characterization of ultra wideband communication channels

Characterization of Ultra Wideband Communication

Channels

by

Ali Hussein Muqaibel

Time Domain and RF Measurement Laboratory The Mobile and Portable Radio Research Group

Dissertation Submitted to the Faculty of

The Bradley Department of Electrical and Computer Engineering Virginia Polytechnic Institute and State University

in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY in

Electrical Engineering

Co-Chairmen

Dr. Sedki M. Riad Dr. Brian D. Woerner Dr. Ahmad Safaai-Jazi

Committee Members Dr. Ioannis M. Besieris Dr. William Tranter

Dr. Norris S. Nahman Dr. Werner Kohler

March 5, 2003 Blacksburg, Virginia

© 2003 by Ali Hussein Muqaibel

Characterization of Ultra Wideband Communication Channels

by

Ali Hussein Muqaibel

Abstract

Ultra-wideband (UWB) communication has been the subject of extensive research

in recent years due to its unique capabilities and potential applications, particularly in

short-range multiple access wireless communications. However, many important aspects

of UWB-based communication systems have not yet been thoroughly investigated. The

propagation of UWB signals in indoor environments is the single most important issue

with significant impacts on the future direction, scope, and generally the extent of the

success of UWB technology. The objective of this dissertation is to obtain a more

thorough and comprehensive understanding of the potentials of UWB technology by

characterizing the UWB communication channels. Channel characterization refers to

extracting the channel parameters from measured data. The extracted parameters are used

to quantify the effect of the channel on communication UWB systems using this channel

as signal transmission medium. Data are measured in different ways using a variety of

time-domain and frequency-domain techniques. The experimental setups used in channel

characterization effort also include pulse generators and antennas as integral parts of the

channel, since the pulse shape and antenna characteristics have significant impact on

channel parameters.

At a fundamental level, the propagation of UWB signals, as any electromagnetic

wave, is governed, among other things, by the properties of materials in the propagation

medium. One of the objectives of this research is to examine propagation through walls

made of typical building materials and thereby acquire ultra-wideband characterization of

these materials. The loss and the dielectric constant of each material are measured over a

frequency range of 1 to 15 GHz. Ten commonly used building materials are chosen for

this investigation. These include, dry wall, wallboard, structure wood, glass sheet, bricks,

concrete blocks, reinforced concrete (as pillar), cloth office partition, wooden door, and

styrofoam slab. The work on ultra-wideband characterization of building materials

resulted in an additional interesting contribution. A new formulation for evaluating the

complex dielectric constant of low-loss materials, which involves solving real equations

and thus requiring only one-dimensional root searching techniques, was found. The

results derived from the exact complex equation and from the new formulation are in

excellent agreement.

Following the characterization of building materials, an indoor UWB

measurement campaign is undertaken. Typical indoor scenarios, including line-of-sight

(LOS), non-line-of-sight (NLOS), room-to-room, within-the-room, and hallways, are

considered. Results for indoor propagation measurements are presented for local power

delay profiles (local-PDP) and small-scale averaged power delay profiles (SSA-PDP).

Site-specific trends and general observations are discussed. The results for pathloss

exponent and time dispersion parameters are presented. The analyses results indicate the

immunity of UWB signals to multipath fading. The results also clearly show that UWB

signals, unlike narrowband signals, do not suffer from small scale fading, unless the

receiver is too close to walls. Multipath components are further studies by employing a

deconvolution technique. The application of deconvolution results in resolving multipath

components with waveforms different from those of the sounding pulse. Resolving more

components can improve the design of the rake receiver. The final part of this research

elaborates on the nature of multiple access interference and illustrates the application of

multi-user detection to improve the performance of impulse radio systems. Measured

dispersion parameters and their effects on the multiple access parameters are discussed.

Dedication

All praises goes to Allah, the Creator and Lord of the Universe.

O’ Allah, Have mercy and accept it as a dedication to you

O’ Allah,

Show it to me in the last day with the good deeds

O’ Allah, Great thanks to you for this accomplishment.

Your bounties;

my mother, the memory of my father,

the memory of my lovely brother Abdu-Allah, a lovely wife, a great family,

and special friends, I cannot deny, led to this accomplishment

Dedication إهداء

بسم االله الرحمن الرحيم ..الحمد الله رب العالمين والصلاة والسلام على أشرف الأنبياء والمرسلين وعلى آله وصحبه الطيبين الطاهرين

اللهم تقبل هذا العمل وإجعله في ميزان حسناتي وإغفر لي ما آان فيه من زلل وتقصير

: اللهم لك الحمد على ما أنعمت علي من نعم

رحيمة صابرة،أم

ذآرى عطرة لوالدي رحمه االله وجعل الفردوس مثواه،

وأخي الحبيب عبداالله رحمه االله وأسكنه فسيح جناته،

زوجة طيبة،

أخوة وأخوات وعائلة آريمة،

وصحبة صالحة،

.نعمك الكثيرة وفضلك يارب سبب آل نجاح

iv

Acknowledgments

My unreserved praises and thankfulness are for Allah, the Most Compassionate,

and the Most Merciful. He blessed me with his bounties. May his peace and blessing be

upon the prophet Muhammad, and his family.

I would like to express my high gratitude to my advisors, Dr. Sedki Riad, Dr.

Brian Woerner, and Dr. Safaai-Jazi for their guidance and support. It is my luck to work

closely with them thereby combine their expertise in time domain and RF measurements,

communications, wave propagations and antennas under the topic of “Characterization of

UWB Communication Channels”. They will always be greatly appreciated. I am greatly

indebt to Dr. Riad and Dr. Safaai-Jazi for the many challenges they have placed upon me

as well as their invaluable guidance, support and time they have so kindly offered me

over the past three years during my work in the Time Domain and RF Measurements

Laboratory (TDL).

My dissertation committee comprised of Dr. Ioannis Besieris, Dr. William

Tranter, Dr. Werner Kohler and Dr. Norris Nahman has been very helpful in improving

my proposal and dissertation. I am grateful to them for sharing their time and expertise.

I would also like to thank Dr. Ahmed Attiya, research associate in the Time

Domain and RF Measurement Laboratory, my co-worker Mr. Ahmet Bayram, and Dr.

Jason Yoho from Picoseconds Pulse Labs for their help in establishing and conducting

the measurements. They are also greatly appreciated for improving my dissertation.

It is my pleasure to acknowledge the support and facilities provided by the Time

Domain and RF Measurements laboratory (TDL) and the Mobile and Portable Radio

Research Group (MPRG), both at Virginia Tech.

My mother, my wife, my sisters, my brothers (Muhammed, Abdu-Allah, and

Ahmed) and my uncle Hassan Al-Attas have always been supportive of my educational

endeavors. My friends to name but a few: Dr. Zain Yamani, Dr. Saad Al-Shahrani, Dr.

Maan Kousa, Mr. Yaser Al-Ghahtani, Mr. Samir Al-Ghadhban, Mr. Bassam Al-Dossary,

Mr. Ibraheem Al-Ghahtani, Mr. Abdul-Aziz Al-Saadi, Mr. Ali Al-Ghamdi, Mr. Obaid

Al-Modaf, Mr. Ghassan Al-Regib and Mr. Khalid Al-Attas are truly exceptional people.

I am grateful to them for their support and assistance during my work towards this goal.

v

Acknowledgments (Arabic Translation) شكر وعرفان

نعم والآء عظيمة ، والصلاة والسلام على أشرف الأنبياء الحمد الله رب العالمين، على ما أنعم علي من

. والمرسلين نبينا محمد وعلى آله وصحبه أجمعين

يشرفني أن أتقدم بالشكر الجزيل لمشرفي الأستاذ الدآتور صدقي رياض، والأستاذ الدآتور براين ورنر و

آما شرفني وأسعدني أن أعمل معهم ، فقد . الأستاذ الدآتور أحمد صفائي جزي على ما أسدوا من توجيه وإرشاد

منحني ذلك فرصة رائعة لأن أجمع خبرة الدآتور رياض في مجال القياسات الزمنية وقياسات موجات الراديو،

) الأنتنا(وخبرة الدآتور ورنر في مجال الإتصالات ، بالإضافة إلى خبرة الدآتور صفائي جزي في مجال الهوائيات

، سأظل "تقييم قنوات الاتصال ذات المجال الواسع جدا " رومغناطيسية في وموضوع واحد وانتقال الموجات الكه

آما يسعدني أن أنوه بشكري للدآتور رياض و الدآتور صفائي جزي على ما وضعوا . دائما مدينا لهم ولما قدموه

سعدت فيها بالعمل معهم في معمل أمامي من تحديات بناءه ، وإرشادات غالية، ودعم آريم خلال السنوات الثلاث التي

.قياسات الموجات الزمنية وترددات الراديو

آما يسعدني أن أعبر عن شكري وامتناني للجنة الإشراف المكونة من الأستاذ الدآتور إيانوس بسيريز،

لما قدموه من والأستاذ الدآتور ويليوم ترينتر، والأستاذ الدآتور ويرنر آوهلر، والأستاذ الدآتور نوريس ناهمان،

.وقت وخبرة لتوجيه الدراسة والرسالة التي قمت بها

آما أتقدم بجزيل الشكر والإمتنان للدآتور أحمد عطية سالم، الباحث بمعمل قياس الموجات الزمنية

وترددات الراديو، وآذلك زميل العمل أحمد بايرام، والزميل الدآتور جاسون ياهو، الباحث في مختبرات موجات

بيكوثانية، على ما قدموه من مساعدة للتجهيز والقيام بالقياسات اللازمة لإنجاز هذه الدراسة، وما قدموه من ال

.إقتراحات وتصحيحات للرقي بمستوى البحث إلى الأفضل

إنه لمن دواعي سروري أن أشكر معمل قياس الموجات الزمنية وترددات الراديو، ومجموعة أبحاث

.قل والجوال على ما قدموه من تسهيلات وأجهزةاتصالات الراديو المتن

، ) محمد ،عبداالله ، وأحمد (والدتي الصابرة، وزوجتي الفاضلة تحملن الكثير من أجل دراستي، أخواني

.أخواتي الكريمات ،و خالي الفاضل حسن العطاس أبدوا آل التعاون والتشجيع في جميع مراحل الدراسة

. زين يماني، د . د: ء الذين آان لهم دور مباشر في نجاح هذه الرسالة آما يشرفني أن أشكر بعض الزملا

معن آوسا ،الفضلاء ياسر القحطاني، بسام الدوسري، سمير الغضبان، إبراهيم القحطاني، . د سعد الشهراني،

جميعا عبدالعزيز السعدي، علي الغامدي، عبيد المضف ، غسان الرقب، والسيد خالد عبدالرحمن العطاس، شكرا لكم

.وجزاآم االله خيرا

vi

Contents

Abstract ii

Acknowledgments v

1 Introduction 1

1.1 Background and Scope of Research ....................................................................1

1.2 Literature Survey of UWB Channel Measurements ...........................................5

1.3 Dissertation Organization ...................................................................................9

2 Ultra Wideband vs. Narrowband Communication Schemes 11

2.1 Background and Historical Evolution................................................................ 11

2.2 Definition and Band Allocation......................................................................... 12

2.3 Communication Signal (Shape and Spectrum) .................................................. 14

2.3.1 Multiple Access Impulse Radio System ................................................ 17

2.4 Coding and Modulation ..................................................................................... 24

2.4.1 Coding.................................................................................................... 24

2.4.2 Modulation............................................................................................. 25

2.5 Interference ........................................................................................................ 27

2.5.1 Interference from Other Radiators to UWB Systems ............................ 27

2.5.2 UWB Interference to Other Systems (Narrowband).............................. 28

2.6 Security .............................................................................................................. 30

2.7 Hardware............................................................................................................ 31

2.8 Applications ....................................................................................................... 34

vii

3 Time Domain and Frequency Domain Channel Measurement Techniques and

Setup 37

3.1 Introduction........................................................................................................ 37

3.2 Channel Impulse Response and Measurable Parameters................................... 38

3.2.1 Pulse-Shape and Frequency Distribution............................................... 38

3.2.2 Material Penetration/Reflection Capability ........................................... 39

3.2.3 Multipath Profile Parameters ................................................................. 39

3.3 Measurement Setups .......................................................................................... 43

3.3.1 Time Domain Measurement Setup ........................................................ 43

3.3.2 Frequency Domain Measurement Setup................................................ 46

3.4 Source Characterization and Conducted Measurements ................................... 48

3.5 Antenna Characterization and Radiated Measurements .................................... 53

3.5.1 Relation between Multipath Angle and Pulse Shape............................. 63

3.6 Calibration Schemes .......................................................................................... 66

3.7 Pros and Cons .................................................................................................... 67

3.7.1 Time and Frequency Resolutions .......................................................... 70

3.7.2 Dynamic Range and Spectral Occupancy.............................................. 72

3.8 Conclusive Remarks .......................................................................................... 73

4 Through-the-Wall Propagation and Material Characterization 75

4.1 Introduction........................................................................................................ 75

4.2 Propagation of Electromagnetic Waves in Dielectric Materials........................ 77

4.3 Measurement Procedures ................................................................................... 79

4.4 Analysis Techniques .......................................................................................... 82

4.4.1 Single-Pass Technique........................................................................... 83

4.4.2 Multiple-Pass Technique ....................................................................... 84

4.5 Comparison of Various Techniques................................................................... 90

4.6 Signal Processing and Parameters Extraction.................................................... 94

4.6.1 Data Acquisition .................................................................................... 94

4.6.2 Time Delay and Initial Guess for Permittivity....................................... 98

viii

4.6.3 Time Gating ........................................................................................... 98

4.6.4 Propagation and Material Parameters ...................................................100

4.7 Description of Samples and Wall Materials .....................................................103

4.8 Measurement Results ........................................................................................104

4.9 Related Issues ...................................................................................................115

4.9.1 Distance from the Sample.....................................................................115

4.9.2 Wall Thickness and Multi-layer Study .................................................115

4.9.3 Repeatability Analysis ..........................................................................115

4.9.4 Variability Analysis ..............................................................................116

4.10 Remarks on Pulse Shaping, UWB Receiver Design, and Modeling Hints.......119

4.10.1 Receiver Design and Pulse Shaping .....................................................119

4.10.2 Modeling and Large-Scale Path-losses.................................................120

4.11 UWB Partition Dependent Propagation Modeling ...........................................120

4.12 Concluding Remarks.........................................................................................124

5 Indoor UWB Channel Measurements 125

5.1 Introduction.......................................................................................................125

5.2 Description of Measurement Procedure and Locations 126

5.2.1 Measurement Procedure and Setup.......................................................126

5.2.2 Description of Measurement Locations ................................................128

5.3 Signal Processing and Data Analysis................................................................129

5.4 Results and Analysis .........................................................................................132

5.4.1 Small-Scale Fading and signal Quality.................................................132

5.4.2 Pathloss and Large-Scale Analysis .......................................................135

5.4.3 Time Dispersion Results .......................................................................140

5.5 Summary and Conclusions ...............................................................................148

6 UWB Channel Model-Deconvolution 149

6.1 Introduction.......................................................................................................149

6.2 Deconvolution...................................................................................................150

ix

6.2.1 Model Deconvolution ...........................................................................151

6.2.2 Multi-Template Model-Deconvolution.................................................154

6.3 Results and Analysis .........................................................................................155

6.4 Summary and Conclusions ...............................................................................158

7 UWB Multi-User Detection 160

7.1 Introduction.......................................................................................................160

7.2 UWB Multi-User Interference ..........................................................................161

7.3 System Model ...................................................................................................163

7.4 Performance Evaluation....................................................................................164

7.5 Simulated System and Parameters ....................................................................165

7.6 Multi-User Detection Schemes.........................................................................166

7.7 Simulation Results ............................................................................................167

7.8 Summary and Conclusive Remarks ..................................................................173

8 Conclusions 175

8.1 Summary of Findings........................................................................................175

8.2 Suggestions for Research Continuation............................................................178

Appendix A1 Single-Pass Time-Duration-Limited Transient Measurement

Derivation 181

Appendix A2 Multi-Pass Complex Dielectric Constant Equation 187

Appendix A3 Proof of Equation (4.42) being Valid with Negative Sign 189

Appendix B1 Antennas Configuration and Structure 193

Appendix B2 Material Pictures 199

Appendix B3 Blueprints and Photos for Measurement Locations 202

References 207

Vita 220

x

List of Tables

2.1 Definitions of NB, WB & UWB signals ................................................................. 14

3.1 Component specifications used for the time domain measurement setup .............. 46

3.2 Total peak power, average power, and emission bandwidth for the used sources . 52

4.1 Errors in dielectric constant and loss tangent obtained form the approximate

formulation............................................................................................................... 93

4.2 Summary of analysis techniques and required equations 100

4.3 Sample building materials, dimensions, and parameters at 5 GHz ........................103

5.1 Measurement locations and scenarios.....................................................................130

5.2 Large-scale pathloss parameters for both TEM horn and biconical antennas ........140

5.3 Parameters for small-scale averaged PDP (SSA-PDP) with TEM horn antennas..143

5.4 Parameters for small-scale averaged PDP (SSA-PDP) using biconical antennas ..144

7.1 Simulated Models (Coherence detection)...............................................................165

B3.1 Measurement locations and scenarios (reproduced) ...............................................206

xi

List of Figures

2.1 Gaussian, Gaussian monocycle, and doublet waveforms and their corresponding

normalized frequency spectrum............................................................................... 16

2.2 Illustration of four users time hoping multiple access format (impulse radio)........ 19

2.3 The basic Gaussian monocycle and its frame-shifted version................................. 20

2.4 Uniform train of Gaussian monocycles in time and frequency domains................. 21

2.5 Non-uniform train of Gaussian monocycles in time and frequency domains ......... 23

2.6 Typical received signal for bit=0, bit=1 and the typical waveform used by the

receiver correlator.................................................................................................... 26

2.7 Higher-level block diagram for a UWB transmitter and receiver ........................... 32

3.1 Example of discrete-time impulse response model for a multipath indoor channel 40

3.2 Time domain measurement setup ............................................................................ 45

3.3 Frequency domain measurement setup.................................................................... 47

3.4 The setup used for the conducted source measurements ......................................... 49

3.5 The resulting waveforms for the Picosecond 4050A generator............................... 50

3.6 The resulting waveforms for the Picosecond 4100 generator.................................. 51

3.7 Effect of connection coaxial cables ......................................................................... 53

3.8 The distances required for far-field approximation vs. frequency........................... 55

3.9 Received and time-gated waveforms for antenna #1 at a distance of 2.5 m............ 57

xii

3.10 Frequency domain transfer function for antenna #1 using the time domain

measurements........................................................................................................... 58

3.11 Received and time-gated waveforms demonstrated for the antenna #2 with the

4050A generator at a distance of 7 m ...................................................................... 60

3.12 Frequency domain transfer functions of the two antennas ...................................... 61

3.13 Received signal for antenna #2 with the 4100 generator......................................... 62

3.14 Received waveforms at different receiver elevation angles (E-scan)...................... 64

3.15 Received waveforms at different azimuth receiver angles (H-scan) ....................... 65

4.1 Incident, transmitted, and reflected waveforms observed in time-domain

measurements........................................................................................................... 80

4.2 Two required measurements, without layer (free space) and with layer in place ... 81

4.3 Comparison between the different measurement and analysis techniques.............. 92

4.4 Two-dimensional search example, illustrating the possibility of reducing it to a one-

dimensional search................................................................................................... 92

4.5 Through the wall and material characterization procedure flowchart ..................... 95

4.6 Illustration of the frequency domain measurements................................................ 96

4.7 Illustration of the Time domain measurements ....................................................... 97

4.8 Three different time domain gating windows.......................................................... 99

4.9 Time domain representation of the six different measurements for the sample door

.................................................................................................................................101

4.10 Comparison for the sample door parameters extracted using different measurement

techniques ...............................................................................................................102

xiii

4.11 Insertion transfer function plotted versus frequency for different materials .........106

4.12 Dielectric constant plotted versus frequency for different materials......................107

4.13 Attenuation constant plotted versus frequency for different materials...................108

4.14 Blocks wall and wallboard ‘free-space’ and ‘through’ measurements...................109

4.15 Cloth office partition and structure wood ‘free-space’ and ‘through’ measurements

.................................................................................................................................110

4.16 Door and wood ‘free-space’ and ‘through’ measurements.....................................111

4.17 Glass and styrofoam ‘free-space’ and ‘through’ measurements.............................112

4.18 Bricks wall and reinforced concrete pillars ‘free-space’ and ‘through’ measurements

.................................................................................................................................113

4.19 TDL reinforced concrete pillar ‘free-space’ and ‘through’ measurements ............114

4.20 Repeatability and variability of frequency domain measurements.........................117

4.21 Repeatability and variability of time domain measurements..................................118

4.22 Gaussian (TEM horn input signal) and Gaussian monocycle (TEM horn radiated

signal) waveforms and their corresponding normalized spectra ............................122

4.23 Illustrative example for UWB partition dependent modeling ................................123

5.1 Measurement grid ...................................................................................................128

5.2 Effect of filtering the measured profile (using the biconical antenna) ...................131

5.3 Comparison between small-scale averaged power delay profiles (SSA-PDP) and

local power delay profiles.......................................................................................134

5.4 The cumulative distribution of the signal quality based on 9 spatial sample points

.................................................................................................................................136

xiv

xv

5.5 Scatter plot for the relative pathloss versus frequency for all locations................. 138

5.6 Scatter plots for the pathloss versus distance, for LOS and NLOS scenarios........ 139

5.7 Cumulative distribution functions (CDF)s for the RMS delay spread (20 dB) ..... 142

5.8 Scatter plot for the mean excess delay versus the RMS delay spread.................... 145

5.9 Scatter plots to characterize the correlation between the RMS delay spread with

each of the distance and the pathloss ..................................................................... 147

6.1 Typical received LOS multipath profile (Whittemore 2nd floor Hallway)............. 152

6.2 Illustration of the measurements of an ideal channel and a multipath indoor channel

................................................................................................................................ 153

6.3 Setup and received waveform with both transmitter and receiver antennas pointing

to the reflection surface .......................................................................................... 153

6.4 Different received waveforms at locations when scanning on the E-plane and the H-

plane ..................................................................................................................... 156

6.5 Improvement in captured energy............................................................................ 157

7.1 Illustrative effect of multi-user interference as a result of high frequency ringing 162

7.2 Multi-stage multi-user detection schemes for two synchronous users................... 168

7.3 Performance of multi-user detection for equal power synchronous case............... 169

7.4 Unequal performance for equal power user with successive cancellation............. 171

7.5 Multi-user detection performance for equal-power asynchronous case................. 172

7.6 Performance of multi-user detection for asynchronous unequal power case......... 174

A2-1 Bounce diagram for propagation through a slab.................................................... 188

B1.1 Schematics and dimensions of the TEM horn antenna (antenna#1).......................195

B1.2 TEM horn antenna#1 in anechoic chamber and measure pattern at 5 GHz ...........196

B1.3 Schematics and dimensions for the TEM horn antenna array (antenna#2) ............197

B1.4 Picture and pattern at 2.5 GHz for the omnidirectional biconical antenna.............198

B2.1 Pictures for the bricks, blocks, styrofoam and walls built out of them...................200

B2.2 Pictures for the wallboard, door, wood, structure wood, cloth office partitions, glass,

and reinforced concrete pillars................................................................................201

B3.1 Whittemore blueprints to illustrate measurement locations and environments......203

B3.2 Whittemore site photos ...........................................................................................204

B3.3 Measurement locations for Durham Hall................................................................205

xvi

Does it (UWB) work? Bob Lucky, Chairman of the FCC's technical advisory committee

Of course it works! Bob Scholtz, Director of UltraLab, University of Southern California

[Sch99]

Chapter 1

Introduction

1.1 Background and Scope of Research

Ultra wideband (UWB) communication has been the subject of extensive research

in recent years due to its unique capabilities and potential applications particularly in

short-range multiple access wireless communications. However, many important aspects

of UWB-based communication systems have not yet been thoroughly investigated. There

have been strong views expressed in the literature regarding the suitability of UWB

technology to wireless communications. Research generated in industrial organizations is

biased towards the advantages of UWB technology. On the other hand, the research work

down playing the viability of this technology is not based on rigorous and exhaustive

analyses. The objective of this study is not to resolve this debate, but instead to obtain a

more thorough and comprehensive understanding of the potentials of UWB technology

by characterizing the UWB communication channels. In particular, channel modeling,

interference effects, and the role of antennas require careful examinations before an

actual implementation of UWB systems can be undertaken. The propagation of UWB

signals in indoor and indoor-outdoor environments is the single most important issue with

significant impacts on the future direction, scope, and generally the extent of the success

1

Chapter 1. Introduction

of UWB technology. An accurate characterization of the propagation channel is essential

to many communication system design issues.

Channel characterization refers to extracting the channel parameters from the

measured data. Extracted parameters are used to quantify the effect of the channel on

communication systems operating in this channel. Data can be measured in different

ways using a variety of experimental setups. Channel characterization can be achieved by

performing measurements in the time domain or in the frequency domain. Channel

characterization for narrowband communication systems is well established. The

accuracy and proper interpretation of measured data have a direct impact on the accuracy

of the channel model to be developed based on such data. Thus, it is important to first

examine the performance of time-domain and frequency-domain measurements systems

intended for channel characterization studies. This includes characterizing pulse

generators and antennas. The purpose of considering different alternatives for data

measurements is to come up with reliable and efficient measurement techniques. In this

regards, frequency-domain versus time-domain techniques for characterizing UWB

communication channels are compared. The following issues are addressed: measurables,

measurement approaches and setups, and calibration schemes. The pros and cons of each

technique are discussed. Characterizing the pulse generators, the antennas and the

receiver is an integral part of UWB channel characterization efforts. The effect of UWB

antennas cannot be easily factored out from the characterization results.

At a fundamental level, the propagation of UWB signals, as any electromagnetic

wave, is governed, among other things, by the properties of materials in the propagation

medium. Thus, the information on electromagnetic properties of building materials in the

UWB frequency range would provide valuable insights in appreciating the capabilities

and limitations of UWB technology for indoor and indoor-outdoor applications. Although

electromagnetic properties of certain building materials over relatively narrow frequency

ranges are available, ultra-wideband characterization of most typical building materials

for UWB communication purposes has not been reported.

2


One of the objectives of this research is to examine propagation through walls

made of typical building materials and thereby acquire ultra wideband characterization of

these materials. The loss and the dielectric constant of each material are measured over a

frequency range of 1 to 15 GHz. Ten commonly used building materials are chosen for

this investigation. These include, dry wall, wallboard, structure wood, glass sheet, bricks,

concrete blocks, reinforced concrete (as pillar), cloth office partition, wooden door, and

styrofoam slab. The characterization method is based on measuring an insertion transfer

function, defined as the ratio of two signals measured in the presence and in the absence

of the material under test. The insertion transfer function is related to the dielectric

constant of the material through a complex transcendental equation that can be solved

using numerical two-dimensional root searching techniques. The insertion transfer

function can be obtained either through frequency-domain measurements using a vector

network analyzer, or by performing time-domain measurements using a pulse generator

and a sampling oscilloscope and then Fourier transforming the measured signals into the

frequency domain. In this research, both frequency-domain and time-domain

measurement techniques are used to validate the results and ensure the accuracy of

measurements by capitalizing on the advantages of each technique. The material

characterization data can be used in studying channel modeling problems.

The work on ultra wideband characterization of building materials resulted in an

additional interesting contribution. As mentioned above, the dielectric constant is

determined by solving a complex transcendental equation, a process which is often time

consuming due to slow convergence and the existence of spurious solutions. A new

formulation for evaluating the complex dielectric constant of low-loss materials, which

involves solving real equation and thus requiring only one-dimensional root searching

techniques, was found. The results derived from the exact complex equation and the new

formulation are in excellent agreement. This formulation reduces the computation time

significantly and is highly accurate for the characterization of low-loss materials.

After studying ultra-wideband propagation properties of typical building

materials, an indoor UWB measurement campaign is undertaken. The measurements are

performed using Gaussian-like pulses with a duration less than 100 ps. Two sets of

3


measurements are performed, one set with directional TEM horn antennas and another

with omni-directional biconical antennas. A total of about 400 signal profiles are

collected. The TEM horn and biconical antennas may be considered as representative

examples for stationary line-of-sight and mobile applications, respectively. The

measurements are carried out in two buildings on Virginia Tech Campus; namely,

Whittemore Hall and Durham Hall. Typical indoor scenarios, including line-of-sight

(LOS), non-line-of-sight (NLOS) in room-to-room, within-the-room, and in hallways are

considered.

Results for indoor propagation measurements are presented for local power delay

profiles (local-PDP) and small-scale averaged power delay profiles (SSA-PDP). Site-

specific trends and general observations are discussed. Some statistical analyses of the

measured data are presented and compared with the previously published UWB and

narrowband results. The results for pathloss exponent and time dispersion parameters are

presented. The analyses results indicate the immunity of UWB signals to multipath

fading which occurs in narrowband signals. Furthermore, the measurement results clearly

show that UWB signals, unlike narrowband signals, do not suffer from small scale

fading, unless the receiver is too close to walls.

A further step is taken by employing a deconvolution technique to extract more

information about the channel, particularly the number of multipath components.

Characterization of UWB channels can be performed by sounding the channel with

pulses, and thereby obtain the impulse response. Multipath components have different

waveforms depending on the type of transmitter and receiver antennas used and the

angles of transmission and reception. A modified deconvolution technique is introduced

to extract the UWB channel response. The application of deconvolution techniques

results in resolving multipath components with waveforms different from those of the

sounding pulse. Resolving more components can improve the design of the rake receiver.

Accurate characterization of the impulse response of a UWB communication system

facilitates performance evaluation studies.

4


The final part of this research work is devoted to illustrating an example of

utilizing measurement results to improve the receiver design. So far, receivers designed

for multiple access ultra-wideband communication systems, known as impulse radio, are

based on conventional single-user matched-filter detectors. Here, we elaborate on the

nature of multiple access interference and illustrate the application of multi-user detection

to improve the performance of impulse radio systems. Measured dispersion parameters

and their effects on the multiple access parameters are discussed.

1.2 Literature Survey of UWB Channel Measurements

Many researchers have studied the propagation of narrowband electromagnetic

waves through walls and floors at 900 MHz, 1.8 GHz, 2.4 GHz, 5.85 GHz, 60 GHz, and

other dedicated narrowband frequency ranges [Zha94], [Dur98], [And02a]. In many

narrowband measurements, only the magnitude of insertion transfer function (i.e., loss)

has been the quantity of interest. However, in UWB communication systems, as in

ground penetrating radars, in addition to the signal magnitude, the phase information

(delay) is an equally important factor that needs to be taken into account in studying the

propagation effects [Dan96]. Therefore, narrowband measurements, although helpful in

providing some general understanding, are not adequate for UWB propagation analyses

and channel modeling.

Some results on ultra wideband characterization of building materials have been

reported during the past decade. Hashemi has presented an excellent review and

comparison of published results for different indoor penetration losses in the UWB

frequency range [Has93a]. However, these results are often inconsistent, making the

assessment of indoor UWB propagation effects unreliable. For example, in [Has93a]

significantly different measured values of 7 dB, 8.5-10 dB, 13 dB and 27 dB for the

insertion loss of concrete blocks are cited. Moreover, in some cases, the relationship

between the expected loss and the operating frequency has not been satisfactorily

addressed. Though it is a common observation that the loss increases with frequency,

some published data indicate a decrease in loss when frequency increases. In a ground

floor experiment, de Toledo and Turkmani [Tol92] report measured average penetration

5


losses of 14.2 dB, 13.4 dB, and 12.8 dB at 900 MHz, 1800 MHz, and 2300 MHz,

respectively. Another example is the data reported by Zhang and Hwang [Zha01] who

performed measurements in the frequency range of 900 MHz to 18 GHz. According to

their investigation, the indoor penetration loss increases with frequency for reinforced

concrete wall but this trend is not generally realized for plasterboard.

At the statistical level, there has been some research on narrowband indoor

channel characterization. Of special value is a series of publications by Saleh and

Valenzuela [Sal87], Hashemi [Has93a], [Has93b], Anderson et al. [And02a], Durgin and

Rappaport [Dur00], and Rappaport [Rap92], [Rap89], [Rap96]. The primary objective of

these researchers has been to develop models that describe the system performance

adequately. A successful characterization requires extensive and accurate measurements.

The accuracy of the model depends mainly on the accuracy of the measurements.

Due to the rapidly growing interest in UWB communications, researchers are

nowadays devoting considerable efforts and resources to develop robust channel models

that allow for reliable and accurate ultra-wideband performance simulation. At present,

the amount of available measurement data is very limited and more are needed to support

a comprehensive channel modeling study. The issue becomes more complicated due to

the fact that UWB pulse measurements are antenna dependent. The spectrum and the

shape of the pulse also affect the measurements.

The analysis of indoor communication systems based on simulation of the entire

transmission link using statistical methods is most useful in assessing the system

performance [Has93a]. This approach, however, requires extensive propagation

measurements. Some research work on both deterministic [Ugu02] and statistical

modeling [Zhu02], [Cas01] has been reported. More recently, Cassioli et al. [Cas02] have

presented simulation results for UWB indoor communications, while Chalillou et al.,

[Cha02] have discussed the main structure of a general simulator for UWB

communication systems. However, there still remain many unresolved issues and hence

the need for more UWB propagation measurements. Different measurement conditions,

insufficient measurement data, and the effect of different excitation pulses are among the

6


priority issues that demand additional measurements in order to formulate comprehensive

and robust models before designing simulators.

The most notable UWB time-domain measurement campaign is that by the Ultra

Lab group conducted at the University of South California (USC), in collaboration with

the Time Domain Corporation [Sch97]. Their measurements were performed using a

sampling oscilloscope, a pulse generator and wideband antennas. The results of these

measurements were used to develop further models [Cra02], [Cas01], but no information

on the pulse shape and the characteristics of the antennas used in their measurements are

provided. Only in a separate study they mention that a diamond dipole antenna has been

used [Sch01]. A disadvantage of this antenna is that it covers a small frequency band.

Also, they used a wireless device in their triggering system. Multiple reflections from the

surroundings may cause mis-triggering in such a wireless triggering system. One further

step was taken with the aim of characterizing a more realistic UWB communication

rather than simple periodic pulses [wit99], [Dic99]. In this scenario, the pulses were

modulated and time dithered to emulate real communication environments. However, at

this stage of the UWB technology evolution, more fundamental investigations are

required in order to achieve a better understanding of the channel characteristics.

Acquiring the transfer function or the impulse response of the channel will help

communication engineers to study the effects of time dithering or any other techniques

through simulation.

Another approach for UWB channel characterization is to perform propagation

measurements in the frequency domain and convert the results to the time domain by

means of inverse Fourier transform. The advantage of this approach is that the sensitivity

of the equipment used, particularly the vector network analyzer, is much higher than that

of the time-domain measurement equipment such as sampling oscilloscope. The chief

disadvantage of frequency-domain measurements is that long high-quality RF cables are

required for connecting the network analyzer to both transmitting and receiving antennas

[Gha02]. Furthermore, double shielding of these cables is often required in order to avoid

coupling of radiated signals from the air through the cable to the receiver. These cables

represent a major limitation for long distance measurements. On the other hand, in direct

7


time-domain measurements, it is only required to use a cable for carrying the triggering

signal from the source at the transmitting side to the sampling oscilloscope at the

receiving end. The bandwidth of the triggering signal is usually much less than the

bandwidth of the pulse. Thus, long cables with moderate attenuation and dispersion levels

are adequate for triggering purposes, making time-domain measurements of UWB signals

at larger distances between the source and the observation point much easier.

A number of researchers have studied UWB channel propagation using

frequency-domain measurements, including Ghassemmzadeh et al. [Gha02], Prettie et

al. [Pre02], Keignart and Daniele [Kei02], Kunisch and Pamp [Kun02], Street et al

[Str01], and Hovinen et al. [Hov02]. Only Ghassemmzadeh et al. [Gha02] used

substantially long cables; up to 45 m, while most others who have described their

measurement setups have used cables of up to nearly 10 m. They have also used different

bandwidths in their measurements. Ghassemmzadeh et al. [Gha02] have performed their

measurements in the Unlicensed National Information Infrastructure (UNII) band. The

bandwidth in their measurements was either 1 GHz or a maximum of 2.5 GHz centered at

a frequency of 5 GHz. Prettie et al. [Pre02] and Hovinen et al. [Hov02] have used a

frequency range of 2 GHz to 8 GHz. Keignart and Daniele [Kei02] have used a smaller

frequency range from 2 GHz to 6 GHz, while Kunisch and Pamp [Kun02] conducted

UWB channel measurements in the range of 1 GHz to 11 GHz. The time-domain

resolution and the time delay that can be obtained from frequency-domain measurements

depend on the minimum and the maximum frequencies in a given bandwidth and the

number of frequency points at which measurements are taken.

Ghassemmzadeh et al. [Gha02] have presented extensive frequency-domain

measurements in 23 residential homes. No multipath component was observed in their

measurements beyond 70 ns of excess delay with a 30 dB threshold. These measurements

were used by Turin et al. [Tur02] to develop an autoregressive model for an indoor UWB

Channel. The generated model allows for simple simulations. However, the parameters of

the model are location dependent. Kunisch and Pamp [Kun02] observed that the channel

gain tends to decrease with frequency. But, details of their measurement system are not

revealed, yet the authors report that all results account for frequency dependent antenna

8


characteristics. Cramer [Cra99] demonstrated that multipath components result in an

extended channel impulse response. The study suggested a window of 300 ns to account

for all multipath components which contain an appreciable amount of power. Diversity is

also considered in this work to extract the structure of the channel from the received array

of data. Beamformer response at different projection angles has also been studied. Prettie

et al. [Pre02] have presented spatial correlation of their UWB measurements. Another

direction for evaluating UWB channels is based on narrowband models. However, such

models are formulated for specific frequency bands and are suitable for interference

studies, but cannot be directly used for UWB channel characterization [Kis01], [Dep].

To illustrate the deficiency of models based on these measurements, Cramer et al.

[Cra99] suggested modeling the UWB channel as a summation of the Hermite

polynomials. Their justification was based on the heuristic approach that the signal

driving the antenna is often modeled as Gaussian and some of the propagation and

reflection effects tend to have the characteristics of the signal derivatives. Later studies

revealed that derivative behavior is a result of the specific antenna transfer function

[Muq02]. Recently, Lee [Lee00] presented a deterministic multipath analysis using a

two-ray model and characterized the time of arrival using Saleh-Valenzuela model

[Sal87]. Another statistical model was also reported by Foerster [Foe01]. The presented

results by both Lee and Foerster are based on simplified assumptions and call for further

experimental support.

