Technologies and Systems for Access and Transport Networks (Artech House Mobile Communications)

Technologies and Systems for Access and Transport Networks

For a listing of recent titles in the Artech House Mobile Communications Series, turn to the back of this book.

Technologies and Systems for Access and Transport Networks Jan A. Audestad

artechhouse.com

Library of Congress Cataloging-in-Publication Data A catalog record for this book is available from the U.S. Library of Congress.

British Library Cataloguing in Publication Data A catalogue record for this book is available from the British Library.

ISBN-13: 978-1-59693-299-9

Cover design by Igor Valdman

© 2008 ARTECH HOUSE, INC. 685 Canton Street Norwood, MA 02062 All rights reserved. Printed and bound in the United States of America. No part of this book may be reproduced or utilized in any form or by any means, electronic or mechanical, including photocopying, recording, or by any information storage and retrieval system, without permission in writing from the publisher. All terms mentioned in this book that are known to be trademarks or service marks have been appropriately capitalized. Artech House cannot attest to the accuracy of this information. Use of a term in this book should not be regarded as affecting the validity of any trademark or service mark.

10 9 8 7 6 5 4 3 2 1

To my wife Synnøve

Contents Preface CHAPTER 1 Introduction 1.1 Evolution of Telecommunications 1.2 What Is Important Knowledge: Generic Technologies or Detailed System Overviews? 1.3 Composition of the Text CHAPTER 2 Networks and Services 2.1 Access, Transport, and Platform 2.1.1 Transport of Bits 2.1.2 Routing 2.1.3 Mobility 2.2 Types of Networks 2.2.1 Transport (or Backbone) Network 2.2.2 Access Networks 2.3 Stupid and Intelligent Networks 2.3.1 Concept 2.3.2 A Note on the Protocol Structure of the Internet 2.3.3 The Line of Demarcation Between Network and Application in the Internet 2.3.4 Network Neutrality 2.3.5 The Commercial Life Below the Demarcation Line 2.3.6 Is There Any Business for the Network Operator Above the Demarcation Line? 2.4 Overlay Access 2.5 Domains and Interworking 2.6 Heterogeneity 2.7 Real-Time and Nonreal-Time Systems 2.8 Backward Compatibility 2.8.1 Commercial Reasons 2.8.2 Technological Reasons 2.8.3 Political Reasons

xv

1 1 3 5

9 9 9 10 10 11 11 11 13 13 14 16 18 19 22 23 24 25 27 28 29 30 30

vii

viii

Contents

2.9 Standards 2.10 Access to the Common: Regulation of the Utilization of the Frequency Spectrum CHAPTER 3 Synchronization 3.1 Definitions 3.1.1 Synchronous 3.1.2 Asynchronous 3.1.3 Plesiochronous 3.1.4 Isochronous 3.1.5 Anisochronous 3.2 Reality Is Not So Simple: Bits, Words, Envelopes, and Frames 3.3 How to Acquire Synchronism: Phase-Locked Loop 3.3.1 Description of the Loop 3.3.2 Applications 3.4 Synchronization in Synchronous Networks 3.4.1 What Type of Synchronization Is Required? 3.4.2 Clock Hierarchies 3.4.3 Master-Slave (Link-by-Link) Synchronization 3.4.4 Signal Restoration: Elastic Store 3.5 Interconnection of Plesiochronous Networks: Application of Elastic Store 3.6 Synchronization of Envelopes of Constant Length 3.6.1 Direct Acquisition and Tracking of Envelopes 3.6.2 Acquisition and Tracking Using Error Detection: ATM Synchronization 3.7 Synchronization of Radio Systems 3.7.1 General Synchronization Sequences in TDMA and Random Access Systems 3.7.2 GSM: Timing Advance Procedure 3.7.3 Wireless LAN: Finding the Information in Sporadically Transmitted Frames 3.7.4 Satellite Systems: Managing Long Delays 3.7.5 Application of Scrambling and Interleaving

30 33

37 37 37 37 38 39 39 39 43 43 45 50 50 51 52 53 54 55 55 57 60 60 62 66 67 71

CHAPTER 4 Multiplexing

73

4.1 Multiplex Structures 4.2 Static Multiplexing: Frequency Division Multiplexing 4.2.1 Principle 4.2.2 Translation of Channels 4.2.3 Multiplexers and Demultiplexers 4.2.4 Distortion in FDM and WDM Systems: Intermodulation 4.2.5 Frequency Division Multiplexing in ADSL 4.3 Static Multiplexing: Time Division Multiplexing 4.3.1 Principle

73 74 74 74 77 78 80 82 82

Contents

4.3.2 Interleaving Patterns 4.3.3 European First-Order Multiplex 4.3.4 Higher-Order Multiplex 4.3.5 TDM Frame Alignment in Higher-Order Systems 4.4 Static Multiplexing: Synchronous Digital Hierarchy 4.4.1 Background 4.4.2 Multiplexing Structure 4.4.3 Compromise: Large Overhead Versus Flexibility 4.4.4 Pointer Mechanisms and Floating Payloads 4.4.5 Rate Adjustment of Plesiochronous Signals 4.4.6 Control Headers 4.5 Statistical Multiplexing 4.5.1 Invariant Frame Structure 4.5.2 Delimitation by Frame-Length Indicators 4.5.3 Delimitation by Flags CHAPTER 5 Multiple Access 5.1 5.2 5.3 5.4 5.5

Multiple Access Techniques Frequency Division Multiple Access Time Division Multiple Access Slow Frequency Hopping Code Division Multiple Access Direct Sequence Code Division Multiple Access 5.5.1 Coding Gain 5.5.2 Autocorrelation Properties 5.5.3 Composition of a DS-CDMA Transceiver 5.5.4 Interference and Channel Capacity 5.5.5 Power Control 5.5.6 Autocorrelation, Acquisition, and Tracking 5.5.7 Multipath Diversity 5.5.8 Application of DS-CDMA 5.6 Fast Frequency Hopping CDMA 5.7 Comparison of FDMA, TDMA, and DS-CDMA 5.8 Space Division Multiple Access 5.9 Random Access: Basic Theory and Applications 5.9.1 Aloha Techniques 5.9.2 Application of Simple Aloha Techniques: INMARSAT and GSM 5.9.3 Application of Carrier Sense Multiple Access: Ethernet 5.9.4 Application of Carrier Sense Multiple Access: WLAN 5.10 Random Access: Stochastic Behavior and Dynamic Control Procedures 5.10.1 Stochastic Behavior 5.10.2 Control Procedures 5.10.3 Application of the Control Procedures

ix

82 83 85 86 87 87 88 89 90 93 94 94 95 96 99

101 101 102 104 106 109 109 111 112 113 114 115 116 116 116 117 118 119 119 121 124 125 127 127 131 132

x

Contents

CHAPTER 6 Switching 6.1 Switched Networks 6.1.1 Terminology and Definitions 6.1.2 Switching Services 6.1.3 Circuit Switching 6.1.4 Connection-Oriented Packet Switching 6.1.5 Connectionless Packet Switching 6.1.6 General System Requirements 6.1.7 Number Analysis and Routing 6.1.8 Signaling Systems 6.2.2 Connectionless Networks 6.2 Switching Technologies 6.2.1 Introduction 6.2.2 Space-Division Switching: Crossbar Switches 6.2.3 Space-Division Switches Using Buffers for Cross-Connect 6.2.4 Time-Division Switching 6.2.5 Particular Switching Networks: Clos-Benes Networks 6.2.6 Particular Switching Networks: Application of Binary Switching Element 6.2.7 Construction of Switching Systems CHAPTER 7 Elements of Protocol Theory 7.1 7.2 7.3 7.4 7.5 7.6

Introduction Purpose of the Protocol Layer Services and Protocol Data Units Specification of Primitives Layering Hardcoding or Softcoding of the Protocol Data Unit 7.6.1 Hardcoding 7.6.2 Softcoding 7.7 Example 1: Layering and Encapsulation in the Internet 7.7.1 Layering 7.7.2 Network Layer: Encapsulation and Tunneling 7.8 Example 2: Protocol Structure of SS7 7.8.1 Signaling Network Architecture 7.8.2 Protocol Stack 7.8.3 Signaling Data-Link Layer (Layer 1) 7.8.4 Signaling Link Control (Layer 2) 7.8.4 Signaling Network Layer (Layer 3) 7.8.5 User Parts and Applications 7.8.6 Performance Requirements 7.9 Example 3: Protocol Structure of the Mobile Network 7.9.1 General Radio Interface

135 135 135 136 138 142 144 144 146 150 151 152 152 152 156 157 159 160 167

171 171 171 173 178 180 184 185 186 188 188 190 191 191 193 193 193 195 196 198 199 199

Contents

xi

7.9.2 Radio Resource Management, Mobility Management, and Media Management 7.9.3 Protocol Stacks CHAPTER 8 Cellular Land Mobile Systems 8.1 What Is a Cellular Network? 8.2 A Brief History of Public Land Mobile Telecommunications 8.3 Radio Propagation in Land Mobile Systems 8.3.1 Large-Scale Variations: Basic Wave Propagation Theory 8.3 Small-Scale Signal Variations: Fading 8.4 The PLMN Architecture 8.4.1 Objectives 8.4.2 Topology 8.4.3 Architecture of GSM 8.4.4 Location Management and Call Handling in GSM 8.4.5 Architecture of GPRS 8.4.6 All-IP UMTS 8.4.7 Mobile IP, Location Updating, and Packet Transfer in GPRS and All-IP UMTS 8.4.8 Paging, Location Updating, and the Size of the Location Area 8.5 Composition of the Radio Interface 8.5.1 Packet Radio Systems 8.5.2 Channel Coding in GSM 8.5.3 Logical Channels in GSM 8.5.4 Traffic and Control Channels in GPRS 8.5.5 Radio Interface of UMTS 8.6 Handover 8.6.1 Soft Handover 8.6.2 Hard Handover 8.7 Subscriber Identity Module 8.8 Adaptive Access 8.9 Smartphones and Information Security CHAPTER 9 Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites 9.1 9.2 9.3 9.4

Introduction Fixed Radio Access Networks Radio Relays Telecommunications Satellite Services 9.4.1 A Brief History 9.4.2 Satellite Orbits 9.4.3 Frequency Bands 9.5 Architecture of Communication Satellite Networks 9.5.1 Broadcast Satellite Systems 9.5.2 Fixed Satellite Systems 9.5.3 Mobile Satellite Systems

200 201

203 203 204 206 206 210 212 213 213 214 216 218 219 220 221 222 222 223 228 231 231 235 235 237 240 242 245

247 247 247 248 251 251 254 258 259 259 260 262

xii

Contents

9.5.4 Very Small Aperture Terminal Systems 9.6 Telecommunications Components of a Satellite 9.7 Propagation Characteristics, Noise, and Link Budgets 9.7.1 Attenuation 9.7.2 Noise 9.7.3 Example 1 9.7.4 Link Budget 9.7.5 Example 2 9.8 Tradeoffs 9.8.1 Cost 9.8.2 Other Tradeoffs 9.9 Mobile Satellite Communication 9.9.1 The INMARSAT System 9.9.2 Frequency Bands 9.9.3 Basic Architecture and Procedures 9.9.4 Antenna Tracking in INMARSAT-A, INMARSAT-B, and INMARSAT-Aero 9.9.5 A Note on Link Budgets

262 263 265 265 267 270 271 271 272 272 273 274 274 275 275 277 278

CHAPTER 10 Optical Communication Systems

281

10.1 Why Optical Systems? 10.2 Composition of Optical Networks 10.3 Optical Transmission Components 10.3.1 Fibers 10.3.2 Splitters, Combiners, and Couplers 10.3.3 Filters 10.3.4 Lasers 10.3.5 Modulation 10.3.6 Detectors 10.3.7 Amplifiers 10.3.8 Wavelength Converters 10.4 Optical Switching 10.4.1 Switching Devices 10.4.2 Packet Switching

281 282 283 283 286 288 290 292 293 293 294 295 295 296

Appendix: Loop Mathematics and Loop Components

299

A.1 A.2 A.3 A.4

299 304 307 309

Loop Mathematics Loop Components Acquisition Devices Numerical Example: Satellite System

Acronyms

311

Contents

xiii

Bibliography

317

About the Author

319

Index

321

Preface Telecommunications has undergone a huge evolution during the last decade. The evolution has taken place in the political, commercial, and technological arenas at the same time. In the political arena, the most important change took place in Europe in 1998, when all telecommunications was opened for free competition. Mobile communication had already been commercialized. Before 1998, telecommunications had been monopoly business, where the monopolies (or cartels) were operating alone in given geographical areas (e.g., a country or a city). Moving from monopoly to competition required strong market regulation by the government to prevent the incumbent from utilizing the market power it had built up on public investments for more than a century. The liberalization of the market had, of course, commercial implications. However, building traditional telecommunications systems required huge investments, so initially the market liberalization had only minor effects on competition. The forces that actually led to competition were the Internet and the technological evolution in computer science. The computational capacity per unit volume of silicon has doubled approximately once every 18 months for the last 30 years (Moore’s law). The amount of software available on computers has grown even faster. This allows us to construct more complex applications that are able to perform tasks that were impossible just a few years ago. As explained in Section 2.3, the Internet has altered the business model of telecommunications in a different and unexpected way. Prior to the Internet, the telecom business was managed by companies consisting of a single vertical structure offering access, transport, services, and even terminal equipment in a single subscription. The Internet has changed this business model entirely by splitting the business into independent parts. Operators called Internet service providers (ISPs) offer access and transport of bits while independent content and application providers (ASPs) offer the services. Of course, one company may be an ISP and an ASP at the same time, but being an ISP or an ASP are nevertheless two different business models. Voice and video streaming on Internet protocol (IP) removes the last traditional services from the telecom and broadcast providers. The Internet allows anyone owning a computer and a Web camera to produce television programs and distribute them to anyone who cares to view them. A lot of people do this already, and, deeming from the number of hits on various Web sites containing homemade video films, these services are popular.

xv

xvi

Preface

Despite of this complex evolution, the basic telecommunications technologies such as switching, access, multiplexing, wireless communication, and synchronization are almost unchanged. Moore’s law has allowed us to implement old ideas such as code division multiple access (or spread spectrum). When the concept was studied during the late 1940s and early 1950s, devices did not exist on which the idea could be implemented. It took more than 50 years before it was feasible to implement the idea on a large scale. The first real-time simulations of the modulator, the fading radio channel, and the demodulator of global system for mobile communications (GSM) had to be run on a Cray computer. The simulation of a sample lasting 4 seconds took about 1 hour on a workstation. This was in 1986. In 1991—thanks to Moore’s law—the modulator and demodulator were contained in small handheld GSM phones. When I see the specification of an entirely new system concept, I often get the feeling that I have seen it before. The reason, of course, is that the new concept makes use of methods and solutions that, possibly in a slightly different guise, have been used in previous systems. This does not imply that the system does not include entirely new ideas that have never before been exploited. Every idea has a first time. After that the idea may be reapplied, altered, expanded, and so on in order to construct new concepts. Understanding this dynamics of system design is important: if you can reuse something that has been done before, you will save time and sometimes even improve reliability of the new system. When I took over the course on access and transport networks at the Norwegian University of Science and Technology (NTNU) 5 years ago, I found that the students knew much about how concrete systems were composed and functioned but did not understand how the same technology was reused in different systems. For this reason, I changed the focus of the course away from the description of systems such as asynchronous transfer mode (ATM), GSM, universal mobile telecommunications service (UMTS), and Ethernet to the description of the baseline technologies that were used in the design of these systems. The response from the students—at least the most clever ones—was encouraging. Having participated in standardization and development of systems for almost 40 years, I could also show in a rather convincing way how we had “stolen” ideas from previous designs and put them together in new ways. GSM is almost a compendium in this way of working. The present text has been developed, in particular, from dialogues with my students. They have offered many suggestions concerning what is important and what can be excluded from the text. Therefore my foremost acknowledgments go all these students. I am also grateful for many suggestions from colleagues in Telenor and NTNU concerning the contents of this book. My final acknowledgments go to all the colleagues I have had during more than 30 years of international standardization and system development. For me, this represented a vast arena of knowledge that is the basis for this book!

CHAPTER 1

Introduction 1.1

Evolution of Telecommunications There are two particular events that have changed telecommunications during the last 25 years. These are the introduction of automatic mobile communications around 1980 and the commercialization of the Internet in the early 1990s. The evolution is illustrated in Figure 1.1. The Internet gradually replaced the telex service during the 1990s. The telex service offered a method by which text could be transferred between teletypewriters. The system operated at a speed of 50 bits per second (bps) and each symbol consisted of five information bits, one start bit, and one and a half stop bit—or 7.5 bits altogether. The telegraph service using the Morse alphabet lasted until 2000 because the service then was no longer mandatory for ships in international waters by the Safety of Life at Sea (SOLAS) convention of the United Nations. The service is now entirely replaced by the more reliable maritime satellite services, as well as maritime VHF and HF telephony. The telecommunications operators developed the packet data transmission service called X.25 [named after the International Telecommunication Union (ITU) recommendation where the service is specified]. This service was replaced by the Internet during the 1990s, even though the telecommunications operators had invested large sums in the implementation of the service. The Internet was a cheap alternative to X.25 that moved the control of the service away from the 1970

1980

1990

2000

2010

Broadcast Telex Telegraph

Convergence

Telephone Mobile telephone X. 25 Data transmission ARPA

Internet

Internet + Web WWW

Telecommunications

IT

Sensors/RFID

Content

µelectronics sensors

Four eras and four dominating industries

Figure 1.1

Evolution of telecommunications.

1

2

Introduction

bureaucratic telecommunications operators. With the Internet, the users can configure their own services. One of the most important evolutions taking place at the moment is the expansion of sensor technology, including the radio frequency identification (RFID). It is expected that the microelectronic industry will take the lead in the future evolution of telecommunications. One reason for this is that it is estimated that there are more than 1,000 times as many autonomic devices containing central processing units (CPUs) as there are personal computers, databases, servers, and mainframe computers. The volume of autonomous machine-machine interactions increases rapidly and is expected to constitute a large part of the future telecommunications traffic, both locally and remotely. Some characteristics of the new traffic may include support of frequent and very short transactions, micropayment for use of computation facilities (grids and agent networks), and information security related to anonymity, nonrepudiation, global access control, and nondisclosure of processing algorithms in an open processing environment. Until about 1985, the telecommunications operators were in charge of the telecommunications business. The business was then mainly concerned with telephony and broadcast. For the next 10 years the information technology industry determined much of what was taking place in telecommunications. During these years, data communication rose and matured. The Web allowed everyone to create and distribute content. This put the content industry in the driver’s seat from about 1995. The content industry has changed telecommunications from being a pure telephone system (or person-to-person interaction) to becoming a system that supports all kinds of data communication and, in particular, dissemination of content and information (person-machine interaction). Now the new role of machine-machine interactions enabled by the microelectronic industry may shape the telecommunications industry further. Still, there are three separate telecommunications networks: • • •

The telephone network supporting fixed and mobile telephone services; Broadcast networks; The Internet.

This situation is about to change. Since the early 1970s, the telecommunications operators have been studying different ways in which all telecommunication services could be supported by a single digital network. The first attempt was the integrated services digital network (ISDN) developed during the late 1970s and the early 1980s. This attempt failed because the ISDN is a circuit-switched network not capable of incorporating packet-switched data communication. However, the ISDN specified how subscribers can be connected to the transport network on a digital access circuit supporting all types of digital services, including Internet access, and allowing several types of terminals to share the same subscriber line. ATM was developed during the late 1980s and the early 1990s in order to support any mixture of circuit-switched and packet-switched communications in a single system. ATM integrates all services in a single network, but it never became a success because it cannot compete with the Internet in terms of switching costs.

1.2 What Is Important Knowledge: Generic Technologies or Detailed System Overviews?

3

Furthermore, asymmetric digital subscriber line (ADSL) offers sufficient bandwidth on the user access so that the ATM technology also became too expensive on the subscriber line. If large bandwidth is required in the access network, for example, to a local area network, this can simply be supported by an optical link and standard Internet switching both in the LAN and in the network. A separate technology such as ATM is not required for this purpose. The evolution taking place now is that service integration is finally being achieved by merging all data services, telephone services, and broadcast services in the Internet, as illustrated in Figure 1.2. The connectionless IP network is capable of offering reliable and high-quality, real-time services. This has resulted in voiceover-IP (VoIP) and video-over-IP services. In addition, the 3G mobile network evolves toward an all-IP version, where all information is sent as IP packets on the radio path. This evolution will have deep impact on the telecommunications business as we shall see in Section 2.3.

1.2 What Is Important Knowledge: Generic Technologies or Detailed System Overviews? One question is as follows: do we need to understand all the basic technologies of telecommunications since the convergence is leading toward a simple network? The answer is affirmative because the technologies are not becoming obsolete even if some of the earlier networks are being replaced by the Internet. This leads directly to the motivation for writing a text that focuses directly on the technologies, thereby shoving the system knowledge to the background. What then about the details and the functioning of systems in actual use? A book on access and transport networks can, of course, describe each system in terms of architecture, design, and details concerning protocols and information exchange. In other words, the focus may be on the detailed description of each system: how it is made, how it works, and what it does. Internet, ISDNs, ATM networks, 2G land mobile systems, 3G land mobile systems, and wireless local area network (WLAN) systems may all be described separately. Systems that are still on the drawing board may also be included in order to avoid overlooking a future evolution that may come. 1970

1980

1990

2000

2010

Broadcast Telephone ISDN

Video over IP Voice over IP All-IP

ATM Data

Figure 1.2

Convergence.

4

Introduction

One problem with this approach is that systems are replaced by new systems. A few years ago, the ATM technology was a central issue in a course focusing on actual systems. Today, detailed knowledge of ATM must be regarded as rather peripheral, though the technology is still used. GSM is still an important topic because the system is in widespread use all over the world, but in a few years time GSM will become obsolete and, therefore, no longer of general interest. Having taken part in the early phases of the development of several complex systems such as maritime satellite communications, automatic land mobile systems, and intelligent networks, my experience is that the most important element leading toward successful design is the understanding of the basic technologies that may or may not be useful in the new system. It is not always evident that the same basic technology is often applied in different and unrelated systems. For the system designer, it is important to understand how a technology can be reused fully or partly in a new design and when an entirely new approach must be found. Therefore, the focus I prefer is to consider the basic technologies rather than the systems in which they are used. The systems, or rather particular features or components of the system, are described in order to illustrate how a particular technology may be exploited in order to achieve a certain result. A technology that is 30 to 50 years old often appears in current system designs. Code division multiple access (CDMA) is a good example. The detailed mathematical description of direct sequence and fast frequency hopping CDMA was fully developed during the 1950s. At that time, the technology was usually known as spread spectrum multiple access (SSMA). The principles had also been demonstrated experimentally. However, it is just recently that this technology has become mature for large-scale production (3G mobile systems). The reason is the tremendous amount of computation required for detection and synchronization of such signals. This is now possible in even small devices such as mobile phones. The reason that a particular advanced technology is not applied is often commercial rather than technical: the technology may simply be too expensive. One example is the smart-house technologies that require remote control and sensing of room temperature, which can thereby reduce heating expenses. The technology is simple and has been available for more than 20 years but the equipment at the user site has been to expensive compared to the reduction of the electricity bill. It took more than 20 years before public key infrastructure (PKI) and electronic signature became commercially feasible. The way in which PKI can be implemented has been known since the early 1980s. The first time I heard about trusted third parties and public key escrows was in the early 1980s. In the early 1990s, one hot item—even resulting in Ph.D.s—was the discussion concerning who had enough trust to own such devices. Still it took 10 years before the first PKI infrastructure was realized. The question of trust is still unresolved. WiMax offers an alternative implementation of the fixed subscriber line. This is a modern realization of a radio access technology being studied and tested during the 1970s. However, it soon became evident that the technology was too expensive at that time compared to alternative subscriber lines. In many implementations, the WiMax technology is still too expensive, even for a new operator that wishes to establish an independent access network. The alternative of leasing access from an incumbent operator may be cheaper because the telecommunications regulatory

1.3 Composition of the Text

5

authorities are fixing maximum prices for so called local loop unbundling (LLUB) that quite often are cheaper than building a new access. Random access, including the control procedures applied in WLAN and Ethernet, was analyzed by Kleinrock and others some 30 years ago. All the “modern” switching and multiplexing methods of digital signals were developed almost 40 years ago, and some of these technologies are recently being extended to optical switching and multiplexing. The IP technology has been with us for the last 35 years. Automatic (or cellular) mobile communication with full roaming capabilities and handover was put into operation in 1981. What is really new is that the perpetual evolution in microminiaturization and computing enables us to implement more and more complex systems. Furthermore, particular systems such as GSM and telephone switching are at one stage becoming obsolete and replaced by new (but not necessarily better) systems. The detailed knowledge of these systems is thus only of limited value. However, the technology on which they are based does not become obsolete but may be reused in entirely different systems. Frequency division multiplexing (FDM) has been regarded as an obsolete technology in the telecommunications network for a long time. The technology is now reappearing in a slightly new guise in optical networks, where it is called wavelength division multiplexing (WDM), and in broadband access networks, where FDM is used to increase the effective bandwidth that can be supported by the twisted pair (direct multitone ADSL). ATM is a technology that is disappearing from the network and as such should not be of interest in a course like this. However, ATM contains some features that may be reused in new designs. Two such features are the use of error detection to synchronize the data stream and the use of length indicators to multiplex different information streams into a common cell structure. The latter method is used in several systems, but the way it is done in ATM is easier to describe and simpler to understand. For these reasons I have focused on the basic technologies applied in access and transport networks rather than the actual systems. Actual systems are used as examples in order to show how the technology is used in particular circumstances. The description of particular systems is contained in numerous textbooks and standards documents, and the reader is referred to such literature in order to study the details of these systems.

1.3

Composition of the Text The book consists of 10 chapters as follows. Chapter 2 contains general definitions and explains some particular features of telecommunications systems, such as the distinction between intelligent networks (ISDNs) and stupid networks (Internets), domain structures, and overlay access and virtual networks. The chapter is also concerned with problems such as real-time operation, heterogeneity, backward compatibility, and standardization. Chapter 3 is about synchronization. One important item is the description of the phase-locked loop (PLL). The PLL is one of the most important components in digital networks and is used for bit timing acquisition, carrier acquisition and coherent demodulation, frequency synthesis, and many other applications. PLLs are

6

Introduction

included in multiplexing equipment, signal regenerators, satellite systems, radio relays, land mobile terminals, and so on. A general description of the loop is contained in the main text. The loop mathematics and construction details of analog loops are contained in the Appendix. A large number of applications of synchronization are described. These include the interconnection of synchronous and plesiochronous networks, synchronization in ATM where the error correction mechanism is used for maintaining cell synchronism, synchronization of TDMA satellite systems, timing advance in GSM, and signal detection in WLANs. Chapter 4 describes several multiplexing methods used for static and statistical multiplexing. Static multiplexing includes frequency division multiplexing, time division multiplexing, and the synchronous digital hierarchy (SDH). Statistical multiplexing methods include systems with constant length envelope (ATM), use of length indicators (also used in ATM), and variable length envelopes using flag delimitation and transparency mechanisms. Chapter 5 is concerned with multiple access; that is, techniques that allow several sources to share a common medium. The basic methods of frequency division, time division, and code division multiple access are explained in detail. The chapter also contains an introduction to random access explaining how the method is applied in satellite systems, GSM, Ethernet, and WLAN. One important part is concerned with the stability of random access channels and the methods that can be applied to avoid channel saturation. The particular methods used in WLAN, Ethernet, the Internet, and other systems are explained. Chapter 6 is concerned with switching systems. The chapter consists of two parts. The first part is concerned with network aspects of switched networks explaining how routing and switching takes place in circuit switched networks (ISDN), connection-oriented packet switched networks (ATM), and connectionless packet switched networks (Internet). Features such as number analysis are also explained in relation to the different technologies. The second part describes in general terms how space and time division switches function. Then a more detailed description of particular switching networks is provided, including the general Clos-Benes network and the application of binary switching matrices in fast switches for ATM and optical networks. Chapter 7 contains the basic elements of protocol theory. Protocol theory is basic knowledge required for understanding signaling systems and data transfer protocols. The chapter contains three examples: •

•

•

Embedding and tunneling in the Internet in order to support mobile IP and particular network related protocols. The structure of Signaling System No. 7 (SS7). This signaling system is used in the ISDN/telephone network and in mobile networks in order to support the interaction between the different entities making up the mobile network. The protocol structure of GSM shows how complex protocol stacks may be required in order to support a large number of functions in the different environments in which the system is embedded.

1.3 Composition of the Text

7

Chapter 8 describes public land mobile systems. The chapter contains both general information such as radio propagation phenomena and generic network architecture, as well as details concerning GSM (2G), general packet radio service (GPRS) (2.5G), and UMTS (3G). One particularly important goal is to show similarities and differences between these systems. The new evolution toward software radio, led by the terminal manufacturers, is also explained. Chapter 9 is concerned with line-of-sight radio communication systems. The chapter contains two brief sections on WiMax and radio relays, respectively. WiMax is a technology based on the Institution of Electrical and Electronics Engineers (IEEE) WLAN standards that is about to be implemented in the access network. WiMax may change the telecommunications market entirely. The major part of the chapter is concerned with fixed and mobile satellite communication in the transport network and the access network. Though optical fibers have replaced several intercontinental satellite systems, there are areas where satellite communication cannot be replaced easily by other technologies. Chapter 10 describes briefly the various components of optical communication systems. Several of these components and systems are still not commercially available because of cost, size, and manufacturing complexity.

CHAPTER 2

Networks and Services 2.1

Access, Transport, and Platform The basic composition of telecommunications networks is shown in Figure 2.1. This configuration applies to all types of networks: the telephone network, the ISDN, data networks, broadband networks, the Internet, and so on. There is no structural difference between networks for different purposes at this level of abstraction. The difference is apparent when we consider construction details. The simple subdivision consists of three elements: •

•

•

The access network connects the terminals or users to the transport network and transports bits across the user-network interface. The access may be complex and contain much functionality, such as in mobile systems, or it may be simple, such as in the fixed network. The access network may consist of several technologies in tandem (e.g., Ethernet, WiMax, and local optical fiber). The transport network connects one access to another access via switching devices and other machines in the platform. The main purpose of the transport network is to transfer bits between two access networks. The platform routes the call from the origin to the destination. The software and hardware in the network and in the terminals cooperate in performing the processing of services and applications. The need for processing in the network depends on whether the network is stupid or intelligent (we’ll discuss these terms later) and the types of services offered.

The network consists of routing devices (e.g., ISDN exchanges in the telephone network or routers in the Internet) that offer three types of services: transport of bits, routing, and mobility. 2.1.1

Transport of Bits

This is the main function of the access network and the transport network. The technologies used for transport of bits are usually different in the access network and in the transport network. The most common technology used in the transport network is optical fibers with bandwidths of several gigabits per second (Gbps). But other technologies offering less bandwidth— coaxial cables, radio relays, and satellites—are also used extensively, though some of them are gradually being taken out of use and replaced by optical fibers where possible.

9

10

Networks and Services

Terminal

Terminal Platform

Access network

Transport network

Access network

Network

Figure 2.1

Elements of the telecommunications networks.

The technologies used in the access network are much more diverse: twisted pairs, coaxial cables, optical fibers, WiMax, GSM, UMTS, WLAN, Bluetooth, satellite systems, and several other technologies. Some of these technologies are considered in more detail in other chapters. 2.1.2

Routing

Routing is the core function of the switching process. The routing function selects the output of the switch on which a datagram or a telephone call is to be forwarded in order to reach the destination. How this is done is explained in Chapter 6. With the Internet, routing is the functionality of IP and is a rather simple process requiring simple switches. With ISDN, the routing requires complex switching functions. One reason that the Internet and ISDN are different is that IP is a connectionless protocol that applies statistical switching, while ISDN is a connection-oriented protocol requiring separate signaling functions for establishment, management, and release of connections. Switching also includes several processes in addition to routing that can be applied to the call: access control such as barring access to or from certain users, nondisclosure of source addresses, and measuring call data for determining charges and for building traffic statistics. These processes are generally more complex with ISDN than the Internet. This is another reason why ISDN exchanges are so complex. 2.1.3

Mobility

Mobility allows terminals or users to roam between different access points. Mobility is offered by GSM, UMTS, WLAN, and mobile IP. GSM and UMTS offer continuous or nondisruptive mobility, while mobile IP offers discrete mobility. WLAN offers something in between: continuous mobility within a WLAN zone and discrete mobility between zones. Personal computers, personal digital assistants (PDAs), smart phones, and mobile phones are equipped for several access technologies.

2.2 Types of Networks

11

Personal computers support any selection of access interfaces: cable connection, GSM/GPRS/UMTS, WLAN, Bluetooth, and WiMax.

2.2

Types of Networks 2.2.1

Transport (or Backbone) Network

The most important technologies applied in the transport network are as follows: •

•

•

Optical fibers. These systems have replaced most other long distance systems during the last few years. Geostationary satellites. Some of the largest systems have been (or are about to be) replaced by optical systems because the latter are cheaper, offer larger bandwidth, and are more reliable than satellite systems. However, the satellite systems are still used on several intercontinental routes where the traffic is low or where it is too expensive to install optical fiber. In addition, there are several domestic systems covering areas that are otherwise impossible to reach or where alternative systems are too expensive (such as Australia, Canada, Indonesia, and Brazil). Radio relays. The radio relay systems are in general cheap, reliable, and easy to establish or rearrange. The bandwidth can be any multiplex rate between 2 Mbps and 640 Mbps. These systems are used in applications such as feeder links in the local part of the network, in mountainous areas where it is too expensive to provide optical fiber systems, and in earthquake areas since it is easy to reestablish the link after an earthquake.

These systems are described in later chapters. 2.2.2

Access Networks

The access networks can be divided into three classes: public fixed access systems, public mobile access systems, and local area access systems. Public fixed access systems include the following: •

•

1.

Twisted pairs (or copper lines) are by far the most common access system in the fixed network. The twisted pair has a very long technical lifetime (in excess of 50 years). The twisted pair supports broadband up to 10 Mbps in the form of ADSL and other digital subscriber line of type x (xDSL)1 technologies. Optical fibers are installed in areas where twisted pairs do not exist and are also used to replace twisted pairs elsewhere. These systems are still rather expensive, which is one reason for the survival of the “obsolete” twisted pair technology.

DSL stands for digital subscriber line. The A in ADSL stands for asymmetric because different bandwidth is allocated to the uplink (from the user) and the downlink (toward the user).

12

Networks and Services •

•

•

•

Coaxial cables are still abundant in cable television systems, though more and more systems are being replaced by optical fiber. However, coaxial cables have a very long technical lifetime (50 years or more), so there must be strong commercial arguments for replacing them by fibers—wider bandwidth and support of duplex services (telephony and the Internet) are such arguments. Coaxial cable can support a bandwidth in excess of 200 Mbps. Broadcast satellite systems can offer broadband Internet services on the downlink to the user. Narrowband Internet may be used in the opposite direction for providing full duplex services. WiMax is a fixed point-to-point radio system (also capable of mobile communication) doing much the same job as twisted pairs but with a higher bandwidth (100 Mbps). WiMax is still an expensive technology, though the equipment cost is dropping. WiMax has become competitive in several applications, such as connecting WLAN hotspots to the transport network and as access network in new housing regions. Electricity modems are used to provide telecommunications services over the local electrical grid. Bandwidth of a few megabits per second is possible. However, the technology is still expensive and is used only rarely.

Public mobile access systems comprise the following: •

•

•

GSM, GPRS, and UMTS (3G) land mobile communications. Competing technologies based to the UMTS system exists in the United States and Japan. These systems are also introduced as competitors to 3G in Europe—for example, in the 450-MHz band previously used for the Nordic Mobile Telephone System (NMT) and other early mobile systems. Maritime and aeronautical satellite systems employing four geostationary satellites to cover all ocean areas and flight routes except the Polar Regions (i.e., coverage between latitudes approximately +70 degrees and −70 degrees). The same satellites are also used for land mobile communication to remote areas, for relief and rescue operations, and for expeditious establishment of broadband access systems (e.g., on-the-spot television reporting). Low orbit satellite (LEO) systems have been tried (Iridium, Globalstar, and other systems) but were not competitive with GSM or other public land mobile systems for general mobile communication. The Iridium consortium and Globalstar went bankrupt after having commenced full service in 1999 and 2003, respectively. However, the satellites were later sold to other companies reestablishing the services. The systems offer telecommunications to governments, the oil industry, scientific explorations, relief operations, and travelers. By the end of 2005, Iridium had 142,000 subscribers. The Teledesic, originally planned with 824 LEO satellites, later downscaled to about 300 satellites, offering “IP in the sky,” was never realized because the company was scared off by the bankruptcies of the other companies. However, Teledesic is still regarded as an alternative to the terrestrial Internet in the future.

2.3 Stupid and Intelligent Networks

13

Local area access networks comprise the following: •

•

•

•

2.3

Ethernet is a fixed local area access network using twisted pair, coaxial cable, or optical fiber as transmission medium. Wireless LANs in one shape or another (e.g., IEEE 802.11) offering short distance communication are emerging rapidly in the unlicensed frequency band around 2.4 GHz. Several related technologies providing larger bandwidth are under development. VSAT systems are local area networks interconnected by geostationary satellites. Bluetooth interconnects devices locally in addition to offering a communication port to external networks (e.g., a WLAN or an Ethernet).

Stupid and Intelligent Networks 2.3.1

2

Concept

The Internet is a stupid network. By this we mean that the Internet is offering very few services to the users. The Internet offers primarily routing and transfer of datagrams. In addition, the Internet manages multicast addressing and routing, allocates bandwidth, and supports mobile IP (tunneling). The routing and delivery of datagrams is based on “best effort”; that is, there is no guarantee that a datagram ever reaches the destination or is lost because of congestion in the network. With the Internet, all intelligence in terms of myriads of applications is in the hosts (PCs, databases, servers, sensors, or other computing devices). The network is stupid but the terminal is intelligent. In other words, the Internet contains a stupid core but an intelligent periphery. On the contrary, the ISDN/telephone network is intelligent, while most of the terminals connected to it are stupid, supporting very few applications. The network is offering several intelligent services, such as barring of incoming and outgoing calls based on time, cost, origin or destination, call redirection, call waiting, recall services, premium rate charging, toll-free services, alternate charged party, charge sharing, nongeographic routing, centralized queuing, distributed call desk, auto-answering services, voice mail, conditional routing, and many more. Some of these services are performed in separate devices called intelligent nodes. The core of the ISDN/telephone network is intelligent, while the periphery is stupid. GSM, GPRS, and UMTS offer services to both stupid and intelligent terminals and represent a transition between intelligent and stupid networks. The mobile terminal contains one or more CPUs, and therefore the mobile phones belong to the category of intelligent terminals since several services may be designed in the terminal rather than in the network. As long as these networks offer telephone services in 2.

These terms were introduced in the Telecommunications Information Networking Architecture (TINA) project in order to distinguish between two fundamentally different types of networks. The term stupid network was introduced at about the same time by David Isenberg, “The Rise of the Stupid Network,” Computer Telephony, August 1997.

14

Networks and Services

its core, the mobile network contains both an intelligent core supporting supplementary services for telephony and an intelligent periphery supporting data transmission and additional capabilities. The all-IP 3G network (delivery 5 of the UMTS specification) offers only IP-based communication (including IP telephony) over the access so that the core of this network no longer will contain intelligent functions. VoIP or IP telephony in the Internet is a telephone service where the intelligent services of the network can no longer be supported unless additional functionality is added to the Internet and the Internet protocols. The owner of the IP network (or ISP) may offer particular handling of IP telephone calls similar to that of intelligent network nodes in separate servers owned by the ISP. However, these intelligent services are services above the demarcation line described next and thus belong to the periphery of the network. The intelligent services are implemented in servers and databases. In summary, a stupid network has a stupid core supporting an intelligent periphery (the Internet); an intelligent network has an intelligent core supporting a stupid periphery (the telephone network). The trend is that all telecommunications networks are developing in the direction of stupid networks. The era of intelligent networks may soon be over. 2.3.2

A Note on the Protocol Structure of the Internet

The protocol structure of the Internet is explained in this section so that the main part of the text can be understood without referring to other literature where the Internet protocols are described (see also Chapter 7 for an introduction to protocol theory). This description contains only the elements and details of the protocol that are required to understand the main arguments presented in the following discussion. The protocol structure of the Internet is shown in Figure 2.2. The figure shows two computers containing application software communicating over a network consisting of two routers (R) in order to perform a common task (for example, a Web search where one computer contains the browser and the other computer contains the search engine). The protocol only ensures that the transaction between the two computers can take place. The computers are interconnected by a protocol consisting of several layers, where each layer is a protocol in its own right. The layers are as follows. The lower layers (link in the figure) support reliable transfer of bits between one node and the next. (The link protocol may consist of several layers. This point is not important here.) The link protocol may be different on different links depending on the characteristics of the transmission medium. The link protocols are different in, for example, optical core networks, GSM, UMTS, WLAN, and fixed broadband accesses. The link only analyzes the link header(s) and not the information contained in the information field. The information field of the link layer contains the IP datagram. IP is a network protocol whose primary task is to route bits from one terminal to another. The IP protocol may be identical across the whole connection (e.g., only IP version 4) or consist of sections with IP versions 4 and 6 in tandem. This can be done since the IP header is analyzed at each router so that the router can determine which actions should be taken, including protocol conversion and tunneling (see Section

2.3 Stupid and Intelligent Networks

15

Access

Access

Application software

Application software

Transport network R

R

Application

Application

TCP/UDP

TCP/UDP

IP

IP

IP

IP

Link

Link

Link

Link

(a)

Link header

IP header

TCP/UDP header

Application protocol and application content

(b)

Figure 2.2

(a, b) Protocol hierarchy in the Internet.

7.7.2 for tunneling of IPv6 across IPv4 networks), when the IP packet is forwarded on the next link. When forwarding the packet, the router then creates a new IP header containing information inserted by the router (e.g., a new value of the time-to-live parameter) and header parameters copied from the received packet (e.g., IP addresses). The router does not analyze the content of the information field of the IP packet.3 However, the IP packet contains a parameter identifying which protocol the information field contains—another IP,4 transmission control protocol (TCP), or user datagram protocol (UDP) (in IP version 4 this is the Protocol field; in IP version 6 this information is contained in the Next Header field—see the specialized IP literature for further details). This information is required by the receiving terminal so that it can identify which software must be activated in order to interpret the content in the information field (e.g., the software required for handling TCP or the simpler software for handling UDP). The router may use the next header information so as to handle datagrams containing TCP or UDP differently (routing selection and buffer 3.

4.

Except the port number contained in the TCP/UDP header (shared address). However, this is not important for the discussion that follows, though it allows the ISP to have some knowledge of what the protocol contains. The shared address field allows address extension of the IP number by using one of the port addresses for the extended IP number (see the specialized IP literature for how this is done). In order to support mobile IP or security (IPsec) the information field contains another IP protocol (see Section 7.7.2). The embedded IP protocol may then contain a Protocol/Next Header field indicating that the information field contains a TCP or UDP header or even another IP header (e.g., for embedding an encrypted datagram).

16

Networks and Services

priority), since UDP is used for real-time services while TCP is used for data transmission where there is no timing constraint. The IP packet is embedded in the information field of the link protocol, as shown in Figure 2.2(b). The layer above the network protocol (IP) is called the transport layer. The header of the transport layer protocol contains information (with the possible exception of the port address—see the footnote) that is only read by the terminals and ignored by the routers. The most common protocols are TCP supporting a connection-oriented data transmission service between the two terminals and UDP supporting connectionless transfer of real-time information (voice and video). TCP and UDP contain a parameter called port number. The port number identifies the type of information (application protocol) included in the information field (e.g., port 80 for http and port 23 for telnet). Even if we know the port number, we may still not know the actual applications the port supports. Port 80 is used for diverse services such as Web search, Web management, XML Web services, and even some types of Internet telephony. Therefore, the actual service supported cannot be identified from information contained in the TCP/UDP header alone. The application protocol is designed for managing a particular application, such as Web search. In this particular case, http is used as application protocol. Application protocols exist for all types of services and applications offered on the Internet (e-mail, file transfer, remote procedure call, and so on). Note that the application protocol is not part of the application but assists the application in transferring the application content (information or commands) across the network. The application software (or the middleware if it exists) contains instructions that request the terminal to initiate the protocol stack whenever a remote transfer is required. In this context, a terminal can be any type of equipment containing a CPU: PC, mainframe computer, server, database, mobile phone, sensor, actuator, printer, smart card, and so on. While there are almost 2 billion (2 × 109) PCs in the world today, it is estimated that there are more than 1 trillion (1012) devices (mainly sensors and actuators) satisfying the stated definition of a terminal. Most of these devices are either directly or indirectly connected to the Internet. 2.3.3 The Line of Demarcation Between Network and Application in the Internet

The Internet can be divided into two parts by a demarcation line, as shown in Figure 2.3. Below the line we have the network consisting of routers, the access, and the IP card in the terminal. Above the line we have the applications or the software (including the application protocol) running in the terminals. The demarcation line is in fact the transport protocol (TCP or UDP). On the two sides of the demarcation line, the business and the user charging models are completely different. The telecom operators and ISP reside below the line. They own routers, cables, and support systems for running the network. This is the traditional telecommunications business. In principle, the operator/ISP may use traditional charging, such as charging the customer for being connected (access charge); basing the charge on volume indicators such as the number of bits or IP packets sent or received; basing the charge on the actual bandwidth used; charging the customer for the duration of the connection; charging various content (voice,

2.3 Stupid and Intelligent Networks

17 Applications

TCP/UDP Demarcation line

IP

IP Access

Access IP

IP R

R

IP

IP

R

IP

R

Network

Figure 2.3

Separation of Internet functionality.

picture, video, data files) differently; or letting the usage of the network be free because the operator/ISP earns money from other sources such as advertising. However, as we shall see, the demarcation line makes all this (except levying access charges) difficult or perhaps impossible. Above the line, we have providers of all types of services and information that only need a medium offering sufficient bandwidth over which information can be transferred: films, music, Internet telephony, Web search, Web conferences, electronic newspapers, e-banking, e-mail, e-commerce, remote sensing and control, and so on. The transfer medium just happens to be the IP network. Any network that offers bandwidth adaptable to the data rate of the source could have done the job—this was just why ATM was developed. However, IP did it cheaper, so therefore the effort to replace the transport network with ATM failed. The problem for the service and information provider is getting paid for the service, the application, or the information. This problem has turned out to be rather difficult, since the range of services and applications is large and the user’s willingness to pay for an application is rather unpredictable. For some of these services, charges may be hidden in other fees (e-banking fees, credit card fees) or included in the price of the goods (e-commerce); other services are financed by advertisements: selling customer databases and information on user behavior (Skype); in still other cases, the service is free of charge because it supports a complementary service on which the provider earns money (Google); finally, the customer is not willing to pay for some services (electronic newspaper). The model in Figure 2.3 can also be drawn as shown in Figure 2.4, where the Internet is split up into three independent networks: the IP network containing routers (R) and the IP interface in the terminal (circles); a dynamic network consisting of terminals or other devices—servers (S), personal computers (PC), databases (D), or any other computing device—that communicate over TCP/UDP; and an even more dynamic and complex network of interacting application software (A).

18

Networks and Services

A

A

A

A

A

A

A A

A

A Applications/services

A

S

PC

A

PC

PC D

Terminals or devices that communicate

Demarcation line: TCP/UDP R

R R

R R IP network Figure 2.4

S S

S

Computer software

Networks owned by R operators and ISPs One operator

The Internet is three networks and not just one.

Network operators and ISPs reside in the lower plane. The actors in the plane in the middle are everyone (persons or firms or organizations) owning a terminal. In the upper plane, we find all kinds of people, firms, and organizations selling or giving away any type of information, content, service, software, transaction, support, or anything else that can be coded as strings of zeroes and ones. 2.3.4

Network Neutrality

The demarcation line is the basis for the notion called network neutrality—or as cartoonist Peter Steiner has put it: “on the Internet, nobody knows you’re a dog.” The applications residing above the demarcation line are egalitarian in that everybody’s information packets are treated in the same way by the IP network. This allows business models where everyone may create content and distribute it without being treated differently depending upon the type of content being distributed and what the provider is—be it a broadcast company or an entrepreneur working out of the garage. However, this causes problems for the network providers because old business model based on volume and time charging may no longer be feasible. Network neutrality encompasses the following four freedoms (of course, subject to legal restrictions) for the users of the Internet: •

• •

Freedom to access content on the network (i.e., access to the information can only be regulated by the owner of the information and not by an ISP or another third party not operating on behalf of the owner of the information); Freedom to run applications of any kind alone or together with other users; Freedom to attach any hardware to the network (e.g., routers, servers, PCs) that satisfies the Internet specifications;

2.3 Stupid and Intelligent Networks •

19

Freedom to obtain information about all services and electronic goods available on the network.

Network neutrality is, of course, subject to political debate. The supporters of network neutrality claim that the principle is in favor of competitive market evolution, since many applications and content provides can operate on the same arena and thus increase the total national revenues generated by the network. Network neutrality also favors innovation, experimentation, and provision of services that are too small and too specialized to be considered seriously by the large ASPs. The opponents claim that network neutrality is bad for the network and the national economy, since the revenues from network operation will become too small to support the future evolution of the network. Therefore, it is claimed that the price of the access should depend on the quality of service (QoS) offered by the ISP. Such QoS parameters may include bandwidth, secure delivery of data, real-time operation, privacy and integrity of data, and priority. This includes both fixed access charges and variable charges depending on volume. The supporters of network neutrality claim that this use of QoS will make the network no longer neutral but favor those who will pay more for the access and thus introduce an unfair competition arena. The opponents claim that the ISP should be entitled to recover their investments by charging for the actual use of network resources. It is likely that this debate will continue for a long time. There are strong commercial interests among both the supporters and the opponents of network neutrality. 2.3.5

The Commercial Life Below the Demarcation Line

The telephone network is connection oriented. This means that the connection between the communicating parties is set up at the beginning of the call and released at the end of the call. This allows the network operator to count the number of calls made by the user and the duration of each call, and base the charging on these measurements. If different bandwidth and processing (e.g., premium rate charging, toll-free charging, and shared charges) are associated with the call, this may also be taken into account when computing the charge. In this charging model, it does not matter which of the parties are sending the largest or smallest amount of information (or whether or not they are exchanging any information at all): the charges are usually levied against the user initiating the call (where toll-free and shared charges are two exceptions). On the contrary, this charging principle is not possible with the Internet because the IP network is connectionless, so that all packets must be treated as singular events. There is no way in which the network operator (or ISP) can correlate IP messages in order to determine the start and the end of a transaction, identify which party initiated the transaction, or measure the amount of information exchanged between the communicating parties during the transaction. The only simple charging method is based on subscription. The number of subscriptions reaches maturity when everybody has one or a few subscriptions, each satisfying their overall need for telecommunications. The revenue of the network operator is then independent

20

Networks and Services

of the usage of the network and of the investments required to maintain the traffic demand. Figure 2.5 shows what we may achieve at the network layer below the demarcation line. The network offers essentially four services: • •

• •

Routing and transport of packets of varying length between terminals; Multicast where information may be sent to several terminals on a single address; Bandwidth; Tunneling in order to support services such as mobile IP and security.

Since the IP network is connectionless and only offers a few basic services (it is a stupid network), the Internet creates new and difficult business scenarios for the network operators: •

•

•

Volume charging based on the number of bits exchanged is a viable method if the operator can set the price in accordance with the application. If the price is too high, it will be too expensive to receive films and music; if the price is too low, there is nothing to earn on voice communication. Volume charging may, however, be possible in association with QoS parameters associated with each packet. Static charges for bandwidth is simple in the same way as general subscription charges and is therefore included in the subscription charge. Dynamic charging based on the actual bandwidth used at any time is just another way of volume charging and is just as complicated to measure. Time charging is not possible in the IP network because there is no way to establish the duration of the call (e.g., a voice call, if no other protocol controlled by the network owner is applied). This is because each IP packet is an TCP/UDP Demarcation line IP

IP

R

IP R

Payment models: • Subscription

Network applications:

• Volume (number of bits/packets)

• Routing of packets and transport of bits

• Bandwidth

• Multicast

• Time

• Bandwidth

• Number of access attempts

• Tunneling

• Quality of service • Contents

Figure 2.5

Below the demarcation line.

2.3 Stupid and Intelligent Networks

•

•

•

21

independent message, and it is not possible from the information contained in the header of the IP packet alone to establish the number of packets that are exchanged in a given transaction and thus determine the duration of the transaction. For the same reason, the number of access attempts in IP also equals the number of packets sent, so this is again similar to volume charging. However, in most cases, the majority of the traffic is sent to the user (e.g., Web search, downloading of video or music) so that a true measure of the traffic can only be derived if the number of packets in both directions is counted. Among the QoS parameters that can be subject to differential charging are real-time delivery, guaranteed delivery, priority, data integrity and privacy protection, and upper bounds on the one-way and two-way delay. The user may then either subscribe for a given level of QoS, in which case the charging is part of a fixed access charge, or be associated with individual packets. The latter is a particular instance of volume charging that is easy to implement. Charging based on content means that the ISP must know what type of information is sent. This cannot be found out by simply looking at the IP packets. The ISP must either require that a particular procedure is used via servers owned by the ISP [the standards H.323 and Session Initiation Protocol (SIP) for IP telephony], read the port address of TCP/UDP and from this information estimate the type of service, or get the information from the provider of the information (another user or an ASP). In practice neither of these methods will work because they may be circumvented or even be forbidden because they violate network neutrality. Moreover, it is difficult to put services into simple categories, since there will always be borderline cases where it may be argued that the service may belong to a different charging class. A multimedia service is one such example: sometime the service is used as a pure telephone service, and sometimes the service is a complex mixture of voice, stills, data files, and moving pictures. These service alternatives use different network resources and should perhaps be charged differently.

In addition, most of the information (film, music, Web files) on the Internet is sent from a database to the user on demand. The traffic pattern is then highly asymmetric where most of the traffic load is made up of downlink traffic to the user. The user is generating only a few small data packets. In order to get paid for all the traffic, the network provider may either levy the charge against the user (received traffic charge) or against the information provider (sent traffic charge). If the network owner charges the information provider, the information provider must charge the user in order cover its own telecommunications expenses. This may be impossible for two reasons. First, it will likely be unsatisfactory for the information provider, user, and perhaps also politicians and market regulators, since the cost for retrieving information is determined by the network operator and not by the owner of the information. The actual cost of the information then gives a wrong picture of the market value of the information. Second, many information services are already delivered free of charge (e.g., information on the Web). It is probably not possible to go from low charges to high charges in order to compensate for the use of the

22

Networks and Services

network. Such attempts are likely to be stopped by the regulators, politicians, and interest groups. The ISP may also offer QoS by which different charges may be levied depending upon the value of the QoS parameters. Because of network neutrality and regulations, such attempts may also be unsuccessful. This leaves the network providers with a problem that is hard to solve. The solution is probably wrong or will at least face difficulties any way the ISPs choose to solve it. 2.3.6 Is There Any Business for the Network Operator Above the Demarcation Line?

The answer is no unless the network operator enters into businesses that are not related to network operation. Among the service providers we have banks, newspapers, search engines, municipalities, governments, publishers, and so on. Some of the service providers are big companies or organizations (e.g., governments), while other providers are small firms employing one or two people. The size distribution of service providers is likely to be a Pareto-like distribution, where there are a few very big providers and a very large number of small providers. This makes the marketplace for services extremely complex, dynamic, and unpredictable. Classic economic theory studies homogeneous markets where all providers have the same opportunities. These theories will most likely not apply to the Internet. All services are software applications. Some of the providers use licensed software, while other providers apply software from free sources (open-source software). Therefore, the cost models of service provision are complex and dynamic. Software products, including stored information, share some basic properties: •

•

•

•

The cost of copying software is almost zero. Therefore, the cost of one sample of software depends only on the cost of developing the software divided by the number of copies in which it is distributed. The marginal cost of software goods is thus zero. Much software can be downloaded from the network without charge (free software resources) so that even the cost of developing new software may be very small. The software may be designed by persons not getting paid for developing the software but for other work they do, for example, being an employee or student of a university or a researcher developing software that is a spinoff from the research. This is called peer production. The Internet is itself the result of peer production, where the network was developed by universities, government agencies, and independent research establishments. Much of the software is easy to develop, so that the development of many software products requires the efforts of just a few people. This is also the case for large software products. These products are usually a combination of many small autonomous pieces. The investments required for software development are small—often just a single PC—favoring the growth of many small companies using just a few software products.

2.4 Overlay Access

23

From this discussion it is obvious that the business above the demarcation line is not simple telecommunications but a complex mixture of actors of different kinds and sizes. The traditional telecom operator and the ISP have no particular advantage in this business. This is unlike the traditional telephone service, where the network operator offered not only routing and interconnectivity but also all the services the user needed. The ongoing introduction of IP everywhere—also in mobile systems—is the event that is likely to transform the telecommunications industry. Predicting which type of businesses is emerging out of this transformation is difficult.

2.4

Overlay Access Overlay access means that one access network is using the capabilities of another network, as shown in Figure 2.6. In the example, the Internet (the embedded IP network) is making use of the ISDN/telephone network (the embedding network) to connect to the Internet platform. The embedding network may even be a switched network, where the Internet access passes several exchanges. The telephone network may be fixed or mobile. Another example of overlay access is mobile IP. The tunneling mechanism consists of embedding the original IP datagram in another datagram that can be routed to the new destination. One IP session is thus embedded in another IP session. Regulation in telecommunications has led to solutions as shown in Figure 2.7. The reason for such regulations is to advance competition without building new physical networks—in particular, access networks. It is simply not economically feasible for a competitor to build new physical networks in many cases. Operator 2 does not own a physical network but, by regulation, is allowed access over the physical network of operator 1. Viewed by the user, operator 2 offers full access capabilities though the physical transfer of bits is carried out by operator 1.

Internet access

Internet terminal

Figure 2.6

Telephone access

Internet platform

Transport

Overlay access.

Operator 1 Physical access 1

Platform 1

Transport 1

Access 2 Platform 2

Terminal

Operator 2

Figure 2.7

Virtual network.

24

Networks and Services

Such network access is often referred to as a virtual network, and operator 2 is called a virtual network operator (VNO). If the access is via a mobile network, operator 2 is usually referred to as a mobile virtual network operator (MVNO). The virtual operator owns a platform (platform 2 in the figure) supporting functions such as service provisions (e.g., Internet telephony) and subscription management including allocation of numbers, access passwords, and usage charging. In a mobile 3G network, the MVNO may issue subscriber identity modules (SIMs) and own a home subscription server (HSS), as well as a gateway supporting mobile IP (GGSN) (see Section 8.4.6).

2.5

Domains and Interworking It is feasible to define two types of domains in telecommunications systems: technology domains and administrative domains. As the name suggests, a technology domain consists of a single technology (e.g., GSM, IP, or ISDN) and may comprise the networks of several operators. The global GSM system is a single technology domain. An administrative domain has to do with the administrative management of the network. Two operators always belong to different administrative domains. One operator may also consist of several administrative domains if the operator owns networks that are managed independently of one another (e.g., telephone networks, mobile networks, broadband networks, and the Internet). The technology in the two interconnected administrative domains may be different (for example, one operator owning a fixed network and another operator owning a GSM network) or the same (for example, two GSM networks owned by different operators). The existence of the two types of domain leads to two fundamentally different ways that two systems can be interconnected, as shown in Figure 2.8. The interconnection between domains is taken care of by interceptors. Between two technologies, an inline interceptor is inserted (often referred to as an interworking unit). This is simply a device that converts the formats of one system to the formats of another system. Sometimes this conversion is simple, depending only on encoding formats (syntax); sometimes the conversion is very complex, involving the translation of entire procedures (semantics). Examples are as follows: •

Interconnection of the telephone service of GSM and the fixed digital telephone network (using the PCM encoding technology) requires that the voice encoding formats must be translated (syntactic translation). This interworking is not transparent in the sense that certain services cannot be offered across the interface (for example, end-to-end encryption) (semantic restrictions). Technology domain

In-line interceptor

Technology domain

Split interceptor Administrative domain

Figure 2.8

Interconnection of domains.

Administrative domain

2.6 Heterogeneity •

•

•

25

Interconnection of VoIP terminals and ordinary PCM telephones requires that there exist transit points between the two networks where the coding formats are converted. In addition, the interworking unit must also map between the operation of connectionless and connection-oriented network operation. This requires complex management of signaling procedures (syntactic and semantic conversion). Interconnection of two PCM systems where one system uses A-law encoding while the other uses µ-law encoding (syntactic translation). Support of data communication on GSM or GPRS requires that snoopers are inserted between the fixed and the mobile network in order to ensure that the flow control mechanism of TCP operates properly. The reason for this arrangement is that the bit error performance of a GSM/GPRS link usually is several orders of magnitude poorer than that of the fixed network (semantic conversion).

Interconnection devices between administrative domains are called split interceptors. These devices, or rather functions, are required even if the technology is identical in the two domains. The split interceptors take care of processes such as security, usage measurement and accounting, traffic management, and error supervision and management. Trusted third parties (TTP) offer security functions such as binding the user identity to the public key owned by the user (provision of certificates), producing and storing evidence in nonrepudiation services, and producing session keys. TTPs are split interceptors, where the interceptor is not owned by any of the communicating parties but, as the name indicates, by an independent third party. While the inline interceptors are realized as physical hardware devices or as distinct software processes, the split interceptor is much more complex. Sometimes the split interceptor is equivalent to a hardware device (e.g., a TTP managing a PKI). Sometimes the split interceptor consists of administrative agreements and is realized as components of a complex distributed process, such as roaming between different GSM networks.

2.6

Heterogeneity My favorite telecommunications example of a heterogeneous network is shown in Figure 2.9. The two users are interconnected by three systems in tandem: •

•

•

−3

A mobile access link with bandwidth 16 Kbps, bit error rate (BER) 10 , and propagation delay 1 ms between the terminal and the network access point; −14 An optical network with bandwidth 140 Mbps, BER 10 , and propagation delay 30 ms, corresponding to an optical communication link of about 5,000 km, between the network access points; −6 A geostationary satellite link with bandwidth 1 Mbps, BER 10 , and propagation delay 250 ms between the network access point and the terminal.

26

Networks and Services

User

Land mobile Bandwidth = 16 Kbps BER = 10–3 Delay = 1 ms

Figure 2.9

User

ATM Bandwidth = 140 Mbps BER = 10–14 Delay = 30 ms

Satellite Bandwidth = 1 Mbps BER = 10–6 Delay = 250 ms

Heterogeneous system.

Such communication links exist. The link may represent a connection between a laptop with GSM access and a laptop at a remote site that is reached via a very small aperture terminal (VSAT) system. VSAT systems consist of small satellite Earth stations (antenna diameters between 1m and 2m with the communication electronics contained in small units; see Chapter 9) communicating via geostationary satellites. The point here is that the land mobile link determines the data rate that can be transferred between the terminals. The delay over the satellite determines the speed and efficiency of the end-to-end application protocol. The time that a message is sent from the mobile terminal and until an acknowledgment is received from the VSAT terminal is about 0.5 seconds. A protocol based on a “stop-and-wait” strategy, where an acknowledgment must be received before the next message is sent, will not work in satellite systems: if each message consists of 5,000 bits, the achievable net bite rate across the satellite link is only 10 Kbps (two messages per second). The high BER of the land mobile system shows that precautions must be taken at this interface in order to achieve proper operation of the TCP connection (for example, snooping). If the BER is 10−3, even a minimum length IP datagram consisting of 536 octets (or, 4,488 bits) cannot be transferred in a single frame over the radio connection. The probability that such a message is destroyed by bit errors is almost 1. In GSM, a minimum-length datagram is in fact sliced up into about 25 pieces before it is transferred across the radio link. If the BER is 10−3, the probability that an arbitrary piece is lost is then 0.15, which is still a large number. Using automatic repeat request (ARQ) for error correction on the radio link, a frame is sent until it has been acknowledged. This reduces the frame loss probability to an acceptable level, but there is still a residual probability that the datagram contains errors so that additional mechanisms (snooping) are required in order to ensure proper operation of the Internet connection. This example is not unique. The VSAT link may be replaced by the INMARSAT system providing communication with ships and aircraft. Most systems are heterogeneous, and they are so for several reasons. The network technologies are evolving over time. So are the capabilities of the components used in terminals and network devices (according to Moore’s law, the processing capacity of computers doubles once per year). This means that one important source of heterogeneity is equipment age: it is very expensive to upgrade networks with new technology. It is cheaper to cope with heterogeneity.

2.7 Real-Time and Nonreal-Time Systems

2.7

27

Real-Time and Nonreal-Time Systems Another important characteristic of signals is the timing constraint. Analog speech and moving picture signals are sampled at a constant rate, and each sample is converted into a word of, say, 8 bits. The encoded signal is transferred across the network and finally decoded by the receiver in order to reproduce the original signal. The reproduced signal will have exactly the same waveform as the original signal if all samples are delayed the same amount. If the delay fluctuates over time, the resulting waveform will be distorted, and if the fluctuation is larger than a certain amount, the speech signal is no longer intelligible. Speech, moving pictures, and all other signals that cannot tolerate jitter in the bit rate are called real-time signals. The network used for transferring such signals must thus satisfy the condition that the jitter of received bits or words is within narrow limits. The telephone network and the broadcast networks are designed in such a way that this condition is satisfied. There is no timing constraint on the transfer of data. The only conditions are that data is received without errors, that data is not lost or duplicated, and that data is received in the correct order. The Internet (or more precisely the TCP) is designed such that these conditions are met. The Internet (and derivatives such as broadband networks and packet radio networks) does not guarantee that the differential delay can be maintained within the strict limits required for speech and moving picture signals. VoIP is a real-time service over a nonreal-time network. The quality of the service is sufficient so long as the traffic on the Internet is so small that congestion and call queuing is insignificant. However, the future market for real-time services over broadband networks is assumed to be big (speech, video, radio broadcast, music, and real-time games) so that the Internet in its current form may not be able to offer real-time services because of excessive delays caused by the traffic. An application protocol embedded in the UDP datagram called real-time transport protocol (RTP) handles the problem of duplication of packets, jitter, and out-of-order delivery. For this purpose, each RTP message contains a sequence number and a real-time clock value allowing the receiver to restore the signal. IP version 6 will improve the capability of the Internet to support real-time operation. Real-time operation can be achieved by using a playback buffer as illustrated in Figure 2.10. The voice or video packets are sent at constant rate by user A. Because of differences in transmission delay along the path (for example, caused by buffering delays), the packets arrive at user B with slightly varying rates. If this jitter is not too big and is on average zero over many frames in succession, the jitter can be removed by storing the packets in a playback buffer and then reading them out at the same rate as that of user A. The playback register introduces a delay that comes in addition to the delay introduced by the network. The total one-way delay should be smaller than about 150 ms in order not to reduce the quality of the voice signal too much.

28

Networks and Services A

B Network

Direction of transmission Well-ordered sequence of packets

Packets received with jitter

Well-ordered sequence of packets

Buffer Jitter removed by playback buffer in the receiver

Figure 2.10

2.8

Jitter prevention by playback buffer.

Backward Compatibility Backward compatibility means that systems we install now must be compatible with existing systems in a certain way. Sometimes backward compatibility may be ignored. This is the case if the new system offers services that do not exist in the old system. In this case, there is no backward compatibility problem at all. When the Internet went public in the 1990s, none of the services offered by the Internet were offered by the telephone network. The Internet could then be introduced independent of which services the telephone network offered. The Internet is a data network with characteristics entirely different from the data networks offered by telecommunications operators in the 1990s. However, if the Internet had been introduced by the large telecommunications operators as an extension of the existing X.25 packet data networks, backward compatibility may have been implemented, possibly rendering the Internet useless to the general public. The Internet is an example where backward compatibility is not a good solution. It was undoubtedly an advantage for telecommunications users and society as a whole that the operators had to abandon their X.25 packet networks and implement the Internet instead. In other cases, users connected to one system must be able to exchange information with users connected to other systems. Examples of two services requiring this type of compatibility are VoIP and e-mail. VoIP can survive as a service that in the long run replaces ordinary telephony only if it initially is compatible with the existing telephone service. Different e-mail systems must share common nodes in which format conversion takes place; otherwise, one of the technologies will be phased out. The battle between Microsoft and providers of free software is a commercial battle. The strategy of Microsoft is to avoid compatibility, while the strategy of other providers is to force Microsoft to become compatible.

2.8 Backward Compatibility

29

Compatibility may be achieved by interworking between network elements (VoIP versus ordinary telephone services), by implementing dual-type user equipment (2G versus 3G), or by offering both systems in parallel over a period of time (IPv4 versus IPv6). In almost all cases, backward compatibility is expensive, delays the implementation of new technologies, and constrains the scope of new services. There are commercial, technical, and political reasons why we must base our decisions on backward compatibility even if this reduces the benefits of the new technology. 2.8.1

Commercial Reasons

The commercial value of networks such as the telephone network and the e-mail network is that everyone can make telephone calls to and exchange e-mails with all other customers connected to the network. Therefore, systems offering new types of telephone services and e-mail services will not penetrate the market unless they can exchange traffic with the existing network. It is reasonable to assume that the cost of being compatible is carried by the operator introducing the new service. This cost may render the new service commercially unprofitable. Sometimes the regulatory authorities and the government may decide otherwise so that the costs of being compatible are carried by the incumbents. Several intelligent services (e.g., number portability) in the telephone network were introduced in this way. Compatibility may also be driven by the manufacturers of terminal equipment. The most obvious example is mobile communications. It is a commercial advantage to offer 2G, 3G, and WLAN in the same equipment, permitting the user or the system to choose whichever access is the most convenient. This is no longer an advantage for the manufacturers if all manufactures do the same, but it is undoubtedly a disadvantage for the manufactures not doing it. The existence of terminals with multiple capabilities makes it easier for the operators to introduce the new system and dismantle the old one. The operators may spread the investments in 3G systems over long periods of time by only replacing 2G systems in cities. Phasing out a service also raises other commercial concerns. Equipment in the network may have a commercial lifetime of more than 10 years. Therefore, it is not commercially clever to replace equipment with long residual lifetime if it can be avoided by simple compatibility means. Users also invest in equipment. Some terminals may have a very long economical lifetime, such as ship Earth stations in the Inmarsat system. Other terminals have a very short lifetime, such as mobile phones. This has forced Inmarsat to still offer very narrowband services for certain applications (small vessels, air traffic control, land mobile) while the major part of the users have a demand for wideband services. The first generation Inmarsat system (Inmarsat-A) was scheduled to be taken out of service in 2007 after 30 years of operation. For the last 15 years the system has been living together with more modern digital systems.

30

Networks and Services

2.8.2

Technological Reasons

At present it is not feasible to provide broadband services at bit rates beyond 500 Kbps to small ships and aircraft. Broadband services cannot be provided in all mobile networks. Radio propagation effects constrain the bandwidth that can be offered: the larger the cell, the lower the bit rate; the faster the mobile terminal is moving, the lower the bit rate. For this reason, 3G systems offer different data rates in cells of different size (indoor, urban, and suburban) and for different usage of the mobile network (in vehicles, for pedestrians, in fixed locations). Since several of the technological constraints are physical and cannot be circumvented, this situation will persist in the future: though the usable bandwidth of mobile radio systems has been increasing steadily, the increase of bandwidth of fixed systems has been faster. The compatibility challenge is thus to provide interworking between systems suffering different technical constraints. The heterogeneous system of Section 2.5 is an example. 2.8.3

Political Reasons

Telecommunications is a social good that shall be available to everyone at affordable prices. This is a political prerequisite that is incorporated in the licensing policy for at least some of the telecommunications operators. The technical term for this provision is universal service obligation (USO). Initially, USO only applied to fixed telephony and noncommercial radio and television broadcast systems. Later, mobile services, Internet services, and, finally, broadband services have become subject to USO. USO indirectly implies a large degree of compatibility between networks and services, since the operator cannot switch technology without possibly compensating for the inconvenience this may cause the users. The technological evolution progresses more slowly in some parts of the world than in other parts. One important political requirement is to ensure that telecommunications services are compatible across such technology barriers.

2.9

Standards Standards are defined for all elements making up a telecommunications system. Common standardization bodies are as follows: •

•

•

The International Telecommunication Union (ITU) develops international telecommunications standards. Some of these standards are developed in cooperation with the International Organization for Standardization (ISO). The ISO develops standards for computers, programming languages, data transmission, and computer networking. The European Telecommunications Standards Institute (ETSI) develops standards for Europe. Many of these standards are adaptations of ITU or ISO standards, where the scope of the original standard may be narrowed or made more precise by removing options and choosing one particular set of

2.9 Standards

•

•

31

parameters. Some standards such as GSM are developed by ETSI alone and later adopted for wider use. The Institute of Electrical and Electronics Engineers (IEEE) has developed a large number of standards for worldwide use. The standards for local area networks (IEEE 802 series) are particularly well known. Internet Engineering Task Force (IETF) standardizes protocols and related aspects of the Internet.

In addition, there are a number of interest groups developing standards for particular areas of telecommunications and computer science: Third Generation Partnership Project (3GPP) develops the 3G standards further, Object Modeling Group (OMG) standardizes distributed computer platforms, Bluetooth Interest Group is responsible for advancing the Bluetooth technology, and several groups develop speech and picture coding standards for various applications. Telecommunication is also regulated by national standards covering technical requirements, licensing, service provision obligations, and business relationships (e.g., allowing virtual network operators, supporting number portability, and settling interconnect pricing). The purpose of the standard is to ensure that remote terminals can communicate with one another. This implies that rules must exist for a number of cases: •

•

•

•

•

The interconnection of networks of different types, manufactures, age, and ownership are specified in a number of telecommunications standards. Operational standards related to reliability of communication equipment, fault monitoring and restoration, QoS requirements, traffic management and dimensioning, remuneration between operators, interconnection obligations, and numbering plans are contained in international telecommunications standards and supplemented by national regulations. The access between the terminal and the network is specified in several telecommunications standards and data transmission standards so that terminals of different manufactures can access networks owned by different operators. Some of these standards documents are very big. The GSM standard and the UMTS standard contain 5,000 and 10,000 pages, respectively. End-to-end transport and application protocols, including standards for middleware, are the topic of data transmission standards issued by ISO, IETF, IEEE, and special interest groups. These standards contain the details of protocols and platforms such as common object request broker architecture (CORBA), TCP, http, remote procedure call (RPC), XML Web Services, e-mail protocols, and simple file transfer (SFT). Other standards include standards for keyboards (QWERTY and the keyboard of mobile phones), user profiles and menus, operation of the terminal, and the connectors on external computer interfaces (e.g., between computer and printer, to external networks, and to the main bus of the computer).

The telecommunications industry is, in fact, the only industry that cannot exist without international standardization of almost all aspects of the industry. The provision of global telecommunications services and the existence of global markets for telecommunications equipment and computers are impossible without these

32

Networks and Services

standards, and this is just the reason that some of the biggest standards organizations in the world are dedicated solely to the task of producing and maintaining these standards. It is important to understand this process in order to understand many of the business processes taking place in telecommunications. The most common development process consists of a set of filters as follows: •

•

•

Ideas for new services, protocols, procedures, and networks are developed by individuals or organizations and presented in journals or at international conferences. The recognition of the idea is thus the first filter through which the idea must pass. Some of these ideas catches the interest of a wider audience and are developed further in joint research projects or in standards organizations. The idea may eventually turn out to be the basis for a new product or simply disappear because it led to nowhere, was not believed (sometimes incorrectly) to give rise to a new product, or thought to be too expensive or complex (also sometimes incorrectly). The development phase represents the second filter. Sometimes a potential new product reaches the market. The market may then adopt the product, reject it, or totally ignore it. Sometimes the operators may force the product on the customers (the television standard), and other times the product is realized by awaiting the response from the market (GSM, videophone). GSM became a commercial success while the videophone failed. The market is the last filter the idea must pass through in order to become a success.

There are, of course, exceptions to this process. Some services can be developed and implemented by a single person or a small group of persons and even not publicized before being marketed. Skype is an example where the service consists of software the user may download free of charge from the source, install it on a computer, and start using it as prescribed. The only additional action the provider of the service must take is to make sure that calls generated in the “Skype” network are passed to the telephone network (or another VoIP network) at some interworking point if the called subscriber is connected to the telephone network and vice versa. The World Wide Web is another example of an idea that did not follow the general evolution path. The World Wide Web is already a tremendously big service. Skype may grow to similar heights—nobody can—at the time of this writing—predict what may happen. The most difficult problem in fact is to pass the market filter. Even a service that eventually becomes big may go through a long latency period before it really starts growing because of positive feedback from the market (network externalities). This point is not pursued further here. The time from idea to product may take several years: GSM took 10 years, UMTS took 15 years, maritime satellite services took 7 years, and data transmission took 10 years—just to mention a few examples. The Internet was commercialized after almost 20 years of operation well hidden in the obscurity of research networks. Hence, most of the technology the network contains is already old when it is implemented. Moreover, the technology must have a commercial lifetime of 10 to

2.10 Access to the Common: Regulation of the Utilization of the Frequency Spectrum

33

20 years after implementation in order to ensure full payback. Therefore, the time constants we have to deal with in telecommunications may span two or three decades. In a few cases, radical changes take place making existing systems obsolete long before they have served us long enough to provide enough payback for the investments. One such radical (or disruptive) change took place in the mid-1990s when the Internet made the data networks owned by the telecom operators obsolete in less than a year. Another radical change takes place just now where all-IP applications such as VoIP and video over IP make not just the telephone technology obsolete but also the old business models by changing the way in which charges for the use of the network are levied (see Section 2.3). These radical changes have not come as a result of standardization efforts and consensus among the large and dominating telecom operators and manufacturers but from unexpected sources such as research (CERN) and clever individuals (Skype).

