Smart antenna engineering- 2005- ahmed el zooghby

Smart Antenna Engineering

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

Smart Antenna Engineering

Ahmed El Zooghby

a r t e c h h o u s e . c o m

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

British Library Cataloguing in Publication DataEl Zooghby, Ahmed

Smart antenna engineering.—(Artech House mobile communications series)1. Antennas (Electronics) 2. Software radio I. Title621.3’824

ISBN-10: 1-58053-515-1

Cover design by Yekaterina Ratner

© 2005 ARTECH HOUSE, INC.685 Canton StreetNorwood, MA 02062

All rights reserved. Printed and bound in the United States of America. No part of this book maybe reproduced or utilized in any form or by any means, electronic or mechanical, including pho-tocopying, recording, or by any information storage and retrieval system, without permission inwriting from the publisher.

All terms mentioned in this book that are known to be trademarks or service marks have beenappropriately capitalized. Artech House cannot attest to the accuracy of this information. Use ofa term in this book should not be regarded as affecting the validity of any trademark or servicemark.

International Standard Book Number: 1-58053-515-1

10 9 8 7 6 5 4 3 2 1

The material covered in this book represents the views of the author, and doesnot necessarily reflect those of QUALCOMM Incorporated unless it is soindicated.

Contents

Preface xiii

Acknowledgments xvii

1 Introduction 1

1.1 Wireless Mobile Communications Systems 1

1.2 Global Mobile Market Growth 3

1.3 Alternatives for Meeting Data Demand 4

1.4 Technology Peak Rates and Throughput 6

1.5 Why Smart Antennas? 7

1.6 Benefits of Smart Antennas 7

1.7 Types of Smart Antennas 8

1.8 Switched and Fixed Beam Antennas 9

1.9 Adaptive Arrays 10

References 11

2 Multiple Access Techniques for 2G and 3G Systems 13

2.1 Introduction 13

2.2 Multiple Access Wireless Communications 14

2.2.1 FDMA Systems 14

2.2.2 TDMA Systems 15

v

2.2.3 Frequency Reuse 16

2.2.4 Cochannel Interference 18

2.2.5 CDMA Systems 20

2.3 Fundamentals of CDMA 21

2.3.1 Isolated Cell Capacity 24

2.3.2 CDMA Codes 25

2.3.3 IS-95 CDMA Systems 29

2.4 Third Generation Systems 36

2.4.1 CDMA2000 37

2.4.2 WCDMA 43

2.4.3 HSDPA 45

2.5 Basic CDMA Procedures 49

2.5.1 Acquisition State 49

2.5.2 Idle State 52

2.5.3 Access State and Call Setup 52

2.5.4 Traffic or Dedicated State 53

2.6 CDMA Embedded Cell Capacity 53

2.6.1 Multipath Fading 55

2.7 Coverage Versus Capacity Trade-Off 55

2.7.1 Coverage-Capacity Trade-Off in the Uplink 56

2.8 Conclusion 57

References 57

Selected Bibliography 59

3 Spatial Channel Modeling 61

3.1 Introduction 61

3.2 Radio Environments and Cell Types 63

3.3 The Multipath Channel 64

3.4 Channel Characterization 65

3.5 Path Loss Models 66

3.5.1 Okumura-Hata Propagation Models 66

3.6 Spatial Channel Modeling 67

3.6.1 Spatial Channel Model Parameters 68

vi Smart Antenna Engineering

3.6.2 Number of Clusters 69

3.6.3 Spatial Distribution of Clusters and Scatterers 69

3.6.4 Base Station Azimuth Power Spectrum and Angle Spread 69

3.6.5 Mobile Station Azimuth Power Spectrum and AngleSpread 74

3.7 Spatial Channel Model Application in SystemSimulations 74

3.8 Angle Spread Impact 77

References 80


4 Fixed Beam Smart Antenna Systems 83

4.1 Introduction 83

4.2 Conventional Sectorization 83

4.3 Limitations of Conventional Sectorization 88

4.4 Antenna Arrays Fundamentals 89

4.4.1 Broadside and End-Fire Arrays 91

4.4.2 Impact of Number of Elements 92

4.4.3 Impact of Element Spacing 93

4.4.4 First Null Beamwidth 96

4.4.5 Half-Power Beamwidth 97

4.4.6 Array Directivity 99

4.4.7 Array Gain 100

4.4.8 Trade-Off Analysis 100

4.4.9 Impact of Element Pattern 101

4.4.10 Planar Arrays 101

4.5 Beamforming 105

4.6 The Butler Matrix 107

4.7 Spatial Filtering with Beamformers 110

4.8 Switched Beam Systems 111

4.9 Multiple Fixed Beam Systems 113

4.10 Adaptive Cell Sectorization in CDMA Systems 114

References 116

Contents vii

5 Adaptive Array Systems 117

5.1 Uplink Processing 117

5.1.1 Diversity Techniques 117

5.1.2 Angle Diversity 118

5.1.3 Maximum Ratio Combining 121

5.1.4 Adaptive Beamforming 122

5.1.5 Fixed Multiple Beams Versus Adaptive Beamforming 130

5.2 Downlink Processing 132

5.2.1 Transmit Diversity Concepts 134

5.2.2 Transmit Diversity in 3G CDMA Standards 134

5.3 Downlink Beamforming 142

5.3.1 Spatial Signature-Based Beamforming 145

5.3.2 DOA-Based Beamforming 146

5.3.3 Maximum SNR 147

5.4 Conclusion 149

References 151


6 Smart Antenna Receivers and Algorithms for RadioBase Stations 159

6.1 Reference Signal Methods 159

6.1.1 The Least Mean Square Algorithm 159

6.1.2 The Recursive Least Squares Algorithm 161

6.1.3 Blind Adaptive Beamforming 161

6.1.4 Least Squares 161

6.1.5 Constant Modulus Algorithm 162

6.1.6 Decision-Directed Algorithm 162

6.1.7 Cyclostationary Algorithms 163

6.1.8 Conjugate Gradient Algorithm 164

6.1.9 Lagrange Multiplier Method 167

6.1.10 Comparison of Adaptive Algorithms 169

6.2 Neural Network DOA-Based Beamforming 170

6.2.1 Generation of Training Data 174

6.2.2 Performance Phase of the RBFNN 174

viii Smart Antenna Engineering

6.3 Angle Spread Impact on Optimum Beamforming 175

6.4 Downlink Beamforming 181

6.5 Vector Rake Receivers 182

6.6 Channel Estimation 183

6.7 Beamforming 184

6.8 Conclusion 185

References 186

7 Coverage and Capacity Improvements in 3G Networks 191

7.1 Introduction 191

7.2 Link Budgets and Coverage 192

7.2.1 Mobile Station Parameters 192

7.2.2 Base Station Parameters 193

7.2.3 System Parameters 193

7.2.4 Margins 193

7.2.5 Other Parameters 193

7.2.6 Fade Margin 194

7.2.7 Confidence (Cell Area) 195

7.2.8 CDMA Traffic Loading 196

7.3 Voice Services 197

7.3.1 Uplink Budgets 198

7.3.2 Downlink Budgets 198

7.4 Data Applications 203

7.5 Limiting Links for Coverage and Capacity 209

7.5.1 Coverage Limited Scenarios 210

7.5.2 Capacity Limited Scenarios 211

7.6 Smart Antennas Impact on Uplink Coverage andCapacity 211

7.6.1 Smart Antenna Impact on Downlink Capacity 216

7.7 Conclusions 226

References 227

Contents ix

8 Smart Antennas System Aspects 231


8.2 Third Generation Air Interfaces and Protocol Stacks 232

8.3 Physical Layer 233

8.3.1 Data Multiplexing 233

8.3.2 Transmit Chain UL/RL PN Scrambling/Spreading 235

8.3.3 DL/FL Physical Channel Formatting 235

8.4 Mobile Call States 237

8.4.1 WCDMA 237

8.4.2 CDMA2000 237

8.5 Mobility Procedures to Support High-Speed DataTransfer 238

8.5.1 Cell_FACH State or Control Hold Mode 240

8.5.2 Idle, Cell_PCH, or URA_PCH States 240

8.6 Procedures to Reestablish High-Speed Data Transfer 240

8.6.1 Cell_FACH State or Control Hold Mode 240

8.6.2 Idle Mode, Cell_PCH, or URA_PCH States 240

8.7 Packet Data Services 240

8.7.1 WCDMA Approach 241

8.7.2 CDMA2000 Approach 241

8.8 Pilot Channels 241

8.8.1 CDMA2000 241

8.8.2 WCDMA 243

8.9 Channels Applicable for Downlink Beamforming 243

8.10 Overview of Major Radio Network Algorithms 244

8.10.1 Power Control 244

8.10.2 Initial Power Setting 245

8.10.3 Admission Control 246

8.10.4 Congestion Control 246

8.10.5 Soft/Softer Handoff 246

8.10.6 Hard Handoff 247

8.11 System Impact of Advanced Spatial Techniques 247

8.11.1 Transmit Diversity 247

x Smart Antenna Engineering

8.11.2 Fixed Beam Approach 248

8.12 Beam Steering/Adaptive Beamforming 258

8.12.1 Channel Estimation at the Mobile 259

8.12.2 Advantages and Disadvantages 260

8.12.3 Uplink Beamforming 260

8.13 Conclusion 261

References 262

9 Mobile Stations’ Smart Antennas 265


9.2 Multiple-Antenna MS Design 268

9.3 Combining Techniques 272

9.3.1 Selection (Switched) Diversity 272

9.3.2 Maximal Ratio Combining 272

9.4 Adaptive Beamforming or Optimum Combining 272

9.5 RAKE Receiver Size 278

9.6 Mutual Coupling Effects 279

9.7 Dual-Antenna Performance Improvements 280

9.8 Downlink Capacity Gains 284

9.9 Conclusions 286

References 287

10 MIMO Systems 289


10.2 Principles of MIMO Systems 290

10.2.1 SISO 291

10.2.2 SIMO 291

10.2.3 MISO 292

10.2.4 MIMO 293

10.3 Transmission Strategies 295

10.3.1 Water Filling 296

10.3.2 Uniform Power Allocation 296

Contents xi

10.3.3 Beamforming 297

10.3.4 Beam Steering 297

10.4 MIMO Approaches 297

10.5 MIMO Advantages and Key Performance Issues 298

10.6 RF Propagation Characterization 299

10.7 SINR Environment 299

10.8 Spatial Multiplexing 300

10.9 Conclusion 302

References 303

List of Acronyms 305

About the Author 311

Index 313

xii Smart Antenna Engineering

Preface

Mobile and wireless communications systems are becoming increasingly morecomplex in an effort to cope with the growing demand for more supportablepeak data rates, coverage requirements, and capacity objectives, as well as excit-ing new applications such as wireless multimedia and anywhere-anytime mobileInternet access. Although new air interface standards and access technologiessuch as code division multiple access (CDMA2000), wideband code divisionmultiple access (WCDMA), and their evolutions, including evolution data opti-mized (EV-DO) and high-speed downlink packet access (HSDPA), promise tomeet these requirements with data rates up to several megabits per second, this isoften achievable only under ideal channel conditions—assumptions are rarelyencountered in real systems deployment. Smart antennas have great potential inovercoming the impairments of these systems by exploiting the spatial domainto reduce the effects of interference, extend the range and coverage of wirelessnetworks, increase system capacity, and achievable data throughout.

The area of smart antennas application in wireless communications hasreceived increased attention both in the wireless industry and academia for thepast few years. It is an interdisciplinary topic that requires knowledge and skillsin areas such as antenna arrays, signal processing, digital communications, radiofrequency (RF) engineering, and wave propagation. Today, a large body of liter-ature about the topic exists, although much of this is in the form of complexresearch papers published across a multitude of technical journals, magazines,and conference proceedings, making it very difficult for a practicing engineer todevelop the skills required for a successful design in a reasonable amount oftime. With that in mind, this book attempts to close the gap by consolidatingand presenting the principles of smart antennas along with the issues associated

xiii

with their application in modern communications systems in an easy-to-followformat. The book’s purpose is to explain the principles and techniques of smartantennas application in wireless and mobile communications systems. It pres-ents topics and issues in the design of advanced antennas systems in aneasy-to-follow methodology. The book is intended for graduate students in elec-trical engineering, practicing communications engineers, engineering and prod-uct managers, and wireless systems designers. It is intended to provide a usefuland needed reference in one place and cover a collection of topics necessary forsuccessful application of smart antennas in wireless systems.

The book begins in Chapter 1 with a brief history of wireless communica-tions systems and their drive to achieve increasing demands in terms of coverageand capacity. In Chapter 2, the effects of cochannel interference, multiple accessinterference, and other impairments affecting existing and future multiple accesstechniques of 2.5 and third generation (3G) wireless systems are discussed toshow how they prevent these systems from achieving their full potential of rangeand system capacity. Models for the mobile radio propagation channel are inte-gral tools that allow system designers to evaluate the benefits of different mea-sures for enhancing system performance. The coverage of smart antennas wouldnot be complete without addressing models that take the spatial domain intoaccount. In Chapter 3, shortcomings of conventional models will be outlined,along with a description of spatial directional channel models adopted by theindustry’s standards bodies. Interference reduction with smart antennas offersan efficient way to reduce the interference in mobile communications systemsthrough the use of narrow beams directed to a cluster of users or an individualuser while, at the same time, steering nulls toward interfering users. Smartantennas could be divided into two major types, fixed multiple beams and adap-tive array (AA) systems. A detailed explanation of these two approaches, alongwith their advantages and drawbacks, will be covered in Chapters 4 and 5. First,we will provide an overview of the fundamentals of antenna arrays and thenshow how these concepts tie into schemes like the Butler matrix and adaptivebeamforming. We will also discuss diversity techniques and other methodsapplicable to both the uplink and downlink of wireless mobile communicationssystems. A daunting task facing any smart antennas developer is selecting thereceiver structure and adaptive algorithms most suitable for the application inhand. Today, a large number of proposed methods and technical solutions exist.A comprehensive classification of smart antennas algorithms along with themain implementation issues and trade-offs is presented in Chapter 6, as well assome comparison between the different techniques. In Chapter 7, a section onsystem performance improvements demonstrates the impact of using smartantennas at the radio base station and potential improvements in terms of cover-age and capacity of mobile communications networks. In Chapter 8, we willaddress the systems aspects of smart antennas and their interaction with various

xiv Smart Antenna Engineering

network control algorithms such as admission control, power control, and radioresource management. The application of antenna arrays in handsets is dis-cussed in Chapter 9. Finally, the book concludes with a brief overview of multi-ple input multiple output (MIMO) systems, which combine antenna arrays atboth the receive and transmit side to create parallel spatial channels that dramat-ically increase spectral efficiency and system capacity.

Although practicing engineers and designers as well as engineering andproduct managers are the primary audience for this book, it can be easilyadopted as a graduate course textbook in smart antenna applications in mobilecommunications systems.

Preface xv

Acknowledgments

First, I would like to thank God for the knowledge and strength that made thisproject possible. I would also like to acknowledge and thank my family andfriends for their support throughout this book. In addition, I would like tothank Bo Hagerman, Soren Andersson from Ericsson Research CorporateUnit, and Patrick Lundqvist from Ericsson Wireless Communications, Inc.for their valuable insights and numerous discussions in adaptive antennas forwireless mobile communications. Special thanks go to Professor ChristosChristodoulou, chair of the electrical and computer engineering department atthe University of New Mexico, for his encouragement and inspiration, whichmade this work possible. I would also like to thank Dr. Said El Khamy and Dr.Hassan El Kamshoushi from the University of Alexandria in Egypt for theirguidance in my early work in adaptive antennas. In addition, I would like toexpress thanks to Qualcomm Inc. for permission to use some illustrations in thisbook. I would also like to acknowledge the publishing team at Artech House fortheir guidance and assistance, as well as the reviewer of this project.

I welcome any comments and suggestions for improvement or changesthat could be implemented in possible future editions and can be reached [email protected].

xvii

1Introduction

Adaptive antennas have been used for decades in areas such as radars, satellitecommunications, remote sensing, and direction finding, to name a few. Forinstance, radar and secure communications systems take advantage of the abilityof these antennas to adapt to the operating environment to combat jamming.Satellite communications systems have used multiple beam and spot beamantennas for years to tailor their coverage to specific geographic locations. Eachof these applications is associated with its own unique set of challenges, such asthe channel in which the system operates, the propagation environment, sourcesof interference, and noise or jamming. In addition, the end goal for which theadaptive antenna is used affects the selection of the type of array, size, adaptivealgorithms, and integration with other system components. In this chapter weprovide a summary of the status of current mobile cellular communications sys-tems, their various evolution paths, mobile systems growth potentials, as well asan introductory discussion of the benefits and use of smart antennas in 3Gcellular communications systems.

1.1 Wireless Mobile Communications Systems

In the 1980s and 1990s, wireless cellular and personal communications systems(PCS) began to flourish with the advent of second generation mobile communi-cations systems, or simply 2G, to cope with increasing demands. Early mobilecommunications systems were based on analog technologies that used frequencydivision multiple access (FDMA). In multiple access, a number of users access orshare the resources of a common source. In FDMA systems, the available spec-trum is divided into channels of specific bandwidth [30 kHz in the case of

1

advanced mobile phone service (AMPS), the North American analog standard]and users are assigned a pair of these channels for bidirectional communicationswith a base station (BS). In other words, the resource shared by all users is thebandwidth. Since the available spectrum is finite, there is a fundamental limit onthe capacity or number of users that can be served by a cell. It is possible to reusethe whole available spectrum in each cell to maximize the capacity—this iscalled reuse factor of one. However, the base station transmit power required tocommunicate with all these users plus additional margins to overcome fadingcaused by multipath creates so much cochannel interference to users in neigh-boring cells that the signal quality is significantly degraded. To reduce this inter-ference to acceptable levels that support a given signal quality, the number ofchannels assigned to each cell must be decreased—in other words, the reusefactor must be increased. This, of course, will lower the overall system capacity.

Engineers then turned to technologies based on digital techniques to solvethis trade-off between capacity and interference. In time division multiple access(TDMA), each user is assigned the entire resource at specific time slots. In thiscase, the shared resource is time. Global systems for mobiles (GSM) are based onthis technology and it uses channels with bandwidth of 200 kHz. InTDMA-based systems, frequency planning plays an important role in balancingsystem capacity versus cochannel interference. Another multiple access tech-nique based on spread spectrum technology is CDMA, in which the codedomain is shared among users as defined in the IS-95 standard. One major dif-ference between CDMA systems and other multiple access technologies is theirre-use factor of one, which enables them to offer higher capacities. This is possi-ble because of the unique way in which CDMA handles interference. A combi-nation of pseudonoise (PN) sequences and orthogonal codes are used to spreadand channelize the base station and user’s data. Spreading the signal to a muchwider bandwidth helps reduce the power levels and makes each signal appear asbackground noise to other users. This scheme allows a large number of users tosimultaneously share the same 1.25-MHz carrier. In addition to spreading,CDMA systems use power control techniques to maintain the interference inthe system at the acceptable levels required to satisfy the signal or radio linkquality. Furthermore, CDMA systems take advantage of multipath through theuse of RAKE receivers to combat fading. Due to the explosion of mobile com-munications demand and the increasing shift to offer new and advanced servicesbased on high-speed data rates, third generation technologies were developed.The main goals of 3G systems are to increase the voice capacity, improve mixedvoice and data services, and offer peak data rates of up to 2 Mbps. There are cur-rently two major 3G technologies, both based on CDMA. These are widebandCDMA or WCDMA, also known as universal mobile telecommunications sys-tem (UMTS) [1], and CDMA2000 [2]. Peak data rates of 384 Kbps are beingachieved in commercially deployed WCDMA networks, whereas the WCDMA

2 Smart Antenna Engineering

evolution path with HSDPA and high-speed uplink packet access (HSUPA)extends the peak rate to 14.4 Mbps on the downlink and more than 4 Mbps onthe uplink, respectively, in a 5-MHz carrier. Similarly, peak data rates of 153.6Kbps in a 1.25-MHz carrier are being achieved on the currently deployedCDMA2000 1x networks. The CDMA2000 1xEV-DO standard furtherextends the peak rates to 3.1 Mbps and 1.8 Mbps on the downlink and uplink,respectively. Both 1xEV-DO and HSDPA technologies were developed to sig-nificantly increase the peak data rates to meet the rapidly growing demand forhigh-speed data applications. The basic concept behind both technologies is thesame, namely the introduction of new features such as adaptive modulation andcoding (AMC), short frames, multicode operation, fast L1 hybrid automaticrepeat request (HARQ), and base station scheduling. In fact, these featuresreplace the two basic CDMA features, namely variable spreading factor (VSF)codes and fast power control by adaptive rate control based on channelconditions.

AMC is a fundamental feature of HSDPA and 1xEV-DO. It consists ofcontinuously optimizing the code rate, the modulation scheme, the number ofcodes employed, and the transmit power per code based on the channel qualityreported [channel quality indicator (CQI) feedback] by the mobile station. Toachieve very high data rates, higher order modulation schemes such as 16 QAMis added to the existing quadrature phase shift keying (QPSK) modulation usedfor R’99 WCDMA and CDMA20001x channels. Different combinations ofmodulation and the channel coding-rate can be used to provide different peakdata rates. Essentially, when targeting a given level of reliability, users experienc-ing more favorable channel conditions (e.g., closer to the base station) will beallocated higher data rates. According to industry bodies, at the beginning of2005, global subscriptions to 3G/UMTS networks reached 16 million on morethan 60 networks, whereas more than 180 million subscribers are usingCDMA2000 on approximately 120 networks.

1.2 Global Mobile Market Growth

At the end of 2004, worldwide cellular subscriptions passed the 1.4 billion markand the rapid growth is expected to last for many years. The chart in Figure 1.1shows that the number of worldwide cellular users is expected to reach nearly2.5 billion by 2010 [3], while Figure 1.2 provides a breakdown of this forecastfor CDMA technologies. Note that this breakdown does not include GSM andEDGE subscribers.

This continued growth and evolution in mobile usage is driven by dataservices such as short message service (SMS), multimedia messaging service(MMS), downloadable ring-tones, images and games, news and information

Introduction 3

sources, mobile chat sites, and Web portals. It is anticipated that voice serviceswill still significantly contribute to revenue streams along with new 3G enabledservices, including personalized access to information and entertainment ser-vices, mobile access to the Internet and corporate networks, location based ser-vices, and rich voice, which is the simultaneous transmission of photos,graphics, video, maps, documents, and other forms of data with pure voice. Thechart in Figure 1.3 shows how the worldwide mobile voice traffic is expected toincrease during the next few years to nearly three times the current levels by2010.

1.3 Alternatives for Meeting Data Demand

Different wireless service providers have different evolution paths with differenttechnology choices to upgrade their 2G networks to third generation systemsdefined in the IMT-2000 standard of the International Telecommunications


3000

2500

1500

2000

1000

500

0

Wordwide cellular users (millions)

2004 2005 2006 2007 2008 2009 2010

Rest of world

Central and Eastern Europe

Central and Latin America

Asia Pacific

Western Europe

North America

Figure 1.1 Worldwide cellular users forecast [3].

CDMA worldwide cellular users (millions)1400

1200

1000

800

600

400

200

02004 2005 2006 2007 2008 2009 2010

WCDMA

CDMA2000 1xEV

CDMA2000

CDMAOne

Figure 1.2 Worldwide CDMA cellular users forecast [3].

Union (ITU). The main evolution paths for GSM and CDMAOne operatorsare shown in Figure 1.4. A number of GSM operators have chosen a migrationpath that involves upgrading their networks to GPRS and EDGE as an interimstep before a full WCDMA migration while others have chosen to evolve theirnetworks directly to WCDMA. CDMAOne operators have a somewhatsmoother migration path with CDMA2000. Eventually, to meet the growingdemand for voice and data capacity, most current 2G networks will be upgradedto use CDMA. Figure 1.5 shows the global 3G cellular users forecast bytechnology until 2010.

Introduction 5

Voice traffic growth350%

300%

250%

200%

150%

100%

50%

0%2004 2005 2006 2007 2008 2009 2010

Figure 1.3 Projected voice traffic growth [3].

2G 2.5G 3G Evolved 3G

Voice centric9.6 Kbps Data

40 Kbps

Data120 Kbps

Voice + data384 Kbps

DataDL: 14.4 Mbps/UL: 384Kbps

DataDL:14.4/UL: 4.3 Mbps

Voice centric9.6/14.4 Kbps

GSM GPRS WCDMAR'99 HSDPA HSUPA

EDGE

cdmaOne CDMA20001xRTT

1xEV-DORev. 0

1xEV-DORev. A

1xEV-DV

Voice + data153.6 Kbps

DataDL: 2.4 Mbps/UL: 153.6 Kbps

DataDL: 3.1 Mbps/UL: 1.8 Mbps

Figure 1.4 2G evolution paths toward 3G.

1.4 Technology Peak Rates and Throughput

As we can see from Figure 1.4, different technologies support different peak datarates. The peak rate is the maximum transmission speed an individual user mayexperience under ideal conditions (i.e., it only affects the user experience). Datathroughput, on the other hand, is a far more important metric for performance.Sector throughput is the average total capacity available to multiple users,whereas user throughput is the average data rate a user may experience. As thesector throughput increases, each sector can handle higher volumes of data, thenetwork requires fewer sites, and, consequently, the capital and operationalexpenses are also reduced. Table 1.1 compares the peak data rates andthroughput for different 3G technologies.


3G worldwide users (millions)1400

1200

1000

800

600

400

200

02004 2005 2006 2007 2008 2009 2010

WCDMA

CDMA2000 1x

CDMA2000 1xEV

Figure 1.5 3G cellular users forecast by technology.

Table 1.13G Technology Comparisons

Technology

CarrierBandwidth/Spectrum (MHz)

DownlinkPeak DataRate (Kbps)

Average UserThroughput(Kbps)

CDMA2000 1x 1.25/1.25 153.6 60–80

CDMA20001xEV-DO

Rev.0

1.25/1.25 2,458 300–500

WCDMA 3.84/5 384 220–320

HSDPA 3.84 -/- 5 14,400 550–1100

For CDMA2000 1x and CDMA2000 1xEV-DO, user throughputs listedin Table 1.1 are based on promotional material from North American operatorsand on real network deployments. WCDMA and HSDPA user throughputs arebased on results from [4-8]. Unlike EV-DO systems, there are no commercialdeployments of HSDPA systems yet; these systems are expected to deploy in late2005 and into 2006. The user throughput for HSDPA is based on simulationdata [5]. Moreover, the choice of scheduler significantly affects the throughputof both 1xEV-DO and HSDPA because of the adaptive modulation and codingnature of the technologies. For instance, one popular scheduler called the pro-portional fairness (PF) schedules users according to the ratio between theirinstantaneous achievable data rate and their average served data rate. This resultsin all users having equal probability of being served even though they may expe-rience very different average channel quality. This scheme provides a good bal-ance between system throughput and fairness. Other schedulers will bediscussed in Chapter 2.

1.5 Why Smart Antennas?

Achieving the peak data rates specified in each standard in a real system remainsvery unlikely because it would require an unloaded system serving a single userto be extremely close to the base station. This leads to two questions: why theincreased interest in smart antennas—a more attractive name for adaptiveantennas—and how are they being considered as a viable technology for applica-tions such as mobile communications? As we have seen, operators are faced withincreasing capacity demands for both voice and data services. Although various3G technologies offer higher data rates and double voice capacity comparedwith their 2G counterparts, their actual performance is still susceptible to inter-ference, and adverse channel conditions created by multipath propagation andsystem loading. As such, smart antennas techniques can complement 3G sys-tems and improve their performance by alleviating and reducing the degrada-tion caused by the aforementioned factors. In fact, because of their nature,technologies such as HSDPA and 1xEV-DO can greatly benefit from smartantennas since any improvement in the SNR experienced by the users woulddirectly translate to better throughput for individual users as well as increasedsector throughput that can support higher capacities.

1.6 Benefits of Smart Antennas

It is a fact that current technologies have nearly maximized the use of temporaland spectral techniques to improve capacity and data transfer speeds. This leaves

Introduction 7

an additional parameter that has not been fully tapped yet, namely space. Inspace division multiple access (SDMA), a user or cluster of users are assigned adedicated narrow beam that tracks their movement across the cell, adapting tothe constantly changing radio environment. The obvious advantage of thisapproach is its applicability to any multiple access technique. Wireless systemdesign and planning involve the optimization of two major components, cover-age and capacity through the manipulation and control of power, interference,and noise. To that extent, smart antennas offer substantial benefits to the designof wireless mobile communications systems, which can be summarized asfollows:

• Increased antenna gain: this helps increase the base station range andcoverage, extends battery life, and allows for smaller and lighter handsetdesigns.

• Interference rejection: antenna pattern nulls can be generated towardinterference sources. On the reverse link or uplink this reduces theinterference seen by the base station. It also reduces the amount ofinterference spread in the system on the forward link or downlink. Suchimprovements in the carrier to interference ratio C/I lead to increasedcapacity.

• Diversity: composite information from the array can be used to mini-mize fading and other undesirable effects of multipath propagation. Inaddition to spatial and polarization diversity, antenna arrays also allowthe use of angular diversity.

As with any other adaptive antennas application, the nature of the systemin which they are employed, the conditions under which they operate, and theresults they are intended to achieve all have to be considered when a smartantenna system design is incorporated in a specific wireless system. Figure 1.6shows a system overview that describes some of the involved factors when weconsider a smart antenna design for mobile communications systems. Subse-quent chapters will provide more details and analysis regarding these areas andhow they affect the selection, design, and performance of a smart antennasystem.

1.7 Types of Smart Antennas

Sectorization schemes, which attempt to reduce interference and increase capac-ity, are the most commonly used spatial technique that have been used in cur-rent mobile communications systems for years. Cells are broken into three or six


sectors with dedicated antennas and RF paths. Increasing the amount ofsectorization reduces the interference seen by the desired signal. One drawbackof current sectorization techniques is that their efficiency decreases as the num-ber of sectors increases due to antenna pattern overlap. Furthermore, increasingthe number of sectors increases the handoffs the mobile experiences while mov-ing across the cell. Compare this technique to that of a narrow beam beingdirected towards a desired user. It is clear that some interference that would havebeen seen by the existing 120° sector antenna will be outside the beamwidth ofthe array. Any reduction in the interference level translates into system capacityimprovements. Smart antennas could be divided into two major types, fixedmultiple beams and AA systems. Both systems attempt to increase gain in thedirection of the user. This could be achieved by directing the main lobe, withincreased gain, in the direction of the user, and nulls in the directions of theinterference [9, 10].

1.8 Switched and Fixed Beam Antennas

The switched beam method is considered an extension of the current cellularsectorization scheme. The switched beam approach further subdivides themacro-sectors into several micro-sectors. Each micro-sector contains a predeter-mined fixed beam pattern with the greatest gain placed in the center of the

Introduction 9

Network planning

Radio network control

Structure andalgorithms

Networkdependentparameters

Air interfaceparameters

Radionetworkprotocols

..

.

..

.

Propagation environmentspatial channel modelinginterference environment

Transmitter Receiver

Channel

Structure andalgorithms

Networkdependentparameters

Air interfaceparameters

Radionetworkprotocols

Figure 1.6 Smart antenna system overview.

beam. When a mobile user is in the vicinity of a micro-sector, the switchedbeam system selects the beam containing the strongest signal. During the call,the system monitors the signal strength and switches to other fixed beams ifrequired. Better performance can be achieved with integrated embedded systemsof fixed multibeam antennas, which can enhance signal detection on the uplinkby making use of the signals from all the available paths in the beams followedby maximum ratio combining (MRC) [11]. The beam receiving the most powerin the uplink can be used to transmit to the desired mobile on the downlink.

1.9 Adaptive Arrays

The main advantage of adaptive antenna arrays compared with switched beamantennas is their ability to steer beams towards desired users and nulls towardinterfering signals as they move around a sector. Several beamformingapproaches exist with varying degrees of complexity. A conventionalbeamformer or delay-and-sum beamformer has all the weights of equal magni-tudes. To steer the array in a particular direction, the phases are selected appro-priately. In order to be able to null an interfering signal, the null-steeringbeamformer can be used to cancel a plane wave arriving from a known directionproducing a null in the response pattern at this direction. When the number ofinterferers becomes large, such as in the case of IS-95 based systems, thisbeamformer might not be a practical approach. The well-known minimum vari-ance distortionless response (MVDR) beamformer attempts to minimize thetotal output noise while keeping the output signal constant in the direction ofthe desired user. This is the same as maximizing the output SNR. For an M-ele-ment array with M degrees of freedom, the number of interferers must be lessthan or equal to M – 2, since one has been used by the constraint in the lookdirection. This may not be true in a mobile communications environment withmultipath arrivals, and the array beamformer may not be able to achieve themaximization of the output SNR by suppressing every interference source.Some a priori knowledge of the desired signal such as the direction of arrival(DOA) is required by the MVDR beamformer. Since in the MVDR approachthe weight vector that minimizes the output power is a function of the spatialcorrelation matrix, some degree of coherency between the uplink and downlinkis needed to provide an estimate of the correlation matrix for transmission. Inthe minimum mean square error (MMSE) approach a minimization of thesquare of the difference between the array output and a reference signal results inthe weight vector that maximizes the signal quality. Since this approach relies onthe inversion of the covariance matrix, its complexity is very high. The maxi-mum likelihood (ML) principle attempts to estimate the data sequence that wasmost likely sent based on the received or observed data. Other spatial techniques


include transmit diversity and MIMO systems. In MIMO systems, antennaarrays are used in the transmitter as well as in the receiver, and the system createsmultiple parallel channels that significantly increase the supportable data rates.Figure 1.7 compares the performance improvement expected from major smartantenna techniques with their complexity.

References

[1] Third Generation Partnership Project, http://www.3gpp.org.

[2] Third Generation Partnership Project2, http://www.3gpp2.org.

[3] “Worldwide Cellular User Forecasts (2004–2010),” Strategy Analytics, December 2004.

[4] “The Economics of Wireless Mobile Data,” Qualcomm Inc, http://www.qualcomm.com.

[5] “Data Capabilities: GPRS to HSDPA,” Rysavy Research, September 2004, http://www.rysavy.com.

[6] Holma, H., and A. Toskala, WCDMA for UMTS: Radio Access for Third GenerationMobile Communications, 3rd ed., New York: John Wiley & Sons, 2004.

[7] “HSDPA for Improved Downlink Data Transfer,” Qualcomm CDMA Technologies,October 2004, http://www.cdmatech.com.

[8] “Nokia High Speed Packet Access Solution,” ZD Net UK, http://whitepapers.zdnet.co.uk/.

Introduction 11

System complexity

Basicsectorization

Higher ordersectorization

Transmitdiversity

Fixed multi-beamantennas

Beam steering(beam shaping,adaptive nulling)

MIMO systems

Perf

orm

ance

imp

rove

men

ts

Figure 1.7 Comparison of major spatial techniques.

[9] Rappaport, T. S., (ed.), Smart Antennas: Adaptive Arrays, Algorithms and Wireless PositionLocation, New York: IEEE Press, 1998.

[10] Tsoulos, G.V., (ed.), “Adaptive Antennas for Wireless Communications,” IEEE Press,2001.

[11] Göransson, B., B. Hagerman, and J. Barta, “Adaptive Antennas in WCDMA Sys-tems—Link Level Simulation Results Based on Typical User Scenarios,” IEEE VehicularTechnology Conference, Boston, MA, September 2000.


2Multiple Access Techniques for 2G and3G Systems

2.1 Introduction

Evaluating the various design choices of different smart antennas architectures,algorithms, and performance trade-offs when applied to modern mobile cellularcommunications systems requires knowledge and understanding of core accesstechnologies as well as the impairments facing different systems. This chapterpresents the concepts of FDMA, TDMA, and CDMA and describes the maindifferences between these access technologies. An overview of the frequencyreuse concept and cochannel interference, critical to the network design of somesecond generation mobile communications systems, is provided. Since all 3Gtechnologies are based on CDMA, there will be greater emphasis on this tech-nology. When evaluating performance issues, two main components are usuallyconsidered, the link level performance and system level performance. In linklevel performance, we are mainly concerned with a single link between a mobilestation and the base station; this link is typically based on the physical layerstructure of the air interface. The physical layer is the layer that carries the actualRF transmissions. On the other hand, in system level performance the impact ofthe upper layers and their interactions with the physical layer has to be takeninto consideration. Functions performed by the upper layers include radioresource management, admission control, and so on.

This chapter has been divided as follows. First, in Section 2.2 we discussthe concepts of FDMA and TDMA systems and how frequency reuse is appliedin the design of mobile networks also briefly cover cochannel interference. Thefundamentals of CDMA technologies are discussed in Sections 2.3 and 2.4,

13

along with systems aspects such as RAKE receiver, power control, and softhandoff, as well as an overview of the IS-95 air interface. In Section 2.5, weintroduce third generation systems and summarize the CDMA2000 andWCDMA standards. An overview of how a CDMA phone works and the differ-ent procedures employed to acquire the system and complete a mobile call areintroduced in Section 2.6. Since the main motivation for using smart antennaswith 3G systems is to improve their coverage and capacity performance, thechapter concludes with Section 2.7, in which the factors affecting CDMAcapacity are presented and the coverage versus capacity trade-off is discussedusing simple models. In a later chapter, a more complex and specific discussioninvolving this trade-off will be presented to provide tools to evaluate the differ-ent smart antennas gains.

2.2 Multiple Access Wireless Communications

In cellular and PCS wireless communications systems a multitude of users accessand share network resources (frequency bandwidth) to obtain different types ofservices, including voice, messaging, and data. The goals of these multiple accesscommunications systems are to provide communications services in a near-uni-versal geographical coverage while minimizing both subscriber stations and net-work equipment, deployment, and operational costs. Because regulatoryagencies have allocated limited bandwidth to these services, a crucial goal ofthese solutions is to achieve high spectral efficiency, traditionally measured inErlangs/megahertz/unit service area for voice applications and in bits/sec-ond/megahertz/unit service area for data applications. The cellular concept pio-neered by Bell Labs in the 1970s makes use of multiple fixed stations, or cellsthat each serve a number of mobile subscribers within a limited geographicalarea. When a subscriber moves between cells, over-the-air messaging is used tohandoff the call between cells, ensuring its continuity. The first such system inNorth America was called AMPS. Similar analog systems were also deployed indifferent parts of the world, including the Nordic Mobile Telephone (NMT) inScandinavia, and the Total Access Communications System (TACS) used in theUnited Kingdom, China, and other countries. The spectrum chosen for thesesystems was in the 800–900-MHz band. The frequency band allotted to eachsystem was then divided according to a scheme called FDMA.

2.2.1 FDMA Systems

In wireless mobile communications systems subscribers share a commonresource such as time, frequency spectrum, power, or code. This is referred to asaccess technology or channelization. This leads to the generation of interferencein the system, which affects signal quality. The degree to which system


performance is affected by interference actually depends on the access technol-ogy used to separate the users in the network. In FDMA, the available spectrumis divided among users by assigning different frequencies to various users, asshown in Figure 2.1. With FDMA systems, a user is assigned a 30-kHz or a25-kHz pair of frequencies for the forward link (downlink) and the reverse link(uplink) throughout a call. To maintain the interference between the two linksat a minimum, the frequency pair is separated by, for example, 45 MHz and 80MHz in North American cellular and PCS systems, respectively. The FDMAscheme could be equally applied to analog and digital communications systems.

2.2.2 TDMA Systems

TDMA is a digital transmission technology that allows a number of users toaccess a single RF channel while reducing interference by allocating unique timeslots to each user within each channel. In TDMA systems channelization is pro-vided first by dividing the frequency among the users, just like in FDMA, andthen again by dividing users in time by assigning users different time slots. Thistransmission scheme multiplexes three signals over a single channel. TheTDMA standard for cellular divides a single channel into six time slots, witheach signal using two slots, providing a 3 to 1 gain in capacity over AMPS. Eachcaller is assigned a specific time slot for transmission, shown in Figure 2.2. InUS TDMA (IS-54), a 30-kHz channel is further divided into three time slots,

Multiple Access Techniques for 2G and 3G Systems 15

f1 f2 f3 f4 fn

Power

Frequency

Time

Figure 2.1 The FDMA concept.

which increases the number of simultaneous users per channel to three. In theEuropean TDMA version, or GSM, a 200-kHz channel is divided among eightusers. TDMA relies on the fact that the audio signal has been digitized; that is,divided into a number of milliseconds-long packets. It allocates a single fre-quency channel for a short time and then moves to another channel. The digitalsamples from a single transmitter occupy different time slots in several bands atthe same time. One of the disadvantages of TDMA is that each user has a prede-fined time slot and users handing off from one cell to another are not allotted atime slot. Thus, if all the time slots in a cell are already occupied, no additionalcalls are allowed. This represents a hard limit on the cell capacity. Anotherproblem with TDMA is that it is subjected to multipath distortion.

2.2.3 Frequency Reuse

In cellular and PCS systems, a cell’s coverage is typically represented by a hexa-gon when omnidirectional antennas with constant transmit power are used atthe base station. As we have seen with FDMA and TDMA systems, the availablefrequency spectrum is divided among the users in the network. Now, let usassume two adjacent cells with two users assigned frequency f1. As these mobilestations move closer together, their use of a frequency f1 will begin to createinterference.To overcome this problem, a process called frequency planning isimplemented, where a group of frequencies are reused in cells that are separatedfrom one another by distances large enough to maintain the interference at


Frequency

Power

Time

Figure 2.2 The TDMA concept.

acceptable levels. Frequency reuse is the term that describes how frequencies areallocated throughout the system as a result of frequency planning. Assume a cel-lular system has F total frequency pairs or duplex channels available for users. Byallocating each cell a group of k channels and dividing the F channels among Ncells, we get

F kN= (2.1)

It follows that the cluster of N cells use the complete available band of fre-quencies. By replicating this cluster several times across the whole system, wecan see that the system capacity will be proportional to N, which is also referredto as cluster size. Since each cell is assigned 1/N of the total channels, this factoris called the frequency reuse factor. Since the available spectrum is finite, there isa fundamental limit on the capacity or number of users that can be served by acell. It is possible to reuse the whole available spectrum in each cell to maximizethe capacity; this is called reuse factor of one. However, the base station transmitpower required to communicate with all these users plus additional margins toovercome fading caused by multipath creates so much cochannel interference tousers in neighboring cells that the signal quality is significantly degraded. Toreduce this interference to acceptable levels that support a given signal quality,the number of channels assigned to each cell must be decreased; in other words,the reuse factor must be increased. This, of course, will lower the overall systemcapacity. Typical cellular reuse assumes N = 7 sets of channels are used, one setin each cell. This seven-cell building block is then repeated over the service area,as shown in Figure 2.3. The design ensures that there are no adjacent cells usingthe same channel (frequency). Several N-way reuse patterns have been deployedin different networks, including the above seven-way reuse. To calculate thecapacity of an N-way reuse pattern, let us consider a 12.5-MHz band in whichwe need to deploy a cellular AMPS system. The total number of available chan-nels with K = 7 becomes

CapacityMHz

KHzchannels=

∗=

12 5

30 757

.

That is, there are approximately 57 AMPS channels available per cell.TDMA systems use the same frequency reuse concept as well but their capacityis higher than that provided by analog systems.

The capacity derived above assumes that the cells are usingomnidirectional antennas. In practice, cell sites are sectorized, usually into threesectors (i.e., each site is equipped with three sets of directional antennas, withtheir azimuths separated by 120°). In practice, sectorization does not lead to an


increase in a sector’s capacity in AMPS. This is because the sector isolation,often no more than a few decibels, is insufficient to guarantee acceptably lowinterference. However, an increase in coverage is possible with sectorizationbecause of the increased gain of the directional antenna but there is no gain inthe reuse. The total cell capacity remains at 57 and the sector capacity becomes19 channels. With this scheme the overall reuse factor (sector-based) becomes K= 7 *3 = 21.

2.2.4 Cochannel Interference

In FDMA and TDMA-based systems, when signals from cells using the samefrequency group interfere with each other they create cochannel interference,which affects the signal quality and system performance. Therefore, these cellsmust be separated by some distance, which is referred to as cochannel separationD and is given by [1]

D NR= 3 (2.2)

Under the assumption that the cell sizes and cell transmit powers are thesame, cochannel interference becomes a function of the ratio of the separationdistance to the cell’s coverage distance or D/R, where R is the cell radius [2, 3].


57

6

1

2 4

3

57

6

1

2 4

3

57

6

1

2 4

3

57

6

1

2 4

3

57

6

1

2 4

3

RD

Figure 2.3 N = 7 frequency reuse plan.

This shows that reducing the cochannel interference requires larger cochannelseparations. Let K be the number of cochannel interfering cells, then the signalto interference ratio (SIR) could be approximated as

( )SIR

SI

D Ri

n

i

K= =−

=∑

1

1

(2.3)

where n is the path loss exponent. It can also be shown that most of thecochannel interference results from cells in the first tier. Based on the hexagonalcell shape, we get K = 6, assuming that the cochannel separations are the same,and using (2.2) we can rewrite (2.3) as follows

( ) ( )SI N N

n n= =− −

1

6 3

1

6 32

(2.4)

From (2.4) we can clearly see the trade-off that exists between the systemcapacity and cochannel interference. To illustrate this trade-off let us assumethat we have a 12.5-MHz spectrum available and a 30-kHz channel bandwidth.Figure 2.4 shows the relation between the cell’s capacity in terms of the numberof voice channels and the SIR versus N for n = 4. We can clearly see the trade-offbetween achieving a high-capacity design versus maintaining an acceptable SIR.Thus, smart antennas become a crucial tool in dealing with such issues, as wewill see in subsequent chapters.


Figure 2.4 Capacity and SIR versus cluster size.

2.2.5 CDMA Systems

As we have seen in previous sections, the most fundamental issue in wirelessmobile systems design is how to deal with interference between users. Oneapproach to mitigating interference is using the concept of slotting, in whicheach mobile user is assigned a frequency or time slot that he, and he alone, useswhile he is active, such as in FDMA- and TDMA-based systems. The drawbackof this approach is the reduced spectral efficiency inherent in the frequency reuseapproach because only a portion of the available spectrum can be used in a givencell at any given time. Another drawback is the need to change the frequencyplan when new base stations are added to cope with increased capacity demands.In CDMA, users are divided by the assignment of a unique code to each.Because users can be identified by their unique code, there is no need to dividethe spectrum in either frequency or time and all users in a CDMA system aregiven access to the system at the same time and on the same frequency. This isshown in Figure 2.5, where a number of users share the same RF band using dif-ferent codes.

One major difference between CDMA systems and other multiple accesstechnologies is their reuse factor of one, which enables them to offer highercapacities. This is possible because of the unique way by which CDMA handlesinterference. A combination of PN sequences and orthogonal codes are used tospread and channelize the base station’s and user’s data. Radio receivers based onother digital technologies separate channels by filtering in the frequencydomain. CDMA receivers separate channels by means of the pseudo-randommodulation that is applied and removed in the digital domain, not on the basisof frequency. Spreading the signal to a much wider bandwidth helps reduce the


f

C1

C2

C3

Cnt

Figure 2.5 CDMA access technology.

power levels and makes each signal appear as background noise to other users.This scheme allows a large number of users to simultaneously share thesame 1.25-MHz carrier. In addition to spreading, CDMA systems use powercontrol techniques to maintain the interference in the system at the acceptablelevels required to satisfy the signal or radio link quality. Furthermore, CDMAsystems take advantage of multipath through the use of RAKE receivers tocombat fading. There are currently two major 3G technologies, both basedon CDMA, namely, WCDMA and CDMA2000. Let us consider the linkbetween a mobile station and a base station in a CDMA system communicat-ing using a unique code. Because of the characteristics of these codes,namely, orthogonality, the communication is successful despite the interfer-ence generated in the system from other mobiles. This is possible because ofthe way CDMA is designed, where the signals from the other links are filteredout as background noise. So in a way, CDMA mitigates interference betweenusers by accepting the fact that interference is present and optimizing the sys-tem to operate in an environment of interference. To achieve this goal,CDMA uses spread spectrum technology. One form of spread spectrum isdirect sequence spread spectrum, in which special spreading codes are used tospread out the signal over a wide bandwidth while reducing its power at thesame time, as shown in Figure 2.6. A spreading code is applied to thenarrowband data at the transmitter, resulting in a signal with a much widerbandwidth. Since the total signal power remains the same, the signal level dropsto the noise floor level. After passing through the channel, the signal at thereceiver will consist of the wanted signal, multiple access interference, and noise.By applying the same spreading code used in the transmitter to the combinedsignal, a pulse-like peak results for the wanted signal and a small residual signallevel for all interferers.

The major advantage of CDMA technology is the potential of extraordi-nary capacity increase over narrowband multiple access wireless technologies.Idealized models show that the capacity improvement may be as high as 20times that of the narrowband cellular standards, such as AMPS in North Amer-ica, NMT in Scandinavia, TACS in the United Kingdom, and 13 times that ofTDMA. However, in practice coverage areas are highly irregular, the load is notspatially uniform and is time variant throughout the day, leading to less but stillsignificant capacity improvements.

2.3 Fundamentals of CDMA

The key to CDMA high capacity is the use of noise-like carrier waves. Instead ofassigning frequency or time slots, different users are assigned different nearlyorthogonal instances of the noise carrier. This alters the system sensitivity to


interference, from having to design a system based on the worst-case interferenceto the average interference. Traditional time or frequency slotted systems must bedesigned with a reuse ratio that satisfies the worst-case interference scenario,which is experienced by only a small fraction of users. Use of pseudonoise carri-ers, with all users occupying the same spectrum, makes the effective noise thesum of all other-user signals. The CDMA receiver correlates its input with thedesired noise carrier, enhancing the signal-to-noise ratio at the detector andovercoming the summed noise enough to provide an adequate SNR at thedetector. Because the interference is summed, the system is sensitive to the aver-age interference instead of the worst-case interference. Frequency reuse is uni-versal, that is, multiple users use the same CDMA carrier frequency. Capacity isdetermined by the balance between the required SNR for each user, and thespread spectrum processing gain, defined as the ratio between the carrier chip rateto the user’s data rate. The figure of merit of a well-designed digital receiver isthe dimensionless Eb/Nt, defined as


Spreading Channel Despreading

C

C

C

I

C

I

Data

Spreading code

Spread signal

Interference + noise + signal

Figure 2.6 Direct sequence spread spectrum fundamentals.

E

N

Energy per bit

Noise Power Spectral Density Interfeb

t

=+ rence Power Spectral Density

(2.5)

The noise part of Eb/Nt, in a spread spectrum system is the sum of thermalnoise and all the other-user interference. Assuming the spectrum of the signals isrectangular, with a bandwidth W, then the noise + interference power spectraldensity is

N NP

Wt o

iotherusers= +∑

(2.6)

where the first term represents the thermal noise level of the receiver. We canthen rewrite Eb/Nt in terms of the data rate and the spread-spectrum bandwidthas:

E

N

P R

NP

W

b

t j

j

o

iotherusers

=

+∑

(2.7)

The interference in this equation is the sum of the signals from all usersother than the one of interest. This equation is the key to understanding howand why CDMA works. Early arguments against CDMA were centered on whatis termed the near-far problem. In the mobile radio environment some users maybe located near the base station while others may be located at the cell edge. Thepropagation path loss difference between those extreme users can be on theorder of several tens of decibels. Consequently, the difference in the receivedpower and the SNR at the base station from users in those two extreme casescould be as high as 50 or 60 dB, if the users are all transmitting at the same con-stant power. Hence, for the base station to accommodate users at the cell edge,the spreading bandwidth would have to be on the order of 40 dB or so, that is10,000 times the data rate. Using a bandwidth of 100 MHz to support a datarate of 10 Kbps would lead to a much worse spectral efficiency than comparedwith a narrowband system. Choosing a more reasonable bandwidth would denyservice to remote users. The key to the high capacity of commercial CDMA wasa simple solution; instead of using constant power, the transmitter’s power can becontrolled in such a way that the received powers from all users are roughly equal.This works because by controlling the received power, the total interference seenat the base station cannot be dominated by any single user as long as all usershave similar data rates. Assuming perfect power control, the interference can be


given by Io = (N – 1)P where N is the total number of users and P is the receivedsignal power from each user. The uplink Eb/Nt now becomes

( ) ( )E

N

P R

N N P W

W R

N W P Nb

t o o

=+ −

=+ −1 1

(2.8)

NW R

E

N

N

Pb

t

o= − +1 (2.9)

NW R

E

N

as Ppoleb

t

= → ∞ (2.10)

Equation (2.10) shows the fundamental dependence of CDMA capacitynot only on power control but also on interference reduction techniques such assmart antennas. Capacity can be maximized if we adjust the power control, ormore broadly P, so that the SNR is exactly what is needed for an acceptable errorrate.

2.3.1 Isolated Cell Capacity

Using (2.10) to solve for N with the assumption that power in unlimited P → ∞and a nominal SNR target of 4.5 to 5 dB for IS-95 CDMA with 9.6-Kbps datarate, we obtain an uplink pole capacity of 46 to 42, respectively. The pole capac-ity of a cell is defined as the maximum number of users a cell can support ifthere is no constraint on the peak received power. In practice, the pole capacitycannot be reached since it implies that the interference is allowed to grow tosuch high levels that the coverage shrinks to zero. Typically CDMA networksare designed and planned to operate at uplink loads of 50%–60%, levels consid-ered to provide good coverage versus capacity trade-off. Ideally, that leads to21–23 users on the uplink with IS-95A CDMA. The actual number of subscrib-ers that 50% or 60% translates to in real networks may vary depending on thedata rate selected and fade margin expected, among other factors. Note thatsince capacity and SNR are reciprocal, a reduction in the required SNR or Eb/Nt

leads to improvement in capacity, and vice versa. CDMA capacity will be dis-cussed in more details in the next sections, along with additional factors thatcontribute to the actual performance, where we will see that overall there ismajor improvement over narrowband technologies. Recall that in the same


1.25-MHz bandwidth, a single sector of a single AMPS cell has only two chan-nels available.

2.3.2 CDMA Codes

Since in CDMA systems all mobiles need to share the same frequency carrier,orthogonal codes called Walsh codes are used to separate between users and dif-ferent communications channels within a cell; that is, they providechannelization on the forward link. This is essential in CDMA to avoid or atleast minimize multiple access interference in the forward link. Walsh codes areorthogonal binary sequences generated using the Hadamard matrix as follows[4, 5]:

WW W

W WN

N N

N N2 =

(2.11)

Figure 2.7 shows how Walsh codes are generated based on (2.11). Simi-larly, Walsh codes of any length 2N where N is an integer can be generated. Bychanging 0s to -1s, Walsh codes can be rewritten as

[ ] [ ]W W12

221 1 1 1= − − = −,

where W mn denotes the mth Walsh code of length n. To illustrate how Walsh

codes are used in CDMA, let us consider three users with messages given by


011 =W

0021 =W 102

2 =W

000041 =W 10104

2 =W 110043 =W 01104

4 =W

Figure 2.7 Walsh code generation.

[ ][ ][ ]

m

m

m

1

2

3

1 1 1

1 1 1

1 1 1

= −

= −

= −

(2.12)

Now let us assign each of the users a Walsh code of length eight,respectively,

[ ][ ][ ]

W

W

W

28

4

8

68

1 1 1 1 1 1 1 1

1 1 1 1 1 1 1 1

1 1 1 1 1 1 1 1

= − − − −

= − − − −

= − − − −

(2.13)

Since the chip rate for the Walsh code in this case is eight times the mes-sage bit rate, spreading each signal with its assigned code will result in wideningthe band from 1/Tb to 1/Tc where Tb and Tc are the bit and chip periods, respec-tively. The spread spectrum signals of the three users Sn(t) and the combined sig-nal C(t) are then given by, respectively,

( ) ( )( ) ( )( ) ( )( ) ( ) ( ) ( )

S t m t W

S t m t W

S t m t W

C t S t S t S t

1 1 28

2 2 4

8

3 3 68

1 2 3

==

== + +

The resultant signals are shown in Figures 2.8 through 2.11, respectively.Now, in order to recover a user’s original message, the receiver spreads the


Figure 2.8 User 1 spread spectrum signal.

received composite signal with the Walsh code assigned to that user. This opera-tion is shown in Figure 2.12 for user 1, where the receiver integrates all the val-ues over a bit period. The original message is reconstructed using the followingdecision criterion

( ) ( )( ) ( )

$

$

m t if C t W

m t if C t Wmn

mn

= ⋅ >= − ⋅ <

1 0

1 0(2.14)





Figure 2.11 Composite spread spectrum signal.

(a)

(b)

Figure 2.12 (a) Effect of spreading received signal with first user’s code; (b) User 1 recovered signal(Tb = 8Tc).

If the receiver attempts to spread the composite signal with a code that wasnot assigned to the user, (e.g., with [ ]W 8

8 1 1 1 1 1 1 1 1= − − − − , we

get all zeros after the integration, as we can see in Figure 2.13, which means thesignal cannot be recovered.

In addition to using Walsh codes to separate different users and differentchannels on the forward link within a sector, a CDMA system needs to separatetransmissions from different sectors within a network. This is accomplishedusing PN codes, as described in Figure 2.14. Some important key differencesbetween Walsh codes and PN codes, which greatly impact the interference levelin a CDMA system, are illustrated in Table 2.1.

2.3.3 IS-95 CDMA Systems

The TIA IS-95 CDMA system is a 2G mobile wireless system that operates inthe cellular 800-MHz band [6, 7]. Another version of this system that operatesin the PCS 1,900-MHz band is defined in J-STD-008 [8]. Both systems use a1.25-MHz wide carrier and a chip rate of 1.2288 Mcps. On the forward link, afamily of 64 Walsh codes is used to separate the different channels and differentusers. Short PN codes of length 215-1chips with a period of 32768 chips or26.67 ms are used to separate transmissions from different sectors. This isaccomplished by using the same PN sequence for all sectors and then identifyingeach sector by a unique time offset in increments of 64 chips, resulting in 512possible PN sequences. On the reverse link, long PN codes of length 242-1chipsare used for channelization, that is, to distinguish different users. In addition,


Figure 2.13 Effect of spreading received signal with wrong Walsh code.

the reverse link signal is further spread by short PN codes of length 215-1chips toidentify the sector to which the transmission is intended.

2.3.3.1 Forward Link Channels

The IS-95 and J-STD-008 standards define two types of forward link channels,namely common channels broadcast to all mobiles in a sector and dedicatedchannels to specific mobiles. Note that in addition to assigning different Walsh


Filtering Modulator

Data stream 1

Data stream 2

Data stream n

Walsh 1

Walsh 2

Walsh n

Sector specificPN code

Demodulator Filtering

Walsh 1Sector specificPN code

Demodulator Filtering

Walsh n

Data stream 1

Data stream n

CDMA transmitter

CDMA receiver

MS1

MS2Sector specificPN code

Figure 2.14 CDMA transmitter and receiver block diagrams.

codes to each channel, all channels are spread with the same PN sequence associ-ated with the transmitting sector. The set of channels defined in the standard arelisted here:

• Pilot channel: The pilot channel is continuously transmitted sector-wideto provide timing and phase references to all users to aid in systemacquisition, signal strength comparison, and demodulation operations.The pilot channel is assigned Walsh code 0 or W0, which is a sequenceof 64 zeros with 1.2288 Mcps chip rate. Note that no baseband infor-mation is carried by the pilot channel.

• Sync channel: The sync channel is also continuously transmitted sectorwide to provide timing information to the mobile during system acqui-sition and power up. Baseband information contained in the sync chan-nel message is used to inform mobiles of system synchronizationinformation and other system parameters. The sync channel is assignedWalsh code W32 and it is transmitted in groups of superframes at the bitlevel. Each superframe lasts for 80 ms and consists of three 26.67 mssync channel frames that are synchronized with each period of the shortPN sequence. Hence, once the mobile acquires synchronization withthe pilot channel, the sync channel frame boundaries are immediatelyknown.

• Paging channel: The paging channel is used to transmit overhead infor-mation to a mobile such as pages and other commands. Call setup com-mands and traffic channel assignments are also sent over the pagingchannel. Based on the standard specifications, there can be up to sevenpaging channels but there must be at least one. At the bit level, eachpaging channel frame lasts for 20 ms, four of which are combined intoan 80-ms paging channel slot.


Table 2.1Comparison Between Walsh and PN Codes

Transmit and ReceiveWalsh CodesCorrelation

PN CodesCorrelation

Same codes, same time offsets 100% 100%

Different codes 0% Low noise-like

Same codes, different time offsets > 0%

< 100%

Low noise-like

• Forward traffic channels: Forward traffic channels carry voice, data, andsignaling once a call has been established. There are two rate setsdefined in the standard. Rate set 1, or RS1, with data rates 1.2, 2.4, 4.8,and 9.6 Kbps, and RS2 with data rates 1.8, 3.6, 7.2, and 14.4 Kbps. Insystems with only one paging channels, there are 61 available Walshcodes that could be assigned to traffic channels. Traffic channel frameslast for 20 ms.

2.3.3.2 Reverse Link Channels

In IS-95 and J-STD-008 standards there are only two reverse link channels:

• Access channel: This channel is used by the mobiles to access the systemfor registration, call origination, page responses, and overhead transmis-sion to the base stations.

• Reverse traffic channels: Similar to the forward link, the reverse trafficchannels carry voice, data, and signaling once a call has beenestablished.

2.3.3.3 RAKE Receiver

The presence of buildings, trees, hills, and other objects in the areas served bymobile systems cause signal reflection, diffraction, and scattering. This createsmultiple replicas of the transmitted signal with different attenuations and timedelays at the receiver. The interaction of the incoming waves at the receiverantenna results in deep and rapid fading or fluctuations in the signal strength.This significantly degrades the system performance. IS-95 based CDMA sys-tems actually take advantage of the multipath components through the use ofRAKE receivers. Multiple correlators are used to detect the strongest multipathcomponents using a searcher finger designed to compare the incoming signalswith the PN code used. This operation detects multipath arrivals by producing aseries of correlation peaks at different times. The magnitude of each peak is pro-portional to the envelope of the signal in a particular path, whereas the time ofeach peak relative to the time of arrival of the first path gives that path’s delay.With the amplitudes and time delays of the strongest multipath componentsknown, a RAKE receiver compensates for the delays and combines the signalsbased on their strengths. This produces a diversity gain at the CDMA receiver,which helps combat fading. The block diagram of a CDMA RAKE receiver isshown in Figure 2.15.

2.3.3.4 Power Control

Recall that in CDMA systems all users share the same RF carrier through the useof PN codes, therefore each user appears like random noise to other users and


contributes to the system noise. If the power of each user is not properly con-trolled and allowed to increase unnecessarily, other users would suffer frominterference that could severely degrade system performance. Consider aCDMA system where all users transmit at the same power. A user close to thebase station will result in a high SNR1 at the receiver, whereas another user fur-ther away from the base station would yield a lower SNR2. Obviously, this dis-parity results in different signal quality between users. This is the classicalnear-far problem. Assume that the required SNR necessary to maintain thedesired signal quality is given by SNRreq. When new users are added to the cell,the interference level in the cell increases, thus reducing the SNRs of existingand new users up to the point at which the SNR of a new user would not be ableto reach SNRreq. Therefore, no more users can be added to the cell and the capac-ity is reached with only a few users. Hence, power control is essential to over-come the near-far problem and maximize the capacity. Power control is the


Delayelement(chips)

t

Correlator

Correlator

Correlator

Strongest multipathcomponents

A1

A2

A3

Sum

Path 1+ interference



Path

1

Path

2

Path

3

Delayelement(chips)

Delayelement(chips)

Figure 2.15 RAKE receiver block diagram.

process by which the transmit power of each user is controlled such that thereceived powers at the base station are equal. The capacity is then maximized byonly allowing each user to transmit just enough power to achieve SNRreq.

2.3.3.5 Reverse Link Open Loop Power Control

As in any communications system, there is always a propagation loss thatimpairs the signal on the forward and reverse links. In addition to the regulardistance-dependent path loss, other factors such as shadowing and multipathproduce fading in mobile communications systems. Basically, there are twotypes of fading, slow fading and fast fading. Slow fading is modeled by alognormal distribution and it manifests itself by slow power variation over severalwavelengths, as shown in Figure 2.16. This type of fading is typically caused bythe signal being partially blocked by buildings, trees, and other obstacles. On theother hand, when multipath components with different amplitudes, phases, andarrival times add up at the receiver, they combine constructively and destruc-tively, forming a standing wave pattern with a half wavelength period. As themobile moves through this pattern, the received power will experience fast fad-ing with an envelope distribution characterized by a Rayleigh distribution.When a mobile is in idle state, that is, a state where it monitors the overheadchannels but no call has been established yet, the base station cannot control thepower of the mobile. To solve this problem, the IS-95 standard defines the openloop power control process, which ensures that each mobile starts its initial trans-missions, also called access probes, with a power level that depends on thereceived power from the base station pr or


Slow fading

Fast fading

Time

Sign

alst

reng

thin

dB

Figure 2.16 Fading as function of time.

P p K NOM PWR INIT PWRt initial r, _ _= − − + + (2.15)

where K is a constant equal to 73 in the cellular band and 76 in the PCS band.NOM_PWR and INI _PWR are system parameters that are broadcast from thebase station to all mobiles. The reason this is called open loop power control isthat if the mobile does not receive an acknowledgement from the base stationafter sending an access probe, it waits for a random time period before sendingthe next access probe with a slightly higher power. The mobile repeats this pro-cess until an acknowledgement is received but there is no feedback from the basestation about the signal quality. Since this process is slow, it only compensatesfor the slow lognormal fading.

2.3.3.6 Reverse Link Closed Loop Power Control

The closed loop power control attempts to balance losses between the link dueto Rayleigh fading or fast fading at slow mobile speeds and interference varia-tions due to loading once the mobile is on a traffic state. It also improves theperformance of mobiles at the cell edge where the signal is weak and the interfer-ing signals from other cells are strong. As briefly described previously, powercontrol adjusts the transmit power of each mobile to maintain the required SNRgiven a specific signal quality. To achieve that, the power control process mustbe able to determine the value of the SNRreq to maintain the signal quality. Theouter loop power control performs this function by adjusting the target SNRaccording to the prevailing environment to achieve the desired end-user qualityof service. Let us define Eb as the energy per bit and No as the interference plusnoise power spectral density, then we get

( )E

N

PR

N IW

WR

PN I

WR

SNRb

o

=+

= ⋅+

= ⋅ (2.16)

where W is the RF carrier bandwidth, R is the signal data rate, and W/R isdefined as the processing gain. It is clear from (2.16) that adjusting Eb/No isequivalent to adjusting the SNR. The closed loop power control is summarizedin Figure 2.17. Based on the target Eb/No, the base station controls the mobiletransmit power. The power control commands are sent from the base station onthe forward link in the form of power control bit (PCB); each power controlgroup lasts for 1.25 ms in IS-95 based systems. Hence, the power of the mobilecan be adjusted up to 800 times per second. This is performed using the innerloop power control as follows: The base station monitors the reverse link Eb/No

and compares it to (Eb/No)Target. If Eb/No > (Eb/No)Target, the base station commandsthe mobile to decrease the transmit power by sending a power down command,


⇒PCB = 0. If Eb/No < (Eb/No)Target, the base station commands the mobile toincrease the transmit power by sending a power up command, ⇒PCB = 1.