1.3 Dissertation Organization

In Chapter 2, a theoretical background for understanding UWB communications

is presented. In particular, important attributes of narrowband and UWB communication

systems are compared, including historical evolutions, bandwidth requirements and

definitions, shapes and spectra of information signals, coding schemes and modulation

techniques, interference, security issues, hardware aspects, and applications. Emphasis is

placed on UWB systems, assuming that the reader is familiar with narrowband systems.

Details on narrowband system can be found in [Hay83],[Pro89], and [Skl88]. Time-

Domain and Frequency-domain techniques and measurement setups are discussed and

9


compared in Chapter 3. Understanding the details of measurement system is very

important for properly interpreting the experimental results. Sources and antennas are

characterized and their effects are examined. Chapter 4 devoted to studying the

propagation of UWB signals through walls and common materials encountered in indoor

communication environments. Ten typical building materials are characterized. Results

for an indoor UWB measurement campaign are presented in Chapter 5. Measurement

results for two typical office buildings are analyzed. Pathloss and time dispersion

parameters are studied using directive and omnidirectional antennas. More advanced

analysis based on extracting the channel impulse response using deconvolution

techniques is discussed in Chapter 6. Multi-template subtractive deconvolution is used to

estimate the number of significant multipath components and the percentage of power

associated with them. Motivated by the experimental view a proposal for utilizing multi-

user detection techniques for UWB communication systems is presented in Chapter 7.

Finally, summary and conclusions are provided in Chapter 8. Appendix A is dedicated to

theoretical analyses and derivations, while Appendix B is devoted to additional details

and miscellaneous items.

10

“An intriguing alternative which may eventually become practical, and even legal, for short-range communication between static terminals is ultra-wideband impulse radio.”

“In fact, the principle of impulse radio is firmly grounded in information theory: maximum power efficiency is achieved by pulse-position modulation in an infinite bandwidth channel”

“although the whole band occupied by the transmission, say, from DC to a few gigahertz is “owned” by other systems, much of it is unused at any given time. Thus, reasonable receiver sensitivity can indeed be achieved with very low transmitted power”

Sergio Verdú [Ver00]

Chapter 2

Ultra Wideband vs. Narrowband

Communication Schemes

2.1 Background and Historical Evolution

Ultra wideband (UWB) systems use precisely timed, extremely short coded pulses

transmitted over a wide range of frequencies. Although UWB technology had some old

roots, ultra wideband communication is a relatively new technology. The technology is

radical departure from current wireless communication methods.

Ultra wideband technology originated from work in time-domain

electromagnetics begun in 1962 [Fon]. The concept started with the objective of

characterizing linear time invariant systems by measuring the output as a result of an

impulse excitation, instead of using the more conventional means of swept frequency

response. However, that was not possible until the developments in the techniques for

subnanoseconds (baseband) pulse generation, which are needed to approximate the

impulse excitation and to make the measurements feasible.

11

Chapter 2. Ultra Wideband vs. Narrowband Communication Schemes

It became obvious that short pulse radar and communication systems can be

developed in the same way. In 1978 efforts turned toward the communication using

UWB signals. An experiment was successfully presented where intelligible voice signals

could be communicated over hundreds of feet without the need for synchronization. Six

years later, greater ranges were possible. Harmuth has published two books about the

transmission of information using orthogoanal functions and the applicability of using

nonsinusoidal waves for radar and radio communication [Har].

The term “ultra wideband” was not used until around 1989. By that time, the

theory of UWB has experienced thirty years of developments. Further historical

information about UWB technology can be found in [Bar01].

In this chapter UWB communication is compared with narrowband

communications. The comparison includes: definition and frequency band allocation,

communication signal (shape and spectrum), coding and modulation, interference,

security, hardware, and applications.

2.2 Definition and Band Allocation

In principle UWB technology is the use of short pulses instead of continuous

waves to transmit information. The pulse directly generates a very wide instantaneous

bandwidth signal according to the time-scaling properties of the Fourier transform

relationship between time, t, and frequency, f.

Before presenting a formal definition for ultra wideband signals and systems, it

should be noted that different terms are used in the literature which essentially refer to the

same thing such as impulse radio, orthogonal functions, Walsh waves, nonsinusoidal,

sequency theory, carrier-free, video-pulse transmission, large relative bandwidth, time-

domain techniques, baseband, large-relative bandwidth and ultra wideband [Har], [Bar].

Researches from Russia and China have been actively developing and testing

UWB impulse generators [Kis92]. The Soviets developed the “superwideband” signal

definition. All RF signals with a low frequency bound, fl, and high frequency bound, fh,

12


have a corresponding index of breadth of band, bµ , or relative bandwidth [Har81],

[Ast97],

)(5.0 lh

lhb ff

ff+

−=µ (2.1)

Researchers from Russia and China term an impulse-like signals

“superwideband” or UWB when 0.1≥bµ . Although, spread spectrum communications

can be designed with 0.1≥bµ , they are not called UWB because they do not possess the

transient behavior.

There is no general definition of UWB in the IEEE dictionary. However US

Defense Advanced Research Project Agency (DARPA) defined UWB for EM waves

with instantaneous bandwidth greater than 25% of center frequency [Pan99].

It is important to note that some technical terms can have different meaning based

on the subject where they are used. Narrowband (NB), wideband (WB) or broadband

(BB) and ultra wideband (UWB) can have different definitions based on the application,

i.e. communication, radar, electromagnetic interference (EMI) / electromagnetic

cancellation (EMC), etc. In mobile communication it is common to refer to the system’s

bandwidth as being narrow or wide relative to the coherence bandwidth. However, the

terminology used here is the one used by RF engineers based on the ratio of the

bandwidth relative to the carrier frequency [Cas02].

Table 2.1 presents both general and percentage bandwidth definitions. Percentage

bandwidth (%BW) – which is directly related to the breadth factor – is the bandwidth of

interest divided by the center frequency,

13


14

%100

2

% ×

+

−=

lh

lh

ffff

BW . (2.2)

2.3 Communication Signal (Shape and Spectrum)

Table 2.1. Definitions of NB, WB & UWB signals

Definitions NB WB UWB

% BW

EM waves with

instantaneous bandwidth

less than 1% of center

frequency (%BW <1%)

EM waves with instantaneous

bandwidth greater than 1%

and less than 25% of center

frequency (1% < %BW<25%)

EM waves with

instantaneous bandwidth

greater than 25% of center

frequency (%BW>25%)

General

Any radio or

communication signals

which is not wideband. 5

kHz for telephones and

AM radio, 25 kHz for FM

radio (military)

Any radio or other

communication signal which

is wider than narrow band.

High bit rate telephone data

circuits, (25 kHz for 9600

BPS) and TV channels (6-10

MHz) are generally considered

wideband signals

None in the IEEE dictionary

or other sources. However

US Defense Advanced

Research Projects (DARPA)

panel on UWB technology

published same as

percentage bandwidth.

Note: table is reproduced from [Pan99]

Narrowband communication is usually achieved by modulating a sinusoidal

carrier with the information to be transmitted. The resultant signal possesses the

sinusoidal nature and occupies a narrow band in the frequency domain. On the other

hand, for UWB applications, any waveform that satisfies the definition of UWB signal

can be used. The choice of a specific waveform is driven by system design and

application requirements. There has been many attempts to choose a signal waveform

suitable for UWB applications and yet has minimal interference with proximity systems

[Ham01a],[Ham01b].

The basic theoretical model for impulse radio uses a class of waveforms known as

“Gaussian waveforms”. They are called Gaussian waveforms because they are very


similar to the Gaussian function. In the time domain a Gaussian waveform, vg(t), is given

by

2

)()( τ

t

g etv−

= . (2.3)

where t is the time and τ is a parameter which represents the temporal width of the pulse.

Another waveform can be created by filtering or differentiating the Gaussian pulse to get

the Gaussian Monocycle∗. For the Gaussian monocycle, τ is the time between minimum

and maximum amplitudes and it defines the time decay constant that determines the

monocycle’s duration. The Gaussian monocycle has a single zero crossing and it is given

by

2

)()( τ

τ

t

m ettv−

= . (2.4)

By increasing the order of differentiation, the number of zero-crossings increases,

the bandwidth decreases and the center frequency increases. Utilizing the following

Fourier transform identity

22

22)(

)()( ττ ττ

ft

ejffVettv −−−=⇔= , (2.5)

and assuming A to be the peak amplitude of the monocycle, and fc to be the center

frequency, then, the Gaussian monocycle in the time domain is given by

2)(22),,( ctf

ccm eAtfeAftv ππ −= , (2.6)

and in the frequency domain it is given by

2

21

2222),,(

−

−= cff

ccm e

fejfAAftV π

π. (2.7)

∗ Some authors refer to the first derivative as doublet and some refer to it as Gaussian monocycle and they refer to the second derivative as doublet. To avoid confusion, we will refer to the first derivative as Gaussian monocycle and to the second derivative as doublet consistently.

15


16

a) Gaussian, Gaussian monocycle, and doublet waveforms

b) Normalized spectrum for Gaussian, Gaussian monocycle, and doublet

waveforms

Figure 2.1. Gaussian, Gaussian monocycle, and doublet waveforms and their

corresponding normalized frequency spectrum

0 2 4 6 8 10 12

10 -8

10 -6

10 -4

10 -2

10 0

Frequency (GHz)

Nor

mal

ized

Mag

nitu

de

GaussianGaussian Monocycle Doublet

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 -1

-0.8 -0.6 -0.4 -0.2

0

0.2 0.4 0.6 0.8

1

Time (ns)

Nor

mal

ized

Am

plitu

de

GaussianGaussian Monocycle Doublet


Gaussian excitation pulse provides excellent radiation properties [Bor99]. Other

possible waveforms include pulse like triangular, trapezium and other shaped pulses.

Figure 2.1 illustrates the basic Gaussian, and its first and second derivatives, which are

the Gaussian monocycle and the doublet waveforms. The corresponding normalized

frequency spectrum is shown in the same figure. Practical implementation of such

Gaussian monocycle remains an important issue.

To get more insight on the signal shape and spectrum, one has to address the

multiple access system. The following subsection is dedicated to the discussion of the

signal waveform and spectrum as applied to multiple access UWB system (impulse

Radio).

2.3.1 Multiple Access Impulse Radio System

Multiple access techniques for narrowband systems include: time, frequency, and

code division techniques. The multiple access UWB system model proposed by Scholtz

[Sch93] is based on time hopping codes. The typical hopping format will be given first,

followed by detailed explanation of the terms and the way they affect the signal

waveform and the spectrum.

In a typical hopping format for impulse radio with pulse position modulation, the

time access is divided into frames, Tf, and every frame is subdivided into time slots, Tc.

Every transmitter send a pulse per frame at different time slots from frame to another.

The signal transmitted by the kth user is given by [Sch97]

∗, (2.8) ( ) ( ) ( )∑∞

−∞=

−−−=j

kNjc

kjf

ktr

kktr s

dTcjTtwts )(/

)()()( δ

where wtr(t) is the transmitted pulse. Superscript k indicates transmitter related quantity.

Tf is the period of the frame or average pulse repetition time. Each user is assigned a time

hopping sequence shift pattern c . This hopping sequence provides an additional shift )(kj

∗ Note that j is not used to indicate imaginary terms. The variable j is used as a counting variable for the summation. It is consistently used in the literature.

17


of . The transmission rate, Rs, determines Ns which is the number of monocycle to

be modulated by a given binary symbol. Pulse position modulation is used with δ added

delay if the modulated bit is one. The term dn refers to the nth binary symbol.

ck

j Tc )(

The assumed channel model is that Nu users are active during transmission. The

received waveform is generally different from the transmitted one. Signal undergoes

constant amplitude attenuations and waveform deformation because of the antennas and

the propagation channel. Because of the antenna, the received signal is related to the

derivative of the transmitted signal. For example, for TEM horn antennas in boresight

configuration, if a Gaussian pulse is transmitted then a Gaussian monocycle is received.

On the other hand, if a monocycle is transmitted, the expected received waveform will

have a doublet shape. This is because the response of the radiating system will act as a

differentiator.

When the number of users is Nu, the received signal is [Sch97]:

(2.9) ( ) ( )∑=

+−=uN

kk

kreck tntsAtr

1

)( )(τ

where Ak is the attenuation of the propagation path of the signal, , received

from the kth transmitter. The time delay between the kth transmitter and the receiver is

represented by

( )k τ−)(krec ts

kτ and the Gaussian noise at the receiver input is represented by n(t).

The signal emitted by the kth transmitter consists of a large number of pulses

shifted to different times. Figure 2.2 illustrates an example for four users impulse radio.

Four time frames are presented with each frame divided into four time slots. Each user is

coded with different color. It is important to note that (2.9) and this discussion assumes

no multipath components.

The jth monocycle nominally starts at time

( ) ( )kjc

kjf dTcjT ++ . (2.10)

18


kk ==11

kk ==22

kk ==33

kk ==44

Tf Nu =4 Tc

J=0 J=1 J=2 J=3

Figure 2.2. Illustration of four users time hoping multiple access format (impulse radio)

To explain the previous terms in (2.10) and their effects on the spectrum, each time shift

will be examined separately assuming Gaussian monocycle as the pulse. The quantity

represents the monocycle waveform that nominally begins at time zero on the kth

transmitter’s clock. In regard to the first term in (2.10), the quantity represents

the monocycle in the j frame.

)( )(ktw

)( fjTtw −

Figure 2.3 illustrates the transmitted monocycle

and the shifted version of it,

)( )(ktw

)( fjTtw − .

A single bit of information is generally spread over multiple monocycles to form

a train of pulses. The quantity represents a uniform pulse train. The frame

time, Tf, may be 100 to 1000 times the monocycle width, resulting in a signal with very

low duty cycle. Both time-domain and frequency-domain plots for a uniform monocycle

pulse train are presented in

∑∞

−∞=

−j

fjTtw )(

Figure 2.4.

The second term in (2.10) is now considered. Multiple access signals composed of

uniformly spaced pulses are vulnerable to occasional catastrophic collisions. To eliminate

19


20

0 0.5 1 1.5 2-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

Time (ns)

Nor

mal

ized

Am

plitu

de

a) Gaussian monocycle )(tw

0 1 2 3 4 5 6 7 8 9 10 -1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

Time (ns)

Nor

mal

ized

Am

plitu

de Frame Time

b) Shifted Gaussian monocycle )( fjTtw −

Figure 2.3. The basic Gaussian monocycle and its frame-shifted version


21

0 2 4 6 8 10 12 14 16 18 20 -1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

Time (ns)

Nor

mal

ized

Am

plitu

de

a) Uniform train of Gaussian monocycles ∑∞

−∞=

−j

fjTtw )(

0 2 4 6 8 10 12

10 -8

10 -6

10 -4

10 -2

10 0

Frequency (GHz)

Nor

mal

ized

Mag

nitu

de

Gaussian Monocycle Monocycle Pulse Train

b) Normalized spectrum for the uniform train of Gaussian monocycles

Figure 2.4. Uniform train of Gaussian monocycles in time and frequency domains


catastrophic collisions, each link (k) uses a distinct pulse shift pattern . The hopping

sequence is periodic with period Np, which means that

( )kjc

jiNj ccp=+ , i is an integer. (2.11)

Each element of the time-hopping sequence is an integer satisfying the following bound

(2.12) hk

j Nc <≤ )(0

Hence the additional time shifts caused by the time-hopping sequence are discrete values

between 0 and . It is assumed that chTN

fch TTN ≤ (2.13)

Since the pseudorandom time-hopping sequence has period Np , then the waveform is

periodic with period

Tp=NpTf (2.14)

It can be shown that the time-hopping sequence effectively reduces the power

spectral density (PSD) of the uniformly spaced pulse train from a line spectral density

(1/Tf apart) down to a spectral density with finer lines 1/Tp. The time domain and the

normalized frequency spectrum representations for the non-uniform pulse trains are

presented in Figure 2.5.

The last term in the time index is for the pulse position modulation (PPM). PPM

is expected to distribute the RF energy across the band by smoothing the spectrum of the

signal [Sch93]. This should make the signal less detectable. The spectral smoothing effect

of PPM is relatively small. This is because the PPM only moves the pulse a very small

fraction of the pulse width. For example, a bit representing an information bit “1” will be

delayed by 0.156 ns compared with a bit representing “0” for a total pulse width of 1.5

ns. More details about PPM are presented in Section 2.4.2.

22


23

0 2 4 6 8 10 12 14 16 18 20 -1

-0.8

-0.6

-0.4

-0.2

0 0.2 0.4 0.6 0.8

1

Time (ns)

Nor

mal

ized

Am

plitu

de

a) Non-uniform train of Gaussian monocycles ∑∞

−∞=

−−j

ck

jf TcjTtw )( )(

0 2 4 6 8 10 12

10 -8

10 -6

10 -4

10 -2

10 0

Frequency (GHz)

Nor

mal

ized

Mag

nitu

dfe

Gaussian Monocycle Monocycle Pulse Train

b) Normalized spectrum for the non-uniform train of Gaussian monocycles

Figure 2.5. Non-uniform train of Gaussian monocycles in time and frequency domains


To give the reader a practical view of the signal characteristics, an example of a

typical signal used by Time Domain Corporation have the following characteristics:

Bandwidth of 2 GHz (>1 GHz). fc typically in the range 650 MHz – 5GHz. Tightly controlled pulse-to-pulse interval. Pulse width 0.2 –1.5 ns. Pulse-to-Pulse interval 100-1000 ns.

2.4 Coding and Modulation

Coding and modulation were discussed briefly in the previous section. In this

section more details about these topics are presented.

2.4.1 Coding

All source and channel coding applicable for narrowband systems are also

applicable for UWB systems. The advantage for UWB communication is the fact that

UWB signals seems to be easier to deal with because the signal is readily presented in a

digital form. UWB technology can be considered as the modulation layer of the

communication system. Thus, the remaining coding principles for the higher-level

communication layers, which are used in narrowband communications, are also valid for

UWB communication

Pulse position coding or “dithering” [Ful91] is a basic building block of the

proposed multiple access UWB system. Pseudo-random noise coding (PN Code) is used

for channelization. The code is used to apply a relatively large time offset at every frame.

Each user has a different code. Only the receiver with the same code can decode the

transmission. In the frequency domain the PN code makes the signal like noise. The code

is essential to suppress multiple access interference [Mar00]. PN codes must be

orthogonal to one another. They must effectively smooth the energy distribution and

allow fast signal locking [Ful00].

In addition to channelization and energy spectrum smoothing, the PN code makes

the UWB signal highly resistant to jamming as explained in Section 2.5.1. It is worth

24


mentioning that time dithering is also essential in radar applications for decorrelating the

ambiguous returns because of the high repetition rate [Ful91].

As was previously mentioned all coding concepts that are applicable for

narrowband can be extended for UWB systems. [For00] presented a practical low rate

convolutional error-correcting code for UWB communication. As expected, the coded

scheme outperforms the uncoded one. In other words, at a given bit error rate the coded

system increases the number of users by a factor which is logarithmic in the number of

pulses used by the time hopping spread spectrum system. [For00]

2.4.2 Modulation

Narrowband modulation includes amplitude modulation (AM), frequency

modulation (FM), phase modulation (PM) and many other variations [Skl88]. UWB

impulse radio can be modulated in analog form or digital form. [Win97a] presented a

comparison between analog and digital impulse radio for wireless multiple-access

communications.

UWB signals are usually time domain modulated using pulse position

modulation. This modulation allows for the use of an optimal receiving matched filter

technique [Pul00]. Pulse position modulation is accomplished by varying the pulse

position about a nominal position. For example in a 10 Mpps (Mega pulse per second)

system, pulses would be transmitted nominally every 100 ns. If the information bit is “0”,

the pulse would be transmitted 100 ps early. For a digital bit of “1”, the pulse would be

transmitted 100 ps late.

As it was mentioned before, PPM is expected to distribute the RF energy across

the band by smoothing the spectrum of the signal [Sch93]. However, the spectral

smoothing is small because the pulse position modulation only moves the pulse a very

small fraction of the pulse width.

[Ram98a], [Ram98b], and [Ram99] discussed higher order time domain M-ary

pulse position modulation. It was shown that by increasing M to a value more than 2, it is

25


possible to reduce the bit error rate, BER, or to increase the number of users at the same

bit error rate. [Sus00] presented a novel chaotic secure modulation. It is different than the

PN code because the sequence need not be periodic. In a chaotic transmitter, a message

signal undergoes two levels of pulse modulation. First, a frequency modulation is used to

modulate the message into subcarrier to be used as the clock pulses of a chaotic circuit.

The modulated clock pulses drive the coaotic circuit to generate the positions of the

carrier impulses. The objective is to guarantee that the time interval between the pulses is

chaotic. Thus the spectrum is smoother. Demodulation is done in two stages. First, the

timing between the pulses is recovered. Second, with the knowledge of the transmitter the

locations of the inner clock pulses are used to demodulate the message signal. No special

synchronization at any level is needed in this chaotic modulation. The level of security

depends on the hardware parameters of the chaotic circuit and the inner clock pulse train.

A single bit of information is generally spread over multiple monocycles. Thus, to

demodulate the received signal, the receiver sums the proper number of pulses to recover

the transmitted information. The receiver is based on decorrelating the received impulse

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-1.5

-1

-0.5

0

0.5

1

1.5

Time (ns)

Ampl

itude

received signal bit=0received signal bit=1template signal v(t)

Figure 2.6. Typical received signal for bit=0, bit=1 and the typical

waveform used by the receiver correlator

26


with a template signal as the one shown in Figure 2.6 assuming that the received signal

has the doublet shape. The template signal is the difference between the pulse that

represent an information bit=1 and the pulse used for an information bit=0.

2.5 Interference

Interference on UWB systems can result from other UWB users - multiple access-

, multipath, other narrowband systems, etc. Similarly, interference to narrowband system

could be multiple access, co-channel, multipath, leakage from other narrowband systems,

or UWB interferences. First, the interference from other radiators to UWB systems is

reviewed. Second, the interference on other narrowband systems as a result of UWB

system will be discussed.

2.5.1 Interference from Other Radiators to UWB Systems

UWB receiver has to deal with many narrowband radiators. The external

interference to the UWB receiver strongly depends on the antenna. Measurements of the

received power across the spectrum using UWB antenna give an illustrative image of the

interfering signals. A significant amount of lower frequency interference power (TV, FM,

and land mobile radiators) can get through the antenna’s frequency side lobs below the

main pass band of the UWB antenna system.

A specific UWB antenna tested by [Sch00] resulted in an interference power of

about -33.5 dBm when the entire system spectrum was utilized. The level of the

interfering power was reduced greatly by using a bandpass filter at the front end of the

receiver. With 97% bandwidth usage (780 MHz, 2.05 GHz) the interference power level

reduced to -40 dBm. A further reduction to 86% bandwidth usage (960 MHz, 1.93 GHz)

removed the strong interference at the edges (900 MHz). The captured signal approached

the noise floor of the spectrum analyzer which is nearly -60dBm.

The previous analysis suggests the possibility of incorporating a notch filter to

remove the strongest interferer. It is important to note that since the noise is highly

dominated by specific interferers, the level of the interferer power would be sensitive to

27


the location of the measurements. This is because the interferer signal may suffer from

multipath enhancement or fading [Sch00].

UWB Multipath interference is not a problem due to the high time resolution.

Signals reflected from different objects can be easily resolved from the line of site signal.

Practical multiuser interference rejection performance is not guaranteed. The processing

gain is used as a theoretical measure to the system ability to reject interference.

Processing Gain and Interference Resistance

The inherent pseudo-random code that is usually associated with UWB

communication system makes the system highly resistant to interference. All other

signals act as jammers to the UWB communication system. UWB signals are designed to

share the same band as other existing systems.

The processing gain reflects the ability of the system to resist interference. It is

defined as the ratio of the RF bandwidth of the signal to the information bandwidth of the

signal. A UWB system that transmits 8 kHz of information using 2GHz of bandwidth has

a processing gain of 250,000 or 54 dB. For example, a 2 GHz with 10 Mpps transmitting

8 kbps would have a processing gain of 54 dB, because 0.5 ns pulse width with a 100 ns

pulse repetition interval represents 0.5% duty cycle (23dB) and 10 Mpps/8,000 bps=1250

pulses per bit (31 dB). The total processing gain then becomes 31+23=54 dB.

2.5.2 UWB Interference to Other Systems (Narrowband)

Communication applications require signals free from interception and

interferences. The electromagnetic (EM) environment due to the UWB systems can

occupy the entire used EM spectrum band of 100 kHz to 10 Ghz. A careful study needs to

be carried out to mitigate the intra system and inter system electromagnetic aspects of the

UWB technology.

The instantaneous power from an UWB source could be very high even with a

very low average power due to the low duty cycle. It is the question of which one is more

dominant to interference: average or instantaneous power? According to [Ful00], the

28


driver for local interference is the average power rather than the peak power in most

scenarios. However, other researchers are opposing this conclusion [Bar01].

There are already thousands of devices that operate across the spectrum like hair

dryers and computers [Ang99]. Some researchers claim that UWB transmitters could be

designed to radiate no more than a hair dryer! [Ful00] claims that tested UWB prototypes

cause only one-quarter of interference of laptop computers or other comparable electronic

devices. Another important issue is the noise generated as an aggregate level that would

cause interference when multi UWB users are in operation.

In addition to the level of interference, the possibility of narrowband front-end

electronic upset should be studied. Those front ends were not designed with UWB

transient signals in mind.

Interference and Regulations

Within the United States, much of the work in the UWB field was carried out

under classified U.S. Government programs. Since 1994, however, much of the work has

not been restricted to the Government and the developments accelerated greatly. The

work continues today with more researchers in the area. The business interests in UWB

technology are growing exponentially.

In Aug. 1998, an FCC note for UWB was released in order to "investigate the

possibility of allowing the operation of UWB radio systems on an unlicensed basis under

Part 15 of its rules". There is a wide agreement that this technology is very promising,

and that there is a very broad applications range. However, there is strong concern to

allow the UWB devices to operate below 2 GHz or even below 3 GHz. Some researchers

are proposing that UWB technology should be licensed and some say that this technology

should not use the spectrum for free. A more moderate opinion is that this technology is

still immature and interference problems may or may not arise. Many researchers are

proposing to extend the period of time to complete the interference tests. On February 14,

2002, the Federal Communications Commission (FCC) issued a first report and order for

29


UWB technology, which authorizes the commercial deployment of UWB technology

[UWB].

Studies of UWB Interference to Existing Narrowband Systems

Interference to GPS receivers is a cause for concern and the results of interference

analyses will strongly influence regulatory decisions. This is because FCC is committed

to ensuring that GPS is protected from interference. Significant amount of work is being

carried out on measuring and analysing the interference to GPS receivers [Sch00],

[Ham01a]. Other important existing systems to be protected from UWB interference are

also initiating measurements and analysis projects [Bee]. Below are some of these

studies:

GPS band: [Sch00], [Ham01a].

GSM 900 uplink band: [Ham01a].

UMTS(Expand)/WCDMA frequency bands: [Ham01b].

CDMA PCS 1.9GHz band: [Dep].

Navigation devices in aircraft: [Ada01].

NTIA report 01-383, submitted to the FCC [Kis01].

2.6 Security

UWB technology has great value in the development of low probability of

intercept and detection (LPI/D) communication systems [Fon]. UWB employs baseband

pulses of very short duration, typically about a nanosecond, and spread the signal energy

through the entire used spectrum. Due to the low energy spectrum, probabilities of

intercept and detection are both low. According to Time Domain Corporation, 5

milliwatts spread over more than 2 gigahertz of bandwidth is virtually indistinguishable

from noise

Time hopping codes is a communication security feature integrated in the UWB

multiple access system. Encryption will add to the security of the system. A new

30


development in the area of secure UWB communications utilizes fractal mathematical

characteristics. It is based on chaos processes and identifying signals by their attractors

[Ang99]. The advantage of the low average power UWB system over narrowband

systems is the communication security rather than just the information security.

2.7 Hardware

The hardware requirements for UWB communications is based on sophisticated

digital electronics. The basic components for UWB communications have so far been

used are the following: pulse train generator, pulse train modulator, switch pulse train

generators, detection receivers, leading edge detector, ring demodulators, monostable

multivibrator detectors, correlation detectors, signal integrators, synchronous detectors

and wideband antenna.

Figure 2.7 presents a higher-level block diagram for a UWB transmitter and

receiver. Amplifiers are not shown explicitly. The power level of the pulse generator is

assumed to be sufficient. The receiver is similar to the transmitter except that the

generated pulses are fed to the correlator. Signal processing is required to extract the

modulated information and control signal acquisition and tracking [Pul00].

Conventional narrowband transmitter/receiver block diagram includes

intermediate frequency (IF) stages. Considering the number of building blocks, UWB

radios are simpler to build than equivalently sophisticated conventional narrowband

radios. The key elements include the pulse generator, the antenna and the receiver front

end.

31


Pulse Generator

Time Hopping Code Generator

Modulation

Data Source Programmable Delay Control

Clock

Transmitter

X∫T

0Sample and Hold

Correlator

Pulse Generator

Programmable Delay Control

Clock

Signal Processing

Data Sink

Time Hopping Code Generator

Receiver

Figure 2.7. Higher-level block diagram for a UWB transmitter and receiver.

Acquisition

Tracking

Pulse Generator (Source)

The pulse generator consists of a single transistor that operates in a digital mode

and switches between “0” and “1” states. The step waveform at the transition is filtered to

produce Gaussian-like pulse shape. Cost and power are reduced by eliminating the need

for linear amplifier as in narrowband conventional communications.

Antenna

Antennas should be specifically designed for the UWB task. Ordinary

“wideband” antennas do not transmit fast transients because they are not corrected for

dispersion [Pra99]. Both frequency and spatial dispersions should be taken into

consideration. Frequency dispersion can be overcome by TEM horn or biconical

antennas. Both structures are frequency independent. Spatial dispersion can be reduced

by using lenses and reflectors [Pra99].

The total frequency domain gain transfer behavior on transmitting and receiving

ends must be designed to offset the path loss effects. Knowledge of the transmit antenna

characteristics such as gain, impedance matching, and radiation pattern are particularly

32


important for UWB impulse signal collection. When an UWB transient signal is

collected, the receiving antenna transfer function should also be documented clearly so

that UWB signals can be accurately compared if collected with different antenna. If the

receiving antenna transfer function does not match across UWB collection system, the

collected signals will have different time and frequency domain characteristics, leading to

invalid and/or multiple signal designations to one transmitter type.

More than one antenna could be used to account for different bands [Pra99].

Designing antennas for UWB communications is an important asset to the success of

UWB communication technology. More details about Ultra wideband sources and

antennas are discussed in [Pra99], [Age98], and [Bor99].

Receiver Front End

An important requirement at the receiver is the ability to sample at very high rate.

At the receiver side, no intermediate frequency (IF) stages are required. Synchronization

with the transmitter is a key requirement for a successful communication. Timing

requirements are more constrained in UWB technology. Performance degradation due to

time jitter is discussed in [Win99]. Very low jitter electronic circuits are required.

Proposed applications require no more than 15 ps timing accuracy and stable time bases

[Ful00]. Timing requirements make building discrete components very difficult.

Application Specific Integrated Circuits (ASIC) implementations are relatively simple

and inexpensive [Dic99]. Time Domain Corporation is developing three chips, namely: a

timer, a correlator and a CMOS logic chip. Their timer chip is capable of handling the 10

ps timing required by the UWB radio. The correlator consists of a variable gain amplifier

and three correlating circuits. Both the timer and the correlator are built using the Silicon

Germanium process. The CMOS logic chip consists of analog to digital converters, about

300,000 logic gates and relatively reduced instruction set (RISC) microprocessor [Pul00].

When produced in substantial volume, these chips are expected to cost few dollars.

According to Time Domain Corporation, a full-featured radio would cost under $20.00

[Pul00].

33


2.8 Applications

The original proposal for UWB is based on the assumption that UWB is immune

to multipath, can support higher rate communications, the signal has a better wall

penetration capabilities, and may consume much less power. These features support

many applications including indoor static wireless LANs [Vre00], home-networking,

asynchronous transfer mode (ATM) multimedia [Win99], un-centralized multiple access

communication and secure military applications [Ang99].

Besides communication applications, UWB technology has many applications in

military, aviation, and space fields. UWB technology applies to both civil and military

applications. UWB systems can be used for navigation and communication. It can be

used for range measurements [Ada01]. UWB receivers can time the transmitted pulses to

within a few thousand billions of seconds. By measuring the round trip delay to within

that level of accuracy, it can be determined whether an aircraft’s wing-flaps are up or

down [Intro].

UWB technology supports integration of services. Some of the services and

applications that can be integrated with UWB communications are described briefly

below.

Precision Geo-location Systems

UWB can work as an augmentation to the Global Positioning Satellite System

(GPS). GPS uses a constellation of satellite to transmit radio signals that carry timing

information. By utilizing multiple satellite signals, GPS receivers can calculate the time

to within 20 ns. Standard GPS could locate objects to within 100 meters horizontal and

300 meters vertical. Improvements to within 1.5 meter could be achieved using

differential GPS [intro]. Time modulated UWB signals are superior to the GPS resolution

by three orders of magnitude. This could result in sub-centimeter range resolution. UWB

signals are immune to multipath, which is a major problem for GPS receivers in the

vicinity of buildings and large topographic features.

34


If reference stations are equipped with UWB technology, precise location could

be determined especially within buildings and areas were GPS fails to operate! [Pul00].

Precision positioning allows aircraft to locate monitor their position relative to other

aircrafts and relative to the ground. A GPS augmented with UWB technology will allow

an instrument landing even under inclement weather.

Imaging

UWB sends many pulses in a very short time. Millions of pulses a second are sent

to provide a near-perfect picture of what the target looks like [Intro]. UWB radar can

reduce post detection signal processing. Traditional narrowband radar requires

computationally intensive fast Fourier transforms (FFT) and inverse fast Fourier

transforms (IFFT) to remove Doppler shifts and to improve the emerging image. Due to

the inherent time resolution in time modulated UWB signals, such computationally

intensive processes are not required [Pul00]. The technology could also be used to predict

and monitor fatigues and failures within smart aerospace structures [Ful00].

Through –Wall Sensing Radar

In order to track multiple small targets, high resolution is required. On the other

hand, low operating frequency is demanded so that the signals can propagate effectively

through walls (cement and brick). Precision time gating is required to separate signals

from outside range of interest. This allows targets at longer ranges (and lower signals

level) to be detected [Pul00]. The technology is also being tested to look inside closed

rooms that might be harboring suspects before battering the door down [intro].

Narrowband radar relies on high-frequency radio waves to achieve high

resolution. Unfortunately, high-frequency radio wave has short wavelength and cannot

penetrate effectively through materials. On the other hand, UWB receiver can time the

transmitted pulses to within a few thousand billions of seconds, and still promise good

penetration through materials.

35


Underground Penetrating Radars

UWB underground penetrating radars could help detecting barely breathing

bodies under several meters of rubble after an earthquake [Ang99]. Moreover, they could

be used to check for power cables and pipes beneath the plaster before drilling [Intro].

UWB radars are promising for space applications to study the soil of different planets

without the need to drill. Water and ice layers could be identified using this technology

[Ful00]. D.J Daniels [Dan96] discussed in detail the system design, modulation

technique, signal processing and other application of surface-penetrating radar. Other

applications of the UWB underground penetrating radars include: target specific

applications, nondestructive testing, geophysical applications, civil engineering

applications, roads, and remote sensing.

UWB technology is also promising in other fields such as automobile collision

avoidance, computational fluid dynamics [Ful00]. Bennett and Ross have reviewed

different applications of UWB technology [Ful00]. More sophisticated applications are

expected due to the recent developments in application specific integrated circuits

[Dic99].

Based on the previous discussion, there is a strong relation between the term time

domain and the term UWB. On the other hand, narrowband communication lends itself to

frequency domain techniques. In the next chapter time domain and frequency domain

techniques are compared as alternatives to characterize the UWB communication

channel.

36

“When you can measure what you are speaking about,

and express it in numbers, you know something about it.”

Lord Kelvin

Chapter 3

Time Domain and Frequency Domain

Channel Measurement Techniques and

Setups

3.1 Introduction

UWB characterization can be achieved by performing measurements in the time

domain or in the frequency domain. The goal of studying different alternatives for

measurements is to come up with measurements that are reliable, repeatable, not overly

complicated, cost effective and reflects the real characterizing parameters.

Channel characterization may refer to extracting the structure of the channel from

the measured data. Data can be measured in different ways using variations of setups.

The objective of this chapter is to compare frequency-domain vs. time-domain techniques

to characterize the UWB communication channel. In this regard the following issues are

addressed: measurables, measurement approaches and setups, and calibration schemes.

The pros and cons of each technique are highlighted. Moreover, in this chapter the

measurement setups including the antennas are characterized to serve as base for

37

Chapter 3. Time Domain and Frequency Domain Channel Measurement Techniques and Setups

interpreting the results to follow in the next chapters. Characterizing the pulse generators,

the antennas and the receiver is an integrated part of the UWB channel characterization

efforts, as their effect cannot be completely deconvolved from the presented results.

3.2 Channel Impulse Response and Measurable Parameters

In time domain, a system -being it a communication channel or not - is described

by its impulse response. The transfer function is the corresponding frequency domain

alternative for describing the system. Though, one can transfer from one domain to

another, some parameters are easily measured in the domain in which they are defined. It

is not always convenient to present the channel by its impulse response or transfer

function because of the large data storage and processing requirements. Parameters based

on the received time domain waveform or spectrum distribution are used instead.

In the following section, measurable parameters are reviewed. At the fundamental

level, pulse parameters including pulse risetime and bandwidth should be accurately

measured. A measure that gained special importance for the UWB channel is the material

penetration/reflection capability. This is because the proposal of UWB promised an

excellent through-the-wall communication capability. Multipath response of the channel

is an important measure that helps characterizing the performance of communication

systems. The multipath response is well represented using the multipath power delay

profile.

3.2.1 Pulse-Shape and Frequency Distribution

Occupied bandwidth represents a main feature of communication systems. When

characterizing UWB signals, it is important to be able to accurately measure signal

spectrum or the corresponding pulse shape and transient durations. As the pulse width

decreases, the risetime becomes a significant fraction of the total signal pulse duration.

For this reason, the bandwidth for transient pulses is related to pulse risetime instead of

pulse width [Kis92]. UWB communication results in no generic pulse shape, all are

damped transient. The effect of transient RF signals, as opposed to steady state signals on

material and circuits is a complex subject [Bar01].

38


An important measure is pulse risetime or alternatively the occupied bandwidth.

Manufacturers of transient, single-shot collection equipment recommend that the analog

input (3dB) bandwidth should support at least three samples along a signal’s rising edge

to reduce error in risetime measurements.

3.2.2 Material Penetration/Reflection Capability

Materials encountered in indoor wireless communication channels need to be

characterized. The reflection of the ceiling, floors, inner walls and external walls are

different due to the difference in the effective permittivity. Effective Permittivity for the

floor, ceiling, external walls, and inner walls are reported to be 6.2, 6.2, 4.2, and 5.0

respectively [Kon99][Rap96]. Permittivity is frequency dependent. Due to the ultra

wideband proposal, the effective permittivity needs to be re-evaluated for indoor

construction materials.