2.10 Access to the Common: Regulation of the Utilization of the Frequency Spectrum The frequency spectrum is the most important shared resource in telecommunications. The spectrum must be shared between a large number of services and between different commercial and public interests. The main constraint is that the same part of the spectrum cannot be used for more than one service at the same place at the same time. The usable radio frequency spectrum extends between 9 kHz and 275 GHz. The lowest frequency corresponds to a wavelength of 33 km, while the bandwidth of the highest frequency is only a trifle more than 1 mm. The majority of applications are found at frequencies between 30 MHz and 30 GHz. Note that the bandwidth of the frequency band from 3 GHz to 30 GHz (the SHF band) is about ten times larger than the frequency band below 3 GHz. This means that the SHF band supports 10 times more information than the lower bands. The EHF band from 30 GHz to 300 GHz is again about 10 times larger than the SHF band and can therefore support 10 times more information, but the band is used to a lesser extent than the SHF band for two reasons. First, it is still difficult to design radio transmitters and receivers at EHF frequencies because of the small bandwidth; second, the propagation loss through the atmosphere is extensive at frequencies above 30 GHz, allowing only short-range applications or applications outside the atmosphere where the high atmospheric loss is an advantage (e.g., satellite-to-satellite communication and space research). The frequency sharing must be strictly organized such that different applications do not interfere with one another. This task is coordinated by the ITU, and the frequency allocation plan is published in the ITU Radio Regulations. The allocation plan is discussed and possibly amended at the meetings of the World Administrative Radio Conference (WARC) under the ITU patronage. National or regional regulation authorities are responsible for local applications of the frequency plan. There are three frequency allocation regions: Region 1 consists of Europe (including Siberia) and Africa; Region 2 consists of the Americas; and Region 3

34

Networks and Services

consists of Asia (except Siberia), Australia, and Oceania. Some frequency bands are allocated for all three regions in order to support global services such as land, aeronautical, and maritime mobile services; intercontinental satellite communication services; broadcast satellite services; terrestrial broadcast services; mobile radar services; navigation systems; dissemination of standard frequencies and time signals; amateur radio; and radio astronomy. Other frequency bands are allocated for different purposes in different regions. In addition, there is a large number of footnotes to the frequency plan allowing particular applications in certain countries on a permanent or temporary basis. Table 2.1 shows an overview of the frequency spectrum. The frequency bands, the name of the band and the range of wavelengths are shown in the three first columns. The relationship between frequency f and wavelength is f = c/ , where c is the speed of light in a vacuum (300,000 km/s). The last column indicates the most important applications of the band. The very low frequency (VLF), low frequency (LF), medium frequency (MF), and high frequency (HF) bands are used for global low-frequency navigation services, AM broadcasting, dissemination of time signals (UTS standard clocks), and for communication with embassies. The upper half of the very high frequency (VHF) band and the lower third of the ultra high frequency (UHF) band are used for radio and television broadcasting. The bands also contain subbands allocated for land mobile, maritime mobile, and aeronautical communication. The upper two thirds of the UHF band are allocated to terrestrial and satellite mobile services (e.g., 2G, 3G, and Inmarsat), including the unlicensed frequency bands for wireless LAN and WiMax. This part of the spectrum contains a large number of other services, such as radio relays, fixed satellite systems, radio navigation, radar, and radio astronomy. The super-high frequency (SFH) band contains bands for radio relays, fixed satellite services, navigation services, radar, and satellite broadcasting. The band also contains subbands allocated to land mobile services still to be exploited. Table 2.1 Radio Frequency Spectrum and Allocations to Services Frequency Band Name Wavelength Main Applications <3 kHz >100 km Not allocated but important in thunderstorm research (whistlers), Earth science, and cosmology to study low-frequency phenomena 3–30 kHz VLF 100–10 km Communication with submarines and beacons; no allocation below 9 kHz 30–300 kHz LF 10–1 km Navigation, time signals, AM broadcast 300 kHz–3 MHz MF 1 km–100m AM broadcast, government (embassies) 3–30 MHz HF 100–10m Broadcast, government (embassies), amateurs 30–300 MHz VHF 10–1m Radio and television broadcast, mobile 300 MHz–3 GHz UHF 1m–10 cm TV, land mobile, WLAN, mobile satellites, navigation satellites 3–30 GHz SHF 10–1 cm Radars, fixed and mobile satellites, radio relays, land mobile, satellite broadcast 30–300 GHz EHF 1 cm–1 mm Radio astronomy, radio relay, satellites, intersatellite; no allocation above 275 GHz >300 GHz <1 mm Not allocated; atmosphere opaque to electromagnetic radiation up to the so called infrared window

2.10 Access to the Common: Regulation of the Utilization of the Frequency Spectrum

35

The extremely high frequency (EHF) band is still sparsely exploited but contains bands for satellite communications, radio relays, short-distance mobile communication (e.g., pico-cells in buildings), navigation, and satellite-to-satellite communication. The EHF technology is still expensive but will become cheaper as these frequencies are taken into use. The band from 1 GHz to 300 GHz is also called the microwave band. Radio astronomy is a particularly important case because the location of the bands depends only on the frequency of the cosmic radio source. Therefore, the allocation of frequency bands for radio astronomy consists of numerous narrow bands allocated on a global basis for this purpose. Examples of bands are: 73–74.6 MHz, 1.400–1.427 GHz, 2.69–2.70 GHz, 42.5–43.5 GHz, and 200–231.5 GHz. The radio frequency band also contains several subbands for scientific exploitation of satellites such as space research, satellite astronomy, meteorology, Earth science, and Earth resource surveillance.

CHAPTER 3

Synchronization 3.1

Definitions 3.1.1

Synchronous

This term applies to a comparison between two or more signals. The ITU defines this term as follows: The essential characteristic of time-scales or signals such that their corresponding significant instants occur at precisely the same average rate. Synchronous signals are locked to the rate of a common clock in order to assure this property. Signals locked to different clocks are not synchronous. The definition means, for instance, that the average bit rates of two synchronous digital signals are exactly the same if the signals are compared bit by bit during the same period of time. Even if the timing of the common clock is subject to a systematic drift, so that the frequency of the clock in one year from now has increased by, say, 0.01%, the signals locked to that clock are still synchronous, though their common average rate has increased. The short-term variation in the bit rates of synchronous signals may be significant. Let us see why. A signal may be split in two and sent along two different propagation paths and then compared. If the propagation velocities along the paths are constant, the two signals will still be in perfect synchronism when compared. Only their relative phase may be different. However, the propagation velocities and the propagation delays along the two paths are generally not constant but may vary independently of each other. Such variations are caused by short-term jitter of the clock (since the propagation delay of the two paths are different, the jitter of the two signals will be uncorrelated), by filters and regenerators along the path, by fluctuations in temperature and moisture, by mechanical vibrations causing variations in the electrical length of the path,1 by thermal and nonthermal noise generated along the path, and so on. When the signals are then compared, the signals are no longer in perfect bit-by-bit synchronism, but the bit rates will fluctuate independently around the same average. These short-term fluctuations may cause displacement of the two signals that may amount to several bit positions. However, the signals are nevertheless synchronous since the average rate is not changed. 3.1.2

Asynchronous

This term is the antonym of what we just described. Between two asynchronous signals, there is no relationship whatsoever between bit rates and other significant frequencies of the signals. 1.

Electrical length le equals the physical length l divided by the wavelength of the signal: le = l/ . The phase of the signal after having propagated a distance l is 2 le = 2 l/ . If we compare the phases of two signals with different wavelengths, the phase difference between the signals after propagating the same distance l is 2 l/ 1 − 2 l/ 2 = 2 (le1 − le2).

37

38

Synchronization

Doppler shift causes synchronous signals to become asynchronous. The Doppler shift is the change in the frequency of a signal from a sender that moves relative to the receiver. If the relative velocity of the sender and receiver is v then this causes a shift in the frequency equal to ∆f/f = ± v/c where c is the speed of light in vacuum (300,000 km/s), and the plus sign applies if the transmitter and the receiver are moving toward each other (increased frequency) and the minus sign applies if they are moving away from each other (reduced frequency). If the relative velocity is 900 −7 km/h (the speed of an aircraft), the Doppler shift is 8.3 × 10 . We see that a signal at bit rate of 1.2 Mbps drifts about one bit period every second; if the bit rate is 120 Mbps, the drift is 100 bits per second. Since the drift usually takes place over short periods of time, Doppler shift is generally not a problem. However, the effect must sometimes be taken into account in system design, in particular if the data rate is high. Two examples are as follows: 1. The transmit clocks in terminals are usually synchronized to the rate of the network clock. The rate of the clock in the terminal is then f ± fv/c, where f is the rate of the clock in the network. The plus sign applies if the sender and receiver move toward each other; the minus sign applies if they move away from each other. Signals sent by the terminal are also Doppler shifted so that the rate received by the network is f ± fv/c ± (f ± fv/c)v/c = f ± 2fv/c + f(v/c)2 ≈ f ± 2fv/c—that is, the rate of the received signal is inaccurate by twice the Doppler shift (ignoring the final term which is of second order in v/c). If these signals are sent further into a synchronous multiplex system, rate adaptation methods such as elastic store (see Section 3.4.4) or bit stuffing (see Sections 4.3.4 and 4.4.5) must be applied in order not to lose or duplicate information. 2. If a transmission section includes a satellite link, the signals in each direction will be subject to Doppler shifts, even if the satellite is in the geostationary orbit, for two reasons: first, satellites cannot be placed in perfect circular orbits, and, second, a perfect circular orbit is unstable so small perturbations cause the satellite to drift into a noncircular orbit (see Section 9.4.2). The Doppler shift may then determine the actual frequency stability of the network. If the satellite interconnects plesiochronous systems, it is this stability and not the stability of atomic clocks that determines which bit stuffing technique must be used at each end of the link. 3.1.3

Plesiochronous

This term also applies to the comparison of two or more signals. Plesiochronous (or almost synchronous)2 signals are asynchronous signals where the frequency difference between the signals is small. Plesiochronous signals are locked to independent high-precision clocks (e.g., atomic clocks). The ITU (and other international standards organizations) specifies the accuracy of these clocks both with regard to short-term fluctuations and long-term drift (about 10−11). The stability measure 2.

Plesio is Greek for close or near (plission in Modern Greek).

3.2 Reality Is Not So Simple: Bits, Words, Envelopes, and Frames −11

39 −11

10 means that the frequency error of the clock is less that ±10 × f for some specified probability (usually 95%), where f is the nominal (or theoretical or specified) frequency of the clock. 3.1.4

Isochronous

This term applies to a single signal. ITU defines the term as follows: The essential characteristic of a time-scale or signal such that the time intervals between significant instants either have the same duration or durations that are integral multiples of the shortest duration. All signals consisting of bits of the same duration are isochronous. This means that all digital signals we are considering in this book are isochronous at this level of organization. However, as we shall see, there are other organizing elements of the signal beside bits. The signal may not be isochronous with regard to these elements. Therefore, we must be careful when defining what is meant by the term in different contexts. It is important to note that, for the same reasons as explained earlier for synchronous systems, the bit rate will fluctuate along the isochronous signal. 3.1.5

Anisochronous

An anisochronous signal does not meet the requirement just defined. There is no “order” along an anisochronous signal. Note again that a signal that is isochronous with regard to one organizational unit (e.g., bit) may be anisochronous with regard to larger organizational units (e.g., words), as we shall see in a moment.

3.2

Reality Is Not So Simple: Bits, Words, Envelopes, and Frames The concepts of synchronization may sometimes be confusing. Let us see why. All signals we shall encounter in this book are isochronous at bit level; that is, all durations of the signal are an integral multiple of the duration of a bit. On the other hand, information is usually organized in octets so that the information itself is an integral multiple of eight bits. This is the number of bits representing one sample of pulse code modulation (PCM)–encoded speech, and since about 1975, the organization of all computers is based on particular multiples of 8 bits: 16, 32, and 64. The Internet (IP and TCP/UDP) uses 32 bits as an organizing unit. The fundamental unit of 8, 16, 32, or 64 bits is called a word, and what we would expect is that the signal used to transfer information is isochronous with regard to the duration of a word. As we shall see, this requirement is not necessarily always fulfilled. In many cases (data transmission), isochronous transfer of words is not important at all. It holds that the signal is isochronous at bit level. However, in real-time systems (speech and moving pictures), isochronism must be maintained at the level of words. Buffer size, bit error rate, transfer delay, and real-time requirements put constraints on the number of bits that can be transferred as a single unit across the network. In PCM, the basic unit is just one octet. The signal must be isochronous with

40

Synchronization

regard to this unit; otherwise, it will be impossible to transfer speech at proper quality. The payload of an ATM cell is only 48 octets long, allowing at most 384 information bits per cell. From a data transfer point of view, this number is very small, but it is well suited for multiplexing and transfer of real-time signals such as speech: it was the real-time requirement that determined the cell size. Moreover, 48 octets is a multiple of 32 bits (12 × 32), so that it is apparently isochronous with the word structure of the Internet. However, this is not the case because the first few octets of the payload is always a header of the ATM adaptation layer (AAL) so that the actual usable part of the cell is no longer an integer multiple of 32. Data communication has weak (if any) constraints concerning delay and real-time operation. The constraints are rather determined by the need for buffer size in senders and receivers, and the BER along the connection. The amount of information to be transferred between computers may vary from a single word to an infinite number of words. If the information string is longer than a specified maximum, the information is segmented (or fragmented) into pieces of this maximum length before being transmitted. The receiver reconstructs the original information string (e.g., an IP datagram) by reassembling the pieces in correct order. The maximum segment length is specified in the standard, taking into account practical buffer sizes in the receiver and the error performance of the links on which the information will be sent. The effect of BER is as follows. It is easy to detect that a signal is received with bit errors by appending a number of parity bits to the message before it is sent. The receiver removes these bits after analyzing them. If the receiver finds that a signal contains bit errors, the whole signal is discarded, since the receiver is not able to correct the error. If the signal consists of n bits, the probability that the signal contains at least one bit error is P = 1 − (1 − p)n, where p is the BER. The formula is derived as follows. If the probability that a bit is an error is p (the BER), the probability that the bit is correct is 1 − p. The probability that all n bits of the block are correct is then (1 − n p) , since the bits are independent, and it follows that the block error probability is P n = 1 − (1 − p) ; this is the probability that not all bits in the block are correct. P is called the block error rate or word error rate. Solving for n gives n = ln(1 − P)/ln(1 − p) = − ln(1 − P)/p if p is small (p < 1) and n = P/p if both P and p are small. In this equation, P is the required block error rate, and p is the actual bit error rate. Then n is the number of bits that can be contained in a block (or message) for given bit error rate p and block error rate P. It is easy to see that we must have n << 1/p if only very few blocks shall contain bit errors. BER is thus putting constraints on the length of the signals that can be transferred over a connection. This is illustrated in Table 3.1 for different values of BER and block error rate. The table clearly illustrates the effect of the BER on the −3 length of signals that can be sent on the channel. If the bit error rate is 10 and we accept a block error rate of 0.1 (every tenth block contains an error on average), then each block may contain 106 bits. If the acceptable block error rate is only 0.001, each block can only contain a single bit; that is, it is not possible to design a system satisfying both conditions simultaneously. Therefore, if the BER of the

3.2 Reality Is Not So Simple: Bits, Words, Envelopes, and Frames

41

Table 3.1 Maximum Number of Bits a Block May Contain Bit Error Rate Block Error Rate 0.1 0.001 –2 11 0 10 –3 106 1 10 –4 1,060 10 10 –8 5 6 10 10 10.6 × 10

−3

system is 10 , we must tolerate block error rates of 0.1 or more in order to pass a reasonable amount of error-free information. In data transmission, all blocks must be received without errors. This is achieved by employing an error correction technology called automatic retransmission upon request (ARQ). TCP makes use of this mechanism.3 The data link layer in most systems (e.g., GSM) also corrects errors by employing the ARQ algorithm. The method is as follows. The transmitter marks all messages by a sequential serial number. The receiver acknowledges all messages received without errors from the transmitter.4 If a message with a higher serial number than expected is received, the receiver assumes that one or more messages have been corrupted in one way or another (e.g., by bit errors or buffer overflow). The receiver then requests the transmitter to retransmit the lost messages. This means that the rate by which information can be sent across the connection depends on the BER. The signal then becomes anisochronous with regard to the information flow, or, in other words, the information does not flow smoothly anymore. This causes no problem for data transmission because this is an anisochronous service where information is stored in buffers and consumed by the receiver when the computational process in the receiver needs the received information. In addition, the receiver makes use of flow control to indicate that the receive buffer is full and that information transfer should be postponed until buffer space is again available. This also has an effect on the smoothness of the information flow. The essence is that in data transmission, the source (transmitter) and the sink (receiver) are not synchronous at the level of information transfer. On the other hand, the source and the sink in real-time systems are synchronous. ARQ cannot be applied to signals with real-time constraints such as speech and moving pictures because, as we just saw, ARQ destroys the synchronization relationship between the source and the sink. One way to achieve an acceptable quality of the received signal even if it is subject to bit errors is to reduce the segment length so that the word error rate is small. In real-time systems, lost information must be replaced by something because the timeslot exists even if there is no 3. 4.

ARQ is primarily not used in TCP to correct bit errors but to recover information lost because of congestion in the network. The receiver cannot acknowledge messages containing one or more bit errors because these messages are unreadable and are therefore discarded. The reason is that the receiver will not know how many errors the message contains and where the errors are located in the message.

42

Synchronization

information to fill it up with. For speech and moving pictures, the signal quality can be further improved by replacing the lost segment by the previous one. This is possible because of the high autocorrelation that exists between adjacent speech samples or picture pixels. There are essentially two ways in which information can be organized in frames, as shown in Figure 3.1. In the first example, the information bits are contained in envelopes of constant length that appear at constant rate. The envelope consists of header bits marking the start of the envelope (and sometimes also tail bits, though this is not shown in the figure). The information is inserted between two successive headers or between header and tail of the same envelope. The envelope structure is recursively used in the synchronous multiplex hierarchies (e.g., the PCM hierarchies), where envelope bits are added at each level of the hierarchy. The cell of ATM is also an envelope structure of constant length. These signals are isochronous with regard to both bit duration and envelope duration (and any divisors of the size of the envelope). Figure 3.1(b) shows the frame structure of high-level data link control (HDLC) where the envelope is not of fixed length in order to make the format suitable for data communication. The frames are delimited by flags consisting of the fixed 8-bit sequence 01111110. The number of information octets that can be inserted between two flags is arbitrary, possibly subject to maximum-length constraints of the frames. The receiver simply looks for the flag pattern in order to detect where one frame ends and the next begins. If there is no information to send, a string of contiguous flags is sent. In order to prevent flags from being simulated in the information stream, a zero is inserted after every five contiguous ones. This action is called transparency stuffing. The receiver removes the inserted 0s by the inverse algorithm. The resulting signal can be regarded as both isochronous and anisochronous depending upon what we are looking for. The signal is isochronous, since all

Bits at fixed beat

Envelope Envelope header at fixed beat (a)

Flag

Information at octet beat

Flag

…0111111010111111100000101000111111100101011111101… …011111101011111011000001010001111101100101011111101… After transparency stuffing (b)

Figure 3.1

(a, b) Envelopes and delimitation.

3.3 How to Acquire Synchronism: Phase-Locked Loop

43

intervals in the signals are still an integer multiple of the bit duration after insertion of the dummy zeros. Hence, we may operate the system with a fixed clock rate. On the other hand, the insertion of dummy zeros means that the frame length is no longer an integral multiple of octets. So if the duration of an octet is the minimum duration of a meaningful information element, the signal is anisochronous. This observation is important since most information sent on data circuits is organized as integral multiples of octets. Since the bit rate is constant and dummy zeros sometimes must be inserted, the number of information bits sent per unit time depends on the density of contiguous ones in the information. In fact, the number of bits sent varies from just the number of information bits if the occurrence of ones is sparse, to 20% more bits if all information bits are ones. In the latter case a dummy zero must be inserted for every 5 information bits. This means that the rate by which information is transferred is reduced by a factor between 1 and 1/1.2 = 0.83 depending upon the pattern of ones in the information signal. Or, in other words, the instantaneous information rate becomes a stochastic variable fluctuating between 0.83 times the nominal bit rate and the full nominal bit rate. HDLC is a suitable format for data communication because the original octet structure is recovered by the decoding process and the stochastic information transfer rate has no impact on the processing of the data. HDLC is not suitable for real-time transfer of deterministic information such as speech and moving pictures.

3.3

How to Acquire Synchronism: Phase-Locked Loop 3.3.1

Description of the Loop

The most versatile device for acquiring synchronism of signals is the phase-locked loop. Synchronism may be obtained by other means (e.g., injection locking of gallium-arsenide diodes and tunnel diodes), but such cases will not be considered here. Injection locking is the only available method in optical systems where synchronization is achieved by stimulated emission of radiation (see Chapter 10). Phase-locked loops are contained in all types of systems—not only for synchronizing digital signals but also for many other purposes. Examples are coherent demodulation of analog or digital signals, speed measurement (Doppler radar), precise distance measurement, programmable frequency synthesizers, filtering and regeneration of signals, satellite antenna tracking, accurate measurement of propagation time, and signal level control (coherent automatic gain control, often abbreviated coherent AGC). It is a basic component in almost all telecommunications systems. Therefore, it is important that the operation of the loop is understood. It is not easy to understand phase-locked loops from intuitive arguments. Their behavior is often surprising. Therefore, phase-locked loops can only be understood by analyzing them mathematically. However, this is still extremely difficult because the analysis of even such a simple device entails the solution of intractable nonlinear differential equations (analog loops) or intractable nonlinear difference equations (digital loops). We shall only consider the analog loop here because it is simpler to explain. The behavior of analog and digital loops are essentially the same. A mathematical description of the loop is given in the Appendix. The Appendix also

44

Synchronization

contains examples of how the loop components are designed. What follows is a verbal description of the loop. The phase-locked loop shown in Figure 3.2 is a very simple device. It consists of three components and a feedback loop. The purpose of the loop is to lock the frequency of a voltage controlled oscillator (VCO) to that of the input signal. The input signal can be the signal from a reference clock or the carrier wave or the digital signal received from another entity; see the following examples. Some characteristics of this component are as follows: •

•

•

•

•

Frequency acquisition. The main purpose of the loop is to produce an output signal with exactly the same frequency as the input signal. In other words, the VCO shall track the frequency of the input signal accurately. Noise reduction. Noise in the input signal and in the loop components produces noise in the output signal. The frequency of the output signal will then jitter slightly around the nominal frequency. One purpose of the loop is to make this frequency jitter on the output signal much smaller than the jitter on the input signal. The loop can then be used as a filter (e.g., in digital regenerators) to reduce the phase noise of the signal. Acquisition range. The loop shall acquire lock over a range of input frequencies and stay locked when acquisition has been achieved. The acquisition range must be larger than the sum of intrinsic instability of the VCO and the range of variation of the input frequency. Separate acquisition devices may be required in order to achieve this objective (see the Appendix). Synchronization range. When the loop has acquired synchronization, the loop shall stay locked to the input frequency over a larger frequency range than the acquisition range. This applies also if acquisition devices are used. Operating threshold. The nonlinearity of the phase-locked loop (PLL) causes certain types of unwanted behavior of the loop, such as cycle slipping. The onset of these nonlinear effects is a rather abrupt function of the noise level at the input and in the loop itself, producing a sharp threshold below which the loop no longer behaves as it should. In practical design, it must be checked that there is a safe margin to this threshold.

The loop operates as follows. Both the input signal and the output signal are fed to a phase detector that produces an error voltage proportional to the phase difference between the two signals. This is a dc signal if the loop has acquired lock; it is a slowly varying signal with frequency equal to the frequency difference (the beat frequency) between the input and the output signal if the loop is not locked. The phase

Input signal

Figure 3.2

Phase detector

Phase-locked loop.

Loop filter

Output Voltage controlled signal oscillator (VCO)

3.3 How to Acquire Synchronism: Phase-Locked Loop

45

detector also includes filters that eliminates higher harmonic frequency components and other mixing products of the input and output signal. Since the phase of a signal is a periodic function with period 2π (with sinusoidal, saw tooth, triangular, or other shape), the loop is obviously a nonlinear device. Different types of phase detectors are shown in the Appendix. The loop filter determines the phase tracking characteristic of the loop. It can be shown that only first-order lowpass filters with two independent time constants produce a loop with proper tracking capabilities. Loops without filters and loops with first-order filters containing only one time constant have inadequate tracking performance and stability. Filters of higher order do not improve the tracking performance and may cause instability. There are therefore only two types of loop filters that produce loops with satisfactory performance: F ( s) = (1 + τ 2 s) / τ1 s and F ( s) = (1 + τ 2 s) / (1 + τ1 s)

where s can be interpreted either as the imaginary variable jω, where ω is the angular frequency, or as the differential operator d/dt (see the Appendix), and τ1 and τ2 are the time constants of the filter. In the Appendix, the differential equation for the loop is derived for a loop filter of the second type. The signal from the loop filter is fed to the VCO. The VCO is an oscillator containing a resonance circuit that is electronically tunable such that the frequency of the VCO is proportional to the control voltage applied to it (the output of the loop filter). The oscillator can be designed as shown in the Appendix. 3.3.2

Applications

The phase-locked loop has a number of applications. A few of them are described here. The two most common applications of the phase-locked loop are to recover the clock rate of the duty signal in any digital system and, in addition, to recover the frequency of the radio carrier wave in radio systems. The methods applied in these two cases are essentially the same. The recovery of the clock signal is as shown in Figure 3.3. The input signal is first put through limiters and other devices for shaping the binary pulses into a square form. This process does not remove the phase jitter of the binary pulses produced along the transmission path. The input signal is then fed to one of the inputs of the phase detector, and the output signal from the VCO is presented to the other input, producing an error signal proportional to the phase difference between the two signals. This allows the VCO to produce a clean clock signal with the same nominal clock rate as the input signal but containing less jitter. In clock synchronization, it is the ability of the loop to reduce the phase jitter on the input signal that is exploited (the noise-filtering capability). The bit rate clock can be recovered in this way even if the input signal contains information as illustrated in the figure; that is, the signal consists of an irregular pattern of ones and zeros. This is the only method available in most communication systems. The operating condition is that there are enough state shifts between zero and one and vice versa in the input signal to make the error signal from the phase

46

Synchronization Input signal

(1) Pulse shaping

Output signal (3)

(2) PD

VCO

Loop filter

Jitter (1)

(2)

(3)

Figure 3.3

Recovery of the binary clock.

detector large enough. In this case, one major problem is that the distance to the operating threshold, and thus the probability for cycle slips, is a complex function of the density of state shifts, the jitter on the input signal, and the intrinsic stability of the VCO. Another complex case exists in satellite systems where we must incorporate the one-way delay of a geostationary satellite (about 250 ms delay) in the feedback loop. Intuitively, we see that the fastest varying component in the control signal to the VCO must be slower than a frequency corresponding to the two-way delay (i.e., the loop bandwidth must be less than 4 Hz). It turns out that the bandwidth of the loop must be narrower than this. It can be shown that the loop bandwidth should correspond to four times the round trip delay (i.e., it should be less than 1 Hz). The configuration of a TDMA system and the phase-locked loop required for recovering the clock rate (and indirectly the instant for transmitting the burst) is shown in Figure 3.4. One of the Earth stations acts as a reference station and the clocks (including the TDMA burst timing) of all other Earth stations are tuned to the clock of the reference station. The feedback signal (1) is derived from the signal sent from the Earth station and received 250 ms later from the satellite. The other input signal (2) is the signal from the reference station. This allows the Earth station to determine the exact position of its own burst in the TDMA signal from the satellite (see also Section 3.7.4). The loop can now be modeled with a delay in the feedback equal to the one-way delay over the satellite. Very high frequency oscillators (in the microwave band) can, if the frequency is not too high, be designed as ordinary transistors oscillators. At higher frequencies, devices called tunnel diodes and gallium-arsenide diodes can be used to produce the frequency directly. These oscillators are not very stable. Figure 3.5 shows how phase-locked loops can be designed to produce stable output signals from a high-frequency VCO. In the figure, the VCO is locked to a

3.3 How to Acquire Synchronism: Phase-Locked Loop

(2)

47

(1) Reference station

Earth station Configuration (Duty signals) Reference burst (1) RX

Own burst (2)

(1)

(Duty signal)

PD

Loop filter

VCO

TX

(2) Delay (250 ms) Equivalent loop

Figure 3.4

Burst and bit rate recovery in a TDMA satellite system.

frequency that is an integer multiple n of the frequency fref of a stable oscillator (e.g., a VCO at 10 GHz locked to a crystal oscillator at 200 MHz giving n = 50). The spikes are generated by nonlinear diode devices (e.g., step-recovery diodes) producing a voltage step with very short rise time whenever the signal voltage from the stable oscillator passes the conduction threshold of the diode. The spike opens the sampling gate for an interval equal to the width of the spike, letting through a narrow sample of the sine wave produced by the VCO. The amplitude of the sample is then a measure for how much the VCO frequency deviates from a multiple of the frequency of the stable reference oscillator. The sampling gate is thus a phase detector producing a voltage that is a function of the instantaneous phase difference between the VCO frequency and the wanted multiple of the crystal oscillator frequency. The requirement is then that the intrinsic stability of the VCO must be better than the frequency of the crystal oscillator. Otherwise, the VCO may lock to a wrong multiple of the crystal frequency. In other words, if the wanted frequency of the VCO is nfref, then the free running frequency of the VCO must be between (n − 1/2)fref and (n + 1/2)fref in order that the VCO shall lock to the wanted frequency and not to one of the other possible frequencies. For a VCO at 10 GHz locked to a crystal oscillator at 200 MHz, this means that the free running frequency of the VCO must be between 9.900 GHz and 10.100 GHz. This requirement is usually easily met. Again using 10 GHz as an example, the duration of an oscillation cycle is 100 picoseconds (ps) at 10 GHz or 50 ps for a half-wave. In order to produce a sample containing enough phase information, the width of the sampling spike should be

48

Synchronization

Output signal VCO

Loop filter

Sampler

Spike generator

Stable oscillator

Stable oscillator

Spike generator

VCO

From sampler

Figure 3.5

Stabilization of a high frequency oscillator.

about 10% to 20% of this value, or just 5 to 10 ps. It is a challenge in itself to produce such narrow spikes. This type of phase-locked loops is used in several satellite systems and radio relay systems. Figure 3.6 shows a frequency synthesizer that can produce any of the frequencies fo = nfi/km by selecting integers m, n and k. The divide-by-i circuits (i = k, n or m) are such that they can be set to produce any frequency of the type f/i for i in the range imin to imax, where imin and imax are two chosen numbers. Such circuits are made of a series of flip-flops with feedback. In telecommunications, frequency synthesizers are used to generate carrier frequencies for frequency division multiplexed (FDM) systems, frequency division multiple access (FDMA) systems, and frequency hopping code division multiple access (FH-CDMA) systems. Together with frequency multipliers, frequency dividers, and ring modulators, frequency synthesizers are the major ingredients in the musical instrument carrying the same name. In coherent demodulation of a phase-shift keyed (PSK) signal, the phase-locked loop is used to remove the modulation (the phase-shift keying) and produce a clean carrier wave that is used to extract the duty signal by correlating the modulated

3.3 How to Acquire Synchronism: Phase-Locked Loop Input frequency fi

Phase detector

÷m

49

Loop filter

VCO

÷k

Output frequency fo

Divide-by-n circuit ÷n

Selection circuit

Select output frequency

Figure 3.6

Frequency synthesizer.

signal with the pure carrier wave. PSK is the most used modulation technique for digital signals for transmission in the radio spectrum. The principle is shown in Figure 3.7 for a 2PSK signal, where the binary information is modulated as the two phase angles = 0 representing the bit value +1 and θ = representing the bit value −1. The modulated signal has then the form sin(ωt + θ), where θ takes any of the two values. After doubling the frequency (×2), we get sin(2ωt + 2θ) = sin2ωt; that is, a pure sine-wave at twice the original frequency. After halving the frequency (÷2), the recovered waveform is sinωt. The phase detector multiplies the two waveforms presented to it producing the following output signal: sin(ωt + θ)sinωt = cosθ − cos(2ωt + θ) = cosθ after the high-frequency term is

PLL

SPLL= sin(ωt)

Phase correlator

Si = sin(ωt + θ)

So= cos θ = ± 1

θ = 0 and π for 2-phase-shift keying (2PSK)

PLL sin(ωt)+θ

×2

sin(2ωt)

Phase detector

Loop filter

VCO sin(2ωt)

Figure 3.7

Coherent demodulator.

÷2

sin(ωt)

50

Synchronization

filtered out. Since cos 0 = +1 and cos = −1, the binary signal modulated on the carrier wave is recovered. n The signal can be modulated on the carrier using 2 phases producing the phase n−1 n angles θ = kπ/2 , where k is a number between 0 and 2 − 1. This allows n bits to be sent simultaneously. For example, if n = 2, then we may code (+1, +1) as θ = 0, (+1, −1) as θ = π/2, (−1, +1) as θ = π, and (−1, −1) as θ = 3π/2. This modulation is called n 4PSK or quadrature PSK (QPSK). In the general case with 2 phases, the modulation is called 2nPSK. Coherent demodulation is used in most digital radio systems (e.g., GSM, UMTS, WLAN, and Bluetooth). Demodulation takes place in these systems using methods similar to those of a 2PSK system.

3.4

Synchronization in Synchronous Networks 3.4.1

What Type of Synchronization Is Required?

Before entering into the detailed discussion of synchronization, it is useful to identify what is meant by synchronization in various applications. We may distinguish between end-to-end synchronization, link-by-link synchronization, and sporadic synchronization. The telephone network and the digital broadcast networks offer strict synchronization of real-time services, such as voice communication and transfer of moving pictures. However, such strict end-to-end synchronization is not always required. Upper and lower limits to the stability of the clocks of the receiver and transmitter must be defined. If the stability is kept within these limits, the subjective quality—that is, the quality that a person perceives—is not significantly reduced. If, on the other hand, the stability of the clocks is too poor, the speech becomes unintelligible or the picture can no longer be synchronized to the screen. The handling of asynchronous real-time signals is based on a particular type of redundancy inherent of the human brain. If the speech samples arrive with higher speed than the decoder clock, now and then a speech sample must be removed from the received signal in order to avoid buffer overflow. Similarly, if the speed of the received signal is too slow, an occasional sample must be presented twice to the decoder in order to avoid that the buffer becomes empty. If these events occur with very low probability, the deterioration of the speech quality is not subjectively recognizable. The reason is that the difference in the amplitudes of adjacent speech samples usually is very small, so that the absence of or the duplication of a sample does not cause significant noise. The same applies to moving pictures: the deletion or duplication of pixels is not recognizable by the eye if this takes place rarely. However, the most important deterioration of the quality of speech or video signals is caused by fluctuations in the differential delay. The differential delay is the difference in propagation time of adjacent speech or video samples. If this fluctuation becomes too big, the speech signal or the moving picture can no longer be reproduced properly. The challenge for IP telephony is to keep the differential delay fluctuations within an acceptable range. The differential delay of packets sent over

3.4 Synchronization in Synchronous Networks

51

the Internet depends on the traffic: the closer routers and transmission links are to saturation, the larger is the fluctuation in the differential delay. Data transmission is by nature asynchronous in the sense that the clocks of the receiver and the transmitter are completely independent. The only requirement is that the speed difference does not cause loss of data because of buffer overflow. Provision of flow control procedures in one or more layers of the data transmission protocol takes care of this problem. The flow control mechanism allows the receiver to request the transmitter to halt the transfer of data temporarily until buffer space again is available. In conclusion, end-to-end synchronization is strictly not required—and is in fact not universally offered—to support telecommunications services. However, large parts of the international telecommunications network (for example, the network of an operator) are strictly synchronous. The reason is that multiplexing synchronous signals is simpler and more efficient than multiplexing asynchronous signals. Link-by-link synchronization is a different matter altogether. A link is the connection between two entities in the network (e.g., two routers, two exchanges, or an exchange and a terminal). The link may contain transmission equipment such as transmitters, receivers, amplifiers, and regenerators. Link-by-link synchronization implies that the transmitter and the receiver clocks at both ends of a link are synchronous. Almost all systems are synchronous on a link-by-link basis, where the receiver clock of one entity is synchronized to the transmitter clock of the other entity. Sometimes the main clock serving both the transmitter and the receiver of one entity is synchronized to the transmitter clock of the other entity. There are systems where link-by-link synchronization serves no purpose. The two most important examples are Ethernets and wireless LANs, where the transmitters send data packets at sporadic instants. Then the receiving entity must synchronize the clock to the bit rate of each individual packet received from the medium. 3.4.2

Clock Hierarchies

All digital networks need stable clocks synchronized to a common ultra-stable clock. The reason is that signals from many sources are multiplexed and switched in common devices. Furthermore, a signal from any source may appear anywhere in the network—this is simply the connectivity requirement for networks. It is a trivial observation that if the rate of one such signal is faster than that of the other signals at a point in the network where the signals are multiplexed into a common stream, then there will be an accumulation of bits from the fast channel that never will be sent. The only way in which accumulation of bits is avoided is when all signals are synchronous and the network contains buffers that pick up short-term variations. Figure 3.8 shows two ways in which networks can be synchronized. The network may contain a single master clock and all other clocks (the slaves) are phase-locked loops that lock to the frequency provided by the master clock. A slave may derive the frequency from other slaves in a hierarchal manner. This is shown in the left-hand side of the figure. Several master clocks may also form a cluster of clocks that are mutually synchronized to one another, as shown in the right-hand side of the figure. In this case all the master clocks are phase-locked loops. Such a system is more reliable than systems with a single master clock, since the cluster can still operate if some of the

52

Synchronization

Master clock

Master clock

Slave clock

Master clock

Slave clock

Slave clock

Slave clock

Slave clock

Slave clock

Master-slave hierarchy

Figure 3.8

Master clock

Slave clock Slave clock

Slave clock

Master clock

Slave clock

Slave clock

Mutually synchronized master clocks

Hierarchies of clocks.

clocks fail. However, it is much more difficult to design such systems because of stability. If the loop parameters of each phase-locked loop are not properly aligned, the common frequency produced by the cluster may start to vary erratically. Atomic clocks based on a particular resonance in caesium atoms are used as high precision clocks in telecommunications networks. The stability of these clocks is 3 parts in 1012, which corresponds to a frequency drift of 1 second in 10,000 years. The master clock in networks owned by different operators is either such a clock or a clock locked to an atomic clock of another organization. Note that for several reasons (political, security, autonomy, economy, and so on), networks owned by different operators are locked to different clocks and are thus plesiochronous. 3.4.3

Master-Slave (Link-by-Link) Synchronization

The most common configuration of two systems in the network is shown in Figure 3.9. From a synchronization viewpoint, one of the systems is the master and the other a slave. The master can be an exchange (or router), and the slave can be another exchange (or router) or a terminal. The master can be the slave of other masters; the slave can be the slave of several masters and the master of several slaves. Each communications interface will then be configured as shown in Figure 3.9. The system works as follows. At the slave, a phase-locked loop recovers the clock rate of the master from the information stream. The recovered clock signal is used to restore the information stream (shaping the bits and reducing the jitter) and for timing of the signals sent to the master. The latter is achieved by letting the bits stored in the output buffer be read by the recovered clock rate. The regenerated

3.4 Synchronization in Synchronous Networks

Signal restoration

PLL

Processing

Buffer Information stream

53

Information stream

PLL

Signal restoration

Restored in-clock

Processing

Buffer Master clock

Slave

Figure 3.9

Master

Master-slave synchronization.

clock may also be used as local clock in the slave. One example where this is done is a regenerator on a communication link. The purpose of the regenerator is to restore the signal, amplify it, and retransmit it on the next section of the link. The regenerator makes use of the property by which phase-locked loop reduces the phase noise (jitter) on the input signal. At the master, the procedure is a little different. First, a phase-locked loop regenerates the clock rate of the slave from the information stream; that is, the loop reduces the phase noise generated on the transmission link (the filtering property of the loop). The regenerated clock is used to sample and restore the received signal. The restored in-clock rate may jitter relative to the clock of the master, even though the clock rate of the slave was derived from that of the master. This jitter is caused by intrinsic instability of the oscillators and short-term variation in the electrical length of the connection and in the signal propagation velocity along the transmission system, as explained in Section 3.1.1. In order to suppress this jitter, the signal is delivered to a buffer before it is read into the processing units of the master using the master clock. 3.4.4

Signal Restoration: Elastic Store

One way of designing an elastic store is shown in Figure 3.10. The device consists of an n-bit shift register (e.g., n = 8), a modulo-n up-down counter, and an output position selector. Bits are fed into the shift register at the input rate, and, when a new bit enters the register, the bits already in the registers move simultaneously one place to the right. The bit in the last position has nowhere to go, so it simply disappears. Suppose now that the next bit to be fed to the output circuit is at position i in the shift register. If a new input bit arrives when the output position is i, all bits in the shift register are moved one place to the right and the new output position is set to i + 1. If a pulse from the output clock arrives in output position i, the bit in this position is fed to the output circuit and the new output position is set to i − 1. The output position thus moves one position to the right for each input bit and one position to the left each time a bit is read from the register. All bits to the left of and including bit i are information bits that eventually will be fed to the output. The bits to the right of bit i are not significant because they will never be read.

54

Synchronization

In-rate

Input clock

Restored in-bit

Up-down counter

New position

...

n-bit shift register

Output position selector Out-bit Out-rate

Figure 3.10

Output clock

Elastic store.

The correct output position, i, is computed by the up-down counter. The counter is incremented by one for every clock pulse from the input clock (in-rate) and decremented by one for every clock pulse from the output clock (out-rate). The counter is initialized such that the number held by the counter is always equal to the number of the cell of the shift register that contains the output bit. The output position selector is a combination of logical gates controlled by the up-down counter such that it is always the bit in the current output position that is fed to the output circuit when a pulse is received from the output clock. In the start position, the first bit to be read is in the middle position of the shift register, allowing equal amount of fluctuations of the clocks in both directions. The n-bit elastic store therefore introduces a delay of the signal equal to n/2 bits, since this is the number of bits that must be read into the register before the first bit is outputted. If the register is full and a new input bit arrives, the new position of the up-down counter is the first bit in the register. The bit in the last position is lost. If the register is empty after an output bit is read, the new position is the last position in the register. An error will now occur if a new output pulse arrives before a new bit has arrived.

3.5 Interconnection of Plesiochronous Networks: Application of Elastic Store Plesiochronous networks are asynchronous by definition, though they are tuned to stable clocks. The networks of different telecom operators are in general plesiochronous. Therefore, we must be able to interconnect such networks. For this purpose we may use elastic stores using the same arrangement as in Figure 3.10. What is different from the synchronous case described earlier is that the input and output clocks are not synchronous. The purpose of the elastic store is to capture both long-term drift of the clock and short-term jitter. The elastic store used in PCM systems consists of a shift register containing 16 bits. Suppose that the input clock runs a little faster than the output clock. Sooner or later the register then becomes full. When the register is full and

3.6 Synchronization of Envelopes of Constant Length

55

a new bit arrives, the last bit in the register is pushed out of the register and the new read-out position is reset to bit number 9. A complete PCM sample of 8 bits is lost. This is called a slip. If the input clock is slower than the output clock, the output pointer gradually moves toward the first bit in the register, and if the register becomes empty the pointer is moved to position number 8 and the same octet is read once more. This is also called a slip. If drift is combined with jitter, the pointer may move back and forth in the same way as for synchronous systems. Since the slip is always 8 bits, there will be no mixture of bits from different octets whenever a slip occurs, and we conclude that the signal maintains isochronism on an octet basis. The average time between slips is ∆T = N/∆R, where N is the number of bits lost in each slip and ∆R is the difference in bit rate between the clocks. For N = 8, clock stability 10−12 and bit rate 1 Gbps, we find ∆T = 2 hours and 13 minutes. See also Chapter 4 for the method used for plesiochronous operation in SDH.

3.6

Synchronization of Envelopes of Constant Length 3.6.1

Direct Acquisition and Tracking of Envelopes

This subsection describes how envelope synchronization takes place in the standard digital multiplex hierarchies PDH and SDH described in Chapter 4. The description holds for every system where the transmission format is based on envelopes of constant length and with an invariant format of the synchronization bits. The ATM system does not belong to this category since the envelope does not consist of a stationary synchronization format. ATM synchronization is described in the next subsection. It is assumed that bit synchronization has already taken place as described earlier and that the local clock follows the input signal with sufficient accuracy to avoid bit slips because of clock jitter (if necessary, applying elastic store). The acquisition method is shown in Figure 3.11. The information (or payload) is contained between envelope synchronization words located equidistantly along the bit stream. The purpose of envelope synchronization is to extract the payload from the bit stream for further handling of the information (sending it forward in new envelopes or consuming it for computational purposes). The synchronization word is a sequence of bits that is the same in every envelope.5 The length of the payload in the first-order European multiplex hierarchy is 504 bits. The synchronization word consists of 7 bits preceded by an unused bit, giving a total length of the envelope of 512 bits. The length of the payload is 836 bits in the second order multiplex hierarchy. The synchronization word consists of 10 bits (followed by two bits used for other purposes). See Section 4.3 for details of these systems.

5.

In some systems, the word is inverted in every second envelope for reasons such as faster synchronization (smaller probability of locking to simulated words) or to distinguish between odd and even frames because they may contain different information.

56

Synchronization

Synchronization word

Simulated synchronization word

Payload

Length of envelope Correlation peak

Detection period

Filtered out

One bit

Figure 3.11

Phase characteristic of tracking mechanism

Acquisition mechanism.

The synchronization consists of an acquisition phase where the receiver hunts for the synchronization word of the envelope. If the synchronization word consists of 5 bits, the acquisition mechanism consists of comparing every consecutive series of 5 bits with the fixed format of the synchronization word until a matching sequence is found. When a synchronization word is detected in this way, a new measurement is made on the next expected synchronization word one envelope-duration later, ignoring the bits between the detected synchronization word and the position of the first bit of the next expected word. If there is still a match, the procedure is continued a number of times (say, six times) before it is supposed that proper envelope synchronization has taken place. This is called the presynchronization procedure. When the presynchronization phase is completed, the receiver can extract the payload. If there is no match at the next expected position of the synchronization word during the presynchronization phase, the receiver leaves the presynchronization phase and continues to hunt for a new synchronization word. After acquisition is completed, the receiver continues to track the synchronization word. Phase-locked loops are not required for this purpose since the position of the synchronization word in the envelope is determined by simple counting. As indicated in the figure, tracking must be done with an accuracy of 1 bit, since otherwise there will be no match. If the envelope synchronization word is not detected a number of times in succession (say, six times), the receiver assumes that synchronization is lost and enters a new acquisition phase and starts the hunting procedure.

3.6 Synchronization of Envelopes of Constant Length

3.6.2

57

Acquisition and Tracking Using Error Detection: ATM Synchronization

ATM may at first glimpse be a confusing name for this system: ATM is one of the most isochronous systems that exist. Only cells consisting of 53 octets are sent along the ATM connections; even empty cells have this format. Empty cells are distinguished from other cells by setting all bits in the first three octets of the header to 0, and encoding the fourth octet as follows: 00000001.6 The existence of 1 in this position ensures that the HEC also contains ones so that the header does not only contain zero bits. If this bit also had been zero, then all bits of the header would be zero. This would make it impossible to use empty frames for synchronization, and, consequently, the synchronization method described next would not be possible. ATM contains three formats of constant duration: the bit, the octet, and the cell, which is again an integer multiple of the other durations (8 bits per octet and 53 octets per cell, or 424 bits per cell). The octet structure represents, as in many other systems, the basic unit of information. In this structure, the envelope of constant length is the ATM cell. The basic cell structure consists of a header containing 5 octets followed by a payload of 48 octets, as shown in Figure 3.12. The first 4 octets of the header are used for switching and management purposes. The fifth octet of the header is used to detect and sometimes correct bit errors that may appear in the first 4 octets of the header. This octet is usually referred to as header error correction (HEC). This octet plays a crucial role in cell synchronization, as we shall see. The content of the HEC is computed from the first 32 bits of the header by sending them through a shift register with feedback. The shift register, consisting of 8 flip-flops,7 as shown in Figure 3.13. There is feedback after the first, second, and last flip-flop. 8 2 8 This is usually expressed as a polynomial: 1 + x + x + x . A single feedback signal is

ATM cell Header

Payload

5 octets

48 octets

Information Empty cell … … Header

Figure 3.12 6.

7.

8.

ATM formats.

The final bit of the fourth octet is the cell loss priority (CLP) bit. If this bit is set to zero, the cell shall, if at all possible, not be discarded if there is congestion in the network. If the bit is set to one, there is no such priority. Empty cells can, of course, always be discarded. A flip-flop is an electronic circuit consisting of two transistors. The arrangement is such that only one transistor of the pair can conduct current (or is open) at a time. This gives rise to two states depending upon which of the two transistors is open. The device can then be used to distinguish between and to store binary zeros or ones. The polynomials also provide us with a pure algebraic method to find the suitable shift register for producing the most appropriate sequence. The same method is used for generating encryption codes and pseudorandom sequences for other purposes.

58

Synchronization

produced by combining the feedback signals in exclusive OR (XOR) gates (the circles with a cross). For every clock pulse the content of the shift register is moved one place to the right. The final bit is fed back to the first flip-flop as shown. The state of the first flip-flop is determined by both the input signal and the feedback signal, as shown in Table 3.2. The first 32 bits are written sequentially into the shift register at clock rate. A counter takes care of counting the bits. When the last bit has been written into the shift register, the shift register contains the HEC. The 8 bits of the HEC are then written to an output device that writes the HEC serially into the ATM output bit stream. The counter resets the shift register to zero and blocks the input to the shift register until the header of the next cell appears. Then the cycle starts over again. The counter is a rather complex device in charge of several timing functions. The details of the counting process have been omitted in the figure. The receiver contains a similar shift register as the sender, producing the HEC from the received bits of the header. The computed HEC is then compared with the HEC contained in the header. If they match perfectly, the header is correctly received. If not, the 32 bits are either not a header or they are a header containing bit errors. The first situation is exploited for synchronization. In the second case, the bit errors in the header may be corrected. We also see that the HEC for empty cells consists of 7 zeros and a one in the final bit of the octet. The header of empty cells then consists of 2 ones: in bit positions 32 and 40. Asynchronous transfer mode means that the bit rates of the information streams sent across the connections need not be synchronized to the cell rate. This is, in fact, an advantage since the interfaces between the information system and ATM are not

Clock

Input

1

2

3

4

5

6

7

8

Reset Counter Serial output HEC Write

Figure 3.13

HEC generator.

Table 3.2 Input 0 0 1 1

Binary State of First Flip-Flop Feedback Binary State 0 0 1 1 0 1 1 0

3.6 Synchronization of Envelopes of Constant Length

59

critical with regard to timing. The information streams are buffered before they are inserted into the ATM cells for alignment with the bit rate of the ATM channel. For transmission, the information bits are thus read into the buffer at the rate of the information stream and read out from the buffer at the ATM rate, and vice versa for reception. The pieces of information transferred by an ATM connection are indicated by the shaded areas in Figure 3.12, where each shaded area may be a part of the same information stream or belong to different streams. The information bits (or payload) need not fill up the cells completely as shown. The multiplexing is taken care of by the AAL. AAL also takes care of segmentation and reassembly of information strings that do not fit into the payload of one cell, as well as timing and flow control of the information stream as required. The synchronization at the AAL layer (i.e., the synchronization of the information stream) is taken care of by reading information appearing at certain places in the cell. The first octet of the payload is used for this purpose, together with length indicators and segmentation indicators (more segments/final segment indicators) depending upon the type of traffic the network supports. However, all these methods require that the start of the payload is known. Since only cells 53 octets in length are sent on the ATM connections, it is simple to retain cell synchronism when it is first achieved. However, synchronization may sometimes be lost, and the system needs to be resynchronized. Furthermore, newly installed equipment must also be synchronized to the cell stream before information can be sent or received. ATM is thus a system with envelopes of constant length (one cell). However, the envelope does not contain an invariant synchronization word. Let us then look at how the HEC is used for cell synchronization. Except for the particular use of the HEC, the procedure is the same as just described for envelopes of constant length. The HEC in error detection mode is used for synchronization, and the principle is as follows. Start at an arbitrary bit (let us call it b1) and compute the HEC for this and the next 31 bits (4 octets) and compare the result with the content of the next 8 bits. If they match, it is possible that the correct start of the header has been found. However, it is also possible that this was just a coincidence—a sequence of 40 consecutive bits appearing somewhere in the bit stream that simulates the format of a header. The synchronization must therefore be reconfirmed by the presynchronization procedure. This is done by running through the same computation once more but now starting at the bit b1 + 424 (the same bit position one cell ahead). If the result still matches, it is more likely that synchronism has been achieved. However, the event may still be a coincidence so that the procedure is repeated a number of times—say, 7—before it is assumed that synchronism has been achieved. If the computation started on bit b1 does not match, a new computation may started at bit b2 = b1 + 424 + 1 (the next bit in the next cell). This is called the hunting procedure. If there is still no match, the next computation is started at bit b3 = b2 + 424 + 1 and so on until the first match occurs. Then the presynchronization procedure is followed until synchronization is assured. This procedure is very slow and is included here only in order to make the explanation simpler. The procedure is sped up in practicable systems by doing the hunting calculations in parallel for

60

Synchronization

consecutive bits b1, b2, b3, and so on, until a match occurs and thereafter following the presynchronization procedure as described. This requires several parallel calculations of the HEC (424 calculations may be required in order to find a match, but at most 40 of these calculations must be done in overlapping time intervals). Using this procedure, a match will then occur within the duration of a single cell. Synchronization may also be lost. If synchronization match does not occur for a certain number of consecutive cells, it is assumed that synchronization is lost and a new hunting procedure is initiated.

3.7

Synchronization of Radio Systems Two major problems concerning time division multiple access (TDMA) radio systems (see Section 5.3 for the description of TDMA) are how to achieve initial synchronization and how to maintain synchronization during a communication session. The methods described in Section 3.6 cannot be applied to radio system because every TDMA burst must be synchronized independently while the systems described earlier were synchronized by hunting over several time division multiplexing (TDM) time slots. How synchronization in radio systems can be implemented is shown for three types of systems requiring fundamentally different methods for acquiring and maintaining synchronization. Other systems use methods that are variations of the three principles shown. 3.7.1 General Synchronization Sequences in TDMA and Random Access Systems

In the radio systems considered here, information (payload) is sent in bursts with the general format shown in Figure 3.14. The format of practical systems may be different, but they operate in the same way as the general system. The carrier and clock synchronization fields are encoded as a single field in GSM, Ethernet, and wireless LAN; the unique word in GSM is located in the middle of the burst splitting the payload field into two parts. The carrier synchronization field is used for regenerating the carrier in the receiver in order to provide coherent demodulation of the signal. This may be a sequence of bits of equal parity (e.g., only ones). The clock synchronization field contains enough bit shifts (for example, a sequence of alternating zeroes and ones) to tune the local clock to the input signal. Phase-locked loops are used for carrier and clock acquisition. The unique word determines the start of the payload. The format allows independent handling of adjacent bursts offering a large degree of flexibility, as will be evident from the examples that follow. Adjacent bursts may be of different duration, tuned to different carrier frequencies and clocks, and received at random instants.

Carrier synchronization

Figure 3.14

Clock synchronization

General burst format.

Unique word

Payload

3.7 Synchronization of Radio Systems

61

The composition of the receiver is shown in Figure 3.15. The receiver consists of an antenna, a low-noise amplifier, a mixer and an amplifier with AGC (amplifier/mixer in the figure) followed by a coherent demodulator, a bit shaping unit, and a circuit where the payload is extracted. The start of the payload is provided to the payload extraction circuit by the unique word (UW) correlator. The carrier frequency is extracted by a phase-locked loop and then applied to the input signal as explained in Section 3.3.1. The carrier is extracted and used for signal restoration in the same way as in master-slave synchronization. We may model the unique word correlator as a shift register of length equal to the number of bits in the unique word. The input signal is read into the shift register at the rate provided by the clock (which, of course, is equal to the bit rate of the input signal). The bits in the shift register are compared one by one with the corresponding bits in the unique word for each beat of the clock, thereby producing a voltage that depends on how many bits match the unique word and how many bits do not. Let us denote the two binary states of a bit as +1 and −1, respectively. Let the unique word be U. The correlation Corr between U and a sequence S of the same length as U is obtained by multiplying the corresponding bits in the two sequences and adding the result for all the bits in the sequence; that is, the correlator produces the voltage V = Corr(U,S) = Σuisi, where the sum is taken over all the bits in the sequence and where ui is the ith bit of U and si is the corresponding bit in S. Note that ui and si only take one of the values +1 or −1. The maximum voltage is obtained if there is a perfect match; that is, the sequence in he shift register is exactly the same 2 as the unique word. Hence, since S = U, V = Corr(U,U) = Σuiui = Σui = Σ1 = N, where N is the number of bits in the unique word. It follows also trivially that if there are k bit errors in one of the sequences, then Corr(U,U with k bit errors) = N − 2k. It then follows that if all bits are in error (i.e., the new sequence is the binary inverse of the original sequence) then the correlation becomes − N, as it should be. Let us take the 7-bit sequence (−1,−1,−1,+1,−1,+1,+1) as an example of a unique word preceded by a clock synchronization sequence …,−1,+1,−1,+1. Table 3.3 shows the correlation function calculated for a sequence consisting of k synchronization bits and the first 7 − k bits of the unique word for all values of k from 0 to 7. Apparently there is a good detection margin, since all values of the correlation function (except for k = 6) are much smaller than 7. However, in order to determine the detection margin, we have to consider the absolute value of the correlation function. The reason is as follows. A bit sent as +1 may be demodulated either as +1 or −1 by the receiver. That is, the demodulated bit may have the same or the inverse parity compared to the

Amplifier mixer

Figure 3.15

Receiver.

PLL carrier

PLL clock

UW correlator

Demodulator

Bit shaping

Payload extraction

62

Synchronization

transmitted bit. This is so because the phase of the recovered carrier can take any value corresponding to a modulation state of the PSK signal, since there is no absolute phase reference. This means that it is not known a priori whether the received unique word will appear as the sequence (−1,−1,−1,+1,−1,+1,+1) or the inverse sequence (+1,+1,+1,−1,+1,−1,−1) after demodulation. In the latter case, perfect match occurs if the correlation result is −7. We see from the table that the input sequence corresponding to k = 5 results in a correlation value of −5. This is not a safe detection margin since it is too close to −7. In fact, if the parity is not known a priori, the correlator cannot distinguish between the case corresponding to k = 5 and an inverted unique word containing a single bit error. In fact, the unique word is also used to determine the correct parity of the bits. If the correlation value is −5, we can therefore not discriminate between the cases where the correlation actually corresponds to k = 5 for a noninverted unique word or to k = 0 for an inverted unique word containing one bit error. The sequence in the example is thus not a good unique word. This particular unique word was chosen in this example just to illustrate this point: if parity is known, the margin is safe; if parity is not known, which is usually the case, the margin is not safe. The challenge in practicable systems is to find a unique word that provides a safe detection margin for the particular synchronization sequence preceding the unique word. In most systems, the unique word is considerably longer than 7 bits. A value often used is 16 bits because it is divisible by 8 and thus preserves isochronism with regard to the octet structure. The unique word consisting of the 16 bits 0000110010111101 offers a safe margin. This word is used for WLAN. 3.7.2

GSM: Timing Advance Procedure

Figure 3.16 illustrates the timing problem in GSM. If two mobile terminals located at different distances transmit bursts at the same time (but in adjacent timeslots), the bursts will not be adjacent when they arrive at the base station. The time difference between the two signals is then τ = (d2 − d1)/c, where c is the speed of light. If mobile terminal 2 is instructed to send the message a time earlier, it is evident from the figure that the bursts from the two terminals will be adjacent when they arrive at the base station. The procedure that takes care of the adjustment of the transmission

Table 3.3 U(0)

Correlation of the Sequence (−1,−1,−1,+1,−1,+1,+1) k 0 1 2 3 4 5 6

−1 −1 −1 +1 −1 +1 +1 Correlation Absolute value

−1 −1 −1 +1 −1 +1 +1 +7 7

+1 −1 −1 −1 +1 −1 +1 −3 3

−1 +1 −1 −1 −1 +1 −1 −3 3

+1 −1 +1 −1 −1 −1 +1 −3 3

−1 +1 −1 +1 −1 −1 −1 −1 1

+1 −1 +1 −1 +1 −1 −1 −5 5

−1 +1 −1 +1 −1 +1 −1 +3 3

7 +1 −1 +1 −1 +1 −1 +1 −3 3

3.7 Synchronization of Radio Systems

63

instant is called the timing advance procedure. Before explaining the details of this procedure, let us have a brief look at how the radio channels are organized in GSM. The base stations in GSM are sending continuous streams of TDMA bursts (see Chapter 8) organized in frames consisting of 8 bursts. Each burst, including guard space, is 0.577 ms long. Bursts that are spaced 4.615 ms apart belong to the same communication channel; that is, bursts with the same burst number in Figure 3.17 belong to the same channel. The communication channels are organized on top of the burst in a very complex pattern, where different patterns are used for different types of channels. These complex patterns are the result of difficult tradeoffs between real-time support of speech signals, combinations of speech, data and signaling channels at different rates in a common synchronous structure, provision of time slots allowing the terminal to measure the field strength of signals from different base stations, and severe radio propagation conditions. In this section, we shall only consider the synchronization of bursts. The synchronization of the communication channels is derived from the burst pattern. Different frequency bands are allocated for the transmissions from the mobile station and the base station. The transmission in both bands is organized in identical patterns of TDMA timeslots. In the direction toward the base station an individual timeslot at a given instant is either empty or contains a burst from the mobile station to which it was assigned. The transmit time of the mobile station is three burst times later than the corresponding burst from the base station. The arrangement is called time division duplex and has the advantage that the mobile station need not receive information while sending, making the radio interface of the mobile station simpler (no duplex filter but just a switch isolating the sender from the receiver). Note that radio systems share the same antenna for reception and transmission for reasons of size and cost. The arrangement is shown in Figure 3.17. The figure also shows the bit format of the standard burst used for regular communication, and the random access burst used by the mobile terminal to request the base station to allocate communication resources. The standard burst consists of a 3-bit tail at each end, two data fields consisting of 57 bits each, and a midamble consisting of 2 stealing bits and the training sequence. The 2 stealing flags are used

2

1

d2

1

2

d1

1

2 τ = (d2– d1)/c

2

1

Timing advance τ

Figure 3.16

Timing in GSM.

1

2

Time

64

Synchronization Frame, 4.615 ms 6

7

0

1

2

3

4

5

6

7

0

1

2

(a)

6

7

0

1

2

3

4

5

6

7

0

1

2

6

7

0

1

2

3

4

5

6

7

BS transmits

0

1

2

MS transmits

(b)

T

Data

S

TS

S

Data

T

3

57

1

26

1

57

3

G

T = tail, S = stealing flag, TS = training sequence, G = guard time (8.25 bits) (c)

T

Synchronization

Data

T

Guard space

8

41

36

3

68.25

(d)

Figure 3.17

Frame and burst formats.

when data has to be replaced by signaling information—for example, during handover. The start of the data field of the bursts is determined from the training sequence. This is a known bit pattern (a unique word) having strong autocorrelation properties; that is, it provides a unique output peak when sent through the autocorrelation device described earlier. The duration of this peak is narrower than 1 bit, thus providing accurate timing information.9 The random access burst consists of an initial 8-bit tail, a synchronization word consisting of 41 bits, a data field used for setting initial parameters, an end tail of 3 bits, and a guard space corresponding to the duration of 68.25 bits. The synchronization field is used for synchronization of the burst and an initial computation of the transfer function of the radio path. Mobile systems face two major problems: the dynamics of received power level and the timing accuracy of the burst. The received power level varies over a range of about 70 dB in GSM. Since random access bursts are received from different mobile stations, the receiver of the base station must be able to adjust to this huge level difference during a period of time equal to the 8 tail bits at the start of the random access

9.

The training sequence serves also other purposes. Since it is a known bit pattern, the receiver can use it to calculate the transfer function of the radio channel and use this information for adaptive error correction. Since the characteristics of the radio channel may change significantly even over a single burst, the training sequence is placed in the middle of the burst in order to estimate the average transfer function for the burst.

3.7 Synchronization of Radio Systems

65

burst. The level variation between adjacent normal bursts is smaller because of power control management so that the tail sequence can be shorter. During the initial tail period, the AGC of the receiver must adjust the gain to normal detection level. The duration of a single bit is 3.69 µs. The duration of 63 bits is then 232.5 µs. −6 This corresponds to a distance of 232.5 × 10 seconds × 300,000 km/s = 70 km where 300,000 km/s is the speed of light in vacuum. The magic of this calculation is that the base station can adjust the timing of transmitted bursts from the mobile station by at most 63 bit positions. A field of 6 bits is allocated in the signaling channel from the base station for this purpose. The field is coded such that 000000 equals no adjustment, 000001 equals 1 bit adjustment, 000010 equals 2 bits adjustment,…and 111111 equals 63 bits adjustment. The adjustment is relative to the instant when the corresponding frame from the base station was received, as shown in Figure 3.18. This is the timing advance procedure. Since the delay has to be divided between the base station-to-mobile link and the mobile-to-base station link, the maximum distance between the base station and the mobile station that can be adjusted is then 35 km. This corresponds to the maximum cell diameter in GSM. The maximum size of GSM cells is thus not determined by radio propagation conditions but by synchronization constraints. The normal timing advance procedure is shown in Figure 3.18(a). This procedure applies when the mobile terminal has been synchronized to the burst pattern and is used to adjust the timing for moving mobile terminals. The base station measures how accurately the burst received from the mobile terminal fits into the nominal receive slot. If the deviation equals 1 bit in one or the other direction, the base station instructs the mobile terminal to move the transmission instance 1 bit duration earlier or 1 bit duration later.

TX RX

Base station

Mobile terminal

RX TX

0

1

2

3 3

4

5 Time

0 3

Propagation delay

Time

Timing advance (a) TX RX

Base station

Mobile terminal

RX TX

0

1 RA

0

Propagation delay

2

1

Time

2

RA

Time Guard space

(b)

Figure 3.18

(a, b) Timing advance and sending of random access burst.

66

Synchronization

In the idle position, the mobile station is tuned to the frame structure of the base station bursts. The first signal the mobile station sends both for incoming and outgoing calls is a random access signal. The TDMA bursts are organized such that there is one time slot available for sending random access messages from the mobile terminal per frame. The mobile terminal will not know how far it is from the base station when a call is initiated. In order to avoid the random access burst overlapping with other bursts, these bursts are shorter than normal bursts, as shown in Figure 3.17, and the mobile terminal sends the random access burst as early as possible in the available time slot, as shown in Figure 3.18(b). Receiving this signal, the base station can estimate the distance to the mobile terminal by determining how much the signal is delayed relative to the start of the base station bursts. The base station then sends a correction message instructing the mobile to move its transmission time a number of bits earlier in order to compensate for the delay. Thereafter, the normal timing advance procedure commences. The guard space for the random access burst is 68.25 bits × 3.69 µs/bit = 252 µs. This corresponds to a one-way distance of 38 km. Since 38 km > 35 km (the maximum size of the cell), the random access burst will not collide with the adjacent bursts if the mobile is within the cell. However, collisions may still occur if the propagation condition allows signals to be sent over larger distances. In such cases, the base station does not respond to the request. 3.7.3 Wireless LAN: Finding the Information in Sporadically Transmitted Frames

Wireless LAN systems are based on packet radio technology. The most important characteristics of these systems are as follows: •

•

•

Physical frames can be sent at any instant in time: there is no master-frame synchronization. The terminals are plesiochronous, and the bit rate of each terminal is determined by the local clock of the terminal. This clock is usually not tuned to any system clock. This implies that the stability of the bit rate of different terminals sharing the same medium is not better than that of the clocks of personal computers or similar equipment. The power level of received frames can vary considerably. The dynamic range is, however, smaller than in the GSM system.

These characteristics have led to a set of solutions that is exemplified by that of the wireless LAN known as IEEE 802.11. The structure of the physical frame of this system is shown in Figure 3.19.10

10. Direct-sequence code-division multiple access (DS-CDMA) frames are slightly different, but they contain the same elements. See Sections 5.4 and 5.5 for a description of SFH-CDMA and DS-CDMA, respectively.

3.7 Synchronization of Radio Systems

67

Synchronization

Start frame delimiter

PLCP header

Payload

80

16

32

variable

Figure 3.19 Format of physical frame in IEEE 802.11 for slow frequency-hopping code-division multiple access (SFH-CDMA).

The frame starts with a sequence of 80 bits with the pattern 010101... This sequence enables the receiver both to recover the carrier frequency of the radio signal for coherent demodulation and to synchronize the digital clock of the receiver to the bit rate of the sender. It does not matter whether or not one or more bits at the beginning of this field are lost as long as the clock of the receiver locks to the pattern. The start of the header of the physical frame—the physical layer convergence protocol (PLCP)—and thus the start of the payload is derived from the start frame delimiter. PLCP indicates the length of the payload and the data rate (1 Mbps or 2 Mbps). The start frame delimiter is a unique word with properties as described earlier. The pattern of this particular unique word is 0000110010111101. This pattern gives rise to a sharp autocorrelation peak at the start of the PLCP header field, enabling the receiver to extract the PLCP header field and the subsequent payload. The synchronization field and the start frame delimiter field are often referred to as the PLCP preamble of the frame. The payload contains the media access control (MAC) layer. 3.7.4

Satellite Systems: Managing Long Delays

The methods used for synchronization of satellite systems are similar to those of the GSM system but are more complex. The geometry of a geostationary satellite system is shown in Figure 3.20. The orbit of a geostationary satellite (i.e., a satellite that uses 24 hours on one revolution around the Earth) is located 36,000 km above the equator. The time it takes a radio signal to propagate from the Earth to the satellite is 36,000/c, where c is the speed of light in a vacuum in km/s. This gives a propagation time of 120 ms between satellite and the Earth. The radius of the Earth is 6,378 km so that the distance from the satellite to a tangential point on the Earth is about 42,000 km. The maximum propagation time is then 140 ms. This generates a differential delay of 40 ms between two Earth stations (Earth station-satellite-Earth station) on the equator just below the satellite and between two Earth stations, both at the edge of coverage. The bit rate of an intercontinental satellite system may be several hundred megabits per second. Using a bit rate of 100 Mbps as an example, the duration of a single bit is only 10 ns, corresponding to a propagation distance of approximately 3m for the electromagnetic wave. This means that the differential delay of 40 ms corresponds to 4 million bits. In the GSM system, the differential delay corresponds

68

Synchronization N 6,378 km

42,000 km

36,000 km

S (a)

Burst n

1

2

...

3

n

1

Frame

(b)

Carrier clock

UW

Payload

(c)

Figure 3.20

(a–c) Satellite system.

to about 30 bits, which is more than 100,000 times smaller. Therefore, we need a synchronization method completely different from the one used in GSM.11 The guard time between bursts in a satellite system is just a few bits corresponding to a time interval of less than 0.1 µs (or 10 bits). The path length must then be determined with a precision of 30m or less in order that adjacent bursts shall not overlap. One way of achieving such a formidable accuracy is as follows. An Earth station in a symmetric system such as INTELSAT can monitor the bursts of all stations including its own since these bursts are retransmitted in an unmodified form from the satellite. One of the bursts (burst 1) is used as a reference burst so that all Earth stations have a common burst timing reference. The unique 11. Historically, the synchronization problem of TDMA satellite systems were studied and solved during the late 1960s and the early 1970s. During the mid-1980s, the technology was taken into use by INTELSAT for intercontinental communication. This was long before the method was applied in land mobile systems. The same applies to random access techniques. The ALOHA system of 1971 was developed for satellite applications.

3.7 Synchronization of Radio Systems

69

word of the control station is different from the unique words of the other earth stations (which usually have the same unique word). Each Earth station can thus monitor the position of its own burst relative to the reference burst and correct the position of its own burst using phase-locked loops. In this way, very accurate timing of the bursts is possible, allowing guard space between bursts of just a few bits. Note that the loop delay of this phase-locked loop is 280 ms. The loop bandwidth must then be narrower than 1 Hz in order to ensure proper stability of the loop (a frequency corresponding to approximately four times the path delay). The burst arrangement of a TDMA satellite system is shown in Figure 3.20(b). Figure 3.20(c) shows the composition of a single burst. The burst consists of a sequence for carrier synchronization and bit timing followed by a unique word that is, as we have seen several times now, a sequence of bits with very good autocorrelation properties. The autocorrelation peak determines the start of the payload of the burst with the accuracy of a single bit. If all Earth stations are in synchronism, it is easy to maintain synchronism. All that is needed are phase-locked loops that use the autocorrelation peak to keep the burst within the timeslot assigned to it. Since the loops are very narrow (order of 1 Hz), their short-term stability (fluctuations faster than 1 Hz) must be good. The stability requirement of the clocks can be estimated as follows. If the bit rate is 100 Mbps, then, during 1 second (corresponding to 1 Hz), 100 million bits (108 bits) are received. In order to keep the jitter of this signal smaller than 1 bit (or 0.5 bits in each direction), the stability of the loop oscilla−8 tor must be better than ± 0.5×10 . It must also be possible to bring a new station into operation. This is one of the most difficult procedures of TDMA satellite systems. We see that a guard time of just 10 bits corresponds to 30-m length variation of the propagation path. It is not possible to measure the path lengths between Earth stations and geostationary satellites with such accuracy using surveying techniques. Furthermore, the frame is much shorter than 280 ms so that random access, such as in the GSM system, cannot be used. Therefore, the initial synchronization must be based on different principles. The method most commonly used is based on spread spectrum with a large spreading factor. We may, for instance, use a signal at a rate of 1 Mbps containing a unique word as initial synchronization signal and employ code division multiple access (CDMA) to this signal. The spreading code may consist of 100 chips per information bit. This gives a chip rate of 100 megachips per second (Mcps). The CDMA signal then fills the whole bandwidth of the satellite. This gives a coding gain of 20 dB; that is, the spread spectrum signal can be sent at a level 20 dB below the level of ordinary bursts and still be detected with the same quality as an ordinary burst. CDMA is described in Section 5.5.1. Since the signal-to-noise ratio of digital signals is about 15 dB, this also means that the spread spectrum signal is weaker than the total noise (combined thermal noise and interference noise) of the system by 5 dB so that the detection threshold of other signals is reduced by about 10 log101.3 ≈ 1 dB when the spread spectrum signal is sent through the satellite. This will be well within the propagation margins of the system (see Section 9.7). The procedure is shown in Figure 3.21. Initially, the Earth station monitors the received burst pattern from the satellite and determines the start of the TDMA frame (the reference burst r). From this information, the position of the timeslot that

70

Synchronization Satellite

Time

Received bursts r

x

...

r

x

Earth station Position of x New transmission time

Timing error

n TDMA frames

Figure 3.21

Initial synchronization of burst transmission.

is assigned to the Earth station (x in the figure) is calculated. The Earth station sends the 100 Mbps spread spectrum signal in slot x and measures the instant when the signal is received back again from the satellite as shown in the figure. This allows the Earth station to determine the timing offset from the correct transmission instance of burst x as shown. The procedure may be repeated sending the spread spectrum burst at the derived instant in order to reconfirm the synchronization. This procedure may be repeated several times. However, this measurement can only be done with an accuracy of 100 bits of the 100 Mbps signal, since the length if 1 bit of the 1 Mbps signal corresponds to 100 bits at 100 Mbps. To resolve this inaccuracy, the Earth station may, after initial synchronization is acquired, send a short burst at nominal power level and bit rate in the middle of the assigned burst and then do an accurate determination of the start of the burst. The station can then commence sending in the allocated burst slot and resume normal synchronization procedures as described earlier. The synchronization system also enables allocation of dynamic burst length to Earth stations depending on instant traffic demand. As shown in Figure 3.22, this involves several processes that can be performed accurately based on this synchronization method: •

• •

•

If one of the channels in the burst becomes empty, the burst is reorganized such that this channel appears at the rear of the burst. The burst is then shortened by one channel. If another Earth station has the demand for another traffic channel, this and all earlier bursts are moved one channel toward the burst that has been shortened. Add a new channel to this burst.

3.7 Synchronization of Radio Systems

71

Idle

Demand (a)

(b)

(c)

(d)

(e)

Figure 3.22

Dynamic burst-length allocation.

The process is controlled by one of the Earth stations acting as a master station of the system. This complicated principle was implemented and tested as early as 1972. 3.7.5

Application of Scrambling and Interleaving

The information signal (including frame headers) may consist of long sequences where the density of shifts between zeroes and ones is small. The signal may even contain long sequences of a single parity. The problem in such cases is that the signal does not contain enough parity shifts to keep the phase-locked loops synchronized. Such situations cannot be avoided by simply specifying them away. It is better to solve the problem by scrambling the bits of the sequence. The most common scrambling method is based on adding modulo 2 a pseudorandom sequence bit by bit to the duty signal at the sender and adding the same pseudorandom sequence at the receiver. Mathematically this works as follows. Let M represent the duty signal and P the pseudorandom signal. The scrambled signal S then takes the form: S = M ⊕ P, where ⊕ is addition modulo 2. The action at the receiver is to add the pseudorandom signal to the scrambled sequence. This gives S ⊕ P = M ⊕ P ⊕ P = M since P ⊕ P = 0. The receiver must then contain an identical scrambler accurately synchronized to the scrambler of the transmitter.

72

Synchronization

The challenge is to find long pseudorandom sequences. This problem is exactly the same as the one we encountered for error correcting codes (the HEC in the ATM header). The solution to this problem requires extensive use of abstract algebra and is outside the scope of this book. The application of the method is illustrated in Figure 3.23. This is the scrambler used in the SDH. The scrambler consists of a shift register with feedback. The shift register in the example consists of seven flip-flops with feedback after flip-flops number 6 and 7. This is expressed as the polynomial 1 + x6 + x7. The reset circuit resets the shift register at certain instants in order to retain synchronism. Another scrambling method consists of writing the bits of the duty signal into a buffer memory row by row and reading them out again column by column. This method causes a delay of the duty signal equal to the number of rows multiplied by the number of columns. This method is called block interleaving. A complex variation of the basic interleaving method is used in GSM. Even with scrambling and interleaving, long sequences of bits with the same parity may occur, but such events are much less frequent than with unscrambled signals. In real systems, low bit density is often a consequence of the specification such as in ATM, where empty cells contain only 2 binary ones out of 424 bits.

Duty signal

1

2

3

4

5

6

7

Clock Reset Scrambled signal Addition modulo 2 0 ⊕ 0 = 0 0 ⊕ 1 = 1 1 ⊕ 0 = 1 1 ⊕ 1 = 0

Figure 3.23

Scrambler used in SDH.

CHAPTER 4

Multiplexing 4.1

Multiplex Structures The purpose of multiplexing is to utilize the transmission medium (cable, radio relay, fiber, or satellite) efficiently by combining several communication channels into a single channel. The principle is illustrated in Figure 4.1. The multiplexer (+) combines the individual channels into a single information stream; the demultiplexer (−) splits the information stream into individual channels. We distinguish between static and statistical (or dynamic) multiplexing. In static multiplexing, the same communication channel always appears at the same place in the multiplexed information stream (frequency or timeslot). Statistically multiplexed channels are not organized in this regular pattern. In systems where the individual communication channels are circuit switched at the network layer, static multiplexing is used (telephony). The IP offers connectionless network services, and, therefore, does not require that the individual communications channels appear at predetermined points in time. Statistical multiplexing can then be employed to individual datagrams in order to maximize the throughput of the system. There are two basic types of static multiplexing methods: frequency division multiplexing and time division multiplexing. A special form of FDM is wavelength division multiplexing used on optical cables. WDM is described in Chapter 10. Multiplexed signals may be further multiplexed into larger systems in several ways. FDM signals may be frequency multiplexed forming an FDM hierarchy;

Switch

–

+

+

–

Multiplexed channel

Multiplexed channel

Individual channels

Figure 4.1

Multiplexing.