2.3.3.7 Soft Handoff

One major advantage of having all users in a CDMA system on the same RF car-rier is the ability of maintaining simultaneous connections. When a mobilemaintains simultaneous traffic channels with sectors belonging to different basestations, it is said to be in soft handoff. On the forward link, the mobile’s RAKEreceiver demodulates the signals received from separate sectors and combinesthem to produce a signal with a better quality. On the reverse link, multiple basestations demodulate the mobile’s signal and the demodulated frames are sentback to the base station controller (BSC) to select the best frame. This operationprovides some diversity since the signals on different links are typicallyuncorrelated and do not fade at the same time with the same depth. This resultsin a soft handoff gain, which improves the air interface capacity. As the mobilemoves around the system, it keeps a list of all active pilots from the soft handofflinks in a set called the active set. Other pilots with raw SNR Ec/Io strong enoughto be candidates for soft handoff are kept in a set called the candidate set.Another important set kept by the mobile is the neighbor set, which containsthose pilots that are neighbors to the current serving sector. When the sectors inthe active set belong to the same base station, the mobile is said to be in softerhandoff. The procedure by which these sets are maintained and the pilots areprocessed is defined in the IS-95 standard.

2.4 Third Generation Systems

As second generation systems started to reach their limits in terms of spectralefficiency along with the increasing demands for higher data rate services, a need


Is received signalqualitybetter than requiredquality

Decrease Increase

Targetob NE )(Yes No

Targetob NE )(

Figure 2.17 Closed loop power control mechanism.

emerged for improved networks that can provide these future requirements.This led to the development of 3G systems [5, 9], with the following mainobjectives:

• Provide data rates from 144 Kbps up to 384 Kbps for mobilityscenarios;

• Provide data rates up to 2 Mbps for limited mobility and fixed wirelessscenarios;

• Provide higher spectral efficiency compared with 2G systems;

• Support multiple simultaneous services (e.g., speech, high-speed data).

There are currently two major standards adopted for 3G systems, both ofwhich are based on CDMA, namely CDMA2000 and WCDMA. Anotheremerging technology also based on CDMA is the time division synchronousCDMA (TD-SCDMA).

2.4.1 CDMA2000

The CDMA2000 family of standards is a wideband spread spectrum radio inter-face that uses CDMA technology to meet the objectives of 3G systems whilemaintaining backward compatibility with IS-95 based systems. This means thatmobile handsets designed according to the IS-95 standard are capable of operat-ing in a CDMA2000 system and vice versa. The first component of theCDMA2000 standard is called 1X radio transmission technology (1X RTT)because it uses an RF carrier of 1.25 MHz just like IS-95 based systems, hencethe 1X, which is also referred to as spreading rate (SR)1. The key benefits of the1XRTT technology standardized under the name of IS-2000 [10, 11] comparedwith IS-95A/B standards [7] can be summarized as follows:

• Better forward error correction (FEC). This is achieved through the use ofhigher convolutional coding rates as well as turbo codes for high datarates. The coding rate refers to the number of symbols produced by theencoder for every bit of input data. The greater this number is, themore protection we get against errors because of the increased correc-tion power. A direct impact of the improved coding is a reduction inthe required Eb/No, which directly translates into higher capacity orhigher data rates. A coding gain of up to 2dB can be achieved withIS-2000 systems compared with the IS-95 standard.

• Fast forward link power control mechanism. As described earlier, a powercontrol mechanism is defined for the reverse link of the IS-95 standard,whereby the transmit power of the mobile is controlled up to 800 times


per second. IS-2000 extends the use of this power control process to theforward link as well, where the mobile station can control the powertransmitted by the base station with speeds of 400–800 times per sec-ond through the use of the forward link closed loop power control mecha-nism. This allows the power resources on the forward link to beoptimized and used much more efficiently than in IS-95 systems, yield-ing significant improvements in capacity.

• Multimedia services and improved data services support. In addition to theimprovements listed above, the IS-2000 standard introduces new dedi-cated channel and common channels to support high data rate applica-tions as well as improved diversity techniques. Moreover, the batterylife is extended through the use of a new quick paging channel. Thecombination of these improvements results in a voice capacity increaseof 1.6 to 2 times compared with IS-95A/B as well as data rates of up to307 Kbps.

2.4.1.1 Overview of IS-2000 Forward Link Physical Channels

As we have seen in the IS-95 standards, there are two rate sets, RS1and RS2,with data rates of up to 9.6 Kbps and 14.4 Kbps, respectively. In IS-2000, awider range of data rates are available and are defined in terms of radio configu-rations (RC), which can be summarized as:

• RC1, which supports IS-95A/B backward compatibility for all rate set 1(RS1) based services up to 9.6 Kbps;

• RC2, which supports IS-95A/B backward compatibility for all rate set2 (RS2) based services, up to 14.4 Kbps;

• RC3, which supports data rates from 1,500 bps up to 153.6 Kbps,using rate 1/4 FEC encoding;

• RC4, which supports data rates from 1,500 bps up to 307.2 Kbps,using rate 1/2 FEC encoding;

• RC5, which supports data rates from 1,800 bps up to 230.4 Kbps,using rate 1/4 FEC encoding.

Table 2.2 provides a summary of the forward link physical channels of theIS-2000 standard and a brief description of their functions.

2.4.1.2 Overview of IS-2000 Reverse Link Physical Channels

Since CDMA networks based on the IS-95 standard have been launched in the1990s it became apparent over the years that there are certain inefficiencies in



Table 2.2IS-2000 Forward Link Physical Channels

Channel Function

Common pilot channel(F-CPICH)

Used to broadcast pilot for the entire cell/sector, for channel andphase estimation (coherent demodulation), initial acquisition, andhandoffs.

Common auxiliary pilotchannel (F-CAPICH)

Used for beamforming applications for a group of mobiles.

Dedicated auxiliary pilotchannel (F-DAPICH)

Used for beam steering and beamforming applications for a singlemobile.

Sync channel This is the same channel as in IS-95A/B, containing system synchro-nization information.

Fundamental channel(F-FCH)

This is identical to the 95A/B traffic channel. Used for voice, data,and control. There can be 0-1 channels.

Supplemental channel(F-SCH), (RC 3-9)

This channel was introduced for supporting high data rates. Therecan be 0–2 channels.

Supplemental code channel(F-SCCH), (RC1-2)

This channel was introduced in IS-95B for medium data rate serviceoption. There can be 0–7 channels.

Dedicated control channel(F-DCCH)

Introduced for MAC, data, and signaling. The power controlsubchannel can also be punctured here when F-FCH is absent.

Common assignment chan-nel (F-CACH)

Designed to provide fast response reverse link channel assignmentsto support transmission of random access packets on the reverse link.

Common power controlchannel (F-CPCCH)

Used by the base station for transmitting common power controlsubchannels.

Paging channel IS 95 A/B One common control paging channel (PCH) where broadcast and mo-bile station directed messages are transmitted.

Broadcast control channel(F-BCCH)

Broadcasts only cell-specific overhead messages (e.g., CDMA chan-nel list, extended systems parameters message and neighbor list) at(4.8, 9.6, or 19.2 Kbps). The sync channel is used to let mobiles knowif F-BCCH is supported.

Common control channel(F-CCCH)

Broadcasts mobile station specific messages (e.g., extended channelassignment message, general page message, order message) at 9.6,19.2, and 38.4 Kbps in discontinuous transmit mode.

Quick paging channel(F-QPCH)

Helps decrease the “wake” time of a mobile station, that is the timethe mobile has to periodically demodulate the PCH or F-CCCH. Thisimproves MS standby time and reduces battery consumption. Thesupport of this channel is optional.

the design of the reverse link. This led to an IS-2000 physical layer design thatadds several enhancements to improve the performance of the reverse link.Table 2.3 provides a summary of the reverse link physical channels of theIS-2000 standard and a brief description of their functions.

One of the major drawbacks of the IS-95 reverse link design is thenoncoherent demodulation by rake receiver, which requires a high received sig-nal-to-noise ratio for good performance. The IS-2000 standard solves this prob-lem by introducing the reverse link pilot or R-PICH, which allows the basestation to estimate the carrier phase and makes coherent demodulation possible.This improves both searching and tracking of mobiles. Another major improve-ment is the introduction of the forward link power control by which the mobilecan adjust the forward link power. This is performed by time-multiplexing for-ward power control (PC) information on the reverse pilot channel, as shown inthe frame structure in Figure 2.18.

2.4.1.3 CDMA2000 1x EV-DO

Data services are expected to have a significant growth over the next few yearsand will likely become the dominant source of 3G traffic and revenue. The


Table 2.3IS-2000 Reverse Link Physical Channels

Channel Function

Reverse pilot channel (R-PICH) Used for searching, tracking, and coherent demodulation.Also used by the forward link channels to adjust forward linkpower and maintain the quality of the link.

Access channel (R-ACH) andenhanced access channel (R-EACH)

Used by mobiles to access the system for registration, callorigination, and page responses.

Reverse common control channel(R-CCCH)

Used to support efficient access procedures of packet dataservices.

Reverse dedicated control channel(R-DCCH)

Used for the discontinuous transmission of user traffic,control, and signaling information to the base station whilethe mobile is in the traffic state.

Reverse fundamental channel(R-FCH)

Used to carry user traffic for RC1 and RC2.

Reverse supplemental codechannel

Used to support medium data rates based on services for RC1and RC2. There can be up to seven such channels.

Reverse supplemental channel(R-SCH)

Used to support high data rates for RC3 and RC4. There canbe zero to two channels.

current 3G operators in Japan and Korea as well as in the United States arealready experiencing great success with their data services. KTK in Korea hasreported that 34% of its ARPU was related to data usage for the third quarter of2003, particularly after the deployment of 1xEV-DO. The 1xEV-DO stan-dard [12–20] is optimized for wireless high-speed packet data services. Becauseof the typical asymmetric characteristics of IP traffic, the downlink is the morecritical of the two links [12]. Thus, several techniques were introduced in1xEV-DO to optimize the downlink throughput. The 1xEV-DO downlink usestime-division-multiplexed (TDM) waveform, which eliminates power sharingamong active users by allocating full sector power and all code channels to a sin-gle user at any instant. This is in contrast to code-division-multiplexed (CDM)waveform on the IS-95 downlink, where there is always an unused margin oftransmit power depending on the number of active users and power allocated toeach user. Through power control, this margin is used to account for large varia-tions of the required mobile station transmit power in fading channels to guar-antee a given target frame error rate. Figure 2.19 shows the sector power usage ofthe IS-95 and 1xEV-DO downlinks. Each channel in IS-95 is transmitted theentire time with a certain fraction of the total sector power, while the equiva-lent channel in 1xEV-DO is transmitted, at full power, only during a certainfraction of time. The efficient usage of sector power resource in 1xEV-DOimproves cell coverage as well as signal-to-interference and noise ratio (SINR)for noise-limited users.

Similar to the IS-95 concepts, every mobile station reports to the networkthe strongest downlink pilots it can measure. In turn, the network selects anactive set for each terminal. Each sector in the terminal’s active set maintains aconnection with the terminal. The active set of sectors for any given terminal isalso the set of power controlling sectors for its uplink. However, instead of trans-mitting equal power on all downlink traffic channels in the active set as adopted


Reverse pilot channel

Reverse powercontrolsubchannel

0 1 2 15

1 Frame (20ms)

1 PCG(1.25ms)

Figure 2.18 IS-2000 reverse link pilot frame structure.

in IS-95, the 1xEV-DO network only transmits on the best link and allocates nopower on the others. To accomplish this procedure, a mobile terminal monitorsthe SINR of all the sectors in its active set and informs the network, via a feed-back channel, of the identity of the selected serving sector. Due to the TDMwaveform of the 1xEV-DO downlink, a terminal is allocated the full sectorpower whenever it is served, thus no power adaptation is needed. Instead, rateadaptation is used on the downlink. The highest data rate that can be transmit-ted to each terminal is a function of the received SINR from the serving sector,which is typically a time-varying quantity. To achieve the highest data rate ateach time of transmission, each terminal predicts the channel condition over thenext packet for its serving sector based on the correlation of the channel states. Itselects the highest data rate that can be reliably decoded based on the predictedSINR, and then informs the serving sector of its selected rate over an uplinkfeedback channel. Whenever the network decides to serve a terminal, it trans-mits at the most recent selected rate fed back from the terminal. Since a sectortransmits traffic data to a single user at any instant of time, a scheduling algo-rithm is implemented in each sector to fairly allocate the available time slotsamong the active users, thus maximizing capacity by exploiting the channeldynamics. Because different users experience independent fading processes, it isunlikely that all users’ SINR will fall into deep fades at the same time. In otherwords, when some users experience a deep fade, others reach peaks of theirreceived signal strength. As a mobile user goes through periods of varying fades,the data rate allocated to it by the network will vary. However, since Internet


TimeTime

Max transmit power Max transmit power

Pilot channel

Paging channel

Sync channel

Traffic channels Total data trafficchannels

Unused power margin

Se

cto

rtr

an

sm

itp

ow

er

Se

cto

rtr

an

sm

itp

ow

er

Pilo

tch

an

ne

l

Co

ntr

olch

an

ne

l

Figure 2.19 Sector power usage comparison between 1xRTT(left) and 1xEV-DO (Right).

protocol (IP) traffic can tolerate relatively longer and variable time delays, unlikevoice services, this can be tolerated. The standard does not specify the type ofscheduler to be used. A smart scheduler will attempt to serve an active user nearits peak SINR while maintaining a certain degree of fairness. For instance, thePF scheduling algorithm represents a good balance because it incorporates thetwo important features of a capacity enhancing scheduler: multiuser diversitygain and fairness. The algorithm selects the terminal based on a metric equal tothe ratio of the instantaneous channel state to the long-term average of theserved throughput. Thus, it attempts to serve each terminal at their local peaksof channel conditions and maintain higher average served throughput when theterminal is in better coverage. Another critical concept in the design of the1xEV-DO standard is link adaptation. Link adaptation is achieved by combin-ing several mechanisms designed to improve spectral efficiency while achievingthe required simplicity and robustness for effective operation in a wireless cellu-lar environment. The idea behind link adaptation is to optimize spectral effi-ciency by matching the transmit data rate, modulation, and coding to the timevarying received SINR at the terminal. A variety of modulation schemes, includ-ing QPSK, 8 PSK, and 16 QAM as well as coding rates that best matches thefading channel, are defined in what is commonly called adaptive modulation andcoding techniques. To fully exploit these concepts, the system includes a collec-tion of techniques that consist of incremental redundancy and repetition cod-ing, time diversity adaptation, and HARQ [15–18].

2.4.2 WCDMA

WCDMA is another 3G air interface based on direct-sequence CDMA(DS-CDMA). WCDMA uses a chip rate of 3.84 Mcps, compared with 1.2288Mcps in both IS-95 and IS-2000 standards and requires an RF carrier with5-MHz bandwidth [21–25]. There are two modes of operation in the WCDMAair interface, a frequency division duplex (FDD) mode, where a pair of 5-MHzcarriers are used, and a time division duplex (TDD), where only one carrier isused. Similar to IS-95 and IS-2000 systems, channelization is achieved throughthe use of orthogonal codes referred to as orthogonal spreading factor codes(OVSF), which are Walsh codes of variable lengths, whereas source separation isachieved using gold codes. Table 2.4 shows a comparative summary of theWCDMA, IS-95, and IS-2000 air interfaces.

2.4.2.1 FDD-WCDMA Forward Link (Downlink) Physical Channels

In this section we will summarize the physical channels associated with theWCDMA downlink. The physical channels are those channels that perform theactual transmission of data bits and are distinguished by an RF carrier, achannelization code, a spreading code, and modulation parameters. WCDMA


physical channels can be grouped into four categories: common channels broad-cast to all mobiles in the cell or sector, channels that carry paging information,channels used for random- and packet-access, and dedicated connectionchannels.

• Common pilot channel (CPICH): This channel is a cell-wide channel,which provides a coherent phase reference for the downlink channels,and it uses the gold code specific for that cell. It also aids channel esti-mation for cell selection and reselection as well as handoff proceduresfor the mobiles. The CPICH uses orthogonal codeC 0

256 .

• Primary common control physical channel (P-CCPCH): This channel isused to broadcast cell information; that is, cell system frame number(SFN) and timing reference for all downlink channels necessary for syn-chronization operations. That is why the P-CCPCH is continuouslytransmitted over the entire cell and it always uses the samechannelization codeC1

256 .

• Secondary common control physical channel (S-CCPCH): This channel isused to transmit information related to the forward access channel(FACH) and the PCH and is mainly monitored by the mobiles in idlemode.


Table 2.4Main Differences Between IS-95, IS-2000, and WCDMA

Link Function IS-95 A/B IS-2000 WCDMA

Forward link(Downlink)

Channelization 64-chip Walshcodes

4~256-chipWalsh codes

4~512-chipOVSF codes

Source separation (215–1)-chipshort PN codes

(215–1)-chipshort PN codes

38400 chips of218 Gold code

Power control rate Slow 800 Hz 1,500 Hz

Reverse link(Uplink)

Channelization None 4~256-chipWalsh codes

4~256-chipOVSF codes

Source separation (242–1)-chiplong PN codes

(242–1)-chiplong PN codes

38,400 chipsof 225 Goldcode

Power control rate 800 Hz 800 Hz 1,500 Hz

• Paging indication channel (PICH): This channel is used in conjunctionwith the PCH to provide mobiles with a sleep mode operation, whichsaves the battery in idle mode. Basically, the PICH is used to alertmobiles of an incoming page.

• Dedicated physical channel (DPCH): The DPCH consists of two sepa-rate channels, the dedicated physical data channel (DPDCH) and thededicated physical control channel (DPCCH), which are time multi-plexed onto one time slot. The DPCCH carries control bits such aspilot bits, which are used by the receiver to measure the channel quality,and transmission power control (TPC) bits, which are used to adjustthe power of the mobile. The DPDCH is mainly used to carry user traf-fic as well as some overhead and signaling data.

2.4.2.2 FDD-WCDMA Reverse Link (Uplink) Physical Channels

As with the downlink, there are two categories of channels on the WCDMAuplink, common uplink physical channels and dedicated uplink physicalchannels.

• Physical random access channel (PRACH): This channel is used to carryaccess requests (i.e., control information and short data bursts) anddoes not contain any pilot or TPC bits since it uses only open looppower control.

• Physical common packet channel (PCPCH): This channel is used to carryconnectionless packet data.

• Dedicated physical data channel (DPDCH): The uplink DPDCH is usedto carry dedicated user traffic data generated at an upper layer. Theremay be zero, one, or up to six uplink DPDCHs.

• Dedicated physical control channel (DPCCH): The uplink DPCCH isused to carry control information consisting of pilot bits to supportchannel estimation for coherent detection, TPC commands, and somefeedback information. Unlike the downlink case, the DPDCH andDPCCH are not time multiplexed; instead, they are fed into the I andQ inputs of a complex spreader.

2.4.3 HSDPA

To meet the increasing demand for high data rates in multimedia services overnetworks supporting WCDMA, the Third Generation Partnership Project(3GPP) has released a new high-speed data transfer protocol named HSDPA[26–29]. HSDPA is expected to provide significant improvements over the basic


WCDMA R’99 for downlink asymmetrical and bursty packet data services.HSDPA will offer a peak data rate up to and in excess of 10 Mbps (as opposed tothe currently deployed 384 Kbps), as well as at least threefold sector throughput.For end users, HSDPA will mean lower delays and faster connection andresponse times, particularly for high data rate applications in loaded systems. Thesubstantial increase in data rate and throughput is achieved by implementing afast and complex channel control mechanism based on a short and fixed packettransmission time interval (TTI, fast HARQ, and fast scheduling performed atthe Node B (base station) instead of the radio network controller (RNC), theequivalent of the BSC. The TTI indicates how often data arrives from higher lay-ers to the physical layer and could take any of the values of 10, 20, 40, or 80 msin R’99 WCDMA. R’99 WCDMA already includes three different channels fordownlink packet data transmission: dedicated channel (DCH), downlink sharedchannel (DSCH), and FACH. The FACH is a common channel offering lowlatency. However, it is not spectraly efficient since it does not support fast closedloop power control. It is therefore limited to carrying only small data traffic. TheDCH is the primary data channel and can be used for any traffic class. The DCHis allocated a certain orthogonal variable spreading factor (OVSF: 4-512) accord-ing to the connection peak data rate, whereas the block error rate (BLER) is con-trolled by inner and outer loop power control. The DCH code and powerallocation are therefore inefficient for bursty and low duty cycle data applicationssince channel reconfiguration can be very slow (in the range of 500 ms) [21]. TheDSCH provides the possibility to time-multiplex different users and improve thechannel reconfiguration time and packet scheduling procedure (in the order of10 ms) [5]. The HSDPA concept can be seen as an extension of the DSCH withthe introduction of new features such as AMC, short packet size, multicode oper-ation, and fast L1HARQ. In fact, these features replace the two basic WCDMAfeatures, namely variable spreading factor VSF and fast power control [21]. Next,we will provide an explanation of key HSDPA features.

A new transport channel named high-speed downlink shared channel(HS-DSCH) has been introduced as the primary radio bearer. Similarly to theDSCH, the HS-DSCH is shared between all users in a particular sector. Theprimary channel multiplexing occurs in the time domain, where each TTI con-sists of three time slots (or 2 ms). The TTI is also referred to as a subframe. TheTTI has been significantly reduced from the 10, 20, 40, or 80 ms TTI sizes sup-ported in R’99 to better achieve short round-trip delay between the mobile sta-tion and the Node B and improve the link adaptation rate and efficiency of theAMC. Within each 2 ms TTI, a constant spreading factor (SF) of 16 is used forcode multiplexing with a maximum of 15 parallel codes for the HS-DSCH.These codes may all be assigned to one user during the TTI, or may be splitamongst several users. Note that the more codes allocated to a user, the higherpeak data rate it can achieve. The number of parallel codes allocated to each user


depends on cell loading, quality of service (QoS) requirements, and the mobilestation code capabilities (5, 10, or 15 codes). To support the HS-DSCH opera-tion, two control channels have been added: the high-speed shared control chan-nel (HS-SCCH) and the high-speed dedicated physical control channel(HS-DPCCH). The HS-SCCH is a fixed rate channel used for carryingdownlink signaling between the Node B and the mobile station before thebeginning of each scheduled TTI. This includes the mobile station identity,HARQ-related information, and the parameters of the HS-DSCH transportformat selected by the link-adaptation mechanism. The HS-DPCCH carriesuplink cyclic redundancy check (CRC)-based ACK/NACK signaling for thephysical layer retransmission as well as CQI to be used in the link adaptationmechanism. The CQI is based on the CPICH and is used to estimate the trans-port block size, modulation type, and number of channelization codes fordownlink transmission. The feedback cycle of the CQI can be set as a networkparameter in predefined steps of 2 ms.

2.4.3.1 Adaptive Modulation and Coding

Adaptive modulation and coding is the fundamental feature of HSDPA. It con-sists of continuously optimizing the code rate, the modulation scheme, and thenumber of multicodes employed as well as the transmit power per code accord-ing to the channel quality experienced (CQI feedback) by the mobile station. Toachieve very high data rates, HSDPA adds a higher order modulation (16QAM) to the existing QPSK modulation in R’99. Different combinations ofmodulation and channel encoding can be used to provide data rates rangingfrom 119 Kbps/code with QPSK and 1/4 code rate to 712 Kbps/code with 16QAM and 3/4 code rate (SF = 16). Users with the most favorable channel con-dition (close to the Node B) will get the highest data rates, whereas users withthe least favorable channel condition will get the lowest data rates (located at thecell edge). HSDPA supports the use of 5, 10, and 15 multicodes. A single usercan receive up to 15 multicodes, resulting in a peak data rate of 10.8 Mbps.However, the maximum specified peak data rate with HSDPA is 14.4 Mbps (or960 Kbps/code) when 16 QAM modulation is used with no coding (effectivecode rate of 1) and 15 multicodes. This rate remains very unlikely to achievesince it corresponds to an unloaded system where the served user is extremelyclose to the node B. Another benefit of AMC is better utilization of the Node Bpower. If no power constraints are specified, the leftover power from thededicated channels (R’99) can be allocated to HS-DSCH, resulting innear-maximum power utilization.

2.4.3.2 Hybrid-ARQ with Soft Combining

The retransmission mechanism selected for HSDPA is HARQ with Stop andWait protocol (SAW). HARQ allows the mobile station to rapidly request


retransmission of erroneous transport blocs until they are successfully received.HARQ functionality is implemented at the medium access control (MAC)layer, as opposed to the radio link control (RLC) layer, as is the case in R’99WCDMA. Therefore, the retransmission delay of HSDPA is much lower thanfor R’99 because it does not involve the RNC. In normal circumstances, aNACK may require less than 10 ms at the MAC layer, while it can take up to100 ms at the RLC layer when Iub signaling is involved [9]. This reduces signifi-cantly the delay jittering for Transmission Control Protocol/Internet Protocol(TCP/IP) and delay-sensitive applications. During retransmission, the mobilestation does not discard the original transmission but rather combines it withthe following retransmission(s) to increase the probability of successful decod-ing. This is called soft combining. HSDPA supports both chase combining(CC) and incremental redundancy (IR). CC is the basic combining scheme. Itconsists of the Node B simply retransmitting the exact same copy of the originalpacket. With IR, additional redundant information is incrementallyretransmitted, providing additional coding gain. This can result in fewerretransmissions than for CC. However, the disadvantage of IR over CC is themuch higher memory requirement for the phone.

2.4.3.3 Fast Scheduling

The scheduler is a key element of HSDPA that determines the overall behaviorof the system and, to a certain extent, its performance. For each TTI, it decideswhich terminal (or terminals) the HS-DSCH should be transmitted to and, inconjunction with the AMC, at which data rate. One important change from theR’99 implementation is that the scheduler is located at the Node B as opposedto the RNC. This, with the short TTI (2 ms), enables the scheduler to quicklytrack the user equipment (UE) channel condition and adapt the data rate alloca-tion accordingly. Three main scheduler algorithms have been proposed forHSDPA: round robin (RR), maximum C/I, and PF. RR schedules users accord-ing to a first-in first-out approach. It allows achieving a high degree of fairnessbetween the users at the expense of the overall system throughput (and thereforespectral efficiency) because some users can be served even when they are experi-encing a destructive fading. The maximum C/I schedules users with the highestC/I during the current TTI. This naturally leads to the highest system through-put because most of the served users will likely sustain a high peak data rate witha low probability of error. However, the fairness between the users is minimal.In fact, users at the cell edge will be largely penalized by experiencing excessiveservice delays and significant outage. The PF offers a good trade-off between RRand the maximum C/I. The PF schedules users according to the ratio of theirinstantaneous data rate to average served data rate. This results in all users get-ting an equal probability of being served even though they may experience very


different channel quality. This allows having a good balance between the systemthroughput and the user fairness.

2.5 Basic CDMA Procedures

As previously described, each CDMA base station transmits a different PNsequence (gold code). In WCDMA, there are 512 primary scrambling codesavailable. Each scrambling code has low cross-correlation with any other scram-bling code regardless of the timing offset between the two scrambling codes.This allows the base stations to be deployed asynchronously. In CDMA2000,base stations transmit the same PN sequence (M-sequence) offset by differentamounts of time. There are 512 PN offsets available. An M-sequence transmit-ted with a given PN offset has low autocorrelation with any other PN offset. Toguarantee that each base station transmits a PN offset that is distinct from thosein its vicinity, base stations need to be synchronized so that they have a commonsense of timing. Figure 2.20 shows a typical CDMA cell layout where the colorof each cell indicates a different PN offset. This approach is necessary so userscould identify the different base stations or sectors and to enable a CDMAphone to acquire the system. This acquisition process is one of several possiblestates for a mobile phone or UE. Other states shown in Figure 2.21 include idle,access, and dedicated.

2.5.1 Acquisition State

Acquisition means acquiring the system. It is done upon power up or loss of ser-vice. For example, in WCDMA acquisition consists of a three-step process


Figure 2.20 CDMA cell layout.

designed to simplify the 512 x 38,400 search space, where there are 512 primaryscrambling codes (PSCs) and 38,400 possible chip offsets of the PSC. The UEacquisition begins by searching and finding the synchronization channel (SCH).On the SCH, the primary SCH (P-SCH) and the secondary SCH (S-SCH) aresent simultaneously, as shown in Figure 2.22. These two channels are coded dif-ferently. The P-SCH is coded with the primary synchronization code, which isthe same for every cell in the system. The S-SCH is coded with the secondarysynchronization code, comprising 64 different code groups. When the codegroup is determined (step 2), the PSC is determined on the CPICH. There areeight PSCs per code group; the UE tries these eight combinations.


Power up/initialization statePhone acquires system.

CPICH, SYNC .

Access statePhone accesses the network for callorigination.

RACH,PRACH

Idle statePhone receives overhead informationon the paging channel.

PCH,P-CCPCH,S-CCPCH,PICH

Traffic/dedicated stateA dedicated channel is allocated to thephone.

FCH,SCH,DPCH

Figure 2.21 CDMA call states.

One frame (38400 chips, T =10 ms)

256 chips

Common pilot channel

Tf

One slot (2560 chips)256 chips

PSC PSC PSCPSCSSC SSC SSC SSC

Figure 2.22 WCDMA cell search signals.

When the PSC is found, the UE resolves a 10 ms ambiguity on the broad-cast channel (BCH). Finally, the UE can start to demodulate, or read, the trans-port BCH on the PCCPCH. This process, also know as cell search, issummarized as follows.

Step 1: Slot synchronization

During the first step of the cell search procedure, the UE uses the SCH’s pri-mary synchroniszation code to acquire slot synchronization to a cell. This is typ-ically done with a single matched filter matched to the primary synchronizationcode, which is common to all cells. The slot timing of the cell can be obtainedby detecting peaks in the matched filter output. Figure 2.23 shows a depictionof this step.

Step 2: Frame synchronization and code-group identification

During the second step of the cell search procedure, the UE uses the SCH’s sec-ondary synchronization code to find frame synchronization and identify thecode group of the cell found in the first step. This is done by correlating thereceived signal with all possible secondary synchronization code sequences, andidentifying the maximum correlation value. Since the cyclic shifts of thesequences are unique, the code group as well as the frame synchronization isdetermined.

Step 3: Scrambling-code identification

During the third and last step of the cell search procedure, the UE determinesthe exact primary scrambling code used by the found cell. The primary scram-bling code is typically identified through symbol-by-symbol correlation over theCPICH with all codes within the code group identified in the second step. Afterthe primary scrambling code has been identified, the primary CCPCH can bedetected and the system- and cell-specific BCH information can be read. Thisoperation is shown in Figure 2.24. If the UE has received information aboutwhich scrambling codes to search for, steps 2 and 3 above can be simplified.


2560 chips2560 chips

Indicates timing of PSC

Matchedfilter

Figure 2.23 WCDMA acquisition step 1.

In CDMA2000, the mobile first gains some idea of system timing bysearching for usable pilot signals. The mobile can first align its own timing bycorrelating with the pilot. When this correlation is found, the mobile synchro-nizes with the synchronization channel. It reads the sync channel message, andchanges to absolute time. When the mobile is synchronized to absolute time, itcan then read the broadcast messages on the F-PCH or F-BCCH.

2.5.2 Idle State

Once the phone has acquired the system upon power up or loss of service,it enters into an idle state except if a call needs to be placed. During this state,the phones monitors the PCH or the P-CCPCH to get updated system informa-tion or to receive pages. To save power, the mobile periodically enters alow-power or “sleep” state, during which it cannot receive data from the net-work. To page a mobile while it is in this state, the network assigns times whenthe mobile should “wake up” and receive any paging messages. Both WCDMAand CDMA2000 use a paging indication protocol. A paging indicator indicatesto the mobile whether it has a paging channel message. If the paging indicatorindicates that the mobile has a paging message, the mobile demodulates theentire paging channel message. Otherwise, it returns to sleep until the nextpaging occasion.

2.5.3 Access State and Call Setup

The mobile enters the access state whenever it needs to get dedicated networkresources, respond to a page, or establish a voice or data call. An access channel(ACH) is used by the mobile to send requests to the network to set up a dedi-cated connection. The same procedure is used in both CDMA2000 andWCDMA, although the channel structure and messages contents and sequenceare different. Let us assume that the mobile phone needs to place a call; the fol-lowing summarizes what happens to establish this call:

• The first step in call setup is called origination, in which the phonesends a message to the network over the ACH or PRACH to request aconnection.


Correlatorbank 10 ms

Correct code

10 ms

Common pilot

Figure 2.24 WCDMA acquisition step 3.

• The network then responds to the phone to acknowledge the messagereceipt and inform the mobile about the connection setup and codechannels to be assigned. This is done over the PCH or S-CCPCH.

• Once the network authenticates the mobile, both the phone and thenetwork start negotiating the type of service and data rate required.

• When the service has been negotiated, the network assigns RF andhardware resources to the call and this information is exchanged withthe mobile so that a dedicated channel [fundamental channel (FCH),SCH, or DPCH] could be established.

2.5.4 Traffic or Dedicated State

Call setup is concluded by assigning a dedicated traffic channel to the mobile. InCDMA2000 this could be the FCH, the SCH, or a combination of both,depending on the type of call and data rate. In WCDMA, the network assigns aDPCH for the same purposes. Although the procedures used for call setup aresimilar, the GSM layering structure (RRC, MM, CC) of WCDMA requireslayer 3 signaling to be exchanged at many different layers, resulting in more sig-naling messages. In CDMA2000, the messaging is more streamlined.

2.6 CDMA Embedded Cell Capacity

The isolated CDMA cell capacity derived in Chapter 1 assumes only a singlecell and ignores the interference from users in neighboring cells. The capacity ofan isolated cell in a narrowband system would also be very high since a reuse fac-tor of one can be employed and all channels can be assigned in the 1.25-MHzbandwidth. In fact, CDMA makes a big difference when the impact of neigh-boring cells is taken into account. Let us rewrite the isolated cell pole capacitygiven by

NW R

E

N

as Ppoleb

t

= → ∞ (2.17)

in which the interference was averaged only over the users in the same cell. Toquantify the potential improvements of smart antennas in CDMA systems thecharacteristics of the interference must be understood. On the downlink, severalbase stations are radiating and a mobile unit suffers from interference from othercells as well as from its home cell. On the uplink of a direct sequence CDMAsystem, the capacity is related to the Eb/Nt, as was shown in Chapter 1. If Eb/Nt is


too low, the frame error rate (FER) or BLER will be high and the system perfor-mance will degrade. If Eb/Nt is too high, the interference level will increase andthis will decrease the reverse link capacity. In TDMA and GSM systems, forevery time slot there is one desired signal and a very small number of cochannelinterferers, which makes the application of adaptive interference cancellationpractical. On the other hand, in low-rate CDMA systems (i.e., for voice-domi-nated services), due to the large number of users sharing the channel, the inter-ference in these systems is typically assumed to be statistically close to whiteGaussian noise. And since in most cases the users can be assumed to be uni-formly distributed across the cell, interference can also be considered spatiallywhite in most operating scenarios. This means that isolating individual usersrequires arrays of large size. In CDMA2000 and WCDMA systems, a largenumber of voice users are mixed with a smaller number of high-speed data users.Since high data rate users have a lower processing gain, in order for them tomaintain the same required Eb/Nt as voice users, their transmit power must bemuch stronger than voice users. As a result, on the reverse link high data rateusers will present strong directional interference in the reception of voice usersand the interference observed by voice users will be colored by data users and isno longer approximated as white Gaussian. In these mixed voice and high datarate systems an interference cancellation/reduction algorithm can be effectivelyused to null/reduce the impact of the limited number of high bit rate users,thereby increasing the overall system capacity. In CDMA systems, the uplinkpole capacity of an embedded cell is given by [4]

( )N

W RE

Nf

Gpoleb

t

s=+1 ν

(2.18)

where v is the voice activity factor, Gs is the sectorization gain, Nt is the totalnoise + interference power spectral density, and f is the reuse efficiency definedas

fI

Ioc

sc

= (2.19)

In (2.19), Ioc denotes the other cell interference power and Isc is the samecell interference power. From (2.18) it can be readily seen that reducing theinterference level increases the number of maximum supportable users in a sec-tor. It turns out that the fraction of the uplink interference that comes from theneighbor cells is about 60% of the own-cell interference and this ratio is not verysensitive to the parameters of the model, provided the assumption that the


mobiles are power-controlled in a sensible way still holds. The effective fre-quency reuse factor in CDMA can be calculated as

F f= +1 (2.20)

F plays the same role in the CDMA capacity equation as that of thenarrowband frequency reuse factor K in TDMA and GSM systems.

2.6.1 Multipath Fading

Just as system capacity is affected by interference, it is also affected by propagationphenomena. Fading in a moving vehicle is more rapid than for pedestrians beingcaused by motion of the vehicle through stationary interference patterns, wherethe spatial scale of the interference pattern is the wavelength. We can address theimpact of multipath fading on the performance of CDMA by first understandingunder what circumstances will fading affect CDMA and what is that effect on theCDMA channel. When the multipath components’ delays separated by at leastthe decorrelation time of the spreading, they can be resolved by the CDMA wave-form and can be separated by the despreader in the receiver because each compo-nent correlates at a different delay. When the multipath components areseparated by less than the decorrelation time, then they cannot be separated in thereceiver, and they interfere with one another, leading to flat fading.

The duration of one spreading chip is 1/1.2288 Mcps = 814 ns inCDMA2000 and 1/3.84 Mcps = 260 ns in WCDMA. Multipath differencesless than those will lead to flat fading, whereas greater separations will lead toresolvable multipath, which will be diversity combined by the receiver. Theeffects of fading depend mainly on the fading rate, which in turn depends on thevelocity of the mobile station. Fading increases the average Eb/Nt needed for aparticular error rate, which in turns causes capacity degradation. Coverage is alsoaffected because a certain fading margin has to be built into the link budget.Power control mitigates the effects of fading at low speed and, to a less degree, athigh speed. At high speed, the forward error correction coding and interleavingbecomes more effective in combating fading as the characteristic fade timebecomes less than the interleaver span.

2.7 Coverage Versus Capacity Trade-Off

There is an inherit trade-off in CDMA between the capacity and coveragebecause of the way interference affects performance. On the uplink, interferenceincreases as the load is increased and follows the expression of 1/(1–η), where ηis fractional loading defined as N/Npole. Figure 2.25 shows how the interference


rises over thermal noise as the fractional loading increases. To reach the polecapacity, the power that the mobiles are required to transmit goes to infinity toovercome this interference. As the required power increases, mobiles at the edgeof coverage will begin to run out of transmitter power. That is, they will beasked to transmit more than their capability allows. It then follows that the sys-tem load should be controlled so that the planned service area never experiencescoverage failures because of this phenomenon. This trade-off is not so much aproblem or a limitation of CDMA systems as it is a system design consideration,which implies that maximum capacity and maximum coverage cannot be simul-taneously achieved.

2.7.1 Coverage-Capacity Trade-Off in the Uplink

Simple capacity models of the reverse link show that RF power rises with load-ing to overcome interference, as previously discussed. Real systems must operatebelow the pole capacity because real user stations have an upper bound to theirtransmitter power. As the load is increased, the average interference level alsoincreases. Because mobile stations are transmitting at their maximum power, S,their corresponding received power at the base station is fixed and N is growingrapidly; that is, SNR degrades. This in turns means that those users would needto move closer to the base station to a point where a smaller path loss allowstheir received power and, consequently, SNR to be restored to the target setpoint necessary to achieve the required quality of service. In effect, the cell cover-age shrinks by the same range. The effect of traffic loading on the range from acell site is referred to as cell breathing. For example, for a loading of 50% of thepole capacity η = 0.5, the loss of coverage on the uplink is 1/(1–0.5) = 2, or 3 dBloss. The impact of the interference rise over thermal noise on the CDMA cover-age is illustrated in Figure 2.25. CDMA networks are typically planned with this


Figure 2.25 CDMA cell breathing.

fractional loading level. That is, the network is designed with the appropriatemargins, and sufficient number of sites and the site locations are optimized sothat there is no performance degradation when the system is loaded.

In a poorly planned CDMA system, increasing this loading would lead torange reductions and open coverage holes under high loads. The pole capacity ofa cell depends only on the average Eb/Nt target, the processing gain, and voiceactivity factor. The coverage area of a cell defined as the area over which all usersobtain the target Eb/Nt, depends on the fractional loading relative to the polecapacity. Detailed analysis of the interaction of coverage and capacity is a com-plex process involving power control, soft handoff, fading, the mobility mix ofsubscribers, and other factors, as well as the differences between downlink anduplink.

2.8 Conclusion

Third generation mobile communications systems are already offering signifi-cant improvements over their 2G counterparts in peak data rates and through-put. However, impairments caused by the propagation channel such asmultipath and interference caused by other users in the system still representchallenges for the network and system design. As we have seen, in CDMA sys-tems the performance of most users except those at the cell edge is interferencelimited. Techniques to reduce the average interference would therefore signifi-cantly improve performance in terms of capacity and coverage. As such, thetechnology of smart antennas can be considered as complimentary to existingmultiple access techniques and an extra tool at the disposal of the systemdesigner.

References

[1] Faruque, S., Cellular Mobile Systems Engineering, Norwood, MA: Artech House, 1996.

[2] Garg, V. K., and J. E. Wilkes, “Principles and Applications of GSM, Upper Saddle River,NJ: Prentice Hall, 1999.

[3] Rappaport, T., Wireless Communications, Principles and Practices, New York: IEEE Pressand Prentice Hall, 1996.

[4] Yang, S. C., CDMA RF System Engineering, Norwood, MA: Artech House, 1998.

[5] Garg, V. K., IS-95 CDMA and CDMA2000 Cellular/PCS Systems Implementation, UpperSaddle River, NJ: Prentice Hall, 2000.

[6] TIA/EIA IS-95A, “Mobile Station-Base Station Compatibility Standard for Dual-ModeWideband Spread Spectrum Cellular System,” 1995.


[7] TIA/EIA-95B, “Mobile Station-Base Station Compatibility Standard for WidebandSpread Spectrum Cellular Systems,” 1999.

[8] ANSI J-STD-008, “Personal Station-Base Station Compatibility Requirements for 1.8 to2 GHz CDMA Personal Communications Systems,” 1995.

[9] Ojanpera, T., and R. Prasad, Wideband CDMA for Third Generation Mobile Communica-tions, Norwood, MA: Artech House, 1998.

[10] TIA/EIA/IS-2000.1-A, “Introduction to CDMA2000 Standard for Spread Spectrum Sys-tems,” 2000.

[11] TIA/EIA/IS-2000.2-A, “Physical Layer Standard for CDMA2000 Spread Spectrum Sys-tems,” 2000.

[12] Esteves, E., M. Gurelli, and M. Fan, “Performance of Fixed Wireless Access withCDMA2000 1xEV-DO,” IEEE 58th Vehicular Technology Conference, Vol. 2, October6–9, 2003, pp. 836-840.

[13] Kuenyoung, K., K. Hoon, and H. Youngnam, “A Proportionally Fair Scheduling Algo-rithm with QoS and Priority in 1xEV-DO, Personal, Indoor, and Mobile Radio Commu-nications,” 13th IEEE Intl. Symp., Vol. 5 , September 15–18, 2002, pp. 2239–2243.

[14] Huang, C. Y., et al., “Schedulers for 1xEV-DO: Third Generation Wireless High-SpeedData Systems,” 57th IEEE Semiannual Vehicular Technology Conference, Vol. 3, April22–25, 2003, pp. 1710-1714.

[15] Yavuz, M., and Paranchych, D. W., “Adaptive Rate Control in High Data Rate WirelessNetworks,” IEEE Wireless Communications and Networking, Vol. 2, March 16–20, 2003,pp. 866-871.

[16] Sindhushayana, N. T., and P. J. Black, “Forward Link Coding and Modulation forCDMA2000 1XEV-DO (IS-856),” 13th IEEE Intl. Symp. on Personal, Indoor, and MobileRadio Communications, Vol. 4, September 15–18, 2002, pp. 1839–1846.

[17] Yonghoon, C., and H. Youngnam Han, “A Channel-Based Scheduling Algorithm forCDMA2000 1xEV-DO System,” 5th Intl. Symp. on Wireless Personal Multimedia Commu-nications, Vol. 2, October 27–30, 2002, pp. 621–625.

[18] Yavuz, M., et al., “Performance Improvement of the HDR System Due to Hybrid ARQ,”IEEE VTS 54th Vehicular Technology Conference, Vol. 4, October 7–11, 2001, pp.2192–2196.

[19] Chung, W., W. L. Hong, and M. Jungbae, “Downlink Capacity of CDMA/HDR,” IEEEVTS 53rd Vehicular Technology Conference, Vol. 3, May 6–9, 2001, pp. 1937–1941.

[20] Qualcomm, Inc., http://www.qualcomm.com/technology/1xev-do/whitepapers.html.

[21] Holma, H., and A. Toscala, WCDMA for UMTS, New York: John Wiley & Sons, 2004.

[22] 3GPP Technical Specification 25.211, Physical Channels and Mapping of TransportChannels onto Physical Channels (FDD).

[23] 3GPP Technical Specification 25.212, Multiplexing and Channel Coding (FDD).

[24] 3GPP Technical Specification 25.213, Spreading and Modulation (FDD).

[25] 3GPP Technical Specification 25.214, Physical Layer Procedures (FDD).

[26] Kolding, T. E., et al., “High Speed Downlink Packet Access: WCDMA Evolution,” IEEEVehicular Technology Society News, February 2003, pp. 4–10.


[27] Parkvall, S., et al., “WCDMA Evolved–High Speed Packet Data Services,” EricssonReview, No. 2, 2003, pp. 56–65.

[28] Helmersson, K. W., and G. Bark, “Performance of Downlink Shard Channels inWCDMA Radio Networks,” Proc. IEEE Vehicular Technology Conference, Vol. 4, Spring2001, pp. 2690–2694.

[29] 3GPP TS25.855, “High Speed Downlink Packet Access; Overall UTRAN Description.”2001.

Selected Bibliography

CDMAAndersen, N. P., M. Pecen, and I. Gonorovsky, “GSM/EDGE Evolution, Based on 8-PSK Cir-cuits and Systems,” Proc. of the 2003 International Symposium on ISCAS, Vol. 3, May 25–28,2003 pp. III-598–III-601.

Cai, J., and D. J. Goodman, “General Packet Radio Service,” GSM Communications Magazine,IEEE, Vol. 35, No. 10, October 1997, pp. 122–131.

Christensen, G., et al., Wireless Intelligent Networking, Norwood, MA: Artech House, 2000.

“The Evolution of Digital Wireless Technology from Space Exploration to Personal Communi-cation Services,” IEEE Trans. on Vehicular Technology, Vol. 43, No. 3, August 1994.

“Four Laws of Nature and Society: The Governing Principles of Digital Wireless Communica-tion Networks,” Wireless Communications: Signal Processing Perspective, H. V. Poor and G. W.Wornell, (eds.), Upper Saddle River, NJ: Prentice Hall, 1998, pp. 380–392.

Gilhousen, K. S., et al., “On the Capacity of a Cellular CDMA System,” IEEE Trans. Veh. Tech.,Vol. 40, No. 2, 1991, pp. 303–312.

Glisic, S., and V. Branka, Spread Spectrum CDMA Systems for Wireless Communications,Norwood, MA: Artech House, 1997.

Hallmann, E., and R. Helmchen, “Investigations on the Throughput in EDGE and GPRS RadioNetworks,” IEEE VTS 53rd Vehicular Technology Conference, Vol. 4, May 6–9, 2001, pp.2823–2827.

Jakes, W. C. Jr., Microwave Mobile Communications, New York: John Wiley & Sons, 1974;reprinted by IEEE Press, 1994.

Lee, J. S., and L. E. Miller, CDMA Systems Engineering Handbook, Norwood, MA: Artech House,1998.

Lee, W. C. Y., Lee’s Essentials of Wireless Communications, New York: McGraw-Hill, 2000.

Lee, W. C. Y., Mobile Cellular Telecommunications, 2nd ed., New York: McGraw-Hill, 1995.

Molkdar, D., W. Featherstone, and S. Larnbotharan, “An Overview of EGPRS: The Packet DataComponent of EDGE,” Electronics & Communication Engineering Journal, Vol. 14, No. 1, Feb-ruary 2002, pp. 21–38.

Parsons, D., The Mobile Radio Propagation Channel, New York: John Wiley & Sons, 1992.

Peterson, R. L., R. E. Ziemer, and D. E. Borth, Introduction to Spread Spectrum Communications,Englewood Cliffs, NJ: Prentice Hall, 1995.

Prasad, R., CDMA for Wireless Personal Communications, Norwood, MA: Artech House, 1996.


Pribylov, V. P., and I. I. Rezvan, “On the Way to 3G Networks: The GPRS/EDGE Concept,”Proc. of the 4th IEEE-Russia Conference on MEMIA, December 23–26, 2003, pp. 87–98.

Proakis, J. G., Digital Communications, 2nd ed., New York: McGraw-Hill, 1989.

Simon, M. K., et al., Spread Spectrum Communication Handbook, New York: McGraw-Hill,1994.

Ross, A. H. M., and K. S. Gilhousen, “CDMA Technology and the IS-95 North American Stan-dard,” in The Mobile Communications Handbook, Boca Raton, FL: CRC Press and IEEE Press,1996, pp. 430–448.

Scholtz, R. A., “The Origins of Spread Spectrum Communications,” IEEE Trans. Commun.,COM-30, May 1982 (Part I), pp. 822–854.

Shannon, C. E., “Communication in the Presence of Noise,” Proc. IRE 37, January 1949, pp.10–21.

Turin, G. L., “Introduction to Spread Spectrum Antimultipath Techniques and Their Applica-tion to Urban Digital Radio,” Proc. IEEE 68, 1980, pp. 328–354.

Viterbi, A., CDMA: Principles of Spread Spectrum Communication, Reading, MA: Addison-Wes-ley, 1995.

Viterbi, A. J., “Error Bounds for Convolutional Codes and an Asymptotically Optimum Decod-ing Algorithm,” IEEE Trans. Inform. Th. IT-13, 1967, pp. 260–269.

Viterbi, A. J., A. M. Viterbi, and E. Zehavi, “Performance of Power-Controlled Wideband Ter-restrial Digital Communications,” IEEE Trans. on Comm., Vol. 41, No. 4, 1993, pp. 559–569.

Viterbi, A. J., et al., “Soft Handoff Extends CDMA Cell Coverage and Increases Reverse LinkCapacity,” IEEE J. Selected Areas in Communications, Vol. 12, No. 8, 1994, pp. 1281–1288.

Viterbi, A. M., and A. J. Viterbi, “Erlang Capacity of a Power Controlled CDMA System,” IEEEJ. on Selected Areas in Communication, Vol. 11, No. 6, 1993, pp. 892–900.

Yallapragada, R., V. Kripalani, and A. Kripalani, “EDGE: A Technology Assessment,” IEEEInternational Conference on Personal Wireless Communications, December 15–17, 2002, pp.35–40.

Yang, S. C., CDMA RF System Engineering, Norwood, MA: Artech House, 1998.

GSM/GPRS/EDGEBates, R. J., GPRS: General Packet Radio Service, New York: McGraw-Hill, 2001.

Mehrotra, A., GSM System Engineering, Norwood, MA: Artech House, 1997.

Redl, S. M., et al., GSM and Personal Communications Handbook, Norwood, MA: Artech House,1998.

Redl, S. M., et al., Introduction to GSM, Norwood, MA: Artech House, 1996.

Seurre, E., et al., EDGE for Mobile Internet, Norwood, MA: Artech House, 2003.

Steele, R., et al., GSM, cdmaOne and 3G Systems, New York: John Wiley and Sons, 2001.

Timo, H., GSM, GPRS, and EDGE Performance, 2nd ed., New York: John Wiley & Sons, 2003.


3Spatial Channel Modeling

3.1 Introduction

The detailed knowledge of radio propagation characteristics is essential todevelop a successful wireless system. Measurement studies have been carried outto identify propagation loss, spatial distribution of power, wideband andnarrowband statistics concerning the random variables of received signals at afixed location due to any surrounding movement, and delay spread. A radiochannel is a generally hostile medium in nature. Transmitted signals undergoseveral propagation phenomena such as reflection, diffraction, and scatteringbefore they reach the receiver. This is mainly caused by the existence of objectsin the physical channel between a transmitter and a receiver such as buildings,trees, mountains, hills, and moving objects. Therefore it is rather difficult topredict the channel behavior. Traditionally, radio channels are modeled in a sta-tistical way using real propagation measurement data. Classical propagationmodels, commonly used for narrowband transmission systems, represent a sig-nal in the radio environment using a large-scale path loss component togetherwith a medium-scale slow varying component having a lognormal distribution,and a small-scale fast varying component with a Rician or Rayleigh distribution,depending on the presence or absence of the line-of-sight (LOS) componentbetween the transmitter and receiver [1, 2]. Accordingly, conventional radiopropagation models describing a wireless cellular environment have focused on:

• Area mean power depending on the path loss characteristics betweenthe transmitter and receiver.

• Local mean power within that area, which is slow varying. This can berepresented by a lognormal distribution.

61

• Superimposed fast fading instantaneous power, which follows a Ray-leigh or nonline-of-sight (NLOS) or Rician distribution (LOS).

Figure 3.1 illustrates a typical propagation environment. Large variations inthe transmission path between the transmitter and receiver can be found, rangingfrom direct LOS to severely obstructed paths due to buildings, mountains, orfoliage. The phenomenon of decreasing received power with distance due toreflection, diffraction around structures, and refraction within them is known aspath loss. Propagation models have been developed to determine this path lossand are known as large-scale propagation models because they characterize thereceived signal strength by averaging the power level over large transmit-ter-receiver separation distances, in the range of hundreds or thousands of meters.

On the other hand, medium-scale propagation models determine thegradual changes in the local-mean power if the receiving antenna is moved overdistances larger then a few tens or hundreds of meters. The medium-scale varia-tion of the received signal power is called shadowing, and it is caused by obstruc-tion by trees and foliage. The term local-mean power is used to denote thepower level averaged over a few tens of wavelengths, typically 40λ. Finally,small-scale propagation models characterize the fast variation of the signalstrength over a short distance on the order of a few wavelengths or over shorttime duration on the order of seconds. Small-scale fast fading, also known asshort-term fading or multipath fading, is due to multipath reflections of a trans-mitted wave by local scatterers such as houses, buildings, and man-made


Direct path

Multipath

Multipath

Multipath

Figure 3.1 Mobile radio propagation channel.

structures, or natural objects such as forests surrounding a mobile unit. Typi-cally, detailed models are needed for a complete coverage and capacity analysisin a certain region. On the other hand, if only rough capacity and range calcula-tions are needed, simple and easy-to-use models are sufficient. In addition to themodeling of the propagation environment, the mobility of the wireless terminalsalso needs to be understood in each radio environment. Mobility modeling has asignificant impact on the analysis of radio resource management, channel alloca-tion, and handoff performance.

3.2 Radio Environments and Cell Types

There are a large number of environments where mobile radio systems can oper-ate. These include large and small cities, with variations in building construc-tion, as well as tropical, rural, desert, and mountainous areas. Moreover,antenna design and height impacts the radio environments. Since it is impossi-ble to consider all possible radio environments in the design of a mobile radiosystem, more general models that will consider the essence of different radioenvironments are required. Therefore, the large number of possible radio envi-ronments has to be condensed into a finite set of generic radio environments[3–6]. One approach is to classify a radio environment based on the typical cellsize, which leads to:

• Macrocells: In a macrocell the base station antenna is placed above therooftops and is much higher than the mobile users. Usually macrocellshave a radius of more than 1 km and can be found in rural as well asurban areas.

• Minicells: A minicell can be considered as a small macrocell where thebase station antenna is placed at the same height as the rooftops. Thistype of cell is only used in urban environments with cell radius rangingfrom 700m to 1 km.

• Microcell: In a microcell the base station antenna is placed in street levelwith a typical antenna height in the range of 5m. It has a cell radius oftypically 200 to 500m and is mainly used for increasing the coverageand capacity in a dense urban environment.

• Picocell: A picocell is mainly for indoor usage. The cell radius is about10 to 20m and is limited by the building itself due to high penetrationlosses in the walls, number of floors, and their compositions.

Another approach is to classify radio environments based on the nature ofthe mobile users being served by the system. This leads to the followingclassification:

Spatial Channel Modeling 63

• Vehicular radio environment: The vehicular environment is character-ized by large macrocells and large transmit powers as well as highmobile speeds (fast-moving vehicles). There is typically no LOS com-ponent, and the received signal is mostly composed of reflections. Inthese environments, the average power of the received signal decreaseswith distance raised to some exponent, referred to as the path loss expo-nent, which varies depending on the environment but is typicallybetween three and five. In addition, shadowing is caused by obstructionfrom trees and foliage, and the resulting medium-scale variation in thereceived signal power can be modeled with a lognormal distribution.The standard deviation of that power varies considerably; for example,8 to 10 dB is generally used in urban and suburban areas, whereas alower value is used in rural and mountainous areas. In addition,small-scale fading is characterized by Rayleigh distribution. Typicaldelay spreads in this case are on the order of 0.8 µs but can be as high astens of microseconds.

• Outdoor to indoor and pedestrian radio environment: This radio environ-ment is characterized by small microcells and low transmit powers withthe antennas usually located below rooftops. Both LOS and NLOSmultipath components exist. Indoor coverage can also be providedfrom this outdoor base station. The path loss exponent varies quite a bitand can be anywhere from two in areas with LOS up to six with NLOScases due to trees and other obstructions along the path. Furthermore, amobile station can experience a sudden drop of 15 to 25 dB when itmoves around a corner. The standard deviation of the shadowing inthese environments varies from 10 to 12 dB and the small-scale fadingis either Rayleigh (NLOS) or Rician (LOS), with typical delay spreadson the order of 0.2 µs.

• Indoor office radio environment: In the indoor office radio environmenttransmit powers are small and base stations and users are locatedindoors. Path loss attenuation exponent varies from two to five depend-ing on the scattering and attenuation by walls, floors, and metallicstructures. Note that each one of these environments has individualcharacteristics with respect to path loss attenuation, shadowing, andsmall-scale fading.

3.3 The Multipath Channel

The short-term fluctuations of the received signal caused by multipath propaga-tion are called small-scale fading. The different propagation path lengths of themultipath signal give rise to different propagation time delays. A multipath


channel can be represented by a power-delay profile consisting of different setsof distinct paths, which are also called multipath taps. Depending on the phaseof each multipath signal when arriving at the receiver, they sum either construc-tively or destructively. Consequently, the power of each multipath tap is timevarying, resulting in fading dips. The depth of the fading dips depends on thechannel type. Using a baseband complex envelope representation and modelingthe RF channel as a time-variant channel, we can represent the classical channelimpulse response as a sum of L multipath components given by [1]

( ) ( ) ( ) ( )( )h t A t e t tlj t

ll

Ll, τ δ τϕ= −

=∑

0

(3.1)

where Al(t) is the amplitude, ϕl(t) is the phase, and τl(t) is the time delay of thesignal component. The distribution of the instantaneous power of the channeltaps can be described by a distribution function, which depends on the radioenvironment. A so-called Rayleigh fading channel is the most severe mobileradio channel, with deep fading dips every λ/2. In a Rayleigh fading channel allmultipath taps are independent and there is no dominant path. The envelope ofindividual multipath components in this case can be characterized by a Rayleighdistribution given by:

( )( )

p rr

e r

r

r

= ≤ ≤ ∞

<

−

σσ

2

22 2

0

0 0(3.2)

In a Rician fading channel, the fading dips are shallower due to a domi-nant path in addition to the scattered paths. This is usually the case in microcelland picocell environments due to the existence of LOS.

3.4 Channel Characterization

The multipath fading channel can be characterized based on delay spread,coherence bandwidth, Doppler spread, and coherence time. The root-mean-square (rms) value of the delay spread is a statistical measure that describesthe spread of the multipath components around the mean delay of the chan-nel. The maximum delay spread tells the delay difference between the firstand last multipath components in the power-delay profile. The coherence band-width is the maximum frequency difference for which the signals are stillstrongly correlated. The coherence bandwidth is inversely proportional to thedelay spread (i.e., the smaller the delay spread the larger the coherence band-width). If the transmission bandwidth of the signal is larger than the coherence


bandwidth, the signal will undergo frequency selective fading. On the other hand,a flat fading channel results if the transmission bandwidth of the signal is smallerthan the coherence bandwidth. The coherence bandwidth can be thought of asbeing a measure of the diversity available to a RAKE or equalized receiver. Asmaller coherence bandwidth means a higher order diversity. If the coherencebandwidth is as large as the transmission bandwidth, then the entire receivedspectrum would be observed to fade. The maximum delay spread can be used tocalculate how many resolvable paths exist in the channel that could be used inthe RAKE receiver. In addition, the movement of the mobile station gives rise toa Doppler spread, which is the width of the observed spectrum when anunmodulated carrier is transmitted. If there is only one path from the mobile tobase station, the base station will observe a zero Doppler spread combined witha simple shift of the carrier frequency (Doppler frequency shift). The Dopplerfrequency varies depending on the angle of the mobile station movement rela-tive to the base station. The range of values when the Doppler power spectrumis nonzero is called the Doppler spread. The reciprocal of the Doppler shift is ameasure of the coherence time of the channel. The coherence time is the durationover which the channel characteristics do not change significantly.

3.5 Path Loss Models

Typically, path loss models are derived using a combination of analytical andempirical methods. In the empirical approach, the measured data is modeledusing curve fitting or analytical expressions. The validity of empirical models inother environments and frequencies can only be validated by comparing themodel to data measured from the specific area and for the specific frequency. Itshould be noted that these models present only a snapshot of the real radioenvironment.

3.5.1 Okumura-Hata Propagation Models

The empirical Okumura method [2] is based on exhaustive measurements thatwere performed in the Tokyo metropolitan area. The results were a series ofcurves, plotting recorded field strength as a function of distance from the trans-mitting antenna. The model is valid for distance ranges of 1 km to 100 km, fre-quency bands from 150 MHz to 2,000 MHz, and base station effective antennaheights from 30m to 1,000m.

3.5.1.1 Hata’s Model

Since Okumura’s curves and tables were intended to be used manually as alook-up resource, Hata was able to derive equations from Okumura’s work.This allowed for accurate computation of path loss without having to peer


through any set of graphs. The standard Okumura-Hata model for an urbancity is given by:

Max. Path Loss = 69.55 + (44.9 − 6.55 log Hb ) log R − 13.82 logHb + 26.16 log f −a( Hm ) (3.3)

where Hb is the radio base station (RBS) effective antenna height, Hm is themobile subscriber antenna height, f is the operating frequency, and R is the dis-tance between the RBS and the mobile station (MS), or the “radius” to RBSfrom the measurement point. The propagation model is valid under the follow-ing conditions: 150 < f < 1,500 MHz, 1 < R < 20 km, 30 < Hb < 200m and 1 <Hm < 10m. The mobile height correction factor a(Hm) can be computed asfollows:

a(Hm) = 3.2 ( log (11.75 Hm) )1/2 – 4.97, when f ≥ 400 MHz for urban en-vironments

a(Hm) = (1.1 log f – 0.7) Hm – 1.56 log f + 0.8 for suburban or rural areas.

3.5.1.2 The COST-231 Model (Suburban)

This model has been developed by the European Union’s Forum for Co-opera-tive Scientific and Technical Research (COST). Since the traditionalOkumura-Hata model is restricted to application in the frequency band below1,500 MHz, it is not applicable in either the PCS or IMT-2000 spectrumregions. The COST-231 model [2] was developed based on analysis ofOkumura’s propagation curves in the higher frequency regions with the aim ofimplementing a suitable formula that characterizes radio wave propagation inthe PCS and IMT-2000 bands. The results led to the following adaptation ofthe Okumura-Hata equation:

Max. Path Loss = 46.3 + (44.9 – 6.55 log Hb) log R – 13.82 log Hb

+ 33.9 log f – a(Hm) (3.4)

where:

a(Hm) = 3.2 ( log (11.75 Hm) )1/2, when f ≥ 400 MHz for large cities

a(Hm) = (1.1 log f – 0.7) Hm – 1.56 log f + 0.8 for medium small cities orrural areas.

3.6 Spatial Channel Modeling

As we can see from previous sections, classical propagation models focus on thepower delay profile without taking into account the angular distribution of the


multipath components. To analyze the performance impact of smart antennas atthe link and system level, the spatial domain must also be considered. Channelmodels that characterize the DOA of multipath components are referred to asspatial channel models. Taking into account the angular dependence of thechannel, the directional channel impulse response can be written as [7]:

( ) ( ) ( ) ( ) ( )( ) ( )( )h t A t e a t t t tlj t

l ll

L

ll, , , ,τ θ φ θ φ δ τϕ= −

=∑

0

(3.5)

where α(θ, φ) is the array response vector given by:

( ) ( )[ ]a e ej j M Tθ φ ψ ψ, = − − −1 1K (3.6)

As can be seen from (3.5) the spatial channel can be characterized by anumber of multipath components L, each with a complex amplitude Al, eleva-tion angle Al, azimuth angle θl, and delay τl. We have previously seen how classi-cal channel models define envelope probability distribution functions, delay andDoppler spread ranges for different radio environments, and cell types. Thesame approach can also be adopted for spatial channel modeling to define thecharacteristics of the parameters that make up the directional channel impulseresponse. In order for (3.5) to be used in link level or any other type of simula-tion intended to study the performance of smart antennas, we either have todefine the amplitude, time delay, and angular spread distributions of the differ-ent multipath components or the spatial distribution of the different scatterersin the channel that can then be used to generate different multipathcomponents.

