Sub-NyquistSamplingImpulseRadioUWBReceivers fortheInternet …1045523/... · 2016. 11. 9. · end...

Sub-Nyquist Sampling Impulse Radio UWB Receiversfor the Internet-of-Things

QIN ZHOU

Doctoral Thesis in Information and Communication TechnologySchool of Information and Communication Technology

KTH Royal Institute of TechnologyStockholm, Sweden 2016

TRITA-ICT 2016:23ISBN 978-91-7729-174-9

KTH School of Information andCommunication Technology

SE-164 40 KistaSWEDEN

Akademisk avhandling som med tillstånd av Kungl Tekniska högskolan framläggestill offentlig granskning för avläggande av teknologie doktorsexamen i Informations-och kommunikationsteknik måndagen den 12 december 2016 klockan 09.00 i Sal 205,Electrum, Kungl Tekniska högskolan, Kistagången 16, kista.

© Qin Zhou, October 2016

Tryck: Universitetsservice US AB

iii

Abstract

In the era of Internet-of-Things, the demand for short range wirelesslinks featured by low-power and low-cost, robust communication and high-precision positioning is growing rapidly. Impulse Radio Ultra-Wideband (IR-UWB) technology characterized by the transmission of sub-nanosecond pulsesspanning up to several GHz band with extremely low power spectral densityemerges as a promising candidate. Nevertheless, several challenges must beconfronted in order to take the full advantage of IR-UWB technology. Themost significant one lies in the reception of UWB signals. Traditional receiverrequires Nyquist rate ADC which is overwhelmingly complex and power hun-gry. This dissertation proposes and investigates possible sub-Nyquist sam-pling techniques for IR-UWB receiver design.

In the first part of this dissertation, the IR-UWB receiver based on energydetection (ED) principle is explored. A low-power ED receiver featured byflexibility and multi-mode operation is proposed. The receiver prototype for3-5 GHz band is implemented in 90 nm CMOS. Measurement results demon-strate that 16.3 mW power consumption and -79 dBm sensitivity at 10 Mb/sdata rate can be achieved. To further optimize the receiver performance,threshold optimization is suggested for the on-off-keying modulated signal,and adaptive synchronization and integration region optimization is proposed.Finally, a low complexity burst packet detection scheme is proposed, which isadaptive to the variations of noise background and link distance.

In the second part of this dissertation, the IR-UWB receiver based oncompressed sensing (CS) theory is investigated. Firstly, appropriate sparsebasis, sensing matrix and reconstruction algorithms are suggested for the CSbased IR-UWB receiver. And then, the architectural analysis of the CS re-ceiver with focuses on the random noise processes in the CS measurementprocedure is presented. At last, a novel two-path noise-reducing architecturefor the CS receiver is proposed. Besides the improvement on the receiver per-formance, the proposed architecture also relaxes the hardware implementationof the CS random projection as well as the back-end signal reconstruction.

Keywords: Ultra-Wideband, impulse radio, receiver, energy detection, com-pressed sensing, sub-Nyquist sampling, Internet-of-Things

iv

Sammanfattning

I en tid präglad av Internet-of-Things (IoT), där det blir allt vanligare attbygga in mikrosystem i våra vardagsföremål för identifikation, detektion, kom-munikation och position kommer vår efterfrågan på trådlösa kortavståndsan-slutningar med låg strömförbrukning, låg kostnad, robust kommunikation, ochexakt positionering fortsätta att växa i allt större takt. Här framstår Impul-se Radio Ultra-Wideband (IR-UWB) för överföring av subnanosekundpulseröver flera GHz-frekvensband med extremt lågt effektspektrum som en lovan-de kandidat. Men det finns flera svårigheter som måste lösas innan vi kanutnyttja IR-UWB-tekniken till fullo. En av de största svårigheterna ligger imottagandet av UWB-signalerna. Traditionella mottagare använder Nyquist-frekvensen ADC, som skulle leda till en alltför komplex och energislukandelösning. Denna avhandling föreslår och undersöker möjliga tekniker för sub-Nyquist-sampling som kan användas i IR-UWB-mottagare.

I avhandlingens första del undersöks IR-UWB-mottagare baserade påen energidetekteringsprincip (ED) utifrån systemarkitektur, kretsimplemen-tering och prestandaoptimering. Därefter föreslås en IR-UWB-mottagare medlågeffekts-ED som erbjuder flexibilitet och flerlägesdrift. Mottagarprototypenför 3-5 GHz-bandet tas fram och tillverkas i 90 nm CMOS teknik. Mätre-sultaten visar att 16.3 mW energiförbrukning och -79 dBm sensitivitet kanuppnås vid en överföringshastighet på 10 Mb/s. För ytterligare optimeringav mottagarprestandan föreslås optimering av tröskelvärdena för den OOK-modulerade signalen tillsammans med adaptiv synkronisering och optimeringav integrationsregionerna. Slutligen behandlas problemet med paketdetekte-ring vid skurpulser i kontexten av trådlöst drivna positionsdetekterande UWBRFID-system. Här föreslås ett detekteringssystem med låg komplexitet somkan anpassas till varierande bakgrundsbrus och länkavstånd.

I avhandlingens andra del undersöks en IR-UWB-mottagare baserat påteorin om framväxande så kallat compressed sensing (CS). Först av allt, an-vänds en genomgående studie av CS-teorin som bas för att föreslå lämpligaalgoritmer för gles bas, detekteringsmatriser och rekonstruktion för den CS-baserade IR-UWB-mottagaren. Därefter presenteras den första arkitektoniskaanalysen av CS-mottagaren med fokus på de randomiserade brusprocessernai CS-mätningen. Slutligen föreslås en nytänkande tvåvägs brusreducerandearkitektur för CS-mottagaren. Den föreslagna arkitekturen bidrar till förbätt-rad mottagarprestanda, men också mindre strikt maskinvaruimplementeringav CS-randomiserad projicering och back-end-signalrekonstruktion.

Acknowledgment

First and foremost, I would like to express my sincere gratitude to my advisors:Prof. Hannu Tenhunen, Prof. Li-Rong Zheng and Dr. Qiang Chen. I am heartilygrateful to Li-Rong for providing me the opportunity to carry out this PhD researchat KTH. His broad and insightful vision greatly inspires me along the whole PhDjourney. I am especially thankful to Hannu for his professional guidance and hisconsistent support and encouragement. I am also very grateful to Qiang for sharinghis rich experiences and offering helpful advices to both work and life.

I am particularly indebted to Dr. Zhuo Zou for his valuable suggestions onresearch methodology, scientific writing and presentation. I appreciate all the in-spiring discussions with him. I would also like to thank all other colleagues inthe energy detection UWB receiver project: Dr. Fredrik Jonsson for his excellentproject supervision and for the sharing of his rich experience in chip design; DavidS. Mendoza, Peng Wang and Dr. Jia Mao for their great work and collaboration.

I would also like to thank all my other former and present colleagues in iPackCenter and ICT school. Many thanks go to Dr. Yi Feng for her pleasant friendshipand for helping me acclimate to the new working and living environment duringmy first several months in Sweden; to Dr. Ning Ma for his earnest help during andafter work; to Dr. Jia Mao for all the helpful technical discussions and interestingdaily chats as my officemate; to Chuanying Zhai for sharing her knowledge on UWBpositioning systems; to Iris Jie Gao, Dr. Li Xie, Dr. Geng Yang, Qiansu Wan, Dr.Zhibo Pang, Dr. Jian Chen, Dr. Botao Shao, Dr. Huimin She, Dr. Liang Rong,Dr. Jue Shen, Dr. Shaoteng Liu, Dr. Sha Tao, Dr. Awet Weldezion, KunlongYang and Lebo Wang for all the nice chats and all kinds of help. There are stilltoo many to mention, and I thank you all indeed. I am also grateful to all the ITsupport and administrative staffs in ES department and in ICT school for creatinga pleasant working environment.

I gratefully acknowledge Prof. Jari Nurmi from Tampere University of Tech-nology for kindly accepting the role of advance reviewer as well as the stand-incommittee member. My sincere thanks are also given to Prof. Kari Halonen fromAalto University for acting as my opponent, as well as to Associate Prof. CristinaRusu from Acero Swedish ICT, Prof. Smail Tedjini from Grenoble-inp/lcis, andProf. Bengt Oelmann from Mid Sweden University for serving as committee mem-bers.

v

vi

Last but not least, I wish to express my deepest gratitude to my family: to myhusband Dr. Qi Gao for his unconditional support and understanding throughoutthese years, and to my parents for their endless love and support. This dissertationis dedicated to you.

Qin Zhou

Oct., 2016, Stockholm

Abbreviations

ADC analog to digital converterAFE analog front-endASIC application specific integrated circuitASK amplitude shift keyingAWGN additive white Gaussian noiseBB basebandBER bit error rateBP basis pursuitBPDN basis pursuit de-noisingBPF bandpass filterBPSK binary phase shift keyingCMOS complementary-metal-oxide semiconductorCS compressed sensingDCO digitally-controlled oscillatorDFF D-Flip-FlopDSP digital signal processingED energy detectionEIRP equivalent isotropically radiated powerFCC Federal Communication CommissionFPGA field-programmable gate arrayGPS global positioning systemHSMC high speed mezzanine cardIoT Internet-of-thingsIR-UWB impulse radio ultra-widebandISM industrial, scientific and medicalLFSR linear feedback shift registerLNA low noise amplifierLO local oscillatorLOS line of sightMAE mean absolute errorMB multi-bandNF noise figureNLOS non-line of sight

vii

viii ABBREVIATIONS

OFDM orthogonal frequency division multiplexingOMP orthogonal matching pursuitOOK on-off keyingPG processing gainPLL phase locked loopPN pseudo numberPPM pulse position modulationPRBS pseudo-random binary sequencePRI pulse repetition intervalPSD power spectral densityQAC quadrature analog correlationQoS quality of serviceRF radio frequencyRFID radio frequency identificationRMPI random modulation pre-integrationRTC real-time controlRTLS real-time locating systemsRX receiverSDR software-defined radioSFD start frame delimiterSNR signal to noise ratioTOA time of arrivalTR transmitted referenceTX transmitterUHF ultra-high frequencyUWB ultra-widebandVGA variable gain amplifierWBAN wireless body area networkWPAN wireless personal area networkWSN wireless sensor network

List of Publications

Papers included in this thesis:1. Qin Zhou, Jia Mao, Zhuo Zou, Fredrik Jonsson, and Li-Rong Zheng, “A

mixed-signal timing circuit in 90nm CMOS for energy detection IR-UWBreceivers,” in IEEE International SOC Conference (SOCC 2010), pp. 413–416, September 2010.

2. Zhuo Zou, David Sarmiento Mendoza, Peng Wang, Qin Zhou, Jia Mao,Fredrik Jonsson, Hannu Tenhunen, and Li-Rong Zheng, “A low-power andflexible energy detection IR-UWB receiver for RFID and wireless sensor net-works,” IEEE Transactions on Circuits and Systems I: Regular Papers, vol.58, no. 7, pp. 1470–1482, July 2011.

3. Jia Mao, David Sarmiento Mendoza, Qin Zhou, Jian Chen, Peng Wang, ZhuoZou, Fredrik Jonsson, and Li-Rong Zheng, “A 90nm CMOS UHF/UWBasymmetric transceiver for RFID readers,” in European Solid State CircuitsConference (ESSCIRC 2011), pp. 179–182, September 2011.

4. Qin Zhou, Zhuo Zou, Fredrik Jonsson, and Li-Rong Zheng, “A flexible back-end with optimum threshold estimation for OOK based energy detectionIR-UWB receivers,” in IEEE International Conference on Ultra-Wideband(ICUWB 2011), pp. 130–134, September 2011.

5. Qin Zhou, Zhuo Zou, Hannu Tenhunen, Li-Rong Zheng, “Adaptive synchro-nization and integration region optimization for energy detection IR-UWBreceivers,” in IEEE International Conference on Ultra-Wideband (ICUWB2012), pp. 62–66, September 2012.

6. Qin Zhou, Zhuo Zou, Qiang Chen, Hannu Tenhunen, and Li-Rong Zheng,“Low complexity burst packet detection for wireless-powered UWB RFID sys-tems,” in IEEE International Conference on Ubiquitous Wireless Broadband(ICUWB 2015), pp. 1–5, October 2015.

7. Qin Zhou, Zhuo Zou, Hannu Tenhunen, and Li-Rong Zheng, “Exploration andperformance evaluation of a compressed sensing based IR-UWB receiver,” in

ix

x LIST OF PUBLICATIONS

IEEE International Conference on Ultra-Wideband (ICUWB 2013), pp. 226–230, September 2013.

8. Qin Zhou, Zhuo Zou, Hannu Tenhunen, and Li-Rong Zheng, “Architecturalanalysis of compressed sensing based IR-UWB receiver for communicationand ranging,” in IEEE International Conference on Ultra-Wideband (ICUWB2014), pp. 222–227, September 2014.

9. Qin Zhou, Zhuo Zou, Qiang Chen, Hannu Tenhunen, and Li-Rong Zheng,“Noise-reducing architecture of compressed sensing receiver for IR-UWB rang-ing systems,” accepted for publication in IEEE International Conference onUbiquitous Wireless Broadband (ICUWB 2016), October 2016.

Related publications not included in this thesis:10. Zhuo Zou, Ti Deng, Qin Zhou, David Sarmiento Mendoza, Fredrik Jonsson,

and Li-Rong Zheng, “Energy detection receiver with TOA estimation enablingpositioning in passive UWB-RFID system,” in IEEE International Conferenceon Ultra-Wideband (ICUWB 2010), vol. 2, pp. 1–4, September 2010.

11. David Sarmiento Mendoza, Zhuo Zou, Qin Zhou, Jia Mao, Peng Wang,Fredrik Jonsson, and Li-Rong Zheng, “Analog front-end RX design for UWBimpulse radio in 90nm CMOS,” in IEEE International Symposium on Circuitsand Systems (ISCAS 2011), pp. 1552–1555, May 2011.

12. Zhuo Zou, Botao Shao, Qin Zhou, Chuanying Zhai, Jia Mao, Majid Baghaei-Nejad, Qiang Chen, and Li-Rong Zheng, “Design and demonstration of pas-sive UWB RFIDs: chipless versus chip solutions,” in IEEE InternationalConference on RFID-Technologies and Applications (RFID-TA 2012), pp. 6–11, November 2012.

13. Chuanying Zhai, Zhuo Zou, Qin Zhou, and Li-Rong Zheng, “A software de-fined radio platform for passive UWB-RFID localization,” in IEEE Interna-tional Conference on Wireless Information Technology and Systems (ICWITS2012), pp. 1–4, November 2012.