73

74

Multiplexing

TDM signals may be time multiplexed forming a digital hierarchy; and statistically multiplexed channels may be transferred in TDM systems such as SDH.

4.2

Static Multiplexing: Frequency Division Multiplexing 4.2.1

Principle

The oldest multiplexing technology is FDM. This was the only technology available before digitalization. The telephone signals from different telephone apparatuses are translated to different frequencies by the multiplexer and then added linearly. This can be done because the frequency spectra of individual channels do not overlap. Since each individual telephone channel occupies a bandwidth between 0 and 3.4 kHz,1 12 channels will occupy a bandwidth of 48 kHz, allowing a spacing of 600 Hz between individual channels in order to avoid interference. Each channel then occupies a bandwidth of 4 kHz. The frequency of the lowest channel is shifted to about 60 kHz in order to have a safe margin to the 1/f noise2 of the oscillators. The demultiplexer filters out each individual channel from the multiplexed signal and translates it back to a frequency between 0.3 kHz and 3.4 kHz. The multiplexing principle is shown in Figure 4.2. The task is to translate all the baseband signals such that they form a sequence of spectra with the exact frequency spacing B of 4 kHz. 4.2.2

Translation of Channels

We first observe that 2 sin ω 0 t sin ω 1 t = cos( ω 0 − ω 1 )t − cos( ω 0 + ω 1 )t. If we can find a circuit that acts as a multiplication function, we can translate the angular frequency 1 (the baseband signal) by the fixed frequency 0 (the carrier frequency). Since the total spectrum of a signal is a linear function (discrete or continuous) of individual frequency components, the complete spectrum will translate in the same way: 2 sin ω 0 t ΣAi sin ω 1 t = ΣAi cos( ω 0 − ω 1 )t − ΣAi cos( ω 0 + ω 1 )t and similar for the continuous case where the sum is replaced by an integral.3 The individual spectral lines of the translated spectrum are represented by ω 0 ± ω1 . The sum is replaced by an integral for continuous spectra. We also observe that the frequency i gives rise to two spectral lines: one above and one below the carrier frequency.

1.

2.

3.

The spectrum is cut off at 0.3 kHz in order to avoid the condition that very low frequency components must be transferred, thereby avoiding large capacitors in the transmission system. Frequencies below 0.3 kHz and above 3.4 kHz are not significant in speech. This noise component varies as 1/f and is significant for low frequencies. Causes of this noise are nonthermal phenomena such as quantum fluctuations in semiconductors, chaotic currents in active devices, and nonlinear mixing. This is also called the modulation theorem in Fourier analysis: 2 F[g(t) sin ω0 t] = G( s − s0 ) − G( s + s0 ) where F[.] is the Fourier integral operator, s = j , s0 = j 0, and

[

]

G( s ) = F g(t ) is the Fourier transform of g(t). The spectrum of g(t) is thus just shifted by 0. See, for instance, Korn and Korn, Mathematical Handbook for Scientists and Engineers, Section 4.11, Dover Publications, 2000.

4.2 Static Multiplexing: Frequency Division Multiplexing

75

Frequency

(a)

f0

Frequency

f0 + B f0 + 2B (b)

Figure 4.2

(a, b) Principle of frequency multiplexing.

The ring modulator is a circuit that just does this type of translation of signals. The ring modulator is shown in Figure 4.3. The device is essentially the same as the phase modulator of the phase locked loop described in the appendix. The baseband signal and the carrier signal are fed to two of the ports of the ring modulator, and the translated signal is extracted at the third port. The waveform produced by the ring modulator is shown in Figure 4.4 (bold line). This corresponds to a carrier wave amplitude-modulated by the baseband signal. The ring modulator is thus an amplitude modulator reproducing the baseband signal accurately as an amplitude variation of the carrier signal. If one of the input signals are the sine wave sinω0t and the other input signal is the baseband spectrum, the ring modulator produces an exact copy of the baseband signal above the carrier frequency ω0 and one below ω0, as shown in Figure 4.5. The signal below ω0 is inverted. The two copies of the spectrum contain exactly the same information, and one of them can be deleted by using a high-pass filter that keeps the upper copy and removes the lower copy. This type of modulation is also called single sideband (SSB) amplitude modulation. Note that there is no spectral line at ω0. This means that after the lower copy has been removed, there is no information in the signal that can be used to extract ω0 at the receiver. This problem can be overcome by placing a pilot tone at a known frequency in the multiplex structure. This pilot is used to determine the fixed

Multiplexing

Translated signal

76

D1

Baseband signal

D1’

D2

D2’

Carrier frequency

Figure 4.3

Ring modulator.

translation frequency ω0 and the distance between adjacent channels. At the receiver, ω0 and the channel distance B are found by using phase-locked loops tuned to the pilot. The same arrangement can be used to form higher multiplex hierarchies. The 48-kHz system (12 telephone channels) is called the primary multiplex level or the first-order system. First-order systems are then multiplexed to form second-order systems (60 telephone channels) and so on up to 3,600 telephone channels. The multiplex method explained previously is not only used for analog telephone channels. Television channels are multiplexed on analog cable systems using FDM in a similar way. FDM is also used in ADSL as described later.

Baseband signal

Frequency

Carrier signal

Figure 4.4

Output signal from the ring modulator.

4.2 Static Multiplexing: Frequency Division Multiplexing

Figure 4.5

4.2.3

77

Placing the channel in the spectrum.

Multiplexers and Demultiplexers

The multiplexer is shown in Figure 4.6. The translation frequencies f 0 , f 0 + B, f 0 + 2B, and so on, ( f 0 = ω 0 / 2π ) are generated by a frequency synthesizer and fed to the ring modulators together with each individual baseband signal. The translated signals are then filtered by highpass filters and combined in an amplifier circuit. The pilot signal is added at the combiner amplifier. The demodulator is shown in Figure 4.7. A phase-locked loop produces one or more reference frequencies from the pilot frequency of the multiplexed signal. The reference frequencies are used by the frequency synthesizer to produce translation frequencies. The translated signal is reproduced by a ring modulator that simply subtracts the translation frequency from the translated signal utilizing the fact that 2 sin( ω 0 + ωi )t sin ω 0 t = cos ωit + cos(2ω 0 + ωi ), where the latter signals is removed by the lowpass filter. The lowpass filter also removes all other frequency components beside the baseband spectrum of the wanted signal. On long-distance cable systems, repeaters are inserted at even distances along the cable in order to compensate for the signal attenuation along the cable. The repeaters amplify the FDM signal to its nominal level before the signal is forwarded on the cable. Since the spectrum of the multiplexed signal does not contain the carrier frequencies of the individual channels, it is not possible to derive an absolute signal level reference from the multiplexed signal alone. The pilot signal inserted by the transmitter is therefore used to determine the amplification required by the repeater. This signal has a well-defined level. The previous FDM systems on the coaxial cables across the Atlantic Ocean (6,000 km) had an overall amplification of 16,000 dB provided by approximately 300 repeaters along the cable. In order to keep the level variation of the received signal on the other side of the ocean within

78

Multiplexing

Baseband signal

Ring modulator

Filter

Ring modulator

Filter

Multiplexed signal Combiner

· · · Ring modulator f0

Filter

f0 + nB ·

·

·

f0 + B

Pilot generator Frequency synthesizer

Figure 4.6

Frequency division multiplexer.

an accuracy of 1 dB, each repeater along the cable was equipped with AGC using the level of the pilot tone as reference. Frequency division multiplexing is used on optical fibers. This type of FDM is called wavelength multiplexing to distinguish it from ordinary frequency multiplexing. WDM is described in Chapter 10. 4.2.4

Distortion in FDM and WDM Systems: Intermodulation

A channel in the FDM signal can be represented as Ai exp j( ω i t + θ i ) in complex notation,4 where Ai is the amplitude of the signal in channel i, i the translation frequency of that signal, and i is the baseband signal represented as a phase angle. The n

n

1

1

product of several such signals is then ∏ Ai exp j ∑ ( ω i t + θ i ). This formula enables us to establish a simple expression for the effects nonlinearities have on the signal, 2 3 since the nonlinearity can be expresses as a series a0 + a1x + a2x + a3x + …, where the second- and higher-order terms represent the nonlinearity. The multiplexed signal causes interference on itself if the nonlinearities produce a signal that falls into the bandwidth of the multiplexed system; that is, producing frequencies of the form ωl ≤

k

∑± ω

i

≤ ω n , where k is the number of carrier signals that interfere,

l

is the

1

4.

The signal can then be represented as a sine wave by taking the imaginary part of this expression: Ai sin( ωit + θ i ) = Im Ai exp j( ωit + θ i ) . We have used the notation exp x = ex for simplic-

[

ity of writing.

]

4.2 Static Multiplexing: Frequency Division Multiplexing

79

Translated signal

Modulated signal

Ring modulator

Lowpass filter

Ring modulator

Lowpass filter

Baseband signal

Splitter · · · Ring modulator f0 + B

f0 + nB ·

·

f0

·

Pilot filter

Lowpass filter

Pilot Phaselocked loop

Frequency synthesizer Reference frequency

Figure 4.7

Demultiplexer.

carrier frequency of the lowest channel, and n is the carrier frequency of the highest channel. This type of self-interference is called intermodulation. The sign of each term in the sum is chosen such that the condition is satisfied. It is easy to see that k must be an odd number in order to satisfy the condition; that is, even-order nonlinearities do not contribute to the intermodulation. The lowest-order intermodulation signal is then of the third order (k = 3) produced by the term a3x3. In most systems, this is the strongest intermodulation signal. If three signals are passed through a nonlinear device, third-order intermodulation components of the form

[(

)

]

B ijk (t ) exp j ω i ± ω j ± ω k t + F ijk (t )

appear, where ijk and Fijk are slowly varying real functions of time representing the effects the nonlinearity have on the amplitude and the phase, respectively, of the three signals i, j, and k. These components are caused by the third-order term in the series expansion. If i ± j ± k is within the frequency range of the FDM system, this signal component causes interference to channels in the multiplexed band. This may happen if two of the signs are positive and the third is negative; that is, i + j − k and the resulting frequency is within the frequency range of the FDM signal as shown previously. If we set i = j = k, we see that the previous formula also can be used to investigate the effect of the third-order nonlinearity on the signal itself (i + i − i = i) (distortion). The case i = j ≠ k is also included. This case corresponds to the case where the frequency of one signal is doubled and then mixed with a second signal.

80

Multiplexing

Note also that the more channels that are multiplexed, the more severe the effect of higher-order nonlinearities (fifth, seventh, and so on) is. This is not evident from the theory. 4.2.5

Frequency Division Multiplexing in ADSL

The principle of frequency division multiplexing of asymmetric digital subscriber line is shown in Figure 4.8. The transmission takes place on twisted-pair telephone lines. On the twisted-pair telephone line, signals may be sent both ways simultaneously. The twisted pair is, for this reason, called a balanced transmission line. The network and the terminals support only unbalanced operation; that is, it contains only unbalanced two-wire circuits, in which one conductor is the Earth, for each direction of transmission as shown in the figure. A transformer is used for conversion between the balanced and unbalanced modes of operation (such a device is called a balun). A perfect balun does not reflect any signal (zero coefficient of reflection). A practical balun reflects a small part of the signal. The reflection problem can be avoided in two ways. An echo canceller can be inserted at each balun, suppressing the echo reflected back from the other end of the two-wire circuit. The signals in the two directions of propagation can then occupy the same frequency band. Alternatively, the uplink and downlink signals are frequency division multiplexed occupying different frequency bands. This is the case shown in the figure. The uplink band occupies the band from 25 kHz to 200 kHz, while the downlink occupies the band from approximately 250 kHz to 1,100 kHz. The twisted pair may then also carry a normal analog two-way telephone channel at frequencies below 4 kHz in addition to the broadband digital signals. The

Analog telephone channel

Upstream (and downstream)

4

25

Downstream

200

1,100 Frequency (kHz)

TX

RX

RX

TX Balun

Figure 4.8

Frequency division multiplexed ADSL.

4.2 Static Multiplexing: Frequency Division Multiplexing

81

echo will not affect the analog audio signal significantly provided that the echo delay is small (less than about 30 ms). ADSL may also use a frequency division multiplexing method called direct multitone (DTM). In this case, the digital signal is divided into several parallel bit streams with bit rates, say, between 0 and 60 Kbps. The method is shown in Figure 4.9. The loss along the cable depends on the frequency as shown in Figure 4.9(a), resulting in a bit error rate for the individual channels as shown in Figure 4.9(b) if error correction is not provided. Applying error correction, the BER characteristic may be improved as shown in Figure 4.9(c), providing an acceptable quality of the signals in all the individual channels. This is achieved as follows. The input signal is first split into bit streams of different data rates at the transmitter. Error correction is then provided to each bit stream in order to bring them up to a common bit rate. More error correction bits are inserted in the bit streams that are sent in the upper part of the spectrum than in the bit streams to be sent at lower frequencies. The resulting bit streams are then frequency multiplexed before they are presented to the cable (or twisted pair). At the receiver, the inverse operation takes place, as shown in the figure. Since the error correction is strongest for the bit streams that are attenuated the most by the cable, the resulting BER is equalized as shown in Figure 4.9(c). By clever allocation of bit rates and error correction methods on the different FDM channels, it is possible to increase the downstream bite rate by as much as a factor of 10 to about 10 Mbps on short subscriber lines.

BER = 1

BER = 1

Received signal

BER Noise f

f

(a)

Error coding Error coding

Error corr FDM mux

Cable

Error coding

(a–d) Direct multitone ADSL.

Error corr

Error corr (d)

Figure 4.9

FDM demux

…

…

Split

f

(c)

(b)

Combine

82

4.3

Multiplexing

Static Multiplexing: Time Division Multiplexing 4.3.1

Principle

Figure 4.10 is a simple example illustrating how four channels are time-multiplexed. The 64-Kbps signals are read into separate buffers (basic channel in the figure). The signals are then sequentially read out from the buffer at the multiplexed rate (in the example, 2 Mbps) (speed up) and inserted at the correct position in the output stream (channel mapping). This process requires only buffers and simple combinations of logical gates. 4.3.2

Interleaving Patterns

The bits from the different channels must be organized relative to one another. This is called interleaving. Interleaving is done principally in two ways: word interleaving or bit interleaving. The two interleaving methods are explained for PCM-encoded speech in order to make the example concrete. A PCM-encoded speech signal consists of a contiguous series of blocks of 8 bits, where each such block represents a single speech sample. The block of 8 bits is usually referred to as a word. A first-order TDM signal is composed of 32 channels encoded in this way. Interleaving has to do with how bits from the different channels are placed relative to one another within the repetitive pattern of 32 channels.

Multiplexer 64 kbps

2 Mbps Transmission system

Terminal

Channel Frame Terminal

Terminal 2 Mbps Multiplexed signal

Terminal Frame Basic channel

Figure 4.10

Time division multiplexing.

Speed-up

Channel mapping

4.3 Static Multiplexing: Time Division Multiplexing

83

In a word-interleaved multiplexer, the words from each channel are inserted successively as shown in Figure 4.11(a). The collection of all 32 words is called a frame. The channel may be bit-interleaved as shown in Figure 4.11(b). Now all the first bits of all 32 channels make up a frame, and similarly for the second bits, third bits, and so on. The complete multiplex structure consisting of 8 frames is called a superframe. Note that bit-interleaving is the same as word-interleaving if the word consists of just one bit. The notation may be confusing: a frame in a word-interleaved first-order multiplexer consists of 256 bits, while the frame of a bit-interleaved multiplexer consists of 32 bits. A bit-interleaved superframe contains exactly the same information as a word-interleaved frame, but since bit stream is designed for speech, any signal (e.g., TCP/IP signal) with a data rate less than or equal to 64 Kbps can be adapted to this rate. This is just a question of organizing the information such that the receiver is able to recover it. Space that is not used is filled with dummy bits. 4.3.3

European First-Order Multiplex

This is the system consisting of 32 channels mentioned previously. We shall now see how 64-Kbps voice channels are organized into a first-order multiplex structure. In the European multiplex hierarchy, 30 speech channels plus two control channels (channel number 0 and channel number 16)—that is, a total of 32 channels—are multiplexed forming a new bit stream at of 32 × 64 Kbps = 2,048 Kbps. (The data rate 2,048 Kbps is usually referred to as “2 megabit per second,” or 2 Mbps, for brevity.) Word interleaving is used so that a frame consists of all 32 channels. The control channels are used for purposes such as providing a unique frame-synchronization pattern, in-band signaling, and link management. In order to provide enough space for signaling information, the frames are organized in a

Frame Word

0a

0b

...

0h 1a

1b

...

........

1h

30a 30b

...

30h 31a 31b

...

31h 0a

Channel no. (a)

Bit no.

Super-frame Frame

0a

1a

...

Word

31a 0b 1b

...

31b

........

Channel no. Bit no.

Figure 4.11

(a, b) Interleaving patterns.

(b)

0g 1g

...

31g 0h 1h

...

31h 0a

84

Multiplexing

superframe consisting of 16 frames. The superframe is thus the basic repetitive format of the first-order multiplex. The arrangement is shown in Figure 4.12. The multiplex structure consists of one superframe containing 16 frames. Each frame consists of 32 channels (or words) containing 8 bits. The first channel (channel number 0) of every second frame (frames number 0, 2, 4,...14) in the superframe contains 7 bits that are used for frame synchronization. The bit pattern of channel number 0 is R0011011, where R takes an arbitrary value and may be used for other purposes. Channel number 0 of every odd frame (frames number 1, 3,…15) is reserved for national use but must contain a 1 in the second bit position in order to avoid this channel containing a simulated frame synchronization pattern. In order to provide synchronization of the superframe, channel 16 of the first frame in the superframe (frame number 0) contains a unique 4-bit superframe synchronization pattern in the first 4 bits of the channel. The bit pattern is 0000. The remaining four bits of the word is reserved for other purposes. Channel 16 of the next 15 frames (frame numbers 1 through 15) of the superframe contains two in-band signaling channels consisting of 4 bits each. One in-band signaling channel is allocated to each voice channel. Not all the bits of the signaling channels are zero in order to avoid simulation of the superframe synchronization word. The basic system consists of 30 + 2 channels at the rate of 64 Kbps. However, any number of 64-Kbps signals may be combined to form new channels—for example, a 2-Mbps system containing 5 channels at rate 192 Kbps (combination of three 64 Kbps channels) and 15 channels at rate 64 Kbps keeping frames 1 and 16 for synchronization and administration purposes. The entire capacity of 2,048 Kbps can also be assigned to a single user either as carrier of a broadband service or as an unspecified channel. In these cases, the frame synchronization as described previously need not be implemented: the channel may be configured in whichever way the user decides or subject to a standard for broadband services.

Superframe 0

1

2

3

4

5

Frame 0 ... 1 1 1 ... 0 1 2 5 6 7

R 0011011

0 0 0 0 Adm

Frame synch 7 bits

Superframe synch

Word 8 bits

Figure 4.12

Multiplex format.

6

7

8

9

10

Frame 1 3 3 ... 1 1 1 ... 0 1 2 3 0 1 5 6 7

R1

Adm

Administration

Chan Chan 1 17 In-band signaling

11

12

3 3 0 0 1

13

14

15

0

4.3 Static Multiplexing: Time Division Multiplexing

4.3.4

85

Higher-Order Multiplex

If four 2-Mbps signals are multiplexed, we arrive at the second level in the European digital multiplex hierarchy (bit rate of 8 Mbps). Four such channels are again combined to form the third level of the hierarchy (34 Mbps). Four channels at a rate of 34 Mbps are combined to form the fourth level (140 Mbps), and, finally, four such channels are multiplexed to form the fifth level (565 Mbps). For applications in optical systems, four fifth-order channels may be multiplexed to form the sixth level of the hierarchy at 2,488.320 Mbps. The European multiplex hierarchy is shown in Table 4.1. The American hierarchy is shown in Table 4.2. The first-order system contains 24 channels at a rate of 64 Kbps. The existence of several multiplex hierarchies (and even different encoding types of PCM) requires that at certain national borders the multiplex structure must be converted. This led to the development of the SDH described in the next section. The 64-Kbps channel does not contain additional bits for synchronization. Instead, particular 64-Kbps channels are used for this purpose (channels 0 and 16) in order to maintain synchronization in the first-order systems as explained earlier. The same procedure is not feasible for higher-order systems. In these systems, bits are added for the purpose of frame synchronization, bit rate adjustment of plesiochronous signals, and management of the links. Since 8,448 Kbps − 4 × 2,048 Kbps = 256 Kbps, we see that the bit stream of second-order systems contains an ample amount of extra bits that can be used for these purposes. The same applies for all higher levels of the hierarchy as shown in Table 4.3. The multiplex hierarchy allows interconnection of plesiochronous signals at all levels except at the first. For this reason the hierarchy is called the plesiochronous digital hierarchy (PDH). Multiplexing of plesiochronous signals is illustrated for the

Table 4.1

European Digital Multiplex Hierarchy

Multiplex Multiplication Order Factor

Data Rate (Kbps)

Number of 64-Kbps Channels

1

32

2,048

32

2

4

8,448

128

3

4

34,368

512

4

4

139,246

2,048

5

4

565,148

8,192

Table 4.2

The American Hierarchy

Multiplex Order

Multiplication Factor

Data Rate (Kbps)

Number of 64-Kbps Channels

1

24

1,544

24

2

4

6,312

96

3a)

7

44,736

672

3b)

5

32,064

480

4

3 × 3b)

97,728

1,440

86

Multiplexing Table 4.3 Additional Capacity of the European Hierarchy Order of System

Additional Capacity (Kbps)

1

0

2

256

3

576

4

1,792

5

8,092

second-order European system (8,448 Kbps). In this system, four channels at a nominal bit rate of 2,048 Kbps are combined. 4.3.5

TDM Frame Alignment in Higher-Order Systems

The four channels are bit interleaved. The basic frame consists of 848 bits divided into four subframes of 212 bits each, as shown in Figure 4.13. This subdivision is done in order to organize the channel in a suitable format. Note that the system could have been organized in several other ways—this is just the choice made by the people who standardized the system. The frame starts with a frame synchronization word consisting of 10 bits as shown. Note that the arrow of time points to the left as indicated in the figure. The alarm bit (A) is used to return alarms to the sender if the received frame contains coding errors or is garbled, the frame synchronization is lost, or the transmission system malfunctions in other ways. The presence of bits reserved for national use (the R bit) is common in most systems, allowing operators to include functions that are strictly not required in the international system.

212-bit subframe 4 × 50-bit-interleaved information bits

Frame synchronization 1

1

1

1

0

1

0

0

0

0

First bit

A

R 4 × 52-bit-interleaved information bits

C1 C2 C3 C4 4 × 52-bit-interleaved information bits C1 C2 C3 C4 4 × 51-bit-interleaved information bits C1 C2 C3 C4 F1 F2 F3 F4 A R C F

Figure 4.13

= = = =

alarm channel reserved for national use control bit; three bits for each 2-Mbps channel bit for rate adjustment; one bit for each 2-Mbps channel

Second-order European system.

Last bit

4.4 Static Multiplexing: Synchronous Digital Hierarchy

87

If the bit rate of one channel is larger than the nominal bit rate, the adjustment bit assigned to that channel is used to transfer one additional information bit. The rate can thus be increased by a factor 1 + 1/848 = 1.0012, corresponding to a clock stability of 12 parts in 10,000. This stability includes both short-term and long-term fluctuations. Note that the nominal bit rate of the system must equal the lowest rate that can be expected, since rate adaptation takes place in only one direction (toward increased rate). The three control bits (C) assigned to each channel indicate whether or not the adjustment bit contains information. The coding of these bits is as follows. If the bits are encoded as 000, the adjustment bit contains information; if the bits are encoded as 111, the adjustment bit is empty. Simple error correction based on majority decision is applied: the receiver assumes that the adjustment bit contains information if the control bits consist of two zeroes and that the adjustment bits are assumed to be empty if at least two ones are detected. Note that the number of bits sent per frame per first-order system is 205. This number is not an even number of octets, so there is no synchronization relationship between the frame pattern of the second-order system and the four first-order systems being multiplexed. The second-order system just treats the four 2-Mbps channels as unstructured bit streams. Exactly the same applies to the higher orders of the multiplex hierarchy.

4.4

Static Multiplexing: Synchronous Digital Hierarchy 4.4.1

Background

The American standard for optical networks, synchronous optical network (SONET), was initially developed by the major telecommunications operators in the United States —AT&T and the Regional Bell Operating Companies (RBOCs)—and endorsed by the American National Standards Institute (ANSI) as an American telecommunications standard. The development of SDH was one part of the development of SONET. Later, the standardization work was taken over by the International Telecommunication Union (ITU), also making the standard applicable to European and Japanese multiplex standards. The European Telecommunications Standards Institute (ETSI) is in charge of coordinating the European applications of the standard. For this reason the standard is extremely complex, supporting a large number of options. Since several organizations have been working on the standard, it is not entirely consistent, and definitions and terminology may vary in different versions of the standard valid in various parts of the world. Several purposes of SDH are as follows: • •

• •

Provision of a standard suitable for optical networks. Provision of a system that can combine any standard data rate of the PDH hierarchies. Provide an effective interface for ATM systems. Offer “add-drop” capabilities; that is, the capability of inserting or removing individual channels of any multiplex level without demultiplexing the entire

88

Multiplexing

•

•

signal. This requires that the system must be strictly synchronous. Otherwise, it will not be possible to access the bit stream directly and extract just part of the signal. Offer plesiochronous signals to float within the synchronous bit stream and still be capable of identifying exactly where the signal is located relative to the bit stream at any time. Offer improved operation and maintenance procedures.

4.4.2

Multiplexing Structure

Figure 4.14 is an example of the multiplexing capabilities of SDH. The system also supports the Japanese hierarchy, directly generated ATM signals, and the European second-order system at 8 Mbps. These levels have not been included for the sake of readability of the figure. The figure nevertheless demonstrates the enormous flexibility offered by the system. The functions supported by the different units are described in Table 4.4. These are as follows: • •

•

The container (C) is the payload itself. The virtual container (VC) encapsulates the payload. The encapsulation consists of a header (the path overhead). The tributary unit (TU) contains pointers by which the position of the payload in the multiplex structure is uniquely identified. This entails insertion of stuffing bits in order to bring the input signal to a specified bit rate. The tributary unit also allows the tributary unit group to multiplex plesiochronous channels.

×N STM-N

×1 AUG

AU-4

OR

VC-4

C-4

OR ×3

×3

TUG-3 ×1 AU-3

VC-3

OR means that either alternative may be used but not both at the same time. The number on the arrows is the number of channels of that type that is multiplexed within a tributary group or virtual container.

Multiplexing structure.

×1 TU-3

OR

VC-3

×1 44/34/32

×7 OR

C-3

OR ×7

Figure 4.14

Bit rates in Mbps 140/98

×1

×1 TUG-2

TU-2

OR

VC-2

×3 TU-12

×1

×1 VC-12

6.3 C-2 2 C-12

×4 TU-11

×1 VC-11

1.5 C-11

4.4 Static Multiplexing: Synchronous Digital Hierarchy Table 4.4

89

Structure Elements

Structure Element Name

n

Use

C-n

Container

11, 12, 2, 3, 4

Contains the payload (basic channel of PDH hierarchy)

VC-n

Virtual container

11, 12, 2

Contains containers 11, 12, or 2 plus the path overhead for the C it contains

VC-n

Virtual container

3

Contains path overhead plus either one C-3 or one TU-3 or seven TUG-2

VC-n

Virtual container

4

Contains path overhead plus either one C-4 or three TUG-3

TU-n

Tributary unit

11, 12, 2, 3

Contains VC-n plus tributary unit pointer indicating the start of the VC-n

TUG-2

Tributary unit group

Multiplexer containing either one TU-2 or three TU-12 or four TU-11

TUG-3

Tributary unit group

Multiplexer containing either one TU-3 or 7 TUG-2

AU-n

Administrative unit

AUG

Administrative unit group

STM-n

Synchronous transport module •

• •

3, 4

Contains VC-3 or VC-4 plus administrative unit pointers indicating location of the VCs Multiplexes either one AU-4 or three AU-3

1, 4, 16…

Multiplexes n AUGs and inserts section and path overheads (see also Table 4.5)

The administrative unit (AU) is in charge of similar functions to the tributary unit (pointers and bit stuffing). The administrative unit takes care of the uppermost level of the multiplex hierarchy. The administrative unit group (AUG) is a multiplexer. The synchronous transport module (STM-n) multiplexes the signal further into the optical system. See the following for values of n.

The tributary unit group (TUG) multiplexes tributary units before they are presented to the next level of the hierarchy. This includes the addition of headers. The format of all STM-n is based on the same principle. STS consists of synchronous transmission signals of order n (STS-n) used in the United States. The frequencies in the STS hierarchy is the nth multiple of the bit rate 51.840 Mbps; the European hierarchy (STM) consists of multiples of 155.520 Mbps. Table 4.5 shows the STS levels and the corresponding STM levels used in Europe. 4.4.3

Compromise: Large Overhead Versus Flexibility

We see from Figure 4.14 that the multiplexing makes wasteful use of the available bit rate in the system. Three channels at a rate of 32.064 Mbps may be multiplexed into an STM-1, producing an overhead of 59.3384 Mbps corresponding to 39% of the overall capacity. All together the STM-1 at a rate of 155.520 Mbps may contain 3 × 7 × 3 = 63 first-order (2 Mbps) channels, giving an overhead of 26.496 Mbps (17%). Some, but not all, of this overhead is used for other purposes, such as synchronization and operation and maintenance. This amounts only to about 5% of the overall capacity. The remaining overhead consists just of stuffing bits.

90

Multiplexing Table 4.5 Comparison of STS and STM Levels STS-n

STM-n

Bit Rate(Mbps)

1

155.520

1 3

51.840

9 12

466.560 4

622.080

18

933.120

24

1244.160

36 48

1866.230 16

2488.320

64

9953.280

96 192

4876.640

One advantage of SDH is that all possible multiplex rates can be supported by a single system. In optical systems where capacity is not a big concern, flexibility is much more valuable than efficient utilization of transmission capacity. Another advantage of SDH is that the configuration of the system can be instantiated and changed by software commands only. This means that the internal multiplex structure of the SDH link, including the add-drop capability, can be reconfigured remotely without replacing hardware. 4.4.4

Pointer Mechanisms and Floating Payloads

The composition and operation of the administrative units and the tributary units are based on the same principles, though the solutions specified for various units are different. The SDH specification is full of such apparent inconsistencies. The reasons are, first, that the system has been developed by different standards organizations with a large degree of autonomy over several years, and, second, that the system shall combine a large number of multiplex rates and support various operational requirements that are not the same in different countries. The operation of the STM-1 containing one AU-4 frame is described later in order to illustrate the basic principles of SDH. The TUs and the STM-1 containing three AU-3 frames are organized in similar ways, though certain details are different. The three AU-3 frames are octet-interleaved in order to fit into the same space as the AU-4 frame. Therefore the basic structure of the two alternatives is the same. The AU-4 of STM-1 contains the virtual container VC-4. All virtual containers consist of two parts: •

•

Operation, administration, and maintenance (OA&M) information is contained in a field called path overhead. This information follows the container until it is removed from the SDH system. The payload can either be a single PDH multiplex signal (container C) or a number of tributary unit groups (TUG) multiplexed in accordance with

4.4 Static Multiplexing: Synchronous Digital Hierarchy

91

principles defined for the VC. The TUGs consist of one or more tributary units multiplexed in accordance with rules that apply for that TUG. The TUs contain pointers and other overhead information plus a virtual container as payload. The TU pointer identifies the beginning and the end of the VC frames. This together with Figure 4.14 illustrates the iterative structure of SDH. This structure then contains a set of linked pointers that finally identifies a single container uniquely. The STM-1 envelope is shown in Figure 4.15. The structure of the envelope is drawn in the form of a matrix consisting of 270 columns of octets and nine rows. The envelope contains 9 × 270 = 2,430 octets. The transmission of bits is as follows. The first bit to be sent is the leftmost bit in row number 1. Then follows bit number two in this row and so on until the last bit of row number 1 has been sent. The next bit is then the leftmost bit in row number 2 and when the final bit of row number 2 has been sent, the next bit is the leftmost bit of row number 3 and so on. The final bit is the rightmost bit in row number 9. The only magic about this arrangement is that it is easy to read, easy to understand, and easy to specify. The STM-1 envelope contains the section overhead, the link overhead, and the AU-4 pointer field. The pointer field is in row number 4. If necessary, the AU adds stuffing bits in order to make the payload plus the path overhead consist of 9 × 261 = 2,349 octets. The STM-1 envelope contains the section overhead (3 × 9 octets = 27 octets), the link overhead (5 × 9 octets = 45 octets), the pointer field, and the actual payload (C-4 or three TUG-3 plus the path overhead). The AU-4 pointer consists of 10 bits in row 4. The value of the pointer multiplied by 3 gives the start position of the VC 1 column First bit 261 columns

9 columns

2

Section overhead

3 4 9 rows

Pointer

5

Path overhead

1

Payload

6 7

Line overhead

8 9 Virtual container Last bit

Figure 4.15

Structure of STM-1.

92

Multiplexing

within the envelope. This gives a maximum value of the number of pointer addresses 10 in the VC: 3 × 2 − 1 = 3,071, where 10 is the number of address bits in the pointer. This is larger than the actual maximum length of the VC frame, which is 9 × 261 = 2,349 octets. The pointer field is shown in Figure 4.16. The field consists of nine octets divided into three groups, H1, H2, and H3, where each group consists of identical formats. One H1 plus one H2 contain together the pointer address as shown in the figure, while H3 is used for rate adjustment. If the envelope contains one AU-4, then the first H1 and H2 octets are used as pointers, while all three H3 octets are used for rate adjustment. If the envelope contains tree AU-3, all three sets of H1, H2, and H3 are used in order to assign individual pointers and adjustment fields to each AU-3. The combination of the H1 and H2 octets consists of three parts as shown: •

•

•

The new data flag consists of four bits that are normally encoded as 1001, indicating that the address in this envelope is the same as in the previous one. The value of the new data flag is the same, even if rate adjustment takes place. If a new pointer address is set, the new data flag is encoded as 0110 in the envelope in which the pointer address is changed. Such operations take place when instantiating the system, when the multiplex configuration is changed, and after a disruption of communications. The SS bits indicate whether the envelope contains one AU-4 or one, two, or three AU-3. This thus represents the multiplexing function of STM-1. The actual address of the pointer position is contained in the I-bits and D-bits. If the pointer value has to be increased because of rate adjustment, the I-bits are inverted in the envelope where the adjustment takes place. The pointer value is then increased by one in the next envelope but with the I-bits in the noninverted form. The operation of the D-bit is similar and is used in order to decrease the pointer value by 1. Rate adjustment is explained next.

H1

N

N

N

N

H1

S

S

H1

I

H2

D

I

H2

D

H2

I

D

H3

I

D

H3

I

H3

D

Pointer address (10 bits) N = new data flag (0110 same address, 1001 new address) S = type of AU (AU-4 or AU-3) I = increment bit (normal: no adjustment; inverted: increment address by one) D = decrement bit (normal: no adjustment; inverted: decrement address by one) Figure 4.16

Pointer field.

4.4 Static Multiplexing: Synchronous Digital Hierarchy

93

Observe that a single bit error anywhere in the pointer field can be corrected, since majority decision is used on the new data flag and the value of the SS bits and the address bits are changing slowly in normal operation. The start of the virtual container cannot be earlier than the first bit following the pointer. This implies that one VC-4 will be contained in two envelopes, as shown in Figure 4.17. This allows the VC-4 to “float” within the envelope structure. The pointer that is within the VC-4 is used for rate adaptation of that VC-4. 4.4.5

Rate Adjustment of Plesiochronous Signals

The floating mechanism also allows bit rate adjustment. The adjustment must take place in groups of three octets because of the way in which the pointers are constructed, as explained previously. The principle for bit adjustment is as follows. The input signal is buffered in an elastic store. Since an adjustment takes place in steps of three octets (24 bits), the size of the elastic store must be at least 2 × 24 = 48 bits in order to support adjustment in both directions. If the input rate of the VC-4 signal is slower than the SDH rate, the number of bits in the buffer will be slowly reduced, since the buffer is not filled up as fast as it is emptied. If the buffer has become 3 octets (24 bits) shorter, the envelope that is sent by the STM-1 envelop is made 24 bits longer than normal by letting the first 3 octets after the header be empty. The envelope contains a normal VC-4 frame of 2,349 octets in addition to the 24 dummy bits. This operation is indicated in the pointer

270 columns 1 End of virtual container VC-4 Pointer

Next frame

9 rows

4

Virtual container VC-4 9 1

4

Pointer

Next frame

This pointer is used for rate adjustment of this VC-4. Beginning of virtual container VC-4 9

Figure 4.17

Floating VC-4.

94

Multiplexing

field by inverting the I bits as explained previously. Now 24 bits more than normal are sent during the interval between two pointer values. In other words, the time interval between the pointers is made 24 bit durations longer. During this time the buffer will be filled up again to its nominal value of 24 bits. In the next envelope, the I bits are reverted to normal and the pointer value is incremented by one. If the input rate is larger than the SDH rate, the buffer is filling up faster than it is emptied. When the buffer has become three octets longer than nominal, a VC-4 frame is inserted in an envelope that is 3 octets shorter than nominal but where 3 octets of the frame are inserted in the three H3 octets of the header. The number of information octets in the frame is still 2,349 octets, but they are sent in a time interval that is 24 bit durations shorter than normal. The D bits are inverted in the pointer of the same header that contains the adjustment bits indicating that rate adjustment is taking place. This operation reduces the buffer length by 24 bits. The pointer value is now decremented by one. 4.4.6

Control Headers

Figure 4.15 shows that there are three types of control headers: section overhead, line overhead, and path overhead. The purposes of these headers are as follows: •

•

•

4.5

Section overhead. The first 2 octets of the section overhead is a synchronization word that determines the start of the STM envelope. STMs can be multiplexed into a hierarchical structure as shown in Table 4.5. One octet in the section overhead is used to indicate the location of the STM (or actually the STS; see Table 4.5) in a higher-order STS/STM-n. One octet is also used for error detection of the complete envelope. The section overhead also contains fields for supporting data channels for operation and management, alarms, and control signaling. Line overhead. The line overhead contains a separate field for management of protection switching. Protection switching means switchover between units in redundant systems, where redundancy means that one system is a backup for one or more other systems in case there is a failure in one of them. Such arrangements are used to improve the availability of the system. In case of a duplicated system, the same mechanism may support load-sharing between the systems as well as switchover to single channel operation in case one of the channels fails. The line overhead also contains a data channel for operation and management. Path overhead. This header stays with the payload until it is finally demultiplexed. The header contains alarm and monitoring information, and may contain specific information concerning how the payload is to be treated in the demultiplexer.

Statistical Multiplexing This is the prevailing multiplexing method in data networks. IP networks and ATM networks employ this multiplexing method. Statistically multiplexed signals are

4.5 Statistical Multiplexing

95

usually inserted in larger transport system using static multiplexing. This method is commonly used in data networks, where the data messages are first multiplexed statistically on, say, one 2-Mbps channel. Several such channels are then combined using time division multiplexing in the usual way. The statistically multiplexed signal may also use a higher multiplex level of the PCM hierarchy. ATM uses fourthorder and fifth-order systems directly. ATM may also be multiplexed on third-order systems allowing a bit rate of 34 Mbps. The general principle is as follows (see Figure 4.18). Information arrives at the multiplexer on several input channels in the format of data packets (or frames or cells). The received packets are placed in a queue before they are retransmitted. The queuing algorithm may be the commonly used first-in-first-out (FIFO) queuing discipline, or the queuing algorithm may be FIFO combined with priority. Packets with higher priority will then always be served before packets with lower priority; the algorithm for packets with equal priority is FIFO. Priority queuing is required in IP systems offering both data transfer and speech. If the queue is empty, dummy information is sent on the transmission system. One important observation is that it is straightforward to combine statistical multiplexing with switching. This principle is utilized in ATM and IP networks. Switching may be done, for example, by providing one queue per output circuit. The system then requires a management function that directs the input message to the right output queue depending on address information. The switching is based either on address information contained in each input message or on separate procedures establishing the path before information is sent (X.25 and ATM). There are principally three ways in which statistically multiplexed signals can be organized. 4.5.1

Invariant Frame Structure

The frame structure may be fixed, consisting of a synchronization header and an information field of fixed length. If no information is sent, empty frames are sent. These frames are identical to information frames except that a parameter in the header is used to indicate that the frame does not contain information.

Individual channels Data storage

Transmission system

Queue Channel formatting

FIFO or priority

Figure 4.18

Statistical multiplexer.

Provide next message

96

Multiplexing

This principle is used in ATM where each frame (or cell) consists of a header containing 5 octets and a payload field of 48 octets, as shown in Figure 4.19. The payload of individual cells may contain different information channels. The layer above ATM keeps track of which cells contain information belonging to the same and different communication sessions. In some applications, forward signaling is used to set up a connection between terminals in the same way as in the telephone network or ISDN. ATM provides a unique numbering of cells in order to support this routing function. ATM supports information streams as shown in the upper part of the figure where cells (or frames) originating from different sources can be statistically multiplexed. The space between cells containing information is, if necessary, filled with empty cells (E). The cell header contains the following parameters. The identifier field contains information that is used by the switching equipment to route the call through the network. The payload type is used to indicate whether the cell contains user-to-user information or network management information. CLP is the cell loss priority indicator. If CLP = 0, the cell must, if at all possible, not be discarded if the network is congested. If CLP = 1, the cell may be discarded in case of congestion. The header error correction field serves two purposes: correcting bit errors appearing in the first 4 octets of the header in order to avoid misrouting of cells (this operation is called the error correction mode), and supporting synchronization of cells (this operation is called the synchronization mode), as explained in Section 3.6.2. All bits of the first 4 octets of the header, except the CLP bit, are set to zero in empty frames. The CLP bit is set to 1. 4.5.2

Delimitation by Frame-Length Indicators

The header is the same for all frames, but the length of the frame varies. For this scheme to work, it is necessary to keep track of the header of each multiplexed signal by use of length indicators, more data indicators, end of message indicators, and alignment with an underlying envelope. This is the multiplexing scheme of the ATM adaptation layers named AAL2. We shall use AAL2 as a simple example in order to explain the principle (even though AAL2 is no longer used in ATM). The principle is also used in GSM, but the procedures are more complex and intertwined with other functions.

E

I2

I1

I3

E

E

E

I4

I1

I4

E

E is an empty cell Time I1 is information from source 1 to sink 1

4.5 Statistical Multiplexing

I1

I2 is information from source 1 to sink 2 I3 is information from source 2 to sink 3 I4 is information from source 3 to sink 1 ATM cell

Header 5 octets

Identifiers 28 bits

ATM frame format.

Payload type 3 bits

CLP 1 bit

Header error correction 8 bits

97

Figure 4.19

Payload 48 octets

98

Multiplexing

In the case of AAL2, the underlying envelope is the ATM cell structure. AAL2 supports low-speed data channels with variable bit rates. More than one AAL2 channel may be multiplexed within one cell. This is done as shown in Figure 4.20. The individual channels are segments of low-rate information streams (for example, several 28-Kbps data channels and 64-Kbps voice channels multiplexed in the 155-Mbps ATM stream). The streams are multiplexed within cells, as shown using the cell header (H3) as fundamental timing reference. The header of the common part sublayer protocol data unit (CPS-PDU), H2, consists of 8 bits and contains a 6-bit pointer indicating where the H1 header of the first channel starts in the CPS-PDU. This is the major purpose of the CPS-PDU header. In the example, the pointer value in the first PDU indicates the start of channel 1, and the pointer value in the second PDU indicates the start of channel 4. The bits following channel 5 in the example are padding bits containing no information and are used to encode empty bit positions at the end of the frame. These bits are set to binary zero. The header H1 consists of 3 octets and contains several fields, the two most important of which are the channel identifier field (the first 8 bits of the header) and the length indicator field (6 bits). The length indicator specifies the total number of octets in the payload field of the channel (maximum 63 octets). The channel identifier enumerates the individual channels in the multiplexing structure. The first bit of the identifier field cannot be zero in order to distinguish it from a padding field. In other words, if the first bit following a channel is zero (channel 5 in the figure), the remaining bits of the CPS-PDU are padding bits. The multiplexing format of AAL2 is very flexible, allowing the size of individual channels, the order in which the channels appear, and the time between fragments of the same channels to vary during a communication session. Note that the arrangement allows channels to be distributed over two ATM cells (channel 3).

Individual channels H1

H2

1

H1

1

2

2

H1

H1 4

3

3

H2

3

4

5

H1

5

Common part sublayer protocol data unit (CPS-PDU)

H3

H3

ATM cells H1 = packet header (3 octets) containing among others a channel identifier and a length indicator H2 = CPS header (1 octet) containing among others a pointer to the first CPS packet in the PDU H3 = cell header

Figure 4.20

AAL2 multiplexing.

4.5 Statistical Multiplexing

4.5.3

99

Delimitation by Flags

In these systems, a delimiter that cannot be simulated by the information signals is inserted between frames. This is the method used in HDLC5-based networks (at the physical layer of some data networks and in the physical layer of SS7). The algorithm is as follows. Frames can be of arbitrary length and contain arbitrary strings of binary information. The delimiter between frames consists of the 8-bit pattern 01111110. This pattern is called a flag. Whenever the receiver detects a zero followed by 6 ones and then a zero, the receiver will interpret this as the end of one frame and the start of the next one. If there is no information to be sent, only flags are sent (corresponding to frames of zero length). In order to avoid a flag appearing in the information stream and thus being misinterpreted as the end of the frame, the sender insert a dummy zero after every sequence of 5 contiguous ones (i.e., the sequence 011111 becomes 0111110). The algorithm in the receiver is then such that whenever the sequence 0111110 is detected, the final 0 in the sequence is deleted and the remaining bits in the sequence are kept as part of the information stream. Isochronism relative to octets cannot be maintained in systems employing this technology, as explained in Section 3.2.

5.

High-level data link control is one of the most basic methods used for adaptive transfer of data. Data are sent in frames that are adapted to the length of the information field, and automatic repeat request is used for error control so that the rate by which information is transferred is adapted to the bit error rate. HDLC also offers data rate adaptation managed by the receiver in order to avoid congestion.

CHAPTER 5

Multiple Access 5.1

Multiple Access Techniques Multiple access is concerned with the way in which the signals are organized in systems where several independent sources are sharing a common medium: satellite systems, land mobile systems, Ethernet, and wireless LAN. There are essentially five different multiple access techniques: frequency division multiple access (FDMA), time division multiple access (TDMA), code division multiple access (CDMA), space division multiple access (SDMA), and random access (RA). Slow frequency hopping code division multiple access (SFH-CDMA) may be regarded as a variation of CDMA, FDM, and TDMA. SFH-CDMA is treated as a separate access technique here. In many cases, several multiple access techniques are combined for serving a single purpose. The GSM system makes, for example, use of FDMA, TDMA, RA, SFH-CDMA, and SDMA simultaneously. UMTS combines FDMA, TDMA, CDMA, RA, and SDMA in a single system. The Ethernet makes use of only RA, while wireless LAN uses CDMA or SFH-CDMA in addition to RA. We may also distinguish between systems where several channels are multiplexed on one multiple access frequency or timeslot and systems where just one narrowband channel (for example, 10 Kbps or 64 Kbps) occupies a multiple access channel. An example of the first type is intercontinental satellite systems where TDM channels may be organized in a TDMA or FDMA configuration. An example where a single narrowband channel occupies one FDMA frequency is the maritime satellite communications system (INMARSAT). In the satellite literature, such systems are often referred to as single channel per carrier (SCPC). Channels may be assigned a frequency, a slot, or a code in a multiple access system on a permanent or preassigned basis (PA). This is sometimes referred to as fixed assigned multiple access (FAMA) (sometimes we see abbreviations such as FDMA-FAMA, FDMA-PA, TDMA-FAMA, and so on). This is the way it is usually done in intercontinental satellite systems and satellite broadcast systems. However, in most cases (e.g., GSM, Ethernet, INMARSAT, WLAN), the multiple access channels are allocated on demand. This is called demand assignment multiple access (DAMA). This leads to system types such as FDMA-DAMA and TDMA-DAMA. Even SCPC-DAMA is sometimes seen if it is not obvious from the context that the capacity of the SCPC system is assigned on demand.

101

102

5.2

Multiple Access

Frequency Division Multiple Access In FDMA, the different sources are assigned specific frequency bands that do not overlap. The frequency slot may be allocated permanently to the source (FDMA-FAMA) or temporarily on demand (FDMA-DAMA). The bandwidth of the frequency slots allocated to different sources may be the same for all sources or be different. The latter configuration is common in the fixed satellite service in order to support different traffic demands for different Earth stations. It is also possible to allocate several frequency slots to each source. The sources send signals in the assigned frequency slots simultaneously on separate carrier waves, one for each frequency slot. When these signals arrive at the destination (say, a base station or a satellite) their spectra will not overlap and the different sources will not interfere with one another. In a duplex system, a terminal is transmitting in one frequency band and receiving in another frequency band.1 Between these bands there is a guard band (which may be used by other radio systems). This arrangement is called frequency division duplex (FDD). The arrangement is shown in Figure 5.1. Since the device is sending and receiving signals at the same time on the same antenna, the receiver chain must be protected against the transmit signal that may otherwise damage the sensitive low-noise amplifier or mixer at the input of the receiver. Since the power difference Channel ...

Receive

...

Guard band

Bandpass filter

High power

Send

Sensitive low-noise amplifier

Bandpass filter Duplexer

Transmitter

Figure 5.1

1.

Receiver

Duplex arrangement.

Note that there are systems that send and receive in the same frequency band. Such systems are either operated in the half-duplex mode, where the device can either receive or transmit but not both at the same time, or operated in full duplex mode using a complex system of filters, coherent receivers, and frequency synthesizers for discriminating between the channels. Walkietalkies are half duplex.

5.2 Frequency Division Multiple Access

103

between sender and receiver may be more than 100 dB, filters are required to provide a signal attenuation of considerably more than 100 dB over the bandwidth of the guard band in order to provide sufficient protection. Filters introduce signal attenuation that is proportional (in decibels) to the number of resonators in the filter. Therefore, if the guard band is narrow, a filter containing many resonators, and thus producing significant attenuation, may be required. More power must then be generated by the high-power amplifier of the transmitter in order to compensate for the losses. In many systems, output power is expensive and must therefore not be lost before it is transmitted. This is the case in satellites systems where energy is produced by solar panels and in mobile terminals that depends on battery power. Attenuation of the received signal before it is demodulated reduces the signalto-noise ratio of the received signal and, hence, increases the bit error rate. The transmitter must transmit more power in order to compensate for this deterioration. Hence, the filters of the transmitter and the receiver increase the demand for power in the transmitter. The effect the losses in the receiver may have on the received signal is described in Section 9.7.4, where the link budgets of satellite systems are explained. A device called a duplexer is inserted between the antenna, the transmitter, and the receiver as shown in the figure. This device contains transmit and receive filters and a directional device (e.g., microwave or magnetic circulator, or balanced transformer) that guides most of the power from the transmitter to the antenna and most of the received signal to the receiver. The duplexer is one of the most expensive components in FDMA systems because of stringent filter requirements and the use of analog components. Note also that in satellite systems, the lower frequency band is used for reception at the Earth stations (satellite-to-Earth link). This is done in order to reduce the demand for power generation in the satellite since free space loss increases proportionally with the square of the frequency of the signal (∼ 20 log f in dB where f is the carrier frequency) (see Section 9.7.4). In the INMARSAT system, the center frequency of the lower band is about 1.54 GHz and that of the upper band is 1.64 GHz. Using the lower band rather than the upper band for the satellite-to-Earth link, the transmitted power of the satellite can be reduced by about 13%. Altogether this may correspond to a total saving in satellite power of about 7% (assuming 50% efficiency of the satellite—the remaining power is used for receivers, stabilization motors, and telemetry systems). This is much in a satellite system where power is obtained from solar cell panels. The area and the weight of the solar panel may then be reduced by 7%. There is not only a guard band between transmit and receive bands but also between individual frequency slots in the same direction of transmission. This guard band is required in order to keep the interference between adjacent channels at a reasonably low level. This interference is added to the thermal noise of the receiver and reduces thus the signal-to-noise ratio (and, hence, increasing the bit error rate) of the received signal. The narrower the guard band, the more efficiently the frequency spectrum is utilized. The width of the guard band depends on the spectrum of the signal being

104

Multiple Access

transmitted, the stability of the clock frequencies of the sources, and the Doppler shift if the sources are moving relative to on another. Note that all satellites, even the geostationary ones, are moving sources. The geostationary satellites move in complex helical orbits around the circular (or rather slightly elliptical) geostationary orbit for several reasons (nonuniformity of the Earth’s gravitational field, and the gravity of the Moon and the Sun). Therefore, Doppler shift must be taken into account when planning satellite systems. Digital signals cause more interference between adjacent channels than analog signals. However, digital signals are more robust against interference than analog ones so that digital channels may be packed more tightly than analog channels. How densely the frequency spectrum may be filled depends on how accurately the clocks of the sources are running. In many satellite systems, one Earth station is used as a frequency reference for the other Earth stations. This is a simple application of phase-locked loops described in Chapter 3.

5.3

Time Division Multiple Access In TDMA, each source is assigned a specified timeslot or burst in each frame. The frame represents a repetitive pattern where the same channel appears at fixed intervals in time. While FDMA divides the spectrum into different bands, TDMA chops the time axis into bursts in which information can be sent. All sources then send at the same frequency but not at the same time, and the spectrum of the TDMA signal spans the entire frequency spectrum allocated to the TDMA system. The timeslots assigned to different sources must not overlap in time when they arrive at the destination (say, a base station or satellite). The arrangement is shown in Figure 5.2. The farther away from the base station (satellite) the source is located, the earlier the source must send the burst. The way in which this timing is carried out in GSM and satellite systems was explained in Section 3.7. The synchronization

Time Frame Burst1

Burst2

Burst3

Burst4

Burst5

Burst6

Burst1

At the transmitters of the sources

Burst2 Burst6

TDMA system.

At the base station receiver

Burst1

Burst5 Burst4

Figure 5.2

Burst1

Burst7

Burst3

Distance from base station

Burst7

5.3 Time Division Multiple Access

105

procedure consists of two phases: first, the source determines the time at which the bursts are to be sent relative to a reference signal received from the base station (satellite) so that the bursts from the source reach the base station (satellite) in the correct time slot; second, the source must retain synchronization afterward by adjusting for variations in delay and timing accuracy. Each burst consists of a sequence for carrier and clock synchronization, a unique word that determines the start of the payload, and the payload itself. In most systems the unique word follows directly after the synchronization pattern. In GSM, the unique word (called the training sequence in the specification) appears in the middle of the burst. The unique word in GSM serves two purposes: the unique word determines the start of the payload in the normal way but is also used to estimate the transfer function of the radio path. This can be done since the word consists of a known sequence of bits, and the appearance of bit errors (and even the pulse shape) in this sequence can be used in the estimation. The estimated transfer function can then be used to shape the receive filters and to set the parameters of the adaptive error correction decoders. This is the reason behind the name training sequence—it “trains” the receiver in a dynamic and adaptive manner to adjust to the variation in propagation conditions. The training sequence is placed in the middle of the burst. The reason is that the propagation conditions may change considerably over the duration of a single burst (0.577 ms). Placing the training sequence in the middle then allows the receiver to determine a better estimate of the transfer function of the radio channel. This arrangement offers better performance of the receiver but requires much processing. Timing accuracy is very important and difficult in TDMA systems. The guard time between bursts must be short for efficient utilization of the available spectrum. Because of the long differential delay (40 ms) in geostationary satellite systems, TDMA is not suitable for mobile narrowband SCPC systems such as INMARSAT. The reason is that the guard space between bursts must be long in order to avoid adjacent bursts from overlapping because of the timing inaccuracy. TDMA can be used for the equally narrowband channels of GSM because it is possible to manage the differential time delay by use of timing advance as was shown in Section 3.7.2. Similar methods cannot be applied in narrowband mobile satellite systems because the differential delay is too big. TDMA is suitable for intercontinental satellite systems where several primary channels are TDM-multiplexed in each TDMA burst. The configuration of such systems is considered in Chapter 9. Applying time division duplex (TDD), the device sends and receives at different times as shown in Figure 5.3. The design of the duplexer is simple because it just consists of an electronic switch controlled by the timing and channel coding subsystem opening and closing the paths between the antenna and the transmitter and the receiver, respectively. During the interval in which the burst is transmitted, the switch opens the path from the transmitter to the antenna and closes the path toward the receiver protecting the receiver from overload. During the remainder of the frame period, the path toward the receiver is open and the path from the transmitter is closed, leading all received power toward the receiver.

106

Multiple Access

6

7

0

1

2

3

4

5

6

7

0

1

2

6

7

0

1

2

3

4

5

6

7

BS transmits

0

1

2

MS transmits

(a)

Timing

Electronic gate

Transmitter Receiver

Channel coding

(b)

Figure 5.3

5.4

(a, b) Duplex arrangement in GSM.

Slow Frequency Hopping Code Division Multiple Access Slow frequency hopping can be regarded as a special case of FDMA. Frequency hopping means that the source is changing transmit frequency dynamically in accordance with a predetermined timing and frequency arrangement. The spectrum of an SFH system consists of several frequency channels with a bandwidth adapted to the information rate of the duty signal in the same way as FDMA systems. The sources then hops form one such frequency channel to another at a rate that is slower than the bit rate of the duty signal. The hop sequence is a pseudorandom sequence.2

2.

The technique is called code division multiple access because each duty signal occupies the whole frequency spectrum over a whole pseudorandom hop sequence, and several duty signals occupying different frequencies for each hop sequence can be present in the system simultaneously. It could just as well be called frequency hopping FDMA or, for that matter, frequency hopping TDMA because the signal may be regarded as a TDMA system where different bursts are using different frequency slots. It may be confusing to treat slow frequency hopping and code division multiple access together because they are using the frequency spectrum in different ways.

5.4 Slow Frequency Hopping Code Division Multiple Access

107

The time-frequency domain of SFH is shown in Figure 5.4 for two signals. The pattern is a pseudorandom sequence that repeats itself after a certain time called the frame length of the SFH sequence. The time that the system stays at a given frequency is called the dwell time. In slow frequency hopping, the dwell time is larger (usually much larger) than the duration of a bit of the duty signal. The dwell time is the same for all hops. The portion of the signal that appears in the dwell time is called a burst. The purpose of SFH is to make the propagation conditions for all sources as equal as possible. The reason is that in many systems such as land mobile systems, interference and propagation conditions depend on the frequency band in which the source operates. Cells using the same frequencies may interfere with one another. The signals from the base station in one cell may interfere with the signals from the base station in another cell, thereby increasing the noise level in the cell. Similarly, mobile stations may cause interference in remote base stations. The frequency planning in mobile systems is such that these problems are rare. However, in cities or other areas with high traffic demand, the problem cannot be eliminated entirely. The frequency reuse distance may then not be large enough to eliminate the problem entirely. Sometimes the interference may be so severe that the channel on which it takes place may be unusable. This puts an upper limit on the number of simultaneous calls in a geographical area. This limit can be increased by use of SFH. SFH ensures that the interference level is equalized for all calls, and the total traffic in the system can be increased still maintaining sufficient quality of each call. This application of SFH is called interferer diversity: the interference is spread over the entire system and not restricted to some frequencies. The radio signals are subject to multiple reflections from the terrain and from buildings and other objects. The signals may also be subject to refraction by obstacles in the propagation path. When rays belonging to the same signal that have followed different paths meet, they will interfere with one another causing a reduced

Frequency

One frame Frequency channel

Time Longer than one bit

Figure 5.4

SFH in the time-frequency domain.

Dwell time

Burst

108

Multiple Access

or an increased power level of the signal depending upon the phase and the amplitude differences between the rays. This phenomenon is called multipath propagation (see Section 8.3, where radio wave propagation in mobile systems is explained). The phase of a ray is equal to 2 times the physical length of the path the wave has propagated divided by the wavelength of the ray. The phase difference between two rays with different frequencies following the same physical paths is thus frequency dependent, and we may conclude that the variation in received signal level—or fading—is frequency dependent. In fact, the frequency-dependent level variation over the frequency bands allocated to GSM and WLAN may be significant as the following example shows. The effect of frequency dependence caused by multipath propagation (sometimes called frequency selective fading) is illustrated in Figure 5.5. The phase difference between the reflected and direct signal component when they interfere in the receiver is 125.5 × 2π = π( modulo 2π ) at 890 MHz, the lowest frequency in the GSM band. The direct and reflected components interfere destructively in the receiver. The relative phase difference at 915 GHz, the highest GSM frequency, is 128 × 2π = 0( modulo 2π ). The direct and reflected components then interfere constructively. This means that a signal transmitted at 890 MHz may be extinguished if the direct wave and the reflected wave have equal amplitude when they interfere, while the amplitude of a signal transmitted at 915 MHz is doubled. Since the geometry shown in the figure is well within what can be expected for GSM, we see that frequency selective fading will take place over the GSM band. Application of SFH prevents a call from staying all the time at a frequency that has low signal-to-noise ratio. The quality of the call will rather vary randomly between poor and good as the signal hops between different frequencies. The application of SFH will thus improve the quality of a signal received with low signal-to-noise ratio. On the other hand, a call on a frequency with high signal-to-noise ratio will, for the same reason, on average experience a poorer quality. SFH thus equalizes the average signal-to-noise ratio over all calls. This application of SFH is called frequency diversity. If the dwell time is short, it is also possible to use error correction and interleaving (see Section 8.5.2) to compensate for the variation in bit error rate of adjacent bursts. GSM applies slow frequency hopping of TDMA bursts; that is, bursts in adjacent frames may appear at different frequencies. The dwell time is equal to the burst length (0.577 ms). On the other hand, one channel makes one hop per TDMA frame so that the TDMA pattern is preserved. The dwell time is then equal to the duration

Reflection

Reflected signal 145m

1000m Transmitter

Figure 5.5

Path geometry.

Direct component Receiver

5.5 Direct Sequence Code Division Multiple Access

109

of one TDMA frame (4.615 ms). Only 0.577 ms of the dwell time is occupied by the signal. Wireless LAN defined in IEEE 802.11 may utilize SFH with a dwell time of 400 ms, implying that many data packets may be sent during one dwell time. The maximum-length Ethernet packet is 1,518 octets. The duration of this packet is 12 ms so that more than 30 such packets may be sent within a single dwell time. Important: Slow frequency hopping is not the same as fast frequency hopping CDMA (FFH-CDMA) described later. Slow frequency hopping is usually referred as a spread spectrum technology, since one channel uses the whole frequency spectrum but the implementation and performance of these systems are more similar to FDMA systems than to other CDMA techniques. Therefore, I prefer to consider slow frequency hopping and fast frequency hopping as multiple access techniques that must be described separately.

5.5

Direct Sequence Code Division Multiple Access DS-CDMA is also often called spread spectrum multiple access (SSMA). The term DS-CDMA is now more common. CDMA exploits another way of arranging signals on the radio path so that they can be distinguished from one another. One principle is that each symbol, 0 or 1, of a binary signal is replaced by a string of bits—say, 1,000 bits. In order to make the notation less confusing, we call the bits of the spreading code “chips,” though they are binary pulses. The new signal will then have a chip rate that is 1,000 times bigger than the bit rate of the original signal, since each symbol of the original signal is coded as 1,000 chips. If the original bit rate is 100 Kbps, the coded bit rate will be 100 Mcps, where cps stands for chips per second. The code consisting of 1,000 chips is called the spreading code, and the factor 1,000 (or 30 dB = 10 log 1,000) is called the spreading ratio. This ratio is also called the coding gain or processing gain, a notation that is justified next. 5.5.1

Coding Gain

Let be a string of bits where each bit takes either of the two values +1 and −1; and let C be the chip sequence where each chip also is represented as binary state +1 or −1. The signal is encoded by the chip sequence C by multiplying chip by chip the bit sequence by the chip sequence C, producing a new sequence E at the same rate as the chip sequence: E = σ × C. A binary +1 of the signal is then represented as a portion of the direct sequence C, while a binary −1 of the signal is represented as a portion of the binary inverse sequence C*. The encoding is shown in Figure 5.6. In the example, the duration of the chip sequence is equal to the bit length for simplicity of drawing. This requirement is not necessary: there may be any ratio between bit length and the duration of the chip sequence. C is a pseudo-noise sequence generated by a shift register with feedback in exactly the same way as was explained for generating scrambling sequences (see Section 3.7.5). The bit pattern that the shift register produces is determined by the position of the feedback loops in the register and the initial condition of the shift

110

Multiple Access

Bit sequence (σ)

Chip sequence (C)

Encoded sequence (E)

Figure 5.6

Encoding of DS-CDMA signal.

register (i.e., the bit pattern in the shift register when it starts running). The pseudo-noise sequence is deterministic so that exactly the same sequence can be produced by the receiver when the initial conditions are known. The chip sequence C is thus represented as a series of binary symbols with binary levels +1 and −1 such as for example the sequence shown in Figure 5.6: +1,+1,−1,+1,−1. C* is then the sequence −1,−1,+1,−1,+1. Multiplying C with C chip by chip and adding (actually integrating), symbolizing this combined multiplication and addition/integration operation by the symbol ⊗, we get C ⊗ C = +K for a sequence of K chips. We also see that C ⊗ C ∗ = −K. We may call the operation ⊗ the correlation between chip streams of length K. In other words, if we receive a stream of bits consisting of K chips each, and correlate the received signal with the locally generated chip stream C, then the received signal is either + K for a binary +1 bit or −K for a binary −1 bit. In other words, if the received power level of a single chip is p, the received power level of a bit after correlation is Kp. This means that the chips can be sent at a power level of 1/K of that of a single bit. The correlator at the receiver will then bring the power of each bit back to its original value. Hence, it is justified to call K (or10 log10 K dB) the coding gain. If one bit consists of 1,000 chips, the coding gain is 30 dB. If a signal-to-noise ratio of 5 dB is needed for demodulating the signal, then the chips may be sent at a level that is 25 dB below the level of the thermal noise of the system. After correlation we then get a signal-to-noise ratio of − 25 dB + 30 dB = 5 dB. Note also that because of the particular correlation operation involving integration over bit durations, the chip stream and the bit stream need not be in mutual synchronism; that is, the duration of one bit need not be an integer multiple of the chip duration, or the transition from one binary state to the other state need not take place simultaneously for bits and chips. However, in the most common implementations, each bit of the duty signal corresponds to one complete chip sequence. The different chip sequences used in a system are chosen such that they are orthogonal. This means that the correlation of the two sequences Ci and Cj is: C i ⊗ C j = 0 if i ≠ j

5.5 Direct Sequence Code Division Multiple Access

111

This means that the decoding of another signal than the wanted signal produces no signal from the correlator. In other words, the correlator picks out only the wanted signal and suppresses all other signals encoded by an orthogonal chip sequence. If n signals encoded with orthogonal chip sequences Cj(i) are superimposed on one another, the resulting signal is S (i ) =

n

∑ C (i ) j

j =1

After correlation over the length of 1 bit of the duty signal (i.e., over K chips where K is the length of the chip sequence), there results in the following signal: K

∑ S (i ) ⊗ C 1 (i ) = i =1

K

n

∑ ∑ C j (i ) ⊗ C 1 (i ) = i =1 j −1

K

∑ C (i ) ⊗ C (i ) = K 1

1

i =1

K

since

∑ C (i ) ⊗ C (i ) = 0 if j ≠ 1 i =1

j

1

because of orthogonality. C1(i) is the chip

sequence of the wanted signal. The most important task is thus to find orthogonal sets of chip sequences. The task can be solved exactly using abstract algebraic methods far beyond the scope of this book. 5.5.2

Autocorrelation Properties

The autocorrelation properties of a suitable chip sequence is illustrated for the 7-bit sequence C(i) −1,−1,−1,+1,−1,+1,+1, where i refers to the position of the chip in the sequence. The autocorrelation is the voltage produced by the correlator if the chip sequence is compared to a replica of itself that is offset by a certain number if chips. The autocorrelation function Aut is therefore defined by the expression Aut( k) =

6

∑ C(i)C(i − k) 0

where k is the replacement of the chip sequences relative to each other. Aut(k) is then the voltage produced for a replacement of k chips. If k = 0, Aut(0) = 7 since 6

6

0

0

∑ C(i)C(i) = ∑ 1 = 7 Table 5.1 shows the autocorrelation function for all values of k from 0 to 6. We see that this particular chip sequence produces a correlation Aut(k ≠ 0) = –1 to any displacements (k ≠ 0) of itself. This is exactly one of the conditions the chip sequences of a CDMA system must satisfy. The other property is that there must be several chip sequences of the same length that are mutually orthogonal.

112

Multiple Access Table 5.1

Autocorrelation of the Sequence −1,−1,−1,+1,−1,+1,+1 k

0

1

2

3

4

5

6

−1

−1

+1

+1

−1

+1

−1

−1

−1

−1

−1

+1

+1

−1

+1

−1

−1

−1

−1

−1

+1

+1

−1

+1

+1

+1

−1

−1

−1

+1

+1

−1

−1

−1

+1

−1

−1

−1

+1

+1

+1

+1

−1

+1

−1

−1

−1

+1

+1

+1

+1

−1

+1

−1

−1

−1

Correlation

+7

−1

−1

−1

−1

−1

−1

C(i)

5.5.3

Composition of a DS-CDMA Transceiver

A DS-CDMA terminal is shown in Figure 5.7. The data stream is encoded (or spread) with the chip sequence of the transmitter, amplified and fed to the antenna for transmission. The receiver contains a similar chip sequence generator that must be synchronized to the chip sequence produced by the remote transmitter. The receiver contains a coherent receiver and a chip sequence correlator that may be integrated as a single function. One principle is to modulate the chip sequence on a reference carrier (the box “reference signal” in the figure) producing one or more signals that are presented to the correlator in order to extract the duty signal using coherent phase shift demodulation. The purpose of the coherent receiver and the correlator is to extract the wanted signal and to suppress interference and other noise. The orthogonality of the chip sequences ensures that the synchronizer locks to the wanted signal and not to signals from other sources. High-frequency spectral components are finally suppressed by a lowpass filter before the data signal is processed further.

Data stream

Chip stream

Transmitter

Chip sequence generator Coherent receiver and correlator Data stream

Lowpass filter

Chip sequence generator

Reference signal

Synchronization

Figure 5.7

DS-CDMA terminal.

Duplexer

5.5 Direct Sequence Code Division Multiple Access

5.5.4

113

Interference and Channel Capacity

We shall now consider the properties of the coherent receiver of Figure 5.7. The coherent receiver extracts the chip sequence of the wanted signal. However, it also produces noise and interference signals. The interference may be generated by jamming and other unwanted interference, signals from other users, multipath propagation, and so on. If the interference is caused by a pure carrier wave, the correlator multiplies this signal with the chip sequence and thus spreads the energy of the carrier over the entire bandwidth of the chip signal; that is, the amount of energy received within the bandwidth of the duty signal is reduced by the spreading factor K. This effect is called the jamming resistance and is one reason that CDMA is used for radars: the receiver cannot be jammed even by powerful signals. In other words, the correlator increases the wanted signal by a factor K and suppresses a pure carrier wave with the same factor. If there are M users and all of them are transmitting carriers with power p, the received signal power is Mp. Even though the different users employ orthogonal chip codes, the wanted signal and the received signals are not orthogonal at the receiver (for obvious and practical reasons, different terminals are not synchronized to one another). The orthogonality property of the spreading signal is used only in order to extract the wanted signal. The receiver will experience interfering carriers containing arbitrary binary sequences. The correlation between such pseudorandom sequences and the spreading sequence (which is itself pseudorandom) produces another pseudorandom sequence with bandwidth equal to the bandwidth of the spreading signal, while that of the wanted signal has been concentrated by the correlator to fill the bandwidth of the duty signal only. This is the same as saying that the wanted signal has been amplified by the factor K (the spreading factor), while the interfering signals have not been affected by the correlation process. The result is then a duty signal of power Kp and M − 1 interfering signals each with power p [i.e., producing an interference or noise level of (M − 1)p]. The signalto-noise ratio after demodulation is then S/N = Kp/(M –1)p = K/(M – 1) = K/M if M is large. The maximum number of carriers that can be present simultaneously is then Mmax = K/(S/N)min, where (S/N)min is the minimum signal-to-noise ratio required for demodulation of the signal. If K = 1,000 and (S/N)min = 5 (or, in decibels, 7 dB), we find that altogether 200 channels can be fitted into the bandwidth. Several other factors must be taken into account when designing such systems, the most important of which is power control. This calculation was made under the assumption that all carriers are equally strong at the receiver. If this is not the case, the capacity of the system is reduced. Let us assume that the power of the wanted carrier is p and that of the other carriers are 2p, then we find that Mmax = K/2(S/N)min since now the noise has increased to 2Mp. In this case the channel capacity is reduced to only 100 carriers. This illustrates that the performance of DS-CDMA depends critically on the power of each carrier. The maximum number of simultaneous carriers is achieved only if all carriers are equally strong. This requires that strict power control is implemented in such systems.

114

Multiple Access

5.5.5

Power Control

Power control can be achieved in mobile systems such as UMTS in the following way. Let us assume that the propagation loss L (in decibels) is the same in both directions. Then we see that (all parameters in decibels): (S/N)BS = SBS – N = PMS – L – N

where N is the total noise and interference, S is the received signal power, P the transmitted power, and (S/N) is the signal-to-noise ratio. BS is base station and MS is mobile station. Similarly, SMS = PBS – L or

L = PBS – SMS

Inserting this value for L in the first equation gives (S/N)BS = PMS – PBS + SMS – N or

PMS = (S/N)BS + PBS + N – SMS

target

Replacing PMS by P MS, PBS by the nominal transmitter level of the base station PnominalBS, (S/N)BS by the signal-to-noise ratio required for proper demodulation of the signal (S/N)target, and N by the maximum expected interference and noise in the system Nmaximum, we get the following expression: target

P

MS

= (S/N)

target

+P

nominal BS

+ Nmaximum – SMS = constant – SMS

The constant is a design parameter of the system. The variation of the transmit power is then the negative of the received power measured in the receiver. This power control method is called open-loop control since it does not entail feedback loops. It assumes that the propagation loss is the same in the two directions. The power adjustment should take place slowly (order of several milliseconds in UMTS) if the output power is increased. Then we avoid producing sudden power spikes causing interference for the other users. The adjustment can be rapid (order of one millisecond) if the power is to be reduced, also in order to reduce interference. We may put it this way: we avoid increasing the interference level—and thus making the quality worse for all other sources—by accepting a poorer quality of the wanted signal. The base station may also measure the received power level and instruct the mobile station to increase or lower the output. This is a closed-loop control mechanism. The method takes into account that the fading in the two directions may be independent—as it usually is. The mobile station may also measure the received power level and request the base station to increase the power level if the signal is received with poor quality. The base station will then decide to increase the power level or to keep it as it is depending upon the number of active calls in the cell and other quality measures.

5.5 Direct Sequence Code Division Multiple Access

5.5.6

115

Autocorrelation, Acquisition, and Tracking

One of the most difficult problems in CDMA has to do with initial acquisition of the chip stream. The autocorrelation of the pseudorandom chip sequence is such that the condition C(t) ⊗ C(t – ) = –1 if ≠ 0 and C(t) ⊗ C(t – ) = K (the number of chips) if = 0. The autocorrelation is significantly different from one only over the duration of a single chip. Initial acquisition in the receiver is done by comparing the received sequence bit by bit until acquisition takes place. This is of course possible because the chip stream is known at the receiver. When initial acquisition has been achieved, a phase-locked loop tracks the chip sequence with accuracy better than one chip duration. The phase detector of this loop has the form shown in Figure 5.8(a). If the loop is one chip out of phase, the initial acquisition procedure is required in order to relock the chip oscillator. The correlator produces the voltage shown in Figure 5.8(b) when comparing the incoming chip stream with the chip stream generated in the receiver. When the two sequences match, then the correlator outputs a peak as shown in the figure indicating the beginning of the next chip sequence. The electronics of the receiver then uses this information to recover the bit sequence by computing the correlation product of the sequence for the duration of a bit. If the length of the bit corresponds to the number of chips in the sequence, then the bit value +1 will produce the voltage +K, and the bit value −1 will produce the voltage −K. Note that CDMA systems can be designed in principally two ways. Each source can use a separate chip code for discrimination between them. This method is easy in FAMA systems because then the code can be preassigned for each channel and used for all future time. The problem in DAMA systems is that in these systems a channel is assigned on demand. If, in addition, the system can be accessed by

Voltage

Phase displacement (a)

Voltage

K

Phase displacement

–1 One chip duration

Length of chip code (b)

Figure 5.8

(a, b) Phase detector characteristic of autocorrelator.

116

Multiple Access

arbitrary sources such as in a public land mobile system, preassignment is not possible so that a chip code, Ci, must be assigned whenever a source requests access to the network and is announced to the source. A systemwide chip code known to all sources must be assigned for the first access to the network. Otherwise, newly arriving sources cannot access the system. 5.5.7

Multipath Diversity

The autocorrelation property of DS-CDMA can also be used to suppress multipath interference. A multipath signal consists of several reflected components and possibly one line-of-sight component. These signal components are received at slightly different times. If the time difference between them is one chip duration or more, only one component will be amplified by the correlation process. The signal variation will then depend only on the variation of one of the multipath components and not on the fading pattern created by all the components. It is also possible to use several parallel receivers that track different multipath components. The demodulated signals can then be combined in order to produce a signal with a smaller bit error rate. This receiver technology is often referred to as the Rake receiver concept. It is far beyond the scope of this book to consider the details of such receivers. 5.5.8

Application of DS-CDMA

The wireless LAN of IEEE 802.11 and systems based on or are similar to UMTS may use DS-CDMA.

5.6

Fast Frequency Hopping CDMA Fast frequency hopping CDMA is such that there are many hops per bit of the duty signal; that is, the dwell time is short compared with one bit. The hop sequence is a pseudorandom sequence similar to the sequence used in DS-CDMA. The receiver extracts the duty signal by correlating the received signal by the known pseudorandom hop sequence. This gives a coding gain in the same way as in DS-CDMA systems. The major problem with FFH-CDMA is that the frequency synthesizers must be extremely fast with exceptionally short switching times. At the receiver, this implies that coherent demodulation cannot be applied because there is not enough time to both synchronize to the chip frequency and to perform demodulation by direct phase comparison. Noncoherent frequency demodulation requires higher signal-to-noise ratio than coherent demodulation, and FFH-CDMA will therefore have smaller channel capacity than DS-CDMA. The hop pattern is illustrated in Figure 5.9 for three sources. The pattern is the same as for slow frequency hopping.