3.6.1 Spatial Channel Model Parameters

It has been known that multipath components tend to cluster in groups thatcould be exploited in modeling the structure of the directional or angularimpulse response. For instance, a cluster can be viewed as a collection ofmultipath components that experience the same small-scale variations sincesmall–scale variations such as fast fading, caused by the instructive or destructiveinteractions of multipath components, occur on the scale of a wavelength. Thisclustering is mainly caused by the fact that the physical structures that causescattering, reflections, and shadowing of the radio signals can be grouped intothose in the vicinity of the mobile, those located near or around the base station,and a third group of distant objects that might exist in the channel. This spatialdistribution of objects in the channel will cluster the multipath components intogroups of signals with similar time delay and angular properties.


3.6.2 Number of Clusters

Clearly, there must be at least one cluster present in the channel that occurs dueto local scattering around the mobile or base station. The existence of more clus-ters will depend on whether there is any scattering due to other distant objects inthe channel, such as buildings, hills, and so on. Therefore, the number of clus-ters, NC, has to be determined based on measurements. In [8] the appropriatevalues for the number of clusters for different radio environments for the COST259 model has been identified based on power delay profile measurementsthroughout Europe. It was found that in macrocell urban environments, NC

ranges from one to two. This is due to the presence of scattering from local aswell as far objects. In suburban and rural environments, NC is typically aroundone, implying that only those objects in the vicinity of the receiver contribute tothe channel. On the other hand, [9] considers the number of clusters in anyurban environment to be six.

3.6.3 Spatial Distribution of Clusters and Scatterers

Since the local objects in the vicinity of the MS or BS will significantly contrib-ute to the multipath, it is reasonable to seek a model for their spatial distribu-tions from which the components of the channel impulse response could bederived. Let us consider the macrocell case where the BS antenna is mounted ona rooftop higher than any of the local scatterers; as a result, the scattering isdominated by those objects around the MS. In fact, most of the relevant scatter-ers are those located closer to the MS since they will have the greatest impact onthe channel. In addition, the signals received at the BS will mainly arrive from acertain angular region, as shown in Figure 3.2. A distribution function thatapproximates this physical behavior is the Gaussian distribution given by

( )f rR

er r

R

MS

= −−ζ

π2 22

2

2

r r

(3.7)

whererr is the position vector,

rrMS is the MS location vector, assuming the BS as

the origin, ζ is a normalization constant, and R is the radius of the scattering cir-cle, shown in Figure 3.2. Recall that with this Gaussian shape, f (r) will be largerfor objects closer to the mobile (i.e., for small r rMS– ) and will decrease forincreasing r rMS– or for objects further away from the MS.

3.6.4 Base Station Azimuth Power Spectrum and Angle Spread

Once the spatial distribution of the scatterers is known, it is possible to computethe azimuth power spectrum, that is, the distribution of the received power ver-sus the azimuth angle. Various PDFs have been proposed in the literature for the


azimuth distribution, including a cosn (φ) distribution, a uniform distribution,and a normal distribution [10]. However, based on field measurements, it wasfound that Laplacian distribution is a better representation for the azimuthpower spectrum [9, 11]. From the azimuth power spectrum and angle of arrival(AOA) distribution, we can derive an expression for the azimuth angle spread.Let us consider the geometry shown in Figure 3.3. Here we assume that eachcluster will contribute Np paths to the channel. To characterize each of thesepaths, we need to define their AOA as the mean angle with which an arrivingpath’s power is received by the BS array with respect to the bore site and theirangle spread defined as the rms of angles with which an arriving path’s power isreceived by the base station array and azimuth power spectrum. The Laplaciandistribution was adopted by the COST 259 model [8] and the spatial channelmodels jointly developed by the third generation partnership programs ThirdGeneration Partnership Project 2 (3GPP2) and 3GPP to model the azimuthpower spectrum of both paths and clusters.

On a cluster basis, the azimuth power spectrum can then be written as

( )( )

PP

el

l

N p

l o

φ

φ

φ φ

σφ

φ

σφ= =

−−∑

1

2

2(3.8)


MS

BS

Scatterers

Figure 3.2 Macrocell base station model.

where ( )P ll

N p

φ=∑

1

is the total power received from the cluster. On a path basis, we

can write the azimuth power spectrum as

( ) ( )P N e Gp l norm

l p

p

φ

φ φ

σφ φ=

−−2

(3.9)

where Nnorm is a normalization constant, φp is the mean path AOA, σp is the anglespread, and G(φ) is the BS antenna gain at angle φ. The normalized azimuthpower spectrum is shown in Figure 3.4 for various spread angles for a mean pathAOA of 45°. Now, let us look at the angle spread given by [12]

[ ]S E Eφ φ φ= −2 2(3.10)

Using (3.8) we can write

( )( )

( )E

P

P

i l i ll

N

i

N

i ll

N

i

N

pc

pcφ

φ φ

φ

φ

φ

2

2

1

1

=∑∑

∑∑=

=

, ,

,

(3.11)


MS

Cluster

cLOS Path

BS

γ

σφ

φ

oφ

Figure 3.3 Spacial channel model parameters.

From Figure 3.3 we can see that the path AOA is given by

φ φ φ γi o c= + + (3.12)

Substituting (3.12) in (3.11) we then get

( )( ) ( )

( )E

P

P

c,i i,l i ll

N

i

N

i ll

N

i

N

pc

pcφ

φ + φ + γ φ

φ

ο φ

φ

2

2

1

1

=

=

∑∑

∑∑=

=

,

,

( ) ( )

( ) ( ) ( )

φ φ φ

γ φ φ φ

φ

φ φ

o c i i ll

N

i

N

i l i l o c i

P

P P

pc

+ +

+ +

∑∑=

, ,

, , ,

2

1

22 ( )

( )

φ γ

φφ

i l i ll

N

i

N

l

N

i

N

i ll

N

i

N

pcpc

pc

P

, ,

,

∑∑∑∑

∑∑

==

=

11

1

(3.13)

Similarly


1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

PDF

0 20 40 60 80 100Azimuth Angle

Angle spread:10

Angle spread:20

Angle spread:30

Figure 3.4 Azimuth power spectrum as a function of φ and σp for φp of 10°, 20°, and 30°.

( )[ ]( ) ( )

( )E

P

P

o c i i,l i ll

N

i l

N

i ll

N

i

N

pc

pcφ

φ φ γ φ

φ

φ

φ

2

1

=+ +

∑∑

∑∑=

=

, ,

,

( ) ( ) ( )

=+ +∑ ∑∑

=

2

1

φ φ φ γ φφ φo c i i ll

N

i l i ll

N

i

N

P Pp pc

, , , ,

( )i l

Nc

i ll

N

i

N

Ppc

=

=

∑

∑∑

φ φ ,

1

2(3.14)

We can see from (3.13) that Sφ is a random variable that is a function ofthe LOS angle φo, the clusters angles φc, the path angles within a cluster γ, thenumber of clusters, the number of paths Np, and the azimuth power spectrum.Note that the AOA of the clusters and paths have PDFs that could, for example,be given by [13]:

( )f ec

c

c

cφπσ

φ

σ=−

1

2

2

(3.15)

and

( )f ep

pγπσ

γ

σ=−

1

2

2

(3.16)

where σc is the cluster’s AOA spread and σp is the path AOA spread. Let us con-sider the special case where all clusters have the same AOA spreads σc and allpaths have the same power. It follows that the expected value of Sφ is givenby [11]

( )( )

E SP

Pc p

ii

N

l

N

i

N

c

pcφ

φ

σ σ

φ

= + −

=

==

∑

∑∑2 2

2

1

11

1 (3.17)

In the spatial channel model adopted by the 3GPP and 3GPP2 [9], twovalues are considered for the path angle spread at the BS, 2° and 5° correspond-ing to path AOAs of 50° and 20°, respectively.


3.6.5 Mobile Station Azimuth Power Spectrum and Angle Spread

Due to the nature of the scatterers that are mostly located around the mobile sta-tion, it is reasonable to assume that the azimuth power spectrum has a uniformdistribution [–180°, 180°] for which the angle spread is given by [10]

S φ = ° = °180 3 104

However, in some cases where the scatterers are not uniformly distributedaround the mobile, a Laplacian power azimuth spectrum can be a better approx-imation for the path’s power as follows:

( )P N ep l norm

l p

p

φ

φ φ

σφ =

−2

(3.18)

where the MS is assumed to have an omnidirectional antenna. In this case, theangle spread is be given by [10]

( )S N enorm p p pp

φ

π

σσ π σ πσ= − + +

−

2 4 83 14

2 2 2 (3.19)

In such a case, the path angle spread is expected to be lower compared withthe uniform azimuth power spectrum.

3.7 Spatial Channel Model Application in System Simulations

To evaluate the performance of a given smart antenna algorithm or implementa-tion, a system level simulation needs to be carried out. In this approach, multi-ple base stations or sectors and mobile stations are considered in addition tomobility modeling to account for the fast fading resulting from a user’s motion.A general procedure to generate a spatial channel model for this type of systemlevel simulation may consist of the following steps:

1. Select a radio environment such as urban, suburban or rural.

2. Select a cell type, such as macrocell, microcell, or picocell.

3. Define the path loss model based on the radio environment and celltype selected above. For instance, the COST 231 model can be usedfor suburban macrocells.


4. Define the antenna pattern and gain both at the base station andmobile station. This will directly impact the azimuth power spectra forthe multipath components.

5. Determine the locations and orientations of all sectors and users in thesystem. Based on this, LOS AOAs as well as distances between theusers and sectors can be computed.

6. Determine the delay spread, angle spread, and lognormal shadow fad-ing parameters. The distributions of these parameters are assumed tohave the form

( )σ αε µDS

DS DS= +10 for the delay spread (3.20)

where µDS is the logarithmic mean of the distribution of the delayspread and εDS is the logarithmic standard deviation of the distribu-tion of the delay spread. Similarly, for the angle spread we assume thedistribution

( )σ βε µAS

AS AS= +10 (3.21)

where µAS is the logarithmic mean of the distribution of the anglespread and εAS is the logarithmic standard deviation of the distribu-tion of angle spread. The shadow fading distribution is given by

( )σ σ γLN

SF=10 10 (3.22)

The parameters α, β and γ are defined in [9, 14] and σSF is thewell-known lognormal shadow fading standard deviation.

7. Determine the number of clusters NC and their random delays τn suchthat τ τ τN Nc c

> > >– ...1 1 . These delays are assumed to follow the clas-sical exponentially decaying profile based on the Laplaciandistribution:

( )P em

ττ

τ τ

στσ

τ=−

−1

2

(3.23)

where στ is the well-known rms delay spread.

8. Determine the random average power for each cluster. It is expectedthat the cluster power is a function of the distance between that clusterand the BS or MS. Furthermore, the longer the cluster’s delay is


(implying additional path loss), the more attenuation there is. A sim-ple model similar to that adopted by the COST 259 would then setthe power of the first cluster equal to the transmit power attenuatedby the path loss computed from the appropriate Hata or COST231 empirical formulae, taking into account shadow fading as P1 = PPL

σLN nand additional cluster powers as

( )

P Pn PL LN

k

n

n

= ⋅−

στ τ τ τ

101

10

min , max

(3.24)

where k is shown in Figure 3.5 and the Sn is the shadow fading gaindefined above. This implies that we have an exponentially decayingpower profile up to a maximum delay after which the power receivedis almost negligible.

9. Determine the powers and phases of the Np paths within each cluster.Here we can assume that all paths have identical powers (Ppath = Pn/Nc).The phases can be drawn from a uniform distribution [0° to 360°].

10. Determine the AOA of the clusters and paths. This is achieved by firstlocating the clusters according to their spatial distribution and thencomputing their AOA relative to either the BS or MS. The paths AOAare calculated based on offsets from the clusters AOA such that thedesired angle spread is obtained.

11. Associate the clusters and paths with the BS and MS. This will allowthe calculation of the antenna gains at the respective AOAs of each ofthe multipath components.


τk−

maxττk−

τP

maxτ τ

Figure 3.5 Power delay profile.

3.8 Angle Spread Impact

Due to multipath, shadowing, and mobile speed the wireless propagation chan-nel causes the transmitted signal to appear spread in time, frequency, and angleat both the base station and mobile user sides. As we discussed earlier, this dis-persion can be characterized in terms of delay spread, Doppler spread, and anglespread [15]. These three parameters determine frequency fading, temporal fad-ing, and spatial fading, respectively. The impact of the angle spread in terms ofspatial fading is illustrated in Figures 3.6 through 3.9. In a channel with a


(a)

(b)

Figure 3.6 (a, b) Fading envelope, Laplacian Azimuth power spectrum, AS = 10°, angle of arrival 90°.

narrow angle spread the fading envelope across an antenna array’s elements isrelatively constant in space, as can be seen in Figure 3.6 (AS = 10°). This isdue to the fact that the signals across the antennas are correlated. This sce-nario is beneficial to the performance of beamforming techniques. A narrowangle spread helps maintain a focused and narrow beam for better interferencereduction. On the other hand, in Figures 3.7, 3.8, and 3.9 we see the impact of


(a)

(b)

Figure 3.7 (a, b) Fading envelope, Laplacian Azimuth power spectrum, AS = 60°, angle of arrival 0°.

wide angle spread (60° and 360°). In these cases, the signal experiences fadingin space, where we can clearly see peaks and valleys in the fading envelope.This results from paths with low cross-correlation and is beneficial for spa-tial diversity applications that result in higher diversity gain as the signalsbecome uncorrelated, but it will degrade the performance of transmitbeamforming.


(a)

(b)

Figure 3.8 (a, b) Fading envelope, Laplacian Azimuth power spectrum, AS = 360°, angle of arrival90°.

References

[1] Rappaport, T., Wireless Communications, Principles and Practices, NJ: IEEE Press andPrentice Hall, 1996.

[2] Siwiak, K., Radiowave Propagation and Antennas for Personal Communications, Norwood,MA: Artech House, 1995.

[3] Lee, W. C. Y., Mobile Communications Engineering, New York: McGraw-Hill, 1982.

[4] Parsons, J. D., The Mobile Radio Propagation Channel, New York: John Wiley & Sons,1992.

[5] Lee, W. C. Y., Mobile Cellular Telecommunications Systems, New York: McGraw-Hill,1989.

[6] Fleury, B. H, and P. E. Leuthold, “Radiowave Propagation in Mobile Communications:An Overview of European Research,” IEEE Communications Magazine, February 1996.

[7] Ertel, R. B., et al., “Overview of Spatial Channel Models for Antenna Array Communica-tion Systems,” IEEE Personal Communications, February 1998.

[8] Correia, L. M., Wireless Flexible Personalized Communications, COST 259: EuropeanCo-operation in Mobile Radio Research, New York: John Wiley & Sons, 2001.

[9] 3GPP-3GPP2 SCM-121, “Spatial Channel Model Text Description,” March 14, 2003.


Figure 3.9 Fading envelope, AS = 360°, angle of arrival 0°.

[10] Fuhl, J., A. F. Molisch, and E. Bonek, “Unified Channel Model for Mobile Radio Systemswith Smart Antennas,” IEE Proc. Radar, Sonar Navigation, Vol. 145, No. 1, February1998.

[11] 3GPP-3GPP2 SCM-027, “Note on the Angle Spread Distribution,” Motorola, May 22,2002.

[12] Fleury, B. H., “Direction Dispersion and Space Selectivity in the Mobile Radio Channel,”IEEE VTS Fall 52nd Vehicular Technology Conference, 2000, Vol. 2, 2000.

[13] 3GPP-3GPP2 SCM-025, “RMS Angle Spread,” Motorola, May 3, 2002.

[14] 3GPP-3GPP2 SCM-029, “Correlated System Level Spatial Channel Model,” June 5,2002.

[15] Buehrer, R. M., “Generalized Equations for Spatial Correlation for Low to ModerateAngle Spread,” Proc. 10th Va. Tech. Symp. Wireless Communication, Blacksburg, VA, June2000, pp. 123–130.


Beach, M., B. Allen, and P. Karlsonn, “Correlation of Power Azimuth Spectrum for Vary-ing Frequency Division Duplex Spacings,” EPMCC 2001, Centre for CommunicationResearch, University of Bristol, Vienna, February 20–22, 2001.

Bertoni H. L., et al., “Sources and Statistics of Multipath Arrival at Elevated Base StationAntenna, IEEE 49th Vehicular Technology Conference, Vol. 1, 1999.

Chen, M, and Asplund, H., “Measurements and Models for Direction of Arrival of RadioWaves in LOS in Urban Microcells,” 12th IEEE Int. Symp. on Personal, Indoor, and MobileRadio Communications, Vol. 1, September 2001.

Greenstein, L. J., et al., “A New Path-Gain/Delay-Spread Propagation Model for DigitalCellular Channels,” IEEE Trans. on Vehicular Technology, Vol. 46, No. 2, May 1997.

Hata, M., “Empirical Formula for Propagation Loss in Land Mobile Radio Services,”IEEE Trans. on Vehicular Technology, Vol. VT-29, No. 3, August 1980.

Kuchar, A., J. P. Rossi, and E. Bonek, “Directional Macrocell Channel Characterizationfrom Urban Measurements,” IEEE Trans. on Antennas and Propagation, Vol. 48, No. 2,February 2000.

Liberti, J. C., and T. S. Rappaport, “A Geometrically Based Model for Line-of-SightMultipath Radio Channels,” Proc. IEEE VTC’96, Atlanta, GA, May 1996.

Lu, M., T. Lo, and J. Litva, “A Physical Spatio-temporal Model of Multipath PropagationChannels,” Proc. IEEE VTC´97, Phoenix, AZ, May 1997.

Pajusco, P., “Experimental Characterization of D.O.A at the Base Station in Rural andUrban Area,” Proc. IEEE VTC’98, Ottawa, Canada, May 1998.

Pedersen, K. I., P. E. Mogensen, and B. H. Fleury, “A Stochastic Model of Temporal andAzimuthal Dispersion Seen at the Base Station in Outdoor Propagation Environments,”IEEE Trans. on Vehicular Technology, Vol. 49, No. 2, March 2000.

Pedersen, K. I., P. E. Mogensen, and B. H. Fleury, “Spatial Channel Characteristics inOutdoor Environments and Their Impact on BS Antenna System Performance,” 48thIEEE Vehicular Technology Conference, Vol. 2, 1998.


Pedersen, K. I., P. E. Mogensen, and B. H. Fleury, “Spatial Channel Characteristics inOutdoor Environments and Their Impact on BS Antenna System Performance,” Proc.IEEE VTC’98, Ottawa, Canada, May 1998.

Petrus, P., J. H. Reed, and T. S. Rappaport, “Geometrically Based Statistical Model forMacrocellular Mobile Environments,” Proc. IEEE Globecom’96, London, November 1996.

Saleh, A. A. M., and R. A. Valenzuela, “A Statistical Model for Indoor Multipath Propaga-tion,” IEEE Journal on Selected Areas in Communications, Vol. SAC-5, No. 2, February1987.

Turin, G. L., et al., “A Statistical Model of Urban Multipath Propagation,” IEEE Trans.on Vehicular Technology, Vol. VT-21, No. 1, February 1972.


4Fixed Beam Smart Antenna Systems

4.1 Introduction

In wireless system design and planning, one has to deal with two main prob-lems, coverage and capacity. Coverage designs include, among other parameters,the selection of the number of base station sites, locations, heights, antenna ori-entation, and transmit power required to provide service in a given geographicarea. When the system is first deployed, the number of subscribers is low and theaim is to maximize the coverage area of each site. As the number of networkusers grows, the number of users per site and sector also grows. As more users areadded to a sector, the base station transmit power as well as the total powertransmitted by those users increases, which effectively increases the interferenceon both the forward and reverse links. The system then becomes capacity orinterference limited. One of the most common ways to deal with this interfer-ence problem is sectorization.

4.2 Conventional Sectorization

In base stations employing omnidirectional antennas, the transmit power isequally radiated in all directions. The equal distribution leads to a portion of thepower being transmitted throughout the cell but not received by the user. Thiswasted power then becomes forward link interference to other base stations orusers in other cells. Similarly, each new user added to a cell increases the interfer-ence and noise levels on the reverse link. This results in a reduction in the sig-nal-to-noise ratio, which in turn degrades the performance of the detection anddemodulation operations. One way to reduce interference is to divide the cell

83

into a number of smaller sectors using directional antennas. The most commonscheme is the three 120°-sectors; however, two and six-sector cells also havesome practical applications. As we can see from Figure 4.1, in a three-sector sitethe radiation pattern of the directional antenna allows it to receive substantiallyhigher power levels from its own sector compared with that received from theother two sectors. It is obvious that sectorization is an effective technique thatcan increase capacity.

In fact, since CDMA capacity is noise or interference limited, assumingideal antenna radiation patterns, the capacity of an Ns-sector cell will be Ns timesthat of an omnidirectional cell. This capacity gain is often referred to assectorization gain (SG). In practice, overlapping sector coverage areas due tononideal antenna radiation patterns will increase the multiuser interference,reducing SG [1–3]. It is well known that CDMA has a soft capacity, which isdetermined by the balance between the required SNR for each user and thespread spectrum processing gain (PG) given by

PGWR

= (4.1)

where W is the bandwidth of the spreading signal and R is the user’s data rate.The figure of merit of the digital receiver is the dimensionless SNR given by


150

180

210

240

270

300

330

0

30

60

90

120

1

0.8

0.6

0.4

0.2

Figure 4.1 Three-sector patterns.

E

N

Energy per bit

Noise plus Interference power spectrab

0

=l density

(4.2)

where the energy per bit is given by

EPRb = (4.3)

To derive an approximate SG for a CDMA system, let us consider thereverse link capacity of an omni system given by

( )N

W RE

If

omnib

oomni

=+1 ν

(4.4)

where v is the voice activity factor, Nomni is the number of users, Io is the totalinterference density, and f is the reuse efficiency

fI

Ioc

sc

= (4.5)

In (4.5) Ioc denotes the other cell interference power and Isc is the same cellinterference power. From (4.4) it can be readily seen that reducing the interfer-ence level increases the number of maximum supportable users in a cell. Now,let us consider the capacity of a single sector given by:

( )N

W RE

If

tb

ot

sec

sec

=+1 ν

(4.6)

Hence the sectorization gain for an N-sector cell becomes

SG NN

NN

f

fst

omnis

omni

t

= ⋅ = ⋅++

sec

sec

1

1(4.7)

It is clear that the sectorization gain is highly dependent on the amount ofreduction in interference provided by the antenna, which in turn is a function ofthe antenna beamwidth and the size of the overlap region. Based on simulationresults for reuse efficiency provided in [4], Figure 4.2 shows SG as a function of

Fixed Beam Smart Antenna Systems 85

NS. For three-sector sites, SG becomes 2.4, which is comparable to 2.55, thesectorization gain measured in actual CDMA network deployments. To showthe effect of the overlap area on the sectorization efficiency, consider Figure 4.3,where a three-sector site is shown with the areas of overlap or softer handoffdetermined by the angle θSH. Recall that in CDMA systems, there are mainlytwo types of handoffs, namely hard handoffs and soft handoffs, as discussed inChapter 2. Soft handoff can be further divided into handoff between two ormore sectors of different cells and handoff between two sectors of the same cell,which is referred to as softer handoff.

As a result of the three overlapping sectors, instead of reducing the inter-ference by three times, as in the ideal case, only (120° + θSH)/360° of the inter-ference is blocked. Hence, we can define the sectorization efficiency by

E sec tSH

=+

120

120 θ(4.8)

Figure 4.4 shows the relation between the sectorization efficiency and θSH.Sectorization and soft/softer handoffs also affect the capacity of the forward linkof a CDMA system.

On the forward link, a simplified form for capacity is given by

NP

P HFLoverhead

traffic

=−⋅ ⋅

1

ν(4.9)


8

7

6

5

4

3

2

1

Sect

oriz

atio

nga

in

2 3 4 5 6 7 8 9Number of sectors

Figure 4.2 Sectorization gain.


SHθ

Soft

erha

ndof

f Reg

ion

Figure 4.3 Softer handoff overlap areas.

0.98

0.96

0.94

0.92

0.9

0.88

0.86

0.844 6 8 10 12 14 16 18 20

Overlap angle

Effic

ienc

y

Figure 4.4 Sectorization efficiency.

where Poverhead denotes the total power in the common channels, Ptraffic is the aver-age traffic channel power that depends on the forward link SNR or Eb/Nt, and Hdenotes the handoff reduction factor. When mobile users are in soft or softerhandoff, additional power is required on the forward link, hence the forwardlink capacity is reduced. However, the Eb/Nt required by those mobiles toachieve a given FER will be lower than that required without soft/softerhandoff, so sectorization will also provide some capacity gain on the forwardlink. [2–6] provide detailed analysis of the impact of sectorization andsoft/softer handoff on CDMA systems.

4.3 Limitations of Conventional Sectorization

The most common form of sectorization uses spatial diversity antennas for sig-nal reception and a single antenna for transmission, as shown in Figure 4.5,where all antennas use vertical polarization. Another form of diversity, calledpolarization diversity, uses either 0°/90° antennas, otherwise known as verti-cal/horizontal polarization or ±45°, also known as cross polarization. In eithercase, two replicas of the received signal are available to the receiver for diversitycombining to combat fading. One drawback of conventional sectorization isthat the signals cannot be separated in the spatial domain, which makes spatialinterference cancellation or reduction impossible to carry out.

Another fundamental problem at the heart of network optimization is thatof traffic loading imbalance when cellular traffic is distributed unevenly amongdifferent geographical areas of the network or among the sectors of a site leadingto increased blocking. This imbalance is often time dependent, for instanceduring rush hour traffic on highways, business districts, or sport avenues.


Duplexers and LNAs

Rx Rx Tx

Rx RxTx

Figure 4.5 Conventional sectorization scheme.

Alleviating such imbalance would require sectors with flexible orientations orbeamwidths, which are not available with conventional sectorization. As a result,unused capacity on other sectors/sites is locked and wasted. As discussed earlier,on the forward link handoff zones have an immediate impact on capacity.Reducing the size of handoff zones and shifting those zones from high- tolow-traffic areas can minimize the negative impact of soft/softer handoff oncapacity. However, with conventional sectorization both the size and orienta-tion of these handoff zones are fixed. One way to deal with these issues and pro-vide means for spatial signal separation for further processing is to replace theconventional base station antennas with antenna arrays. The additional degreesof freedom provided by antenna arrays can offer more effective techniques todeal with multipath and interference and improve signal quality, leading toimproved coverage and/or capacity. The fundamental advantage of arrays istheir ability to generate one or more main beams with tailored beamwidths,with radiation pattern nulls and increased gain. Two main approaches exist—fixed multiple beam antennas and fully adaptive antennas. The remainder ofthis chapter is devoted to the first approach.

4.4 Antenna Arrays Fundamentals

Assume that we have a linear array composed of M identical antenna elementsarranged along some axis, with interelement spacing d, as shown in Figure 4.6.In the simplest form, the array elements are fed with equal amplitudes Am andconstant phase delay β.

The radiation pattern of the array excluding the element pattern is referredto as the array factor. A general form for the array factor is given by

( ) ( ) ( ) ( )( )AF A e e ej kd j kd j M kd= ⋅ + + + ++ + − +1 2 1cos cos cosγ β γ β γ βL (4.10)

where k =2π

λ. Without loss of generality, we can assume the signal amplitudes

as follows:

A m Mm = =1 1 2, ,K (4.11)

Hence, the array factor can be rewritten as

( )AF e j m

m

M

= −

=∑ 1

1

ψ (4.12)


where

ψ γ β= +kd cos (4.13)

The total radiation pattern of an antenna array in the far field E(θ, ϕ) isrepresented by a product between two factors, the array factor AF (θ, ϕ) and theelement factor EF (θ, ϕ). The element factor depends on physical dimensionsand electromagnetic characteristics of the radiating element, whereas the arrayfactor depends on the amplitude, phase, and position of each of the elements inthe array antenna. In (4.13) γ is the angle between the array axis and the vectorfrom the origin to the observation point. For an array along the z-axis, we have:

( )cos sin cos sin sin cos cγ θ φ θ φ θ= ⋅ = ⋅ + + =$ $ $ $ $ $a a a a a az r z x y z os θ (4.14)

It follows that

ψ θ β= +kd cos (4.15)

When the array is placed along the x-axis, we get:


x

y

z

θ

φ

d

1

2

3

M

ar

Figure 4.6 Coordinate system for linear antenna arrays.

( )cos sin cos sin sin cos

s

γ θ φ θ φ θ= ⋅ = ⋅ + +

=

$ $ $ $ $ $a a a a a ax r x x y z

in cosθ φ(4.16)

and

ψ θ φ β= +kd sin cos (4.17)

Finally, for an array along the y-axis, we get:

( )cos sin cos sin sin cos

s

γ θ φ θ φ θ= ⋅ = ⋅ + +

=

$ $ $ $ $ $a a a a a ay r x x y z

in sinθ φ(4.18)

and

ψ θ φ β= +kd sin sin (4.19)

When the reference point or origin is chosen at the physical center of thearray, it can be shown [7, 8] that the normalized array factor of a uniformlyexcited, equally spaced linear array is reduced to

( )( )AF

M

M

= ⋅1 2

2

sin

sin

ψψ

(4.20)

4.4.1 Broadside and End-Fire Arrays

Note that the AF in (4.20) has a maximum at ψ = 0. To find the conditionsunder which the maximum radiation occurs, let us assume we have a linear arrayplaced along the z-axis; it follows that

ψ θ β= + =kd ocos 0 (4.21)

therefore β = –kdcosθo, where θo is the direction of the maximum radiation.When θo = 90°, the array is called broadside, and it follows that β = 0. The maxi-mum of the radiation pattern of broadside arrays is always directed normal tothe array axis.

From the above we can see that a broadside array requires equal magnitudeand phase excitation. For θo = 0° or θo = 180° the resulting array is called anend-fire array. The maximum of the radiation pattern in this case is directedalong the array axis. For θo = 0°we get ψ = kdcosθ + β = kdcos(0) + β = kd + β = 0.

Hence, the progressive phase shift required for an end-fire array with max-imum radiation directed at 0° is


β = −kd (4.22)

As can be seen, the array factor given by (4.20) is a function of the numberof elements M, element spacing d, and the phase shift β. Therefore, it is impor-tant to investigate the impact of these parameters on the radiation pattern of anantenna array.

4.4.2 Impact of Number of Elements

Figures 4.7 and 4.8 show the effect of increasing the number of elements M onthe radiation pattern of a broadside array along the z-axis.


θ

Figure 4.7 Radiation patterns of broadside array along the z-axis, d = λ/2.

θ

Figure 4.8 Radiation patterns of broadside array along the z-axis, d = λ/2.

We can observe that increasing M has the following effects on the radia-tion pattern:

• The width of the main lobe decreases; in other words, it becomes nar-rower. This is crucial for the applications of smart antennas when a sin-gle narrow beam is required to track a mobile or cluster of mobiles.

• The number of sidelobes increases. In addition, the level of the first andsubsequent sidelobes decreases compared with the main lobe. Sidelobesrepresent power radiated or received in potentially unwanted direc-tions. So in a wireless communications system, sidelobes will contributeto the level of interference spread in the cell or sector by a transmitter aswell as the level of interference seen by a receiver when antenna arraysare used.

• The number of nulls in the pattern increases. In interference cancella-tion applications, the directions of these nulls as well as the null depthshave to be optimized.

4.4.3 Impact of Element Spacing

The element spacing d also has a significant impact on the shape of the radia-tion pattern. It is evident that the more elements an array has or alternativelythe larger the array gets, the better the characteristics of the radiation patternas far as its shape and degrees of freedom. Another way of achieving alarger array would be by increasing d. The major drawback of this approachlies in the behavior of the array factor function in (4.20), namely the appearanceof replicas of the main lobe in undesired directions, referred to as grating lobes.Figure 4.9 shows the polar radiation pattern of a broadside six-elementarray along the z-axis with element spacing of d = λ/2. We can see that forthis element separation, aside from a few sidelobes, we only have a main lobedirected toward 90°. When we increase the spacing to d = λ, we get the radia-tion pattern shown in Figure 4.10. Notice the appearance of a grating lobe at0°. Not only have we wasted power in the grating lobe, we also spread orreceive more interference from the broader lobe. In practice, the optimum ele-ment spacing for beamforming and adaptive interference cancellation applica-tions is d = λ/2.

However, in specific applications such as transmit diversity, we intention-ally design an array with much larger spacing to combat fading effects, as will bedescribed in detail in a later chapter. A typical transmit or receive diversityantenna array has two elements separated by up to 10λ.

The radiation patterns of a two-element array with element spacing ofλ/2, 5λ, and 10λ are shown in Figures 4.11, 4.12, and 4.13, respectively. The



150

180

210

240

270

300

330

0

30

60

90

120

1

0.8

0.6

0.4

0.2

Figure 4.9 Polar pattern, broadside array, M = 6, d = λ/2.

150

180

210

240

270

300

330

0

30

60

90

120

1

0.8

0.6

0.4

0.2

Figure 4.10 Polar pattern, broadside array M = 6, d = λ.


150

180

210

240

270

300

330

0

30

60

90

120

1

0.8

0.6

0.4

0.2

Figure 4.11 Polar pattern, broadside array, M = 2, d = λ/2.

150

180

210

240

270

300

330

0

30

60

90

120

1

0.8

0.6

0.4

0.2

Figure 4.12 Polar pattern, broadside array, M = 2, d = 5λ.

sidelobes and grating lobes generated in Figures 4.11 and 4.12 makes this designunsuitable for applications seeking to improve system performance when thedegradation is mainly caused by interference or jamming.

4.4.4 First Null Beamwidth

The null-to-null beamwidth (NNBW) of the array has a significant impact onthe performance of a smart antenna system and is considered one of the impor-tant parameters that need to be considered in the antenna design. For a broad-side array on the z-axis, the null-to-null beamwidth is given by [7]

θπ λ

N Md= −

−22

1cos (4.23)

The behavior of the NNBW is shown in Figures 4.14 and 4.15 as a func-tion of d and M, respectively. Note that the larger the array, the smaller theNNBW becomes and the narrower the main lobe gets.


150

180

210

240

270

300

330

0

30

60

90

120

1

0.8

0.6

0.4

0.2

Figure 4.13 Polar pattern, broadside array, M = 2, d = 10λ.

4.4.5 Half-Power Beamwidth

Another very important beamwidth measure to consider is the half-power or3-dB beamwidth. The 3-dB beamwidth of a broadside array on the z-axis isgiven by [7]

θπ λ

ππ λH Md

d= −

<<−22

139111cos

.for (4.24)


θ

λ

Figure 4.14 NNBW as a function of element spacing d.

θ λλ

λλ

Figure 4.15 NNBW as a function of number of elements M.

A more general formula for the 3-dB beamwidth of a linear-phased arrayantenna is

θ θλ

θλ

H o oMd Md= −

− +

− −cos . cos .1 10 443 0 443cos cos

(4.25)

This is valid for a range of scanning angles but not for end-fire arrays. InFigures 4.16 and 4.17, we notice the same behavior for the 3-dB beamwidthwhen we increase M or d.

Figure 4.18 demonstrates that the 3-dB beamwidth of a linear-phasedarray of a given size is not constant but rather it depends on the scanning angle.


θ

λ

Figure 4.16 3-dB bandwidth as a function of element spacing d.

θ λλ

λ

λ

Figure 4.17 3-dB bandwidth as a function of number of element spacing M.

4.4.6 Array Directivity

The radiation intensity of an antenna array can be defined as:

( ) [ ]U AFθ = 2 (4.26)

Antenna arrays have the ability to direct or concentrate the radiated powerin a particular angular direction in space. This ability is measured by what iscalled the directive gain, defined as [9]:

( )Dpower radiated per unit solid angle in the directio

θ ϕπ

, =4 ( )n

Total power radiated by the antenna

θ ϕ,(4.27)

The directive gain in the direction of the maximum radiation density isreferred to as the directivity and is given by

DU

Porad

=4 π max (4.28)

For a broadside array and small element spacing (d < λ), the directivity canbe approximated by [7]

D Md

o = 2λ

(4.29)


θ

θ

Figure 4.18 Effect of the scanning angle on the 3-dB beamwidth.

4.4.7 Array Gain

The gain of an antenna array is the ratio of the radiation density in a particularangular direction in space to the total input power to the array or

( )Gpower radiated per unit solid angle in the directio

θ ϕπ

, =4 ( )n

Total input power to the antenna

θ ϕ,(4.30)

Note that we can define the antenna array efficiency as

η =P

Pin

rad (4.31)

It follows that

G D= η (4.32)

4.4.8 Trade-Off Analysis

We have seen from previous discussions the effect of different parameters on thecharacteristics of the array and potential system performance impacts. Table 4.1presents a trade-off analysis summarizing the impact of increasing each of theseparameters.


Table 4.1Impact of Array Parameters on System Performance

Parameter Pros ConsSmart AntennaPerformance Impact

Number ofelements M

Lower sidelobe levels

More and deeper nulls

Narrower beams

Higher gain

More sidelobes

Larger arrays may be morecostly

Physical limitations oninstallation

Better interferencecancellation capabilities

Improved performancebecause of higher gainand narrower beams

Element spacingd

Narrower beams

Higher gain

Grating lobes Grating lobes havenegative impact oninterference nulling

Scanning angleθo

Smaller 3-dBbeamwidths.

— Improved performancebecause of narrowerbeams

4.4.9 Impact of Element Pattern

As we indicated earlier, the total radiation pattern of an antenna array in the farfield E(θ, ϕ) is represented by a product between two factors, the array factorAF(θ, ϕ) and the element factor EF(θ, ϕ). Consider two of the most widely usedantennas, namely, the short dipole and a half-wave patch antenna, with elementpatterns given by

( ) ( )( )( )

E θ ϕπ θ

θ,

cos cos

sin=

22

half wave dipole

( ) ( )EL

θ ϕπ

λθ, cos sin=

E plane, half wave patch

where Lr

≈ 0 49.λ

εis the resonant length of a half-wave patch [10].

The complete radiation pattern of the array is obtained by pattern multi-plication, as shown in Figures 4.19 through 4.22, where we have consideredtwo- and four-element arrays of half-wavelength dipoles as well as half-wavepatches. It is obvious how the element pattern can affect the total array patternby changing the shape of the pattern and the direction of the main beam, as inFigure 4.19, by changing the beamwidth, as in Figure 4.21, and by changing themaximum gain, as in Figure 4.22.

4.4.10 Planar Arrays

In planar arrays, elements are placed in a planar or rectangular grid. Let us con-sider a general planar array antenna with elements located at arbitrary positions(yn, zn) in the yz-plane, as shown in Figure 4.23.


150

180

210

240270

300

330

0

30

6090

1201

0.80.60.4

0.2

150

180

210

240270

300

330

0

30

6090

1201

0.5150

180

210

240270

300

330

0

30

6090

1201

0.80.60.4

0.2

Array factor Element pattern Total pattern

Figure 4.19 Total pattern, two-element array of half-wave dipoles, β = 180°, d = λ/2.

The array factor for this array antenna is given by

( ) ( )AF a en

jk y z

m

Mn y n z= = + + +

=∑θ φ

θ φ β θ β,

sin sin cos

1

(4.33)

where an is the complex excitation of element n. Assuming we have M elementsin the y-direction and N elements in the z-direction, we can rewrite (4.33 ) as


150

180

210

240270

300

330

0

30

6090

1201

150

180

210

240270

300

330

0

30

6090

1201

0.5150

180

210

240270

300

330

0

30

6090

1201


0.5 0.5

Figure 4.21 Total pattern, four-element array of half-wave patch, β = 180°, d = λ/2.

150

180

210

240270

300

330

0

30

6090

1201

0.80.60.4

0.2

150

180

210

240270

300

330

0

30

6090

1201

0.5150

180

210

240270

300

330

0

30

6090

1201

0.80.60.4

0.2


Figure 4.20 Total pattern, two-element array of half-wave dipoles, β = 0°, d = λ/2.

150

180

210

240270

300

330

0

30

6090

1201

150

180

210

240270

300

330

0

30

6090

1201

0.5150

180

210

240270

300

330

0

30

6090

1201


0.5 0.2

Figure 4.22 Total pattern, four-element array of half-wave patch, β = 0°, d = λ/2.

( ) ( )( ) ( ) ( )[ ]AF a emn

jk m d n d

n

N

m

My y z zθ φ

θ φ β θ β,

sin sin cos= − + + − +

==∑ 1 1

11∑ (4.34)

The above array factor can then be separated into the product of two termsas follows:

( ) ( ) ( ) ( )( )AF e a ejk n dmn

jk m d

m

M

n

z z y yθ φ θ β θ φ β, cos sin sin= − + − +

==∑1 1

11

N

∑ (4.35)

where dy and dz are the element separations in the y and z directions, respectively,and amn is the complex excitation of the element at position (mdy; ndz). For sim-plicity, let us assume the excitation distribution over the array to be uniform andequal to ao, we can then rewrite (4.35) as a product of two sums as

( ) ( ) ( ) ( )( )AF a e eojk n d jk m d

m

M

n

z z y yθ φ θ β θ φ β, cos sin sin= − + − +

==∑1 1

11

N

∑ (4.36)

Comparing (4.36) and (4.12) and using (4.20), it follows that


x

y

z

θ

φ

dz

dy

1

2

3

2 3

..

.

M

N

Figure 4.23 Planar array geometry.

( )AF

N

N

z

zMθ φ

ψ

ψ,

sin

sin

=

⋅1 12

2

sin

sin

N y

y

ψ

ψ

2

2

(4.37)

To get an expression for the radiation pattern of the planar linear array, wemultiply the array factor by the element factor:

( ) ( ) ( )E AF EFθ φ θ φ θ φ, , ,= (4.38)

An example of the 3D array factor of a planar array of 16 elements is plot-ted in Figure 4.24.

4.4.10.1 Directivity of Planar Arrays

Using the expression in (4.27) we can write the directivity as

( ) ( )

( )D

E

E d d

θ φπ θ φ

θ φ θ θ φ0

π

π

π,,

, sin

=

∫∫−

42

2

(4.39)


1

0.8

0.6

0.4

0.2

0200

150

100

50

0 050

100

150

200

Figure 4.24 4 x 4 planar array pattern, dz = dy = λ/2

Notice that the directivity is defined for arbitrary angles θ and φ. We areoften interested in the maximum directivity Do of a pattern. For large planarbroadside arrays, the elevation half-power beamwidth is small. We can then usethe approximation sin(θ) = 1 and rewrite the directivity as

( ) ( ) ( ) ( )

( ) ( ) ( ) ( )D

AF EF AF EF

AF EF AF EF d do =

∫∫−

42

2

π θ θ φ φ

θ θ φ φ θ θ ϕ0

π

π

π

sin

(4.40)

which can be further separated into

( ) ( )

( ) ( )

( ) ( )

( ) ( )D

AF EF

AF EF d

AF EF

AF EF do = =

∫ ∫

2 22

2

0

2

2

θ θ

θ θ θ θ

π φ φ

φ φ ϕπ

−π

π

sin

D Dθ φ (4.41)

From (4.41) we can see that with large planar arrays, the directivity can becalculated as the product of the directivities of two linear arrays.

4.5 Beamforming

As we have previously seen, an array’s main beam direction depends on the ele-ment spacing as well as the phase difference between adjacent elements. From(4.12) the phase difference between element m and the first element consideredas the array reference would be

( ) ( )( )ψ ψ β γm m m kd1 1 1= − = − + cos (4.42)

For half-wavelength spacing, the phase difference between two adjacentelements can then be written as

ψ β θ βπ

λ

λθ β π θ= + = + ⋅ ⋅ = +kd cos cos cos

2

2(4.43)

It then follows that for a broadside array with θo = 90°, ψ becomes zero.Now, if we wish to direct the array main beam toward some other angle θb,where 0° ≤ θb ≤ 180°, then

ψ β π θb b b= + cos (4.44)


Since the direction of the main beam occurs when ψb = 0, we obtain thefollowing relation

β π θb b= − cos (4.45)

Consider a linear antenna array connected to a signal generator (or areceiver); as we have seen earlier, this will produce a main beam at a specificangle and nulls at other directions. Therefore, in order to produce multiplebeams at different directions, we need to feed the array with multiple signal gen-erators (or connect it to multiple receivers) simultaneously. This can be accom-plished using a feed network referred to as a beamformer, as shown inFigure 4.25. In an M x M beamformer, M input ports are connected to Mantenna elements, whereas M output ports are connected to signal generators orreceivers. The presence of a signal at one of the output ports will induce a phaseshift between adjacent input ports and array elements, resulting in a radiationpattern with a main beam and nulls along specific directions.

Now, when different signals are fed to or applied at all output ports, corre-sponding radiation patterns will be produced, the superposition of which willresult in multiple simultaneous beams along different angles. When the peak ofa radiation pattern is directed along the nulls of other patterns, the beamformeris called orthogonal. Note that an M x M beamformer produces M beams. In a


45° 45°

1 2 43

Figure 4.25 4 × 4 Butler matrix beamformer.

symmetric beamformer, M/2 beams are produced on each side of the array’sboresight. In an asymmetric beamformer, a broadside beam and M/2-1 beamsare formed on one side of the array’s boresight, whereas M/2 beams are pro-duced on the other side. The resulting phase difference between adjacent arrayelements will be given by [20]

( )( )β πb b M b b M M= − = −2 1 2 1 2, , ,where K K (4.46)

for symmetric beamformers and

( )β πb b M M b M= − ≤ ≤2 2 2, where (4.47)

for asymmetric beamformers. Using (4.47), we can determine the angle θb offthe array’s axis at which the bth beam is pointing as

( )θ ψ πb b= −cos 1 (4.48)

4.6 The Butler Matrix

The Butler matrix is the most commonly used beamforming network that, inconventional form, is capable of producing M beams, where M is any integralpower of 2. The Butler matrix uses passive hybrid power dividers and fixedphase shifters to produce the desired progressive phase shifts at the elements ofan antenna array necessary to form simultaneous multiple beams [11, 12]. Afour-element Butler matrix is shown in Figure 4.25, where phase-lag directionalcouplers and fixed phase shifters are used to produce four orthogonal beams.

The directional couplers have outputs that are equal in power but are 90°out of phase, as shown in Figure 4.26. Then, from Figure 4.25, we can see thatwhen the input signal e j0 is applied at each of the matrix ports, the resultingphases at the array elements are given as shown in Table 4.2.

Now let a four-element array be connected to the input ports of the 4 x 4Butler matrix. Then for a symmetric beamformer, the locations of the fourbeams and the corresponding adjacent element phase shifts would be given inTable 4.3.

Let us assume that a signal source is located along one of the main beams.The array factor, which in this case could also be thought of as the transfer func-tion between the signal source and the corresponding port, is given by:

( ) ( )( )

( )G eM

j Mθψ

ψ

ψ= ⋅ −sin

sin2

2

1 2 (4.49)



0°0°

–90° –90°

Figure 4.26 Hybrid coupler.

Table 4.2Phases at Array Elements in a 4 × 4 Butler Matrix

Input Port

Output Port A1 A2 A3 A4

P1 0° –45° –90° –135°

P2 –90° 45° –180° –45°

P3 –45° –180° 45° –90°

P4 –135° –90° -45° 0°

Table 4.3Beam Locations

BeamIndex b

PhaseShift b

BeamLocation b

–2 –135° 138.6°

–1 –45° 104.5°

1 45° 75.6°

2 135° 41.4°

Using (4.44), the magnitude of the set of beam patterns are then given by

( )[ ]( )AF

M

kd

kd

M

12 1

1

1= ⋅

++

sin cos

sin cos

β θ

β θ(4.50)

( )[ ]( )AF

M

kd

kd

M

22 2

2

1= ⋅

++

sin cos

sin cos

β θ

β θ(4.51)

( )[ ]( )AF

M

kd

kd

M

3

2 3

3

1= ⋅

+

+

sin cos

sin cos

β θ

β θ(4.52)

( )[ ]( )AF

M

kd

kd

M

4

2 4

4

1= ⋅

+

+

sin cos

sin cos

β θ

β θ(4.53)

Figure 4.27 shows the beam patterns of a 4 × 4 beamformer.


0 20 40 60 80 100 120 140 160 180−20

−18

−16

−14

−12

−10

−8

−6

−4

−2

0

AOA Degrees

Arr

ayFa

ctor

(dB)

Figure 4.27 Radiation pattern of 4 × 4 Butler matrix beamformer.

4.7 Spatial Filtering with Beamformers

As we have previously seen, in an orthogonal beamformer, the peak of any beamcoincides with nulls of all other beams. Considering the 4 × 4 beamformer alreadydiscussed, it then follows that if only port 1 was fed by a signal generator while allother ports were match terminated, a pattern will be formed with a main beampointing at 104.5° off the array axis and nulls at 138.6°, 75.6°, and 41.4°.

Similarly, if a signal source was located at 104.5° off the array axis, then asignal will appear only at port 1. Hence, the beamformer network allows for thespatial separation of the signal sources. Even though all signals at thebeamformer input ports are transported to all output ports, the signal from asource located along one of the main beams will only appear at the correspond-ing output port. Let us assume we have four signal sources, s1(t), s2(t), s3(t), s4(t)located along the directions of the four main beams in addition to L interferingsignal sources Il (t) located at arbitrary angles θl. Let the transfer functionsbetween the signal sources along the main beams and their corresponding out-put ports be denoted by Gi and the transfer function between interference signall and port i be denoted by Gli. Moreover, assume all the signals are uncorrelated.It follows that the total signal appearing at port i is given by

( ) ( ) ( ) ( )y t s t G I t Gi i i l li ll

L

= ⋅=∑ θ

1

(4.54)

The total output power at port i will then be given by

( ) ( ) ( ) ( )

( ) ( ) ( )

y t s t G I t G

s t G I t G

i i i l lil

L

l

i i l li ll

2

1

2

2

= + ⋅

= + ⋅

=∑ θ

θ

( )( ) ( ) ( )=

=

∑

∑

+ ⋅

1

2

1

2

L

i i l li ll

L

s t G I t GRe θ

(4.55)

However, since all the signals are uncorrelated, the total output powerreduces to

( ) ( ) ( ) ( )y t s t G I t Gi i i l li ll

L2 2 2 2 2

1

= +=∑ θ (4.56)

Now let us consider the more general case where we have L signal sourceslocated at arbitrary angles throughout the array field of view represented by the


vector s(t) = [s1(t) s2(t) … sL(t)]T. Let the signals induced at the array elements bedenoted by the vector

( ) ( ) ( ) ( )[ ]x t x t x t x tM

T= 1 2 L (4.57)

As we have seen earlier, the signals at the array elements (or beamformingnetwork input ports) will be transported to all output ports being modified inthe process with the transfer functions between the input and output ports. Letthe matrix T denote the transfer functions between the input and output portsof the beamforming network and be given by

[ ]T = w w w M

T

1 2 L (4.58)

Note that since these transfer functions or weights are fixed, this is oftenreferred to as fixed beamforming. It follows that the beamforming network out-put can be written as

( ) ( )y T xt tH= (4.59)

where the superscripts T and H denote the transpose and complex conjugatetranspose of a vector or matrix, respectively. It follows that the output power isgiven by

( ) ( )[ ]( ) ( )[ ]

P E t t

E t t

H

H H

H

=

=

=

y y

T x x T

T RT

(4.60)

where E[.] denotes the expectation operator and R is the array correlation matrixwhose elements are the correlation between various elements. Once the signalsare transported from the array elements to the output ports of the beamformingnetwork, a scheme must be developed to extract the desired signal and deal withinterfering sources, either by reducing or canceling their signals. With this typeof fixed beamforming, there are two general approaches that can be used,namely switched beam systems and multiple fixed beam systems.

4.8 Switched Beam Systems

The switched beam method is considered an extension of the current cellularsectorization scheme. In the switched beam approach, the sector coverage is


achieved by multiple predetermined fixed beam patterns with the greatest gainplaced in the center of a beam. When a mobile user is in the vicinity of a beam,then the signals at the output ports will be given as in (4.54). This enables theswitched beam system to select the signal from the output port corresponding tothat beam [13, 14]. As the mobile moves to the coverage of another beam dur-ing the call, the system monitors the signal strength and switches to other outputports as required. A basic switched multiple beam antenna architecture is shownin Figure 4.28.

Switched beam systems offer several advantages, including:

• Low complexity and cost. Since switched beam systems only require abeamforming network, RF switches, and simple control logic, they arerelatively easy and cheap to implement.


Duplexer

MxM beamforming network

Switch

Rcvr Rcvr Rcvr Rcvr

Combining and demodulation

Beamselection

TxDuplexer

Duplexer

Duplexer

Figure 4.28 Switched beam system architecture.

• Moderate interaction with base station receivers. In practice, switchedbeam systems can simply replace conventional sector antennas with-out requiring significant modifications to the radio base stationantenna interface or the baseband algorithms implemented at thereceiver.

• Coverage extension. The antenna array aperture gain will boost the linkbudget, which could be translated to a coverage extension.

However, there are several limitations to switched beam antennas that canbe summarized as follows:

• Susceptibility to interfering signals or multipaths arriving from anglesnear that of the desired signal since, in that case, those signals wouldalso appear at the same output port as the desired signal, making it hardto separate them.

• Scalloping. From Figure 4.27, we can see that the antenna patterndrops or rolls off as a function of AOA as we move away from the centerof a beam to the intersection of two beams by as much as 3.9 dB. Forclarity, this is shown in Figure 4.29. As a result, the mobile’s signalstrength will fluctuate by the same amount as it moves across the differ-ent beams’ coverage.

• Lack of path diversity. Since the switched beam system will only selectthe signal from one port, it will be unable to combine coherentmultipath components that might be arriving from directions of beamsother than the main beam.

4.9 Multiple Fixed Beam Systems

In multiple fixed beam antennas, rather than selecting the signal from a spe-cific port, the signals from all ports are combined, making use of path diversity.This approach, which we can refer to as an integrated embedded system offixed multibeams, can achieve better performance since it enhances the receivedsignal detection on the uplink by making use of the signals from all the availablepaths in the beams followed by some kind of diversity-combining techniquesuch as maximum ratio combining. On the reverse link, or uplink, the beamreceiving the most power can be used to transmit to the desired mobile on thedownlink.


4.10 Adaptive Cell Sectorization in CDMA Systems

As we have seen earlier, traffic imbalance can lead to locking unused capacity onsome sectors or sites, prompting operators to deploy additional carriers beforethey are truly needed networkwide. Hence, balancing the traffic load among thesectors of the cell can help reduce peak loading and increase the cell’s capacity.We have also seen how the forward link capacity is reduced due to soft handoff.Reducing the size of handoff zones and shifting those zones from high- tolow-traffic areas can minimize this negative impact of soft/softer handoff oncapacity. Another common problem in CDMA networks is that of pilot pollu-tion, where multiple pilot signals of approximately equal strength exist at somelocations, resulting in unreliable call originations and terminations, unreliablehandoffs, and decreased capacity. Reducing the size of the coverage footprint ofsuch pilots by decreasing the elevation of offending antennas, introducing downtilt, or reducing the transmit power can help overcome pilot pollution. Adaptivecell sectorization allows an operator to control cell site sectorization for increasedCDMA capacity and improved network performance. It can provide CDMA ser-vice providers with flexible tuning options for controlling interference sources,creating dominant servers or pilots, managing handoff activity and handoffzones, and dealing effectively with nonuniform and time-varying traffic distribu-tions. In this sense, a phased array antenna at the base station can be consideredas an advanced optimization tool that uses beamforming technology to manageand control a specific network’s hot spots and help reduce blocking, rather than


0

–0.5

–1

–1.5

–2

–2.5

–3

–3.5

–440 50 60 70 80 90 100 110 120 130 140

AOA (Degrees)

Arr

ayga

in(d

B)

Figure 4.29 Scalloping.

an actual smart antenna system that attempts to improve capacity and coverageon a systemwide basis. The basic idea is to replace the conventional base stationantenna with a multiple beam antenna system of M beams of θBW beamwidth. Itthen becomes possible to use the individual beams or a group of beams to adaptthe size and position or orientation of sectors to the load distribution. There areseveral ways to combine the beams; for instance, one can combine M/3 beams toform the equivalent of three 120° sectors. Alternatively, load balancing can beachieved by combining different number of beams per sector to customize sectorazimuth pointing angle and beamwidth in θBW increments to balance trafficloading across sectors. Average busy hour traffic distribution based on BSC orswitch data such as channel assignments, successful calls, blocking, and for-ward/reverse power overload control duration can be used to study the networkperformance over time. Based on this evaluation, sectors or sites near or at theblocking thresholds that are contributing to load imbalances can be identified[15]. By reducing peak loading in those most heavily used sectors or sites, the sys-tem can create capacity headroom for traffic growth in the cell and reduce theaccess failure rate. It is important to note that the actual system capacity is notincreased, rather the traffic is merely redistributed in a more balanced way, asillustrated in Figure 4.30. One advantage of this approach is the ability to createany number of sectors and tailor that to the network traffic demand; however,


Sector capacity limit

Sector capacity limit

Alpha

Beta

Gamma

Alpha

BetaGamma

(a)

(b)

Traf

fic/

Sect

orTr

affic

/Se

ctor

Figure 4.30 (a) Load distribution with conventional sectorization; and (b) load balancing with adap-tive cell sectorization.

excessive sectorization is inefficient. The only way to increase the system capacitywith this approach is to eventually transform it into a switched beam system.Furthermore, beam gains and phase settings can be continuously adjusted to cre-ate sharper sector rolloff, thus reducing the size of softer handoff zones andreducing the impact of the handoff reduction factor on capacity.

References

[1] Yang, X., S. Ghaheri, and N. R. Tafazolli, “Sectorization Gain in CDMA Cellular Sys-tems,” First International Conference on 3G Mobile Communication Technologies,2000, pp. 70–75.

[2] Uc-Rios, C. E., and D. Lara-Rodriguez, “On the Effect of Directional Antennas on theReverse Link Capacity of CDMA Cellular Systems,” IEEE 54th Vehicular TechnologyConference, Vol. 1, 2001, pp. 217–221.

[3] Lee, C. C., and R. Steele, “Effect of Soft and Softer Handoffs on CDMA System Capac-ity,” IEEE Trans. on Vehicular Technology, Vol. 47, No. 3, August 1998.

[4] Chan, G. K., “Effects of Sectorization on the Spectrum Efficiency of Cellular Radio Sys-tems,” IEEE Trans. on Vehicular Technology, Vol. 41, No. 3, August 1992, pp. 217–225.

[5] Lee, T. S., and Lee, Z. S., “A Sectorized Beamspace Adaptive Diversity Combiner forMultipath Environments,” IEEE Trans. on Vehicular Technology, Vol. 48, No. 5, Sep-tember 1999.

[6] Wong, T. W., and Prabhu, V. K., “Optimum Sectorization for CDMA 1900 Base Sta-tions,” IEEE 47th Vehicular Technology Conference, Vol. 2, May 4–7, 1997, pp.1177–1181.

[7] Balanis, C.A., “Antenna Theory: Analysis and Design,” John Wiley & Sons, New York,N.Y., 1982

[8] Kraus, J.D., “Antennas,” Second edition, McGraw Hill, New York, N.Y., 1988.

[9] Litva, J., K.Y. Lo, Titus, “Digital Beamforming in WIreless Communications, ArtechHouse, Norwood, MA, 1996.

[10] Stutzman, W. L., and G. A. Thiele, “Antenna Theory and Design,” John Wiley and Sons,1998.

[11] Shelton, J., K Kelleher, “Multiple Beams from Linear Arrays,” IRE Transcriptions onAntennas and Propogation, Vol. 9, Issue 12, 1961, pp. 154–161.

[12] Foster, H. E., R. E. Hiatt, “Butler Network Extension to any number of antenna ports,”IEEE Transaction Antennas and Propogation, Vol. 18, Issue 6, pp. 818–820.

[13] Dong, L., M. A. Ingram, “Beam Selection Algorithum based on PTR metric and its Synchro-nization Performance,” Radio and Wireless Conference, August 10–13, 2003, pp. 115–118.

[14] Matumoto T., S. Nishioka, D. J. Hodder, “Beam Selection Performance Analysis of aSwitched Multi-beam Antenna System in Mobile Communications Environments,” IEEETransactions on Vehicular Technology, Vol. 46, Issue 1, 1997, pp. 10–20.

[15] Mahmoudi, M., E. S. Sousa, and H. Alavi, “Adaptive Sector Size Control in a CDMASystem Using Butler Matrix,” IEEE 49th Vehicular Technology Conference, Vol.2, No.,July 1999, pp. 1355–1359.


5Adaptive Array Systems

Chapter 3 showed how the multipath channel causes significant impairments tothe signal quality in mobile radio communication systems. As signals travelbetween the transmitter and receiver, they get reflected, scattered, diffracted,and shadowed. In addition, user’s mobility gives rise to Doppler shift in the car-rier frequency. As a result, those signals experience fading because of the attenu-ation and phase shifts when they are combined at the receiver. Another sourcethat degrades the performance of signal reception in a mobile environment isinterference, especially in interference-limited systems such as those based onCDMA. Techniques that overcome these impairments and improve system per-formance are examined in this chapter, namely, diversity and adaptivebeamforming, both of which are spatial techniques. Diversity techniques providea diversity gain or a reduction in the margin required to overcome fading. In adigital communication system, this results in an improvement in the requiredSNR or Eb/No necessary to achieve a given quality of service in terms of bit errorrate (BER) or FER. Similarly, beamforming provides several types of improve-ments in terms of array gain, interference reduction, and spatial filtering, whichhave the cumulative effects of improving Eb/No as well. We will show how thesetechniques are applied on both the uplink and downlink and discuss thelimitations of each approach.

5.1 Uplink Processing

5.1.1 Diversity Techniques

As we have shown in Chapter 3, a mobile user can experience both slow and fastfading. The effects of slow fading can easily be overcome using closed loop

117

power control, as is the case in all CDMA systems. Fast power control is used onboth the uplink and downlink at speeds of 800 Hz in CDMA2000 and 1,500Hz in WCDMA. However, even at these speeds, power control may not be fastenough to deal with fast fading, especially at high mobile speeds. Diversityreception is a more effective solution to increase the receiver’s immunity againstfast fading, especially in Raleigh fading channels. The diversity concept relies onthe fact that when multiple replicas or multipaths of the transmitted signal fadeindependently as they go through distinct channels, the probability of a deepfade occurring in all channels at the same time is significantly reduced. Assumethe probability of deep fade in one channel to be given by p, then for L channelsthe probability becomes pL. Improvements in signal statistics depend on twomain conditions, namely, uncorrelated signals with comparable local meanpowers or signal strengths. If the signals from the different multipaths are highlycorrelated then they will fade together and no diversity is provided. Cross-corre-lation coefficients of 0.7 or lower between the multipaths have been shown toprovide reasonable improvements [1]. If the mean signal powers of themultipaths are widely unequal or unbalanced, that would affect the performanceof the diversitycombining techniques. Diversity gain is generally computed bycomparing the cumulative distribution function (CDF) of the Rayleigh fadingon a single antenna to the CDF of the fading signal at the output of the diversityreceiver at a specified probability. An empirical formula for this diversity gainwas derived in [2] for various combining techniques. For maximum ratiocombining, discussed shortly, the two-branch diversity gain at 90% signalreliability is given by

( )G ediv =− −714 0 59 0 11. . .ρ ∆ (5.1)

where ρ is the cross-correlation coefficient between signals and ∆ is the meansignal level difference. Figure 5.1 shows the effect of ρ and ∆ on the diversitygain. It is also possible to define the diversity gain at other percentiles, such as95% and 99%. Obviously, the diversity gain is strongly dependent on thedefinition.

5.1.2 Angle Diversity

Several antenna schemes exist to create the diversity channels necessary toachieve the diversity gain above, including spatial and polarization diversity. Inspatial diversity, a popular technique used in many wireless communication sys-tems, the system uses antennas with vertical polarization horizontally spaced byat least 10λ in order to achieve the required low cross correlation between thediversity branches. Spatial diversity systems are designed such that the signals atthe different antennas of the receiver have low cross correlation with the


maximum gain achieved for uncorrelated signals. The correlation between thesignal replicas is a function of both the antenna spacing and the angle spread ofthe wireless propagation channel. For example, on the base station side, theangle spread is typically narrow—in the order of 10°–20°. Therefore, highdecorrelation between the signals requires large antenna separation, as much as10 wavelengths in order to achieve a high diversity gain. On the other hand, atthe mobile station side, the angle spread is very wide, typically close to 360°, inwhich case the decorrelation of the signals can be achieved with separations onthe order of a quarter wavelength. In the case of multiple beam systems, thearray is typically designed with half wavelengths interelement separation, thus itmight not provide any significant improvements based on spatial diversityunless the angular spread of the received multipaths is such that the lowcross-correlation condition is still met. Polarization diversity uses antennas withorthogonal polarizations to achieve the same conditions necessary to obtain agood diversity gain. The very nature of multiple beam systems offers anotherscheme, referred to as angle diversity, in which the beams are used to create thediversity channels as shown in Figure 5.2, where the different paths are receivedthrough different beams. Note that angle diversity would be expected to providesignificant diversity gains in urban and dense urban environments where richmultipaths always exist but would be of little use in rural environments charac-terized more with line-of-sight Rician channels.

Recall that in urban and dense urban environments a large number of scat-terers typically exist and the mobile station’s signal is received at the base station

Adaptive Array Systems 119

Figure 5.1 Diversity gain as a function of correlation coefficient.

from several DOAs. Therefore, these types of environments tend to have largeangular spread. On the other hand, in rural environments the received signal atthe base station is dominated by an LOS component and very few multipathcomponents. This yields a small angular spread. Therefore, using narrow beamsin the base stations allows the receiver to capture different replicas of the signalfrom the multitude of DOAs. The narrower the beam, the better the spatial res-olution becomes and in rich multipath environments, angular diversity has beenshown to yield diversity gain comparable with space diversity. This is because inrich scattering channels, all beams are expected to receive signal paths with lowcross correlation and comparable mean powers. By contrast, those same narrowbeams will yield unbalanced mean signal powers in the diversity branches inRician fading LOS channels, causing a degradation to the diversity gain. In [2],multiple beam antennas with 24 15° beams and 12 30° beams with selectioncombining were used to compare the performance of angular diversity and spacediversity. The reported results show an angular diversity gain of 8.5 dB with the15° multibeam antenna and 7.5 dB with the 30° multibeam antenna versus 8dB and 7 dB, respectively, of space diversity gain at 99% reliability level in anurban environment. In a rural environment, space diversity outperformed angu-lar diversity by 2.5 dB with 15° multibeam antennas and 4.5 dB with 30°multibeam antenna at a 99% reliability level due to imbalance in the meanreceived signal strength between the diversity branches.