14. Chuanying Zhai, Zhuo Zou, Qin Zhou, Jia Mao, Qiang Chen, Hannu Ten-hunen, Li-Rong Zheng, and Lida Xu, “A 2.4-GHz ISM RF and UWB hybridRFID real-time locating system for industrial enterprise Internet of things,”Enterprise Information Systems, pp. 1–18, March 2016.

15. Jia Mao, Qin Zhou, David Sarmiento Mendoza, Jian Chen, Peng Wang,Fredrik Jonsson, Lida Xu, Li-Rong Zheng, and Zhuo Zou, “A hybrid readertransceiver design for industrial Internet of things,” Journal of IndustrialInformation Integration, vol. 2, pp. 19–29, June 2016.

xi

Other non-review papers and presentations:16. Qin Zhou, Zhuo Zou, Jia Mao, Fredrik Jonsson, and Li-Rong Zheng, “A

flexible energy detection IR-UWB receiver for RFID and WSN,” in SwedishSystem-on-Chip Conference (SSoCC 2010), March 2010. (Non-review).

17. Qin Zhou, Zhuo Zou, Fredrik Jonsson, and Li-Rong Zheng, “A flexible energydetection IR-UWB receiver for RFID and wireless sensor networks” in 6thIEEE UWB Forum on Sensing and Communications, May 2011. (InvitedPresentation).

Contents

Abbreviations vii

List of Publications ix

Contents xiii

List of Figures xv

List of Tables xvii

1 Introduction 11.1 Background and Motivation . . . . . . . . . . . . . . . . . . . . . . . 1

1.1.1 The Vision of Internet-of-Things . . . . . . . . . . . . . . . . 11.1.2 Short Range Wireless Link for the IoT . . . . . . . . . . . . . 21.1.3 IR-UWB: Opportunities and Challenges . . . . . . . . . . . . 3

1.2 Thesis Contributions and Organization . . . . . . . . . . . . . . . . . 41.2.1 Contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.2.2 Thesis Organization . . . . . . . . . . . . . . . . . . . . . . . 10

2 Overview of UWB Technology 132.1 Fundamentals of UWB . . . . . . . . . . . . . . . . . . . . . . . . . . 13

2.1.1 History, Definition and Regulations . . . . . . . . . . . . . . . 132.1.2 Benefits and Potential Applications . . . . . . . . . . . . . . . 14

2.2 Generation of UWB Signals . . . . . . . . . . . . . . . . . . . . . . . 162.3 IR-UWB Modulations . . . . . . . . . . . . . . . . . . . . . . . . . . 182.4 UWB Channels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 202.5 IR-UWB Receiver Overview . . . . . . . . . . . . . . . . . . . . . . . 22

2.5.1 Review of IR-UWB Receiver Architectures . . . . . . . . . . 222.5.2 Promising Rx Candidate for IoT applications . . . . . . . . . 24

2.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25

3 Sub-Nyquist Sampling IR-UWB Receiver I: Energy Detection 293.1 Basic Principle of Energy Detection . . . . . . . . . . . . . . . . . . 29

xiii

xiv CONTENTS

3.2 A Flexible and Low-Power ED Receiver . . . . . . . . . . . . . . . . 303.2.1 Receiver Architecture . . . . . . . . . . . . . . . . . . . . . . 313.2.2 Implementation and Measurement Results . . . . . . . . . . . 32

3.3 Highly Programmable Timing Circuit . . . . . . . . . . . . . . . . . 333.3.1 Proposed Timing Specification . . . . . . . . . . . . . . . . . 343.3.2 Circuit Design and Implementation . . . . . . . . . . . . . . . 35

3.4 Optimum Threshold Estimation for OOK Modulated Signal . . . . . 363.4.1 Theoretical Derivation . . . . . . . . . . . . . . . . . . . . . . 383.4.2 Proposed Method for Threshold Optimization . . . . . . . . . 39

3.5 Adaptive Synchronization and Integration Region Optimization . . . 413.5.1 Numerical Study of Signal Energy Capture . . . . . . . . . . 413.5.2 Adaptive Synchronization . . . . . . . . . . . . . . . . . . . . 42

3.6 Low Complexity Burst Packet Detection . . . . . . . . . . . . . . . . 423.6.1 Proposed Method . . . . . . . . . . . . . . . . . . . . . . . . . 443.6.2 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . 46

3.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

4 Sub-Nyquist Sampling IR-UWB Receiver II: Compressed Sensing 514.1 Overview of CS Theory . . . . . . . . . . . . . . . . . . . . . . . . . 514.2 CS for IR-UWB . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534.3 Noise Driven Architectural Analysis . . . . . . . . . . . . . . . . . . 56

4.3.1 Correlated Noise versus Uncorrelated Noise . . . . . . . . . . 564.3.2 Architectural Analysis . . . . . . . . . . . . . . . . . . . . . . 57

4.4 Noise-Reducing Architecture for CS Ranging Receiver . . . . . . . . 584.4.1 Two-Path Noise-Reducing Architecture . . . . . . . . . . . . 594.4.2 Performance Evaluation . . . . . . . . . . . . . . . . . . . . . 61

4.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62

5 Conclusions 655.1 Thesis Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 655.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66

Bibliography 69

List of Figures

1.1 The vision of Internet-of-Things. . . . . . . . . . . . . . . . . . . . . . . 21.2 UHF/UWB hybrid RFID system with asymmetric links. . . . . . . . . . 41.3 Navigation diagram of the dissertation. . . . . . . . . . . . . . . . . . . 11

2.1 FCC spectral mask for UWB systems. . . . . . . . . . . . . . . . . . . . 152.2 IR-UWB signal versus traditional narrowband signal. . . . . . . . . . . 152.3 UWB application scenarios. . . . . . . . . . . . . . . . . . . . . . . . . . 172.4 Waveforms of the first 12 derivatives of Gaussian pulse with α = 0.714ns. 192.5 PSD of the first 12 derivatives of Gaussian pulse with α = 0.714ns and

pulse repetition interval Tf = 100ns. . . . . . . . . . . . . . . . . . . . . 202.6 Data modulation options for IR-UWB signals. . . . . . . . . . . . . . . 212.7 Example of one channel realization for CM1 - CM8, respectively. . . . . 262.8 Different architectures of IR-UWB receiver. . . . . . . . . . . . . . . . . 27

3.1 Block diagram of an energy detection IR-UWB receiver. . . . . . . . . . 303.2 Signaling of energy detection with OOK modulated signal. . . . . . . . 303.3 The architecture of the proposed flexible energy detection IR-UWB re-

ceiver. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 313.4 Chip micrograph of the proposed IR-UWB receiver. . . . . . . . . . . . 333.5 Measurement setup. (a) Link connections, and (b) test environment. . . 333.6 BER versus average Rx input power (10 Mb/s, OOK modulation, un-

coded, wired connection). . . . . . . . . . . . . . . . . . . . . . . . . . . 343.7 Architecture of the proposed timing circuit. . . . . . . . . . . . . . . . . 373.8 Timing diagram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 373.9 BER performance for the proposed threshold optimization in IEEE

802.15.4a UWB channels. . . . . . . . . . . . . . . . . . . . . . . . . . . 403.10 (a) Channel impulse response, (b) BER with respect to integration inter-

val, (c) normalized signal energy capture in terms of integration interval,for CM1, CM2, CM7, and CM8, respectively. . . . . . . . . . . . . . . . 43

3.11 BER performance for the proposed synchronization scheme. . . . . . . . 443.12 Packet format. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443.13 Flowchart of the proposed packet detection scheme. . . . . . . . . . . . 45

xv

xvi List of Figures

3.14 Preamble detection algorithm. . . . . . . . . . . . . . . . . . . . . . . . 463.15 UWB SDR testbed. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.16 Simulation and field test results of Pfa (numComp stands for the num-

ber of comparisons in the preamble detection process and it determinesthe required length of the preamble). . . . . . . . . . . . . . . . . . . . . 48

3.17 Simulation and field test results of Pmd (15-bit SFD is adopted). . . . . 493.18 Simulation results of Pe with proposed method and conventional threshold-

based method. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49

4.1 The procedure of compressed sensing. . . . . . . . . . . . . . . . . . . . 524.2 The RMPI architecture. . . . . . . . . . . . . . . . . . . . . . . . . . . . 554.3 BER comparison between correlated noise situation and uncorrelated

noise situation (AWGN channel and Ts = 100 ns). . . . . . . . . . . . . 584.4 MAE comparison between correlated noise situation and uncorrelated

noise situation (CM1 channel and Ts = 200 ns). . . . . . . . . . . . . . . 594.5 Possible architectures of the CS based IR-UWB receiver: (a) parallel ar-

chitecture with single antenna and single LNA; (b) parallel architecturewith single antenna and multiple LNAs; (c) parallel architecture withmultiple antennas and multiple LNAs; and (d) serial architecture. . . . 60

4.6 The proposed noise-reducing architecture for CS based IR-UWB rangingreceiver. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61

4.7 Signaling comparison between conventional and proposed architecture. . 624.8 Power spectral density before and after the mixing stage. . . . . . . . . 634.9 MAE of TOA estimation in AWGN channel. . . . . . . . . . . . . . . . 634.10 MAE of TOA estimation in CM1 channel. . . . . . . . . . . . . . . . . . 64

List of Tables

1.1 Comparison of Short Range Wireless Technologies. . . . . . . . . . . . . 3

2.1 IEEE 802.15.4a Channel Models. . . . . . . . . . . . . . . . . . . . . . . 21

3.1 Comparisons of the Proposed IR-UWB Receiver with Previous Pub-lished Works. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32

3.2 Performance of the Timing Circuit at 900 MHz Reference Clock. . . . . 363.3 Optimum Normalized Threshold. . . . . . . . . . . . . . . . . . . . . . . 39

4.1 Noise Situation in CS Measurement Process Subject to Receiver Archi-tecture and Noise Source. . . . . . . . . . . . . . . . . . . . . . . . . . . 58

xvii

Chapter 1

Introduction

1.1 Background and Motivation

1.1.1 The Vision of Internet-of-Things

Along with the advances in microelectronics, communications and information tech-nology over the past decades, we’re now entering a new era of the Internet-of-Things(IoT): a vision of connectivity for anybody and anything, at anytime and from any-where [1], as depicted in Figure 1.1. The IoT extends the current Internet into thereal world of physical objects, and has been recognized as the next technologicalrevolution after the World Wide Web and universal mobile accessibility [2]. The po-tential applications enabled by the IoT penetrate every aspect of our lives, to namea few, logistics, healthcare, smart environment, and industrial automation, whichwill dramatically improve the quality of our lives and contribute to the growth ofworld’s economy [3].

In the vision of IoT, all daily objects become smart and interconnected. Theyare able to sense the surrounding environment, communicate with each other, accessInternet services and interact with people [4]. To make this vision become a reality,micro-devices enabling identification, sensing, communication and positioning aredesired to be integrated and embedded into any daily objects, such as automobiles,food and medicine packages, furnitures and even paper documents. It is expectedthat there will be 50 to 100 billion smart objects connected to the Internet in thenear future [2]. The development of IoT requires technical innovations in a numberof different fields, for instance, sensor design and integration with radio frequencyidentification (RFID), wireless links for robust communication and high-precisionpositioning, low-power and low-cost electronic system design, system integrationand miniaturization [5].

1

2 CHAPTER 1. INTRODUCTION

Figure 1.1: The vision of Internet-of-Things. Figure from Internet [6].

1.1.2 Short Range Wireless Link for the IoTShort range wireless technology is one of the key enablers for the IoT. In the IoTscenario, radio frequency identification (RFID) tags [7] or smart wireless sensornetwork (WSN) nodes [8] that are scattered in the local area, personal area oreven body area will be networked by wireless links and bridged to the Internet.Such wireless links have a range of several to tens of meters and desire a moderatedata rate (∼ Mb/s) with ultra-low energy consumption (∼ nJ/bit). Besides, theradio transceivers should be low power and low cost which can be highly integratedinto self-powered or wireless-powered tags [9]. Furthermore, sub-meter ranging andpositioning capability is highly anticipated for location-aware applications.

To date there are several wireless technologies dedicated to short range commu-nications, passive ultra-high frequency (UHF) RFID, ZigBee [10], Bluetooth [11],Wi-Fi [12] and the emerging Impulse Radio Ultra-Wideband (IR-UWB). Table 1.1summarizes the main characteristics of those wireless technologies [13–16]. PassiveUHF RFID has been widely considered for the IoT applications due to its low tagcost and long lifetime. However, it suffers from multi-user interference and limitedcoverage range and data rate [17–19]. ZigBee provides highly reliable mesh net-works with long battery life [20]. The network coverage range is typically abouttens of meters and is able to support essentially an unlimited number of sensornodes due to its meshing capability. However, the data rate can not exceed 250kb/s. Compared to ZigBee, Bluetooth has a higher data rate up to 723 kb/s but

1.1. BACKGROUND AND MOTIVATION 3

at the expenses of higher power consumption and less coverage range (only 10meters). Besides, it only support star network topology with limited number ofnodes. Wi-Fi, on the other hand, offers both high data rate (10-105 Mb/s) andsufficient coverage range (10-100 m). Nevertheless, the high power consumption isunacceptable for most IoT applications. One common drawback of passive UHFRFID, ZigBee, Bluetooth and Wi-Fi lies in the ranging/positioning capability, onlymeter level positioning accuracy can be achieved which is insufficient for indoorapplications. IR-UWB technology has recently attracted much attention due toits unique features like centimeter positioning capability and low power low costimplementation. It is increasingly accepted as one promising candidate for the nextgeneration RFID and WSN towards the IoT [21].

Table 1.1: Comparison of Short Range Wireless Technologies.

Technology Standard Frequency band Maximum Range Positioning Power

data rate accuracy865-868 MHz (Europe)

Passive UHF RFID Gen2 (ISO 18000-6C) 902-928 MHz (U.S.) 640 kb/s 10 m Meter level Low950-956 MHz (Japan)2.4-2.4835 GHz 250 kb/s

ZigBee IEEE 802.15.4 901-928 MHz 40 kb/s 10-300 m Meter level Low868-868.6 MHz 20 kb/s

Bluetooth IEEE 802.15.1 2.4 GHz 723 kb/s 10 m Meter level LowWi-Fi IEEE 802.11 a/b/g 2.4 GHz; 5 GHz 10-105 Mb/s 10-100 m Meter level High

UWB IEEE 802.15.4a 3.1-10.6 GHz tens of Mb/s 10-100 m Centimeter level Low

1.1.3 IR-UWB: Opportunities and ChallengesEver since the Federal Communications Commission (FCC) approved UWB forcommercial use in the 3.1 GHz to 10.6 GHz frequency band in 2002 [22], IR-UWBhas drawn explosive attention as a new type of short range wireless technologyfor communication and localization. In contrast to conventional continuous-wavemodulated narrow band communication, the transmission of IR-UWB relies onextremely-short duration pulses (on the order of nanosecond) with low duty cycle,spanning up to several GHz frequency band with very low power spectral den-sity. The short pulses promise both robust communication and high-precision po-sitioning. The carrier-less UWB signal can be generated without any prior RFmixing stage which potentially allows low-power and low-cost transceiver imple-mentation [23, 24]. Those unique features make IR-UWB a promising solution forIoT applications.