5.7 Comparison of FDMA, TDMA, and DS-CDMA

Comparison of FDMA, TDMA, and DS-CDMA The three multiple access techniques FDMA, TDMA, and DS-CDMA are equivalent when we compare the amount of information that can be sent in a given bandwidth provided that the best possible technology is applied in each case. This is an intuitive corollary of Shannon’s fundamental theorem on channel capacity. What makes the methods different is how efficient the system can be made in practice, where efficiency takes into account cost, size, power consumption, computational requirements, how mature the technology is, and environmental constraints such as propagation conditions and spatial distribution. Therefore, it is not possible to range the different technologies in a unique manner. It is generally possible to identify cases where one technology is superior over another technology and cases where the situation is opposite. The main difference between the three access technologies is shown in Table 5.2. The characteristics that discriminates them are how they use the frequency spectrum, the time, and the code space. The last column is concerned with the capacity limits of the system. The capacity limits of FDMA and TDMA are hard in the sense that only a maximum number of sources corresponding to the case where all frequencies or all timeslots are occupied can access the system simultaneously. The capacity limit of CDMA is soft since

Frequency

5.7

117

2

1

3 3

3

1

2

2

3

3

1

2

2

2 2

3

1

1

3

1

3 3

1

1

3

2

1

2

1

2 Time

Figure 5.9

Table 5.2

Frequency hopping.

Characteristics of Multiple Access Techniques Use of Spectrum

Use of Time

Use of Code Space

Capacity Limit

FDMA

Unique frequency channel

All

All

Hard

TDMA

All

Unique timeslot

All

Hard

CDMA

All

All

Unique code

Soft

118

Multiple Access

as more and more sources access the system, the quality of the communication is gracefully degraded until the bit error rate becomes too large. The capacity limit is soft also in the sense that channels with different chip rates may coexist in the same system. This implies on the one hand that the users can occupy the channel in agreement with their capacity requirements. On the other hand, the number of users of the channel can be traded with the communication bandwidth allocated to each user. These are the major reasons why CDMA is of particular interest in mobile systems.

5.8

Space Division Multiple Access Earlier we considered how communication entities can share a common transmission medium by dividing the common frequency spectrum into frequency bands, dividing the time axis into frames, or operating in a shared code space. SDMA has to do with the way in which space is used to discriminate between communication sessions. SDMA is used in satellite systems and land mobile systems. In land mobile communications, the radio cells makes up an SDMA system (see Chapter 8 for more details). In spot-beam satellite systems, the Earth is illuminated by several narrow beams as illustrated in Figure 5.10. The beams are shaped by several antennas in the satellite or by one antenna capable of producing several beams in different directions simultaneously. The advantage of this system is that the frequency spectrum can be reused in coverage areas that are sufficiently far away from one another, thus increasing the communications capacity of the satellite. The technology requires that the two-way communication satellites must be equipped with onboard switching arrays in order to route calls between different spot-beams. The principle is illustrated in Figure 5.11. The diversity circuit shapes the radiation pattern of the antenna into several spot-beams.

Figure 5.10

Spot-beam satellite system.

5.9 Random Access: Basic Theory and Applications

119 ...

Diversity circuits

Duplexer

TX

Duplexer

RX

TX

Duplexer

RX

TX

RX

... Demultiplexers

Switching fabric

Multiplexers ...

Figure 5.11

5.9

Transponder of spot-beam satellite.

Random Access: Basic Theory and Applications 5.9.1

Aloha Techniques

The Aloha technique consists simply of transmitting packets of information when the sender is ready to do so independently of whether there are other senders in the system doing the same. If each packet has a length T and the traffic intensity is , then the probability that there is no other packet in the collision interval 2T where the packets may overlap is e−2 T. The probability that a packet becomes ready for transmission after a given time interval is assumed to follow the negative exponen−2G tial distribution. The throughput of this channel is then S = Ge , where G = T is the offered traffic. The throughput is thus the fraction of time that the channel is occupied by packets that are successfully delivered to the receiver. This method is called pure Aloha. The throughput has a maximum for G = 1/2. The throughput is then 1/2e ≈ 0.18; that is, it is not possible to utilize more than 18% of the total capacity of the channel. For all other values of the offered traffic the throughput is smaller than 18%. In slotted Aloha, the channel is divided into slots of duration T. A packet that becomes ready to be sent during a slot interval must wait until the end of this interval before it is sent. Collision takes place only if other packets become ready for −G transmission during the same slot interval. The throughput now becomes S = Ge , since the interval in which two or more packets may overlap is now equal to the slot

120

Multiple Access

duration T. As before, G = T is the offered traffic. This expression has a maximum for G = 1, giving a maximum throughput of 1/e ≈ 0.37. The performance of the random access system can be improved by letting the sources listen to the channel before they decide to transmit a packet. This method is called carrier sense multiple access (CSMA). In the simplest version of CSMA, called 1-persistent CSMA, the source will send the packet immediately, provided that the medium is idle when the packet is ready to be sent. The contention probability is then the probability that another source senses the medium to be idle during the same short sensing interval. If the medium is busy when the packet is ready, the source will wait until the channel becomes idle and then transmit the packet. The contention probability is now given by the probability that other packets become ready for transmission during the time the channel was busy. This procedure can be improved as follows. In p-persistent CSMA, the source senses the channel and follows the following algorithm after finding the channel idle: • •

•

The packet is sent in the first idle slot with probability p. With probability 1 − p the source will wait until the next slot. If this slot is idle, the procedure is repeated: the packet is either sent with probability p or deferred with probability 1 − p, and so on until the packet has been sent or the source has given up. If the channel is busy after rescheduling, the same procedure may be repeated or a back-off algorithm may be applied as if the packet actually collided. The same back-off procedure is used if the packet actually collides.

Note that a slot in this context is just a time interval not necessarily related to the length of the packets. In nonpersistent CSMA, the source senses the channel and sends the packet immediately if the channel is idle. If the channel is busy, the source waits a random time after it senses the channel to be idle before the packet is sent in order to reduce the probability of colliding with other sources that have become ready for transmission during the same interval. If the channel again becomes busy during this interval, the source draws another random interval and repeats the procedure. In practical applications of nonpersistent CSMA, the random interval will consist of a limited number of slots. The slot in which the packet is sent is usually drawn from the uniform distribution; that is, the probability is the same for choosing any number between zero and the maximum number. Nonpersistent CSMA (both with finite and infinite number of slots) is not a fair method in the meaning that some sources may wait a very long time before they can access the channel. A more fair method is used in IEEE 802.11, where the countdown of the random time is stopped if the channel becomes busy during the countdown period. Instead of drawing a new random time, the source continues the countdown from the value at which the countdown stopped when the channel again becomes idle. The method used in IEEE 802.11 is called collision avoidance (CA), and the access method is called CSMA/CA. All the CSMA methods described here allow packets of variable length to be sent on the medium.

5.9 Random Access: Basic Theory and Applications

5.9.2

121

Application of Simple Aloha Techniques: INMARSAT and GSM

In the INMARSAT system, using geostationary satellites, the best way for a ship to get the first contact with a coast Earth station is to use pure Aloha. The collision interval for pure Aloha is just twice the packet length. The length of the access signal in the INMARSAT-B system is 31 ms. The overlap interval is then 62 ms. Slotted Aloha will waste time because the differential delay between signals sent from directly beneath the satellite, and signals sent from the edge of the coverage area is about 20 ms. The minimum slot size if signals from different locations shall not overlap in the satellite is then twice the maximum differential delay plus the length of the access signal (packet length) as shown in Figure 5.12. The slot length must then be 71 ms (i.e., 9 ms longer than the overlap interval for pure Aloha). Exactly the same calculations can be made for the GSM system. From Figure 3.14 (in Chapter 3) we see that the duration of the request packet is 0.325 ms (88 bits of duration 3.69 µs) and the two-way differential delay corresponding to a

Satellite

Directly beneath satellite

Contention interval

Periphery

Differential delay = 20 ms

Timing at periphery

Timing at equator directly beneath satellite

Figure 5.12

Timing diagram in a geostationary satellite system.

122

Multiple Access

distance from the base station of 35 km is 0.233 ms. Since the duration of the request packet is longer than twice the differential propagation delay, slotted Aloha is the optimum choice for GSM. The slot duration corresponds, of course, to the duration of a TDMA timeslot (i.e., 0.577 ms). Because of the long differential delays in satellite systems and GSM, the simple pure or slotted Aloha techniques outsmart even the most sophisticated CSMA method. The GSM system takes advantage of the capture effect as follows. The demodulator of the receiver is a nonlinear device that will suppress the weaker of two radio signals interfering in the same frequency band. As a rule of thumb, a radio signal that is about 5 dB stronger than another radio signal will suppress the weaker signal and be decoded without errors. It is, therefore, likely that the difference in power level of two packets colliding in the same timeslot is large enough so that one of them will be decoded correctly. The performance of the Aloha system in GSM is, therefore, considerably more robust than that predicted by the formula for slotted Aloha: in most cases, random access signals sent in the same timeslot will not destroy one another because of the capture effect so that one of the access messages will survive. The capture effect is not strong enough in the INMARSAT system since the packets will arrive at the satellite with almost the same power (about 2 dB difference at most) and both colliding packets will be lost. The use of the random access procedure for establishing dedicated communication channels is illustrated in Figure 5.13. The normal access procedure is shown in Figure 5.13(a). The access request from the mobile station (MS) contains a unique identification of the MS. The access grant message can then be addressed directly to the MS as shown. The procedure used in GSM is shown in Figure 5.13(b). GSM utilizes the capture effect. The random access messages in GSM are too short to contain the unique identity of the mobile terminal. The “identity” contained in the access request message consists of only 5 bits and is just a reference number assigned to the random access event. The remaining bits in the access message are assigned for other purposes. The five “identification” bits carry no relationship to the identity of the MS. The probability that two MSs choose the same “identity” is only 3%. Although this probability is small, a procedure resolving possible contention events must be devised. The procedure is as follows. The MS identifies itself with an arbitrary number in the range 0 to 31 as explained previously. In Figure 5.13(b), two of the MSs chooses the same identity 5 while the third chooses the identity 2. The signal from MS1 is the strongest signal suppressing the access messages from MS2 and MS3.3 The BS then returns an access grant message to the MS with identity 5. This message is received by both MS1 and MS2. After a while MS3 will decide that the access request was 3.

Since the time required by the base station to process the access message may vary slightly depending upon traffic load, and since (as will be explained in Chapter 8) one paging channel corresponds to four random access channels, the contention window is larger than one timeslot. Two MSs sending random accesses with the same identity within this window cannot decide to which station the access grant message is returned.

5.9 Random Access: Basic Theory and Applications MS

123

Access request

BS

Access grant

(a)

BS MS1 MS2 MS3 Access request (5) Access request (5) Access request (2)

Access grant (5)

SABM (5, MS-ID1) SABM (5, MS-ID2)

UA (5, MS-ID2)

Signaling message = suppressed

(b)

Figure 5.13

(a, b) Contention resolution in GSM.

unsuccessful and enter the retransmission procedure. MS1 and MS2 then synchronize to the frequency and TDMA timeslot indicated in the access grant message and both send the message SABM4 containing the number 5 and the identity of the MS.

4.

Set asynchronous balanced mode (SABM) is the name given to the setup message in the most common data link protocol, HDLC. UA is the unnumbered acknowledgment confirming that SABM has been received.

124

Multiple Access

In the example, MS2 wins the contention this time and the BS returns an unnumbered acknowledgment (UA) containing the identity of MS2. MS2 then continues to establish the call while MS1 either abandons the call or enters the retransmission procedure. 5.9.3

Application of Carrier Sense Multiple Access: Ethernet

Ethernet utilizes a cable (coaxial, twisted pair, or optical) as common transmission medium. The sources are connected to the cable by transducers particularly designed for the type of cable and data rate. The transducer allows signals to be sent on the cable and the signals to be extracted from the cable. Sources can be connected to the cable at arbitrary points along the cable. Ethernet applies a CSMA method called collision detection (CSMA/CD). If the medium is idle, a source (let us call it source 1) having a packet ready for transmission sends that packet on the cable applying 1-persisten Aloha. The receiver of source 1 monitors the cable for collisions. A collision takes place if there are other sources sending packets that overlap in time with the packet sent by source 1 anywhere along the cable. The worst case is shown in Figure 5.14. In this case, source 4 at the other end of the cable starts sending a packet just before it detects the packet sent by source 1. These packets will overlap in time all the way along the cable, provided that the packets are long enough. If the packet is somewhat longer than twice the propagation delay along the cable, source 1 will also detect the collision and from that deduce that the transmission was unsuccessful. The receiver of source 1 will experience this as a sudden increase in the voltage on the cable. The time it takes before a collision is detected is called the slot time. The slot time chosen in Ethernet corresponds to a minimum packet length of 64 bytes. This choice of slot time determines the length of the Ethernet cable. If there is no collision during the slot period, source 1 has seized control of the cable and can continue to send information. In order to prevent a source from holding the medium for unduly long periods of time, a maximum size of the packets is (arbitrarily) chosen to be 1,518 bytes.

1

2

3

4

1 starts sending Minimum packet length

Collision at 3 Collision at 2

1 detects collision and aborts

Figure 5.14

Collision events in Ethernet.

4 starts sending 4 detects collision and aborts

5.9 Random Access: Basic Theory and Applications

125

If a collision is detected, source 1 sends a jamming signal on the cable in order to announce that a collision is taking place. Source 1 then reschedules the transmission of the destroyed packet in accordance with a back-off algorithm. Source 4 in Figure 5.14 also detects that the packet it is sending experiences a collision and also aborts the transmission and initiate a similar back-off algorithm. If this is the first collision source 1 experiences, the source retransmits the packet in either the next timeslot (delay = 0 timeslot) or the following timeslot (delay = 1 timeslot) with equal probability. Transmission takes place if the medium is idle at the scheduled retransmission time; otherwise, the source waits until the cable is idle and repeats the procedure. If the packet collides once more, the interval in which retransmission is scheduled is twice as long (delays equal to 0, 1, 2, or 3 timeslots). The retransmission interval doubles for every collision until 15 attempts have been made. Then the transmission is regarded to have failed and the source gives up. The minimum length of the packets—or slot time—defined for Ethernet is 64 bytes. This corresponds to a slot time of 51.2 µs and 5.12 µs for 10 Mbps and 100 Mbps Ethernets, respectively. The roundtrip delay of the cable must be smaller than the slot time. The roundtrip delay is composed of a number of events such as propagation delay along the cable, time required to detect the signals, signal rise times, collision detection time, feeder delays, and so on. Taking all this into account, the maximum cable length for these two systems are about 2,000m and 200m, respectively, allocating about half of the slot time to processing delays in repeaters and interfaces. If we increase the transmission speed to 1 Gbps, the cable length will be only about 20m for the minimum packet length of 64 bytes. The cable can be made just as long as for the 100-Mbps Ethernet if the minimum packet length is increased to 512 bytes, but then all terminals accessing Ethernets have to implement two rate-dependent minimum packet lengths. This is not regarded to be a good idea so instead this limitation is overcome by specifying that if the packet is shorter than 512 bytes, random bits are appended to the packet by the MAC layer so that the overall packet length becomes 512 bytes. This is called carrier extension. Collisions are then detected if the cable length is about 200m. However, the transmission efficiency of this system is reduced, in particular if the amount of short packets is large. This loss is partly recovered by a method called frame bursting where several packets (up to about 8,000 bytes) may be sent contiguously. The Ethernet is now being extended to 10 Gbps. In this case, collision detection is no longer used, and the Ethernet operates in full duplex mode on long cable (up to 40 km). 5.9.4

Application of Carrier Sense Multiple Access: WLAN

Collision detection cannot be used in radio systems using frequency division duplex. In FDD systems, the source cannot monitor its own packets and thus detect collisions. Even if the radio system is not using FDD, there are other problems. The devices cannot receive and send at the same frequency at the same time, so the system cannot detect the presence of a colliding packet. There may also be hidden sources: two sources may both be visible to a third source but not to each other.

126

Multiple Access

Such situations may arise because of propagation loss in radio systems. If two such sources that are hidden for each other send simultaneously to a third source, their signals may interfere at the third source, a situation that is impossible to detect by the sending sources even if they could listen to the medium while sending. Collision avoidance CSMA (CSMA/CA) is a control procedure that can be used in radio systems. The algorithm is as follows. If a source is ready to send and the medium is sensed idle during a certain short time interval, the source starts the transmission at the end of this short interval. If the channel is busy, the source waits until the medium becomes idle, draws a random back-off time (also known as contention window), and continues to monitor the medium while counting down the back-off time. If the medium becomes busy again, the source repeats this procedure until the message is sent. This may take several cycles. The method used is thus nonpersistent CSMA with limited retransmission interval. A more fair method is to give sources that have already been waiting a smaller back-off window the next time the medium becomes idle. One way to do this is to stop the back-off counter when the medium becomes busy again and restart it when the medium becomes idle. Then the source continues to count down the back-off counter. When the back-off counter expires, the packet is sent immediately (by definition, the medium is always idle when the back-off counter expires). The load is further reduced by applying exponential back-off time where the collision window is doubled each time a collision is detected. This control mechanism is the same as in CSMA/CD described earlier. Since the sources cannot detect collisions directly, collisions must be detected using other means. This may be handled by higher layers of the protocol (e.g., TCP), but then the source will not be aware of the collision since a retransmission by TCP will be regarded as a fresh packet. However, whenever TCP generates a retransmission, the control procedure of TCP will reduce the load on the radio channel.5 Alternatively, acknowledgment packets may be sent directly on the medium. This procedure is possible by the WLANs of the IEEE 802.11 specification and is based on the particular interframe spacing shown in Figure 5.15. The time intervals in the IEEE 802.11 WLANs using direct sequence CDMA6 are as follows: short interframe space (SIFS) = 10 µs, slot time = 20 µs, point coordination function IFS (PIFS) = 30 µs, and distributed coordination function IFS (DIFS) = 50 µs (PIFS plus a slot time). Normal messages are sent after a waiting period equal to DIFS after the medium was sensed idle. Acknowledgment message is sent after a waiting time of SIFS (i.e., during the first slot time) and will therefore not collide with any other packets. The acknowledgment packet must reach all other sources in the area before the end of the normal waiting period DIFS in order to prevent the acknowledgment packet from colliding with regular packets. The maximum roundtrip delay (the maximum time from sending the last bit of the information

5.

6.

This procedure works only if the bit error rate on the radio channel is small. Otherwise, TCP packets may be lost also because of bit errors and not only because of congestion. The traffic regulation mechanism of TCP may then be triggered even if there is no congestion in the system, thereby reducing the traffic handling capability of the system. Other access methods (e.g., slow frequency hopping) have other values for these time intervals.

5.10 Random Access: Stochastic Behavior and Dynamic Control Procedures

127

DIFS PIFS Slot time

SIFS

Slot time

Packet

Medium idle

Ready to send IFS = Interframe space SIFS = Short IFS PIFS = Point coordination function IFS DIFS = Distributed coordination function IFS

Figure 5.15

Interframe space (IFS).

packet until detecting the first bit of the acknowledgment packet) must thus be shorter than two slot times (40 µs). This corresponds to maximum distance between sources of 3,000m, which is far more than the diameter of the coverage areas of practical WLANs. The collision probability can be further reduced if a reservation technique is employed where a source is using normal collision avoidance procedures to send a short request-to-send (RTS) packet. The receiver of the RTS packet returns a clear-to-send (CTS) packet during the collisionless interval (i.e., just after the SIF in the same way as for acknowledgment packets). Since all sources will be aware of the reservation, the information packets for which reservation is made will not meet collisions. This method is specified as an option in IEEE 802.11 WLANs. Several reservation methods have been proposed, tested, and implemented in various radio communications systems (e.g., HIPERLAN). The aim of these methods is to increase the throughput, to reduce the probability that packets are lost due to contention, and to reduce the waiting time before packets are successfully sent. However, this point will not be pursued further here.

5.10 Random Access: Stochastic Behavior and Dynamic Control Procedures 5.10.1

Stochastic Behavior

The purpose of this section is to study methods by which saturation effects and instability in random access channels can be handled. Before we can do that we must understand the dynamic behavior of the random access channel. The simplest case to study mathematically is slotted Aloha. A model for the Aloha channel is shown in Figure 5.16. There are M sources altogether in the system. At a given instant M − i of them can generate fresh traffic while there are i sources in backlog waiting for retransmission. Fresh traffic is generated at rate . Packets colliding in the channel are transferred to the backlog. Packets in backlog are retransmitted with intensity . A practicable value for the intensity

128

Multiple Access

Fresh packets intensity µ

Fresh traffic M – i sources

Successful packets

Channel

Retransmitted packets intensity γ

Unsuccessful packets Backlog i sources

Figure 5.16

Channel model.

can be estimated as follows. After collision, it takes the source a certain number of slots R to discover that the packet was corrupted (e.g., awaiting an acknowledgment at a higher protocol layer). In order to avoid that the packet again collides with the same packets as before, the source waits a random number of slots k before the packet is retransmitted. The total waiting time for retransmission is then R + k, where k is a number in the interval between 1 and K. The average delay before a packet is retransmitted is then R + (K + 1)/2 if the number k is drawn with uniform probability. In the model we may then put = 1/[R + (K + 1)/2]. This allows us to investigate the impact of K on the performance of the channel. The retransmitted packet may also collide and again be placed in backlog. Figure 5.17 shows the possible state transitions that can take place in the backlog. The state transitions are as follows: a packet may be successfully retransmitted reducing the backlog by 1; a fresh packet may be transmitted successfully, or zero or at least two packets from the backlog my collide leaving the backlog unaltered; a single fresh packet may collide with one or more packets from backlog increasing the backlog by 1; or a number of fresh packets may collide, increasing the length of the backlog with the number of fresh packets taking part in the collision. Note that the number of packets in backlog can never be decreased by more than one packet. The number of packets in backlog can increase by an arbitrary number. This particular fact makes Aloha channels particularly sensitive to bursty behavior. If a burst saturates the channel it may take considerable time before the channel again operates properly.

Etc.

i–2

Figure 5.17

i–1

i

State transitions in the backlog.

i+1

i+2

i+3

5.10 Random Access: Stochastic Behavior and Dynamic Control Procedures

129

Let the total number of sources in the system be M and let i sources be backlogged. The transition probability from state i to state j is then for all 0 ≤ i, j ≤ M: j ≤ i – 2: pij = 0 j ≤ i – 1: pij = exactly one packet from backlog and no fresh ones = i (1 – )i –1(1 – M–i ) j = i: pij = (no packets form backlog and exactly one fresh packet) or (zero or at i M–i–1 least two packets from backlog but no fresh packets) = (1 – γ) (M – i)µ(1 – µ) i –1 M–i + [1 – iγ(1 – γ) ](1 – µ) j = i 1: pij = exactly one fresh packet and at least one packet from backlog = (M – i)µ(1 – µ)M – i – 1[1 – (1 – γ)i] j ≥ i + 2: pij = j – i fresh packets and arbitrary many from backlog =  M − i  j −i M−j  µ (1 − µ)   j −i  The throughput of the channel, Sout(n, ), is given by the two cases where exactly one packet is present in the channel; that is, either one fresh packet or one retransmitted packet. From these equations we get immediately for n packets in backlog: S out ( S, n ) = (1 − γ )

n

( M − n ) µ(1 − µ)

M − n −1

+ nγ(1 − γ )

n −1

(1 − µ )

M −n

where S = (M − n) is the fresh traffic. In the limit of many sources, we may use the Poisson approximation: if M is large and also much larger than n, then we may put (1 –µ)M = e–µM and M – n ≈ M. This gives: S out ( S, n ) = (1 − γ ) Se − S + nγ(1 − γ ) n

n −1

Se − S

where S is now approximately M . Also assuming that n is large (but still much smaller than M; that is, M >> n >> 1), we may use the Poisson approximations once n more: (1 − γ ) = e − nγ . The equation now becomes S out = Se

− ( nγ + S )

+ nγ e

− ( nγ + S )

Setting G = n + S, the total traffic in the channel, we find the usual expression −G for the throughput of slotted Aloha: Sout = Ge . (Note that the derivation of this formula also shows that the combination of two Poisson processes with different intensities and populations is a new Poisson process—an observation that is useful in many contexts.) The equilibrium contour of the channel is shown in Figure 5.18 as a relationship between fresh traffic and the number of sources in backlog for one particular value of the intensity, , of the backlog traffic. For each value of there will be one such contour. The fresh traffic S as a function of the number of sources in backlog is the straight line S = (M − n) . This line is called the channel load line. The equilibrium contour expresses the case where the fresh traffic equals the expected throughput of

130

Multiple Access Fresh traffic (packets/slot)

 Overloaded channel 0.37

‚ ƒ

Unstable channel

Equilibrium contour

Stable channel

Number of packets in backlog (n) Stable point Unstable point Saturation point

Figure 5.18

Stability of a slotted Aloha channel.

the channel, Sout, given by the previous equation. At a point below the equilibrium contour, the fresh traffic is smaller than the possible throughput of the channel, while at a point above the contour the fresh traffic is larger than the possible throughput. Figure 5.18 shows three load lines. The uppermost load line (1) corresponds to a saturated channel. In this case M exceeds the capacity of the channel given by the equilibrium contour. In such a channel almost all sources will be backlogged and will remain so indefinitely. The lower load line (3) corresponds to a stable channel. This load line intersects the equilibrium contour in just one point. If this channel is overloaded in a short interval, the traffic performance on average will be such that the load on the channel will move back toward the stable equilibrium point. The load line in the middle (2) corresponds to an unstable channel. The characteristic of this load line is that it intersects the equilibrium contour in three points. The point closest to the ordinate is a stable equilibrium point: after a small perturbation away from this point, the system is most likely to drift back to equilibrium. The point in the middle is an unstable or rejecting equilibrium point. A small perturbation toward smaller n will cause the system to move toward the stable equilibrium

5.10 Random Access: Stochastic Behavior and Dynamic Control Procedures

131

point. But a small perturbation toward larger n will cause the system to accelerate toward the saturation point where the system will stay for a very long time if let alone. Figure 5.19 shows a dynamic model that approximates the behavior of the system. The model is called the fluid approximation. In this model it is assumed that the sources are sending at a constant rate during different intervals. During such an interval where the rate of transmission is larger than the nominal rate, the number of backlogged sources will increase along the corresponding load line. When the traffic is again reduced, the number of backlogged sources will continue to increase or start decreasing depending upon the direction of flow along the new load line. Two examples are shown in the figure. In the first case (solid line), the channel is stable, but a series of fluctuations in the amount of fresh traffic takes it over the saturation point. When the traffic quiets down, the system will relax toward the stable point. In the second case (dashed line), the channel is unstable. If the traffic perturbation brings the channel into saturation, the channel may continue toward the saturation point after the load is relaxed. 5.10.2

Control Procedures

The control procedures for slotted Aloha first studied by Lam and Kleinrock in 1975 can be used to assess how these systems may be controlled in order to avoid saturation (Figure 5.20). This discussion suggests two methods by which the channel can be kept in the stable state. The most direct method is to reduce the average traffic per source or the Fresh traffic (packets/slot)

Unstable channel

Flow lines

Stable channel Number of packets in backlog The arrows indicate the direction of flow

Figure 5.19

Behavior of the channel: fluid model.

132

Multiple Access Input load (fresh traffic) packets/slot

Alternative 2: keep traffic load but increase K

Small K Large K

Reduced traffic Number of packets in backlog Alternative 1: keep K but reduce the input traffic

Figure 5.20

Control procedures.

number of sources in the system that are allowed to make calls in a certain period so that the new load line is entirely in the safe area. This is called the input control procedure (alternative one). The second method is to increase the randomization interval K for sources in backlog. This is called the retransmission control procedure (alternative two). The shape of the equilibrium curve depends on K as shown in the figure: if K is increased (or, equivalently, is reduced), the equilibrium contour is moved to the right increasing the stable region. The original load line may then be inside the stable region for its entire length. Of course, both methods may be used at the same time. The method is then called the input-retransmission control procedure. 5.10.3

Application of the Control Procedures

Where do we then find systems with traffic characteristics that resemble those of slotted Aloha? Here are a few examples: •

•

•

•

Telecommunications networks such as the Internet and ATM networks are likely to have traffic handling characteristic that includes unstable regions. The equilibrium contour is, of course, much more complex than that of slotted Aloha, but the study of the stability of Aloha may suggest some simple control procedures, some of which are actually in use. Mobile systems such as GSM and INMARSAT use slotted Aloha or pure Aloha on the access channel from mobile terminals. Ethernet and wireless LANs apply carrier sense techniques having traffic characteristics similar to those of slotted Aloha. Some systems used for distress communication are packet radio systems.

5.10 Random Access: Stochastic Behavior and Dynamic Control Procedures

133

The regulation in ATM is based on a contract between the user and the network specifying the operational range (total amount of cells per unit time, burstiness, burst length) allocated to the user. Cells not satisfying the contract are discarded by the network in case of congestion. One practicable method is to use the leaky bucket mechanism. In this case, the local switch of the user defines a rate at which tokens are created for the user in accordance with the traffic contract. The tokens are kept in a buffer (the bucket) until they are used. When an ATM cell arrives at the switch, one token is removed from the bucket. If the cell rate from the user is faster than the rate by which the bucket is filled, the bucket will become empty. Cells that arrive when the bucket is empty may then be discarded, thereby reducing the traffic load. The ATM regulation is thus an input control procedure reducing the offered traffic. A Swedish mobile messaging system called Mobitex (also implemented in Norway) offered a combination of public traffic and distress traffic. The system has now been taken out of use but is mentioned here because it made use of an efficient control procedure in order to support distress traffic and reject other traffic if the system becomes saturated. The method is simple: every mobile terminal has a digital identity assigned at the time of subscription of the service. The terminals listen to the control information disseminated by the base stations. The control information contains a bit field with the following function. If the first bit is binary 1, then all sources with a binary 1 as the first bit in the identification field are not allowed to access the channel. If the second bit in the control field is binary 1, then all terminals having 1 in the second position of the identity are denied access, and so on. In case of a rescue operation, the authorities may then deny access for all terminals not taking part in the operation, and in this way ensure that the system is not saturated by other traffic. The reason why the system was developed was that the telephone network cannot be used effectively if a major disaster takes place because then people start calling relatives and friends, and the network is likely to get overloaded with traffic. This happened after a huge landslide in Gothenburg around 1980, making the rescue operation difficult. The Swedish authorities then decided to develop the system. GSM is using a similar method based on a two-bit class mark allocated arbitrarily to the mobile terminals. The control procedure is such that in case of overload of the random access channels, one or more classes of mobile stations are not allowed to access the system until the situation has been resolved. TCP uses the combined method where both fresh and backlogged traffic are slowed down from individual sources. If TCP packets are not acknowledged within a given time, the source assumes that congestion has occurred and slows down the transmission of packets by first sending one packet and waiting for acknowledgment; if acknowledgment is received, two packets may be sent and so on, slowly building up the traffic. It is thus unlikely that the Internet will become saturated. However, the traffic capacity and the delay to send information may become unsatisfactory. Ethernet and WLAN use a retransmission control procedure where the K in the backlog queue increases exponentially whenever a source experiences collisions. The control is again acting on individual sources.

CHAPTER 6

Switching 6.1

Switched Networks 6.1.1

Terminology and Definitions

In this chapter we are concerned with how a communication path can be established between users for the purpose of transferring or exchanging information. The interchange of information between two or more users will be referred to as a call or a conversation in the rest of this chapter in order to simplify the text. Conversation and call are used synonymously; however, in some contexts it is more convenient to use only the term call. For example, the term call is generally used in order to describe actions such as call setup, call release, call routing, call management, and so on. The terms conversation and call are used for information exchange in all networks: Internet, mobile networks, telephone network, and local area networks. A call (or conversation) may consist of a single message sent from one user to one or several users, or the call (or conversation) may consist of several messages exchanged between two or more users. A user in the context of this chapter can be a person, a machine, or a computer program. The user is the source and/or sink of the information flow. If the user is a computer (e.g., personal computer, server, database), the term host is sometimes used. A connection-oriented network is a network in which a communication path (or connection) is first established between the users before information is sent. The communication path consists of a series of links interconnected by switching equipment. The communication path is retained as long as the conversation exists. When the conversation is terminated, the path is released and the different sections of the path can be reused for other calls. Note that a connection-oriented call may time-share the channel resources with several calls. See the definition of connection-oriented packet switching later in the chapter. A connectionless network is a network where every unit of information (or packet) is routed individually, even if the packets are part of a conversation defined at a higher protocol layer (e.g., TCP in the Internet). A call (or conversation) in a connectionless network consists only of a single information packet. Circuit switching is a connection-oriented technology. A circuit switched call occupies the connection for the entire duration of the call. The circuit switched exchanges interconnect links in tandem to form a single circuit between the endpoints on which information can flow. The telephone network is a circuit switch network.

135

136

Switching

The information transfer in packet-switched networks is organized in packets consisting of a signaling and management header followed by the data payload. Packet-switched networks may be connection-oriented or connectionless. In the connection-oriented mode, the same communication path is allocated to the call so long as the call exists but the call occupies the channel only when information is sent. The channel is usually time-shared with other calls. The terms virtual call, virtual circuit, or virtual connection designate a particular call, circuit or connection in such networks. In the connectionless mode, the packets are independent units of information not requiring a preestablished connection in order to be correctly routed. ATM is a connection-oriented packet switching technology while the Internet (IP) is connectionless. SS7 supports both connection-oriented operation and connectionless operation (see Section 7.8). The term exchange is often used as common name for switches in both connection-oriented and connectionless networks in this chapter in order to make the text easier to read. This is so even if the term router is commonly used for switches in the Internet. The term router is used if the text concerns the Internet only. The term signaling is used for the transfer and exchange of internal information between exchanges in both connection-oriented and connectionless networks. In connection-oriented networks, signaling is required for establishing, managing, and releasing connections. In connectionless networks, the signaling information required for the routing and management of the packet is contained in each packet. Signaling in connectionless networks thus consists only of information sent in the forward direction, while signaling in connection-oriented networks consists of information sent in both directions. Signaling is also used to describe the exchange of information between the user and the network (i.e., information exchange on the access). The Internet in the context of this chapter is only concerned with IP and control procedures that support routing and network management functions, such as open shortest path first (OSPF), Internet Control Message Protocol (ICMP), and Resource Reservation Protocol (RSVP). The full protocol hierarchy of data networks is described in Chapter 7. The acronym PSTN (public switched telephone network) is used to designate the telephone network (including the ISDN). 6.1.2

Switching Services

The foremost purpose of the switching platform is to route calls from one user to other users. The switching platforms in the telephone network, the Internet (IP), broadband networks, mobile networks, and local area networks offer essentially the same set of services. This set can be classified as follows: •

Connection services. The connection service connects the calling user with other users or resources in the network. Components of the connection services are routing the call to the destination; establishing, controlling, and releasing the connection; and providing call progress information such as called user free, called user busy, network engaged, network failure, and access not allowed. Ringing and busy tones are examples of call progress

6.1 Switched Networks

•

•

•

•

•

•

•

•

•

137

information in the PSTN. The connection services also include connection and management of conference calls, multicast services on the Internet, and management of the service components of multimedia calls. IP is connectionless, mainly offering routing services, while the PSTN is connection-oriented, offering all the services in the list. Broadcast and multicast services are particular connection services where one user may send messages to or have conversations with several other users simultaneously. This is a simplex service. Conference call is also a particular connection service where several users are taking part in the same duplex call. Quality of service (QoS) management. This may include selection of data rates, bit error performance, priority, flow control, management of terminal capabilities, delay management, real-time and differential delay performance, resilience, and provision of special selection rules such as minimum cost routing and minimum delay routing. Existing networks offer static QoS; that is, the parameters are design elements of the platform. Dynamic or user-controllable QoS is studied for future platforms. IP version 6 will offer such services per datagram. Supplementary services. This category contains mainly services available on the ISDN and GSM/UMTS; that is, on networks designed primarily for speech services. However, some of these services must also be implemented for voice over IP. Examples of supplementary services are alternate routing, call forwarding, multiple addressing, hunting services, universal address, call barring services, unavailable services, and completion of call services. Charging services. Examples are toll-free services, shared charges, premium rate services, fixed selection of charged party, and conditional selection of charged party. These services are also mostly associated with telephone-type applications where charges are levied on a per call basis. Security services. This may include authentication, authorization and access control, confidentiality, integrity protection, nonrepudiation, untraceability, anonymity, and protection against fraud and malicious attacks. Mobility services. Possibilities are terminal mobility, personal mobility, session and application mobility, and mobility sensitive services (e.g., fleet management based on location or tracing of sporting dogs). Translation services. These are services required for interconnecting networks designed with different technologies. The translation service may comprise translation of protocols, translation of signaling messages, rate adaptation, and translation of message formats (e.g., suppressing the transfer of still pictures and video when Web pages are transferred over narrowband GSM accesses and interworking between voice-over-IP and the ordinary telephone network). Management services. These services include traffic measurement and charging management, retrieval of statistics, fault management, change management, and restoration management.

138

Switching

Historically, most of these services have been offered by the switched telephone network because the terminals connected to these networks have no (or almost no) processing capability. In IP-based networks the situation is the opposite: the terminals have much processing capability while the network has virtually none. However, which services and solutions will emerge in the network when the IP technology takes over more and more of the traffic of the telephone network are still to be seen. 6.1.3

Circuit Switching

The switching process for a circuit switched system (ISDN, PSTN) is illustrated in Figure 6.1. The switch connects on demand the input where the call exists to an output in the direction of the called user. This connection is kept for the duration of the call and is disconnected when the call is terminated. Information exchanged between the interconnected users will then flow along exactly the same path for the duration of the call. If the users are not connected to the same switch, the connection will consist of several exchanges in tandem, as shown in the figure. The switches are interconnected by transport sections containing multiplexers and transmission systems such as cables, fiber, satellites, and radio relays. Two calls between the same users initiated at different times may follow different routes; that is, they may pass through different transit exchanges, where the actual route is selected on the basis of measured traffic load (congestion) and other information concerning the operating status of the network. This can be done

Switch Circuit switched platform

Access

Transport

Access

Figure 6.1

Switching process.

6.1 Switched Networks

139

because the network is a mesh network where several routes will exist between the different nodes (exchanges) in the mesh. The route selection is based on a preprogrammed algorithm. The processing of call setup in the PSTN is subject to strict timing constraints. The reason is psychological rather than technical. The call setup must be completed—that is, the ringing tone must be heard—on the order of a few seconds after the called number is presented to the network; otherwise, the user may become impatient and clear the call prematurely. The call setup involves processing in several exchanges in tandem, allowing a processing time of the order of 200 ms for each exchange. This time also includes the time required for actuating the switching electronics. The switching equipment in the PSTN must also process many calls at the same time (several thousand). The type of programs, the initiation data, and the order in which the programs are executed are often different for separate telephone calls, and synchronization between the processing events of different calls in the exchange is not possible. This implies that a separate asynchronous program must be initiated for every call. The timing of the programs is illustrated in Figure 6.2. The computers in the switching platform must therefore be equipped with timesharing capabilities. Timesharing means that several computer programs are executed at the same time either on different CPUs or in quasi-parallel on one computer. Quasi-parallelism means that the computer allocates small timeslots for each program and executes parts of the program in each such slot. At the end of the slot, the state of the computation and preliminary results are stored and made ready for further processing at the next timeslot. The processing time is then approximately equal for each program with the same number of execution events. Real-time performance has to do with how quickly the programs must be executed. Digital coding of the speech samples of a telephone channel requires strict timing. Exchanges in the PSTN support real-time performance. This is also the case if transcoding of the speech signal takes place in the network (e.g., between PCM and compactly encoded speech in mobile networks). This is the major problem of voice-over-IP (VoIP) since the Internet cannot guarantee the same delay for all packets belonging to a VoIP conversation. The switching process also requires distributed processing. The call in connection-oriented systems is established over several switches, where all the Allocation

P1

P2

P3

Time

Figure 6.2

Timesharing of processes.

140

Switching

switches process the call in parallel; that is, the execution at the different exchanges overlaps in time. The exchanges are not only processing call setup and release but a number of other tasks related to operation and management of the network so that each exchange must contain an array of computers. The execution of a single task is then usually distributed over several of these computers requiring complex software synchronization of the computers. Timesharing and real-time requirements makes the distributed processing extremely complex. 6.1.3.1

Example: Simple Switched Network

The basic features of distributed, parallel processing in telecommunications are illustrated by a simple example. Figure 6.3 shows a simple two-party telephone call. The system consists of three exchanges (outgoing exchange, transit exchange, and incoming exchange). For simplicity, only four processes at each exchange are included: (1) user access control, (2) interexchange control, (3) number analysis and route selection, and (4) switching. Practical systems are, of course, more complex, but the example contains just those features that are required for explaining the nature of the distributed processing involved. Any two exchanges, independent of age, technology, or manufacturer, can be interconnected provided they satisfy the same interface conditions and a set of rules applying to operations invoked on them. In the figure, two types of interface are indicated, I1 between the terminal and a local exchange (user interface) and I2 between two exchanges (interoffice interface). All that is required for interconnecting exchanges in the same or in different networks is to implement the rules that apply at interface I2. This requirement is independent of how the networks are designed or implemented. The purpose of standardization is to ensure that this is possible. Standardization of interface I1 ensures that terminal equipment designed by different manufacturers can interface the exchange. Process 1 controls the subscriber access in the outgoing exchange. It is invoked when an outgoing call request is received from the user. The user provides the address of the called user. The address is passed to process 3 (number analysis) in Interoffice interface I2 Signaling

User interface I1

1

3

2′

2

4

Outgoing exchange

3′

4′

Transit exchange Information

Figure 6.3

Interoffice interface I2

Simple model of a two-party call.

User interface I1

2″

3″

1″

4″

Incoming exchange

6.1 Switched Networks

141

order to determine which outlet or route is selected. If there is no such route, the processing will stop at this point and the remaining processes will not be invoked. Processes 1″ and 3″ represent similar functions in the incoming exchange. Process 3″ determines which outlet corresponds to the address of the called user, and process 1″ alerts the user about the incoming call. The number analysis process consists essentially of a table relating address information received and the identity of the outlet lines that can be selected for forwarding the call. Processes 2, 2′, 2″, 4, 4′, and 4″ are essentially the same in all three exchanges and perform all functions required for interconnecting the exchanges. Processes 2, 2′, and 2″ send and receive signaling information such as addressing and management information. Processes 4, 4′, and 4″ are the switching processes connecting an input line with the required output line. Figure 6.4 illustrates how the processes interact through exchange of messages. Some of these interactions take place across the external interfaces I1 and I2 (e.g., between calling user and process 1, between process 2 and process 2′, and between process 1″ and the called user). These interactions are called signaling in connection-oriented networks (see Chapter 7). Connectionless networks such as the Internet contain equivalent functions for passing information along the forward path of the call. Signaling—including the equivalent functions of the Internet—across external interfaces must be designed in accordance with a standard developed for that purpose. Equipment produced by different manufacturers and installed in different networks cannot otherwise cooperate. The standards may be developed by one of the international

I1 Call

1

3

2

4

I2

2’

3’

4’

1’’

I2

2’’

3’’

Find route Call

Establish

Find route Establish

Find route Check

Address complete Address complete Connect Connect Connected Connected Alert Alerting

Ringing

Answer Answered

Release

Answer

Answered

Answer

Connect

Release Disconnect Release Disconnect

Release Release

Figure 6.4

Lifetime and concurrency of processes.

Disconnect

4’’

I1

142

Switching

standardization bodies (ITU, ISO, ETSI), by a single manufacturer (IBM, Sun Microsystems, Netscape), or by independent organizations (ECMA, IEEE, ACM). Interactions taking place over internal interfaces in an exchange need not follow any standard. Sometimes it is useful to standardize such interfaces also because the equipment may consist of parts produced by different manufacturers. 6.1.4

Connection-Oriented Packet Switching

Connection-oriented packet switching is based on essentially the same mechanisms as circuit switching. The major difference is the way in which the links making up the end-to-end connection are shared between several calls that overlap in time. While a circuit switched call occupies the links alone, the links are shared by several simultaneous packet switched calls. Let us use ATM switching at the cell level as an example to explain how connection-oriented packet switching operates. The ATM cell is shown in Figure 6.5. The cell consists of a header containing 5 octets and a payload of 48 octets. The total size of the cell is thus 53 octets. The fields of the header are as follows. The virtual path identifier (VPI) and virtual channel identifier (VCI) are address fields identifying the connection at each switch. At the switch, all cells belonging to the same call have the same VPI and VCI. The combination of VPI and VCI allocated to a call is known as the channel identifier (CI). The CI thus consists of the pair (VPI, VCI). It is the CI that is significant for the switching process. 1

2

3

4

5

6

7

8 VPI = virtual path identifier (12 bits) VCI = virtual channel identifier (16 bits) PTI = payload type identifier (3 bits) C = cell loss priority (1 bit) HEC = header error control (8 bits)

1 VPI 2

3

VCI

4

5

PTI

HEC

6

.

.

Payload (48 octets)

.

53

Figure 6.5

ATM cell.

C

6.1 Switched Networks

143

The payload type identifier (PTI) is used for congestion control and to distinguish between cells used for operation and maintenance and cells carrying normal traffic. The cell loss priority (C) is also used for congestion control by indicating which cells can be discarded when congestion arises. The header error control has two purposes: assisting in cell synchronization as explained in Chapter 3 and correcting bit errors appearing in the header to avoid, in particular, misrouting of cells. Figure 6.6 shows a switched ATM network consisting of three exchanges and five terminals. The connections in the network are established by a common channel signaling system prior to establishing the ATM connections. The signaling system is also interconnected via the ATM switches using higher layer protocols. So-called permanent virtual connections (PVC) are used for this purpose.1 The CIs of these connections are permanently assigned by administrative procedures, and the exchanges can then easily distinguish between signaling cells and cells containing ordinary traffic, since the signaling cells will always appear on the same PVC. A connection identifier on the connection carrying ordinary traffic between two exchanges or between the user and the exchange is assigned to each call as part of the connection establishment procedure. All cells belonging to the same call have the same CI. Different calls on the same connection must have different CIs. After the call has been terminated, the CI may be assigned to another call. The CI is only significant on a particular connection. The exchange contains a routing table that equates the CI on the input line to the CI on the output line, as shown in the figure. When a cell arrives at an exchange, the exchange will identify on which port the cell is received and then read the CI of the cell. The port on which the cell is to be forwarded and the new CI to be inserted in the cell header are then read directly from the routing table.

Signaling (32,77)

A

(36,56)

1

8

ATM 7 switch 1 3

(6,98)

(11,11) 2

8

(1,23)

1

(6,98)

(4,12)

B

(34,17)

(18,27)

(78,5) C

ATM 9 switch 2

(13,98) 4 D

(4,12),7 (11,11),3

(11,11),8 (78,5),9

(6,98),8

(1,23),7

(13,98),4 (6,98),2

...

... Switch 1

Figure 6.6 1.

2 (18,27),1

(36,56),1 (18,27),8

...

(11,45) ATM 11 switch 3

(32,77),11

(34,17),11

(6,98),1

(6,98),2

(11,45),11

...

... Switch 3

... Switch 2

Switched ATM network.

PVCs may also be used for other purposes such as network administration and management. Virtual circuits assigned on demand are called switched virtual circuits (SVCs).

E

144

Switching

6.1.5

Connectionless Packet Switching

The switching process in connectionless networks is simpler. With the Internet, routing and switching take place at the IP layer, and the switching device is usually called a router. The transport protocols (TCP and UDP) are not involved in the switching process and the content of the TCP/UDP headers are not analyzed by the routers except the fields containing the port addresses. These addresses are used for address extension in IPv4 networks using a method called network address translation (see any standard textbook on Internet protocols for details). By definition, TCP and UDP are there to support proper operation of the end systems, such as ensuring delivery of all data in correct order and without duplication. However, there are cases, for example, in mobile systems, where particular network units are handling the TCP/UDP packets but such cases are exceptions (see Section 8.4.7). The switching process is as follows. Each datagram is a single call so that a connection is not established at the IP layer. When an IP datagram arrives at the input of a router, the router places the datagram in an input buffer, reads the destination address and from that determines on which output port the datagram is to be forwarded, and then places the datagram in an output buffer ready to be forwarded when the selected output becomes idle. In IPv6, the router may also use additional information (addresses of intermediate routers and real-time priority information) to support real-time operation and select the same route for a succession of packets with the same real-time priority. This is done for speech and video packets, since presentation of these services at the decoder is critically dependent on real-time operation. IP offers no guarantee that datagrams are actually delivered. To overcome this problem in certain computer applications, TCP supports end-to-end connectionoriented operation offering services such as acknowledgment, recovery of lost information, and rejection of duplicated information. 6.1.6

General System Requirements

Figure 6.4 illustrates that, for a connection-oriented call, the various processes in the exchange run in parallel and have different lifetimes. In connection-oriented networks, interacting processes run in parallel on several exchanges. Connectionless switches are autonomous and the switching process is simpler, as they do not require parallel processing of several routers. The processes in connection-oriented networks are described next, since this is the most complex case. Let us use Figure 6.4 as a basis for describing what goes on in the connectionoriented switch. A switch for connectionless services is not much different. Some processes are required in order to perform a single task of short duration, such as number analysis (process 3). Other processes are running for most of the time the call exists (processes 1, 2, and 3). Most of the time these processes are in a waiting state; that is, they are not actively processing but waiting to be invoked for processing new tasks. One example of this type of process is the switching process. This process is active when the call is through-connected in the exchange. Activities taking place in the switching process are: selecting the path through the exchange, marking the input circuit and the selected output circuit as busy, actuating the switching

6.1 Switched Networks

145

devices, and updating the tables containing switching path information. After these activities have been completed, the process waits for the command to release the connection. Release activities include disconnecting the switching path, removing the path from switching path tables, and marking the corresponding input and output circuits as idle. Some of the processes terminate by themselves when they have finished their task (processes 2, 2′, 2″, 3, 3′, and 3″), while other processes terminate on command by other processes (processes 4, 4′, and 4″). Each process is started either by another process or by some external stimulus (e.g., the user answering the call). Another observation, not apparent from the figures, is that many users are connected to each exchange so that several independent calls may exist at the same time. This implies that many of the processes described earlier are executed simultaneously. These processes are not synchronized to one another; that is, independent copies of the software are initiated with different data at arbitrary times so that many such copies are running at the same time. The way each copy of the software is executed is different each time the process is executed, since the program execution path often depends on initiation data and events occurring during execution. Moreover, each program is usually a component of a bigger construct, which may have different composition whenever a task is performed. Different processes may also have different execution priorities, and there may be various constraints on execution time depending on initiation data and call event. This shows that the switching software is complex, requiring sophisticated and complex algorithms for managing priority, queuing, and timesharing. In summary, both connection-oriented and connectionless telecommunication systems must meet some basic requirements: •

•

•

•

•

Concurrency. Telecommunications applications require processing of concurrent processes, both within one machine (or a complex of machines at one site) and in machines at different locations. This is the only way an exchange and a network can handle thousands of calls at the same time. Weak coupling. Only a small number of messages are exchanged between any two pairs of processes within an exchange or between exchanges. The coupling between the processes is therefore weak. In Figure 6.4, few messages are exchanged between each process. Information hiding. The interchange of messages between processes within an exchange and between exchanges requires very little knowledge of how the processes are actually designed. This means that the interchange in most cases is based on a small but consistent set of knowledge about the cooperating process (the interface specification). In Figure 6.4, only the syntax and semantics of the messages passing the external interface (I1 and I2) need to be known for designing systems that can cooperate. Weak synchronization. The processes are weakly synchronized in the sense that a process has no a priori knowledge of when and with which initiation parameters it will be invoked by another process. Therefore, the process must be allowed a wide range of processing times depending on context. Fault recovery. Each process must be able to recover from faults on its own. If a process fails, all processes depending on it must be able to return to a safe

146

Switching

state independent of the faulty process. If, for example, process 2′ in Figure 6.4 fails and the connection across interface I2 disappears, then process 2 must be able to terminate the call and stop processes 1 and 4 without awaiting further instructions. All these items represent different aspects of complex systems. They all contribute strongly to dependability, fault tolerant behavior, and flexibility in design. Weak coupling ensures that there are few design constraints imposed by cooperation and that it is rather easy to understand (and thus test) the behavior of a cluster of processes. In cases where strong coupling exists, the design should be reconsidered in order to find a more appropriate decomposition into processes where the principle of weak coupling is retained. This may involve placing strongly interacting software components in the same process. Weak synchronization and autonomous fault recovery ensure independent processing. If a process or machine is overloaded, it is often better that it ignores new requests and leaves it to the requesting process to detect and act on lack of response. Similarly, if an acknowledgment that a message has been received is not strictly required, it should not be sent in order to reduce processing load. When specifying and designing telecommunications systems, much work is put in weakening synchronization, removing unnecessary acknowledgments, and defining procedures for autonomous fault handling. This is probably the most important—and usually the most complex—activity when designing a system. This was, for example, an important issue when designing the network protocol of GSM called the mobile application part (MAP). Information hiding is a deed of necessity in telecommunications in order to allow cooperation between networks operated by different network operators. Within each network, there is equipment designed by different manufactures at different times and built with different technologies. The networks are also subject to a continuous evolution in order to meet demands for new services, reduce running costs, improve performance, and increase the efficiency of operation and administration. Telecommunications systems require, therefore, well-defined interfaces that are independent of current technology, actual design, and network structure. Information hiding is the most important mechanism to handle these types of heterogeneity. 6.1.7

Number Analysis and Routing

We distinguish between geographic and nongeographic numbers. The information contained in a geographic number is the location of the network access point (NAP) where the user is physically connected to the network. The telephone number of a fixed telephone or the IP number (possibly plus extension information in IPv4 networks) is a geographic number that identifies the subscriber interface module in the exchange belonging to that user. Most IP addresses are geographic numbers because they designate a fixed location of a user or a local network to which the user is connected. The general format of IP addresses consists of a network identity followed by the host identity. Sometimes the host identity refers to a single computer; sometimes the identity refers to an

6.1 Switched Networks

147

ISP using the network address translation facility for identifying the actual host; and sometimes it is associated with an organization or company that may be distributed over a large geographic area. In the latter case, the call is routed to the physical destination of the user on a local area network2 of that organization or company. This is not the same as mobile IP, though similar functionality may be used in order to support the internal routing. Mobile IP is described in Section 8.4.7. In nongeographic numbers there is no direct association between NAP and number. Examples of nongeographic numbers are numbers used for toll-free services,3 shared rate services, premium rate services, and mobile users. These numbers contains two types of information elements where one element identifies the type of service and the other element is a unique identification of the owner of the nongeographic number. The number usually does not contain any information about the network access point to which the user is physically connected. Let us first describe how routing to geographic numbers are done. Routing based on nongeographic numbers is described toward the end of this section. Routing based on geographic numbers in the PSTN is illustrated in Figure 6.7. The called number, xxxyyyzzz, is presented to the outgoing exchange by the user initiating the call. In the example, the number consists of three fields of digits (xxx, yyy, and zzz) having different significance at different places in the network. Telephone and IP numbers with a geographical structure is designed according this principle (but may contain more than three fields). The table in the outgoing exchange uses part of the address (xxx) to determine the identity of the transit exchange to which the call shall be forwarded. The exchange also knows on which line the user initiating the call is connected (l1). In some systems, the input line is determined by monitoring the state of the access line (electric current is or is not flowing); in other systems (ISDN), the state is signaled in a separate signaling system together with the number xxxyyyzzz. The number analysis table of the outgoing exchange contains the mapping function of the digits xxx into the identity of the group of lines (g1) on which the call can be forwarded. The outgoing exchange chooses one arbitrary idle line in this group and signals forward which line has been chosen. The called number or part of it is also forwarded. (If only part of the number is forwarded, such as in the PSTN, the routing table is preprogrammed to know the number of digits to be forwarded.) The identity of the input line and the identity 2.

3.

These networks are sometimes referred to as private wide area networks, since the private network may be implemented on top of several public networks using leased lines and switched connections to interconnect the different parts of the private network. To be pedantic, a local area network designates a local area network at one site. The term metropolitan area networks (MAN) designates, as the name suggests, an intermediate size network. Most often all these solutions are referred to as LANs for simplicity. The distinction between them is only important when it comes to the technical realization of them. A small LAN may use Ethernet or WLAN technology; a MAN may be realized as an optical ring network; a WAN will require more complex resources. At first, toll-free numbers and premium rate numbers were actually geographic numbers. During the early 1990s, the United States ran out of geographic numbers for such applications. This led to the development of the intelligent network with the main purpose to translate nongeographic numbers to geographic ones. The use of nongeographic numbers for mobile applications goes back to about 1980 when the first cellular systems commenced operation (NMT for land mobile services and MARISAT for satellite communications with ships).

148

Switching

Transit exchange

Outgoing exchange

Incoming exchange

l1 xxx(yyyzzz) g1

l2 ( xxx)yyy(zzz) g2

l3 ( xxxyyy)zzz l4

Information transferred

xxxyyyzzz = called number

Switching relationship

li = line identification

gi = group/line identification

Figure 6.7

Number analysis.

of the output line (l1 and l2) are the information needed by the switching process of the outgoing exchange to select the path through the exchange. The basic functions of the routing table are thus to identify an appropriate output line from the called number and to provide the identities of input and output lines to the switching fabric for selecting an appropriate path through the exchange. The exchange may also perform other functions, such as identifying the signaling circuit to the next exchange and restrictions related to the call (barring of certain numbers, cost constraints, and so on). The transit exchange uses again part of the number (yyy) to determine on which group of output lines the call can be forwarded. The process is identical to that of the outgoing exchange. The incoming exchange also receives the called number (or part of it). Part of the number (zzz) determines the unique line (l4) connected to the called user. The routing table at the destination exchange may also be associated with processes determining call restrictions and particular call handling procedures that apply to the call (e.g., forwarding). The PSTN is designed such that a shortest possible route will be chosen for each call. Since the telephone network is a mesh network, a call can also reach the destination via other routes. The possible route that can be chosen is based on a prioritized selection scheme where an exchange selects a route with lower priority only if the routes with higher priority cannot be chosen because of traffic load. Routing is simpler in the Internet. The routers compares the number, or part of it, with the routing table and selects in which output buffer the call is to be placed. Sometimes the output buffer is chosen almost arbitrary because the router actually has no information that can be used in the routing decision. Therefore, a datagram may pass a large number of routers before it eventually is closing in on the target.

6.1 Switched Networks

149

Sometimes the target is never reached and the datagram is simply dropped because it has passed too many routers (usually 256). Routing based on nongeographic numbers in the telephone network is illustrated in Figure 6.8. A geographic telephone number is assigned to every network access point in the telephone network. If the call contains this number, the call will be routed to the access point in the normal way. If, on the other hand, the call contains a nongeographic number, the network must first translate this number to the corresponding geographic number of the access point. The nongeographic number and the corresponding geographic number (e.g., a toll-free number) are contained in a database that the network accesses in order to obtain the geographic number of the access point. Such networks are called intelligent networks (INs). The first intelligent networks were implemented in the early 1990s (1991 in Norway, one of the first INs to be built). The network routes a toll-free number to an exchange that is able to retrieve the physical address from the database as shown in the figure. The call can then be routed forward in the usual manner. Since the call is handled by a separate database, several other features can be implemented, such as the following: • • •

•

•

Routing the call to different physical numbers depending upon time and date; Queue management and priority handling toward busy numbers; Distribution of traffic to, for example, call centers equalizing the load on the operators; Automatic generation of additional information (for example, number information) provided in terms of short messaging service (SMS) messages or e-mails; Managing conference calls (adding new participants, removing participants, forming subgroups, putting groups of participants on hold, and so on).

SCP

Data structure toll-free number → physical number

Routing information

Provide routing information

SSP

Physical number

Figure 6.8

Intelligent network.

150

Switching

The database in intelligent networks is called the service control point (SCP), while an exchange capable of accessing the SCP is called a service switching point (SSP). The telephone network of one telecommunications operator may contain several SSPs but only a single SCP handling all calls requiring number translation or other particular treatment. The SCP is thus a common resource that may control the switching process in a number of exchanges. The routing tables in the telephone network and in ATM are updated by administrative means. In the Internet, this is also done, but in addition the routers use the OSPF protocol to maintain an updated list of adjacent routers. In addition, a number of other protocols are available for routing management (for example, protocols for distance metrics measurements and identification of intermediate and end systems—see any textbook on the Internet for details). 6.1.8

Signaling Systems

6.1.8.1

Connection-Oriented Networks

The purpose of the signaling system is to pass information between exchanges. Such information may be as follows: •

•

Forward signals. This is information required for establishing or releasing the connection. Forward signals are also sometimes used to provide additional information during the call, such as changing the call state (e.g., suspension, reconnection) or altering call conditions (e.g., charging rate). Backward signals. This is information, such as call progress signals (e.g., user free, user busy, congestion, subscriber line faulty), answer signal, or backward release signal. Backward signals may also be used to indicate particular changes of the call state during the conversation.

The signaling procedure is exemplified in Figure 6.9. The example shows a simple telephone call between users connected to analog subscriber lines. The procedure is as follows. When the calling subscriber goes off-hook, the event is detected by the exchange (current starts flowing toward the user’s telephone) that returns the dial tone, inviting the user to present the called number. The number (or sometimes just a part of it) is forwarded to the next exchange in a message called the initial address message (IAM). When this message reaches the destination, the terminating exchange will return the address complete message (ACM) containing the status of the called number (e.g., user free, user busy, line out of order, no user with this number, or access to this user is barred). In all other cases than the user free indication, the call is terminated at this point without further signaling taking place. If the user is free, the terminating exchange will start alerting the called user and the outgoing exchange will send the ringing signal to the calling user. The off-hook condition of the called user is indicated by the answer signal. At this point all the exchanges through-connects the call. The call is terminated by one of the users going on-hook. The release guard signal acknowledges that the circuit has been released and can be reused for another call.

6.1 Switched Networks

151

User

User Off-hook Dial tone Digits

IAM

IAM ACM

ACM Ringing

Answer

Answer

T

T

Alerting T

Off-hook

Conversation On-hook

Forward release

Forward release

Release guard R

Release guard R

On-hook R

IAM = initial address message ACM = address complete message T = through-connect R = released

Figure 6.9

Simple signaling.

Signaling system no. 7 is described in Section 7.8. Signaling in land mobile systems is contained Chapter 8. 6.2.2

Connectionless Networks

In IP, all information required is contained in the header of the IP datagram. IP does not contain backward messages; all signaling information flows in the forward direction. The IP header contains the following signaling information that is used by routers in order to forward the call: • •

• •

IP addresses of the source and the destination. Time-to-live parameter indicating the number of routers the datagram has passed (maximum value 255). The parameter is incremented by one by each router the datagram passes. If the time-to-live parameter of a datagram equals the maximum number, the datagram is discarded, thereby preventing the datagram from circulating indefinitely in the network. Optional QoS parameters, such as priority, delay, and reliability. Optional information concerning preferred routing and real-time preference, and information that can be used for statistical purposes.

The other information that the header contains is concerned with the handling of the datagram at the destination—information related to fragmentation, type of

152

Switching

information (protocol) contained in the payload, length of datagram, and error check.

6.2

Switching Technologies 6.2.1

Introduction

Historically, the switching devices used in the telephone network were designed to mimic the way manual operators connected calls: connecting the plug on which an incoming call exists to the output jack of the called user line. The switches developed by Strowger (1896) and L. M. Ericsson consisted of a rotating dial containing the plug of the input lead and a fixed circular segment containing the output jacks. The action of the device is to rotate the plug to the position of the required output jack and insert it in the jack. The angle of rotation is determined from the called number, and the rotation is actuated by an ingenious combination of relays, gears, brakes, and clutches. After almost 40 years, the rotary switch was replaced by the more efficient electromechanical crossbar matrix. The electromechanical switches were the dominant design for more than 70 years. Most electromechanical switches were replaced by electronic exchanges during the 1970s, when the electromechanical switching matrix was replaced by computer-controlled electronic matrices. The principle is called space-division switching because the connections represent different points in space. In time-division switching, the channel is moved from one timeslot to another in the same multiplexed signal. Digital networks apply a combination of space-division switching and time-division switching. 6.2.2 6.2.2.1

Space-Division Switching: Crossbar Switches Basic Principle

Space-division switching using crossbar switching matrix is illustrated in Figure 6.10. The space-division switch is used for switching individual channels at 64 Kbps, or multiplexed signals at 2 Mbps, or any other multiplex rate of the PCM or optical hierarchies. However, the same bit rate is used on all circuits in the matrix. Otherwise, it is not possible to connect an arbitrary input to an arbitrary output. The figure shows a matrix with 4 input circuits and 6 output circuits. Every input circuit can be connected to every output circuit. A space switch can have any number of input circuits and any number of output circuits. The interconnection takes place where the input and the output lines cross each other. An example where input 1 is connected to output 5 and input 4 is connected to output 1 is shown. Control signals determine which input is connected to which output. The value of the control signal is derived from the result of the number analysis function described earlier. 6.2.2.2

Number of Bits in the Control Signal

The most compact method of expressing which input circuit is connected to which output circuit is the following. If the matrix has n input circuits, we need log2 n bits

6.2 Switching Technologies

153

1

Input circuits 2 3

4

Control circuits

2 3 4

Output circuits

Control signals

1

5 6

Switching matrix

Figure 6.10

Space-division switching.

to enumerate all the input circuits, where log2 is the logarithm to base 2. The symbol x defines a number that is equal to x if x is an integer or is equal to the smallest integer greater than x if x is not an integer (the universal rounding up operator). This is the number of bits required to enumerate all the input circuits that cross a given output circuit. If the number of output circuits is m, the total number of bits required to enumerate all cross-points in the matrix is then m log2 n. In Figure 6.10 there are 4 input circuits and 6 output circuits so that the total number of bits is 6 × log2 4 = 12. If we have a matrix consisting of 1,024 input circuits and 1,024 output circuits, the number of bits required is 1,024 × log2 1,024 = 10,240 bits. Even for very large switching matrices, this is a rather small number of bits, requiring only modest memory capacity. 6.2.2.3

Crossbar Matrix Configuration

Figure 6.11 shows one way in which the crossbar matrix can be constructed. Each cross-point consists of an AND gate (& in the figure) with the input circuit and the control signal as input to the gate. If the control signal at one cross-point is set to logical 1, the gate is open and the digital signal on the vertical circuit (the input circuit) is routed to the horizontal circuit (the output circuit). The control signals on the other gates on the same horizontal circuit are set to logical 0, ensuring that the gate is closed for signals from the other input circuits since one output circuit can carry the signal from only one input circuit. Note that, on the other hand, the same input signal can be routed to several output circuits, allowing broadcast, multicast,

154

Switching Input circuits 1

2

3

4

S/P converter &

&

Control signals

Channel configuration information

Decoder

&

&

1 0 0 0

&

&

Figure 6.11

&

&

&

&

&

&

&

&

&

&

&

&

+

1

+

2

+

3

+

4

+

5

+

6

&

&

&

Output circuits

0 0 0 1

&

&

&

Realization of switching matrix.

and conference services. The encircled gates in the figure are open so that the input signal on input circuit 1 is routed to output circuit 5, and the input signal on input circuit 4 is route to output circuit 1. The decoder receives the binary number (channel configuration information) defining all the cross-points in the matrix and extracts the information concerning the state (connected or not connected) of each output circuit encoded as explained earlier. Whenever there is a change in the state of the matrix (that is, which input channel is connected to which output circuit), the decoder is updated with the new information. Based on the channel configuration number, the decoder and the series-to-parallel (S/P) converter provide a unique bit to the AND gate at each cross-point such that binary 1 indicates that the input and output circuits are connected at that cross-point while binary 0 indicates that the circuits are not connected. 6.2.2.4

Switching of Individual Channels of Time-Division Multiplexed Signal

The configuration is shown in Figure 6.12. The circles indicate which input circuits are connected to which output circuits at a given moment. The numbers in the circles show which time slot of the input circuit is passed to the indicated output circuit. Timeslot 1 of input circuit 1 is sent to output circuit 1, timeslot 3 to output circuit 2, and timeslots 2 and 4 to output circuit 3. Input circuit 1 is not connected to output circuit 4. The figure does not show which input timeslot is mapped onto which

6.2 Switching Technologies

155

output timeslot. This is shown in Table 6.1, containing a snapshot of the complete switching arrangement. The number of control bits required to indicate all paths through the switching matrix has now increased to rm log2 n , where r is the number of timeslots of each Input circuits

S/P converter

4 2

3

4

4 3

3

4

3

3

2

2

2

2

1

1

1

1 4

& &

1

& &

Control signals

4

2

1

2

1

2

& &

3

4

& &

4

3

&

1

4

& &

+

&

4

3

+

&

2

&

2,4

2

Figure 6.12

3

+

1

2,4

1

& 1,3

3

2 1

+

&

3

3

Space-division switching of multiplexed signal.

Table 6.1

Switching Table

Input Channel

Input Timeslot

Output Channel

Output Timeslot

1

1 2 3 4

1 3 2 3

4 1 1 2

2

1 2 3 4

2 3 1 4

2 3 3 1

3

1 2 3 4

3 4 2 4

4 2 3 4

4

1 2 3 4

1 4 1 2

2 3 1 4

3

2 4

1

Output circuits

4 1

156

Switching

input and output system, m is the number of output circuits, n is the number of input circuits, and x is the universal rounding up operator defined earlier. The number of control bits required for a 1,024 × 1,024 space matrix switching 2-Mbps bit streams is 327,680 bits since each 2-Mbps circuit contains 32 64-Kbps channels (r = 32). Even this is a small number for a computer with regard to both storage capacity and database search capability. However, in the old electro-mechanic switches, the switching path information had to be stored in electro-mechanical relays, where each relay could store only one bit. The part of the exchange computing and storing the switching-path information was huge. The control signals indicate the switching path with regard to circuit and timeslot. The output units marked with the + sign in Figure 6.12 are the storage units and the multiplexers required for time alignment of the 64-Kbps input and output channels. 6.2.3

Space-Division Switches Using Buffers for Cross-Connect

The switching arrangement is shown in Figure 6.13. This arrangement is only possible in packet switched networks. The switch in Figure 6.13 contains a buffer on each input port and a buffer on each output port. A packet arriving at the input circuit is placed in the buffer. If the input buffer is full when the packet arrives, the packet is lost. If the packet is accepted, the switch reads the destination address of the packet and from this information decides in which output buffer the packet shall be placed. The packet is transferred to the correct output buffer when there is an empty position in that buffer. The memory management function decides which packet from which input buffer shall be transferred to the output buffer in case of contention. The algorithm may be based on the arrival time of the packets at the switch or an arbitrary choice function. The algorithm may also depend on the priority of the packet: a new packet will be placed before packets with lower priority but after packets with the same priority that arrived at the switch at an earlier time. Packets are sent to the output port as soon as the port becomes idle. In the Internet, each buffer location must be capable of storing packets of different length so that both the beginning and the end of the packet must be recognized for proper buffer management. In ATM, all buffer locations are equally long (one cell). The queuing discipline applied to the buffer is FIFO, possibly with priority management. This means that packets with the same priority are handled in the same order as they arrive at the queue. Packets with lower priority are handled after all packets with higher priority have been handled. The switching table takes care of the management of the different buffers. The input queue is strictly not needed. It suffices to implement one buffer at each output port. When a packet arrives at the switch, the packet can be placed in the correct output buffer directly after number analysis has been completed. The packet is lost if the output buffer on which the packet is to be forwarded is full when the packet arrives.

6.2 Switching Technologies

157

Switching table and memory management

Input 1

Output 1

Input 2

Output 2

Input 3

Output 3 Memory

From

To

Input

Packet

Output

Packet

1

5

3

5

3

7

1

1

1

2

2

4

2

2

1

3

Table

Figure 6.13

Switching by use of buffers.

A more sophisticated principle is to use a single buffer. Such switching architectures are often referred to as shared memory systems. All packets that arrive at the switch are placed in linked lists, one list for each output port. When the port is idle, it reads the location of the next packet in the memory from the linked list associated with that port, sends the packet, and marks the memory position as idle. A new packet is linked to the appropriate list when it arrives. 6.2.4

Time-Division Switching

Time-division switching consists of rearranging the timeslots of a multiplexed signal as shown in Figure 6.14. In the example, a frame consisting of four timeslots (or information channels) is switched. Time-division switching can be done on any type of time division multiplexed signal of one of the standard hierarchies (or of a nonstandard hierarchy used in a particular switch configuration if this is cheaper for the manufacturer). The principle is as follows. The frames of the input signal are stored in a memory in the sequence in which they are received. The output port reads the input signals from the memory in the order indicated by the switching instruction. In the example, output channel 1 reads the content of input channel 4, output channel 2

158

Switching Frame

4

3

Frame

2

1

4

Time switch

Input signal Original frame

Figure 6.14

3

3

Store and rearrange

1

2

4

3

1

Output signal Rearranged frame

Rearranging timeslots.

reads the content of input channel 2, output channel 3 reads the content of input channel 1, and output channel 4 reads the content of input channel 3. The output frame is thus a rearrangement of the input frame. However, if a timeslot in the input channel is empty, the timeslot is not switched. After rearrangement there may also be empty slots in the output frame. An input timeslot may be copied on several timeslots of the output, thus supporting broadcast and multicast operation. The equivalence between time-division switching and space-division switching is shown in Figure 6.15. In the time-division switch, the input channels are multiplexed at the input of the switch into a single TDM signal. The switch then rearranges the timeslots of the TDM signal in accordance with the switching instructions, and the rearranged signal is finally demultiplexed into individual channels. The result is a spatial rearrangement of the input channels. If the input channels

1

3

Time-division switch

2 Users 3

4

4

1

1

2

2 Space-division switch

Users

Figure 6.15

Demultiplexer

Users

Multiplexer

2

1

Users

3

3

4

4

Equivalence of time-division switching and space-division switching.

6.2 Switching Technologies

159

are fed directly into a space switch with the same number of inputs and outputs as the time-division switch, the same spatial rearrangement can be made. This proves the equivalence between time-division switching and space-division switching. This equivalence allows us to design a complex exchange consisting of several switching matrices in a cost-effective way. Some switching matrices may then be space switches, while other matrices may be time-division switches. 6.2.5

Particular Switching Networks: Clos-Benes Networks

A large exchange will contain several switching matrices in a configuration as shown in Figure 6.16. The individual switching matrices may be space-division crossbar switches or time-division switches. An N × N switch, where N = nr, is realized by interconnecting three rows of smaller matrices. The number of leftmost n × m and rightmost m × n matrices is r, and the number of r × r matrices in the middle row is m. Note that there is exactly one connection from each rightmost matrix to each midmost matrix, and exactly one connection from each midmost matrix to each leftmost matrix. This switching network is called a Clos-Benes network. Each of the smaller matrices may also be Clos-Benes networks allowing a large exchange to be designed as an iterative structure. A Clos-Benes network can be blocking or strictly nonblocking. There are also conditionally nonblocking and rearrangeable nonblocking switches. Strictly nonblocking means that it is always possible to establish a connection between any idle

r n × m matrices

m r × r matrices

nr inputs

nr outputs

In the example, n = 3, m = 3, r = 4

Figure 6.16

r m × n matrices

Matrix of switching matrices.

160

Switching

input circuit and any idle output circuit irrespective of how all other inputs and outputs are connected. The N × N (N = nr) network is strictly nonblocking if the number of midmost matrices is at least m = 2n − 1. This is easily seen. The task is to connect a call on an arbitrary input port of one input matrix to an arbitrary output port of an output matrix irrespective of how many other calls are connected through the switch. If all other n − 1 input ports of the input matrix and all other n − 1 output ports of the output matrix are already occupied but none of these connections are between the chosen input and output matrices, then 2n − 2 midmost matrices are already occupied. Therefore, one more matrix in the middle row is required in order to connect the new call; that is, m = 2n − 1. In Figure 6.16, m = 3, which is less than 2n − 1 = 5. The network in Figure 6.16 is thus not strictly nonblocking. Conditionally nonblocking means that there are fewer than 2n − 1 matrices in the middle row, but there exists an algorithm describing how the internal matrices are to be occupied when a new call arrives. It can be shown that the switch is nonblocking if m ≥ 3n / 2, where x is the integer part of x, and a midmost matrix that is not already connected to an input matrix is not used for the new connection unless there are no other available routes through the switch. The switch in Figure 6.16 is not conditionally nonblocking since 3 × 3 / 2 = 4 > m = 3. In a rearrangeable nonblocking switch, the existing calls may be rearranged if a new call destined for an idle output arrives and there is no free path through the switch. A network with m = n can obviously always be rearranged so that it is nonblocking. This value of m is the smallest number of middle matrices in any nonblocking Clos-Benes network. The switch in Figure 6.16 is rearrangeable nonblocking. 6.2.6 6.2.6.1

Particular Switching Networks: Application of Binary Switching Element Binary Switching Matrix

The simplest switching element is the binary 2 × 2 matrix shown in Figure 6.17(a). The switch consists of just two inputs and two outputs. These matrices are used in optical switches (see Chapter 10) and in switches for high bit rate signals such as in ATM and other high-speed packet data networks. The matrix has two states as shown in the figure: the bar state interconnecting input 1 with output 1 and input 2 with output 2; and the cross state interconnecting input 1 with output 2 and input 2 with output 1. Obviously, a conflict then exists if simultaneous calls on the two input ports are destined to the same output port. A simple routing algorithm suitable for packet switching is called self-routing. The algorithm does not involving explicit computations in the switching element and thus offers an efficient path selection method in fast networks. The method is employed in the ATM networks. The principle is as follows. A binary index field is affixed at the front of the input packets as shown in Figure 6.17(b). The index field is computed when the packet is presented to the switch as shown in Figure 6.17(c). The destination address of the packet is either provided to the switch in a separate signaling message (connection-oriented system) or derived from information contained in the packet itself (connectionless system).

6.2 Switching Technologies

161

Output

Input 1

1

1

1

1

1

2

2

2

2

2

2×2 2 Matrix

Bar state

Cross state

(a) Index field Packet

First bit of packet

First bit of index field

(b)

Address information

Compute index Packet Insert index

Packet

110

(c)

Figure 6.17

Packet

11 Packet

1

(a–c) Binary matrix and self-routing algorithm.