Figure 5.2 Angular diversity concept.

5.1.3 Maximum Ratio Combining

As discussed in Chapter 4, multiple fixed beam antennas systems generate anumber of signals in all ports, making use of the signals from all the availablepaths in the beams. Diversity schemes can then be used to combine these signals.Since classical switched and selective diversity only selects one signal out of allthe possible diversity branches, they are not suitable for the application at hand.On the other hand, the equal-gain combining technique applies different phaseshifts to all signals (cophasing) and then combines them with equal magnitudes.The drawback of this technique is that overall system SNR is reduced when oneof the branches has a significantly lower SNR than the remaining branches. Themost optimal diversity scheme is the MRC. In MRC the received signals are firstco-phased (aligned in phase), weighted by their instantaneous SNRs, and thencombined. Assume that we have L diversity branches, then the signal received bythe ith branch will be given by

x sh ni i i= +

where s is the transmitted signal, αi is the complex channel attenuation, and ni isadditive noise. The optimum weight for each branch is given by [1, 3].

wh

i

i

i

=2

2σ(5.2)

where σ i2 is the noise power at the branch. It follows that the SNR at the

combiner output is

SNRw h

wMRC

i ii

L

i ii

L= =

=

∑

∑1

2

2 2

1

σ(5.3)

When the noise power at individual branches are all equal to σ i2 , it then

follows that

SNR

hh

h

hMRC

i

ni

i

L

i

nn

i

L

i

n

=

==

=

∑

∑

2

21

2

2

2

2

2

1

21

2

σ

σσ

σ 211

===∑∑ SNRii

L

i

L

(5.4)


5.1.4 Adaptive Beamforming

Consider a linear array composed of M elements and let L be the number ofnarrowband plane waves, centered at frequency ω0 impinging on the array fromdirections θ1 θ2 … θL, as shown in Figure 5.3. Using complex signal represen-tation, the received signal at the mth element can be written as

( ) ( ) ( ) ( )x t s t e n tm ij i k

mi

Li= +− −

=∑ 1

1

(5.5)

where si(t) is the signal of the ith source at the mth element, nm(t) is the noise sig-nal received at the mth element, and

( )kd

i i=2π

λθsin (5.6)

Using vector notation, we can write the array output (M elements) in thematrix form

( ) ( ) ( )X AS Nt t t= (5.7)

where

( ) ( ) ( ) ( )[ ]( ) ( ) ( ) ( )[ ]( ) ( ) ( ) ( )[ ]

X

S

N

t x t x t x t

t s t s t s t

t n t n t n t

M

T

M

T

M

T

=

=

=

1 2

1 2

1 2

L

L

L


lth source

l

d

θ

Figure 5.3 Uniform linear array.

The superscript “T” indicates the transpose of the matrix.The steering matrix whose columns are the steering vectors representing

the array response to a signal at angle θi is given by

( ) ( ) ( )[ ]A a a a= θ θ θ1 2 L L (5.8)

and the steering vector of a linear array

( ) ( )[ ]a θ ijk j k j M k T

e e ei i i= − − − −1 2 1L (5.9)

In wireless mobile communications systems the array usually receivesmany multipath components of the transmitted signal with different DOAs inaddition to the signal received along the direct path. Hence, we can rewrite thetotal signal vector (without the noise and interference components) as [4]

( ) ( ) ( ) ( ) ( ) ( )a a aθ α θ θ1 1 1 1 1 12

s t s t s tl ll

L

+ ==∑ (5.10)

where a1 is called the spatial signature and αl is the phase and magnitude differ-ence between the lth component and the direct path. In adaptive arrays, com-plex weights are applied to the element outputs represented by theM-dimensional vector:

[ ]W = w w w M

T

1 2 L (5.11)

Then the array output can be written as:

( ) ( ) ( )y W Xt w x t tii

M

iH= =∗

=∑

1

(5.12)

The mean output power is thus given by:

( ) ( ) ( )P E y t y t Hs

HNw W R W W R W= = +∗ (5.13)

where * denotes the conjugate and E [.] denotes the expectation operator and Ris the array correlation matrix given by

( ) ( ) R X X R R= = +E t tHs N (5.14)


where H is the complex conjugate transpose of a vector or a matrix and RS andRN represent the signal and noise plus interference correlation matrices, respec-tively. Several techniques exist to obtain the weight vector, which vary in com-plexity and limitations, as discussed next. Those can be classified as suboptimal,such as beam steering and null steering, and optimal, which seek to optimize aselected performance criterion.

5.1.4.1 Beam Steering

Beam steering is the simplest form of beamforming that can be achieved by adelay-and-sum beamformer. The weights of the beamformer are all made equalin magnitude, whereas the phases are selected to steer the main beam of the arrayin a particular direction θ0 [5]. The array weights are given by

( )W aBS =1

0Mθ (5.15)

where M is the number of elements and a(θ0) is the steering vector. In otherwords, the array main beam is steered toward the DOA of the desired source.Thus, beam steering requires the knowledge of the desired signal location. TheSNR at the output of the beamformer is given by:

SNR BS

Hs

HN

=W R

W R

W

W(5.16)

In the special case where the system is dominated by uncorrelatednoise (RN = σ2I ) and no dominant interference exists, the output SNR thenbecomes

( ) ( ) ( ) ( )

( ) ( )SNR M

P

M

P

M

MPBS

H H H

n H n n

= = =

12 0 0 0 0

2

2 0 0

2

a a A A a a

a a

θ θ θ θ

σθ θ

σ σ 2(5.17)

5.1.4.2 Null Steering Beamforming

In a null steering beamformer, the signal arriving from a known directioncan be cancelled by placing nulls in the array response at the DOA of that signal[5]. This can be accomplished by choosing the beamformer weights so that abeam with unity gain is created toward the DOA of the desired signal whilenulls are created at the directions of interference. This can be formulated asfollows


( ) ( ) ( )[ ]W A W a a aH Hk= =

θ θ θ0 1

1

0

0

LM

(5.18)

where a(θ0) is the steering vector associated with the desired signal and a(θ1),…a(θk) are those associated with the interfering signals. To null the M – 1 interfer-ing signals we then get

[ ]W ANSH T= −1 0 0 1L (5.19)

One disadvantage of null steering techniques is that it requires the knowl-edge of the DOA of all interference sources. In addition, beam steering and nullsteering do not result in the maximum output SNR except in special cases. Toobtain the optimal performance, several criteria can be used to derive the opti-mal weight vector. These include maximum SINR, minimum mean squareerror, minimum variance, and maximum likelihood [6–8].

5.1.4.3 Maximum Signal-to-Interference and Noise Ratio

As we have seen earlier, the SINR at the beamformer output is given by:

SINRH

SH

N

=W R W

W R W

where [ ]R SSSHE= is the desired signal’s correlation matrix and R N E=

NN H is the noise and interference correlation matrix. If we maximize the

quantity WHRSW subject to the constraint that WHRNW = 1, then we canachieve the maximum SINR. Using the method of Lagrange multipliers and set-ting the gradient with respect to W to zero we then get

( )[ ]∇ + − = − =W R W W R W R W R WHs

HN s Nλ λ1 0 (5.20)

This results in the eigenvalue problem

R W R Ws N= λ (5.21)

This can be further written as

R R W WN− =1

s λ (5.22)


The solution of (5.22) requires that W be an eigenvector of R RN s−1 . Hence,

the eigenvector corresponding to the maximum eigenvalue (λmax) would maxi-mize the SINR and we get SINRmax = λmax. Therefore

R W R W R Ws N NSINR= = ⋅λmax (5.23)

Solving for the optimum weight and using the definition of the signal cor-relation matrix we then get:

W R Aopt SINR NK= −1 (5.24)

5.1.4.4 Minimum Mean Square Error

In the MMSE criterion, the error between the desired signal or a reference signald(t) closely representing the desired signal and the output of the beamformer isminimized. The mean square error is given by:

( )[ ] ( ) ( )[ ] ( ) ( )[ ] E e t E d t t d t tH H H2 = − −W X W X (5.25)

which can be expanded as

( ) ( ) ( ) ( ) ( )

( ) ( ) ( ) ( )E e t E

d t d t d t t

t d t t t

H H

H H H H

2 =− −

+

X W

W X W X X W

The gradient of the MMSE with respect to W can then be written as

( ) ( ) ( ) ( ) ( ) ∇ = − +E e t E d t t t tH H2 2 2X X X W (5.26)

Setting this gradient to zero, we get the optimum weight given by

W R ropt =−1 (5.27)

where r is ( ) ( ) E d t tH X . Using (5.25) and letting d(t) = s(t), then

( ) ( )( ) r S AS N A= + =E t tHsσ 2 (5.28)

When the interference is orthogonal to the desired signal, the correlationmatrix in (5.14) can be expressed as


R AA R= +σ sH

N2 (5.29)

using the matrix inversion lemma

If F B CD C= +− −1 1 H then ( )F B BC D C BC C B− −= − +1 1H H

Hence using F R B R C A D= = = =−, , ,N s1 21 σ , we get

R R A A R A A R− − −

−

−= − +

1 1

21

1

11N

s

HN

HNσ

(5.30)

R RA R A

A R A

− −−

−= −

+

1 11

211N

HN

s

HN

I

σ

(5.31)

which could be simplified to

R R1A R A

− −−=

1 12 1Ns

HN1+ σ

(5.32)

Hence,

WA R A

R A R Aopts

sH

NN MMSE NK=

+

=−

− −σ

σ

2

2 11 1

1(5.33)

One disadvantage of the MMSE scheme is that it requires the knowledgeof the desired signal or a closely correlated replica to use as the reference signal.

5.1.4.5 Minimum Variance Distortionless Response

Another optimum performance criterion involves minimizing the array outputso that the desired signals are passed with specific gain while minimizing thecontributions due to noise and interference. In other words:

min W RW W A dH Hsubject to r= (5.34)

In (5.34) Ad is the steering matrix pointing to the desired signals and r isthe V × 1 constraint vector, where V is the number of desired signals. When the


elements of r are all 1s, the criterion is known as the MVDR. From (5.34) wecan see that minimizing the variance of the beamformer output power is equiva-lent to minimizing WHRNW. The method of Lagrange multipliers can be usedto solve this constrained minimization problem by taking the gradient withrespect to W and setting it to zero as follows

( ) ∇ + − = − =W R W W A R W AHN

Hd N dλ λ1 0 (5.35)

Hence,

W R Aopt N d= −λ 1 (5.36)

Multiplying through by A dH and making use of the fact that WHAd = 1, we

get

A W A R AdH

dH

N d= =−λ 1 1 (5.37)

yielding

λ = −

11A R Ad

HN d

(5.38)

Therefore, the optimum weight is given by

WR A

A R AR Aopt

N d

dH

N dMVDR N dK= =

−

−−

1

11 (5.39)

One advantage of the MVDR beamformer is that it does not require anyknowledge of the directions of the interference, rather only those of the desiredsignal(s).

5.1.4.6 Optimum SINR

We have shown in the previous sections that the optimum weight has the form

W R Aopt NK= −1 (5.40)

Note that in (5.40), A refers to the steering vector/matrix or it could alsobe substituted for by the spatial signature. It follows that the optimum SINR


can be obtained by substituting the optimum weight into the SINR equation,which results in

( ) ( )( ) ( )

SINRK K

K Kopt

N

H

s N

N

H

N N

sH

N= =− − −

− − −

R A R R A

R A R R AA R

1 1 1

1 1 1

2σ −1 A (5.41)

For simplicity let us rewrite the array output assuming we have adesired signal Sd, an interference signal Sl, and thermal background noise [9].This could, for instance, represent a simplified multirate model for the CDMAuplink where we have a low data rate voice user, a high data rate user, anda white noise term that includes the thermal background noise plus all otherlow data CDMA multiple access interference terms. In other words, we areassuming that enough low data CDMA users are present, who each areindependently Rayleigh fading that they can be modeled as white spatialnoise. The equivalent white noise term on each antenna element is complexGaussian and the noise on all antenna elements has zero mean and varianceequal to σ2.

( ) ( ) ( ) ( )X = A A Nt S t S t td d I I+ + (5.42)

We further assume that the equivalent white noise term is uncorrelatedwith Sd, Sl, Ad, and AI. The power in signal Sd is E S Sd d

Hs= σ 2 and the power

in signal Sl is E S SI IH

I= σ 2 . It then follows that noise and interference correla-

tion matrix is given by

R A A RNI I I IH

N= +σ 2 (5.43)

where RN = σ2I. Using the matrix inversion lemma we can write

[ ]R R R A A R A RNI N N I I I IH

N I I I IH

NA− − − − − −= − +1 1 1 1 1 11σ σ σ σ (5.44)

R A A

A ANI

II I

H

IIH

I

I− = −+

1

2

2

2 2

2

1 1

1σ

σ

σ σ

σ

(5.45)

which could be reduced to


RA A

A ANI

I IH

IIH

I

I− = −

12 2

2

1

σ σ

σ

(5.46)

Substituting into the expression for SINR we get

SINR Iopt sdH

I IH

IIH

I

= −+

σ

σ σ

σ

22 2

2

A A A

A A

A d (5.47)

SINR opts H

ddH

I IH

d

IIH

I

= −+

σ

σ σ

σ

2

2 2

2

A AA A A A

A A

(5.48)

Since ( )A A A AIH

d IH

d S= cos θ , where θS is the angle between the sig-

nal and interference steering vectors, we can then write

( )SINR opt

s d I s

II

= −+

σ

σ

θ

σ

σ

2 2

2

2 2

2

2

2

1A A

A

cos

(5.49)

which can be rearranged as

( )SINR opt

s dI s

I

II

=+

+

σ

σ

θσ

σ

σ

σ

2 2

2

2 22

2

2

2

2

AA

A

sin

(5.50)

5.1.5 Fixed Multiple Beams Versus Adaptive Beamforming

To compare the performance of adaptive beamforming to that of the fixed mul-tiple beams followed by the MRC approach, we must first derive a generalized


form for the SINR in the MRC case. Let us rewrite the general form of theMRC weight as

WA

MRCd=

σ 2(5.51)

Substituting into the SINR expression we get:

( ) ( )( ) ( )

SINR MRCd

H

s d

d

H

NI d

dH

s d dH

d

dH

NI d

= =A R A

A R A

A A A A

A R A

σ 2

(5.52)

( )( )SINR

+MRC

s dH

d

dH

N T I IH

d

=σ

σ

2 2

2

A A

A R A A A

Since RN = σ2I, it then follows that

( )SINR MRC

s dH

d

dH

d I dH

I IH

d

s d

I

=+

=+

σ

σ σ

σ

σ σ

2 2

2 2

2 2

2

A A

A A A A A A

A

( )2 2 2A I scos θ

(5.53)

Now that we have derived expressions for the SINR achievable using mul-tiple beam antennas with MRC and optimum adaptive beamforming, generalcomparisons can be drawn regarding the performance of both approaches andthe environments for which they are most suitable.

• Noise-dominated systems: In systems where background noise is domi-nant, such as the case of no or very little directional interference in thechannel, we then end up with

SINR SINRopt MRCs d= =

σ

σ

2 2

2

A(5.54)

This is also the case when the interference is spatially white, suchas in 2G CDMA (IS-95A) based systems, where all users have low datarates and no interference spatial coloring is experienced. This wouldalso be the case in 3G CDMA systems (based on either IS-2000 orWCDMA) when only low data rate users are present in the sector.Hence, in such systems, optimum adaptive beamforming will perform


as well as multiple fixed beam antennas followed by MRC since there isno interference to cancel.

• Spatially collocated desired and interference signals: In this case the anglebetween the desired and the interference signals steering vectors or spa-tial signatures θs = 0°. It then follows that

SINR SINRopt MRCs d

I I

= =+

σ

σ σ

2 2

2 2 2

A

A(5.55)

In this case, since the desired and interference signal have thesame DOA, both approaches would perform similarly.

• Interference-limited systems: When we have one or more low data rateusers along with one or more high data rate users, then the interfer-ence experienced by the lower data rate users will be directional orspatially colored. Such is the case in multirate services, where both lowdata rate speech service and high data rate applications are simulta-neously supported. It becomes apparent by comparing the SINR forboth approaches that optimum adaptive beamforming will provide again over multiple fixed beam antennas with MRC because of the abil-ity to deal with the directional interference more effectively. Theamount of this gain will depend on the ratio of the interference-to-noisevariance, θs, and the spatial signature or steering vector of the interfer-ence signal.

5.2 Downlink Processing

Just as fading and interference degrade the performance of mobile communica-tions systems on the reverse link (uplink) they also impair the performance ofthe forward link or downlink, reducing the capacity and affecting the signalquality. In Chapter 4, we presented expressions for the reverse link capacityof a CDMA system that showed that reducing the Eb/Nt that denotes theenergy per bit to noise and interference ratio necessary to maintain therequired signal quality measured for instance by the FER would increase thecapacity. To understand the impact of fading and interference on the perfor-mance of the downlink, we need to calculate the capacity of that link. Unlikethe uplink case, interference on the downlink of a CDMA system has manydifferent sources, including overhead channels such as the pilot, sync, and pag-ing. Since the base station power is limited, the downlink capacity will dependon the strength of the overhead or common channels, as well as all the supportedactive traffic channels. Recall that the traffic channels are power controlled; that


is, the transmit power allocated to each traffic channel is adjusted by theouter and inner loop power control to achieve the required Eb/No in the pres-ence of fading, interference, and other channel impairments. We can thenwrite the following for the total base station power of an IS-2000 CDMAsystem:

Ptot = Total Overhead Power (paging, pilot, and sync channels) + Number ofTraffic Channels * Traffic Channel Power.

It is then straightforward to obtain the downlink pole capacity of anIS-2000 system as [10]

N

PilotE

IPaging

E

ISync

E

I

h E TrMAX

c

or

c

or

c

or=− + +

⋅

1

afficE

Ic

or

(5.56)

In general, the downlink CDMA capacity is given by

N

Common ChannelsE

I

h E TrafficE

I

MAX

c

or

c

or

=−

⋅

1

(5.57)

where NMAX is the maximum number of supportable users, h is the handoffreduction factor, and Ec /Ior is the average channel transmit energy, relative to thetotal CDMA channel transmit energy. The handoff reduction factor accountsfor the soft handoff links that consume additional downlink power for a givenuser. The traffic channel Ec /Ior is a function of the voice activity factor, chiprate, data rate, and traffic Eb /Nt, which is the required traffic channel receivedSNR corresponding to rate Rb at nominal FER. Examining (5.57) one mightargue that the CDMA downlink pole capacity can be improved simply byincreasing the total transmitter power. However, this argument ignores the factthat when using conventional sectorized antennas, increasing the transmitpower would simply raise the interference seen on the downlink, and the trafficchannel Ec /Ior required to maintain the same FER or BLER would also increaseto combat the higher interference level. The only and most effective way toimprove the capacity would then be to improve the traffic channel Ec /Ior, that is,adopt techniques that would reduce the required Ec /Ior for a given FER orBLER.


5.2.1 Transmit Diversity Concepts

As we have shown in the previous sections, receive diversity helps combat fadingand is widely used in 2G systems to improve the base station receiver perfor-mance. Similarly, transmit diversity can improve the MS receiver performanceby providing it with multiple independent copies of the transmitted signal [11,12]. Note that at the mobile side, receive diversity can be used to achieve perfor-mance gains similar to those obtainable with transmit diversity, as will bedescribed in detail in Chapter 9. There are two categories of transmit diversity(TD) schemes, open loop techniques, which do not require any channel informa-tion at the transmitter, and closed loop techniques, which rely on feedback chan-nels from the MS or UE that provide necessary channel information. Commonmethods for TD employ two spatially separated antennas (several wavelengthsapart) and use delay or frequency diversity. In delay diversity, copies of the samesymbols are transmitted from multiple antennas at different times (i.e., withdelays). This has the effect of producing multipath-like distortion at the receiverthat could be exploited to obtain some diversity gain. A major drawback of delaydiversity is the reduced throughput because we have to send the same symbolsmultiple times.

Let us assume that the signal received at the MS without TD is given by

x sh n= + (5.58)

where s is the transmitted signal, h is the complex channel attenuation, and n isadditive noise whose variance is σ2. Let the signal variance be given by σ s

2 so the

SNR at the mobile station is given by:

SNR hs= ×σ

σ

2

2

2 (5.59)

Bad channel conditions mean that |h|2 is small, in which case the base sta-tion has to increase the transmit power to achieve the same quality of service(same SNR) at the mobile station. This in turn increases the interference level inthe system since the transmit power increases and, consequently, the forwardlink (downlink) capacity decreases. To reduce this fading impact, the followingtransmit diversity schemes were proposed for the 3G CDMA standards.

5.2.2 Transmit Diversity in 3G CDMA Standards

5.2.2.1 Open Loop Transmit Diversity

Since open loop TD (OLTD) techniques do not require any feedback from theMS, there is no additional signaling overhead and the receiver complexity is not


increased by much. Three OLTD techniques were adopted in the 3G standards,orthogonal TD (OTD), space-time spreading (STS) for the IS-2000 standard[13], and space-time TD (STTD) for the WCDMA standard [14]. To improvediversity gain in a multiple access environment, mutual interference between thetransmitting antennas at the receiver is minimized through the use of orthogonalsignals on each antenna. The orthogonal signals are constructed using a pair ofcomplex symbols every two-symbol (2T) period.

Orthogonal Transmit Diversity

The principle of OTD is illustrated in Figure 5.4. OTD is a transmit diversitytechnique using two spatially separated antennas. After encoding and interleav-ing, the symbols are split into two different streams, even and odd. Even sym-bols are transmitted on the first antenna, whereas odd symbols are transmittedon the second antenna simultaneously each at half power using two orthogonalWalsh codes W1 and W2. Let the transmission matrix be defined by X that repre-sents the set of two transmitted signals over two antennas over 2T duration. Thetransmission matrix X will then represent the output of the antenna pair every2T time interval. The rows of X indicate the transmitted signal every 2T timeinterval per antenna. In the OTD case, the matrix is

XS W S w

S w S we e

o o

=−

(5.60)


Encoder Demux 1W

2W

1pilot

2pilot

es

os

Figure 5.4 Orthogonal transmit diversity.

where Se,So denote the even and odd symbol with 2T length and w is anorthogonal Walsh code with length T. We can also write X in terms of effectiveorthogonal sequences of period 2T as

XS W

S We

o

=

1

2

(5.61)

where W1 = [w, w] and W1 = [w, w ] and w, w are complementary Walsh codes.The received SNR at the mobile station is:

For even symbols:

SNR hes= ×

σ

σ

2

2 1

2(5.62)

and for odd symbols:

SNR hos= ×

σ

σ

2

2 2

2(5.63)

Space-Time Spreading (STS)

The STS technique is another OLTD scheme based on Alamouti’s idea [15]. InSTS, both symbols are transmitted on both antennas over a 2T time interval byusing different orthogonal Walsh codes, as shown in Figure 5.5. The transmis-sion matrix is given by

XS W S W

S W S We o

e o

=−+

1 2

2 1

*

*(5.64)

where W1,W2 are two orthogonal codes with 2T length. Since this schemeachieves intra- antenna orthogonality using Walsh codes (i.e., in the spreadingcode domain), we can denote it as a CDM technique [16].

Space-Time Transmit Diversity

STTD is another OLTD technique that implements Alamouti’s space-timeblock code. In STTD, both symbols are transmitted on both antennas over twoconsecutive T duration time slots, as shown in Figure 5.6. The output matrixbecomes

XS w S w

S w S wo e

e o

=−

* * (5.65)


where w is an orthogonal code with length T. Therefore, the total transmissiontime remains 2T. The same Walsh code is used in STTD as opposed to the STScase since the intra-antenna orthogonality is achieved in the time domain (i.e.,by TDM signaling). Both STTD and STS schemes are similar and they achievethe same performance in terms of diversity gain. The received SNR at themobile station in the case of STS and STTD is given by [17]

( )SNR h hs= × +σ

σ

2

2 1

2

2

2

2(5.66)

As seen earlier, OTD showed a diversity gain compared with thenontransmit diversity case. But with OTD, if one of the channels experiencesdeep fading, half the symbols are received with a bad SNR since they are onlytransmitted from one antenna. This problem is solved with STS or STTD sinceeach symbol is received with the same SNR that depends on the two channels.

5.2.2.2 Closed Loop Transmit Diversity

In closed loop transmit diversity (CLTD), the base station uses the channelknowledge it receives from a feedback channel from the MS to calculate trans-mit weights for the signals on the different antennas such that the SNR at theMS receiver is maximized. In CLTD the same symbol is transmitted from thetwo antennas with the same spreading sequences or Walsh codes. Since with thistechnique the weights adapt with the MS movement, this CLTD scheme isadaptive in nature and is often referred to in the literature as transmit adaptive


Encoder Demux

--------

1W

2W

2W

1W

Figure 5.5 Space-time spreading.

antennas (TXAA), with the goal to enhance the forward link capacity of CDMAsystems [18]. Besides being adaptive, another difference between CLTD andOLTD schemes discussed earlier is that whereas STS and STTD cannot be eas-ily extended to a higher number of antennas, TXAA can theoretically be usedfor any number of antennas, however, certain implementation and performanceissues bound the maximum number of antennas. The block diagram of TXAAfor two antennas is shown in Figure 5.7. Each antenna at the base station isweighted by a complex coefficient so as to maximize the SNR received at themobile station. The data on the forward traffic channel or dedicated channelsare transmitted simultaneously with the same code on each antenna but withantenna-specific amplitude and phase weighting. The complex weight coeffi-cients depend on the fading channels corresponding to each antenna. So eachmobile station estimates the fading channels using the pilots transmitted on eachantenna by the base station. The MS then computes the optimal weights thatthe base station must use to maximize the power received at the mobile station.Once it has computed the coefficients, the mobile station feeds them back to thebase station, which applies them to transmit data to the mobile. In other words,each user computes a set of weights for the antenna array and feeds them back tothe base station. To aid the MS with the demodulation, the base station cantransmit an additional pilot: a dedicated per-user pilot. In the presence of errorsin the transmission of the optimal weights computed by the mobile station, thebase station applies corrupted coefficients and the mobile station can use thisdedicated pilot to estimate the weights that were used and demodulate the datareceived on the traffic channel.

Now let hi be the channel coefficients for each antenna i, i ∈ [1 … M],where M is the number of transmit antennas. Let wi be the weight for a givenuser. To keep the same transmit power, we impose the following constraint:


Encoder DemuxW 1pilot

2pilot

eo ss ,

** , oes–s

W

Figure 5.6 Space-time transmit diversity.

w ii

M2

1

1=∑ =

The received signal at the mobile station is then given by:

x s w h ni i ii

M

= ⋅

+

=∑

1

(5.67)

where s is the transmitted signal. Let’s define H = [h1 h2 … hM] and W = [w1w2 … wM]T. We can then rewrite the received signal as

X HWs n= + (5.68)

The SNR at the mobile station is:

( )SNR W H HWs H H=σ

σ

2

2(5.69)

The optimal weights are those that maximize the received SNR orWHHHHW subject to the constraint W =1. It follows that the solution to this is


Commonchannels

Trafficchannel

)(1 tw

w

Sector pilot P0

Auxiliary pilot P1

User specific pilotPu

Decodeweights

RX TX

RX

Matchedfilter for

P0

Matchedfilter for

P1

Compute/encodeweights

)(ˆ1

th

)(1 th

)(ˆ2

th

)(2 th

)(2 t

Figure 5.7 TXAA architecture.

the eigenvector corresponding to the largest eigenvalue of the matrix HHH.When H is a vector, the optimal weight becomes:

WH

HHopt

H

H= (5.70)

which means that the weights are the complex conjugates of the channel coeffi-cients. Substituting the expression for the optimal weight into the SNR we thenget for the TXAA

SNRH

HHH H

HHHTXAA

sH

H

H

HH

Hs=

=

σ

σ

σ2

2

2

2

2

σH (5.71)

For the two antenna cases, the above equation reduces to

( )SNR h h hTXAAs

ii

Ms= = +

=∑σ

σ

σ

σ

2

21

2 2

2 1

2

2

2(5.72)

Comparing (5.72) and (5.66) we can see that TXAA increases the receivedSNR by a factor of two. This means that the transmit power required to achievea given FER or BLER is halved, which corresponds to a gain of 3 dB in thetransmit power. Note that with TXAA, just like the STS and STTD schemes,the diversity gain is a function of both channels. Furthermore, by examining theSNR in (5.72) we can conclude that with TXAA, the performance gain willincrease as the number of channels and antennas increase. This extra gain com-pared with OLTD schemes can be explained by the fact that in TXAA theweights are chosen in such a way that the signals transmitted on each antennaadd constructively at the mobile station because the optimal weights are the con-jugates of the channel coefficients. So by applying those weights, we compensatefor the phase differences between the channels corresponding to each antenna,and then each is weighted in amplitude as a function of the quality of thechannels.

TXAA Operation

The first step in the TXAA operation is to estimate the channel from eachantenna to the mobile. Matched filters to the orthogonal user-specific pilotsare used to perform this channel estimation. Based on this estimate, the MScan compute the weights, which are up to a scaling factor the conjugate ofthe channel. Now, since we can only transmit a limited number of bits perupdate on the feedback channel—otherwise the overhead is increased—we need


to quantize the weights that the mobile station feeds back to the base station.Let’s define the following notations for the weights corresponding to the twoantennas:

w w e

w w e

j

j

1 1

2 2

1

2

=

=

φ

φ

The following quantization is adopted in the 3G standards, one bit is usedfor the power ratio of the two antennas w w1 2/ and three bits for the phasedifference between the two antennas: φ2 – φ1. For instance, in WCDMA thereare two modes of operation defined for CLTD, closed loop mode 1 (CL1) andclosed loop mode 2 (CL2) [14]. In CL1 the amplitude ratio between the weightsis kept at one and only the phases are changed. In odd-numbered slots, themobile can choose from a phase difference of 0 or π between the antennas,whereas in even-numbered slots the phases are either π/2 or –π/2. In CL2 theweights are chosen based on Tables 5.1 and 5.2.

Once the MS have transmitted the weights back to the base stationthey are applied, after transmission and reception delays, to the antennas.Note that in fast fading conditions, the channel may change faster than theweights are applied because of those delays and this would result in outdatedweights.


Table 5.2Phase Ratios in CL2 Mode

Index 000 001 010 011 100 101 110 111

Phase Ratio 180° –135° –45° –90° 135° 90° 0° 45°

Table 5.1Gain Ratios in CL2 Mode

Index 0 1

w 1 0 8. 0 2.

w 2 0 2. 0 8.

Gain ratio (dB) 6 –6

5.3 Downlink Beamforming

While CLTD or TXAA techniques can improve the downlink performance byreducing the impact of fading they are not capable of addressing interferenceissues. Moreover, while TXAA performance gains can be improved by increas-ing the number of antennas, there are some practical implementation issues thatwould limit their performance. Recall that TXAA require wide antenna spacing(∼5–10 λ) to achieve low cross correlation between the elements and conse-quently provide independent channels at the MS. Obviously, increasing thenumber of antennas would be impractical due to space and tower limitationsand would consume more of the overhead signaling capacity because of theincreased feedback information requirements. Furthermore, increasing thenumber of elements at these spacings would generate grating lobes in theantenna array radiation patterns, which would spread more interference in thedownlink, as shown in Figures 5.8 and 5.9. Hence, the most promising


150

180

210

240270

300

330

0

30

6090

1201

0.8

0.6

0.4

0.2

150

180

210

240270

300

330

0

30

6090

1201

150

180

210

240270

300

330

0

30

60

90

1201

0.80.6

0.4

0.2

(a) (b)

(c)

0.8

0.6

0.4

0.2

Figure 5.8 Two-element array radiation pattern. (a) d = λ/2, (b) d = 5λ, (c) d = 10λ.

technique to combat interference on the downlink would be to use narrowbeams that would be transmitted to a desired user or cluster of desired users andnulls in directions of interference.

For beamforming applications, the interelement spacing required is λ/2since they require high correlation between the antenna elements. The spatialcorrelation between antenna elements for small angular spreads (below 25°) canbe approximated as [19, 20]

( )ρπ θ

λ

π σ θ

λd e ej

dd p

=−

2

22

2

sincos

(5.73)

where σp is the angular spread AS. Equation (5.73) is shown to be valid forGaussian, uniform, and Laplacian power azimuth distributions in [20]. Further-more, it is shown to provide very accurate approximation for high correlation


150

180

210

240270

300

330

0

30

6090

1201

0.8

0.6

0.4

0.2

150

180

210

240270

300

330

0

30

6090

1201

150

180

210

240270

300

330

0

30

60

90

1201

0.80.6

0.4

0.2

(a) (b)

(c)

0.8

0.6

0.4

0.2

Figure 5.9 Four-element array radiation pattern. (a) d = λ/2, (b) d = 5λ, (c) d = 10λ.

values but introduces some errors for low correlation values below about0.4. Since good diversity performance is achievable for correlation valuesbelow about 0.7, as described earlier, this approximation can be used to evalu-ate the impact of angular spread on the spatial correlation between a pair ofantenna elements regardless of the angular distribution function. Figure 5.10illustrates the effect of the AS on the antenna elements spatial correlation. Wecan see that for narrow AS, high correlation values result with small separations,whereas low correlation values require larger separations from which thewell-known rule of thumb of 10λ at the base station is derived. On the otherhand, for large AS, low correlation values can still be obtained even with smallseparations.

This has a significant impact on the performance of spatial techniquessuch as transmit diversity TD and beamforming. In the case of TD, low correla-tion is necessary to achieve large diversity gains. This can be easily achievedin environments with large angular spreads without the need for large separa-tions, whereas it would require separations on the order of 10λ with narrowangular spreads. This implies that TD techniques are more suitable for


Spat

ial c

orre

latio

n

1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

00 1 2 3 4 5 6 7 8 9 10

Element separation (wavelength)

AS=3 degAS=5 degAS=10 degAS=15 deg

Figure 5.10 Spatial correlation versus element spacing at the base station side for various AS.

channels with large AS. Similarly, since beamforming performance depends onhaving high correlation values between the antenna elements, we can con-clude from Figure 5.10 that their performance will outperform that of TDin narrow AS environments and show some degradation as the AS isincreased. On the mobile side, the AS is much higher than that at the basestation side and, as we see from the correlation coefficient plot in Figure 5.11,low correlation values can be achieved even for very small separations. Thisimplies that diversity techniques can provide good diversity gain even withthe physical constraints of mobile stations, as we will see in more detail inChapter 9.

5.3.1 Spatial Signature-Based Beamforming

Let us assume that we have K sources each represented by their spatial signatureSS given by


Spat

ial c

orre

latio

nco

effic

ient

1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

00 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5

Element separation (wavelength)

AOA=0 degAOA=10 degAOA=30 degAOA=60 deg

Angle spread = 360 degrees

Figure 5.11 Spatial correlation versus element spacing at the mobile side with various AOAs andAS of 360°.

( ) ( ) ( ) ( ) ( ) ( ) ( )x a a at s t s t s tl ll

L

= + ==∑θ α θ θ1 1 1 1 1 1

2

(5.74)

then the total received signal at the base station antenna array from all sources isgiven by

( ) ( ) ( )X a nt s t tk kk

K

= +=∑

1

(5.75)

One approach for the beamforming weight vector design [4, 21] is tocapture the uplink SS and transmit its complex conjugate back on thedownlink. This is equivalent to maximizing the SNR at the MS. Note that thistechnique does not attempt to cancel any interference on the downlink.This technique is highly effective and suitable for TDD systems in whichthe uplink and downlink share the same carrier frequency. However, in FDDsystems like those based on CDMA the uplink and downlink carrier frequen-cies are different and the SS captured on the uplink is not the same as thedownlink SS. Spatial signatures change as a function of time, physical displace-ment, and frequency. For a stationary MS, variation of the spatial signatureis primarily a result of changes in the local scatterers distribution. This wouldbe vehicle movement, surrounding pedestrian movement, or any other physi-cal changes of the scattering objects in the vicinity of the MS. SS variationsversus frequency for an FDD system reported in test results [4, 22] show thatfrom 874 to 924 MHz the relative amplitude change of the spatial signaturewas on the order of 14 dB peak to peak and the relative angle change was closeto 100%. The spatial signature variation is thus quite significant for smallchanges in frequency, mainly due to the wavelength differences relative to thepaths traveled by the received signal. Additional test results [23] at 1.6 GHz alsoconfirm significant changes in spatial signature as a result of change in fre-quency. Given the experimental results available in the literature, it can be con-cluded that the spatial aspects of the RF channel (urban, suburban) changesignificantly with the frequency differences on the order of CDMA FDD sys-tems (IS-2000, WCDMA), so an adaptive antenna system that bases downlinkbeamforming weights solely on uplink spatial signatures will likely besuboptimal.

5.3.2 DOA-Based Beamforming

In the DOA approach the uplink channel correlation matrix or the uplink SSare first captured and then some DOA estimation technique is applied to


determine the signal’s DOAs and their associated amplitudes. The DOA θmax

with the maximum amplitude, which would indicate the strongest path, isselected and its array response vector a(θmax) is chosen as the downlink beam-forming weight. Note that this is equivalent to the beam steering techniqueused earlier on the uplink. Recall that the spatial signature is a linear combina-tion of the array response vectors from the direct and multipath compo-nents of the signal, which are dependent upon the DOA of the signal andthe angular spread (AS). Let us assume that the main signal’s DOA is the peakof the power azimuth spectrum (PAS), which is valid for sufficient averagingof the PAS in urban and suburban outdoor RF environments as the shapetends toward a Laplacian distribution. DOA variation versus frequency wastested in [24]. The results show negligible variation in DOA for this fre-quency change compared with the spatial signature variations in the 899 to924 MHz range. In [25] the averaged channel characteristics of uplink anddownlink duplexes under fast fading conditions is investigated. The test bedwas a base station uniform linear array of eight elements with half wave-length spacing at 1,780 MHz and a single MS transmitting antenna. Resultsshow that the main lobes of the PAS are almost identical (within a fewdegrees) for uplink and downlink frequencies. The major differences betweenthe two are in the tails of the PAS. We can then conclude based on the pub-lished measurements results that, provided we apply sufficient averaging, theMS DOA and AS characteristics over the uplink and downlink duplexesshow a high degree of correlation. These results show that in an adaptiveantenna system (where no MS channel feedback is available), pointing beamsin the general direction of the received DOAs has the potential to besuccessful.

5.3.3 Maximum SNR

Let us assume that we have L signal sources, each with K multipath components.The downlink channel vector can then be written as

( ) ( ) ( )h aDLj f t

ii

LK

tP

LKe i i= +

=∑ ϕ π θ2

1

(5.76)

where P is the channel power, ϕi is the random phase of the ith component, fi isthe Doppler frequency shift of the ith component, and θi is the angle of depar-ture of the ith component distributed over θ σ θ σ− +p p/ , /2 2 . One

beamforming criterion is to select the weights that optimize the array gain sothat the SNR or SINR at the MS is maximized subject to the constraint that the


total transmitted power is equal to that of a single antenna. Let the powerreceived by the kth MS be denoted by

Pk kH

DLSUM

k= w R w (5.77)

where the sum of all the mean channel correlation matrices corresponding to alltaps given by

( ) ( )[ ]R R h hDLSUM

i i iH

i

LK

i

LK

E t t= ===∑∑

11

(5.78)

Hence, the maximization problem can be formulated as follows:

Maximize SNR kH

DLSUM

k=w R w

σ 2subject to w k

21= (5.79)

As we have seen in a previous section, the solution to this problem is theunit norm principal eigenvector (corresponding to the maximum eigenvalue) ofR DL

SUM

w uopt DL= ,max (5.80)

The issue now is that the base station needs information about thedownlink correlation matrix to calculate the eigenvectors. One possible solutionis to capture the uplink correlation matrix, calculate its principal eigenvectoruDL, max, and use it as the weight vector. As long as the FDD gap is small enoughthere is a strong correlation between the uplink and downlink average statisti-cal properties. Results reported in [26] show that for two-, four-, and eight-ele-ment arrays with σp ∼ 10° the correlation between the uplink and downlinkprincipal eigenvectors ρul, dl ≥ 0.9 for FDD gaps up to 500 MHz with a perfor-mance loss of 0 to -1 dB. For larger FDD gaps, there will be a performanceloss that can be as high as -6 dB and increases with the number of elements.This implies that current CDMA systems with FDD gaps of 45 MHz or 80MHz in the North American IS-2000 bands and 190 MHz in the EuropeanWCDMA bands would be able to apply this technique without sacrificingsignificant performance. To avoid or at least minimize the performance losscaused by using the principal eigenvector of RUL

SUM instead of R DLSUM , we must find

a way to estimate the downlink correlation matrix with reasonable accuracy.


One such approach described in [26] uses the uplink correlation matrix to esti-mate the PAS as

( ) ( ) ( )PAS f f fULH

UL ULSUM

ULθ θ θ, , ,= a R a (5.81)

From the estimated PAS, the directions and amplitudes Pi of the mainpeaks, peaks within 10 dB from the maximum peak, can be determined andused in constructing the downlink spatial correlation matrix by transforming thearray response vectors from the uplink to the downlink frequency as

( ) ( )[ ]R a aDLSUM

i i DLH

i DLi

N

P f f==∑ θ θ, ,

1

(5.82)

where Pii

N

==∑ 1

1

(i.e., the amplitudes) are normalized.

5.4 Conclusion

Smart antenna technology uses sophisticated signal processing techniques tomanipulate signals at the base station or the mobile side and dynamically controltransmission and reception. Conventional radio systems indiscriminately broad-cast energy, creating interference for other users. Using adaptive beamforming,radio transmission and reception is optimized by selectively amplifying signalsto and from users of interest and rejecting unwanted and interfering signals.This substantially increases the signal quality and suppresses and mitigates inter-ference on both the uplink and downlink radio channels, resulting in increasedcoverage and spectral efficiency or system capacity. A comparison of the differ-ent spatial techniques discussed in this chapter is provided in Table 5.3. Manyapproaches are equally applicable to both the uplink and downlink, whereasothers are only designed for a specific link. Some approaches like TD performwell in large angular spread environments, which make them good candidatesfor microcells, whereas the performance of beamforming tends to degrade insuch environments. On the other hand, beamforming tends to perform better insmall angular spread situations such as in macrocells.



Table 5.3Comparison of Beamforming Approaches

SpatialTechnique Aspects Pros Cons

ReceiveDiversity

Requires low correla-tion and comparablesignal strength be-tweenantenna pair.

Provides diversity gainand fading reduction.

Spatial diversity requires largeantenna separation (10λ) but inlarge AS environments λ/2separation could still providesome decorrelation.

Open Loop TD Improves the mobilestation’s receiverperformance by provid-ing it with multipleindependent copies ofthe transmitted signal.

Does not require anyfeedback from the MS,so there is noadditional signalingoverhead.

Since channel is unknown to thetransmitter, diversity gain islow.

Provides the most gain for usersat cell edge (low geometry).

Closed LoopTD

Improves the mobilestation’s receiverperformance by provid-ing it with multipleindependent copies ofthe transmitted signal.

Provides higher gainsthan open loop TD tech-niques.

Requires feedback from the MS,so there is additional signalingoverhead, which limits perfor-mance at high speed.

BeamSteering

Simplest form ofbeamforming.

Main beam is directedtoward desired user.

Weight is the same assteering vector.

Provides gain of M forspatially whiteinterference environ-ments.

Requires direction of arrivalestimation of the desired user.

Suboptimal performance interms of SNR or SIR.

Null Steering Pattern nulls aredirected towardinterference sources.

Provides gain in SIR asa result of interferencereduction.

Requires direction of arrival esti-mation of all interfering sources.

Suboptimal performance interms of SNR or SIR.

OptimumBeamforming

The most adaptivescheme.

SNR and SINR areoptimized based onsome given criterion.

Optimal weight vectorand optimalperformance.

Computationally intensive andmay require adaptive algorithmsfor implementation.

Fixed MultipleBeams

Set of predefined fixedmultiple beams.

Performs as well asoptimum beamformingwhen interference isspatially white (whenall users have low/similar data rates).

Underperforms in systems withmultirate services where bothlow data rate speech serviceand high data rate applicationsare simultaneously supportedsince beams cannot track users.

References

[1] Saunders, S. R., Antennas and Propagation for Wireless Communication Systems, John Wiley& Sons, 1999.

[2] Perini, P. L., and C. L. Holloway, “Angle and Space Diversity Comparisons in DifferentMobile Radio Environments,” IEEE Trans. on Antennas and Propagation, Vol. 46, No. 6,June 1998, pp. 764–775.

[3] Jakes, W. C., Microwave Mobile Communications, New York: IEEE Press, 1974, p. 316.

[4] Jeng, S. S., et al., “Experimental Studies of Spatial Signature Variation at 900 MHz forSmart Antenna Systems,” IEEE Trans. on Antennas and Propagation, Vol. 46, No. 7, July1998, pp. 953–962.

[5] Godara, L. C., “Application of Antenna Arrays to Mobile Communications, Part II:Beamforming and Direction-of-Arrival Considerations,” Proc. of the IEEE, Vol. 85, No. 8,August 1997, pp. 1195–1245.

[6] Litva, J., and T. K. Lo, Digital Beamforming in Wireless Communications, Norwood, MA:Artech House, 1996.

[7] Ertel, R., Ph.D., Dissertation Thesis Submittal, “Antenna Array Systems: Propagation andPerformance,” Virginia Polytechnic Institute and State University, July 1999, pp. 16–21.

[8] Liberti, J. C., and T. S. Rappaport, Smart Antennas for Wireless Communications: IS95 andThird Generation CDMA Applications, Chapter 6, Upper Saddle River, NJ: Prentice Hall,1999.

[9] Burke, J. P., and J. R. Zeidler, “CDMA Reverse Link Spatial Combining Gains: Optimalvs. MRC in a Faded Voice-data System Having a Single Dominant High Data User,”IEEE Global Telecommunications Conference, Vol. 1, No., 2001, pp. 264–268.

[10] Lee, J. S., CDMA Systems Engineering Handbook, Norwood, MA: Artech House, Septem-ber 1998.

[11] Derryberry, R. T., et al., “Transmit Diversity in 3G CDMA Systems,” IEEE Communica-tions Magazine, April 2002, pp. 68–75.

[12] Parkvall, R. T. et al., “Transmit Diversity in WCDMA: Link and System Level Results,”IEEE 51st Vehicular Technology Conference Proceedings, 2000. VTC 2000–SpringTokyo. Volume 2, May 15–18 2000 pp. 864–868.

[13] TIA/EIA IS-2000 Physical Layer Specification for CDMA Spread Spectrum Communica-tions Systems, June 2000.

[14] TS 25.214 3GPP TSG RAN WG4, v.3.2.0, Physical Layer Procedures (FDD), 2000-03.

[15] Alamouti, S. M., “A Simple Transmit Diversity Technique for Wireless Communica-tions,” IEEE Trans. on Selected Areas in Communications, Vol. 16, October 1998, pp.1451–1458.

[16] Kogiantis, A., et al., Downlink Improvement Through Space-Time Spreading, 3GPP2-C30-19990817-014.

[17] Vucetic, B., and J. Yuan, Space-Time Coding, John Wiley & Sons, 2003.

[18] Harrison, M., Transmit Adaptive Array Description, New Results and Feedback Channel,3GPP2/TSG-C3 #1, 1999.

[19] Bae, K., J. Jiang, and W. H. Tranter, “Downlink WCDMA Performance Analysis withDiversity Techniques Combined with Beamforming,” IEEE


Wireless Communications and Networking, 2003, Vol. 1, March 16–20, 2003, pp.202–206.

[20] Buehrer, R. M., “The Impact of Anular Energy Distribution on Spatial Correlation,” Proc.of the Fall 2002 Vehicular Technology Conference, Vancouver, Canada, September 2002.

[21] Winters, J. H., J. Salz, and R. D. Gitlin, “The Impact of Antenna Diversity on the Capac-ity of Wireless Communications Systems,” IEEE Trans. on Communications, February1994, pp. 1740–1751.

[22] Lin, H. P., et al., “Experimental Studies of SDMA Schemes for Wireless Communica-tions,” 1995 International Conference on Acoustics, Speech, and Signal Processing, Vol. 3,1995, pp. 1760–1763.

[23] Xu, G., et al., “Experimental Studies of Space-Division-Multiple-Access Schemes for Spec-tral Efficient Wireless Communications,” IEEE International Conference on Communica-tions, Conference Record, Serving Humanity Through Communications, 1994, pp.800–804.

[24] Bigler, L., et al., “Experimental Direction of Arrival and Spatial Signature Measurementsat 900 MHz for Smart Antenna Systems,” IEEE 45th Vehicular Technology Conference,Vol. 1, 1995, pp. 55–58.

[25] Pedersen, K. I., P. E. Mogensen, and F. Frederiksen, “Joint-Directional Properties ofUplink and Downlink Channel in Mobile Communications,” Electronics Letters, Vol. 35,No. 16, August 5, 1999, pp. 1311–1312.

[26] Koutalous, A. C., and J. S. Thompson, “Effect of Frequency Division Duplex on OpenLoop Downlink Beamforming in WCDMA Systems,” Proceedings, 56tj VTC-2002, Vol.2, 24–28 Sept. 2002, pp. 686–690.


Abu-Dayya, A., and N. C. Beaulieu, “Outage Probability of Diverse Cellular Systems withCo-Channel Interference in Nakagami Fading,” IEEE Trans. Veh. Technol., Vol. 41, 1992, pp.343–355.

Adachi, F., et al., “Cross Correlation Between the Envelopes of 900 MHz Signals Received at aMobile Radio Base Station Site,” Proc. Inst. Elect. Eng., Vol. 133, Pt. F, 1986, pp. 506–512.

Agee, B. G., “Blind Separation and Capture of Communication Signals Using a MultitargetConstant Modulus Beamformer,” Proc. IEEE MILCOM, Bedford, MA, 1989, pp. 340–346.

Agee, B. G., S. V. Schell, and W. A. Gardner, “Spectral Self Coherence Restoral: A NewApproach to Blind Adaptive Signal Extraction Using Antenna Arrays,” Proc. IEEE, Vol. 78,1990, pp. 753–767.

Alamouti, S. M., “A Simple Transmit Diversity Technique for Wireless Communications,” IEEEJournal on Select Areas in Communications, Vol. 16, No. 8, October 1998, pp. 1451–1458.

Alouini, M-S., and A. J. Goldsmith, “Capacity of Rayleigh Fading Channels Under DifferentAdaptive Transmission and Diversity-Combining Techniques,” IEEE Trans. on Vehicular Tech-nology, Vol. 48, No. 4, July 1999, pp. 1165–1181.

Anderson, S., et al., “An Adaptive Array for Mobile Communication Systems,” IEEE Trans. Veh.Technol., Vol. 40, 1991, pp. 230–236.


Aste, T., et al., “Downlink Beamforming Avoiding DOA Estimation for Cellular Mobile Com-munications,” ICASSP, 1998, pp. 3313–3316.

Astely, D., and B. Ottersten, “The Effects of Local Scattering on Direction of Arrival Estimationwith MUSIC,” IEEE Trans. on Signal Processing, Vol. 47, No. 12, December 1999, pp.3220–3234.

Balaban, P., and J. Salz, “Dual Diversity Combining and Equalization in Digital Cellular MobileRadio,” IEEE Trans. Veh. Technol., Vol. 40, 1991, pp. 342–354.

Balaban, P., and J. Salz, “Optimum Diversity Combining and Equalization in Digital DataTransmission with Application to Cellular Mobile Radio—Part I: Theoretical Considerations,”IEEE Trans. Commun., Vol. 40, 1992, pp. 885–894.

Bäro, S., G. Bauch, and A. Hansmann, “Improved Codes for Space-Time Trellis-Coded Modula-tion,” IEEE Commun. Lett., Vol. 4, January 2000, pp. 20–22.

Beach, M. A., et al., “Adaptive Antennas for Third Generation Systems,” Proc. Inst. Elect. Eng.Colloquium Mobile Communications Toward Year 2000, London, England, 1994, pp. 10/1–10/6.

Bengtsson, M., and B. Ottersten, “Uplink and Downlink Beamforming for Fading Channels,”1999, pp. 350–353.

Brennan, D. G., “Linear Diversity Combining Techniques,” Proc. IRE, Vol. 47, 1959, pp.1075–1102.

Bresler, Y., V. U. Reddy, and T. Kailath, “Optimum Beamforming for Coherent Signal andInterferences,” IEEE Trans. Acoust., Speech, Signal Processing, Vol. 36, 1988, pp. 833–843.

Buehrer, R. M., R. A. Soni, and Q. Li, “Transmit Diversity with More Than Two Antennas,”Proceedings of the 10th Virginia Tech Symposium Wireless Personal Communication, 2000, pp.153–165.

Cardoso, J. F., and A. Souloumiac, “Blind Beamforming for Non-Gaussian Signals,” Inst. Elect.Eng. Proc.—F, Vol. 140, 1993, pp. 362–370.

Chang, L. F., and P. J. Porter, “Performance Comparison of Antenna Diversity and Slow Fre-quency Hopping for the TDMA Portable Radio Channel,” IEEE Trans. Veh. Technol., Vol. 38,1989, pp. 222–229.

Chiba, T. T., and Y. Karasawa, “Transmitting Null Beamforming with Beam Space AdaptiveArray Antennas,” Proc. IEEE 44th Vehicular Technology Conf., Stockholm, Sweden, 1994, pp.1498–1502.

Choi, S., and T. K. Sarkar, “Adaptive Antenna Array Utilizing the Conjugate Gradient Methodfor Multipath Mobile Communication,” Signal Process., Vol. 29, 1992, pp. 319–333.

Choi, H., Y. Kim, and J. Park, Use of More Than 2 Tx Antennas for Closed Loop Tx Diversity Sys-tems, Samsung Electronics.

Chujo, W., and K. Yasukawa, “Design Study of Digital Beam-Forming Antenna Applicable toMobile Satellite Communications,” IEEE Antennas and Propagation Symp. Dig., Dallas, TX,1990, pp. 400–403.

Compton, R. T., Jr., Adaptive Antennas: Concepts and Performances, Englewood Cliffs, NJ:Prentice-Hall, 1988.

Cox, H., R. M. Zeskind, and M. M. Owen, “Robust Adaptive Beamforming,” IEEE Trans. onASSP, Vol. 35, October 1987, pp. 1365–1375.


Eggers, P. C. F., “Angular Dispersive Mobile Radio Environments Sensed by Highly DirectiveBase Station Antennas,” Proc. IEEE Int. Symp. Personal, Indoor Mobile Radio Communications,Toronto, Canada, 1995, pp. 522–526.

Farsakh, C., and J. A. Nossek, “Channel Allocation and Downlink Beamforming in an SDMAMobile Radio System,” Proc. IEEE Int. Symp. Personal, Indoor Mobile Radio Communications,Toronto, Canada, 1995, pp. 687–691.

Farsakh, C., and J. A. Nossek, “Spatial Covariance Based Downlink Beamforming in an SDMAMobile Radio System,” IEEE Trans. on Communications, Vol. 46, No. 11, November 1998, pp.1497–1506.

Fernandez, J., I. R. Corden, and M. Barrett, “Adaptive Array Algorithms for Optimal Combin-ing in Digital Mobile Communication Systems,” Proc. Inst. Elect. Eng. Int. Conf. Antennas Propa-gation, Edinburgh, Scotland, 1993, pp. 983–986.

Forschini, G. J., “Layered Space-Time Architecture for Wireless Communication in a FadingEnvironment When Using Multi-Element Antennas,” Bell Labs Technical Journal, Autumn1996, pp. 41–59.

Forschini, G. J., and M. J. Gans, “On Limits of Wireless Communications in a Fading Environ-ment When Using Multiple Antennas,” Wireless Personal Communications, No. 6, 1998, pp.311–335.

Friedlander, B., and B. Porat, “Performance Analysis of a Nullsteering Algorithm Based onDirection-of-Arrival Estimation,” IEEE Trans. Acoust., Speech, Signal Processing, Vol. 37, 1989,pp. 461–466.

Gardner, W. A., “Exploitation of Spectral Redundancy in Cyclostationary Signals,” IEEE SignalProcessing Mag., Vol. 8, 1991, pp.14–37.

Gerlach, D., “Transmit Antenna Beamforming for the Advanced Mobile Phone System,”Asilomar, 1996, pp. 1162–1166.

Gerlach, D., and A. Paulraj, “Adaptive Transmitting Antenna Arrays with Feedback,” IEEE Sig-nal Processing Letters, Vol. 1, No. 10, October 1994, pp. 150–152.

Gerschman, A. B., U. Nickel, and J. F. Bohme, “Adaptive Beamforming Algorithms withRobustness Against Jammer Motion,” IEEE Trans. on Signal Processing, Vol. 45., No. 7, July1997, pp. 1878–1885.

Goldberg, J., and J. R. Fonollosa, “Downlink Beamforming for Cellular Mobile Communica-tions,” Vehicular Technology Conference, 1997, pp. 632–636.

Golden, G. D., G. J. Forschini, and R. A. Valenzuela, “Detection Algorithm and Initial Labora-tory Results Using V-BLAST Space-Time Communication Architecture,” Electronic Letters, Vol.35, No. 1, January 1999, pp. 14–16.

Goldsmith, J., and P. P. Varaiya, “Capacity of Fading Channels with Channel Side Informa-tion,” Vol. 43, No. 6, November 1997, pp. 1986–1992.

Griffiths, J., “A Simple Adaptive Algorithm for Real-Time Processing in Antenna Arrays,” Proc.IEEE, Vol. 57, 1969, pp. 1696–1704.

Harrison, M., and K. Kuchi, “Open and Closed Loop Transmit Diversity at High Data Rates on2 and 4 Elements,” 3GPP2-C30-19990817-017, Portland, OR, 1999.

Hassell Sweatman, C. Z. W., et al., “A Comparison of Detection Algorithms Including BLASTfor Wireless Communication Using Multiple Antennas,” IEEE, 2000, pp. 698–703.

Haykin, S., (ed.), Array Signal Processing, Englewood Cliffs, NJ: Prentice-Hall, 1985.


Heath, R. W., Jr., and A. Paulraj, “A Simple Scheme for Transmit Diversity Using Partial Chan-nel Feedback,” Asilomar Conference, 1998, pp. 1073–1078.

Henry, P. S., and B. S. Glance, “A New Approach to High Capacity Digital Mobile Radio,” BellSyst. Tech. J., Vol. 60, 1981, pp. 1891–1904.

Hottinen, A., and R. Wichman, Transmit Diversity Using Filtered Feedback Weights in theFDD/WCDMA System, Proceedings of the 2000 Internationsl Zurich Seminar on BoradbandCommunications, Zurich, Feb. 2000: Broadband Communications, 2000, pp. 15–21.

Huang, H., H. Viswanathan, and G. J. Forschini, “Achieving High Data Rates in CDMA Sys-tems Using BLAST Techniques,” Globecom, 1999, pp. 2316–2320.

Hugl, K., J. Laurila, and E. Bonek, “Downlink Beamforming for Frequency Division DuplexSystems,” Advanced Signal Processing for Communications, Globecom, 1999, pp. 2097–2101.

Hugl, K., J. Laurila, and E. Bonek, “Downlink Performance of Adaptive Antennas with NullBroadening,” 1999, pp. 872–876.

Jalloul, L. M. A., et al., “Performance Analysis of CDMA Transmit Diversity Methods,” Vehicu-lar Technology Conference, 1999, pp. 1326–1330.

Jones, M. A., and M. A. Wickert, “Direct Sequence Spread Spectrum Using Directionally Con-strained Adaptive Beamforming to Null Interference,” IEEE J. Select. Areas Commun., Vol. 13,1995, pp. 71–79.

Katz, M., and J. Ylitalo, “Extension of Space-Time Coding Beamforming WCDMA Base Sta-tions,” Vehicular Technology Conference, 2000, pp. 1230–1234.

Kawala, P., and U. H. Sheikh, “Adaptive Multiple-Beam Array for Wireless Communications,”Proc. Inst. Elect. Eng. 8th Int. Conf. Antennas and Propagation, Edinburgh, Scotland, 1993, pp.970–974.

Kim, J., and J. M. Cioffi, “Spatial Multiuser Access with Antenna Diversity Using Singular ValueDecomposition,” IEE International Conference on Communications, ICC 2000, Vol. 3, June18-22, 2000, pp. 1253–1257.

Kim, B., and H. S. Lee, “Transmit Diversity System with Linear Prefilters Exhibiting Near-Opti-mal Performance,” Electronic Letters, Vol. 36, No. 18, August 2000.

Krim, H., and M. Viberg, “Two Decades of Array Signal Processing: The Parametric Approach,”IEEE Signal Processing Mag., July 1996, pp. 67–94.

Lee, W. C. Y., “Effects of Correlation Between Two Mobile Radio Base Station Antennas,” IEEETrans. Commun., Vol. COM-21, 1973, pp. 1214–1224.

Lehnert, J. S., and M. B. Pursely, “Multipath Diversity Reception of Spread Spectrum MultipleAccess Communications,” IEEE Trans. Commun. Vol. COM-35, 1987, pp. 1189–1198.

Lopez, R., “Performance Predictions for Cellular Switched Beam Intelligent Antenna System,”IEEE Commun. Mag., Vol. 34, October 1996, pp. 152–154.

Loyka, S., and J. Mosig, “Channel Capacity of n-Antenna BLAST Architecture,” Electronics Let-ters, Vol. 36, No. 7, March 2000, pp. 660–661.

Marr, J. D., “A Selected Bibliography on Adaptive Antenna Arrays,” IEEE Trans. Aerosp. Elec-tron. Syst., Vol. AES-22, 1986, pp. 781–788.

Monzingo, R. A., and T. W. Miller, Introduction to Adaptive Arrays, New York: John Wiley &Sons, 1980.


Naguib, A., et al., “A Space-Time Coding Modem for High Data Rate Wireless Communica-tions,” IEEE Journal on Selected Areas in Communications, Vol. 16, No. 8, October 1998, pp.1459–1478.

Naguib, A. F., A. Paulraj, and T. Kailath, “Capacity Improvement with Base-Station AntennaArrays in Cellular CDMA,” IEEE Trans. Veh. Technol., Vol. 43, 1994, pp. 691–698.

Naguib, F., and A. Paulraj, “Performance of CDMA Cellular Networks with Base-StationAntenna Arrays,” Proc. IEEE Int. Zurich Seminar on Communications, 1994, pp. 87–100.

Narula, A., et al., “Efficient Use of Side Information in Multiple Antenna Data TransmissionOver Fading Channels,” IEEE Journal on Selected Areas in Communications, Vol. 16, No. 8,October 1998, pp. 1423–1436.

Narula, A., M. D. Trott, and G. Wornell, “Performance Limits of Coded Diversity Methods forTransmitter Antenna Arrays,” IEEE Trans. on Information Theory, Vol. 45, No. 7, November1999, pp. 2418–2433.

Nilsson, M., B. Volcker, and B. Ottersten, “A Cluster Approach to Spatio-Temporal ChannelEstimation,” IEE International Conference on Acoustics, Speech, and Signal Processing,ICASSP 2000, Vol. 5, June 5-9, 2000, pp. 2757–2760.

Passerini, C., et al., “Adaptive Antenna Arrays for Reducing the Delay Spread in Indoor RadioChannels,” Electron. Lett., Vol. 32, 1996, pp. 280–281.

Raghothaman, B., et al., “Performance of Simple Space Time Block Codes for More than TwoAntennas,” Proc. Allerton Conf. Commun., Control and Comp., October 2000.

Raghothaman, B., R. T. Derryberry, and G. Mandyam, “Transmit Adaptive Array WithoutUser-Specific Pilot for 3G CDMA,” Proc. ICASSP, Istanbul, Turkey, 2000.

Raghothaman, B., G. Mandyam, and R.T. Derryberry, “Performance of Closed Loop TransmitDiversity with Feedback Delay,” Proc. Asilomar Conf. Sig., Sys. Comp., 2000.

Rajan, D., E. Erkip, and B. Aazhang, “New Spread Spectrum Techniques for Multiple AntennaTransmit Diversity,” Globecom, 1999, pp. 2321–2325.

Rajan, D., and S. D. Gray, “Transmit Diversity Schemes for CDMA 2000,”IEEE Wireless Com-munications and Networking Conference, 1999, pp. 669–673.

Raleigh, G., et al., “A Blind Adaptive Transmit Antenna Algorithm for Wireless Communica-tions,” Proc. IEEE ICC, Seattle, WA, 1995, pp. 1494–1499.

Rayleigh, G. G., and J. M. Cioffi, “Spatio-Temporal Coding for Wireless Communication,”IEEE Trans. on Communications, Vol. 46, No. 3, March 1998, pp. 357–366.

Riba, J., J. Goldberg, and G. Vasquez, “Robust Beamforming for Interference Rejection inMobile Communication,” IEEE Trans. on Signal Processing, Vol. 45, No. 1, January 1997, pp.271–275.

Salz, J., and J. H. Winters, “Effect of Fading Correlation on Adaptive Arrays in Digital MobileRadio,” IEEE Trans. Veh. Technol., Vol. 43, 1994, pp. 1049–1057.

“Special Issue on Adaptive Processing Antenna Systems,” IEEE Trans. Antennas Propagat., Vol.AP-34, March 1986.

“Special Issue on Adaptive Systems and Applications,” IEEE Trans. Circuits Syst., Vol. CAS-34,July 1987.

“Special Issue on Beamforming,” IEEE J. Oceanic Eng., Vol. OE-10, July 1985.


Suard, B., et al., “Performance of CDMA Mobile Communication Systems Using AntennaArrays,” IEEE Int. Conf. Acoustics, Speech, and Signal Processing (ICASSP), Minneapolis, MN,1993, pp. 153–156.

Swales, S. C., et al., “The Performance Enhancement of Multibeam Adaptive Basestation Anten-nas for Cellular Land Mobile Radio Systems,” IEEE Trans. Veh. Technol., Vol. 39, 1990, pp.56–67.

Swales, S. C., et al., “The Realization of a Multibeam Adaptive Base-Station Antenna for CellularLand Mobile Radio Systems,” Proc. IEEE Veh. Technol. Conf., San Francisco, CA, 1989, pp.341–348.

Swindlehurst, and J. Yang, “Using Least Squares to Improve Blind Signal Copy Performance,”IEEE Signal Processing Lett., Vol. 1, 1994, pp. 80–82.