Nevertheless, to fulfill the expectations of IR-UWB technology, there are stillsome challenges to overcome. One of the most significant challenges is the IR-UWBreceiver design. Comparing with the narrowband system, the UWB transmitter canbe extremely low-power and low-complexity, however, the receiver design is muchmore complex and power-consuming due to the large signal bandwidth. Besides, theultra-short pulses also increase the difficulty of synchronization process especiallyfor burst manner short packet transmissions, such as in RFID and WSN applica-


UHF RxWake-up Radio

UWB TxMain Radio

Radio Tag 1UHF Tx

UWB Rx

ReaderEnergyClock

Request

Data

Figure 1.2: UHF/UWB hybrid RFID system with asymmetric links. Figure adaptedfrom Paper VI.

tions. There are already several works devoted to solve this problem and attempt toemploy IR-UWB radio in battery-less RFID systems. For example, a UHF/UWBhybrid radio architecture with asymmetric links was proposed in [25,26]. As shownin Figure 1.2, it uses the UHF radio as the downlink (from reader to tag) to powerup and control the tags, whereas IR-UWB radio is employed in the uplink (fromtag to reader) for massive data transmission. This hybrid radio architecture cantake full benefits of IR-UWB signal from the uplink, and at the same time avoidthe complex and power hungry UWB receiver on resource-limited tags by shiftingthe burden to the reader side [26]. With an eye on the whole wireless link, this solu-tion relaxes the power and cost constraints on the IR-UWB receiver design, solvingthe problem by a devious path. On the other hand, research continues to exploredifferent architectures for IR-UWB receiver design, seeking promising solutions forlow-power and low-complexity implementations.

This dissertation is focused on the IR-UWB receiver design for IoT applications.Due to the GHz signal bandwidth, traditional receivers requiring Nyquist rate ADCwill be overwhelmingly complex and power hungry, even for the UHF/UWB hybridRFID systems. Consequently, possible sub-Nyquist sampling techniques need to beexplored and investigated for the IR-UWB receiver design.

1.2 Thesis Contributions and Organization

1.2.1 ContributionsThis dissertation studies two types of sub-Nyquist sampling principle for IR-UWBreceiver design, the Energy Detection (ED) principle and the Compressed Sens-ing (CS) principle. Multidisciplinary research results covering different aspectsof the receiver design are provided, from system architecture, application specificintegrated circuit (ASIC) fabrication and prototyping, to signal processing for syn-chronization and detection.

In the first part of this dissertation, a low-power energy detection IR-UWBreceiver featured by flexibility and multi-mode operation to better fit into various

1.2. THESIS CONTRIBUTIONS AND ORGANIZATION 5

IoT applications is proposed. The receiver prototype for the 3-5 GHz band isimplemented and fabricated in 90 nm CMOS. Measurement results demonstratethat 16.3 mW power consumption and -79 dBm sensitivity at 10 Mb/s data ratecan be achieved. A highly programmable timing circuit is the key enabler for theflexibility and multi-mode operation of the proposed receiver. To cope with thehigh frequency reference clock as well as the high programmability, a mixed-signaldesign flow is adopted in the implementation of the timing circuit. To furtheroptimize the receiver performance, threshold optimization is suggested for the on-off-keying (OOK) modulated signal, and adaptive synchronization and integrationregion optimization based on the study of signal energy capture is proposed. Finally,the problem of burst pulse burst packet detection in the context of wireless-poweredlocation-aware UWB RFID systems is addressed. A low complexity burst packetdetection scheme based on sensing the preamble signal characteristic instead of thereceived signal strength is proposed. As a result, the detection is adaptive to thevariations of noise background and link distance.

In the second part of this dissertation, the IR-UWB receiver based on the emerg-ing compressed sensing theory is investigated. Unlike the ED receiver which is at-tractive due to its low power and low complexity implementation, the CS receiverallows sub-Nyquist sampling without sacrificing any time domain resolution, whichis desirable in some IoT applications to meet the stringent ranging/positioningrequirements. First of all, based on the throughout study of the CS theory, appro-priate sparse basis, sensing matrix and reconstruction algorithms are suggested forthe CS based IR-UWB receiver. And then, the first-ever architectural analysis ofthe CS receiver with focuses on the random noise processes in the CS measurementprocedure is presented. At last, a novel two-path noise-reducing architecture for theCS receiver is proposed. Besides the improvement on the receiver performance, theproposed architecture also relaxes the hardware implementation of the CS randomprojection as well as the back-end signal reconstruction.

The related work and contributions are mainly included in the following ap-pended publications, which can be partitioned into two groups.

Sub-Nyquist Sampling IR-UWB Receiver I: Energy Detection

[Paper I] Qin Zhou, Jia Mao, Zhuo Zou, Fredrik Jonsson, and Li-Rong Zheng, “Amixed-signal timing circuit in 90 nm CMOS for energy detection IR-UWBreceivers,” in IEEE International SOC Conference (SOCC 2010), pp. 413–416, September 2010. [27]Paper Contribution: This paper presents a flexible timing circuit with 1.1ns delay resolution for energy detection IR-UWB receivers. Referenced at 900MHz input clock, the circuit generates multi-phased integration windows andreset signals that enable/disable the operation of analog blocks. The designis highly programmable, adapting the receiver to pulse level synchronization,symbol level synchronization, different data rates and various channel envi-ronments. Mixed-signal design flow is adopted to avoid the complexity of full


custom design and the large power consumption of full synthesized digitaldesign. The timing circuit is implemented in UMC 90 nm CMOS process,with 219 µW power consumption and 190*295 µm2 die area.The author’s contribution: The author came up with the idea, definedthe system specification, designed and implemented the circuit in CMOS,conducted the circuit measurement, and wrote the manuscript.

[Paper II] Zhuo Zou, David Sarmiento Mendoza, Peng Wang, Qin Zhou, Jia Mao,Fredrik Jonsson, Hannu Tenhunen, and Li-Rong Zheng, “A low-power andflexible energy detection IR-UWB receiver for RFID and wireless sensor net-works,” IEEE Transactions on Circuits and Systems I: Regular Papers, vol.58, no. 7, pp. 1470–1482, July 2011. [24]Paper Contribution: This paper presents an energy detection IR-UWBreceiver for RFID and WSN applications. An ASIC consisting of a 3-5 GHzanalog front-end, a timing circuit and a high speed baseband controller isimplemented in a 90 nm standard CMOS technology. A Field-ProgrammableGate Array (FPGA) is employed as a reconfigurable back-end, enabling adap-tive baseband algorithms and ranging estimations. The proposed architectureis featured by high flexibility that adopts a wide range of pulse rate, process-ing gain, correlation schemes, synchronization algorithms, and modulationschemes (PPM/OOK). The receiver prototype was fabricated and measured.The power consumption of the ASIC is 16.3 mW at 1 V power supply, whichpromises a minimal energy consumption of 0.5 nJ/bit. The whole link is eval-uated together with a UWB RFID tag. Bit error rate (BER) measurementdisplays a sensitivity of -79 dBm at 10 Mb/s with 10−3 BER achieved bythe proposed receiver, corresponding to an operation distance over 10 metersunder the FCC regulation.The author’s contribution: The author participated in defining the systemarchitecture and specification, involved in the circuit design and implementa-tion, conducted the chip measurement, and wrote parts of the manuscript.

[Paper III] Jia Mao, David Sarmiento Mendoza, Qin Zhou, Jian Chen, PengWang, Zhuo Zou, Fredrik Jonsson, and Li-Rong Zheng, “A 90 nm CMOSUHF/UWB asymmetric transceiver for RFID readers,” in European SolidState Circuits Conference (ESSCIRC 2011), pp. 179–182, September 2011.[28]Paper Contribution: This paper presents an integrated asymmetric transc-eiver in 90 nm CMOS technology for RFID reader. The proposed reader usesUHF transmitter to power up and inventory the tags. In the reverse link, anon-coherent energy detection IR-UWB receiver is deployed for data recep-tion with high throughput and ranging capability. The transmitter delivers160 kb/s ASK modulated data by an integrated modulator and a DigitalControlled Oscillator (DCO) in UHF band with 11% tuning range. The DCO


consumes 6 mW with 0.12 mm2 area. On the other side, adopting two inte-gration channels, the 3-5 GHz energy detection receiver supports maximum33 Mb/s data rate both in OOK and PPM modulations. The receiver front-end provides 59 dB voltage gain and 8.5 dB noise figure (NF). Measurementresults show that the receiver achieves an input sensitivity of -79 dBm at 10Mb/s, with power consumption of 15.5 mW.The author’s contribution: The author participated in defining the systemarchitecture and specification, involved in the circuit design and implementa-tion, conducted the chip measurement.

[Paper IV] Qin Zhou, Zhuo Zou, Fredrik Jonsson, and Li-Rong Zheng, “A flexibleback-end with optimum threshold estimation for OOK based energy detectionIR-UWB receivers,” in IEEE International Conference on Ultra-Wideband(ICUWB 2011), pp. 130–134, September 2011. [29]Paper Contribution: This paper presents an on-off keying (OOK) basedenergy detection IR-UWB receiver with focus on the back-end design. In orderto optimize the receiver performance according to different multi-path envi-ronment, variable integration interval and adaptive threshold optimization areconsidered in the proposed back-end which is composed by a programmabletiming circuit and a reconfigurable baseband processor. The timing circuitis able to generate multi-phased integration windows with wide-range vari-able integration interval and is implemented in 90 nm CMOS process. Novelschemes on synchronization and optimum threshold estimation are suggestedfor baseband processing. The proposed synchronization scheme is based onmaximum energy variance (between symbol ‘0’ and ‘1’) detection, coveringboth the pulse level and symbol level synchronization. And the scheme for op-timum threshold estimation is based on look up table, enabling low complex-ity implementation. System simulation with IEEE 802.15.4a channel modelsshows an appreciable improvement on the bit error rate (BER) performancecompared with the conventional scheme.The author’s contribution: The author came up with the idea, conductedthe theoretical derivation and system simulation, analyzed and evaluated thesimulation results, and wrote the manuscript.

[Paper V] Qin Zhou, Zhuo Zou, Hannu Tenhunen, Li-Rong Zheng, “Adaptivesynchronization and integration region optimization for energy detection IR-UWB receivers,” in IEEE International Conference on Ultra-Wideband (ICU-WB 2012), pp. 62–66, September 2012. [30]Paper Contribution: The performance of energy detection receivers is usu-ally suffered from the noise enhancement due to the large time-bandwidthproduct. The integration region of the receiver integrator significantly affectsthe bit error rate (BER) performance. This paper presents a method of syn-chronization and estimating the optimal integration region (i.e., the starting


point and the length of the integration window), which is based on the analysisof received signal energy capture and combined with a time of arrival (TOA)estimation. The proposed scheme is based on the symbol rate sampling anddoes not require a priori information about the channel delay profile. Besides,it can adapt to various indoor channel environments. The algorithm has amoderate accuracy but a very low complexity and fast synchronization speed.The validity of the proposed approach is demonstrated by numerical resultsusing IEEE 802.15.4a channel models.The author’s contribution: The author came up with the idea, conductedthe system simulation, analyzed and evaluated the simulation results, andwrote the manuscript.

[Paper VI] Qin Zhou, Zhuo Zou, Qiang Chen, Hannu Tenhunen, Li-Rong Zheng,“Low complexity burst packet detection for wireless-powered UWB RFID sys-tems,” in IEEE International Conference on Ubiquitous Wireless Broadband(ICUWB 2015), pp. 1–5, October 2015. [31]Paper Contribution: This paper addresses the issue of UWB signal acquisi-tion in the context of wireless powered UWB RFID systems. In this scenario,the data transmission is based on short packet so as to meet the micro-powerbudget of autonomous power harvesting. The burst short packet transmis-sion as well as the low duty cycling UWB pulse modulation places a stringentchallenge at the UWB receiver for timing acquisition and packet detection.Besides, in a positioning enabled RFID system where variable signal-to-noiseratio (SNR) due to the variable link distance and noise background is unavoid-able, conventional packet detection schemes rely on predefined threshold canhardly achieve good performance. In this study, we propose a low complexitymethod for burst packet detection. It is performed by sensing the preamblesignal characteristic instead of the received signal strength, and thus bypass-ing the necessity of detection threshold. The validity of the proposed approachand its adaptivity to SNR variations is demonstrated by simulation results aswell as field test with a UWB software defined radio (SDR) platform.The author’s contribution: The author came up with the idea, conductedthe system simulation and field test, analyzed and evaluated the results, andwrote the manuscript.

Other publications not included in this thesis but related are [32,33].

Sub-Nyquist Sampling IR-UWB Receiver II: Compressed Sensing

[Paper VII] Qin Zhou, Zhuo Zou, Hannu Tenhunen, and Li-Rong Zheng, “Explo-ration and performance evaluation of a compressed sensing based IR-UWBreceiver,” in IEEE International Conference on Ultra-Wideband (ICUWB2013), pp. 226–230, September 2013. [34]


Paper Contribution: This paper provides an exploration of the CS-basedIR-UWB receiver from different aspects: front-end hardware architectures,back-end signal processing algorithms as well as application scenarios. Andthe performance of the CS receiver regarding the number of CS measurementand different CS recovery algorithms is evaluated and compared against theconventional sub-Nyquist sampling receiver based on energy detection (ED)scheme. Moreover, a strategy to improve the CS receiver performance in han-dling UWB signals with heavy noise and multipath propagation is proposed.The author’s contribution: The author came up with the idea, conductedthe system simulation, analyzed and evaluated the simulation results, andwrote the manuscript.