The first bit of the index field (the rightmost bit in the figure) is used by the first binary switching element in order to select the output port of that switching element. If the value of the index bit is 0, then the upper output port—port 1 in Figure 6.17(a)—is selected; and if the index bit is 1, the lower output port—port 2—is chosen. The index bit is then removed and the next binary switching element uses the bit that now appears as the first bit in the index (bit number two in the original index). The index bits are used in the same way at all stages in the switch. When the packet appears from the switch, the packet will no longer contain the index field, since all the bits in it has been consumed. The principle is illustrated in Figure 6.17(c) for a switch consisting of three stages (only two of them are shown in the figure). Later, self-routing in the 8 × 8 Banyan network is explained. 6.2.6.2

Clos-Benes Networks

Switching networks can be constructed iteratively from binary matrices arranged in several rows and columns called stages. The advantages of these networks are that they can be constructed using purely iterative procedures and that the modules of each iterative step can be implemented in VLSI. The switch will then be both small and cheap. One particular example of a network of this type is the rearrangeable Clos-Benes networks constructed from binary matrices. Figure 6.18 shows how 4 × 4 and 8 × 8 networks can be implemented. Each box is a binary matrix. The

162

Switching n

n

n

construction is iterative for 2 × 2 matrices. The number of binary matrices in the 2 n n−1 × 2 matrix is two columns of 2 matrices each (the input and output columns) plus n−1 n−1 twice the number of matrices in the 2 × 2 matrices that make up the interior of n n the new matrix. If we call the number of matrices Cn in the 2 × 2 matrix, then this gives the simple iteration equation: C n = 2 ⋅ 2 n −1 + 2C n −1 = 2 n + 2C n −1

Since C1 = 1, we find easily that C n = 2 n −1 (2n − 1) = ( N / 2 )(2 log 2 N − 1) n

where N = 2 is the number of input or output ports of the switch. We see that the complexity of the switch is almost a linear function of the size N of the switch. In comparison, the number of cross-points in an N × N crossbar 2 switch is N and thus increases much faster. The 4 × 4 and the 8 × 8 networks of binary matrices consist of 6 and 20 binary matrices, respectively. A 1,024 × 1,024 matrix requires 9,728 binary matrices. Even such a big switch can be made in VLSI.

1

1

2

2

3

3

4

4 4 × 4 matrix

1

1

2

2

3

3

4

4

5

5

6

6

7

7

8

8 8 × 8 matrix

Figure 6.18

Iterative productions of 2 × 2 matrices. n

n

6.2 Switching Technologies

6.2.6.3

163

Banyan Networks

Figure 6.19 shows an 8 × 8 Banyan network consisting of three stages. All stages (or columns) contain four binary matrices. The Banyan network is particularly suitable for fast packet switching of packets that are equally long and is, for this reason, used in ATM switches. The most important characteristic of the Banyan network is that there is only a single path between any pair of input and output ports. It can be shown that the number of stages in a 2n × 2n network must be n in order to fulfill this condition. For the same reason, there is also a unique way (except for symmetries and translations) in which the binary matrices in the network are connected. Self-routing as explained earlier can be used in Banyan networks. The binary address of the destination port (shown to the right in Figure 6.19) is used to construct the index field in the front of the packet. The index field is in fact the binary address inserted in the reverse order. For example, the out port addresses 001, 101, and 110 appear in the index field as 100, 101, and 011, respectively. The first bit of this address is then examined by the binary matrix in the first stage. The second bit is examined by the second stage, and so on. The packet is routed to the upper outlet if the bit is 0 and to the lower outlet if it is 1, as explained for self-routing in the binary matrix.

Stage 1

Stage 2

Stage 3

1

1

(000)

2

2

(001)

3

3

(010)

4

4

(011)

5

5

(100)

6

6

(101)

7

7

(110)

8

8

(111)

Figure 6.19

8 × 8 Banyan network.

164

Switching

If a packet arriving at input port 7 is destined for output port 5 (address 100 or reverse 001), stage 1 moves the packet to the lower outlet, stage 2 moves it to the upper outlet, and stage 3 moves the packet to the upper outlet as illustrated in the figure. Banyan networks are not nonblocking. In the figure, there is no free path between input port 4 and output port 6 (address 101). The path is blocked by the packet from input port 7 to output port 5 in stage 2. The blocking performance of the network can be improved in several ways. Every input contains a buffer in the switch in Figure 6.20. If the buffer is full when a packet arrives, the packet is lost. The access to the switching fabric is slotted. This means that a buffer can start sending a packet only in the beginning of a slot. If the buffer is not empty, the first packet in the buffer is sent. If it collides with another packet in one of the binary switches, one of the packets will be put through while the other packet is stopped. This event is signaled back to the appropriate buffer. The buffer will then attempt sending the packet in the next slot. A more efficient scheme is the buffered Banyan network. A buffer is then placed at the input ports or the output ports or at both ports of each binary switch in the network. The binary switch containing both input and output buffers is shown in Figure 6.21. The other alternatives are designed in a similar way. All input packets are stored in the input buffer until they are successfully delivered to the appropriate output buffer. If the input buffer of the first stage is full when a packet arrives, the packet is lost.

Figure 6.20

Banyan network with input buffer.

6.2 Switching Technologies

165

If the two input buffers attempt to deliver a packet in the same slot, only one of the packets will be accepted by the output buffer. The other packet remains in the input buffer until the next slot. Similarly, if the output buffer is full, the packets will be kept in the input buffer until there is a free position in the output buffer. The output buffer forwards the packet to the input buffer of the next stage in the same way. If contention occurs and the packet is not accepted by the next input buffer, the packet remains in the output buffer until the next slot. There is then no internal congestion in the switch. Packets are only lost if an input buffer of the first stage is full when a packet arrives. The algorithm is called a backpressure algorithm since congestion is moved to the front of the switch. The switch may consist of several Banyan networks in parallel as shown in Figure 6.22. The input buffer will select one of the parallel Banyan networks and attempt to send the packet on that network. If this attempt is not successful, the input buffer will try another Banyan network. The output buffer just orders the sequence of transmission of the packets on the output connection. Figure 6.23(a) shows the Batcher-Banyan network. The purpose of the Batcher sorter is to make the exchange nonblocking. The Batcher sorter is a complex combinatorial network that orders the arriving packets according to the destination port address. The network is composed of binary switching elements but does not contain buffers. The composition of the network is not shown here. However, the Batcher sorters are rather large networks: an 8 × 8 Banyan switch requires a Batcher sorter containing 24 binary switching elements. Batcher sorters can be made for Banyan switches of any size. The packets arrive at the input of the switch in an arbitrary order independently of destination port address. There is thus no ordering of the packets before they arrive. The Batcher sorter orders the packets such that the packets appear at the output of the Batcher sorter in an ascending order of their destination address. That is, the packet with the lowest address will appear at the top of the sorter network; the packet with the next lowest address will appear at the second outlet; and the packet with the highest address will appear a number of outlets down the network equal to the number of packets to be switched. Packets with the same address will appear at adjacent ports. The shuffle stage just adapts the ordered sequence to the switching structure of the Banyan network. Figure 6.23(b) shows a Batcher-Banyan switch where external conflicts are resolved. An external conflict means that more than one packet with the same destination address arrive at the switch simultaneously. Packets with the same

Figure 6.21

Buffered binary switch.

166

Switching

Figure 6.22

Several Banyan networks in parallel.

destination address are called conflicting packets. All other packets are nonconflicting packets. Conflicting packets appear at adjacent channels of the Batcher sorter. The trap network then just picks out and marks conflicting packets based on this criterion. All packets are then fed to the concentrator. The concentrator picks out one packet from each conflicting set and forwards these packets together with nonconflicting

Shuffle stage

Batcher sorter

Banyan network

(a)

.

. Recirculation queue

. .

. .

Batcher sorter

. .

.

Trap network

.

. . .

.

Concentrator

. . .

(b)

Figure 6.23

.

(a, b) Batcher-Banyan networks.

. Banyan network

. .

6.2 Switching Technologies

167

packets to the Banyan network. All these packets will be successfully switched, since all packets are destined for separate outlets and the sorter has put the packets in an order in which internal conflicts in the Banyan switch have been eliminated. The concentrator concentrates the trapped packets to the upper outlets of the concentrator. These packets are then sent to the recirculation queue from which the packets are again fed into the Batcher network as shown. The rate by which packets are sent to the recirculation queue is called the recirculation rate. This switch is called the starlight switch. A variation of this method called the sunshine switch contains K parallel Banyan networks with an output buffer at each output port. This arrangement allows K conflicting packets to be handled at the same time, thereby reducing the recirculation rate. The purpose of complex arrangements such as the starlight and the sunshine switch is to offer high throughput for very bursty traffic. There are several other ways in which high-speed packet switches have been designed. A deeper description of all proposed methods is outside the scope of this book. 6.2.7

Construction of Switching Systems

The basic components of a circuit switched exchange are shown in Figure 6.24. The exchange may consist of units as follows: •

•

•

• •

• •

•

•

•

•

User modules manage the interface between the users and the exchange, including signal coding, protocol management, and fault monitoring. User signaling modules send, receive, and interpret user signaling at the user interface. The common channel signaling module is in charge of the signaling between exchanges. Timing units provide all clock and timing signals required in the exchange. Trunk modules manage the state of the trunk circuits (idle, busy, faulty, under test, and so on). Junction modules manage the switching path through the switching fabric. The number analysis module interprets addressing information (particular call handling information), computes the switching path through the exchange, and provides the required information to the junction modules and other modules involved in the switching process. Service processing units handle services (call forwarding, call barring, number presentation, charged party selection, and many other supplementary services). The console provides local access to the operations system for management and maintenance purposes. The operations system is responsible for operation, management, and maintenance of the exchange. The unit for remote access allows remote operations and management (O&M) centers and remote personnel to access the local operations system of

168

Switching

User module

Common channel signaling

User signaling

Trunk module

Timing

Switching fabric

Operations system

Junction module

Number analysis

Service processing

Console

Remote access to operations system

Figure 6.24

•

Structure of a circuit-switching exchange.

the exchange for fault monitoring, software updating, retrieval of data, and testing. The switching fabric consists of the switching matrices and all control functions required to select, establish, and release the switching path.

The information sent between the different units can be routed via the common switching fabric, via dedicated switching matrices, via bus circuits, or by other means. The modules of the exchange are computers, and all processing of connections, services, and operations are implemented in software. Some equipment in the exchange may be duplicated in order to improve the reliability of the exchange. The flow of the processing is supervised in order to detect anomalies. This supervision is built into the application programs and is supported by software and hardware for supervision of the exchange. The exchange also contains programs in the operations system that can restart the exchange after a failure. These programs read the temporary data of the exchange and load the data into backup storage devices, detect when and where failures occur, and reload software and data for automatically rebooting failing computers. The packet switched exchange is simpler but several of the same functions are present: • • • •

Input modules receiving and buffering IP datagrams; Output modules buffering and forwarding datagrams; Number analysis module; Switching management functions such as output selection, buffer allocation, priority management, real-time management, and time-to-live management;

6.2 Switching Technologies

•

• • •

169

Route management functions supporting shortest-path-first or other algorithms; Timing unit; Operation and management module; Operations system including console and remote access.

CHAPTER 7

Elements of Protocol Theory 7.1

Introduction This chapter offers a brief introduction to formal protocol theory as specified by ISO and ITU. The different layers of the protocol stack is explained but not treated in detail. The Internet, signaling system no. 7 (SS7), and the mobile radio interface are included as examples in order to illustrate different protocol structures. The example from the Internet describes the case where a network layer protocol is encapsulated in another network layer protocol. The structure of SS7 is described in some detail because it is an important supplement of Chapter 6. The example from GSM contains two significant elements. First, it shows that the application layer protocol may be mapped directly into the data field of the data link layer, and, second, it shows how different protocol stacks can be interconnected in tandem.

7.2

Purpose of the Protocol Protocols manage the dialog between two entities. Dialogs can be of two kinds: the exchange of unstructured information such as speech or video streams, or the exchange of structured information, where there exist semantic and syntactic rules governing the organization and the content of the dialog. We shall mainly consider dialogs consisting of the transfer of structured information. Note that the distinction between the two types of information transfer is floating. Unstructured information is often put into “containers” for routing, integrity management, and other functions. These containers are indeed protocols in the ordinary sense. In SDH (see Section 4.4), the information streams are embedded in containers providing identification of different information channels in the multiplexing hierarchy and other structural information. The containers in SDH may then be regarded as protocols, while SDH, at the same time, may be viewed as a multiplex system transferring unstructured information. The same applies to packet data networks such as the TCP/IP network. Here protocol or structured information (the protocol header) is sent together with unstructured information (the payload of the protocol). The principle is shown in Figure 7.1. Two processes taking part in a common task requiring exchange of information. We call this the logical information exchange. The protocol then supports the physical transfer of this information (the physical information exchange). The protocol is embedded in the channel

171

172

Elements of Protocol Theory

Logical information exchange Process

Process

Physical information exchange

Protocol

Channel

Figure 7.1

Information transfer.

connecting the machines in which the processes are accommodated. The protocol receives the information to be transferred from the application process and instructions concerning the organization of the information transfer. Such instructions are not part of the logical information exchange and may include addressing, timing requirements, protection of integrity, and delivery conditions. Another way of looking at this is that the logical information exchange is concerned with the semantics, or meaning, of the information. The protocol stack used for transferring the information is the syntax by which this is done, and as syntax it should be independent of the information, or semantics, actually transferred. This syntax may also be called the transfer syntax of the system. The distinction between syntax and semantics is made clear in the Open Systems Interconnection (OSI)1 protocol specification, but this is not the case in all protocols. Without this distinction, the application cannot always be altered without altering the protocol, or, vice versa, an alteration in the protocol stack cannot be done because it affects the design of the application. Each layer of a protocol stack will contain its own semantics and syntax independent of the semantics and syntax of the application. The semantics then describes how the layer performs its task: establishment, release and recovery of the connection, addressing, sequencing of data, and several other functions. The syntax is the representation of formats and parameters used by the layer. Examples of protocols are as follows: •

1.

Packet data protocols such as IP and TCP designed for the purpose of establishing connections between users for transfer of structured or unstructured information;

Open System Interconnection was a standard developed jointly by the standards organizations ITU and ISO during the 1980s. The theoretical design principles, including protocol syntax, protocol semantics, service access points, protocol services, and layering, suggested in this standard are mathematically well-founded concepts that are still valuable in practical protocol design. However, most of the specific protocols defined in the OSI suite of standards are no longer used.

7.3 Layer Services and Protocol Data Units

•

•

• • • •

• • • •

173

Protocols in the Internet supporting routing and management functions such as OSPF; Signaling protocols in communication networks such as digital subscriber-line signaling 1 (DSS.1) and SS7 for management of call establishment and release; Transaction protocols for database management and electronic payment; E-mail protocols; Protocols like http of the World Wide Web; Protocols for remote monitoring of sensors and control of actuators in industrial production systems; MAC protocols of Ethernet and WLAN; The Bluetooth source identification protocol; Protocols between wireless smart cards or RFIDs and reading devices; Protocols for organizing distributed processing over several computers—for example, remote procedure call (RPC), remote operations (RO), and distributed file management.

Protocols are used in every communication system for a large number of purposes. This ubiquity makes it so important to understand the basic principles on which protocols are built. Even localized systems such as PC platforms at home require protocols between the PC and peripherals like the mouse, keyboard, printer, scanner, and display. Within the PC there are protocols connecting together the CPUs, memories, disk drives, and CD-ROM drives. In the GSM mobile terminal, there is a protocol between the smart card containing the subscriber identity module (SIM) and the rest of the mobile terminal. This protocol is defined in the GSM specification, making it possible to use the same SIM card in mobile terminals made by different manufacturers.

7.3

Layer Services and Protocol Data Units All protocols are layered or hierarchical as shown in Figure 7.2. The layered protocol is a hierarchy consisting of n layers, where n is a small number. In the original OSI specification, n is seven. In all practical systems, n is smaller than seven (e.g., five in the Internet). The uppermost layer interfaces the distributed application process, but the protocol is not a part of the application process itself even though the n-th layer usually is called the application layer. The purpose of the application layer is to provide an interface through which the application processes can exchange information. Layer 1 interfaces the physical medium that can be an optical fiber, a radio transmission medium, an Ethernet cable, or any other transmission medium. Layer 1 is usually called the physical layer. A layer provides services to the layer above it, or if it is the top level of the protocol (the application layer), the service is provided to the application process. Figure 7.3 shows a segment of the protocol stack illustrating how the (N)-layer interfaces the (N + 1)-layer and the (N − 1)-layer.

174

Elements of Protocol Theory

Entity

Distributed application

n

Entity

n n–1

n–1 Layered protocol

1

1

Physical medium

Figure 7.2

Layered protocol connecting application entities.

The (N + 1)-layer receives services from the (N)-layer through one of the service access points (SAPs) of the (N)-layer—the (N)-SAP. There may be several (N)-SAPs in the (N)-layer interfacing different (N + 1)-entities. The SAP is identified by an address called the SAP-address. The services offered to the (N + 1)-entity are provided by the (N)-entity. An (N + 1)-protocol exists between (N + 1)-entities defined entirely within the context of the (N + 1)-layer. The (N + 1)-protocol is realized by

(N + 1)-protocol (N + 1)-entity

(N + 1)-entity

(N)-SAP

(N)-CEP

(N + 1)-layer

(N)-connection (N)-layer (N)-protocol (N)-entity

(N)-entity

(N – 1)-SAP

(N – 1)-layer

Figure 7.3

Protocol layer.

7.3 Layer Services and Protocol Data Units

175

mapping it onto the (N)-layer entity. Similarly, the (N)-layer contains an (N)-protocol, again entirely defined within the context of that layer. The (N + 1)-protocol is conveyed via an (N)-connection as viewed from the (N + 1)-layer. The connection endpoint (CEP) terminates the (N)-connection within the SAP. The SAP addresses are associated with the connection endpoints. The (N + 1)-protocol is thus embedded in and conveyed by the (N)-layer protocol. The (N)-layer protocol is similarly embedded in and conveyed by the (N − 1)-protocol via the CEP terminating the (N − 1)-protocol. This structure is repeated all the way down to the physical layer, where bits are transferred. Note that in this way the resulting n-layer protocol can be viewed as being composed of several individual protocols, one for each of the n layers. The application protocol is then mapped onto the uppermost layer of the protocol—the (n)-layer. The (n)-layer protocol is mapped onto the (n − 1)-protocol and so on until the lowest layer of the hierarchy (the (1)-layer protocol). The (1)-layer interfaces the physical transmission medium. The principle is shown in Figure 7.4. Note that, in this discussion, (n) indicates the highest level of the protocol, while (N) denotes an arbitrary layer of the protocol. Note again that there may be several SAPs between one layer and the layer above. There may also exist a number of (N)-connections between two peer entities. The layering architecture is flexible in this respect. Note that this way of looking at protocols is purely theoretical and is strictly not necessary for designing the protocol. However, the advantage lies in reducing the

Application process

Application protocol

Application process

Mapping

(n)-layer protocol (n)-layer

(n)-layer

(n – 1)-layer protocol (n – 1)-layer

(1)-layer

Figure 7.4

Mapping between layers.

(n – 1)-layer

(1)-layer protocol

(1)-layer

176

Elements of Protocol Theory

complexity and making it easier to see where changes have to be made in order to adapt the protocol to new environments. The principle allows the designer to concentrate on developing a single layer without taking into account how the adjacent layers are designed. The only requirements that must be observed are the rules that apply at the service access point. The protocol at any layer consists of the exchange of packets of information called protocol data units (PDUs). Each PDU consists of a header (H), the payload, and possibly a tail (T). The payload of the PDU of one layer is the PDU of the layer above, and so on in an iterative manner as shown in Figure 7.5. The PDU may also contain protocol parameters and streams belonging to the same layer. The iterative PDU structure is another way of expressing the mapping shown in Figure 7.4. The header contains the name of the PDU and other control parameters. The payload may be empty. If so, all information in the PDU is carried in the header. The (N)-layer offers services to the (N + 1)-layer. These services are primarily to connect (N + 1)-layer entities, to transfer data between the (N + 1)-entities, and to handle exception or error conditions. The (N)-layer then offers an (N)-protocol implemented in terms of (N)-PDUs. The services offered by the (N)-layer are supported by primitives provided at the service access point, as shown in Figure 7.6.

H1

H2

H3

Etc.

T3

T2

Payload 3 Payload 2 Payload 1

Figure 7.5

Services.

Protocol service user (N + 1)-layer entity

Request

(N + 1)-layer entity

Confirmation

Indication

(N)-SAP

(N)-SAP (N)-layer (N)-protocol (N)-PDU Protocol service provider

Figure 7.6

Response

Format of protocol data units.

T1

7.3 Layer Services and Protocol Data Units

177

The types of primitive are: request, indication, response, and confirmation. This provides a simple and repetitive structure of the protocol stack. The four primitives are defined as follows: •

•

•

•

Request (req) by which the (N + 1)-layer requests the (N)-layer to perform a specified task such as sending a message to the peer entity of the (N + 1)-layer. The primitive may also be used to request the (N)-layer to perform local tasks. Confirm (cnf) by which the (N)-layer acknowledges the receipt of a request primitive. The confirmation may be received from the peer (N + 1)-layer or be generated by the local (N)-layer. Indication (ind) by which the (N)-layer delivers a protocol data unit to the remote (N + 1)-layer. The indication message may be generated either by the sending entity of the (N + 1)-layer or locally by the (N)-layer. Response (rsp) by which the (N + 1)-layer acknowledges the receipt of an indication primitive.

This arrangement gives rise to the set of nine elementary services shown in Figure 7.7. The most common services are the ones shown in Figure 7.7(a) and Figure 7.7(e). In Figure 7.7(a), the (N)-layer transfers the acknowledgment generated by the receiving entity. In Figure 7.7(e), there is neither local acknowledgment nor

(N + 1)-layer

(N)-layer

(N + 1)-layer req

req

ind

ind cnf cnf

rsp

rsp (a)

(b)

req

req ind

ind

cnf

rsp (d)

(c)

req

req ind cnf (e)

(f)

ind

req

ind

rsp

(h)

(g)

Figure 7.7

(a–i) Elementary protocol services.

(i)

178

Elements of Protocol Theory

end-to-end acknowledgment. This service is used if the acknowledgment either is not important or is part of another procedure. In Figure 7.7(b), the (N)-layer acknowledges the receipt of the request locally and the receiving (N + 1)-layer entity acknowledges the receipt of the indication, but the acknowledgment is not passed between the entities. This service may be used in several cases (e.g., when it is impossible to transfer the acknowledgment because of interworking between dissimilar systems). Service in Figure 7.7(c) is only acknowledged locally by the (N)-layer. Service in Figure 7.7(d) is acknowledged by the receiving (N + 1)-layer entity, but the acknowledgment is not transferred to the sending (N + 1)-layer entity. The remaining cases are local services between the (N)-layer and the (N + 1)-layer. They are used when an (N + 1)-layer entity is requesting the connection to be aborted, or when the (N)-layer has to indicate some exception condition to the layer above (transmission fault, abortion, excessive load, and so on). The general services and their primitives are used to design (N)-layer services, such as transfer of data and connection establishment. The general format of the service is NAME.type where NAME indicates the type of service (for example, DATA for transfer of data and ESTABLISH for establishing a connection), and type is either of the four primitives req, ind, cnf, and rsp. Figure 7.8 shows the procedures at the (N)-layer: the services offered to the (N + 1)-layer, the PDUs of the (N)-layer protocol, and how the services and the PDUs are interrelated. In Figure 7.8(a), the (N)-layer offers a connection-oriented service consisting of a connection establishment phase, a data transfer phase, and a disconnection phase. The establishment service is a type (a) service of Figure 7.7. The disconnect service is a type (c) service; and the data transfer service is a type (e) service. However, the (N)-protocol acknowledges the receipt of messages, (N)-PDU, so that error correction may take place within the (N)-protocol. These activities are not visible to the (N + 1)-layer. If the data transfer procedure cannot be completed, the (N)-layer may provide a notification to the (N + 1)-layer that the data transfer was unsuccessful. TCP is a variant of this protocol. Figure 7.8(b) shows the simple connectionless data transfer service provided by IP or UDP. The service is of type (e) and there is no acknowledgment services within the (N)-protocol.

7.4

Specification of Primitives The primitives can be defined using a method prescribed by ITU/ISO. The method is simple and accurate. Each service consists of one, two, three, or four primitives—request, indication, response, and confirm, as explained earlier. Which primitives are required depend on the semantics of the service. Each primitive may contain parameters where the parameters are the actual information transferred by the primitive. Furthermore, the primitives and the set of parameters contained in the primitive can be written as abstract data types using a high-level language such as abstract syntax notation no. 1 (ASN.1) also developed by ITU/ISO. The first example is the ESTABLISH service shown in Table 7.1.

7.4 Specification of Primitives

179

(N)-service

(N)-protocol

ESTABLISH.req

(N)-PDU<establish>

(N)-service

ESTABLISH.ind ESTABLISH.res

(N)-PDU ESTABLISH.cnf DATA.req

(N)-PDU

DATA.req

DATA.ind

(N)-PDU

DISCONNECT.req

DATA.ind

(N)-PDU

DISCONNECT.cnf

(N)-PDU DISCONNECT.ind

(a)

(N)-service

(N)-protocol

DATA.req

(N)-PDU

DATA.req

(N)-PDU

(N)-service

DATA.ind DATA.ind

(b)

Figure 7.8

(a, b) Examples of information exchange in a protocol.

Table 7.1

Definition of the ESTABLISH Service

Parameter

Request

Indication Response

Confirm

Application context name Destination address Destination reference Origination address Origination reference Specific information Responding address Result Refuse-reason Provider error

M M U U U U

M(=) M(=) C(=) O C(=) C(=)

U

C(=)

U U M C

C(=) C(=) M(=) C(=) O

The table describes which information or parameters the different primitives must or may contain. An M means that this parameter is mandatory in the primitive. M(=) means that the parameter must have exactly the same form as in the corresponding request or response primitive. U indicates a user option. A user in this context is the layer above that is making use of the ESTABLISH service. For instance, the origination reference may or may not be included in the message. If this parameter is present, it must be provided to the peer entity. This is shown by the C, which means that the parameter (e.g., the destination reference) is a conditional

180

Elements of Protocol Theory Table 7.2

Delimeter

Parameter Request

Table 7.3

Indication

Data

Parameters

Request

Indication

Protocol user User data Priority Do not fragment Source address Destination address

M M U U M M

M(=) M(=)

M(=)

parameter. These parameters usually appear in the form C(=), indicating that the parameter must have the same form as the corresponding user option if the user option is applied. C may also be used in other cases. One example is shown in the table, namely the refuse-reason parameter, which is used by the remote entity if the indication primitive is refused for some reason. O is a provider option, for example, reporting a provider error (such as line interrupted). The second example is a delimiter service, which, for instance, is used to indicate the end of a sequence of messages. This service does not contain any parameters and consists of just a request and an indication primitive, as shown in Table 7.2. The corresponding PDU consists then of just the message name. Even if the table is empty, it must be contained in the specification. Otherwise, there is a risk that the delimiter service is not implemented. The IP protocol consists of just one service—namely, the data transfer service. The service consists of only the request and the indication primitive. The parameters are the information contained in the datagram header plus the payload field. The datagram is shown in Table 7.3. The user of this service can be TCP, UDP, or management protocols (the protocol service users). Some of the parameters are only provided to the service provider (IP) at the sending end. The PDU contains additional information, such as version number, length indicator, checksum, fragmentation indicator, time-to-live, next header information, and network options. This information is used by the IP network and is neither received from the user nor provided to the user.

7.5

Layering Practical protocol stacks contain at most five layers: physical layer, data link layer, network layer, transport layer, and application layer, as shown in Figure 7.9. The reason why just these layers are required can be justified as follows.

7.5 Layering

181

A physical layer is needed in order to adapt the protocol to the physical medium. There will be different physical layer protocols for accessing different physical media. At the physical layer, universality is not possible. The layer may include data fields and functionality for the following: • •

• •

Synchronization of clocks (GSM, WLAN, and Ethernet); Frame delimitation by use of flags and bit stuffing for transparency (HDLC), unique words (WLAN and Bluetooth), fixed envelope structures (PDH and SDH), or particular patterns (use of the error detection field in ATM cells); Length indicators of physical frames (WLAN, Ethernet, and GSM); Data rate indication (WLAN).

The physical layer is usually connectionless. The data link layer is a more general protocol than the physical layer protocol that may be implemented on a variety of physical channels. This is also a link-by-link protocol terminated at each physical link endpoint. The data link layer and the physical layer shall provide a basic communication capability to the layers above; that is, keeping the error rate at a low level by adaptive error control and adapting the rate of data flow on the link to the capabilities of the equipment on each side of the link (data rate and buffer size). Data links with these capabilities were among the first protocols to be developed. The most common data link protocol is the high-level data link control protocol. This protocol is used in one or another appearance in a number of systems: the digital subscriber line connecting the user to the digital backbone network, data connections over satellites, GSM, SS7, and local area networks (WLAN and Ethernet). The information transfer in HDLC is shown in Figure 7.10. The figure also shows the primitives received from and provided to the layer above.

Router

Terminal

Terminal

Application process

Application process

Application

Application Transport

Transport Network

Network

Network

Data link

Data link

Data link

Physical

Physical

Physical

Physical medium

Figure 7.9

Protocol stacks.

Physical medium

182

Elements of Protocol Theory

HDLC is connection-oriented. The data link is established by a setup sequence consisting of a setup frame (set asynchronous balanced mode) and an acknowledgment (unnumbered acknowledgment). After the connection has been established, data transfer may commence using information frames (I) and control frames receiver ready (RR), reject (REJ), and receiver not ready (RNR), which is not shown in the figure. The RNR control frame is used for flow control. When the sender receives this frame, transmission of information frames is suspended until an RR frame is received. The RNR is sent by the receiver if the data link buffer is full. The information frames and the control frames contain an address field (1 or 2 octets) identifying which type of equipment or process shall receive the frame (e.g., a specific terminal on a multiterminal digital subscriber line, or the software managing the data transfer across the link). These frames also contain a control header consisting of 1 or 2 octets that identify the type of frame (I, RR, REJ, or RNR). Information frames are numbered with a forward frame number identifying the frame and a backward frame number used for acknowledging the receipt of a frame. The control frames contain only the backward frame number. The backward frame number is the forward frame number in the next information frame that is expected (usually one more than the frame number of the frame just received). The receiver acknowledges all received frames that are in sequence with an RR frame. If one or more frames are lost [frame I(1,0) in the figure], the sequence number of the next received message will be too large. The receiver will notify the sender that the transfer is out of sequence by returning an REJ frame, REJ(1) in the figure, with the sequence number of the next frame the sender shall deliver to recover the correct

EST.req

SABM

EST.cnf UA DATA.req

I(0, 0)

DATA.req

I(1, 0)

DATA.req

I(2, 0)

DATA.ind RR(1)

I(1, 0)

REJ(1) DATA.ind

I(2, 0)

DATA.ind

Figure 7.10

Operation of the HDLC protocol.

RR(2)

DATA.ind

I(0, 3)

DATA.req

7.5 Layering

183

sequence. The sender starts retransmitting the frames from this sequence number onward. Information frames can also be acknowledged by other information frames (I(0,3) in the figure). The important point here is the way in which retransmissions are managed. The method used in HDLC is called automatic repeat request. ARQ is an adaptive error correction method: if the bit error rate is high, the repetition rate is large and the information transfer slow; if the bit error rate is small, the repetition rate is also small and the information transfer rate is large. This method is much more efficient than forward error control (FEC) because FEC offers the same information transfer rate irrespective of the bit error rate. The data link layer ensures that the error performance of the link is good. The network layer is in charge of routing the information from the origin to the destination. The most commonly used network layer protocol is the Internet protocol. For routing purposes, the network needs the IP number of the destination. In order to route the call efficiently, the network may also require other information such as priority and real-time requirements. The parameters that can be contained in the request primitive are shown in Table 7.3. One important parameter is the protocol user parameter indicating which higher layer protocol uses the IP connection. This may be a transport protocol, a management protocol, or another embedded IP protocol. The latter is the case for mobile IP and IP security (IPsec). This feature is explained later. IP is a connectionless network layer protocol offering only transfer of data. Several connection-oriented network protocols have been developed, but they are rarely used in current networks. One exception is the transfer of huge information files between databases where connection-oriented protocols are still used in many applications. Such applications are also likely to be replaced by the simple Internet protocols. The network layer brings the data from one terminal to another. IP offers only a best effort service, meaning that there is no guarantee that all information has been delivered, that some information is not duplicated, or that the information has been received in the correct order. Files and other ordered data structures are meaningless if the data are incomplete or mixed up. Transfer of ordered data on Internet requires a transport layer protocol to ensure correct receipt of data. TCP is the connectionoriented transport protocol on the Internet. TCP establishes and releases the connection between the terminals and contains mechanisms for recovering lost information (using ARQ), sorting the received information in the correct order, and eliminating duplicated information. The header of TCP contains a parameter called the port address. This address indicates which application is contained in the information field so that the receiving processor can invoke the correct software in order to interpret the content of the information field. All transport protocols must contain a parameter serving the same purpose. The transport protocol is not needed in some systems, even though structured data is transferred. One such example is SS7, where the network layer offers sufficient data integrity. The network layer of SS7 then contains address information

184

Elements of Protocol Theory

similar to the port address of TCP. In the network layer protocol of SS7, this information is called the subsystem number (SSN). Not all applications of the Internet need the integrity features offered by TCP. Retransmission of lost information is meaningless for real-time streams such as speech and video. Therefore, the simpler transport protocol UDP is sufficient. UDP is connectionless and does not contain essential information other than the port address. The port address is, of course, required in order to distinguish between different types of streams. The transport protocol contains the application protocol in the payload. There are many application protocols designed for a multitude of purposes, such as remote procedure call, http, file transfer, security, and real-time management. Some of these protocols are connectionless, while others are connection-oriented. The OSI stack also contained separate layers for session control (management of synchronization points, task organization, and allocation and transfer of execution control) and presentation management (negotiation of syntax, indication of high-level specification language (compiler version), version handling, and so on). Such protocols are still required, but they are usually implemented as separate application protocols or as elements of application protocols. Note that the application protocols may form an iterative structure; that is, one application protocol may be embedded in the user data field (or information field) of another protocol. Examples are given in Sections 7.8 and 7.9, where the MAP is embedded in the transmission control application part (TCAP). TCAP consists again of the remote operation protocol embedded in the transaction sublayer. The structure at the application layer may thus be very complex, allowing embedding of application protocols as explained, multiplexing of several application protocols within another application protocol, and constructing application sessions consisting of different protocols in series in order to perform different tasks at different times. Any attempt to define an orderly or hierarchic model for the application layer is of little value, and no serious attempt has been made to do so.

7.6

Hardcoding or Softcoding of the Protocol Data Unit The coding of the PDU can be done in two ways. All the parameters the data unit consists of can be described in terms of concrete bit fields. A particular value of the parameter then corresponds to a set of values of the bits in the field. The field may consist of any number of bits. A field consisting of a single bit may represent a Boolean variable, a yes/no event, or an included/not included indication. The bit field may be of constant length (such as the one-bit field just mentioned) or be of variable length where length indicators or other delimiters are used to indicate the length of the field. This way of coding the parameters are called hardcoding. Hardcoding is used for the transport layer and all layers below. Softcoding is used on the application layer.

7.6 Hardcoding or Softcoding of the Protocol Data Unit

7.6.1

185

Hardcoding

Let us take the IP protocol as example in order to show how hardcoding is done. Figure 7.11 shows the format of the IPv4 datagram and the format of the service subfield. The general format of the datagram consists of 32-bit words. A parameter may consist of a whole word (e.g., source IP address), a part of a word (e.g., version, flags, protocol), or several words (e.g., user data). In the figure, the length of the parameters is given as the number of bits it contains (in brackets). The coding of the parameter is sometimes described in the text in the form of a binary number (the version parameter is coded 0100 for an IPv4 datagram and 0110 for an IPv6 datagram) or a particular bit pattern (the source and destination addresses are divided into four blocks of 8 bits each, where each block is written as a number between 0 and 255). The coding of the service field is shown in Figure 7.11(b). The field consists of 8 bits, where the rightmost bit is not used. The remaining part of the field is then divided into several parts of different length. The precedence parameter consists of 3 bits allowing seven levels of priority of the datagram. The priority indicates the priority by which a datagram may be discarded in case of congestion. This field has never been used. The bits D, T, R, and C are used to set other priorities such as minimizing delay (D) or cost (C) or maximizing throughput (T) or reliability (R). Different services are assigned different values of these bits, for example: •

TELNET (remote logon) is encoded DTRS = 1,000, where the major concern is to minimize the delay.

Version (4)

Header Length (4)

Service (8)

Total length (16)

Identification (16) Time to live (8)

Flags (3)

Protocol (8)

Fragmentation offset (13)

Header checksum (16)

Source IP address Destination IP address Option (0–10 32-bit words)

User data (0–65,535 octets minus header length)

(a)

Precedence

D

T

R

C

D: Minimize delay T: Maximize throughput R: Maximize reliability C: Minimize cost (b)

Figure 7.11

(a, b) IPv4 datagram and service subfield.

186

Elements of Protocol Theory

•

•

Data transfer using the file transfer protocol (FTP) is encoded 0100 for maximum throughput while the subprotocol for FTP control is encoded 1000 for minimum delay. Simple network management protocol (SNMP) is encoded 0010 for maximum reliability.

The coding of the user data parameter is, of course, unrestricted. The number of octets in this field is given by the difference between the total length of the datagram and the header length. The header length is variable between five 32-bit words (20 octets) (no options) and fifteen 32-bit words (60 octets). Therefore, 4 bits is sufficient to indicate the number of 32-bit words in the header (the header length parameter). Hardcoding exists mainly because this was the way in which protocol specification started. The bit pattern of the protocol is then independent of any translation from abstract to concrete representation of the protocol. The hardcoding is also close to the assembler code of the computer so that it is easy to implement the specification on the computer. 7.6.2

Softcoding

Softcoding means that the protocol code is written in a high-level computer language and compiled in the normal way into machine instructions. Any high-level language such as Java, C, and C++ can be used for this purpose. A specification of this kind is possible only if the communicating computers translate the specification into exactly the same bit pattern. However, some high-level languages are developed for expressing data types and procedures suitable for protocol design. One such language is ASN.1, developed jointly by the ITU and the ISO. The advantage of using a syntax language such as ASN.1 is that the protocol specification is independent of particular programming languages. This makes it easier to compile the specification into any language or operating system. Standards organizations such as ISO, ITU, and ETSI publish some of these standards on CD-ROM in compiled form. A number of basic data types are defined in ASN.1: INTEGER, BOOLEAN, ENUMERATED, NULL, OBJECT IDENTIFIER, SEQUENCE, SEQUENCE OF, and a number of different unstructured strings (bit string, ASCII string, numerical string, octet string, and so on). These basic types are then used in more complex constructs such as remote operations and message formats. The compilation of the abstract syntax into the user data format of the application protocol is specified in a separate recommendation (X. 209 in ITU and 8825 in ISO). The syntax itself is specified in X. 208 in ITU and 8824 in ISO. The ITU and the ISO recommendations are identical. A brief extract of an ASN.1 specification is shown in Figure 7.12. The example is taken from the specification of the MAP in GSM/UMTS (see also Sections 7.8 and 7.9). The extract shows the “update” operation used by a visitor location register (VLR) to update the location of a mobile terminal in the home location register (HLR).

7.6 Hardcoding or Softcoding of the Protocol Data Unit

187

…

MAP_ProtocolOperations {root (0)identifiedOrganization (4) etsi (0) mobileDomain (0) gsmNetwork (1) modules (3) mapProtocol (4) operations (3) version4 (4)} DEFINITIONS :: = IMPORTS … N1 (0.4.0.0.1.3.4.2.4), N2 (0.4.0.0.1.3.4.2.4), … EXPORTS … …

BEGIN update OPERATION ARGUMENT updateArgSEQUENCE imsi mscNumber vlrNumber tmsi

{ OCTET STRING (SIZE (3..8)), [1] IMPLICIT OCTET STRING (SIZE (1..20)) OPTIONAL, OCTET STRING (SIZE (1..20), [9] IMPLICIT OCTET STRING (SIZE (4)) OPTIONAL

} RESULT updateResult SEQUENCE { hlrNumber OCTET STRING (SIZE (1..20)), authenticationData SEQUENCE (SIZE 5) OF SEQUENCE { r INTEGER (SIZE (1..N1)), sres INTEGER (SIZE (1..N2)), kc BIT STRING (SIZE (64)) } } ERRORS {–systemFailure – localValue: 34, --dataMissing– localValue: 35, --unexpectedDataValue – localValue: 36, --unknownSubscriber– localValue: 1, --roamingNotAllowed– localValue: 8} :: = localValue : 2

Figure 7.12

Example of an ASN.1 specification.

The module containing the MAP operations is provided with a globally unique identification code called an object identifier. The object identifier is a tree that identifies a particular organization, project, document, and even a single data type. In this structure the operation “update” is identified by the string (0.4.0.0.1.3.4.2.4.2), where the final number 2 is the identity that is assigned to the operation within the module called “MAP protocol, operations, version 4” (localValue :: = 2). The object identifier of the “MAP protocol, operations” module starts with the number 0, which identifies the ITU. OSI is identified by the number 1. Joint specifications of the ITU and ISO start with the root number 2. The next level of the object identifier tree says that this specification is done in a recognized organization outside the ITU (4) and not in an ITU study group. The 0 that follows shows that this organization is ETSI. The specification continues down the tree in a similar manner until version 4 of the “MAP-protocol, operation” specification is identified. By presenting the sequence (0.4.0.0.1.3.4.2.4.2) to a database of object identifiers, the database can identify (possibly assisted by several other databases) where the source code (or specification) of the operation “update” in version 4 of the MAP specification is stored. The statement IMPORTS contains data types or other specifications from which parameters, data types, and other constructs are imported. An object identifier (0.4.0.0.1.3.4.2.2) shows where the data types N1 and N2 can be found. The

188

Elements of Protocol Theory

EXPORTS statement contains data types that can be imported by other specifications. The “update” operation consists of the three parts: the invoke (ARGUMENT) containing the instructions and information that is sent to the peer; return result (RESULT), which specifies the results of the computation at the peer; and the different error messages the originator of the operation may receive if the procedure fails (ERRORS). The error messages are identified by a local reference value if they are contained in the same module as the operation. If not, they can be identified by a global object identifier—for example, (mapProtocol.4.2) referring to version 2 of module 4 of the MAP specification containing the error specifications. The coding of the data types is self-explanatory except for a few cases. Optional parameters are assigned an implicit label (e.g., [1] IMPLICIT) in order to distinguish between the cases where the parameter is or is not included. The size of every data type (except SEQUENCE, ENUMERATED, and other data types where the length is derived implicitly from the components of the data type) is stated explicitly in the definition. INTEGER (SIZE (1..20)) indicates that a valid integer must be in the range 1–20. OCTET STRING (SIZE (3..8)) specifies that this is an otherwise unspecified string of bits consisting of any number of octets between 3 and 8. SEQUENCE (SIZE 5) OF means that the following data type (in the example a SEQUENCE) is repeated five times (but with different values of the parameters). The particular construct in the example says that five different triplets of the security parameters (r, sres, kc) are provided to the VLR initiating the updating. The way the module is written has no other significance than that it is easy to read (for example, identifying corresponding parentheses). The advantage of softcoding is that it is easy to construct complicated protocols. Most protocols on the application layer are just complex in this way, containing optional parameters, complex sequences of parameters, repetition of the same set of parameters, embedding of several layers of sequences or sets of data, and so on. It would have been impossible to apply hardcoding to such situations. For example, some complex protocol structures used in invocation and management of supplementary services require a coding space of several thousand octets.

7.7

Example 1: Layering and Encapsulation in the Internet 7.7.1

Layering

The general format of the Internet protocol stack is shown in Figure 7.13. The protocol consists of four layers: •

•

Lower layers consist functionally of a physical layer and a data link layer. These layers are in some implementations distinct layers performing functions as described in Section 7.5; in other implementations the demarcation line between the physical layer and the data link layer is not so distinct. Therefore, the layers are shown as a single box in the figure. Network layer consisting either of IPv4 or IPv6.

7.7 Example 1: Layering and Encapsulation in the Internet

Application layer

Transport layer

Network layer

Lower layers

Figure 7.13

•

•

189

Applications: file transfer, http, e-mail …

TCP

UDP

IPv4

SCTP

IPv6

Physical layer and data link layer depending upon type of physical implementation: WLAN, Ethernet, fixed subscriber line, UMTS, etc.

Protocol layers of the Internet.

Transport layer supporting either the connection-oriented TCP for structured data transmission, the general connectionless UDP for transfer of single unordered data streams, or the combined connection-oriented and connectionless SCTP for multimedia. These protocols can be embedded in both IPv4 and IPv6. Application layer containing a large number of protocols.

The reader is referred to the rich literature concerning the details of the Internet protocols. Only a few features will be considered here. The IPv4 and IPv6 packets contain the address (IP number) of the receiver of the packet. This supports the routing function of the network layer. The packet may also contain additional information used for routing, such as priority and other QoS parameters, and indications concerning which routers the packet has to pass (IPv6). The header also contains a parameter called protocol. This parameter indicates which type of protocol is embedded in the information field of IP: TCP, UDP, Stream Control Transmission Protocol (SCTP), a particular network management protocol (e.g., ICMP), or another IP protocol. TCP is called a byte-oriented protocol. This implies that TCP preserves the byte structure of the information message. TCP may segment a message into several pieces that are sent as individual messages that are reassembled at the receiver. TCP is connection-oriented and offers reliable services. Reliable services include the mechanisms by which lost messages or message segments are recovered by the application of the ARQ procedure, duplicated messages are removed, and messages received out of order are put in the correct order before the information is provided to the application. TCP also provides flow control and congestion control. TCP is used for transfer of structured data. UDP is message-oriented in the sense that there is no relationship between any two UDP messages: each message is an independent entity of information. UDP supports no control services and as such is an unreliable protocol. UDP is used for real-time applications such as speech and video.

190

Elements of Protocol Theory

SCTP is message-oriented but the protocol supports the control procedures of TCP and is thus a reliable protocol. SCTP operates in full duplex and may support multiple message streams. The protocol is used for new applications such as VoIP and multimedia services containing both structured and unstructured data. The port address contained in the header of the transport protocols of the Internet identifies what type of application protocol is embedded in the information field. 7.7.2

Network Layer: Encapsulation and Tunneling

In some applications, the information field of IP does not contain a transport protocol but another network protocol. Sometimes, the information field contains another IP protocol. Here are five examples. The ICMP is used for error reporting and diagnosis. If a router cannot forward a call for some reason (destination unreachable, time-to-live exceeded, parameter problems, and so on), the router returns the reason in an error reporting ICMP packet. ICMP can also be used for diagnostics where, for example, a router may measure the time it takes to pass a datagram from one machine to another or investigate the condition of the path to a remote machine. ICMP messages consist of an ICMP header and information field. The ICMP PDU is encapsulated in the IP datagram as shown in Figure 7.14. The Internet group management protocol (IGMP) is used for managing the membership of multicast groups. This IGMP PDU is also encapsulated in the IP datagram as shown in Figure 7.14. In both cases, the IP header is used for routing the ICMP/IGMP message. This implies that the routers need not handle messages of other formats than IP, thus simplifying the network. The router will just deliver the data field, including the ICMP/IGMP header and data field, to the destination associated with the IP number. Figure 7.15 shows three other applications of encapsulation. In the first example, Figure 7.15(a), security in the form of authentication, data integrity, and privacy is provided by the IPsec header and the IPsec tail. In the encrypted mode, only the sender and the receiver are able to read the transport layer in clear text. The purpose of the encapsulation is that the routers need not implement a new algorithm for routing the IPsec packets. Figure 7.15(b) is an example of tunneling. The IP number is a geographic number that, by definition, identifies both the access point and the terminal connected to it (see Section 6.7). The network always routes the datagram to the network access point associated with the IP number. If the user moves to another network access point, the terminal is still identified by its IP number, but this number no longer identifies the network access point. The router at the new location will therefore inform the home router of the user about

IP header

Figure 7.14

ICMP/IGMP header

Encapsulation.

ICMP/IGMP data field

7.8 Example 2: Protocol Structure of SS7

IP header

191

Transport layer

IPsec header

IPsec tail

(a)

New IP header

Original IP header

Transport layer

(b)

IPv4 header

Transport layer

IPv6 header

(c)

Figure 7.15

(a–c) Encapsulation of IP in IP.

the new location and allocate a so called care-of number to the user that can be used to forward the datagram. The home router will then insert the original datagram, including the original IP header, in a new datagram where the destination address is the care-of number. The new router can then retrieve the original datagram and forward it to the user. Another example of tunneling is shown in Figure 7.15(c). This method can be employed in order to introduce IPv6. If an IPv6 datagram has to be routed via a network only supporting IPv4, the original IPv6 datagram is encapsulated in an IPv4 datagram. The address in the IPv4 datagram is then an IP number that signifies a router in another IPv6 network. This router will then remove the IPv4 header and forward the datagram as an IPv6 datagram. All these examples show that it is often simpler and cheaper to use encapsulation rather than altering the network to accept new network layer protocols to support new applications.

7.8

Example 2: Protocol Structure of SS7 7.8.1

Signaling Network Architecture

SS7 was developed by the ITU during the late 1970s and the early 1980s. The purpose was to replace the current in-band signaling systems based on tone signaling within the same frequency band as the voice signal. These signaling procedures were slow (about 8 digits per second, equivalent to about 32 bits per second, in the fastest of these systems) so that little information could be transferred within the time available for call establishment (2 to 3 seconds). The early systems were designed from electromechanical devices where all “computation” in the exchange was done by magnetic relays arranged in complex patterns.

192

Elements of Protocol Theory

SS7 was designed in order to meet the requirements of new telecommunications networks, such as the ISDN and mobile networks that required transfer of more signaling information, faster switching actions, and support of entirely new functions compared to previous systems. SS7 was also designed for improved remote management of computer-controlled exchanges and support systems. The design was based on the earlier signaling system no. 6 (SS6) that was soon judged to be too slow, too unreliable, and too expensive. SS6 was only taken into use on a few intercontinental telephone connections to avoid ring routing of calls and to support time-dependent routing rearrangement of the network in an effort to avoid overload (for example, during Christmas, traffic from French-speaking Canadian regions was routed via Sydney to Paris, while traffic from the English-speaking regions was routed directly across the Atlantic to London). SS7 must meet extremely severe performance and reliability requirements since an interruption of one signaling link may cause outage of a large part of the telephone network. These requirements are shown at the end of this section. SS7 is a common channel signaling system. This means that all signaling information is exchanged on a separate data network designed for just this purpose. The network is shown in Figure 7.16. The signaling network consists of nodes called signaling points (SPs) interconnected by signaling links. The signaling point is a router in the SS7 network. A signaling link is an ordinary digital connection operating at 64 Kbps, 2 Mbps, or another digital data rate. Several signaling links may be multiplexed on the same digital connection using a combination of time division multiplexing and statistical multiplexing. The SPs are connected to the signaling units in the exchange where all signaling processing takes place. The purpose of the SP is only to route and transport signaling messages. A signaling point may not be connected to an exchange but just act as a router in the signaling network.

SS No. 7 network SP

SP SP

SP

SP SP

Exchange Exchange

Exchange Exchange

PSTN SP = signaling point

Figure 7.16

Signaling network.

7.8 Example 2: Protocol Structure of SS7

7.8.2

193

Protocol Stack

The protocol stack of SS7 consists of four layers, as shown in Figure 7.17: • • •

•

The signaling data-link layer is essentially the physical layer. The signaling link-control layer is the actual data link layer. The network layer consists of the message transfer part (MTP) and the signaling connection control part (SCCP). The SCCP is not required in all applications. The application layer contains a number of protocols called signaling user parts (UPs), transaction capability parts, or application parts (APs).

For historical reasons which we shall not explain here, the network layer and the application layer have a rather complex structure. 7.8.3

Signaling Data-Link Layer (Layer 1)

This is essentially the physical layer of the protocol, despite its name. The layer specifies the data rate of the signaling link. The default rate is 64 Kbps, but other rates such as 2 Mbps and 8 Mbps may be used on connections requiring much signaling. The usual 8-bit flag (01111110) is used as delimiter between signaling messages. Transparency bit stuffing is used in order to avoid simulation of the flag inside a signaling message. 7.8.4

Signaling Link Control (Layer 2)

The signaling link control is one of several variants of the HDLC protocol. The format is shown in Figure 7.18. Three different signaling units can be sent on the signaling link. The message signal unit (MSU) contains a signaling message in the signaling information field (SIF). The link status signal unit (LSSU) is used for flow control, for management purposes, and for reporting problems at the signaling

AP Application layer

User part TCAP

SCCP Network layer Message transfer part

Signaling link control

Signaling data link

Figure 7.17

Protocol layers in SS7.

Data link layer Physical layer

194

Elements of Protocol Theory

CK

SIF

SIO

16

8n(n > 2)

8

LI 2

6

F I FSN B 1

7

B I B 1

BSN 7

(a)

CK 16

SF

LI

8 or 16

2

6

F I FSN B 1

7

B I B 1

BSN First bit 7

(b)

CK 16 BSN Backward sequence number

2

LI

F I FSN B

B I B

BSN

6

1

1

7

7

(c)

BIB Backward indicator bit FSN Forward sequence number FIB Forward indicator bit LI

Length indicator

SIO Service information octet SIF Signaling information field CK Checksum bits SF Status field

Figure 7.18

Signal unit formats.

point, such as processor outage. The information submitted by the LSSU is contained in the status field (SF). The fill-in signal unit (FISU) is sent when there is no other information (MSU or LSSU) to be sent on the link. This means that signal units are sent on the link all the time. The receiver discards received signal units containing bit errors. The check sum (CK) bits are used for this purpose. ARQ is used for error correction. Forward signal units are numbered sequentially modulo 127 (i.e., in the sequence …, 126, 127, 0, 1, …, 127, 0, 1, …) by the forward sequence number (FSN). The forward signal units are acknowledged by the receiver by inserting the FSN of the forward signal unit in the backward sequence number (BSN) field of the acknowledgment. The acknowledgment may be returned in an MSU, an LSSU, or an FISU. If there is a gap in the forward sequence number, the receiver inverts the value of the backward indication bit (BIB) and returns a negative acknowledgment containing the FSN of the last correctly received signal unit and the new value of the BIB. The new value of the BIB is maintained in the acknowledgment messages until a new error occurs. Then the BIB is again inverted. The sender will then start retransmitting signal units with the FIB inverted. Since signal units are sent all the time (SIFU if nothing else is to be sent), a particular procedure must be implemented concerning the handling of the sequence numbers. Handling of the BSN is simple. If a message does not acknowledge a new forward message, the BSN is the FSN of the last accepted message. Multiple acknowledgments of the same message do no harm. The FSN is significant only in

7.8 Example 2: Protocol Structure of SS7

195

MSUs. Therefore, LSSUs and SIFUs contain the same forward sequence number as the last transmitted MSU. Then the FSN is incremented by 1 only when an MSU is sent. The length indicator discriminates between the three types of signal units: LI = 0 is the FISU, LI = 1 or 2 is the LSSU with one or two octets in the status field, and 63 ≥ LI > 2 is the MSU. The service information octet (SIO) discriminates between international and national signaling messages and different user parts. 7.8.4

Signaling Network Layer (Layer 3)

The structure of the network layer of SS7 may be confusing. The lowest part of the layer can support the oldest and most primitive applications of SS7—namely, interconnecting signaling points with the sole purpose of serving exchanges in the network. This layer is called the message transfer part, often referred to just as layer 3. The MTP messages are composed of two fields: an address field and a payload field. For routing of signaling messages, unique numbers are allocated to the telephone/ISDN exchanges. This number is called a point code. The number consists of only 14 bits providing a total of about 16,000 addresses. First, point codes are assigned to exchanges on the international level (i.e., exchanges interconnecting national networks). Then the point codes are reused in the national networks. This implies that the national point codes must be derived from the telephone number of the called used at the national border. The address field consists of the destination point code, the origination point code, and information used for managing load sharing if more than one signal link exist between two signal points. The MTP, developed around 1980 for simple telephone applications, had to be replaced by a network layer providing flexible routing. A new sublayer called the signaling connection control part was added above the MTP, providing the capabilities of the packet switched data networks. Without the SCCP, SS7 would have been useless for all applications other than signaling in telephone networks. The SCCP enables equipment of any type to be interconnected by the signaling network. This includes support systems such as databases and computers for subscription and management, intelligent network nodes, and entities required in mobile networks (e.g., location registers and SMS centers). The SCCP offers four classes of services: •

•

• •

The basic connectionless class that works in the same way as IP (best effort delivery of messages); The sequenced connectionless class where the delivery of messages is acknowledged (secure delivery of messages); The basic connection-oriented class; The flow-control connection-oriented class.

The SCCP contains a subsystem address (SSN) offering more or less the same functions as the port number in TCP. In addition, SCCP contains other numbers for routing the call (e.g., a telephone number or an IP number of the destination).

196

Elements of Protocol Theory

Almost all signaling applications—also in GSM—use the basic connectionless class. Application processes will take care of the rare case where SCCP packets are lost. The connection-oriented classes are used for transfer of large amounts of data for purposes such as operation and maintenance, as well as database management. The connection-oriented version is also used on the signaling links between the base stations and the exchanges in GSM. This solution makes it simpler to manage several mobile terminals simultaneously at the base station, since one SCCP connection is assigned for each active mobile terminal. SCCP in the connectionless mode is used for remote cell management by the exchange. This application of SS7 solves, in a simple way, an otherwise awkward problem. 7.8.5

User Parts and Applications

The actual signaling information and the signaling procedures are realized in a signaling user part—for example, the telephone user part (TUP) and the ISDN user part (ISUP) implemented between ordinary telephone exchanges and ISDN exchanges, respectively. The user parts can be included directly in the payload field of the message transfer part or in the payload field of the SCCP as shown in Figure 7.18. The SIO of the signal unit (see Figure 7.18) indicates which of the two methods are used. Hard-coding is used for encoding the user parts. The user parts define the various signaling messages and the semantics of signaling procedures (i.e., how the different messages are interpreted and used). See Section 6.8.1, where an example of the signaling procedure is shown. TCAP was developed in order to support the application protocols required for handling intelligent network services and exchange of information between functional units in the GSM system. TCAP and the application parts are encoded using soft-coding (ASN.1—see Section 7.6.2). TCAP supports association establishment, transfer of remote operations, multiplexing, and explicit binding of communicating entities for the duration of the association. Unfortunately, TCAP does not support session management functions such as software synchronization, alternation between the roles as initiator or responder, and grouping of operations belonging to the same computational task. The absence of these functions makes the application protocol and the application procedures using TCAP more complex than necessary since the session management functions must be explicitly programmed in the application software. TCAP consists of two sublayers, as shown in Figure 7.19. The component sublayer supports the RO protocol defined jointly by ITU and ISO. This is an efficient and simple protocol for remote procedure calls. The procedure call is sent in an invoke message. The protocol offers four alternative methods for peer response: •

•

Class 1 provides a return result message containing the outcome of the execution or an error message if execution cannot take place for some reason (e.g., a wrong parameter value in the invoke message, or the command is out of sequence with the processing state of the peer). Class 2 provides no return result message if the execution is successful but an error message if execution cannot take place.

7.8 Example 2: Protocol Structure of SS7

197

Application

Protocol RO Primitives

Component Sublayer

Invoke return result error Reject

Transaction primitives

Transaction sublayer

Begin Continue End Abort Expedited Data

TCAP

Figure 7.19

•

•

TCAP.

Class 3 provides a return result message if the processing is successful but no error message if it is not. Class 4 provides no return message whatever the result of the processing.

Complex transaction procedures can be specified, combining messages of the various classes in clever ways.

198

Elements of Protocol Theory

An invoke identifier associates the invoke message with the corresponding return result and error messages. There is no association between different invokes. Reject messages can be generated both by the protocol and by the peer user. The purpose of the reject message is to indicate a situation that cannot be resolved in other ways (e.g., disruption of the connection or processing failure in the peer). Dialogs are handled by the transaction sublayer. The sublayer offers a simple connection-oriented protocol for this purpose. The dialog messages are Begin, Continue, and End supporting obvious functions. The peer entity responds to a Begin message with either a Continue or an End message: Continue if this is not the final message of the dialog or End if it is the final message. The procedure to be followed (i.e., selecting the appropriate type of message) is stated explicitly in the application software. RO messages are embedded in the transaction messages. Usually the transaction messages (also Begin and End messages) contain one or more RO messages. The transaction message may, however, be empty; for example, an empty Begin message can be used to establish a dialog, and an empty End message can be used to terminate the dialog. There are no rules disallowing this type of procedure, and there may be cases where the procedure is useful. At the beginning of the dialog, a transaction indicator is allocated to the dialog. All subsequent messages of the dialog sent in either direction carry this indicator. It is then possible to multiplex several simultaneous dialogs between the same peer entities by allocating different transaction indicators to them. The application must indicate the start and the end of the dialog to the transaction layer. The transaction sublayer offers separate service primitives to the application for these purposes. Several RO messages belonging to the same dialog can be grouped together and sent in the same transaction message. The transaction sublayer also supports the expedited data service, which allows sending of information that is not part of a dialog (e.g., exchange of management information such as flow control and buffer management). The Abort message can be used both by the protocol and by the application to disrupt an ongoing dialog. 7.8.6

Performance Requirements

Since a single SS7 link may serve several thousand calls simultaneously, the availability of the link must be very good. For this reason, the outage of a single signaling link should not exceed 10 minutes per year. This corresponds to an availability of 99.998%, where the availability is defined as A = Tup/(Tup + Tdown), where Tup and Tdown are the expected uptime and expected downtimes (e.g., per year), respectively. The availability of exchanges is the same as for the signaling system. This establishes the reliability that the equipment and the transmission systems must fulfill. There should not be more than 1 in 1010 signal units containing undetected bit errors; less than 1 in 107 signal units should be lost because of failure of the message transfer part; not more than 1 in 1010 signal units should be received out of sequence or be duplicated. This establishes the error performance of the transmission links.

7.9 Example 3: Protocol Structure of the Mobile Network

199

The processing time of the most processing-intensive signaling messages should with 95% probability be less than about 2 seconds. This is the sum of the processing times of the end systems and all signaling nodes along the signaling link. The corresponding processing time per signaling node must then be about one order of magnitude smaller (about 200 ms). This establishes the requirements for processing speed and buffer capacity of the signaling nodes and the end systems.

7.9

Example 3: Protocol Structure of the Mobile Network 7.9.1

General Radio Interface

A model of the protocol stack between the mobile terminal and the base station is shown in Figure 7.20. A similar model can be constructed for the network but the different functions above the MAC are distributed among several network entities, as we shall see later. The purpose of the MAC layer is to establish and manage the information transfer between the mobile terminal and the base station subsystem. A MAC layer is not explicitly specified in the GSM system. However, the MAC functions are taken care of by the physical layer, the data link layer, and the radio resource management function described later. The MAC offers an application programming interface (API) to the functions above it. The API may support standard interface specifications (e.g., digital subscriber line or the IP interface) or be designed for the particular application. MAC consists essentially of the physical layer and the data link layer of the radio interface protocol. MAC offers multiplexing and demultiplexing of signaling and user data, conversion of format, segmentation, error control, media synchronization, and several other functions.

User data

Signaling

Management

API

MAC

Physical channel

Figure 7.20

Basic protocol structure of the mobile terminal.

200

Elements of Protocol Theory

The functions above MAC include protocols for information transfer (user data), signaling protocols, and management protocols. These protocols may consist of several layers starting at the network layer or, as we shall soon see for the GSM system, consist only of the application layer directly interfacing the data link layer of MAC. 7.9.2 Radio Resource Management, Mobility Management, and Media Management

The GSM system is chosen as an example because the subdivision of the functionality of the signaling system is particularly clearly specified in this system. The architecture of GSM is found in Section 8.4.3 (Figure 8.10). The essential network elements are the base station transceiver (BTS), the base station controller (BSC), the mobile-services switching center (MSC), the VLR, and the HLR. The figure can easily be modified to describe any other mobile system. If we replace BTS by node B, BSC by radio network controller (RNC), MSC plus VLR by serving GPRS support node (SGSN), HLR by HSS, and add GGSN to the media management system of the network we get the configuration of an all-IP UMTS system. If we keep the structure of GSM and add RNC, SGSN, HSS, and GGSN, the architecture is that of UMTS offering both circuit-switched services and IP services. The signaling architecture of the GSM system is shown in Figure 7.21. The circles indicate the entities analyzing and acting upon signaling messages of the three types: media management, mobility management, and radio resource management. The first important observation is that these functions are distributed among several entities in the network. The three functions also interact in a complex way in the network, but this is not shown for simplicity. The ellipse denoted “other entities” includes authentication and encryption escrows, servers for SMS and similar services, and equipment identity registers. Several entities of the same kind may be involved in executing a single task (for simplicity of drawing, this is not shown in the figure). A single handover in GSM, for example, may involve as many as two BTSs, two BSCs, three MSCs, and two VLRs. A location updating may involve two VLRs,

Mobile terminal

BTS

BSC

Media management

Mobility management

Radio resource management

Figure 7.21

Functional architecture of GSM.

MSC

VLR

HLR

Other entities

7.9 Example 3: Protocol Structure of the Mobile Network

201

one HLR, and one authentication center, and sometimes also a mobile equipment register. What is not shown in the figure is that all the entities in the network will interact, sometimes even in real time, with the operations and management systems of the mobile operator. This adds even more complexity to the distributed processing system the GSM network entails. The architecture is of the “plug-and-play” type. The interfaces between all entities and functions are defined by primitives, as described earlier. Any one of the entities may be replaced by other entities, and the system may still operate properly provided that the interface is not changed. The three entities in the figure offer services as follows. Radio resource management is responsible for establishing, maintaining, and releasing radio connections on demand. This includes management of frequencies, time slots, channel configuration, power control and synchronization, spectrum spreading codes, and frequency hopping codes. Dissemination of system information that the mobile terminals need in order to adjust the radio parameters may include the identity of neighboring cells, identity of location area, power level indication, channel configuration, and several other system-dependent parameters. Radio resource management is also responsible for executing handover. Other system-dependent functions include power-saving functions such as sleep mode management and discontinuous transmission of speech signals. Sleep mode implies that, when in the idle state, the mobile station need listen to not every paging channel, but only to some of them where the scheduling is derived from the international identity of the mobile terminal. Discontinuous transmission implies that the mobile station does not send any signal during speech pauses. The power saving is considerable, since on average the speech channel is active only 40% of the time in one direction. Mobility management is responsible for management functions such as location updating, identity management, access rights management, anonymity and untraceability management, encryption of the radio path, terminal and network authentication, and selection of mobile network. Most of these functions are not required in fixed telecommunications systems. The SIM card is the most important component of the mobility management function. Media control is a composite function handling telecommunications services such as speech, data communication (IP), video, MMS, and SMS. The point is that these services are implemented in more or less the same way in all fixed and mobile networks so that standard software can be exploited. The media management function must first request the mobility management function to establish the radio connection and execute the mobility management services before the call is set up in the standard way. 7.9.3

Protocol Stacks

The protocol structure of GSM is shown in Figure 7.22. All protocols above the dotted line are application layer protocols. The protocols within the box (e.g., MC and MM) are not hierarchical protocol layers but independent protocols on the same level of the application layer. All other protocols are layered (e.g., MAP, TCAP,

202

Elements of Protocol Theory MS

BTS

MSC/VLR

BSC

HLR

MC

MAP

MM

TCAP

RR

RR BTSM

BSSMAP DTAP SCCP

SCCP

LAPDm

LAPD

MTP

MTP

Radio

Any

L1

L1

MC = media control

MAP = mobile application part

MM = mobility management

TCAP = transaction capabilities AP

RR = radio resource management

SCCP = signaling connection control part

BTSM = BTS management

MTP = message transfer part

BSSMAP = BSS management part

L1 = layer 1

DTAP = direct transfer AP

LAPDm = LAPD mobile LAPD = link access protocol D

Figure 7.22

Protocol structure in GSM.

SCCP, MTP, and L1 form a protocol consisting of five layers, while MM, LAPDm and the physical radio protocol form a three-layered hierarchical structure). The media control protocol (MC) and the mobility management protocol (MM) are extended transparently to the MSC and the VLR. The radio resource management is analyzed by both the BSC and the MSC, depending upon the task that is to be performed. Information on MC, MM, and RR that are to be sent transparently between the MS and the MSC are embedded in the direct transfer application part (DTAP) on the link between the BSC and the MSC. The VLR cooperates with the HLR in order to manage several procedures such as location updating, call establishment, and management of certain supplementary services. These procedures are included in MAP. MAP makes use of TCAP, described in Section 7.8. MAP is also used between the MSC/VLR/HLR and other entities—for example, SMS centers; authentication centers; operations, management, and maintenance (OAM) centers, and equipment registers. This is not included in the figure. The base station subsystem (BSS) is controlled by separate procedures implemented in the protocols BTS management (BTSM) and BSS management part (BSSMAP). These procedures include assignment of frequencies and time slots, management of handover, and management of broadcast information to be passed to the mobile terminals. Below the dotted line there is a number of different protocol stacks. On the radio path and between the BTS and the BSC, the application protocols MC, MM, and RR are embedded directly into the information field of the data link protocol. LAPD is a data link protocol of the HDLC type numbered D to distinguish it from other data link protocols numbered LAPA (or just LAP), LAPB, and so on. LAPDm indicates that this is a LAPD protocol particularly adapted to the TDMA burst structure of GSM. In the rest of the network, SS7 is used with the SCCP as network protocol.

CHAPTER 8

Cellular Land Mobile Systems 8.1

What Is a Cellular Network? The term cellular network denotes a system where mobile terminals (in the figures usually abbreviated as MS—mobile station) can roam between cells controlled by different base stations and receive and initiate calls independently of where the mobile terminal is located in the area covered by the network. Roaming is the trademark of cellular systems. Without the roaming capability, the system is not cellular. A wireless LAN is not a cellular network because it does not support roaming. The cellular network and the mobile terminal—but not the user of the terminal or anyone calling the mobile terminal—keep track of where the mobile terminal is located in the network at any time. Location in this context is not geographical coordinates but is defined relative to an abstract coordinate grid where base stations are the coordinate points. This automatic location service is the basis for roaming. In addition, the access technology and other call procedures are the same everywhere in the system, allowing all mobile terminals to access the system at any base station. The radio interface and the access and calling procedures are then in accordance with a systemwide specification or standard. A cell is served by a single base station. Therefore, the terms base station and cell are used synonymously throughout the text. The radio interface at the base station is the transceiver; that is, the combination of a transmitter and a receiver. A sectorial cell is just a sector of the area that can be covered from a single base station site. Each sectorial cell is served by a separate transceiver and is regarded, from a system viewpoint, as a separate base station. The form of the sector is determined from the radiation pattern of the antenna serving the sector. The base station site may thus contain a single transceiver or several transceivers. The concept is illustrated in Figure 8.1. The base station supports the radio interface toward the mobile terminals and the interface toward the telephone network or the Internet. The base station is part of the access network. Handover is an additional feature where a call is not disrupted if the user moves out of the coverage area of one cell and into the coverage area of another cell. The call is automatically moved from the original cell to the new cell without involvement of the user. We shall see later how this can be achieved. Handover takes place only when the mobile terminal is active in a call. Calling procedures in the cellular network are automatic and performed in the same way as in the fixed telephone network or the Internet.

203

204

Cellular Land Mobile Systems

Base station site

Transceiver base station

Sectorial cell

Cell

Figure 8.1

Base stations and cells.

Public land mobile network (PLMN) was the name given by ITU in 1980 to cellular networks offering services to everyone on an international basis. The PLMN is owned by many independent operators offering the same set of services to visiting users and their own customers. Examples of PLMNs are GSM and UMTS. Ownership shall not restrict the roaming capabilities in the cellular system. One exception to this rule is that GSM and UMTS operators may deny access from mobile terminals of competitors operating in the same geographic area. The regulatory authorities may decide otherwise, in particular, allowing small operators not having national coverage to access the competitor’s network in geographical areas where they do not have coverage in order to avoid undue execution of competitive power. Regulation of the mobile market is a complex matter that is outside the scope of this book. The frequency bands used for land mobile communication are in the vicinity of 1 GHz (second and third generation) and 2 GHz (third generation).

8.2

A Brief History of Public Land Mobile Telecommunications The first cellular system to become operational was NMT. In 1981 the system commenced operation in Denmark, Finland, Norway, and Sweden. Before that, all land mobile communication had been manual (the manual service in Norway was put into operation in 1967). During the early 1980s, similar mobile systems were put into operation in several countries in Europe and elsewhere. Some countries chose the NMT system, while other countries developed their own systems (C-Netz in Germany, TACS in Britain, Radiocom 2000 in France, AMPS in the United States). The technical invention that made these systems possible was the development of the microprocessor in 1975. Before the invention of the microprocessor, automatic mobile terminals would be too large and heavy for practical use. An attempt was made in 1981 by France, Britain, the Netherlands, Switzerland, and the Nordic countries to introduce NMT as a pan-European standard, but this

8.2 A Brief History of Public Land Mobile Telecommunications

205

failed because Britain introduced competition on mobile communications in early 1982. This forced the U.K. to choose a system concept with which none of the competitors were familiar, and the first attempt to develop a pan-European network came to an end. The Netherlands then proposed to the European standardization and cooperation organization for telecommunication and postal services CEPT that a new system should be developed for Europe. During the summer of 1982, representatives from the Netherlands and the Nordic countries wrote the mandate for a new study group (GSM) in CEPT dedicated to this task. The meeting also drafted the important document later known as GSM 2/82 specifying the baseline elements of the GSM system. The GSM specification was pioneering work, and several of the methods and concepts invented there are used, in slightly modified forms, in new systems. However, several of the basic constituents of a land mobile network were defined by ITU a few years earlier (1980–1981). This included the basic architecture of land mobile systems adopted for GSM and later the universal mobile telecommunication system, the procedures for location management, the specification of a global identification plan for mobile terminals, and the handover method presently applied in GSM. The first outline of a protocol interconnecting the different parts of the network was also the result of this early work. The protocol became later known as the mobile application part. The United States and Japan did not take part in the GSM development and, unlike most of the world, did not implement the system in general. It was suggested already in 1986 to specify a global land mobile system that could replace GSM in the future. A long and crooked development resulted in UMTS. The system was ready for commercial implementation around 2001 but was delayed by the general problems the telecommunications industry faced after the dot.com boom and the following collapse of the ICT economy. The early versions of UMTS are based on telephone applications. However, a new version based on an all-IP technology is now about to be realized. UMTS is labeled the third generation public land mobile system (3G), while GSM is the second generation system (2G) to distinguish them from the earlier systems NMT, TACS, AMPS, and so on. GPRS was developed during the second half of the 1990s as an enhancement of GSM, offering Internet services at data rates up to about 100 Kbps by combining the GSM channels in a new way. Internet services to mobile terminals could then be offered before the UMTS was ready for operation. GPRS is also known as the 2.5G system, signifying that it is a system that belongs to the transition generation between GSM and UMTS. EDGE is also a technology based on GSM but where the bit rate is increased by more efficient modulation methods. The first versions of UMTS were base on the GPRS/GSM architecture. The major difference between UMTS and GPRS/GSM is the way in which the radio interface is designed. New trends now are to combine WLAN (WiFi) and PLMNs (2G/2.5G/3G) in the same mobile terminal, enabling the terminal to choose WLAN in certain “hotspots” and PLMNs elsewhere. This provides the user with cheaper services and possibly larger bandwidth.

206

8.3

Cellular Land Mobile Systems

Radio Propagation in Land Mobile Systems 8.3.1

Large-Scale Variations: Basic Wave Propagation Theory

If there is no obstacle between the sending and receiving antennas, the signal power drops off at a rate proportional to the square of the distance between the antennas. The reason is that the signal power is distributed evenly over the surface of a spherical sector (which is the entire sphere if the antenna is omnidirectional). It is well-known from geometry that the area of a spherical sector of any shape is proportional to the square of the radius of the sphere; that is, the power at a distance d from the transmitter is P/4 d2, where P is he transmitted power. It can be shown from electromagnetic wave theory that the gain of an ideal omnidirectional receiver antenna for a signal with wavelength is 2/4 so that the received power is

(

)

Preceived = P λ2 / 4π / 4πd 2 = P( λ / 4πd )

2

The free-space loss is then L free − space = Preceived / P = ( λ / 4πd )

2

If the antennas are close to the ground, part of the wave energy will be reflected by the ground and the reflected and the direct waves will interfere at the receiver as shown in Figure 8.2. It is easily seen from the figure that the distances the direct 2 wave and the reflected wave are traveling are D 2 = d 2 + (h1 − h 2 ) and

R 2 = d 2 + (h1 + h 2 ) , respectively. This gives rise to a phase difference between the 2

interfering waves equal to φ = k(R − D), where k = 2π / λ is the wave number. If h1 and h2 are much smaller than d, then the phase difference becomes φ = 2 kh1 h2 / d

This is easily seen since R − D = (R 2 − D 2 ) / (R + D) ≈ 4h1 h 2 / 2d = 2h1 h 2 / d and D ≈ R ≈ d far away from the transmitter. When the two waves interfere, the resulting wave is à cos( ωt + θ ) = A D cos ωt + ΓA R cos( ωt + φ)

h1 h2

d

Figure 8.2

Geometry of propagation path.

8.3 Radio Propagation in Land Mobile Systems

207

where AD and AR are the received amplitudes of the two waves, is the reflection coefficient, AD and AR are the amplitudes of the signal at a distance D and R from the transmitter, respectively, subject only to free space attenuation, and is the phase difference calculated perviously. The phase angle is not important in the calculation, since it does not have an effect on the received signal power. In the frequency bands used for mobile communication, the reflection coefficient of the Earth 1 = −1 means that the surface reflects the is = −1 for small angles of incidence. wave perfectly but rotates the phase angle of the wave by 180°. The amplitude reduction over a distance x is the square root of the free space loss derived previously since the power of the signal is proportional to the amplitude squared. The amplitude reduction is then λ / 4πx = 1 / 2kx. It then follows that AD = A × (amplitude reduction over distance D) = A / 2kd ≈ A / 2kd and, similarly, AR = A / 2kR ≈ A / 2kd, since D ≈ R ≈ d far away from the transmitter. A is the amplitude of the transmitted signal. After adding the two waves and calculating the new amplitude, we get à = ( A / 2 kd )φ = Ah1 h2 / d 2

since sin ≈ if the phase difference between the interfering waves is small ( << 1). The power of the received signal falls off at a rate L grazing = Preceived / P = Ã / A 2 = ( h1 h2 ) / d 4 2

2

as a function of the distance d between the transmitter and the receiver, since the signal power is proportional to the square of the amplitude of the signal. The conclusion is that multipath interference at small angles of incidence (grazing angles) causes the power of the received signal to drop off at rate equal to the fourth power of the distance. We may assume that the antenna height of the mobile terminal is small (order of 1 or 2 meters). From the footnote we see that if the height of the base station antenna is large, free-space propagation may be expected in a considerable part of the cell. On the other hand, if the antenna height at the base station is small, the fourth power law holds for the entire cell except for a very short distance from the base station. This is shown in Figure 8.3.