The insertion transfer function is first measured across the spectrum of interest,

then different parameters can be extracted from it. Parameters of interest include the

complex effective permittivity or alternatively the dielectric constant and the loss tangent.

Propagation through different materials can also be characterized in the form of

attenuation constant and phase constant.

3.2.3 Multipath Profile Parameters

When the channel is exited with a pulse, the received waveform is a summation of

modified pulses with different attenuation factors and different time delays. The received

waveform is referred to as multipath profile and the individual pulses are referred to as

multipath components because they arrive to the receiver through different paths.

The pulse travels from the transmitter to the receiver through different paths

having real positive gain, , and propagation delays, ka kτ , where k is the path index.

Assuming no-dispersion within individual pulses, the channel impulse response is real

and can be represented as a superposition of these paths as in [Win97b]

39


∑ −=k

kk tath )()( τδ , (3.1)

where (.)δ is the Dirac delta function. An example of a discrete-time impulse response of

a multipath indoor channel is depicted in Figure 3.1. The excess delay is a measure of

time delay relative to the first arriving component

Excess Delay τ ∆

Maximum Excess Delay

N ∆τ

τ

) ( τ r P

Figure 3.1. Example of discrete-time impulse response model for a multipath indoor channel

The signal at the input of the receiving antenna is the time convolution of the

radiated pulse, , and the channel impulse response, h(t), as follows )(tp

)()( kk

k tpatr τ−= ∑ . (3.2)

The signal to noise ratio can be improved by different techniques. For example,

the signal can be matched-filtered. The filtered signal is given by

)()( kk

kf tatr τγ −= ∑ , (3.3)

where )(tγ is the convolution of p(t) with p(-t). Assuming that there is no overlap of

pulses, i.e., >− lk ττ resolution, e.g. 2 nanoseconds when lk ≠ , the filtered power

profile after passing through a square envelope detector is

40


∑ −== 222)()()( kkff tatrtP τγ . (3.4)

The unfiltered power delay profile can be expressed as

∑ −=k

kk tpatP )()( 22 τ (3.5)

Based on (3.5) some different time dispersion parameters can be defined.

Time Dispersion Parameters

All time dispersion parameters are measured relative to the time of arrival of the

first component. The profile energy is normalized and all signals below a specific

threshold X dB relative to the maximum are forced to zero. Presenting the time dispersion

parameters for a specific threshold eliminates the noise that varies from measurement

setup to another.

The maximum excess delay (X dB), maxτ , of a power delay profile is defined as

the time required for the energy to fall X dB relative to the maximum [Rap96]. The mean

excess delay is the first moment of the power delay profile [Rap96]

∑

∑∑

∑==

kk

kkk

kk

kkk

P

P

a

a

)(

)(

2

2

τ

ττττ , (3.6)

and the RMS delay spread is the square root of the second central moment of the power

delay profile [Rap96]

22 )(ττστ −= , (3.7)

where

∑

∑∑

∑==

kk

kkk

kk

kkk

P

P

a

a

)(

)( 2

2

22

2

τ

ττττ . (3.8)

41


The ratio of the mean excess delay to the RMS delay spread can be used as a measure of

the time dispersion for UWB signals.

In the frequency domain, similar quantities are defined and widely used in

narrowband communications. For example, the coherence bandwidth, BC, is a statistical

measure of the range of frequencies over which the channel is considered flat. Within this

band different frequency components have strong amplitude correlation. If the coherence

bandwidth is defined as the bandwidth over which the frequency correlation function is

above 0.9, then the coherence bandwidth is approximately [Rap96]

τσ50

1≈CB . 90% (3.9)

For 0.5 frequency correlation, the coherence bandwidth is approximately

τσ5

1≈CB . 50% (3.10)

Pathloss and Power Attenuation

Another useful parameter is the total multipath power gain. It describes the energy

characteristics of the multipath profile P(t) and it is given by

∑=k

kaP 2 . (3.11)

The spatial average of the power gain, P as a function of the distance d from the

transmitter is generally a decreasing function of the distance d. The logarithmic value of

this attenuation is

−=

)()(log10)(

010 dP

dPdPL (3.12)

where is a reference distance. The last quantity can be compared to the free space

propagation loss for different values of pathloss exponent, n, like 2 and 3,

0d

. (3.13) )(log10)( 10n

n ddPL −−=

42


Although all these measurable parameters were developed for narrowband

systems, many researchers are extending their use to UWB characterization problems.

However, new parameters may be more efficient in characterizing the UWB multipath

propagation channels.

3.3 Measurement Setups

Wideband measurements are performed using setups assembled and tested in the

Time Domain Laboratory at Virginia Tech. The objective of the following sections is to

characterize the measurement setup to be used for wideband material characterization and

indoor UWB measurements. Measurements are performed using both frequency-domain

and time-domain techniques. The measurements presented in this section help de-

embedding the effect of the sources and sampling scopes. The effect of the antennas and

the phase shifters associated with them is also studied to allow for a partial de-

embedding. Details about the equipment are given to ensure repeatability of

measurements and corresponding results.

3.3.1 Time Domain Measurement Setup

One of the essential properties of time domain measurement techniques is the

ability to distinguish discontinuities and time separations between them. A schematic

illustration of the components of a time domain measurement system is shown in Figure

3.2a. The pulse generator triggers a sampling oscilloscope as the pulse enters the system

under test. The figure displays the measurement locations of both time domain reflection

(TDR) and transmission (TDT) measurements. In the case of the Gaussian-like pulse

input, which has frequency spectra that extend to dc, the time-domain resolution depends

on the risetime of the pulse, which is inversely related to the pulse’s bandwidth.

An experimental setup was established to evaluate the performance of impulse

radio. The setup is shown in Figure 3.2b. It consists of a pulse generator that furnishes

pulses to the transmitting antenna. The antenna is preceded by a balun which converts the

unbalanced coaxial terminals to two terminals feeding the signal to the balanced antenna

terminals. The baluns and the antennas, on both the transmitting and receiving sides, have

43


sufficient bandwidth such that the spectral characteristics of the pulse signal are not

degraded. Radiated pulses propagate through the channel and are captured by the

receiving antenna. Received pulses are acquired by means of a sampling oscilloscope and

a data acquisition unit. The received signal suffers attenuation and dispersion whose

degree depends on the characteristics of the channel as well as the radiation patterns of

both transmit and receive antennas. Matched load calibration is used to cancel the

oscilloscope offset.

Synchronization is achieved through an external circuit. The sampling

oscilloscope requires a pre-trigger. The oscilloscope has to receive the pre-trigger 80 ns

before the trigger signal to the transmitter. This is achieved by using a step generator

driver that can supply the required trigger and pre-trigger. The trigger setup on the

sampling oscilloscope is set to the negative slope with about 350 mV. It is seen that

lower or higher trigger levels resulted in higher jitter for the used setup.

As the distance between the receiver and the transmitter increases, the need for a

wideband amplifier becomes more pronounceable. Different sets of antennas or different

sizes in conjunction with different various sources for generating pulses can also be used.

The setup can be manipulated to handle multiple antennas for diverse communication.

Table 3.1 below exhibits a more detailed list of the apparatus used for the ultra wideband

experiments.

44


45

RF PULSE GENERATOR

SAMPLING OSCILLOSCOPE

BRIDGING T NETWORK

Transmission Reflection

Channel or DUT

(a)

(b)

Figure 3.2. Time domain measurement setup,

a) General TDR and TDT measurement setup, b) Illustration of the time domain measurement setup.

Radiated Measurements

Conducted Measurements

2*10 dB amplifier, up to 15 GHz

trigger input

trigger

pre-trigger

Running LabView® 6.0i

Channel

Balun and wideband horn receiving antenna/ biconical antenna

Balun and wideband horn transmitting antenna/ biconical antenna

Data Acquisition Unit

Step Generator Pulse Generator PSPL-4100/4050A

Digitizing Oscilloscope


Table 3.1. Component specifications used for the time domain measurement setup

Equipment Description Pulse Generator Picosecond Pulse Generator 4050A / 4100 Digital Sampling Oscilloscope Tektronix 11801 or HP 54120A with HP

54124A Four Channel Test Set. Trigger Box (Controlled Delay)

TDL home made

Thin Semi-Ridged Cables To connect the signal BNC Cable Pre-trigger and trigger connection, low noise

coaxes 50Ω and connectors. 20dB Attenuator DC-26.5 GHz HP 33340C, for conducted measurements Data Acquisition Unit Collect Data, Laptop or Desktop with GBIP card BNC to male SMA Connect the pre-trigger cable to the scope Antennas (Horn) #1 A pair of horn antennas with baluns Horn array Antenna #2 A pair of horn antennas array with wider

bandwidth and 180 phase shifter at the input Biconical Antennas#3 A pair of wideband omnidirectional biconical

antennas Balun or 180 phase shifter Required to change to/from co-axial

configuration, The PSPL Model 5320A BALUN is used with antenna#1, [PSPL-5320A] and the Krytar 4010124, 1-12.4 GHz is used with antenna#2

Ultra wideband Amplifier 10dB inverting, 15GHz, , two amplifiers may be cascaded I\5828 Ultra-Broadband Amplifier, [PSPL-5828]

3.3.2 Frequency Domain Measurement Setup

The frequency domain measurement system is displayed in Figure 3.3a. The

general block diagram for both the frequency domain reflection (FDR) and frequency

domain transmission (FDT) are shown on the same figure. It is important to note the

presence of the vector analyzer, which is needed for recovering the amplitude and phase

of the system response. Vector analyzers employ calibration standards to reduce the

reflection and transmission errors at the expense of an increased number of measurements

resulting in increased measurement time. System bandwidth determines the limits to

pulse and spectral line widths as well as line separation in the frequency spectra, which

determines the time duration.

46


47

(a)

(b)

Figure 3.3. Frequency domain measurement setup;

(a) General FDR and FDT measurement setup, (b) Illustration of the frequency domain measurement setup

Directional Coupler

DirectionalCoupler

Reference Reflection

DirectionalCoupler

Directional Coupler

Reference

Reflection

SYNTHESIZEDSWEEPER

Channel orDUT

VECTOR ANALYZER

Data Acquisition unit

Port 1

X(ω)

Port 2

Y(ω)

Balun and wideband transmitting antenna

)()()()(21 ω

ωωωXYHS =∝

S-parameters test set

Vector Network Analyzer withSwept Frequency Oscillator

Balun and wideband receiving antenna

Channel

)(ωH

[ ])()( ωHIFFTth =

Inverse FFT Processing


Figure 3.3b is a more detailed and illustrative view of the frequency-domain

characterization. A network analyzer is used for performing swept frequency

measurements. Port 1 is connected to the transmitter and the receiver is connected

through port 2. The network analyzer sweeps the frequency within the measured band of

interest. For UWB characterization, one has to trade off the frequency resolution with the

required number of measurements, which is directly proportional to the time it takes to

perform the experiment as well as data storage requirements. Decision has to be made on

the measured bandwidth, number of points and the sweep time. The inverse Fourier

transform is used to obtain the impulse response of the UWB channel. Neither pulse

generator nor sampling oscilloscope is needed for the frequency domain measurements.

The additional equipment for the frequency domain measurements is the HP 8510

network analyzer.

3.4 Source Characterization and Conducted Measurements

In time domain, a full understanding of the characteristics of UWB propagation

requires two different types of measurements. First, individual pulses are captured

directly in time domain at the output of the pulse generator device. This procedure is

most commonly referred to as “conducted” measurements [Kis01]. Secondly, received

pulse is captured using a specific transmitter and receiver antennas. The following

section presents the results for such measurements.

Two pulse generators are used. The first pulse generator is the Picosecond Pulse

Labs® 4050A which has time jitter of ±2.5 ps. A 10 V, 45 ps pulse rise time is generated

in a remote fast pulser head module. A high quality 50 Ω cable is used to connect the

step generator to the pulse head. The pulse head allows us to connect the fast 45 ps pulse

directly to the input before the antenna. This approach eliminates the rise time slowing

effects of interconnecting coaxial cables. The generated pulse is a clean pulse with a

manufacturer nominal value of 0.3% for the precursor and 2% overshoot [PSPL-4050A].

To add variability dimension another source is also used. The second pulse generator to

be used is the Picosecond Pulse Labs® 4100 with ±2.5 ps time jitter. The generated

48


pulse has 5 V amplitude and an 85 ps rise time. More details about the second source can

be found in [PSPL-4100].

The output of the pulse generator is directly captured using a sampling

oscilloscope. A 20 dB attenuator is used to protect the input of the sampling

oscilloscope. The setup used for the conducted source measurements is further clarified

in Figure 3.4.

Pulse Generator

Delay Generator (Trigger Box)

Digitizing Oscilloscope

Data Acquisition unit

pretrigger

trigger

trigger input

trigger input

Pulse Head (and/or) Atten.

Figure 3.4. The setup used for the conducted source measurements

The measured pulse for the 4050A generator is shown in Figure 3.5. The shape of

the pulse is far from being Gaussian due to the smooth falling edge. It is apparent that

the pulse is smooth and ringing effect is not pronounced. The precursor and the

overshot are within the manufacturer specifications as can be seen from the zoomed view

in Figure 3.5b. The spectral occupancy is shown in the normalized frequency plot in

Figure 3.5c. It is worth to mention that due to the periodic nature of the pulse, the

frequency representation is discrete. However, the plot is represented with continuous

line. Such continuous representation is used for the sake of clarity. Only positive

frequency values are plotted due the symmetry of the frequency response.

For the 4100 source, the output pulse waveform and the normalized spectrum are

both shown in Figure 3.6. The main pulse duration is shorter compared with the 4050A

pulse, but damped ringing effects are much larger.

49


)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

0 2 4 6 8

10

Time ns

Ampl

itude

in V

Picosecond 4050A

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.2

-0.1

0

0.1

0.2

Time ns

Ampl

itude

in V

Picosecond 4050A, Zoomed

-20

0

dB

FFT

)

)

)
(c)(c
(b)(b

(a(a

50

Figure 3.5. The resulting waveforms for the Picosecond 4050A generator, (a) The generated waveform, (b) Zoomed version of the generated waveform, (c) Spectrum of the generated waveform.

0 5 10 15-60

-40

Frequency GHz


)

)

)

itude

i

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

0 2 4 6 8

10

Time ns

Ampl

itude

in V

Picosecond 4100

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

-0.2

0

0.2

0.4

Time ns

Ampl

n V

Picosecond 4100, Zoomed

-20

0

dB

FFT
(c(c)
(b(b)

(a(a)

51

Figure 3.6. The resulting waveforms for the Picosecond 4100 generator, (a) The generated waveform, (b) Zoomed version of the generated waveform, (c) Spectrum of the generated waveform.

0 5 10 15-60

-40

Frequency GHz


For both measurements mentioned above, two disparate quantities can be

computed. The first of which is the total peak power; ‘instantaneous’ power embedded

in the kth sample point on the transmitted or received waveform. The kth element in the

waveform should be the maximum absolute value in the array in order to calculate the

direct peak power as follows:

0

2 )max(Z

aP k

peak = (3.14)

where ak represents the instantaneous measured value point in the waveform sequence

and Z0 is the input impedance of the measurement equipment which is 50 in the

present case. The total average power is another quantity which can be computed using

the acquired time domain information as follows:

Ω

ta

PRRP k

kavg ∆

= Σ 50

2

, (3.15)

where Pavg is the average power, PRR is the pulse repletion rate, and ∆t is the sample

interval. The PRR is 1/177.3E-6 where 177.3E-6 is the pulse repetition interval as

determined by the trigger signal. For signal generated by the 4050A, the total peak

power is 33.46 dBm while the total average power is –22.30dBm. For signal generated

by the 4100, the total peak power is 29.58 dBm and the average power is -34.98 dBm.

The –10 and –20 dB bandwidth can be extracted from the frequency domain power

spectrum. These specific spectral characteristics are summarized below in Table 3.2.

Table 3.2 Total peak power, average power, and emission bandwidth for the used sources Device Peak Power

(dBm)

Average

Power (dBm)

-10 dB

Bandwidth

(GHz)

-20 dB

Bandwidth

(GHz)

-40 dB

Bandwidth

(GHz)

PSPL 4050A 33.46 –22.30 0.781 2.148 9.961

PSPL 4100 29.58 -34.98 7.813 9.765 12.891

52


53

The effect of the coax connecting cables is illustrated in Figure 3.7. Due to the

mismatch at the connection points there are some reflections in time domain plots. High

quality cables are used to minimize those reflections.

0 2 4 6 8 10 12 14 16 18 20 - 0.1

- 0.08

- 0.06

- 0.04

- 0.02

0

0.02

0.04

0.06

0.08

0.1

Time (ns)

Ampl

itude

(V)

Picosecond 4050A, Zoomed

Figure 3.7. Effect of connection coaxial cables, indicated by the circle

3.5 Antenna Characterization and Radiated Measurements

Our case consists of three pairs of antennas. Two of them are based on TEM horn

structure and the third pair has biconical design. The second TEM horn pair is wider in

bandwidth than the first TEM horn pair. Using different antennas enables variability

study and sheds more light on the importance of antennas to the UWB system.

For impulsive free-space measurements a TEM horn is suggested by [Law78].

TEM horns are quite broadband in receiving mode, both in magnitude and phase. The

suggested antenna was reproduced and tested in the Time Domain and RF Measurements

Laboratory at Virginia Tech. Structures and configurations for the antennas in

experimentation are described in Appendix B1.


Radiated measurements are used to characterize the antennas. The two antennas

are placed facing each other at a separation distance that ensures far field approximation.

If the time domain signals are acquired with the same receiver settings, and over a short

frame for good instrument stability, the received signal can be used as a reference. Under

same measurements settings, there would be no need to specify the transfer function of

the receiving system [Aur96]. The setup for the radiated measurements is similar with

those introduced in Figures 3.2 and 3.3.

The approximate separation requirement in order to achieve far-field or

Fraunhofer region characterization is given by the Fraunhofer distance which is as

follows [Rap96]:

λ

22Dd f = (3.16)

where λ is the wavelength and D is the largest physical linear dimension of the antenna.

To reside in the far field region two other conditions on df must be met, which can be

written as

(3.17) Dd f >>

λ>>fd (3.18)

For the two antenna pairs under consideration, D is 0.213 m and 0.279 m,

respectively. The required far field distances versus frequency are shown for the two

TEM antenna pairs in Figure 3.8. In view of the plot, to cover the frequency range of

interest up to 15 GHz, the far distances are approximately 4.5 m and 8 m, respectively for

the two antennas.

54


55

0 5 10 15 0

1

2

3

4

5

6

7

8

Frequency GHz

Far d

ista

nce

(m)

Antenna#1 Antenna#2

Figure 3.8. The distances required for far-field approximation vs. frequency

The antennas were placed facing each other as shown earlier in Figure 3.2. The

distance was fixed to 2.5 m and 7 m based on the antenna pair due to far-field

requirements. The measurements should be done in a reference channel that has minimal

multipath effect, such as anechoic chamber. However, a large enough chamber was not

at the time available for use. Therefore, the measurements were taken in an environment

with minimum reflectors. The main reflection is due to the ceiling and the floor. The first

antenna pair was characterized in the Time Domain Lab at Virginia Tech, with a total

height of 2.42 m. The antennas were kept at 1.03 m from the floor. A laser pointer was

used to direct the antennas. This setup results in a main reflection from the floor with a

time delay 2.46 ns relative to the line of sight path. Time gating can be applied safely if

the main received pulse width is less than 2.46 ns. For the second antenna pair the

experiment was carried out on the 3rd floor of Whittemore building on the open area

where there is no ceiling. The antenna elevation was increased to 1.42 m to increase the


delay of the reflection from the floor. The time gating capability for this setup then

becomes 1.81 ns.

The radiated measurements for the first antenna pair were carried out using time

domain and frequency domain techniques. The time domain radiated measurements were

performed using two sources. The received waveforms together with the gated

waveforms are shown in Figure 3.9 for source1 and source2. Time gating is less accurate

using source#2, because the ringing effect cannot be isolated from the multipath

reflection. The time-gated waveforms are then converted to frequency domain data using

FFT and divided by the corresponding frequency transform of the input signal to obtain

the transfer function. A set of 1024 points was acquired in a 5 ns window, which resulted

in a 5 ps sampling time. The acquired data are based on 128 averaged measurements.

Furthermore, the frequency domain measurements using the same antenna setup

were considered using network analyzer, as demonstrated in Figure 3.3. The S21

parameter was measured in the range of 45 MHz-15 GHz, where 801 points were

acquired in order to complete the experiment.

The magnitude of the transfer function for antenna#1 is presented in Figure 3.10

acquired by both the time and frequency domain measurements. The frequency plot is

based on 16+1 points moving average. The acquired frequency domain data within the

aforementioned range is first transformed to time domain to allow for time gating and

then transformed back to frequency domain. A Kaiser window with β=20 is used to

smoothly gate the required duration. However, time gating is not perfect because the

incomplete frequency information introduces error in the time domain representation.

The difference between time and frequency domain responses could be a result of un-

gated multipath components.

56


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 -0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Time (ns)

Ampl

itude

(V)

Received

Time gated

0 0.5 1 1.5 2 2.5 3 3.5 4-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

Time (ns)

Am

plitu

de (V

)

ReceivedTime-gated

(b)

(a)

Figure 3.9. Received and time-gated waveforms for antenna #1 at a distance of 2.5 m

(a) with the 4050A generator.

(b) with the 4100 generator.

57


0 5 10 15 -65

-60

-55

-50

-45

-40

-35

Frequency (GHz)

Tran

sfer

Fun

ctio

n Am

plitu

de (d

B)

4050A4100HP 8510 Network Analayzer

(a)

0 5 10 15 -3

-2

-1

0

1

2

3

Frequency (GHz)

Phas

e fo

r Ant

enna

#1 (r

adia

ns)

4050A 4100

(b)

Figure 3.10. Frequency domain transfer function for antenna #1 using the time domain measurements (a) Magnitude and (b) phase

58


The difference between the two time-domain measurements could be due to the

different spectral occupancy or errors in time gating as source2 has the ringing effect

which lasts longer than the time gating capability of the setup. The linearity of the phase

for the same antenna pair is shown in Figure 3.10. The relevant phase information could

not be extracted from the network analyzer measurements because of uncontrolled

multipath reflections. The frequency domain result pertaining to the antenna #1 shown in

Figure 3.10 confirms that the response is reasonably flat (± 3 dB) up to 7 GHz. [Law78]

showed that the far-field time-domain response of the transmitting antenna is a constant

times the derivative of the time domain response of the receiving antenna. In other

words, in the transmit mode, the low frequency roll off with decreasing frequency is

nearly 6 dB/octave steeper than in the receive mode.

The large distance required for far field performance limits the time gating

capability. Furthermore, frequency domain measurements are much harder to be

performed as inconveniently long wideband cables are needed to connect to the network

analyzer. Thus, antenna#2 is characterized only by source 4050A. The received and time

gated waveforms are shown in Figure 3.11. The magnitude and phase of the transfer

function for antenna#2 compared with antenna#1 are shown in Figure 3.12. Antenna#2

covers more bandwidth up to 12 GHz with some variation on the pass band. The phase is

linear in this band as can be seen from Figure 3.12. Linear phase components could be

added to the plots to account for the exact delay.

The biconical antenna has a wideband that covers 0.1-18 GHz. It has the

advantage of being omnidirectional. Time gating with biconical antenna is more difficult

because the antenna is omnidirectional and it is not optimized for impulsive radiation.

59


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

Time (ns)

Ampl

itude

(V)

received time gated

Figure 3.11: Received and time-gated waveforms demonstrated for the antenna #2 with the 4050A generator at a distance of 7 m.

60


0 5 10 15 -65

-60

-55

-50

-45

-40

-35

Frequency (GHz)

Mag

nitu

de (d

B)

Antenna#1 Antenna#2

(a)

0 5 10 15 -3

-2

-1

0

1

2

3

Frequency (GHz)

Phas

e i(r

adia

ns)

Antenna#1 Antenna#2

(b)

Figure 3.12. Frequency domain transfer functions of the two antennas using their time domain measurements with 4050A generator at a distance of 7 m

(a) Magnitude, and (b) phase.

61


0 1 2 3 4 5 6 7 8 9 10

0

0.0

0.0

0.0

0.0

0.1

Time (ns)

Ampl

itude

(V)

Figure 3.13. Received signal for biconical antenna (antenna#3) with the 4100

generator at a distance of 1 m.

62


3.5.1 Relation between Multipath Angle and Pulse Shape

It should be noted that previously presented radiated measurements are for

boresight direction. However, in practical cases, the antenna pattern plays an important

role for both the receiver and the transmitter. To illustrate this, the effect of the angle of

arrival using the antenna#1 with source#1 (PSPL-4050A) is considered in this subsection.

Vertical Rotation Measurements

The first scenario is to investigate the effect of vertical rotation of an antenna (E-

plane scan). In this case, the transmitting antenna is kept fixed at a height of 1.03 m

above the ground, whereas the receiving antenna at the same height is rotated along the

elevation angle, θ. Measurements are performed at different angles; namely ±150, ±300,

±450, ±600.

The schematic diagram embedded in Figure 3.14 demonstrates the first two

multipath components that are expected to be captured in time domain. Spherical lines

covering the receiver represent the rotation in the vertical direction as well as the antenna

far-distance electric field lines. Negative angles refer to rotation towards floor, and on the

other hand, positive angles in the reverse direction. When the receiver antenna is tilted

more towards floor – i.e. negative angles, – the line-of-sight pulse and the reflection off

the floor tend to move toward each other and vice versa. Another interesting feature is the

broadening of the original pulse when compared directly to the boresight reception.

Horizontal Rotation Measurements

The second scenario is to investigate the effect of horizontal rotation of an

antenna (H-plane scan). Similarly, the transmitting antenna is kept fixed at a height of

1.03 m above the ground, whereas the receiving antenna at the same height is rotated

along the spherical angle, φ, – i.e. in the horizontal plane. Measurements in time

63


0 1 2 3 4 5-0.02

0

0.02

0.04

0.06

0.08

Time ns

Am

plitu

de in

V

015-30-45-60-

0 1 2 3 4 5-0.02

0

0.02

0.04

0.06

0.08

Time ns

Am

plitu

de in

V

015+30+45+60+

ceiling

floor

Rx Tx

receiving antenna

60o

-60o

-45o

-30o

-15o

0o

15o

30o

45o

(a)

(b)

(c)

Figure 3.14. Received waveforms at different receiver elevation angles (E-scan), using antenna#1 and source#1 (a) experimental view, (b) negative elevation angles comparison, (c) positive elevation angles

64


i i

60o

-45o

-30o

-15o

0o

15o

30o

45o

Side Wall

Side Wall

Rx Tx

(a)

-60o

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.02

0

0.02

0.04

0.06

0.08

Time ns

Am

plitu

de in

V

015-30-45-60-

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.02

0

0.02

0.04

0.06

0.08

Time ns

Am

plitu

de in

V

015+30+45+60+

(b)

(c)

Figure 3.15. Received waveforms at different azimuth receiver angles (H-scan), using antenna#1 and source#1 (a) experimental view, (b) negative elevation angles comparison, (c) positive elevation angles

65


domain are repeated for various angles; namely ±150, ±300, ±450, ±600.

The schematic diagram embedded in Figure 3.15 demonstrates placement of the

antennas in the laboratory room. Spherical lines covering the receiver represent both the

rotation in the horizontal plane and antenna far-field lines. Negative angles refer to left

rotation, positive angles to right. In view of Figure 3.15 either rotation reveals identical

results when compared to each other. That is because the left and the right rotations are

symmetrical about the boresight direction. Another important feature is the swift

broadening of the original pulse when compared directly to the boresight reception.

A more comprehensive characterization can be done in an anechoic chambers.

Usually, antenna characterization is done in the frequency domain for only specific

frequencies. For accurate time domain applications, such task requires a huge number of

frequency measurements. The setup and the radiation pattern at 5GHz are displayed in

Appendix B1.

More discussions about the effect of the pulse shape on estimating the channel

impulse response and receiver design are presented in Chapter 6.

3.6 Calibration Schemes

Imperfections of the system components used in the measurement process can

result in errors. As an example, the calibration of the measurement system may require

some type of interfacing with the system under test. The identical interfacing used to

allow for the calibration standards is also used to connect the system under test to the

measurement system. Some effects of this interfacing can be included as part of the

system components as an imperfection. Other imperfections could force a proposed

characterization technique not to converge, or become unstable.

Another systematic measurement error, which is always common in measurements, is

the noise content of the measured waveforms. Noise can result from imperfections in the

calibration or from the effects of not calibrating all error components within the system.

Noise can also be introduced from external sources near the measurement system or the

66


system under test. Repeating the experiment and taking the average can reduce the

effects of the noise. [Yoh01]

For the frequency domain measurements the system can be calibrated using short,

open and matched load standard kits at both ports. A standard is a physical device having,

precisely defined characteristics, that is used as a reference for a unit of measurements

[Su92]. Through calibration can be done by connecting port 1 to port 2 directly. The

network analyzer adjusts the measurements based on the calibration.

For time-domain calibration, early calibration technique for networks involved the

use of a short circuit standard to produce a reference waveform. Two transmission lines

can be used as time-domain isolators (to isolate secondary reflections due to impedance

mismatch between the scope and the device under test (DUT). It is important to note that

the commercially available standards for frequency calibration might not be applicable

for time domain measurements due to the large bandwidth requirements. However, a

coaxial precision line can simulate a known impedance over a relatively wideband of

frequencies due to the ability to gate time domain waveforms. [Su92].

The channel can be measured with no input to account for any persisting signals

due to the setup or other sources. If the received signal does not average to zero while the

input is zero, it means that this measured signal has to be subtracted from the channel

measurements when the input is applied. Another process that might add to the

calibration is the deconvolution process where the input or conducted measurements is

deconvolved from the radiated measurements. Such procedure should result in an ultra

high-resolution channel characterization [Vau99] [Mov98]. In principle, this should make

the channel characterization independent of the input signal assuming the signal spectrum

cover the entire frequencies of interest. Ultra wideband application requires more

accurate and wideband measurement techniques.

3.7 Pros and Cons

Advantages and disadvantages of both the time domain and the frequency domain

will be discussed in this section. Rather than providing the reader with a list of pros and

67


cons, they are discussed to engage the reader in a thoughtful study of both techniques.

Making a decision on the technique to be used does not avoid the engineering tradeoffs.

Next paragraphs emphasize the time domain pros and cons followed by an emphasis on

the frequency domain.

One advantage of time-domain measurements is that they require only a single

measurement where the frequency domain techniques require frequency sweeping, which

involves a measurement at each discrete frequency. Sounding a channel with time

domain pulse requires relatively a simple implementation. Another advantage of the time-

domain technique is that the resulting waveform immediately gives a physical insight into

the characterization problem, whereas calculations (i.e. pulse development, convolution,

inverse Fourier transforms) need to be performed to the frequency domain equivalent

measurement before information is visually realizable by the user. Time-domain methods

are capable of operating directly on the time domain data, as the name suggests. The

advantage of performing the measurements in the time domain is a reduction of error

caused by the Fourier and inverse Fourier transform operators used to transform the data

to and from the frequency domain, respectively.

One problem associated with time-domain technique is that the method is more

susceptible to noise compared with frequency-domain techniques due to the wide variety

of noise reducing algorithms in the frequency domain. Moreover, typical measurements

that provide low noise results are designed for continuous signals rather than transient

signals [Bar01]. Also, time-domain measurement is limited by the pulse generating

equipment with respect to the pulse characteristics that may be used.

Synchronization requirements remain an issue for time domain measurements.

Any time jitter that exists on the system will deteriorate the quality of measurements.

Especially when the distance between the transmitter and the receiver increases, the

trigger signal will suffer more attenuation and dispersion. On the other hand, the vector

network analyzer (VNA) must be physically connected to the transmitter and receiver.

This limits the application of VNA measurements to channel length where the cable is

practical and does not cause dramatic attenuation at the frequency band of interest.

68


As mentioned earlier, frequency domain technique has the advantage of an

improvement in signal noise and jitter due to the ease of application of noise reduction

algorithms and the use of calibration techniques. Another advantage is that the limitation

on waveforms is reduced; waveforms can be created and applied through frequency

domain convolution before the Fourier transformation operation is applied.

Frequency domain technique is a good candidate for interference characterization.

Network and spectrum analyzers are industrial standard pieces of equipment. On the

other hand, digital sampling oscilloscopes are not common and more difficult to deal

with. Moreover, narrowband linear amplifiers are easier to build “cheaper” when

compared to the low noise ultra-wide amplifier required for time domain measurements.

The time required for data acquisition limit the application of the measurement

technique to slow varying channels. When the network analyzer is used, the sweeping

time represents a limiting factor since the impulse response can be calculated at the end

of the sweep. In time domain, real time measurements can be performed using real time

oscilloscopes. However, when digital sampling oscilloscopes are used, sampling and

averaging will limit the application to relatively slower channels.

It should be noted that power spectrum or power density spectrum does not easily

apply to single transient events unless careful attention is given to the sampling rate.

State-of-the art spectrum analyzers do not sample fast enough to capture the peak power

of the individual UWB. Even fast (>20 GHz) scopes will not sample the peak from

different emitters in the case of multiple access or multiuser UWB systems. Only real-

time oscilloscopes are capable of capturing UWB aperiodic transient noise. [Bor01]

With a constant energy, J, greater field strengths and powers can be obtained by

shortening the duration of the pulse. This is why the use of power spectral density as a

measure for the UWB emission should be questioned. [Bor01]

On one hand, the frequency domain method allows for convenient noise reduction

algorithms to be used. Filtering algorithms may be application specific and can be easily

implemented using frequency domain techniques. On the other hand, the application of

69


filters in the frequency domain as well as the Fourier transform errors could result in a

non-causal response and instability. Time domain processing thus preserves the causal

and stable nature of the signal.

An argument that has to be clarified that practical time domain and frequency

domain techniques are not equivalent. Transformation from one domain to the other is

not a reversible process in the digital domain. Leakage errors are a redistribution of

energy due to data windowing, which result from the Fourier or inverse Fourier

transforms. Leakage error occurs when the values on each side of the time or frequency

domain window are not continuous before performing the transformation operation.

These discontinuities are normally caused by windowing of obtained data or by the time

duration limitation or bandwidth limitation of the measurement equipment.

To clarify the difference between the two techniques, we will consider the

resolution, spectral occupancy, and dynamic range of both techniques.

3.7.1 Time and Frequency Resolutions

When conducting measurements in the discrete domain one is limited by

complexity and hardware to a maximum of N points. The relationship between limited

transition duration, τ, and frequency bandwidth, BW, is an inverse relation. The number

of points in the time domain transition duration, Nτ, and the number of points in the

frequency domain bandwidth, NBW, are given as

t

N∆

=τ

τ , (3.19)

f

BWNBW ∆= , (3.20)

where ∆t is the time spacing, or time resolution, and ∆f is the frequency spacing, or

frequency resolution. Multiplying Nτ and NBW results in the following relation

70


ft

BWNN BW ∆∆⋅

=τ

τ (3.21)

For passive networks, which are not under-damped, the bandwidth for transient

pulses is related to pulse risetime [Kis92] as

BW

35.0≅τ . (3.22)

Applying the same methodology to the number of points, N, results in the

following [Yoh01]

NNN BW 35.0≅τ (3.23)

Substituting (3.22) and (3.23) into (3.21), the number of points is related to the

time and frequency domain resolution as follows

N

ft 1=∆∆ (3.24)

In (3.22), the time duration is shown to be related to the reciprocal of the

bandwidth. It is obvious that higher speed transition duration results in a wideband signal

in the frequency domain. So faster pulses are needed to extend the bandwidth of the

measurements.

In the case of time and frequency domain resolution, the ideal situation is to have

the time spacing, ∆t, equal to zero which result in an infinite time duration, T = ∞, and an

infinite number of samples, N = ∞. This unrealistic situation is even more noticeable

noting that the measurement equipment would have to have an infinite bandwidth. The

equation that relates the time and frequency domain resolution to the number of samples

is shown in (3.24) where ∆t is the time spacing, or time resolution, and ∆f is the

frequency spacing, or frequency resolution. The time resolution depends on the risetime,

which is inversely related to the bandwidth of the pulse excitation function. A faster

excitation risetime gives higher time resolution while compromising the frequency

71


domain resolution as a result of the corresponding increase in bandwidth. The number of

samples, N, is limited by the measurement equipment, so through investigating (3.24),

there is an obvious trade off between frequency domain resolution and time domain

resolution. Higher frequency domain resolution will result in lower time domain

resolution and vise versa. The application determines the weight of the resolution of each

domain and how they need to be modified to obtain desired results.

Some methods to increase time and frequency domain resolution, which are

usually limited by the measurement equipment, will be investigated. The obvious

method of increasing the resolution is by increasing the number of points, which will

result in higher resolutions in both domains. An increase in the time domain resolution

can be obtained by increasing the sampling frequency. As in relationship (3.24), when

increasing the time domain resolution by increasing the sampling frequency, the

frequency domain resolution will be compromised,

sf

t 1=∆ . (3.25)

Another method for increasing resolution is through interpolation. Interpolation

can be accomplished through zero padding before the Fourier or inverse Fourier

transforms. This process manually increases the number of points which increases

resolution. One has to tradeoff the resolution when conducting the experiment. [Yoh01]

3.7.2 Dynamic Range and Spectral Occupancy

Dynamic range is defined as the ratio between the maximum to the minimum

input level which the system provides with reasonable signal quality. The input pulse of

the system compromises the dynamic range in the time domain. The power applied by

the input pulse is non-uniform or spread throughout the frequency range, which restricts

the amount of power applied over the frequency spectra of the input pulse. At

frequencies where the power of the input pulse is compromised, the system noise could

cause measurement difficulties. The level of power supplied by the input pulse generator

is also limited by the tolerances of the sampling head; therefore, introducing a high-

72


powered pulse to increase the dynamic range of the time domain measurement system is

limited. Since the frequency measurements are taken using discrete methods one

frequency at a time, the amount of power applied for each frequency point is fairly

uniform with respect to the system noise. Since uniform power can be applied at all

frequencies, the dynamic range of the frequency domain measurement system is much

greater than that of the time domain counterpart. [Yoh01]

One disadvantage of time domain characterization is that partial channel sounding

is achieved. Full channel sounding requires the use of an input signal that has high

spectral occupation. Spectral occupation efficiency is a measure of how similar the

spectrum of the pulse and the channel are [Vau99]. It could be written as,

ωωω

ωωωη

dgg

dggsoe

))()((

)()(222

21

21

+=

∫∫ (3.26)

where )(1 ωg is the channel spectral support region and )(2 ωg is the pulse power

spectrum. An obvious solution might be to decrease the rise time which results in more

details and more accurate to model. Unfortunately, this results in a more difficult

characterization [Fid90].