[Paper VIII] Qin Zhou, Zhuo Zou, Hannu Tenhunen, and Li-Rong Zheng, “Ar-chitectural analysis of compressed sensing based IR-UWB receiver for commu-nication and ranging,” in IEEE International Conference on Ultra-Wideband(ICUWB 2014), pp. 222–227, September 2014. [35]Paper Contribution: This paper presents an architectural analysis of theCS-based IR-UWB receiver with focuses on investigating the random noiseprocesses in the CS measurement procedure. We find that different noisesources (sky noise or amplifier noise) and different receiver architectures (par-allel or serial) will results in different noise situation (correlated or uncorre-lated) in the CS measurement procedure. Bit error rate (BER) simulation fora communication system and time-of-arrival (TOA) estimation for a rangingsystem in additive white Gaussian noise (AWGN) channel as well as IEEE802.15.4a CM1 channel are performed. It shows that CS-based signal detec-tion in uncorrelated noise situation outperforms the correlated noise situation.This noise driven architectural analysis can be used as a design guideline forthe CS-based IR-UWB receiver regarding different application scenarios.The author’s contribution: The author came up with the idea, performedthe noise-based architectural analysis, conducted the system simulation, an-alyzed and evaluated the simulation results, and wrote the manuscript.

[Paper IX] Qin Zhou, Zhuo Zou, Qiang Chen, Hannu Tenhunen, and Li-RongZheng, “Noise-reducing architecture of compressed sensing receiver for IR-UWB ranging systems,” accepted for publication in IEEE International Con-ference on Ubiquitous Wireless Broadband (ICUWB 2016), October 2016.Paper Contribution: A compressed sensing (CS) based impulse radio ultra-wideband (IR-UWB) receiver with two-path noise-reducing RF front-end ar-chitecture is proposed in this paper. By adding an identical input path (an-tenna and gain stage) together with a mixer, the noise in the received signalbefore feeding into the CS sampling block is alleviated comparing with theconventional CS receiver. Moreover, the mixing stage shifts the signal fre-quency spectrum to the lower band which eases the CS sampling hardware as


well as the complexity of back-end signal reconstruction. Simulation resultsfor a ranging system validate that the proposed CS receiver significantly out-performs the conventional one in both additive white Gaussian noise (AWGN)channel and IEEE 802.15.4a multi-path channel.The author’s contribution: The author came up with the idea, conductedthe system simulation, analyzed and evaluated the simulation results, andwrote the manuscript.

1.2.2 Thesis OrganizationThis dissertation is comprised of two parts, a cover essay and a collection of papersin appendices. The cover essay as the first part gives the readers the researchbackground and motivations. It also introduces the conducted research work ina concise manner. Figure 1.3 shows the navigation diagram of this dissertation,which is organized as follows.

Chapter 2 gives a comprehensive introduction of UWB technology. It coversthe key features and benefits of UWB signals, UWB transmission issues as well asan overview on the IR-UWB receiver design.

Chapter 3 explores the sub-Nyquist sampling IR-UWB receiver based on theenergy detection principle. A flexible and low-power energy detection receiver witha highly programmable timing circuit is designed and implemented. Moreover,back-end signal processing algorithms for optimum threshold estimation, adaptivesynchronization and integration region optimization, as well as low complexity burstpacket detection are proposed.

Chapter 4 investigates another type of sub-Nyquist sampling IR-UWB receiverwhich is based on the compressed sensing theory. The performance of the CSreceiver regarding different design parameters is evaluated and analyzed, followedby a noise driven architectural analysis. Besides, a noise-reducing architecture forimproved ranging performance is proposed.

Chapter 5 concludes this dissertation by providing a summary of main contri-butions and some suggestions for future research.


IR-U

WB

: Rel

iabl

e C

omm

unic

atio

n an

d Pr

ecis

e Po

sitio

ning

Rad

io fo

r IoT

UW

B T

echn

olog

y B

asic

s &

IR-U

WB

Rec

eive

r Ove

rvie

w

Sub-

Nyq

uist

Sam

plin

g IR

-UW

B

Rec

eive

r I: E

nerg

y D

etec

tion

Sub-

Nyq

uist

Sam

plin

g IR

-UW

B

Rec

eive

r II:

Com

pres

sed

Sens

ing

Cha

pter

2

ED R

ecei

ver D

esig

n an

d Im

plem

enta

tion

in C

MO

S

Cha

pter

3

Pape

r I, I

I, II

I, IV

, V, V

I

Para

met

ric A

naly

sis

Noi

se D

riven

A

rchi

tect

ural

Ana

lysi

sN

oise

-Red

ucin

g A

rchi

tect

ure

for C

S R

ecei

ver

Cha

pter

4

Pape

r VII

, VII

I, IX

Tim

ing

Circ

uit D

esig

n

Opt

imum

Thr

esho

ld

Estim

atio

n

Ada

ptiv

e Sy

nchr

oniz

atio

n an

d In

tegr

atio

n R

egio

n O

ptim

izat

ion

Low

Com

plex

ity B

urst

Pa

cket

Det

ectio

n

Figu

re1.3:

Navigationdiagram

ofthedissertatio

n.

Chapter 2

Overview of UWB Technology

Ultra-Wideband (UWB) technology has attracted much attention for applicationsin wireless communication and localization ever since the Federal CommunicationsCommission (FCC) approved UWB for commercial use in the 3.1 GHz to 10.6 GHzfrequency band in 2002 [22]. This chapter first presents the fundamentals of UWB,including its definition, unique benefits and potential applications. Followed by theintroduction of UWB signals with focuses on IR-UWB pulse generation, modulationand channel characteristic. Finally, an overview on the IR-UWB receiver designis provided. The design challenges and the pros and cons of the most commonreceiver architectures are discussed.

2.1 Fundamentals of UWB

2.1.1 History, Definition and RegulationsUWB communication spreads signals across a very wide range of bandwidth, insteadof broadcasting on separate frequencies as the conventional narrowband wirelesstransmission. Although the universally adopted method of wireless communica-tion nowadays is the continuous sinusoidal wave based narrowband transmission,the very first wireless transmission in 1894 is actually an impulse based UWBtransmission. The transmission is via the Spark Gap Emitter invented by Italianscientist Guglielmo Marconi. Within the past few decades, accelerating develop-ments in analog and digital electronics and UWB signal theory have opened a newera for practical UWB communication systems. On February 14, 2002, the UnitedStates FCC adopted the First Report and Order allowing commercial deploymentof UWB devices with a given spectral mask requirement [22]. This historical event‘sparked’ exponential increase of interest in UWB in both academia and industry.

According to the proposed definition by FCC [22], a signal is called UWB if ithas a fractional bandwidth equal to or greater than 0.2, or an absolute bandwidth ofat least 500 MHz regardless of the fractional bandwidth. The absolute bandwidthis the frequency band between the upper frequency fH of the -10 dB emission point

13

14 CHAPTER 2. OVERVIEW OF UWB TECHNOLOGY

and the lower frequency fL of the -10 dB emission point, given by

B = fH − fL, (2.1)

which is also refer to as “-10 dB bandwidth”. On the other hand, the fractionalbandwidth is calculated as

Bf = B

fc= fH − fL

(fH + fL)/2 , (2.2)

where fc is the center frequency defined as fc = (fH + fL)/2.UWB signals span a very large portion in the spectrum. To avoid significant

interference with existing users, the FCC specifies the power emission limits incertain frequency bands for various types of UWB applications. The emission limitsare assigned in terms of equivalent isotropically radiated power (EIRP)1. Figure 2.1shows the spectral mask for indoor and outdoor applications in the United Statesunder Part 15 of the commission’s rules [22]2. A wide spectrum up to 7.5 GHz isavailable between 3.1 GHz and 10.6 GHz at the EIRP limit of -41.3 dBm/MHz,which becomes the main spectrum target for most UWB systems.

2.1.2 Benefits and Potential ApplicationsUltra-Wideband technology3, usually refers to as Impulse Radio Ultra-Wideband(IR-UWB). Figure 2.2 illustrates an example of IR-UWB signals. Based on thetransmission of ultra-short pulses (on the order of subnanoseconds) spanning upto several GHz frequency band with very low power spectral density, IR-UWBhas several unique features and benefits that make it different from traditionalnarrowband systems. Those key benefits are listed as follows.

• High speed communication. Thanks to the large signal bandwidth, UWBhas a great potential to provide high data rate wireless communication overshort and medium range.

• Scalability. The duty cycled pulse transmission enables the scalability of ra-dio links in terms of data rate, communication range and power. For example,longer communication range can be achieved by lowering the data rate.

• Multipath and interference immunity. The large bandwidth promisesa large frequency diversity as well as a very high time domain resolution ofthe multipath, together with low duty-cycled transmission, improves the linkrobustness against multipath and interference [40].

1EIRP is defined as “the product of the power supplied to the antenna and the antenna gainin a direction relative to an isotropic antenna” [36].

2The spectral masks have some differences depending on applications and regions of the world.3At present, there are two types of UWB technology. One is based on orthogonal frequency

division multiplexing (OFDM), namely Multi-Band (MB) OFDM UWB [38], which will be brieflyintroduced in the next section. And the other is based on extremely-short duration pulses, namelyImpulse Radio UWB (IR-UWB) [39]. This dissertation is focused on the IR-UWB technology.

2.1. FUNDAMENTALS OF UWB 15

14 © NOKIA FUTURA_WS.PPT / 16-08-2004 / KKa

Underlay Systems (UWB)• UWB: standardization IEEE 802.15.3a, technology selection

ongoing process but there are no globally agreed rules nor spectrum masks for UWB

• Research still required in co-location and interference to victim terminals

Figure 2.1: FCC spectral mask for UWB systems. Figure from [37].

time

frequency

time

frequency

Time-domain behavior Frequency-domain behavior

IR-U

WB

si

gnal

Nar

row

band

si

gnal

Figure 2.2: IR-UWB signal versus traditional narrowband signal.

• High-precision ranging and positioning. The high time domain resolu-tion (on the order of subnanoseconds) of UWB pulses also enables rangingand positioning with sub-meter precision.

• Superior penetration capability. Spanning over a wide range of frequen-cies, UWB signals are able to penetrate effectively through various materials.


This is particularly true for the low frequencies of UWB signals, which havelong wavelengths leading to superior penetration capability through differentmaterials [41].

• Low-power and low-cost transmitter implementation. The UWBpulse can be generated and propagate without any RF mixing stage andthus allows extremely low-power and low-cost transmitter implementation.

Due to the above-mentioned features and benefits, UWB technology is attractivein a wide range of application scenarios as depicted in Figure 2.3 and summarizedas follows.

• High data rate applications (IEEE 802.15.3a [42]). UWB links areexpected to replace cable connections with high data rate that range from100 kb/s to 100 Mb/s. For example, wireless connectivity between a host PCand associated peripherals (mouse, keyboard, USB, printer, etc.), home andprofessional media networking (digital TV, audio CD, DVD, digital camera,etc.).

• Low data rate applications (IEEE 802.15.4a [43]). IR-UWB is widelyconsidered for low data rate applications combined with precise tracking andpositioning capability, such as WSN and RFID. In fact, the low-rate wire-less personal area network (WPAN) standard IEEE 802.15.4a is primarilyfocused on tracking and positioning applications. For instance, personneland asset tracking, mobile inventory, environmental monitoring and controlof home/office. Industry pioneers have already launched several commercialproducts for identification and real-time location system (RTLS), such as De-caWave [44], Time Domain [45], Ubisense [46], Zebra [47] and BeSpoon [48].

• Radar applications. The superior penetration capability makes UWB at-tractive to radar imaging systems, such as through-wall radar imaging [49],medical imaging [50] and ground penetration radar [51]. Besides, due to thehigh time-domain resolution and accurate ranging capability, UWB can beadopted for vehicular radar systems for road anti-collision [52], parking guid-ance [53], etc.

2.2 Generation of UWB Signals

Generally, there are two types of UWB signaling scheme: 1) Multi-Band orthogonalfrequency division multiplexing (MB-OFDM) UWB, and 2) Impulse Radio (IR)UWB.

The MB-OFDM scheme divides the large UWB spectrum into multiple sub-bands with a -10 dB bandwidth of at least 500 MHz, and in each sub-band operateas a conventional narrow-band radio. Thus the information bit stream is interleaved

2.2. GENERATION OF UWB SIGNALS 17

1k

10k

100k

1M

10M

100M

1G

10m 100m 1km

Scalability and trade-off between data rate and range

High speed data communication

Positioning and target detection

Dat

a ra

te (

bits

/s)

Operation range

Figure 2.3: UWB application scenarios.

in several parallel low-data-rate streams, with different and orthogonal modulationcarriers (subcarriers) for each of them [54]. MB-OFDM UWB has been primarilyused for short range high speed wireless communications such as wireless USB andstreaming video. It has been proposed for the physical layer standard of futurehigh speed WPAN (IEEE 802.15.3a). In this MB-OFDM WPAN proposal, the 3.1to 10.6 GHz frequency band is divided into 14 sub-bands with 528 MHz bandwidthfor each that may be added or dropped according to the interference from or toother systems [55]. It supports high data rates up to 480 Mb/s, and the desiredoperation range is 2 m for 480 Mb/s and 10 m for 110 Mb/s [55]. With hundreds ofmilliwatt level power consumption, the MB-OFDM UWB systems are not suitablefor power-constrained RFID and WSN applications.

On the other hand, IR-UWB transmits ultra-short pulses spanning up to sev-eral GHz frequency band without dividing it into sub-bands. The very low dutycycle leads to low average transmission power. Generally there are two differentapproaches for IR-UWB signal generation, carrier-based approach (e.g., [56, 57])and carrier-less approach (e.g., [58, 59]).

The carrier-based IR-UWB signal can be generated by up-converting a basebandenvelope to a center frequency in the targeted UWB band (mixing the envelope witha local oscillator (LO)) [60]. This architecture offers better spectrum utilizationenabling spectrum diversity and tunability [61]. Besides, the generated signal withphase information relaxes the signal processing at the receiver side to some extent.


However, the mixer and the LO operating at the pulse center frequency bring abouthigh circuit complexity and power consumption, which is unaffordable for passivetags.

The carrier-less approach directly generates the IR-UWB pulse in basebandwithout any frequency conversion, eliminating the necessity of LOs and mixers.Thus, the implementation of the carrier-less IR-UWB transmitter can be extremelylow-complexity and low-power. The impulse is usually generated using logic gateswith sub-nanosecond delay elements defining the pulse width, followed by a pulseshaping filter [62].

A bell shape Gaussian pulse can be generated in the easiest way by a pulsegenerator. The Gaussian pulse can be expressed as:

p(t) = ± 1√2πσ2

e−( t22σ2 ) = ±

√2αe−

2πt2α2 (2.3)

where α2 = 4πσ2 is the shape factor and σ2 is the variance. The Gaussian deriva-tives which remove the DC offset of the basic Gaussian pulse and better fall intothe UWB band are ideal to be adopted as the carrier-less IR-UWB signals. Anexample of waveforms and power spectral density (PSD) of the first 12 derivativesof Gaussian pulse are shown in Figure 2.4 and Figure 2.5, respectively [63].