1.

For vertically polarized waves (the commonly used polarization in land mobile systems), the reflection coefficient is Γ( θ) = ( sin θ − X) / ( sin θ + X), where ε2 X2 = ε− cos 2 θ, ε is the dielectric constant of the ground, and is the angle of incidence of the reflected wave. From Figure 8.2 we see that tgθ = (h1 + h2 ) / d . The dielectric constant for the ground varies between 3 and 25 depending upon the texture and humidity of the ground (approximately 3 for loose gravel and 25 for wet Earth). X then varies between 0.2 and 0.5 for small angles of incidence (cos ≈ 1). We see from the formula that = −1 as claimed for = 0. The grazing condition is then equivalent to sin θ ≈ θ << X = 0.2. This holds for grazing angle around 1° (≈ 0.02 radians) and less. Since tgθ ≈ θ for small angles, we find that the fourth-power law holds for d > 50(h1 + h2 ), where the lower limit corresponds to the grazing angle of 1°. For heights of the two antennas of 50m and 3m, respectively, we find that the fourth-power law holds for distances longer than about 2.6 km. If both antennas have the height of 2m, the distance is less than 200m before the fourth-power law becomes effective.

208

Cellular Land Mobile Systems

2

Low base station antenna

Received power level

1/d

High base station antenna

Receiver threshold

4

1/d

4

1/d

Distance d from base station Cell radius

Figure 8.3

Effect of antenna height on cell size.

From the figure we observe that if we want to make the cell large, we shall find high locations for the base station: high buildings and hilltops. If we want small cells, the base station antenna should be placed close to the ground. This provides us with an efficient tool for engineering the cell size just by choosing the appropriate base station antenna height. The total traffic capacity of a geographic area is roughly proportional to the number of cells covering the area, since each cell can carry the same traffic capacity irrespective of its size. The capacity of the cell depends only on the frequency band (or in DS-CDMA systems on the code space) assigned to it. On the other hand, the total investment cost of the mobile system is proportional to the number of cells. In urban areas where the traffic capacity is the concern, the cells can be made small by placing the base stations close to the ground; in rural areas where the cost is the concern, cells can be made large by placing the base station on hilltops. It is also easy to increase the traffic capacity of an area by finding new locations for the base station antennas. This has been done many times during the evolution of the land mobile service since the start in 1981. Already after one year of operation of the NMT system, the cell structure in the Nordic capital cities were altered in this way. The packing density of cells depends on the distance between cells that is required in order to reuse the same frequency. This distance depends on the interference produced in one base station by mobile terminals in the other cell and vice

8.3 Radio Propagation in Land Mobile Systems

209 2

versa, which is a factor of both propagation and system parameters such as modulation method and multiple access technique. Since the ground is not a perfect dielectric medium but has nonzero conductivity, one component of the wave will propagate along the surface. This wave component is called the surface wave. The surface wave also gives rise to a fourth power law, but the effect of the surface wave is much weaker than the interference caused by the reflected wave at the frequencies used for mobile communications. If there is an obstacle between the antennas such that there is no direct path between them as shown in Figure 8.4, diffraction makes the signal still detectable below the horizon created by the obstacle. The obstacle does not produce an absolute shadow. The effect is called edge diffraction. It turns out that the signal power drops off approximately at a rate of 1/d4 in this case also. In addition, other terms do not depend directly on the distance between the antennas but on small-scale parameters such as height of buildings and the distance between them. We may then conclude that the fourth-power law is the predominating propagation mechanism in urban areas. The propagation models for real cells are much more complex than described here because of the shape of the terrain, reflections from buildings, effect of foliage, variation of reflection coefficient due to moisture, and so on. In the simple Lee model, the signal strength Preceived at a distance d (in km) from the transmitter can be expressed approximately as Preceived = P − A − B log d + C log h

Edge diffraction

d

Figure 8.4

2.

Edge diffraction.

Observe that if the signal had dropped off at a rate proportional to the distance squared for all distances, then the interference from all other cells would approach infinity provided that the density of users would be the same in all cells. The number of users at a distance r from a base 2 station would then be proportional to 2 r . These users would cause interference proportional 2 2 to 2 πr / r = 2 π = constant. Integrating over all distances, this will result in an interference power ∫ constant dr → ∞. Land mobile communication is impossible in this case. The fourth power law gives an interference proportional to 2 πr 2 / r 4 = 2 π / r 2 . The interference is now finite and equals ∫ c / r 2 dr ~ 1 / d , where d is the diameter of the cell.

210

Cellular Land Mobile Systems

where P is the transmitted power and h is the base station antenna height in meters. Preceived, P, A, B, and C are given in decibels. C is approximately 20 dB. A and B are parameters depending on the shape of the terrain, location of reflecting surfaces, and blockages in the transmission path. A is approximately 130 dB for GSM, and B is approximately 40 dB in urban areas. In linear units this is Pr ≈ Pt Eh 2 / d 4 , where 20 log E = − A. B is between 20 dB and 40 dB in rural cells, where the first value is a good approximation near the base station while the second value applies to large distances. Several other propagation models for estimating the size of the cell have been proposed, but none of them are accurate enough to prevent that the actual coverage of a cell often must be measured explicitly. However, it is beyond the scope of this book to pursue the complex propagation problems of land mobile systems any further.

Small-Scale Signal Variations: Fading The complex propagation of the radio waves described previously gives rise to a spatial variation as illustrated in Figure 8.5 for one direction of propagation. All other directions of propagation are similar, giving rise to a complex spatial distribution of the power level consisting of high peaks and deep troughs quite close to one another. What makes this distribution so complex is that the radio signal at any point is composed of a large number of direct, refracted, and reflected signal components arriving with different amplitudes and phases at the receiver as illustrated in Figure 8.6. This phenomenon is called multipath interference. The receiver adds linearly all the received signal components or, using different terminology, the signal components interfere linearly in the receiver. The form of the signal (in complex variables) received at the point p in space can then be expressed as follows:

Average propagation loss

Received signal level

8.3

Multipath loss

Distance

Figure 8.5

Signal level as a function of distance.

8.3 Small-Scale Signal Variations: Fading

211

Σ

Figure 8.6

Multipath channel.

A( p)e

j ( ωt + θ ( p ) )

= ΣA i ( p)e

j ( ωt + θ i ( p ) )

A(p) and (p) are the amplitude and the phase of the resulting signal at point p, and Ai(p), and i(p) are the amplitude and the phase of the individual carriers that reach the receiver at point p. As a function of p—for example, represented by the Cartesian coordinates (x,y,z)—the spatial field pattern is an irregular pattern of peaks corresponding to constructive interference and troughs corresponding to destructive interference, where the distance between peaks and troughs can be as small as a quarter of the wavelength (7.5 cm at 1 GHz). A(p) and (p) are stochastic functions of the point p. The propagation channel is called a multipath channel, since the signal components follow a large number of different paths from the transmitter to the receiver. The spatial variations of the signal are converted into temporal variations called fading when the mobile terminal moves in the spatial field-strength pattern. The fading (as well as the spatial distribution) is given by the probability density function called the Rayleigh distribution if there is no direct wave component present. If a direct wave component is present, the fading is shallower. This is called Rician fading. The formulas for these distributions are not important for us. What is important for us is the impact the velocity of a terminal moving through the spatial field pattern has on the bit error performance of the channel. Figure 8.7 shows the impact of fading on the received signal if the mobile terminal moves with small velocity. Figure 8.8 shows the fading pattern if the velocity of the mobile terminal is large. If the terminal moves slowly, the error pattern consists mostly of error bursts; if the terminal moves rapidly, the error pattern consists mostly of single errors. The GSM system is designed for velocities between 0 and 500 km/h (rapid trains) and must therefore be able to detect and correct both burst errors and single errors. The error correction method used on the radio channel must then be capable of handling both burst errors and single errors with reasonable quality in both cases. The maximum density of fades can be estimated as follows. If two equally strong waves received from opposite directions are 180° out of phase at a point p, they will also be out of phase at the point p + /2. This gives rise to a fading pattern with minima spaced /2 apart. This corresponds to a distance of 15 cm at 1 GHz. If

212

Cellular Land Mobile Systems

Detection threshold

Bit errors

Figure 8.7

Rayleigh fading for low speed.

a mobile terminal moves at a speed of 2 m/s (7.2 km/h), a fade will occur every 75 ms; if the mobile terminal moves with a speed of 30 m/s (108 km/h), a fade will occur every 5 ms. Deep fades causing transmission problems are much less frequent than this. The problem arises at the edge of coverage where the average field strength is low.

8.4

The PLMN Architecture The PLMN architecture that all public land mobile systems should support was developed by the ITU already in 1981. The architecture was developed for supporting telephone applications and the connection-oriented packet switched data service called X.25 specified by the ITU. The first system applying this architecture was the GSM system. The GSM architecture supports circuit switching. The same architecture is then applied for circuit switching in UMTS. GPRS requires a different architecture that is suitable for packet switching. GSM/GPRS and UMTS therefore support both architectures at the same time in order to interface both the circuit-switched telephone

Detection threshold

Bit errors Time

Figure 8.8

Rayleigh fading for high speed.

8.4 The PLMN Architecture

213

network and the packet-switched Internet. However, the all-IP version of UMTS requires only the packet-switched architecture. The main elements of the architectures of GSM, GSM/GPRS, and UMTS are shown next. Only the part of the architecture playing an essential role in call management is included. 8.4.1

Objectives

The PLMN architecture must support the following general objectives: •

•

•

•

•

•

Universality. A user of the network shall have access to all parts of the network irrespective of who owns the particular part of the network. Single subscription. In order to achieve universality, the user needs just a single subscription contract with only one operator. Continuous radio coverage. GSM and UMTS offers continuous radio coverage, while wireless LANs do not. The access of WLANs is confined to isolated cells. We are also referring to GSM and UMTS as systems offering continuous mobility while WLAN and mobile IP only offers discrete mobility.3 Automatic location updating. This means that it is the responsibility of the network to keep track of where the mobile terminal is located in the network at any time. Location updating should not require any intervention by the user of the mobile terminal. Automatic handover. This means that if a mobile terminal moves form one cell to a neighboring cell during conversation, the call shall be handed over to the new cell automatically without disrupting the ongoing conversation. Automatic call routing. A user anywhere in the telecommunications network calling a mobile user shall only know the telephone number (or IP number) of the mobile user in order to place the call. A mobile user shall not need additional procedures in order to call another user in the telecommunications network.

Later, we shall come back to how some of these functions are implemented in GSM, GPRS, and UMTS. 8.4.2

Topology

The topology of public land mobile system is shown in Figure 8.9. The purpose of this topology is to define system coordinates from which the mobile terminals can determine where they are located within the network. The system is divided into networks that are the entities owned by the operators. A globally unique identity code is assigned to each network. Each network is subdivided into location areas also identified by a unique identity code within the 3.

WLAN may offer continuous mobility since it is a radio technology not very different from All-IP UMTS. WLAN and PLMN technologies may merge in the 4G development or in initiatives such as unlicensed mobile access (UMA) and open broadband access network (OBAN).

214

Cellular Land Mobile Systems

System Network

Location area Cell

Figure 8.9

PLMN topology.

network. The location area is the accuracy by which the location of the mobile terminal needs to be known in order to route calls to the mobile terminal. Finally, the location area is divided into cells, each being identified by a unique code within the location area. The location coordinates of a mobile terminal are then given by three numbers: network identity, location-area identity, and cell identity. If all three numbers are provided to the mobile terminal, the terminal will know where it is located relative to a coordinate system where cells are the coordinate points. The unique coordinates identifying a single cell globally may be written in the following coordinate format: (network identity, location-area identity, cell identity). The network identity must be globally unique. Usually the location area coordinate must also be globally known or derivable from other information on a global basis (for example, the VLR identity in GSM). Otherwise, global roaming is not possible. 8.4.3

Architecture of GSM

GSM is a telephone (ISDN) system. Interworking with the data networks takes place in the same way as in the ISDN. The interface, not shown in the figure, between data networks and GSM is implemented in the MSC. The architecture of GSM is shown in Figure 8.10. The various entities are as follows: • •

•

•

The BTS supports the radio interface in a single cell. The BSC supports radio signaling management, signal reformatting, and access management of several BTSs. The MSC is a telephone exchange. The MSC also controls a number of BSCs and supports handover between BTSs of the same BSC, between BSCs connected to the same MSC, and between BSCs connected to different MSCs. The handover procedures are explained in Section 8.7. Any exchange in the telephone or GSM network that recognizes mobile telephone numbers and is equipped with the capability to access the HLR for routing assistance may act as a gateway exchange (GTW). However, in all existing

8.4 The PLMN Architecture

215 Auxiliary units HLR BSC

VLR GTW

MS BTS BTS

BTS BTS Permanent association Temporary association

Figure 8.10

•

•

4.

BSC

MSC

Fixed telephone network

MS Mobile terminal BTS Base transceiver station BSC Base station controller MSC Mobile-services switching center GTW Gateway VLR Visitor location register HLR Home location register

GSM architecture.

networks, the GTW is an MSC in the home network of the mobile terminal, since telephone exchanges other than MSCs are not equipped with the capability to access the HLR. The GTW is required only for routing calls to the MSC where the mobile terminal is located. The VLR is in charge of all functions related to mobility management. This includes storing subscriber information (number and identity, service options, subscription information, identity of the HLR in which the mobile terminal is registered, call restrictions), location updating of the HLR, management of temporary identities, authentication of the SIM in the mobile terminal, supporting routing of calls to mobile terminals, and forwarding short messages. The VLR remotely controls the switching functions of the MSC. The VLR may control several MSCs, but in the current network the VLR is colocated with a single MSC. The VLR usually controls just one location area, but it may be in charge of several location areas. This is an implementation detail. The HLR is a subscription database storing all information required for call management. Information contained in the HLR is the telephone number allocated to the user, the identity of the SIM—called the international mobile subscriber identity (IMSI)—service options and usage restrictions,4 and the location of the mobile terminal (the network identity and the location-area identity coordinates defined previously).

For example, call forwarding, roaming restrictions, barring of outgoing calls to certain destinations, barring of incoming calls, and charging tariff and method.

216

Cellular Land Mobile Systems

•

•

8.4.4

Auxiliary units include an equipment identity register where the identity (serial number) of the mobile terminal hardware is stored,5 one authentication center for each HLR storing secret keys and algorithms for authentication of SIMs and encryption of the radio path, and units supporting the short message services (SMS and MMS). One particular set of auxiliary entities are related to the so called customer applications for mobile network enhanced logic (CAMEL) services. These entities have the same functionality as the SCP in the fixed telephone/ISDN network (see Section 6.1.7), thus extending intelligent services to mobile terminals. CAMEL services are also offered in UMTS systems supporting circuit-switches services. Location Management and Call Handling in GSM

The location coordinates (network identity, location-area identity, cell identity) are sent continuously on the downlink (toward the mobile terminal) from the base station. When the mobile terminal discovers that the network identity and/or locationarea identity have a new value, the mobile terminal knows that it has entered a cell controlled by a new VLR. The mobile terminal then—without informing the user of the terminal—sends a request for location updating to the new VLR. The VLR checks that the mobile identity is not already registered. This is the case if the VLR controls a number of location areas and the updating takes place between two such areas. The HLR of the mobile user is then not updated. If the mobile terminal is not registered, the VLR must update the HLR about the new location of the mobile terminal. From the identity of the mobile terminal, the VLR deduces in which HLR the mobile terminal is registered. As part of the updating procedure, the HLR returns all information the VLR needs in order to handle calls to and from the mobile terminal. This information may include particular call handling such as conditional call barring. The location updating procedure is somewhat more complex in order to support security. There are two types of security mechanisms that contribute to this complexity. Access is not allowed before the mobile terminal has been authenticated. Therefore, the mobile terminal must be authenticated before the HLR is updated. The second security mechanism is that the international identity (IMSI) is not sent on the radio path in order to prevent tracing of the mobile terminal. Instead, the VLR allocates a temporary identity (TMSI) sent to the mobile terminal in encrypted form. The mobile terminal uses the TMSI as identity on the radio path. When the mobile terminal updates itself to a new VLR, it sends the TMSI allocated by the previous VLR plus the identity of this VLR in the updating message. The new VLR then obtains the international identity (IMSI) and the set of

5.

The IMSI identifies the SIM and not the terminal in which the SIM is used. Therefore, a unique international mobile equipment identity (IMEI) is assigned to each mobile terminal for identification of the hardware unit. The IMEI can be used for identifying faulty equipment, prohibiting unlicensed hardware equipment to access the network, identifying particular types of hardware (e.g., test facilities), and tracking stolen equipment.

8.4 The PLMN Architecture

217

1 Update location

5 Update location New VLR

4 Authentication

MS

HLR

6 Ack<service para, auth para>

7 Ack 2 Provide IMSI

3 Ack 8 Cancel location

Previous VLR

Figure 8.11

Location updating procedure in GSM.

parameters required for authenticating the mobile user from the previous VLR. The new VLR authenticates the mobile user before the update location message is sent to the HLR. The procedure is shown in Figure 8.11. The number associated with each message shows the order in which the messages are sent and <⋅⋅⋅> represents the set of parameters in the message. The authentication parameters received from the HLR in message 6 are used by the new VLR for authentication of the mobile terminal the next time it attempts a call. A new TMSI is allocated to the mobile terminal in message 7. The HLR requests the previous VLR to cancel the location information (event 8). The location information is used for establishing calls to mobile users. The procedure is shown in Figure 8.12. The procedure is used not only in GSM but also in other circuit-switched mobile networks (circuit-switched versions of UMTS). The call contains the telephone number of the called mobile user. This number allows the telephone network to route the call to a GTW that can identify the HLR in which the mobile user is registered. The gateway exchange is usually an MSC. The HLR is identified by the first few digits of the telephone number of the mobile user. The gateway exchange then requests the HLR to provide a number that can be used for forwarding the call to the MSC where the mobile user is located. The HLR must get this information from the VLR controlling this MSC. The HLR knows the

HLR 3 Provide routing number

2 Provide routing info

4 Ack

7 Provide TMSI

5 Ack VLR GTW 1 Call

8 Ack Mobile user (MU) 9 Page

Figure 8.12

MSC

Telephone network 6 Call

Establishment of calls to mobile user.

Calling user

218

Cellular Land Mobile Systems

identity of the VLR from the location data. The HLR then sends a request for routing number to the VLR using the IMSI of the called mobile user as reference. The VLR allocates a routing number temporarily to the mobile user and returns it to the HLR (messages 3 and 4 in the figure). The routing number is an ordinary telephone number drawn from a pool of telephone numbers uniquely identifying the telephone network access point of the MSC. The VLR stores the routing number against the IMSI of the mobile user. The HLR forwards the routing number to the gateway exchange that uses this number in the call message toward the MSC (message 5). When the call arrives at the MSC, the MSC requests the VLR to provide the temporary mobile subscriber identity (TMSI) to be inserted in the paging message (sequences 7, 8, and 9 in the figure). The VLR then removes the routing number from the location table and hands it back to the pool of routing numbers. If the call does not arrive at the MSC within a timeout (say, one minute), the roaming number is also handed over to the pool. Routing of calls from mobile users is straightforward, since the MSC is an ordinary telephone exchange that delivers the call directly to the network. Before the call is accepted, the MSC requests the VLR to check whether the call shall be handled in a particular way (for example, call barring, number translation, call waiting) or be set up in the normal manner. 8.4.5

Architecture of GPRS

GPRS is an abbreviation for general packet radio service. A packet radio service is an IP-type service on the radio path, where information is sent in individual packets. In Section 8.5 it is explained how the information transfer is organized on the radio path in the GSM system and in the GPRS system. GPRS offers a direct interface to the Internet so that data packets are not routed via the MSC.6 The maximum data rate supported by GSM is 14.4 Kbps on a single TDMA channel. GPRS combines several TDMA channels so that data rates up to 115.2 Kbps, corresponding to all eight TDMA channels, can be provided. The net bit rate available to the user is a little less because of coding overhead required for organizing the time slots. Practical data rates available to the user are, for example, 13.4 Kbps (one TDMA channel), 40.2 Kbps (three TDMA channels), 53.6 Kbps (four TDMA channels), and 107.2 Kbps (eight TDMA channels). The data rates on the downlink and uplink may be different providing an asymmetric service (e.g., for a Web access requiring little capacity on the uplink but larger bit rate on the downlink). This is just a question of organizing the TDMA time slots in the two directions. The data rates and the channel configuration are settled during the initial signaling between the mobile terminal and the network. The architecture of GPRS is shown in Figure 8.13. The purpose of GPRS is to provide an Internet connection to the mobile terminal. This is done by adding three

6.

By using more complex coding, higher data rates are achieved such as in EDGE. The overall structure of the radio channel is very complex, offering much flexibility for ingenious channel coding.

8.4 The PLMN Architecture

219 Auxiliary units HLR

BSC PCU

VLR

GTW

MS BTS BTS

BSC

MSC

PCU

Fixed telephone network GGSN

SGSN

Internet

BTS BTS Permanent association Temporary association

Figure 8.13

PCU Packet control unit SGSN Serving GPRS support node GGSN Gateway GPRS support node See Figure 8.12 for other acronyms.

GPRS/GSM architecture.

new entities to the original GSM architecture: the packet control unit (PCU), the SGSN, and the GGSN. The support nodes are Internet routers containing additional control procedures. The PCU supports the channel encoding required in order to format the packet radio channel. The PCU is a supplement to the BSC and must cooperate with the BSC in order to allocate radio channels. The SGSN is an Internet router supporting the same functions as the VLR, including communication with the HLR. The SGSN obtains instructions concerning how the calls are to be handled from the HLR. The HLR must then, of course, be enhanced with subscription information required for Internet communication. The SGSN also instructs the PCU concerning required data rate and other information for managing the radio path. The GGSN is also an Internet router. The GGSN is located in the home network of the mobile terminal, and the IP number of the mobile terminal is allocated to that router. The Internet directs all calls destined to the mobile terminal to the GGSN of that terminal. SGSN and GGSN apply mobile IP for location updating and call handling. Mobile IP is described later. 8.4.6

All-IP UMTS

The all-IP UMTS or 3GPP Release 5 is shown in Figure 8.14. 3GPP releases prior to Release 5 were essentially the same as GPRS/GSM, where the telephone procedures were supported by an infrastructure similar to that of GSM while the Internet services were handled in the same way as in GPRS. The difference between the early versions of UMTS and GSM is mainly on the radio interface applying other multiple access and channel coding techniques offering larger bandwidth per channel and more flexible channel coding. The role of the RNC is also different from the BTSs in GSM. The RNC offers a combination of the functionalities of the BSCs and the MSCs. The RNCs of a region are interconnected so that calls can be passed transparently between them. The

220

Cellular Land Mobile Systems Auxiliary units HSS Fixed telephone network

RNC

MGW

UE Node B

RNC

SGSN

GGSN Internet

Node B Permanent association Temporary association

Figure 8.14

UE Node B HSS RNC MGW SGSN GGSN

User equipment Base station transceiver Home subscription server Radio network controller Media gateway Serving GPRS support node Gateway GPRS support node

All-IP UMTS.

RNC network plus the node Bs are called UMTS terrestrial radio access network (UTRAN). The all-IP version is, in the same way as GPRS, based on the mobile IP system consisting of the SGSN and the GGSN. Note that the location updating takes place between the SGSN and GGSN, as explained later for mobile IP. The HLR is now called the home subscription server (HSS), containing all subscription data including identities and IP/telephone numbers, service provisioning information, and security support. The HSS does not contain the location information. The location information is held by the GGSN. The interface with the telephone network is supported by the media gateway (MGW), where translation takes place between the formats used in the telephone network and the Internet. The auxiliary units support a number of functions related to multimedia handling (e.g., session control, multicasting, broadcast services, and conference facilities). The RNC controls the radio resources. The mobile terminals are renamed as user equipment (UE) and the base station transceivers are called node B in UMTS. 8.4.7 Mobile IP, Location Updating, and Packet Transfer in GPRS and All-IP UMTS

The application of mobile IP in GPRS and UMTS is shown in Figure 8.15. The method of defining location coordinates is essentially the same as in the GSM system. The location coordinates of a cell are broadcast to all mobile user equipment (MUE) in the cell. The decision to initiate location updating is thus that the MUE discovers that the network identity and/or the location-area identity have changed. When the MUE enters an area controlled by a new SGSN, the MUE sends its identity (shown as the MUE IP number in the figure) possibly together with a temporary identity to the SGSN. In the same way as in the GSM system, the new SGSN may get information from the previous SGSN required for the initial handling of the

8.4 The PLMN Architecture

221 Update <MUE IP number, tunnel id>

Update <MUE IP number> MUE

GGSN IP<MUE IP number>

SGSN

IP<MUE IP number>

UE

Internet IP

Updating Information transfer

Figure 8.15

Mobile IP in GPRS and all-IP UMTS.

MUE (e.g., temporary identities, security parameters). The SGSN allocates an IP number called the tunnel identity that the Internet can use in order to route the IP datagram to the SGSN. The identity of the MUE and the tunnel identity are provided to the GGSN. The GGSN stores the tunnel identity together with the MUE IP number. The location of the MUE in the UMTS network is then uniquely determined by the tunnel identity since the tunnel identity is unique for each SGSN. The same tunnel identity can be used for all MUEs located in the area of the SGSN. This is possible because of the tunneling method shown Figure 8.16. The GGSN embeds the original IP datagram containing the MUE IP number in the payload of a datagram that contains the tunnel identity as the destination IP address as shown. This datagram will be received by the appropriate SGSN, since it is an IP number allocated permanently to that SGSN. The SGSN forwards the original datagram to the mobile user equipment using the original MUE IP address. 8.4.8

Paging, Location Updating, and the Size of the Location Area

The accuracy by which the location of the mobile terminal is known is the location area. In order to service an incoming call to a mobile user, the user must be paged in all the cells (corresponding to a BTS in GSM or node B in UMTS) in the location area. This gives rise to tradeoffs concerning the size of the location area. If the location area is large (that is, consisting of many cells), much capacity must be used on the paging channel. In GSM, the maximum number of paging/access grant messages that can be sent on the common control channel per second is about 36 messages. If there are N cells in the location area, the maximum capacity of each paging channel is reduced to 36/N since the same paging message

IP header (tunnel id)

Figure 8.16

Tunneling.

IP header (MUE IP number)

Payload

222

Cellular Land Mobile Systems 7

must be sent in each of the N cells. The expected number of paging/access grant messages per second then determines the size of the location area. On the other hand, if the location area consists of few cells, the location of the mobile terminals must be updated frequently. Five signaling messages are sent in each direction on the standalone dedicated control channel for every location updating event. The number of dedicated control channels in a cell is four or eight, depending upon channel configuration. These channels are shared between location updating, mobile originating calls, mobile terminating calls, and SMS, each event requiring exchange of several signaling messages. Therefore, the more cells in the location area, the more capacity is available for call handling. The topology of the mobile network is thus a tradeoff between paging over large location areas and frequent location updating in small location areas.

8.5

Composition of the Radio Interface 8.5.1

Packet Radio Systems

The radio interfaces of packet radio systems and circuit switched systems such as GSM are different. The complex methods used in GSM and GPRS are described later. Though GPRS is a packet radio system, it is designed on top of the frame structure used in GSM and must, therefore, have a channel format adapted to the channel format of GSM. GPRS and GSM represent one rare example where packet switching and circuit switching share the same resources. UMTS versions supporting both circuit-switched telephone services and packet data services are similar to GSM/GPRS except at the radio interface, where different multiple access techniques are used. In pure packet radio systems, all information, including user data and control messages, is sent as packets that contain the address of the destination. There is no need to organize these transmissions other than using carrier sense multiple access (CDMA) described in Section 5.9 for WLANs. The packets can be divided into logical types, such as packets containing user information and packets used for control and organization purposes. Control packets may include broadcast of system information such as location area and cell identities for roaming purposes, operational constraints such as bit rate allocation, and maximum allowed power level. Control packets also comprise packets exchanged with individual mobile terminals used for reservation of traffic capacity, location updating, and dynamic power control. Sometimes the terms “handover” and “handoff” are used instead of the term “location management” in packet radio systems. This may be confusing since handover (in the United States, handoff) in circuit-switched systems means that the call is handed over form one cell to another during conversation. The reason is, of course, that the duration of a circuit-switched call may be very long (several minutes) so that there is a significant probability that the user may move from one cell to 7.

In a particular configuration of the cell where two timeslots per TDMA frame are used for paging/access grant, this number is doubled, giving 72/N messages per second.

8.5 Composition of the Radio Interface

223

another during this time. In packet radio systems, each “conversation” consists of a single packet with duration less than a second. The probability that such a “conversation” is disrupted because of the motion of the terminal is negligible. 8.5.2

Channel Coding in GSM

The main channel coding principles of GSM are described in this subsection. The main purpose of the section is to visualize some of the concerns that must be taken into account when designing a complex system. These concerns are as follows: •

•

•

How to arrange the physical data streams into organized frame structures from which the information can be identified and extracted; The type of error correction coding that should be applied in order to ensure proper operation at low carrier-to-noise ratios for different types of information; How to disperse the bits of the encoded data stream so that the system can operate properly in environments producing single errors and in environments where burst errors are dominating.

The GSM system is an FDMA/TDMA system. The frequency spectrum allocated to the system is first divided into frequency bands or carriers, each 200 kHz wide. Each carrier is divided into eight TDMA timeslots. The eight timeslots together are called a TDMA frame. The duration of the frame is 4.615 ms (= 120/26 ms). The duration of each timeslot is then 0.577 ms (= 15/26 ms). The timeslot is the basic period of the channel structure: all other durations are multiples of 15/26 ms. The frame structure of the TDMA system is organized as shown in Figure 8.17. The dotted line enclosing the TDMA frame indicates that the multiframe consists of timeslots only occupying 1/8 of the time in each TDMA frame. Each channel occupies one timeslot per TDMA frame (e.g., timeslot number 4). The traffic channels are organized in 26-multiframes. A 26-multiframe consists of 26 timeslots taken from 26 consecutive frames. The duration of the 26-multiframe is then 26 × 8 × 15/26 ms = 120 ms. The control channels are organized in 51-multiframes consisting of 51 timeslots each. These structures are merged in superframes consisting of either 51 26-multiframes or 26 51-multiframes. In addi11 tion, a hyperframe consisting of 2 = 2,048 superframes = 2,715,648 timeslots is defined. The duration of this beast is 3 hours, 28 minutes, and 54 seconds. The hyperframe number (from 0 to 2,715,647) is used as a reference number in the encryption algorithm over the radio interface8 and for generation of the slow frequency hopping sequence.

8.

Stream ciphering is used as encryption method. The cipher text is produced by feeding the 64-bit encryption key and the hyperframe number to the encryption algorithm call A5 in the GSM specification. The cipher text consists of 114 bits and is added modulo 2 to the 114 information bits of the burst (see Section 3.7.2 for the burst format). Since the hyperframe number changes from burst to burst, each burst is then encrypted individually. This results in a virtually unbreakable encryption method. See also Section 8.7.

224

Cellular Land Mobile Systems Hyperframe = 2,048 superframes = 3 hours, 28 minutes, and 54 seconds

Superframe = 51 × 26 multiframes = 6.12 seconds

Superframe = 26 × 51 multiframes = 6.12 seconds 51 multiframes = 235 ms

26 multiframes = 120 ms

TDMA frame = 8 timeslots = 4.615 ms

Timeslot = 15/26 ms

Figure 8.17

Frame formats.

The two multiframe lengths of 26 slots for traffic channels and 51 slots for control channels, respectively, are deliberately chosen to be different in such away that they have no common divisors (they are, in other words, relatively prime). The reason for this choice is that when a mobile terminal is busy on a traffic channel (on a 26-multiframe), the terminal will always be able to find a burst containing the synchronization channel (on a 51-multiframe) of an adjacent base station that does not overlap with the burst in which the terminal is busy either sending or receiving information. The complex coding schemes of the speech channel and the combined broadcast and common control channel are shown in Figure 8.18. Except for the frequency correction channel, the synchronization channel and the random access channel, each timeslot consists of 114 information bits. In order to provide a unit of organization of the information that is isochronous with regard to the octet structure, four timeslots are combined to form a codeword consisting of 456 bits (or 57 octets). These 57 octets contain both the information bits and bits required for forward error correction. The error correction method is different on different types of channel. For example, the error correction method used on the common control channel (except the random access channel) and the broadcast channel (except the frequency control channel and the synchronization channel) leaves 184 bits (23 octets) available for information.

8.5 Composition of the Radio Interface

x = 3 cyclic redundancy check bits y = 4 zero bits for resetting the codec

225

class Ia 50 bits

50 bits

class Ib 132 bits

x

132 bits

378 convolutionally encoded bits (½ rate)

class II 78 bits

y

78 bits

(a)

z = 4 stuffing bits (value 0)

signaling information 184 bits

184 bits

Fire 40 bits

z

456 convolutionally encoded bits (½ rate) (b)

Figure 8.18

(a, b) Coding of code words: two examples.

Broadcast and common control channels [Figure 8.18(a)] contain 184 information bits (23 octets). A fire code consisting of 40 bits is employed in order to correct and detect burst errors that remain after the signal is passed through the half-rate convolutional decoder. Signals containing uncorrected errors are discarded by the fire decoder. The 4 stuffing bits are just useless bits, since they do not form a complete octet: all information on the signaling channels is encoded as octets. One speech sample is contained in one codeword [Figure 8.18(b)]. The speech sample consists of 260 information bits divided into three parts (class Ia, class Ib, and class II bits) containing different parameter sets. The most important information is contained in class Ia. These bits encode the parameters of speech that represents the human vocal tract (filter coefficients) and other particularly important parameters. Three parity bits are computed for this sequence. If the parity check fails at the receiver, the speech block is discarded and replaced by the previously received block. The next 132 bits are not that critical but are still sensitive to bit errors, though the speech sample is not discarded if these bits contain errors. Class Ia and Ib bits are protected by half-rate convolutional forward error-correction

226

Cellular Land Mobile Systems

coding in order to remove most of the bit errors. Finally, the 78 least significant bits of the encoding process are added without protection. The channel encoding is an excellent example of tradeoffs between compact encoding and application of different protection methods depending upon the nature of the signal. The propagation model of Section 8.3 indicates that we may expect that bit errors often occur in bursts. One way to combat such error performance is to use interleaving. Interleaving will spread out the bits in such a way that an error burst is broken up into single bit errors that can be corrected by simple forward error correction methods. Interleaving does not change the statistics of single bit errors: it is very unlikely that the interleaving combines the single errors into a burst; the most likely event is that they occur in a different sequence of single bit errors after interleaving. The principle of the interleaving method used for speech channels in GSM is illustrated by the following example corresponding to one of several interleaving methods used in GSM. Two consecutive codewords of the speech channel are interleaved on eight subblocks of 57 bits, each numbered from 1 to 8. The codeword is then spread on eight consecutive timeslots as follows. First, each timeslot is divided into two subblocks, where one subblock consists of all the odd-numbered bits in the timeslot (bits 1, 3, 5, ..., 455) and the other subblock consists of the even-numbered bits (bits 0, 2, 4, ..., 454). Then bit number 0 goes into subblock number 1, bit number 1 goes into subblock number 2, ..., bit number 7 goes into subblock number 8, bit number 8 goes into subblock number 1, bit number 9 goes into subblock number 2, and so on until all bits have been placed in one of the eight subblocks. Finally, the first four subblocks are put into the even-numbered bits of four consecutive timeslots and the next four subblocks are put into the odd-numbered bits of the next consecutive four timeslots. The final pattern is as shown in Figure 8.19. The interleaving causes a delay of the speech signal corresponding to eight TDMA frames (i.e., 120/26 ms = 37 ms).

Even-numbered bits

Odd-numbered bits

Timeslot number

0

1

2

3

Codeword 1

Codeword 2

Codeword 3

Figure 8.19

Interleaving of speech channels.

4

5

6

7

8

9

10

11

8.5 Composition of the Radio Interface

227

Though this procedure is straightforward, it is useful to consider a simpler example in order to see intuitively how interleaving works. Figure 8.20 shows a simple example where the input bits are written into a memory matrix consisting of 3 × 3 memory cells columnwise (the in-sequence). The output reads the cells linewise (the out-sequence) and sends them on the radio medium where an error burst may corrupt four consecutive bits. The receiver recovers the correct order of the bits by the inverse procedure. An error burst in the out-sequence consisting of four consecutive bits will then be dispersed after the receiver has recovered the original in-sequence as shown.

In

11

12

13

21

22

23

32

33

31

Dispersed errors

In-sequence

11 21 31 12 22 32 13 23 33

Out-sequence 33 32 31 23 22 21 13 12 11

Error burst

Figure 8.20

Simple interleaver.

Out

228

Cellular Land Mobile Systems

The GSM system is indeed complex. The frame structure consists of two formats (26-multiframes and 51-multiframes). Several interleaving techniques and error correction and error detection methods are applied, depending upon the type of information contained in the multiframe. In addition, the two types of multiframe support and multiplex a number of logical channels, some of which are described here. 8.5.3

Logical Channels in GSM

The channel configuration of UMTS and GSM are almost identical. In order to illustrate the principles, GSM is used as an example. All types of information are sent on logical channels that are organized on the multiframe. One logical channel may occupy all the timeslots of the multiframe or just some of them. Some of these structures will be explained here. Three basic types of logical channel are required to organize the GSM system: broadcast channels, common control channels, and dedicated channels. Broadcast channels and common control channels are organized in 51multiframes, while traffic channels are organized in 26-multiframes. Figure 8.21 shows the main principles of organization in terms of two examples. There are also other ways in which the 51-multiframe or 26-multiframe are organized. Each rectangle represents a single timeslot belonging to the same TDMA channel. All eight channels of the TDMA frame support a similar pattern. The broadcast channel consists of three of the subtypes of channels shown in the figure: •

The frequency correction channel consists of only one timeslot in which all bits are set to binary zero. With the modulation technique used, this produces

0

10

20

30

40

Frequency correction channel Synchronization channel

Broadcast channel

Broadcast control channel Paging channel/access grant channel/stand-alone dedicated control channel/slow associated control channel Empty (timeslot contains no signal) (a)

0

12 Traffic channel Slow associated signaling channel Empty (timeslot contains no signal) (b)

Figure 8.21

(a, b) Examples of multiframe organization.

25

50

8.5 Composition of the Radio Interface

•

•

229

a pure sine wave that enables the clocks of the mobile terminal to acquire lock. Having detected this particular beacon signal, the mobile terminal knows that the next timeslot of the same TDMA channel contains the synchronization channel. The frequency correction channel appears in timeslots 0, 10, 20, 30, and 40 in the 51-multiframe. The synchronization channel consists also of a single timeslot. The synchronization channel has a longer training sequence than the other timeslots (64 bits versus 26 bits) since it is the first timeslot the mobile terminal must be able to read when it synchronizes to a new base station. The channel contains the frame number and the cell identity code. The synchronization channel appears in timeslots 1, 11, 21, 31, and 41 in the 51-multiframe. The broadcast control channel consists of a complete codeword and is used to inform the mobile terminal about specific system parameters such as location area and network codes, the frequencies on which neighboring cells can be found, particular channel configuration information, and maximum power level. This channel is sent only once per 51-multiframe.

The remainder of the 51-multiframe contains the combination of the common control channel (paging and access grant channels) and the standalone dedicated control channel (SDCCH) combined with the slow associated control channel (SACCH). The SDCCH is a duplex signaling channel used for exchange of signaling messages associated with call establishment, location updating, and other management functions. SMS messages are also sent on the SDCCH. An SACCH is always associated with dedicated control channels and traffic channels. The SACCH is used for timing advance and power control on the downlink and for transfer of field strength measurements on the uplink as input parameters to the handover decision procedure. The SACCH is also used for SMS on speech channels. This is why an SMS may be received while the mobile terminal is busy in a telephone call. In the direction toward the mobile terminal (downlink), the common control channel consists of the paging channel on which an incoming call is announced to the mobile terminal and the access grant channel where communication resources are allocated to the mobile terminal upon request. The paging channel and the access grant channel are randomly mixed in the available timeslots depending upon the instantaneous signaling needs. The uplink of the 51-multiframe corresponding to the paging channel and access grant channel is used for random access from the mobile terminals. The multiplexing of the 51-multiframes is complex and can be done in several ways depending upon the traffic intensity in a given cell. The configuration may be different in adjacent cells. The GSM specification defines seven channel combinations numbered from I to VII where different types of control and traffic channels are multiplexed. In the broadcast control cannel, the base station indicates which channel combinations are used in the cell so that the mobile terminal can find the different information elements multiplexed on the channel. Combinations I, II and III are concerned with the configuration of traffic channels in the 26-multiframes and their slow associated control channels, while the remaining configurations are concerned with the multiplexing of control channels in the 51-multiframe.

230

Cellular Land Mobile Systems

In one configuration (combination IV), the 51-multiframe contains only the broadcast channel and the paging and access grant channels on the downlink. The entire capacity of the uplink is then allocated to the random access channels. In such cells, the SDCCH/SACCH used for signaling prior to the allocation of traffic channel and sending of SMS messages is multiplexed on other timeslots (called channel combination VII). The configurations on the uplink and downlink are identical except that the corresponding channels of the uplink and downlink are spaced 12 frames apart in order to allow time for processing of the information. The configuration where combinations IV and VII are used together is common in location areas containing many BTSs since the configuration supports high traffic capacity on the paging channel. The eight SDCCH/SACCH pairs in combination VII are used by different mobile terminals so that the SDCCH/SACCH channels are time-division multiplexed in the TDMA sequence. The multiple access procedure can thus be denoted TDMA/TDMA where one TDMA procedure is used in organizing the timeslots of the primary TDMA frame and another TDMA procedure is used to pick out the timeslots containing the wanted channel. In channel combination V, the same 51-superframe contains the broadcast channel, three common control channels for paging and access grant (3 × 4 timeslots), and four pairs of SDCCH/SACCHs (4 × 4 + 4 × 4 timeslots) on the downlink. The uplink contains the four SDCCH/SACCHs and 27 timeslots for random access. Combination VI contains only the broadcast control channel and common control channels. Frequency correction channels and synchronization channels are not included (the corresponding timeslots does not contain information). The combination is used together with combination IV in order to provide more capacity for paging and access grant. The traffic channels are organized on the 26-multiframe. The format is the same on both the uplink and the downlink. A telephone channel occupies the channel as shown in Figure 8.21(b) (combination I).9 The channel contains a single timeslot for the slow associated control channel. This channel is used for signaling during the conversation and for transfer of SMS. For example, measuring reports concerning the received field strength are sent to the base station on the slow associated control channel as part of the handover decision procedure. The base station sends power control and timing advance messages to the mobile terminal on this channel. Note that timeslot number 25 is empty (i.e., nothing is sent in this timeslot). The mobile terminal can use this interval to do field strength measurements of adjacent cells.

9.

Combination II and III are used for half-rate speech channels where configuration II allows two traffic channels (e.g., two speech channels, two data channels, or one speech and one data channel) to coexist in the same 26-multiframe, while in combination III one of the half-rate channels is idle.

8.5 Composition of the Radio Interface

8.5.4

231

Traffic and Control Channels in GPRS

GPRS is a packet data service sharing resources with the GSM system. Different data rates are obtained by distributing the communication over several timeslots. The GSM system assigns one timeslot to each user, while GPRS assigns one or more timeslots to each user. GPRS may use exactly the same broadcast control channels and the common control channels as GSM. However, as an option, a dedicated channel may be used as common control channel for GPRS. The traffic channels of GPRS combine several timeslots in the same TDMA frame. The structure associated with each timeslot is organized in 52-multiframes (i.e., two 26-multiframes in tandem but where the different parts of the multiframe may have different functions). A complete GPRS packet channel thus consists of a number of such 52-multiframes. In this structure, where timeslots are numbered from 0 to 51, timeslot 12 (see Figure 8.21) and timeslot 38 are used for timing advance control and power control on the downlink. The mobile terminal sends field strength measurement reports for preparation of handover in timeslots 12 and 38, thereby making GSM and GPRS operation similar in the mobile terminal. Earlier, we found that only 23 bytes of information can be sent per codeword (four timeslots). This means that a datagram containing 1,500 bytes must be sent in 261 timeslots. This takes only 223 ms if four TDMA channels are combined (53.6 Kbps). The maximum size of an IP datagram is about 64,000 bytes. This datagram will occupy about 2,800 codewords, but it takes only 9.5 seconds to send the datagram. This shows that handover is strictly not required in GPRS for Internet applications since every datagram is an independent “conversation” of short duration. Another enhancement of GSM is called enhanced data rates for GSM evolution (EDGE), where the modulation scheme of GSM is altered in order to provide higher data rates per timeslot without changing the frequency plan of GSM. GSM/GPRS and EDGE can coexist on the same base station sharing the same frequency band. The base stations can switch between the different technologies. Furthermore, the mobile terminals may support all technologies. This is possible since most of the complex tasks are implemented in software rather than in hardware. 8.5.5

Radio Interface of UMTS

The details of the radio interface of UMTS and similar systems are considerably more complex than the GSM interface for several reasons. The specification contains several options or alternative ways of implementing certain functions, there are several versions of the specification that must coexist in order to support worldwide roaming, and the system must support both circuit-switched and packetswitched procedures. The ITU has allocated the frequency bands 1,885–2,025 MHz and 2,110–2,200 MHz for the 3G systems worldwide. UMTS is the name adopted by ETSI for the 3G mobile system that is now replacing GSM/GPRS/EDGE. CDMA2000 and wideband packet CDMA (WPCDMA) are names used in the United States and other countries for a similar, but not exactly the same, system. The description here is based on the ETSI standard.

232

Cellular Land Mobile Systems

The 3G networks will most likely not be identical everywhere—the issue of selecting a single worldwide standard has become political rather than technical, which is often the case when the direction of specification takes different routes in Europe and the United States. The evolution of the 3G standard is confusing, and it is difficult to assess whether or not handsets offering hybrid solutions that can be used in all 3G systems worldwide will emerge. This is a question of cost, complexity, and size of the multifunctional terminal as well as the size of the market for the different standards. The next generation systems (4G) are already planned with the aim to provide a single worldwide standard. The newest all-IP UMTS is a step in this direction. UMTS is designed to offer user bit rates depending on the applications: • • •

2 Mbps for fixed environments and indoors; 384 Kbps for pedestrian use and in urban cells; 144 Kbps for use in vehicles and in large cells.

The reason for this inhomogeneity has to do with radio propagation. The longer the propagation path and the higher the velocity of the mobile terminal, the narrower the usable bandwidth is (see Section 8.3). The reason is that the signal is subject to multipath propagation that may cause significant frequency selective fading across the signal bandwidth. High data rates can only be achieved for communication over short distances where the multipath delays are short (or in cases where the propagation conditions are stationary such as in WiMax, allowing accurate adaptation of the propagation conditions to the environment). WLAN may offer high data rates because the coverage area is small. Ethernet can offer data rates of more than 1 Gbps because the transmission is confined to a single stable medium without multipath interference. UMTS is a compromise where the goal is to offer worldwide roaming. UMTS must therefore support large cells in rural areas (low system cost per unit area), small cells in urban areas (high traffic capacity per unit area), and pico-cells for indoor applications (competition with WLAN and other local solutions). The basic multiple access method of UMTS is CDMA combined with FDMA, allowing different operators to provide the service in nonoverlapping frequency bands. Each carrier of the FDMA system may in addition be divided into timeslots combining CDMA, FDMA, and TDMA in a single system. CDMA is described in Section 5.5. We saw earlier that GSM uses much of the capacity (55%) on error correction and channel coding. Provision of control channels increases the overhead to 63%. An access method based on CDMA will have smaller overhead than the TDMA method of GSM. This is one motive for choosing CDMA in the new systems, though pure TDMA is also an option in UMTS. Another motive is to exploit the graceful degradation trading capacity for quality (see Sections 5.5 and 5.7). The larger the spreading factor of CDMA (see Section 5.5.1), the less sensitive the signal is to multipath interference, since the coding can be used to suppress multipath signals arriving at the receiver with a delay of one chip duration or more. Therefore, large cells and urban cells where multipath interference is significant should have large spreading factors.

8.5 Composition of the Radio Interface

233

Spreading rates between 1 and 16 are used in UMTS. Spreading rates and spreading codes are assigned by the base station subsystem and disseminated to the mobile terminals on one or more of the control channels. This includes possible choices of spreading codes for the random access channel. The maximum possible data rate on one carrier is about 480 Kbps in pico-cells. The spreading rate is then 4. A user data rate of 2 Mbps is obtained by combining six parallel channels each with a spreading rate of 4. One advantage of CDMA is that channels with different spreading rates may exist in the same frequency spectrum simultaneously. UMTS requires control channels of the same type as the GSM system: broadcast control channel for dissemination of system related data; paging and access grant channels for network access to the mobile terminal; random access channels; a dedicated control channel for power control; timing advance and other call related information; and various types of traffic channels. The same basic functions are required in the UMTS system and the GSM system. UMTS provides duplex services (i.e., supports simultaneous transmission of information on the uplink and downlink). There are two possibilities that are exploited in UMTS: FDD where uplink and downlink channels use different carrier frequencies, and TDD where the uplink and downlink occupies different timeslots that do not overlap in time. The uplink and the downlink of the TDD system may occupy the same or different frequency bands. Corresponding TDD timeslots on the uplink and the downlink may not follow a strict timing schedule as in GSM in order to support a two-way traffic channel. The timing is easily adapted to the type of traffic the channel is carrying (random packet traffic or real-time streams). The TDD mode of UMTS uses wideband TDMA/CDMA sharing the same frequency on the uplink and the downlink. The FDD mode uses wideband CDMA (WCDMA). The capacity of the FDD mode and the TDD mode is approximately the same. It is even possible to do a handover from one mode to the other. The chip rate in both modes is 4.096 Mchips/s. FDD is the preferred method for large-cell applications because it is more likely that the traffic is more symmetric in large cells (mainly voice traffic) than in small cells (a large percentage of the traffic is data transmission and wideband streaming). TDD is preferred for asymmetric operations such as Web search since the method allows flexible allocation of capacity (number of timeslots) in the two directions. Figure 8.22 shows an FDMA/CDMA system using different frequencies on the uplink and the downlink (i.e., FDD). Systems with different data rates and spreading factors may coexist on the same frequency. The power allocated to each channel depends on the spreading rate as illustrated. The data rates in the two directions may also be different. The figure illustrates the flexibility that can be achieved in UMTS. The frame configurations of the FDD and TDD modes are shown in Figure 8.23. Each FDD carrier is divided into frames of duration 10 ms. Each frame contains 16 TDMA slots of duration 0.625 ms. The slots may contain any combination of signaling control channels (power control, format identifier, synchronization information) and user data channels. The frames are combined into superframes

234

Cellular Land Mobile Systems Uplink f2

f1

f3

Downlink f2

f1

f3

f Information rate

Chip rate Power level

Figure 8.22

FDMA/CDMA.

consisting of 72 frames. The frame number is used to keep track of the different logical channels multiplexed on the same carrier. This structure allows different information to be time division multiplexed on the channel. CDMA synchronization, multiplexing, frame synchronization, and segmentation of user data is taken care of by a MAC entity located between the physical layer and the control functions and the media management functions. The frame of the TDD mode consists of 16 TDMA timeslots. The timeslots consists of user data or signaling control information and a midamble (or training sequence) for estimating the channel quality. The guard time between timeslots is 23.4 s.

Superframe 0

1

2

69 70 71

720 ms

Frame 0

1

Frame

2

13 14 15

10 ms

1

2

13 14 15

Timeslot

Slot 625 µs Multiplexed user data and control signaling (a)

Figure 8.23

0

(a, b) FDD and TDD frame structures.

D

M

D

G

625 µs

D = user data or signaling M = midamble (training sequence) G = guard period (23.4 µs) (b)

10 ms

8.6 Handover

8.6

235

Handover Handover is the process of transferring a call in progress from one cell to another without interrupting the call. Handover may take place for two reasons: the communication channel is taken out of service because of equipment failure, interference, or maintenance; or the mobile terminal moves out of one cell and into another. In the first case, the call is handed over to another resource in the same base station (intracell handover); in the second case the call is handed over to another base station (intercell handover). Handover is not always necessary in IP networks since the duration of a single datagram is very short. However, handover may nevertheless be required in IP networks if a communication resource is allocated or reserved for a long time (e.g., Internet telephony or video). On the other hand, handover is an essential feature in mobile systems offering circuit-switched speech or video communication. Since the duration of a speech conversation or video transmission is several minutes, the probability that a handover is required is not negligible. Handover can be classified further as hard handover and soft handover. An intercell soft handover implies that the base stations of the two cells receive the same information form the mobile terminal and can prepare the handover without notifying the mobile terminal. Soft handover is possible in CDMA systems where the base stations share the same frequency band. 8.6.1

Soft Handover

The principle of soft handover for a packet switched call is illustrated in Figure 8.24. The RNCs under the same SGSN are interconnected. The RNCs contain, in fact, Internet router functionality. In the upper picture, the mobile terminal is served by RNC1. In the border zone between the two cells, the signals from the mobile terminal are demodulate by both RNCs. RNC2 sets up a connection to RNC1 and transfers the message received from the mobile terminal together with field strength information. RNC1 then decides if and when a handover should take place. When the call is handed over to RNC2, the call is still routed via RNC1 so that the management of the call is still with RNC1. The control of the call may be handed over to RNC2 if RNC1 finds this more appropriate, particularly if several handovers have taken place in succession and the distance between the RNCs is long. Note that the RNCs of UMTS have a more central role in controlling the network than the base station controllers of GSM. Note that the SGSN does not take part in the handover. The procedure for soft handover of circuit switched calls (or relocation as it is called in the UMTS specification) is shown in Figure 8.25 for the case where new radio resources need not be allocated (the two BSCs are receiving the same CDMA signal). In the figure, the RNC is replaced by a BSC and the SGSN is replaced by an MSC supporting circuit-switched radio connections, since, as explained earlier, UMTS supporting both packet switching and circuit switching must provide for a separate architecture for each of the two switching modes. The procedure does not require any signaling between the mobile terminal and the network. However, the

236

Cellular Land Mobile Systems

RNC 2

SGSN

1 RNC 1

RNC 2

SGSN

2 RNC 1

RNC 2

SGSN

3 RNC 1

Figure 8.24

Soft handover in an all-IP UMTS.

BSC may receive channel quality measurements from the mobile terminal, for example, as part of the power control procedure. If new radio resources are allocated (e.g., changing to another frequency band), the BSC1 will signal this to the mobile terminal. The procedure is as follows for a handover taking place in the same MSC.10 When BSC1 finds that the call has to be handed over to BSC2, BSC1 sends the relocation required signal to the MSC, which forwards the request to BSC2. Upon receiving the acknowledgment from BSC2, the MSC sends the relocation command to BSC1. BSC1 then sends the relocation commit signal to BSC2, requesting BSC2 to take over the control of the communications path. When BSC2 has received this signal and is otherwise tuned to the information stream from the mobile terminal, the relocation detected signal and the relocation complete signal are sent to the MSC. 10. Handover to a BSC on another MSC requires first that a connection is established between the MSCs using MAP signaling. Then the procedure commences in the same way as explained.

8.6 Handover

237

MS

MSC

BSC1

BSC2

Relocation required Relocation request Ack Relocation command Relocation commit Relocation detected Relocation complete Release channel

Figure 8.25

Circuit-switched handover procedure in UMTS.

The MSC can now switch the circuit over to BSC2 and release the connection to BSC1. This is done without interrupting the information stream. 8.6.2

Hard Handover

A hard handover is a discontinuous transfer of the information channel from one cell to another. Information may be lost in this discontinuous process. GSM employs hard handover. Hard handover may also be used in UMTS if the handover takes place between parts of the network controlled by different MSCs/SGSNs. Hard handover is also used for handover between GSM and UMTS. In this case, a mixed procedure supporting the soft handover procedure explained earlier between the mobile terminal and the UMTS BSC and the hard handover procedure described later between the mobile terminal and the GSM BSC is applied. In GSM, the mobile terminal measures the quality (field strength and bit error rate) of the signal received in its own cell and in adjacent cells. The identity of the adjacent cells and the carrier frequency on which measurement may take place is signaled to the mobile terminal in the broadcast control channel. We saw in Section 8.5.2 that the mobile terminal when locked to a traffic channel neither sends nor receives information in the last timeslot of the 26-multiframe. During this interval, the mobile terminal tunes to the appropriate frequency of one of the neighboring cells and measures the field strength of the received signal.11 The mobile terminal scans all the adjacent cells in some cyclic order. The measurement results are sent to

11. The mobile terminal may also measure the quality of adjacent cells during any time the terminal is not sending or receiving. Whether this method is used in addition to measure the quality during the empty timeslot is a construction detail decided by the manufacturer.

238

Cellular Land Mobile Systems

the BSC together with the error rate and field strength of the traffic channel to which the terminal is locked.12 The BSC compares the measurements and decides whether a handover shall take place. The procedure is shown in Figure 8.26 for handover between base stations on the same MSC. The procedure for handing over the call from BSC1 to BSC2 is as follows. When receiving the field strength and bit error rate measurements from the mobile terminal, BSC1 will determine if handover is required and, if so, that it actually can be performed. The message from the mobile terminal also contains the identity of the cell in which the measurement is done. BSC1 also measures the quality on the uplink. Both the measurements of the uplink and the downlink are used in order to establish the handover criterion. From the information received from the mobile terminal, BSC1 can also determine to which BSC (BSC2 in the figure) the call should be handed over. BSC1 then requests the MSC to initiate the allocation of a radio channel at BSC2. The MSC does this by sending the handover request message to BSC2. BSC2 returns the frequency and timeslot allocated to the call in the acknowledgment

MS

BSC1

Measurements

MSC

BSC2

Handover required Handover request

Handover command

Handover command

Ack

Access message Access grant Handover detected SABM UA Handover complete Handover complete Release channel Measurements

Figure 8.26

Handover in GSM.

12. In order to avoid misunderstandings, note that the measurement and decision procedures taking place when the mobile terminal changes cells when it is idle (i.e., performs a location updating) is different. The terminal may now search for and identify the presence of stronger carriers at any time. While the BSC decides that a handover shall take place, the terminal decides that a location updating shall take place. Note further that location updating does not take place as the result of a handover. Location updating may take place afterward if the mobile terminal stays tuned to the new base station after the call has been released and the base station is in a new location area.

8.6 Handover

239

message if a channel is available. The MSC then sends the handover command to the mobile terminal (via BSC1) containing the channel number of the traffic channel and the channel on which the random access message from the mobile terminal can be sent. The handover command also contains other information such as power level in the new cell, identification of the channel containing the synchronization channel, and a handover reference that is used in the following signaling messages. Having received the handover command message, the mobile terminal must follow a procedure that establishes the correct burst timing in the new cell (timing advances). Timing advance is described in Section 3.7.2. This is done by sending a random access message to BSC2 on the indicated channel. BSC2 will then return the access grant message identifying the new traffic channel on which the call is going to be continued and the timing advance information. BSC2 may inform the MSC about this event, but this is strictly not necessary. Since contention may take place on the random access channel, the contention resolution procedure described in Section 5.9.2 (Figure 5.13) must be followed; that is, BSC2 sends a SABM containing the full identity of the mobile terminal. The mobile terminal acknowledges this message by the UA message, as explained in Section 5.9.2. Finally, the mobile terminal sends the handover complete message to BSC2 and commences normal operation, including measuring the quality of the new channel. The handover complete event is signaled to the MSC that orders BSC1 to release the radio channel. In Figure 8.26, the random access message, the access grant message, the SABM, and the UA are exchanged between the mobile terminal and BSC2. This introduces unnecessary delay. Therefore, these messages are treated autonomously by the BST during handover. In all other cases, these messages are handled by the BSC. The handover command, the handover complete, the SABM, and the UA are carried by the traffic channel. During normal call setup, all signaling information is conveyed over a standalone dedicated control channel. By using the traffic channel for passing signaling information during handover, the disruption of the call is so short that it is not noticeable. For this purpose, the traffic channel may for a single burst (or several consecutive bursts) be converted into a signaling channel called the fast associated control channel. This is done by setting the two stealing flags (S) to binary 1 in the burst normally carrying a traffic channel. The burst format is shown in Figure 8.27. The fast associated control channel may also be used for other purposes where signaling information must be exchanged during the conversation. Handover to a cell under another MSC is similar. The additional procedure is that the MSCs must establish a connection between the MSCs in the telephone

T

Data

S

TS

S

Data

T

3

57

1

26

1

57

3

T = tail, S = stealing flag, TS = training sequence, G = guard time (8.25 bits)

Figure 8.27

Format of the traffic burst.

G

240

Cellular Land Mobile Systems

network before the target BSC is requested to allocate the new channel. After the handover has been completed, the call is still controlled by the original MSC; that is, all signaling exchanged between the network and the mobile terminal is handled by the old (or controlling) MSC. This also permits the call to be handed over to a third MSC or back to the original MSC without involving a long chain of MSCs. In the first case, the controlling MSC sets up a separate connection to the new MSC. When the handover is completed, the connection to the second MSC is released. In this way, no more than two MSCs will take part in the handover. Similarly, if the call is handed over to the original MSC, the connection to the second MSC is released as soon as handover is completed. The loss of information during the handover is at most one codeword. Note that the error correction method and the interleaving technique used in GSM are so strong that a complete burst may be lost and the content of it may still be recovered.

8.7

Subscriber Identity Module Since the mobile terminal may roam between different networks, it must contain its unique international identity—or rather the identity of the user of the mobile terminal—in order to be served properly. The identity is required in order to determine in which HLR or HSS the user is registered. It was decided early in the development of the GSM specification that the IMSI should be assigned to the user and not to the terminal. The IMSI consists of 15 decimal digits where the first three digits indicate the home country of the user and the next two or three digits indicate a mobile network within that country. The remaining digits identify the user within the network. One country may have more than one country code, and each network may have more than one network code. Furthermore, the IMSI and other personal information should be contained in a smart card that is inserted in the mobile terminal. There were several reasons behind this decision. First, the subscription is always with a person and not with a terminal. Therefore, if the IMSI is assigned to the subscriber, the IMSI can be used as an identifier in the subscription database. Second, it should be possible to replace the mobile terminal without changing the registration in the HLR or HSS. This is possible by removing the smart card from the old terminal and inserting it in the new one. Then the registration is independent of the terminal in which the card is used. Third, from a market point of view it is advantageous to offer personalized services. Personalization includes storage of address lists, storage of sent and received messages, personal selection of operational mode and setting of technical parameters, and logging of call events. Finally, the smart card is a tamper-resistant device for storage of security parameters. SIM cards have been used in all public land mobile networks since GSM. The security issue is particularly important. In GSM and UMTS, the SIM contains the encryption key and the algorithm used for authenticating the mobile user. The SIM card also contains the algorithm for computing the stream cipher key used for encryption on the radio path. In earlier systems (for example, the NMT), the identity of the user was encoded in the mobile terminal. The contents of these terminals could then be copied to

8.7 Subscriber Identity Module

241

another terminal (an activity often referred to as identity cloning). The user of the cloned terminal could then make free calls since they would be charged on the owner of the identity. But, more importantly, the terminal could be used for criminal purposes (for example, in drug trafficking), since the identity of the actual user of the terminal could not be derived from the identity of the mobile terminal. Eavesdropping of the radio connection would reveal what was going on but not disclose the identity of the parties involved. Identity cloning became a lucrative business for organized crime. The interface between the SIM and the terminal is standardized so that the SIM and the terminal can be produced by different manufacturers. The SIM is used for security services (among other services) and contains the authentication key, the authentication algorithm, and the algorithm generating the encryption key. These parameters can be neither read nor modified by external command. The security mechanisms in GSM are shown in Figure 8.28. GSM is used as an example of how security can be implemented in a public land mobile system since the method applied there is intuitively simple. The same cannot be said about UMTS, where much more complex and less intuitive methods are applied. The figure shows which information each entity contains in order to authenticate the SIM and encrypt the information stream on the radio path. The procedure is as follows. The authentication key Ki is a secret shared by the SIM card and the authentication center (AUC). The AUC is a secure database only allowing access from the HLR. Upon request from the VLR, the AUC selects a random number R and inputs this number together with the authentication key in the algorithm A3, producing the number S. AUC uses algorithm A8, the key Ki, and the random number R to produce the encryption key Kc to be used for encryption of messages on the radio path. Five sets of R, S, and Kc are produced this way and sent to the VLR. Then the VLR need not request new authentication parameters every time the mobile terminal accesses the network. This was done simply to reduce the signaling traffic between VLRs and HLRs.

SIM

MS

BSC

Ki A3 A5 A8

A5

A5

VLR

HLR

Ki A3 A8

D

Provide auth para Preparation

5

(R, S, Kc)

R Authentication Encryption

S′ Provide Kc Kc

S′ = A3(Ki, R) Kc = A8(Ki, R)

Figure 8.28

Encryption command Kc

Security functions in GSM.

Compute auth para (R, S, Kc)5

D

D(S = S′) = c = A5(n, Kc) ⊕ m m = A5(n, Kc) ⊕ c

AUC

yes no

S = A3(Ki, R) Kc = A8(Ki, R)

242

Cellular Land Mobile Systems

Authentication of the SIM card then takes place as follows. The random number R is sent to the mobile terminal that forwards it to the SIM. The SIM calculates the response S′ using the A3 algorithm as shown. S′ is returned to the VLR, and the VLR simply checks that S′ equals S (the algorithm D in the figure). If they match, the authentication is successful. 13 We see that only random numbers R and S are sent openly on the radio path. The SIM also calculates the encryption key Kc associated with the random number R using the A8 algorithm and provides the key to the hardware of the mobile terminal the next time an encrypted message is to be sent on the radio path. The VLR sends Kc to the BSC. The public algorithm A5 is used to calculate the cipher-text c from Kc and the current hyperframe number n of the timeslot in which the next segment of the message is sent. The cipher-text is obtained by taking the output of the algorithm and XORing this bit stream (represented by the symbol ⊕) to the clear-text message m. The addition rules of XOR are 1 ⊕ 1 = 0 ⊕ 0 = 1 and 1 ⊕ 0 = 0 ⊕ 1 = 0. Decryption is then done by adding the same output of the algorithm to the cipher-text since A5 ( n , K c ) ⊕ A5 ( n , K c ) = 1 in accordance with the addition rules. Note that the algorithms A5 and A8 are public algorithms that are used by all operators. Otherwise, roaming would not be possible. On the other hand, the operators may choose their own algorithm A3 for calculating S, since the algorithm is only contained in the AUC owned by the operator and in the SIMs produced by the same operator. This enhances the security of the system, since the operator can easily change the algorithm if it has been compromised and, if necessary, revoke earlier SIM cards independently of all other operators.

8.8

Adaptive Access A large number of radio technologies are in use in various frequency bands: GSM, several G3 technologies, WLAN (WiFi), Bluetooth, WiMax, and so on. Even more systems are on the workbenches of several laboratories around the world. These systems apply different access methods, modulation technologies, and channel coding; offer different data rates; and are intended to be used in different environments. This is of no advantage for the user who, in the worst case, must possess several mobile terminals in order to gain universal access. One important evolution is toward the development of multipurpose mobile platforms, where several access technologies are supported by a single terminal. Most units (PCs, printers, earpieces, smart phones) now support Bluetooth, enabling them to register and communicate in very small local networks (pico-networks). In addition, several types of mobile phones and PDAs can switch between WLAN operation and public networks (GSM or 3G) and between GSM/GPRS and 3G. The latter is required if there shall be a smooth transition from GSM to 3G systems. 13. Since R is a random number, S also appears as a random number since S can only be computed from R if Ki is known. The algorithm A3 is such that it is hard to calculate Ki even if R, S and A3 are known.

8.8 Adaptive Access

243

The need for adaptive access will persist into the future because it is Utopian to believe that the world will ever agree on a single system concept. The situation now and in the future is that different systems reach the market at different times, coexist for a while, and then disappear again. This evolution is possible because most of the functions in the mobile terminal are realized in software. It is possible though not trivial to implement the signaling procedures and management of the information streams of several systems in one processor. One problem is the lack of standardized operating systems for this purpose so that automatic downloading of new software may be difficult. However, this problem can be solved in a number of ways using the virtual machine concept. The software structure of the mobile terminal can then be as shown in Figure 8.29. The mobile terminal contains a basic set of user applications. In addition, the terminal must support the download of different types of software for enhanced operation. Such software may be related to user applications and hardware configuration. The download may be done by the user or automatically by the operator. The API supports the download and interfaces the software controlling the

User applications (voice, video, Web) Software download

API support

Programmable radio interface (2G/3G/WiFi)

Virtual machine Programmable service modules (2G/3G/WiFi)

Resident software

Hardware (2G/3G/WiFi)

Figure 8.29

Software structure of programmable mobile terminal.

244

Cellular Land Mobile Systems

residential software and the hardware of the terminal. The programmable radio interface selects and instantiates the required hardware controlling software based on certain criteria—for example, automatic choice of WiFi (WLAN) over public services if WiFi is detected at the radio interface. The virtual machine is the software interface between the residential software in the terminal and the downloaded software modules. The mobile terminal consists of several basic modules as exemplified in Figure 8.30: •

•

•

•

Software platform supporting the download and multiple management of the terminal. The platform software should be remotely programmable. The platform may consist of SIM for storing of security parameters as well as personal profiles and procedures (e.g., electronic wallet); Java modules for remote programming; and software download, security services, and display and keyboard functions. Service module supporting different user applications such as voice, keyboard, text display, still picture display and camera control, video presentation and video camera control, Web services, e-mail, security services, mobility services (e.g., location), and so on. This module should be remotely programmable. Multiple software and hardware for particular operations of the terminal (for example, 2G/3G/WiFi) and Bluetooth and other interfaces for personal area networks (PANs). Each of these technologies contains software and hardware controlling a particular technology and consists of submodules for channel coding and modulation. The mobile terminal may choose the appropriate mode automatically

Programmable software platform

SIM

Java

Programmable service module -

Programmable hardware module

-

Telephone

Internet

Video

Mob

Mail

Stills

Figure 8.30

Keyboard display

Security

Multiple-mode mobile terminal.

2G

3G

Bluetooth (PAN)

WiFi

8.9 Smartphones and Information Security

245

(for example, 2G or 3G), or the mode of operation may be chosen manually by the user. The biggest challenge is the radio interface of mobile telephone–type terminals, where the support of several systems may require diverse hardware to be implemented in a limited space. Some of these functions may be realized in programmable hardware modules, while the terminal must be replaced if the changes are too extensive. The problem is simpler in PCs, where new hardware may be inserted in the PC in order to support new applications.

8.9

Smartphones and Information Security Most mobile terminals connected to the GSM/GPRS/EDGE and UMTS are computers capable of handling not only speech but also data applications such as Web search and Web services, e-mail, voice over IP, video over IP, and downloading of executable programs and data files of any kind. The operating system of most PCs is Microsoft Windows. In order to attack most PCs, it is then sufficient to write malicious software (or malware, for short) that can infect programs running on this platform. The situation is similar for the smart phones, since at least 70% of them use the Symbian operating system. The producer of malware need then only produce malware for one type of operating system in order to do much damage. An infected mobile terminal may also be an intermediate host for the malware. The malware may be passed to a PC when the user synchronizes the smart phone and the PC. The malware may then not be detected if the port used for synchronization (e.g., the bus) is not protected by antivirus software. The mobile terminal may also be the target for the malware. One example of such malware is programs that generate streams of SMS messages sent to premium rate numbers in the network, thus generating huge bills for the user of the mobile terminal and income for the owner of the premium rate service. The mobile terminal is usually infected via Bluetooth. Investigation has shown that the Bluetooth receiver in most mobile terminals is open most of the time. This enables other mobile terminals to transfer information even without the user being aware that this is going on. Sometimes the messages are camouflaged as offers of ringing tones, background pictures, and new games. Malware may also be attached to picture files, music files, and e-mail. A terminal can only be infected via Bluetooth if it is within a certain range of an infected terminal, determined by the propagation conditions for the Bluetooth signal. This makes the epidemiology of mobile system similar to the spreading of a disease by air in medical epidemiology. Studies concerning the spreading of malware in this way have just commenced.

CHAPTER 9

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites 9.1

Introduction A line-of-sight system is a radio system where the transmitter and receiver antennas are arranged such that they point directly at each other and where multipath interference can be kept at controlled levels. There is no obstacle in the transmission path between the two antennas1 and the signal attenuation follows the law of free space propagation (one divided by the square of the distance between the antennas). Three types of system are described later: fixed radio access (WiMax), radio relay, and satellite. The major part of the chapter is concerned with satellite systems, since these systems are considerably more complex than the other systems. Line-of-sight systems are used both for access and transport. Fixed radio access systems are, of course, only applied in the access network. Radio relays are transport technologies, while satellite systems are used both for transport and access, as explained later.

9.2

Fixed Radio Access Networks The development of fixed radio access networks started more than 30 years ago, but it is only recently that such networks have been established, as the technology has previously been too expensive to compete with twisted pair and optical fiber. WiMax has altered this situation. A fixed radio access network is illustrated in Figure 9.1. The system consists of base stations (TX/RX) connected to the fixed network and user equipment connected to either single user terminals or local area networks. Wireless LANs accessible for everyone (for example, at a hotel or an airport) are often referred to as hotspots. The base station antenna is radiating in a sector. Every user terminal in the line-of-sight of the base station will receive the signal. The antenna at the user terminal is directional. The prevailing fixed radio access technology is WiMax. WiMax is not just a single standard but a set of standards that may be used for different purposes. The basic standards were developed by IEEE and are contained in the 802.16 series of specifications. The systems are designed for operation in the frequency range 2 to 66 GHz. The most popular bands are the unlicensed bands between 2.4 and 2.7 GHz

1.

Edge diffraction may be used in certain radio relay systems. One example is a radio relay link in Greenland using edge diffraction across a nunatak on which it would be too expensive to place a repeater station.

247

248

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites Directional antenna Hotspot Sectorial antenna Fixed network

LAN (Ethernet or WiFi)

TX/RX user

TX/RX Base station

TX/RX user

User terminal

TX/RX user

User terminal

Broadcast

Figure 9.1

Fixed radio-access network.

and between 3.4 and 3.6 GHz. WiMax offers bit rates in the range from 1.5 to more than 100 Mbps and is an alternative to ADSL on fixed connections. The range is usually between 7 km and 10 km but may be longer depending on the topology. The large bandwidth can be achieved over long distances because the link is stationary. The most interesting aspects of WiMax are commercial. WiMax enables new operators to build cheap broadband access networks instead of leasing the access from the incumbents—called local loop unbundling (LLUB) in the telecommunication regulatory parlance—though the major commercial obstacle has been the cost of the base stations and the transceivers at the user site. WiMax is also an alternative to optical fibers and ADSL (e.g., in new housing areas because of flexibility with regard to the number of users that can be connected). WiMax is connected to the Internet and may offer any Internet service including VoIP. The radio channel is organized in an orderly manner from the base station. From the user terminal, CSMA and collision avoidance contention control combined with capacity reservation is used in more or less the same way as in the WLAN systems defined in IEEE 802.11.

9.3

Radio Relays In line-of-sight radio relay systems, the signal is sent from one antenna to another by use of narrow beams of radio signals. Radio relays are used in the transport network and are not a competitor but a supplement to optical fibers. Radio relays are used in mountainous and other terrain where optical fibers are too expensive. The configuration of a typical connection containing radio relays is shown in Figure 9.2. The narrow radio beam is obtained by using parabolic dish antennas at both the transmitter and the receiver. The radio signal is called the carrier. The frequency range used for carrier frequencies is in the microwave range from 2 GHz to 40 GHz. This corresponds to a wavelength between 20 cm and 0.75 cm. Higher frequencies are used for communication over very short paths.

9.3 Radio Relays

Fiber Ex

Figure 9.2

249

TX RX

Radio relay path

Active repeater

TX RX

Ex

Connection between exchanges (Ex) containing fibers and radio relays.

The digital signal—or baseband signal—to be transferred is modulated on this carrier wave by use of phase shift keying (PSK) or amplitude-phase shift keying (APSK). The bit rates may range from 2 Mbps to 644 Mbps. The latter frequency corresponds to one of the modulation rates used on optical fibers (see SDH in Section 4.4). The basic component of a radio relay terminal is shown in Figure 9.3. The baseband signal is delivered to the transmitter by an optical fiber (or coaxial cable). The transmitter modulates the signal onto the carrier and feeds the modulated signal to the antenna. The signal in the opposite direction is picked up by the antenna and demodulated by the receiver. The resulting baseband signal is delivered to the fiber as shown. The simplest antenna2 consists of a parabolic dish, a feeder, and a horn. The horn radiates the signal as a spherical wave. The parabolic dish then collimates the wave so that when the wave leaves the dish, it is confined to a thin pencil of rays along the axis of the parabola (almost plain wave). The radiating source (horn in the figure) is located in the focal point of the parabola. Similarly, a plane wave received by the parabolic dish is focused into a spherical wave, having maximum intensity at the focal point where it is picked up by the horn. The radio relay contains a single antenna that both transmits and receives signals. The signal therefore propagates in two directions in the antenna feeder. This is possible in waveguides. A duplexer is used to separate the two directions of wave propagation in the guide so that the transmitted signal is fed to the antenna and not sent into the receiver. The duplexer also directs the received signal to the receiver so that none of it is lost into the transmitter chain.

2.

More complex antennas exist: Cassegrain antennas with hyperbolic subreflectors and where the feed is a hole in the center of the parabolic dish; offset-feed reflector antennas, where the dish is an off-axis part of the parabola so that the feeder horn is outside the path of the radiated wave but still at the focal point of the parabola; and phase-array antennas, consisting of many small antenna elements where the radiation pattern is built up by controlling the phase of each antenna element.

250

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites Plane wave

Spherical wave

Horn Parabolic dish

Microwave feeder

Receiver

Transmitter Duplexer Baseband signal Optical fiber

Figure 9.3

Components of a radio relay.

A transmission system may consist of several radio relays in tandem. The system then consists of terminals at the end of the transmission chain and active repeaters along the path where the signal is amplified before it is retransmitted. Sometimes passive reflectors are used to lead the signal past obstacles as shown in Figure 9.4. The passive reflector may either be a plain metallic sheet or two parabolic antennas placed back to back and connected by waveguide. The passive reflector does not contain devices amplifying the signal and is a cheap way of constructing a complex signal path (e.g., across a hill or along a winding valley). Without the passive reflector, a full active repeater station had to be built. The line-of-sight path is limited by several constraints. The path is naturally limited by the curvature of the Earth. The (geometric) line-of-sight distance between two antennas placed in towers 30m above the surface of the Earth is 40 km. The distance is about 70 km if the towers are 100m high. The refractive index of the atmosphere tends to bend the wave toward the surface of the Earth so that the path becomes longer than predicted previously. In the earlier examples, we may add 10% to 20% to the geometric distance to determine the radio distance.