3.8 Conclusive Remarks

In this chapter, the measurement techniques of ultra-wideband characterization

are discussed. Two different kinds of measurements are carried out, namely: source

conducted and radiated measurements. The first type can be used to learn more about

sources in experimentation. We also included structures of three sets of UWB antennas

in hand along with their main operational characteristics. Additionally, radiated

measurements were carried out as to figure out more closely the effects of these antennas.

Based on the previous discussion and the availability of equipment, the proposed

measurement systems for material characterization and through-the-wall propagation are

based on both time-domain and frequency-domain setups. Indoor measurements are

73


based on time-domain measurements because of the current distance limitation on the

wideband frequency measurements.

74

Chapter 4. Through-the-wall Propagation and Material Characterization

75

the electrical parameters (conductivity and permittivity) also vary with frequency. This area has not received much attention to date, and publications that are available are form the field of microwave heating, so tend to deal with foodstuffs and human tissue rather than building materials ..

[Div91]

Chapter 4

Through-the-Wall Propagation and Material Characterization*

4.1 Introduction

The introduction of UWB communication promises an excellent indoor alternative due to

the expected through-the-wall propagation capabilities. The main reason is low signal attenuation

at low frequencies. However, to avoid interference with existing systems the bandwidth of

operation should be shifted to frequencies above 3 GHz. In this chapter, quantitative results

versus frequency are given for the delay and loss associated with the propagation through typical

walls encountered in indoor environments. Results of this research provide valuable insights into

the transient behavior of a pulse as it propagates through typical construction materials and

structures. Some research work has been performed at the statistical level with the aim of

characterizing the UWB communication channel. The results of these studies cannot be validated

or explained due to the lack of understanding of basic characteristics of pulse propagation in

typical UWB communication environments. A study of propagation through different materials

and scatterers would facilitate the development of a basic theory for pulse shaping, receiver

design, and channel modeling. Though there has been some research on pulse propagation in the

radar field and related electromagnetic aspects, this study has a unique communication flavor.

* [Muq03a],[Muq03b],[Muq03c],[Muq03d]


76

Generally speaking, material characterization is performed using different techniques,

including capacitor, resonator, and coaxial cavities methods, and radiated measurements as well

[Bak98], [Muq02b], [Muq02c]. This work concentrates on ultra wideband signal propagation

through different materials and structures and measures their characteristics as they are

encountered in the actual UWB communication applications. Radiated measurements allow for

non-destructive and broad-band applications [Zha01]. The importance of the measurements in

hand stems from the fact that the data available above 1 GHz are not adequate for UWB material

characterization. Some data at specific frequencies are available through studies of wireless

communication inside buildings. Many researchers have examined propagation through walls

and floors, but these data are often limited to specific frequency ranges and are also limited to

only few materials, thus not adequate for the proposed ultra wideband applications. Moreover,

with the inconsistency in published results, understanding and characterization of UWB

propagation through walls and in building environments become more compelling.

In this chapter, the effects of structures as well as materials on UWB propagation are

investigated. Some typical obstacles and materials encountered in the indoor wireless

propagation channel are studied. These include wooden doors, concrete blocks, reinforced

concrete pillars, glass, brick walls, dry walls, and wallboards. In addition to the time-domain

transient response, for some materials the following information is also presented: insertion

transfer function (impulse response), relative permittivity, loss tangent, attenuation coefficient,

and time delay. Measured data are provided for a frequency range of 1 to 15 GHz. Both

frequency-domain and time-domain measurement techniques are used to validate the results and

also capitalize on the advantages of each technique.

In section 2 the electromagnetic theory of wave propagation through a material slab is

reviewed. Then, the measurement procedure and the techniques used to relate the acquired data

to the electrical parameters of materials are discussed in sections 3 and 4. Comparison of results

obtained from various techniques for a sample material is presented in section 5. Section 6 is

devoted to a comprehensive discussion of the signal processing required to extract the

parameters. Detailed descriptions and dimensions for sample materials to be tested are given in

Section 7. Main results and observations are presented in Section 8. Additional measurement

issues such as distance between the antennas and the sample, repeatability, and variability are


77

also addressed in this section. Finally, section 10 includes some remarks on pulse shaping, UWB

receiver design, and modeling hints. Appendices provide details of some mathematical analyses

and additional results.

4.2 Propagation of Electromagnetic Waves in Dielectric Materials

In this section, propagation of electromagnetic waves through a lossy dielectric material

is reviewed and important parameters are defined. Assuming steady-state time-harmonic

electromagnetic fields, a TEM (transverse electromagnetic) plane wave propagating in the +z

direction can be represented using the phasor expression ztj eeEzE γωω −= 0),( , where fπω 2= is the

radian frequency (f is the frequency in Hz) and γ is the complex propagation constant given as

( ) ( ) ( )j jγ ω α ω β ω ω µε≡ + = . ( 4.1)

where α is the attenuation constant in Np/m, β denotes the phase constant in rad/m, andε and

µ are, respectively, permittivity and permeability of the material. For non-magnetic

materials 00 µµµµ ≅= r can be safely assumed.

The dielectric polarization loss may be accounted for by a complex permittivity

( ) ( ) ( )jε ω ε ω ε ω′ ′′≡ − , where 0εεε r=′ is the real permittivity with rε being the relative

permittivity constant (≥1). The imaginary part of the complex permittivity, ε ′′ , represents the

dielectric loss. The dielectric loss is also represented by a parameter referred to as loss tangent

and is defined as ( ) tan ( ) / ( )p ω δ ε ω ε ω′′ ′= ≡ . It should be noted form (4.1) that the attenuation

constant and the phase constant are both functions of the complex permittivity.

The conductivity loss can be modeled by an additional imaginary term in the complex

permittivity, ( )( ) j σε ω ε ε ω′ ′′= − + , where σ (ω) is the macroscopic conductivity of the

material of interest. The conductivity loss cannot be easily separated from the dielectric loss but

the two losses may be combined and represented by an effective loss tangent [Poz98],

( )epσε ε σωωε ε ωε

′′ + ′′= = +

′ ′ ′. ( 4.2)


78

A complex effective relative permittivity can now be defined as

[ ]( ) ( ) 1 ( )re r ejpε ω ε ω ω= − . ( 4.3)

To characterize any subsurface material, two parameters should be measured:

• the dielectric constant, ( )rε ω

• the effective loss tangent ( )ep ω , or a directly related parameter.

The complex propagation constant is then given by

( ) (1 )re r ej j jpc cω ωγ ω ε ε= = − , ( 4.4)

where 800 103/1 ×≅= εµc m/s is the speed of light in vacuum. For a TEM plane-wave

propagating inside the material, the attenuation coefficient and the phase constant can be

separated in the exponent

( ) ( ) ( )0 0( , ) z j z zE z E e E e eγ ω β ω α ωω − − −= = . ( 4.5)

The attenuation constant is given by

12

2( ) 1 12

rep

cεωα ω = + −

Np/m, ( 4.6)

while the phase constant becomes

12

2( ) 1 12

rep

cεωβ ω = + +

rad/m. ( 4.7)

A more widely used unit for the attenuation constant α is dB/m. The conversion to dB/m

is simply made using the relationship

α (dB/m)= α)log(20 e . (Np/m) = α686.8 (Np/m). ( 4.8)


79

At low frequencies, the loss due to ionic conductivity is dominant. But for the frequency

range of interest here, which is above 1 GHz, water dipolar relaxation becomes the significant

loss mechanism. It is pointed out that if the water content of the material is high then the

dielectric property is dominated more by the moisture content rather than by the material itself.

In general, practical subsurfaces can be considered as mixture of a variety of materials

contributing to the effective permittivity [Aur96]. In the next section, the procedures and

corresponding setups for measuring the complex dielectric constant are presented.

4.3 Measurement Procedures

Radiated transmission measurement is used because it allows one to find both the

attenuation constant and dispersion of the material under test. Moreover, it provides direct

insight into how a critical role through-the-wall propagation plays in UWB communications. The

measurement can be performed in time domain or in frequency domain. In the time-domain

approach, an electromagnetic pulse, Ei(t, z), is applied to a homogenous, isotropic material layer

of thickness d. The incident pulse gives rise to a reflected pulse, Er(t,z), and a transmitted pulse,

Et(t,z). The diagram of the experiment is illustrated in Figure 4.1. The transmission scattering

parameter is then related to the incident and transmitted signals by,

)(

)()(21 tvFFTtvFFTjS

i

t=ω , ( 4.9)

where vt is the voltage at the output terminals of the receive antenna and is proportional to Et ,

while vi is the voltage at the input terminals of the transmit antenna and is proportional to Ei and

fast Fourier transform (FFT) is used to convert the sampled signal to the frequency domain data.


80

Figure 4.1. Incident, transmitted, and reflected waveforms observed in time-domain measurements.

If the material slab is symmetric, then

)()( 2112 ωω jSjS = . ( 4.10)

Instead of measuring the transmitted and received voltage signals, it is more convenient

to measure the following two signals on the receive side:

• a transmit through signal, vt(t), which is received with the material layer in place, and • a free-space reference signal, ( )fs

tv t ,which is the received signal without the layer.

Therefore, two measurements as illustrated in Figure 4.2 should be carried out with

exactly the same distances and antenna setup. The free-space measurement is used as a reference

to account for all the effects that are not due to the material under test; for example, the antennas,

the receiver, and the signal generator. Assuming a fictitious layer of free-space of the same

thickness as the material slab, the propagation through this layer involves simply a delay equals

cd≡0τ , where d is the layer thickness and c is the speed of light in free space. In other words,

Z0=50 ohmS11 S22 S21

S12

Reference plane 1 Reference plane 2

t

vt(t) vi(t)

t

vr(t)

t


81

Figure 4.2. Two required measurements, without layer (free space) and with layer in place

0( )( )

fsjt

i

E j eE j

ωτωω

−= . ( 4.11)

The insertion transfer function is defined as the ratio of two radiation transfer functions,

)()(

))(())((

)()(

)()(

)()(

)(ωω

ωω

ωω

ωω

ωjV

jVtvFFTtvFFT

jEjE

jEjE

jEjE

jH fst

tfs

t

tfs

t

t

i

fst

i

t

===≡ . ( 4.12)

Combining (4.9), (4.11), and (4.12), the scattering parameter 21S can be related to the insertion

transfer function as

021( ) ( ) jS j H j e ωτω ω −= . ( 4.13)

Region I

(air)

Region II

(Material)

0 z

d

Region III

(air)

Ei fs

tE

Er

Ei Et


82

In the frequency-domain method, the pulse signals are replaced with sinusoidal signals

and a vector network analyzer is used to monitor the received waveforms. Otherwise, the

measurement procedure is the same as that for the time-domain approach illustrated in Figure

4.2. Thus, in summary, first the time-domain signal ( )tfsv t is measured with a sampling

oscilloscope or the frequency-domain signal ( )tfsV jω is measured with a network analyzer in the

absence of the material layer. Then, the time-domain signal ( )tv t or the frequency-domain signal

Vt(jω) is measured with the material layer in place. The insertion transfer function is then

calculated using (4.12). Care must be taken to ensure that the conditions for the free-space

measurement are as closely identical as possible to those for the measurement through the

material slab. Once the insertion transfer function (or 21S ) is obtained, numerical methods are

used to extract the attenuation coefficient and dielectric constant of the material as detailed in the

next section.

Using the time-domain waveforms, the delay between the two pulses can be measured to

obtain an approximate value for the dielectric constant. The total signal power can also be

measured in the free-space case and through the material to estimate the power loss due to

propagation through the material. In the following sections, the procedure and signal processing

required to extract the material parameters are discussed.

4.4 Analysis Techniques

The free-space and through-the-wall measurements would be most accurate if performed

inside an anechoic chamber where all the multipath components and reflections from the floor

and ceiling are absorbed. Ideally, the sample to be measured should be infinitely large to avoid

scattering from the edges. Samples under test have to be at a far-field distance from the antenna,

typically several meters for the frequency range of interest and the antenna dimensions.

Maintaining these requirements is not a convenient task, keeping in mind that absorbers and

chamber environment do not allow easy movements of large samples. Fortunately, time gating

can be used to reduce significantly the undesired effects such as reflections from the surrounding

walls and scattering from edges. For time gating to be efficient three conditions have to be met.

First, the transmitter and the receiver antennas should be positioned away from the reflecting


83

surfaces. Second, samples should have relatively large surface dimensions in order to minimize

the edge effects. Finally, there should be flexibility in adjusting the distance between the antenna

and the sample. Time gating can also be used to isolate a desired portion of the received signal;

namely, the first single-pass of the signal transmitted through the slab. In this application, the

sample thickness should be large enough to yield sufficient delay. Thus, not only zooming on

and extracting the first pulse, but also removing all delayed pulses due to multiple reflections

inside the slab become possible. In the following subsections analysis techniques based on

single-pass, multiple-pass, and approximate solutions for low loss materials are presented.

4.4.1 Single-Pass Technique

Single-pass technique can be used if the duration of the test pulse is sufficiently short or

the wall or material slab under study has a thickness that is large enough to allow gating out the

portions of the signal due to multiple reflections inside the slab.

A short-duration electromagnetic pulse, Ei(t), is applied to a homogenous and isotropic

material layer of thickness d. The transmitted signal, Et(t), results in a voltage at the receiver

antenna terminals. To simplify the problem we assume that the wave is normally incident on the

material surface and the duration of the pulse is smaller than the pulse travel time through the

material. Then, multiple reflections inside the layer, which are delayed more than the pulse

width, can be eliminated by means of time gating. The same technique can be used to eliminate

antenna ringing and extraneous paths signal components. To sum up, this is a single-pass

duration-limited transient measurement procedure based on a one-dimensional model of plane-

wave propagation through a planar layer.

The derivations pertaining to the short-pulse propagation measurements are available in

[Aur96] and represented in appendix A1. The results are summarized below.

2 2

0 0

( )( ) 1( ) 1 12

spr

d fffdf

τετ πτ

Φ ∆≅ + = −

, ( 4.14)

2

0

1 ( )1( ) ln ( )( ) 4 ( )

re sp

r r

fp f H f

f f f

ε

π τ ε ε

+ ≅ −

, ( 4.15)


84

where ( ) | ( ) | exp[ ( )]sp sp spH f H f j f= Φ is the single-pass insertion transfer function. It is the

ratio of the Fourier transform of the single-pass received signal when the slab is in place to the

Fourier transform of the single-pass received signal in the absence of slab. It should be noted that

the derivative term ( ) /spd f dfΦ in (4.14) is based on the assumption that the phase varies

linearly with frequency. The advantage of using the derivative of the phase is to avoid tracking

the unwrapped phase function. As a function of the unwrapped phase, the dielectric constant is

given by

2 2

0 0

( )( )( ) 1 12

spr

ffff

τετ πτ

Φ ∆≅ + = −

. ( 4.16)

4.4.2 Multiple-Pass Technique

If single-pass signal cannot be gated out satisfactorily, multiple reflections from the slab

interior that constitute part of the received signal must be accounted for. This situation

particularly arises when the transit time through the thickness of the slab is small compared to

the pulse duration. In this case, an insertion transfer function that accounts for multiple

reflections is needed. This insertion transfer function, denoted as H(jω), has a definition similar

to spH (jω) except that single-pass signals should be replaced with signals containing all multiple-

pass components in the Fourier transform calculations. Thus, time-domain measured data can be

used to find H(jω). However, H(jω) can be conveniently calculated using a frequency-domain

technique. In this method, measured data are obtained using a sweep generator and a vector

network analyzer. For frequency-domain measurements too, ideally the slab has to be infinitely

large and should be measured in an anechoic chamber in order to avoid capturing the scattering

from the edges of slab and reflections from the floor, ceiling, and adjacent walls. However, if the

undesired scattering and reflection signals can be removed through time gating mechanisms, as

explained later, the signal received in a relatively large time window provides sufficiently

accurate results.

In order to obtain an expression for the insertion transfer function H(jω), let us assume

that an x-polarized uniform plane-wave, representing the local far-field of a transmit antenna is

normally incident on a slab of material of thickness d. The material has an unknown complex


85

dielectric constant rrr jεεε ′′−′= . The incident plane-wave, as depicted in Figure 4.2, establishes

a reflected wave in region I (air), a set of forward and backward traveling waves in region II

(material), and a transmitted wave in region III (air). The electric and the magnetic fields in

region I can be written as

)( 00001

zjr

zjix eEeEaE ββ +− +=

r, ( 4.17)

)(1 0000

11

zjr

zjiy eEeEaH ββ

η+− −=

r, ( 4.18)

where cfπ

λπεµωβ 22

000 === and Ω== πεµη 120

0

01 .

In region II, the fields are given by

)( 222zz

x eEeEaE γγ +−−+ +=r

, ( 4.19)

2 2 22

1 ( )z zyH a E e E eγ γ

η+ − − += −

r, ( 4.20)

where 0 0 ( )r rj j jγ α β ω µ ε ε ε′ ′′= + = − and 20 ( )r rj

µηε ε ε

=′ ′′−

. Similarly, in region III the

fields are expressed as, zj

x eEaE 033 β−+=

r, ( 4.21)

zjy eEaH 0

31

31 β

η−+=

r. ( 4.22)

Boundary conditions require continuity of the tangential components of the Er

and Hr

fields at z=0 and z=d. These conditions are summarized as

−+ +=+⇒= 220021 )0()0( EEEEEE ri , ( 4.23)

)()0()0( 222

10021

−+ −=−⇒= EEEEHH ri ηη

, ( 4.24)

djdd eEeEeEdEdE 032232 )()( βγγ −++−−+ =+⇒= , ( 4.25)


86

djdd eEeEeEdHdH 03

1

22232 )()( βγγ

ηη −++−−+ =−⇒= . ( 4.26)

We need to find T=0

30

i

dj

EeE β−+

which is equivalent to S21 in the scattering parameters terminology.

Manipulating the boundary conditions, we obtain

(4.23)+(4.24) ! )1()1(22

12

2

120 η

ηηη −++= −+ EEEi , ( 4.27)

(4.25)+(4.26) ! )1(21

232

0

ηηβγ += −+−+ djd eEeE , ( 4.28)

(4.25) - (4.26) ! )1(21

232

0

ηηβγ −= −++− djd eEeE . ( 4.29)

Substituting for +2E from (4.28) and for −

2E from (4.29) into (4.27) yields,

)1)(1(21)1)(1(

212

2

1

1

2)(3

2

1

1

2)(30

00

ηη

ηη

ηη

ηη βγβγ −−+++= −−+−++ djdj

i eEeEE . ( 4.30)

The transmission coefficient is now readily obtained as

−−+

++

==−

−+

1

2

2

1

1

2

2

10

3

22

40

ηη

ηη

ηη

ηη γγ

β

ddi

dj

eeEeET ( 4.31)

Based on the definition of the insertion transfer function given in (4.12), we can write

)(0

0ωβ

β jHTee

TEEEE dj

dji

fst

it === − . ( 4.32)

Thus,

)2()2(

4)(

1

2

2

1

1

2

2

1

0

ηη

ηη

ηη

ηηω

γγ

β

−−+++=

− dd

dj

ee

ejH . ( 4.33)

4.4.2.1 Exact Solution

When the complex insertion transfer function H(jω) is determined by measurements as

described in Section 4.3, equation (4.33) can be solved for the complex dielectric constant


87

rrr jεεε ′′−′= . In terms of the scattering parameter 21S , (4.33) with the help of (4.13) can be

easily cast into the following form [Alq96],

02)cosh(2)sinh(1

21

=−+

+

SxPxP

xx , ( 4.34)

where rx ε= and djP 0β= . An alternative derivation of (4.34) based on bounce diagram is

presented in Appendix A2. This equation can be solved numerically using two-dimensional

search algorithms. The convergence of this algorithm is not always guaranteed taking into

account possible multiple solutions and noise in the measurements. In the next section, using

reasonable assumptions, (4.33) is reduced to a one-dimensional problem involving real equations

only.

4.4.2.2 Approximate Solution *

When the material occupying region II is low loss, rε ′′ / 1rε ′ << and the following

approximations can be used,

)211(

)211()(

0

0000

r

rr

r

rrrr

jj

jjjjj

εεεβ

εεεεµωεεεµωβαγ

′′′

−′=

′′′

−≅′′−′=+=

and

rrrr j ε

ηεεµ

εεεµη

′=

′≅

′′−′= 1

0

0

02 )(

.

Then, r

r

rr ε

εε

εηη

ηη

′+′

=′

+′≅+11

1

2

2

1 and (4.33) reduces to

)12()12(

4)()()(

0

r

rdj

r

rdj

dj

ee

ejH

εε

εεω

βαβα

β

′+′−+

′+′+

=+−+

. ( 4.35)

Rewriting the insertion transfer function in terms of magnitude and phase, we obtain

* This formulation is submitted for publication, refer to [Muq03a], [Muq03b] by Ali Muqaibel and A. Safaai-Jazi.


88

21

22

2

2

2 14)2cos(21212

16)(

′+′−+

′+′−+

′+′+

=−

r

r

r

rd

r

rd dee

jH

εεβ

εε

εε

ωαα

( 4.36)

and

φβω −=∠ djH 0)( ( 4.37)

where

⋅

′+′−+

′+′+

′+′

−−

′+′

+

=−

−

− )tan(1212

12

12

tan 1 d

ee

ee

r

rd

r

rd

r

rd

r

rd

β

εε

εε

εε

εε

φαα

αα

. ( 4.38)

Equation (4.38) can be written in a more compact form as

⋅

+−= −

−− )tan(

11tan 2

21 d

QeQe

d

d

βφ α

α

( 4.39)

where

2

2

2

11

)1()1(

1212

12

12

+′−′

−=+′−′

−=+′+′−′−′

=

′+′+

′+′−

=r

r

r

r

rr

rr

r

r

r

r

Qεε

εε

εεεε

εεε

ε

. ( 4.40)

For most applications of interest Q has a small value. For example, for the relative

permittivities of 2.0, 4.0, and 8.0, Q is about 0.02, 0.1, and 0.3, respectively. Later we will use

this fact to further simplify the solution. For the time being no assumption is made about Q.

Letting Xe d =− α2 , then

( ) ( ) ( )244

2

1)2cos(211116)(

−′−−′++′

′=

rrr

r

dXX

jHεβεε

εω , ( 4.41)

or


89

( ) ( ) ( ) 01)(

81)2cos(214

2242 =+′+

′+−′−−′ r

rrr X

jHdX ε

ωεεβε ,

which is a quadratic equation in terms of X. Solving this equation or X, we have

( ) ( ) ( )

( )44

2

22

22

2

1

1)(

81)2cos()(

81)2cos(

−′

−′−

′+−′±

′+−′

== −

r

rr

rr

r

djH

djH

d

eXε

εω

εεβω

εεβα

( 4.42)

Only the solution with negative sign in (4.42) is valid (the proof is given in Appendix A3.

Substituting for X from (4.42) in the phase expression (4.39), we obtain the following equation,

which is only in terms of rε ′ .

[ ] 0)tan(11)(tan 0 =+−+∠− d

QXQXjHd βωβ ( 4.43)

Solving this equation numerically, rε ′ is readily determined. Then, X and subsequentlyα are

found from (4.42). Finally, rε ′′ is calculated using

ωεα

ε rr

c ′=′′

2. ( 4.44)

Special Case

If it can be further assumed that 12 <<− de α , then (4.39) becomes

dd ββφ =≈ − ))(tan(tan 1 ,

and

rddddjH εββββω ′−=−=∠ 000)( , where rεββ ′≈ 0 and 2

0

2

0

0 )(1)(

∠−=

∠−=′djH

djHd

r βω

βωβε . ( 4.45)

Once rε ′ is determined, α and then rε ′′ can be found from 2)( ωjH using the following

relationships,


90

′+′

−+

′+′

+

≈22

2

2

14)2cos(2

12

16)(

r

r

r

rd de

jH

εεβ

εε

ωα

( 4.46)

′+′

′−′+

=

r

r

r

rdjH

dε

ε

εεβ

ωα

4

2

2

)1(

)1()cos(2)(

16

ln21

. ( 4.47)

This simplified analysis reduces to the single pass case as in [Aur96], where the wall is assumed

to be thick and single transmitted pulse can be time gated. This is because the assumption

12 <<− de α has the implication that the multiple-pass components of the received signal are very

small, as for αd>>1 these components are attenuated significantly more than the single-pass

signal.

4.5 Comparison of Various Techniques

In Section 4.4.1 two sets of expressions for the calculation of dielectric constant and loss

tangent, based on single-pass insertion transfer function ( )spH f , were presented. These are

equations (4.14) and (4.15) for single-pass involving phase derivative, and equations (4.16) and

(4.15) for single-pass involving the phase itself. Similarly, in Section 4.4.2 two sets of

expressions for the calculation of dielectric constant and loss tangent or attenuation coefficient,

based on multiple-pass insertion transfer function ( )H jω , were presented. These are equations

(4.34) or (4.35) for exact solutions (material need not be assumed low loss), and equations

(4.42), (4.43), and (4.44) for approximate solutions applicable to low loss materials. Here, the

results obtained from these solutions are calculated and compared in order to better appreciate

the accuracy as well as the applicability of each method. Measurements are carried out for a

sample wooden door representing the slab. The results for the dielectric constant obtained from

the four sets of solutions mentioned above are shown in Figure 4.3. It is noted that the results


91

from other three solutions are in excellent agreement. This agreement is due to the fact that for

this specific sample (wooden door), multiple reflections inside the door are very small compared

to the first single-pass. It is further noted that the results obtained from the exact and

approximate solutions (multiple-pass) are nearly identical, indicating that the door material is

low loss. The fact that the search for a complex solution problem can be reduced to a one-

dimensional problem is illustrated in Figure 4.4. This figure illustrates that the complex search

problem is separable, as any cut on a constant rε ′′ plane results in the same minimum. For better

visualization, a vertical cut at rε ′′ =0.14 is also shown.

Both the exact complex and approximate real equations have spurious solutions that can

be avoided by starting with an initial guess obtained from the single-pass solution at a high

frequency and by using a constrained search. At high frequencies the wavelength is smaller and

the assumption of thick slab become more reasonable. The solution obtained at a high frequency

point is then used as an initial guess for the next frequency point, because variations of the

dielectric constant versus frequency are slow over a narrow frequency range.

Whenever single-pass time gating is possible, the single-pass analysis technique can be

used. Using time-domain measurements, this technique is applicable if one of the following two

requirements is met: (i) the pulse has a width shorter than the transit time through the slab, (ii)

the material has a sufficiently high loss so that the second and higher-order reflections are

attenuated much more than the first single-pass signal. If single-pass time gating is not possible,

the multiple-pass analysis technique should be used. First the approximate solution is attempted,

but the result has to be validated by comparing with the exact solution to see if the low loss

requirement is met. Whenever possible, the results from both time-domain and frequency-

domain measurements should be obtained and compared to ensure the validity of measurements

and also avoid any spurious results.


2 3 4 5 6 7 8 9 101.9

1.95

2

2.05

2.1

2.15

2.2

2.25

2.3

Frequency GHz

Die

lect

ric C

onst

ant

Exact, Two Dimensional SearchNew Formulation, One Dimensional SearchSingle-Pass

Figure 4.3. Comparison between the different measurement and analysis techniques

Figure 4.4. Two-dimensional search example, illustrating the possibione-dimensional search

rε ′′rε ′

Mag

nitu

de

02468100

1

2

3

4

5

6

'

Mag

nitu

de

rε ′′ =0.14

92

lity of reducing it to a

rε ′


93

In order to assess the accuracy of the approximate method presented above, we take a

data point from the measured insertion transfer function, H(jω), for the sample wooden door at a

frequency of 5.00 GHz and artificially increase the loss by decreasing the magnitude of H(jω).

This is achieved by multiplying H(jω) by a constant a as given in the first row of Table 4.1. A

zero-loss case is also included in the table by choosing a such that a|H(jω)|=1. As noted, the

errors are less than 1% for loss tangents about 0.2, a representative upper-limit loss factor for

most dielectric materials of practical interest. Only for very high loss cases (tan δ>0.9) the errors

become significant, about 15% for this example. As expected, for the zero-loss case

(corresponding to a=1.23 in this example) the approximate solution for the dielectric constant

becomes exact. Although this error analysis is for a specific example, it provides a realistic

measure of the accuracy of the approximate method presented here.

Table 4.1. Errors in dielectric constant and loss tangent obtained form the approximate formulation. (The data point used in this analysis is H(jω)= -0.3317 - j0.7418).

a 1.23 1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1

rε ′ 2.178 2.176 2.174 2.171 2.166 2.159 2.147 2.128 2.095 2.033 1.885

Exac

t

tan δ 0 0.063 0.096 0.133 0.176 0.226 0.286 0.363 0.466 0.624 0.940

rε ′ 2.178 2.176 2.176 2.175 2.175 2.174 2.174 2.174 2.173 2.173 2.173

%Error 0 0.036 0.972 0.2104 0.406 0.730 1.265 2.162 3.747 6.895 15.312

tan δ 0 0.063 0.095 0.132 0.174 0.223 0.281 0.352 0.445 0.575 0.799 App

rox.

% Error 0 -0.438 -0.485 -0.609 -0.845 -1.243 -1.883 -2.914 -4.631 -7.772 -15.02


94

4.6 Signal Processing and Parameters Extraction

In this section, the procedure for processing the measured data and extracting the material

parameters is presented. The procedure is summarized in the flowchart shown in Figure 4.5. As

mentioned earlier, two measurements are performed; the first is the free-space reference in the

absence of the material, while the second is the through measurement that is performed with

the material in place. Measurements are carried out in both the frequency and time domains.

Transformations between the two domains are possible by means of Fourier or inverse Fourier

transform.

4.6.1 Data Acquisition

Frequency-domain measurements can be made in a frequency range from 45 MHz to 15

GHz. Over a given frequency range 801 complex data points (magnitude and phase) can be

collected which is the maximum number of points allowed by the network analyzer. This

limitation on the number of points imposes a limit on measurement resolution and the accuracy

of transformation to time domain using inverse Fourier transform. To obtain the un-gated

insertion transfer function, a band-pass finite-impulse-response (FIR) filter is used to remove the

noise at low frequencies and beyond the antenna bandwidth. The cutoff frequencies of the filter

are adjusted to remove the noise regions. The order of the filter is chosen to be 100. A sample

frequency measurement is shown in Figure 4.6 which illustrates the magnitude and the

unwrapped phase of the free-space and through signals, the FIR filter characteristic, and the

obtained impulse responses. Zeros are padded to get higher resolution in the transformed time

domain.

For time-domain measurements, 128 traces are averaged and acquired in a 5 ps sampling

time using the sampling oscilloscope. Offset adjustment is achieved through load calibration and

post-processing. The signal acquiring window spans more than 10ns and consists of 2048 points.

An illustrative measurement is shown in Figure 4.7a.


95

Figure 4.5. Through the wall and material characterization procedure flowchart

Time Domain Measurements (128 averaged)

Frequency Domain Measurements

FIR filter, ungated insertion transfer function (optional)

ifft with proper zero padding

FFT with proper zero padding

Determine the time domain peak-to-peak

differential delay using sliding correlator.

Time-gating to remove undesired reflections

First estimate of

delay and loss

Insertion Transfer Function, or transmission coefficient (complex)

Use approximate equations or search algorithms to extract

unknowns

rε ′ , and rε ′′ ,

0 1 2 3 4 5 6 7 8 9 10-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

0.12

Time ns

Am

plitu

de V

Free-SpaceThrough

0 1 2 3 4 5 6 7 8 9 10-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

0.12

Time ns

Am

plitu

de V

Free-SpaceThrough

0 5 10 15-90

-80

-70

-60

-50

-40

-30

-20

Frequency GHz

Ampl

itude

(dB)

Free-SpaceThrough

0 5 10 15-90

-80

-70

-60

-50

-40

-30

-20

Frequency GHz

Ampl

itude

(dB)

Free-SpaceThrough

0 5 10 15-80

-70

-60

-50

-40

-30

-20

-10

0

10

20

Frequency GHz

Ampl

itude

(dB)

Insertion Transfer FunctionFilterFilter Insertion Transfer Function

0 5 10 15-80

-70

-60

-50

-40

-30

-20

-10

0

10

20

Frequency GHz

Ampl

itude

(dB)


0 5 10 15-1200

-1000

-800

-600

-400

-200

0

200

Frequency GHz

Ang

le (r

ad)

Free-SpaceThrough

0 5 10 15-1200

-1000

-800

-600

-400

-200

0

200

Frequency GHz

Ang

le (r

ad)

Free-SpaceThrough

0 0.5 1 1.5 2 2.5 3-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

time ns

Am

plitu

de

Source 1 WindowSource 2 WindowWindow for NA

0 0.5 1 1.5 2 2.5 3-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

time ns

Am

plitu

de


0 0.5 1 1.5 2 2.5 3-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

time ns

Am

plitu

de


0 10 20 30 40 50 60-8

-6

-4

-2

0

2

4

6

8

10

12x 10-5

time ns

Ampl

itude

Free-Space Impulse ResopnseThrough Impulse Response

0 10 20 30 40 50 60-8

-6

-4

-2

0

2

4

6

8

10

12x 10-5

time ns

Ampl

itude


0 500 1000 1500 2000 2500 3000 3500 4000 4500-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

Sample n

Ampl

itude

0 500 1000 1500 2000 2500 3000 3500 4000 4500-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

Sample n

Ampl

itude

0 5 10 151

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

3

Frequency GHz

Die

lect

ric C

onst

ant

Source 1 Antenna 1Source 2 Antenna 1Source 1 Antenna 2Source 2 Antenna 2Network Analyzer Anteena 1Network Analyzer Anteena 2

0 5 10 151

1.2

1.4

1.6

1.8

2

2.2

2.4

2.6

2.8

3

Frequency GHz

Die

lect

ric C

onst

ant


0 5 10 15Frequency GHz


0 5 10 15Frequency GHz


Attenuation Constant Dielectric Constant & Loss Tangent

0 1 2 3 4 5 6-0.06

-0.04

-0.02

0

0.02

0.04

0.06

Time (ns)

Am

plitu

de (V

)

Free-spaceThrough0.5*Free-space filter0.5*Through filter

0 1 2 3 4 5 6-0.06

-0.04

-0.02

0

0.02

0.04

0.06

Time (ns)

Am

plitu

de (V

)


0 5 10 15-6

-5

-4

-3

-2

-1

0

Frequency (GHz)

Inse

rtion

Los

s (d

B)

FitS1A1S2A1S1A2S2A2NA,A1NA,A2

0 5 10 15-6

-5

-4

-3

-2

-1

0

Frequency (GHz)

Inse

rtion

Los

s (d

B)


0 5 10 150

20

40

60

80

100

120

140

Frequency (GHz)

Atte

nuat

ion

Con

stan

t (dB

/m)


0 5 10 150

20

40

60

80

100

120

140

Frequency (GHz)

Atte

nuat

ion

Con

stan

t (dB

/m)


96

(a)0 5 10 15

-90

-80

-70

-60

-50

-40

-30

-20

Frequency GHz

Am

plitu

de (d

B)

Free-SpaceThrough

(b)0 5 10 15

-1200

-1000

-800

-600

-400

-200

0

200

Frequency GHz

Ang

le (r

ad)

Free-SpaceThrough

(c)0 5 10 15

-80

-70

-60

-50

-40

-30

-20

-10

0

10

20

Frequency GHz

Am

plitu

de (d

B)


(d)0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

-8

-6

-4

-2

0

2

4

6

8

10

12x 10

-5

time ns

Am

plitu

de

Free-Space Impulse ResopnseThrough Impulse Response1E-4*Free-Space Filter1E-4*Through-Filter

Figure 4.6. Illustration of the frequency domain measurements, (a) Measured magnitude, (b) Measured phase, (c) Filter and filtered un-gated insertion transfer function, and (d) Impulse response and weighted gating window

Mag

nitu

de (d

B)


97

0 1 2 3 4 5 6-0.06

-0.04

-0.02

0

0.02

0.04

0.06

Time (ns)

Am

plitu

de (V

)


(a)

0 500 1000 1500 2000 2500 3000 3500 4000 4500-0.2

-0.1

0

0.1

0.2

0.3

0.4

0.5

Sample n

Am

plitu

de

(b)

Figure 4.7. Illustration of the Time domain measurements, (a) Measured through and free space signals (b) Illustration of the correlation function to estimate the time delay.


98

4.6.2 Time Delay and Initial Guess for Permittivity

The through and free-space time-domain measured signals or the corresponding

impulse responses obtained from frequency-domain measurements are correlated using a sliding

correlator to obtain the first guess on the delay and effective dielectric constant. The shape of the

correlator output is illustrated in Figure 4.7b. An estimate of the average dielectric constant could

also be obtained through peak-to-peak impulse time delay, τ∆ . An average value of the

dielectric constant that does not reflect the frequency dependence is given by

2

1

∆+≅′cdrτε , ( 4.48)

where d is the thickness of the slab and c is the speed of light in free-space.

4.6.3 Time Gating

Time gating is required to remove multi-pass components in received signals, as they are

not accounted for in the calculation and extraction of material parameters. Multiple reflections

should be gated out too if single-pass technique is used. In multiple-pass technique, perfect time

gating cannot be achieved because, strictly speaking, infinite acquisition time is required to

capture infinite number of multiple reflections. However, because higher-order multiple

reflections die out quickly for materials of interest in this research, satisfactory time gating is still

achievable.

Time gating capabilities are enhanced with shorter pulse durations and longer distances

between the test material and reflectors and scatterers. If single-pass is desired, pulse duration

should be shorter than the twice the travel time through the slab to avoid pulse overlapping.

To avoid abrupt changes on the signal level, the gating-window should have smooth

transition from zero to the flat level. This window is based on the modified Kaiser window with

a flat region in the middle. However, the results for material parameters should be essentially

independent of the used window. Various parameters of the window are changed to make sure


99

that the results are not sensitive to the details of the window. A Kaiser window of length M has a

time domain sequence h(n) given by [Pro96],

2 2

0 0

0 0

1 12 2

12

M MI n

MI

α

α

− − − −

−

, 10 −≤≤ Mn , ( 4.49)

where I0 is the modified Bessel function of order zero and 0α is a design smoothing factor set

equal to 25. The window size was chosen to be 2 ns for source#1, 0.5 ns for source#2, and 3ns

for the Fourier-transformed data measured with the network analyzer. These values were chosen

to allow for nearly optimum time gating. The windows have two symmetrical transition regions

and a flat region defined by the intervals (0.2,1.6,0.2), (0.1,0.3,0.1), and (1,1,1) ns, parameters

within parentheses refer to (risetime, width of flat region, fall time), respectively as illustrated

in Figure 4.8.