2.3 IR-UWB Modulations

Different modulation schemes can be used to encode data onto IR-UWB pulses.Figure 2.6 depicts the most widely used modulation schemes in IR-UWB systems:binary phase shift keying (BPSK), on-off keying (OOK) and pulse position modu-lation (PPM).

BPSK modulation encodes binary data onto the polarity of IR-UWB pulses.The modulated signal can be formulated as

s(t) =+∞∑

k=−∞dkptx(t− kTf ), dk = 1,−1 (2.4)

where ptx represents the transmitted pulse waveform and Tf is the pulse repetitioninterval. BPSK promises 3 dB gain in power efficiency comparing with OOK andPPM. Besides, it provides a smooth power spectrum since the change of polaritycan remove the PSD spectral lines [55]. However, it only works for carrier-basedIR-UWB systems where accurate phase generation and coherent demodulation isfeasible.

OOK modulation uses the presence and absence of the transmitted pulses torepresent a data bit “1” and “0”, respectively. The OOK modulated signal is givenby

s(t) =+∞∑

k=−∞dkptx(t− kTf ), dk = 1, 0 (2.5)

2.3. IR-UWB MODULATIONS 19

−1

0

1

−1

0

1

−1

0

1

−2 0 2

x 10−9

−1

0

1

−2 0 2

x 10−9

−2 0 2

x 10−9

Time (s)

Am

plitu

de (

V)

Figure 2.4: Waveforms of the first 12 derivatives of Gaussian pulse with α =0.714ns.

This modulation scheme has no restrictions on the type of IR-UWB signals, nomatter carrier-based or carrier-less. And it only requires non-coherent demodula-tion. OOK is the simplest scheme for IR-UWB modulation, resulting in the lowestachievable complexity of transceiver design.

PPM is another modulation scheme which allows non-coherent detection. Ituses a time shift δPPM with respect to the position reference to distinguish a databit “1” and a data bit “0”. The modulated signal can be expressed as

s(t) =+∞∑

k=−∞ptx(t− kTf − dkδPPM ), dk = 1, 0 (2.6)

Comparing with OOK modulation, PPM is efficient in minimizing the spectralpeaks caused by pulse repetitions in time [55]. However, it suffers from its sensitivityto timing mismatch caused by imperfect synchronization.


0 2 4 6 8 10 12−400

−350

−300

−250

−200

−150

−100

−50

0

Frequency (GHz)

PS

D (

dBm

/MH

z)

1st derivative

12th derivative

FCC UWB indoor emission mask

Figure 2.5: PSD of the first 12 derivatives of Gaussian pulse with α = 0.714ns andpulse repetition interval Tf = 100ns.

2.4 UWB Channels

After passing through a propagation environment, the UWB signal arrived at thereceiver side is very much different from the original transmitted signal. Due to theobstacles in the environment, e.g., walls, ceilings and furniture, the received signalare multiple versions of the transmitted signal with different attenuated amplitudesand delays. Besides, it also suffers from noises and possibly interferences. Thereceived signal can then be modeled as [55]

r(t) =L−1∑l=0

αls(t− τl) + n(t), (2.7)

where L is the total number of multi-path components. αl and τl denotes the fadingcoefficient and the delay of the l-th path, respectively. The term n(t) is the additivewhite Gaussian noise (AWGN) with double-sided power spectral density N0/2.

In the UWB engineering community, IEEE 802.15.3a/4a channel models arewidely adopted. IEEE 802.15.3a channel models [64] are characterized for highdata rate applications while IEEE 802.15.4a channel models [65] are characterizedfor low data rate applications. With the focus of this dissertation on the low data

2.4. UWB CHANNELS 21

Data 1 0 1

BPSK

OOK

PPM

PPM

Figure 2.6: Data modulation options for IR-UWB signals.

rate IoT applications, Table 2.1 gives a summary of the 802.15.4a channel modelsfor different propagation environment. An example of one channel realization foreach channel model is shown in Figure 2.7.

Table 2.1: IEEE 802.15.4a Channel Models.

Channel Model Propagation Environment LOS/NLOSCM1 Residential (7-20 m) LOSCM2 Residential (7-20 m) NLOSCM3 Indoor Office (3-28 m) LOSCM4 Indoor Office (3-28 m) NLOSCM5 Outdoor (5-17 m) LOSCM6 Outdoor (5-17 m) NLOSCM7 Industrial (2-8 m) LOSCM8 Industrial (2-8 m) NLOSCM9 Open Outdoor NLOS


2.5 IR-UWB Receiver Overview

2.5.1 Review of IR-UWB Receiver ArchitecturesOver the last decade, various architectures were introduced in literature for the IR-UWB receiver design. Those different candidates range from high performance alldigital receivers [66–68], analog correlation receivers [69–72], low complexity trans-mitted reference (TR) receivers [73,74] and energy detection (ED) receivers [75–78],to the recent emerging compressed sensing (CS) receivers [79–82]. They mainlydistinguished from each other in two aspects. One aspect is where to locate theanalog-to-digital converter (ADC) or indeed the sampling rate of the ADC. And theother is how to implement the signal correlation in the analog domain: with whichsignal, random sequence or pulse template, and in a coherent or non-coherent way.This section reviews the basic architecture of those receiver alternatives, and themain pros and cons of each architecture are summarized.

All Digital Receiver

There are many works, e.g., [66–68] target on the all digital or mostly digital im-plementations of IR-UWB receiver in order to reduce the analog complexity. Inthose architectures, the ADC is placed as close to the antenna as is reasonable.Figure 2.8a shows an example of the all digital receiver architecture. The incomingsignal is directly digitized after amplification (and quadrature down-conversion forthe carrier based IR-UWB signals), moving most of the signal processing, such asmatched filtering and RAKE reception, to the digital domain. This architectureguarantees an optimal receiver performance and reserves the high time domainresolution of the incoming signal. Besides, it provides a large design flexibility tothe resulting receiver, enabling fast prototyping and reconfiguration. However, theNyquist-rate sampling ADC is required in this architecture, which is extremelycomplex and power hungry. This is even more unacceptable for the carrier-less im-pulse radio, where no frequency down-conversion can be performed ahead, leadingto a big challenge for the state-of-the-art ADC design.

• Pros: optimal performance, high flexibility.

• Cons: Nyquist-rate ADC, high complexity implementation and power con-suming.

Analog Correlation Receiver

To avoid the complex and power consuming Nyquist-rate ADC, analog correlationreceiver as shown in Figure 2.8b performs the matched filtering in the analog do-main. The receiver performance is however very much depend on the accurateanalog generation of the pulse template and is very sensitive to timing mismatchand clock jitter. The quadrature analog correlation (QAC) receiver [70, 71] con-quers these challenges by using a quadrature windowed sine-wave as the matching

2.5. IR-UWB RECEIVER OVERVIEW 23

template [72]. The QAC receiver is usually adopted in carrier-based IR-UWB sys-tems. The windowed sine-wave which can be easily generated in the analog domainis then well matched with the transmitted signal [26]. Accurate frequency com-ponent, such as phase-locked loop (PLL), is critical for the QAC receiver. Thus,to relax the complexity and power consumption of the PLL, QAC receiver usuallyworks at a relatively low UWB frequency band with limited bandwidth (e.g., 500MHz).

• Pros: superior performance.

• Cons: require accurate pulse template, sensitive to timing mismatch andclock jitter.

Transmitted Reference Receiver

The architecture of the transmitted reference (TR) receiver is given in Figure2.8c [73, 74]. The basic principle of the TR scheme is to transmit a pair of UWBpulses per frame, one serves as the reference (unmodulated) pulse and the otheris the data carrying (modulated) pulse. The reference pulse is delayed throughan analog delay line to serve as the correlation template for the data pulse [72].As a result, the TR receiver automatically captures all the multi-path componentsand avoids the necessity of accurate channel estimation which is highly importantfor an analog correlation receiver. It is especially beneficial when transmittingthrough an unknown channel with severe distortion or in a dense multi-path envi-ronment [83, 84]. However, this advantage comes at the expense of several dB biterror rate (BER) performance degradation due to the accumulated noise power.Another drawback of the TR scheme is the energy wasted for transmitting the ref-erence pulses. Besides, it is very challenging to integrate wideband delay lines withlarge delay values on silicon [85].

• Pros: low complexity sub-Nyquist sampling, avoid stringent channel estima-tion.

• Cons: a few dB performance degradation, energy wasted for reference pulse,implementation challenge of wideband delay lines with large delay values.

Energy Detection Receiver

Energy detection (ED) is the most simple receiver architecture suggested for IR-UWB systems, as depicted in Figure 2.8d [75–78]. It is based on the autocorrelationof the incoming signal [72]. Thus, neither accurate template generation nor analogdelay lines are required in ED receivers. Although lowest power consumption andimplementation cost can be achieved with the ED scheme, it suffers from a degra-dation of BER performance and time domain resolution compared with coherentreceivers.


• Pros: sub-Nyquist sampling, lowest power consumption and implementationcost.

• Cons: degradation of BER performance and time domain resolution.

Compressed Sensing Receiver

Compressed sensing (CS) receiver shown in Figure 2.8e is recently proposed aimingat reducing the ADC sampling rate without sacrificing the time domain resolutionof the UWB signal [79–82]. It is based on the compressed sensing theory whichsuggests that, a signal can be acquired and recovered with far fewer samples thanthat required by Nyquist sampling theorem as long as the signal has a sparserepresentation in some domain [86,87]. For the recovery of each transmitted pulse,multiple CS measurement processes are required which either results in repeatedtransmission of the information pulse or increased parallel branches of the front-endhardware. So far the CS-based IR-UWB receiver has not been throughly studied,especially in the hardware implementation perspective.

• Pros: sub-Nyquist sampling without sacrificing time domain resolution.

• Cons: multiple CS measurements for each pulse recovery.

2.5.2 Promising Rx Candidate for IoT applicationsIdeally, a receiver that can provide optimal performance, high operation flexibility,and at the same time enable low power and low cost integrated hardware implemen-tation is desired for future RFID and WSN applications towards the IoT. However,accommodating all these advantages in one single receiver architecture might not bepossible. In Section 2.5.1, several most common receiver architectures are reviewed.It reveals that each Rx option has its own trade-offs in terms of operation flexi-bility, performance, hardware complexity, power consumption, etc. Furthermore,the applicability of the Rx candidate also subjects to the type of IR-UWB signal(carrier-based or carrier-less), modulation scheme, channel characteristic (sparse ordense multi-path) as well as application scenario.

Concentrated on low power and low cost RFID and WSN applications, thisdissertation mainly copes with the carrier-less IR-UWB signals. As mentioned inSection 2.2, the carrier-based signal requires the deployment of the mixer and thelocal oscillator operating at the pulse center frequency for both Tx and Rx. Theresulting high circuit complexity and power consumption is unaffordable for passivetags. Currently, the implementation of all digital receiver is not feasible for carrier-less IR-UWB signals. Without pre-down-conversion, direct sampling of the signalwith several GHz bandwidth and center frequency is beyond the capability of thestate-of-the-art ADC design. On the other hand, the QAC receiver only preferablefor the reception of the carrier-based IR-UWB signal with limited bandwidth (e.g.,

2.6. SUMMARY 25

500 MHz). The transmitted reference receiver enables low complexity sub-Nyquistsampling, nevertheless, to the best of the author’s knowledge, it hasn’t been imple-mented in hardware so far due to the difficulty of integrating wideband delay lineswith large delay values on silicon.

The sub-Nyquist sampling energy detection receiver appears as a promisingcandidate. Recently, several literatures [76–78] about the ASIC design of the EDreceiver have published, which demonstrate the feasibility of low-power and low-complexity implementation. Besides, research efforts continue to improve the per-formance of ED receivers so as to close the performance gap with respect to thecoherent ones.

The compressed sensing based IR-UWB receiver emerged recently as anotherpromising candidate. The most attractive feature of the CS receiver is sub-Nyquistsampling without sacrificing any time domain resolution, which is highly desired instringent ranging/positioning applications. Several research pioneers [79, 88] haveintroduced the compressed sensing theory into the IR-UWB receiver design eversince the CS theory proposed in 2006 [86, 87]. Nevertheless, the research oppor-tunities for the design of CS receiver are still wide open. Many issues, especiallyin the hardware implementation perspective, need to be addressed before enteringpractical applications.

2.6 Summary

This chapter provides a comprehensive introduction of the UWB technology. Thefundamentals of the UWB signal as well as transmission related issues such asregulation, modulation and channel modeling were studied, which serves as thebackground knowledge for the following chapters. The highlight of this chapterwas given to the overview of IR-UWB receiver design. Various receiver architec-tures were investigated and the pros and cons of each architecture were discussed.Based on the discussion of all the trade-offs of the receiver design as well as theconsideration of the whole IR-UWB system for IoT applications, we narrowed ourfocuses to the two most promising sub-Nyquist sampling candidates: the energydetection receiver and the compressed sensing receiver. The design issues of theED and CS receiver will be addressed in the next two chapters respectively.


0 50 100 150 200 250−0.3

−0.2

−0.1

0

0.1

0.2

0.3

0.4

0.5CM1 Channel Impulse Response

Time (ns)

Am

plitu

de G

ain

0 50 100 150 200 250−0.2

−0.15

−0.1

−0.05

0

0.05

0.1


Time (ns)

Am

plitu

de G

ain

0 50 100 150 200 250−0.35

−0.3

−0.25

−0.2

−0.15

−0.1

−0.05

0

0.05

0.1


Time (ns)

Am

plitu

de G

ain

0 20 40 60 80 100 120−0.2

−0.15

−0.1

−0.05

0

0.05

0.1


Time (ns)

Am

plitu

de G

ain

0 50 100 150 200 250 300−0.25

−0.2

−0.15

−0.1

−0.05

0

0.05

0.1

0.15


Time (ns)

Am

plitu

de G

ain

0 50 100 150 200 250 300 350−0.2

−0.15

−0.1

−0.05

0

0.05

0.1

0.15


Time (ns)

Am

plitu

de G

ain

0 20 40 60 80 100 120 140−0.4

−0.3

−0.2

−0.1

0

0.1

0.2

0.3


Time (ns)

Am

plitu

de G

ain

0 20 40 60 80 100 120 140 160 180 200−0.08

−0.06

−0.04

−0.02

0

0.02

0.04


Time (ns)

Am

plitu

de G

ain

Figure 2.7: Example of one channel realization for CM1 - CM8, respectively.

2.6. SUMMARY 27

ADCDigital

Back-end LNA

Nyquist Rate

(a) All digital receiver.

ADCDigital

Back-end LNA

Pulse Rate

Pulse Template

(b) Analog correlation receiver.

ADCDigital

Back-end LNA

Pulse Rate

Delay

(c) Transmitted reference receiver.