Active repeater (transceiver) Passive reflector

Terminal

Figure 9.4

Path containing both active repeater and passive reflector.

9.4 Telecommunications Satellite Services

251

The second constraint is the free space attenuation. The signal is attenuated by a factor that is proportional to both the square of the distance and the square of the frequency (see Section 9.7). Rain attenuation is also important in the higher frequency bands used for radio relay systems (above about 10 GHz). The statistics for the rain attenuation is directly given by the statistics of rain intensity. Radio relay systems used in the long haul part of the transport network carry many connections and require high reliability. Probability of outage due to rain attenuation of only 0.01% is common for a radio relay hop. In this case, we may have to take into account rain attenuation of 20 dB or more. Multipath attenuation occurs when the radio signal can follow several paths between the transmitting and receiving antenna. Multipath propagation is usually caused by reflection of the signal from the terrain and buildings. The multiple signals meet at the receiving antenna with different time delay, causing the signals to interfere. Multipath attenuation is avoided by shielding the antennas so that only the main beam is allowed to pass. This can be done by placing screens close to one of the antennas or using mountain ridges or rooftops as natural shields. Typical radio relay hops are about 50 km at frequencies in the frequency range 4 to 6 GHz. The hop length is about 25 km for frequencies in the range of 10 to 12 GHz. In the range 18 to 23 GHz, the hop length is about 10 km. A communication path of 500 km may then consist of 10 to more than 50 transceivers depending upon frequency and terrain.

9.4

Telecommunications Satellite Services 9.4.1

A Brief History

The satellite systems exploit radio frequencies in the microwave range. The microwave range spans the frequencies from about 2 GHz to about 200 GHz or, in terms of wavelength, from 20 cm to 2 mm. Because of the maturity of the technology, frequencies above about 40 GHz are rarely used. The first satellite put into the geostationary orbit (SYNCOM II) was launched in 1963. The large distance between the Earth and the geostationary orbit gives rise to two problems: it is difficult to launch a satellite into the orbit that is so far away from the Earth, and the propagation path suffers from huge propagation loss requiring large antennas, high-power microwave amplifiers, and low-noise microwave receivers. The SYNCOM II experiment proved the feasibility of the geostationary orbit in practical systems. SYNCOM II thus inaugurated the telecommunications satellite area and laid the foundation for the huge spacecraft industry that rapidly started evolving from about 1965. Early Bird was the first commercial satellite of the telecommunications satellite era. It was launched in 1965 about half a year after the International Telecommunications Satellite Consortium (after 1973, renamed as the International Telecommunications Satellite Organization), or INTELSAT, had been founded. The charter of INTELSAT was to provide and operate international satellite communication for the member organizations (governments). Since 2001, INTELSAT (now Intelsat Ltd.) is a private company with headquarters in Bermuda. This is just another result

252

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

of the general privatization of the telecommunications industry. INTELSAT was an organization owned and run by the old telecommunications monopolies. The number of shares and hence the investment liability of an operator at any time were determined from how much the operator actually used the satellites. This is no longer a viable business model in the competitive telecommunications market. Eight generations of INTELSAT satellites were put into orbit before the organization was sold. The evolution of the INTELSAT system is shown in Table 9.1. How international satellite communication will evolve in the future is too early to predict. Intelsat Ltd. offers intercontinental telecommunications services in competition with terrestrial optical fiber systems. Therefore, we may expect that the need for intercontinental communications satellites will be smaller in the future as the optical network is built out further. Broadcast distribution on the terrestrial Internet may also reduce the need for satellite broadcast services. The most noticeable evolution in the communication satellite business is the weight of the satellite. The weight of the Early Bird was only 39 kg. This was the carrying capacity of the original Delta rocket. The Atlas Centaur rocket, having a carrying capacity of about 2,000 kg was used between 1971 and 1988. Since 1989, the weight of the satellites has been in excess of 4,000 kg, matching the carrying capacity of the European Ariane rocket. The telecommunications capacity of Intelsat VIII is about 120,000 64-Kbps telephone channels3 or 22,000 telephone channels combined with 3 television broadcast channels. The design lifetime of the early satellites was between 7 and 10 years. The design lifetime of Intelsat VIII is 14 to 17 years. In 1969, a study in Norway showed that it was both technically and economically feasible to offer geostationary satellite communication to ships. A study in the United States at the same time also proved the feasibility of providing a similar service to aircraft. In both cases, the main motive was improved safety. The only radio communication system that could be used with ocean-going ships and aircraft on intercontinental flights was Morse telegraphy at HF (the frequency band between 3 and 30 MHz). HF propagation makes use of the reflection of the radio waves by the ionosphere, thereby providing worldwide coverage. The region between the ionosphere and the ground can be modeled as a huge waveguide. However, the propagation conditions at HF are complex and far from stable. Dead zones with which longrange communication is not possible may form. When and where such zones will

3.

Since each party of a speech conversation is actively speaking for only 40% of the time on average, only about 120,000 × 0.4 = 48,000 physical channels are required for carrying 120,000 simultaneous conversations if a multiplexing technique called speech interpolation is used. Speech interpolation multiplexing means that the satellite channel in one direction is assigned to a conversation only during the 40% of the time the channel actually contains a speech signal in this direction. During speech pauses, the channel is reallocated to other active conversations. When speech is again detected in the original conversation, the system allocates a new channel for this conversation. If there is no idle channel, this may result in a phenomenon called freeze-out, where one or two phonemes are lost at the beginning of the first utterance. However, the dynamics of speech is such that freeze-out is no problem even if the satellite supports 2.5 times as many conversations than there are physical channels.

9.4 Telecommunications Satellite Services Table 9.1

253

The INTELSAT System

System

Launch Time

Number of Satellites

Approximate Weight (kg)

Intelsat I (Early Bird)

1965

1

39

Intelsat II

1967

3

87 and 192

Intelsat III

1968–1970

6

293

Intelsat IV

1971–1978

11

1,500

Intelsat V

1980–1989

14

2,000

Intelsat VI

1989–1991

5

4,200

Intelsat VII

1993–1996

8

4,200

Intelsat VIII

1997–Present

No information

4,200

form can actually be predicted from atmospheric and meteorological observations. A dead zone may persist for several days at a time. The event that triggered the study of maritime satellite communications was the accident of the Norwegian ship Etnefjell in 1968, where 30 seamen lost their lives. When the ship started burning, it was located in the ocean south of Africa in a region where there are no commercial air-traffic routes, so the distress message was not picked up by aircraft (aircraft and ships share the same distress frequency at HF). In addition, there were no other ships within radio distance and the radio-propagation conditions on HF were such that it was impossible to send messages to the distress centers in South Africa or elsewhere. The first maritime-mobile satellite system (MARISAT) was put into service in 1976–1977 by the private U.S. company COMSAT General. COMSAT General put an additional commercial transponder on a set of satellites they were to operate on behalf of the U.S. Navy. These satellites had the capacity to carry the commercial transponder in addition to their main mission. The MARISAT system was taken over by the organization INMARSAT in 1982, the same year the Norwegian coast Earth station was put into operation at Eik in Rogaland. INMARSAT was inaugurated in 1979 as an intergovernmental organization for offering satellite communication to commercial vessels. INMARSAT later entered into aeronautical and land mobile satellite services and is now about to introduce broadband services with bit rates up to 432 Kbps. As of 2005, INMARSAT was traded on the stock exchange and, as INTELSAT, had become a private company. Since 1975, a large number of national satellite systems have been put into operation for serving particular parts of the world. The first one was put in operation in 1976 in Norway (NORSAT) as a communication service for the oil installations in the North Sea. The system used leased satellite transponders owned by INTELSAT. This system was later expanded to provide telecommunications and broadcast services to Spitsbergen. Later, national satellite systems were built in several countries (including Indonesia, Australia, Canada, and Japan). Several of these systems have met severe competition from optical fibers, but some of them will continue into the future for the simple reason that all other telecommunications systems are likely to be too expensive. One particular system is called VSAT offering services to remote locations such as hydropower stations and dams. VSAT systems are easy to reconfigure. This makes these systems particularly suitable as local area networks in environments that are temporary or likely to

254

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

change rapidly such as construction sites. In particular in the United States, VSAT is used for interconnecting the offices of companies spread over large areas, such as shopping chains and mining companies. The idea behind the Iridium system was to employ low Earth orbit (LEO) satellites in polar orbits. The system was planned with 66 satellites in six polar orbits containing 11 satellites each.4 The business idea was to offer worldwide land mobile communication. The business idea was totally unrealistic and based on market assessments without any rational foundation. The price of the Iridium terminals and the communication charges could not compete with terrestrial mobile systems such as GSM. The traffic in areas not covered by GSM and similar systems was too small to form a sound commercial basis for Iridium. The system also didn’t offer any particular features that would make it attractive for potential users. Therefore, the Iridium Consortium went bankrupt before it commenced operation. However, the satellites were later sold to other companies reestablishing the services. The systems offer telecommunications to governments, oil companies, scientific explorations, relief operations, and travelers. By the end of 2005, Iridium had 142,000 subscribers. The new Iridium system is claimed to be profitable because most of the original debts could be written off. An even more ambitious project was Teledesic, where a constellation of 824 LEO satellite in polar orbits was hoped to offer global broadband data services independently of and in competition with terrestrial systems. The project was scaled down to encompass 288 satellites before it was eventually given up because of totally unrealistic market estimates: the project was even more pretentious and farther from reality than the Iridium project (the two projects were even competing for the same customers). Projects like Iridium and Teledesic must be viewed as offspring of the overheated dot.com economy. However, systems like Teledesic may become feasible in the future. Therefore, the idea has not been abandoned completely. Satellites are applied in numerous areas other than telecommunications. Examples are astronomy and astrophysics, meteorology, Earth resource management and surveillance, navigation, space research, and so on and so forth. 9.4.2

Satellite Orbits

Satellite orbits can be classified as follows. •

•

4.

Low Earth orbit (LEO) satellites are at an altitude above the surface of the Earth between 500 and 1,000 km. The orbit periods of these satellites are between 1.6 and 1.8 hours. Medium Earth orbit (MEO) satellites are at an altitude between 10,000 and 30,000 km. The GPS satellites are MEO satellites at an altitude of 20,200 km.

Originally the system concept consisted of 77 satellites. The system was christened Iridium because element No. 77 of the periodic system is Iridium. When the number of satellites was reduced to 66 satellites, the name was not changed because element No. 66 carries the more prosaic name dysprosium.

9.4 Telecommunications Satellite Services

•

255

Geostationary (GEO) satellites are located in an equatorial orbit at an altitude of 36,000 km. These satellites make one revolution in 24 hours, and, since they rotate in the same direction as the Earth, they are always at a fixed location relative to a point on the surface of the Earth. There is only one geostationary orbit.

The inclination of the satellite orbit is the angle the orbit makes with the equatorial plane of the Earth. The geostationary orbit has 0° inclination, while a polar orbit has 90° inclination. The inclination of the GPS satellites is 55°. The satellite orbit is elliptical, where the Earth is located in one of the foci of the ellipse. The eccentricity is the deviation from circularity of the orbit: the larger the eccentricity the more elliptical is the orbit. The eccentricity of an ellipse is a number in the interval 0 ≤ ε < 1, where 0 corresponds to a circle.5 The ideal geostationary orbit is circular. However, the ideal geostationary orbit is unstable and gravitational perturbations caused by the Sun and the Moon and inhomogeneities in the gravitational field of the Earth converts the orbit into a helix winding around a slightly elliptical path. Therefore, the position of the satellite is not entirely geostationary. The satellite will drift along the orbit and move up and down vertically relative to the equatorial plane. The actual geometry of the geostationary orbit is very complex. The orbit position and the attitude of the satellite must be adjusted from time to time. This activity is called station keeping. For this purpose, the satellite is equipped with several small booster rockets that are used to move the satellite along the orbit, adjust the altitude and inclination of the orbit, and alter the rotational speed of the body of the satellite. Intercontinental satellite systems (INTELSAT), INMARSAT, broadcast satellites, and many national satellite systems exploit the geostationary orbit. The geostationary orbit is very crowded and contains also much waste (remnants after earlier satellites). Now satellites placed in the geostationary orbit shall have a small remainder of propulsion fluid at their end of life so that they can be pushed out of the geostationary orbit (if the telemetry system still works). There are no satellites at altitudes between 2,000 an 8,000 km because of the intense electromagnetic radiation and high particle activity (electrons, protons, ions, and various force particles such as mesons) in this region (the Van Allen radiation belt). Even a short stay at these altitudes may knock out the electronics in the space craft. Transfer orbit is the orbit in which the satellite is placed before it is maneuvered into the final orbit via the deployment orbit. For simplicity, the deployment orbit in Figure 9.5 is shown as a single arch from the transfer orbit to the geostationary orbit. In reality, the orbit may spiral toward the final orbit for several revolutions around the Earth. The transfer orbit of a GEO satellite is highly elliptical with the apogee (highest point of the orbit) close to the geostationary orbit (altitude of 5.

(

)

The general formula for a conic can be written in the form y2 + 1 − ε2 x 2 = 4 px, where is the eccentricity and p is a constant. If 0 ≤ ε < 1, the conic is an ellipse; if > 1, the conic is a hyperbola; if = 1, the conic is a parabola.

256

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

36,000–37,000 km). The lowest point (the perigee) is almost touching the atmosphere at an altitude of about 300 km. The booster rockets are used to push the satellite from the deployment orbit and into its final position in the geostationary orbit. This operation may take several weeks. During this time, the satellite will pass through the Van Allen belt several times. The orbital parameters for the geostationary orbit, the launch orbit, and the deployment orbit are shown in Figure 9.5. Placing the satellite in orbit is a complex process consisting of a number of stages: • •

•

•

•

• •

Launching; Separation of the satellite from the launch vehicle and transfer of control from the launch control center to the satellite control center; Orienting the satellite in the transfer orbit so that it is ready for injection into the final orbit; Accurately controlling the apogee boost rockets to maneuver the rocket into the final orbit and moving it along the orbit to the desired location; Spin stabilizing the satellite so that the communication antennas are pointing toward the Earth; Unfolding the solar panels and activating the Sun acquisition subsystem; 6 Testing all telemetry functions and the telecommunications platform.

During an eclipse, the satellite is in the shadow of the Earth and does not receive light from the Sun for its solar panels. The time the satellite is in eclipse depends on the orbit. The maximum time the GEO satellite is in the shadow of the Earth is approximately 1 hour per 24 hours, or about 4% of the time. The maximum occurs around equinox, while around summer and winter solstice there is no eclipse. The maximum eclipse of a LEO satellite is about 30 minutes every 100 minutes (the revolution time of the LEO) or 30% of the time. If the LEO orbit rotates such that the orbital plane is vertical ±8° to the direction from the Sun, there will be no eclipse. Such orbits are called Sun synchronous. Sun synchronous orbits are exploited in several scientific satellite missions to provide maximum solar energy to the batteries. These orbits must obviously be polar orbits. Only one or two polar orbits of a global system such as Iridium can be Sun synchronous. All other orbits will experience eclipse between 0% and 30% of the time. Full operation during an eclipse is ensured by battery backup. The batteries are then loaded when they receive sunlight.

6.

One amusing event that really showed what could be done to a satellite took place in 1976. The telecommunications transponder of the first MARISAT satellite did not work properly. Investigations of the second satellite that was ready for launching showed that the problem could be caused by gold dust in one of the microwave cavities in the transponder. The satellite was then shaken by firing booster rockets on opposite sides of the satellite in rapid succession. This apparently caused the dust to settle somewhere in the cavity, where it no longer caused signal loss. After that the satellite worked without any further problems for many years.

9.4 Telecommunications Satellite Services

257 Deployment orbit

Geostationary orbit

Apogee 35,000–37,000 km

Altitude 36,000 km

Transfer orbit Perigee 300–500 km

Launch orbit

Figure 9.5

Bringing the satellite into the geostationary orbit.

The battery capacity of a LEO satellite must be such that it allows full operation during the maximum eclipse period of 30 minutes. The battery must be fast-charging, since it must be fully charged in about 1 hour. A final point is concerned with how the satellite can direct the communication antennas toward the Earth and not point it in an arbitrary direction in space. This is called attitude control. If the satellite body is spinning, the principle is simple. A set of optical sensors consisting of a telescope and a photo detector are mounted on the satellite body. As the body rotates, the sensors will detect a low temperature (a few K) if the detector is pointing toward empty space and a high temperature (about 300K) if the sensor is pointing toward the Earth. This gives rise to pulse patterns from the sensors that can be used to determine exactly where the Earth is located relative to the sensor assembly. In fact, just two such sensors called south-pointing and north-pointing Earth sensors are required for attitude control of a spinning satellite. The satellite body may also contain sensors for gravitational sensing and Sun tracking to support the attitude control. The principle is shown in Figure 9.6 for four attitudes of the satellite. Pairs of small booster rockets on the body of the satellite are fired for very short periods of time, causing the satellite to rotate vertically to its spinning axis and thereby altering the attitude of the satellite until the sensors point directly at the Earth. Once the satellite body is in the correct position, servo loops ensure that the satellite retains its proper direction in space by firing the booster rockets when necessary. The attitude of the satellite can also be altered by the satellite control center (the TT&C center) on Earth. The antenna is stabilized by letting electrical motors spin the antenna in the opposite direction of the satellite body and, in case of spot beam operation, even use

258

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

Attitude

Figure 9.6

Earth

Sensor signal

Attitude control.

standard antenna tracking methods to rotate the antenna into position toward one Earth station antenna. The body of a three-axis stabilized satellite does not rotate regularly relative to the Earth. Three-axis stabilization means that the attitude control keeps the whole body of the satellite in a fixed orientation in space. In such systems, either spinning sensors can be used, thus providing the same type of attitude control as spinning satellites, or a sensor array can be used to scan the environment electronically, thereby producing a signal pattern determining the location of the Earth. 9.4.3

Frequency Bands

The most commonly used frequency bands for satellite communication are (using the convention uplink/downlink) as follows: •

•

INMARSAT and other international mobile satellite communication systems 1.6/1.5 GHz (uplink, 1.610 to 1.660 GHz; downlink, 1.525 to 1.610 GHz; These bands contains subbands for maritime services, aeronautical services, land mobile services, and distress services); Intelsat, 6/4 GHz (uplink, 5.925 to 6.425 GHz; downlink, 3.700 to 4.200 GHz) and 14/12 GHz (uplink, 14.000 to 14.500 GHz; downlink, 11.700 to 12.200 GHz);

9.5 Architecture of Communication Satellite Networks

• •

•

259

Fixed domestic systems, 30/20 GHz; NSTAR (Japanese mobile system) and other domestic and international systems (INMARSAT), 2.5/2.0 GHz; Broadcast, 12 GHz (only downlink—the uplink may be in the 6-GHz band).

These bands are allocated on a global basis. Observe that the frequency of the downlink is smaller than that of the uplink. The reason is that the higher the frequency, the larger is the propagation loss. This is will become evident in Section 9.7.1.

9.5

Architecture of Communication Satellite Networks 9.5.1

Broadcast Satellite Systems

In broadcast satellite systems, television, audio, and data signals are fed to the satellite and broadcast to the users. The primary mission of broadcast satellite systems is to offer point-to-multipoint services. Several video, audio, and data channels are multiplexed on each FDMA carrier or TDMA timeslot assigned to each feeder Earth station. The users receive the signals directly from the satellite. The user terminal consists of a parabola for receiving the signals from the satellite and an electric module containing low noise amplifiers, demodulators, video, and audio channel demultiplexers and possibly decryption or descrambling equipment for extracting the individual channels. The advantage offered by the satellite is that the satellite covers a large area and many users. Though the satellite is expensive, the broadcast satellite system is much cheaper than a terrestrial broadcast system. The system configuration is shown schematically in Figure 9.7. In the figure, P represents the producer of the information (for example, a broadcasting company) and B bundles and multiplexes the broadcast channels received from the producers and presents them to the satellite system. The receiver antennas at the homes are small parabolas with a diameter of about 60 cm. The most commonly used frequency band on the downlink is in the 12-GHz range. The common broadcast channel may be radiated in global beams or, what is more common, in spot beams. As we shall see soon, the higher the frequency, the smaller the antenna is needed to detect the signal. The 12-GHz band was chosen some 30 years ago for satellite broadcasting services because the frequency is suitable for small antennas and is within the technological range where simple and cheap equipment can be designed (wavelength of about 2.5 cm require microwave components with linear dimensions of about 1 cm). One particular application of broadcast satellites is shown in Figure 9.8. The satellite link offers a one-way broadband channel but no return channel from the user via the satellite. Since the user usually only requires narrowband services in the direction toward the network, the return link can be offered via the terrestrial network. This is one simple way for the broadcast provider to become an ISP in competition with IPSs offering terrestrial Internet services. The receiver equipment at home (set-top) must then be equipped with additional electronics that pick out the data channel and be able to establish the return

260

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

FDMA carrier with multiplexed broadcast channels

FDMA carrier with multiplexed broadcast channels Common broadcast channel

P

P B

B P P

Figure 9.7

Broadcast satellite system.

connection via the fixed network. This technique is similar to that used by some cable television companies to avoid investment in new infrastructure. 9.5.2

Fixed Satellite Systems

Satellites are used in the fixed telecommunications services in the transport network (point-to-point systems), where one advantage is that the cost of the transmission

Broadband link

User terminal

Server Narrowband link

Network

Figure 9.8

Internet services over broadcast satellite.

9.5 Architecture of Communication Satellite Networks

261

path is largely independent of distance and the number of Earth stations connected to the satellite. It is simple and cheap to add new Earth stations to the network and to reconfigure the traffic between the Earth stations. In these systems the satellite acts as a repeater and sometimes also as a router or switch in the communication link (see Section 9.6). Earth stations in the intercontinental system of Intelsat Ltd. are usually called teleports. The configuration of such a system is shown in Figure 9.9. The configuration is the same for satellite systems used in the national transport networks. The satellite links are interconnecting the telecommunications network in the same way as optical fibers in the terrestrial network. The satellite systems are particularly important in areas where it is too difficult or too expensive to build optical fiber networks. This and the simple configurability of the network are the major advantages of satellites in the transport network. The disadvantages are as follows: •

•

The propagation delay over geostationary satellites is about 260 ms from one Earth station to another. The two-way delay, being equal to the time between question and reply in a conversation, is then about 520 ms. Such a long delay reduces the subjective quality of the conversation because the reactions of the other party appears to be very sluggish. Therefore, two-way satellite connections are avoided if possible. The second disadvantage has to do with the overall reliability of the service. The average lifetime of satellites is approximately 15 years. This means that stringent contingency plans, including spare satellites in orbit, must be

Transport network

Teleport

Figure 9.9

Transport network containing satellite links.

262

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

devised. These plans must also take into account the event that the launch is unsuccessful and potential interruptions of the launching services because of general problems with the launchers. A fault in one satellite may then cause long outages of service. The antenna diameter of the Earth stations in the fixed satellite service is usually between 10m and 30m. Thirty meters was the antenna size of the early systems in the 6/4-GHz bands. These antennas are still in operation. These huge antennas must be pointed at the satellite with an accuracy better than 0.02°. This accuracy is achieved by servo systems and huge gears rotating the antenna around two axes, one vertical and one horizontal. The propagation characteristics and link planning of satellite systems is considered in Section 9.7. 9.5.3

Mobile Satellite Systems

Mobile satellite systems offer access to users located in areas not feasible for terrestrial technologies. Applications include maritime and aeronautical communication and land mobile services for distress operation and access to isolated areas. The particular characteristics of these systems are considered in detail in Section 9.9. 9.5.4

Very Small Aperture Terminal Systems

Local area networks can be implemented using very small aperture terminal (VSAT) technology. VSAT networks have a number of applications where other solutions are impractical or impossible. Examples are as follows: •

• •

Interconnecting remote sites such as hydropower plants and dams with central operation, management, and control systems; Providing temporary telecommunications capabilities for construction sites; Setting up local area networks for companies with operations widely distributed over large distances, such as shopping chains.

The most important advantage of VSAT systems is that they are easy to install, expand, and reconfigure. The most common configuration of a VSAT system is a star network consisting of a central station or hub and a number of outstations. Communication between outstations must then be relayed via the hub. The VSAT system can also be configured as a mesh network (or as a combination of star network and mesh network), where one of the stations is designated as the network controlling station. In this configuration, the outstations may communicate directly with one another. The VSAT system may offer transmission of data, voice, and video signals. The typical configuration of a VSAT network is shown in Figure 9.10. PBX designates a private branch exchange interconnecting the local telephone system of outstations and hubs.

9.6 Telecommunications Components of a Satellite

263

TDMA is employed on the direction from outstation to hub, where one or more timeslots are assigned to each outstation. TDM on a single carrier is used in the direction from the hub. The most common frequency bands are in the 14/11-GHz range; that is, in a band around 11 GHz for the downlink and around 14 GHz for the uplink. The antenna size is in the range of 1.2m to 1.8m.

9.6

Telecommunications Components of a Satellite The satellite is primarily a repeater (amplifier and signal regenerator) placed in orbit around the Earth. The telecommunication payload of the satellite consists of equipment such as antenna, duplexer, low-noise receiver (LNR), signal regenerator (REG), possibly routing and switching circuits, frequency converter (FQC), and high-power amplifier (HPA). The configuration is shown in Figure 9.11. Each path through the satellite is called a transponder. The satellite may contain several transponders. The telemetry and control subsystem allows the satellite to send measurement reports concerning the operational status of all its components to the telemetry, tracking, and control (TT&C) station monitoring the operation of the satellite. The subsystem also receives and distributes commands to actuators and control devices required for station keeping (orbit repositioning), reconfiguration of the transponders, updating software, and other control functions.

Video conference center

Video conference center Outstation

PBX

Operation and management

LAN

Telephone network

Hub

PBX

LAN Outstation

LAN

Router Internet

Figure 9.10

VSAT system.

264

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

Transponder

LNR

REG

FQC

Optional switch

Parabola, phase array One or more antennas

Figure 9.11

Duplexer

HPA

Telemetry and control

Actuators Controls Measurements

Transponder.

The satellite may contain one or several communications antennas. These may be parabolas or multibeam phase array antennas. The antenna may offer a single beam covering approximately one-third of the Earth or several spot beams only covering a small area each. Spot beam systems are also referred to as SDMA systems (see Section 5.8). The multiple access technique may be FDMA or TDMA, or a combination of these methods. In some systems (e.g., those of INMARSAT), there may also be random access channels offering particular services such as signaling for allocation of communication resources. Many complex combinations of multiple access techniques are often exploited together to optimize the satellite to the purpose it serves. Channels may be assigned on a permanent basis or on demand (DAMA). See Chapter 5 for details. The low-noise receiver amplifies the weak microwave signal before it is handled further. The further signal handling depends on the configuration of the demand assignment system and the antenna beam configuration. In the simplest case, the signal is fed to a signal recovery device (REG), where the signal is amplified, resynchronized, and reshaped. Then it is fed to the FQC, where the frequency is shifted to the downlink frequency. Finally, the signal is amplified by the high-power amplifier before if is fed to the antenna system. If the satellite offers a complex multiple access scheme consisting of a mixture of TDMA, FDMA, and SDMA, the regenerator may demodulate and demultiplex the received signal in each transponder and feed it to a switching device, where the individual channels are distributed among all the transponders, remultiplexed, and remodulated before the signal is presented to the transmitter chain.

9.7 Propagation Characteristics, Noise, and Link Budgets

9.7

265

Propagation Characteristics, Noise, and Link Budgets 9.7.1

Attenuation

The commonly used antenna in satellite communication is the parabola with a radiation pattern as shown in Figure 9.12. The parabola collimates the radiated signal into a narrow beam as shown in the figure. The antenna gain expresses how effectively the antenna collimates the beam. The figure also shows that the shortest distance from the Earth to the geostationary satellite is 36,000 km, while the longest path corresponding to the edge of coverage is about 42,000 km. In the link budgets, the calculations are done for the longest path and for the antenna gain in this direction (indicated by min in the figure). Figure 9.13 shows possible sources of signal degradation along the propagation channel. The power of the signal pt fed to the antenna by the high power amplifier (HPA) multiplied by the antenna gain gt is called the equivalent isotropic radiated power 7 (EIRP); eirp = ptgt or in decibels: EIRP(dBW ) = Pt (dBW ) + Gt (dB), where we dBW means dB relative to 1W. Small letters will be used for the absolute value of a variable, and the corresponding capital letter will be used for the same variable expressed in decibels. If the antenna gain is 13 dB and the power from the HPA is 4W, then the EIRP = 13 dB +10 log 4 dBW = 19 dBW or eirp = 80W. 2 The free space loss is ( λ / 4πd ) , where d is the distance between the transmitter and the receiver, and is the wavelength. In decibels: L fs = −147.5 + 20 log d + 20 log f, where f is the frequency in Hz (f = c/ , where c is the speed of light in vacuum) and the distance d is given in meters. Absorption by the troposphere is shown in Figure 9.14. The path-length through the troposphere is less than 10 km. For frequencies below 20 GHz, the

Max 42,000 km

6,378 km

Min 36,000 km Gain contour

Geometry of geostationary satellite

Radiation pattern of parabolic antenna

Figure 9.12

7.

Antenna pattern and link configuration.

The EIRP is the power an isotropic antenna would have to radiate uniformly in all directions of space in order that the signal strength produced at the receiver is the same as that of the collimated antenna.

266

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

Absorption and scattering by air and precipitation EIRP

Environmental noise from air and precipitation

Receiver noise Environmental noise from ground Receiver loss

Figure 9.13

Preamplifier

Loss and noise in radio systems.

attenuation is therefore less than 10 km × 0.05 dB/km = 0.5 dB. Thus, 0.5 dB may then be used as a conservative figure for the atmospheric loss in the link budget. The attenuation curves for rain and snow will differ from place to place and have a form as illustrated roughly in Figure 9.15. The best case corresponds to a dry climate, while the worst case corresponds to a tropical climate with frequent and intense rain. It is impossible to draw universal curves for the rain attenuation because of the large variation of climate between different locations: measurements over several years must be made at the particular antenna location to establish accurate statistics. However, there exist theoretical models where fairly accurate rain attenuation curves can be estimated from climatologic data. Here are a few general rules. The higher the frequency, the more severe the rain and snow attenuation are. The higher the rain or snow intensity is, the larger the loss is. The attenuation in snow depends strongly on how wet the snow is and how big the snowflakes are. The rules are that the wetter the snow, the larger the attenuation (high dielectric constant); and the bigger the snowflakes, the bigger the attenuation (large radar cross section). Note that 0.1% and 0.01% of a year is 8.8 hours and 0.88 hour (or 53 minutes), respectively. Selecting a rain margin is thus the same as selecting the availability of the link. If an availability of 99.9% is sufficient, then 1 or 2 dB may be a safe margin. If the availability shall be 99.99%, the margin in a dry climate may be 5 dB while that in a wet climate may be as high as 15 dB or more. For a system with global coverage, a single rain margin is often chosen for the entire system. The tradeoff is then between offering less availability or increasing the size of the receiving antenna in wet climate. The antenna gain is proportional to the area, or, equivalently, to the square of the radius of the parabolic reflector. This means that a 3 dB increase of the antenna gain corresponds to about 40% increase in antenna radius. If a parabola with 20m diameter is enough in dry climate, the diameter of an antenna in wet climate offering the same availability may then be 28m or larger. Interference from systems sharing the same part of the radio spectrum is another cause of signal deterioration. Radio relay systems and satellite systems share the 6/4-GHz band. Under particular circumstances, these systems may cause significant

9.7 Propagation Characteristics, Noise, and Link Budgets

267

100

Attenuation dB/km

O2

10

1 H2O

0.1

20

Figure 9.14

40

60 Frequency GHz

80

100

Absorption in the troposphere.

interference on each other. Furthermore, circular polarization is used in many satellite systems. Deviation from circularity causes loss in signal power. In systems with linear polarization, Faraday rotation of the polarization angle when the signal passes through the ionosphere causes loss in the same way. A particular loss component in maritime satellite systems is multipath interference caused by reflections from the sea and the superstructure of the ship. Let us call the total loss caused by all the sources discussed earlier (and possibly a few more) for M dB. Then the received signal is Pr = EIRP − Lfs − M. The loss M is a variable loss that will be treated in a particular way in the link budget. EIRP and Lfs are calculated for the longest path at the edge of coverage. 9.7.2

Noise

To calculate the signal-to-noise ratio at the receiver, we must first determine the noise. The noise power N in a bandwidth B caused by thermal noise is given by N = kTB where k is Boltzman’s constant (= 1.38 × 10 −23 W / sK (watt per second per kelvin)) and T is the temperature in kelvin of the device producing the noise. T is also called the noise temperature. In this formula, B is often referred to as the noise bandwidth to distinguish it from the signal bandwidth. In a linear system, the signal bandwidth and the noise bandwidth are equal but in nonlinear systems such as

268

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

Percent of period attenuation exceeded

10

1

0.1

Worst 0.01

Best

0.001

5

10

15

20

25

30

Attenuation in dB

Figure 9.15

Typical rain attenuation curves.

phase-locked loops and coherent demodulators signal bandwidth and noise bandwidth are different. The signal-to-noise ratio (S/N) is defined as the ratio between the signal power and the noise power at the same point in the system. The noise figure of an electrical circuit expresses the change in S/N ratio when the signal passes through the circuit; that is n f = ( S / N ) in / ( S / N ) out

A signal S that passes through an attenuator with loss l and embedded in a medium with temperature T is reduced by a factor l. Since the medium has a constant temperature, the noise power is not affected by the loss but remains at the same value kTB at both the input and the output. This means that the noise figure for an attenuator with loss l is nf = l since (S/N)out = (Sin/l)/Nin = (1/l) (S/N)in. Note that the noise produced by the medium is Nin = kTB. We now define an equivalent input noise temperature Tin by the following formula Nout = (input noise produced by the medium + input noise equivalent to that produced by the attenuator)/loss = (kTB + kTinB)/l. Inserting this expression for Nout, the expression for the input noise Nin = kTB and Sout = Sin/l in the formula for the noise figure we get the following expression for the noise figure and the equivalent input noise temperature of a circuit with loss l and temperature T

9.7 Propagation Characteristics, Noise, and Link Budgets

269

n f = l = 1 + Tin / T or Tin = T(l − 1)

Similarly, we define the equivalent output noise temperature as Tout = Tin / l = T (1 − 1 / l ). The equivalent input and output noise temperatures allow us to calculate the equivalent noise temperature at any reference point in the system. These formulas will now be used to derive an expression for the noise in the receiver of the satellite system. The configuration of the receiver system is shown in Figure 9.16. To calculate the S/N in the Earth station, we must determine the signal power and the noise power at the same reference point. The reference point chosen in Figure 9.16 is the output of the antenna. The universe and the dry atmosphere contribute with a noise temperature Tc of approximately 100K and 20K at elevation angles of about 5° and 90°, respectively. The temperature may vary slightly with the solar activity. The antenna will always experience this temperature whether the sky is clear or there are clouds, rain, or snow. The noise temperature of the Moon is about 300K. Therefore, there is usually no problem if a high gain receiving antenna points directly at the Moon’s surface. The noise temperature of the Sun at microwave frequencies is between 100,000K and 1 million K, depending on the sunspot activity. If the receiver antenna points directly at the Sun, communication may be completely blocked. However, the noise contribution from the Sun depends on the gain of the receiving antenna. An antenna with gain exceeding 55 dB will have an opening angle smaller than the diameter of the Sun. The noise contribution is then between 100,000K and 1 million K as long as the antenna points directly at the Sun. It is easy to see from the orbital geometries of the satellite and the Earth and the seasonal rotation of the equatorial plane that such situations only occur seasonally for certain locations of the Earth station antenna. The interference takes place only as long as the Sun is within the antenna beam (i.e., only for a few minutes a day for a few days per year). However, if the antenna gain is less than 55 dB, the noise temperature of the Sun decreases by the same amount. If the antenna gain is 25 dB, the noise temperature is reduced by a factor corresponding to 55 − 25 = 30 dB = 1,000; that is, the noise temperature is between approximately 100K and 1,000K. The noise contribution of the Sun in an omnidirectional (isotropic) antenna is less than 3K. Noise generated by the Sun is then only a problem for high-gain antennas. Around equinox, the Sun may cause increase in the noise temperature of the satellite. This is also easily seen from geometrical arguments. This interference is not severe.

Figure 9.16

Equivalent antenna temperature of the Earth station.

270

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

If there is rain, the ambient temperature is Te, producing the equivalent noise temperature at the antenna of Te(1 − 1/le), where le is the loss due to the rain. If there is no rain, le = 1 and this noise component disappears. A typical value for Te is 290K (20°C). We see that the loss through rain enters twice in the performance calculation of the satellite link; that is, both in the calculation of the noise temperature and as the variable loss component M defined earlier. The feeder produces a noise temperature of Tf(lf − 1) at the reference point. The feeder temperature Tf is set to 290K (20°C). Finally, the equivalent noise temperature of the low noise amplifier at the reference point is lfTr, where Tr is the noise temperature of the receiver.8 Usually the noise produced by the preamplifier is given in terms of the noise figure of the amplifier: Tr = (nr – 1)T0, where T0 is a reference temperature usually taken to be 290K and nr is the noise figure conventionally given in dB as Nr = 10 log nr. The equivalent noise temperature of the antenna is thus

(

)

Ta = Tc + Te (1 − 1 / l e ) + Tf l f − 1 + l f Tr

In this formula, all components except the rain contribution do not vary over time and can be regarded as system constants. The noise temperature without rain contribution is referred to as the system temperature Ts

(

)

Ts = Tc + Tf l f − 1 + l f (n r − 1)T0 and Ta = Ts + Te (1 − 1 / l e )

where the noise produced by the preamplifier has been expressed in terms of the noise figure of the amplifier. The rain contribution is usually treated in a particular way in the link budget just as the loss caused by rain. The satellite is looking at the Earth having a noise temperature of 290K. There is, of course, no increase because of rain, since rain does not change the temperature of the Earth viewed from the satellite. However, the path loss caused by rain must be included in the total loss also on the uplink. 9.7.3

Example 1

Let us calculate the noise temperature for the following case: Tc = 100K Tf = 290K Lf = 1 dB or lf = 1.26 Nr = 1 dB or nr = 1.26 (high-performance low noise receiver) T0 = 290K Te = 290K Le = 3 dB or le = 2

8.

An attenuator with loss l reduces a noise signal Ni at the input in the same way as any other signal: No = Ni/l . Do not confuse this with the noise produced by the attenuator.

9.7 Propagation Characteristics, Noise, and Link Budgets

271

We find Ts = 100 + 290 × (126 . − 1) + 126 . × (126 . − 1) × 290 = 270K and Ta = 270 + 290 × (1 − 1 / 126 . ) = 331 K. This figure is typical for a large Earth station in the Intelsat system. 9.7.4

Link Budget

We found earlier that the signal received at the remote antenna is given as pr = eirp/lfs in linear terms (we will go back to decibels in a moment). At the reference point at s the output of the receiver antenna, the signal is received with the strength p r = gr eirp/lfs, where gr is the gain of the receiver antenna. The system noise power at the reference point is kTsB. The carrier-to-noise ratio is the ratio between the received signal power, and the noise in decibel at the reference point is then C/N = 10 log (gr eirp/kTsBlfs), or C / N = EIRP − L fs + (Gr / Ts ) − 10 log B + 2286 . or C / N = EIRP − 20 log d − 20 log f + (Gr / Ts ) − 10 log B + 376

where we have inserted L fs = −147.5 + 20 log d + 20 log f for the path loss. In this formula we have introduced the figure of merit of the receiver antenna (G r / Ts ) = G r −10 log Ts . EIRP is the equivalent isotropic radiated power of the transmitter in dBW, d is the longest distance between the satellite and Earth in meters, f is the carrier frequency in hertz, and B is the noise bandwidth of the system in hertz. The formula is used to calculate the carrier-to-noise ratio of the received signal. From this figure we then can derive the bit error rate of the signal when we know the modulation and demodulation methods used. From the formula we see that the carrier-to-noise ratio decreases as the carrier frequency increases. Since the available power is limited in the satellite, the lowest frequency is always used on the downlink (satellite-to-Earth), thereby reducing the path loss in this direction. Furthermore, the uplink is designed such that the carrierto-noise level in the satellite receiver is much larger than that of the downlink. Therefore, we may usually neglect the noise contribution from the uplink in the overall link budget. 9.7.5

Example 2

Let us calculate the satellite EIRP at beam edge for a downlink with the following characteristics: f = 4 GHz d = 40,000 km B = 1 MHz Ts = 270K from Example 1 Gr = 60 dB C/N = 12.5 dB (bit error rate of 10−6 for 4PSK without error correction) Margin M = 5 dB (rain and other effects such as depolarization and pointing error)

272

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

First we find that (Gr/Ts) = 60 − 10 log 270 = 35.7 dB/K. Inserting all the values in the equation EIRP = C/N + 20 log d + 20 log f –(Gr/Ts) + 10 log B – 376 + M we find EIRP = 10 dBW or eirp = 10W.

9.8

Tradeoffs 9.8.1

Cost

Geostationary satellites are used for broadcasting. This is why the parabola used for receiving satellite television can be kept pointing in the same direction all the time (i.e., pointing at the fixed location of the geostationary satellite). With nongeostationary systems the antenna must track the satellite to maintain communications. This is an example of tradeoff between using an expensive satellite orbit and offering a simple receiving system at each home. If a cheaper orbit (less expensive with regard to satellite launching) had been used, the Earth segment would have become so expensive that the system would hardly be used. This can be seen from the following equation. The cost per user (Cu) of a satellite system is given by C u = C sat / N + C rec

The cost for the satellite system is Csat and the cost of the user terminal is Crec. N is the number of users. Both cost elements must be small. The cost of the satellite per user, Csat/N, is small if there are many users (large N). The cost of the receiver, Crec, is small if the electronics and in particular the mechanics of the receiver is simple. The cost of the receiver is also small if the receiver is equipped with omnidirectional antennas. An omnidirectional antenna receives the signal equally well from all directions in space. The gain of such an antenna is 1 (0 dB). However, from the link budget in Section 9.7.4, it follows directly that the satellite must then produce more energy to produce a high enough carrier-to-noise ratio in the receiver. Producing more energy makes the space segment more expensive. Land mobile terminals, terminals for small vessels, and terminals mounted on aircraft employ antennas that are almost omnidirectional. In order to not make the first INMARSAT system too expensive, the antenna at the ship terminal is a directive parabola with gain between 20 and 24 dB pointing at the satellite. An electronic control system compensates for roll, sway, and pitch of the ship so that the antenna always points in the direction of the satellite independently of the motions of the ship. The EIRP of the satellite is 20 to 24 dB smaller than systems employing omnidirectional antenna at the ship. Let us now look at the cost of the satellite segment, Csat. It is particularly expensive to place a geostationary satellite in orbit. However, only three or four satellites are required for a global system. It is comparatively cheap to place the satellite in a low-Earth orbit. However, many satellites are required for global coverage. The Iridium system required 66 satellites, and the Teledesic system was planned with 840 satellites. The cost estimate for the satellites of the Teledesic system is $10 billion. In comparison, the geostationary satellite system owned by INMARSAT offering worldwide communication to ships and aircraft costs about $2 billion. The higher cost of the LEO system is compensated for by a higher business potential than

9.8 Tradeoffs

273

the GEO system. The failure of the Iridium system showed that this is certainly not always the case. The cost of a GEO satellite amounts to about one-third of the overall cost. The cost of the launch is also one-third. The final one-third includes insurance, management, and operation of the satellite. The cost of a satellite is roughly proportional to its weight. The weight also represents a complex tradeoff concerning the accuracy of station keeping (i.e., how accurately the satellite is kept in its position in the orbit). The reason is that the spacecraft must contain rocket fuel for repositioning the satellite as it drifts away from its intended position because of the gravitation pull of the Sun and the Moon and anomalies in the Earth’s gravity caused by mountains and ocean depths. The more rocket fluid, the more often the orbit can be corrected, but the more expensive the spacecraft is. On the other hand, repositioning may prolong the usable lifetime of the system and hence reduce the time between investments. 9.8.2

Other Tradeoffs

There is also a tradeoff between weight, transmitted power level, antenna size, and beam shaping. Large radiated power level may mean larger solar cell panels and more battery capacity, again increasing the weight of the satellite. In the worst case, a larger rocket may be required to get the satellite into orbit. On the other hand, larger solar cells may reduce the cost of the Earth segment, including user terminals since the receiver can be made simpler, as explained earlier for the satellite broadcast service. There is also a tradeoff between the cost of the infrastructure of the satellite system and the cost of the user terminal. The LEO systems will have a high infrastructure cost and a low terminal cost. They need many users to be competitive with terrestrial systems. The INMARSAT system requires expensive user terminals. There is only a small number of users so that both the investment cost and the usage cost are high for this system. On the other hand, satellite communication is the only available communication technology for ships in international waters and aircraft on intercontinental routes, and the satellite operator can make use of monopoly advantages. The tradeoffs is also concerned with choice of operating frequencies, on-board processing, and several other items too numerous to include here. The total economy of the system will include elements such as the following: •

•

• • • •

The cost of the different components of the system: satellite, launching, operation, management, insurance, spare equipment, and Earth segment; The usage forecast including increase in number of users and churn rate (users lost to or gained from competing systems); The charges and charge fluctuations due to competition and market feedback; Market segmentation and penetration rate depending on usage cost; Variation of interest rate on debt; The cost of in-orbit repair if applicable;

274

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

•

•

The cost of destroying satellites no longer in use (e.g., pushing it out of the geostationary orbit); The cost of delays.

Note that the cost of the satellite system is a sunk cost in the sense that the system can hardly be used for purposes other than those for which it was designed, though some of the cost of the Iridium system was recovered after the bankruptcy. Since the investment is huge, the economy of satellite communication requires deep and sober analysis before the decision to implement it is taken.

9.9

Mobile Satellite Communication 9.9.1

The INMARSAT System

The INMARSAT system is in all respects a very marginal system. Ship Earth stations with very small antennas are to be served via geostationary satellites. The bandwidth per channel, or equivalently the information rate, must then be small since the amount of power that the satellite must radiate is proportional to the bandwidth of the signal (see the link budget in Section 9.7.4). As mentioned earlier, the first study of maritime satellite communication was undertaken in Norway in 1969 where it was demonstrated that communication with ships via geostationary satellites was indeed feasible both technically and economically. At about the same time, an American study showed that communication with aircraft via geostationary satellite was also feasible. The first system, called Marisat, came into operation in 1976. The system was developed by the American company COMSAT and offered initially services to the Atlantic Ocean and the Pacific Ocean. The whole operation was cheap because a set of satellites the company had developed for military purposes had spare capacity to accommodate the maritime transponders. The Indian Ocean was covered in 1977. The Marisat system was taken over by the newly established INMARSAT in 1982. This system was named the INMARSAT-A system. The system is analog, using frequency modulation for the telephone and telex carriers. This was an unfortunate choice, severely hampering the further development of the system. The Norwegian feasibility study and the satellite system to the North Sea put in operation in 1975 had shown clearly that a digital system with moderate speech encoding had superior performance. INMARSAT-A was designed for large ships in the commercial fleet. INMARSAT-A was scheduled to be phased out during 2007 after 30 years. Note that it usually takes a long time to phase out a system. The reason is that the technical and commercial lifetime of an INMARSAT-A terminal is at least 15 years. Replacing the system earlier would be expensive for those ships that still used INMARSAT-A. INMARSAT commenced immediately the development of a digital version called INMARSAT-B for large ships and a terminal called INMARSAT-C that could be fitted on very small ships (e.g., fishing vessels, leisure boats). Soon afterward, INMARSAT-Aero for communication with commercial aircraft and INMARSAT-M for land mobile communication were developed. The newest

9.9 Mobile Satellite Communication

275

narrowband system, INMARSAT-Fleet, also supports bit rates of 64 Kbps and 128 Kbps. The 64-Kbps service is used for extending GSM to commercial aircraft. Presently, INMARSAT is introducing broadband services up to 432 kbps in its broadband global area network (BGAN) made possible with its newest generation satellites (fourth generation). The first of these satellites was launched March 11, 2005. These satellites will offer global and regional BGAN services on spot beams and global beams in addition to the current narrowband services on global beams. All these systems share the same satellite resources and the same cost of Earth stations and control station infrastructure. The INMARSAT systems are summarized in Table 9.2. The table contains the main areas of usage of the different systems, the main services offered by each system, and a selection of the most important data rates. Note that the telex service operating at a transfer rate of only 50 bps is no longer used in the telecommunications network. This simple text service has been replaced by e-mail. 9.9.2

Frequency Bands

The INMARSAT systems are using frequency bands around 1.53 GHz and 1.63 GHz for the link between satellite and ship Earth station (SES) (downlink) and SES and satellite (uplink), respectively. The uplink and downlink frequencies between the coast Earth station (CES) and the satellite are in the 6-GHz and 4-GHz bands, respectively. These are the most commonly used frequency bands for satellite communication. The satellite transponder thus translates the frequencies from 1.63 GHz to 4 GHz and from 6 GHz to 1.53 GHz. Different frequency slots are allocated for maritime and aeronautic systems (19 MHz and 10 MHz, respectively). This allows the two systems to evolve independently of each other. There is also a separate narrow slot between the maritime and aeronautical bands (1 MHz) allocated to common distress and safety services. 9.9.3

Basic Architecture and Procedures

The architecture of the INMARSAT systems (except BGAN) is shown in Figure 9.17. The system consists of coast Earth stations (CES) or land Earth stations (LES) Table 9.2

The INMARSAT Systems

System

A

B

C

Aero

M/M4

Fleet

Broadband

Start of service

1982 (1976)

1993

1991

1990

1993/ 2003

2004

2006

User category

Large ship Large ship

Small vessel Land mobile

Aircraft

Small vessel Land mobile Aircraft (M4)

Small and Maritime large ships Land Offshore

Services

Voice and telex

Voice, telex, and data

Messages and data

Voice and data

Voice and data

Voice and data

Data rates

Analog and 50 bps

16 Kbps, 9.6 Kbps, 300 bps, 50 bps

600 bps

9.6 Kbps, 4.8 Kbps, 4.8 Kbps 4.8 Kbps, 2.4 Kbps 2.4 Kbps 300 bps 64 Kbps (M4) 64 Kbps

IP Up to 432 Kbps

276

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

connecting ship Earth stations (SES), aircraft Earth stations (AES), or mobile Earth stations (MES) to the public network via the satellite. All CESs can communicate with all mobile Earth stations (SES, AES, MES). In addition, one station acts as a network control station (NCS) for each satellite coverage area in charge of allocating communication resources to the CESs. The CES communicates with the NCS over the satellite using the same frequency bands as between the CES and the SES. The reason for this arrangement is that the number of communication channels is so small that all resources must be shared by all CESs on demand assignment within one satellite coverage area. Frequency division multiple access is used for all channels with a single channel per carrier (SCPC). This represents the most efficient way in which to utilize the limited number of communications channels. This applies both on the links between the mobile terminal and the satellite (at 1.6/1.5 GHz) and between the coast Earth stations and the satellite (at 6/4 GHz). Pure Aloha is used for the request message from the ship (the reason is explained in Section 5.9.2). The procedure is as follows. The mobile terminal sends the request message on the random access channel. The request is addressed to one particular coast Earth station. The request also contains the required type of communication resource (voice, data) and priority to support preemptive handling of distress communication. Preemptive operation means that if there is no idle channel, a call without distress priority is forcedly released and the channel is allocated to the distress call.

NCS

CES CES Fixed network

SES

CES

NCS

Request (random access) Assignment request Assignment Conversation

Figure 9.17

INMARSAT architecture and request procedure.

Release

9.9 Mobile Satellite Communication

277

The coast Earth station sends a request for channel assignment to the network coordination station that assigns the channel, marks the channel as busy, and returns the assignment message. Since this message is sent on the 1.5-GHz downlink from the satellite, the message is received by both the coast Earth station and the mobile terminal. The Earth station and the mobile terminal tune to the assigned channel and commence conversation. When the call is released, the coast Earth station hands back the channel to the network coordination station that marks the channel as idle and available for future assignment. 9.9.4

Antenna Tracking in INMARSAT-A, INMARSAT-B, and INMARSAT-Aero

The antenna gain of INMARSAT-A and INMARSAT-B ship Earth stations is between 20 and 24 dB giving an aperture angle between 3 dB points somewhere in the range from 10° to 20°. An antenna with this narrow beam must be mounted on a stable platform to compensate for the motion of the ship in rough sea. Antenna stabilization is done in two steps. First, the platform on which the antenna is mounted must be fixed relative to a Cartesian coordinate system x, y, and z, where z is in the direction of zenith and x and y are the coordinates of the tangential plain of the Earth at the point where the ship is located. The y axis may point in an arbitrary direction, in the direction in which the ship is heading or, say, toward the North Pole if an absolute reference system is provided. An absolute reference system is required if program tracking of the antenna beam is applied. The second step consists of directing the antenna beam toward the satellite. The angle between the x-y-plane and the direction in which the antenna is pointing is called the elevation angle. The exact pointing direction of the antenna is thus given by the elevation angle and the angles between the beam and x and y axes. A stable Cartesian reference system can be achieved in two ways. In the original system designed in 1976, a spinning wheel platform was implemented. The platform consisted of a plate on which the antenna is mounted. The plate balances on a pivot fastened to the superstructure of the ship and contains two flywheels rotating in opposite directions. Because of the inertia, the flywheels act as a gyro stabilizing the plate in the horizontal plane whichever movement the ship makes, provided that the friction between the pivot and the plate is small. The technology is simple but bulky and expensive. In more modern designs, the platform is stabilized by a servo system that uses the gyrocompass of the ship to determine the Cartesian reference coordinates. The gyrocompass provides both the coordinates for the course of the ship and the direction of zenith. Equipping the platform with motors that rotate the platform around the three coordinate axes and sensors such as accelerometers and rate-integrating gyros that measure the instantaneous deviation from the reference coordinates, the servo system will keep the platform in a stable horizontal position. After having provided a stabilized platform for the antenna mount, we have to point the antenna toward the satellite. The pointing direction will change due to the ordinary movements of the ship so that a servo system is required to compensate for these movements. The movement of the ship is usually a very slow process. The beam can be kept in the direction of the satellite by employing either program tracking or step tracking.

278

Line-of-Sight Systems: Fixed Radio Access, Radio Relays, and Satellites

The principle of program tracking is as follows. The direction toward the satellite can be determined accurately from the orbital coordinates of the satellite and the geographic coordinates of the ship. The satellite coordinates are the same during the entire lifetime of the satellite and the coordinates of the four satellites need only to be read into the terminal computer during installation. The geographical coordinates of the ship are provided by the navigation system of the ship. It is then simple trigonometry to calculate the pointing angles of the antenna and simple mechanical maneuver to make the antenna point in the desired direction. Program tracking is also called open-loop tracking since a servo system with feedback is not required to preserve the pointing direction. The second method is called step-track. Step-track requires a servo system operating as follows. Suppose that the mobile Earth station is pointing directly at the satellite. As the ship moves, the pointing becomes more and more inaccurate and the received power level eventually drops with a certain amount, say, 0.5 dB. The servo system then moves the antenna a certain angle in an arbitrary direction. If the level drops further, the antenna is moved in the opposite direction until a maximum power level is detected. Then the process is continued in the direction vertical to the initial one until the maximum power is detected. In a geostationary system, the time between adjustment and the detection of the result of that adjustment is about 250 ms (corresponding to the one-way delay over the satellite). It can be shown that if the loop shall remain stable, the time between adjustment and measurement must be longer than 1 second (four times the delay). The time between adjacent steps must then be longer than one second. INMARSAT-C does not require antenna tracking. INMARSAT-Aero uses program tracking based on the position of the satellite and data from the course computer of the aircraft. The antenna gain is only 12 dB corresponding to a 3 dB opening angle of 50°. This implies that pointing adjustment is not frequently required for normal motion of the aircraft. The Aero antenna is an electronically steerable phase array antenna. Land mobile terminals may also use program tracking combined with GPS. 9.9.5

A Note on Link Budgets

The EIRP of the satellite is then (see the previous link budget): EIRP = C / N + 20 log d + 20 log f − (G r / Ts ) + 10 log B − 376 + M

We shall compare the power requirements of INMARSAT-B and -C and see that the systems are compatible in this way, though they are very different in most other respects. Table 9.3 shows the main characteristics of the INMARSAT-B and two versions of INMARSAT-C, one with zero antenna gain and one with gain corresponding to a crossed dipole. INMARSAT-B applies 3/4-rate convolutional error correction coding, while INMARSAT-C applies 1/2-rate coding, reducing the carrier-to-noise threshold by about 3 dB for the same bit error rate (10−3). INMARSAT-B supports telephony at 9.6 Kbps, giving a bandwidth of 24 kHz. INMARSAT-C supports only message services at bit rates up to 600 bps, giving a total bandwidth of 1,200 Hz.

9.9 Mobile Satellite Communication Table 9.3

279

Characteristics of INMARSAT-B and -C INMARSAT-C INMARSAT-C (3 dB Gain) INMARSAT-B (0 dB Gain)

G/T

−4 dBK

Bandwidth B

24 kHz

1,200 Hz

1,200 Hz

C/N

8 dB

5 dB

5 dB

Carrier frequency f

1.5 GHz

1.5 GHz

1.5 GHz

Distance d

40,000 km

40,000 km

40,000 km

Margin, M

3 dB

3 dB

3 dB

Table 9.4

−23 dBK

−21 dBK

EIRP and Transmitted Power INMARSAT-B

INMARSAT-C (0 dB Gain)

INMARSAT-C (3 dB Gain)

EIRP

18.3 dBW

21.3 dBW

18.3 dBW

Transmitted power

67W

134W

67W

Table 9.4 shows that INMARSAT-B and INMARSAT-C require approximately the same satellite EIRP and no particular precaution needs to be taken when allocating communication resources to the two types of system. Similar considerations apply for INMARSAT-Aero and INMARSAT-M. Equal EIRP in the satellite was an important design criterion for these systems.

CHAPTER 10

Optical Communication Systems 10.1

Why Optical Systems? Optical systems are used for both long-distance and short-distance communication. The obvious benefits of long-distance optical communication systems are large bandwidth and low loss, allowing long distances between amplifiers and regenerators. Overcapacity is cheap, so the time between investments in more capacity is long. Cheap overcapacity also makes the introduction of new broadband applications economically feasible. This is particularly important for the evolution of technologies based on IP, since many of these technologies are independent of the distance over which the information has to travel. A high-capacity connection across the globe may replace a short low-capacity link if necessary. A less obvious advantage is that optical fibers can be tightly packed in the same casing and run alongside one another for long distances without generating crosstalk or interference. This is difficult with electrical transmission systems such as coaxial cables even over short distances. Coaxial cables in the same duct must be equipped with additional shielding in order to avoid that the signals in the cables disturb one another. If the coaxial cable were made of perfect conductors with infinite conductivity, there would be no signal at the outside of the cable: all energy would be confined to the space between the inner and outer conductor. However, finite conductivity of the outer conductor will cause a small proportion of the electrical field to penetrate the outer conductor (the skin effect) and thus generate an induced signal in the neighboring cable. For exactly the same reason it is difficult to eavesdrop a fiber optical cable. There is no field outside the cable that the eavesdropper may pick up. On the other hand, it is simple to monitor coaxial cables that are not shielded. Optical fibers are expensive in the access network because they need to be buried in ducts or be suspended on poles. For the time being, other alternatives such as extending the capabilities of the twisted pair (ADSL) and radio systems such as WiMax are cheaper. However, even these systems are threatened to be replaced by public and local area mobile systems. Optical systems have several applications over short distances. Examples are as follow: •

They can be used in systems where immunity to electromagnetic interference (EMI) is important. This includes immunity to electromagnetic pulses generated by nuclear bombs.

281

282

Optical Communication Systems

•

• •

•

•

10.2

They can be used in low-weight applications in large sensing and control systems in spacecraft, aircraft, and elementary particle detectors. They can interconnect the parts of supercomputers. Fibers can be wound on electrical high-voltage lines, providing cheap mounting of fiber optic systems. Fibers can be used to separate areas with different electrical potentials, since the fibers are effective electrical isolators. Fibers can be used in inflammable environments, since sparks will not be generated by optics alone.

Composition of Optical Networks The telecommunications network may include only optical technology on the transmission sections. The end systems and the switches are still based on nonoptical technology. This is the configuration of the existing telecommunications network. The structure of these networks is shown in Figure 10.1. The optical transmission section consists of optical modulators, multiplexers, fibers, amplifiers along the fiber, demultiplexers, and, finally, optical demodulators. Each of these components is described in Section 10.3. The optical transmission system may be used both in the transport network and in the access network (fiber to the home, fiber to the curb, fiber ring). An all-optical network is shown in Figure 10.2. These systems contain optical switches in addition to optical transmission systems. Optical switches are briefly described in Section 10.4. All-optical telecommunications systems are still to be implemented, since the switching technologies are still not mature for large-scale production. However, the technology is evolving at great speed. Some of the components described later are mature technologies (lasers, filters, MEMS, switches, and fibers), while other

Nonoptical end system

Nonoptical end system

Optical transmission system Optical transmission system

Nonoptical switching system

Optical transmission system

Nonoptical switching system

Optical transmission system

Nonoptical end system

Optical transmission system

Nonoptical end system

Mod

Dem

Mod

Dem Mux

Amp

Fiber

Amp

Mod Mod

Figure 10.1

Demux Dem

Optical transmission system

Network with optical transmission only.

Dem

10.3 Optical Transmission Components

Transmission system

End system

Switching system Transmission system

End system

283

Transmission system

Transmission system

End system

Transmission system

End system

Switching system

Optical system

Figure 10.2

All-optical network.

technologies are still under development (electrically controlled switches, switches based on gratings, and packet switches).

10.3

Optical Transmission Components 10.3.1

Fibers

The composition of an optical fiber is shown in Figure 10.3. The fiber consists of the core surrounded by the cladding. The fiber is protected by an outer buffer and jacket layer of mechanically tough resins and plastic. The buffer and jacket layer does not contribute to the optical properties of the fiber. The fiber may be further reinforced by steel and other material if the fiber is used in particularly harsh environment (e.g., submarine cables). A separate copper sheath may be added for provision of electric power to amplifiers and regenerators along the cable. The optical properties of the fiber are entirely given by the characteristics of the core and the cladding and the boundary between them. The light is transmitted through the core and ideally no light is propagated into the cladding. Light propagating into the cladding is lost. Light may be lost to the cladding if the boundary between the core and the cladding is not very smooth (impurity scattering). Light may also propagate into the cladding by refraction as explained later. The refractive index of a medium is the ratio between the speed of light in vacuum (c = 300,000 km/s = 3 × 108 m/s) and the speed of light in the medium. The

Buffer and jacket

Cladding Core

Figure 10.3

Optical fiber.

284

Optical Communication Systems

speed of light in vacuum is a constant of nature and is the highest propagation velocity electromagnetic waves (and any other wave or object) can attain. The speed cm in a medium with refractive index nm is then cm = c/nm, where c is the speed of light in vacuum. The refractive index for glass is about 1.5 so that light propagates at a speed of approximately 200,000 km/s in glass. A geometric model for propagation of waves in a tube consisting of two layers with uniform refractive index is shown in Figure 10.4. Rays 1 and 2 meet the boundary at small angles of incidence and are reflected back into the core with the same angle between the ray and the boundary as the incident wave (total reflection). Ray 3 meets the boundary at a larger angle and is refracted into the cladding. Ray 4 represents the limiting case where the angle of incidence is such that the ray is refracted horizontally along the boundary between the cladding and the core. This angle is called the critical angle of incidence. By convention, the angle between the vertical to the boundary and the ray is used in the calculations of refraction. This angle is 90° minus the angle of incidence referred to earlier. The geometry is shown in Figure 10.5. The angles between the incident and refracted rays and the vertical axis and the refractive indices are related by Snell’s law: n core sin θ core = n clad sin θ ckad

The critical angle corresponds to clad = 90°, since at larger angles the ray is reflected back to the core. Since sin 90° = 1, sin core = nclad/ncore at the critical angle.

3 4

Refractive index nclad

1

Refractive index ncore

2

Figure 10.4

Propagation of light in the fiber.

θcore ncore

nclad θclad

Figure 10.5

Refraction.

10.3 Optical Transmission Components

285

Since sin ≤ 1 for all angles, total reflection can only take place if ncore > nclad (i.e., the refractive index of the core must be higher than that of the cladding). The refractive index of the core is usually only 1% or less higher than that of the cladding. The critical angle of incidence between the axis of the fiber and the ray is 8° if this difference is 1%, since cos 8° = nclad/ncore = 0.99 ncore /ncore = 0.99. This means that all rays at angles of incidence smaller than 8° will propagate along the fiber and eventually reach the opposite end. A ray that is incident at 8° will travel 1% longer than an axial ray, or, equivalently, requie 1% more time to reach the end of the remote fiber. The delay-spread, defined as the difference in propagation time between a wave following the longest possible path and the axial ray, is thus 1%. The delay-spread gives rise to dispersion of light pulses sent along the fiber. Dispersion implies that the pulse is smeared out over a wider and wider range, causing interference to adjacent pulses. Dispersion thus limits the range of optical fibers before signal regeneration must take place. This theory is based on a fiber geometry called step-index fiber. The refractive index of the core is then the same everywhere in the core. In a graded-index fiber, the refractive index of the core decreases continuously between the axis of the core and the cladding. The dispersion of the graded-index fiber is less than that of the step-index fiber because rays with high angle of incidence will eventually leak out of the core. The index profile can be shaped in such a way that the dispersion of the signal is minimized. The optimum shape is almost a parabola as a function of the distance from the axis. This theory is based on simple geometrical optics. However, electromagnetic wave analysis (Maxwell’s equations) is required in order to understand exactly what is going on in the fiber. The fiber is in fact a cylindrical dielectric waveguide. If the diameter of the fiber is large compared to the wavelength of light, light may propagate in several transversal modes. A transversal mode is a standing wave pattern of minima and maxima of the electromagnetic field over the cross section of the core. Transversal means that the pattern is vertical or transversal to the direction of the propagation of light, which takes place along the fiber. The existence of several transversal modes simultaneously also gives rise to dispersion and signal degradation because the various modes propagate with different speeds. The multimode dispersion of a graded-index fiber is minimized if the refractive index of the core has the optimum near-parabolic shape. If the diameter of the fiber is just a few wavelengths wide, only a single mode is supported. Such a fiber is called single-mode or mono-mode. Electromagnetic wave analysis shows that a considerable fraction of the energy of the electromagnetic wave in a single-mode fiber is in fact carried by the cladding. This is called the evanescent wave. The evanescent wave is used in optical couplers. The core diameter of a single-mode fiber is about 10 µm (µm = micrometer = 10−6m). The diameter of the multimode fiber is usually 50 µm. Optical fibers can be used for wavelengths between about 1,200 nm (nm = −9 nanometer = 10 m) and about 1,650 nm, except in a small region around 1,400 nm, where there is an anomalous maximum. The loss curve is illustrated in Figure 10.6. The usable bandwidth of an optical fiber is formidably 50 THz or 50,000

286

Optical Communication Systems Loss dB/km Usable bandwidth (50 THz)

1

200 nm

200 nm

0.5

1,100

1,400

1,310

Figure 10.6

1,700

Wavelength nm

1,550

Loss in optical fibers.

GHz. The minimum losses at 1,310 nm and 1,550 nm are less than 0.5 dB/km and 0.2 dB/km, respectively. There is also a band around 850 nm that is used for short distance communications. Transmitters and receivers at 850 nm are cheap. The peak at 1,400 nm is caused by hydroxyl ions (OH−), while the main cause of loss is scattering from the irregular patterns of vibrating molecules in the glass (Rayleigh scattering). However, the maximum transmission distance for optical fibers is not limited by loss but by dispersion. The maximum transmission distance for a single-mode fiber is typically between 80 and 140 km. 10.3.2

Splitters, Combiners, and Couplers

Figure 10.7 shows a 1 × 2 splitter, a 2 × 1 combiner, and a 2 × 2 coupler. More generally, we may design N × 1, N × 1, and N × N devices. N = 2 is the most common design. In the N × 1 splitter with identical splitting ratios on all outputs, the power of each output signal is 1/N of the input power. The splitting ratio of a symmetric binary splitter is 1/2 or 50%. The signal on each of the two outputs is then 3 dB smaller than the input power. However, 1 × 2 splitters (and more generally 1 × N

λ1 Splitter

Figure 10.7

λ1

λ1

λ1

λ2

λ1+ λ2 Combiner

Splitter, combiner, and coupler.

λ1

λ1 or λ2

λ2

λ2 or λ1 Coupler

10.3 Optical Transmission Components

287

splitters) with any splitting ratio can be made (e.g., 10% of the power is guided into one branch and 90% into the other branch). In the splitter, the input signal 1, where 1 is the wavelength of the optical signal, appears on both outputs. In the combiner, the input signals with different wavelengths 1 and 2 are combined to form the multiplexed signal 1 + 2. By joining 2 × 1 combiners and 1 × 2 splitters, any number of wavelengths may be multiplexed and demultiplexed. Figure 10.8 shows how five signals can be multiplexed by 2 × 1 combiners. This multiplexing method is called wavelength division multiplexing (WDM). The demultiplexer consists of binary splitters and filters (F) as shown in the figure, where each filter is tuned to the appropriate wavelength. The splitter consists of a junction where the core of one fiber is fused with the core of the two output fibers. This is a rather simple operation. The combiner is a splitter run in the inverse direction. This is possible because the optical junctions are passive devices allowing transmission of light in both directions. Add-drop multiplexers can easily be made as shown in Figure 10.9. The add-drop configuration may be fixed or dynamic. In the latter case, a control signal is required in order to set the multiplexing configuration in each node. The dynamic multiplexer must also contain devices such as couplers, tunable filters, and switching devices. The coupler utilizes the evanescent wave in the fiber. The evanescent wave is the part of the electromagnetic wave carried in the cladding of the fiber, as explained λ1

F λ1 + λ2 + λ3 + λ4 + λ5

λ2

λ1

F

λ2

λ3

F

λ3

λ4

F

λ4

λ5

F

λ5

Multiplexer

Figure 10.8

Transmission

Demultiplexer

Use of combiners and splitters in multiplexing.

Mux

λ1 + λ2 + λ3 +···

Demux

Control signal

λ1 λ2 Drop

Figure 10.9

Add-drop multiplexer.

λ2 λ1 Add

λ1 + λ2 + λ3 +···

288

Optical Communication Systems

earlier. By placing two fibers very close together over a certain length, the evanescent wave of the signal in one fiber sets up a wave in the other fiber. At a characteristic length, complete coupling takes place (i.e., the waves in the two fibers change place). If the length is somewhat shorter or longer than the characteristic coupling length, no coupling takes place. The coupling is thus critically dependent on the wavelength and the distance over which the fibers are close together. If the refractive index of the fiber can be altered by an external electrical potential, a coupler with two states can be designed, as shown in Figure 10.10. Altering the refractive index implies that the effective coupling length is altered. If no voltage is applied to the coupler, the waves 1 and 2 change place (the cross state). If a voltage is applied, the coupling length is altered so much that no coupling takes place (the bar state). The coupler is thus a binary switching element with bar and cross states, as described in Section 6.2.6.1. 10.3.3

Filters

The design of optical filters is based on interferometry. A few examples are shown in Figure 10.11. The Mach-Zehnder (MZ) interferometer is based on the interference between a direct wave and a delayed wave. The interferometer will have a pass-band for frequencies interfering constructively. It is easy to see that this occurs if the delay corresponds to an integer number of wavelengths of the signal. For a fixed delay D, the direct wave sin t and the delayed wave sin (t + D) interfere at the output of the filter producing the signal sin ωt + sin ω(t + D) = 2 sin ω(t + D/ 2 ) cos( ωD/ 2 )

Voltage

No voltage

λ1

λ1

λ1

λ2

λ2

λ2

λ2

λ1

Bar state

Figure 10.10

Electrically controlled coupler.

Cross state

10.3 Optical Transmission Components

289 Delay

Mach-Zehnder interferometer I

T1 R1

T2

R2

T3 Fabry-Perot interferometer λ1 + λ2 + λ3 +···

λ1 λ2 λ3 λ4 Grating

Figure 10.11

Example of filter elements.

This is a new sine wave with amplitude 2cos( D/2). The carrier wave is sin ω(t + D/2), which is identical to the incoming wave except that it is phase-shifted by = D/2. The light intensity from the filter is proportional to the square of the amplitude; that is, the intensity variation as a function of angular frequency and delay is: I = I 0 cos 2 ( ωD / 2 )

where I0 is the intensity for D = 0. This signal has a maximum equal to I = I0 for ωD / 2 = 2nπ, where n is an integer, and a minimum equal to I = 0 for ωD / 2 = (2n + 1)π. The angular frequency of a signal with wavelength is ω = 2πc / λ,where c is the speed of light in vacuum. Inserting this expressing for we find that a signal with wavelength λ = cD / 2n will pass unperturbed through the filter, while a signal with wavelength λ = cD / (2n + 1) is stopped by the filter. By selecting an appropriate value for the delay D and n, the filter will let through a signal with a given wavelength and suppress other wavelengths. The filter characteristic of the MZ filter is shown in Figure 10.12. The filter has a pass-band for every λ = cD / 2n for integer n. This gives rise to the filter characteristic of repetitive pass-bands shown in the figure. Spurious bands can be removed by another MZ filter. Filters with steeper flanks and narrower pass-bands are produced by placing several MZ filters with slightly different delays in tandem.

290

Optical Communication Systems

Amplitude

3-dB bandwidth

....

.... λ Desired wavelength

Spurious bands removed by another MZ filter

Figure 10.12

Filter characteristic of the MZ filter.

The Fabry-Perot (FP) interferometer consists of two partially transparent plates placed a distance from each other. The space between the plates can be filled with air or a transparent dielectric medium (e.g., glass). The wave may then pass directly through the interferometer or be reflected back and forth between the plates several times. This gives rise to constructive interference if twice the electrical distance between the plates is an integer multiple of the wavelengths. The light may be perpendicular to the plates of the interferometer. The angle of incidence is oblique in the figure in order to illustrate the principle. The filter characteristic of the FP interferometer is derived in a similar way as for the MZ filter. However, since the light is reflected back and forth between the plates several times, the calculation of the filter characteristic is complex but not difficult. The filter characteristic is similar to that of the MZ interferometer, though a little more complex since it depends on two parameters (the distance between the plates and the transmittance of the plates) rather than one (the delay in the MZ interferometer). Mach-Zehnder and Fabry-Perot interferometers can be made from materials where the refractive index is altered by applying an electrical field to the interferometer. Such filters are tunable, since the pass-band can be adjusted by altering the strength of the applied electrical field. The most common optical grating is a flat sheet of glass with parallel grooves etched into it. This is the reflective grating. There are also transmissive gratings. Waves reflected by the grating (or is passed through it) interfere with one another in such a way that only waves of one wavelength will interfere constructively in a given direction, as shown in Figure 10.11. All other wavelengths will interfere destructively in that direction. An optical fiber placed at a given angle will pick up signals with that particular wavelength as shown. 10.3.4

Lasers

Laser is an acronym for light amplification by stimulated emission of radiation. Einstein predicted from pure thermodynamic arguments in 1917 that when light passes through a transparent medium, two components of radiation may result: one

10.3 Optical Transmission Components

291

spontaneous component with a wavelength uncorrelated to the wavelength of the input signal and one stimulated component with a wavelength and phase identical to those of the input signal. The principle was verified experimentally early, but the stimulated component was always so weak that it had not practical applications. In 1958, Charles Towns discovered materials where the stimulated radiation was significant. The physics of the laser cannot be fully comprehended without the aid of quantum mechanics. In popular terms, the laser can be described as follows. The energy levels of an atom are given by the energy states (or orbits) of the electrons. The ground state of an atom is the state where all the electrons are in their lowest possible energy state subject to Pauli’s exclusion principle.1 In an excited state of the atom, the energy of one or more electrons is at a higher energy level than the ground state. If the electron is in an excited state with energy E1, it will eventually move to a lower state with energy E2 and transmit a photon with angular frequency proportional to the energy difference between the two states: hω = E1 − E 2 , where is the angular frequency of the photon and h (pronounced h-stroke) is Planck’s constant divided by 2 . Similarly, the electron moves from the lower state E2 to the higher state E1 by absorbing a photon with the angular frequency corresponding to the energy difference. The electrons in the laser are raised to an excited state by an external energy source (either light or electrical current). In most substances, the excited state is unstable, and the electrons return to the lower state quickly without producing significant amounts of stimulated emission. In particular substances, however, the excited state lives for a long time. Such excited states are called quasi-stable. Provided enough excitation energy, the number of electrons in the quasi-stable state can be higher than that in the ground state. This effect is called population inversion. The laser is composed as shown in Figure 10.13. The device consists of a quasi-stable medium (or lasing medium) placed between two mirrors, where one of the mirrors is partly transparent (or semitransparent) and an external excitation device provides enough energy to cause population inversion. When a photon with an angular frequency equal to the energy of the excited state interacts with an excited electron, the electron returns to the ground state and

Lasing medium Mirror

Semitransparent mirror Excitation device

Figure 10.13

1.

Laser.

Pauli’s exclusion principle states that two electrons with exactly the same set of quantum numbers (e.g., spin, angular momentum) cannot occupy the same energy level. This gives rise to the particular energy level structure of atoms and to the periodic system itself. See any introductory book on quantum mechanics.

292

Optical Communication Systems

releases a photon with exactly the same direction of motion, angular frequency and phase as the interacting photon. This phenomenon is called stimulated emission. The new photon and the original photon create more photons that create more photons and so on, as the photons are reflected back and forth between the mirrors. Some of the light is transmitted through the partly transparent mirror. If the excitation energy is kept constant, the light output from the laser will quickly reach a stable intensity. The light produced by the laser is called coherent since all the light waves have the same frequency and phase. The coherency of the light emitted from a laser is the particular characteristic of the laser that makes it useful in so many applications, including optical transmission in fibers. The space between the mirrors is a resonant cavity. The length of this cavity must be an integer number of wavelengths in order to create constructive interference and build up the intensity of the light. The laser will then only produce light at a single frequency (or more accurately, within a narrow bandwidth around this frequency) determined by the length of the cavity. By varying the length of the cavity, it is possible to adjust the wavelength of the light within a narrow bandwidth. The lasers applied in telecommunications are made from p-n semiconductors (laser diodes). The frequency of the emitted light then corresponds to the energy difference between electrons in the n region and holes in the p region. The excitation of the junction takes place in the usual way by applying a voltage across the junction moving electrons from the p region to the n region. The mirrors of the laser cavity are fixed to the ends of the p-n junction. Tunable diode lasers are of particular importance in WDM because a single type of laser can then be used for all the sources. The individual channel is selected by tuning the laser to the particular frequency of that channel. 10.3.5

Modulation

The simplest modulation technique for optical systems is on-off keying of the signal. The signal then consists of light pulses corresponding to, say, binary 1 and darkness corresponding to binary 0. The modulation method is called intensity modulation. Bit rates of more than 20 Gbps can be obtained in this way. Modulation may take place by turning the laser on an off at the required bit rate. This method is not good for bit rates much in excess of 2 Gbps. Fast modulators can be designed by placing a tunable Mach-Zehnder interferometer (see Figure 10.11) in the fiber following the laser. Modulation takes place by turning the light on and off by applying the binary signal to be modulated to one of the branches of the interferometer. If no voltage is applied, light is admitted through the interferometer; if the voltage is applied, the refractive index over a certain length of the fiber is changed so that the interference is destructive and no light is transmitted. The method is illustrated in Figure 10.14. The laser and the interferometer can be integrated in a single device.