Tarokh, V., et al., “Space-Time Codes for High Data Rate Wireless Communication: Perfor-mance Criterion in the Presence of Channel Estimation Errors, Mobility, and Multiple Paths,”IEEE Trans. on Communications, Vol. 47, No. 2, February 1999, pp. 199–207.

Tarokh, V., N. Seshadri, and A. R. Calderbank, “Space-Time Codes for High Data Rate WirelessCommunication: Performance Criterion and Code Construction,” IEEE Trans. on InformationTheory, Vol. 44, No. 2, March 1998, pp. 744–765.

Tarokh, V., S. M. Alamouti, and P. Poon, “New Detection Schemes for Transmit Diversity withNo Channel Estimation,” IEEE International Conference on Universal Personal Communica-tions, 1998, ICUPA, ’98 Vol. 2, 1998, pp. 917–920.

Tarokh, V., and H. Jafarkhani, “A Differential Detection Scheme for Transmit Diversity,” IEEEJournal on Select Areas in Communications, Vol. 18, No. 7, July 2000, pp. 1169–1174.

Tarokh, V., H. Jafarkhani, and A. R. Calderbank, “Space-Time Block Codes from OrthogonalDesigns,” IEEE Trans. Info. Theory, Vol. 45, July 1999, pp. 1456–1467.

Telatar, E. I., Capacity of Multi-Antenna Gaussian Channels, AT&T Bell Labs, technical report,June 1995.

Thompson, J. S., P. M. Grant, and B. Mulgrew, “Downlink Transmit Diversity Schemes forCDMA Networks,” Vehicular Technology Conference, 1999, pp. 1382–1386.

Tong, L., Y. Inouye, and R. W. Liu, “Waveform Preserving Blind Estimation of Multiple Inde-pendent Sources,” IEEE Trans. Signal Processing, Vol. 41, 1993, pp. 2461–2470.

Turin, G., “The Effect of Multipath and Fading on the Performance of Direct Sequence CDMASystems,” IEEE J. Select. Areas Commun., Vol. 2, 1984, pp. 597–603.

Turin, G. L., “Introduction to Spread-Spectrum Antimultipath Techniques and Their Applica-tion to Urban Digital Radio,” Proc. IEEE, Vol. 68, 1980, pp. 328–353.

Van Veen, B., and K. Buckley, “Beamforming: A Versatile Approach to Spatial Filtering,” IEEEASSP Magazine, April 1988, pp. 4–24.

Vaughan, R. G., and N. L. Scott, “Closely Spaced Monopoles for Mobile Communications,”Radio Sci., Vol. 28, 1993, pp. 1259–1266.

Vaughan, R. G., “On Optimum Combining at the Mobile,” IEEE Trans. Veh. Technol., Vol. 37,1988, pp. 181–188.

Van Veen, B. D., and K. M. Buckley, “Beamforming: A Versatile Approach to Spatial Filtering,”IEEE Aerosp. Electron. Syst. Mag., Vol. 5, 1988, pp. 4–24.

Veen, J. V. D., S. Talwar, and A. Paulraj, “Blind Estimation of Multiple Digital Signals Trans-mitted Over FIR Channels,” IEEE Signal Processing Lett., Vol. 2, 1995, pp. 99–102.


Wang, Y., and J. R. Cruz, “Adaptive Antenna Arrays for Cellular CDMA Cellular Communica-tion Systems,” IEEE ICASSP, Detroit, MI, 1995, pp. 1725–1728.

Wang, Y., and J. R. Cruz, “Adaptive Antenna Arrays for the Reverse Link of CDMA CellularCommunication Systems,” Inst. Elect. Eng. Electron. Lett., Vol. 30, 1994, pp. 1017–1018.

Weerackody, V., “Diversity for the Direct-Sequence Spread Spectrum System Using MultipleTransmit Antennas,” IEEE, 1993, pp. 1775–1779.

Widrow, B., and S. D. Stearns, Adaptive Signal Processing, Englewood Cliffs, NJ: Prentice-Hall,1985.

Winters, H., “Optimum Combining in Digital Mobile Radio with Cochannel Interference,”IEEE J. Select. Areas Commun., Vol. SAC-2, 1984, pp. 528–539.

Winters, H., “Optimum Combining for Indoor Radio Systems with Multiple Users,” IEEETrans. Commun., Vol. COM-35, 1987, pp. 1222–1230.

Winters, J. H., “On the Capacity of Radio Communication Systems with Diversity in RayleighFading Environment,” IEEE J. Select. Areas Commun., Vol. SAC-5, 1987, pp. 871–878.

Winters, J. H., “The Diversity Gain of Transmit Diversity in Wireless Systems with RayleighFading,” IEEE Trans. on Vehicular Technology, Vol. 47, No. 1, February 1998, pp. 119–123.

Winters, J. H., “Signal Acquisition and Tracking with Adaptive Arrays in the Digital MobileRadio System IS-54 with Flat Fading,” IEEE Trans. Veh. Technol., Vol. 42, 1993, pp. 377–384.

Winters, J. H., J. Salz, and R. D Gitlin, “The Impact of Antenna Diversity on the Capacity ofWireless Communication Systems,” IEEE Trans. Commun., Vol. 42, 1994, pp. 1740–1751.

Wolniansky, P. M., et al., “V-BLAST: An Architecture for Realizing Very High Data Rates overthe Rich-Scattering Wireless Channel,” IEEE, 1998, pp. 295–300.

Wornell, G., and M. D. Trott, “Efficient Signal Processing Techniques for Exploiting TransmitAntenna Diversity on Fading Channels,” IEEE Trans. on Signal Processing, Vol. 45, No. 1, Janu-ary 1997, pp. 191–205.

Wu, Q., and K. M. Wong, “Blind Adaptive Beamforming for Cyclostationary Signals,” IEEETrans. Signal Processing, Vol. 44, 1996, pp. 2757–2767.

Wu, Q., K. M. Wong, and R. Ho, “Fast Algorithm for Adaptive Beamforming of Cyclic Sig-nals,” Inst. Elect. Eng. Proc.—Radar, Sonar Navigation, Vol. 141, 1994, pp. 312–318.

Xu, G., and H. Liu, “An Effective Transmission Beamforming Scheme for Frequency DivisionDuplex Digital Wireless Communication Systems,” Proc. IEEE ICASSP, Detroit, MI, 1995, pp.1729–1732.

Yamada, Y., K. Kagoshima, and K. Tsunekawa, “Diversity Antennas for Base and Mobile Situa-tion in Land Mobile Communication Systems,” IEICE Trans., Vol. E74, 1991, pp. 3202–3209.

Zetterberg, P., “A Comparison of Two Systems for Downlink Communication with Base StationAntenna Arrays,” IEEE Trans. on Vehicular Technology, , Vol. 48, No. 5, September 1999, pp.1356–1370.

Zetterberg, P., and B. Ottersten, “The Spectrum Efficiency of a Base Station Antenna Array Sys-tem for Spatially Selective Transmission,” IEEE Trans. on Vehicular Technology, Vol. 44, No. 3,August 1995, pp. 651–660.

Zoltowski, M. D., “On the Performance Analysis of the MVDR Beamformer in the Presence ofCorrelated Interference,” IEEE Trans. Acoust., Speech, Signal Processing, Vol. 36, 1988, pp.945–947.


6Smart Antenna Receivers and Algorithmsfor Radio Base Stations

In Chapter 5, several adaptive beamforming criteria were discussed and a generalform for the optimum array weight vector was derived that would require largeamounts of computational load. Hence, it is necessary to find techniques thatcan find this optimum solution in real time, adapting to the time variant chan-nel while keeping the computational load to a reasonable level. A classificationof such adaptive array algorithms is shown in Figure 6.1. A brief description ofthese algorithms can be found in [1]. In the remainder of this chapter we willdescribe some of these algorithms in detail and investigate issues related to theirperformance.

6.1 Reference Signal Methods

6.1.1 The Least Mean Square Algorithm

As we can see from Figure 6.1, the least mean square (LMS) algorithm belongsto the trained algorithms category in which a reference signal is used to updatethe weights at each iteration as follows

( ) ( ) ( )w wn n MSEw+ = − ∇1 µ (6.1)

where ∇w is the gradient of the MSE, which is the mean square error between thereference signal r(n) and the array output given by

159

( )( ) ( ) ( ) ( )[ ]( )[ ] ( ) ( ) ( ) ( ) ( )[ ]

MSE n E n n n

E n n n n E n n

H

H H

w r w x

r w Rw w x r

= + − +

= + + − + +

1 1

1 2 1 1

2

2

(6.2)

In the LMS algorithm, we are searching for the optimal weight that wouldmake the array output either equal or as close as possible to the reference signal,which is the weight that minimizes the MSE. Since the MSE has a quadraticform, moving the weights in the negative direction of the gradient of the MSEshould lead us to the minimum of the error surface. The gradient can be calcu-lated as

( )( ) ( ) ( ) ( )[ ]( ) ( ) ( ) ( ) ( )

∇ = − + +

= + + − + +=

w

H

MSE n n E n n

n n n n n

n

w Rw x r

x x w x r

x

2 2 1 1

2 1 1 2 1 1

2 ( )+ ∗1 ε

(6.3)

where ε is the error given by

( ) ( ) ( )ε = + − +w x rH n n n1 1 (6.4)


Trained Blind

Decisiondirected

Constantmodulus

Spectralcoherence

Least meansquareLMS

Recursive leastsquaresRLS

Propertyrestoral

Conjugategradient

Leastsquares

Lagrangemultiplier

Adapative arrayalgorithms

Figure 6.1 Classification of adaptive array algorithms.

( ) ( ) ( )w w xn n n+ = − + ∗1 1µ ε (6.5)

The constant µ, also called the step size, determines how close the weightsapproach the optimum value after each iteration and it controls the convergencespeed of the algorithm. Typical values for the step size are 0 < µ < Trace(R).

6.1.2 The Recursive Least Squares Algorithm

One of the drawbacks of the LMS algorithm is its slow convergence speed undercertain conditions, for example when the eigenvalue spread of R is large. Thisleads to the development of the recursive least squares (RLS) algorithm, whichreplaces the step size µ with the inverse of R. The algorithm is initialized by firstsetting

( )R I− = >1 01

0δ

δ, (6.6)

The weights are then updated using

( ) ( ) ( ) ( )w w Rn n n x n+ = − + +− ∗1 1 11 ε (6.7)

where the update of the inverse of the correlation matrix is given by

( ) ( ) ( ) ( ) ( ) ( )( ) ( ) ( )R R

R x x R

x R x− −

− −

−+ = −+ +

+ + +

1 11 1

11

1 1 1

1 1n n

n n n n

n n n

H

Hξ ξ (6.8)

6.1.3 Blind Adaptive Beamforming

Self-adaptive or blind beamforming algorithms have been gaining a lot of atten-tion, especially with the introduction of digital technology in wireless cellularand PCS systems. The term blind refers to the fact that no array calibration isnecessary (i.e., the array manifold knowledge is not required). Some blind algo-rithms can be classified as property restoral techniques; that is, they rely on refer-ence signals that satisfy known properties of the desired signal, whereasothers, like the least squares, are based on using the temporal characteristics ofthe desired digital signals to determine the array response and transmittedsequence [1].

6.1.4 Least Squares

Using the standard array model, we can write the received signal at the arrayoutput as

Smart Antenna Receivers and Algorithms for Radio Base Stations 161

( ) ( ) ( )X = AS Nt t t+ (6.9)

where

( ) ( ) ( ) ( )[ ]X t x t x t x tM

T= 1 2 L

( ) ( ) ( ) ( )[ ]S t s t s t s tM

T= 1 2 L is the signal vector

( ) ( ) ( ) ( )[ ]N t n t n t n tM

T= 1 2 L is the noise vector.

The LS algorithm minimizes the ML criterion

min S ∈ −Ω X AS (6.10)

A flowchart of the LS algorithm is shown in Figure 6.2.

6.1.5 Constant Modulus Algorithm

Some modulation techniques used in modern cellular communications systemssuch as frequency modulation (FM), phase-shift keying (PSK), frequency-shiftkeying (FSK), or quadrature amplitude modulation (QAM) produce signalswith constant or low modulus variation. Assuming that the transmitted signalshave a constant envelope, the array output should also have a constant envelope.However, due to multipath fading effects, the array output will not have a con-stant envelope. The constant modulus algorithm (CMA) can therefore be usedto restore the array output to a constant envelope signal on average. A cost func-tion, which measures the signal modulus variation, is minimized to adjust thearray weights. Although a simple search algorithm such as the steepest-descentmethod can easily implement the CMA, a major drawback is that the conver-gence is not guaranteed because the cost function is nonconvex and may havefalse minima. Because CDMA systems use power control, that is, every user’ssignal power is adjusted to meet the quality of service criteria, this algorithm isnot suitable for CDMA signals.

6.1.6 Decision-Directed Algorithm

In this algorithm, a reference signal is generated based on the outputs of athreshold decision device. The beamformer output y(n) is demodulated toobtain the signal q(n). The decision device then compares q(n) to the knownalphabet of the transmitted data sequence and makes a decision in favor of theclosest value to q(n) denoted by r(n). The reference signal is obtained by modu-lating r(n), then the cost function for the beamformer is established. The deci-sion-directed algorithm convergence depends on the ability of the receiver to


lock onto the desired signal. Since it may not always be able to do that, theconvergence is not guaranteed.

6.1.7 Cyclostationary Algorithms

Some commonly used communications signals such as amplitude modulation(AM), binary phase-shift keying (BPSK), binary frequency-shift keying (BFSK),


)()()(

)(nn

nn H

H

SSXS

A =

Set initialrandom guess

H0A

)1()(

−=

nn

H XAS

Has symbolsequenceconverged

Project elements of ( )S nonto the nearest value inthe alphabet

ENDYESNO

)1( −nHA )1( −nA

Figure 6.2 Flowchart of the least squares method.

and Gaussian minimum-shift keying (GMSK) possess some unique nontrivialcycle frequency. A signal is said to be cyclostationary if its cyclic autocorrelationR(τ, α) is nonzero at some time delay τ or at some frequency shift α, that is,

( ) ( ) ( ) R E s n s n e j kτ α τ πα, = +∗ − 2 (6.11)

Therefore, based on its cycle frequency, a particular signal can be extractedfrom a mixture of signals in time or frequency domain. The reference signal isgenerated based on the knowledge of both the time delay τ and the cycle fre-quency α of the wanted signal. The least squares spectral self-coherence restoral(SCORE) algorithm is one of the beamforming techniques that makes use ofthis cyclostationarity property. The minimization leads to a solution for theweights vector that is similar to the Wiener solution. Another class of blind algo-rithms takes the approach of iteratively solving for the optimum beamformingweight based on maximizing the SINR or SNR [1–12]. For convenience, let usrewrite the expression for the SNR at the beamformer output as:

SNRH

sH

=w R w

w wσ n2

(6.12)

where σ N2 is the noise variance. The maximization leads to the eigenvalue

problem

R w ws = λ

where the optimal weight is given by the eigenvector corresponding to the maxi-mum eigenvalue. However, this solution requires large amounts of computa-tions and is often impractical. One technique that can be used to solve this eigendecomposition iteratively, the conjugate gradient (CG) algorithm, is describednext.

6.1.8 Conjugate Gradient Algorithm

In some CDMA systems, the interference can become spatially white (e.g., in2G CDMA (IS-95A) based systems where all users have low data rates and nointerference spatial coloring is experienced). This would also be the case in 3GCDMA systems (based on either IS-2000 or WCDMA) when only low data rateusers are present in the sector. In that case, the correlation matrix can take theform

R ASA I= +Hnσ 2 (6.13)


Let the eigenvectors and eigenvalues of R be denoted by e1 e2 … eM andλ1 λ2 … λM, respectively. Assuming we have L signals received at the array, itthen follows that the set of eigenvectors e1 e2 … eL spans the signal subspace,whereas the set of eigenvectors eL+1 e2 … eM spans the noise subspace. Note thatsince the two subspaces are orthogonal, each signal eigenvector can be written asa linear combination of the signal direction vectors. We can then write theeigenvector corresponding to the maximum eigenvalue as

( )e a1 11

==∑ ζ θll

L

(6.14)

where ζl is a constant that depends on the magnitude and angle distribution ofevery signal component. When the optimal weight is set to e1, which is a func-tion of all the signals incident on the array, the resulting beam pattern producesmain lobes in the directions of θ1 θ2 … θL, that is, in the direction of thedesired and interfering signals. The adaptive algorithm introduced in [2, 9] tosolve the eigenvalue problem makes use of a unique CDMA characteristic,namely, the fact that the desired signal power becomes much larger than that ofthe interference after the correlation of the desired and interfering signals withthe chip sequence corresponding to the desired user. More specifically, thepower of the desired signal becomes PG times greater than that of the interfer-ence where PG is the processing gain, which could be quite high for low datarate users. We can then use the approximation

( )e1 1 1≈ ζ θa (6.15)

This would lead to a main lobe in the direction of the desired user withmuch smaller gain in the direction of the interfering signals. To estimate thecorrelation matrix, the CG algorithm uses the approximation

( ) ( ) ( )$R x xnN

n nH

n

N

==∑1

0

(6.16)

In the CG algorithm [2] subsequent updates to R are computed as

( ) ( ) ( ) ( )R R x xn f n n nH= − +1 (6.17)

where f is called a forgetting factor. The weights are updated using

( ) ( ) ( ) ( )w w ui i z i i+ = +1 (6.18)


where the adaptive gain or step size z(i) and the search direction vector u(i) aredetermined by maximizing the Raleigh quotient

( )( ) ( ) ( ) ( )( ) ( )

J ii R n i

i i

H

Hw

w w

w w= (6.19)

This maximization leads to [9]

( )z iB B AC

A=− − −2 4

2(6.20)

and

( ) ( ) ( )( )

( )u rr

rui i

i

ii+ = + −

+1 1

12

2(6.21)

where

( ) ( ) ( ) ( )( ) ( ) ( ) ( )( ) ( ) ( )( ) ( ) ( )( ) ( )[ ] ( )

a i i n i

b i i n i

c i i i

d i i i

A b i c i d i a

H

H

H

H

====

= −

w R u

u R u

w u

u u

Re Re ( )[ ]( ) ( ) ( ) ( )

( ) ( ) ( )( )[ ] ( ) ( )

i

i i n i

B b i i d i

C a i i d i

Hλ

λ

λ

== −= −

w R w

Re

( ) ( ) ( ) ( ) ( )r w R wi i i n i+ = + + − +1 1 1 1λ is the instantaneous error. Theprocedure is started by setting the following initial conditions:

( ) ( )( )

wx

x0

1 0

0=

M, ( ) ( ) ( ) ( )λ 0 0 0 0= w R wH , ( ) ( ) ( ) ( )u w R0 0 0 0= −λ ,

( )w 0 , and ( ) ( )r u0 0= . At the end of each iteration, the weight vector is normal-

ized as follows ww

w=

2. Although this algorithm eliminates the need for the

direction solution of the eigenvalue problem, it still requires lots of matrix


multiplications with a computational load of O(3N2 + 12N), where O (N)denotes the order of computational load required for a scalar product of two N

1 complex-valued vectors. This load can be significantly reduced if the forget-ting factor f is set to zero. It then follows that

( ) ( ) ( )R x xi i iH= (6.22)

and

( ) ( ) ( ) ( ) ( ) ( )( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )( ) ( ) ( ) ( )

λ i i i i i i

a i i i i i i i i

b i i i i

H H

H H H

H H

= =

= =

=

w x x w y

w x x u y x u

u x x u

2

( ) ( ) ( )( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( )

i i i

i i i i i i i i i i

H

H

=

+ = + + − + = + + − ∗

u x

r w x x w w x y

2

1 1 1 1 1 1λ λ

The computational load is thus reduced to O(11N). This reduction incomplexity is mainly because the correlation matrix is computed based on theinstantaneous received signal rather than averaging the matrix over a period oftime using previous instances of R, as shown in (6.17). This added simplicitycomes at the expense of less accurate estimates of the correlation matrix and itseigenvalues and eigenvectors resulting in less accurate weights. In practice, onehas to consider this trade-off between the need for faster convergence along withreduced computational complexity and therefore implementation costs versusaccuracy of the optimal weights and any possible performance degradationresulting from using less accurate weights.

6.1.9 Lagrange Multiplier Method

When we have one or more low data rate users along with one or more high datarate users, then the assumption of spatially white interference breaks down sincethe interference experienced by the lower data rate users will be directional orspatially colored. Such is the case in multirate services, where both low data ratespeech service and high data rate applications are simultaneously supported. Letx(n) denote the input signal to the correlator and let the despread signal vector atthe output of the correlator be given by

( ) ( ) ( )y s sn n nd I= + (6.23)

where sd(n) and sI(n) are the desired and interfering signal vectors. It then fol-lows that the SINR can be written as


SINRH

dH

I

=w R w

w R w(6.24)

where ( ) ( ) R s sd d dHE n n= and ( ) ( ) R s sI I I

HE n n= . Note that the signals

sd(n) and sl(n) cannot be separately obtained from x(n), however, [8, 13] demon-strate that Rd and Rl can be obtained from x(n) and y(n), as shown next

R R Rx d I= + (6.25)

R R Ry d IPG= + (6.26)

We can then use this information to write

( )( )

w R w

w R w

w R R w

w R R w

Hy

Hx

Hd I

Hd I

PG=

++

(6.27)

w R w

w R w

w R w

w R w

w R w

w R w

Hy

Hx

Hd

HI

Hd

HI

PG

=

+

+

1

1

(6.28)

From (6.28) we can see that for PG > 1 the weight vector that maximizesw R w

w R w

Hy

Hx

also maximizesw R w

w R w

Hd

HI

or the SINR, which leads to

R w R wyy xx= λ (6.29)

In [8, 10], the method of the Lagrange multiplier is used to solve thiseigenvalue problem using the following procedure:

First, the multiplier

( ) ( )f Hy

Hxw w R w w R w= + −γ 1 (6.30)

is introduced for the constraint w R wHx =1. The weight updates are carried out

based on

( ) ( ) ( )( )w w wn n fw+ = + ∇1 0 5. µ (6.31)


where

( )( ) ∇ = −w y xf w R w R w2 γ

It then follows that

( ) ( ) ( )w w R w R wn n y x+ = + −1 µ γ (6.32)

The multiplier is then found as [8]

γ =− −b b ac

a

2

(6.33)

where

( )a H= µ δ 2 2x x

( ) ( ) ( )[ ]b H H H H H H= + ⋅ ∗w x x x x x x y w x w y

2µ Re

( ) ( ) ( )[ ]c H H H H H H= + ⋅ ∗µ µx y x x x x x y w x w y

2Re

6.1.10 Comparison of Adaptive Algorithms

As we indicated earlier, adaptive algorithms were developed to overcome thecomputationally intensive task of finding optimal weights in real time. Many ofthese techniques trace their roots to equalization schemes and vary in their per-formance. Performance criteria most relevant to the mobile communicationsapplications include convergence rate, mean error between optimal and derivedweights, computational complexity in terms of the number of complex multipli-cations, additions that directly contribute to the receiver complexity, and cost.Most of these techniques use some iterative implementation to find the solutionto the cost function, which minimizes some error between the desired signal, itsreplica, and the received signal. Hence, convergence becomes highly dependenton the step size as well as the initial conditions of the algorithm. A small step sizeleads to good estimates of the optimal weights at the expense of a slower conver-gence rate. A large step size, on the other hand, would speed up the convergencerate but the price paid for this is less accurate estimates of the weights. The char-acteristics of the correlation matrix also may affect the convergence rate. Forinstance, in the LMS algorithm the rate is a function of the eigenvalue spread of


the correlation matrix, whereas the convergence rate of the RLS is independentof the eigenvalue spread. Convergence speed is critical for mobile applicationsbecause the signals are fast-changing in time due to fading and mobile speed. Acomparison of the various algorithms described in this section is shown inTable 6.1.

6.2 Neural Network DOA-Based Beamforming

Some efforts have been made to use neural networks in adaptive beamforming.In [48], J. Litva and T. Lo propose the use of a recurrent neural network,namely, the Hopfield network. They demonstrated that Lyapunov functions,


Table 6.1Comparison of Weight Adaptation Algorithms

AdaptiveAlgorithm Approach Convergence References

LMS Minimizes the mean squareerror between the receivedsignal and a reference signal.

Could be slow.

Depends on the correlationmatrix eigenvalue spread.

[14–27]

RLS Minimizes the mean squareerror between the receivedsignal and a reference signal.

Faster than LMS.

Independent of the correlationmatrix eigenvalue spread.

[28–33]

CMA Restores signal envelope byminimizing interference im-pact on the modulus.

Suitable for constant enve-lope modulations.

Nonconvex cost function mayhave false minima.

[34–37]

Decision-Directed

Minimizes the error betweenthe received signal and thecloset member of a knownalphabet.

Depends on receiver ability tolock on the desired signal.

[38–42]

Cyclostationary Exploits cyclic nature of thedesired signal to extract itfrom interference and noise.

Solution requires eigenvaluedecomposition.

[43–47]

ConjugateGradient

Iterative solution to theeigenvalue decompositionproblem.

[1–12]

which are functions that become smaller for any change in the state of the net-work until a stable state is reached, exist for the Hopfield network. This makesthem suitable for solving optimization problems. To minimize the minimumsquare error (MSE) in adaptive beamforming, the MSE is mapped into theenergy function of the network. This is a DOA-based beamforming techniquethat does not require a reference signal to construct an error function to mini-mize. The only a priori knowledge about the desired signals are their angles ofarrivals. The optimum weight vector is a nonlinear function of the correlationmatrix and the constraint matrix. Therefore, it can be approximated using asuitable architecture such as a radial basis function neural network (RBFNN)[49-58]. Note that a radial basis function neural network can approximate anarbitrary function from an input space of arbitrary dimensionality to an outputspace of arbitrary dimensionality. The block diagram of an RBFNN used forbeamforming is shown in Figure 6.3. RBFNNs [59, 60] are a member of a classof general-purpose methods for approximating nonlinear functions. TheRBFNN can be considered as designing neural networks as a curve fitting (orinterpolation) problem in a high-dimensional space. The mapping from theinput space to the output space may be thought of as a hypersurface Γ represent-ing a multidimensional function of the input. During the training phase, theinput-output patterns presented to the network are used to perform a fitting forΓ. The generalization phase represents an interpolation of the input data pointsalong the surface built as an approximation for Γ. The architecture consideredin this chapter consists of three layers—the input layer (sensory nodes), a hiddenlayer of high dimension, and an output layer, as shown in Figure 6.3.


h1 h2 hL

. . .

. . .

. . .

Input layer

Hidden layer

Output layer

Z 1 Z 2 Z 2M(M 1)−

W1 W2 WK

Figure 6.3 Radial basis function neural network for adaptive beamforming.

The transformation from the input space to the hidden-unit space is non-linear, whereas the transformation from the hidden layer to the output space islinear. The network represents a mapping from the p-dimensional input spaceto the m-dimensional output space: p→ m . The radial-basis functions (RBF)technique consists of choosing a function F that has the following form:

( ) ( )F w i i

N

x x x= −∑ ϕ1

(6.34)

where denotes the norm, N is a set of arbitrary functions, and xi are the cen-

ters of the radial-basis functions. One of the common and most useful forms forϕ is the Gaussian function defined by:

( )ϕ σx ex

=− 2

22 for σ > 0, and x ≥ 0 (6.35)

Different learning strategies exist for training RBF networks:

1. Fixed centers selected at random: The locations of the centers areselected randomly from the training set. The standard deviation of theGaussian functions is fixed according to the spread of the centers. If Nc

is the number of centers and d is the maximum distance between thechosen centers, then

σ =d

N c2

This choice avoids too peaked or too flat functions. Thus, only the lin-ear weights w of the output layer need to be learned.

2. Self-organized selection of centers: In this approach, the radial-basisfunctions can move the locations of their centers using a standard rulesuch as the k-nearest neighbor [61]. Then, a supervised learning rulesuch as the LMS algorithm can be used to compute the linear weightsof the output layer.

3. Supervised selection of the centers: This is the most generalized formof RBF networks. A gradient-descent procedure is used to find theweights, the centers, and their spreads, as described in [60]. The bene-fit of such an approach is a minimal network configuration; in otherwords, the same generalization performance can be achieved with asmaller network. As can be seen from Figure 6.3, the RBFNN consists


of three layers of nodes, the input layer, the output layer, and the hid-den layer. The input layer is the layer where the inputs are applied,whereas the output layer is the layer where the outputs are produced.In our application, the input layer consists of J = 2M nodes for anM-element array to accommodate both the real and imaginary part ofthe input vector [i.e., X(t)]. The output layer consists of 2M nodes toaccommodate the output vector (i.e., Wopt). As is the case with mostneural networks, the RBFNN is designed to perform an input-outputmapping trained with examples (Xl(t); Wl

opt) ;l = 1,2,…,NT, where NT

stands for the number of examples contained in the training set. Thepurpose of the hidden layer in a RBFNN is to transform the input dataX(t) from an input space of dimensionality J to a space of higherdimensionality L (see Figure 6.3). The rationale behind this transfor-mation is based on Cover’s theorem [62], which states that aninput/output mapping problem cast in a high-dimensionality spacenonlinearly is easier to solve. The nonlinear functions (the h’s inFigure 6.3) that perform this transformation are usually taken to beGaussian functions of appropriately chosen means and variances.There are a lot of learning strategies that have appeared in the literatureto train an RBFNN. We use the learning strategies where an unsuper-vised learning algorithm (such as the K-means) is initially used to iden-tify the centers of the Gaussian functions comprising the hidden layer.Then, an ad hoc procedure is used to determine the widths (standarddeviations) of these Gaussian functions. According to this procedure,the standard deviation of a Gaussian function of a certain mean is theaverage distance to the first few nearest neighbors of the means of theother Gaussian functions. The aforementioned unsupervised learningprocedure allows us to identify the weights (means and standard devia-tions of the Gaussian functions) from the input layer to the hiddenlayer. The weights from the hidden layer to the output layer are identi-fied by following a supervised learning procedure, applied to a singlelayer network (the network from hidden to output layer). This super-vised rule is referred to as the delta rule. The delta rule is essentially agradient decent procedure applied to an appropriately defined optimi-zation problem. Once training of the RBFNN is accomplished, thetraining phase is complete and the trained neural network can operatein the performance mode (phase). In the performance phase, the neuralnetwork is supposed to generalize, that is, respond to inputs (X(t)’s)that it has never seen before but are drawn from the same distributionas the inputs used in the training set. One way of explaining the gener-alization exhibited by the network during the performance phase is byremembering that after the training phase is complete, the RBFNN


has established an approximation of the desired input/output map-ping. Hence, during the performance phase the RBFNN producesoutputs to previously unseen inputs by interpolating between theinputs used (seen) in the training phase.

6.2.1 Generation of Training Data

It is quite clear that the performance and accuracy of the weight vectors gener-ated using neural networks depends on the type and amount of data used duringthe training phase. One approach for the training data design is to simulate alarge number of scenarios in terms of the number of desired users, relative powerlevels, azimuth distributions, angular separations, and so forth. The step-by-stepprocedure to produce the training data Zl(t); Wl

opt ;l = 1,2,…,NT for theRBFNN in this application is provided next.

1. Generate array output vectors Xl(t); l = 1,2,…,NT.

2. Normalize each one of the above array output vectors by its norm. Forsimplicity of notation we still refer to these vectors by Z(t)’s.

3. Evaluate the correlation matrix Rl (l = 1,2,…, NT) for each of the arrayoutput vectors generated in step 1. Using the calculated Rl ’s, calculatethe vectors Wl

opt; l = 1,2,…, NT based on the Wiener solution.

4. Produce the required training input/output pairs of the training set,that is (Zl(t); Wl

opt) ;l = 1,2,…,NT.

Another approach to the selection of the training data is to use real net-work data collected from field measurements, for example during the spatialchannel modeling. This data collection could be conducted at different times ofthe day, corresponding to periods of various traffic demands. Recall that thistraining phase is performed such that once the RBFNN is trained with a repre-sentative set of training input/output pairs it is ready to function in the perfor-mance phase or the real-time operating mode. In the performance phase, theRBFNN produces estimates of the optimum weights for the array outputsthrough a simple, computationally inexpensive, two-step process, describednext.

6.2.2 Performance Phase of the RBFNN

Once the training phase of the RBFNN is completed offline, the neural networkis ready for online processing of the array data in real time using the followingsimple steps:


1. Collect the normalized array output vector ( )$X t and perform thenormalization.

2. Present the normalized array output vector at the input layer of thetrained RBFNN. The output layer of the trained RBFNN will pro-duce as an output the estimates of optimum weights for the array out-puts (i.e., $W opt ).

Unlike the LMS, RLS, or SMI algorithms, where the optimization is car-ried out whenever the directions of the desired or interfering signals change, inthe RBFNN approach the weights of the trained network can be used to pro-duce the optimum weights needed to steer the narrow beams of the adaptivearray to the direction of desired users. Knowing that the response time for neuralnetworks (i.e., the time that it takes a trained neural network to produce an out-put if it is excited by an input) is very small, this leads us to believe that the pro-posed adaptive beamforming technique will track the mobile users as they move.

6.3 Angle Spread Impact on Optimum Beamforming

A number of adaptive algorithms discussed earlier use a fundamental character-istic of CDMA systems with low data rate users, namely the fact that under theseassumptions the processing gain is large, therefore resulting in a much strongerdesired signal compared with other interfering signals after despreading. Thisenables the algorithm to approximate the optimum vector solution given by theprincipal eigenvector of the spatial correlation matrix, which corresponds to themaximum eigenvalue, as shown in (6.15). This approximation holds when themagnitude of the maximum eigenvalue is much larger than the remainingeigenvalues. One of the most important factors affecting the performance of anyadaptive solution is the impact of the channel on the structure of the spatial cor-relation matrix. It is therefore important to understand the relation between theeigenvalues under different channel conditions. In this section, we analyze theeffect of the angular spread on the eigenvalues of R to gain insight on the perfor-mance of some of the algorithms discussed in this chapter. Let us consider anantenna array with M elements transmitting to L users on the downlink of aCDMA system and assume that the azimuth power spectrum follows a Gaussianprobability density function. Figure 6.4 plots the relative eigenvalue amplitudesfor M = 6, L = 4, angle of departure (AOD) = 0°, and AS of 0° with relativepowers of 0, –6, –10, and –13 dB. It is clear that for small AS, the principaleigenvalue is dominant and the approximation can hold. However, as the ASstarts to increase, the amplitudes of the remaining eigenvalues also increase andbecome comparable to that of the principal eigenvalue. Figure 6.5 shows thesame behavior for AOD of 30°.



AS (Degrees)

0 10 20 30 40 50 60

Rela

tive

Eige

nval

ueam

plit

ude

(dB)

0

–2

–4

–6

–8

–10

–12

–14

–16

–18

Eigenvalue 1Eigenvalue 2Eigenvalue 3Eigenvalue 4

Figure 6.4 Eigenvalues versus AS, M = 6, AOD = 0°, unequal signal powers.

AS (Degrees)

0 10 20 30 40 50 60

Rela

tive

Eige

nval

ueam

plit

ude

(dB)

0

–2

–4

–6

–8

–10

–12

–14

–16

–18


Figure 6.5 Eigenvalues versus AS, M = 6, AOD = 30°, unequal signal powers.

In Figure 6.6 we plot the eigenvalues for equipower signals with AODof 0°. We can observe that although the curves tend to shift up or downacross the three figures, they hold the same behavior as far as the impact of theAS on the relative eigenvalue amplitudes. We can then conclude that the solu-tions that depend on approximating the weight vector by (6.15) are valid forlarge PG and low AS. The question then becomes how would these solutionsbehave for large AS and low PG. To investigate this we compare thebeampattern corresponding to the maximum SINR criterion for an array ofeight elements transmitting to one desired and one interfering user at 30° and60°, respectively, with AS of 0°.

In Figure 6.7, the signals are assumed to have equal powers, whereas inFigure 6.8 the desired signal is assumed to be 128 times larger than the interfer-ing signal (i.e., PG = 128 or about 21 dB). We can see that when the signals havecomparable powers, the beampattern results in two main lobes at both thedesired and interfering user (Figure 6.7). This introduces unnecessary interfer-ence on the forward link or the downlink, limiting that link’s capacity.

In contrast, Figure 6.8 shows that when the PG is large enough only onemain lobe is generated toward the desired user, whereas the gain toward theinterference is significantly reduced. Recall that the weight vector in these caseswas approximated by the principal eigenvector of the correlation matrix.


AS (Degrees)

0 10 20 30 40 50 60

Rela

tive

Eige

nval

ueam

plit

ude

(dB)

0

–2

–4

–6

–8

–10

–12

–14

–16

–18


Figure 6.6 Eigenvalues versus AS, M = 6, AOD = 0°, equal signal powers.


0

–5

–10

–15

–20

–25

Arr

ayp

atte

rn(d

B)

–100 –80 –60 –40 –20 0 20 40 60 80 100AOD (Degrees)

Figure 6.7 Array beampattern, M = 8, equal signals at 30° and 60°, AS = 0°.

0

–5

–10

–15

–20

–25

Arr

ayp

atte

rn(d

B)

–100 –80 –60 –40 –20 0 20 40 60 80 100AOD (Degrees)

Figure 6.8 Array beampattern, M = 8, desired signal (30°) 21 dB higher than interference (60°), AS= 0°.

Figure 6.9 shows the beampattern for the same array with the desired signalpower being 21 dB higher than the interfering signal power and AS of 10°,whereas Figure 6.10 assumes the signals have equal powers and AS of 15°. Wecan then observe that as the AS is increased the beampatterns show degradationin performance. For example, in Figure 6.9 the main lobe is slightly off thedesired user, whereas in Figure 6.10 we can observe a peak at 80° almost equalin gain to that at the desired user.

As we have seen earlier, the difference between the amplitude of the largesteigenvalue corresponding to the principal eigenvector and those of the remain-ing eigenvalues diminishes as the AS increases. To investigate how this affectsthe beamwidth of the main lobe (and consequently the array’s ability to focusenergy in a narrow angular section), we plot the radiation pattern of a six-ele-ment array in Figure 6.11, where we consider one desired signal at 30° and oneinterferer at 60° that is 21 dB weaker than the desired signal, assuming the azi-muth power spectrum follows a uniform probability density function with vari-ous angular spreads. For zero AS, we observe a narrow beam at the direction ofthe desired user. As we start increasing the AS we observe that the main beamstarts to become wider; that is, more power is received from directions other


0

–5

–10

–15

–20

–25

Arr

ayp

atte

rn(d

B)

–100 –80 –60 –40 –20 0 20 40 60 80 100AOD (Degrees)

–30

–35

Figure 6.9 Array beampattern, M = 8, desired signal (30°) 21 dB higher than interference (60°), AS= 10°.


0

–5

–10

–15

–20

–25

Arr

ayp

atte

rn(d

B)

–100 –80 –60 –40 –20 0 20 40 60 80 100AOD (Degrees)

–30

Figure 6.10 Array beampattern, M = 8, equal signals at 30° and 60°, AS = 15.

As=0 degrees

Arr

ayp

atte

rn(d

B)

0

–5

–10

–15

–20–100 –50 0 50 100

AOD (Degrees)

As=10 degrees

Arr

ayp

atte

rn(d

B)

0

–5

–10

–15

–20–100 –50 0 50 100

AOD (Degrees)

As=20 degrees

Arr

ayp

atte

rn(d

B)

0

–5

–10

–15

–20

–100 –50 0 50 100AOD (Degrees)

–25

As=35 degrees

Arr

ayp

atte

rn(d

B)

0

–5

–10

–15

–20

–100 –50 0 50 100AOD (Degrees)

–25

–30

Figure 6.11 Array beampattern, M = 6, desired signal (30°) 21 dB higher than interference (60°), AS= 0°, 10°, 20°, 35°.

than the desired user’s on the uplink and more power is spread in the system onthe downlink (i.e., more interference is created).

6.4 Downlink Beamforming

Because of the asymmetric nature of the traffic demands of data applications,it is expected that the forward link or the downlink will be the limiting linkas far as capacity is concerned in CDMA systems. In TDD-based CDMA sys-tems, the same carrier frequency is used on both the uplink and downlink.Hence, the channel conditions on the two links can be assumed to be identicalbetween consecutive time slots and the weights generated on the uplink can beused on the downlink. However, the situation is different in FDD systemsbecause there is a large separation between the uplink and downlink frequencies.Several techniques have been published in the literature that attempt to over-come this problem. On the uplink, the channel is first estimated; that is, we canestimate the spatial correlation matrix, which can then be employed inbeamforming. The problem in the downlink is that we have no knowledge ofthe downlink channel or the downlink correlation matrix. The simplestapproach is then to simply use the weights computed based on uplink measure-ments directly on the downlink. Clearly, this can result in suboptimal perfor-mance due to the large separation between the frequency bands. Anotherpromising technique is based on the fact that the DOAs of the desired user sig-nal should be independent of the frequency. Since the array steering vector is afunction of the DOA, interelement spacing, and wavelength, we can use thedirections of the dominant paths estimated from the uplink to calculate thedownlink weights. However, this ignores the fact that the amplitudes and phasesof the paths on the downlink will be different from those on the uplink due tothe different carrier frequency, even though the DOAs can be assumed to be thesame. Of course, a feedback technique can be used so that the mobile users canestimate the downlink channel and feed the information back to the base stationto be used in calculating the adaptive weights. As we discussed in an earlierchapter, this method has several drawbacks. First, the amount of feedback willincrease as the number of elements increase. This will reduce the uplink capacitybecause we are increasing the overhead signaling. However, a moderate hit tothe uplink capacity may be tolerable in some cases in exchange of a muchneeded higher capacity in the downlink. Another drawback is that this methodwould be inefficient in high-speed situations where the fading experienced bythe mobiles might be too fast for the feedback loop to compensate. Finally, sucha technique would require changes to be introduced to the current standards.Hence, a more optimal approach would be to try to estimate the downlinkspatial correlation matrix from the uplink measurements. [63, 64] propose a


technique called the frequency calibrated (FC) algorithm, in which an uplinkcovariance matrix is used to estimate the downlink covariance matrix. Reference[63] uses the FC algorithm to estimate the downlink correlation matrix RDL

from the uplink measurements, then uses this information to compute the opti-mum weights using eigen decomposition. On the other hand, [64] uses thesame FC technique to calculate the uplink weights, form the covariance matrixof those weights, and then use that to estimate the covariance matrix of thedownlink weights.

6.5 Vector Rake Receivers

In 3G CDMA systems, reverse link or uplink pilots were introduced as a refer-ence to aid in channel estimation. In the WCDMA system case, the data carriedon the DPDCH channel is modulated on the I-branch, whereas the DPCCHchannel carrying control and signaling information, including pilot bits, is mod-ulated on the Q-branch. The two channels are spread using different PSCs. Atthe base station, the received pilot bits of the DPCCH channel are used to esti-mate the channel parameters, namely the complex fading coefficients. Thisinformation is then used in the demodulation of the DPDCH channel in theRake receiver. The operation of a 1D Rake receiver in a CDMA system can besummarized as follows:

• The time delay positions of the paths with the most significant energiesare identified and correlators or Rake fingers are assigned to thosepeaks.

• Within each Rake finger the fast-changing amplitude and phase aretracked and compensated for.

• Finally, the demodulated and phase-adjusted symbols from all activefingers are MRC combined and the output is fed to the decoder.

Since pilot bits are predefined, they can be used to perform channelestimation. This is performed by first despreading the pilot channel and thenmultiplying the output by the conjugate of the pilot pattern. This effectivelyproduces an instantaneous estimate of the complex fading channel coefficient,which is then filtered and averaged to remove the effects of noise. Thesame principle can be extended to combine spatial processing via beamformingand Rake temporal processing, resulting in the so-called 2D or vector Rakereceivers.


6.6 Channel Estimation

Figure 6.12 shows a block diagram of the channel estimation process based onthe DPCCH channel. From [65] we can write the desired signal at the output ofthe correlator as

( ) ( ) ( ) ( ) ( ) ( )y a S N1 1 11

1 1 1

1 1

τ α θ δ τ τ τ τϕ= − + +=∑R

b n eb

kk

Kj

k k Il k

, , ,,

where αl,k and ϕl,k are the amplitude and phase of the kth multipath componentof the lth user, respectively. SI(τ) is the multiple access interference (MAI) vectorand N(τ) is the noise vector. The first step in channel estimation is to obtain thedelay profile for each pilot bit. Each DPCCH 10 ms frame has 15 slots with Np

pilot bits per slot. Since each user’s pilot is scrambled using a different codesequence, after despreading, the MAI can be considered as independentGaussian noise. Let the output of the correlator at the mth antenna elementwhen the nth pilot bit is received be denoted by y n

m1 , .


CorrelatorCoherentintegration

Correlator

Correlator

Noncoherentintegration

Threshold/decision

DOA Esitimation

sin wot

sin wot

Scrambling code

Scrambling code

Scrambling code

Coherentintegration

sin wot

Coherentintegration

Figure 6.12 Channel estimation in WCDMA beamforming.

Signals at each array element will have certain delay profiles that, whencoherently integrated and averaged, will result in a mean delay profile that canbe written as

( ) ( )gN

ymn

m

n

N

1 11

1τ τ=

=∑ ,

Since the correlator is matched to the desired user’s signal whose multipathwill be phase-aligned, it follows that the magnitudes of the paths belonging tothe desired user will be enhanced or increased by Np times after coherent integra-tion, whereas the paths at time delays belonging to other users will be signifi-cantly reduced because of the randomness introduced by scrambling. Now if thedelay profiles from all M elements were combined, the MAI effect can be furtherreduced, thereby improving the overall SINR. This results in a mean delay pro-file that can be written as [65]

( ) ( )gM

g m

m

M

1 11

1τ τ=

=∑

The mean delay profile can be further processed to estimate the DOAs ofthe incoming signals. This delay profile is essentially a discrete time signal thatcan be Fourier transformed, yielding peaks in the spatial domain representingthe DOAs with magnitudes representing the amplitudes and phases of themultipath components.

6.7 Beamforming

Now that the channel delay profile and the multipath DOAs have been deter-mined, this information can be employed in the beamforming weight calcula-tions, which are then applied to the DPDCH data. The main goal here is thatwe need to invert the channel effects on the transmitted signals. Using the esti-mated mean delay profile, this can be accomplished by simply multiplying thecorrelator output of each antenna element by the complex conjugate of the cor-responding profile or ( )( )g m A

1 τ . The output of the array elements can then be

averaged and fed to the Rake combiners, which perform the MRC scheme.Alternatively, [65] proposes another approach in which the estimated DOAs ofthe multipath components can be used to construct the array response or steer-ing vectors and use those vectors as the weights of the array antenna. Thebeampattern resulting from this operation will then have main lobes toward theDOA directions. The output of the beamformers are then fed to the Rake


receiver for temporal processing, which further removes the effects of chan-nel fading by multiplying its input by the conjugate of the complex fading coef-ficients. On the downlink several beamforming techniques can be used.Based on the assumptions that the downlink multipath channel will retain thesame DOAs as the uplink but with different amplitudes and phases, aDOA-based beamforming technique can be effective. Three choices exist whenthis DOA-based approach is employed, which can be summarized in thefollowing:

• Select the direction of the dominant path and use its DOA to form thearray response vector, which can then be used as the beamformerweight vector, that is w = a(θmax). This will create a single beam towardthe direction of the strongest path.

• Select the directions of the Km strongest paths and use their correspond-ing DOAs to form multiple beams. The multiple beams can be assignedequal energy ratios, in which case the weight vector becomes

( )w m kk

K m

==∑ a θ

1

, or they can be weighted by the corresponding paths

magnitudes, resulting in ( )w m k kk

K m

==∑ α θa

1

. The weights are then nor-

malized to unity to make the total transmitted energy equivalent to theone antenna case.

6.8 Conclusion

A number of training-based and blind adaptive algorithms have been presentedto compute the optimum weight vector for a smart antenna system with anantenna array at the base station. These include classical techniques such as theLMS and RLS originally developed for adaptive filtering, which rely on compar-ing the array output to a reference signal. Other approaches belonging to theso-called blind beamforming attempt to exploit specific signal structures orproperties such as constant envelope or cyclostationarity. Other members of thisclass include methods based on well-known optimization techniques such as theconjugate gradient and Lagrange multipliers. Several methods especially suitablefor CDMA systems take advantage of the unique characteristics of spread spec-trum signals and the resulting correlation matrix structure. Moreover, it hasbeen shown that the computational load of these techniques can be reducedusing reasonable approximations. An emerging technique based on using neuralnetworks to approximate the nonlinear Wiener solution was also discussed. Theimpact of angular spread on the eigenvalue spread of the spatial correlation


matrix was also analyzed, along with its effect on the beampattern of the maxi-mum SINR beamforming criterion. Results presented show how large angularspread can degrade the array performance.

References

[1] Godara, L. C., and D. B. Ward, “A General Framework for Blind Beamforming”, Proc. ofthe IEEE Region 10 Conference, Vol. 2, No., December 1999, pp. 1240–1243.

[2] Alam, F., D. Shim, and B. Woerner, “Comparison of Low Complexity Algorithms forMSNR Beamforming,” IEEE 55th Vehicular Technology Conference, Vol. 4, 2002, pp.1776–1780.

[3] Bengtsson, M., and B. Ottersten, Uplink and Downlink Beamforming for Fading Chan-nels,” IEEE 2nd Workshop on Signal Processing Advances in Wireless Communications, 1999,pp. 350–353.

[4] Chang, T., J. Kim, and C. Kim,” Investigation of the Trade-off Characteristics ofBeamforming Performance in DS-CDMA System,” IEEE 52nd Vehicular Technology Con-ference, Vol. 1, No., 2000 pp. 110–115.

[5] Shim, D., and F. Alam, “A New Adaptive Downlink Beamforming Method for WCDMASystem,” IEEE 53rd Vehicular Technology Conference, Vol. 1, No., 2001 pp. 157–161.

[6] Koutalos, A. C., J. Thompson, and P. Grant, “Downlink Adaptive Antenna Techniquesfor WCDMA,” IEEE 55th Vehicular Technology Conference, Vol. 3, No., 2002, pp.1135–1139.

[7] Kwon, S., I. Oh, and S. Choi, “Adaptive Beamforming from the Generalized EigenvalueProblem with a Linear Complexity for a Wideband CDMA Channel,” IEEE 50th Vehicu-lar Technology Conference, Vol. 3, No., 1999, pp. 1890–1894.

[8] Choi, S., et al., “A Novel Adaptive Beamforming Algorithm for Antenna Array CDMASystems with Strong Interferers,” IEEE Trans. on Vehicular Technology, Vol. 51, No. 5,September 2002, pp. 808–816.

[9] Choi, S., and D. Yun, “Design of an Adaptive Antenna Array for Tracking the Source ofMaximum Power and its Application to CDMA Mobile Communications,” IEEE Trans.on Antennas and Propagations, Vol. 45, No. 9, September 1997, pp. 1393–1404.

[10] Shim, D., and S Choi, “A New Blind Adaptive Algorithm Based on Lagrange Formula fora Smart Antenna System in CDMA Mobile Communications.”

[11] Choi, S , and D. Shim, “A Novel Beamforming Algorithm for a Smart Antenna System ina CDMA Mobile Communication Environment,” IEEE Trans. on Vehicular Technology,Vol. 49, No. 5, September 2000, pp. 1793–1806.

[12] Godara, L. C., “Application of Antenna Arrays to Mobile Communications, Part II:Beamforming and Direction-of-Arrival Considerations,” Proc. of the IEEE, Vol. 85, No. 8,August 1997, pp. 1195–1245.

[13] Naguib, A. F., “Adaptive Antennas for CDMA Wireless Networks”, Ph.D. dissertation,Stanford University, 1996.


[14] Widrow, B., and J. M. McCool, “A Comparison of Adaptive Algorithms Based on theMethods of Steepest Descent and Random Search,” IEEE Trans. Antennas Propagat., Vol.AP-24, 1976, pp. 615–637.

[15] Iltis, R. A., and L. B. Milstein, “An Approximate Statistical Analysis of the Widrow LMSAlgorithm with Application to Narrow-band Interference Rejection,” IEEE Trans.Commun., Vol. COM-33, 1985, pp. 121–130.

[16] Clarkson, P. M., and P. R. White, “Simplified Analysis of the LMS Adaptive Filter Usinga Transfer Function Approximation,” IEEE Trans. Acoust., Speech, Signal Processing, Vol.ASSP-35, 1987, pp. 987–993.

[17] Gardner, W. A., “Comments on Convergence Analysis of LMS Filters with UncorrelatedData,” IEEE Trans. Acoust., Speech, Signal Processing, Vol. ASSP-34, 1986, pp. 378–379.

[18] Foley, J. B., and F. M. Boland, “A Note on the Convergence Analysis of LMS AdaptiveFilters with Gaussian Data,” IEEE Trans. Acoust., Speech, Signal Processing, Vol. 36, 1988,pp. 1087–1089.

[19] Solo, V., “The Limiting Behavior of LMS,” IEEE Trans. Acoust., Speech, Signal Processing,Vol. 37, 1989, pp. 1909–1922.

[20] Feuer, A., and E. Weinstein, “Convergence Analysis of LMS Filters with UncorrelatedGaussian Data,” IEEE Trans. Acoust., Speech, Signal Processing, Vol. ASSP-33, 1985, pp.222–229.

[21] Jaggi, S., and A. B. Martinez, “Upper and Lower Bounds of the Misadjustment in theLMS Algorithm,” IEEE Trans. Acoust., Speech, Signal Processing, Vol. 38, 1990, pp.164–166.

[22] Boland, F. B., and J. B. Foley, “Stochastic Convergence of the LMS Algorithm in Adap-tive Systems,” Signal Process., Vol. 13, 1987, pp. 339–352.

[23] Ko, C. C., “A Fast Adaptive Null-steering Algorithm Based on Output Power Measure-ments,” IEEE Trans. Aerosp. Electron. Syst., Vol. 29, 1993, pp. 717–725.

[24] Ko, C. C., G. Balabshaskar, and R. Bachl, “Unbiased Source Estimation with an AdaptiveNull Steering Algorithm,” Signal Process., Vol. 31, 1993, pp. 283–300.

[25] Benesty, J., and P. Duhamel, “A Fast Exact Least Mean Square Adaptive Algorithm,”IEEE Trans. Signal Processing, Vol. 40, 1992, pp. 2904–2920.

[26] Godara, L. C., “Improved LMS Algorithm for Adaptive Beamforming,” IEEE Trans.Antennas Propagat., Vol. 38, 1990, pp. 1631–1635.

[27] Ohgane, T., et al., “BER Performance of CMA Adaptive Array for High-Speed GMSKMobile Communication—A Description of Measurements in Central Tokyo,” IEEETrans. Veh. Technol., Vol. 42, 1993, pp. 484–490.

[28] Eweda, E., and O. Macchi, “Convergence of the RLS and LMS Adaptive Filters,” IEEETrans. Circuits Syst., Vol. CAS-34, 1987, pp. 799–803.

[29] Fabre, P., and C. Gueguen, “Improvement of the Fast Recursive Least-squares Algorithmsvia Normalization: A Comparative Study,” IEEE Trans. Acoust., Speech, Signal Processing,Vol. ASSP-34, 1986, pp. 296–308.

[30] Mantey, P. E., and L. J. Griffiths, “Iterative Least-squares Algorithm for Signal Extrac-tion,” in 2nd Int. Hawaii Conf. System Science, Honolulu, HI., 1969, pp. 767–770.


[31] Eleftheriou, E., and D. D. Falconer, “Tracking Properties and Steady State Performance ofRLS Adaptive Filter Algorithms,” IEEE Trans. Acoust., Speech, Signal Processing, Vol.ASSP-34, 1986, pp. 1097–1110.

[32] Mueller, M. S., “Least Squares Algorithms for Adaptive Equalizers,” Bell Syst. Tech. J.,1981, pp. 1905–1925.

[33] Cioffi, J. M., and T. Kailath, “Fast Recursive-Least-Square, Transversal Filters for Adap-tive Filtering,” IEEE Trans. Acoust., Speech, Signal Processing, Vol. ASSP-32, 1984, pp.998–1005.

[34] Chiba, I., W. Chujo, and M. Fujise, “Beamspace Constant Modulus Algorithm AdaptiveArray Antennas,” Proc. Inst.Elect. Eng. 8th Int. Conf. Antennas and Propagation, Edin-burgh, Scotland, 1993, pp. 975–978.

[35] Godard, D. N., “Self-recovering Equalization and Carrier Tracking in Two-dimensionalData Communication Systems,” IEEE Trans. Commun., Vol. COM-28, 1980, pp.1867–1875.

[36] Treichler, J. R., and B. G. Agee, “A New Approach to Multipath Correction of ConstantModulus Signals,” IEEE Trans. Acoust., Speech, Signal Processing, Vol. ASSP-31, 1983, pp.459–472.

[37] Shynk, J. J., and C. K. Chan, “Performance Surfaces of the Constant Modulus AlgorithmBased on a Conditional Gaussian Model,” IEEE Trans. Signal Processing, Vol. 41, 1993,pp. 1965–1969.

[38] T. E., Reed, J. H., and W. H Biedka Tranter, “Mean Convergence Rate of a DecisionDirected Adaptive Beamformer with Gaussian Interference,” Proc. of the 2000 IEEE SensorArray and Multichannel Signal Processing Workshop, March 16–17, 2000, pp. 68–72.

[39] Jwa, O. Hyunseo, and K. Mungeon, “Decision-directed Chip-level Beamforming inWCDMA Antenna Array System,” IEEE 55th Hyekyung Vehicular Technology Conference,Vol. 1 , May 6–9, 2002. pp. 322–326.

[40] Duhamel, P., M. Montazeri, and K. Hilal, “Classical Adaptive Algorithms (LMS, RLS,CMA, Decision Directed) Seen as Recursive Structures,” IEEE International Conference onAcoustics, Speech, and Signal Processing, Vol. 3, April 27–30, 1993, pp. 496–499.

[41] Povey, G. J. R., P. M. Grant, and R. D. Pringle, “A Decision-directed Spread-spectrumRAKE Receiver for Fast-fading Mobile Channels,” IEEE Trans. on Vehicular Technology,Vol. 45, No. 3, August 1996, pp. 491–502.

[42] Swindlehurst, A. L., S. Daas, and Y. Jiankan, “Analysis of a Decision DirectedBeamformer,” IEEE Trans. on Signal Processing, Vol. 43 , No. 12 , December 1995, pp.2920–2927.

[43] Altuna, J., and B. Mulgrew, “A Comparison of Cyclostationary Blind Equalisation Tech-niques Multipath Countermeasures,” IEE Colloquium on , May 23, 1996, pp. 8/1–8/6.

[44] Abed-Meraim, K., et al., “Blind Source-separation Using Second-order CyclostationaryStatistics,” IEEE Trans. on Signal Processing, (see also IEEE Trans. on Acoustics, Speech, andSignal Processing), Vol. 49, No. 4 , April 2001, pp. 694–701.

[45] Enserink, S., and Cochran, D., “On Detection of Cyclostationary Signals,” InternationalConference on Acoustics, Speech, and Signal Processing, Vol. 3, May 9–12, 1995, pp. 2004–2007.


[46] Castedo, L., et al., “Linearly-constrained Adaptive Beamforming Using CyclostationarySignal Properties,”IEEE International Conference on Acoustics, Speech, and Signal Processing,Vol. iv, April 19–22, 1994, pp. IV/249–IV/252.

[47] Castedo, L., C. Y Tseng,.and L. J.Griffiths, “A New Cost Function for AdaptiveBeamforming Using Cyclostationary Signal Properties,” IEEE International Conference onAcoustics, Speech, and Signal Processing, Vol. 4, April 27–30, 1993, pp. 284–287.

[48] Litva, J., and T. K. Yeung Lo, Digital Beamforming in Wireless Communications, Norwood,MA: Artech House, 1995.

[49] El Zooghby A. H., C. G. Christodoulou, and M. Georgiopoulos, “Neural Network-BasedAdaptive Beamforming for One and Two Dimensional Antenna Arrays,” IEEE Trans. onAntennas and Propagation, December 1998.

[50] El Zooghby A. H., C. G. Christodoulou, and M. Georgiopoulos, “A Neural Net-work-based Linearly Constrained Minimum Variance Beamformer,” submitted to TheMicrowave and Optical Technology Letters. Microwave and Optical Technology Letters,1999, p.p. 451–455.

[51] El Zooghby A. H., C. G. Christodoulou, and M. Georgiopoulos, “Neural Network basedBeamforming for Interference Cancellation,” Proc. of the SPIE’s AeroSense Conference,Orlando, FL, 1998.

[52] El Zooghby A. H., C. G. Christodoulou, and M. Georgiopoulos, “A Novel Approach toAdaptive Nulling with Neural Networks,” Proc. SoutheastCon 98, Orlando, FL, 1998.

[53] El Zooghby A. H., C. G. Christodoulou, and M. Georgiopoulos, “Adaptive InterferenceCancellation with Neural Networks,” Proc. of the 8th Virginia Tech Symposium on WirelessPersonal Communications, Blacksburg, VA, June 1998, pp. 281–292.

[54] El Zooghby A. H., C. G. Christodoulou, and M. Georgiopoulos, “Neural Network basedSmart Antennas for Mobile Satellite Communications,” Proc. of the International Sympo-sium on Electromagnetic Theory, Thessaloniki, Greece, May 1998, pp. 336–339.

[55] El Zooghby A. H., C. G. Christodoulou, and M. Georgiopoulos, “Adaptive InterferenceCancellation in Circular Arrays with Radial Basis Function Neural Networks,” IEEEAP/URSI Symposium, Atlanta, 1998.

[56] El Zooghby A. H., C. G. Christodoulou, and M. Georgiopoulos, “Radial Basis FunctionNeural Network Algorithm for Beamforming in Cellular Communication Systems,”IEEE-APS Conference on Antennas and Propagation for Wireless Communications, Waltham,MA, 1998, pp. 52–56.

[57] El Zooghby A. H., C. G. Christodoulou, and M. Georgiopoulos, “Multiple Sources Neu-ral Network Direction Finding with Arbitrary Separations,” IEEE-APS Conference onAntennas and Propagation for Wireless Communications, Waltham, MA, 1998, pp. 57–60.

[58] El Zooghby A. H., C. G. Christodoulou, and M. Georgiopoulos, “Multiple Mobile UserTracking with Neural Network-based Adaptive Array Antennas,” accepted/to be publishedin the SPIE’s AeroSense 99 Conference, Orlando, FL.

[59] Haykin, S., Neural Networks: A Comprehensive Foundation, Macmillan College Publishing.

[60] Mulgrew, B., “Applying Radial Basis Functions,” IEEE Signal Processing Magazine, March1996, Vol. 13, No. 2, pp. 50–65.

[61] Tou, J. T., and R. C. Gonzalez, Pattern Recognition Principles, Reading, MA: AddisonWesley, 1975.


[62] Moody, T. J., and C. J. Darken, “Fast Learning in Networks of Locally Tuned ProcessingUnits,” Neural Computation, Vol. 1, 1989, pp. 281.

[63] Liang, Y. C, and F. P. S. Chin, “Downlink Channel Covariance Matrix Estimation and ItsApplication in Wireless DS-CDMA Systems,” IEEE Journal on Selected Areas in Communi-cations, Vol. 19, No. 2, February 2001, pp. 222–232.

[64] Liang, Y. C, and F. P. S. Chin, “FDD DS-CDMA Downlink Beamforming by ModifyingUplink Beamforming Weights,” IEEE 52nd Vehicular Technology Conference, Vol. 1, Sept.24–28, 2000, pp. 170–174.

[65] Li, H. J, and T. Y Liu; “Comparison of Beamforming Techniques for W-CDMA Com-munication Systems,”IEEE Trans. on Vehicular Technology, Vol. 52, No. 4, July 2003, pp.752–760.


7Coverage and Capacity Improvements in3G Networks

7.1 Introduction

CDMA2000 and WCDMA have many similarities as well as some differences.They use different types of channelization and spreading codes, pilot transmis-sion schemes, coding, interleaving and modulations, physical channel structure,frame structure, power control, and handoff algorithms, as well as higher layeralgorithms. These differences, however, have little impact on the system’s rela-tive coverage or capacity because they are optimized up to certain limits, mainlyas a result of the required backward compatibility with their respective 2G sys-tems (IS-95 and GSM). On the other hand, the fundamental differ-ence—spreading bandwidth—affects the dimensioning process moresignificantly. Each CDMA2000 carrier uses 1.25 MHz of the total availablebandwidth with a chip rate of 1.2288 Mcps, whereas WCDMA spreads the sig-nals into a 5M-Hz band, using a chip rate of 3.6864 Mcps. A wide bandwidthgives additional diversity (i.e., improved capacity for WCDMA). A narrowbandwidth gives reduced own-cell interference (i.e., improved capacity forCDMA2000). Thus, the overall impact is small in most cases (i.e., CDMA2000and WCDMA have similar basic performance). Since in general CDMA is aninterference-limited system, it is always capacity limited. However, when thecapacity is not the main issue, radio design will be based on the link budget andcoverage. This is typically the case in the initial deployment of a new system orin the design of isolated rural cells, where the goal is to cover the area with aminimum number of base stations. Similarly, when reliable coverage for high bitrate indoor users from outdoor base stations is a requirement, cell sizes will be

191

small, which provide more capacity per km2 than is needed. This chapter com-pares the CDMA2000 and WCDMA radio links. The impact of issues such asavailable bandwidth, detailed environment characteristics (which are importantin network optimization), the suitability or compatibility of either WCDMA orCDMA2000 for overlaying on, or co-existing with (in a gradual migration), aparticular deployed 2G system are not taken into account. In other words, weassume that the choice would be based only on the related performances of thesetwo systems in terms of capacity and coverage.

7.2 Link Budgets and Coverage

For narrowband systems such as FDMA or TDMA, the cell loading does notaffect coverage design. However, for CDMA, due to universal frequency reuse,capacity and coverage analyses are tied together as more users are admitted intothe network, the interference level is increased, and coverage area of the cellsshrinks. Therefore, to have reliable coverage at the cell boundary, link budgetand coverage analysis are usually based on a certain cell loading (that is, a certaininterference margin is considered in the link budget). For the uplink (UL), a cellloading of 50% is commonly considered, which corresponds to a noise rise of 3dB. This might be regarded as the design objective in the initial deployment of asystem. Therefore, the need for additional base stations would occur only afterpenetration has reached this limit (50% load). An uplink budget is employed topredict the coverage of a CDMA RF network based on the limitations of thesubscriber stations to communicate back to the base station. This fundamentallimitation on coverage in most cases will be discussed later. Therefore, the limi-tation of service is based on the uplink budget(s). The uplink budget can varywith the subscriber type, subscriber configuration, subscriber mobility, andmorphology classification. A separate link budget should be created for eachpermutation of these parameters. The parameters included in the link budgetare provided below.