Figure 4.8. Three different time domain gating windows

If the through and free-space signals both return to the zero level in the window, the

gating can be implemented easily. If the signal does not become exactly zero in the window, the

window opening for the received signal is delayed by an amount equal to 0τ . After time gating

0 0.5 1 1.5 2 2.5 3-0.2

0

0.2

0.4

0.6

0.8

1

1.2

1.4

time ns

Am

plitu

de



100

the signals with proper zero padding, the fast Fourier transform (FFT) and (4.12) are used to

calculate the insertion transfer function.

4.6.4 Propagation and Material Parameters

Form the complex insertion transfer function, the dielectric constant and loss tangent of

the material under test can be extracted. Table 4.2 summarizes the analysis techniques and the

corresponding equations required to extract the material parameters. The choice of the analysis

technique is based on how satisfactorily time gating can be implemented. In many cases,

multiple reflections inside the slab decay rapidly so that single-pass or multiple-pass techniques

essentially yield the same results.

Six different measurements are performed for the characterization of each material. These

include four time-domain measurements, with two different pairs of antennas and two pulse

generators, and two frequency-domain measurements using two pairs of antennas. The results for

six different time-gated measurements for a sample door are given in Figure 4.9.

Table 4.2 Summary of analysis techniques and required equations

Analysis Technique Equations

Single-Pass (low -loss) (4.15), (4.16)

Multiple-pass (exact solution) (4.34) or (4.35)

Multiple-pass (approximate solution, low-loss) (4.42), (4.43), (4.44)

The other parameters for the sample door are given in Figure 4.10. The confidence on the

obtained result is strengthened when the time-domain and frequency-domain measurements

agree, because different equipments, calibrations, and processes are involved. There are two

advantages for the frequency-domain method, one is that characterization over a higher

frequency range can be achieved by using multi-band filters, and another advantage is that no

external synchronization is required. Both input and output are centrally processed, thus reducing

the synchronization errors. On the other hand, the time-domain method offers higher resolution

and more data points. Examining Figure 4.10 reveals that the time-domain measurements

performed with short pulses (source #2) exhibit less variation than the frequency-domain results

across the frequency band of interest.


101

Antenna 1 Antenna 2

Sour

ce 1

0 0.5 1 1.5 2 2.5 3-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

0.12

time ns

Am

plitu

de V

Free-Space ReferenceThrough

0 0.5 1 1.5 2 2.5 3

0

0.05

0.1

0.15

0.2

0.25

time ns

Am

plitu

de V


Sour

ce 2

0 0.5 1 1.5 2 2.5 3-0.06

-0.04

-0.02

0

0.02

0.04

0.06

time ns

Am

plitu

de V


0 0.5 1 1.5 2 2.5 3

-0.1

-0.05

0

0.05

0.1

time ns

Ampl

itude

VFree-Space ReferenceThrough

Net

wor

k A

naly

zer (

ifft)

0 0.5 1 1.5 2 2.5 3

-5

0

5

10

15

x 10-5

time ns

Am

plitu

de


0 0.5 1 1.5 2 2.5 3

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

2.5

3x 10-4

time ns

Ampl

itude


Figure 4.9. Time domain representation of the six different measurements for the sample door


102

(a) 0 5 10 15

-6

-5

-4

-3

-2

-1

0

Frequency (GHz)

Inse

rtion

Los

s (d

B)


(b) 0 5 10 15

0

20

40

60

80

100

120

140

Frequency (GHz)

Atte

nuat

ion

Con

stan

t (dB

/m)


(c) 0 5 10 15

1.5

1.6

1.7

1.8

1.9

2

2.1

2.2

2.3

2.4

2.5

Frequency (GHz)

Die

lect

ric C

onst

ant


(d) 0 5 10 15

0

0.05

0.1

Frequency (GHz)

Loss

Tan

gent


Figure 4.10. Comparison for the sample door parameters extracted using different measurement techniques, (a) time-gated insertion transfer function , magnitude, (b) attenuation constant, (c) dielectric constant, and (d) loss tangent.


103

4.7 Description of Samples and Wall Materials

Ten different wall materials commonly encountered in building environments are

selected for UWB characterization. These include drywall, glass, wallboard, styrofoam, cloth

office partition, wooden sample door, wood, structure wood, brick, concrete block, and

reinforced concrete column. Table 4.3 lists the selected samples and their dimensions. Thickness

measurements are taken as an average of 6 to 8 repeated measurements for best accuracy.

Table 4.3. Sample building materials, dimensions, and parameters at 5 GHz

Insertion Loss (dB),

af+b, f(GHz) Material Dimensions (cm) εεεεr

Loss at 5 GHz (dB)

Frequency Range (GHz) a b

Wallboard 1.16992 × 121.8 × 196.9 2.44 0.45 0.6-13.9 0.0132 0.3940

Cloth Partition 5.9309 × 140.7 × 153.1 1.23 2.55 0.5-13.8 0.6601 -0.7418

Structure Wood 2.06781 × 121.5 × 197.8 2.11 1.35 0.8-14.1 0.1138i 0.8656

Wooden Door 4.44754 × 90.70 × 211.8 2.08 2.0 1.0-14.3 0.3777 0.1258

Plywood 1.52146 × 121.9 × 197.51 2.49 1.75 2.0-14.6 0.1769 0.9198

Glass 0.235661 × 1.44 × 111.76 6.40 1.25 1.0-14.3 0.2895 0.1005

Styrofoam 9.90702 × 121.8 × 197.7 1.11 0.02 0.5-13.8 ≈0 0.0418

Bricks 8.71474 × 5.82676 × 19.8 4.22 6.45 1.0-7.0 1.0702 0.9757

Concrete Block 19.45 × 39.7 × 19.5 2.22 13.62 2.0-6.8 0 13.62

Reinforced Concrete (TDL) 60.96 × 121.92 × - - - - -

Note: The first number in the dimensions column is the thickness of the sample; i.e. the propagation path length through the material.

The requirement that free-space and through measurements should be performed with

the same antenna spacings, makes in-situ measurements very difficult. After the in-situ through

measurements are performed, the free-space measurements should be made at a different

location but with the same distance between the transmiting and receiving antennas as in the

through measurement setup. Since it is impossible to have exactly the same distances between

the antennas for measurement setups at two different locations, errors will inevitably result in the


104

calculation of insertion transfer function. For example, at 10 GHz the wavelength is about 3cm,

then1mm change in the spacing between the two antennas results in 12 degrees phase error. This

is an extremely tight tolerance requirement that cannot be met easily. To overcome this problem,

a moving platform was constructed, and bricks and blocks were used to build walls on it. This

allows us to move the wall between the two antennas and make repeated measurements while the

setup is kept at a fixed location. Figure B2-1 illustrates the brick and block samples, moving

walls built with bricks, blocks, and styrofoam. Styrofoam slabs are used to secure the walls on

the platform. One styrofoam slab was also measured to estimate its loss and dielectric constant

and hence its impact on the measurement of other materials. It has very low loss and a dielectric

constant close to unity, thus it can be assumed to be effectively air.

A reinforced concrete pillar in the 3rd floor of Whittemore Hall, the building that houses

the Electrical Engineering Department of Virginia Tech, was also measured. Another reinforced

pillar in the Time Domain Lab (TDL) located in the 4th floor of Whittemore Hall was also

measured.

The cloth office partitions that were tested have round edges at the upper ends with

wooden caps for holding the cloth material tight. Each partition has two metal stands and as well

as support pieces inside. Figure B2-2 shows the different materials and walls used in

measurements.

4.8 Measurement Results

The dielectric constant and the loss at 5 GHz as representative values of parameters for

the selected materials are listed in Table 4.3. The frequency of 5-GHz is in a region of bandwidth

where the measurements are believed to be most accurate, because the results obtained from

different techniques agree very well and the amount of power transmitted in this region is

significantly far above the noise floor. It is emphasized that for most materials measuring the loss

is more difficult than measuring the real part of effective permittivity [Gey90]. A straight line is

used to model the insertion loss versus frequency [Gib99]. The fitted insertion transfer functions

for different materials are shown in Figure 4.11 and the corresponding parameters for the linear

fit are also given in Table 4.3. The insertion transfer function for the door is re-plotted in part (b)

of Figure 4.11 for ease of comparison. Cloth partition shows higher loss due to support elements


105

inside the partitions. Similarly, the dielectric constants are presented in Figure 4.12. The results

for the brick wall and the concrete block wall are over smaller bandwidths because of higher

losses of these materials that reduce their useful bandwidths. The dielectric constant versus

frequency can be modeled as a straight line with very small negative slop. However, the

dielectric constant for the brick has a small positive slope that is believed to be due to the non-

homogeneity of the sample. Attenuation constants for the door, wood, and structure wood sample

are given in Figure 4.13. It was possible to extract the attenuation constant for these materials

due to their moderate loss and homogenous structure.

To gain more insight into the effects of various walls on the propagation of UWB pulses,

the free-space and through gated signals for all the measured materials are presented in

Figures 4.14 through 4.19. For the case of blocks and bricks, the un-gated signals are presented

due to the difficulty of gating.

The dielectric constant of the glass sample could not be measured using the time delay

between peaks of the received pulses and the single-pass technique. This is because of the small

thickness of the glass that does not allow multiple reflections to be avoided, thus the multiple-

pass analysis should be used.

The reinforced concrete wall resulted in a very small amount of received power. No

further processing could be done, but an average dielectric constant was obtained by measuring

the time delay between the incident and received pulses. For better viewing, a longer time

window is shown in Figure 4.19. This figure illustrates multiple reflections inside the reinforced

concrete pillar. It is important to note that this window includes multipath components that might

not have traveled through the pillar. A repeated W shape is observed in the receiver signal. For

reinforced concrete, concrete block, and brick walls, a 10 dB gain and 15 GHz bandwidth

amplifier was used at the receiver side to increase the measured signal level.


106

0 5 10 15-10

-8

-6

-4

-2

0

Frequency (GHz)

Inse

rtion

loss

(dB)

door foam structure glass partition

0 5 10 15-15

-10

-5

0

Frequency (GHz)

Inse

rtion

loss

(dB

)

door board bricks blockswood

Figure 4.11. Insertion transfer function plotted versus frequency for different materials


107

0 5 10 15

1

2

3

4

5

6

7

Frequency (GHz)

Die

lect

ric C

onst

ant

doorfoamstructure woodglasspartition

0 5 10 15

1

2

3

4

5

6

7

Frequency (GHz)

Die

lect

ric C

onst

ant

doorboardbricksblockswood

Figure 4.12. Dielectric constant plotted versus frequency for different materials


108

0 5 10 150

50

100

150

200

250

300

Frequency (GHz)

Atte

nuat

ion

Con

stan

t (dB

/m)

d oorstructu re woodwood

Figure 4.13. Attenuation constant plotted versus frequency for different materials


109

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.06

-0.04

-0.02

0

0.02

0.04

0.06

time (ns)Am

plitu

de (V

)


0 1 2 3 4 5 6 7 8 9 10-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

time (ns)

Ampl

itude

(V)


Figure 4.14. Blocks wall and wallboard free-space and through measurements; plots (a) and (b) are ungated time-domain waveforms for blocks as the material using sources #1 and #2, respectively. Plots (c) and (d) are time-gated waveforms for the board using sources #1 and #2, respectively.

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

time (ns)

Ampl

itude

(V)


Board

0 1 2 3 4 5 6 7 8 9 10-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

time (ns)

Ampl

itude

(V)


(a) (c)

(d) (b)

Blocks wall


110

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.02

0

0.02

0.04

0.06

0.08

0.1

time (ns)

Ampl

itude

(V)


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.06

-0.04

-0.02

0

0.02

0.04

0.06

time (ns)

Ampl

itude

(V)


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

time (ns)

Ampl

itude

(V)


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.06

-0.04

-0.02

0

0.02

0.04

0.06

time (ns)

Ampl

itude

(V)


Figure 4.15. Cloth office partition and structure wood free-space and through measurements; plots (a) and (b) are time-gated waveforms for cloth partition as the material using sources #1 and #2, respectively. Plots (c) and (d) are time-gated waveforms for the structure wood using sources #1 and #2, respectively.

(a) (c)

(d) (b)

Cloth Partition Structure Wood


111

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

time (ns)

Ampl

itude

(V)


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

0.12

time (ns)

Ampl

itude

(V)


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.06

-0.04

-0.02

0

0.02

0.04

0.06

time (ns)Am

plitu

de (V

)


Figure 4.16. Door and wood free-space and through measurements; plots (a) and (b) are time-gated waveforms for sample door as the material using sources #1 and #2, respectively. Plots (c) and (d) are time-gated waveforms for the wood using sources #1 and #2, respectively.

(a) (c)

(d) (b)

Door Wood

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

time ns

Am

plitu

de V



112

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

time (ns)

Ampl

itude

(V)


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.06

-0.04

-0.02

0

0.02

0.04

0.06

time (ns)

Ampl

itude

(V)


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.04

-0.02

0

0.02

0.04

0.06

0.08

0.1

time (ns)

Ampl

itude

(V)


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.06

-0.04

-0.02

0

0.02

0.04

0.06

time (ns)Am

plitu

de (V

)


Figure 4.17. Glass and styrofoam free-space and through measurements; plots (a) and (b) are time-gated waveforms for glass as the material using sources #1 and #2, respectively. Plots (c) and (d) are time-gated waveforms for Styrofoam using sources #1 and #2, respectively.

(a) (c)

(d) (b)

Glass Styrofoam


113

0 1 2 3 4 5 6 7 8 9 10-0.04

-0.02

0

0.02

0.04

0.06

0.08

time (ns)

Ampl

itude

(V)


0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

time (ns)

Ampl

itude

(V)


0 2 4 6 8 10 12 14-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

0.25

0.3

time (ns)

Ampl

itude

(V)

Reference10*Through

0 2 4 6 8 10 12 14-0.2

-0.15

-0.1

-0.05

0

0.05

0.1

0.15

0.2

time (ns)

Ampl

itude

(V)

Reference10*Through

Figure 4.18. Bricks wall and reinforced concrete pillars free-space and through measurements; plots (a) and (b) are ungated time-domain waveforms for bricks as the material using sources #1 and #2, respectively. Plots (c) and (d) are ungated waveforms for reinforced concrete pillars (Whittemore and TDL) using sources #1 and #2, respectively.

(a) (c)

(d) (b)

Bricks wall Reinforced Concrete Pillars (c) Whittemore and (d)TDL)


114

0 5 10 15-0.4

-0.2

0

0.2

0.4

time (ns)

Am

plitu

de (V

) Reference

0 5 10 15-0.04

-0.02

0

0.02

0.04

time (ns)

Ampl

itude

(V) Through

0 10 20 30 40 50 60 70 80 90 100-0.04

-0.02

0

0.02

0.04

time (ns)

Ampl

itude

(V) Through, Long Profile

Figure 4.19. TDL reinforced concrete pillar free-space and through measurements; plot (a) is the reference measurement. Plot (b) indicates the measurement through the reinforced concrete pillar (TDL). Plot (c) demonstrates a longer profile for the through measurement.

(a)

(b)

(c)


115

4.9 Related Issues

In the following section, some points related to validity of the measurements are

discussed. These include distance between the samples and antennas, slab thickness and multi-

layer study, repeatability, and variability.

4.9.1 Distance from the Sample

The distance between a sample and the antenna should be long enough to ensure that the

sample is in the far field of the antenna. On the other hand, as the sample is moved away from

the antenna, edge and scattering effects cannot be gated out. Hence, a trade-off has to be made

without degrading the results. Moreover, as the distance increases the signal level decreases and

hence the frequency range over which reliable characterization can be made becomes smaller.

Most of our measurements were performed with a total distance of 1-3 meters. However, the

effect of the distance would not be pronounced if the free-space and through measurements

are carried out with exactly the same setup. An experiment was done by varying the distance

between the antennas and the sample in steps of 0.25m. No significant change was observed

other than the signal level.

4.9.2 Wall Thickness and Multi-layer Study

The thickness of the layer under study is critical to the measurement outcomes. If the

thickness is very small error becomes more pronounced. For example, to estimate the dielectric

constant of a slab of glass with 2 mm thickness, we should be able to measure the delay as result

of passing through this thin layer. If the thickness is larger, there will be a larger delay to

measure and hence less relative error in the thickness of the layer. On the other hand, very thick

slabs may cause high losses, resulting in weak signal levels that cannot be accurately measured.

For the case of glass sample, slabs consisting of one, two, and three layers were tested to confirm

the obtained parameters. The layers were carefully aligned to reduce the air gap. For the case of

the board, two layers were tested to confirm the results.

4.9.3 Repeatability Analysis

Repeatability analysis describes the process of evaluating the precision of measurements

taken at different instances of time. Measurements that have high precision are said to be


116

repeatable [Yoh01]. One important factor is to allow the equipments to warm up for a stable

performance. There is a small drift in the pulse with respect to the time axis when the equipment

warms up. Figures 4.20, and 4.21 illustrate the repeatability of the frequency and time-domain

measurements, respectively. Three different measurements of the wallboard sample are shown.

The wallboard was chosen as they have smaller thickness and low loss compared with other

materials. Examining the plots of insertion transfer functions and dielectric constants and noting

that the differences in the measurement results are minor lead to the conclusion that the

measurements are repeatable. The differences noticed in the plots must include tolerances of the

measurement setup.

4.9.4 Variability Analysis

Variability analysis describes the process of evaluating the precision of measurements of

different samples of the same material. Different measurements of different samples that have

high precision are said to have low variability. In the case of wall measurements, two different

samples of two different wallboards were measured (using the same calibration). The results of

these measurements are given in Figures 4.20 and 4.21. It should be noted that the results for

both repeatability and variability analyses are shown on the same plots. Examining these plots

and noting the differences in the measurement results are minor lead to the conclusion that the

two wallboards have a low variability, yet the differences indicate that there is some degree of

variability in the walls. In indoor environments, wallboards built from different materials and by

different manufacturers are used. But, this should not be a concern as the primary objective of

the material/wall characterization effort is to obtain estimates of the loss and the associated delay

for different construction materials and gain insights into how UWB propagation is affected by

these materials.


117

2 4 6 8 10 12 14-2

-1

0

1

2

Frequency (GHz)

Inse

rtion

Tra

nsfe

r Fun

ctio

n, M

agni

tude

Board 1, Meas. 1Board 1, Meas. 2Board 1, Meas. 3board 2

2 4 6 8 10 12 141

2

3

4

5

Frequency (GHz)

Dile

ctric

Con

stan

t


Figure 4.20. Repeatability and variability of frequency domain measurements


118

2 3 4 5 6 7 8 9-2

-1

0

1

2

Frequency (GHz)

Inse

rtion

Tra

nsfe

r Fun

ctio

n Board 1, Meas. 1Board 1, Meas. 2Board 1, Meas. 3board 2

2 3 4 5 6 7 8 91

2

3

4

5

Frequency (GHz)

Die

lect

ric C

onst

ant


Figure 4.21. Repeatability and variability of time domain measurements


119

4.10 Remarks on Pulse Shaping, UWB Receiver Design, and Modeling Hints

The following section is dedicated to putting the results of measurements into perspective

with regard to receiver design, pulse shaping, and channel modeling.

4.10.1 Receiver Design and Pulse Shaping

The idea of using correlators at the receiver might not work very well in a non-line-of-

site configuration. The pulse seems to undergo shape deformation as it propagates through

structures with small dimension due to inter-pulse interferences. Multiple reflections within the

material and multipath components have a significant impact on the maximum data rate and/or

multiple access capabilities of UWB systems. The claim that UWB has high multipath resolution

works very well in a free-space line-of-site configuration but seems to be less certain in

structures with fine details relative to the pulse duration.

Two seemingly contradicting requirements have to be traded off. One would like to have

the pulse with very short duration and at the same time to have enough low frequency

components. As the pulse gets shorter, its spectral contents are shifted to higher frequencies

which suffer more attenuation as the pulse propagates. When deciding on the spectrum to be

used, it is important to note that as the signal is shift to higher frequencies, the original reasons

for proposing UWB including spectrum reuse and propagation through-the-wall become

irrelevant. Reception based on pulse shape might not be the best approach for indoor

environments. Other means of capturing the signal energy might prove to be more practical.

The originally proposed modulation scheme, in which a bit is demodulated as a zero or

one based on the time delay, is also vulnerable to errors [Sch93]. Walls and barriers can

complicate the demodulation process as they introduce more delays. This might not be a major

problem as the tight synchronization requirement is an integrated part of UWB systems.


120

4.10.2 Modeling and Large-Scale Path-losses

The physical models used to predict pulse propagation in dielectric materials are based on

two techniques; namely, electromagnetic wave theory and geometrical optics. The latter method

is only applicable when the wavelength of the applied electromagnetic signal is considerably

shorter than the dimension of object or medium being excited [Dan96].

One way of modeling large-scale path losses is to assume logarithmic attenuation with

various types of structures between the transmitter and receiver antennas [Has93a]. It has also

been stated that adding the individual attenuations results in the total dB loss [Has93a].

Furthermore, it is important to note that when assuming no dispersion takes place, a narrow band

approximation is implied. This assumption is not as good for UWB because the dielectric

constant decreases slowly with frequency.

Many results for the propagation through walls have been published. A good summary is

given in [Has93a]. However, these results were often obtained at specific frequencies and

measurements were not performed with sufficient care to remove the effects of scattering from

edges. The measurements carried out in our lab have been crosschecked by using both time-

domain and frequency-domain techniques

4.11 UWB Partition Dependent Propagation Modeling

Many of the narrowband channel characterization efforts are performed at specific

frequencies. For UWB characterization, one has to define the pulse shape or its spectrum

occupancy. Results generated for a specific pulse might not be generalized to other UWB

signals.

In this section, the results for the loss of the tested materials are used to develop partition-

dependent propagation models. The partition based penetration loss is defined as the path-loss

difference between two locations on the opposite sides of a wall [And02a]. The penetration loss

is equal to the insertion loss presented earlier. The free space path-loss exponent is assumed to be

n=2. The total loss along a path is the sum of free-space path loss and loss associated with

partitions present along the propagation path.


121

In the narrowband context, the path loss with respect to1 m free space at a point located a

distance d from the reference point is described by the following equation

........)(log20)( 10 ba XbXaddPL ×+×+= , ( 4.50)

where a, b, etc., are the numbers of each partition type and Xa, Xb, etc., are their respective

attenuation values measured in dB [Dur98]. To extend this concept to UWB communication

channels, we introduce the frequency dependent version of equation (4.50),

)........()()(log20),( 10 fXbfXadfdPL ba ×+×+= , ( 4.51)

where Xa(f), Xb(f) are the frequency dependent insertion losses of partitions. Equation (4.51)

gives the path loss at single frequency points. In order to find the pulse shape and the total power

loss we need to find the time domain equivalent of (4.51) by means of inverse Fourier transform

over the frequency range of the radiated signal. In doing so, we start with the radiated pulse

prad(t). In most wideband antennas such as TEM horns, this signal is proportional to the

derivative of the input signal to the antenna. Then, we determine the spectrum of the received

signal at the location of the receive antenna using the following relation ship,

d

fPfPfXbfXa

radrec

ba

⋅

=×+×

20).......)()((

10

)()( ( 4.52)

It is important to note that the attenuation is applied to the radiated signal rather than the input to

the antenna. The transmit antenna alters the spectrum of the input signal as illustrated in Figure

4.22. Starting with a Gaussian pulse, the time-domain received signal )(tprec is obtained by

inverse Fourier transforming )( fPrec . With the received pulse determined, one is able to assess

pulse distortion and the total power loss. It has been assumed that the dielectric constant of the

partitions remain constant over the spectrum of the radiated signal.

Example:

In this example we illustrate how to utilize the material characterization results and apply

them to a partition problem. The objective is to find the power loss through a propagation path


122

and to estimate the pulse shape and the frequency distribution of the received signal. Consider a

line-of-site path with two partitions between two TEM horn antennas as shown in Figure 4.23a .

The first partition is a sheet of glass and the second is a wooden door with the same thickness as

those that have been characterized. The input signal to the antenna and that radiated from it are

displayed in this figure. These signals are obtained through measurements.

To estimate the signal passed through the glass partition, Fourier transform is used to

determine the spectrum of the radiated signal and the frequency dependent loss is applied to this

spectrum. Inverse Fourier transform is then used to obtain the time-domain signal passed the

glass sheet. The same procedure is repeated to estimate the signal passed through the wooden

door partition. Examining The loss in the signal power is evident in Figure 4.23b. It is also noted

that higher frequencies are smoothed out. The change in frequency distributions is more evident

in Figure 4.23c. At lower frequencies, the spectra of the radiated signal, signal after the glass and

signal after the wooden door are very close, whereas at higher frequencies the differences are

more pronounced. This analysis is helpful in link-budget analysis and understanding of potential

interference effects from indoor to outdoor environments.

Figure 4.22 Gaussian (TEM horn input signal) and Gaussian monocycle (TEM horn radiated signal) waveforms and their corresponding normalized spectra.

0 2 4 6 8 10 12

10-8

10-6

10-4

10-2

100

Frequency (GHz)

Nor

mal

ized

Mag

nitu

de

Gaussian and Gaussian Monocycle in Frequency Domain

GaussianGaussian Monocycle

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

Time (ns)

Nor

mal

ized

Am

plitu

de

Gaussian and Gaussian Monocycle Waveforms

GaussianGaussian Monocycle

(a) Gaussian and Gaussian monocycle waveforms (b) Normalized spectra for Gaussian and Gaussian monocycle waveforms


123

Figure 4.23 Illustrative example for UWB partition dependent modeling

(a) Illustration of the partitions setup, (b) Radiated signal, signal after the glass partition, and the signal after the wooden door, (c) Frequency distribution of the signal at different points.

Er E

0 0.5 1 1.5 2 2.5 3-0.1

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

time (ns)

Am

plitu

de (V

)

0 0.5 1 1.5 2 2.5 3-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

time (ns)

Am

plitu

de (V

)

0 0.5 1 1.5 2 2.5 3-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

time (ns)

Am

plitu

de (V

)

0 0.5 1 1.5 2 2.5 3-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

0.05

time (ns)

Am

plitu

de (V

)

Glass Wooden Door

0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

0.04

time (ns)

Am

plitu

de (V

)

Radiated SignalAfter Glass PartitionAfter Wooden Door

0 2 4 6 8 10 12-70

-60

-50

-40

-30

-20

-10

0

10

Frequency (GHz)

Nor

mal

ized

Mag

nitu

de d

B

Radiated SignalAfter Glass PartitionAfter Wooden Door

(a)

(b) (c)


124

4.12 Concluding Remarks

Electromagnetic characterization of materials and walls commonly encountered in indoor

environments was undertaken with the aim of assessing their impacts on UWB propagation.

Measurements were carried out in both time domain and frequency domain. Also, whenever

possible, both single-pass and multiple-pass analysis techniques were used. A new formulation

for the characterization of low-loss materials has been presented which requires solving real

equations only and converges more rapidly, thus requires much less computation time than that

based on solving the complex equation relating the insertion transfer function to the dielectric

constant of the material under test. The new formulation can be used to accurately characterize

many materials of practical applications which are low loss. Results from different techniques

agree well, thus ensuring the reliability and accuracy of the measurements. Ten different

materials were tested and results were presented in terms of insertion loss and dielectric constant.

The presented results should serve as a basis for further studies in developing appropriate models

for UWB channels. The results are also useful in UWB link budget analysis.

Chapter5. Time Domain Indoor Channel Measurements

“… Any realistic channel model…should derive its parameters from actual field measurements rather than basing them on simplified theory”

H. Hashemi [Has93a]

Chapter 5

Indoor UWB Channel Measurements*

5.1 Introduction

The objectives of this report are to present time-domain measurements and

characterization of ultra-wideband (UWB) propagation in indoor environments and to

detail the experimental procedures and measurement setup used to collect data. First,

experimental procedures and locations where the measurements were carried out are

described. Then, post-processing of the acquired data is explained. Finally, the results

pertaining to the signal quality, small-scale effects, large-scale pathloss exponents, and

time dispersion parameters are discussed. Some site-specific trends and observations are

described and channel performances for two types of directive and omni-directional

antennas are compared.

* [Muq03e][Muq03f]

125


5.2 Description of Measurement Procedure and Locations

In this section the details about the measurements procedure are presented.

Locations where the measurements were conducted are also described to allow for

understanding some of the site-specific trends.

5.2.1 Measurement Procedure and Setup

Time domain measurements were performed using a sampling oscilloscope as a

receiver and a Gaussian-like pulse generator as transmitter. Two low noise wideband

amplifiers were used at the receiver side. Each amplifier has a again of 10 dB and a 3dB-

bandwidth of 15 GHz. The width of the excitation pulse is less than 100 ps. Offset

calibration is carried out with a matched load before performing any measurements. The

original data were over-sampled in the time domain with 10 ps/sample which results in a

noise tail in the frequency domain. An acquisition time window of 100 ns was selected by

making sure all observable multipath components are accounted for. This time window is

consistent with the maximum excess delay of 70 ns reported by other investigators

[Hov02]. The sampling oscilloscope allows a maximum of 5K points at a time. The 5K

points correspond to 50 ns time window. Two measured 50 ns time windows were

cascaded to yield a 100 ns acquisition time. The process was semi-automated using

LabView ® software. A total of about 400 profiles were collected. The spatial width of

the used pulse in our measurements is much smaller than the one used in previously

published measurements. The spatial width is small enough to make the line-of-site path

always resolvable from any other multipath component. Information about the excitation

pulse allows for deconvolution and hence generalization of results for use in other

communication applications in the covered frequency range [Vau99].

In indoor environments, the time varying part of the impulse response is typically

due to human movements. By conducting measurements during low activity periods and

by keeping both the transmitter and the receiver stationary, the channel can be treated as

126


being quasi-stationary. This allows us to average 32 measurements, thus effectively

canceling out the noise.

Two different sets of measurements were performed based on TEM horn antennas

(antenna#1) and biconical antennas. Details about the antennas are presented in Chapter

3. Both transmitter and receiver antennas were placed on plastic moving carts at an

elevation of about 1.25 m above the floor. Styrofoam slabs were used to adjust the

elevation without introducing reflectors around the antenna. The TEM horn antennas

were aligned for maximum boresight reception. The advantage of using TEM horn

antennas is that they are ultra wideband radiators designed and optimized for time-

domain impulse response measurements. TEM horn antennas emulate sectored antenna

proposed for gaga-Hertz frequency indoor application. The TEM horn antenna has very

narrow beamwidth and thus is highly directional. With TEM horns fewer multipath

components are received, and almost none from behind the receiver. TEM horns have

been used for channel characterization in the past. For example, [Dav91] and [And02a]

used a TEM horn as the transmitting antenna and a TEM horn or an omnidirectional

antenna as the receiving antenna for channel characterization in the 60 GHz band. The

extent of multipath effects due to directional antennas on measurements is highlighted in

[Dur00]. On the other hand, biconical antennas are omni-directional and are more likely

to be used in mobile applications. The biconical antennas used in this investigation are

not designed as impulse antennas but they are impedance matched over a very wide

bandwidth.

Two levels of measurements are performed to characterize the small-scale and the

large-scale fading parameters of the channel. A 77 × grid with 15 cm spacing between

adjacent points was designed and used in the measurements. Since no small-scale fading

due to phase cancellation was observed, only a 3 3× measurement grid with 45 cm

spacing between adjacent points was used, except for measurements in the 4th floor

corridor in Whittemore Hall where one measurement was performed with the TEM horn

antennas and four measurements were made using the biconical antennas per each large-

scale location. The measurement grid is illustrated in Figure 5.1. The triggering signal

127


was carried by a coaxial cable to the sampling oscilloscope. As the distance between

transmitter and receiver increased the loss and dispersion in the triggering cable increased

too, resulting in a higher jitter. An effort was made to use higher quality cables and

minimum possible length for the triggering cable. A personal computer was used to store

and post process the data.

(1,1) (1,4) (1,7)

(4,1) (4,4) (4,7)

(7,1) (7,4) (7,7)

Figure 5.1. Measurement grid

90cm

90 cm

5.2.2 Description of Measurement Locations

The measurements were carried out in two buildings on Virginia Tech Campus,

namely; Whittemore Hall and Durham Hall. The former building comprised mainly of

offices and classrooms. Most walls are made of drywalls with metallic studs. Some walls

at certain locations including stairwells are made of cinderblock and poured concrete. In

Durham Hall, interior walls are largely made of drywalls and cinderblocks. The floor is

covered with ceramic tiles in hallways and with carpet inside the rooms. An advantage of

performing UWB experiments in these buildings is that they have been characterized for

some narrowband measurements and site-specific ray tracing studies [Sei94], [Haw91],

[Rap92], [And02a], [And02b]. This allows one to compare the narrowband and the UWB

channel characterization results.

128


In Whittemore Hall, the measurements were performed in three different floors;

along the hallways on the second floor, in a narrow corridor on the fourth floor and in a

conference room on the sixth floor. Appendix B3 shows the blue prints and photographs

of measurement locations. In Durham Hall, all measurements were carried on the fourth

floor. Five different transmitter locations were considered. For every location,

measurements at different receiver locations were performed, as indicated on the

blueprints in Figure B3.3a. To visualize the measurements environment, photographs of

the measurement locations are presented in Figures B3.3b and B3.3c.

Different scenarios are considered. Line-of-site (LOS) and non-line-of-site

(NLOS) topographies are of paramount interest. Room-to-room, within the room and

hallways are all typical indoor environments. Shadowing effects can also be assessed in

some scenarios. Table 5.1 summarizes the locations and scenarios measured.

5.3 Signal Processing and Data Analysis

A major challenge in UWB channel measurements is that the measurement

bandwidth is open to any signal. When taking measurements close to utility rooms or

laboratories that are radiating electromagnetic signals, there is an apparent increase in the

noise floor. To reduce the interference from undesirable sources, the acquired profile is

filtered in the time domain to remove some signals that are not part of the transmitted

pulse. The 3 dB bandwidth of the bandpass filter used for interference rejection occupies

a frequency range from 0.1 GHz to 12 GHz. The corner frequencies of the filter were

chosen by observing the spectrum of the radiated pulse and making sure that there is no

significant energy outside the pass band. The filtering process has three advantages. First,

the noise energy is reduced by eliminating out of band noise resulting from over

sampling. Second, any dc offset that has not been taken into account by the calibration is

removed. Finally, the lower frequency signals radiated from the pulse generator’s internal

electronics are eliminated. It was noted that the pulse generator gives off low frequency

components in the 30 MHz range that can be picked up by the biconical antenna. This is

illustrated in Figure 5.2 where a typical profile measured with the biconical antennas and

its filtered version are compared.

129


Table 5.1. Measurement locations and scenarios

# Location Description

# profiles (TEM

)

# profiles (B

iconical) d

W2.A Hallways in 2nd floor LOS, Hallways 9 9 9.3

W2.B Hallways in 2nd floor LOS, Hallways 9 9 17.6

W2.C Hallways in 2nd floor NLOS, Hallways 0 0 -

W2.D Hallways in 2nd floor NLOS, Hallways 0 0 -

W2.E Hallways in 2nd floor LOS, Hallways 9 9 15.4

W2.F Hallways in 2nd floor LOS, Hallways 9 9 30.9

W2.G Hallways in 2nd floor LOS, Hallways 9 9 49.1

W2.H Hallways in 2nd floor NLOS, Hallways 0 0 -

W2.I Hallways in 2nd floor NLOS, Hallways 0 0 -

W4 Corridor in the 4th floor LOS, Small corridor 12 25 Varying

W6.A Conference Room in 6th floor LOS, Within a room 9 9 5.1

W6.B Conference Room in 6th floor LOS, Within a room 9 9 6.2

D1.A Tx (Room 475)-Rx (Room 471) NLOS, Room-to-Room 9 9 3.6

D1.B Tx (Room 475)-Rx (Room 471) NLOS, Room-to-Room 9 9 9.0

D2.A Hallway in 4th floor LOS, Hallways with concrete walls 9 9 8.2

D2.B Hallway in 4th floor LOS, Hallways with concrete walls 9 9 14.6

D2.C Hallway in 4th floor LOS, Hallways with concrete walls 9 9 20.4

D3.A Hallway+ open environment Hallway with open space in the middle 9 9 20.3

D3.B Hallway+ open environment Hallway with open space in the middle 9 9 30.7

D4.A Tx (Room 476)- Rx (internal room) NLOS, Room-to-Room 9 9 5.5

D4.B Tx (Room 476, MPRG Lab.) Cubical office Environment (Obstructed) 0 0 -

D4.C Tx (Room 476, MPRG Lab.) LOS, Cubical office Environment 9 9 11.8

D5.A Tx (Room 423, MPRG Reception), Rx (Room 433)

LOS/NLOS , Room-to-Room 3 3 5.6

D5.B Tx (Room 423, MPRG Reception), Rx (Hallway)

Room-to-Hallway, (obstructed, NLOS) 9 9 5.9

D5.C Tx (Room 423, MPRG Reception), Rx (Hallway)

Through glass wall 9 9 7.5

D5.D Tx (Room 423, MPRG Reception), Rx (Hallway)


130W: Whittemore Hall , D: Durham Hall


0 10 20 30 40 50 60 70 80 90 100-0.06

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

time (ns)

Am

plitu

de (V

)

)

0 10 20 30 40-0.06

-0.05

-0.04

-0.03

-0.02

-0.01

0

0.01

0.02

0.03

Am

plitu

de (V

)

Figure 5.2. Effect of filtering the measured

(a) unfiltered profile (b) filtered profile

(a

50 60 70 80 90 100time (ns)
(b)
profile (using the biconical antenna)

131


Precursor and noise before the arrival of the first component are forced to zero in

order to ensure the causality of results in the processing stage. For energy calculation and

large-scale pathloss analysis, a noise threshold is introduced below which all data are

assumed to be zero. The threshold was set at 6 dB above the noise floor determined as the

maximum level of the profile tail in the last 5 ns of the 100 ns time window.

5.4 Results and Analysis

The results for measurements with the TEM horn and biconical antennas are

presented separately. At locations W2.C, W2.D, W2.H, W2.I and D4.B no signal could

be clearly detected by the receiver. At the first four locations there was no line-of-sight

between the transmitter and the receiver, because the obstructed path is either through

concrete walls or through multiple drywalls with metallic studs. The insertion loss is a

function of frequency. Thick reinforced concrete walls are impenetrable at high

frequencies [Dav91]. At the location D4.B, the line-of-sight is obstructed with office

cubical metallic partitions. The use of the omnidirectional antenna did not improve the

measurements at those locations. The reason is that the pulse spectrum contains

substantial high frequency contents. At higher frequencies, the line-of-sight component is

the most significant part, since diffracted components and through-the-wall propagation

are much weaker. Moreover, indoor path loss generally increases with frequency. Small-

scale effects, large-scale path loss and time dispersion parameters are discussed

separately in the following subsections.

