignment monitoring, 133–134 Two-tap sampling, phase-modulated signal optical performance monitoring techniques, 340–342, 340f, 342f

V Vestigial-sideband (VSB) signals, group delay measurement between signals for chromatic dispersion monitoring, 80–81, 81f, 82f VSB signals, see Vestigial-sideband signals

W Wavelength-division multiplexing (WDM), 21–22, 28–29, 29f, 74, 351–352 long-haul transmission systemoverview, 425 device monitoring, 434 pilot tone monitoring, see Amplitude modulation pilot tones; Frequency modulation pilot tones; Phase-modulated pilot tones Wavelength division multiplexing passive optical network (WDM-PON), 134, 140–141 WDM, see Wavelength-division multiplexing WDM-PON, see Wavelength division multiplexing passive optical network

X XPM, see Cross-phase modulation xPSK, see Multilevel phase-shift keying xQAM, see Multilevel quadrature-amplitude modulation