7.2.1 Mobile Station Parameters

Here are the mobile station parameters:

• Tx power: Transmit power subscriber station;

• Cable loss: Attenuation due to connector and cables to the antenna;

• Mobile station effective antenna gain: Gain of antenna on MS terminal;

• Effective isotropic radiated power (EIRP): The mobile station’s transmitpower less any cable loss and effective antenna gain.


7.2.2 Base Station Parameters

Here are the base station parameters:

• BS Rx antenna gain: Gain of BS antenna;

• BS cable loss: Attenuation for BS cable from antennas to receiver frontend, including any couplers and multiplexers;

• Noise figure: Noise figure of BS receiver.

7.2.3 System Parameters

Here are the system parameters:

• Maximum traffic channel data rate;

• Required BS Eb/No: Required ratio of traffic channel energy per bit tointerference energy. Sometimes this is also referred to as Eb/Io. The valueis based on fading environment of subscribers. Highly mobile subscrib-ers tend to require a higher ratio to overcome more severe fading.

7.2.4 Margins

The following are margins:

• Building or vehicle penetration: RF attenuation of signal passing intobuilding or mobile vehicle. The penetration margin tends to be associ-ated with the building classification or morphology.

• Fade margin: Margin allocated to overcome effects of shadowing andlognormal fading. The amount of margin is determined by the desiredconfidence in coverage at the cell edge. Typically, the fade marginranges from 4.5 dB to 11 dB.

• Soft handoff (SHO) gain: Gain produced by mobile soft handoff diver-sity on two or more RBSs. The gains to the link budget are assumeddue to the statistical independence of the RF fading between the sub-scriber terminal and any two base stations. No handoff gain is assumedfor fixed users since their directional antennas or orientation inside thebuildings tends to increase the correlation in fade paths and reduce theamount of soft handoff.

7.2.5 Other Parameters

The following are other parameters:

Coverage and Capacity Improvements in 3G Networks 193

• Thermal noise density: Ambient RF noise floor of environment set at–174 dBm/Hz.

• Signal bandwidth: Bandwidth of CDMA channel.

• Load margin: Rise in thermal noise floor on uplink due to increasedinterference levels produced by increased traffic. A traffic load of 50%equates to a 3.0 dB rise in the uplink noise floor.

• Receiver sensitivity (unloaded) = Thermal noise density + BSreceiver noise figure + Required BS mean Eb/No + 10 log (maximumdata rate).

• Maximum allowable path loss (MAPL): Maximum path loss for thegiven parameters the subscriber station can tolerate and still communi-cate reliably on the uplink.

• Maximum path loss = Maximum subscriber Tx EIRP – Head Loss –Building penetration loss + Soft handoff gain – Fading margin + BSantenna gain – BS cable loss – BS receiver sensitivity (unloaded) – Loadmargin.

7.2.6 Fade Margin

The confidence of coverage at cell edge and area is characterized statisticallyin terms of a normal distribution curve. To achieve a certain confidencein having service at the predicted edge of coverage, a particular fading mar-gin must be applied to the link budget. The amount of margin addeddepends on the level of confidence required and the standard deviation oflognormal fading. For the case where handoffs occur near the cell edge of cov-erage, the transmitter must only provide enough power to overcome back-ground noise and interference. However, with shadowing, the transmitter hasto increase the transmitted power to overcome the random fluctuation in pathloss due to lognormal fading. The margin set in transmitted power due tocompensate for lognormal fades is known as the lognormal shadow mar-gin. Lognormal shadow margin, µ, can be calculated by the followingequation:

Confidence level (cell edge) ( )= −⋅ − ⋅ ⋅

∞ −

∫11

2 10

2

2

π µ

δ

ex

dxn rlog


n path loss exponent

r Normalized cell radius on the cell b

=

= ( )oundary, r

Log normal shadow standard deviation

==

⋅

∞

1

1

2

δ

π ( )e dx

n rOutage probability ce

x−

∫− ⋅ ⋅

=

2

2

10µ

δ

log ( )ll edge

It is the percentage of time when a receiver experiences

" shadowed" performance during shadowing

(7.1)

For example, for lognormal shadow standard deviation of 8 dB, alognormal shadow margin of 10.3 dB is required for 90% confidence if hardhandoffs occur at the edge between hexagonal cells. The higher the confi-dence, the higher the required margin and the smaller the cell radiuses are. (Forexample, assuming flat earth, no clutter and 35 dB per decade morphology lossslope, 1 dB decrease in margin will result in 6% decrease in cell radius and 14%decrease in cell area). A 90% confidence level indicates that 90% of the time thereceivers on the cell edge can expect to be able to access the RF network.

7.2.7 Confidence (Cell Area)

The cell area confidence level is that percentage of time when those active sub-scribers randomly distributed over the cell sites coverage area can overcome theshadow fading to be received by the RBS receiver. Confidence level within a cellis higher than that of the confidence level on the cell boundary for a given signalstrength threshold xo. The confidence level within the cell, Fu, can be calculatedas follow:

( )F erf x e erfx y

yu

x y

y= ⋅ − + ⋅ −− ⋅

− ⋅ ⋅

1

21 1

11 2

2

(7.2)

where

( )

y n log

erf xp

e dt

10

tx

= ⋅ ⋅⋅

= −⋅

= ⋅ −∫

102

22 2

0

e

x

δµ

δ


µ = Log normal shadow margin for a given xo

(see “log normal shadow margin” section for equationn = Path loss exponentδ = Log normal shadow standard deviation in dB

Figure 7.1 shows the relation between confidence level for cell area and celledge against lognormal shadow margin (hard handoff) of 8 dB standarddeviation.

7.2.8 CDMA Traffic Loading

With CDMA technology, all subscribers are using common frequencies toreceive and transmit, unlike earlier technologies that assigned distinct frequen-cies and/or time slots for each subscriber. Because the spectrum is shared, theprimary factor limiting capacity in CDMA is the interference levels within theallocated CDMA spectrum. The uplink capacity is based on a formula called thepole capacity. The pole capacity represents a theoretical limit of simultaneouscalls a single sector could support if all the subscribers could theoretically main-tain their transmit power levels at their minimal values with perfect power con-trol. But in reality, only a fraction of the pole capacity is practically achievable.


Figure 7.1 Cell edge versus cell area confidence.

The higher the fraction of pole capacity the design attempts to achieve, the morethe voice quality and/or grade of service must be compromised. The effect of ris-ing interference requires the subscriber to increase its transmit power to main-tain an acceptable frame error rate on the reverse link (RL). The maximumaverage traffic loading recommended is 65% of pole capacity. Cell loading isdefined as the ratio of active users to the maximum allowable number of usersexpressed in terms of percentage of pole capacity. Uplink cell loading iscalculated as [1, 2]

ην

ULoc

sc

b oj j

j

NI

I WE N

R

u

= +

+=∑1

1

11

(7.3)

where

Nu: number of users;

vj: activity factor for user j [0.67 for adaptive multirate (AMR) voice inWCDMA and 1 for data];

I

Ioc

sc

: average other cell to same cell interference ratio (0.65 for three-sector

sites).

The downlink (DL) traffic loading factor equation can be written as

( )η ν αDL j

b o j

j

oc

scj

N E N

W R

I

I

u

= + −

=∑ 1

1

(7.4)

where α is the average orthogonality factor (1 fully orthogonal, 0 noorthogonality).

7.3 Voice Services

Table 7.1 summarizes the Eb/No values required to achieve acceptable qual-ity based on the listed BLER targets for different WCDMA data rates [1].This value is used as one of the inputs to the link budget and is one of themost important factors that determine the coverage and capacity for a givenservice.


7.3.1 Uplink Budgets

Tables 7.2 and 7.3 show the link budget for a pedestrian A (Ped. A) channel forurban, suburban, and rural environments for CDMA2000 and WCDMA,respectively. For CDMA2000, the assumptions and, hence, the different entriesin the link budget, such as mobile transmit power, pilot overhead, BS andmobile antenna gains, and different losses at mobile station and BS are takenfrom [3]. For WCDMA, most of the data has been taken from [1, 2]. Weassumed a vocoder output rate of 12.2 Kbps for WCDMA and 9.6 Kbps forCDMA2000.

7.3.2 Downlink Budgets

For downlink, we assumed a maximum BS transmit power of 20W and alsoassumed that the total available traffic power is given only to the voice users.Furthermore, we set a power limit for each user at 10% of the total availabletraffic power (which gives a maximum transmit power around 32 dBm). Thisvalue, although somehow arbitrarily chosen here, has significant effect on thedownlink coverage area. In real scenarios, this value is set for each user depend-ing on the number of users and their bit rates. For CDMA2000, the valuesshown in Table 7.4 are taken from [3]. In general, SHO gain in the forward link(FL) is higher than the reverse link. The reason is due to the use of MRC in the


Table 7.1WCDMA Link Level Performance Summary for Different Data Rates

Data Rate (Kbps) Mobile ChannelUL Eb/No(dB) DL Eb/No (dB)

BLER(%)

12.2 Additive white Gaussian noise(AWGN)

2.9 4.4 1%

12.2 Pedestrian A 3 km/hr 4.2 7 1%

12.2 Case 3 5.5 7 1%

64 AWGN 1 2.5 10%

64 Pedestrian A 3 km/hr 2.2 5.3 10%

64 Case 3 3.4 5.3 10%

384 AWGN 0.6 2.4 10%

384 Pedestrian A 3 km/hr 2 5.1 10%

384 Case 3 3.4 5.1 10%


Table 7.2CDMA2000 RL/UL Voice Link Budget

Parameters CDMA2000

Urban Suburban Rural

RL RL RL

Radio Configuration RC3 RC3 RC3

Mobile Environment Ped. A Ped. A Ped. A

Target FER/BLER 1.00% 1.00% 1.00%

Peak Data Rate (bps) 9,600 9,600 9,600

MS Parameters

MS TX PO (Watts) 0.2 0.2 0.2

MS TX PO (dBm) 23.0 23.0 23.0

MS Combiner Loss (dB) 0.0 0.0 0.0

MS Cable Loss (dB) 0.0 0.0 0.0

MS Antenna Gain (dBi) 0.0 0.0 0.0

EIRP (dBm) 23.0 23.0 23.0

Total Reverse Link Overhead (dB) –1.5 –1.5 –1.5

Traffic EIRP (dBm) 21.5 21.5 21.5

BS Parameters

BS Antenna Gain (dBi) 18.0 18.0 18.0

BS Combiner Loss (dB) 0.0 0.0 0.0

BS RX Cable Loss (dB) –1.0 –1.0 –1.0

BS Noise Figure (dB) 3 3 3

Thermal Noise (dBm/Hz) –174 –174 –174

BS Receiver Noise Density (dBm/Hz) –171 –171 –171

Peak Data Rate (dB) 39.8 39.8 39.8

Required Eb/No Set Point (dB) 5.5 5.5 5.5

Eb/No Std Deviation (dB) 0.0 0.0 0.0

Mean Eb/No (dB) 5.5 5.5 5.5

Effective BS Sensitivity (dBm) –125.7 –125.7 –125.7

Network Design Margins

Body Loss (dB) –2 –2 –2


Table 7.2 (continued)

Parameters CDMA2000


RL RL RL

Building Penetration Loss (dB) –20.0 –15.0 –10.0

Confidence (Cell Edge) 90% 90% 90%

Lognormal Shadow Std Dev (dB) 8.0 8.0 8.0

Path Loss Slope Y 3.5 3.5 3.5

Lognormal Shadow Margin (dB) –10.3 –10.3 –10.3

Confidence (Cell Area) 97% 97% 97%

CDMA Soft Handoff Gain at 50% Correlation (dB) 3.0 3.0 3.0

CDMA Traffic Loading (%) 50% 50% 50%

CDMA Traffic Loading Effect (dB) –3.0 –3.0 –3.0

Maximum Allowable Path Loss (dB) 131.9 136.9 141.9

Table 7.3WCDMA RL/UL Voice Link Budget

Parameters WCDMA


UL UL UL


FEC Code Convolution Convolution Convolution

Target FER/BLER 1.00% 1.00% 1.00%


MS Parameters

MS TX PO (Watts) 0.125 0.125 0.125

widctlparMS TX PO (dBm) 21.0 21.0 21.0




EIRP (dBm) 21.0 21.0 21.0

Traffic EIRP (dBm) 21.0 21.0 21.0



Parameters WCDMA


UL UL UL

BS Parameters










Mean Eb/No (dB) 4.2 4.2 4.2



Body Loss (dB) –2.0 –2.0 –2.0







CDMA Soft Handoff Gain at 50%Correlation (dB)

3.0% 3.0% 3.0%


CDMA Traffic Loading Effect (dB) –3.0 –3.0 –3.0



Table 7.4CDMA2000 FL/DL Voice Link Budget

CDMA2000

Peak Data Rate (bps) Input 9,600 9,600 9,600


Variable Origin DL DL DL

BS Parameters

BS TX Power (Watts) Input: 20.0 20.0 20.0

BS TX Power (dBm) Calc: 43.0 43.0 43.0

Forward Overhead (%) Input: 25% 25% 25%

Power Allocated to All Voice Users (%) Input: 100% 100% 100%

Maximum Power Allocated to a User(%)

Input: 10% 10% 10%

Maximum TX Traffic Power/User(Watts)

Calc: 1.5 1.5 1.5

Maximum TX Traffic Power/User(dBm)

Calc: 32 32 32

BS Combiner Loss (dB) Input: 0.0 0.0 0.0

BS Cable Loss (dB) Input: –1.0 –1.0 –1.0

BS Antenna Gain (dBi) Input: 18.0 18.0 18.0

EIRP/User (dBm) Calc: 48.8 48.8 48.8

MS Parameters

MS Antenna Gain (dBi) Input: 0.0 0.0 0.0

MS Combiner Loss (dB) Input: 0.0 0.0 0.0

MS RX Cable Loss (dB) Input: 0.0 0.0 0.0

FLMS Noise Figure (dB) Input: 8 8 8

Thermal Noise (dBm/Hz) Calc: –174 –174 –174

MS Receiver Noise Density (dBm/Hz) Calc: –166 –166 –166

Peak Data Rate (dB) Calc: 39.8 39.8 39.8

Required Eb/No Set Point (dB) Input: 6.0 6.0 6.0

Eb/No Std Deviation (dB) Input: 0.0 0.0 0.0

Mean Eb/No (dB) Calc: 6.0 6.0 6.0

Effective MS Sensitivity (dBm) Calc: –120.2 –120.2 –120.2

forward link, as opposed to the selection diversity employed in the uplink. TheWCDMA DL budget is shown in Table 7.5. Whether we consider CDMA2000or WCDMA, the coverage versus capacity trade-off remains the same with onlyslight differences in the actual MAPL and cell sizes. In the remainder of thischapter, the analysis will focus on WCDMA networks, although the same meth-odology is equally applicable to CDMA2000 networks. In Figure 7.2, we com-pare the uplink and downlink coverage versus capacity trade-off for 12.2 KbpsWCDMA voice in a macrocell, where we can see that for small loads orthroughputs the uplink is the limiting link for coverage. At very high number ofusers, the downlink may become the coverage limiting link, assuming that sectorcapacity can be supported.

7.4 Data Applications

Tables 7.6 and 7.7 show the uplink and downlink budgets for a pedestrian Achannel for WCDMA. For the downlink (Table 7.6), we have assumed that80% of the total available traffic power is given to data users. Furthermore,the maximum power of each user is limited to 30% of the total traffic power.



CDMA2000





Body Loss (dB) Input: 0.0 0.0 0.0

Building Penetration Loss (dB) Input: –20.0 –15.0 –10.0

Confidence (Cell Edge) Input: 90% 90% 90%

Lognormal Shadow Std Dev (dB) Input: 8.0 8.0 8.0

Lognormal Shadow Margin (dB) Calc: –10.3 –10.3 –10.3

Soft Handoff Gain at 50% Correlation(dB)

Input: 3.0 3.0 3.0

Target Loading (%) 50% 50% 50%

Interference Margin (Noise Rise) (dB) Input –3.0 –3.0 –3.0



Table 7.5WCDMA FL/DL Voice Link Budget

WCDMA




BS Parameters

BS TX Power (Watts) Input: 20.0 20.0 20.0

BS TX Power (dBm) Calc: 43.0 43.0 43.0

Forward Overhead (%)1 Input: 0% 0% 0%

Power Allocated to All Voice Users (%) Input: 100% 100% 100%

Maximum Power Allocated to a User (%) Input: 10% 10% 10%

Maximum TX Traffic Power/User (Watts) Calc: 2 2 2

Maximum TX Traffic Power/User (dBm) Calc: 33 33 33

BS Combiner Loss (dB) Input: 0.0 0.0 0.0

BS Cable Loss (dB) Input: –1.0 –1.0 –1.0

BS Antenna Gain (dBi) Input: 18.0 18.0 18.0

EIRP/User (dBm) Calc: 50 5 50

MS Parameters

MS Antenna Gain (dBi) Input: 0.0 0.0 0.0

MS Combiner Loss (dB) Input: 0.0 0.0 0.0

MS RX Cable Loss (dB) Input: 0.0 0.0 0.0

MS Noise Figure (dB) Input: 8 8 8

Thermal Noise (dBm/Hz) Calc: –174 –174 –174

MS Receiver Noise Density (dBm/Hz) Calc: –166 –166 –166

Peak Data Rate (dB) Calc: 40.9 40.9 40.9

Required Eb/No Set Point (dB) Input: 7.0 7.0 7.0

Eb/No Std Deviation (dB) Input: 0.0 0.0 0.0

Mean Eb/No (dB) Calc: 7.0 7.0 7.0

Effective MS Sensitivity (dBm) Calc: –118.1 –118.1 –118.1

1. Overhead already accounted for in Eb/No values

Figure 7.3 shows the results of coverage versus capacity analysis for data users inan urban area. Similar to voice application, the uplink remains as the limitinglink for coverage for data users. Note that the data rate service with the lowestallowed propagation loss determines the cell range. The voice coverage for theuplink is smaller than downlink for both systems and for all environments. This



WCDMA





Body Loss (dB) Input: 0.0 0.0 0.0

Building Penetration Loss (dB) Input: –20.0 –15.0 –10.0

Confidence (Cell Edge) Input: 90% 90% 90%

Lognormal Shadow Std Dev (dB) Input: 8.0 8.0 8.0

Lognormal Shadow Margin (dB) Calc: –10.3 –10.3 –10.3

Soft Handoff Gain at 50% Correlation (dB) Input: 3.0 3.0 3.0


Interference Margin (Noise Rise) (dB) Input –3.0 –3.0 –3.0


Figure 7.2 WCDMA 12.2-Kbps voice coverage versus capacity in uplink and downlink.


Table 7.6WCDMA UL Data Link Budget

Parameters WCDMA


UL UL UL


ntblTarget FER 10.00% 10.00% 10.00%


MS Parameters

MS TX PO (Watts) 0.2 0.2 0.2

MS TX PO (dBm) 23 23 23




EIRP (dBm) 23 23 23

Traffic EIRP (dBm) 23 23 23

BS Parameters










Mean Eb/No (dB) 2.2 2.2 2.2


Design Margin

Body Loss (dB) 0.0 0.0 0.0

Building Penetration Loss (dB) –20 –15 –10





Parameters WCDMA


UL UL UL




CDMA Soft Handoff Gain at 50%Correlation (dB)

2.0 2.0 2.0


CDMA Traffic Loading Effect (dB) –3 –3 –3


Table 7.7WCDMA DL Budget


Variable DL DL DL


FER 10.00% 10.00% 10.00%

BS Parameters

BS TX Power (Watts) 20.0 20.0 20.0

BS TX Power (dBm) 43.0 43.0 43.0

Power Allocated to All Data Users (%) 80% 80% 80%

Maximum Data Power Allocated to a User (%) 30% 30% 30%

Maximum TX Traffic Power/User (Watts) 4.8 4.8 4.8

Maximum TX Traffic Power/User (dBm) 36.8 36.8 36.8


BS Cable Loss (dB) –1.0 –1.0 –1.0


EIRP/User (dBm) 53.8 53.8 53.8

is favorable to 3G asymmetric services, such that the coverage design can bedone based on the uplink and support higher data rates for downlink. As the bitrate goes high, the coverage shrinks.

From the results shown in Figures 7.2 and 7.3, we can draw the followingobservations:




Variable DL DL DL

MS Parameters



MS RX Cable Loss (dB) 0.0 0.0 0.0

MS Noise Figure (dB) 8 8 8


MS Receiver Noise Density (dBm/Hz) –166 –166 –166


Required Eb/No Set Point (dB]) 5.3 5.3 5.3


Mean Eb/No (dB) 5.3 5.3 5.3

Effective MS Sensitivity (dBm) –104.9 –104.9 –104.9

Design Margins

Body Loss (dB) 0.0 0.0 0.0





Soft Handoff Gain at 50% Correlation (dB) 2.0 2.0 2.0


Interference Margin (Noise Rise) (dB) –3.0 –3.0 –3.0


• The downlink has a better MAPL than the uplink for symmetric ser-vices (e.g., voice services and data services with 64Kbps peak rate onboth links).

• This implies that in those cases, the coverage will be uplink limited.This means that some base station transmit power that would have oth-erwise been used up in a larger cell size on the downlink is now availableand can be employed to support higher data rates on the downlink atthe cell edge.

• In the case of asymmetric services (e.g., 384 Kbps on the DL and 64Kbps on the UL), the downlink coverage can become the limiting fac-tor. The load factor on the DL is much higher than the UL. Moreover,the base station can run out of power much faster than in the case ofvoice services, in which system capacity is more likely to become powerlimited than code limited.

In CDMA systems there is always a trade-off between coverage and capac-ity. To provide large cell sizes, especially in early deployment phases, the net-work is designed with a low UL load factor that limits system capacity.Increasing capacity requires an increase in the target load factor, which results ina smaller cell size. The UL load factors for voice and 64 Kbps data are shown inFigures 7.4 and 7.5.

7.5 Limiting Links for Coverage and Capacity

To determine how to optimize and maximize the impact of employing smartantennas to improve system performance, we first need to identify the areas thatlimit that performance or, more precisely, we need to identify whether the sys-tem is coverage limited or capacity limited and which link is the limiting one.


Figure 7.3 WCDMA 64-Kbps data coverage versus capacity in uplink and downlink.

7.5.1 Coverage Limited Scenarios

In general, for low bit rate services such as voice and for small cell loading (smallthroughput situations), the uplink is the limiting link for coverage. The reason ismainly due to the limitation of the mobile terminal transmit power. This isdespite other factors in favor of the uplink, such as receive antenna diversity andbetter receiver design, available at the base station. On the other hand, as thenumber of users increases and/or for higher bit rates (i.e., as the sector through-put is allowed to increase, for example, by increasing the target uplink load), thedownlink becomes the limiting link. This is shown in Figure 7.6, where for 384Kbps on the downlink and 64 Kbps on the uplink, the coverage becomesdownlink limited. The reason is that in the downlink, the maximum transmitpower is the same, regardless of the number of users or their bit rates, and is


Figure 7.4 WCDMA 12.2-Kbps voice interference rise versus uplink load.

Figure 7.5 WCDMA 64-Kbps data interference rise versus uplink load.

always shared among all downlink users, whereas in the uplink, each additionaluser has its own power amplifier. Therefore, in the downlink, in addition to themaximum base station total power, individual users have power limits thatdepend on the number of users in the cell and their bit rates. This power limitmight become the deciding factor in the coverage.

7.5.2 Capacity Limited Scenarios

The two key factors in determining which link is limiting the system capacityare the base station transmit power and uplink target load factor. As we haveseen earlier, there is a trade-off between system coverage and capacity; that is,increasing the sector throughput (in other words, the capacity or number ofusers) shrinks the coverage. Normally, the initial system design is based on thecoverage requirement, which is why the initial target uplink load factor is low sothat the coverage area is maximized. This would lead to an uplink capacity lim-ited scenario as the maximum uplink load is reached before the base station runsout of transmit power. For a network designed with high uplink load, such as inurban environments, the system may become downlink capacity limited whenthe base station runs out of transmit power before the uplink load is reached.This is especially true when the traffic is asymmetric with higher traffic (bit rate)in the downlink. Table 7.8 lists key factors by which the coverage and capacitylimiting links are evaluated for the downlink and uplink. Techniques that can beused to alleviate the limiting link(s) are listed in Table 7.9.

7.6 Smart Antennas Impact on Uplink Coverage and Capacity

Because coverage is mostly uplink limited, except for very high data rates or lowPA capabilities (e.g., microcells), this section will focus on evaluating the cover-age improvements that can be obtained using antenna arrays at the base station.The sensitivity of the BS receiver is determined by the noise figure, the


Figure 7.6 WCDMA 384/64-Kbps data coverage versus capacity in uplink and downlink.

maximum symbol rate received, the noise floor of the RF environment, and theEb/No set point. If we can reduce this sensitivity requirement this would translateto a capacity increase and/or reduced mobile transmit power. Let C be the car-rier power and N and IMAI be noise and multiple access interference power; thenthe receiver sensitivity is given by [4]


Table 7.8Limiting Link Evaluation

Link Coverage Capacity

LimitingFactors

Reasons Limiting Factors Reasons

Uplink Required mobileERP.

Highest data rate.

MAPL inverselyproportional to datarate.

UL loading.

Number of activeusers/sector.

Low target uplinkload factor.

Symmetric traffic.

Downlink BS power sharedamong all downlinkusers.

BS transmit power.

Traffic channels/sector.

High target uplinkload factor.

Asymmetric traffic(higher traffic onthe DL).

Table 7.9Coverage and Capacity Improvement Techniques

LimitingLink Coverage Capacity

Indications Solutions Indications Solutions

Uplink UL MAPL < DLMAPL

Improve ULlink budget

UL load atmaximum.

Total BSpower belowmaximum.

Improve ULload equation.

Improve BSEb/No.

Downlink DL MAPL < ULMAPL

Improve DLlink budget

UL load belowmaximum.

BS transmitpower atmaximum.

Improve DLload equation.

Improve DLlink budget.

Improve DLEb/No.

SC

N ICIMAI tot

=+

= (7.5)

and the cell load can be shown to be [5]

α =I

IMAI

tot

(7.6)

where the multiple access interference is given by

I I IMAI SC OC= + (7.7)

If the smart antenna can lower the receiver’s sensitivity requirement to S´,a capacity increase of β and a power reduction of δ can be achieved as follows

( )SC

N IC

I IM tot tot

'' '

=+

=− +

δ

α βδα1(7.8)

It follows that the capacity gain and power reduction are given by

( )β

δ α

δα=

+ −SS ' 1

(7.9)

( )δ

α

− βα=

−1S

S '

(7.10)

In Figure 7.7, we plotted δ as a function of the net gain that can beachieved using smart antennas for different load factors. The net gain is definedas the total sum of the directivity gain, diversity gain or loss, and other types oflosses, such as combining loss associated with some fixed beam implementa-tions. In a 75% loaded system, a 1 dB gain in the receiver’s sensitivity could leadto a 3-dB power reduction. When the transmit power is reduced the mobile bat-tery life is extended.

On the other hand, if the same mobile transmit power is maintained, thiscould translate to a range extension. Figure 7.8 shows the expected capacityincrease we can achieve with no power reduction. A 3-dB gain with a 40%loaded system corresponds to about a 250% capacity increase. However, in theabove situation if we allow the capacity to increase, the system load will alsoincrease.


To control and maintain the same system load, the above expression forthe capacity improvement is modified as follows

( )β

δ

1− α+ δα=

SS ' (7.11)

We plotted (7.11) in Figure 7.9 in a 50% loaded system for differentpower reduction factors. It can be seen how capacity increase and power


Figure 7.7 Power reduction versus gain.

Figure 7.8 Capacity increase versus net gain with no power reduction.

reduction can be traded off. For a 3-dB gain, power can be reduced by 50% andcapacity will increase by 33%. A capacity improvement of 67.5 % could beobtained with a 4-dB gain.

Several approaches can be used to evaluate the performance impacts ofsmart antennas [5–34]. They vary from extensive link and system level simula-tions to analytical techniques that look at outage probabilities and statisticalcapacity analysis. An alternative way to evaluate the impact of smart antennason the uplink that is adopted in this chapter is to use the link budgets andload equations. Using smart antennas on the uplink provides two link level ben-efits, the first is a reduction of the required BS Eb/No and the second is anaperture gain, both of which improve the uplink coverage. From (7.3) wecan see that the uplink load factor is a function of the data rate, Eb/No, andnumber of users. Therefore, reducing Eb/No lowers the load factor for a givennumber of users or sector throughput. This reduction in uplink load factorfor a given system throughput improves both coverage and capacity. The cover-age is improved because the lower load factor results in a higher MAPL. Thecapacity is increased because the number of users and, hence, the sectorthroughput can be increased until the load factor reaches the target load. Figure7.10 illustrates the coverage improvements achievable using four fixed beamantennas, assuming an Eb/No reduction of 3 dB [23]. We notice that the gainsare dependent on the number of users (throughput) or, more precisely, on theload factor.

The gains are largest for high loads because, at those levels, the interferencerise over thermal would normally be so high that when smart antennas are usedsignificant reductions can be obtained. This is demonstrated in Figure 7.11,where we can see the greatest impact at high loads. Similarly, capacity gains can


Figure 7.9 Capacity increase and power reduction trade-off in a 50% loaded system.

be obtained, as shown in Figure 7.12. The actual increase in the uplink capacityis a function of the target load factor. The higher the target load the higher thecapacity gains. In summary, with smart antennas at the base station, we canincrease the uplink capacity by raising the uplink target load factor without sac-rificing coverage because the reduction in Eb/No and array gain can limit theinterference level and improve coverage.

7.6.1 Smart Antenna Impact on Downlink Capacity

As we have seen earlier, the capacity is typically downlink limited. In thissection, we will present the link level performance gains of two different


Figure 7.10 Uplink coverage improvements with multiple fixed beam antennas.

Figure 7.11 Uplink interference reduction with multiple fixed beam antennas.

approaches, namely transmit diversity and fixed beam antennas, and show howthe downlink sector capacity can be increased.

7.6.1.1 Transmit Diversity

Transmit diversity techniques were discussed in detail in Chapter 5. We willnow examine the diversity gain that can be obtained using open loop and closedloop transmit diversity techniques and show how this gain can be traded off forcapacity improvements.

Figures 7.13 and 7.14 compare the diversity gains of OTD, STS, andTXAA approaches versus speed and multipath for CDMA2000 9.6-Kbps voiceusers [8, 19].


(a)

(b)

Figure 7.12 (a, b) Uplink capacity improvements versus load factor with multiple fixed beamantennas.

Figure 7.15 shows the diversity gains obtained in WCDMA 12.2-Kbpsvoice with STTD and CLTD techniques [35]. The diversity gain can be definedin different ways; here, we consider the reduction in transmit Ec/Ior (ratio of traf-fic channel power to total transmit power) or Eb/No. We can summarize theresults as follows:

• Closed loop techniques such as TXAA provide the largest gains, fol-lowed by STS and then OTD. This is because the feedback from themobile stations improves channel estimation and the application ofmore accurate weights.


Figure 7.13 Open loop versus closed loop transmit diversity gains in CDMA2000.

Figure 7.14 OTD versus STS transmit diversity gains in CDMA2000.

• Diversity gain is greatest when time and multipath diversity perfor-mance is poor (i.e., for low-mobility users with little multipath). This isbecause such users spend more time in periods with deep fades, thusaffecting the performance.

• High-speed users experience far less diversity gain and almost negligibleperformance improvement. This is in contrast to beamforming tech-niques that offer performance improvements independent of the user’sspeed, as we will discuss later in this chapter.

Figure 7.16 shows the diversity gain as a function of the geometry G,defined as the ratio of the intracell interference power spectral density to the


Figure 7.15 Open loop versus closed loop transmit diversity gains in WCDMA.

Figure 7.16 Transmit diversity gains versus geometry in CDMA2000.

intercell interference power spectral density or I I I Isc oc or oc/ $ /= . We can see

that regardless of the diversity technique used, the diversity gain is higher for lowG values (i.e., when the intercell or other cell interference is dominant). This isusually the case at the cell edge. For high G values (i.e., close to the base station),the intracell interference is more dominant and we do not get as much diversitygain. Diversity gain or, in other words, the reduction in Ec/Ior or Eb/No canimprove both coverage and capacity on the downlink. For instance, inmicrocells where the coverage might be downlink limited due to low total basestation transmit power, this diversity gain can be employed to improve coverage.In macrocells, the coverage is typically uplink limited and the capacity is usuallydownlink limited, in which case we can make use of the diversity gain toimprove the downlink capacity.

Figure 7.17 shows the reduction in downlink load factor and, conse-quently, the interference rise as the sector throughput (load) is increased in aWCDMA system carrying voice traffic at 12.2 Kbps, assuming an STTD diver-sity gain of 2 dB. Note that the maximum capacity gain with this performanceimprovement would be 100.2 or roughly 60%.

However, the actual capacity gain depends on the target load factor to whichthe system is designed. The higher the load factor (smaller coverage), the higherthe capacity gain, as illustrated in Figure 7.18. For example, at 40% target loadthe sector throughput can be increased from 483 Kbps to 724 Kbps, a 50% capac-ity improvement, whereas the capacity can be increased from about 905 Kbps to1,450 Kbps, equivalent to 60% capacity improvement at 90% target load.

Similarly, the downlink coverage can be improved with transmit diversity,as can be seen from Figure 7.19, where an STTD gain of 2 dB was assumed. Wenote the same behavior as well where the coverage gain is dependent on the load


Figure 7.17 Downlink load reduction with STTD.

factor and we do not get the same coverage improvement for all loads. Oneimportant factor that must be taken into account when we discuss the capacityimprovements obtained from spatial techniques such as TD is that even if weassume that all the transmit power reduction is traded off for capacity gains, thesystem may simply become code limited instead of power limited and thosecapacity improvements would not be possible to obtain in practice unless weintroduce secondary scrambling codes into the cell. This is because in voice ser-vices with SF of 128, there are approximately 60–65 users that can be supportedwhen the common channel and handoff reduction factor are taken into account.To achieve the full benefits of TD, we must therefore increase the number ofavailable codes. In WCDMA, a second PSC code with a corresponding group of


Figure 7.18 Downlink capacity improvements versus load factor with STDD macrocell.

Figure 7.19 Downlink coverage improvements with STTD macrocell.

orthogonal codes can be allocated, although this second group will cause someinterference with the group of orthogonal codes associated with the first PSC.Similarly, CDMA2000 overcomes this problem by allocating additional codesusing quasi-orthogonal functions (QOFs). Alternatively, instead of increasingthe voice capacity the extra base station transmit power that is freed up as aresult of the diversity gain can be used to support more high data rate users inthe cell or increase the peak data rate at the cell edge.

7.6.1.2 Beamforming

Although coverage and capacity are coupled in the sense that conventionallywhen one is improved the other will be degraded, when smart antennas are usedfor coverage and capacity improvement it is possible to improve both simulta-neously. Antenna arrays with more than two elements that implement eitherfixed beam or user-specific schemes can provide greater gains than transmitdiversity. The gain has two main components, aperture gain and spatial filteringgain. The aperture gain is proportional to the number of elements M and isgiven by 10Log (M). Spatial filtering confines the interference in a limited AS,therefore the gain is greatest for low AS because the interference is confined to asmall angular region and is decreased as the AS is increased, since more interfer-ence is received or spread into the system. Figure 7.20 presents the gains interms of reduction in downlink Eb/No in a macrocell [2]. We can see that thegains increase as the number of beams (elements) is increased and are greatest forlow AS.

In the previous section we showed how the uplink coverage and capacitycan be improved when beamforming is implemented at the base station. Figures7.21 and 7.22 show the improvement we can achieve when smart antennas are


Figure 7.20 Beamforming gains versus angular spread in WCDMA [2].

used at the base station on both the uplink and downlink for 12.2-Kbps voiceservice.

With four fixed beams on the uplink providing approximately 3 dB ofgain in Eb/No and two beams on the downlink with Eb/No gain of 2.2 dB at an ASof 2°, the sector throughput can be increased from 905 Kbps to 1,208 Kbps,whereas the coverage is improved by about 9 dB. The Eb/No gains for AS of 20°


100

105

110

115

120

125

130

135

140

145

150

0 200 400 600 800 1000 1200 1400

Throughput [kbps]

UL (3-sectors)

UL (4 Fixed beams)

DL (3-sectors)

DL (2 Fixed beams,AS=2)

Capacity gain

MA

PL

[dB

]

Co

ve

rag

eg

ain

Figure 7.21 Downlink/uplink coverage and capacity improvements with multiple fixed beam anten-nas. Macrocell, small AS.

Figure 7.22 Downlink/uplink coverage and capacity improvements with multiple fixed beam anten-nas. Macrocell, large AS.

are slightly less (1.8 dB). Higher AS results in more interference received on theuplink or spread in the downlink since the angular region is wider, whichreduces the performance gains slightly. The gains are a result of improvementsin the uplink budget, uplink load factor (lower interference rise), aperture gain,and lower interference in the downlink. That is an improvement in both thelink budget and load equation. Higher gains are possible if four beams are usedon the downlink as well. For instance, the downlink coverage is significantlyimproved for 384/64 Kbps data, as shown in Figure 7.23, with the antennaarray providing 4.5-dB gain in Eb/No. The actual capacity gains in terms of insector throughput are dependent on the downlink target load. Figure 7.24 com-pares those gains for different target loads.


Figure 7.23 Downlink coverage improvements with multiple fixed beam antennas.

Figure 7.24 Downlink capacity improvements with multiple fixed beam antennas.

7.6.1.3 Microcell Environments

In a microcell the base station transmit power is usually limited and is muchlower than the macrocell case. Therefore, the coverage in a microcell is morelikely to be downlink limited. Another major difference between the two typesof cells is that the angular spread in a microcell is much larger than in amacrocell. As we have seen in previous chapters, as the AS increases the correla-tion between antenna elements is decreased; as a result, the diversity gain alsoincreases. Therefore, transmit diversity might be more effective in these environ-ments because the beamforming performance in terms of interference reductionis affected by the large AS. In [2] STTD and CLTD capacity increases of 50%and 75%, respectively, were reported in a microcell environment. Figure 7.25compares the reduction in the interference rise (gains in capacity) using STTDin both macrocells and microcells. The downlink orthogonality factor and


(a)

(b)

Figure 7.25 Downlink capacity improvements versus interference rise with STTD: (a) macrocell, and(b) microcell.

other-to-same cell interference ratio used in the calculations are 0.9 and 0.4,respectively, for the microcell, as opposed to 0.5 and 0.65 for the macrocell case.For a 40% target load factor (about 2.2 dB noise rise), a 65% improvement inthe throughput can be achieved for the microcell case versus an approximate50% increase in the macrocell case.

7.7 Conclusions

The applications of different smart antenna techniques in 3G networks werediscussed in this chapter, including transmit diversity and beamforming. Table7.10 provides a brief comparison between TD and Beamforming. Methods todetermine the coverage and capacity limiting links and techniques to improvethe overall system performance were described. On the uplink, antenna arrays atthe base station enable the system to tune itself for optimized signal reception.The result is that the received signal level is improved by a factor of M (numberof antenna elements); at the same time, interference is significantly reduced.Similar gains also occur when smart antennas are used on the downlink. Whenthe system is tuned for optimized signal transmission the power level of thetransmitted signal is a factor of M over the power emitted by a single antenna atthe base station. At the same time, less interference is spread on the downlink.The reduction in interference allows an increase either in the number of sub-scribers using the system or in the overall signal quality, which enables higherdata throughput. The benefit of this reduction in interference networkwide is,in both cases, an increase in spectral efficiency. Smart antenna gains are not lim-ited to just providing an aperture gain, but they also provide improvements inEb/No that directly translate into coverage and capacity gains. All wireless systemssuffer some degree of fading. Since the environment is dynamic, this fading istime varying. The consequence for wireless system designers is that the air inter-face must be robust to sudden outages and margins against fading must be intro-duced into link budgets and cell planning, which reduces coverage. Fading issubstantially mitigated when multiple antennas are used. When one antennafades in the array, chances are that others do not. The output of the array is,therefore, much smoother over time. Thus, there is a reduction in the neededmargin against fading, which is often referred to as a “diversity gain” that is inaddition to the aperture gain. The amount of this gain depends on the targetedoutage probability, the detailed assumptions regarding the fading process, andthe number of antennas. Simply put, smart antenna systems fundamentallyimprove the coverage and spectral efficiency trade-offs of wireless systems,although some trade-offs between cost, coverage, and capacity remain in thewireless system design.


References

[1] Holma, H., and A. Toskala, WCDMA for UMTS: Radio Access for Third GenerationMobile Communications, 3rd ed., New York: John Wiley & Sons, 2004.

[2] Laiho, J., T. Novosad, and A. Wacker, Radio Network Planning and Optimization forUMTS, New York: John Wiley & Sons, 2001.

[3] Kim, K. L., Handbook of CDMA System Design, Engineering, and Optimization, PrenticeHall, N. J., 2000.


Table 7.10Transmit Diversity Versus Beamforming Performance Comparison

Transmit Diversity Beamforming

Mobile Speed Performance degradeswith increasing mobilespeed.

Performance independent of mobile speed.

Angular Spread Performance improveswith increasing AS.

Performance degrades with increasing AS.

Interference Not effective againstinterference.

Very effective in interference limited scenarios.Significant interference reduction capabilities.Performance improves with increasing numberof beams/elements.

FLMacrocell Less effective thanbeamforming due torelatively lower AS.

Effective. Largest performance improvements inlow AS cases.

Microcell More suitable thanbeamforming.

Performance affected by the increased AS.

UL Coverage Limited N/A Performance gains could be traded off forcoverage improvement.

UL Capacity Limited N/A Performance gains could be traded off forcapacity improvement.

DL Coverage Limited Performance gainscould be traded off forcoverage improvement.

Performance gains could be traded off forcoverage improvement.

DL Capacity Limited Performance gainscould be traded off forcapacity improvement.

Performance gains could be traded off forcapacity improvement.

Geometry G Better gains at celledge (low geometry).

Performance independent of geometry.

[4] EL Zooghby, A. H., “Potentials of Smart Antennas in CDMA Systems and UplinkImprovements,” IEEE Antennas and Propagation Magazine, Vol. 43, No. 5, October 2001,pp. 172–177.

[5] Lee, H. W., J. Yeom, and D. K. Sung, “Performance Analysis of Downlink Time SwitchedTransmit Diversity (TSTD) in W-CDMA System,” IEEE 51st Vehicular Technology Con-ference Proc., Vol. 1, May 15–18, 2000, pp. 561–565.

[6] Tiirola, E., and J. Ylitalo, “Comparison of Beamforming and Diversity Approaches for theCoverage Extension of W-CDMA Macro Cells,” IEEE 54th Vehicular Technology Confer-ence, Vol. 3, Oct. 7–11, 2001 pp. 1274–1278.

[7] Dabak, A. G., et al., “A Comparison of the Open Loop Transmit Diversity Schemes forThird Generation Wireless Systems,” IEEE Wireless Communications and Networking Con-ference, Vol. 1, September 23–28, 2000, pp. 437–442.

[8] Soni, R. A., and R. M. Buehrer, “On the Performance of Open-Loop Transmit DiversityTechniques for IS-2000 Systems: A Comparative Study,” IEEE Trans. on Wireless Commu-nications, Vol. 3, No. 5, September 2004, pp. 1602–1615.

[9] Rohani, K., “Open-loop Transmit Diversity for CDMA Forward Link,” IEEE EmergingTechnologies Symposium: Broadband, Wireless Internet Access, April 10–11, 2000, p. 5.

[10] Hak-Seong K., W. Lee and Y. Shin, “Performance Analysis of an Open Loop TransmitDiversity for Rician Multipath Wideband CDMA Channels,” IEEE Seventh InternationalSymposium on Spread Spectrum Techniques and Applications, Vol. 2, September 2–5, 2002,pp. 400–404.

[11] Qiang, Y., and D. Li, “Performance Analysis of Several Open-loop Transmit DiversitySchemes for IMT-2000 Systems,” International Conference on Communication TechnologyProc., Vol. 2, April 9–11, 2003, pp. 1107–1110.

[12] Canales, M., et al., “Performance Analysis of Downlink Transmit Diversity SystemApplied to the UTRA FDD Mode,” IEEE Seventh International Symposium on SpreadSpectrum Techniques and Applications, Vol. 2, September 2–5, 2002, pp. 410–414.

[13] Soni, R. A., and R. M. Buchrer, “Open-Loop Transmit Diversity in IS-20000 Systems,”Conference Record of the Thirty-Third Asilomar Conference on Signals, Systems, and Comput-ers, Vol. 1, October 24–27, 1999, pp. 654–658.

[14] Hamalainen, J., and R. Wichman, “Closed-Loop Transmit Diversity for FDD WCDMASystems,” Conference Record of the Thirty-Fourth Asilomar Conference on Signals, Systems,and Computers, Vol. 1, Oct. 29–Nov. 1, 2000, pp. 111–115.

[15] Vanganuru, K., and A. Annamalai, “Analysis of Transmit Diversity Schemes: Impact ofFade Distribution, Spatial Correlation and Channel Estimation Errors,” IEEE WirelessCommunications and Networking, Vol. 1, March 16–20, 2003, pp. 247–251.

[16] Vishwakarma, R., and K. S. Shanmugan, “Performance Analysis of Transmit AntennaDiversity in 3G WCDMA System,” IEEE International Conference on Personal WirelessCommunications, December 17–20, 2000, pp.1–4.

[17] Rohani, K., M. Harrison, and K. Kuchi, “A Comparison of Base Station Transmit Diver-sity Methods for Third Generation Cellular Standards,” IEEE 49th Vehicular TechnologyConference, Vol. 1, May 16–20, 1999, pp. 351–355.

[18] Rajan, D., and S. D. Gray, “Transmit Diversity Schemes for CDMA-2000,” IEEE WirelessCommunications and Networking Conference, September 21–24, 1999, pp. 669–673.


[19] Derryberry, R. T., et al., “Transmit Diversity in 3G CDMA Systems,” IEEE Communica-tions Magazine, Vol. 40, No. 4, April 2002, pp. 68– 75.

[20] Goransson, B., et al., “Advanced Antenna Systems for WCDMA: Link and System LevelResults,” 11th IEEE International Symposium on Personal, Indoor and Mobile Radio Com-munications, Vol. 1, September 18–21, 2000, pp. 62–66.

[21] Sousa, V. A., Jr., et al., “Coverage and Capacity of WCDMA Systems with Beam SteeringAntennas,” IEEE 58th Vehicular Technology Conference, Vol.2, Oct. 6–9, 2003, pp.826–830.

[22] Bae, K. K., J. Jiang, and W. H. Tranter, “Downlink WCDMA Performance Analysis withDiversity Techniques Combined with Beamforming,” IEEE Wireless Communications andNetworking, Vol. 1, March 16–20, 2003, pp. 202–206.

[23] Ylitalo, J., and E. Tiirola, “Performance Evaluation of Different Antenna ArrayApproaches for 3G CDMA Uplink,” IEEE 51st Vehicular Technology Conference Proc.,Tokyo, Vol. 2, May 15–18, 2000, pp. 883–887.

[24] Kogiantis, A. G., “Uplink Capacity Studies for Adaptive Antenna Arrays in Third Genera-tion CDMA Wireless Systems,” IEEE Wireless Communications and Networking Confer-ence, Sept. 21–24, 1999, pp. 679–683.

[25] Lee, J., and R. Arnott, “System Performance of Multisector Smart Antenna Base Stationsfor WCDMA,” Second International Conference on 3G Mobile Communication Technolo-gies, (Conf. Publ. No. 477), March 26–28, 2001, pp. 1–6.

[26] Baumgartner, T., T. Neubauer, and E. Bonek, “Performance of Downlink Beam Switch-ing for UMTS FDD in the Presence of Angular Spread,” IEEE International Conference onCommunications, Vol. 2, April 28–May 2, 2002, pp. 851–855.

[27] Yong Z., and F. Zhenghe, “Capacity Analysis of CDMA Networks with Smart Antenna,”Canadian Conference on Electrical and Computer Engineering, Vol. 2, May 13–16, 2001,pp. 1333–1336.

[28] Soni, R. A., R. M. Buehrer, and R. D. Benning, “Intelligent Antenna System forCDMA2000,” IEEE Signal Processing Magazine, Vol. 19, No. 4, July 2002, pp. 54–67.

[29] Goransson, B., B. Hagerman, and J. Barta, “Adaptive Antennas in WCDMA Sys-tems-Link Level Simulation Results Based on Typical User Scenarios,” IEEE 52nd Vehicu-lar Technology Conference, Vol. 1, September 24–28, 2000, pp. 157–164.

[30] Barta, J., S. Petersson, and B. Hagerman, “Downlink Capacity and Coverage Trade-offs inWCDMA with Advanced Antenna Systems,” IEEE 55th Vehicular Technology Conference,Vol. 3, May 6–9, 2002, pp. 1145–1149.

[31] Barta, J., et al., “Interference Distributions in Mixed Service WCDMA Systems—Oppor-tunities for Advanced Antenna Systems,” IEEE 53rd Vehicular Technology Conference, Vol.1, May 6–9, 2001, pp. 263–267.

[32] Osseiran, A., et al., “Downlink Capacity Comparison Between Different Smart AntennaConcepts in a Mixed Service W-CDMA System,” IEEE 54th Vehicular Technology Confer-ence, Vol. 3, October 7–11, 2001, pp. 1528–1532.

[33] Ericson, M., et al., “Capacity Study for Fixed Multibeam Antenna Systems in a MixedService WCDMA System,” 12th IEEE International Symposium on Personal, Indoor andMobile Radio Communications, Vol. 1, September 30–October 3, 2001, pp. A-31–A-35.


[34] Zhou, Y., et al., “Performance Comparison of Transmit Diversity and Beamforming forthe Downlink of DS-CDMA System,” IEEE Trans. on Wireless Communications, Vol. 2,No. 2, March 2003, pp. 320–334.

[35] Parkvall, S., et al., “Transmit Diversity in WCDMA: Link and System Level Results,”IEEE 51st Vehicular Technology Conference Proc., Tokyo, Vol. 2, May 15–18, 2000, pp.864–868.


8Smart Antennas System Aspects

8.1 Introduction

Air interface standards have recently started to take smart antenna techniquesinto account, making it possible to optimize radio system design for spatial pro-cessing and integrate those advanced antennas into future adaptive modems.The main advantages expected with smart antennas include higher sensitivereception, interference cancellation in uplink and downlink functions, and miti-gation of multipart fading effects. On the system level, this will lead to highercapacity, extended range, improved coverage of current dead spots, higher qual-ity of service, lower power consumption at the mobile, and improved powercontrol (PC). While smart antennas increase system complexity and cost, theyalso provide an additional degree of freedom for radio network control andplanning. To improve radio network performance, the propagation and inter-ference environment as well as expected traffic and user’s mobility in the cellshould be taken into account when optimizing the smart antenna receiver struc-tures and algorithms. While smart antenna receiver parameters are important forcapacity, coverage, and interference planning, they also interact with differentnetwork control protocols. In addition, when evaluating different smart antennareceivers and algorithms one must consider their high degree of dependence onthe air interface and its parameters, such as the multiple access method, the typeof duplexing, pilot availability, modulation, diversity, physical channels split-ting, and frame structure. Smart antenna algorithms should also be compatiblewith radio network protocols because link level control protocols must maintainthe required link quality dynamically while carrying out channel and interfer-ence monitoring. Several smart antennas strategies are available in the radionetwork design and planning, including:

231

• The use of smart antennas at the BS in the uplink, only to increase cov-erage. Figure 8.1 illustrates this concept.

• The use of smart antennas at the uplink and downlink simultaneouslyto improve coverage and capacity.

• The use of smart antennas at mobiles alone without installing them atthe BS, achieving similar improvement in coverage and capacity as withsmart antennas at the BS, although the extent of such an approachwould be limited by the cost and complexity of the mobile stations.

• The use of smart antennas at both ends to allow several parallel chan-nels to be established between the mobile and BS. In this case, higherbit rate transmission can be achieved by splitting data streams betweenparallel channels.

• The use of space-time (ST) coding, which exploits transmit diversitytechniques with a MIMO channel.

Several of these topics are discussed in more detail in the remainder of thischapter.

8.2 Third Generation Air Interfaces and Protocol Stacks

Wireless mobile communications standards are layered architectures defined interms of a protocol stack, which describes different layers that handle different


Sectorization

Fixed multiple beams

Adaptive beamforming

Sectorization

Fixed multiple beams

Adaptive beamforming

Low interferenceenvironment

High interferenceenvironment

Cov

erag

e(c

ellr

ange

/are

a)

Figure 8.1 Coverage improvement comparison.

functions on the physical, logical, and other levels. The interactions andmappings between the layers are also defined. The two major 3G standards arethe CDMA2000 evolution of IS-95 and the WCDMA evolution ofGSM/GPRS. In this section, we will briefly discuss and compare the differencesand similarities between those two technologies, both based on CDMA.

The WCDMA protocol stack has a uniform structure[1–5]. Signaling anduser data (voice, packet-switched data, circuit-switched data) all flow throughthe RLC and MAC layers by using logical and transport channels configured tosupport the appropriate quality of service. Mobility management, call and ses-sion management, short message service, and supplementary services functionsare defined in the nonaccess stratum (NAS) layer specifications. The architec-ture, functionality, and interfaces of the NAS layers are specified in 3GPP speci-fications in a way that is tightly coupled with the access stratum layers. TheCDMA2000 protocol stack is layered, but the layering is not as uniform as inthe WCDMA protocol stack. For example, the link access control (LAC) layercorresponds to the RLC layer of WCDMA, but only signaling flows through theLAC layer. User data flows directly into the MAC layer. The Radio Link Proto-col (RLP) sublayer is architecturally part of the MAC layer, but it is not specifiedin the MAC layer specification. Instead, it is specified as part of the IS-707 dataservices specification. In CDMA2000, upper layer signaling, or layer 3, includesfunctions that in WCDMA would be separated into NAS and RRC layers.Layer 3 is responsible for radio resource functions, such as handoffs and channelconfigurations, as well as mobility management and call control functions, suchas registrations, call origination, and call release [6–10]. A comparison betweenthe WCDMA and CDMA2000 stacks is shown in Figure 8.2.

8.3 Physical Layer

8.3.1 Data Multiplexing

One main difference between WCDMA and CDMA2000 is the way differentdata streams are multiplexed in the physical layer, as shown in Figure 8.3. Datastreams can be voice, signaling, or data at various data rates, quality require-ments, block sizes, traffic patterns, and such. To accommodate the need to carrymultiple data streams, WCDMA uses time multiplexing, whereas CDMA2000uses code multiplexing. CDMA2000 does include time multiplexing of multi-ple data streams at the multiplex sublayer, which is above the physical layer.Because this time multiplexing occurs above the physical layer, however, all themultiplexed data streams receive the same treatment with respect to error correc-tion, interleaving, and so on. In WCDMA, multiplexing occurs in the physicallayer after error correction and interleaving are individually applied to eachtransport channel. A lot of the differences in the details between the WCDMA

Smart Antennas System Aspects 233


NAS

RRC Voiceservices

Dataservices

RLC

Physical layer

Multiplexing and QOS

Layer 3

Layer 2

Layer 1

MACsublayer

Upper layersignaling

LACsublayer

Voiceservices

Dataservices

Physical layer

Multiplexing and QOS

Layer 3

Layer 2

Layer 1

SRBP

RLP

(a) (b)

RLP RLP

MACsublayer

Figure 8.2 3G standards protocol stacks: (a) WCDMA and (b) CDMA2000 protocol stacks.

Voice Data Signaling Data Signaling

TrCh multiplexing

Physicaionl channel segmentation

PN scrambling

Channelization

TrCh TrChTrCh

TrCh

Voice

Multiplexing

Channeliztion

PN scrambling

PhCh PhCh PhCh

WCDMA CDMA2000

Physicallayer

Channeliztion Channeliztion

Figure 8.3 Physical layer interaction with upper layers.

and CDMA2000 physical layers stem from this difference in basic approach.The WCDMA standard allows the time-multiplexed data stream to be splitamong multiple code channels; however, data streams passed from higher layersare always time multiplexed into a single stream first. In almost all knowndeployments, this multicode option is not used and the time-multiplexed data issent over a single code channel.

8.3.2 Transmit Chain UL/RL PN Scrambling/Spreading

QPSK modulation is used for both WCDMA and CDMA2000 air interfaces.QPSK is a four-quadrant signaling pattern, where each signal is represented by asingle point in the quadrant. All possible transitions between constellationpoints are allowed between successive QPSK symbols. HPSK is a variation ofthis, where only certain transitions are allowed. In HPSK, on alternate symbols,180° phase transitions, otherwise known as zero-crossings, are not allowed;these symbols can only move to adjacent quadrants (90° phase transitions). Thishelps reduce the waveforms power peak to average ratio. In WCDMA, goldcodes are used for scrambling, where each MS is assigned one of 224 codes by thenetwork. CDMA2000 employs M-sequences for spreading. Each MS calculatesa mask applied to the long code (of period 242 - 1). The masked long code isXORed with the CDMA2000 I-Q short codes.

8.3.3 DL/FL Physical Channel Formatting

A frame is the basic building block of a physical channel. In WCDMA, a frameis 10 ms long, whereas a CDMA2000 frame is 20 ms long. The frame structuresof the forward and reverse links of the dedicated channels of both CDMA2000and WCDMA are shown in Figure 8.4. In WCDMA, a downlink (forwardlink) frame consists of data bits: transport format combination indicatior (TFCI)bits, used to signal the data format to avoid the need to perform blind transportformat detection; time-multiplexed power control (TPC) bits; and dedicated pilotbits. The TPC bits are essentially the inner loop power commands that informthe transmitter to increase or decrease its power. In WCDMA, channel estima-tion is based on the primary CPICH and can be improved if the dedicated pilotbits are also included, resulting in a better SIR estimation for downlink powercontrol. The most important function of the dedicated pilot bits is their use inuser-specific beamforming, which will be explained in detail later. In addition toTFCI, TPC, and dedicated pilot bits, a WCDMA uplink frame also includesFBI bits. The FBI bits are used only with closed loop transmit diversity to enablethe MS to transmit the necessary set of weights back to the base station based onthe quality of channel estimation. The major difference between the downlinkand uplink frames is that data is time multiplexed with the control bits on the


downlink but sent on the in-phase branch on the uplink, whereas control bits(signaling) are sent on the quadrature-phase branch. In CDMA2000, a forwardpower control subchannel is punctured into the traffic channel and carries theinner loop power control commands, whereas on the reverse link the data bitsare sent on the I branch while power control and reverse link pilot bits are senton the Q branch. Note that the dedicated pilot and TFCI bits improve the sys-tem performance at the expense of increased overhead. This overhead is smallfor high data rates; however, at low data rates (e.g., voice), these add significantoverhead. This necessitates a larger share of the OVSF code space per user. Ifmost users in a cell have favorable downlink conditions, the cell could run out ofOVSF codes even before the cell’s maximum power limit is reached. Additional


Data 2 TPC TFCI Data 2 Pilot

WCDMA 10 ms frame

0.66 ms slot

Data Data

CDMA2000 20 ms frame

1.25 ms power controlgroup (PCG)

FL power controlsubchannel

WCDMA 10 ms frame

0.66 ms slot

CDMA2000 20 ms frame

1.25 ms power controlgroup (PCG)

Pilot TFCI FBI TPCData

DataPilot PC

IQ

IQ

(a)

(b)

Figure 8.4 Physical layer frame structures. (a) FL/DL frame structure. (b) RL/UL frame structure.

OVSF codes can be assigned on one or more secondary scrambling codes, butthese would cause undesirable cross-interference with users on the primaryscrambling code.

8.4 Mobile Call States

8.4.1 WCDMA

WCDMA provides several connected mode substates that conserve air interfaceresources by freeing up physical layer resources. In these substates, the layer 3connection context is preserved in times of low traffic activity to accommodatethe bursty nature of packet data traffic. Transition into these substates is con-trolled by the network, and each substate reflects a progressively lower level oftraffic activity. These include:

• CELL_DCH: dedicated traffic channels are allocated to the MS.

• CELL_FACH: dedicated traffic channels are released. The MS continu-ously monitors the FACH. There is no transmission on the uplink, anda random access procedure is needed to get back into CELL_DCH.

• CELL_PCH and URA_PCH: dedicated traffic channels are released,and the MS operates in a discontinuous receive (DRX) mode to pre-serve battery life (same as idle except the layer 3 connection context ispreserved). URA_PCH typically requires less frequent updates thanCELL_PCH by the MS to notify the network that its location haschanged. The different call states are shown in Figure 8.5, whereas theprocedure by which the mobile station requests access to the systemover the RACH channel is illustrated in Figure 8.6.

8.4.2 CDMA2000

CDMA2000 provides for layer 3 connection preservation, though it is notreflected in the call states defined in the air interface specification. Dedicatedchannels are allocated to the mobile when it is in the mobile control on the traf-fic channel state. Layer 3 connection preservation is handled independently bythe upper layers. The mobile transitions to idle state when the upper layers are inthe preserved state. In the traffic channel substate (control hold mode), dedi-cated channels are still allocated and the mobile continuously monitors them,but the only reverse link transmission is a gated pilot. For the most efficient useof system capacity, the forward link dedicated channel would be DCCH operat-ing in discontinuous transmission (DTX) mode. This substate uses more systemcapacity than the CELL_FACH state of WCDMA because of the gated reverse


link pilot; however, dedicated operations can be resumed more quickly becausea random access procedure is not required. The CDMA2000 access procedure isshown in Figure 8.7.

8.5 Mobility Procedures to Support High-Speed Data Transfer

During periods of low or no data activity, dedicated high-speed data resourcesare released. To reestablish high-speed data transfer, the network must keeptrack of the location of the MS, as described next.


URA_PCH CELL_PCH

CELL_DCH CELL_FACH

Idle mode

WCDMA connected modes

Traffic channelsubstate (activemode)

Release substate

Mobile idle state

CDMA2000 MS control on traffic channel state

System accessstate

Mobile stationinitialization state

Releaseconnection

Releaseconnection

Establishconnection

Establishconnection

(a)

Traffic channelsubstate (controlhold mode)

Traffic channelinit substate

(b)

Figure 8.5 Mobile station call states: (a) WCDMA, and (b) CDMA2000.


Start

Max preamblecycles?

L1 PRACH procedure

Transmitmessage

Persistence pass?

ACK received?Wait10 ms

Wait10 ms

Waitbackoffinterval

Tx unsuccessful

End

Wait10 ms

Yes

No

No

Yes

Yes

No NACK

Figure 8.6 WCDMA access procedure.

Start

Max attempts?

Transmit preambleand message

Done

Persistence pass?

ACK received?Waitbackoffinterval

Tx unsuccessful

End

Wait20 ms

Yes

No

No

Yes

Yes

No

Figure 8.7 CDMA2000 access procedure.

8.5.1 Cell_FACH State or Control Hold Mode

In WCDMA Cell_FACH state, UTRAN keeps track of the location of the MSby using the cell update procedure. Whenever it moves to a new cell, the UEsends a cell update message on the RACH. In CDMA2000 control hold mode,a dedicated channel is allocated to the UE (either FCH or the discontinuousDCCH). Traffic channel handoff procedures allow the network to know inwhich cell or cells the UE is located.

8.5.2 Idle, Cell_PCH, or URA_PCH States

In WCDMA idle state, the UE performs routing area updates when it moves toa new routing area. To reestablish dedicated channels, UTRAN must page theUE over all cells of a routing area. If the UE is in Cell_PCH state, the cellupdate procedure allows UTRAN to page the mobile in a single cell. InURA_PCH state, the URA update procedure allows UTRAN to page themobile over all cells in a UTRAN registration area (URA). In CDMA2000 idlestate, the network must page the UE according to its registration and pagingpolicy (e.g., if zone-based registration is used, a page is sent over the lastregistered zone).

8.6 Procedures to Reestablish High-Speed Data Transfer

8.6.1 Cell_FACH State or Control Hold Mode

In WCDMA Cell_FACH state, UTRAN sends the radio bearer setup messageon the FACH common channel to assign high-speed dedicated channels. InCDMA2000 control hold mode, the network sends a message on the dedicated,discontinuous DCCH to assign high-speed dedicated channels.

8.6.2 Idle Mode, Cell_PCH, or URA_PCH States

In WCDMA idle state, the signaling is similar to initial call establishment (RRCconnection, authentication, and security procedures). In Cell_PCH andURA_PCH, the RRC connection, authentication, and security procedures arenot required because the RRC connection is preserved in these states. InCDMA2000 idle state, signaling is similar to initial call establishment.

8.7 Packet Data Services

Because of differences in their respective protocol stacks, the WCDMA andCDMA2000 standards employ different packet scheduling approaches, as out-lined next.


8.7.1 WCDMA Approach

Channel Allocation

The RNC assigns a DPCH to the UE, which carries both data and signaling.The peak DPCH data rate determines the OVSF code space required. TheRNC reconfigures the DPCH to a lower peak rate when less data needs to besent or when other users have higher priority [11]. The UE sends traffic volumemeasurement reports to the RNC to inform the RNC of the amount of uplinkdata queued in the UE. Reconfiguration of the signaling channel is requiredevery time the data channel configuration needs to be changed. The DPCH datarate can be changed on a TTI basis. This reduces the overall transmit powerused by DPCH. Each transport channel has its own QoS, which includes aBLER target. However, because transport channels are multiplexed together onthe same physical channel, the required SIR is effectively the same for all of thetransport channels.

8.7.2 CDMA2000 Approach

Channel Allocation

The base station scheduler assigns an SCH to a mobile on an as-needed basis.The SCH data rate is a function of loading, scheduler policy, data buffer depth,and channel condition. On the reverse link, the mobile requests a channel, andthe base station determines what rate to allow the mobile to use. In release 0,when an SCH is assigned, the data rate is fixed for the duration of the assign-ment. Only the assigned rate or zero rate is allowed. In release A,frame-by-frame variable data rates are supported. Because the data and signalingcan be sent on separate channels, the power used is variable. The SCH has inde-pendent power control from the FCH/DCCH.