ADCDigital

Back-end LNA

Pulse Rate

(d) Energy detection receiver.

ADCDigital

Back-end LNA

Pulse Rate

Random Sequence

(e) Compressed sensing receiver.

Figure 2.8: Different architectures of IR-UWB receiver.

Chapter 3

Sub-Nyquist Sampling IR-UWB Receiver I:Energy Detection

Energy detection is an attractive scheme for the IR-UWB receiver design. By simplyperforming “square, integrate, and symbol-rate sampling”, it provides a low powerand low complexity solution for the receiver design. In this chapter, a flexible andlow-power energy detection IR-UWB receiver for RFID and WSN applications ispresented, with a highlight on the design of a highly programmable timing circuitwhich is the key enabler for the receiver flexibility and reconfigurability. Followingthat, the back-end signal processing aiming to improve the link performance isdeveloped, which includes the optimum threshold estimation for OOK modulatedsignal, the adaptive synchronization and integration region optimization. Finally, alow complexity method for burst packet detection in the context of wireless poweredUWB RFID systems is proposed.

3.1 Basic Principle of Energy Detection

The block diagram of an energy detection IR-UWB receiver is shown in Figure 3.1.The received signal is first passing through a bandpass filter (BPF), amplified by alow noise amplifier (LNA), and then fed into the square law device and the integra-tor. The analog-to-digital converter (ADC) samples the output of the integrator ateach data (also referred to as symbol) boundary. The digitized samples are thensent into the digital back-end for synchronization and demodulation.

As discussed in Section 2.3, both pulse position modulation (PPM) and on-off keying (OOK) modulated signal can be demodulated by an ED receiver. Andmultiple pulses can be combined to represent one bit information in order to achieveprocessing gain combating noises and interferences. Figure 3.2 sketches an exampleof the signaling of energy detection scheme with OOK modulated signal.

29

30CHAPTER 3. SUB-NYQUIST SAMPLING IR-UWB RECEIVER I: ENERGY

DETECTION

Digital Back-end

Timing Circuit

N-bit ADC( )dt 2

LNASquare Law Integrator

BPF

Figure 3.1: Block diagram of an energy detection IR-UWB receiver. Figure adaptedfrom Paper I.

)(tt x

)(2 trt

rst

int

out_int threshold

t0

td tw

T p

PGTT psym *

t

t

t

t

Figure 3.2: Signaling of energy detection with OOK modulated signal. Figureadapted from Paper I.

3.2 A Flexible and Low-Power ED Receiver

The design and implementation of an energy detection IR-UWB receiver for RFIDand WSN applications is presented in this section. In order to support adaptive linkin dynamic circumstances, e.g., scalable system capacity, different power and energycondition, quality of service (QoS) requirement, coverage, and channel environment,the proposed receiver is highly flexible and reconfigurable in terms of pulse rate(512 kHz-33 MHz), processing gain (0-18 dB), integration window length as wellas modulation schemes (OOK/PPM) [24]. The ASIC prototype of the receiver isimplemented and fabricated in 90 nm CMOS process. The power consumption ofthe ASIC is 16.3 mW at 1 V power supply, which guarantees a minimum energyconsumption of 0.5 nJ/bit. Measurement results exhibit that -79 dBm sensitivity at10 Mb/s with 10−3 BER can be achieved, corresponding to over 10 meters operationrange under the FCC regulation [24].

3.2. A FLEXIBLE AND LOW-POWER ED RECEIVER 31

Figure 3.3: The architecture of the proposed flexible energy detection IR-UWBreceiver. Figure adapted from Paper II.

3.2.1 Receiver Architecture

The architecture of the proposed flexible energy detection IR-UWB receiver isshown in Figure 3.3. To facilitate and enhance its flexibility, the receiver prototypeis partitioned into an ASIC part and a field-programmable gate array (FPGA) part.A 3-5 GHz analog front-end (AFE), a timing circuit as well as a baseband controllerare implemented in ASIC. The AFE is composed by an LNA, a wideband variablegain amplifier (VGA), and two channels of signal path comprised of a square lawdevice, an integrator and a baseband VGA. On the other hand, the digital back-endis implemented in FPGA. It is responsible for baseband processing such as channelestimation, pulse/symbol level synchronization, data demodulation, time-of-arrival(TOA) ranging and so on. Besides, the back-end also handles the real time controlof the ASIC to accommodate different operation modes [24].

The two-interleaved-channel architecture of the AFE is adopted to assist fastsynchronization and TOA estimation where parallel processing is required. Besides,it is also applied to perform PPM demodulation when the integration windows for“1” and “0” are overlapped (δPPM is shorter than the optimal integration windowlength) [24].

The highly programmable timing circuit together with the high speed basebandcontroller is the core of the proposed flexible receiver. It controls the operationof the two channels of integrator bank independently, enabling a wide range ofreconfigurability, such as pulse rate, processing gain, integration window phase andlength, etc. Hence, receiver performance optimization, link adaptation as well asmulti-mode operation can be achieved.


DETECTION

Table 3.1: Comparisons of the Proposed IR-UWB Receiver with Previous PublishedWorks (JSSC 101 [78], TCAS 09 [77], JSSC 102 [70], ISSCC 06 [71]). Table adaptedfrom Paper II.

Ref. JSSC 101 TCAS 09 * JSSC 102 ISSCC 06 This Work

Technology 90 nm 180 nm 130 nm 180 nm 90 nm

Architecture ED ED QAC/CB-IR QAC/CB-IR ED

Frequency band 3-5 GHz † 7.25-8.5 GHz 0-960 MHz 3-5 GHz 3-5 GHz

Max. Data Rate 16 Mb/s 100 kb/s 39 Mb/s 20 Mb/s 33 Mb/s

Power 22.5 mW 5.4 mW 4.2 mW 28.8 mW 16.3 mW ◆

Min. Energy/bit 1.4 nJ/bit 54 nJ/bit 0.108 nJ/bit 1.44 nJ/bit 0.5 nJ/bit

Normalized Sens.@100 kb/s -98 dBm -89 dBm -78 dBm n.a. -99 dBm

† 500 MHz sub-band is used;

* simulated, only mixer and integrator;

◆ the backend and the ADC are not included.

3.2.2 Implementation and Measurement Results

The ASIC part of the receiver prototype is implemented in UMC 90 nm CMOSprocess, and the flexible digital back-end is implemented using Altera Cyclone IIIFPGA development kit. A high speed mezzanine card (HSMC) from Altera isemployed to bridge the ASIC part and the FPGA part. Figure 3.4 shows the chipmicrograph of the receiver ASIC, which has an active area of 1.1 mm2 [24].

Measurement results demonstrate that the power consumption is as low as 16.3mW for single channel operation and 22.3 mW for dual channel operation. Ac-cordingly, 0.5 nJ/bit energy consumption can be achieved at 33 Mb/s data rate.The communication link has been verified by pairing the proposed receiver withthe UWB tag in [25]. The measurement setup is shown in Figure 3.5. A coaxialcable and a variable attenuator are used to connect the UWB transmitter and thereceiver, modeling a multi-path free environment. The BER measurement resultsare plotted in Figure 3.6. As can be seen from the figure, targeting at 10−3 BER,-79 dBm sensitivity at 10 Mb/s data rate is achieved. Comparing with the idealimplementation1, the prototype has a 12 dB loss which is mainly due to the noisefigure of the analog front-end as well as the non-ideal issues of the digital baseband,such as timing offset and non-optimal threshold. The proposed receiver guaranteesover 10 meters operation range at 10 Mb/s data rate working with an ideal 3-5GHz UWB transmitter which provides an irradiated power of -8.3 dBm reachingthe FCC mask at the Tx antenna. The key features of the proposed receiver arecompared with previous works [70, 71, 77, 78] in Table 3.1. For fair comparison, asensitivity of -99 dBm is obtained by scaling the data rate down to 100 kb/s [24].

More detailed explanation on the system design and circuit implementation ofthe proposed receiver can be found in Paper II in appendices.

1The ideal implementation refers to the theoretical simulation results without considering anyloss in hardware implementation.

3.3. HIGHLY PROGRAMMABLE TIMING CIRCUIT 33

Band G

ap Ref.

Figure 3.4: Chip micrograph of the proposed IR-UWB receiver. Figure adaptedfrom Paper II.

RX chip & FPGA

UWB Tag

Attenuators

PC

SPI and USB

SpectrumAnalyzer Reference Clock

Random Data Gen.

FPGAUWB TagRX ASIC

(a) (b)

Figure 3.5: Measurement setup. (a) Link connections, and (b) test environment.Figure adapted from Paper II.

3.3 Highly Programmable Timing Circuit

In an energy detection IR-UWB receiver as shown in Figure 3.1, the timing circuitis responsible for generating accurate timing signals (integration windows and resetsignals) that control the operation of analog blocks. Different design methodologiescan be adopted to implement a timing circuit. A two-stage variable delay line is


DETECTION

-100-98 -96 -94 -92 -90 -88 -86 -84 -82 -80 -78 -7610-4

10-3

10-2

10-1

100

Received Power (dBm)

BE

R

Ideal ImplementationMeasured Results

12 dB Loss

Figure 3.6: BER versus average Rx input power (10 Mb/s, OOK modulation,uncoded, wired connection). Figure adapted from Paper II.

employed in [71] to generate multi-phased timing signals, one for coarse delay steps(1 ns) and the other for fine delay steps (60 ps). Relatively high timing resolutioncan be achieved by using analog delay elements. However, it is sensitive to processand temperature variations. In [89], the timing circuit has four identical cascadedstages and each stage is comprised of a multiplexer and a divide-by-two with dif-ferential outputs. The phase of the timing signal is selected via the control bitsconnected to the multiplexers. Despite of its simplicity, this design may lead tophase uncertainty and thus increase the synchronization difficulty of the receiver.These previous works are aiming at fixed pulse/data rate communications. Besides,none of these works is considering the requirement of integration window optimiza-tion on the timing circuit. In Paper I [27], we propose a highly programmabletiming circuit which overcomes the above mentioned drawbacks and also enablesmulti-mode operations and flexibilities.

3.3.1 Proposed Timing SpecificationLet’s first write the integrated energy of the i-th pulse and j-th symbol as

Ep(i) =∫ t0 +i Tp + td + tw

t0 +i Tp + td

r2(t)dt (3.1)

Es(j) =∑i=(j+1)PG+SS−1

i=jPG+SSEp(i) (3.2)

3.3. HIGHLY PROGRAMMABLE TIMING CIRCUIT 35

where r(t) denotes the received signal. And the parameters t0, Tp, td, tw, PG andSS are used to define the timing specification.

• t0: integration reference time.

• Tp: pulse repetition interval. This parameter is designed to be programmablein order to adapt to different pulse rates.

• td: phase delay of the integration window, introduced for pulse level synchro-nization. The tuning range of this parameter should be able to cover thewhole pulse repetition interval. And the tuning resolution corresponds to thelimit of synchronization and ranging accuracy (e.g., 1 ns phase resolution iscorresponding to 30 cm ranging accuracy).

• tw: integration window width. To achieve optimum receiver performance,this parameter should be adjusted to the optimum integration window widthwhich is related to the delay spread of the UWB channel and varies with thechannel environment.

• PG: processing gain. This parameter is adopted to trade data rate for trans-mission distance and vise versa. It is better to be programmable so as toobtain the best trade-off in different transmission circumstances.

• SS: This variable is introduced for the symbol level synchronization. Whenmultiple pulses are combined to stand for one symbol, it must be employedand tuned during the synchronization procedure searching for the startingpulse of a symbol.

3.3.2 Circuit Design and ImplementationThe proposed timing circuit is referenced at 900 MHz clock2. Considering the highfrequency reference clock and the high programmability, neither the full synthesizeddigital design nor the full custom design is a proper solution for the proposed timingcircuit. Thus, a mixed-signal design flow is adopted and the architecture of thecircuit is shown in Figure 3.7.

The block Div4 which is comprised of two differential D-Flip-Flops (DFF) con-nected in a closed loop converts the 900 MHz reference clock down to 225 MHzwith 0◦, 90◦, 180◦ and 270◦ phases, respectively. The four clocks are then fedinto a multiplexer and the chosen clock Q1 is used to fine tune the phase of theintegration window with 1.1 ns resolution. The digital part of the circuit is mainlycomposed of several programmable dividers and counters. DIV_1 generates thepulse_clk with a variable pulse rate controlled by PR < 5 : 0 >. CNT_1 withcount enable signal pulse_clk is acting as a delay cell, and the delay resolution is

2The proposed timing circuit is employed in the UWB/UHF hybrid RFID system with UWBuplink and UHF downlink [24,25,28]. The 900 MHz reference is chosen to reuse the UHF carrierof downlink.


DETECTION

equal to the period of Q0. The output of CNT_1 is a delayed version of pulse_clkand is used to indicate the rising edge of the integration window. CNT_2 furtherdelays the pulse_clk to specify the falling edge. That is to say, CNT_1 tunes thephase of the integration window and CNT_2 adjusts the window width. Each ofthe two counters is followed by a DFF clocked at Q1 for the phase fine-tuning. Andeventually the integration window signal is generated by a SR − latch as the finalstage. Likewise, DIV_2 and CNT_3 are combined to generate the symbol_clk.DIV_2 adjusts the symbol rate (or equivalently the processing gain) and CNT_3tunes the phase of the symbol_clk for the symbol level synchronization. The resetsignal of the integrator should be triggered right after the last integration windowof each symbol duration. Therefore, two DFFs are used to retime the symbol_clk.And following that, a rising edge triggered glitch generator is adopted to generatethe reset signal. Figure 3.8 displays the diagram of the above mentioned timingsignals [27].

The design has been fabricated in 90 nm CMOS process, with 219 µW powerconsumption and 190*295 µm2 die area. Full custom design is used to implementthe high frequency divider Div4, the multiplexer and the glitch generator. And therest parts are implemented by full synthesized digital design. The performance ofthe proposed timing circuit is illustrated in Table 3.2. Although the timing circuitis intended for running at 900 MHz clock, it has been measured under differentreference clocks and it functions properly up to 4 GHz.

Table 3.2: Performance of the Timing Circuit at 900 MHz Reference Clock.