10.3 Optical Transmission Components

293

Modulation signal

Light pulses Laser Interferometer

Figure 10.14

10.3.6

Modulator.

Detectors

The simplest demodulators consist of photo-detectors demodulating the optical signal directly. Photo-detectors used in optical networks are p-n photodiodes where the received light creates electron-hole pairs both in the p layer and in the n layer. The electrons created in the p layer will cross over to the n layer, and the holes created in the n layer will cross over to the p layer. This sets up an electrical current across the junction, thus recreating the modulation signal. This current is then amplified by electronic amplifiers and provided to the electronic circuits interpreting the signal. 10.3.7

Amplifiers

Optical signals can be amplified by first demodulating the signal, regenerating it, and finally remodulating the regenerated signal on a stronger carrier for further transmission along the fiber. However, direct amplification of the optical signal is also possible. Direct amplification of an optical signal may take place in several ways. The most common amplifier used on optical fibers is the erbium doped fiber amplifier (EDFA). The EDFA can amplify signals with wavelengths in the range 1,525 nm to 1,560 nm. The wavelength of minimum damping is 1,550 nm so that the EDFA is particularly suitable for direct amplification of light along the fiber. Praseodymium-doped fluoride fiber amplifiers are used in a similar way in the band between 1,280 nm and 1,330 nm. Praseodymium (Pr) is a rare Earth element with atomic number 59. Erbium (Er) is also a rare Earth element with atomic number 68. Rare Earth elements have a particularly rich spectrum of energy levels because of a rather unusual structure of the outer electron orbits. The EDFA amplifier is shown in Figure 10.15. The amplifier consists of an erbium doped section of the fiber (a few meters). The pump laser emits a strong signal with wavelength of 980 nm or 1,480 nm into the fiber. The pump signal excites the erbium doped section such that the incoming optical signal is amplified by stimulating the erbium atoms to emit photons of the same wavelength. The photons injected by the pump will suffer considerable attenuation so that the pump signal dies off after a few kilometers of fiber length. The optical semiconductor amplifier is a modified semiconductor laser. The modified laser may contain mirrors with a reflectivity that keeps the laser below the

294

Optical Communication Systems Erbium-doped fiber

Pump laser

Figure 10.15

Amplifier.

lasing threshold or only allows a single passage of the signal through the active region. The weak input signal causes stimulated emission in the active region of the laser, thereby amplifying the signal. The gain of these amplifiers is about 25 dB. With a loss per kilometer of less than 0.2 dB, this allows a distance between amplifiers of more than 130 km. The semiconductor amplifier is small and can be integrated with other components. The semiconductor amplifier is used in optical switches. 10.3.8

Wavelength Converters

A wavelength converter is a device where the wavelength of the input signal is converted to another wavelength at the output. This can be done simply by optoelectronic means (i.e., the duty signal is first converted to an electrical signal by a photo detector and then remodulated on an optical signal with a different wavelength). Such converters can be designed as compact integrated circuits. Wavelength conversion can also be done directly on the optical signal. One technique is to use nonlinear mixing where the new frequency is created as follows: f out = f in + f1 − f 2

where f1 and f2 are pure carrier waves with frequency separation equal to the frequency offset between the input and the output signals. This third-order nonlinearity is achieved in silica fibers. Since all combinations (e.g., − fin + f1 − f2) of the three waves will exist simultaneously in the mixer, the correct mixing product must be selected by an optical filter. Second-order mixing is also possible: f out = f − f in

where f is a laser frequency approximately twice that of the input frequency. Other possibilities are to use semiconductor optical amplifiers where the modulated input signal and a pure carrier signal generated by a local laser are applied at the input of the device. The frequency of the pure carrier is the same as the wanted frequency of the output signal. The intensity-modulated input signal causes the gain of the amplifier to vary in the same way as the duty signal. The pure carrier signal will then be intensity modulated by the amplifier and thus produce an intensity

10.4 Optical Switching

295

modulated output signal identical to the input signal but at the new carrier frequency. The desired output signal is then extracted by a filter at the output of the device. The device is shown in Figure 10.16. Wavelength converters using semiconductor optical amplifiers can be designed in several ways offering different tradeoffs between complexity, bandwidth, energy efficiency, and signal degradation. Much research takes place in this area because wavelength conversion is essential for optic packet switching.

10.4

Optical Switching 10.4.1

Switching Devices

We saw earlier that it is possible to construct an electrically controlled binary switching element from the optical coupler (Figure 10.10). Similarly, we may construct a binary switch using electrically controlled amplifiers as shown in Figure 10.17. Electrical signals are applied to the amplifiers in order to open or close a path through the switch. In this way, bar states and cross states are obtained in the same way as in the coupler switch. However, one advantage of the amplifier switch is that if the signals on the two inputs have different wavelengths, the same signal may appear on both outputs, providing multicall and broadcast capabilities. Several binary switching elements may be combined in order to build larger switches. The principle is the same as described in Sections 6.2.6.2 and 6.2.6.3 for Clos-Benes and Banyan networks.

λin Laser

Figure 10.16

λout

Semiconductor optical amplifier

F

λout

Wavelength converter.

Amplifier I1

O1 Amplifier

Amplifier I2

O2 Amplifier

Electrical control

Figure 10.17

Optical amplifier 2 × 2 switch.

296

Optical Communication Systems

Note that each binary switching element in the optical amplifier switch introduces a 3-dB signal loss because of the splitter at the input. This constrains the size of switches constructed from these switching elements. A rather exotic type of switch is constructed from tiny mirrors smaller than one micrometer. The mirrors can be rotated by electrical control signals, as illustrated in Figure 10.18. The mirror assembly can be designed as a microelectromechanical system (MEMS). where the mirrors and the mirror adjustment control are integrated in a single miniature device. The mirrors are rotated independently by the electrical switching control subsystem in order to obtain all possible paths through the switch. By rotating the mirrors into particular positions, the input signals appear on the desired output port as illustrated. A switch based on tuneble wavelength converters (TWCs) and optical transmission grating is shown in Figure 10.19. The signal input to the grating will leave the grating at an angle determined by the frequency of the signal. The output fibers are located at the points where the particular frequencies are picked up. The switching control device provides an electrical tuning signal to the TWCs. The frequency of the input signal is thus converted to a frequency such that it appears at the desired output as illustrated in the figure. The conversion and switching take place without altering the intensity modulation of the signal. All switching techniques described earlier are suitable for circuit switching. Some of them are too slow for packet switching. Let us next describe one promising approach toward optical packet switching. 10.4.2

Packet Switching

In a circuit switched network, the switched path is established by a signaling system before any information is sent. The signaling information is interpreted by all exchanges that shall take part in the setup of the connection. Optical transmission takes place after the end-to-end connection has been established.

MEMS 1 2 3

2 3 1

Electrical switching control

Figure 10.18

Optical switch using rotatable mirrors.

10.4 Optical Switching

297

Tunable wavelength converters 1

TWC TWC

2

TWC TWC

3

TWC TWC

4

TWC TWC

4 Optical grating 2

Input fiber

1

Electrical signals from switching control device

Figure 10.19

3 Output fiber

Switch using TWCs and optical grating.

In packet switching, all information an exchange or router required for forwarding the call is contained in each individual packet. Therefore, the address and other packet header information the router requires must be read in real time directly from the packet. Optical packet switching has not yet been realized. The European research project STOLAS has come up with a solution to this problem that may turn out to be economically and technically feasible. A brief description of the switching method proposed by STOLAS is given next. The switching process proposed in the STOLAS system is the TWC and grating switch shown in Figure 10.19. The tuning of the wavelength converter is determined from information contained in the packet called the label. The label is modulated on the packet using frequency shift keying at a bit rate of, say, 155 Mbps, while the duty signal is intensity modulated at a much higher bit rate (say, 10 Gbps). The label is inserted at the front of the packet and occupies only a small part of it. The duty signal is present in the whole packet, including the part containing the label. The switch is shown in Figure 10.20. The details of only one input channel are shown. The signal management takes place in a unit called an optical label swapper (OLS). There is one such unit for each channel. The OLS operates as follows. The optical signal is first divided into two parts by an optical splitter where, say, only 10% of the power is sent to the FSK demodulator. The FSK demodulator picks out the label and provides it to the label manager. The label manager determines from the label on which output path the signal shall appear. The manager also constructs a new label that will be consumed by the next switch. The manager then provides the tuning voltage plus the new label to the tunable laser. The label is encoded as a binary amplitude variation on the tuning voltage. The tuning voltage then adjusts the carrier frequency of the tunable laser, and the amplitude variations corresponding to the label appear as an FSK modulated signal on the carrier. Some time is required by the OLS to read the label, calculate the switching path, and construct the new label. The duty signal is delayed for this period of time before it is presented to the semiconductor amplifier.

298

Optical Communication Systems

Label

Optical label swapper (OLS) FSK demodulator

Label management Tuning voltage and new label

Duty signal

Tunable laser

Semiconductor optical amplifier

Delay OLS OLS

Optical grating

OLS OLS OLS Delay Delay

Figure 10.20

STOLAS packet switch.

The semiconductor optical amplifier then converts the frequency as explained earlier and passes the signal to the optical grating. When a new packet arrives, the desired output of the switch may be busy transmitting another packet. This type of contention can be resolved by forwarding the new packet to a recirculation path as shown in the figure. The package is then delayed before it is again presented to the switch. If the desired output is no longer busy, the packet is switched and forwarded in the normal way. If the output is still busy, the packet may be recirculated once more. Packets are only lost if all recirculation paths are busy when the packet is presented to the OLS or if the packet has been recirculated a maximum number of times. The principle is similar to the Batcher-Banyan network shown in Figure 6.23.

Appendix: Loop Mathematics and Loop Components A.1 Loop Mathematics The phase-locked loop is shown in Figure A.1. It consists of a phase detector where the phases of the input and the output signals are compared. This comparison produces an error signal, e(t) (t is time), which is filtered by the loop filter producing a control signal, c(t), that drives the voltage controlled oscillator (VCO). The frequency of the VCO is (ideally) proportional to the control voltage. The VCO contains a resonance circuit where the capacitor is made from semiconductor material that produces a capacitance proportional to the voltage applied to the semiconductor. The output signal from the VCO has the same nominal frequency as the input signal. Let us make the following assumptions in order to explain how the loop works. The input signal is x (t ) = A x cos( ωt + φ1 (t ))

The output signal of the VCO is Input signal x(t)

Phase detector

Error signal e(t)

Loop filter f(t)

Control signal c(t)

VCO

Output signal y(t)

R1

C

τ1 = (R1 + R2)C τ2 = R2C F(s) = (1 + sτ2)/(1 + sτ1)

R2

Second-order loop filter

Figure A.1

Phase-locked loop.

299

300

Appendix: Loop Mathematics and Loop Components

y(t ) = A y cos( ωt + φ 2 (t ))

where ω is a constant frequency (carrier frequency or bit rate). The amplitudes (Ax and Ay) are constant (they can always be made so by amplifier and limiter circuits) so that they can be ignored in the following calculation. The slow temporal variations of the signals are included in the phases i (i =1 or 2). This includes a possible small frequency difference, ∆ , between the input and the output signals producing a phase difference ∆ t. Frequency instability is thus included in the phase difference as a linear function of time. Let us then derive an equation for the dynamic performance of the loop. The change in the output phase of the VCO is proportional to the control signal (constant of proportionality is K1). K1 is called the gain of the VCO: dφ 2 (t ) / dt = K1 c(t )

The phase detector produces a nonlinear error signal of the form: e(t ) = W ( x (t ), y(t )) = W ∗ (φ1 (t ) − φ 2 (t )) = W ∗ (φ(t ))

where φ = φ 1 − φ 2 is the phase difference between the input signal and the VCO signal. The phase detector includes filters that remove frequency components proportional to ω and multiples of ω, leaving only the signal that varies slowly with the phase (the “dc” component). The function W* can have various shapes such as saw tooth, triangular, or sinusoidal. If W* is a sine function,1 then the error signal becomes: e(t ) = K 2 sin φ(t )

where K2 is a constant given by the characteristics of the phase detector. The constant is called the gain of the phase detector since it is the constant of proportionality between the error voltage and the phase difference for small phase differences (sin ≈ for small ). The error signal is fed into the input port of the loop filter shown in Figure A.1. The loop filter will then modify this signal as follows: c(t ) = e(t )∗ f (t ) =

[(1 + τ s) / (1 + τ s)]e(t ) 2

1

where f(t) is the impulse response of the filter, (1 + τ 2 s ) / (1 + τ1 s ) is the Laplace transform of f(t), ∗ is the convolution operator, and s is the differential operator d/dt. Inserting the expressions for e(t) and c(t) we get: dφ 2 / dt = K1 c(t ) = K1

[(1 + τ s) / (1 + τ s)e(t )] = K K [(1 + τ s) / (1 + τ s)] sin φ 2

1

1

2

2

1

Multiplying both sides by 1 + τ1s, setting s = d/dt and K = K1K2 (the overall loop gain) gives: 1.

Since all the derivatives of the sine function are continuous, it is simpler to derive closed form expressions for loops having this phase detector shape. The sinusoidal shape of the W* function occurs in practical systems if the phase detector is a multiplier circuit (ring modulator).

A.1

Loop Mathematics

301

(1 + τ1 d / dt )dφ / dt

= K(1 + τ 2 d / dt ) sin φ

Finally, setting φ 2 = φ 1 − φ, differentiating and ordering the terms gives: τ1 d 2 φ / dt 2 + (1 + Kτ 2 cos φ)dφ / dt + k sin φ = dφ1 / dt + τ1 d 2 φ1 / dt 2

(A.1)

This is an inhomogeneous nonlinear second order ordinary differential equation, where the inhomogeneous term is given by dφ 1 / dt + τ1 d 2 φ 1 / dt 2 . This equation allows us to study the performance of the loop with initial constant frequency offset ( (dφ 1 / dt ) = ∆ω and d 2 φ 1 / dt 2 = 0) and when 1 is an arbitrary function of time (e.g., a phase or frequency modulated signal). We shall not study how the loop reacts to modulation since this theory is rather complex but look only at the acquisition performance of the loop. Equation (A.1) is an initial value problem in the sense that when the loop is turned on, the loop will start operating from an arbitrary initial phase and VCO frequency. The total solution of (A.1) will then consist of two parts: one general solution satisfying the initial conditions derived from the homogeneous equation by setting the left-hand side of (A.1) to zero, and one particular solution satisfying the inhomogeneous term to the right. That is, φ(t ) = φ hom (t ) + φ inhom (t ). If the initial conditions are within a range where the loop will eventually acquire lock, the homogeneous solution φ hom (t ) will be a transient that eventually will die out (with relaxation time depending upon the damping of the loop), and we are left with only the inhomogeneous solution as time approaches infinity. The total solution of (A.1) as time approaches infinity is independent of the homogeneous solution and is given completely by the homogeneous solution. That is, for large t, φ(t ) = φ inhom (t ). The inhomogeneous solution of (A.1) with constant frequency offset φ 1 = ∆ωt is then simply given by K sin φ = dφ1 / dt + τ1 d 2 φ1 / dt 2 = ∆ω or φ = constant = arcsin( ∆ω / K )

Since the maximum value of sin = 1, we conclude that it is possible to keep the loop locked for frequency offsets relative to the free-running frequency of the VCO smaller than ∆ = K. This means that if the loop is already locked, we way increase the frequency deviation at the input slowly toward this limit without losing lock. What we have found is thus the condition that the loop will stay synchronized once it has been locked. This limit is called the synchronization range of the loop. Note that this condition holds for sinusoidal phase detectors. In a similar way it can be shown that the triangular phase detector has a synchronization range that is /2 times larger. Since d / dt = ( dφ / dt )( d / dφ)

and thus

[

]

d 2 φ / dt 2 = ( dφ / dt ) d ( dφ / dt ) / dφ

302

Appendix: Loop Mathematics and Loop Components

the second-order differential equation can be converted to a first-order nonlinear, inhomogeneous differential equation in the dependent variable (d /dt) and the independent variable :

[

]

τ1 ( dφ / dt ) d ( dφ / dt ) / dφ = dφ1 / dt + τ1 d φ1 / dt 2

2

− K sin φ − dφ / dt − Kτ 2 ( dφ / dt ) cos φ

(A.2)

The solution d /dt as a function of is called a first-integral of the original equation, since the original problem is reduced to solving two first-order differential equations—first finding d /dt from (A.2) and then solving this differential equation for (t). The first-integral provides us with a tool to study the acquisition process in details. Figure A.2 illustrates the region of the phase plane (the ( , d /dt) plot) in which the VCO will be stably locked to the input frequency, and the regions where the loop does not acquire acquisition. The dynamics of the loop can be depicted as a vector field as shown. The vector field within the stable region spirals toward the stable point of the loop. The vector field outside this region never enters the stable region. For large frequency offsets, the frequency of the VCO (i.e., the derivative of the phase) will oscillate regularly between a lower an upper frequency. If ( , d /dt) is within the stable region, both and d /dt tend to a stable point as t → ∞. This is the locking condition for the loop. If ( , d /dt) is outside this region, the loop never acquires lock. = t, we can study the acquisition performance of the loop for a If we set fixed initial frequency offset ∆ω. It is found that the maximum frequency deviation where the loop may still acquire lock is given approximately as (this is not a simple derivation and is for that reason omitted here): ∆ω acq = 2( K / τ1 )(1 + Kτ 2 ) = 2 Kζω n

(A.3)

The time it takes to achieve lock for a given frequency deviation ∆ is found to be dφ/dt

No acquisition

Acquisition φ –π +π

Figure A.2

Region of stability of the loop.

A.1

Loop Mathematics

303

[

]

Tacq = ∆ω 2 τ12 / K(1 + Kτ 2 ) = 2 ∆ω 2 τ1 / ∆ω acq

The natural frequency of the loop, follows:

n

(A.4)

, and the damping, , are defined as

2 ω n ζ = (1 + Kτ 2 ) / τ1 and ω n2 = K / τ1

(A.5)

We saw earlier that the synchronization range is ∆ω syn = K. This is the maximum deviation the signal may have before the VCO loses lock. The synchronization range is considerably larger than the acquisition range since K >> ω n and ζ ~ 1 in practical loops. We may study the performance of the loop for small by linearizing the loop equation setting sin φ = φ and cos φ = 1. Equation (A.1) then reduces to a linear differential equation: τ1 d 2 φ / dt 2 + (1 + Kτ 2 )dφ / dt + Kφ = dφ1 / dt + τ1 d 2 φ1 / dt 2

(A.6)

We see from (A.6) that the stationary solution is dφ / dt = 0 and φ = ∆ω / K for a constant frequency offset ∆ between the input frequency and the free-running frequency of the VCO; that is, the loop locks perfectly to the input frequency (random noise is not included in this calculation), but there is a phase error proportional to the initial frequency offset. The phase error can be made small choosing high loop gain K. Again introducing the operator s = d/dt and φ 2 = φ 1 − φ the equation can be written: φ 2 = φ1

[(2ω ζ − ω n

2 n

)

/ K s + ω n2

]/ (s

2

+ 2 ω n ζs + ω n2

) = H( s)φ

1

(A.7)

where and n are loop parameters defined in (A.5), H(s) is the transfer function of the loop, and 1 and 2 are the input and output phases, respectively. If the feedback loop is removed, the transfer function of the resulting circuit is called the open loop transfer function G(s):

[

]

G( s) = G( jω) = K(1 + jωτ 2 ) / jω(1 + jωτ1 )

where s is represented in the usual way as the imaginary variable j where is the angular frequency. We see that the phase is negative and purely imaginary for ω → 0 and ω → ∞; that is, the phase angle is − /2 for these values. The phase response of the loop is

(

tgθ = Im( θ ) / Re( θ ) = − 1 + ω 2 τ1 τ 2

) / ω( τ

1

− τ2 )

Since we always have that (see Figure A.1) τ1 > τ 2 , this expression is always less than zero. The minimum phase angle is then:  2 τ1 τ 2   θ = arctan  −   τ1 − τ 2 

(A.8)

304

Appendix: Loop Mathematics and Loop Components

Since the feedback of the output signal is negative (the phase detector produces the difference between output signal and input signal), the phase angle in the forward open loop must not be less than − or greater than as long as the loop gain is larger than 1, since otherwise the feedback becomes positive and the loop becomes unstable. In practical loops this condition is satisfied with good margin. It can be shown that it is easier to design stable loops using the loop filter in Figure A.1 than with all other shapes of the loop filter. The loop parameters of a practical loop can be calculated in the following way. If the loop is used to reduce the frequency fluctuations of a signal, then ∆ syn must be larger than the maximum frequency instability of the input signal. This determines K. Then it is usual to require that the loop is critically damped as in most control systems (i.e., ζ = 1 / 2). This gives a relation between 1 and 2. Then we may consider reasonable values for ∆ acq, giving an additional equation for 1 and 2. The problem is that we have to find values for K, 1, and 2 that are realizable. Therefore, the calculations are usually repeated iteratively until a reasonable compromise has been found. The loop is also an efficient noise filter. The reason is that the natural frequency of the loop is small. The noise bandwidth of the loop (in hertz) is approximately

(

B n = (2 ω n / 4ζ ) 1 + 4ζ 2

)

which, for a critically damped loop (ζ = 1 / 2), gives approximately Bn = 106 . ωn. This is almost equal to the minimum value of the noise bandwidth Bn = ω n at ζ = 1 / 2. In most practical loops, n is very small, at most a few hundred hertz. Therefore, the noise bandwidth of the loop is also small. This conclusion holds if the input noise or VCO jitter is so low that the small-phase approximation holds (sin = ). It is very difficult to determine the noise performance of the loop if this condition does not hold (i.e., if the loop is embedded in a very noisy environment). In such cases, the loop may lose lock for certain periods of time and then resynchronize. Such events are called cycle slips. One estimate for the time Ts between cycle slips is

[

(

Ts = ( π / 2 B n ) exp 2( S / N ) sig B sig / B n

)]

where (S/N)sig is the signal-to-noise ratio of the input signal, Bsig is the bandwidth of the input signal, and Bnis noise bandwidth of the loop.

A.2 Loop Components Several designs of phase detectors exist. Only three of them will be described here because they represent three detector characteristics: a sine function, a saw tooth function, and a triangular function. A balanced analog phase detector is shown in Figure A.3. It consists of a ring of four diodes pointing in the same direction. For this reason the circuit is called a ring modulator. One voltage (e.g., the output signal from the VCO—the result is the

Loop Components

305

Error voltage

D1’

D1 From input

A.2

D 2’

D2

From VCO

Figure A.3

Analog phase detector.

same if we use the input signal) is fed into the midpoint of the transformers as shown. For simplicity, let us assume that the voltage from the VCO is a square signal with amplitude larger than that of the input signal. We then see that if the VCO voltage is positive and larger than the input voltage, diodes D1′ and D2 will conduct and feed the input signal to the output. If the VCO voltage is negative with absolute value larger than the input value, diodes D1 and D 2′ will conduct and feed the opposite polarity of the input signal to the output. The sequence is shown in Figure A.4. The transitions between negative and positive values will then occur at the phase difference between the signals as shown in the figure. The dc voltage after filtering out the higher-order frequency components is proportional to cos , where is the phase difference between the VCO voltage and the input voltage. We see this from Figure A.4 in the following way for the case where the input signal and the signal from the VCO have acquired the same frequency and differ only by the phase angle . If the frequencies are not the same, it is more difficult to see that the same equation hold because then is a function of time.

From VCO Solid line: error voltage

Positive sign φ = φ2 – φ1

From input Negative sign

Figure A.4

Error voltage.

306

Appendix: Loop Mathematics and Loop Components

If the amplitude of the input signal is A, the dc voltage produced by device is found by integrating the signal along a single period of the signal, observing that the error voltage is: • • •

Proportional to the negative of the input signal in the time interval 0 to / ; Proportional to the input signal in the time interval φ / ω to φ / ω + π / ω; Proportional to the negative of the input signal in the time interval φ / ω + π / ω to 2π / ω.

This gives for the error voltage: e=

ω  − 2 π 

φ /ω

∫ A sin ωt dτ + 0

φ /ω+ π /ω

∫

A sin ωt dτ −

φ /ω

 2A 2A  π  A sin ωt dτ = cos φ = sin  − φ   π π 2  φ /ω+ π /ω 2 π /ω

∫

Note that this circuit can be modeled as an analog multiplier, where the input signal and the VCO signal are multiplied with one another and then the high-frequency components are filtered away. If one signal is sin t and the other signal is sin(ωt + φ ), then the resulting signal is 2 sin ω1t sin(ω2t + φ) = cos((ω2 – ω1)t + φ) – cos((ω2 + ω1)t + φ). After filtering, the signal cos((ω2 – ω1)t + φ) = cos(∆ωt + φ) remains. Two digital phase detectors are shown in Figure A.5. One has a phase characteristic in the form of a saw tooth; the other has the shape of a triangular wave. This is easily seen from the figures. In both cases, the input signal and the VCO signal are binary signals. The first phase detector consists of a flip-flop where the input signal is provided to the port that forces the flip-flop to enter the 1-state, provided that the flip-flop Error voltage

φ 0

2π

Input signal

VCO signal

∫

Flip-flop

φ

Integrator

φ

Error voltage

φ 0

Input signal VCO signal

XOR

φ

Figure A.5

Digital phase detectors.

∫

2π

φ

A.3

Acquisition Devices

307 Table A.1

Logic States of XOR Gate

Input Signal

VCO Signal

Resulting Signal

1

1

1

1

0

0

0

1

0

0

0

1

was in the 0-state when a binary 1-pulse appears on the input signal. If the flip-flop is in the 1-state when a binary 1-pulse appears on the VCO signal line, the flip-flop enters the 0-state. After integration, the resulting error voltage-to-phase characteristic is a saw tooth signal. The phase detector in the other circuit is an XOR gate. The logic state table of this gate is in Table A.1. It is easy to see that this gives rise to the triangular error voltage-to-phase characteristic shown in the figure. Figure A.6 shows two examples of voltage controlled oscillators. An ordinary oscillator is in principle an amplifier where the output signal is fed back with positive sign to the input of the amplifier via a resonance circuit. The resonance circuit ensures that the feedback is positive only for the resonance frequency of the resonance circuit so that the circuit can maintain oscillation at this frequency. A simple resonance circuit consists of an inductor and a capacitor either in series (zero impedance at resonance) or in parallel (infinite impedance at resonance). The frequency of the oscillator is made variable by applying a varactor diode (sometimes called varicap diode). This diode is equivalent to a capacitor with capacitance roughly proportional to the voltage across the diode. Applying the control voltage across the diode, the VCO will oscillate at a frequency proportional to the control voltage. If we need a more stable (or less noisy) oscillator, we may insert a quartz crystal in the resonance path together with the varactor. Such oscillators are called VCXO, where X stands for crystal abbreviated xtal. The crystal oscillates at a very stable frequency but its frequency range can be stretched somewhat by the varactor.

A.3 Acquisition Devices We saw that the synchronization range was much larger than the acquisition range of the loop. This means that the frequency difference between the free-running VCO and the input signal must be small in order to acquire synchronism. On the other hand, the VCO tolerates a large frequency drift after the loop has acquired synchronism. Therefore, it is often hard to acquire synchronism but easy to preserve it. Figure A.7 shows a circuit that allows acquisition over a large frequency band. The sweep generator produces a triangular signal that sweeps the output frequency of the free-running VCO slowly over the entire synchronization range. If the frequencies of the VCO and the input signal differ enough, the error signal will fluctuate rapidly and the lowpass filter produces a low voltage. If the two frequencies are close to each other, the fluctuations are slow and the lowpass filter produces a

308

Appendix: Loop Mathematics and Loop Components Control voltage

L

C

Analog oscillator Varactor diode

Amplifier VCO signal

Control voltage

L

C

Analog oscillator with stable source (crystal)

Crystal

Amplifier VCO signal

Figure A.6

Voltage controlled oscillators.

voltage that is recognized by the decision circuit so that the switch closes the loop for normal operation. It is also possible to switch between two different loop filters where one is a broadband filter used for initial acquisition and the other is narrowband filter reducing the fluctuations of the output signal when acquisition has been completed. While

Input signal

Error signal

Phase detector

Control signal

Output signal

Loop filter

Lowpass filter

VCO

Decision circuit

Sweep generator

Figure A.7

Acquisition circuit.

A.4

Numerical Example: Satellite System

309

the sweep method can be used for all signal-to-noise ratios, varying the loop bandwidth requires good signal-to-noise ratio in order to achieve initial acquisition. The acquisition device is useful in cases where the Doppler shift of the frequency is large or where the stability of the VCO is poor, such as for oscillators in the GHz band. Viterbi found that the maximum sweep rate is approximately given by R = ω 2n / 4π Hz /s .

A.4 Numerical Example: Satellite System Let us take the TDMA satellite system of Section 3.7.4 as an example. The bit rate is 100 Mbps and we saw that the loop bandwidth ( n/2 ) must be smaller than 1 Hz in order to keep the loop stable. This gives ω n = 2π s −1 . Because of the small guard space between bursts, we found that the short-term clock accuracy should be better than 10−8 or 1 Hz. If we suppose that the long-term clock accuracy due to Doppler shift and other fluctuations is smaller than 100 Hz, the loop can have a synchronization range as small as 100 Hz. This gives immediately K = 2π × 100 s −1 for a critically damped loop (ζ = 1 / 2). Equation (A.3) then gives: 1 = 16 seconds and 2 = 0.23 second. Choosing C = 1 F, we find that R1 = 16 MΩ and R2 = 230 kΩ. This filter is realizable. The acquisition range of this loop is approximately 17 Hz (i.e., 17 times wider than the loop bandwidth). With an initial frequency offset of 10 Hz, the acquisition time of this loop is 11 seconds. If we had made the synchronization range ten times larger (1,000 Hz), we would have another realizable loop with C = 10 F, R1 = 16 MΩ, and R2 = 23 kΩ. The acquisition range is only 54 Hz. The reason that the acquisition range increases so slowly is that, for the same loop bandwidth, the acquisition range increases as the square root of the loop gain. Therefore, acquisition becomes more and more difficult as the frequency instability (synchronization range) increases. It is even difficult to use a sweep device. We saw that the maximum sweep rate is given by R = ω 2n / 4π Hz / s . For a loop bandwidth of 1 Hz, the maximum sweep bandwidth becomes R = 3.14 Hz/s. It takes then more than 600 seconds to sweep over the entire synchronization range. This illustrates one particular problem with very narrow loops.

Acronyms 3GPP AAL ACM ACM ADSL AES AGC AMPS ANSI AP APSK ARQ ASN.1 ASP ATM AU AUC AUG BER BGAN BIB BS BSC BSN BSS BSSMAP BTS BTSM C CAMEL CDMA CD-ROM CEP

Third Generation Partnership Project ATM adaptation layer Association of Computer Manufacturers Address complete message Asymmetric digital subscriber line Aeronautic Earth station Automatic gain control Automatic mobile phone system American National Standards Institute Application part Amplitude phase shift keying Automatic repeat request Abstract Syntax Notation No. 1 Application service provider Asynchronous transfer mode Administrative unit Authentication center Administrative unit group Bit error rate Broadband global area network Backward indication bit Base station Base station controller Backward sequence number Base station subsystem BSS management part Base transceiver station BTS management Container Customer applications for mobile network enhanced logic Code division multiple access Compact disc-read only memory Connection endpoint

311

312

Acronyms

CEPT CERN CES CI CLP CORBA CPS-PDU CPU CSMA CSMA/CA CSMA/CD CTS DAMA dc DS-CDMA DSS.1 DTAP ECMA EDFA EDGE EIRP EMI ETSI FAMA FDD FDM FDMA FEC FFH-CDMA FIB FIFO FISU FSN GEO GGSN GPRS GSM GTW HDLC HEC

Conférence Européenne des Postes et Télécommunication Conceil Européen pour la Recherche Nucléaire Coast Earth station Connection identifier Cell loss priority Common object request broker architecture Common part sublayer protocol data unit Central processing unit Carrier sense multiple access Carrier sense multiple access with collision avoidance Carrier sense multiple access wit collision detection Clear-to-send Demand assignment multiple access Direct current Direct sequence code division multiple access Digital subscriber line no. 1 Direct transfer application part European Computer Manufacturer Association Erbium doped fiber amplifier Enhanced data rates for GSM evolution Equivalent isotropic radiated power Electromagnetic interference European Telecommunications Standardization Institute Fixed assigned multiple access Frequency division duplex Frequency division multiplexing Frequency division multiple access Forward error control Fast frequency hopping code division multiple access Forward indicator bit First in, first out Fill-in signaling unit Forward sequence number Geostationary orbit Gateway GPRS support node General packet radio service Global system for mobile communications (originally Groupe Spécial Mobile) Gateway exchange High-level data link control Header error correction

Acronyms

313

HLR HSS html http I IAM ICMP ICT IEEE IETF IGMP IMEI IMSI IN IP IPv4 IPv6 ISDN ISO ISP ISUP ITU LAN LAP LEO LES LLUB LSSU MAC MAN MAP MEO MEMS MMS MS MSC MSU MTP MUE MVNO NAP

Home location register Home subscription server Hypertext Markup Language Hypertext Transfer Protocol Information frame Initial address message Internet Control Message Protocol Information and communication technology The Institution of Electrical and Electronics Engineers Internet Engineering Task Force Internet Group Management Protocol International mobile equipment identity International mobile subscriber identity Intelligent network Internet Protocol Internet Protocol version 4 Internet Protocol version 6 Integrated services digital network International Organization for Standardization Internet service provider ISDN user part International Telecommunication Union Local area network Link access protocol (often with a suffix B, D, Dm, and so on) Low Earth orbit Land Earth stations Local loop unbundling Link status signal unit Media access control Metropolitan area network Mobile application part Medium Earth orbit Microelectromechanical system Multimedia messaging service Mobile station Mobile services switching center Message signal unit Message transfer part Mobile user equipment Mobile virtual network operator Network access point

314

Acronyms

NCS NMT OA&M OBAN OMG OSPF PA PABX PAN PBX PC PCM PCU PDA PDH PDU PKI PLCP PLL PLMN PSK PSTN PTI PVC O&M QoS RA RBOC REJ RFID RNC RNR RO RPC RR RSVP RTS RX SABM SACCH SAP

Network control station Nordic mobile telephone system Operation, administration, and maintenance Open broadband access network Object Modeling Group Open shortest path first Preassigned Public automatic branch exchange Personal area network Public branch exchange Personal computer Pulse code modulation Packet control unit Personal digital assistant Plesiochronuous digital hierarchy Protocol data unit Public key infrastructure Physical layer convergence protocol Phase-locked loop Public land mobile network Phase shift keying Public switched telephone network Payload type identifier Permanent virtual circuit Operations and management Quality of service Random access Regional Bell operating company Reject Radio frequency identification Radio network controller Receive not ready Remote operations Remote procedure call Receive ready Resource reservation protocol Request-to-send Receiver Set asynchronous balanced mode Slow associated control channel Service access point

Acronyms

315

SCCP SCP SCPC SCTP SDCCH SDH SDMA SES SFT SFH-CDMA SGSN SIF SIM SIO SIP SMS SOLAS SS7 SSB SSN SSP STM STS SVC TACS TCAP TCH TCP TDD TDM TDMA TMSI TT&C TTP TU TUG TUP TX UA UDP UMA

Signaling connection control part Service control point Single channel per carrier Stream control transmission protocol Standalone dedicated control channel Synchronous digital hierarchy Space division multiple access Ship Earth station Simple file transfer Slow frequency hopping code division multiple access Serving GPRS support node Signaling information field Subscriber identity module Service information octet Session Initiation Protocol Short message service Safety of life at sea Signaling system no. 7 Single sideband Subsystem number Service switching point Synchronous transport modus Synchronous transmission signal Switched virtual channel Total access system Transmission control application part Traffic channel Transmission control protocol Time division duplex Time division multiplexing Time division multiple access Temporary mobile subscriber identity Telemetry, tracking, and control Trusted third party Tributary unit Tributary unit group Telephone user part Transmitter Unnumbered acknowledgment User datagram protocol Unlicensed mobile access

316

Acronyms

UMTS UP USO UTRAN UW VC VCI VCO VLR VNO VoIP VPI VSAT WAN WARC WDM WLAN WP-CDMA WWW xDSL XML

Universal mobile telecommunications service User part Universal service obligation UMTS terrestrial radio access network Unique word Virtual circuit Virtual channel identifier Voltage-controlled oscillator Visitor location register Virtual network operator Voice over IP Virtual path identifier Very small aperture terminal Wide area network World Administrative Radio Conference Wavelength division multiplexing Wireless local area network Wideband packet CDMA World Wide Web Digital subscriber line of type x Extensible Markup Language

Bibliography Bing, B., High-Speed Wireless ATM and LAN, Norwood, MA: Artech House, 2000. Black, U., Emerging Communications Technologies, Upper Saddle River, NJ: Prentice Hall, 1997. Blanchard, A., Phase-Locked Loops: Application to Coherent Receiver Design, New York: John Wiley & Sons, 1976. Borella, Michael S., et al., “Optical Components for WDM Lightwave Networks,” Proceedings of the IEEE, Vol. 85, No. 8, August 1997. Campanella, S. J., “Satellite Communications,” in Encyclopedia of Telecommunications, R. A. Meyer (ed.), New York: Academic Press, 1989. Comer, D. E., Computer Networks and Internets with Internet Applications, 4th ed., Boston, MA: Pearson Education International, 2004. Do, van Thanh, (ed.), “Future Mobile Phones,” Telektronikk, No. 3/4, 2005. Elbert, B. R., Introduction to Satellite Communication, Norwood, MA: Artech House, 1999. Freeman, R. L., Fundamentals of Telecommunications, New York: John Wiley & Sons, 1999. Forouzan, B. A., Data Communication and Networking, 4th ed., New York: McGraw-Hill International Edition, 2007. Goncalves, M., and K. Niles, IPv6 Networks, New York: McGraw-Hill, 1998. Halsall, F., Computer Networking and the Internet, 5th ed., Reading, MA: Addison-Wesley, 2005. Hipponen, M., “Malware Goes Mobile,” Scientific American, November 2006. Inoue, Y., M. Lapierre, and C. Mosotto, (eds.), The TINA Book: A Co-Operative Solution for a Competitive World, Upper Saddle River, NJ: Prentice-Hall, 1999. Izzo, P., Gigabit Networks: Standards and Schemes for Next-Generation Networking, New York: John Wiley & Sons, 2000. Jain, B. N., and A. K. Agrawala, Open Systems Interconnection: Its Architecture and Protocols, New York: McGraw-Hill, 1993. Kleinrock, L., and S. S. Lam, “Packet Switching in a Multiaccess Broadcast Channel: Performance Evaluation,” IEEE Transactions on Communications, Vol. COM-23, No. 4, April 1975. Lam, S. S., and L. Kleinrock, “Packet Switching in a Multiaccess Broadcast Channel: Dynamic Control Procedures,” IEEE Transactions on Communications, Vol. COM-23, No. 9, September 1975. Lee, J. S., and L. E. Miller, CDMA Systems Engineering Handbook, Norwood, MA: Artech House, 1998. Mouly, M., and M. B. Pautet, The GSM System for Mobile Communication, published by the authors, 1992. Noll, J., and E. Buracchini, “Software Radio-A Key Technology for Adaptive Access,” in Wireless Communications and Mobile Computing, Vol. 2, New York: John Wiley & Sons, 2002, pp. 789–798.

317

318

Bibliography Ohmori, S., H. Wakana, and S. Kawase, Mobile Satellite Communications, Norwood, MA: Artech House, 1998. Olsson, A., Understanding Changing Telecommunications: Building a Successful Telecom Business, New York: John Wiley & Sons, 2003. Perros, H. G., An Introduction to ATM Networks, New York: John Wiley & Sons, 2001. Proakis, J. G., and D. G. Manolakis, Digital Signal Processing: Principles, Algorithms, and Applications, 3rd ed., Upper Saddle River, NJ: Prentice-Hall International, 1996. Prasad, R., W.Mohr, and W. Konhäuser, (eds.), Third Generation Mobile Communication Systems, Norwood, MA: Artech House, 2000. Redl, S. M., M. K. Weber, and M. W. Oliphant, An Introduction to GSM, Norwood, MA: Artech House, 1995. Rheingold, H., Smart Mobs: The Next Social Revolution, New York: Basic Books, 2002. Schiller, J., Mobile Communication, Reading, MA: Addison-Wesley, 2000. Siwiak, K., Radio Propagation and Antennas for Personal Communications, Norwood, MA: Artech House, 1998. Smith, C., and D. Collins, 3G Wireless Networks, New York: McGraw-Hill, 2002. Stallings, W., Business Data Communications, 5th ed., Boston, MA: Pearson Education International, 2005. Stallings, W., ASDN and Broadband ISDN with Frame Relay and ATM, Upper Saddle River, NJ: Prentice-Hall, 1995. Vince, J., and R. Earnshaw, (eds.), Digital Media: The Future, New York: Springer, 2000. Walker, J., (ed.), Advances in Mobile Information Systems, Norwood, MA: Artech House, 1998. Wilkinson, N., Next Generation Network Services: Technologies and Strategies, New York: John Wiley & Sons, 2002. Yang, S. C., CDMA RF System Design, Norwood, MA: Artech House, 1998. Zouganeli, E., (ed.), “Optical Communication,” Telektronikk, No. 2, 2005.

About the Author Jan A. Audestad holds an M.S. in theoretical physics from the Norwegian Institute of Technology in 1965. He has worked in various fields of telecommunications since 1967 to 1971 in Telenor first as a researcher and research manager and later as advisor for the corporate management. Work includes domestic satellite communications to oil installations in the North Sea and for connecting the artic island Spitzbergen to the mainland telecommunications network, maritime satellite communications leading up to the establishment of INMARSAT, development of specifications related to the network aspects of GSM, standardization and implementation of intelligent networks, and the introduction of distributed processing platforms in the telecommunications network. One major activity has been development of new signaling protocols for various telecommunications systems. As an advisor for the corporate management of Telenor, his major work has been in the fields of techno-economic strategy, business development, and postgraduate education programs. Since 1993 Professor Audestad has been adjunct professor in distributed processing at the Norwegian University of Science and Technology (NTNU) where he has been teaching several courses in distributed processing, protocol theory, access and transport networks, and telecommunications business strategy. Since 2003 he has also been adjunct professor in information security at the University College of Gjøvik, Norway, where he has been teaching subjects such as the effects of ICT on the vulnerability of society, non-repudiation, and the economy of information security. He has chaired or cochaired several international groups, particularly in ITU, ETSI and ESA in satellite communications, land mobile communications, and intelligent networks. Professor Audestad has produced more than 100 papers published in scientific journals or presented at scientific conferences.

319

Index 1-persistent CSMA, 120 3GPP release, 5, 219 3GPP (Third Generation Partnership Project), 31

A Absorption (radio signal), 265–67 Abstract syntax notation no. 1 (ASN.1), 178, 186, 196 Access network (defined), 9 Acquisition ATM cells, 57–60 envelopes of constant length, 55–56 in PLL, 44, 302–3 random access burst, 60 Sun acquisition subsystem, 256 TDMA burst, 60 Acquisition device, 44, 307–9 Acquisition phase, 56 Acquisition range, 44, 303, 307, 309 Adaptive access, 242–43 Add-drop multiplexer optical, 287 SDH, 87, 90 Address complete message (ACM), 150 Administrative domain, 24–25 Administrative unit, 88 Administrative unit group, 89 ADSL, 3, 5, 11, 76, 80, 248 Aeronautical satellite system (INMARSAT Aero), 253, 258, 277 All-IP UMTS architecture, 13, 200, 219–21 convergence, 3, 205 effects on business models, 33 evolution toward 3G, 232 handover, 236 as stupid network, 14 All-optical network, 283 Aloha (defined), 119 Altitude of satellite orbit, 257 Ambient temperature (in satellite systems), 270 Amplifier (optical), 293, 295 Anisochronous, 39, 41–43 ANSI, 87

Antenna gain, 265–66, 269, 277–78 Antenna height, effect of, 207–8, 210 Antenna tracking, 258, 277–78 API, 199, 243 Apogee, 255 Apogee boost rocket, 256 Application layer, 173, 181, 184, 189, 193, 200 Application part, 193 Application protocol, 16, 26, 27, 31, 175, 184, 202 Application service provider (ASP), 19, 21 Architecture of mobile systems, 212–20 Architecture of satellite systems, 259–63, 275–77 Architecture of signaling network, 191–92 Asynchronous, 37–38, 50, 54 Asynchronous transfer mode (ATM) ATM adaptation layer (AAL), 40, 96, 98 cell (frame) format, 57, 142 control procedure, 133 convergence, 3 service integration, 2, 17 statistical multiplexer, 95–98 switching, 143 synchronization, 57–60 use of leaky bucket, 133 ATM cell, 57, 72, 96, 133, 142 Atomic clock, 38, 52 Attenuation (radio signal), 103, 207, 247, 251, 266, 284 Attitude control, 257–58 Authentication, 157, 190, 200–1, 215–17, 241 Authentication center (AUC), 201, 202, 216, 241 Autocorrelation in CDMA, 11, 115, 116 device, 64 peak, 67, 69 between pixels, 42 between speech samples, 42 unique word, 64, 69 Automatic gain control (AGC), 43, 61, 65, 78 Automatic repeat request (ARQ), 26, 41, 183, 189, 194

321

322

B Backbone network, 11 Backlog, 127–28, 130–31, 132, 133 Back-off algorithm, 120, 125 Back-off counter, 126 Back-off window, 126 Backward compatibility, 28–30 Backward signals of signaling system, 150 Balun, 80 Bar state, 160, 288, 295 Baseband signal FDM, 74–79 radio relay, 249 Base station controller (BSC) defined, 214 role in handover, 236–40 Base transceiver station (BTS), 214 Banyan network, 163, 295 Batcher-Banyan network, 165, 298 Binary inverse sequence, 61, 109 Binary switching element, 160, 288, 295 Bit error rate (BER) direct multitone ADSL, 81 effect on TCP, 41, 126 effect on throughput, 26, 40–41 Bit interleaved, 83, 86 Block error rate, 40–41 Blocking network, 159 Bluetooth, 10, 13, 242, 244, 245 Bluetooth Interest Group, 31 Boltzmann’s constant, 267 Broadcast control channel, 229, 231, 233, 237 Broadcast satellite, 12, 34, 255, 259 Buffer (applications of) elastic store, 53, 93 leaky bucket, 133 multiplexing, 82 packet switching, 144 playback, 27 switching, 144, 156, 164, 165 synchronization, 53 Burst format, 60, 64, 68, 239 Byte-oriented protocol, 189

C Cable system, 12 Call procedure in GSM, 217 CAMEL, 216 Capture effect, 122 Care-of number, 191 Carrier extension, 125

Index

Carrier sense multiple access (CSMA), 120, 124, 126, 248 collision avoidance, 126 collision detection, 124 Carrier synchronization, 60, 69 Carrier-to-noise ratio (C/N), 271 Cartesian reference coordinate system, 277 CDMA transceiver, 112 Cell (in land mobile systems), 203 Cell diameter in GSM, 65 Cell identity, 214, 229 Cell loss priority (CLP), 57, 96 Cellular network, 203 Cell size, 208 Central processing unit (CPU), 2, 16 Channel capacity, 113 Channel coding, 119, 223, 232, 242, 244 Channel load line, 129 Channel saturation, 130 Charging in Internet, 19–21 Charging services, 137 Chip (definition), 109 Chip rate (definition), 109 Circuit switching, 138–42, 168, 212, 222, 296 Cladding, 283 Clear-to-send, 127 Clock hierarchy, 51 Clock synchronization, 45, 60 Clos-Benes network, 159, 161, 295 Code division multiple access (CDMA), 4, 109–16 applications, 69, 118, 126, 231–35 soft capacity limit, 117 Code word, 225 Coding gain, 69, 109, 116 Coherent demodulation applications, 50, 60, 67, 116 defined, 48 Coherent light, 292 Collision avoidance, 120, 126 Collision detection, 124, 126 Combiner, 286 Common channel signaling, 143, 167, 192 Concurrency, 145 Conditionally nonblocking switch, 160 Conference call, 137, 149 Connection endpoint, 175 Connectionless network, 135, 141, 144, 151, 181, 195 Connection-oriented network, 10, 135, 139, 141, 183 Connection service, 136

Index

Container, 88, 171 Contention probability, 120 Contention resolution, 123, 239 Contention window, 122, 126 Control bit, 86–87, 155 Control procedure, 131–33 input control procedure, 132 input-retransmission control procedure, 132 retransmission control procedure, 132 Control signal, 152, 156 Convergence, 3 Core, 283 Correlation, 61, 110 Correlator, 49, 61, 110, 112, 113, 115 Coupler, 286 Coupling length, 288 Crossbar switch, 152, 159, 162 Cross state, 160, 288, 295 Cycle slip, 44, 46, 304

D Data link layer, 41, 171, 181, 183, 188, 193, 199 Decrement bit, 92 Delimitation, 42, 96, 99, 181 Delimiter, 67, 99, 180, 184, 193 Delimiter service, 180 Demand assignment multiple access (DAMA), 101, 115, 264 Demarcation line, 16–17 implications, 19–23 Demultiplexer defined, 73 design of, 77, 79 optical, 287 satellite transponder, 119, 259 Density of fades, 211 Deployment orbit, 255 Direct multitone ADSL, 81 Dispersion, 285 Distortion (intermodulation), 78 Distributed coordination function IFS (DIFS), 126 Distributed processing, 139, 173 Doppler shift, 38 Domains, 24 Duplexer, 103, 105, 112, 119, 249, 263 Dynamic burst length, 70

E Early Bird, 251, 253 Eccentricity, 255 Eclipse, 256

323

Edge diffraction, 209, 247 Effect of antenna height, 208 Einstein, 290 Elastic store, 38, 53, 54, 93 Electrical length, 37, 53 Electricity modem, 12 Electromagnetic interference (EMI), 281 Encapsulation, 88, 190 Encryption between administrative domains, 24 key escrows, 200 in mobile systems, 201, 216, 223, 240–42 Envelope, 42, 55–56, 91–94 Equilibrium contour, 129 Equipment identity register, 200, 216 Equivalent isotropic radiated power (eirp), 265 Equivalent output noise temperature, 269 Erbium doped fiber amplifier (EDFA), 293 Ethernet access, 124–25 ETSI, 30, 87, 231 Evanescent wave, 285, 287 Evolution of telecommunications, 1–3 Excitation device, 291

F Fabry-Perot interferometer, 289, 290 Fading frequency selective, 108, 232 in land mobile systems, 210–12 multipath diversity, 116 Fast frequency hopping CDMA, 116 Fault recovery, 145, 146 Fiber, 283–86 Figure of merit (G/T), 271 Fill-in signal unit (FISU), 194 First-in-first out (FIFO) queuing discipline, 95, 156 First order European digital multiplex, 83–84 Fixed assignment multiple access (FAMA), 101, 115 Flag (HDLC), 42, 99, 193 Floating payload, 90 Flow control, 25, 41, 51, 137, 182, 189, 193 Fluid approximation, 131 Forward signal unit, 194 Forward signals, 150 Frame alignment, 86 Frame bursting, 125 Frame synchronization, 84, 85, 86 Free space attenuation (loss), 207, 251 Frequency acquisition, 44 Frequency allocations, 33–35

324

Frequency correction control channel, 228 Frequency diversity, 108 Frequency division duplex (FDD), 102, 125, 233 Frequency division multiple access (FDMA), 102–4 comparison with TDMA and CDMA, 117 duplex operation (FDD), 102 GSM, 223 satellite systems, 259, 264 UMTS, 232, 234 Frequency division multiplexing (FDM), 74–81 ADSL, 80–81 application of frequency synthesizer, 77 distortion, 78–79 pilot, 77 translation of frequencies, 74–77 wavelength division multiplexing, 287 Frequency synthesizer, 48

G Gateway exchange (GTW), 214 Gateway GPRS support node (GGSN), 219, 220 General packet radio service (GPRS), 218 Geographic number, 146 Geostationary orbit (GEO), 251, 255 Geostationary satellites in access network, 12, 13 inclination, 255 number of orbits, 255 orbital stability, 38, 104, 255, 257 propagation delay, 25, 46, 104, 261 radio propagation, 265 shape of orbit, 67, 104, 255 synchronization, 46, 67–71 systems, 12, 13, 121, 251, 255, 274 in transport network, 11 Globalstar, 10 GPRS as access technology, 12 architecture, 218–19 channel organization, 231 data rates, 218 location updating, 220–21 mobile IP, 220–21 Graded-index fiber, 285 GSM as access technology, 12 architecture, 214–216 authentication and security, 240–42 call handling, 217–18

Index

cell size, 65 channel coding, 223–28 duplex arrangement (TDD), 63 guard time between bursts, 64 handover, 237–40 history of, 205 interleaving, 226–28 location management, 216–17 logical channels, 228–30 media management, 201 mobility management, 201 protocol stacks, 201–2 radio resource management, 201 random access burst, 64 SIM, 240–42 TDMA frame and burst format, 63–64, 239 timing advance, 65 training sequence, 64

H Handover, 235–40 Hard capacity limit, 117 Hard handover, 237–40 Header error correction (HEC), 57 synchronization mode, 57–60 HEC generator, 58 Heterogeneous system, 25–26 High-level data link control (HDLC), 42, 43, 99, 123, 181–83, 193, 202 Home location register (HLR) defined, 215 role in call handling, 217–18 role in location updating, 216–17 Home subscription server (HSS), 220 Hyperframe definition, 223 role in GSM encryption, 223, 242

I IEEE 802.11, 13, 66, 109, 116, 120, 126, 127, 248 Increment bit, 92 Index field, 160–61, 163 Information hiding, 145–46 Initial address message (IAM), 150–51 INMARSAT antenna tracking, 277–78 architecture, 275–77 frequency bands, 275 link budgets, 278–79 organization, 253 random access, 121 systems, 274–75

Index

Input control procedure, 132 Input-retransmission control procedure, 132 Institution of Electrical and Electronics Engineers (IEEE), 31 Intelligent network, 13, 149 INTELSAT, 68, 251–53 teleport, 261 Interference in CDMA, 113, 114 contribution to total noise, 69 electromagnetic interference (EMI), 281 frequency sharing, 266 interferer diversity, 107 interferometers, 288–90, 292 intermodulation, 78–79 multipath, 116, 207, 210, 232, 247, 251, 267 Interferer diversity, 107 Interferometer, 288–90 Interframe space, 126 Interleaving, 71–72, 82–83, 108, 226–28, 240 Intermodulation, 78–79 International mobile equipment identity (IMEI), 216 International mobile subscriber identity (IMSI), 215, 216, 240 Internet Control Message Protocol (ICMP), 136, 189–90 Internet Engineering Task Force (IETF), 31 Internet Group Management Protocol (IGMP), 190 Internet Protocol (IP) all-IP UMTS, 220 charging in Internet, 19–22 encapsulation of PDU, 190 format of datagram, 185 GPRS as IP network, 218–19, 220 handover, 235 mobile IP, 10, 220–21 overlay access, 23–24 service primitive, 180 stupid network, 13 switching of datagram, 144 tunneling of datagram, 190–91 Internet service provider (ISP), 14–23, 259 Interworking, 24, 29, 30, 137, 178 IPsec, 15, 183, 190 Isochronous, 39, 42, 57, 224 defined, 39 Iridium system, 12, 254, 256, 272, 274 ISO, 30 ITU, 30, 33, 87

325

J Jamming resistance, 113 Jitter, 27, 37, 44, 45–46, 52–53, 54–55, 69, 304 Junction module, 167

L Laser, 290–92 Launch orbit, 256 Layered protocol, 173 Leaky bucket, 133 Lee model, 209 Line overhead, 94 Link budget, 271–72, 278 Link-by-link synchronization, 51, 52 Link status signal unit (LSSU), 193–95 Local loop unbundling (LLUB), 5, 248 Location area defined, 213–14 size constraint, 221–22 VLR control, 215, 216 Location-area identity, 214, 216, 220 Location coordinates, 214, 216, 220 Location updating, 213, 220 mobility management, 201 procedures, 216–17 role of HLR, 200, 202 role of SDCCH, 229 role of SGSN and GGSN, 219, 220 role of VLR, 200, 215, 216 Logical channels, 228, 234 Logical information exchange, 171–72 Loop filter, 45, 299 Loss in optical fibers, 286 Loss in radio systems, 265–67 Low Earth orbit (LEO), 12, 254 cost of, 272–73 eclipse, 256–57 Lower layers, 14, 188–189 Low-noise amplifier, 61, 102, 259, 270 Low-noise receiver, 263, 264, 270

M Mach-Zehnder (MZ) interferometer, 288, 292 Malicious software, 245 Management services, 137 MARISAT, 253, 256, 274 Master clock, 51 Media access control (MAC), 67, 125, 199, 234 Media gateway, 220 Media management, 200–1

326

Medium Earth orbit (MEO), 254 Message-oriented protocol, 189–90 Message signal unit (MSU), 193–95 Message transfer part (MTP), 193, 195 Microelectromechanical system (MEMS), 282, 296 Midamble, 63, 234 Mobile application part (MAP) application in GSM, 201–2 ASN.1 description, 186 weak synchronization property, 146 Mobile IP, 220 discrete mobility, 10, 213 GPRS and UMTS, 219–21 overlay access, 23 tunneling, 20, 220–21 Mobile-services switching center (MSC) defined, 214 as gateway, 214 role in handover, 236–40 routing number, 218 Mobile malware, 245 Mobility management, 200, 201, 202, 215 Mobility service, 137, 244 Mobile virtual network operator (MVNO), 24 Modulation theorem, 74 Modulator (optical), 292 Mono-mode fiber, 285 Moore’s law, 26 Multicast, 13, 20, 137, 153, 158, 190, 220 Multimode fiber, 285 Multipath attenuation, 251 Multipath channel, 211 Multipath diversity, 116 Multipath interference, 116, 207, 210, 232, 247, 267 Multipath propagation, 113, 232, 251 Multiplexer ADSL, 80 defined, 73 direct multitone ADSL, 81 FDM, 77–78 optical, 287 satellite transponder, 119, 259 SDH, 88–89 statistical, 95–96 TDM, 82

N (N)-connection, 175 (N)-entity, 174 (N)-layer, 173

Index

(N)-protocol, 174 Network access point, 146 number assignment, 149, 190, 218 Network control station, 276 Network identity, 214, 215, 216, 220 Network layer, 20, 183, 188, 190, 193, 195, 200 Network neutrality, 18–19 Node B, 220 Noise, 267–71 Noise bandwidth, 267, 304 Noise figure (definition), 268 Noise figure of attenuator, 268 Noise reduction in PLL, 44 Noise temperature, 267–70 ambient, 270 Moon, 269 Sun, 269 Nonblocking switch, 160 Nongeographic number, 146, 149 Nonpersistent CSMA, 120 Nonthermal noise and 1/f noise, 37 NORSAT, 253 Number analysis, 140, 146–50 Number analysis module, 167 Number portability, 29, 31

O Object Modeling Group (OMG), 31 Omnidirectional antenna, 206, 272 Open-loop control, 114 Open-loop tracking, 278 Open system interconnection (OSI), 172 Operation, administration and maintenance (OAM), 90, 202 Operation and maintenance, 88, 89, 143, 196 Operating threshold of PLL, 44, 46 Optical fiber, 283–86 Optical filter, 288 Optical grating, 290 Optical label swapper, 297 Optical semiconductor amplifier, 293, 294 Optical switch, 295–98 Output position selector, 53, 54

P Packet control unit (PCU), 219 Packet radio, 66, 218, 222 Packet switching connectionless, 144 connection oriented, 142 label switching, 297 mobile, 212

Index

network, 136 optical, 297 self-routing, 160 Paging channel, 221, 228, 229 Passive reflector, 250 Path overhead, 88, 89, 90, 94 Pauli’s exclusion principle, 291 PDH hierarchy, 87, 90 Peer production, 22 Perigee, 256 Phase detector, 44, 115, 299, 304–7 Phase-locked loop, 43–50, 299–309 acquisition device, 307 acquisition range, 44 applications, 45–50, 51, 52, 53, 56, 60, 61, 69, 76, 104, 115 defined, 43–50 mathematical theory, 299–309 noise bandwidth, 304 operating threshold, 44 synchronization range, 44, 301 Phase noise, 44, 53 Phase-shift keying, 48 Photo detector, 257, 293, 294 Physical information exchange, 171 Physical layer, 173, 175, 181, 188, 193, 199 Physical Layer Convergence Protocol (PLCP), 67 Physical medium, 173, 181 Platform defined, 9 multipurpose mobile platform, 242 purpose, 136 time-sharing capability, 139 virtual network, 23–24 Playback buffer, 27 Plesiochronous defined, 38 floating payload, 88, 90 interconnection of plesiochronous networks, 54–55 multiplexing of plesiochronous signals, 85, 88, 93–94 plesiochronous terminals, 66 rate adjustment, 85, 93 P-persistent CSMA, 120 Point coordination function IFS (PIFS), 126 Pointer, 90–94, 98 Pointer address, 92 Population inversion, 291

327

Port address (number) switching, 163, 165 transport protocol, 16, 144, 183, 190, 195 Power control CDMA, 113, 114 GPRS, 231 GSM, 65, 201, 229, 230 packet radio, 222 UMTS, 233 Preamble, 67 Presynchronization, 56, 59 Primitive, 176–80, 183, 197, 198, 201 Programmable hardware module, 244 Programmable service module, 243, 244 Programmable radio interface, 243, 244 Protocol data unit AAL2, 98 coding, 184–88 embedding, 176 header, 176 payload, 176 tail, 176 Protocol services, 177 Pseudo-noise sequence, 109, 110 Pseudorandom sequence, 71, 106, 113, 115, 116 Public key infrastructure, 4, 25 Public land mobile network (PLMN) architecture, 212–22 defined, 204 Public switched telephone network (PSTN), 136 routing, 147–49 Pure Aloha, 119, 121, 132, 276

Q Quality of service (QoS), 137, 151, 189 charging, 19–22

R Radio frequency identification (RFID), 1, 2, 173 Radio network controller (RNC) defined, 219 role in soft handover, 235–36 Radio resource management, 200–2, 226 Radio relay, 11, 34, 73, 248–51 Rain attenuation, 251, 266, 268 Rain margin, 266 Random access, 60, 101, 119–133 Random access burst, 63–64, 66 Random access channel, 127, 224, 230, 233, 239, 264, 276

328

Random access message, 66, 122, 239 Rate adjustment, 85, 92, 93 Rayleigh fading, 212 Rayleigh distribution, 211 Rayleigh scattering, 286 RBOC, 87 Real-time effect of ARQ, 41 effect of HDLC, 43 handling lost information, 41 operation, 27, 40, 144 performance, 139 priority, 144 Reflection (of radio waves), 107, 209, 251, 252, 267, 284, 285 Reflection coefficient, 207, 209 Refraction, 107, 283, 284 Refractive index, 283 Remote operations (RO), 184, 196 Repeater, 77, 249, 250, 263 Request-to-send, 127 Resonant cavity, 292 Retransmission control procedure, 132, 133 Rician fading, 211 Ring modulator, 75, 304

S SABM, 123, 182, 238–39 Safety of life at sea (SOLAS), 1 Scrambler, 71–72 Section overhead, 94 Sectorial cell, 203 Security services, 137, 241, 244 Self-routing, 160, 163 Semantics, 24, 145, 172, 178, 196 Service access point (SAP), 174 Service control point (SCP), 150, 216 Service processing unit, 167 Service switching point (SSP), 150 Serving GPRS support node (SGSN), 219, 220 Signaling in circuit-switched networks, 141 in connectionless networks, 151 in connection-oriented networks, 141, 150–51 defined, 136 signaling system no. 7, 191–99 Signaling connection control part (SCCP), 195 Signaling information field, 193 Signaling link control, 193–95 Signaling network, 191–92 Signaling point (SP), 192

Index

Signaling system no. 6, 192 Signaling system no. 7, 192–99 Signal restoration, 53, 61 Signal-to-noise ratio (S/N), 69, 103, 108, 110, 267–68 Single channel per carrier (SCPC), 101, 105, 276 Single-mode fiber, 285 Single sideband (SSB), 75 Slip (in plesiochronous systems), 55 Slotted Aloha, 119, 122, 127 Slot time, 124–27 Slow associated control channel SACCH, 229 Slow frequency hopping CDMA, 101, 106–9, 223 Snell’s law, 284 Snooper, 25 Snooping, 26 Soft capacity limit, 117 Softcoding, 186–88 Soft handover, 235–37 SONET, 87 Space division multiple access, 101, 118, 276 Space-division switching, 152, 156 Splitter, 286 Sporadic synchronization, 50 Spot-beam, 118, 119, 257, 259, 264, 275 Spreading ratio, 109 Spread spectrum, 4, 69, 109 Standalone common control channel (SDCCH), 229 Starlight switch, 167 Start frame delimiter, 67 State transitions, 128 Station keeping, 255, 263, 273 Statistical multiplexing, 73, 94–99 Stealing flag, 63, 239 Step-index fiber, 285 STOLAS switch, 297 Stop-and-wait, 26 Stream Control Transmission Protocol (SCTP), 189–90 Structured information, 171 Stupid network, 13, 20 Subscriber identity module (SIM) defined, 240–42 mobility management, 201 software platform, 244 standard interface, 173 Sunshine switch, 167 Sun synchronous orbit, 256 Superframe, 83–84, 223, 230, 233

Index

Supplementary services, 14, 137, 167, 202 Surface wave, 209 Switching fabric, 119, 148, 164, 168 Switching path, 145, 156, 167, 297 Synchronization channel, 224, 229 Synchronization range, 44, 301, 307, 309 Synchronization word, 55, 64, 84, 86, 94 Synchronous defined, 37 networks, 50–54 SDH, 87–94 Sun synchronous orbit, 256 Synchronous digital hierarchy (SDH), 87–94 bit rates, 89–90 composition of AU–4, 90 control headers 94 decrement bit, 92 flexibility, 89 floating payload, 91 line overhead, 94 multiplexing structure, 88 new data flag, 92 path overhead, 94 overhead and coding efficiency, 89 plesiochronous signals, 93 pointer field, 92 purpose, 87 rate adjustment, 93 section overhead, 94 structure of STM-1, 90–91 Synchronous transmission signal (STS), 89 Synchronous transport module (STM), 89 SYNCOM II, 251 Syntax, 24, 145, 172 Abstract Syntax Notation No. 1, 178, 186 System temperature, 270

T Telex, 1, 274, 275 Technology domain, 24 Teledesic, 12, 254, 272 Telemetry, tracking & control (TT&C), 257, 263 Teleport, 261 Temporary mobile subscriber identity (TMSI), 216–18 Thermal noise, 69, 103, 110, 267 Throughput, 73, 119, 129, 130, 167, 185, 186 Time charging, 20 Time division duplex (TDD), 63, 105, 233–234

329

Time division multiple access (TDMA) burst contraction, 70 burst timing, 46 defined, 101, 104–5 synchronization, 46, 60–62, 70 TDMA burst, 60, 63, 66, 105, 108 TDMA frame, 60, 69, 70, 108, 223, 226, 228, 230, 231 TDMA timeslot, 63, 122, 223, 234, 259 Time division multiplexing (TDM), 82, 87, 105, 158, 263 Time-division switching, 152, 157–159 Timesharing, 139–140, 145 Time-to-live, 15, 151, 168, 180, 190 Timing advance, 62–66, 105, 229, 230, 231, 233, 239 Timing diagram, 121 Timing unit, 167, 169 Towns, Charles, 291 Tracking of antenna, 43, 277–78 Tracking of equipment, 216 Tracking of signal, 45, 55, 57, 115 Tracking of Sun, 257 Tradeoff, 63, 221, 226, 266, 272–74 Traffic channel (GSM), 223, 228, 229, 230 Training sequence, 63–64, 105, 229, 234, 239 Transaction capabilities, 193, 202 Transaction indicator, 198 Transaction message, 198 Transceiver, 112, 200, 203, 220, 248, 255 Transfer orbit, 255, 257 Transfer syntax, 172 Translation services, 137 Translation of channels (FDM), 74–77 Transmission Control Protocol (TCP) ARQ operation, 41 in connectionless packet switching, 144 contention resolution, 126, 133 control procedure, 133 encapsulation of PDU, 15, 190 as line of demarcation, 16 port address/number, 16, 183 protocol hierarchy, 15, 189 role in silly network, 16 Transparency stuffing, 42 Transponder, 119, 263 Transport layer, 16, 183, 189 Transport network (definition), 9 Transversal mode, 285 Tributary unit, 88, 90 Tributary unit group, 89, 90 Trunk module, 167

330

Trusted third party (TTP), 25 Tunable laser, 292, 297 Tunable wavelength converter (TWC), 296, 297 Tunnel identity, 221 Tunneling, 14, 190, 221

U UMTS as access technology, 12 architecture, 219–20 channel organization, 233–34 data rates, 232 handover, 235–37 handover to/from GSM, 237 location updating, 220–21 mobile IP, 220–21 multiple access, 237 radio interface, 232 Unique word, 61–62, 64, 67, 69, 105, 181 Unit for remote access, 168 Universal Service Obligation (USO), 30 Unstructured information, 171, 172 Up-down counter, 53–54 User data field, 184 User Datagram Protocol (UDP) connectionless packet switching, 144 encapsulation of PDU, 15 as line of demarcation, 16 port number, 16 protocol hierarchy, 15, 189 as transport protocol, 16 User module, 167 User part, 193, 196 User signaling module, 167 UTRAN, 220

V Varicap diode, 307 Very high frequency oscillator, 46 Virtual call, 136 Virtual circuit, 136 Virtual connection, 136 Virtual container, 88–89, 90–91 Virtual container VC-4, 90, 93 Virtual operator, 24 Visitor location register (VLR)

Index

call handling, 216 definition, 215 information security, 217–18 location management, 216–17 Voice over IP (VoIP) all-IP, 32 backward compatibility, 28–29 convergence, 3 intelligent network support, 14 interworking, 25, 137 real-time, 27, 139 Skype, 32 smart phones, 245 Video over IP, 3, 33, 245 Voltage controlled oscillator (VCO), 44, 299, 307–8 Volume charging, 20 VSAT, 13, 26, 253, 262–63

W Wavelength converter, 294–95, 296, 297 Wavelength division multiplexing (WDM), 5, 73, 78, 287, 292 Weak coupling, 145, 146 Weak synchronization, 145, 146 Wireless LAN (WLAN) in access network, 13 adaptive access, 242–43 application of CDMA, 126 contention control, 133 frequency band, 34 mobility, 10, 213 physical layer, 181 synchronization, 66–67 WiFi, 205, 243, 244 WiMAX in access network, 12 adaptive access, 242 architecture, 247–48 frequency band, 34 as subscriber line, 4 Word error rate, 40 Word interleaved, 83

X xDSL, 11 XOR gate, 58, 307

Recent Titles in the Artech House Mobile Communications Series John Walker, Series Editor 3G CDMA2000 Wireless System Engineering, Samuel C. Yang 3G Multimedia Network Services, Accounting, and User Profiles, Freddy Ghys, Marcel Mampaey, Michel Smouts, and Arto Vaaraniemi 802.11 WLANs and IP Networking: Security, QoS, and Mobility, Anand R. Prasad, Neeli R. Prasad Advances in 3G Enhanced Technologies for Wireless Communications, Jiangzhou Wang and Tung-Sang Ng, editors Advances in Mobile Information Systems, John Walker, editor Advances in Mobile Radio Access Networks, Y. Jay Guo Applied Satellite Navigation Using GPS, GALILEO, and Augmentation Systems, Ramjee Prasad and Marina Ruggieri CDMA for Wireless Personal Communications, Ramjee Prasad CDMA Mobile Radio Design, John B. Groe and Lawrence E. Larson CDMA RF System Engineering, Samuel C. Yang CDMA Systems Capacity Engineering, Kiseon Kim and Insoo Koo CDMA Systems Engineering Handbook, Jhong S. Lee and Leonard E. Miller Cell Planning for Wireless Communications, Manuel F. Cátedra and Jesús Pérez-Arriaga Cellular Communications: Worldwide Market Development, Garry A. Garrard Cellular Mobile Systems Engineering, Saleh Faruque The Complete Wireless Communications Professional: A Guide for Engineers and Managers, William Webb EDGE for Mobile Internet, Emmanuel Seurre, Patrick Savelli, and Pierre-Jean Pietri Emerging Public Safety Wireless Communication Systems, Robert I. Desourdis, Jr., et al. The Future of Wireless Communications, William Webb Geographic Information Systems Demystified, Stephen R. Galati GPRS for Mobile Internet, Emmanuel Seurre, Patrick Savelli, and Pierre-Jean Pietri GPRS: Gateway to Third Generation Mobile Networks, Gunnar Heine and Holger Sagkob

2

GSM and Personal Communications Handbook, Siegmund M. Redl, Matthias K. Weber, and Malcolm W. Oliphant GSM Networks: Protocols, Terminology, and Implementation, Gunnar Heine GSM System Engineering, Asha Mehrotra Handbook of Land-Mobile Radio System Coverage, Garry C. Hess Handbook of Mobile Radio Networks, Sami Tabbane High-Speed Wireless ATM and LANs, Benny Bing Interference Analysis and Reduction for Wireless Systems, Peter Stavroulakis Introduction to 3G Mobile Communications, Second Edition, Juha Korhonen Introduction to Communication Systems Simulation, Maurice Schiff Introduction to Digital Professional Mobile Radio, Hans-Peter A. Ketterling Introduction to GPS: The Global Positioning System, Ahmed El-Rabbany An Introduction to GSM, Siegmund M. Redl, Matthias K. Weber, and Malcolm W. Oliphant Introduction to Mobile Communications Engineering, José M. Hernando and F. Pérez-Fontán Introduction to Radio Propagation for Fixed and Mobile Communications, John Doble Introduction to Wireless Local Loop, Second Edition: Broadband and Narrowband Systems, William Webb IS-136 TDMA Technology, Economics, and Services, Lawrence Harte, Adrian Smith, and Charles A. Jacobs Location Management and Routing in Mobile Wireless Networks, Amitava Mukherjee, Somprakash Bandyopadhyay, and Debashis Saha Mobile Data Communications Systems, Peter Wong and David Britland Mobile IP Technology for M-Business, Mark Norris Mobile Satellite Communications, Shingo Ohmori, Hiromitsu Wakana, and Seiichiro Kawase Mobile Telecommunications Standards: GSM, UMTS, TETRA, and ERMES, Rudi Bekkers Mobile Telecommunications: Standards, Regulation, and Applications, Rudi Bekkers and Jan Smits

3

Multiantenna Digital Radio Transmission, Massimiliano “Max” Martone Multiantenna Wireless Communicatio ns Systems, Sergio Barbarossa Multipath Phenomena in Cellular Networks, Nathan Blaunstein and Jørgen Bach Andersen Multiuser Detection in CDMA Mobile Terminals, Piero Castoldi OFDMA for Broadband Wireless Access, Personal Wireless Communication with DECT and PWT, John Phillips and Gerard Mac Namee Practical Wireless Data Modem Design, Jonathon Y. C. Cheah Prime Codes with Applications to CDMA Optical and Wireless Networks, Guu-Chang Yang and Wing C. Kwong QoS in Integrated 3G Networks, Robert Lloyd-Evans Radio Engineering for Wireless Communication and Sensor Applications, Antti V. Räisänen and Arto Lehto Radio Propagation in Cellular Networks, Nathan Blaunstein Radio Resource Management for Wireless Networks, Jens Zander and Seong-Lyun Kim Radiowave Propagation and Antennas for Personal Communications, Third Edition, Kazimierz Siwiak and Yasaman Bahreini RDS: The Radio Data System, Dietmar Kopitz and Bev Marks Resource Allocation in Hierarchical Cellular Systems, Lauro Ortigoza-Guerrero and A. Hamid Aghvami RF and Baseband Techniques for Software-Defined Radio Peter B. Kenington RF and Microwave Circuit Design for Wireless Communications, Lawrence E. Larson, editor Sample Rate Conversion in Software Configurable Radios, Tim Hentschel Signal Processing Applications in CDMA Communications, Hui Liu Smart Antenna Engineering, Ahmed El Zooghby Software Defined Radio for 3G, Paul Burns Spread Spectrum CDMA Systems for Wireless Communications, Savo G. Glisic and Branka Vucetic Technologies and Systems for Access and Transport Networks, Jan A. Audestad Third Generation Wireless Systems, Volume 1: Post-Shannon Signal Architectures, George M. Calhoun

4

Traffic Analysis and Design of Wireless IP Networks, Toni Janevski Transmission Systems Design Handbook for Wireless Networks, Harvey Lehpamer UMTS and Mobile Computing, Alexander Joseph Huber and Josef Franz Huber Understanding Cellular Radio, William Webb Understanding Digital PCS: The TDMA Standard, Cameron Kelly Coursey Understanding GPS: Principles and Applications, Second Edtion, Elliott D. Kaplan and Christopher J. Hegarty, editors Understanding WAP: Wireless Applications, Devices, and Services, Marcel van der Heijden and Marcus Taylor, editors Universal Wireless Personal Communications, Ramjee Prasad WCDMA: Towards IP Mobility and Mobile Internet, Tero Ojanperä and Ramjee Prasad, editors Wireless Communications in Developing Countries: Cellular and Satellite Systems, Rachael E. Schwartz Wireless Communications Evolution to 3G and Beyond, Saad Z. Asif Wireless Intelligent Networking, Gerry Christensen, Paul G. Florack, and Robert Duncan Wireless LAN Standards and Applications, Asunción Santamaría and Francisco J. López-Hernández, editors Wireless Technician’s Handbook, Second Edition, Andrew Miceli For further information on these and other Artech House titles, including previously considered out-of-print books now available through our In-Print-Forever® (IPF®) program, contact: Artech House

Artech House

685 Canton Street

46 Gillingham Street

Norwood, MA 02062

London SW1V 1AH UK

Phone: 781-769-9750

Phone: +44 (0)20 7596-8750

Fax: 781-769-6334

Fax: +44 (0)20 7630-0166

e-mail: [email protected]

e-mail: [email protected]

Find us on the World Wide Web at: www.artechhouse.com