8.8 Pilot Channels

In CDMA, a common pilot channel is broadcast throughout a sector to providecell identification, phase reference, and timing information to the mobile sta-tions. When a sector is subdivided into multiple narrow fixed beams, a commonpilot channel cannot be used for channel estimation because the reference signal(pilot) used for channel estimation must go through the exact same path(including antennas) as the data (traffic); consequently, each antenna beamrequires a separate pilot.

8.8.1 CDMA2000

In addition to the common forward pilot channel, the following pilot channelsare specified in the CDMA2000 standard:


Forward Transmit Diversity Pilot Channel

If transmit diversity is supported on the forward CDMA channel, then thetransmit diversity pilot channel is used and is spread using Walsh code W16

128 and

is transmitted at a power level of 0, –3, –6, or –9 dB relative to the power levelof the forward pilot channel.

Auxiliary Pilots

When antenna arrays and antenna transmit diversity schemes are employed onthe forward link, a separate pilot for channel estimation and phase tracking isneeded. In CDMA2000, such pilots, referred to as auxiliary pilots, are codemultiplexed with other forward link channels and use orthogonal Walsh codes.Since a pilot contains no data, auxiliary pilots may use a longer Walsh sequenceto lessen the reduction of orthogonal Walsh codes available for traffic.

Generating Auxiliary Pilots

Code-multiplexed auxiliary pilots are generated by assigning a different Walshcode to each auxiliary pilot. This results in a reduction of the number oforthogonal codes available for traffic channels. To overcome this limitation, thesize of the Walsh code set used for auxiliary pilots is expanded. Any Walsh codeW m

n denoting the mth Walsh code of length n can be used to generate Nw auxil-

iary Walsh codes, where Nw must be a power of 2. A longer Walsh sequence isbuilt by concatenating Nw timesW m

n where each concatenatedW mn may have a

different polarity. The sequence of polarity is selected such that Nw additionalorthogonal Walsh sequences of order Nw× n are generated. For the case Nw = 2,the two possible Walsh functions of order 2 × n areW m

nW mn andW m

nW mn , where

the overbar denotes a polarity change. The CDMA2000 standard specifies 512as the maximum length of the Walsh functions that is allowed for Walsh func-tion spreading or quasi-orthogonal function spreading of an auxiliary pilot.Obviously, Walsh function W n

0 cannot be used to generate auxiliary pilots

because they would interfere with the common pilot channel; in addition, con-catenation of certain Walsh functions such as W16

128 is not allowed by the stan-

dard. Since Walsh sequenceWin is orthogonal to all otherW j

n only if j ≠ i, it

follows that if WalshWin is used to generate an auxiliary pilot it cannot be used

by another traffic channel.

Auxiliary Transmit Diversity Pilot Channel

When transmit diversity is supported on the forward CDMA channel associatedwith an auxiliary pilot channel, then an auxiliary transmit diversity pilot channelspread with a Walsh function or a quasi-orthogonal function is used.


8.8.2 WCDMA

Common Pilot Channel

The CPICH is a fixed rate downlink physical channel that carries a continuouspredefined bit/symbol sequence. The spreading factor of the CPICH is 256 andthe rate of the CPICH is 30 Kbps. If transmit diversity is used in the cell thenthe CPICH is transmitted with both antennas, where a different pilot sequenceis used for the second antenna. There are two types of CPICHs, the primary andsecondary CPICH. They differ in their use and the limitations placed on theirphysical features [12].

Primary Common Pilot Channel

The primary common pilot channel (P-CPICH) is always spread withchannelization code Cch.256.0 and scrambled with the primary scrambling code ofthe cell. Hence, the P-CPICH can be used in the mobile to determine thescrambling code used for scrambling the downlink channels of the cell. TheP-CPICH is broadcast over the entire cell and there is always only one uniqueP-CPICH per cell. The P-CPICH is always the phase reference for the SCH,P-CCPCH, acquisition indication channel (AICH), and PICH, and it is thedefault phase reference for all other downlink physical channels [1].

Secondary Common Pilot Channel

The secondary common pilot channel (S-CPICH) can be spread by anychannelization code of length 256 and can be scrambled by either the primary ora secondary scrambling code. There may be zero, one, or several S-CPICHs percell. A S-CPICH may be transmitted over the entire cell or only into a part ofthe cell. The S-CPICH can be the phase reference for S-CCPCH and DPCH. Ifthis is the case, the terminal is informed by higher layer signaling [5].

8.9 Channels Applicable for Downlink Beamforming

The radio interface of UMTS has been designed to allow the use of smart anten-nas. In the downlink there are basically two possibilities for beamforming. Thetwo options are to transmit an S-CPICH on the beams or not. If an S-CPICH istransmitted on a beam together with the user data, then the terminals can usethe S-CPICH for channel estimation, which will give a good channel estimate.If no S-CPICH is transmitted together with the user data then the dedicatedpilot bits have to be used for channel estimation. Hence, only physical channelswith pilot bits can use this beamforming method. Table 8.1 shows whichbeamforming method, if any, is applicable for the different physical channels.


As was discussed in Chapter 1, smart antenna strategies that do not rely oncooperation or feedback from the mobiles can be classified into fixed beam meth-ods and user-specific beamforming methods that dedicate a single beam or trans-mission pattern to each user. Figure 8.8 illustrates the difference between fixedbeams and user-specific beamforming.

8.10 Overview of Major Radio Network Algorithms

8.10.1 Power Control

The aim of the PC is to keep the UL and DL quality at the required level. ThePC algorithm consists of the inner loop PC and the outer loop PC. The innerloop PC adjusts the power every slot/PCG to fulfill the SIR target. The outerloop PC uses measurements of the quality in both RL/UL and FL/DL transmis-sion and compares the measured quality with a quality target. If the quality istoo low, the outer loop increases the SIR target, and vice versa. The quality mea-surements could be the FER/BLER/CRC but even BER and SIR might be usedin more advanced schemes. The PC measurements and targets operate on eachindividual MS in the system.


Table 8.1WCDMA Downlink Channels Applicable for Beamforming

Physical ChannelMultiple FixedBeams with S-CPICH

Flexible Beams withDedicated Pilot Bits

P-CPICH No No

P-CCPCH No No

SCH No No

S-CCPCH (carryingPCH and FACH)

No No

S-CCPCH (carryingonly FACH)

Yes Yes

S-CCPCH (carryingonly PCH)

No No

DPCH Yes Yes

PDSCH Yes Yes

PICH No No

AICH No No

CSICH No No

8.10.2 Initial Power Setting

The initial power procedure is used to set the initial power of both UL and DLtransmission. The procedure is used both by common and dedicated channels(i.e., the same procedures are used for continuous and discontinuous services).Before the random access procedure, the MS needs to find the appropriate ran-dom access transmit power. This is done by the preamble power rampingscheme. According to [3], the initial preamble power in WCDMA is calculatedby the mobile as follows

Preamble

Primary

_ _

_ _ _ _

Initial Power

CPICH DL TX Power C= − PICH RSCP

UL

_

_+ +interference Constant

(8.1)

The primary CPICH transmit power, Primary_CPICH_DL_TX_Power,the total interference on the UL, UL_interference, and the constant value, Con-stant, are all sent to the MS from UTRAN. The only parameter that is measuredby the MS is the received power of the primary CPICH by the MS,CPICH_RSCP. In CDMA2000, the corresponding initial power settings to beused on the first access probe is given by

Preamble Pr_ _ _ _ _Initial Power NOM PWR INT PWR P CNST= + − −(8.2)


(a) (b)

Figure 8.8 Fixed versus adaptive beamforming. (a) Fixed multiple beams. (b) User tracking withadaptive antennas.

The nominal correction factor for the base station, NOM_PWR, and thecorrection factor for the base station from path loss decorrelation between theFL and RL frequencies, INT_PWR, are signaled to the MS in the access parame-ters message.

8.10.3 Admission Control

The admission control (AC) algorithm accepts or rejects requests to establishnew radio links. This is done by checking if the requested amount of resources isavailable. The algorithm is located in the RNC/BSC, where relevant informa-tion from several sectors can be obtained. The radio link request is admitted ifall UL and DL AC criteria are fulfilled. The specific AC algorithms are not stan-dardized and therefore each vendor will have its own solution. In the literature afew AC strategies are proposed, which are based on the DL power, UL interfer-ence, and throughput [13–18]. A new user is admitted if the resulting total ULinterference level is lower than a predefined threshold. This could be summa-rized by the following equation:

I I IUL UL ULTh+ <∆ (8.3)

where I ULTh is the UL interference threshold, I UL is the existing total UL interfer-

ence, and ∆I UL is the UL interference increase caused by the new user, which is

a function of the bit rateE

Nb

o

and the voice activity factor.

8.10.4 Congestion Control

The purpose of the congestion control (CC) is to quickly stabilize the systemwhen an overload situation occurs. An overload situation is the result of fluctua-tions of I

UL, P

DL, which mainly are caused by fading and intercell interference or

variations in the carried traffic of a link. The CC algorithm uses parametersfrom Node B and events from the RNC as inputs. The input parameters fromNode B are the measurement of UL interference, received signal strength (RSSI),and DL transmitted power, TxCPwr. The CC may use I

UL, P

DLand code usage

for its decision. Thus, a similar reasoning as for AC can be performed. However,CC and AC are interdependent and differ in certain respects.

8.10.5 Soft/Softer Handoff

In the current CDMA standards, soft handoff between cells is based on mea-surements on the primary common pilot channel. There can only be one pri-mary common pilot per sector. When the MS is in the dedicated mode it sends


measurement reports to the base station. These messages contain any of the fol-lowing measurements:

• FL/DL RSCP (for received signal code power);

• FL/DL Ec/No;

• Path loss (i.e., distance gain, antenna gain and fading).

The measurement type and the sectors to be measured are decided by thebase station (UTRAN). The MS will enter the soft handoff state when, forexample, the measured DL Ec/No from a new BS is within the configuredhandoff margin.

8.10.6 Hard Handoff

Hard handoff (HHO) is when the system executes a handoff by removing alllinks in the active set and establishing a new radio link. There are several differ-ent types of HHO:

• Inter BSC/RNC HHO;

• Interfrequency HHO;

• Intersystem HHO.

All HHO types employ the same measurements as the SHO (i.e., Ec/No,RSCP, and path loss). The exception is the intersystem HHO in WCDMA,where only RSSI received signal strength is used.

8.11 System Impact of Advanced Spatial Techniques

While most of the smart antenna literature has focused on algorithm/architec-ture development and performance analysis of using antenna arrays, the interac-tion between the previously described radio network algorithms and advancedantenna systems concepts [19] has received less attention. In this section, theimpact of smart antenna architectures, namely transmit diversity andbeamforming on power control, admission control, and handoff, as well as otherradio network algorithms, are analyzed and discussed.

8.11.1 Transmit Diversity

In both the WCDMA and CDMA2000 standards, a number of transmit diver-sity schemes are defined, as described in an earlier chapter. The schemes can be


divided into open and closed loop schemes. The open loop transmit diversityschemes, STTD and STS, use a space-time code in FL/DL without any MSfeedback. The closed loop schemes, on the other hand, optimize the link byadjusting the phase and amplitude of the transmission according to a feedbacksignaling message. Antennas used for transmit diversity should be placed as farapart as possible to experience as low correlation as possible between the anten-nas. The impact of transmit diversity on different radio network algorithms isdiscussed next.

8.11.1.1 Initial Power Setting

The traffic channels in TD are optimized for each individual user but the initialpower setting is based on measurements of the common pilot. Therefore, theinitial power setting algorithm will not be affected.

8.11.1.2 Admission Control

Similar to the single antenna case, the base station can just average the downlinkpower of both antennas and then use this single measurement in the AC algo-rithm (i.e., the FL/DL power is split equally). Therefore, admission control isnot affected by transmit diversity.

8.11.1.3 Code Allocation

With transmit diversity and favorable channel conditions, the sector capacity isincreased. This could lead to situations where the capacity becomes limited bythe number of available orthogonal codes. In WCDMA, this problem is solvedby allocating a second PSC code with a corresponding group of orthogonalcodes, although this second group will cause some interference with the group oforthogonal codes associated with the first PSC. Similarly, CDMA2000 over-comes this problem by allocating additional codes using QOFs.

8.11.2 Fixed Beam Approach

In this scheme, the fixed beams transmit only channels carrying user-specificdata, while the common channels like P-CPICH, SCH, P-CCPCH, and so onfor WCDMA or pilot, paging, and sync channels in CDMA2000 are transmit-ted into the entire sector using a single antenna element of the antenna array.Typically, the data sent on the different beams is scrambled with the samescrambling code [WCDMA] or assigned the same PN offset; [CDMA 2000]therefore, the channels sent on the different beams are orthogonal to each other.The receive power of the common pilot channel at the mobile determines theserving sector. Within the serving sector, the mobile will be served with thebeam with the highest uplink/reverse link receive power.


8.11.2.1 Additional Equipment at the Base Station

The beamforming for the multiple fixed beam method can be done either with apassive beamforming network, as in Figure 8.9, or digital in baseband, as illus-trated in Figure 8.10. In the passive beamforming network implementation, thefollowing components are needed:

• A signal processing unit capable of processing the signals of the beamsand the common channels sent/received by the sector antenna.

• One transceiver chain per beam plus an additional transceiver for thecommon channels transmitted and received in/from the entire sector.

• A beamforming network and an antenna array per sector.

In case of digital beamforming at baseband, the following componentswould be needed:

• A signal processing unit that is capable for handling all beams plus thecommon channels sent/received with the sector antenna.


Uplink/downlinkpassivebeamformingnetwork

Basebandprocessingfor multiplefixed beams andsector beam

Transceiver

Transceiver

Transceiver

M elements Nb beams

Calibration needed

Figure 8.9 Passive networks fixed multiple beam architecture.

• An antenna array per sector.

• A transceiver chain per antenna element of the antenna array is needed.

8.11.2.2 Scrambling Code/PN Offset Assignment

If there are so many users in a sector that the number of required orthogonalcodes exceeds the codes offered by an OVSF code tree, then a new code tree isused and the new channels are scrambled with another secondary scramblingcode up to 15 secondary scrambling codes per cell, as supported in WCDMA.CDMA2000 uses QOF for the same purpose. These secondary scramblingcodes/QOF are not orthogonal to each other and the PSC/short PN sequence,so it is beneficial if all data channels transmitted with one beam are scram-bled/spread with the same scrambling code/PN offset. To reduce the interfer-ence from the common pilot and the control channels that are sent into thewhole sector and are always scrambled/spread with the PSC/PN offset, as manybeams as possible should use the PSC/PN offset. If a secondary scramblingcode/QOF has to be used it should first be assigned to the outermost beam thatis next to the border of a sector because the neighboring sector always usesdifferent codes.


Basebandprocessing anddigitalbeamformingfor multiplefixed beams plussector beam

Transceiver

Transceiver

Transceiver

M elements Nbbeams

Calibration needed

Figure 8.10 Digital baseband fixed multiple beam architecture.

8.11.2.3 Channel Estimation

By default, mobiles in a CDMA system use the P-CPICH/forward pilot channelfor channel estimation. In the fixed beam method the P-CPICH/forward pilotchannel are transmitted with a single antenna element, whereas the data (trafficchannels) are sent on a beam. Since the data channel and the P-CPICH/forwardpilot channel experience different propagation and multipath channels, thecommon pilot channel cannot be used for channel estimation. The UMTS spec-ification allows the signaling of the mobile via higher layer protocols to useeither the dedicated pilot bits that are time multiplexed with the user data or touse an S-CPICH that is transmitted with the same beam as the user data, forchannel estimation. The S-CPICH has the same structure as the P-CPICH butis spread with a different channelization code and scrambled with the samescrambling code as the data channels on the beam. There are two drawbacks forusing the dedicated pilot bits for channel estimation in the multiple fixed beamcase. First, since the dedicated pilot bits are transmitted with low energy com-pared with a P-CPICH or S-CPICH, the channel estimation in the mobile willbe worse, which would in turn increase the Eb/No to achieve the required servicequality [20]. Second, since a single beam serves more than one user it is morebeneficial to use S-CPICHs. One drawback of using the S-CPICH is that someof the base station transmit power will need to be allocated to the pilots associ-ated with every beam (e.g., 6 dB below the P-CPICH), thereby increasing theoverhead and reducing the power available for dedicated channels. For a smallnumber of beams this should be offset by the increased capacity inherent inusing beamforming. The spreading and scrambling code of the S-CPICH to usefor channel estimation is signaled to the mobile via higher layer messages gener-ated at the RNC. In UMTS it is not possible that a mobile uses two differentS-CPICHs from one sector as phase reference, so each mobile can be served onlyby one beam per sector. In CDMA2000, each beam can be assigned a code mul-tiplexed auxiliary pilot, which decouples the pilot from the actual traffic userdata being sent. Hence, the pilot reference is not bound to a particular user datastream. This permits multiple mobile stations to be placed in the same fixedbeam using a common auxiliary pilot.

8.11.2.4 Beam Transition

For scenarios where the Node B uses multiple fixed beams with one S-CPICHassigned per beam, the mobile’s active set is determined by measuring themobile’s uplink receive power in the different beams. In order for the mobile totransition from one beam to another, it needs to get informed that it should useanother S-CPICH. This requires higher layer signaling between Node B andRNC. For each MS, the Node B can measure the uplink-received power of thepilot symbols in all the beams where an S-CPICH is assigned. These measure-ments are averaged in the Node B and then reported to the RNC. Based on


these measurements, the RNC determines whether a beam transition or handoffis needed or not. Hence, the beam handoff algorithm can be implemented incoherence with the conventional sector handover algorithms.

8.11.2.5 Soft-Handoff Procedures in Fixed Beam Mode

From an operational perspective, each code multiplexed auxiliary pilot used on aparticular sector can be part of the sector’s neighbor list. The mobile station willthen search for that auxiliary pilot in a similar way to the search for other sectors’pilots (the mobile station will search for a different Walsh pilot instead ofsearching for other PN offsets). The mobile station can report the auxiliary pilot(that is, the fixed beam) using the same procedures as used for reporting anotherregular sector. In this way, mobile stations can go in and out of the coverage ofthe beam in a similar way to the soft-handoff mechanism between sectors. Thisallows placement of mobile stations in the coverage of the beam when interfer-ence conditions require it. This extra capability can be made optional by notinstructing the mobile stations to search for auxiliary pilots (e.g., for individualbeam mode).


When an MS moves through a cell it will change the DL beam depending onthe position. This may result in a change of the antenna and channel gain. How-ever, it is likely that the MS cannot separate the beam change from ordinary fastfading and no specific problems will arise.

8.11.2.7 Preamble Initial Power Setting

The initial power algorithm will be affected by the introduction of a fixed beamconcept. To explain how, first the UL and then the DL initial power algorithmis examined.

Uplink Preamble Power Setting

To calculate the correct UL initial preamble power value for a fixed beam sys-tem, the MS ideally needs to measure the UL gain and the corresponding ULinterference per beam for the cell. This will enable the MS to correctly estimatethe necessary initial preamble power. The 3GPP standard states that the MSmust measure the DL gain on the primary CPICH (see 8.1). A solution to thisproblem could be to have the MS measure the received power of the primaryCPICH transmitted by the wide beam, RSCP. However, this leads to some dif-ference in the interference level due to the difference between the wide beamgain and the narrow beam gain. If the total RSSI at the base station is used forUL interference measurement, the measured UL interference is higher than theinterference in a separate beam (due to the spatial suppression/reduction of theinterference in the UL). Thus, there will be some difference between the wide


beam UL interference and the narrow beam UL interference. This differencecan easily be compensated or accounted for by adjusting the constant in (8.1)both for the antenna gain difference and the UL interference difference.

Downlink Initial Power Setting

The DL initial DCH power setting assumes that the MS is able to measure theSIR on the downlink primary CPICH and then transmit the measurement tothe UTRAN. Even though the gain from the wide beam may differ from thenarrow beam gain, the DL interference measured by the MS is correct. If thedifference between the wide beam gain and the envelope of the narrow beamgain is known (equal for all angles or as a function of the angle), this can beexploited by the UTRAN. The power setting algorithm simply adjusts the initialpower with the gain difference.

8.11.2.8 Reverse Link/Uplink Call Admission Control

When the base station receives a new call setup request from a mobile station, itchecks its own resources and then the call is either admitted if there are availableresources or otherwise blocked. System performance in terms of stability is typi-cally measured using blocking and dropping probabilities, that is, the number ofblocked calls and dropped calls over the total number of call attempts. Blockingoccurs when SINR < SINRreq, whereas existing calls are dropped when the SINRdrops below a certain threshold (i.e., in the reverse link/uplink, a conventionalCDMA network estimates the load of a sector cell with the help of the wideband received power at the base station $I tot , which reflects the interference at the

base station). Let the uplink load ηUL of a sector be given as a percentage of thepole capacity (i.e., 0≤ηUL≤1) and the noise rise be given by Nrise where

NI

risetot

N

=$

σ 2(8.4)

and σ N2 is the background and receiver noise. The relation between the sector

load and noise rise is given by

N riseUL

=−

1

1 η(8.5)

As new mobiles attempt to get admission to the system, the interferenceincreases due to the new user, ∆I based on the actual sector load, and therequired Eb/No for the requested service. Admission is granted as long as theexpected resulting interference after admission of the user does not exceed a cer-tain threshold.


I I Itot th+ ≤∆ (8.6)

The value of the threshold is up to the system designer but the thresholdshould be chosen so that at the corresponding load a small increase in systemload does not cause any tremendous interference increase. From Figure 8.11, areasonable threshold for Nrise is around 6 dB. In [21] Hara suggests a new admis-sion control scheme on the reverse link that takes into account adaptivebeamforming at the base station. The algorithm’s flowchart is shown in Figure8.12 and can be summarized as follows:

• When a call is attempted, the mobile’s ID is retrieved and the corre-sponding steering vector/matrix A is calculated.

• The interference plus noise correlation matrix RN is then calculated andthe SINR for the new user is estimated as SINR = σ s

2 AHR N−1 A, where σ s

2

is the signal power.

• The estimated SINR is then compared with the required SINR and thecall is admitted if SINR > SINRreq, otherwise, the call is blocked.

8.11.2.9 Radio Resource Management

The radio resource management includes admission control, load control,packet scheduling, handoff control, and congestion control and is implementedin the RNC/BSC, as the RNC/BSC has knowledge of the load of all sectors inits radio network subsystem. The task of the admission control is to ensure thatthe network stays stable after a new user has been granted access to the network.The handoff control checks if the system stays stable after any update of amobile’s active set, whereas the task of the packet scheduler is to adjust the data


Figure 8.11 Interference rise over thermal noise.

rates of the nonreal-time services so that the system is operating as close as possi-ble at its maximum throughput.

8.11.2.10 Forward Link/Downlink Admission Control

For the downlink/forward link, it is well known that the total transmit power ofa base station, Ior, is a good measure of the load of a CDMA sector [17]. A newmobile is granted access if the estimated average transmit power of a sector afteradmission of the new mobile stays below a certain threshold

I I Ior or orth+ <∆ (8.7)

where I orthis a function of the power amplifier capacity, power control dynamic

range, and power allocated to common channels. The increase of the transmitpower of the sector after admission of the new user, ∆Ior, is estimated using themeasured path loss to the new mobile, the interference at the mobile, and therequired Eb/No of the requested service. The same procedure is also followed forhandoffs. Similarly, the task of the packet scheduler is to adjust the data rate ofthe downlink nonreal-time services so that the actual total downlink transmitpower of the cells is as close as possible to I orth

. The same holds true for theuplink, where the task is to send as much data as possible so that the maximum


Receive a new access probe / RACH message

Retrieve mobile station ID

Calculate array steering vector

Estimate SINR for new MS

Estimated SINR >SINRreq?

Admit newcall

Block newcall

Yes

No

Figure 8.12 Uplink call admission control.

sector load is reached as close as possible while not exceeding the maximumallowed sector load. Since the capacity gains from using adaptive antennas ishighly dependent on the spatial interference distribution, which in turn dependson the user’s spatial distribution and data rates, the AC algorithm applied inconjunction with adaptive antennas must also take the directions of the mobilesinto account, as suggested by Pedersen in [22–24]. From Chapter 5 we knowthat the antenna array power pattern is given by

( ) ( )I I I Gor orH

orθ θ= =A RA (8.8)

where G (θ) is the normalized antenna array gain.As we described earlier, the common channels are all transmitted from a sec-

tor wide beam, whereas the dedicated traffic channels are transmitted from direc-tional narrow beams. In the multiple fixed beam case, we will assume that everyMS is served by only one beam on the downlink and an auxiliary pilot/S-CPICHis transmitted on each beam. Note that the common pilot power, auxiliarypilot/S-CPICH power, and overhead common channels powers are constant,whereas the dedicated traffic channels are power controlled. Assuming we have Nb

beams, we can express the average transmit power per beam as

I P Por nSCPICH

nDPCH

n

tot= + for WCDMA (8.9)

I P Por nAux Pilot

nTraffic

n

tot= +. for CDMA2000 (8.10)

It then follows that the total base station transmitted power in all Nb

beams is given by [24]

( )I MG Por orbeams common= +θ I (8.11)

where

[ ]I orbeams

or or orN

TI I I

b= 1 2 L (8.12)

Hence, a directional power-based AC algorithm can be formulated as fol-lows, the load/power transmitted in every beam is monitored, and a new user isonly admitted to the beam if the following condition is satisfied:

( ) [ ]I orbeams

n th bI n Nθ < ∀ ∈ 1 2, , L (8.13)

That is, before a mobile is admitted, the power increase in all beams mustbe calculated to ensure that the above condition is met to grant admission. This


is due to the fact that in CDMA when a new user is admitted to a sector/beamthe downlink interference level in the system is increased and, consequently, thetransmit power in all other beams will also increase to overcome the interfer-ence. In a conventional CDMA network, when an MS is denied admission to asector because it violates (8.7), the sector capacity cannot increase. When thedirectional power-based AC is applied in conjunction with adaptive arrays, anMS at some AOA θ1 can be denied admission in beam number i if it violates(8.13) but another MS at a different AOA θ 2 requesting admission into beam j,

i j could satisfy (8.13) and gains admission. This is becauseE

Ic

or1

, the required

dedicated channel power to support MS at θ1 , might be much greater thanE

Ic

or2

,

the power necessary to support MS at θ 2 . The admission control algorithm canbe summarized in the following steps.

First, the current system resources [e.g., the base station transmit powerlevel (load)] need to be estimated. The second step involves estimating theamount of transmit power that would be required to support the new user aswell as the amount of power increase needed to support existing users. The rea-son for the power rise is that admitting the new user will raise the interferencelevel in the sector. Let us assume that the new user requests admission to beamnumber k, then an estimate of the power required for each beam can beexpressed as

I P P P nor new nPilot

nTraffic

p n newtraffic

n

tot

, ,= + ⋅ +∆ for = k (8.14)

I P P n kor new nPilot

nTraffic

p nn

tot

, ,= + ⋅ ≠∆ for (8.15)

where ∆ p n, is a factor that expresses the increase in transmit power for existingusers after admission of the new user given by

( )∆ p n

b

t

b

t

E

N PGE

N PGG

, =+ ⋅

+ ⋅

11

2

11

2θ

(8.16)

To estimate the transmit power required to support the new user, we start

with the required trafficE

Nb

t

expressed as


( )E

NWR

PMLG

N I Ib

t Traffic

Traffic

th oc sc

= ⋅ ⋅

+ +θ

(8.17)

Rearranging the above equation we get the following expression for thetransmit power for the new user

( )P

E

NR

W

N I I

MLG

E

newTraffic b

t Traffic

th oc sc=

⋅ ⋅

+ +=

θ

( )b

t Traffic

th oc sc

NR

W MG

N I I

L

⋅ ⋅ ⋅

+ +1

θ

(8.18)

where M is the number of antenna elements and L is the path loss.In CDMA systems, the pilot is continuously measured by the MS and the

measurement is sent to the base station in the measurement report message/pilotstrength measurement message, either on a periodic basis or when specific eventsare triggered. The actual pilot measurement is expressed as

E

NRSCP

RSSIP L

N I IIc

o Pilot

Pilot Pilot

th oc sc

= =

+ += or

tot c

or Pilot th oc sc

E

IL

N I I⋅

⋅

+ +(8.19)

8.12 Beam Steering/Adaptive Beamforming

Beam steering is a user-specific beamforming method, where each user is servedwith an individual beam. In the multiple fixed beams method, users are servedwith the beam with the lowest path loss. Beam steering produces a unique beamfor each user in order to transmit the signal for a user only into the directionwhere the signal experiences the lowest path loss while traveling to the user whilesimultaneously keeping the transmit power into other directions as low as possi-ble. However, this only applies to dedicated traffic channels, although the com-mon channels still have to be transmitted in the entire sector with a singleantenna element of the antenna array. Just as it is done in a conventional UMTSsystem, the serving sectors are determined through measurement of theP-CPICH quality at the mobile. The scrambling code assignment for beampointing is analog to the switched beam method. For radio resource manage-ment, an analog algorithm for the switched beam method can be used where(4.10) is checked for a set of equidistant directions. Figure 8.13 illustrates theequipment needed for user-specific beamforming, including:


• A signal processing unit capable of serving, calculating, and applyingappropriate antenna weights for all users, plus common channels sent/received into/from the entire sector.

• An antenna array.

• A power amplifier per antenna element of the antenna array.

8.12.1 Channel Estimation at the Mobile

Like in systems that use the fixed beam scheme, the mobiles served by a user-specific beam cannot use the common pilot channel for channel estimationbecause this pilot sent with a single antenna element experiences a differentradio channel than that experienced by the user data sent with a narrow beam.When user-specific beamforming like beam steering is used in a UMTS system,then each MS is served by an individual beam. Due to the high number ofbeams and the fact that only a single mobile is served by each beam, it is noteffective to transmit an S-CPICH on every beam because this will significantlyincrease the downlink power allocated to overhead, thus reducing the system


Basebandprocessing anddigital beamformingfor multipleusers plus sectorbeam

Transceiver

Transceiver

Transceiver

M elements

Calibration needed

Figure 8.13 Adaptive beamforming architecture.

capacity. The solution for this is to use the dedicated pilot bits for channel esti-mation. The use of the dedicated pilot bits causes a degraded channel estimate atthe mobile as the energy in the dedicated pilot bits compared with the energy ofthe CPICH is lower. Therefore, the mobiles using the dedicated pilot bits forchannel estimation will require a higher Eb/No to achieve their desired servicequality than mobiles that could use a CPICH [20]. When adaptive beam steeredchannels are employed in CDMA2000 systems, an L3 message directs themobile station to a beam steered channel where an auxiliary pilot channel will beused by the mobile station to coherently demodulate the traffic channel. Themobile station continues receiving the transmission on the beam steered channeluntil an L3 message directs the mobile station to a nonbeam steered channel orterminates transmission on the beam steered channel.

8.12.2 Advantages and Disadvantages

The advantage of beam steering is that the transmit power is essentially concen-trated toward the desired users. Therefore, beam steering should provide ahigher capacity gain than the fixed beam methods. In addition, as the servingbeam tracks the mobile, there is no need to transition the mobile from beam tobeam while it is traveling through the coverage area of the sector, like it is neces-sary in fixed beam schemes. Hence, far less signaling is needed than in fixedbeam methods, where signaling is necessary every time a mobile moves from thecoverage area of one beam to another. The main disadvantage of beam steeringis that the base station has to determine the optimum transmit direction for allactive users in the sector, which is computationally very intensive comparedwith the determination of the serving beam in the fixed beam method.

8.12.3 Uplink Beamforming

For a particular mobile station transmit power, demodulation performance ofthe reverse link can be improved by using narrow beams and increasing theantenna gain in the direction of one or more mobile stations. Adaptivebeamforming of the RL mobile station transmission is one such method. As dis-cussed in a previous chapter, in many beam-forming applications a reference sig-nal is required for adjusting the RL beam pattern. The RL pilot channelprovides the reference for adjusting the RL beam pattern for each mobile sta-tion. Unlike the FL, one RL pilot channel is used for traffic channel demodula-tion regardless of the technique used for adjusting the antenna pattern.


Unlike the case of fixed beams, no changes of antenna and channel gain areanticipated. Therefore, beam steering/adaptive beamforming techniques willimpact the current power control algorithms.


8.12.3.2 Initial Power Setting

As far as the initial power setting for an access probe/RACH, the MS will not beable to differentiate between a steered beam solution and a fixed beam solution.Therefore, the expected impact in both cases is similar.

8.12.3.3 Admission Control

The same AC concepts previously discussed for fixed multiple beam systems areequally applicable with steered beams/adaptive beamforming techniques. Forinstance, a per-beam AC algorithm based on the DL/FL power can beemployed, whereas an SINR-based algorithm can be used on the uplink.

8.12.3.4 Soft/Softer/Hard Handoff

The steering beam architecture interaction with the SHO algorithm is very simi-lar to the fixed beam case. The only difference is that it is not feasible to assignone pilot per beam if a beam is to track a single user.

8.13 Conclusion

In this chapter, system aspects of smart antennas and the interaction of differ-ent techniques with the various layers were provided. Figure 8.14 illustratesthe main system components or processes affected by smart antennas. A sum-mary of the impact of all three smart antennas techniques discussed in this


Networkplanning anddimensioning

Air interface

Receiverstructure andalgorithms

Radio network control(Resourcemanagement, powercontrol, etc...)

Smartantennas

Figure 8.14 Smart antenna integration in system design.

chapter, namely transmit diversity, fixed beamforming, and adaptive (user-spe-cific) beamforming, on various radio network algorithms is also provided inTable 8.2.

References

[1] 3GPP TS 25.211, Physical Channels and Mapping of Transport Channels onto PhysicalChannels (FDD).

[2] 3GPP TS 25.213, Spreading and Modulation (FDD).

[3] 3GPP TS 25.214, FDD: Physical Layer Procedures.

[4] 3GPP TS 25.302, Services Provided by the Physical Layer.

[5] 3GPP TS 25.331, Radio Resource Control (RRC) Protocol Specification.

[6] TIA/EIA/IS-2000.2, Physical Layer Standard for CDMA2000 Spread Spectrum Systems.

[7] TIA/EIA/IS-2000.3, Medium Access Control (MAC) Standard for CDMA2000 Spread Spec-trum Systems—Addendum 2.

[8] TIA/EIA/IS-2000.4, Signaling Link Access Control (LAC) Standard for CDMA2000 SpreadSpectrum Systems—Addendum 2.

[9] TIA/EIA/IS-2000.5, Upper Layer (Layer 3) Signaling Standard for CDMA2000 SpreadSpectrum Systems—Addendum 2.

[10] Vanghi, V., A. Damnjanovic, and B. Vojcic, CDMA2000 System for Mobile Communica-tions, Prentice Hall PTR, March 2004.

[11] Holma, H., and and A. Toskala, WCDMA for UMTS, Radio Access for Third GenerationMobile Communications, New York: John Wiley & Sons, 2000.


Table 8.2Smart Antennas’ Impact on Radio Network Algorithms

Concept Impact

PowerControl

PreamblePowerSetting

AdmissionControl

CongestionControl

SHO/HHO CodeAllocation

TransmitDiversity

None None None None None Minor

FixedMultipleBeams

Minor Minor Significant Significant Medium Significant

AdaptiveBeam-forming

None Medium Significant Significant Medium Significant

[12] 3GPP TR 25.887 V6.0.0, Technical Report Beamforming Enhancements.

[13] Kazmi, M., and P. Godlewski, “Admission Control Strategy and Scheduling Algorithmsfor Downlink Packet Transmission in WCDMA,” Proc. IEEE Vehicular Technology Con-ference, Boston, MA, September 2000.

[14] Yates, R., and C. Y. Huang, “Call Admission in Power Controlled CDMA System,” Proc.IEEE Vehicular Technology Conference, 1996.

[15] Magda El Zarki Zhao Liu, “SIR-based CAC for DS-CDMA Cellular Systems,” IEEE Jour-nal in Selected Areas in Communications, May 1994.

[16] Redana, S., and A. Capone, “Received Power-Based Call Admission Control Techniquesfor UMTS Uplink,” IEEE 56th Vehicular Technology Conference Proc., Vol. 4, 2002, pp.2206–2210.

[17] Perez-Romero, J., et al., “A Downlink Admission Control Algorithm for UTRA-FDD,”4th International Workshop on Mobile and Wireless Communications Network, 2002,pp. 18–22.

[18] Kuri, J., and P. Mermelstein, “Call Admission on the Uplink of a CDMA System Basedon Total Received Power,” IEEE International Conference on Communications, Vol. 3,1999, pp. 1431–1436.

[19] Boukalov, A. O., and S. G. Häggman, “System Aspects of Smart-Antenna Technology inCellular Wireless Communications—An Overview,” IEEE Trans. on Microwave Theoryand Techniques, Vol. 48, No. 6, June 2000, pp. 919–929.

[20] Qaraqe, K. A., and S. Roe, “Channel Estimation Algorithms for Third GenerationWCDMA Communication Systems,” IEEE VTS, 53rd Vehicular Technology Confer-ence, 2001. VTC 2001 Spring. Vol. 4, Mau 6–9, 2001. pp. 2,675–2,679

[21] Hara, Y., “Call Admission Control Algorithm for CDMA Systems with Adaptive Anten-nas,” Proc. IEEE Vehicular Technology Conference, Boston, MA, September 2000.

[22] Ramiro-Moreno, J., K. Pedersen, and P. Mogensen. “Direction Power-Based AdmissionControl for WCDMA Systems Using Antenna Arrays,” Proc. of 53rd IEEE Vehicular Tech-nology Conference, Vol. 1, May 2001, pp. 53–57.

[23] Pedersen, K. I., P. E. Mogensen, and J. Ramiro-Moreno, “Application and Performance ofDownlink Beamforming Techniques in UMTS,” IEEE Communications Magazine, Vol.41, No. 10, October 2003, pp. 134–143.

[24] Pedersen, K. I. and P. E. Mogensen, “Directional Power-Based Admission Control forWCDMA Systems Using Beamforming Antenna Array Systems,” IEEE Trans. VehicularTechnologies, Vol. 51, No. 6, November 2002, pp. 1294–1303.


9Mobile Stations’ Smart Antennas

9.1 Introduction

With the advent of mobile high-speed data applications, it is expected that thedownlink of 3G CDMA systems will be the limiting link as far as capacity isconcerned. Hence, it is important to investigate methods that can increase thedownlink capacity to cope with increasing demand. Performance enhancementsthat are possible with diversity reception in wireless systems are well known andhave been employed in cellular systems for decades on base station receivers [1,2]. Commercial implementation of some form of diversity in wireless deviceshas also been reported [3–11]. Widespread implementation of diver-sity-equipped wireless devices has not been achieved yet for several reasons. Themost important hurdle is that the implementation of diversity receptionincreases the complexity and cost of mobile stations. The second reason, the lackof widespread commercialization of diversity handsets for cellular systems,relates to the fact that the benefits that could be obtained from such devices aretechnology dependent. CDMA-based networks are well suited to benefit fromany reduction in the interference levels in the system even if only a portion ofthe mobiles are equipped with diversity receivers. The capacity improvement inthe downlink is simply proportional to the percentage of the advanced handsets(i.e., those equipped with receiver diversity). This is the case because a dualreceiver unit requires a smaller amount of base station transmit power, thusallowing more simultaneous connections for an equivalent amount of spectrumand average power constraint. The downlink capacity of a CDMA system isinversely proportional to the base station transmit power required to maintain agiven level of service to each user [12]. Under power control, the base station (ormultiple base stations when in soft handoff) adjusts the power fraction

265

transmitted to each user to maintain a given level of service. Let the energy per

information bit over total noise plus interference density be denoted byE

Nb

t

; this

is the key parameter that characterizes the performance of any digital communi-cation system. For a given transmission format and receiver design, the value ofE

Nb

t

determines the error performance of the digital link and is a function of the

channel characteristics. In CDMA systems, this value must be varied to main-tain a given error rate over a variety of channel conditions. This is accomplishedthrough outer loop power control, which sets a target value based on the desiredquality of service. Let the processing gain, defined as the ratio between the

spread spectrum chip rate and the information bit rate, be denoted by PG =WRb

.

For a given signal bandwidth, W is fixed [e.g., 1.2288 Mcps in a 1.25-MHz car-rier (for IS-95 and CDMA2000) or 3.84 Mcps in a 5-MHz carrier forWCDMA systems]. Note that the information bit rate Rb will vary dependingon the type of service (e.g., CDMA2000 voice services can use variable ratespeech coders, resulting in Rb of 9,600 bps, 4,800 bps, 2,400 bps, and 1,200bps, yielding processing gains of 128, 256, 512, and 1024, respectively, orhigher data rate applications up to 153.6 Kbps for 1XRTT and 3.1 Mbps for1XEV-DV systems). Similarly current WCDMA systems can support 64, 128,256, or 384 Kbps on the downlink. Let the total base station transmit power bedenoted Ior and the base station transmit power on the traffic or dedicated chan-nel for the ith user be denoted Ec. Then the fraction of the transmit power dedi-

cated to the ith user is given byE

Ic

or i

. This fraction varies over time subject to

power control. Finally, let SINR represent the received signal to interference +noise ratio, which is a function of the location of the mobile user within the cov-erage area and the associated channel conditions. The relationship between theabove quantities for the ith user then becomes

E

NWR

E

ISINRb

t b

c

or i

= ⋅

(9.1)

from which we can write the fraction of the transmit power dedicated to the ithuser as

E

I

R

W

E

N SINRc

or i

b b

t

= ⋅ ⋅

1(9.2)


For voice services, it also becomes a function of the voice activity factor v:

E

I

R

W

E

N SINRc

or i

b b

t

= ⋅ ⋅ ⋅

1ν (9.3)

Equation (9.3) assumes that each mobile user has only a single trafficchannel from only one sector. When mobiles are in softer handoff (with activelinks coming from sectors belonging to the same base station) or in soft handoff(the active set contains sectors from different base stations), then the total powerrequired by this user from the system perspective would depend on the handoffstate. Assuming the user has Ns sectors in the active set and is being served by Ns

sectors simultaneously, then the expression for the requiredE

Ic

or

should be modi-

fied to

E

I

R

W

E

N SINRNc

or i

b b

ts

= ⋅ ⋅ ⋅ ⋅

1ν (9.4)

The expected or average value ofE

Ic

or

then becomes

EE

I

R

W

E

NE

SINRNc

or i

b b

ts

= ⋅ ⋅

⋅ ⋅

1ν (9.5)

Since the number of active links for all data rates is limited by the base sta-tion transmit power, we can express the average number of users that can besupported in a sector or simply the sector capacity as

( )N

E I

EE

I

users

c or Tot Overhead

c

or i

max__=

−

1(9.6)

It is obvious that the smaller the fraction necessary to support each user,the higher the number of users that can be supported (i.e., the capacity of the

downlink). One way to reduce the average value ofE

Ic

or

is downlink

beamforming using antenna arrays at the base station. As we have discussed in aprevious chapter, the FDD gap in CDMA systems leads to suboptimal

Mobile Stations’ Smart Antennas 267

performance when downlink weights are obtained based on uplink measure-ments. Another approach that overcomes this problem is to use mobile stationsequipped with multiple antennas versus that of a single receiver station. Thistechnique eliminates the need for accurate downlink channel estimation at thebase station because this task can be moved to the MS, which has all the infor-mation necessary to estimate the downlink correlation matrix accurately. Itshould be noted that regardless of the number of antennas used in the receiver

design,E

Nb

t

and PG remain the same. The reduction in EE

Ic

or i

results from

the enhancements achieved in the SINR through the use of antenna arrays ordiversity reception. Let us group currently available commercial mobile stationsinto two groups, the first being handsets that support voice and moderate dataservices and the second including advanced mobile terminals that could physi-cally accommodate multiple antennas. The most practical and obvious solutionwould then be to add a second antenna to form a dual-antenna handset that hasa primary antenna used for both transmit and receive and a secondary antennaused for receive only. This secondary antenna can be designed to occupy muchless volume than the primary antenna. The secondary antenna’s small volumeallows it to be put inside the plastics of even small phones.

9.2 Multiple-Antenna MS Design

Dual-antenna receiver designs offer several gains over single-antenna designs.The first part of these gains can be attributed to diversity. Just as using twoantennas for spatial or polarization diversity improves the uplink performance,using two antennas at the MS should provide some diversity gain. Diversity gainrefers to the improvement in received signal level if the better of two receiveantennas is used (switched diversity) or the signals from both antennas werecombined, as discussed in Chapter 5. As long as the fading is not completelysynchronized between the two antennas, it is possible to improve the average sig-nal level. It is well known that antenna diversity is obtained when the antennashave different reception characteristics so that the signals received by eachantenna have low cross correlation. This can be accomplished using spatialdiversity, which requires the antennas to be separated by multiple wavelengths.Therefore, this type of diversity is largely limited to base stations and renders itimpractical to mobile stations. To understand the requirements for spatial diver-sity we need to look at the correlation between two antennas. A model for thecross correlation between two antennas separated by a distance d was derived in[13], assuming the signals DOAs have a uniform PDF. It was shown that thecorrelation between the real parts of the signals at the nth and mth antenna ele-ments Rr and that between the real and imaginary parts Ri are given by


( ) ( ) ( ) ( )R m n Jd

m n Jd

m n kkA

r k− = −

+ −

0 22 2 2 2π

λπ

λθcos

sin( )( )

S

kASk =

∞

∑1

(9.7)

( ) ( ) ( )[ ]( )

R m n Jd

m n kk

AS

i kk

− = −

+

+

+=

∞

∑2 2 2 12 1

22 1

0

πλ

θsinsin

( )

+

2 12

kAS

(9.8)

where AS is the angle spread in degrees. The envelope correlation is thus givenby

R R Rmn r i= +2 2 (9.9)

Figure 9.1 shows how the envelop correlation changes as the AS isincreased for 0° DOA. We can see that for large AS the correlation is low;


1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

00 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

Antenna separation d in wavelengths

Enve

lop

eco

rrel

atio

n

AS=180 AS=120

AS=80AS=40

AS=20

Figure 9.1 Envelope correlation versus antenna separation (d/λ), DOA = 0°.

however, zero correlation requires separations that are large compared with amobile station size.

For low AS we can observe that the correlation is quite high and diversityperformance degradation should be expected. On the other hand, this high cor-relation improves the performance of optimum combining methods, whichcould help offset some of the degradation in diversity gain. The effect of correla-tion on the performance of diversity systems with two antennas was studied in[2] and it was shown that correlations below 0.3 do not add significant diversitygain, whereas performance degradations are small up to correlations of 0.7. Fig-ure 9.2 shows the minimum required antenna separation to achieve correlationsof 0.3 and 0.5 as a function of the AS for 0°DOA.

This further confirms that improved diversity performance requires largeseparations. From Figure 9.2, a correlation of 0.5 could serve as a good trade-offbetween diversity and optimum combining, which requires high correlationbetween the signals at the antenna elements. Although spatial diversity tech-niques might not be suitable for mobile stations, other forms of diversity that donot require large separations can be employed. These include pattern or polar-ization diversity, as described in [14]. Polarization diversity makes use of the factthat different filed components experience different channel conditions, whereaspattern diversity exploits the fact that antennas with different patterns will


180

160

140

120

100

80

60

AS

(deg

rees

)

0.25 0.3 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75

d in wavelength

Envelope correlation < 0.3Envelope correlation < 0.5

Figure 9.2 Relationship between AS and minimum separation to achieve a given envelope correla-tion, DOA = 0°.

receive different paths and thus are able to achieve some level of signaldecorrelation. Possible designs might include placing the two whip antennas oneither side of a phone, using the standard whip antenna as a primary antennaand a back-mounted antenna such as a planar-inverted F antenna (PIFA) as asecondary antenna. These antenna designs would achieve low signal cross corre-lation based on their separation, polarization, or different patterns. The resultsshown in [14] confirm that such designs can achieve sufficient signaldecorrelation that some diversity gain can be obtained. The second gain compo-nent can be termed array gain. The aperture of M antennas should be about Mtimes as large as the aperture of one antenna, thus an M -element array can pro-vide a gain of M. In practice, this might not always be true in dual-antennahandsets. If the antennas are too close together, then their apertures will overlap,and the total aperture will be less than the sum of the individual apertures.Moreover, mutual coupling between the antennas will tend to cause losses thatwill reduce the overall array gain. In addition, if the secondary antenna is lessefficient than the primary antenna, the total aperture would be expected to beless than twice the aperture of the primary antenna. Finally, the last gain compo-nent that can be achieved results from the antenna array interference reductioncapabilities. Optimum weights could be applied to the signals received by eachantenna according to one of the adaptive beamforming criteria discussed inChapter 5, which could in the case of a strong signal and a single stronginterferer nearly completely cancel the interference out, resulting in a highSINR. With multiple interferers and receiver noise cases, adaptive beamformingwill make the best possible trade-off between signal enhancement and interfer-ence reduction. Unlike antennas at the base station, single- and dual-antennahandsets suffer from a unique problem that should be considered, the meaneffective gain (MEG), and its possible imbalance between the primary and sec-ondary antennas. The MEG [15] is a good way to characterize the effectivenessof an antenna at coupling to the signals it receives. Let us denote the base stationtransmit power by Ior and the total power incident on the antenna (from alldirections) after it has gone through the channel by $I or , while the mean power

received (absorbed) by the antenna is Prec, then the MEG is given by

MEGP

I or

= rec

$(9.10)

The unique situation we have with handset antennas is that the presence ofthe head or hand will affect the ability of the antenna to couple the incidentenergy and therefore will affect its MEG. The MEG of a whip antenna on ahandset held vertically with no head or hand near it is likely to be 0 dB. How-ever, the following factors will reduce the MEG: the antenna will be held with


45°–60° of tilt away from vertical, the head will block some directions andabsorb some radiation, and the hand will provide additional blockage andabsorption. The biggest effect of holding the handset is a reduction in the gain,especially in the direction of the head. The effect of this on the MEG is toreduce it by an amount that depends on the operating frequency. Obviously,this reduction in MEG could impact the receiver performance, so the question isby how much? The receiver performance is mainly dictated by its sensitivity,which is limited by the interference picked and the noise added to the receiverchain. It is well known that receivers in CDMA systems are mostly interferencelimited [i.e., the received interference powers are much stronger than the noisefloor except at locations where the received signal power is too close to the noisefloor (e.g., edge of coverage)]. It then follows that the receiver SINR in interfer-ence-limited situations is dominated by the SIR or signal-to-interference ratio.But since the antenna gain or MEG will affect the received signal and interfer-ence powers equally, any reduction in MEG should have a minimal impact onthe receiver performance.

9.3 Combining Techniques

Several combining techniques have been previously discussed in Chapter 5. Thefollowing is a summary of some methods applicable to mobile stations.

9.3.1 Selection (Switched) Diversity

This diversity combining technique assumes that the better of the two antennasis always and immediately chosen. In a realistic algorithm, there would be areduction in the performance of this approach due to the real overhead of theswitching system.

9.3.2 Maximal Ratio Combining

In the MRC scheme, the antennas are combined to maximize sensitivity underthe assumption that the noise and interference the two antennas are receivinghave low cross correlation. The weights applied to the antennas’ inputs are afunction of the received SNR.

9.4 Adaptive Beamforming or Optimum Combining

This is the most optimal combining scheme, where the antennas are combinedwith full knowledge of the interference correlation between the two antennas.Several adaptive beamforming techniques have been discussed in Chapter 5.


Since the goal is to reduce the interference and improve the SINR, null steeringcan be employed to cancel or reduce some of the interference seen by the mobilestation. However, null steering might not be of much practical use in this casedue to the size constraints of the mobile stations. Optimum combining (OC)techniques such as those based on the Wiener-Hopf type calculations, resultingin an MMSE or max SINR, are thus more suitable. In the remainder of this sub-section we will show why this is the case by means of simulations and by study-ing some relevant array characteristics. It is known that an adaptive antennaarray of M elements can null M-1 interferers in both LOS and multipath envi-ronments. In [16, 17], it is stated that an antenna array can separate and com-bine closely spaced signals such as the case in multipath fading environments aslong as the array beamwidth is smaller than the AS of the received paths. Theperformance of the array improves with M, larger AS, and denser multipaths.From [18], we know that at the mobile station, assuming a power azimuth spec-trum with uniform distribution over 360°, the resulting AS is 104°. Figure 9.3shows the array beamwidth for small element separations for two-, three-, andfour-element broadside arrays. We can see from the plot that for M = 3 and 4,the beamwidth is smaller than the AS (104°), even for small separations d. ForM = 2, we notice that larger values of d (about 0.3λ) are required to achieve thenecessary beamwidth.

To compare the performance of optimum combining techniques such asmax SINR to that of null steering we will consider both LOS and multipath fad-ing cases. Recall from Chapter 5 that the null steering weights are given by

[ ]W ANSH T= −1 0 0 1L (9.11)


θ

λ

Figure 9.3 3-dB beamwidth as a function of interelement separation for M = 2, 3, and 4.

whereas the optimum weight vector obtained from the max SINR criterion isgiven by

W R Aopt SINR NK= −1 (9.12)

where A is the steering matrix whose columns are the array steering vectors(LOS) or spatial signatures (multipath). After null steering is applied, the arrayoutput can be written as

( ) ( ) ( )y t s t N tdH= + w (9.13)

It then follows that the SINR at the array output after null steering is

SINR NSs

NH

=σ

σ

2

2 w w(9.14)

where σ s2 is the signal power and σ N

2 is the noise variance. On the other hand,

the SINR at the array output as a result of the max SINR criterion is

( ) ( )SINR ROC LOS sH

d IN d_ = −σ θ θ2 1a a (9.15)

SINR A R AOC NLOS sH

IN_ = −σ 2 1 (9.16)

where NLOS stands for the multipath case, θd is the DOA of the desired signal,A denotes the spatial signature vector of the desired user, and RIN is the interfer-ence plus noise correlation matrix. To compare the performance of these twotechniques, simulations were run for different environments, element spacing,number of multipaths, and DOA of desired and interfering signals. Figures 9.4and 9.5 compare the SINR achieved using a two-element array in LOS environ-ment for d = 0.5 λ and 0.2 λ, respectively, for one desired signal at 0° and oneinterferer at 3°.

Figures 9.6 and 9.7 show the performance when one desired and oneinterfering signal have two paths each.

Finally, Figures 9.8 and 9.9 compare the performance when the desiredsignal and interfering signal each has four paths.

For all the results shown, the desired and interfering signals have the samepower, 5,000 samples were used for each data point, and the mean SINR wasplotted as a function of the input SNR. For the multipath cases, different pathswere assumed to have random phases between [0,2π]. From these plots we canconclude that the max SINR technique significantly outperforms null steering



SNR (dB)

–20 –10 0 10 20 30 40

M=2, LOS uncorrelated sources at 0, 3 degrees, d=0.5 (wavelength)

Max SINRNull steering

30

20

10

0

–10

–20

–30

–40

SNR

(dB)

Figure 9.4 Optimum SINR versus null steering, M = 2, desired signal at 0°, interference at 3°, d =0.5 λ in LOS.

SNR (dB)

–20 –10 0 10 20 30 40

M=2, uncorrelated sources (LOS) at 0 and 3 degrees, d=0.2 (wavelength)


20

10

0

–10

–20

–30

–40

SNR

(dB)

–50

Figure 9.5 Optimum SINR versus null steering, M = 2, desired signal at 0°, interference at 3°, d =0.2 λ in LOS.


SNR (dB)

–20 –10 0 10 20 30 40

M=2, Multipath fading at 0 and 2 degrees, interference at 4 and 6 degrees, d=0.5 (wavelength)


30

20

10

0

–10

–20

–30

–40

SNR

(dB)

Figure 9.6 Optimum SINR versus null steering, M = 2, desired signal with two paths at 0° and 2°interference with two paths at 4° and 6°, d = 0.5 λ. Equal power signals.

SNR (dB)

–20 –10 0 10 20 30 40

M=2, Multipath fading, desired signal at 0 and 2 interference at 4 and 6, d=0.2 (wavelength)


20

10

0

–10

–20

–30

–40

SNR

(dB)

–50

Figure 9.7 Optimum SINR versus null steering, M = 2, desired signal with two paths at 0° and 2°interference with two paths at 4° and 6°, d = 0.2 λ. Equal power signals.


SNR (dB)

–20 –10 0 10 20 30 40


30

20

10

0

–10

–20

–30

–40

SNR

(dB)

M=2, Multipath fading, desired signal paths at 0, 2, 4, and 6 degrees,and interference at 3, 7, 9, and 11 degrees, d=0.2 (wavelength)

Figure 9.8 Optimum SINR versus null steering, M = 2, desired signal with four paths at 0°, 2°, 4°,5° and interference with four paths at 3°, 7°, 9°, 11°, d = 0.2 λ. Equal power signals.

SNR (dB)

–20 –10 0 10 20 30 40

M=2, Multipath fading, desired signal paths at 0, 2, 4, and 6 degrees,and interference at 3, 7, 9, and 11 degrees, d=0.5 (wavelength)

30

20

10

0

–10

–20

–30

–40

SNR

(dB)


Figure 9.9 Optimum SINR versus null steering, M = 2, desired signal with four paths at 0°, 2°, 4°,5° and interference with four paths at 3°, 7°, 9°, 11°, d = 0.5 λ. Equal power signals.

for low SNR regardless of the environment. This is because maximizing theSINR, which is equivalent to maintaining a high gain at the direction of thedesired signal while taking into account additive noise (maximizing S and mini-mizing N), is more effective than just reducing the interference (reducing I)alone. For high SNR, since S is already high, both techniques achieve similarperformance. In fact, the performance becomes identical for very high SNR.This is due to the fact that as SNR increases, because we are assuming the inter-ference and desired signals have similar power, reducing or nulling the interfer-ence will greatly enhance the SINR (by reducing I). Under these conditions, it isvery beneficial to null or reduce the interference because the signal power isalready much higher than the noise and so it is less important to increase thegain in the direction of the desired signal. Based on the above, it is recom-mended that an optimum combining technique such as max SINR or theMMSE be selected over null or beam steering to achieve the desired perfor-

mance improvements in EE

Ic

or i

.

9.5 RAKE Receiver Size

A RAKE receiver combines multiple copies (fingers) of the received signal tosynthesize a single, higher SINR copy. The multifinger RAKE receiver allowsefficient detection of CDMA signals because it makes use of the availablemultipath. The RAKE allows most of the power from the different paths to becombined in the receiver to enhance the SINR. In soft handoff, the signal to thephone is sent from multiple sectors in the network. If at least one finger isassigned to signals arriving from each serving sector, then the phone receiver cancombine the signal energies from all sectors simultaneously. Existing CDMAphones based on the IS-95 or CDMA2000 standard with a chip rate of 1.2288Mcps use four-finger RAKE receivers. Commercial CDMA deployments haveshown that increasing the number of fingers beyond four would add complexityto the handset with only a small improvement in diversity performance, hence itis widely accepted that this is a good size for the RAKE receiver for sin-gle-antenna phones in narrowband CDMA systems. For dual-antenna phones,the RAKE receiver must assign two fingers for each path arriving at the phonefrom each antenna so that the gains from combining that path as seen on thetwo separate antennas are achieved. Since the four-finger RAKE is limited tocombining only two paths if each path is combined on two antennas, the fullexploitation of dual-antenna diversity requires eight fingers. For WCDMA sys-tems, the chip rate is 3.84 Mcps, which allows multipath components separatedby as little as one chip or 0.26 µs to be resolvable. The increased number of


usable multipath from the channel because of the higher system bandwidthimplies that more fingers can be added to the RAKE receiver to improve thediversity gain.

9.6 Mutual Coupling Effects

Due to the size constraints of mobile stations, placing the diversity antennas veryclosely results in mutual coupling, which can affect the performance of thearray. As previously described, in a dual-antenna diversity mobile station a singleantenna is used to transmit and is connected to one receive chain while the sec-ond antenna is connected to the second receive chain; that is, the secondaryantenna is connected to a receive filter among other parts of the chain. The pres-ence of the second antenna and its associated terminating load will affect theamount of energy coupled from the primary antenna. This in turn will affect theisolation and return loss parameters of the array, leading to coupling and mis-match losses. Let us denote the return loss and isolation by S11 and S21, respec-tively. We can then define the coupling loss as

( )L coupS= − −10 1 1010

1021log (9.17)

and the mismatch loss as

( )L mS= − −10 1 1010

1011log (9.18)

Figure 9.10 plots Lcoup and Lm as a function of S11 and S21. We can see thatany degradation in the return loss or isolation will result in some loss, which willreduce the antenna’s efficiency and hence the gain. The total loss in the antennaefficiency can be written as

L L L Ltot coup m other= + + (9.19)

where Lother accounts for losses such as dielectric and conductor losses. Anotherproblem that arises from mutual coupling is the transmit band coupling, whichleads to reduction in the transmit efficiency. A portion of the transmit powercan be coupled from the primary antenna to the secondary antenna. Dependingon the termination of the secondary antenna this energy can either be absorbedor reflected by the antenna. If the secondary antenna absorbs this energy, thetransmit efficiency of the primary antenna will be reduced and hence its gainwill be reduced. If the energy is reflected, it will be radiated by the secondaryantenna and will cause interference to the primary antenna transmission, whichcould affect its beam pattern.


9.7 Dual-Antenna Performance Improvements

Extensive field and lab testing of dual-antenna handsets have been carried outand reported in [19]. A commercially available CDMA phone with an externalwhip antenna has been modified by adding an internal antenna as the secondaryantenna and several combining techniques, including switched diversity, MRC,and optimum combining, were employed. Figure 9.11 shows the CDF of the

requiredE

Ic

or

necessary to achieve anE

Nb

t

of 3.9 dB (target corresponding to 1%

FER) with eight RAKE fingers. The required traffic power relative to the totaltransmit power available from sectors in the system calculations included theeffect of handoff.

Table 9.1 shows the forward link capacity improvement with respect tothe whip antenna feeding a four-finger RAKE receiver.

Table 9.1 shows that the whip antenna performance sees little improve-ment when the number of fingers is varied from four to eight. The capacitygains reported here are the average reductions in transmit power per user (i.e.,

EE

Ic

or i

. From Table 9.1, we can draw the following set of conclusions:


3.5

3

2.5

2

1.5

1

0.5

0–25 –20 –15 –10 –5 0

S11/S21 (dB)

Cou

plin

g/m

ism

atch

loss

(dB)

Figure 9.10 Coupling loss as a function of return/insertion loss.


Required TX power for 8-finger RAKE1

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2

0.1

0

Prob

abili

tyth

atTX

>x-

axis

WhipIntSwitchedOCMRC

–30 –25 –20 –15 –10 –5TX fraction system, dB

Figure 9.11 EEI

c

or i

required for eight-finger RAKE. (From: [19]. Qualcomm, Inc. Reprinted

with permission.)

Table 9.1Performance Improvements with Different Combining Techniques

Number ofFingers

WhipAntenna

SwitchedDiversity MRC Optimum Combining

Capacity Increase Relative to Whip Antenna (dB)

PCS Band 4 Reference 1 1.8 2.5

6 0.3 1.3 2.4 3.2

8 0.4 1.3 2.7 3.5

Cellular Band 4 Reference 0.9 1.7 2.9

6 0.1 0.9 2.3 3.6

8 0.1 1 2.4 3.8

Source: [19].

• Optimum combining outperforms all other techniques and results inthe most significant gains, which can reduce the average required trans-mit power by about 3 dB, thereby doubling the capacity.

• OC outperforms MRC by 0.7–1.4 dB, regardless of the number ofRAKE fingers used.

• Increasing the number of fingers improves the capacity gains. We cansee significant gains in OC and MRC by going from four to six fingersbut only a small improvement (0.2–0.3 dB) when we add another twofingers.

When the phone is in soft handoff (i.e., the phone is being served by mul-tiple sectors), the load it presents to the system increases. For instance, when inNs-way handoff, the phone has Ns sectors in its active set, which in turn requiresa larger number of RAKE receiver fingers. Since in the dual-receive-chain diver-sity architecture two fingers are devoted to each time-resolvable path beingtracked by the receiver, the diversity performance will be compromised duringhigh handoff states with smaller RAKE sizes. Table 9.2 summarizes the impactof the handoff state on capacity improvements for the PCS band. In any givenhandoff state, we can see some capacity improvements by increasing the number


Table 9.2Impact of Handoff on Capacity Gains, PCS Band

Handoff StateNumber ofFingers Whip Antenna OC MRC Switched

Capacity Increase Relative to Whip Antenna (dB)

Single Sector 4 Reference 3.2 2.4 1.6

6 0.1 3.5 2.7 1.7

8 0.1 3.6 2.8 1.7

Two-way SHO 4 Reference 2.4 1.7 1.0

6 0.3 3.2 2.4 1.2

8 0.4 3.6 2.7 1.3

Three-way SHO 4 Reference 2.0 1.3 0.6

6 0.4 2.7 2.0 1.0

8 0.6 3.3 2.6 1.1

Source: [19].

of fingers because more fingers will allow the receiver to combine gains fromthe dual-antenna architecture and conventional soft handoff. On the otherhand, as the handoff state increases (i.e., as we have more sectors in the activeset), we see a drop in the capacity improvements, as is the case of three-waySHO. This is due to the fact that the receiver is unable to combine all the avail-able paths from both the spatial domain (dual-antenna) and temporal domain(RAKE) because it is limited by the number of fingers (recall that a path fromone sector requires at least two fingers in this architecture to achieve maximumcapacity gains).

The impact of the difference in the received power on both antennason the performance is shown in Figure 9.12. We can observe two distinct per-formance regions, one in which there is a large difference in the powers receivedby each antenna, which corresponds to the right- and left-hand side. In thissituation we see that the dual-antenna improvements are not much better thanthe performance with the single antenna. In the second region, where thereceived powers on each antenna are similar, the capacity gains are significantwith dual-antenna receivers since this would be the optimum conditions toachieve diversity and beamforming gains. Another observation we can drawfrom these results is that OC always outperforms other techniques, followedby MRC.


10

8

6

4

2

0

–2

–4

–6

–8

Cap

acity

rela

tive

toW

hip

, dB

Relative capacity gain

–15 –10 –5 0 5 10 15P /P , dBWhip INT

WhipIntOCMRCSwitched

Figure 9.12 Relative capacity gains versus receive power difference between whip and internal forPCS band, six-finger RAKE. (From: [19]. Qualcomm, Inc. Reprinted with permission.)