Parameters Control Word Value/RangePulse Rate PR<5:0> 112.5 MHz ∼ 3.57 MHz

Phase Resolution P<7:0> 1.1 nsWindow Width W<5:0> 4.4 ns ∼ (Tp-4.4 ns)Processing Gain PG<5:0> 1 ∼ 63 pulses per symbol

3.4 Optimum Threshold Estimation for OOK Modulated Signal

In energy detection of OOK modulated signals, the information is demodulatedby comparing the integrated energy with certain threshold. The performance ofthe receiver heavily depends on the chosen threshold. The optimum thresholdestimation for energy detection receivers is an essential yet challenging issue. In [90],the average of integrated energy of bits ‘0’ and ‘1’ is selected as the threshold,which is quite simple but non-optimal. In [91], the optimum threshold is estimatedby Gaussian approximation of the Chi-square statistics. However, this approachrequires extremely complex computation. Besides, it only works for situations withlarge degree of freedoms (large bandwidth and integration interval). And in [92],the threshold estimation is based on bit error rate (BER) calculation over several

3.4. OPTIMUM THRESHOLD ESTIMATION FOR OOK MODULATED SIGNAL 37

clk ctrl outDIV_1

clk ctrl

outCNT_1en

clk ctrl

outCNT_2en

clk ctrl

outDIV_2

clk ctrl

outCNT_3en

D

CKQDFF

D

CKQDFF

S

RQSR

latch

clkDiv4

D

CK

QDFF

D

CKQDFF Glitch_gen

Digital partMixed-signal part

PR<5:0>: select pulse ratePG<5:0>: select processing gainP<7:0>: select window phaseW<5:0>: select window widthSS<5:0>: select the first pulse in each symbol

00

900

1800

2700

Figure 3.7: Architecture of the proposed timing circuit. Figure adapted from PaperI.

PGTT psym *

Ref_clk

Q0

pulse_clk

Q1

Int

Symbol_clk_raw

Symbol_clk

Rst

tdf

T p

tdf tdc

td tw

offsetsymbol _

Figure 3.8: Timing diagram. Figure adapted from Paper I.

test values of the threshold and the one minimizes the BER is considered to beoptimal. Large amount of training symbols are required in this method and thusleads to high energy consumption per useful bit. In Paper IV [29], we present apractical scheme based on look-up table for the threshold optimization, enablinglow complexity implementation.


DETECTION

3.4.1 Theoretical DerivationFor energy detection of a received OOK signal in additive white Gaussian noise(AWGN), the false alarm probability (error probability for bit ‘0’ denoted by Pe,0)has a centralized Chi-square distribution while the probability of miss detection(error probability for bit ‘1’ denoted by Pe,1) follows the non-centralized Chi-squaredistribution. Similar to [91], the average error probability for equiprobable ‘0’/‘1’can be calculated as

Pe = 0.5 (Pe,1 + Pe,0) (3.3)

Pe,1 = 1−QM

(√4EbwN0

,

√2ξN0

)(3.4)

Pe,0 = e−ξ/N0

M−1∑k=0

1k!

(ξ

N0

)k(3.5)

where QM is the generalized Marcum-Q function of orderM . HereM is the degreeof freedom defined by 2M = 2BT + 1 = 2BNstw + 1, in which B is the signalbandwidth, Ns is the number of pulses per symbol and tw denotes the integrationtime interval. Assume the useful received signal energy averaged over bits ‘0’ and‘1’ is represented by Eb, then Ebw is the averaged useful signal energy covered bythe integration window. And Ebw = Eb when all the multi-path components arecovered by the integration window [29].

The optimum threshold ξ0 is located at the intersection of the two Chi-squaredistributions, which yields

e−ξ0/N0

M−1∑k=0

1k!

(ξ0

N0

)k= 1−QM

(√4EbwN0

,

√2ξ0

N0

)(3.6)

Then we normalize the threshold as

λ0 = ξ0 −m0

m1 −m0(3.7)

where λ0 is the optimum normalized threshold. m0 and m1 are the means ofintegrated energy for bit ‘0’ and bit ‘1’, respectively. Ebw and N0 can be calculatedby m0 and m1 as

Ebw = m1 −m0

2 (3.8)

N0 = m0

BT(3.9)

3.4. OPTIMUM THRESHOLD ESTIMATION FOR OOK MODULATED SIGNAL 39

Using (3.7), (3.8) and (3.9) we obtain

ξ0

N0= BT + 2λ0

EbwN0

(3.10)

Substitute (3.10) into (3.6) reveals that the value of λ0 is only subjected to BT andEbw/N0. That is to say, if B, Ns, tw and Ebw/N0 are available, the value of λ0 canbe obtained by numerically searching for the one that minimizes Pe.

3.4.2 Proposed Method for Threshold OptimizationBased on the theoretical basis in Section 3.4.1, a look-up table of the optimumnormalized threshold can be derived and employed in the process of threshold op-timization. Table 3.3 shows an example of the derived look-up table under theconditions of B = 2 GHz (compatible to 3-5 GHz UWB system) and Ns = 1 (onepulse per symbol). tw and Ebw/N0 are chosen as the variables of the look-up table.tw is ranging from 10 ns to 80 ns corresponding to various optimum integrationintervals for the IEEE 802.15.4a UWB channels [65]. And the range of Ebw/N0 isselected considering the requirement for targeted 10−3 BER. In Table 3.3, for eachcase of tw, 10−2 to 10−6 BER are covered.

Table 3.3: Optimum Normalized Threshold. Table adapted from Paper IV.

tw (ns)EbwN0

(dB) 19 18 17 16 15 14 13 12

10 0.36 0.36 0.38 0.38 0.3820 0.38 0.38 0.4 0.4 0.4230 0.38 0.38 0.4 0.42 0.4240 0.4 0.4 0.42 0.44 0.4450 0.42 0.42 0.42 0.44 0.4460 0.4 0.42 0.42 0.44 0.4670 0.42 0.42 0.44 0.44 0.4680 0.42 0.42 0.44 0.46 0.46

Relying on the predefined look-up table, the optimum threshold ξ0 can be esti-mated by following procedures.

1) Estimate m0 and m1 by training symbols comprised of equivalent number ofbits ‘0’ and ‘1’;

2) Calculate Ebw/N0 = (m1−m0)BNstw2m0

;

3) Look up the value of λ0 in Table 3.3 according to the predefined tw and theestimated Ebw/N0;

4) Calculate ξ0 = m0 + λ0(m1 −m0).


DETECTION

12 13 14 15 16 17 18 1910

−6

10−5

10−4

10−3

10−2

10−1

Eb/No (dB)

BE

R

BER with conventional schemeBER with proposed optimum threshold estimation

(a) CM7 channel.

12 13 14 15 16 17 18 1910

−6

10−5

10−4

10−3

10−2

10−1

100

Eb/No (dB)

BE

R

BER with conventional schemeBER with proposed optimum threshold estimation

(b) CM8 channel.

Figure 3.9: BER performance for the proposed threshold optimization in IEEE802.15.4a UWB channels. Figure adapted from Paper IV.

The proposed method for threshold optimization is verified by system simula-tions with IEEE 802.15.4a channel model CM7 and CM8. For CM7 channel, we usean integration interval of 10 ns while 50 ns is adopted for CM8 channel. As shownin Figure 3.9, the simulation results exhibit an appreciable BER improvement overthe conventional method with λ = 0.5.

3.5. ADAPTIVE SYNCHRONIZATION AND INTEGRATION REGIONOPTIMIZATION 41

3.5 Adaptive Synchronization and Integration RegionOptimization

Frame-level synchronization and integration region optimization is also one criticalissue for energy detection IR-UWB receivers. In IR-UWB systems, high link ro-bustness can be obtained by increasing the amplitude of the transmitted impulsessuch that a high peak signal-to-noise ratio (SNR) can be guaranteed at the receiverside. However, this results in a long pulse repetition interval (PRI) typically rang-ing from 100 ns to 1 µs so as to comply with the FCC regulations [93]. Withoutintegration region optimization, such long PRI will lead to a large proportion ofunnecessarily accumulated noise energy, which in turn significantly degrades theBER performance. Existing works such as [94] usually assume that the channeldelay profile is pre-known to the receiver which is not a general case. In PaperV [30], we propose a novel scheme of synchronization and integration region opti-mization based on the numerical analysis of signal energy capture and combinedwith a time-of-arrival (TOA) estimation. The proposed scheme requires no a priorichannel information and is adaptive to channel variations.

3.5.1 Numerical Study of Signal Energy CaptureIn an energy detection receiver, the integration region has a distinct impact on thesignal energy capture as well as the noise accumulation, and eventually affects theBER performance. In order to optimize the integration region, the signal energycapture in terms of the integration interval and the corresponding BER performanceare investigated through numerical analysis.

As can be seen from the channel impulse response shown in Figure 3.10a, indoorUWB channels are featured by dense multi-path components, of which the relativelysignificant ones arrive in the leading part while the latter part contains only trivialones. Therefore, in this numerical study, during each pulse repetition intervalTp, we fix the integration starting point at the first arriving path and adjust theintegration interval tw incrementally. Matlab simulations using 3-5 GHz OOKmodulated signals are performed regarding various indoor UWB channel models,explicitly, IEEE 802.15.4a CM1, CM2, CM7 and CM8. Figure 3.10b plots the BERperformance in terms of the integration interval for different Eb/N0. And Figure3.10c depicts the normalized signal energy capture, which is defined as

ε =E1(tw) − E0(tw)

E1(Tp) − E0(Tp)(3.11)

where E0(·) and E1(·) denote the captured energy for bit ‘0’ and bit ‘1’, respectively,with (·) indicates the integration interval. As can be observed from Figure 3.10b,with respect to the targeted BER, the optimum integration interval is about 20 nsfor CM1, 30 ns for CM2, 10 ns for CM7 and 40 ns for CM8, respectively, whichis much smaller than the total delay spread. In Figure 3.10c, it can be noticed


DETECTION

that the integration interval required to reach ε = 0.8 is approximately the sameas the optimum integration interval. Although the optimum integration intervalchanges along with different channel models and may even vary from one channelrealization to another, the corresponding normalized signal energy capture remainsnearly the same. This derived result is reasonable and intuitive. On one hand,the optimum integration region is where the captured signal energy is much moresignificant than the noise energy. On the other hand, for indoor UWB channels,the significant multi-paths taking up majority of the signal energy (around 80%)are concentrated in the leading several or tens of nanoseconds. To give a generalconclusion, the integration region from the first arriving path until 80% signalenergy is captured can be adopted as the optimal one.

3.5.2 Adaptive SynchronizationThe proposed adaptive frame-level synchronization includes the estimation of in-tegration starting point and optimum integration interval. The estimation of theintegration starting point is actually the same problem as the TOA estimation ofthe leading path. A coarse synchronization is firstly performed locking onto themaximum energy slot. Following that, a fine synchronization determines the inte-gration starting point by carrying out a back-search prior to the maximum slot. Theintegration interval optimization is based on the conclusion we have drawn fromthe numerical study of the signal energy capture in Section 3.5.1. The detailedsynchronization procedure can be found in Paper V [30].

The validity of the proposed adaptive synchronization and integration regionoptimization is demonstrated by system simulations. OOK modulated 3-5 GHz IR-UWB signals with 320 ns PRI and IEEE 802.15.4a channel models are employed inthe simulations. And the synchronization resolution (slot duration) is 5 ns. Figure3.11 presents the simulation results for CM7 and CM8, respectively. Comparingwith the non-adaptive method which fixes the integration interval to be the entirePRI, an appreciable improvement on the BER performance can be observed.

3.6 Low Complexity Burst Packet Detection

In the context of wireless powered UWB RFID systems as shown in Figure 1.2,the data transmission is performed in a burst manner with very short packet inorder to meet the micro-power budget of autonomous power harvesting. Suchburst short packet transmission plus the low duty cycling UWB pulse modulationplaces a stringent challenge at the IR-UWB receiver for timing acquisition andpacket detection. Previous published literature rarely addressed the detection is-sue of such burst pulse burst packet transmission. Existing works on the burstpacket detection all need certain threshold comparison to declare the presence ofthe packet [90, 95, 96]. Furthermore, only fixed threshold can be applied, since theadaptive threshold for variable signal-to-noise ratio (SNR) always requires SNR es-timation which unfortunately can only be processed after a successful acquisition in

3.6. LOW COMPLEXITY BURST PACKET DETECTION 43

0 20 40 60 80 100 120 140 160 180−0.5

−0.4

−0.3

−0.2

−0.1

0

0.1

0.2


Time (ns)

Am

plitu

de G

ain

0 20 40 60 80 100 12010

−6

10−5

10−4

10−3

10−2

10−1

100

CM1

Integration Interval (ns)

BE

R

10 dB12 dB14 dB16 dB18 dB

0 20 40 60 80 100 1200.55

0.6

0.65

0.7

0.75

0.8

0.85

0.9

0.95

1

1.05CM1


Cap

ture

d C

hann

el E

nerg

y

0 20 40 60 80 100 120 140 160 180 200−0.2

−0.15

−0.1

−0.05

0

0.05

0.1

0.15


Time (ns)

Am

plitu

de G

ain

0 20 40 60 80 100 12010

−6

10−5

10−4

10−3

10−2

10−1

100

CM2


BE

R


0 20 40 60 80 100 120

0.4

0.5

0.6

0.7

0.8

0.9

1

1.1CM2


Cap

ture

d C

hann

el E

nerg

y

0 20 40 60 80 100 120−0.25

−0.2

−0.15

−0.1

−0.05

0

0.05

0.1

0.15

0.2


Time (nS)

Am

plitu

de G

ain

0 20 40 60 80 100 12010

−5

10−4

10−3

10−2

10−1

100

CM7


BE

R

10 dB12 dB14 dB16 dB

0 20 40 60 80 100 1200.65

0.7

0.75

0.8

0.85

0.9

0.95

1

1.05CM7


Cap

ture

d C

hann

el E

nerg

y

0 20 40 60 80 100 120 140 160 180 200−0.06

−0.04

−0.02

0

0.02

0.04


Time (ns)

Am

plitu

de G

ain

(a)

0 20 40 60 80 100 12010

−5

10−4

10−3

10−2

10−1

100

CM8


BE

R


(b)

0 20 40 60 80 100 1200.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1CM8


Cap

ture

d C

hann

el E

nerg

y

(c)

Figure 3.10: (a) Channel impulse response, (b) BER with respect to integrationinterval, (c) normalized signal energy capture in terms of integration interval, forCM1, CM2, CM7, and CM8, respectively. Figure adapted from Paper V.

burst communications [97]. However, in the targeted location-aware UWB RFIDsystems, where variable SNR is inevitable due to the variable link distance andnoise background, fixed threshold scheme can hardly achieve good performance.In Paper VI [31], we propose a low complexity method for burst packet detectionwhich bypasses the necessity of predefined threshold. It is achieved by sensing thecharacteristic of the preamble signal, which is designed with a specific pattern, in-stead of the received signal strength. In this way, the detection is not sensitive tothe signal/noise strength but to the signal existence and thus is adaptive to thevariations of noise background and link distance.