5.4.1 Small-Scale Fading and signal Quality

In narrowband communication systems, small-scale fading describes the

fluctuation due to constructive and destructive interference of the multipath components

at the receiver when sub-wavelength changes are made in the receiver position

[Rap96][Dur00]. Such definition can be extended to UWB communication as the

constructive and destructive interference of the multipath components at the receiver due

to a change in its position in the order of sub-spatial-width of the transmitted pulse.

132


Sample results are presented for measured delay profiles, referred to as local

power delay profiles (local-PDP). The remaining results are presented for small-scale

averaged power delay profile (SSA-PDP). In the SSA-PDP the nine measurements are

properly delayed and averaged. Figure 5.3 illustrates how SSA-PDP are different from

the local-PDP. The first three plots, Figure 5.3a to 5.3c, are local-PDPs and the last one,

Figure 5.3d, is the average of all 9 local measurements. When delay-and-average is used,

the line of site components tend to prevail and the other components spread out on the

time axis such that they do not add coherently because of the high resolution of the

transmitted pulse. Small-scale processing shows the capability for using a delay-and-sum

beamformer to process a received array of signals from different antennas located in a

very small area [Cra98]. It is important to note that small-scale and large-scale

terminologies are used as relative measures of distances between the receiver locations

compared with the wavelength. These terminologies do not fit well to our analysis

because any movement tends to be large compared to the effective wavelength.

In narrowband measurements, the spacing between the local measurements is

related to the wavelength, λ . It is reported in [Dur98] that one can cancel out small-scale

effects by averaging power along 20λ linear or circular paths independent of the signal

bandwidth. An established fact is that local fading results from the destructive

interference of multipath components [Rap89]. However, for UWB signals there is no

single frequency or single wavelength, thus no destructive interference can occur over the

entire bandwidth of the pulse. Our observations of received signals at different points in

a grid confirm the absence of small-scale fading. To quantify this effect, let us consider

the signal quality as defined in [Win98b]

01010 log10log10 EEQ −= (5.1)

where E is the received signal energy given by

(5.2) dttrET

∫=0

2 )(

133


134

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

2

4

6x 10-3

time (ns)

v2

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

2

4

6x 10-3

time (ns)

v2

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

2

4

6x 10-3

time (ns)

v2

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 50

2

4

6x 10-3

time (ns)

v2

(a)

(b)

(c)

(d)

Figure 5.3. Comparison between small-scale averaged power delay profiles (SSA-PDP) and local power delay profiles. Measurements were performed at location W2 (TEM horn antennas, Whittemore, 2nd floor)

(a) measured power delay profile (PDP) at location (1,1) (b) measured power delay profile (PDP) at location (1,4) (c) measured power delay profile (PDP) at location (1,7) (d) small-scale averaged power delay profiles (SSA-PDP) for the nine

measurements on the grid.


where is the measured multipath profile and T is the observation time. E0 is the

energy measured at a reference location, which is usually chosen to be at a 1 m distance

from the transmitter. The cumulative distribution functions (CDF)s for the signal quality

for all measured grids are shown in Figure 5.4. There is almost no fading as a result of

interference. All local PDPs are within 3 dB of the average level unless some profiles are

obstructed with some objects in the channel or they belong to points close to walls. If

transmitter and/or receiver locations are close to walls, the received profile is affected

significantly [Rap89]. Robustness of UWB communication systems, insofar as multipath

is concerned, is manifested by small variations in signal quality at various grid locations

[Sch97].

)(tr

5.4.2 Pathloss and Large-Scale Analysis

The energy in the received profile, statistically speaking, decreases with the

distance between the receiver and the transmitter. The pathloss exponent, n, is a measure

of decay in signal power with distance, d, according to 1 . A reference measurement

is performed at a distance of 1 m form the transmitter. Subsequent energy measurements

are performed with respect to the reference measurements. Using the log-normal

shadowing assumption, the path loss exponent, n, is related to the received energy at

distance d and the reference measurement by

/ nd

σXddndPLdPL +

+=

0100 log10)()( (5.3)

where is the reference distance,0d )( 0dPL is the average measured energy at the

reference distance and is a zero-mean Gaussian distributed random variable in (dB)

with standard deviation equals

σX

σ [Rap96]. The path loss exponent is obtained by fitting

a line on the logarithmic scatter plot of energy versus distance. The standard deviation for

the Gaussian random variable is obtained by calculating the deviation from the obtained

fit. The reference measurement is very important as it defines the intercept with the

vertical axis and hence affects the fitted slope. Many reference measurements can be

averaged together to reduce the effect of the measurement environment.

135


-35 -30 -25 -20 -15 -1010

20

30

40

50

60

70

80

90

100

Signal Quality (dB)

Per

cent

age

less

than

Abs

ciss

a

TEM Horn Antennas

-35 -30 -25 -20 -15 -1010

20

30

40

50

60

70

80

90

100

Signal Quality (dB)

Per

cent

age

less

than

Abs

ciss

a

Biconical Anteenas

(a)

W2W6D1D2D3D4D5

W2W6D1D2D3D4D5

(b)Figure 5.4. The cumulative distribution of the signal quality based on 9 spatial sample

points;

(a) with TEM horn antennas (b) with biconical omnidirectional antennas

136


In narrowband characterization, the local PDPs, small-scale averaged PDPs (SSA-PDP)

are usually used to eliminate any small-scale effect. The same technique is implemented

by Cassioli et al. [Cas01] to generate local and global parameters.

In the present analysis, the UWB pulse delay time is used to find the distance

between the receiver and the transmitter. First, the distance between the transmitter and

the receiver is measured at a reference position. Then, other distances separating the

receiver from the transmitter are calculated using the pulse delay time. This allows us to

take measurements at locations with small separation distances and reduces the error

associated with distance measurements. Measured points are distributed across the entire

scattering plots rather than being clustered. This distribution reduces the error associated

with reference measurements. The scatter plots for global data are presented in Figure

5.5.

Scatter plots for LOS and NLOS scenarios are shown separately in Figure 5.6.

The extracted parameters for each scenario are listed in Table 5.2. The minimum path

loss exponent is 1.27 for the case of the narrow corridor which has nearly the behavior of

a lossy waveguide structure. The maximum pathloss exponent is 3.29 in the obstructed

scenario D1. The global line-of-sight parameters are n=1.61 and σ =1.58 dB for the

TEM horns and n=1.58 andσ =1.91 dB the biconical antennas. The NLOS scenarios

have pathloss exponents greater than 2 and also have larger σ values compared with LOS

scenarios.

In general, there is a close agreement between the results obtained with directive

antennas and the results obtained with omni-directional antennas. This similarity of

results obtained with the two types of antennas de-emphasizes the contribution of the

back reflection components. A notable difference is lower σ value when directive

antennas are used, especially in NLOS scenarios as directive antenna can be easily

shadowed by any object in the channel while omni-directional antennas can still receive

some components.

137


15

20

25

30

35 TEM Horn Antennas

1m

free

spa

ce p

ath

loss

(dB)

W2W4

Fig

(a)

10 0 101

10 2 0

5

10

Transmitter-receiver separation distance (m)

Path

loss

with

resp

ect t

o W6D1D2D3D4D5Fit, n=1.8274n=2

20

25

30

35 Biconical Antennas

free

spa

ce p

ath

loss

(dB )

W2W4

u

(b)

10 0 101

10 2 0

5

10

15

Transmitter-receiver separation distance (m)

Path

loss

with

resp

ect t

o 1m W6

D1D2D3D4D5Fit, n=1.7482n=2

re 5.5. Scatter plot for the relative pathloss versus frequency for all locations

(a) using TEM horn antennas, (b) using biconical antennas.

138


100 101 1020

5

10

15

20

25

30

35LOS using the TEM Horn Antennas

Transmitter-receiver Separation Distance (m)

Pat

hlos

s w

ith re

spec

t to

1m fr

ee s

pace

pat

h lo

ss (d

B) LOS

Fit, n=1.6077n=2

100 101 1020

5

10

15

20

25

30

35

40NLOS using the TEM Horn Antennas


Pat

hlos

s w

ith re

spec

t to

1m fr

ee s

pace

pat

h lo

ss (d

B)

NLOSFit, n=2.6039n=2

100 101 1020

5

10

15

20

25

30

35

40NLOS using the Biconical Antennas

Transmitter-receiver separation sistance (m)

Pat

hlos

s w

ith re

spec

t to

1m fr

ee s

pace

pat

h lo

ss (d

B)

NLOSFit, n=2.4118n=2

100 101 1020

5

10

15

20

25

30

35LOS using the Biconical Antennas


Pat

hlos

s w

ith re

spec

t to

1m fr

ee s

pace

pat

h lo

ss (d

B)

LOSFit, n=1.5826n=2

(b) (a)

(d) (c)

Figure 5.6. Scatter plots for the pathloss versus distance, for LOS and NLOS scenarios:

(a) LOS using TEM horn antenna, (b) NLOS using TEM horn antenna, (c) LOS using biconical antenna, (d) NLOS using biconical antenna.

139


Table 5.2. Large-scale pathloss parameters for both TEM horn and biconical antennas

Antenna TEM Horn Biconical Location n σ n σ

W & D 1.8274 5.7291 1.7482 4.2585 W 1.5602 1.7196 1.5653 2.0095 W2 1.5454 1.6763 1.5807 1.3494 W4 1.2744 0.4763 1.3035 1.9032 W6 1.7845 0.7160 1.8192 1.0235 D 2.0401 6.5007 1.9103 4.8007 D1 3.2883 2.6456 2.9655 1.8769 D2 1.6591 1.0542 1.5365 1.5381 D3 1.7986 2.3358 1.7983 2.2394 D4 1.6478 1.3841 1.7870 3.9154 D5 2.6701 5.6894 2.2455 1.8009 LOS 1.6077 1.5816 1.5826 1.9135 NLOS 2.6039 6.0840 2.4118 3.2698

W: Whittemore Hall, D: Durham Hall

The reported pathloss exponents for narrowband systems are 1.6-1.8 for in-

building line-of-sight environments and 4-6 for obstructed in-building environments

[Rap96]. As noted from Table 2, The pathloss exponents for UWB are comparable with

pathloss exponents for narrowband LOS scenarios but are smaller for NLOS scenarios.

The results for pathloss exponent and the standard deviation introduced by Ghassem-

zadeh et al. [Gha02] are also comparable to the results obtained from our measurements.

[Gha02] performed UWB frequency domain measurements around 5 GHz, which is close

to the center frequency in the spectrum of the pulse used in our experiments. Their

parameters are n=1.7, σ =1.6 dB for LOS scenarios and n=3.5,σ =2.7 dB for NLOS

scenarios.

5.4.3 Time Dispersion Results

The time dispersion parameters shed some light on the temporal distribution of

power relative to the first arriving components. Delay spreads restrict transmitted data

rates and could limit the capacity of the system when multi-user systems are considered.

140


The time dispersion of UWB pulses can be presented as the ratio of the average arrival

time to the spread of the arrival time. Time dispersion parameters formulation can be

found in Chapter 3.

The ratio of the mean excess delay to the RMS delay spread can be used as a

measure of the time dispersion for UWB signals. If τστ = , then the multipath delay

profile decays exponentially. The same situation corresponds to two multipath

components with equal power where the second path is 2τ away from the first

component. High concentration of power at small excess delay values is reflected by

τστ / <1. When the energy arrive at the mid point of the power delay profile and not at

the earliest part then τστ / >1 [Rap89].

The cumulative distribution function (CDF) for the RMS delay spread is plotted

in Figure 5.7. All multipath components within 20 dB of the maximum are included.

Obstructed and non-line-of-sight scenarios resulted in higher time dispersion. The

variations between different scenarios and buildings are less for the omnidirectional

antennas when compared with the directive TEM horns. When the biconical antennas are

used the values are higher because receive bicone can receive more multipath

components. The results for SSA-PDPs are presented in Table 5.3 for the TEM horn and

in Table 5.4 for the biconical antenna. It should be noted that the instantaneous delay

spreads cannot be averaged to give the delay spread. Instead, for the SSA-PDPs, the

power delay profiles are averaged and then the delay spread is calculated. Individual

power delay profiles are averaged, weighted by their own power [Vau99].

Time Dispersion Parameters Correlation with channel Parameters

Now, the correlation between the channel time dispersion parameters is examined.

The relation between the mean excess delay and the RMS delay spread is illustrated in

Figure 5.8. The ratio τστ / is mostly in the range of 0.25-1. The small values for this

ratio imply high concentration of power at small excess delay. Obstructed measurements

141


0 5 10 15 20 250

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

RMS Delay Spread (ns)

Pro

babl

ilty

RM

S D

elay

Spr

ead<

Abs

ciss

a

TEM Horn

D1D2D3D4D5

0 5 10 15 20 250

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


Pro

babl

ilty

RM

S D

elay

Spr

ead<

Abs

ciss

a

Biconical Antennas

W2W4W6

0 5 10 15 20 25 300

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


Pro

babl

ilty

RM

S D

elay

Spr

ead<

Abs

ciss

a

Biconical Antennas

D1D2D3D4D5

0 1 2 3 4 5 60

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1


Pro

babl

ilty

RM

S D

elay

Spr

ead<

Abs

ciss

aTEM Horn

W2W4W6

(b)(a)

(c) (d)

Figure 5.7. Cumulative distribution functions (CDF)s for the RMS delay spread (20 dB), for Whittemore and Durham Halls:

(a) Whittemore Hall using TEM horn antenna, (b) Whittemore Hall using TEM horn antenna, (c) Durham Hall using biconical antenna, (d) Durham Hall using biconical antenna.

142


Table 5.3. Parameters for small-scale averaged PDP (SSA-PDP) TEM horn antennas

Threshold 10 dB 20 dB Location d τ σ maxτ % τ σ maxτ %

W2.A 9.26 0.232 0.473 1.29 52.4 0.633 1.109 13.13 81.4

W2.B 17.70 0.310 0.306 1.21 67.1 0.843 2.079 14.11 85.3

W2.E 15.24 0.229 0.379 1.16 54.0 0.838 1.570 13.71 81.5

W2.F 30.61 0.469 0.349 1.73 71.0 1.139 2.276 14.01 89.0

W2.G 48.94 0.459 0.459 2.95 73.3 0.969 1.943 26.23 90.2

W6.A 5.11 0 0.058 0.09 68.4 0.101 0.955 13.13 73.1

W6.B 6.37 0.014 0.090 0.33 74.4 0.073 0.627 10.54 83.3

D1.A 3.60 0.597 0.518 1.84 27.9 6.745 7.846 36.91 80.4

D1.B 9.00 0.053 0.169 0.69 21.7 3.259 5.102 28.34 54.0

D2.A 8.21 0.014 0.133 0.78 52.0 0.464 1.176 13.13 78.1

D2.B 14.59 0.160 0.301 1.00 54.3 0.412 0.809 13.11 79.2

D2.C 20.43 0.215 0.266 0.78 59.8 0.582 1.400 13.94 81.6

D3.A 20.31 0.165 0.250 0.64 59.1 0.770 2.445 14.06 81.7

D3.B 30.73 0.191 0.183 0.48 60.1 0.932 2.767 13.88 79.1

D4.A 5.45 0.012 0.057 0.10 49.0 0.618 1.466 10.53 68.8

D4.C 11.76 0.039 0.088 0.32 56.9 0.201 0.427 3.00 74.9

D5.A 5.59 0.007 0.054 0.09 69.2 0.115 0.675 13.11 77.6

D5.B 5.89 1.020 1.215 4.00 33.5 7.272 8.591 45.77 81.3

D5.C 7.53 0 0.054 0.08 67.5 0.216 1.719 13.15 77.4

D5.D 9.51 0 0.053 0.08 63.9 0.198 1.659 13.16 72.0

min 3.6 0 0.053 0.08 21.7 0.0730 0.4270 3.00 54.0

max 48.94 1.02 1.215 4 74.4 7.2720 8.5910 45.77 90.20

mean 14.29 0.2093 0.2727 0.982 56.78 1.3190 2.3320 16.85 78.45

median 9.39 0.1625 0.2165 0.735 59.45 0.6255 1.6145 13.44 79.80

143


Table 5.4. Parameters for small-scale averaged PDP (SSA-PDP) using biconical antennas

Threshold 10 dB 20 dB Location d (m) τ σ maxτ % τ σ maxτ %

W2.A 9.27 0.661 0.590 1.68 46.93 2.717 3.656 16.63 76.63

W2.B 17.41 1.294 2.269 10.10 37.93 6.355 8.410 92.35 76.33

W2.E 15.63 1.080 1.082 3.99 50.21 3.361 4.266 56.37 83.53

W2.F 31.2 0.335 0.285 0.89 47.81 1.956 2.859 14.93 79.27

W2.G 49.34 0.527 1.042 6.19 44.14 1.916 2.942 26.29 70.10

W6.A 5.12 0.102 0.030 0.00 16.91 3.678 4.377 21.37 62.18

W6.B 6.05 0.224 1.521 8.62 20.16 5.485 5.447 25.23 58.40

D1.A 3.54 4.078 3.862 27.31 42.98 15.546 15.960 83.41 94.19

D1.B 8.98 18.622 13.698 64.52 63.67 28.781 22.086 88.67 99.88

D2.A 8.22 0.708 0.768 2.03 39.30 2.724 3.560 24.49 81.16

D2.B 14.58 0.579 0.560 2.59 53.26 2.009 3.964 24.31 82.66

D2.C 20.39 0.818 0.700 3.03 60.67 2.928 5.392 27.78 85.49

D3.A 20.34 0.529 0.489 1.89 61.95 1.409 2.880 26.42 78.75

D3.B 30.73 0.413 0.459 1.76 62.45 1.139 2.356 26.07 79.47

D4.A 5.53 2.079 2.251 16.30 36.32 9.674 9.812 59.23 88.75

D4.C 11.82 0.564 0.609 2.13 42.82 3.172 4.509 26.29 78.55

D5.A 5.6 0.257 0.825 2.81 30.73 4.183 6.236 29.50 65.48

D5.B 5.84 9.291 7.471 43.61 75.38 16.310 16.440 91.00 98.23

D5.C 7.45 0.000 0.033 0.00 23.30 1.945 4.120 29.05 50.84

D5.D 9.44 0.148 0.035 0.07 21.76 1.802 3.240 25.71 53.93

min 3.54 0.000 0.030 0.00 16.91 1.139 2.356 14.93 50.84

max 49.34 18.622 13.698 64.52 75.38 28.781 22.086 92.35 99.88

mean 15.43 2.770 2.377 12.00 44.13 6.682 7.134 41.93 77.02

median 9.44 0.579 0.768 2.81 44.13 3.172 4.377 26.42 78.75 All time variables are in ns

144


0 5 10 15 20 250

5

10

15

20

25


Mea

n E

xces

s D

elay

(ns)

TEM Horn

W2W4W6D1D2D3D4D5slope=1

0 0.5 1 1.5 2 2.5 3 3.5 40

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2


Mea

n E

xces

s D

elay

(ns)

TEM Horn


0 5 10 15 20 25 300

5

10

15

20

25

30


Mea

n E

xces

s D

elay

(ns)

Biconical Antenna


0 1 2 3 4 5 6 7 8 9 100

1

2

3

4

5

6

7

8

9

10


Mea

n E

xces

s D

elay

(ns)

Biconical Antenna


(b) (a)

(c) (d)

Figure 5.8. Scatter plot for the mean excess delay versus the RMS delay spread, for the TEM horn and the biconical antennas,

(a) TEM horn antenna, all scenarios, (b) TEM horn antenna, zoomed view, (c) Biconical antenna, all scenarios, (d) Biconical antenna, zoomed view.

145


tend to have τστ / =1 which means that the power decays exponentially with time. For

the LOS scenarios the mean excess delay is close to zero, indicating that only the LOS

component is within the specified level of power. The number of dominant multipath

components is limited to two in LOS scenarios. This is consistent with the measurement

carried out at the same location in Durham Hall by [And02b].

Scatter analysis of our UWB measured data indicates that there is no relationship

between delay spread and transmitter-receiver (T-R) separation. This is in agreement with

that reported in [Rap89] and [Sal87] for narrowband systems. On the other hand, when

considering the relation between the received energy and the delay spread, lower energy

signals might seem to have large excess delay. However, this is because the locations

where the received energy is low are usually obstructed and signals arrive at the receiver

through many multipath components. In general received power is not correlated with the

excess delay parameters. In [Rap89] and [Sal87] scatter plots of RMS delay spread versus

pathloss indicate no correlation. The scatter plots relating the RMS delay spread to each

of the T-R separation and the received energy are presented in Figure 5.9.

Comparison with UWB and Narrowband Published Results

In the 5-30 m range, indoor channels are expected to have an RMS delay spread

of 19-47 ns [Fos01] and mean values in the range of 20-30ns [Has93b]. Keignart and

Daniele [Kie02] presented their measurements for a maximum range of 10 m in an indoor

UWB channel. They found that their measured RMS delay spread varies between14 to 18

ns which is a lower than that reported by Hashemi [Has93b]. They also found that the

mean excess delay increases when transmitter/receiver antenna separation increases. The

mean excess delay in their experiment was 4-9 ns for LOS and 17-23 ns for NLOS

scenarios.

146


0 5 10 15 20 250

5

10

15

20

25

30

35

40

45

50


Dis

tanc

e (m

)

TEM Horn

W2W4W6D1D2D3D4D5

0 5 10 15 20 25 300

5

10

15

20

25

30

35

40

45

50


Dis

tanc

e (m

)

Biconical Antenna

W2W4W6D1D2D3D4D5

0 5 10 15 20 25 30-35

-30

-25

-20

-15

-10

-5

0


Pat

hlos

s w

ith re

spec

t to

1m fr

ee s

pace

pat

h lo

ss (d

B)

Biconical Antenna

W2W4W6D1D2D3D4D5

0 5 10 15 20 25-35

-30

-25

-20

-15

-10

-5

0


Pat

hlos

s w

ith re

spec

t to

1m fr

ee s

pace

pat

h lo

ss (d

B)

TEM Horn

W2W4W6D1D2D3D4D5

(a) (b)

(c) (d)

Figure 5.9. Scatter plots to examine the correlation between RMS delay spread and distance/pathloss.

(a) TEM horn antenna, distance versus the RMS delay spread, (b) TEM horn antenna, pathloss versus the RMS delay spread, (c) Biconical antenna, distance versus the RMS delay spread, (d) Biconical antenna, pathloss versus the RMS delay spread.

147


When comparing the results for the published narrowband and UWB propagation

experiments, one has to consider the difference between the used pulse-shape and the

associated frequency spectrum. Time dispersion parameters are functions of the noise

floor, thus without considering the noise power level, time dispersion parameters lose

their significance. For the presented results, we used 10 dB and 20 dB from the maximum

instantaneous signal power. Unless stated otherwise, down to 20 dB below the maximum

instantaneous power is considered.

5.5 Summary and Conclusions

Time-domain measurements were presented for indoor channel characterization.

The performed measurements have high resolution thus suitable for developing accurate

UWB communication channel models. The high-resolution pulses used in these

measurements are good candidates for small cells scenarios, such as single-cell-per-room

where few obstructions exist. Directive TEM horn antennas were compared with the

omnidirectional biconical antennas. Site-specific trends and general observations were

also discussed. Some statistical analyses of the measured data were presented and

compared with the previously published UWB and narrowband results. These

measurements and their corresponding statistical analysis clarified the immunity of UWB

signal to multipath fading compared with the narrowband signals. The calculated pathloss

exponent was as low as 1.27 for a narrow corridor. For LOS and NLOS scenarios the

global pathloss exponents were found to be nearly 1.6 and 2.7, respectively. The

calculated time dispersion parameters for the measured results indicate high

concentration of power at low excess time delays.

Combining the results of penetration loss presented in a previous chapter with the

results of pathloss and time dispersion parameters presented in this report should

facilitate the development better UWB communication channel models. Results might

also prove to be useful for narrowband characterization.

In the next chapter, deconvolution is applied to extract more information about the

UWB channel from the performed measurements.

148

Chapter 6. UWB Channel Model-Deconvolution

“Frequently, knowledge from electrophysics will be employed to formulate a model which

is consistent with physical fact (not mathematical fancy)”

Norris S. Nahman [Nah78]

Chapter 6

UWB Channel Model-Deconvolution*

6.1 Introduction

Before UWB impulse radio can be implemented for indoor applications, the UWB indoor

channel must be accurately characterized and modeled. The importance of accurate channel

characterization cannot be underestimated. The early measurement attempts reported in the

literature extend the narrowband measurement scenarios to the UWB case. Both the approach

and the results need to be verified. For narrowband characterization, usually no deconvolution is

needed and the excitation signal is assumed to be close to an ideal Dirac-delta impulse which

means that the received signal can approximate the impulse response. For narrowband channels,

deconvolution was only used when super-resolutions were required [Vau99], [Mor98].

Deconvolution is most needed for the characterization of wideband devices and channels due to

the limited bandwidths of available test signals as compared to the bandwidths of devices and

channels themselves [Par83]. Since the channel under study is wideband, deconvolution

techniques are needed to estimate the UWB channel impulse response. Moreover, with

deconvolution the estimated channel impulse response is independent of the excitation signal,

which allows for the simulation of different waveforms for wave-shaping studies.

* [Muq02d]

149


In the previous chapter some non-model characterization were presented, namely: signal

quality, pathloss exponent, and time dispersion parameters. In this Chapter, a modified model is

proposed for the UWB channel. It is based on the fact that UWB antennas result in different

impulse response depending on the angle of transmission and angle of arrival. Model-

deconvolution is then used to estimate the number of significant multipath components. The

results are compared with the usual technique where the delay axis is discretized into bins of τ∆

seconds. The selection of the bin size should be equivalent to the measurement resolution

[Has93a] [Has93b]. A multipath component with magnitude ia is said to exist at ττ ∆= ii , if

the integrated power within the ith delay bin interval of the received signal exceeds the

minimum detectable signal threshold. The dominant paths are the paths with the largest

amplitudes [Win97c]. If the number of dominant paths components is small, e.g. 5, ray tracing

can be used as modeling technique. However, one has to employ statistical analysis as the

number of dominant paths increases and ray tracing becomes more complicated and site-specific

[Win97c].

2ia

In section 2 the deconvolution problem is formalized. The incentive for the proposed

multi-template model deconvolution is given through experimental means. Improvements and

results of applying the multi-template deconvolution algorithm on estimating the number of

multipath components and the energy associated with them are presented in section 3

6.2 Deconvolution

Channels can be characterized by their transfer function in the frequency domain or by

their impulse response in the time domain. The measurements under investigation are conducted

in the time domain. Deconvolution of the time-domain waveforms can be used to determine the

impulse response of a linear time-invariant system. The indoor channel is assumed to be time-

invariant if the transmitter and the receiver are static and no motions take place in the channel. If

h(t) is the impulse response of such a channel whose input is x(t), then the output y(t) is given by

the convolution integral,

∫+∞

∞−

−⋅=∗= τττ dthxthtxty )()()()()( , (6.1)

150


where * denotes the convolution operation. In the frequency domain, convolution transforms into

multiplication as follows,

)()()( ωωω jHjXjY ⋅= , (6.2)

where Y(jω), X(jω) and H(jω) are the frequency-domain representations of y(t), x(t), and h(t),

respectively.

The process of obtaining h(t) knowing both x(t) and y(t) is called deconvolution. Ideally,

deconvolution can also be performed in the frequency domain using the Fourier transform. Thus,

from (6.2)

)(/)()( ωωω jXjYjH = . (6.3)

Due to measurement and signal processing limitations, simple division will result in noise-like

error around the zeros of X(jω). Filtering should be used to improve the estimation of the

impulse response [Ria86].

An example of a typical channel profile, y(t), using source#2 and the TEM horn antenna

(Antenna#1) is shown in Figure 6.1. Let x(t) be the received line-of-sight reference gated-pulse

using source#2 and antenna#1 presented in Chapter 3. The received profile, as shown in Figure

6.1, is not a simple summation of delayed pulses. It is evident from the received profile that

multipath components have different waveforms compared with the reference pulse. Though, the

transfer function and the impulse response give full channel description, only few parameters can

be used by the receiver for channel estimation. Model deconvolution is usually used to

characterize the channel with few parameters [Nah81].

6.2.1 Model Deconvolution

Recall that the impulse response of the propagation channel is often modeled as a

summation of effective scatterers,

∑ −=k

kk tath )()( τδ (6.4)

151


where ak is the magnitudes of the kth scatterers. The model in (6.4) is widely used and can

adequately represent the channel for many narrowband communication purposes. This model

does not perfectly fit the UWB channel because the delta function at the receiver implies an

infinite channel bandwidth, which is not possible or acceptable approximation. To make the

model more accurate, the reference pulse used is the convolution of the sounding pulse with the

impulse response of the transmitter antenna, receiver antenna, and the sampling oscilloscope.

This reference pulse is measured in a well-behaved channel where the multipath reflections can

be gated out, as shown in Figure 6.2. The transmitter and the receiver antennas are facing each

other with a distance that guarantees far field reception for the antenna under use. Both the

transmitted and received pulses are presented in Chapter 3.

Though this technique is widely used, the assumption that the received pulses through

different paths have the same waveform is not justified. This assumption requires that both the

transmitter and the receiver antennas have isotropic radiation patterns at all frequencies. It was

noted in [Cra02] that if the antenna is electrically large compared to the wavelength of the center

frequency of the received signal, the waveforms radiated in different directions from the

transmitter antenna look considerably different in the far field region. In Chapter 3, it was also

demonstrated that the signal received at different angles, have considerably different waveforms.

Another illustrative experiment is performed with the two antennas directed towards a reflecting

surface (floor). The setup for this experiment and the received waveforms are displayed in Figure

6.3. For the setup shown in the figure, the waveform associated with the direct path is totally

different from the reference waveform.

5 1 0 1 5 2 0 2 5-0 . 0 8

-0 . 0 6

-0 . 0 4

-0 . 0 2

0

0 . 0 2

0 . 0 4

0 . 0 6

0 . 0 8

T im e (n s )

Am

plitu

de (V

)

Figure 6.1. Typical received LOS multipath profile (Whittemore 2nd floor Hallway)

152


153

Tx Horn Rx Horn Ideal Channel

Sampling O’Scope

Pulse Gen.

x(t)

h(t) Tx Horn Rx Horn

Sampling O’Scope

Pulse Gen.

y(t)

Figure 6.2. Illustration of the measurements of an ideal channel and a multipath indoor channel

RRX x

ceiling

floor

Tx

3 4 5 6 7 8-0 .03

-0 .02

-0 .01

0

0 .01

0 .02

0 .03

0 .04

0 .05

Tim e ns

Am

plitu

de in

V

Figure 6.3. (a) Setup and (b) received waveform with both transmitter and receiver antennas pointing to the reflection surface.


On the other hand, the reflection from the floor, which has the same angle of arrival and

transmission relative to the antennas, has the same shape as the reference waveform. The

reference waveform is reproduced as an inset of the plot for comparison.

6.2.2 Multi-Template Model-Deconvolution

Based on the previous experiments, it is evident that the assumption that multipath

components have shapes similar to that of the reference line-of-sight template is far from being

valid. The same conclusion can be extended to other practical antennas. With this perception, the

model can be modified to allow for more than one received pulse waveform. The proposed

model is antenna specific and is given by

∑ −=k

kj

k thath )(~)( τ (6.5)

where is the impulse response of a system, whose output is the jth template jh~ jp~ , when excited

by the line-of-sight pulse 1~p . When the received signal is the line-of-sight, corresponds to jh~

)(~1h = tδ . Assuming k different templates, the subtractive deconvolution algorithm is modified

from that described in [Vau99] as follows:

1. Initialize the dirty map with the received waveform r(t), d(t)=r(t) and the clean map

with c(t)=0;

2. Form the correlation coefficient functions (normalization is

understood and means correlation) for j=1,2,…k;

)()(~)( tdtp jj Θ=Γ τ

Θ

3. Find the peaks (max , j=1,2,…k ), and their positions, jiΓ iτ , in the ; )(τjΓ

4. If all Γ < threshold, go to step 8; ji

5. Clean the dirty map by inserting zeros in place of the detected multipath component;

6. Update the clean map by using . )()()( ijj

i thtctc τ−Γ+=

7. Go to step 2;

8. The impulse response is then . )()(ˆ tcth =

154


Note that in step (5) updating the dirty map is done by inserting zeros in place of the

detected component rather than updating it by replacing d(t) with )(~)( ijj

i tptd τ−Γ− as in

[Vau99]. This inherently assumes that multipath components do not overlap. This assumption is

justified by the dispersive nature of the channel. If the dirty map is updated as in [Vau99],

invalid multipath components will be produced as a result around the previously detected

components. As a result of zeroing the window of the detected component, not all the energy

can be captured. In the next section, the modified model deconvolution is applied to indoor

channel measurements presented in Chapter 5.

6.3 Results and Analysis

The energy in the multipath profiles from the measurement campaign presented in

Chapter 5 is now captured using a correlator with a fixed template and compared with that

obtained using the proposed multi-template correlator. For optimal selection of the receiver

templates a full antenna characterization should be performed. Unfortunately, antennas were

usually characterized in the frequency domain and for small frequency bands. The TEM horns

used in the presented measurements were characterized in Chapter 3. Research on developing

UWB antenna characterization is under development. In this section, we present a simple

experiment for characterizing the received pulse to illustrate the idea. Figure 6.4 illustrates the

experiment. The separation between the two antennas is 3 m and the distance, d, shown in Figure

6.4 is increased in steps of 20 cm which result in changing the elevation angle (E-scan). The

antennas are then rotated for H-scan where the azimuth angle changes with d. For our purpose of

experimenting the model-deconvolution, the reference templates are based on antenna

measurements at different elevation angles because the directivity of the antenna cause the more

reflections to result from the floor and the ceilings. Figure 6.5 presents the improvement in the

captured energy versus the number of captured multipath components. Different traces are

shown for different number of templates. For the case of single template, the reference template

was measured at d=0. For the case of two templates, the reference templates were measured at

d=0 and d=120 cm. For the case of four templates, the two templates and their inversion are

considered because reflection from different objects at different angles causes the received pulse

to change sign. The choices of templates were not fully optimized but rather were based on the

155


0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

-0.1

-0.05

0

0.05

0.1

time (ns)

volta

ge (v

)

d=0d=20d=40d=60d=80d=100d=120

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2

-0.1

-0.05

0

0.05

0.1

0.15H scan

time (ns)

volta

ge (v

)

d=0d=20d=40d=60d=80d=100d=120

(b)

(a)

d

d increases

d increases

0.15E scan

(c)

Tx-Rx separation 3m

Figure 6.4. Different received waveforms at locations when scanning on the E-plane and the H-

plane (a) received waveforms with E-scan (b) received waveform with H-scan, and

(c) experimental setup (antennas are rotated 90 degrees for the H-scan)

156


0 2 4 6 8 10 12 14 16 18 20

20

30

40

50

60

70

80

90

Number of Single-Path Correlators

% E

nerg

y C

aptu

reLOS, TEM Horn

1 Template2 Templates4 Templates

0 2 4 6 8 10 12 14 16 18 20

20

30

40

50

60

70

80

90

Number of Bins

% E

nerg

y C

aptu

re

LOS, TEM Horn

100 ps200 ps300 ps

0 2 4 6 8 10 12 14 16 18 20

20

30

40

50

60

70

80

90

Number of Single-Path Correlators

% E

nerg

y C

aptu

re

NLOS, TEM Horn

1 Template2 Templates4 Templates

0 2 4 6 8 10 12 14 16 18 20

20

30

40

50

60

70

80

90

Number of Bins

% E

nerg

y C

aptu

reNLOS, TEM Horn

100 ps200 ps300 ps

(d) (c)

(b) (a)

Figure 6.5. Improvement in captured energy

(a) versus number of multipath bins for different bin sizes (LOS), (b) versus number of single-path correlator for different number of templates. (LOS), (c) versus number of multipath bins for different bin sizes (NLOS), (d) versus number of single-path correlator for different number of templates

(NLOS).

157


shape of the reference waveform to allow capturing more energy. All templates are normalized

to have the same energy.

As depicted in Figure 6.5, using two templates resulted in more than 10% increase in the

captured energy for the LOS scenarios as opposed to 5% for the NLOS scenarios. As the

number of templates is increased to four, the energy capture improved by more that 10% again in

the LOS case. In the NLOS scenario less gain is achieved by increasing the number of templates

as the pulse shape undergoes significant change and the energy is distributed through many

components. With 20 single-path correlators, the performance saturated with four templates at

about 72% of the received energy. The limit in the maximum captured energy is a direct result of

the assumption that multipath components are not allowed to overlap.

Using the traditional technique of discretizing the delay access into bins and evaluating

the energy in every bin, similar conclusions can be drawn. Three different bin sizes are

presented because the duration of the pulse is different based on the path. Figure 6.5c and 6.5d

illustrate the percentage energy capture for bin size of 100 ps, 200 ps, and 300 ps. By comparing

the results when using the correlator receiver and the traditional technique, one can asses the

performance of the correlator receivers compared to the total energy capture. It is also important

to note that the plots in Figure 6.5 have sharp peaks at the first five bins or correlators which

suggest that rake receiver need not have a complexity more than 5. Ray tracing is usually a good

candidate for channel modeling when the number of dominant multipath components is small as

in the presented results

6.4 Summary and Conclusions

In this chapter, a modified model was presented based on experimental results which

illustrates that UWB multipath components may have dramatically different waveforms at

different angles relative to the transmitter and receiver antennas. A multi-template UWB

propagation model was proposed to account for the received components at different angles.

Subtractive deconvolution was modified and used to extract the model parameters from

measured channel profiles. The resultant impulse response is antenna-specific. It was shown that

the captured energy increases by more than 10% when using two reference waveforms. For rake

receiver design only 5 correlators would be necessary to capture the energy without high

158


complexity. NLOS pulses undergo dramatic changes and adding more reference template does

not increase the captured energy significantly. Ray tracing is a good technique for modeling

LOS scenarios as the dominant paths are less than 5. Further extension of this work can include

optimizing the choices of reference templates based on extensive antenna measurements.

159

Chapter 7. UWB Multi-User Detection

“Multi-access noise has considerable structure, and certainly much less randomness

than white Gaussian background noise. By exploiting that structure, multi-user detection can

increase spectral efficiency, receiver sensitivity, and the number of users the system can

sustain.”

Sergio Verdú [Ver00]

Chapter 7 UWB Multi-User Detection* 7.1 Introduction

In multiple access systems, interference caused by users, who share the same channel,

limits the system capacity. Multiuser detection is a valuable technique that can increase the

system capacity with a complexity trade off. In multi-user detection techniques, the receiver

demodulates all users rather than a single user. Making a joint decision improves the

reliability of the detection process. Thus, multi-user detection can increase the system

capacity. The objective of this chapter is to evaluate the application of multi-user detection

to ultr-wideband (UWB) communication, commonly known as impulse radio. Impulse radio

technology seems to be promising and it has some potential for multi-user applications in

indoor wireless especially with static terminals [Ver00].