9.8 Downlink Capacity Gains

As we have seen earlier, a reduction inE

Ic

or

results in a capacity increase that can

be used in several ways. First, in parts of the system where capacity is limited bythe downlink, dual-antenna phones will raise the downlink capacity to matchthe uplink capacity or until the higher uplink capacity becomes the limit. Sec-ond, new high-speed data services demand more resources on the downlink thanthe uplink. By introducing dual-antenna phones, downlink capacity is freed-upfor use in data services. Finally, the downlink capacity increase can be matchedby an uplink capacity increase using antenna arrays at the base station. Let us

assume that all the improvement inE

Ic

or

is traded for increased capacity on the

downlink; we will then proceed to show how much capacity gain can beachieved by using dual-antenna mobile stations with dual-receive chain. From(9.6) the downlink capacity of a CDMA system with conventional mobile sta-tions is given by

( )C

E I

EE

I

conv

c or Tot Overhead

c

or i conv

=−

1_ (9.20)

This expression is derived based on the premise that the downlink capacityis limited by the total base station transmit power, part of which will be devotedto the common or overhead channels with the remaining part allocated to thetraffic or dedicated channels of the users in the sector. Similarly, the capacityachievable with dual-antenna or diversity mobile stations is given by

( )C

E I

EE

I

div

c or Tot Overhead

c

or i div

=−

1_ (9.21)

It then follows that

C

C

EE

I

EE

I

conv

div

c

or i div

c

or i

=

conv

(9.22)


where we have assumed that both types of mobile stations experience the samesize of active sets [i.e., the handoff reduction factor (see Chapter 2) is equivalentin both cases]. Now, let us assume that the sector is supporting mixed mobilestations with Nconv and Ndiv representing the number of conventional and diver-sity enabled stations, respectively. We can then write

N EE

IN E

E

Iconvc

or i conv

divc

or i

+

=

div

c

or i conv

convEE

IC (9.23)

where we use the fact that the portion of the base station transmit power avail-able for traffic in the case of mixed and conventional mobile stations is the same.Equation (9.23) can be rewritten as

N NE

ICconv div

c

orconv+

=∆

where

∆E

IE

E

IE

E

Ic

or

c

or i div

c

or i

=

conv

Defining the capacity of the sector with mixed mobile stations as Cmixed weget

C

C

N N

N NE

I

mixed

conv

conv div

conv divc

or

=+

+ ⋅

=

−∆

1

1 α( )div divc

or

E

I+ ⋅

α ∆

(9.24)

where αdiv is the percentage of diversity-enabled mobile stations in the sector.Equation (9.24) actually represents the achievable capacity gain and it is shown

in Figure 9.13 for different values of αdiv and ∆E

Ic

or

. We can see that as the

percentage of diversity-enabled phones is increased, the capacity gain increases.

Notice that the capacity is doubled when ∆E

Ic

or

= -3 dB.


9.9 Conclusions

The downlink capacity of a CDMA sector can be doubled if all the mobile sta-tions in that sector use dual-antenna receivers. Capacity increase is achievedbecause the average signal sensitivity of the phone receiver is doubled throughadaptively combining the two antenna-receiver chains. With twice the sensitiv-ity, average base station transmit power per phone can be cut in half. Thus,twice as many downlink active calls can be served.

The downlink capacity is increased incrementally in proportion to thepenetration of dual-antenna phones into the system. The capacity increase canbe used in three ways. First, where capacity is limited by the downlink,dual-antenna phones will raise total network capacity. Second, dual-antennaphones free up capacity for new downlink capacity-intensive data services.Finally, the downlink capacity increase can be matched by an uplink capacityincrease using antenna arrays at the base station. The primary mechanisms bywhich the sensitivity is improved are aperture gain, interference reduction, anddiversity gain.


2

1.9

1.8

1.7

1.6

1.5

1.4

1.3

1.2

1.1

1

Rela

tive

cap

acity

gain

0 10 20 30 40 50 60 70 80 90 100Percentage of dual-antenna mobile stations

Delta Ec/Ior = –1 dBDelta Ec/Ior = –2 dBDelta Ec/Ior = –3 dB

Figure 9.13 Relative sector capacity versus percentage penetration of dual-antenna mobile sta-

tions for different ∆ EI

c

or

.

References

[1] Lee, W. C. Y., Mobile Cellular Communications Systems, New York: McGraw-Hill, 1989.

[2] Jakes, W. C., (ed.), Microwave Mobile Communications, New York: IEEE Press, 1994.

[3] Taga, T., and K. Tsunekawa, “Performance Analysis of a Built-In Planar Inverted-FAntenna for 800-MHz Band Portable Radio Units,” IEEE Journal on Selected Areas inCommunications, Vol. JSAC-5, June 1987, pp. 921–929.

[4] Fuhl, J., P. Nowak, and E. Bonek, “Improved Internal Antenna for Hand-Held Termi-nals,” Electronics Letters, Vol. 30, October 1994, pp. 1816–1818.

[5] Erätuuli, P., et al., “Performance of Internal Microstrip Handset Antennas,” 46th IEEEVehicular Technology Conference, Atlanta, GA, April 28–May 1, 1996, pp. 344–347.

[6] Haruki, H., and A. Kobayashi, “The Inverted-F Antenna for Portable Radio Units,” Conv.Rec. IECE Japan (in Japanese), March 1982, p. 613.

[7] Winters, J. H., “Optimum Combining in Digital Mobile Radio with Co-Channel Inter-ference,” IEEE Trans. Veh. Technol., Vol. VT-33, No. 3, 1984, pp. 144–155.

[8] Erätuuli, P., and E. Bonek, “Diversity Arrangements for Internal Handset Antennas,” 8thInternational Symposium on Personal, Indoor and Mobile Radio Communications, Helsinki,Finland, September 1–4, 1997, pp. 589–593.

[9] LeFevre, M., M. A. Jensen, and M. D. Rice, “Indoor Measurement of HandsetDual-antenna Diversity Performance,” IEEE VTC’97, Phoenix, AZ, May 1997.

[10] Braun, C., G. Engblom, and C. Beckmam, “Antenna Diversity for Mobile Telephones,”IEEE AP-S’98, Atlanta, GA, July 1998, pp. 2220–2223.

[11] Cox, D. C., “Antenna Diversity Performance in Mitigating the Effects of Portable Radio-telephone Orientation and Multipath Propagation,” IEEE Trans. Commun., Vol.COM-31, May 1983, pp. 620–628.

[12] Viterbi, A. J., CDMA Principles of Spread Spectrum Communication, Reading, MA: Addi-son-Wesley, 1995.

[13] Salz, J., and J. Winters, “Effect of Fading Correlation on Adaptive Arrays in DigitalMobile Radio,” IEEE Trans. on Vehicular Technology, Vol. 43, No. 4, November 1994,pp. 1049–1057.

[14] Colburn, J. S., et al., “Evaluation of Personal Communications Dual Antenna HandsetDiversity Performance,” IEEE Trans. on Vehicle Technology, Vol. 37, August 1998, pp.737–746.

[15] Taga, T., “Analysis of Mean Effective Gain of Mobile Antennas in Land Mobile RadioEnvironments,” IEEE Trans. on Vehicle Technology, Vol. 39, May 1990, pp. 117–131.

[16] Winters, J., “Smart Antennas for Wireless Systems,” IEEE Personal Communications, Feb-ruary 1998, pp. 23–27.

[17] Winters, J., “Optimum Combining in Digital Mobile Radio with Cochannel Interfer-ence,” IEEE JSAC, Vol. 2, No. 4, July 1984, pp. 528–539.

[18] 3GPP-3GPP2 SCM-121, Spatial Channel Model Text Description, March 14, 2003.

[19] Wengler, M. J., et al., “Capacity_Increase_CDMA2000.pdf,” found at: http://www.qualcomm.com/technology/1xev-do/publishedpapers/Capacity_Increase_CDMA2000.pdf,2003.


10MIMO Systems

10.1 Introduction

In conventional wireless communication systems, one end of a link, typically thebase station, is equipped with more than one antenna (e.g., two antennas forreceive diversity or, more recently, an antenna array with more than two ele-ments for beamforming or beam steering). Several chipset and handset vendorshave recently started developing and designing phones with dual antennas formobile receive diversity (MRD). In a MIMO system the communication linkhas both a transmitter as well as a receiver that are equipped with multipleantenna elements, as shown in Figure 10.1. The idea behind MIMO is to useAA at both the transmitter and receiver in combination with space-time modu-lation and coding techniques to achieve very high spectral efficiencies. Themotivation for using such a setup is to achieve improvements that can be used toincrease both the quality of service and the revenues significantly. Examples ofapplications that can benefit from MIMO systems include cellular/PCS systemdeployment based on fixed user terminals, microcells, picocells, and wirelessLANs, as well as portable applications such as laptops and PDAs. Coding andmodulation are essentially temporal techniques. A MIMO system is aspace-time signal processing approach in which the time dimension is comple-mented with the spatial dimension through the use of multiple spatially distrib-uted antennas. As such, MIMO systems can be viewed as an extension of smartantennas. Therefore, we can classify multiantenna schemes as a family thatincludes spatial techniques such as adaptive antennas, beamforming, and spatialdiversity, as well as spatial multiplexing (SM) and space-time coding (STC) suchas MIMO. Multiantenna schemes have great potential for significant informa-tion theoretic capacity increase. As we have seen in previous chapters, the main

289

impairments to the performance of a wireless communication system are fadingdue to multipath and interference. The different multiantenna techniques dealwith those impairments in different ways to either mitigate their impact orexploit their presence to improve the link and system level performance. A keyfeature of MIMO systems is the ability to exploit multipath by taking advantageof random fading [1–5], effectively extending the benefits of smart antennas astep further by achieving a multifold increase in transfer rates. This chapter pro-vides an introductory overview of the most important aspects of MIMO sys-tems, points out their relation to other spatial processing approaches, andsummarizes the key issues and factors that affect their performance. An excellentdiscussion on MIMO systems can be found in [6] and the references therein.

10.2 Principles of MIMO Systems

Table 10.1 shows a comparison of all the possible combinations of multiantennatechniques. In general, we define the number of transmit antennas as MT andthe number of receive antennas as MR. We can think of the wireless channel as avector channel with dimensionality MR × 1 or MT × 1. If we use only oneantenna for transmission, the data rate will always be limited by the performanceof that single antenna. One way to recover from the channel impairments wouldbe by using a multiantenna at the receive end to reverse all or part of the channeleffects. In this case the single transmit antenna becomes the bottleneck in a way.However, the performance is often superior to that obtained with single anten-nas at both ends. The same effect can be accomplished using multiple antennasat the transmitter end in situations where implementation issues prevent theiruse at the receiver end, such as in mobile handsets to try to create signal condi-tions at the receiver similar or close to those present had it been equipped withmultiple antennas to enable it to take advantage of additional degrees of free-dom, such as higher diversity orders. The resulting performance is roughly thesame when the same number of elements is used. On the other hand, in MIMOsystems data are transmitted over a matrix channel created by MT transmit andMR receive antennas rather than a vector channel, creating new types of gainsbeyond just diversity or array gain benefits. It was shown in [2] how one may,


Tx Rx

MT MR

Matrixchannel

Figure 10.1 MIMO wireless communications systems.

under certain conditions, transmit independent data streams simultaneouslyover the eigenmodes of a matrix channel. Let us denote the transmit signal by s(t), the received signal by y(t), the received noise by n(t), and the channel matrixby H(τ, t). We can then write

( ) ( ) ( ) ( )y s nt t t t= +H τ, (10.1)

10.2.1 SISO

In single input single output (SISO) systems, shown in Figure 10.2, MT = MR =1, y(t) and s(t) are 1 × 1 vectors, and the channel matrix H(τ, t) = h is also a 1 × 1vector. The normalized Shannon capacity in this case is given by

( )C h SINR= + ⋅log 2

21 (10.2)

The limitation of SISO systems is that the capacity increases very slowlywith the log of SINR and in general it is low. Moreover, fading can cause largefluctuations in the signal power level, increasing the variance. Only temporaland frequency domain processing are possible but the spatial domain isneglected.

10.2.2 SIMO

In this case, both y(t), h are MT × 1 vectors, whereas s(t) is 1 × 1, as shown inFigure 10.3 Example applications of single input multiple output (SIMO) sys-tems include receive diversity, beamforming, beam steering and null steering.The normalized Shannon capacity in this case is given by

C SINR hii

M R

= + ⋅

=

∑log 2

2

1

1 (10.3)

Similar to the SISO case, capacity increases logarithmically as both MT andSINR are increased. The average channel capacity is, however, higher than the

MIMO Systems 291

Tx RxChannel

Figure 10.2 SISO scheme.

SISO case. The actual performance depends on the nature of the channel andthe correlation across the antenna elements. For low correlation, SIMO systemsprovide diversity gain, which helps reduce fading effects when coherent combin-ing is employed [7]. When the signals across the antennas are highly correlated(e.g., in the beamforming case), the system provides an array gain in addition tointerference reduction or cancellation.

10.2.3 MISO

In this case, both s(t) and h are MT× 1 vectors and y (t) is a 1 × 1vector, as shownin Figure 10.4. Transmit diversity and beamforming are example multiple inputsingle output (MISO) schemes. The normalized channel capacity in the casewhere the channel knowledge is unknown (open loop) is given by

CSINRM

hT

ii

M T

= + ⋅

=∑log 2

2

1

1 (10.4)

We can see that the SINR is normalized by MT, ensuring fixed total trans-mit power. It is also clear that there is no array gain in this case and the capacityincreases logarithmically with SINR. Assuming the channel coefficients aregiven by

h Mii

M

T

T2

1=∑ = (10.5)


Tx Rx

MR

Vectorchannel

Figure 10.3 SIMO scheme.

Tx Rx

MT

Vectorchannel

Figure 10.4 MISO scheme.

then we get

( )C SINR= +log 2 1 (10.6)

That is, the capacity does not increase when the number of transmitantennas is increased. However, that does not mean we cannot increase capacityusing open loop transmit diversity schemes such as STTD and STS. In fact, wehave seen in Chapter 7 that STS and STTD provide diversity gain, which can betraded off for capacity improvements. This capacity increase is not a result ofincreasing the number of antenna elements but is rather due to SINR improve-ments for a given BLER or FER target compared with the SISO case. On theother hand, when the channel is known to the transmitter, such as the case inclosed loop transmit diversity (TXAA and CLTD), then we get from [6] that

C SINR hii

M T

= + ⋅

=

∑log 2

2

1

1 (10.7)

It then follows that for channel coefficients given by (10.5), the capacity isgiven by

( )C SINR M T= + ⋅log 2 1 (10.8)

This is consistent with the results in Chapter 7, where CLTD outper-formed STTD.

10.2.4 MIMO

In the case of MIMO, s(t) is a MT × 1 vector, y(t) is a MR × 1 vector, and H is aMT × MT matrix. An example MIMO setup is shown in Figure 10.5. Let thechannel matrix be given by

H =

h h h

h h h

h h h

M

M

M M M M

R

R

T T T R

11 12 1

21 22 2

1 2

L

L

M M O M

L

(10.9)

In the absence of channel knowledge at the transmitter, the capacity isgiven by

CSINRMM

T

H

R= +

log det2 I HH (10.10)

MIMO Systems 293

which can rewritten as [6, 8]

CSINR

Mi

Ti

L

= +⋅

=∑ log 2

1

1λ

(10.11)

Equation (10.11) clearly shows that the MIMO channel capacity can beexpressed as the sum of the capacities of L SISO channels or multiple spatialdata pipes. To maximize the MIMO channel capacity, given a fixed total chan-

nel power y such that H ii

L2

1

= ==∑ λ γ,H, H must be orthogonal [8] (i.e.,

given by)

H =

h

h

hM MT R

11

22

0 0

0 0

0 0

L

L

M M O M

L

(10.12)

Hence, the channel capacity is maximized when the channel matrix isdiagonal (i.e., when the subchannels are uncorrelated, such as parallel independ-ent subchannels). Any correlation between the different subchannels results inincreased fading and a reduction in channel capacity. To achieve this very highcapacity, the channel matrix must be made diagonal through signal processing atthe receiver. The capacity can then be rewritten in the simple form

C MSINRMT

T

= +

log 2 1 (10.13)

Consider the multiantenna system diagrams in Figure 10.6. A digitalinput signal is fed to a serial to parallel splitter after error control coding andmapping to complex modulation symbols. The splitter produces several separatesymbol streams and each are then mapped onto one of the multiple transmit


Tx Rx

MT

Vectorchannel

MR

Figure 10.5 MIMO scheme.

antennas, which may include spatial weighting of the antenna elements orantenna space-time precoding. At the receiver, the signals are captured by multi-ple antennas and the signals are recovered after demodulation and demapping.This can be considered as an extension to conventional smart antenna applica-tions. The intelligence of the multiantenna system lies in the weight selectionalgorithm and can offer a more reliable communications link in the presence ofadverse propagation conditions such as multipath fading and interference.

Figure 10.7 compares the performance of all four schemes versus the SINRand number of receive and transmit array elements. For the adaptive antennaarray case (beamforming), only the effect of the increased array gain was consid-ered in the comparison. The actual performance is better when the interferencereduction capabilities are factored in.

A comparison between the different spatial techniques already discussed isshown in Table 10.1

10.3 Transmission Strategies

It is important to realize that the performance improvements achievable withMIMO systems by going to multiple transmit antenna do not derive from

MIMO Systems 295

Signalprocessing

MT MR

S/P P/SModulation/weighting

Figure 10.6 Spatial multiplexing with MIMO.

Figure 10.7 Performance comparison between different spatial techniques.

increased transmit power, which would be a very inefficient approach to over-come interference or improve spectral efficiency. Rather, in MIMO systems apeak power constraint of Pmax is imposed on each transmit antenna so that thetotal power available at the transmitter is Ptot = MT Pmax and is equivalent to thesingle transmit antenna case. It is possible to allocate this total power over all Nnonzero eigenmodes of the channel in a variety of ways, as long as theper-antenna power limit and the total power limit are not exceeded. Commonmethods include water filling, uniform power allocation, beamforming, andbeam steering.

10.3.1 Water Filling

Assuming a total transmit power constraint, Ptot = MT Pmax, the optimum, capac-ity achieving power allocation strategy for the N parallel channels is found bywaterfilling [9]. The waterfilling method performs a distribution of the availablepower over the eigenmodes in such a way that the mode with the lowest noisevariance receives the greatest fraction of total power. The waterfilling powerallocation technique is optimal under constraint of total power. However, per-formance degrades when the per-antenna power limit is taken into account.

10.3.2 Uniform Power Allocation

One transmission method is to allocate the total power evenly over all modes.This uniform power allocation method assigns power Ptot /N to each mode; this


Table 10.1Multielement Spatial Schemes

Scheme MT MR Example Benefits

SISO 1 1 No transmit or receive diversity No diversity

SIMO 1 > 1 Receive diversity, beamforming,beam steering

Diversity proportional to MR.

Array gain interference reductionwith beamforming.

MISO > 1 1 Transmit diversity, beamforming,beam steering

Diversity proportional to MT.

Array gain interference reductionwith beamforming.

MIMO > 1 > 1 Use of multiple antennas at boththe transmitter and receiver

Diversity proportional to theproduct of MT and MR.

Array gain (coherent combiningassuming prior channel estimation).

power allocation results in equal power on each antenna. When the number ofmodes is less than the number of transmit antennas, a scaling coefficient can beused to meet the constraint. It is worth noting that the effective result is thesame as if the maximum power Pmax were allocated to each mode.

10.3.3 Beamforming

The beamforming power allocation strategy places all of the available power on asingle eigenmode. To approach capacity, the total transmit power is assigned tothe eigenmode corresponding to the highest eigenvalue. Thus, the SINR is max-imized given the constraint of using a single mode.

10.3.4 Beam Steering

In the beamforming transmission strategy already described, both the amplitudeand phases of the principal eigenmode are used at the transmitter. The beamsteering transmission strategy also places the total available power on the singledata stream (eigenmode); however, instead of using both the amplitude andphase information of the principal eigenmode, only the phase information isused. The amplitude information is discarded by normalizing the principaleigenvector such that all coefficients of the vector have unity amplitude. Toensure that the power across each transmit antenna is Pmax a rescaling coefficientequal to Pmax/Ptot is applied. As discussed in Chapter 5, the beam steeringapproach better uses the total available power by increasing the transmittedpower in the direction of the desired user. In the beamforming strategy, the scal-ing ensures that the highest antenna power is equal to the per-antenna powerlimit, whereas the beam steering scheme forces the power on all transmit anten-nas to equal Pmax, thus, resulting in a higher overall transmit power and a highereffective SINR.

10.4 MIMO Approaches

There are several approaches to implement MIMO systems that are based on thepresence or absence of channel information at the transmitter. A summary of thedifferent MIMO schemes is shown in Table 10.2.

The best performance can be achieved with fixed terminals where thereceive array size is not severely constrained by the physical dimensions. Lowfading rates or, more precisely, lack of mobility allows for accurate channel esti-mation, hence the full CSI-based approach is possible. Moreover, the use ofdirectional antennas in the receive array helps improve performance. For porta-ble terminals, which are usually stationary during usage (e.g., laptop), the arraysize is somewhat more constrained but the use of directional and/or omni

MIMO Systems 297

receive antennas is still possible. In this case, a mixed approach is more suitablewhere the mode can be adaptively selected between full, partial, or no CSI.Finally, for mobile terminals, the array size is severely constrained (e.g., phone)and the terminals are typically restricted to omni receive antennas. This resultsin a low to moderate SINR environment that favors a partial or no CSI-basedapproach. This is because in high mobility the fade rates are also high, leading toless accurate channel estimates. Although the feedback frequency can beincreased so that the transmitter has meaningful channel estimates faster thanchanges in the channel, this would significantly increase the overhead.

10.5 MIMO Advantages and Key Performance Issues

The biggest advantage of MIMO systems is their ability to provide tremendouscapacity gains under certain conditions compared with other spatial techniques[7, 10–20]. In SIMO/MISO systems capacity improves by about 1 bps/Hzwhen the SINR is doubled, whereas in a MIMO system doubling SNRimproves capacity by ~ N bps/Hz, N = min(MT, MR). Another benefit of MIMOsystems that derives from the increased diversity order is improved link reliabil-ity. As diversity increases, the probability that a given data rate cannot be reli-ability sustained is reduced:

( )P SINRoutage

N= ⋅ −κ (10.14)

where κ is a constant and N is the diversity order. In a SIMO or MISO systemN = MR & MT, respectively, whereas in a MIMO system N = MR * MT.


Table 10.2MIMO Schemes

MIMO Scheme Pros Cons Applications

Transmitter equippedwith channel stateinformation (full CSI)

Best performance Increased overheadsince a CSI feedbackchannel is required

Fixed terminals

Transmitter does not haveCSI (non-CSI)

No CSI feedbackchannel required

Worst performance Mobile terminals

Transmitter has limitedCSI (partial CSI)

A CSI limited feedbackchannel is till required→ reduced overheadrelative to CSI

Performance betweenCSI and non-CSIschemes

Portable terminals

10.6 RF Propagation Characterization

MIMO performance advantages occur when the RF propagation channel isrichly scattered (i.e., multipath, fading) such as in nonLOS applications, whichtypically have good scattering. In LOS cases, the use of cross-polarized antennascan preserve capacity by capturing more multipath. To achieve theorthogonality required in (10.12), which maximizes the capacity, the signalsacross the array elements must be uncorrelated. Therefore, the correlationbetween array elements has a big impact on performance. As we have discussedin previous chapters, this correlation is a function of the array element spacingand angular spread.

10.7 SINR Environment

MIMO gains over diversity systems increase with the SINR, as was demon-strated in Figure 10.7. Figure 10.8 shows how the performance compares forlow to moderate SINR. It follows that in situations where the system is interfer-ence limited (low SINR), MIMO gains will be reduced. In those situations,enhanced interference management techniques, such as power control anddirectional antennas, can be employed to improve a MIMO systemperformance.

Effect of SINR on MIMO Capacity

MIMO capacity gain over that of a SISO system also increases as the SINR isincreased. This is shown in Figure 10.9, where we can see that the gain is mar-ginal for SINRs below 0 dB, but significant gains can be achieved at high

MIMO Systems 299

Figure 10.8 Performance comparison between different spatial techniques at low to moderateSINR.

SINRs. It can be shown that as the SINR → ∞, CMIMO/CSISO → N, where N =min(MT, MR), assuming the same total power for both schemes. This can also beseen in Figure 10.9, where the gain converges toward two and four for the 2 × 2and 4 × 4 cases, respectively. Figure 10.10 also illustrates how MIMO andreceive diversity performances compare for the same total number of antennaelements. We can see that MIMO starts outperforming receive diversity only athigh SINRs, above 10 dB, and this gain improves as the SINR is increased. Wecan summarize the performance of MIMO systems with respect to the SINR inTable 10.3

10.8 Spatial Multiplexing

The combined use of transmit and receive arrays as seen earlier offers capacitiesthat increase linearly with the number of array elements. SM in MIMO systems


Figure 10.9 Benefits of MIMO systems over SISO systems.

Figure 10.10 Comparison between MIMO and receive diversity for four total antennas.

can be employed in cases where high peak data rates and very low service outageprobability are required. The operating requirements necessary to achieve thosegains with SM can be summarized as follows:

• Sufficiently rich signal scattering;

• Data rate much higher than maximum Doppler spread (e.g., fixed orstationary users);

• Moderate to high SNRs.

A comparison between the performance of MIMO systems andbeamforming, receive, and transmit diversity is reported in [21, 22]. Transmitdiversity gains and capacity improvements increase as the number of array ele-ments or diversity branches is increased but the gains diminish beyond four ele-ments [23]. Beamforming performance improves with increasing the number ofelements as the beamwidth decreases and the array gain increases. However,since decreasing the beamwidth below the angle spread does not provide addi-tional gains, there is an upper bound on the number of elements. This is shownin Figure 10.11, where we can see that in macrocells where the AS is generallybelow 15°, the upper bound on the number of elements is 20 for AS of 5° andabout eight for AS of 12°. A depiction of the performance of the differentapproaches discussed in this chapter is shown in Figure 10.12. We can see thatfor users with high SINR (i.e., those close to the base station), SM outperformsall other techniques. In moderate SINR, both beamforming and SM have simi-lar performances, whereas beamforming outperforms all others for users withlow SNR (e.g., at cell edge).

MIMO Systems 301

Table 10.3MIMO Performance Comparison

Low SINR region Some diversity order.

Reduced outage probability → lower link margin required→better coverage.

No significant increase in average data rate.

Moderate to high SINR region Significant increase in average data rates over SIMO (byexploiting parallel channels to increase link throughput).

High SNR→ throughput increases with linkdimensionality (see Figure 10.9).

Large diversity reduces variability of link data rate.

10.9 Conclusion

MIMO techniques are being widely proposed as a key technique to enhance theradio channel capacity of cellular systems. This technology can enhance theSINR and improve spectral efficiency by enabling multistream transmission. Tofully use these potential features, MIMO technology should be combined withflexible link adaptation, also known as adaptive modulation and coding mecha-nisms, which can map high SINR values into high user data rates. That is why it


Figure 10.11 Upper bound on number of elements versus AS.

Diversity

Beamforming

Diversity

Beamforming

SpatialMultiplexing

MIMO

Diversity

High SNRuser close to BS

Moderate SNR Low SNRuser close to cell edge

Beamforming

SpatialMultiplexing

MIMO

SpatialMultiplexing

MIMO

Incr

easi

ngp

eak

data

rate

s

Figure 10.12 Comparison between MIMO, beamforming, and diversity.

is natural to combine MIMO together with the HSDPA, a new feature that ispart of the Release 5 specifications of the 3GPP WCDMA/UTRA-FDD stan-dard. HSDPA takes the maximum peak data rate in a 5-MHz WCDMA carrierto 14.4 Mbps from the currently commercial 384 Kbps. Combining MIMOwith HSDPA improves the maximum data rate up to 21.6 Mbps in the same5-MHz bandwidth [24]. Even though it might appear that few mobile applica-tions could require this very high peak data rate, the real motivation for usingMIMO is to actually increase the sector throughput. In currently deployedWCDMA networks, the sector throughput is typically around 0.9–1 Mbps,which improves by about twofold with HSDPA [25]. Fixed beamforming withHSDPA has been studied in [26]. A capacity gain of at least 2.5 was reportedusing four beams. That would further improve the sector throughput withHSDPA from around 2 Mbps to ~ 5 Mbps. It is expected that by using MIMOschemes, based on the comparisons we have shown earlier in this chapter, thesector throughput can be increased much further.

References

[1] Foschini, G. J., and M. J. Gans, “On Limits of Wireless Communications in a FadingEnvironment When Using Multiple Antennas,” Wireless Pers. Commun., Vol. 6, March1998, pp. 311–335.

[2] Foschini, G. J., “Layered Space-Time Architecture for Wireless Communication in a Fad-ing Environment When Using Multielement Antennas,” Bell Labs Tech. J., Fall 1996, pp.41–59.

[3] Telatar, E., “Capacity of Multiantenna Gaussian Channels,” AT&T Bell Laboratories,Tech. Memo, June 1995.

[4] Raleigh, G., and J. M. Cioffi, “Spatial-Temporal Coding for Wireless Communications,”IEEE Trans. Commun., Vol. 46, 1998, pp. 357–366.

[5] Bölcskei, H., D. Gesbert, and A. J. Paulraj, “On the Capacity of OFDM-Based SpatialMultiplexing Systems,” IEEE Trans. Commun., Vol. 50, February 2002, pp. 225–234.

[6] Gesbert, D., et al., “From Theory to Practice: An Overview of MIMO Space-Time CodedWireless Systems,” IEEE Journal on Selected Areas in Communications, Vol. 21, No. 3,April 2003, pp. 281–302.

[7] Shiu, D., et al., “Fading Correlation and Its Effect on the Capacity of MultielementAntenna Systems,” IEEE Trans. Commun., Vol. 48, March 2000, pp. 502–513.

[8] Paulraj, A., et al., Introduction to Space-Time Wireless Communications, New York: Cam-bridge University Press, May 2003.

[9] Smith, P. J., and M. Shafi, “Waterfilling Methods for MIMO Systems,”Proc. 3rd Austra-lian Communication Theory Workshop, Canberra, Australia, 2002.

[10] Paulraj, A., and C. B. Papadias, “Space-Time Processing for Wireless Communications,”IEEE Signal Processing Mag., Vol. 14, November 1997, pp. 49–83.

MIMO Systems 303

[11] Winters, J. H., “On the Capacity of Radio Communication Systems with Diversity in aRayleigh Fading Environment,” IEEE J. Select. Areas Commun., Vol. SAC-5, June 1987,pp. 871–878.

[12] Foschini, G. J., et al., “Simplified Processing for Wireless Communication at High Spec-tral Efficiency,” IEEE J. Select. Areas Commun.—Wireless Commun. Series, Vol. 17, 1999,pp. 1841–1852.

[13] Lozano, A., and C. Papadias, “Layered Space Time Receivers for Frequency SelectiveWireless Channels,” IEEE Trans. Commun., Vol. 50, January 2002, pp. 65–73.

[14] Foschini, G. J., et al., “Analysis and Performance of Some Basic Space-Time Architec-tures,” IEEE VTS 54th Vehicular Technology Conference, 2001. VTC Fall 2001 Vol. 2pp. 1220–1224.

[15] Blum, R., J. Winters, and N. Sollenberger, “On the Capacity of Cellular Systems withMIMO,” Proc. IEEE Vehicular Technology Conf., Atlantic City, NJ, October 2001.

[16] Catreux, S., P. F. Driessen, and L. J. Greenstein, “Attainable Throughput of an Interfer-ence-Limited Multiple-input Multiple-Output Cellular System,” IEEE Trans. Commun.,Vol. 48, August 2001, pp. 1307–1311.

[17] Dai, H., A. Molisch, and H. V. Poor, “Downlink Capacity of Interference LimitedMIMO Systems with Joint Detection,” IEEE Trans. Wireless Commun., submitted forpublication. Vol. 47, Issue No. 2, pp. 173–176, 1999.

[18] Driessen, P. F., and G. J. Foschini, “On the Capacity Formula for Multiple-input Multi-ple-output Wireless Channels: A Geometric Interpretation,” IEEE Trans. Commun., Vol.47, February 1999, pp. 173–176.

[19] Andersen, J. B., “Array Gain and Capacity for Known Random Channels with MultipleElement Arrays at Both Ends,” IEEE J. Select. Areas Commun., Vol. 18, November 2000,pp. 2172–2178.

[20] Lizhong Zheng; Tse, Diversity and Multiplexing: A Fundamental Trade-Off in Multi-ple-Antenna Channels,” IEEE Trans. on D.N.C. Information Theory, Vol. 49, No. 5, May2003, pp. 1073–1096.

[21] Lozano, A., F. R. Farrokhi, and R. A. Valenzuela, “Lifting the Limits on High-SpeedWireless Data Access Using Antenna Arrays,” IEEE Commun. Mag., Vol. 39, September2001, pp. 156–162.

[22] “Emerging Wireless SIG–New Multiple-Input-Multiple-Output (MIMO) Smart AntennaTechnology to Boost Wireless Bandwidth, Capacity and Range,” General Atomics,http://www.sdtelecom.org/comdir/Documents.cfm?NID=3572&ckid=2.

[23] 3GPP, Tx Diversity Solutions for Multiple Antennas Release 5, Tech. Rep. 3G TR 25.869 v1.0.0, 2001.

[24] 3GPP, Multiple-Input Multiple Output in UTRA, Tech. Rep. TR 25.876 V1.5.1, 2004.

[25] Holma, H., et al., WCDMA for UMTS: Radio Access for Third Generation Mobile Commu-nications, 3rd ed., New York: John Wiley & Sons, 2004.

[26] Pedersen, K. I., and P. E. Mogensen, “Performance of WCDMA HSDPA in aBeamforming Environment Under Code Constraints,” IEEE 58th Vehicular TechnologyConference, Vol. 2, October 6–9, 2003, pp. 995–999.


List of Acronyms

2G second generation

3G third generation

3GPP Third Generation Partnership Project

3GPP2 Third Generation Partnership Project 2

AA adaptive antenna

AC admission control

AF array factor

AMPS advanced mobile phone service

AMR adaptive multirate

AOA angle of arrival

AOD angle of departure

ARQ automatic repeat request

AS angle spread

AWGN additive white Gaussian noise

BCCH broadcast control channel

BER bit error rate

BLER block error rate

BPSK binary phase shift keying

305

BS base station

BSC base station controller

CAPICH common auxiliary pilot channel

CDF cumulative distribution function

CDM code division multiplexing

CDMA code division multiple access

C/I carrier-to-interference ratio

CLTD closed loop transmit diversity

COST European COoperation in the field of Scientific and Technicalresearch

CPICH common pilot channel

CRC cyclic redundancy check

DAPICH dedicated auxiliary pilot channel

DCCH dedicated control channel

DCH dedicated channel

DL downlink

DOA direction of arrival

DPCCH dedicated physical control channel

DPCH dedicated physical channel

DPDCH dedicated physical data channel

DTCH dedicated traffic channel

EIA electronic industry alliance

EIRP effective isotropic radiated power

EV-DO EVolution Data Optimized

FACH forward access channel

F-CAPICH forward common auxiliary pilot channel

FCC Federal Communications Commission

FCH fundamental channel

FDD frequency division duplex


FDMA frequency division multiple access

FEC forward error correction

FER frame error rate

FL forward link

GSM global system for mobile communication

HHO hard handoff

HSDPA High Speed Downlink Packet Access

IMT International Mobile Telecommunications

IR incremental redundancy

IS interim standard

ITU International Telecommunication Union

Kbps kilobits per second

L1 OSI layer 1: physical layer

LMS least mean square

LOS line of sight

LS least squares

MAC medium access control

MAPL maximum allowable path loss

MEG mean effective gain

MIMO multiple input multiple output

MISO multiple input single output

MMSE minimum mean square error

MRC maximal ratio combining

MS mobile station

MVDR minimum variance distortionless response

NF noise figure

OC optimum combining

OLTD open loop transmit diversity

OTD orthogonal transmit diversity


OVSF orthogonal variable spreading factor

PAS power azimuth spectrum

P-CCPCH primary common control channel

PCG power control group

PCH paging channel

PG processing gain

PN pseudorandom noise

PSC primary synchronization code

QOF quasi-orthogonal function

QPCH quick paging channel

QPSK quadrature phase shift keying

R99 Release 99

RBFNN radial basis function neural network

RBS radio base station

RF radio frequency

RL reverse link

RLS recursive least squares

RMS root mean square

RNC radio network controller

SCCH supplemental code channel

SCH supplemental channel

S-CPICH secondary common pilot channel

SDMA space division multiple access

SF spreading factor

SHO soft handoff

SIMO single input multiple output

SINR signal-to-interference plus noise ratio

SISO single input single output

SM spatial multiplexing


SNR signal-to-noise ratio

SS spatial signature

STS space time spreading

STTD space time transmit diversity

TD transmit diversity

TDD time division duplex

TDM time division multiplexing

TDMA time division multiple access

TIA Telecommunication Industry Association

TXAA transmit antenna array

UE user equipment

UL uplink

UMTS universal mobile telecommunication system

VA voice activity factor

WCDMA wideband code division multiple access


About the Author

Ahmed El Zooghby received his Ph.D. in electrical engineering from the Uni-versity of Central Florida in 1999 and his B.S. and M.S. in electrical engineeringfrom Alexandria University in Egypt in 1991 and 1994, respectively. He wasa faculty member at the Arab Academy for Science and Technology and Mari-time Transport from 1992–1995. Dr. El Zooghby is currently a UMTS prod-uct manager with QUALCOMM CDMA Technologies, working on thedevelopment of advanced chipset solutions for current and future 3G wirelesscommunications systems. Dr. El Zooghby has also held previous positions atEricsson Wireless Communications, Inc., where he worked on CDMA infra-structure. At Ericsson, he was responsible for managing and supervising techni-cal issues related to defining, configuring, planning, and evaluating requirementoutlines of wireless CDMA infrastructure systems for key Ericsson customers.Dr. El Zooghby’s main research interests include smart antennas, neuralnetwork–based adaptive array processing, and direction finding. Dr. ElZooghby has published more than 25 transaction and conference papers onsmart antennas. He has also lectured and offered several seminars about smartantennas.

311

Index

1xEV-DO, 40–43downlink, 41, 42link adaptation, 43optimization, 41sector power usage comparison, 42

1xRTTdefined, 37sector power usage comparison, 42

2G networks, 53-dB beamwidth. See Half-power beamwidth

3Gcellular users forecast, 6evolution paths towards, 5technology comparisons, 6

3G networks, 36–49, 191–227coverage/capacity limitations, 209–11data applications, 203–9impact on uplink coverage/capacity,

211–26link budgets and coverage, 192–97voice services, 197–203

AAccess channel (ACH), 52Adaptive arrays, 10–11, 117–50

downlink beamforming, 142–49downlink processing, 132–41uplink processing, 117–32See also Antenna arrays

Adaptive beamforming, 122–32architecture illustration, 259

beam steering, 124blind, 161fixed multiple beams vs., 130–32maximum SINR, 125–26MMSE, 126–27MVDR, 127–28null steering, 124–25optimum SINR, 128–30See also Beamforming

Adaptive cell sectorization, 114–15defined, 114load balancing, 115See also Sectorization

Adaptive modulation and coding (AMC), 3,43

Admission control (AC) algorithm, 246forward link/downlink, 255–58reverse link/uplink, 253–54system impact, 248uplink beamforming, 261

Advanced mobile phone service (AMPS), 2,14

Angle diversity, 118–20concept illustration, 120defined, 119performance, 120

Angle of arrival (AOA)of clusters and paths, 73, 76computing, 76distribution, 70spread, 73

313

Angle spread, 147base station, 69–73beamforming gains vs., 222impact, 77–80impact on optimum beamforming,

175–81mobile station, 74

Antenna arrays, 89–105array factor, 89broadside, 91–92coordinate system, 90directivity, 99element spacing, impact of, 93–96end-fire, 91–92first null beamwidth, 96–97fundamentals, 89–105gain, 100, 271half-power beamwidth, 97–99interference reduction capabilities, 271number of elements, impact, 92–93planar, 101–5radiation pattern, 89, 90trade-off analysis, 100

Asymmetric beamformers, 107Auxiliary pilots, 242Azimuth power spectrum, 69–74

base station, 69–73mobile station, 74

BBase stations

additional equipment at, 249–50angle spread, 69–73azimuth power spectrum, 69–73parameters (link budget), 193

Beamformers, 105–74 x 4 Butler matrix, 106asymmetric, 107defined, 106orthogonal, 106spatial filtering, 109–11symmetric, 107

Beamforming, 105–7, 184–85, 222–24,296

adaptive, 122–32approach comparison, 150blind adaptive, 161channel estimation, 183–84DOA-based, 146–47, 170–75downlink, 142–49, 181–82

gains vs. angular spread (WCDMA), 222MIMO comparison, 301, 302optimum, 175–81performance, 145spatial signature-based, 145–46transmission strategy, 297transmit diversity performance

comparison, 227See also Antenna arrays

Beam pointing, 258Beam steering, 124, 258–61

advantages/disadvantages, 260channel estimation at mobile, 259–60defined, 258transmission strategy, 297

Beam transition, 251–52Beamwidths

half-power, 97–99null-to-null (NNBW), 96–97tailored, 89

Blind adaptive beamforming, 161Broadside arrays, 91–92

defined, 91polar pattern, 94–96radiation patterns along z-axis, 92See also Antenna arrays

Butler matrix, 107–94 x 4 beamformer, 106defined, 107phases at array elements, 108phase shifters, 107power dividers, 107

CCapacity

CDMA, 21, 24, 84cell, 54coverage trade-off, 55–57downlink, 132, 133downlink, of CDMA sector, 286downlink gains, 284–86downlink improvements, 221, 224embedded, 53–55gain, 220, 282, 283improvements vs. load factor, 217improvement techniques, 212increase vs. gain, 214isolated, 24–25limited scenarios, 211limiting links for, 209–11


MIMO, 294, 299–300normalized Shannon, 291power reduction trade-off, 215relative sector, 286smart antennas impact, 211–26STTD, 220voice coverage vs., 205WCDMA cell, 54See also Coverage

CDMA2000, 3, 5, 7, 37–44, 1911x-EV-DO, 40–431XRTT, 37, 42access procedure, 239cell capacity, 54defined, 37fast forward link power control

mechanism, 37–38FEC, 37mobile call states, 237–38M-sequences, 235multimedia services, 38multiplexing, 233open loop vs. closed loop transmit

diversity gains, 218OTD vs. STS transmit diversity gains,

218packet data services, 241pilot channels, 241–42protocol stack, 234radio links, 192transmit diversity gains vs. geometry, 219voice link budget, 199–200, 202–3See also Code division multiple access

(CDMA); Third-generationsystems

CDMAOne operators, 5Cell breathing, 56Cells

area confidence, 195–96CDMA layout, 49edge, 196embedded capacity, 53–55isolated capacity, 24–25macrocells, 63, 70, 221microcells, 63, 225–26minicells, 63picocells, 63

Channel estimation, 183–84, 251Channel quality indicator (CQI), 3

Closed loop power control, 35–36Closed loop transmit diversity (CLTD), 134,

137–40capacity increases, 225CL1 mode, 141CL2 mode, 141defined, 137OLTD schemes vs., 138TXAA, 137–38TXAA architecture, 139TXAA operation, 140–41See also Transmit diversity

Cochannel interference, 18–19Cochannel separation, 18Code allocation, 248Code division multiple access (CDMA), 2,

20–36access state, 52–53access technology, 20acquisition state, 49–52advantage, 21basic procedures, 49–53call setup, 52–53call states, 50capacity, 21, 24, 84carrier frequency, 22cell breathing, 56cell layout, 49codes, 25–29coverage vs. capacity trade-off, 55–57defined, 20direct-sequence (DS-CDMA), 43downlink capacity, 133embedded cell capacity, 53–55forward link channels, 30–32fundamentals, 21–36idle state, 52IS-95, 24, 29–36isolated cell capacity, 24–25multipath fading, 55power control, 32–34RAKE receivers, 32receiver block diagram, 30receivers, 22reverse link channels, 32reverse link closed loop power control,

35–36reverse link open loop power control,

34–35

Index 315

Code division multiple access (CDMA)(continued)

sectorization gain (SG), 85–86soft capacity, 84soft handoff, 36systems, 20–21TDD-based, 181traffic loading, 196–97traffic or dedicated state, 53transmitter block diagram, 30See also Multiple access techniques

Coherence bandwidth, 65–66Combining

maximal ratio, 272optimum, 272–78performance improvement, 281selection diversity, 272techniques, 272

Common pilot channel (CPICH), 44Confidence, 195–96Congestion control (CC), 246Conjugate gradient (CG) algorithm, 164–67

approximation, 165computational load, 167forgetting factor, 165subsequent updates, 165See also Reference signal methods

Constant modulus algorithm (CMA), 162Conventional sectorization, 83–89

limitations, 88–89scheme illustration, 88three-sector pattern, 84

COST-231 model, 67Coupling

loss, 280mutual effects, 279–80

CoverageCDMA trade-off, 55–57downlink improvements, 221, 224improvements with multiple fixed beam

antennas, 216improvement techniques, 212limited scenarios, 210–11limiting links for, 209–11link budgets and, 192–97smart antennas impact, 211–26voice, vs. capacity, 205See also Capacity

Cumulative distribution function (CDF),118

Cyclic redundancy check (CRC), 47Cyclostationary algorithms, 163–64

DData applications, 203–9Data multiplexing, 233–35Decision-directed algorithm, 162–63Dedicated channel (DCH), 46Dedicated physical channel (DPCH), 45Dedicated physical control channel

(DPCCH), 45Dedicated physical data channel (DPDCH),45, 184Dedicated pilot bits, 235, 243Delay spread, 65Delta rule, 173Direction on arrival (DOA), 10

characterization, 68downlink beamforming and, 181multipath, 184

Directivityarrays, 99planar arrays, 104–5

direct-sequence CDMA (DS-CDMA), 43Diversity

angle, 118–20gain, 118, 119, 220, 268mobile receive (MRD), 289polarization, 270–71selection, 272space, 120techniques, 117–18transmit, 93, 134–41, 217–22

DOA-based beamforming, 146–47neural network, 170–75PAS, 147uplink channel correlation matrix, 146See also Beamforming

Doppler shift, 66, 147Doppler spread, 66Downlink

admission control, 255–58beamforming, 142–49, 181–82budgets, 198–203capacity, 132, 133capacity gains, 284–86capacity improvements, 221, 224capacity of CDMA sector, 286


coverage improvements, 221, 224initial power setting, 253load reduction with STTD, 220smart antenna capacity impact, 216–26

Downlink processing, 132–41transmit diversity concepts, 134transmit diversity in 3G CDMA

standards, 134–41See also Adaptive arrays

Downlink shared channel (DSCH), 46Dual-antenna

diversity mobile station, 279performance improvements, 280–83relative sector capacity vs. percentage

penetration, 286

EElement pattern, 101Element spacing

half-power beamwidth as function, 98impact of, 93–96spatial correlation vs., 144

End-fire arrays, 91–92Envelope correlation, 269, 270Equal-gain combining, 121EV-DO systems, 7

FFade margin (link budget), 194–95Fading, 34–35

envelope, 77, 78, 79, 80fast, 34, 35flat, 55, 66frequency, 77frequency selective, 66as function of time, 34multipath, 55Rayleigh, 35, 118Rician, 120slow, 34, 35spatial, 77temporal, 77

Fast fading, 34Fast forward link power control, 37–38Fixed beam antennas, 9–10, 83–115

adaptive cell sectorization, 114–15arrays, 89–105beamforming, 105–7Butler matrix, 107–9conventional sectorization, 83–89

multiple, 113system impact, 248–58See also Smart antennas

Fixed beamforming, 110Flat fading, 55, 66Forward error correction (FEC), 37Forward link channels, 30–32

forward traffic, 32paging, 31pilot, 31sync, 31types of, 30

Forward traffic channels, 32Four-finger RAKE receivers, 278Frame error rate (FER), 54Frequency calibrated (FC) algorithm, 182Frequency division duplex (FDD), 43, 146

defined, 43gaps, 148

Frequency division multiple access (FDMA),1

concept illustration, 15spectrum division, 15systems, 14–15

Frequency planning, 16Frequency reuse, 16–18

defined, 17plans, 18

Frequency selective fading, 66Fundamental channel (FCH), 53

GGain(s)

array, 100, 271beamforming, angular spread vs., 222capacity, 220, 282, 283capacity increase vs., 214diversity, 118, 119, 220, 268downlink capacity, 284–86mean effective (MEG), 271–72MIMO, 299power reduction vs., 214

Gaussian minimum-shift keying (GMSK),164Gold codes, 235Grating lobes, 93

HHalf-power beamwidth, 97–99

defined, 97

Index 317

Half-power beamwidth (continued)formula, 98as function of element spacing, 98scanning angle effect, 99

Handoffshard (HHO), 247impact on capacity gains, 282soft, 36, 246–47softer, 86, 87, 246–47zones, 89

Hard handoff (HHO), 247, 261Hata’s model, 66–67High-speed data transfer, 239–40

mobility procedures, 238–40reestablish procedures, 240

High-speed dedicated physical controlchannel (HS-DPCCH), 47

High Speed Downlink Packet Access(HSDPA), 7, 45–49

adaptive modulation and coding, 47defined, 45–46for end users, 46fast scheduling, 48–49hybrid-ARQ with soft combining, 47–48peak data rate, 46scheduler algorithms, 48

High-speed downlink shared channel(HS-DSCH), 46, 47

High-speed shared control channel(HS-SCCH), 47

High-speed uplink packet access (HSUPA), 3Hybrid-automatic repeat request (HARQ), 3

defined, 47–48functionality implementation, 48

Hybrid couplers, 108

IIndoor office radio environment, 64Initial power setting, 245–46

downlink, 253fixed beam approach, 252–53preamble, 252–53system impact, 248uplink, 252–53uplink beamforming, 261

Interferenceaverage, 22cochannel, 18–19reduction capabilities, 271uplink reduction, 216

worst-case, 22Interference-limited systems, 132International Telecommunications Union

(ITU), 4–5IS-95 CDMA, 24, 29–36

defined, 29forward link channels, 30–32IS-2000 vs., 44RAKE receivers, 32reverse link channels, 32WCDMA vs., 44See also Code division multiple access

(CDMA)IS-2000, 38–40

forward link physical channels, 38, 39IS-95 vs., 44reverse link physical channels, 38–40reverse link pilot, 40reverse link pilot frame structure, 41WCDMA vs., 44

LLagrange multiplier method, 167–69Least mean square (LMS) algorithm, 159–61Least squares (LS) algorithm, 161–62Limiting link

capacity scenarios, 211coverage scenarios, 210–11evaluation, 212

Line-of-sight (LOS), 61, 274Link adaptation, 43Link budgets, 192–97

base station parameters, 193datalink, 206–7fade margin, 194–95margins, 193maximum allowable path loss (MAPL),

194mobile station parameters, 192receiver sensitivity, 194signal bandwidth, 194system parameters, 193thermal noise density, 194voice services, 197–203

Load reduction, 220Lyapunov functions, 170–71

MMacrocells, 63

base station model, 70


STTD, 221Margins, 193

fade, 194–95list of, 193

Maximal ratio combining, 272Maximum likelihood (ML) principle, 10Maximum ratio combining (MRC), 10, 131,

132Mean effective gain (MEG), 271–72

defined, 271reduction in, 272whip antenna, 271

Microcells, 63, 225–26MIMO, 11

advantages, 298approaches, 297–98beamforming and diversity comparison,

302capacity, 299–300channel capacity, 294channel matrix, 293conclusion, 302–3data transmission, 290gains, 299introduction, 289–90mixed approach, 298performance, 297performance comparison, 301performance issues, 298principles, 290–95receive diversity comparison, 300RF propagation characterization, 299scheme illustration, 294sector throughput, 303SINR environment, 299–300SISO systems vs., 300spatial multiplexing, 295, 300–302systems, 289–303transmission strategies, 295–97wireless communication systems, 290

Minicells, 63Minimum mean square error (MMSE), 10,

125–26criterion, 126disadvantage, 127gradient, 126

Minimum variance distortionless response(MVDR), 127–28

beamformer, 10, 128

defined, 128Mobile call states, 237–38

CDMA2000, 237–38illustrated, 238WCDMA, 237

Mobile market, growth, 3–4Mobile radio

environments, 63–64indoor office, 64outdoor to indoor environment, 64propagation channel, 62vehicular environment, 64

Mobile receive diversity (MRD), 289Mobile station parameters (link budget), 192Mobile stations

angle spread, 74azimuth power spectrum, 74diversity-enable, 285smart antennas, 265–86SNR, 134, 139

Mobility procedures, 239–40Multimedia messaging service (MMS), 3Multipath channels, 64–65Multipath DOAs, 184Multipath fading, 55

channel characterization, 65–66impact, 55

Multiple access techniques, 14–21CDMA, 20–36cochannel interference, 18–19FDMA, 14–15frequency reuse, 16–18TDMA, 15–16

Multiple-antenna MS design, 268–72Multiple fixed beam systems, 113

adaptive beamforming vs., 130–32additional equipment at base station,

249–50beam transition, 251–52channel estimation, 251digital baseband architecture, 250downlink capacity improvements, 224downlink coverage improvements, 224downlink/uplink coverage/capacity im-

provements, 223forward link/downlink admission control,

255–58passive networks architecture, 249power control, 252

Index 319

Multiple fixed beam systems (continued)preamble initial power setting, 252–53radio resource management, 254–55reverse link/uplink call admission control,

253–54scrambling code/PN offset assignment,

250soft-handoff procedure, 252uplink capacity vs. load factor, 217uplink coverage improvements, 216uplink interference reduction, 216

Multiple input multiple output. See MIMOMultiple input single output (MISO),

292–93defined, 292scheme illustration, 292

Mutual coupling effects, 279–80

NNear-far problem, 23Neural network DOA-based beamforming,

170–75Noise-dominated systems, 131–32Nonaccess stratum (NAS) layer specification,

233Nonline-of-sight (NLOS), 62, 274Nordic Mobile Telephone (NMT), 14Normalized Shannon capacity, 291Null steering beamforming, 124–25

defined, 124disadvantages, 125optimum SINR vs., 275–77See also Beamforming

Null-to-null beamwidth (NNBW), 96–97

OOkumura-Hata propagation models, 66–67

COST-231 model, 67Hata’s model, 66–67

Open loop power control, 34–35Open loop transmit diversity (OLTD),

134–35Optimum combining (OC)

conclusions, 280–82defined, 273performance, 282

Orthogonal beamformers, 106Orthogonal spreading factor codes (OVSF),

43

Orthogonal transmit diversity (OTD),135–36

defined, 135illustrated, 135See also Transmit diversity

Outdoor to indoor radio environment, 64

PPacket data services, 240–41

CDMA2000 approach, 241WCDMA approach, 241

Paging channel, 31Paging indication channel (PICH), 45Path loss models, 66–67Personal communications systems (PCS), 1Physical common packet channel (PCPCH),

45Physical layer, 233–37

data multiplexing, 233–35formatting, 235–37frame structures, 236interaction with upper layers, 234–35transmit chain, 235

Physical random access channel (PRACH),45

Picocells, 63Pilot channels, 31, 241–43

auxiliary transmit diversity, 232CDMA2000, 241–42common (CPICH), 243forward transmit diversity, 242primary common (P-CPICH), 243secondary common (S-CPICH), 243WCDMA, 243

Pilot measurement, 258Pilot pollution, 114Planar arrays, 101–5

array factor, 102defined, 101directivity, 104–5geometry, 103See also Antenna arrays

Planar-inverted F antenna (PIFA), 271Polarization diversity, 270–71Power azimuth spectrum (PAS), 147, 149Power control, 32–34

fast forward link, 37–38fixed beam approach, 252inner loop, 35reverse link closed loop, 35–36


reverse link open loop, 34–35uplink beamforming, 260

Power control bit (PCB), 35Power control (PC) algorithm, 244Power reduction, 214–15

capacity increase trade-off, 215gain vs., 214

Primary common control physical channel(P-CCPCH), 44

Protocol stacks, 232–37CDMA2000, 234WCDMA, 234

QQuadrature phase shift keying (QPSK), 3Quasi-orthogonal functions (QOFs), 222

RRadial basis function neural network

(RBFNN), 171, 172–73block diagram, 171contents, 172–73delta rule, 173input-output mapping, 173performance phase, 173, 174–75training data, 174

Radial-basis functions (RBF), 172Radiation patterns, 142–43

four-element array, 143two-element array, 142

Radio configurations (RC), 38Radio environments, 63–64Radio network algorithms, 244–47

admission control (AC), 246congestion control (CC), 246hard handoff (HHO), 247initial power setting, 245–46power control (PC), 244smart antennas’ impact on, 262soft/softer handoff, 246–47

Radio network controller (RNC), 46Radio resource management, 254–55Rake combiners, 184RAKE receivers, 32

block diagram, 33eight-finger, 281fingers, adding, 279four-finger, 278searcher finger, 32size, 278–79

vector, 182Rayleigh fading, 35, 118Recursive least squares (RLS) algorithm, 161Reference signal methods, 159–70

blind adaptive beamforming, 161CG algorithm, 164–67CMA algorithm, 162comparison, 169–70cyclostationary algorithms, 163–64decision-directed algorithm, 162–63Lagrange multiplier method, 167–69LMS algorithm, 159–61LS algorithm, 161–62RLS algorithm, 161SCORE algorithm, 164

Reverse linkadmission control (AC) algorithm,

253–54channels, 32closed loop power control, 35–36open loop power control, 34–35physical channels, 38–40, 45pilot, 40

Rician fading, 120

SScalloping, 112, 113SCORE algorithm, 164Scrambling, 235, 250Secondary common control physical channel

(S-CCPCH), 44Sectorization

adaptive cell, 114–15conventional, 83–89defined, 83efficiency, 87techniques, 8–9

Sectorization gain (SG), 84, 85–86CDMA, 85–86defined, 84illustrated, 86

Selection diversity, 272Short message service (SMS), 3Signal-to-interference and noise ratio (SINR)

maximum, 125–26mean, 274MIMO, 299–300optimum, 128–30, 275–77

Signal-to-interference ratio (SIR), 19

Index 321

Signal-to-noise ratio (SNR)high, 278maximum, 147–49mobile stations, 134, 139

Single input multiple output (SIMO)systems, 291–92

defined, 291illustrated, 292low correlation, 292normalized Shannon capacity, 291

Single input single output (SISO) systems,291, 300

Slow fading, 34, 35Smart antennas

advanced spatial techniques, impact of,247–58

benefits, 7–8downlink capacity impact, 216–26fixed beam, 9–10, 83–115impact on radio network algorithms, 262impact on uplink coverage and capacity,

211–26integration in system design, 261performance impacts, 68, 215sectorization techniques, 8–9switched beam, 9–10, 111–13system aspects, 9, 231–62types of, 8–9on uplink, 215why?, 7

Softer handoff, 246–47defined, 86overlap areas, 87uplink beamforming, 261See also Handoffs

Soft handoff, 36, 246–47basis, 246candidate set, 36defined, 36in fixed beam mode, 252neighbor set, 36state, 247uplink beamforming, 261See also Handoffs

Space diversity performance, 120Space division multiple access (SDMA), 8Space-time coding (STC), 289Space-time spreading (STS)

defined, 136

illustrated, 137transmission matrix, 136

Space-time transmit diversity (STTD),136–37

capacity increases, 225defined, 136downlink capacity improvements, 221downlink coverage improvements, 221downlink load reduction with, 220gain, 220illustrated, 138macrocell, 221Walsh code, 137See also Transmit diversity

Spatial channel modeling, 67–74AOA, 71–73application in system simulations, 74–76base station azimuth power spectrum,

69–71distribution of clusters/scatterers, 69number of clusters, 69parameters, 68

Spatial filtering, with beamformers, 110–11Spatial multiplexing (SM), 289, 295,

300–302Spatial signature-based beamforming,

145–46Spreading, 235Spread spectrum

composite signals, 28processing gain, 22, 84signal effect, 28, 29signals, 26, 27

Stop and Wait (SAW) protocol, 47Superframes, 31Switched beam antennas, 9–10, 111–13

advantages, 111–12architecture, 112defined, 111limitation, 112–13

Symmetric beamformers, 106–7Synchronization channel (SCH), 50System parameters (link budget), 193

TThird Generation Partnership Project 2

(3GPP2), 70Third-generation systems, 36–49

CDMA2000, 37–44HSDPA, 45–49


WCDMA, 43–45Time division duplex (TDD), 43, 146Time division multiple access (TDMA), 2

concept illustration, 16defined, 15disadvantages, 16systems, 15–16

Time-division-multiplexed (TDM)waveform, 41, 42

Time-multiplexed power control (TPC), 235Total Access Communications System

(TACS), 14Traffic loading, 196–97Transmission strategies, 295–97

beamforming, 296beam steering, 297uniform power allocation, 296–97water filling, 296

Transmit adaptive antennas (TXAA),137–38, 218

architecture, 139diversity gain, 140operation, 140–41use of, 138

Transmit diversity, 93, 134–41, 217–22in 3G CDMA standards, 134–41beamforming performance comparison,

227closed loop, 134, 137–40concepts, 134gains vs. geometry (CDMA2000), 219open loop, 134–35open loop vs. closed loop gains

(CDMA2000), 218open loop vs. closed loop gains

(WCDMA), 219orthogonal, 135–36OTD vs. STS gains, 218space-time, 136–37system impact, 247–48

Transport format combination indicator(TFCI), 235

UUniform power allocation, 296–97Uplink

call admission control, 253–54capacity improvements vs. load factor,

217coverage improvements, 216

interference reduction, 216preamble power setting, 252–53smart antennas on, 215

Uplink beamforming, 260–61admission control, 261initial power setting, 261power control, 260soft/softer/hard handoff, 261See also Beamforming

Uplink budgets, 198Uplink processing, 117–32

adaptive beamforming, 122–32angle diversity, 118–20diversity techniques, 117–18maximum ratio combining, 121See also Adaptive arrays

User-specific beamforming methods, 244

VVector RAKE receivers, 182Vehicular radio environment, 64Voice services, 197–203

CDMA2000, 199–200, 202–3coverage vs. capacity, 205downlink budgets, 198–203uplink budgets, 198voice coverage vs. capacity, 205WCDMA, 200–201, 204–5

Voice traffic, projected growth, 5

WWalsh codes, 25

complementary, 136generation, 25PN codes comparison, 31spreading signal effect, 29STTD, 137

Water filling, 296Weight selection algorithm, 295Weiner-Hopf type calculations, 273Wideband CDMA (WCDMA), 2, 43–45,

191access procedure, 239acquisition steps, 51–52beamforming, channel estimation, 183beamforming gains vs. angular spread,

222cell capacity, 54cell search signals, 50data coverage vs. capacity, 209

Index 323

Wideband CDMA (WCDMA) (continued)data interface rise vs. uplink load, 210datalink budget, 206–7defined, 43DL budget, 207–8downlink channels applicable for

beamforming, 244European bands, 148FDD forward link physical channels,

43–45FDD reverse link physical channels, 45gold codes, 235IS-95 vs., 44IS-2000 vs., 44

mobile call states, 237, 238multiplexing, 233–35open loop vs. closed loop transmit

diversity gains, 219packet data services, 241pilot channels, 243protocol stack, 233, 234radio links, 192voice interface rise vs. uplink load, 210voice link budget, 200–201, 204–5See also Code division multiple access

(CDMA)Wireless communications systems, 1–3,

14–21, 290


Recent Titles in the Artech HouseMobile 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 WirelessCommunications, 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 AugmentationSystems, 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 andLeonard E. Miller

Cell Planning for Wireless Communications, Manuel F. Cátedra andJesús Pérez-Arriaga

Cellular Communications: Worldwide Market Development,Garry A. Garrard

Cellular Mobile Systems Engineering, Saleh Faruque

The Complete Wireless Communications Professional: A Guide forEngineers and Managers, William Webb

EDGE for Mobile Internet, Emmanuel Seurre, Patrick Savelli, andPierre-Jean Pietri

Emerging Public Safety Wireless Communication Systems,Robert I. Desourdis, Jr., et al.

The Future of Wireless Communications, William Webb

GPRS for Mobile Internet, Emmanuel Seurre, Patrick Savelli, andPierre-Jean Pietri

GPRS: Gateway to Third Generation Mobile Networks,Gunnar Heine and Holger Sagkob

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 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 MobileCommunications, 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 DebashisSaha

Mobile Data Communications Systems, Peter Wong andDavid Britland

Mobile IP Technology for M-Business, Mark Norris

Mobile Satellite Communications, Shingo Ohmori, HiromitsuWakana, and Seiichiro Kawase

Mobile Telecommunications Standards: GSM, UMTS, TETRA, andERMES, Rudi Bekkers

Mobile Telecommunications: Standards, Regulation, andApplications, Rudi Bekkers and Jan Smits

Multiantenna Digital Radio Transmission,Massimiliano “Max” Martone

Multiantenna Wireless Communications Systems,Sergio Barbarossa

Multipath Phenomena in Cellular Networks, Nathan Blaunstein andJørgen Bach Andersen

Multiuser Detection in CDMA Mobile Terminals, Piero Castoldi

Personal Wireless Communication with DECT and PWT, John Phillipsand Gerard Mac Namee

Practical Wireless Data Modem Design, Jonathon Y. C. Cheah

Prime Codes with Applications to CDMA Optical and WirelessNetworks, Guu-Chang Yang and Wing C. Kwong

QoS in Integrated 3G Networks, Robert Lloyd-Evans

Radio Engineering for Wireless Communication and SensorApplications, Antti V. Räisänen and Arto Lehto

Radio Propagation in Cellular Networks, Nathan Blaunstein

Radio Resource Management for Wireless Networks, Jens Zanderand Seong-Lyun Kim

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 RadioPeter 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

Third Generation Wireless Systems, Volume 1: Post-Shannon SignalArchitectures, George M. Calhoun

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 andJosef Franz Huber

Understanding Cellular Radio, William Webb

Understanding Digital PCS: The TDMA Standard,Cameron Kelly Coursey

Understanding GPS: Principles and Applications, Elliott D. Kaplan,editor

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 Ojanperäand Ramjee Prasad, editors

Wireless Communications in Developing Countries: Cellular andSatellite Systems, Rachael E. Schwartz

Wireless Intelligent Networking, Gerry Christensen,Paul G. Florack, and Robert Duncan

Wireless LAN Standards and Applications, Asunción Santamaríaand 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

Smart antenna engineering- 2005- ahmed el zooghby

Engineering

cdma systems

types of smart antennas

g systems

benefits of smart antennas

fdma systems

generation systems

tdma systems

spatial channel modeling