Most of the research in the UWB receiver design area is based on conventional

detectors where other users are assumed to have the Gaussian noise form [Sch93], [Win98b],

[Win98b], [Win00]. Capacity estimates and performance evaluation is based on this

assumption. The research presented in this chapter is an effort to exploit the performed

channel characterization to get more insight on multiple access interference and methods to

mitigate them. Different multi-user detection techniques are considered to improve the

performance and increase the capacity.

* [Muq02a]

160


In section 2, different sources of multi-user interference on light of the measurement

results are presented. An incentive for multi-user detection is given through an experimental

view of multiple access interference. The system model used for simulation is given for

proper assessment. Some multi-user detection techniques are then tested with impulse radio.

Simulation results are analyzed and conclusions are drawn with an assessment to the validity

of the Gaussian approximation. The advantages of using multi-user detection techniques are

illustrated through comparison with the conventional detection and the single user bound.

7.2 UWB Multi-User Interference

Although every user is assigned a distinct time slot, multi-users can still interfere with

each other. Multi-user interference is caused by the dispersive nature of the channel,

multipath components, ringing effects, time jitter, and non-orthogonal code assignment.

Pulse dispersion occurs when the pulse propagate through or reflect from different materials.

Pulse dispersion can take place due to the nature of the material or due to multiple reflections

inside the structure that causes extended channel response as presented in Chapter 4. A more

significant source of multi-user interference is the delayed multipath components. In channel

measurements presented in Chapter 5 and Chapter 6 it was shown that there are about five

dominant delayed pulses and the energy spreads with an RMS delay spread of up to 22 ns for

NLOS scenarios. To exploit the capacity of the channel one has to transmit before the other

users multipath components die out and mitigate the interference.

Another source of interference is the reflection as a result of incomplete matching in

the design of the transmitter, which is illustrated by the acquired signals in Figure 7.1. This

non-ideality cannot be avoided. A completely time-limited signal cannot be generated

especially at very high frequency. Reflections of the original pulse are modified version of

the original pulse. Figure 7.1b shows the case when two pulses in two consequent time slots

have the same power. In this case, the two pulses can easily be distinguished. Figure 7.1c

shows the case when the two consequent pulses have different power as in the near-far effect.

In the illustrated case the first user has amplitude which is nearly six times greater than the

161


0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45-1

-0.8

-0.6

-0.4

-0.2

0

0.2

0.4

0.6

0.8

1

time (nano-seconds)

User 1 A=1User 2 A=1super position

0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45-6

-4

-2

0

2

4

6

time (nano-seconds)

User 1 A=6User 2 A=1

0.21 0.22 0.23 0.24 0.25 0.26 0.27 0.28 0.29

-1

-0.5

0

0.5

1

time (nano-seconds)

User 1 A=6User 2 A=1

0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2-1

0

1

2

3

4

5

6

7

8

time (ns)

Am

plitu

de (V

)

(b) (a)

(c) (d)

Figure 7.1. Illustrative effect of multi-user interference as a result of high frequency ringing

(a) Generated Gaussian-Like pulse and internal reflections

(b) Scenario I : Equal power

(c) Scenario II: User 1 amplitude equal to 6* user 2 amplitude

(d) Zoomed view for case II

162


second user, In this case, the reflection of the first user is very comparable to the pulse from

the second user. This problem is made clearer in the zoomed view in Figure 7.1d. Ringing

effect can also happen in the antenna and transmission media.

Non-orthogonal pseudo-random hopping sequences due to overloaded system or un-

centralized code assignment results in multi-user interference. Moreover, timing is relative to

the receiver. At certain locations pulses from different users can arrive at the same time.

Time jitter and errors in timing are all causes of multi-users interference.

For time hoping multiple access applications this could limit the capacity and the

performance of the system. Knowledge of the interfering pulses could improve the

performance. This motivates the application of multi-user detection techniques.

7.3 System Model

The multiple access system model presented in Chapter 2 is used here. Recall that the

typical hopping format for impulse radio with pulse position modulation is given by

( ) ( ) (∑∞

−∞=

−−−=m

kNmc

kmf

ktr

kktr s

dTcmTtwts )(/

)()()( δ ) (7.1)

The assumed channel model is that Nu users are active during transmission. Signal undergoes

constant amplitude attenuations and waveform deformation. Pulse position modulation

(PPM) is used with bits of 1 delayed by 0.156 ns. Starting with a source that generates

monocycle pulses, a typical received waveform is shown in Figure 2.6 and is given by

])/(41[)35.0( 2mrec ttw τπ−=+ with 2877.0 =mτ (7.2)

When the number of users is Nu, the received signal is:

( ) [ ]∑ ∑=

∞

−∞=

+−−−−=u

s

N

k j

kNjc

kjfkrec utnudTucjTutwts

1

)(/

)( ),())()()(( δτ (7.3)

163


A single bit of information is generally spread over multiple monocycles. The

receiver sums the proper number of pulses to recover the transmitted information. The

receiver is based on correlating the received impulse with the template signal shown in

Figure 2.6. The template signal is the difference between the pulse that represents an

information bit=1 and the pulse used for an information bit=0.

7.4 Performance Evaluation

Most of the research conducted in the area of impulse radio is based on the

assumption of Gaussian noise approximation to the multiple access interference. The signal

to noise ratio (SNR) for the first user when invoking the Gaussian approximation is given by:

( )

∑=

+=

uN

kkasn

psu

AN

mANNSNR

2

222

21)(σσ

(7.4)

where

∫ ∫∞

∞−

∞

∞−

−

−= dsdttvstwT recfa

2

12 )()(σ , (7.5)

and mp is the output when a single impulse is correlated with the template signal. The

numerator of the SNR expression is basically the useful power in the signal, which is related

to the amplitude A, and the number of pulses per bit Ns. The second expression in the

denominator is the approximation of the multiple access interference to Gaussian noise.

Single user bound can be found by setting this expression to zero. v(t) is the template signal

used by the correlator and shown in Figure 2.6. The error probability is then given by the

binary phase-shift keying (BPSK) formula with coherence detection is given by [Skl88],

( )SNRQPe = . (7.6)

Both the Gaussian approximation and the single user bound are used as reference

evaluation measures in subsequent analysis.

164


7.5 Simulated System and Parameters

In the following section, the parameters used for the simulation purpose are listed.

The signal used is as given in Figure 2.6. The pseudo-random hopping code is generated

randomly without coordination between different users to account for the correlation between

the users. The receiver is assumed to have access to these codes and some channel

information as needed by some of the detection techniques. Minimum mean square error

(MMSE) detection algorithm is assumed to have the value of Gaussian noise variance and

the amplitudes of the signals from different users are also assumed to be known if needed by

the detection technique.

Two simulation cases were considered one for synchronous and the other for

asynchronous multiple accesses. The asynchronous case corresponds more to the practical

use of impulse radio where pulses cannot be synchronized for mobile users since few

centimeters corresponds to more than one chip time, Tc. For the asynchronous simulation,

different parameters are chosen due to the complexity associated with the designed detection

code. Table 7.1 lists the different parameters for the two cases. In both cases coherence

detection is assumed and delay for all users can be estimated accurately.

Table 7.1. Simulated Models (Coherence detection)

Parameters Synchronous (1) Asynchronous (2)

Tc 1 ns 1 ns

Tf 10 ns 4 ns

Rs 10 Mbit/second 50 Mbit/second

Nu 5 users 5 users

Ns 10 pulse/bit 5 pulse/bit

165


166

The huge number of users that can be accommodated by impulse radio is based on

Gaussian approximation analysis. Before selecting the system to be simulated, we assess the

capacity of the suggested system for simulation to assure that the system is not saturated.

With perfect power control, the number of users that can be supported for a given

data rate (Rs) is represented as a function of the required additional power [Win98b] by

( ) 1101)( )10/(11 +−=∆ ∆−−− Pspecu SNRMPN , (7.7)

where SNRspec is the required SNR for single user to achieve a specific BER and,

sRMM

11

~ −− = (7.8)

2

21~

af

p

Tm

Mσ

=− (7.9)

Taking the limit as the additional power goes to infinite the maximum number of users that

can be accommodated at the given data rate is calculated as

111max += −−

specSNRMN (7.10)

Using capacity analysis given in [Win98b], for the synchronous case with a rate of 10

Mbit/second, the maximum number of users that can be accommodated with BER =10-3, 10-4

and 10-5 is estimated to be 53, 37 and 28 respectively. This gives an approximate idea of the

expected system performance with 5 users. It would be shown later that Gaussian

approximation give a very optimistic estimate.

7.6 Multi-User Detection Schemes

Different multi-user schemes are proposed. For the purpose of evaluation the

performance of multi-user detection with impulse radio, the following algorithms are tested:

decorrelator, MMSE, successive cancellation, 4 stages parallel cancellation, 4 stages parallel

cancellation with decelerator first stage, and conventional detectors. These techniques are

evaluated against the single user bound and the Gaussian approximation. The basic idea

behind multi-user detection is to make a joint decision on the received signal. This is


167

different than the case of conventional detector where the users are demodulated assuming

other users are Gaussian-noise like interferers. The conventional single user and other

implemented single-stage multi-user detection schemes are demonstrated for two users in

[Ver98]. Multi-stage multi-user detection schemes are illustrated in Figure 7.2.

The decorrelating detector is a simple detector based on decorelating the contribution

of other users before making the decision. The decorrelating receiver does not require

knowledge of the received power. However, it requires inversion of the correlation matrix

which could be singular and computational extensive. A better performance is achieved with

the MMSE detector where knowledge of the received power and noise variance is exploited.

In successive cancellation a decision is made about the demodulated user. A signal

corresponding to this decision is subtracted from the received signal before proceeding with

the next user. Equal power users are treated un-equally as users decoded last tend to have

higher probability of correct decision. The order of the canceled users greatly affects the

performance of successive cancellation. To overcome this shortcoming, parallel cancellation

is introduced where successive cancellation is applied to all users in a symmetric way.

Multistage receiver can be used to improve the average performance. A four-stage parallel

detector is demonstrated in Figure 7.2a. A simpler version with decorrelator first stage is

shown in Figure 7.2b. Further discussion of these algorithms can be found in [Ver98].

Single user is used as an optimum bound. Performance in the existence of multi-

users cannot be better than the case of a single user if the users are sending independent

information. For the case of multi-stage cancellation (Parallel Cancellation) four stages were

used to examine the effect of multi-stages processing.

7.7 Simulation Results

In the following section, simulation results for the two cases under consideration are

discussed. First we start with the equal power case and then the unequal power.

When synchronous case is simulated with equal power users the performance of

different users is shown in Figure 7.3. It is apparent that, though decorrelator based detectors

have bad performance at low SNR, they improve sharply at higher SNR. When the


168

∫T

0x

∫T

0x

1b

2b

1y

2y

)(1 ts

)(2 ts

)(ty x ∫T

0

+

+

+

+

-

-

ρ x

x

1A

2A

+ +

+ +

-

-

x

x

1A

2A

+ +

+ +

-

-

x

x

1A

2Aρ ρ

∫T

0x

∫T

0x

1b

2b

1y

2y

)(1 ts

)(2 ts

)(ty x ∫T

0

+

+

+

+

-

-

ρ x

x

+ +

+ +

-

-

x

x

1A

2A

++

++

-

-

x

x

1A

2A ρ ρ

(a) four-stages parallel cancellation

(b) four-stages parallel cancellation with decorrelator in the first stage

Figure 7.2. Multi-stage multi-user detection schemes for two synchronous users


0 5 10 15 20 25

10-4

10-3

10-2

10-1

100

Bit

Erro

r Rat

e

SNR(dB)

Bit-Error Rate in a five-user channel with equal-power user

Single UserGaussian ApproximationConventional DetectorSuccessive CancellationDecorrelatorMMSEParallel CanellationMultiStage ,Decorrelator first

Figure 7.3. Performance of multi-user detection for equal power synchronous case

169


performance of all users is averaged out successive cancellation perform the worst after the

conventional detector. MMSE seems to perform very well at the cost of added channel

information. It is important to note that Gaussian approximation, which supposed to estimate

the performance of the conventional detector, is very optimistic at least for the considered

case.

Though successive cancellation performed the worst when the performance of all

users is averaged out, it can result in a promising performance for some specific users

cancelled at later stages. This means unequal performance for equal power users. This is

recommended when the receiver is not interested in all users which could be the case for

impulse radio receivers. Figure 7.4 illustrates the pronounceable different in the performance

of the five detected users.

The effect of multi-stages on both the parallel cancellation with conventional first

stage and with decorrelator first was also studied. The results indicate that more than two

stages do not add much to the performance especially with hard decisions made at the

previous stages.

The more practical asynchronous case is considered next. For this case only parallel

cancellation and successive cancellation are examined. Parallel cancellation outperforms the

successive canceller for equal power case as shown in Figure 7.5.

An effort was made to evaluate the performance with the use of the decorrelator

however the singularity of the solution made the detector more difficult to simulate. It is

important to note that whenever decorrelator is used one has to switch to conventional

detector when the correlation matrix cannot be inverted or the inversion is close to singular.

The determinant of the correlation matrix can be compared to a specific value and decision

on singularity should be made as a result. This value is critical to the performance of the

decorrelator detectors and has to be optimized for better performance

170


0 5 10 15 20 25 10 -4

10 -3

10 -2

10 -1

10 0

Bit E

rror R

ate

SNR (dB)

Improvement in Performance with Successive Cancellation

Last cancelled=5 432First Cancelled=1

Figure 7.4. Unequal performance for equal power user with successive cancellation

171


0 5 10 15

10-4

10-3

10-2

10-1

100

Bit

Erro

r Rat

e

SNR(dB)

Bit-Error Rate in a five-user channel Asynchronous with equal-power users

Single UserGaussian ApproximationConventional DetectorSuccessive CancellationParallel Canellation

Figure 7.5. Multi-user detection performance for equal-power asynchronous case

172


Unequal power was also consider with the interferers having power levels as -3 dB, -

4 dB, -5 dB and -6 dB relative to the first user. The performance of the asynchronous case is

shown in Figure 7.6a and 7.6b. The first plot shows the performance for the first user and the

second plot shows the average performance. An important fact is revealed by comparing the

two plots that multi-user detection might not be recommended when the strongest user is the

only desired user as in the studied example.

7.8 Summary and Conclusive Remarks

Multi-user detection was shown to have a potential for application with impulse radio

multiple access technology. In fact the simple model that assumes pulses arrive at distinct

time slots is not practical due to multipath components, time jitter, non-orthogonal time

hopping codes, ringing, and high frequency switching.

The output of the simulation work results in emphasizing that the selection of the

proper multi-user detection technique is based on the operation region in terms of SNR and

BER. Moreover, for applications where all users have equal importance, as in base stations,

the parallel cancellation technique is a good candidate. Successive cancellation works well

when only single user is of interest, as in mobile units if the delay and complexity factors

could be tolerated. Decorrelator detectors have a bad performance at low SNR. However,

they exhibit a very sharp waterfall like performance curve as the SNR is increased.

Gaussian approximation is shown to be over optimistic for impulse radio, which

utilizes very large bandwidth. Though similar results are published for narrowband systems

The performed work opens the doors for further investigations for multi-user detection

applied to impulse radio especially at high data rate or loaded systems.

173


0 5 10 15

10-5

10-4

10-3

10-2

10-1

Bit

Erro

r Rat

e

SNR(dB)

Bit-Error Rate in a five-user channel Asynchronous (First User A=1)

Single UserGaussian ApproximationConventional DetectorSuccessive CancellationParallel Canellation

(a) Performance for the strongest users

2 4 6 8 10 12 14

10-2

10-1

Bit

Erro

r Rat

e

SNR(dB)

Bit-Error Rate in a five-user channel Asynchronous (Average)

Conventional DetectorSuccessive CancellationParallel Canellation

(b) Averaged performance

Figure 7.6. Performance of multi-user detection for asynchronous unequal power case

174

Chapter 8. Conclusions

“Each advantage offered by UWB is offset by a disadvantage, the cure for which is

another disadvantage. The engineering goal remains –as always- balanced optimization”

Terence W. Barrett [Bar01]

Chapter 8 Conclusions 8.1 Summary of Findings

The proposed UWB communication holds great promises with radical departure from

traditional narrowband systems. Before UWB communication is materialized, the

characteristics of propagation medium have to be well understood. At a fundamental level,

we have studied the basic differences between UWB and narrowband communications. Both

frequency-domain and time-domain methods were used in order to exploit the advantages of

each method and also have means of cross checking the measured results. Measurements

were performed using two types of pulse generators producing different pulse shapes and

three pairs of antennas, two pairs of TEM horns as directional antennas and one pair of

biconical antennas as omnidirectional radiators.

Electromagnetic characterization of materials and walls commonly encountered in

indoor environments was undertaken with the aim of assessing their impacts on UWB

propagation. Different analysis techniques for extracting material parameters; namely, loss

coefficient and delay, were reviewed. Ten commonly used building materials were selected

for investigation. These include dry wall, wallboard, structure wood, glass sheet, bricks,

concrete blocks, reinforced concrete (in a pillar form), cloth office partition, wooden door,

and styrofoam slab. The characterization was based on measuring an insertion transfer

175


function for each material. The insertion transfer function was obtained through both

frequency-domain measurements using a vector network analyzer, and time-domain

measurements using a pulse generator and a sampling oscilloscope. The loss and the

dielectric constant of the selected materials were measured over a frequency range of 1 GHz

up to 15 GHz [Muq03c], [Muq03d]. The results can serve as a basis for further studies in

developing appropriate models for UWB channels. They are also very useful in link budget

analysis. In this investigation, also the relevance to pulse shaping, receiver design, and

channel modeling were highlighted.

A new formulation for evaluating the complex dielectric constant of low-loss

materials was presented. This formulation involves solving real equations and thus requiring

only one-dimensional root search techniques. The results derived from the exact complex

equation and the new formulation are in excellent agreement. The new formulation reduces

the computation time significantly and is highly accurate for the characterization of low-loss

materials [Muq03a] [Muq03b].

After successfully carrying out the ultra-wideband characterization of building

materials, time-domain characterization of typical indoor channels were performed. Cases

such as line-of-sight (LOS), non-line-of-sight (NLOS) topographies, room-to-room, within-

the-room and hallways were studied. The measurements were carried out in two buildings

on Virginia Tech Campus: Whittemore Hall and Durham Hall. Results governing the

pathloss exponent and excess delay parameters were presented for two types of antennas.

The first antenna is a directive TEM horn and the second is an omnidirectional biconical

antenna. The performed measurements have high resolutions, thus suitable for developing

accurate UWB communication channel models. The high-resolution pulses used for the

experiments are appropriate candidates for small cells scenarios, such as single-cell-per-room

where few obstructions exist. Results obtained by directive TEM horn antennas were

compared to those obtained by the omnidirectional biconical antennas. Site-specific trends

and general observations were also discussed. A statistical analysis of the measurements was

presented and compared with the previously published UWB and narrowband results. These

measurements and their statistical analysis clarified the immunity of UWB signals to

multipath fading compared with the narrowband signals. The calculated pathloss exponent

176


was as low as 1.27 for narrow corridors when directive antennas were used. For LOS and

NLOS scenarios the global pathloss exponents were found to be nearly 1.6 and 2.7,

respectively. The calculated time dispersion parameters for the measured results indicate

high concentration of power at low excess time delays. Possible correlations between

different parameters were examined. Results might also prove useful for narrowband

channel characterization [Muq03e] [Muq03f].

A further step was taken by investigating deconvolution technique to extract more

information about the channel. Multipath components have different waveforms depending

on the type of transmitter and receiver antennas used and the angles of transmission and

reception. Combining the results of insertion loss with the results of pathloss and time

dispersion parameters can help one develop better UWB communication models. A

modified deconvolution model was introduced to extract the UWB channel response.

Subtractive deconvolution was modified and used to extract the model parameters from

measured channel profiles [Muq02d]. It is necessary to note that the resultant impulse

response is antenna-specific. It was shown that the captured energy increases by more than

10% when using two reference waveforms. For designing a rake receiver only five

correlators were found to be necessary to capture the energy of the pulse without high

complexity. NLOS pulses undergo dramatic changes and adding more reference templates do

not increase the captured energy significantly. Ray tracing is a suitable technique for

modeling LOS scenarios as the dominant paths are less than five. The application of

deconvolution techniques resulted in resolving multipath components with waveforms

different from that of the sounding pulse. Resolving more components should improve the

design of the rake receiver. Accurate characterization for the impulse response of a UWB

communication system facilitates performance evaluation studies such as simulating the

effect of pulse shaping.

The research work is concluded with an illustrative example of incorporating

measurement insight into improving receiver design. So far, receiver designed for multiple

access ultra-wideband communication systems known as impulse radio is based on

conventional single-user matched filter detectors. Here, we illustrated the nature of the

multiple access interference based on the performed experiments. Multi-user detection was

177


shown to have potential applications in impulse radio multiple access technology. In fact,

the simple model that assumes pulses arrive at distinct time slots is not practical due to issues

related to multipath components, time jitter, non-orthogonal time hopping codes, ringing, and

high frequency switching. The results for the simulation put an emphasis on the selection of

a proper multi-user detection technique based on the operation region in terms of SNR and

BER. Moreover, for applications where all users have equal importance, the parallel

cancellation technique is a good candidate. Successive cancellation works well when only

single user is of interest and the delay and complexity factors could be tolerated. Decorrelator

detectors, though perform badly at low SNR, exhibit a very sharp waterfall like performance

curve as the SNR is increased. Gaussian approximation was shown to be over optimistic for

impulse radio. Though similar results have been published for narrowband systems, the

performed work holds the key to further investigations for multi-user detection applied to

impulse radio especially at high data rate or loaded systems [Muq02a].

8.2 Suggestions for Further Research

It would be very interesting to develop frequency-domain measurement techniques

for characterization of UWB channels involving distances in the 50-meter range. The

limitation on the use of vector network analyzers may be solved with the use of optical links.

It would be then interesting to compare indoor channel measurements in frequency domain

with the presented time-domain measurements, as we have performed in the material

characterization section. There are other research issues to be dealt with including the

limitation on the number of frequency-domain measurements and the resolution that can be

achieved. New developments in designing UWB antennas seem to be a major factor in the

success of UWB communications.

It would also be interesting to extend the measurement campaign to a wider variety of

materials and building layouts, containing a larger number of partition types and using

different pulses. Based on the measurement presented and the transient perspective gained

from this research, different reliable statistical and deterministic models can be developed.

Also, UWB site-specific simulators can be developed based on ray-tracing concepts.

178


More advanced communication characterization rather than channel characterization

should consider the mobility of the transmitter and the receiver. It should involve measuring

instantaneous signals as well as the average power. Spectrum analyzers should be used as

supportive equipment. Measurements should be supported by simulation. Interference of

narrowband systems on the UWB communication link should be investigated for accurate

system performance evaluation.

Further extension of the deconvolution model would include optimization among the

choices of reference templates based on extensive antenna measurements. On the multi-user

detection research, experimenting with multiple transmitters is recommended for a better

understanding of the multiple access interference and then using the obtained parameters to

perform simulation with real channel rather than assuming a fixed received reference

waveform. Some of the suggestions for research developments are already being initiated in

both the Time Domain Laboratory and the Mobile and Portable Radio Research Group at

Virginia Tech. Among such suggestions are developing extended distance frequency domain

setup, site specific ray-tracing and statistical modeling.

179

Appendix A

Appendix A

180

Appendix A1 Single-Pass Time-Duration-Limited Transient Measurement Derivation

A1 Single-Pass Time-Duration-Limited Transient Measurement Derivation

This appendix is dedicated for presenting the derivation of the short-pulse propagation

measurements. The analysis presented here follows closely the analysis given by [Aur96]. A

short duration electromagnetic pulse, Ei(t) is applied to a homogenous, isotropic material layer of

thickness d. The transmitted signal, Et2(t) is measured at the other side of the material. To

simplify the problem we assume normal incidence to the material surface and we assume that the

duration of the pulse is less than the transient time through the material. Multiple reflections

inside the layer as a result of the finite-sized layer can be eliminated. Same technique can be used

to eliminate antenna ringing and extraneous paths. In summary, this is a single-pass time-

duration-limited transient measurement procedure which assumes 1-D model of plane-wave

propagation through a planer layer.

To get more insight into the problem, the lattice or bounce diagram is shown in Figure

A2-1. Note that this is not steady state harmonic analysis which include the internal ringing. Two

signals are measured:

• the transmitted ‘through’ pulse, Et2(t), with the layer in place; and • a free-space reference pulse, ,which is the received wave without the layer. )(2 tE fs

t

The two measurements should be done with exactly the same distance and antenna setup.

The free space measurement is used as a reference to account for all the effects, which are not

due to the material under test, for example, the antennas and the receiver.

The partial reflection coefficient at the first boundary is defined as the ratio of the first

reflection, Er(t), to the incident pulse, Ei(t):

)()()( 1

fEtEf

i

r≡Γ (A1-1)

Like the TEM transmission line theory, the partial reflection coefficient in terms of the

wave impedance in the layer and the free-space wave impedance is given by:

181


)(1)(1

)()()(

0

0

ff

fff

re

re

εε

ηηηη

+

−=

+−

=Γ (A1-2)

In the same way the wave transmitted across the first boundary, Et1(f), is determined by

the partial coefficient across the boundary, )()()( 11 fEfEf itT ≡ . From the boundary condition:

, The wave incident to the second boundary, E)(1)(1 ffT Γ+= i2(f), is related to the wave

transmitted through the first boundary by . The added complex

exponential factor is due to the propagation through the material. The partial reflection

coefficient at the rear boundary is defined to be

df )(γ−ti efEfE 12 )()( =

)()(2 ffi Γ−=)()( 22 EfEf r≡Γ , and the

transmission coefficient is given by

)(1)(1)()()( 2222 fffEfEfT it Γ−=Γ+=≡ (A1-3)

The emerging pulse, Et2(f), has spectrum given by

. The transfer function in terms of the first

reflection coefficient is given by

)()()( 1212222 fEeTTeETETfE idd

titγγ −− ===

ddd

i

t eeeTTfEfE γγγ −−− Γ−=Γ+Γ−== )1()1)(1()()( 2

122 (A1-4)

Similar measurements are made in free space without the layer, . The free space

measurements contain the non-idealities in the source, antennas and receiver. Assuming an

imaginary layer of free space then the propagation is through the air passed through two

boundaries of transmission coefficient of 1. This is approximated by a propagation delay through

the imaginary layer given by:

)(2 fE fst

0)1)(1()()(

122 ωτωγ jcdjd

i

fst eeeTT

fEfE −−− === (A1-5)

where cd≡0τ , where d is the layer thickness and c is the speed of light in free space.

Finally the insertion transfer function is defined as

182


dfs

t

t efEfE

fH )(2

2

2 0)1()()(

)( γγ −−Γ−== (A1-6)

The insertion transfer function is formed as the ratio between the two radiation transfer

functions,

)()(

)()(

)()(

)(2

2

2

2

fEfE

fEfE

fEifE

fH fst

t

i

fst

t

=≡ (A1-7)

This is similar to studying the effect of shielding. In linear system analysis this model is a

two-port representation of lossy, causal network, and it requires measurements of scattering

parameter S21(f).

Substituting Equations (A1-2) & (4.4) for the complex propagation constant and the

reflection coefficient, we obtain the layer transfer in terms of the relative complex effective

permittivity as

( ))1(

20

1

4)( −−

+= rej

re

re efH εωτ

ε

ε (A1-8)

The last equation needs to be transcendentally solved to get the relative complex effective

permittivity from the measured transfer function.

In summary, we measure the received pulse with no layer and with the material,

E

)(2 fE fst

t2(f). FFT is used to convert the sampled signal to the frequency domain equivalent. Care must

be taken to make sure that the free space measurement is similar to the measurement through the

material. Numerical methods are used based on the previous equations to extract the unknown

parameters.

In time domain, we may estimate the delay between the two pulses to get an estimate

value for the dielectric constant. The total signal power can also be measured in the free space

case and through the material to quantify the loss in power when propagating through the

material.

183


Low-Loss Analysis Method

The requirement of iterative solution may be relaxed if the material is assumed to be low

loss. Equations (A1-6) and (A1-8) can be written as two equations if the transmission mismatch

at the two boundaries are approximately real valued. That is the loss tangent , then it can

be dropped out of the square roots:

1<<ep

( ) [ ] [ ] ( )2222

121

4

)1(1

)1(4

)(1

)(41

r

r

er

er

re

re

jp

jp

f

fTT

ε

ε

ε

ε

ε

ε

+≅

−+

−=

+=Γ−= (A1-9)

The resulting expression is real-valued. [Aur96] to give an idea of how good is this

approximation, assume very poor condition very lossy, but lower dielectric constant for such

loss): 10=rε and pe=0.2. The approximate calculation for T2T1 is 0.7301, which is only 0.3%

greater than the magnitude of the exact value T . With only 3-degree phase error,

this approximation is acceptable compared with the accuracy of the time-domain hardware.

o37277.012 ∠=T

The second simplification is done to the exponential part of (A1-6) and (A1-8) in order to

make the phase and the attenuation constant separable. The exponent is given by

[ ]1)()()( 000 −−−+−−− == fjdjjd reeee εωτββαγγ (A1-10)

The Taylor-series expansion for small loss tangent is given by:

......81

2111 422 +−+≈+ eee ppp (A1-11)

For the attenuation constant we retain the first two terms,

[ ]c

pp

cf er

er

211

2)(

21

2 εωεωα ≅

−+= (A1-12)

For the phase constant only one term is kept, yielding

[ ]c

pc

f re

r εωεωβ ≅

++=

21

2 112

)( (A-13)

184


Substituting the simplified constants into the exponent

[ ]1)()()( 0 −−−−+− ≅ fjdfdjj reee εωταββα .

Inserting the two simplifications into the original transfer function result in a low-loss

version of the layer transfer function:

( )( )1

20

1

4)( −−−

+≅ rjd

r

r eefH εωτα

ε

ε (A1-14)

The low-loss layer transfer function has separable magnitude and phase components as

follows:

( )([ 12

1

2)( 0

0

1

4)()( −−−Φ

+≅= r

er jp

r

rfj eeefHfH εωτεωτ

ε

ε ) ] (A1-15)

where (f) is the phase component given by the second term. Again to examine the

validity of the combined approximation, consider the previous scenario with

Φ

10=rε , pe=0.2,

and 01 τ=f (like 10 =τ ns, and f=1 GHz). The approximate formula yields a magnitude of

0.7230 compared to the exact value of 0.7277; the error is 0.6%. Again compared with the

accuracy of the hardware, this accuracy is adequate.

In summary, using the low-loss analysis to characterize the material under test requires

two steps. First, the dielectric constant is determined from the differential delay time, τ∆ , of the

layer transfer function: [ ]1)()( 0 −≡∆ ff rεττ . The phase shift versus frequency is given by

)(2)( ff τπ∆−≅Φ , and the derivative is given by )(2 fdfd τπ∆−≅Φ . The differential time delay

is then written as df

fdf )(21)( Φ

−≅π

τ∆ . Solving for the dielectric constant, we have

2

0

2

0

)(2

11)(1)(

Φ−=

∆+≅

dffdffr πττ

τε (A1-16)

185


The second step is to calculate the effective loss tangent from the result of (A1-15) and

the magnitude of the measured transfer function. Loss tangent is given by

[ ]

+−≅ )(

)(4)(1

ln)(

1)(2

0

fHff

fffp

r

r

re ε

εετπ

(A1-17)

The previous formulation was tested by [Aur96] and works will for dielectric constants in

the range 2.5-8; and loss tangent <0.2.

186

Appendix A2. Multi-Pass Complex Dielectric Constant Equation

A2. Multi-Pass Complex Dielectric Constant Equation

The derivation of the complex dielectric constant equation based on a multi-pass analysis

is presented in this appendix. Assuming a uniform plane-wave normally incident on an infinite

material slab, the partial reflection coefficient at the first boundary, denoted as ρ , is given by

12

12

ηηηη

ρ+−

= (A 2-1)

where η1 and η2 are the intrinsic impedances of air (essentially free space) and the material of the

slab under test, respectively. The transmission coefficient at the first boundary is obtained from

the relationship 1τ =1+ ρ . At the second boundary, when propagation is in the direction of the

material toward air, the partial reflection coefficient is equal to − ρ , while the partial

transmission coefficient is 2τ =1− ρ . Thus, the first partial transmitted wave through the slab is

T 1τ 2τ Ei =T(1- 2ρ )Ei, where Ei is the incident field and

T (A 2-2) de γ−=

accounts for propagation through the slab thickness, with γ being the complex propagation constant of the slab. Using the bounce diagram shown in Figure A2-1, the overall transmission coefficient through the slab, which is the same as the scattering parameter , is obtained from 21S

( )( ) ( )22

244222

21 1111

TTTTTS

ρρρρρ

−−

=⋅⋅⋅+++−= (A 2-3)

In case of the free-space measurements 0=fsρ , and S21 is given by

dja

fs aeTS β==21 (A 2-4)

Relating the two measurements, one can write the insertion transfer function as

( ) ( )( )22

2

21

21

11

TTT

SSjH

afs ρ

ρω−−

== (A 2-5)

By substituting (A2-1) and (A2-2) into (A2-5) one can write

02)cosh(2)sinh(1

21

=−+

+

SxPxP

xx (A 2-6)

187

Appendix A2. Multi-Pass Complex Dielectric Constant Equation

where rx ε= and cdfjP π2= , which is the equation used for multi-pass complex search.

( )21 ρ−T

η1 η2 η1

( )232 1 ρρ −T

( ) 423 1 Tρρ −−

deT γ−= , 12

12

ηηηη

ρ+−

=

1

ρ

( ) 221 Tρρ −−

Figure A2-1. Bounce diagram for propagation through a slab

188

Appendix A3. Proof of Equation (4.42) being Valid with Negative Sign

A3. Proof of Equation (4.42) being Valid with Negative Sign

Here, it is proved that the solution for X with a negative sign in front of the square root is

the only valid solution. Whenever a solution exists, we must have

( ) ( ) 01)(

81)2cos( 4

2

22 >−′−

′+−′ r

rr

jHd ε

ωε

εβ (A 3-1)

(A3-1) is rewritten as

( ) ( ) ( ) ( ) 01)(

81)2cos(1

)(8

1)2cos( 22

222

2 >

−′+

′+−′⋅

−′−

′+−′ r

rrr

rr

jHd

jHd ε

ωε

εβεωε

εβ

(A 3-2)

The second square bracket in (A3-2) is always positive, because

( ) ( ) ( ) ( ) 0)(

8)2cos(111

)(8

1)2cos( 222

22 >

′++−′=

−′+

′+−′

ωε

βεεωε

εβjH

djH

d rrr

rr , (A 3-3)

Thus, the first square bracket should be positive too. That is,

( ) 01 2 >−′rε , ( 0)2cos(1 >+ d )β , and 0)(

82 >

′

ωεjH

r .

Thus,

( ) ( ) 01)(

81)2cos( 22

2 >−′−′

+−′ rr

rjH

d εωε

εβ

or

( ) ( ) 01)(

81)2cos( 22

2 >−′>′

+−′ rr

rjH

d εωε

εβ (A 3-4)

Hence, the solution for X with a negative sign in front of the square root is always >0, i.e.,

189


( ) ( ) ( )

( ) 01

1)(

81)2cos()(

81)2cos(

4

4

2

22

22

>−′

−′−

′+−′−

′+−′

=r

rr

rr

rjH

djH

d

Xε

εω

εεβ

ω

εεβ

(A 3-5)

On the other hand, X should be less than 1 (otherwise instead of attenuation we have

amplification). Now, we show that the solution for X with positive sign in front of the square

root is greater than 1 and thus not acceptable

( ) ( ) ( )

( ) 11

1)(

81)2cos()(

81)2cos(?

4

4

2

22

22

>−′

−′−

′+−′+

′+−′

r

rr

rr

rjH

djH

d

ε

εω

εεβ

ω

εεβ

(A 3-6)

Using the condition in (A3-4) in (A3-6), we have

( ) ( ) ( )

( ) 11

1)(

81)2cos(1?

4

4

2

222

>−′

−′−

′+−′+−′

r

rr

rrjH

d

ε

εω

εεβε

(A 3-7)

or

( )( )

( ) ( )

( ) 11

1)(

81)2cos(

1

1 ?

4

4

2

22

4

2

>−′

−′−

′+−′

+−′

−′

r

rr

r

r

rjH

d

ε

εω

εεβ

ε

ε (A 3-8)

The second term is positive so it sufficient to prove that

( ) 11

1 ?2

2 >

−′

−′

r

r

ε

ε (A 3-9)

190


Or

( ) 11

1 ?

2 >−′

−′

r

r

ε

ε (A 3-10)

which leads to rrr εεε ′−+′>−′ 211?

or which is obviously true. Consequently, the

solution of X with “+” sign in front of the square root is not valid. The correct solution is the one

with the “-” sign in front of the square root.

1?>′rε

191

Appendix B

Appendix B

192

Appendix B1. Antennas Configuration and Structure

B1. Antennas Configuration and Structure

In the conducted experiment three pairs of antennas were used. Two antennas are

based on TEM horn structure and the third pair has a biconical design. The second TEM

horn pair is wider in bandwidth than the first TEM horn pair. Using different antennas

enables variability study and shed some light on the importance of the antenna to the

UWB system.

Antenna #1 (TEM Horn) Structure and Design

For impulsive free-space measurements a TEM horn is suggested by [Law78].

TEM horns are quite broadband in receiving mode, both in magnitude and phase. The

suggested antenna was reproduced and tested in the Time Domain and RF Measurements

Laboratory at Virginia Tech.

The antenna shown in Figure B1.1 was made by cutting the proper triangle on a

standard 1/16” glass epoxy printed circuit board. Holes drilled on the ground plane are

used to fasten the antenna with nylon bolts. The wedge spacer was cut from polystyrene

foam. The structure can be regarded as two half-horn ground plane models placed back-

to-back. The ground plane is longer than the antenna, and has the same width as the

antenna printed boards. Ferrite beads around the coax and absorbing foam can be used to

damp the reflections on the ground plane that propagate on the outer surface of the coax

lines.

The antenna is driven through a balun which splits the unbalanced signal into two

identical outputs of opposite polarity. A better scenario is to use a generator that

produces the two balanced signals directly. In the receiver side, the output of the two

half-horn can be combined directly using a balun. Still a better scenario is to look at the

two outputs directly on two channels of the sampling oscilloscope (A and B), and since

they are of opposite polarity, they can be combined as A-B. [Law78]

193


The balun is a simple broadband power divider with one output delayed and

inverted while the other is merely delayed. It is mounted in aluminum utility box. The

extra space is filled with foam to reduce any probable internal reflections.

A picture of the antenna is displayed in Figure B1.2a. The antenna is shown inside

anechoic chamber where the radiation pattern was measured at different frequencies. The

radiation pattern at 5 GHz is presented in Figure B1.2b. Both the electric field and the

magnetic field patterns illustrate the directive nature of the antenna. The asymmetry in

the radiation pattern in the E-plane is due to the holding structure.

Antenna #2 (TEM Horn array) Structure and Design

The second pair of antennas has a broader band (500MHz– 10GHz) and TEM can

be assumed at sufficiently far distances. Each antenna consists of an array of two smaller

horns. The dimensions and details are shown in Figure B1.3. 180-degree phase shifters

are used at the input of the transmitting antenna and at the output of the receiving

antenna. The phase shifter (Krytar 4010124) has a frequency range of 1-12.4 GHz. The

amplitude imbalance is 0.4 dB while the phase imbalance is 10 degrees. Resonant effects

are reduced by resistive loading. The resistance increases exponentially with distance

(Approach: semi-distributed Resistive, strip array). The resistive loads are printed using

thin film technology.

Biconical Antennas

Lately, an omnidirectional wideband biconical antenna was made available by the

Mobile and Portable Radio Research Group (MPRG) at Virginia Tech. The antenna is

displayed in Figure B1.4. The antenna is not designed for impulsive measurements;

however, it has a wideband that extends from 0.1-18 GHz. The antenna has a height of

15 cm inches and a diameter of about 31 cm. The normalized antenna patterns for the H-

plane (azimuth angle) and the E-plane (elevation angle) at 2.5 GHz are displayed in

Figure B1.4b.

194


5 cm

20 cm46 cm

36 cm

semi-rigid coax

30 cm copper cladding 10 cm 21.3

cm

microwave absorbing material polystyrene foam

coax center conductor soldered to apex of copper foil triangle

1/16” glass epoxy printed circuit board

5 cm

5 cm

double-sided board for ground plane

outer conductors soldered to ground plane and each other

Figure B1.1. Schematics and dimensions of the TEM horn antenna (antenna#1)

195


(a)

(b)

Figure B1.2. TEM horn antenna#1 in anechoic chamber and measure pattern at 5 GHz

(a) TEM horn antenna#1 in anechoic chamber (b) Electric and magnetic fields radiation pattern at 5 GHz

(Measurement is done at VTAG: Virginia Tech Antenna Group by R. Nealy)

196


Figure B1.3. Schematics and dimensions for the TEM horn antenna array (antenna#2)

∆ ∑

Upper Antenna (shown tilted for clarification)

Lower Antenna

Stripline Feed

Phase Shifter

3 inch

12 inch.

Top View

17.4o

3.714 inch Side View

w

h

lx lh

3 in

5 in

15 in

197


(a) Biconical Antenna

(b) Normalized antenna patterns for the 2.5 GHz. Azimuth (left) and elevation (Right) angles. [And02b]

Figure B1.4. Picture and pattern at 2.5 GHz for the omnidirectional biconical antenna

198

Appendix B2. Material Pictures

B2. Material Pictures

Figure B2-1. Pictures for the bricks, blocks, styrofoam and walls built out of them

Figure B2-2. Pictures for the wallboard, door, wood, structure wood, cloth office

partitions, glass, and reinforced concrete pillars

199


(

(e) mov

w = 8.5354h = 5.8267b = 4.1503d2 = 2.159

d

Figure B

a) bricks moving wall

(b) blocks moving wall

(c) single brick (d) single block

ing bricks wall between two

Styrofoam

(f) Styrofoam (two slabs are shown)

a = 12.2 cm b = 12.5 cm c = 4.8 cm d = 3.7 cm e = 3.2 cm

e

d

c

a

b

2 cm l = 19.8 cm 6 cm a = 3.5179 cm 6 cm d1 = 1.905 cm cm

l w

h1 d2

a b

2.1. Pictures for the bricks, blocks, styrofoam and walls built out of them

200


(a) wallboard (two are shown)

(b) door

(c) wood

(d) structure wood

(e) cloth office partition

(f) glass

(g) concrete pillar (TDL)

(h) reinforced concrete pillar (Whittemore)

201

Figure B2.2. Pictures for the wallboard, door, wood, structure wood, cloth office partitions, glass, and reinforced concrete pillars

Appendix B3. Blueprints and Photos for Measurement Locations

B3. Blueprints and Photos for Measurement Locations

Figure B3.1. Whittemore blueprints to illustrate measurement locations and environments

Figure B3.2. Whittemore site photos

Figure B3.3. Measurement locations for Durham Hall

Table B3.1. Measurement locations and scenarios (reproduced)

202


(a)

Transmitter

RxA RxB

Tx Rx

(c) (b) Figure B3.1. Whittemore blueprints to illustrate measurement locations and environments

(squares represent transmitter locations, circles represent receiver locations)

(a) Hallways in the 2nd floor. (b) Corridor in the 4th floor. (c) Conference room in the 6th floor.

203


(c)

(b)(a)

Figure B3.2. Whittemore site photos

(a) Hallways in the 2nd floor. (b) Corridor in the 4th floor.

204(c) Conference room in the 6th floor.


)

Figure B3.3. Me rement locations for Durham Hall (square present transmitter loca

) )

(a) (b) (c)

asu(b

tions, circles represent receiver locations)

Blueprint for the fourth floor of Durham Photo for location 4C with cubical partitiHallways in Durham Hall, location 2B.

s re(c

(a

Hall ons

205


Table B3.1. Measurement locations and scenarios

# Location Description

# profiles (TEM

)

# profiles (B

iconical) d

W2.A Hallways in 2nd floor LOS, Hallways 9 9 9.3 W2.B Hallways in 2nd floor LOS, Hallways 9 9 17.6

W2.C Hallways in 2nd floor NLOS, Hallways 0 0 -

W2.D Hallways in 2nd floor NLOS, Hallways 0 0 -

W2.E Hallways in 2nd floor LOS, Hallways 9 9 15.4

W2.F Hallways in 2nd floor LOS, Hallways 9 9 30.9

W2.G Hallways in 2nd floor LOS, Hallways 9 9 49.1

W2.H Hallways in 2nd floor NLOS, Hallways 0 0 -

W2.I Hallways in 2nd floor NLOS, Hallways 0 0 -

W4 Corridor in the 4th floor LOS, Small corridor 12 25 Varying

W6.A Conference Room in 6th floor LOS, Within a room 9 9 5.1

W6.B Conference Room in 6th floor LOS, Within a room 9 9 6.2

D1.A Tx (Room 475)-Rx (Room 471) NLOS, Room-to-Room 9 9 3.6

D1.B Tx (Room 475)-Rx (Room 471) NLOS, Room-to-Room 9 9 9.0

D2.A Hallway in 4th floor LOS, Hallways with concrete walls 9 9 8.2

D2.B Hallway in 4th floor LOS, Hallways with concrete walls 9 9 14.6

D2.C Hallway in 4th floor LOS, Hallways with concrete walls 9 9 20.4

D3.A Hallway+ open environment Hallway with open space in the middle 9 9 20.3

D3.B Hallway+ open environment Hallway with open space in the middle 9 9 30.7

D4.A Tx (Room 476)- Rx (internal room) NLOS, Room-to-Room 9 9 5.5

D4.B Tx (Room 476, MPRG Lab.) Cubical office Environment (Obstructed) 0 0 -

D4.C Tx (Room 476, MPRG Lab.) LOS, Cubical office Environment 9 9 11.8

D5.A Tx (Room 423, MPRG Reception), Rx (Room 433)

LOS/NLOS , Room-to-Room 3 3 5.6

D5.B Tx (Room 423, MPRG Reception), Rx (Hallway)

Room-to-Hallway, (obstructed, NLOS) 9 9 5.9

D5.C Tx (Room 423, MPRG Reception), Rx (Hallway)


D5.D Tx (Room 423, MPRG Reception), Rx (Hallway)


W: Whittemore Hall , D: Durham Hall

206

References

207

References

[Ada01] J. C. Adam, W. Gregorwich, L. Capots, and D. Liccardo, “Ultra-wideband for navigation and communications,” IEEE Proc. Aerospace Conf., vol. 2, 2001, pp. 2/785 -2/792.

[Age98] F . J. Agee, C. E. Baum, W. D. Prather, W.D., J. M. Lehr, J. P. O'Loughlin, J. W. Burger, J. S. H. Schoenberg, D. W. Scholfield, R. J. Torres, J. P. Hull, and J. A. Gaudet, “Ultra-wideband transmitter research,” IEEE Trans. Plasma Science, vol. 26, pp. 860-873, Jun. 1998.

[Alq96] I. L. Al-Qadi, and S. M. Riad, “Characterization of Portland Cement Concrete: Electromagnetic & Ultrasonic Measurement Techniques”, Virginia Polytechnic Institute and State University, submitted to National Science Foundation, September 1996.

[And02a] C. R. Anderson, T. S. Rappaport, K. Bae, A. Verstak, N. Ramakrishnan, W. Tranter, C. Shaffer, and L. Watson, “In-Building Wideband Multipath Characteristics at 2.5 & 60 GHz,” Proceedings of IEEE 56th Vehicular Technology Conference, vol.1, 2002. pp. 97-101.

[And02b] Christopher R. Anderson, Design and Implementation of an Ultrabroadband Millimeter-Wavelength Vector Sliding Correlator Channel Sounder and In-Building Multipath Measurements at 2.5 & 60 GHz, Thesis Submitted to the Faculty of the Virginia Polytechnic Institute and State University, May 2002.

[Ang99] Amara D. Angelica, “Powered by Pulse, More than an Pipe Dream,” TechWeek , May.1999

http://www.timedomain.com/Files/PDF/news/TechWeek.pdf

[Ast97] L. Y. Astanin and A. A. Kostylev, Ultrawideband Radar Measurements, Analysis and Processing, The institution of Electrical Engineers, London, United Kingdom, 1997.

[Aur96] F. J. Aurand, “Measurements of Transient Electromagnetic Propagation Through Concrete and Sand”. Sandia National Laboratories, Sandia Report SAND96-2254 UC-706, Livermore, California 94550, Sep. 1996.

[Bak90] J. Baker-Jarvis, E. J. Vanzura, and W. A Kissck, “Improved Technique for Determining Complex Improved Technique for Determining Complex Permittivity with the Transmission/Reflection Method,” IEEE Transactions on Microwave Theory and Techniques, vol. 38 , no. 8, pp. 1096-1103, August 1990.

208

http://www.timedomain.com/Files/PDF/news/TechWeek.pdf

[Bak98] J. Baker-Jarvis, R. G. Geyer, J. H. Grosvenor Jr., M. D. Janecic, C. A. Jones, B. Riddle, C. M. Weil, and J. Krupka, “Dielectric Characterization of Low-loss Materials: A Comparison of Techniques,” IEEE Transactions on Dielectric and Electrical Insulation, vol. 5, no. 4, pp. 571-577, August 1998.

[Bar] Terence Barrett, “History of Ultra Wideband (UWB) Radar & Communications: Pioneers and Innovators”

http://www.multispectral.com/history.html

[Bar01] Terence Barrett, “ History of Ultra Wideband Communications and Radar: Part I, UWB Communications,” Microwave Journal, Jan. 2001.

[Bee] W. Beeler, “UWB Measurement Techniques and Issues” Technical report, Time Domain Corporation.

http://www.ero.dk/EROWEB/SRD/UWB/Time%20Domain/UWB%20Measurement.pdf

[Ben90] A. Bennia and S. M. Riad, “Effect of noise and bandwidth on homomorphic deconvolution,” Conference on Precision Electromagnetic Measurements, CPEM '90 Digest., 1990 , pp. 224 –225.

[Bor99] A. Boryssenko, “Time domain studies of ultra-wide band antennas,” IEEE Canadian Conf. Electrical and Computer Engineering, vol. 1, 1999, pp. 95-100.

[Cas01] D. Cassioli, M. Win, and A. Molisch, “A Statistical Model for the UWB Indoor Channel,” IEEE Conference on Vehicular Technology, 53rd VTC 2001, vol. 2, 2001, pp. 1159-1163.

[Cas02] D. Cassioli, M. Win, A. Molisch, “The Ultra-Wide Bandwidth Indoor Channel: From Statistical Model to Simulations,” IEEE Journal on selected areas in Communications, vol. 20, no. 6, pp. 1247-1257, August 2002.

[Cha02] S. Chalillou, D. Helal, and C Cattaneo, “Timed Simulator for UWB Communication Systems” IEEE Conference on Ultra Wideband Systems and Technologies, May 2002, pp. 6-11.

[Cra98] R.J.-M. Cramer, M.Z. Win, and R. A. Scholtz, “Impulse radio multipath characteristics and diversity reception,” IEEE Int. Conf. on Comm. , Conference Record, vol. 3, 1998, pp. 1650-1654.

[Cra99] J.M. Cramer, R.A. Scholtz, and M.Z. Win, “On the analysis of UWB communication channels,” in IEEE Proc. Military Communications Conf., vol. 2, 1999, pp. 1191-1195.

209


http://www.ero.dk/EROWEB/SRD/UWB/Time Domain/UWB Measurement.pdf

http://www.ero.dk/EROWEB/SRD/UWB/Time Domain/UWB Measurement.pdf

[Cra02] Cramer R., Scholtz R., and Win M. “Evaluation of an Ultra-wide-Band Propagation Channel,” IEEE Transactions on Antennas and Propagation, vol. 50, no. 5, pp. 561-570, May 2002.

[Cui00] I. Cuinas and M. Sanchez, “Building Material Characterization from Complex Transmissivity Measurements at 5.8GHz,” IEEE Transaction on Antennas and Propagation, vol. 48, no. 8, pp. 1269-1271, August 2000.

[Dan96] D.J. Daniels, Surface-Penetrating Radar, The institution of Electrical Engineers, London, United Kingdom, 1996.

[Dav91] R. Davies, M. Bensebti, M. A. Beach ,a dn J. P. McGeehan, “Wireless Propagation Measurements in indoor multipath environments at 1.7GHz and 60GHz for Small Cell Systems” Proceedings of IEEE 41st Vehicular Technology Conference, 19-22 May 1991, pp. 589 –593.

[Dep] R. DePierre, “Analysis of the Impact of UWB Emissions on a 1.9 GHz CDMA PCS System,” Technical Report, Time Domain Corporation.

http://www.ero.dk/EROWEB/SRD/UWB/Time%20Domain/UWB%20&%20IS-95.pdf

[Dic99] D. Dickson and P. Jett, “An application specific integrated circuit implementation of a multiple correlator for UWB radio applications,” in IEEE Proc. Military Communications Conf., vol. 2, 1999, pp. 1207-1210.

[Dur98] G. Durgin, T.S. Rappaport, and H. Xu, “Measurements and Models for Radio Path Loss and Penetration Loss in and Around Homes and Trees at 5.85 GHz”, IEEE Transactions on Communications, vol. 46, no. 11, pp. 1484-1496, Nov. 1998.

[Dur00] G. D. Durgin, and T. S. Rappaport, “Theory of Multipath Shape Factors for Small-Scale Fading Wireless Channels,” IEEE Transactions on Antennas and Propagation, vol. 48, no. 5, pp.682-693, May 2000.

[Fid90] K. M. Fidanboylu, A. Bennia, S. M. Riad, “Effect of measurement bandwidth on time domain device modeling,” Conference on Precision Electromagnetic Measurements, CPEM '90 Digest., 1990, pp. 226 –227.

[Foe01] J. R. Foerster, “The Effects of Multipath Interference on the Performance of UWB Systems in an Indoor Wireless Channel,” IEEE VTS 53rd, vol.2, 2001, pp. 1176-1180.

[Fon] Robert Fontana, “A Brief History of UWB Communications”


210

http://www.ero.dk/EROWEB/SRD/UWB/Time Domain/UWB & IS-95.pdf

http://www.ero.dk/EROWEB/SRD/UWB/Time Domain/UWB & IS-95.pdf


[For00] A. R. Forouzan, M. Nasiri-Kenari, and J.A. Salehi, “Low-rate convolutionally encoded time-hopping spread spectrum multiple access systems,” The 11th IEEE Int. Symposium on Personal, Indoor and Mobile Radio Comm., vol. 2, 2000, pp. 1555-1558.

[Ful91] L. Fullerton, “UWB waveforms and coding for communications and radar,” Proc. National Telesystems Conf., 1991, pp. 139 –141.

[Ful00] L. W. Fullerton, “Reopening the electromagnetic spectrum with ultrawideband radio for aerospace,” IEEE Proc. Aerospace Conf., vol. 1, 2000, pp. 201-210.

[Gey90] R. G. Geyer, “Dielectric Characterization and Reference Materials” National Institute of Standards and Technology (NIST): NIST Technical Note 1338, 1990.

[Gha] Saeed Ghassemzadeh, Chris Rice, and Matthew Sherman, “Characterization of the Ultra-Wideband Channel At 5 GHz”, AT&T presentation

http://www.research.att.com/~saeedg/supercomm2001.pdf

[Gha02] S. Ghassemzadeh, R. Jana, C. Rice, and W. Turin, “A Statistical Path Loss Model for In-Home UWB Channels” IEEE Conference on Ultra Wideband Systems and Technologies, May 2002, pp. 59-64.

[Gib99] T. Gibson and D. Jenn, “Prediction and Measurements of Wall Insertion Loss” IEEE Transactions on Antennas and Propagation, vol. 47, no. 1, pp. 55-57, Jan. 1999.

[Goo91] J. M. Goodman, HF Communications Science and Technology, Van Nostrand Reinhold, New York, 1992.

[Gro00] J. Gronkvist, A. Hansson, and J. Nilsson, “A comparison of access methods for multi-hop ad hoc radio networks,” IEEE 51st Proc. Vehicular Technology, VTC 2000-Spring Tokyo, vol. 2, 2000, pp. 1435 –1439.

[Ham01a] M. Hamalainen, J. Iinatti, V. Hovinen, and M. Latva-aho, “In-band interference of three kind of UWB signals in GPS L1 band and GSM900 uplink band Personal,” 12th IEEE Int. Symposium on Indoor and Mobile Radio Comm., vol. 1, Sept. 2001, pp. 76 –80.

[Ham01b] M. Hamalainen, V. Hovinen, J. Iinatti, and M. Latva-aho, “In-band interference power caused by different kinds of UWB signals at

211

http://www.research.att.com/~saeedg/supercomm2001.pdf

UMTS/WCDMA frequency bands,” IEEE Conf. on Radio and Wireless, RAWCON’01, 2001, pp. 97-100.

[Har] Henning Harmuth, “An early History of Nonsinusoidal Electromagnetic Technology”


[Har81] Henning F. Harmuth, Nonsinusoidal Waves for Radar and Radio Communication, Academic Press, Inc. 1981.

[Has93a] H. Hashemi “The Indoor Radio Propagation Channel” Proceeding of the

IEEE, vol. 81, no. 7, pp 943-968, July 1993.

[Has93b] H. Hashemi, “Impulse Response Modeling of Indoor Propagation Channels,” IEEE Journal on Selected Areas in Communications, vol. 11, No. 7, Sep. 1993.

[Haw91] D. A. Hawbaker, Indoor wide band radio wave propagation measurements and models at 1.3 GHz and 4.0 GHz, M.S. thesis, Dep. Elec. Eng., Virginia Polytech. Inst. And State Univ., May 15, 1991.

[Hay83] S. Haykin, Communication systems, John Wiley & Sons Inc., 2nd ed, 1983.

[History] Multispectral Solutions, Inc. “History of UWB Technology”, multi authors, see link


[Hov02] V. Hovinen, M. Hämäläinen, and T. Pätsi, “Ultra Wideband Indoor Radio Channel Models: Preliminary Results,” IEEE Conference on Ultra Wideband Systems and Technologies, May 2002, pp. 75-79.

[Intro] “Bandwidth from thin air, Two new ways of transmitting data by wireless exploit unconventional approaches to create valuable additional capacity”, The Economist, Nov. 6, 1999, p. 85.

http://www.timedomain.com/Files/HTML/news/TheEconomist.htm

[Kei02] J. Keignart and N. Daniele, “ Subnanosecond UWB Channel Sounding in Frequency and Temporal Domain,” IEEE Conference on Ultra Wideband Systems and Technologies, May 2002, pp. 25-29.

212



http://www.timedomain.com/Files/HTML/news/TheEconomist.htm

[Kis92] E.C. Kisenwether, “Ultrawideband (UWB) impulse signal detection and processing issues,” Proc. of Tactical Communications Conf., Tactical Communications: Technology in Transition., 1992, p. 87 –93.

[Kis01] William A. Kissick, “The Temporal and Spectral Characteristics of Ultrawideband Signals”, U.S. Department of Commerce, NTIA Report 01-383, January 2001.

[Kon99] G.D. Kondylis, F. De Flaviis, G.J. Pottie, and Y. Rahmat-Samii, “Indoor channel characterization for wireless communications using reduced finite difference time domain (R-FDTD),” IEEE Conf. Vehicular Technology, VTC‘99 - Fall. VTS 50th, vol. 3 , 1999 , pp. 1402 –1406.

[Kra93] D. Kralj and L. Carin, “Ultra-Wideband Characterization of Lossy Materials: Short-Pulse Microwave Measurements” Microwave Symposium Digest, IEEE MTT-S International, vol. 13, 1993, pp.1239 -1242.

[Kun02] J. Kunisch and J. Pamp, “Measurement Results and Modeling Aspects for the UWB Radio Channels,” IEEE Conference on Ultra Wideband Systems and Technologies, May 2002, pp. 19-23.

[Law78] R.A. Lawton and A.R. Ondrejka, “Antennas and Associated Time Domain Range Measurement of Impulsive Fields” U.S. Department of Commerce / National Bureau of Standards, NBS Technical Note 1008, Boulder, Colorado 80303, November 1978.

[Law86] R.A Lawton, S.M. Riad and J.R. Andrews, “Pulse and Time-Domain Measurements,” Proc. of the IEEE, vol. 74, no. 1, Jan. 1986.

[Lee00] Hojoon Lee, Byungchil Han, Yoan Shin, and Sungbin Im, “Multipath characteristics of impulse radio channels,” IEEE 51st Proc. on Vehicular Technology, VTC 2000-Spring Tokyo. 2000, vol. 3 , 2000, pp. 87 –2491.

[Mal90] J. Maloney, G. Smith, and W. Scott, “Accurate computation of the radiation from simple antennas using the finite-difference time-domain method,” IEEE Trans. Antennas Propagat., pp. 1059-1068, July 1990.

[Mar00] C.J. artret and G.B. nnakis, “All-digital PPM impulse radio for multiple-access through frequency-selective multipath,” IEEE Proc. of Array and Multichannel Signal Processing Workshop ‘00, 2000 , p. 22 –26.

[Mat00] J. Matlacz, and K. D.Palmer, “Using offset Parabolic Reflector Antennas for Free Space Material Measurements,” IEEE Transaction on Instrumentation and Measurements, vol. 49, no. 4, August 2000.

213

[Mor98] G. Morrison and M. Fattouche, “Super-Resolution Modeling of the Indoor Radio Propagation Channel,” IEEE Transactions on Vehicular Technology, vol. 47, no. 2, pp. 649-657, May 1998.

[Muq02a] A. Muqaibel, B. Woerner, and S. Riad, "Application of multi-user detection techniques to impulse radio time hopping multiple access systems," IEEE Conference on Ultra Wideband Systems and Technologies, 2002, pp. 169 –174.

[Muq02b] A. Muqaibel, A. Safaai-Jazi, and S. Riad, "Characterization of Dielectric Substrate Material Using Fork Resonator," 2002 IEEE AP-S International Symposium and USNC/URSI National Radio Science Meeting, San Antonio, Texas.

[Muq02c] A. Muqaibel, A. Safaai-Jazi, and S. Riad, "Fork Resonators for High Frequency Characterization of Dielectric Substrate Materials," submitted to the IEEE Transaction on Instrumentation and Measurement.

[Muq02d] A. Muqaibel, A. Safaai-Jazi, B. Woerner, and S. Riad "UWB Channel Impulse Response Characterization Using Deconvolution Techniques," 45th IEEE International Midwest Symposium on Circuits Systems, Tulsa Oklahoma, 2002, August 4-7.

[Muq03a] A. Muqaibel, and A. Safaai-Jazi, “A New Formulation for Characterization of Materials Based on Measured Transfer Function” submitted to the IEEE Transactions on Microwave Theory and Techniques (MTT paper #1384).

[Muq03b] A. Muqaibel, A. Safaai-Jazi, “New Formulation for Evaluating Complex Permittivity of Low-Loss Materials”, submitted to IEEE-APS 2003

[Muq03c] A. Muqaibel, A. Safaai-Jazi, A. Bayram and S. Riad, “UWB Through-the-Wall Propagation and Channel Characterization,” submitted IEEE Transactions on Wireless Communications. (reference number TW03-058).

[Muq03d] A. Muqaibel, A. Safaai-Jazi, A. Bayram and S. Riad, “Ultra Wideband Material Characterization for Indoor Propagation,” submitted to the IEEE-APS 2003.

[Muq03e] A. Muqaibel, B. Woerner, A Safaai-Jazi, and A. Attiya “Pathloss and Time Dispersion Parameters for Indoor UWB propagation” in preparation to be submitted to the IEEE Transactions on Wireless Communications.

[Muq03f] A. Muqaibel, B. Woerner, A Safaai-Jazi, and A. Attiya “Experimental Characterization of Indoor UWB Propagation” in preparation to be submitted to the 2nd IEEE Conference on Ultra Wideband Systems and Technologies, Nov. 2003.

214

[Nah78] N. Nahman, Introduction to Time Domain Measurements and Applications, course notes EE593, The University of Colorado, Boulder, Colorado. Spring 1978.

[Nah81] N. Nahman and M. Guillaume “Deconvolution of Time Domain Waveforms in the Presence of Noise,” National Bureau of Standards, Technical Note 1047, Oct. 1981

[Par83] Parruck B. and Riad S., “An Optimization Criterion for Iterative Deconvolution” IEEE Transaction on Instrumentation and Measurement, vol. IM-32, no.1, Mar. 1983.

[Par00] J. D. Parsons, The mobile Radio Propagation Channel, John Wiley & Sons Ltd, second edition 2000

[Pan99] D.C. Pande, “Ultra wide band (UWB) systems and their implications to electromagnetic environment ,” Proc. of the Int. Conf. on Electromagnetic Interference and Compatibility '99, 1999, pp. 50 –57.

[PSPL-4050A] Model 4050 A Picosecond Step Generator, Instruction Manual, P.O. Box 44, Boulder, Colorado 80306, U.S.A.

[PSPL-4100] Model 4100C & 4600C Pulse Generator, Instruction Manual, P.O. Box 44, Boulder, Colorado 80306, U.S.A.

[PSPL-5320A] Balun Data Sheet, Picosecond Pulse Labs, AN-3048, Revision 1, 11/99.

http://www.picosecond.com/objects/AN-8.pdf

[PSPL-5828] Amplifier Data Sheet, Picosecond Pulse Labs, Spec-4040052, Revision 3, 7/01.


[Poz98] D. M. Pozar, Microwave Engineering, 2nd ed., John Wiley & sons, Inc., 1998.

[Pra99] W. D. Prather, F.J. Agee, C. E. Baum, J. M. Lehr, J.P. O'Loughlin, J. W. Burger, J. S. H. Schoenberg, D. W. Scholfield, R. J. Torres, J. P. Hull, and J. A. Gaudet, J. A. “Ultra-wideband sources and antennas,” Ultra-Wideband Short-Pulse Electromagnetics 4, 1999, pp. 119 –130.

[Pre02] C. Prettie, D. Cheung, L. Rusch, and M. Ho, “Spatial Correlation of UWB in a Home Environment,” IEEE Conference on Ultra Wideband Systems and Technologies, May 2002, pp. 65-69.

[Pro89] J.G. Proakis, Digital Communications, McGraw-Hill Inc., 2nd ed., 1989.

215


[Pro96] J. G. Proakis and D. G. Manolakis, Digital Signal Processing: Principles, Algorithms, and Applications, Prentice-Hall, Inc. , 3rd edition, 1996.

[Pul00] Time Domain Corporation, “PulsON© Time Modulated Ultra-Wideband For Wireless Applications,” May 2000 Rev 2.

http://www.timedomain.com/Files/downloads/techpapers/PulsONoverview.html

[Ram98a] F. Ramirez-Mireles and R.A. Scholtz, “Multiple-access with time hopping and block waveform PPM modulation,” IEEE Int. Conf. on Comm., ICC 98. Conference Record, vol. 2, 1998, pp. 775 -779.

[Ram98b] F. Ramirez-Mireles, M. Z. Win, and R. A. Scholtz, “Performance of ultra-wideband time-shift-modulated signals in the indoor wireless impulse radio channel ,” Conference Record of the Thirty-First Asilomar Conference on Signals, Systems & Computers, vol. 1 , 1997, pp. 192 –196.

[Ram99] F. Ramirez-Mireles, “Performance of ultrawideband SSMA using time hopping and M-ary PPM,” IEEE Journal on Selected Areas in Communications, vol. 19, no. 6 , pp. 1186 –1196, June 2001.

[Rap89] T.S. Rappaport, “Characterization of UHF multipath radio channels in factory buildings ,” IEEE Transactions on Antennas and Propagation , vol. 37, no. 8 , pp. 1058 –1069, Aug. 1989.

[Rap92] T. S. Rappaport and D. A. Hawbaker, “Wide band microwave propagation parameters using circular and linear polarization antennas for indoor wireless channels,” IEEE Trans. Commun., vol. 40, no. 2, pp 1-6, Feb. 1992.

[Rap96] T.S. Rappaport, Wireless Communications, Principles & Practice, Prentice Hall, Inc. 1996.

[Ria86] Riad S., “The Deconvolution Problem: An Overview,” Proceedings of the IEEE, vol. 74, no. 1, pp. 82-85, Jan. 1986.

[Sal87] A. A. Saleh and R. A Valenzuela, “A statistical model for indoor multipath propagation,” IEEE J. Selected Areas Commun., vol. 5, pp.128-137, 1987.

[Sch93] R. A. Scholtz, “Multiple access with time-hopping impulse modulation,” Conference record Military: Communications Conference, Communications on the Move, MILCOM '93, vol.2, 1993, pp. 447 –450.

[Sch97] R. A. Scholtz , and M. Z. Win , “Impulse Radio,” Invited paper , IEEE PIMRC’97- Helsinki, Finland. pp. 245-267.

216

http://www.timedomain.com/Files/downloads/techpapers/PulsONoverview.html

[Sch98] R.A. Scholtz, R.J.-M. Cramer, and M.Z. Win, “Evaluation of the propagation characteristics of ultra-wideband communication channels,” IEEE International Symposium: Antennas and Propagation Society, vol. 2 , 1998, pp. 626 –630.

[Sch99] "Answers to questions posed by Bob Lucky, Chairman of the FCC's technical advisory committee" an e-mail communication from R. A. Scholtz (June 29, 1999)

http://ultra.usc.edu/New_Site/papers/lucky.pdf

[Sch00] R.A. Scholtz, R. Weaver, E. Homier, J. Lee, P. Hilmes, A. Taha, and R. Wilson ,”UWB radio deployment challenges,” The 11th IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, PIMRC 2000, vol. 1, 2000, pp. 620-625.

[Sch01] H. G. Schantz and L. Fullerton, “ The Diamond Dipole: A Gaussian Impulse Antenna”, IEEE International Symposium Antenna and Propagation Society, vol.4, 2001, pp.100-103.

[Sei94] S. Sediel, and T. Rappaport, “Site-Specific Propagation Prediction for Wireless In-Building Personal Communication System Design,” IEEE Transactions on Vehicular Technology, vol. 43, no. 4,pp. 879-891, nov. 1994.

[Siw01] Siwiak Kazimierz, and Alan Petroff, “A Path Link Model for Ultra Wide Band Pulse Transmissions” IEEE VTC’01, pp.1173-1175.

[Skl88] B. Sklar, Digital Communications Fundamentals and Applications, Prentice Hall, 1988.

[Str01] A. Street, L. Lukama, and D. Edwards, “Use of VNAs for wideband propagation measurements,” IEE Proc.-Commun., vol. 148, no. 6, Dec. 2001.

[Su92] Wansheng Su, Calibration of Time Domain Network Analyzers, Ph.D. Dissertation submitted to the Faculty of the ECE department at Virginia Tech.

[Su96] W. Su, R. Mostafa, S. M. Riad, and I.L. Al-Qadi. 1996, “ Characterization of Concrete Material Using TEM Horn Antennas.” Presented at the URSI National Radio Science Meeting, Boulder, CO, January 11, 1996. Bradley Dept. of Electrical Engineering, Virginia Polytechnic Institute and State University, Blacksburg, VA.

[Sus00] M. Jr. Sushchik, , N. Rulkov, L. Larson, L. Tsimrin, H. Abarbanel, K. Yao, and A. Volkovskii, “Chaotic pulse position modulation: a robust method of communicating with chaos,” IEEE Communications Letters, vol. 4 Issue: 4 , pp. 128 –130, April 2000.

217




[Tol92] A. F. de Toledo and A. M. D. Turkmani, “Propagation into and within Buildings at 900, 1800, and 2300 MHz,” IEEE 42nd Vehicular Technology Conference, vol. 2, 1992, pp. 633-636.

[Tur02] W. Turin, R. Jana, S. Ghassemzadeh, and C. Rice, “Autoregressive Modeling of Indoor UWB Channel,” IEEE Conference on Ultra Wideband Systems and Technologies, May 2002, pp. 71-74.

[Win97a] M.Z. Win and R.A. Scholtz, “Comparisons of analog and digital impulse radio for wireless multiple-access communications,” IEEE International Conference on Communications: Towards the Knowledge Millennium, ICC '97 Montreal, vol. 1 , 1997, pp. 91–95.

[Win97b] M.Z. Win, F. Ramirez-Mireles, R.A. Scholtz, and M.A. Barnes, M.A. “Ultra-wide bandwidth (UWB) signal propagation for outdoor wireless communications,” IEEE 47th Conference on Vehicular Technology, vol. 1 , 1997, pp. 251 –255.

[Win97c] Win M. and Scholtz R., “Energy Capture vs. Correlator Resources in Ultra-Wide Bandwidth Indoor Wireless Communications Channels,” MILCOM '97 Proceedings, vol. 3 , 1997, pp. 1277-1281.

[Win98a] M.Z Win and R.A. Scholtz, “Impulse radio: how it works,” IEEE Communications Letters, vol. 2, no. 2, pp. 36 –38, Feb. 1998.

[Win98b] M. Z. Win, Ultra-Wide Bandwidth Spread-Spectrum Techniques for Wireless Multiple-Access Communications, Ph.D. dissertation, University of southern California, Electrical Engineering, Los Angeles, 1998.

[Win99a] M.Z. Win, Xiaoxin Qiu, R.A. Scholtz, V.O.K. Li, “ATM-based TH-SSMA network for multimedia PCS,” IEEE Journal on Selected Areas in Communications, vol. 17 , no. 5 , pp. 824 –836, May 1999.

[Win99b] M.Z. Win, “Spectral density of random time-hopping spread-spectrum UWB signals with uniform timing jitter,” IEEE Proc. On Military Communications Conference, MILCOM’99, vol. 2 , 1999, pp. 1196 -1200.

[Win00] M.Z Win and R.A. Scholtz, “Ultra-wide bandwidth time-hopping spread-spectrum impulse radio for wireless multiple-access communications,” IEEE Transactions on Communications, vol. 48, no. 4, pp. 679 –689, April 2000.

[Wit] P. Withington, “Impulse radio overview,” Time Domain Corporation.

[Wit99] P. Withington, R. Reinhardt, and R. Stanley, “Preliminary results of an ultra-wideband (impulse) scanning receiver ,” IEEE Conference Proceedings on Military Communications, MILCOM’99, vol. 2 , 1999, pp. 1186 –1190.

218

[Ugu 02] B. Uguen, E. Plouhinec, and G. Ghassay, “A Deterministic Ultra Wideband Channel Modeling,” IEEE Conference on Ultra Wideband Systems and Technologies, May 2002, pp. 1-5.

[UWB] Ultra-Wideband Working Group

http://www.uwb.org

[Vau99] R.G Vaughan and N. Scott, “Super-Resolution of Pulsed Multipath Channels for Delay Spread Characterization,” IEEE Transactions on Communications, vol. 47, no.3, pp. 343-347, Mar. 1999.

[Ver98] S. Verdo, Multiuser Detection, Cambridge University Press, 1998.

[Ver00] S. Verdu, “Wireless bandwidth in the making,” IEEE Communications Magazine , vol. 38 , no. 7, pp. 53 –58, July 2000.

[Yoh01] J. J. Yoho, Physically-Based Realizable Modeling and Network Synthesis of Subscriber Loops Utilized in DSL Technology, Ph.D Dissertation submitted to the ECE dep. at Virginia Tech, 2001.

[Zha94] Y. P. Zhang and Y. Hwang, “Measurements of the Characteristics of Indoor Penetration Loss,” IEEE 44th Vehicular Technology Conference, vol. 3, 1994, pp. 1741-1744.

[Zha01] J. Zhang, M. Nakhash, and Y. Huang, “In-Situ Characterization of Building Materials,” 11th International Conference on Antennas and Propagation, Conference Publication no. 480, April 17-20, 2001, pp. 269-274.

[Zha02] H. Zhang, T. Udagawa, T. Arita, and M. Nakagawa, “A Statistical Model for the Small-Scale Multipath Fading Characteristics of Ultra Wideband Indoor Channel,” IEEE Conference on Ultra Wideband Systems and Technologies, May 2002, pp. 81-85.

[Zhu02] F. Zhu, Z. Wu, and C Nassar, “Generalized Fading Channel Model with Application to UWB” IEEE Conference on Ultra Wideband Systems and Technologies, May 2002, pp. 13-17.

219

http://www.uwb.org/

Vita

Ali Muqaibel was born on June 6, 1974 in Dammam, Saudi Arabia. In 1996, he

was awarded Bachelor of Science degree with highest honors from King Fahd University

of Petroleum and Minerals (KFUPM). In 1999, he received his Master of Science degree

in Electrical Engineering from KFUPM with a focus on communication systems. In

2000, Ali began work towards a Doctor of Philosophy degree in Electrical Engineering

with the Time Domain and RF Measurements Laboratory (TDL). He also joined the

Mobile and Portable Radio Research Group (MPRG) at Virginia Tech.

Mr. Muqaibel has worked for three years as an electrical engineer and one year as

a lecturer at KFUPM. While at Virginia Tech, he participated in research as well as

teaching. He was awarded over 40 honor certificates through his academic life and

career.

Mr. Muqaibel is a member of IEEE, Ultra Wideband (UWB) Working Group, and

IEEE Communication Society. His research interest includes: channel characterization,

UWB communication, time domain and RF measurements, and channel coding. He has

co-authored many technical reports and over 20 conference and journal papers.

220

characterization of ultra wideband communication channels

Documents