DETECTION

10 11 12 13 14 15 16 1710

−4

10−3

10−2

10−1

100

Eb/No (dB)

BE

R

CM7 Non−adaptive SchemeCM7 Proposed SchemeCM8 Non−adaptive SchemeCM8 Proposed Scheme

Figure 3.11: BER performance for the proposed synchronization scheme. Figureadapted from Paper V.

M-bit '1' '0' sequence N-bit PN sequence

Figure 3.12: Packet format. Figure adapted from Paper VI.

3.6.1 Proposed MethodThe packet format is designed as shown in Figure 3.12. The preamble and startframe delimiter (SFD) are known to the receiver and used for timing acquisition andpacket detection. The header specifies the length of the payload which carries theuseful information bits. Figure 3.13 illustrates the main procedure of the proposedpacket detection scheme.

Step 1. Preamble Detection

The purpose of the preamble detection is to detect whether desired signal or purenoise is received and at the same time roughly estimate the symbol boundary. Thepacket we considered here is using OOK modulation. The preamble is a sequenceof ‘1’ and ‘0’ alternating bits, thus, the pulse is on in one frame but off in the

3.6. LOW COMPLEXITY BURST PACKET DETECTION 45

Figure 3.13: Flowchart of the proposed packet detection scheme. Figure adaptedfrom Paper VI.

next frame. When the preamble is present, and the receiver starts the energy in-tegration of each symbol at a proper instant, it can be observed that, of every twosuccessive integration outputs, the former one is always larger than the latter one.On the other hand, if the incoming signal is pure noise, this result does not holdsince the noise is a stochastic process and the integration outputs are statisticallyindependent. Based on this preliminary idea, the algorithm of the preamble detec-tion is developed and described in Figure 3.14. Assume that the receiver starts tointegrate at a random time t0, and we set the integration interval to be the frameduration Tf . The parameter numComp denotes the sufficient number of successiveenergy pairs to distinguish the expected preamble signal from pure noise. Betterdetection performance can be achieved by increasing the value of numComp buton the trade-off of a larger communication overhead. The integration windows areshifted by 1/2Tf for every increment of i, so as to avoid the situation when the pulseenergy is equally distributed in two successive frames because of the randomnessof integration starting point t0. The sliding phase with higher resolution can beadopted in order to achieve more accurate synchronization.


DETECTION

E11 E12 E13 E14 E15 E1(2j-1) E1(2j)...

E21 E22 E23 E24 E25 E2(2j-1) E2(2j)...

E31 E32 E33 E34 E35 E3(2j-1) E3(2j)...

.

..

(E12) (E13) (E14) (E15) (E16)(E1(2j)) (E1(2j+1))

i=1, j=1

Ei(2j-1)>Ei(2j) ?

j=numComp ?

Yes

Preamble Detected

j=j+1

Yes

No

i=i+1, j=1No

'1' '0' '1'Tf

t0

Figure 3.14: Preamble detection algorithm. Figure adapted from Paper VI.

Step 2. OOK Threshold Estimation

After a successful preamble detection, coarse timing information of the symbolboundary is acquired. Aside from this, to demodulate the incoming OOK signal, ademodulation threshold is needed. Taking advantage of the ‘1’ and ‘0’ alternatingpattern of the preamble signal, the threshold for OOK demodulation can be easilyestimated by averaging an even number of successive integrated energies. To combatthe noise variance, we use 10 bits of the preamble signal which guarantees anacceptable estimation accuracy.

Step 3. SFD Detection

Based on the timing information and threshold obtained in previous steps, thesubsequent incoming signal is decoded and checked to see if it matches to theknown SFD sequence. If the SFD is detected, the packet presence is declared andthe header and payload reception begins.

3.6.2 Performance Evaluation

The performance of the proposed packet detection scheme is evaluated via Matlabsimulations in AWGN channel and field tests in real channel environment. The fieldtests are performed using the UWB software defined radio (SDR) testbed as demon-strated in Figure 3.15. Transmitted IR-UWB signals with 3-5 GHz bandwidth at40 MHz data rate corresponding to a frame duration Tf (also referred to as pulserepetition interval) of 25 ns are adopted for both simulations and field tests. We se-lect the detection error probability Pe as the main performance metric. It includes

3.7. SUMMARY 47

Figure 3.15: UWB SDR testbed. Figure adapted from Paper VI.

the false alarm probability Pfa defined as the probability of declaring packet pres-ence when actually no packet is transmitted, and the missed detection probabilityPmd which is the probability of not declaring packet existence when transmissiontruly happened. The performance of Pfa and Pmd is subjected to the length ofthe preamble and the SFD, as shown in Figure 3.16 and Figure 3.17, respectively.Figure 3.18 compares the performance of the proposed method with the conven-tional scheme based on fixed detection threshold. As can be seen, the conventionalscheme achieves excellent performance under certain SNR conditions, however, therobustness is very poor against the SNR variations. The proposed method, rely-ing on the preamble characteristic instead of the received signal strength, exhibitsquite good adaptivity to the SNR variations. As shown in Figure 3.18, above 15dB which is corresponding to the concerned range of SNR that ensures 10−3 BER,the proposed method guarantees a robust detection performance.

3.7 Summary

This chapter is focused on the sub-Nyquist sampling IR-UWB receiver designbased on the energy detection scheme. A flexible and energy-efficient ED receiverwas designed and implemented in ASIC. The measurement results of the ASICdemonstrated superior performance to other IR-UWB receivers in recent publishedworks [70, 71, 77, 78], in terms of energy efficiency, sensitivity and data rate. Inaddition, back-end signal processing to further optimize the performance of the EDreceiver was discussed and developed, i.e., optimum threshold estimation for OOKmodulated signal, adaptive synchronization and integration region optimization, ro-


DETECTION

6 8 10 12 14 1610

-4

10-3

10-2

10-1

numComp

Fals

e A

larm

Pro

babi

lity

SIM,11-bit SFDSIM, 13-bit SFDSIM 15-bit SFDFT,11-bit SFDFT, 13-bit SFDFT 15-bit SFD

Figure 3.16: Simulation and field test results of Pfa (numComp stands for thenumber of comparisons in the preamble detection process and it determines therequired length of the preamble). Figure adapted from Paper VI.

bust burst packet detection. Significant improvement on the receiver performancewas observed through simulations and field tests.

3.7. SUMMARY 49

6 8 10 12 14 1610

-3

10-2

10-1

100

numComp

Mis

sed

Det

ectio

n P

roba

bilit

y

SIM, Eb/No=15 dBSIM, Eb/No=17 dBSIM,Eb/No=20 dBFT, distance=1 mFT, distance=0.5 m

Figure 3.17: Simulation and field test results of Pmd (15-bit SFD is adopted). Figureadapted from Paper VI.

12 14 16 18 20 22 2410-4

10-3

10-2

10-1

100

Eb/No (dB)

Det

ectio

n E

rror P

roba

bilit

y

Proposed method, numComp=16, 15-bit SFDConventional method, optimized threshold @ 17 dBConventional method, optimized threshold @ 18 dBConventional method, optimized threshold @ 19 dB

Figure 3.18: Simulation results of Pe with proposed method and conventionalthreshold-based method. Figure adapted from Paper VI.

Chapter 4

Sub-Nyquist Sampling IR-UWB Receiver II:Compressed Sensing

The recent emerging compressed sensing theory provides a new sub-Nyquist sam-pling solution for the IR-UWB receiver design. IR-UWB signals can be acquiredat a sampling rate far below that required by Nyquist sampling theorem, mostattractively without losing any time domain resolution. This chapter first givesan overall introduction about the compressed sensing theory. Then, the issues ofapplying the CS theory in UWB systems are discussed. Following that, the noisedriven architectural analysis of the CS based IR-UWB receiver is provided. Finally,a novel noise-reducing architecture is proposed for the CS based IR-UWB receiver.

4.1 Overview of CS Theory

Traditionally in many applications, the signal acquisition is based on the princi-ple of Nyquist-Shannon sampling theorem [98] which suggests that the samplingrate must be at least twice the maximum frequency of the underlying signal. Inthe absence of a priori knowledge of the signal other than being band-limited, theNyquist sampling is an optimal solution. However, when further restrictions areimposed upon the signal, the Nyquist sampling may no longer be necessary. Inmany cases, the signal of interest is structured and has an “information” rate muchsmaller than suggested by its bandwidth [99]. That is, the signal has a conciserepresentation if expressed in an appropriate basis. For such kind of sparse or com-pressible signals, the recent emerging theory of compressed sensing (also known ascompressive sampling or CS) provides a new framework that holds a great promisefor full reconstruction at a sub-Nyquist sampling rate. In this section, we providean overview of the CS theory and its basic principles. For more theoretical andthorough introduction of compressed sensing, please refer to [86,87,99–101].

Figure 4.1 shows the procedure of compressed sensing. The CS theory is basedon the following key ingredients which will be explained hereafter: sparse represen-

51

52CHAPTER 4. SUB-NYQUIST SAMPLING IR-UWB RECEIVER II: COMPRESSED

SENSING

Sampling Reconstruction

CS Measurement

Find Sparse Solution

Signal Reconstruction

x y

x

1 1

M NM N

y x

y x

y

1 1ˆ

ˆˆN NN N

x

x

1min

y subject to

Figure 4.1: The procedure of compressed sensing. Figure adapted from Paper VII.

tation, incoherent measurements and signal reconstruction.

A. Sparse Representation

The first and foremost principle of sampling a signal using compressed sensing isthat the signal has a sparse representation on some basis. Mathematically speaking,a signal x ∈ RN is called K-sparse on the basis Ψ ∈ RN×N if x can be representedby certain linear combination of K vectors from Ψ with K � N , written as

x = Ψθ (4.1)

where θ = [θ1, θ2, · · · , θN ]T is the coefficient vector with only K nonzero entries.The signal is not necessarily to be strictly sparse but must be compressible, thatis, most of the entries of θ are insignificant so that they can be discarded withoutmuch perceptual loss [99].

Fortunately, many natural signals have a sparse representation when expressedin a proper basis. For example, a continuous wave signal is sparse in the frequencydomain while a UWB signal is sparse in the time domain.

B. Incoherent Measurements

In the CS framework, samples are taken not by direct sampling but by measuringa few linear projections of the signal of interest. The projection process can beformulated as

y = Φx = ΦΨθ = Θθ (4.2)

where Φ ∈ RM×N is the measurement matrix, also called as the projection matrixor sensing matrix, and y = [y1, y2, · · · , yM ]T is the generated measurements. Notethat the number of measurements M is far less than the dimension of the signal N .

Based on the M -dimensional measurements y, the original signal x can bereconstructed by seeking a sparse solution via nonlinear optimization. In order

4.2. CS FOR IR-UWB 53

to successfully reconstruct the signal with minimum number of measurements, themeasurement matrix Φ must be incoherent with the sparse basis Ψ. A quantitativedefinition of the coherence between Φ and Ψ is given in [99] as

µ(Φ,Ψ) =√N · max

1≤i,j≤N|〈φi, ψj〉| (4.3)

where the value of µ(Φ,Ψ) can range from 1 to√N . The smaller the coherence, the

fewer measurements are needed to reconstruct the signal [99]. Fortunately, randommatrices with independent and identically distributed (i.i.d.) entries exhibit a verylow coherence with any fixed basis [100]. When the random matrix is employed asthe measurement matrix, a K-sparse signal with length N can be reconstructed byM ≥ CKlog(N/K)� N measurements as suggested in [86]. The C here is a smallconstant.

C. Signal Reconstruction

A practical and commonly used approach to reconstruct the sparse signal is to solvethe l1-norm minimization problem via convex optimization [99], formulated as

θ = arg min ‖θ‖1 subject to y = Θθ (4.4)

where ‖θ‖1 =∑i |θi| is an efficient measure of the signal sparsity. This reconstruc-

tion approach is also known as the basis pursuit (BP) algorithm [102], which can beimplemented by linear programming with computational complexity ofO(N3) [103].

When the signal is corrupted by noise, to achieve robust signal reconstruction,we can relax the constraint of l1-norm minimization as

θ = arg min ‖θ‖1 subject to ‖y−Θθ‖2 ≤ ε (4.5)

where ε depends on the noise level. This algorithm is also called the basis pursuitde-noising (BPDN) [102].

Besides the convex optimization, other alternatives widely used to solve thesparse reconstruction problem are the greedy algorithms, such as orthogonal match-ing pursuit (OMP) [104]. The OMP algorithm has a computational complexity ofO(KMN) [103].

4.2 CS for IR-UWB

In this section, we start to explore the application of compressed sensing theoryin the IR-UWB receiver design. The main design issues and considerations arediscussed as follows.

Sparse Basis

IR-UWB signals are naturally sparse in the time domain such that no furthersparse transform is required and simply Ψ = I can be adopted. The authors of [79]


SENSING

suggested a more appealing sparse basis to further increase the sparsity of the IR-UWB signal. The proposed basis is composed by atoms which are shifted versionsof the pulse waveform, written as

dj(t) = p(t− (j − 1)δ), j = 1, 2, · · · (4.6)

where δ denotes the minimum shifting step. Ideally, the pulse distortion that maybe occurred during the transmission should be considered in defining the pulseshape dj(t). Let dj stands for the discrete-time representation of dj(t), the sparsebasis is then defined as

Ψ = {d1, d2, d3 · · · } (4.7)

The above basis has been widely adopted in recent works, such as [81, 88, 105,106]. In many literatures, the sparse basis is also called the dictionary. The dictio-nary facilitated signal reconstruction not only reduces the CS measurements dueto the increased signal sparsity, but also helps to filter out the noise in the recon-structed signal. The performance comparison between the signal reconstructionwith and without dictionary can be found in Paper VII.

Front-end CS Measurement

As mentioned in Section 4.1, whatever the sparse basis is, random sensing matricescan satisfy the incoherence requirement, e.g., a Gaussian matrix with each entrybeing i.i.d. taken from N(0, 1). However, the Gaussian matrix is not hardware-implementation-friendly. To facilitate efficient hardware implementation, the ran-dom Bernoulli matrix with ±1 entries can be adopted. The pseudo-random binarysequence (PRBS) can be easily generated by a linear feedback shift register (LFSR).Aside from the hardware consideration, the Bernoulli matrix is also beneficial forthe signal reconstruction under the noisy situation. This is due to the matrix en-try being either +1 or -1, the row vectors have the equal norm, leading to equallyweighted measurements.

The random modulation pre-integration (RMPI) architecture shown in Figure4.2 can be used as the front-end CS measurement of the IR-UWB receiver [99].The architecture performs a random modulation followed by an integrate-and-dumpoperation, hence the name RMPI. The random modulation is implemented by timedomain correlation of the incoming signal x(t) with analogue rectangular waveformsof the pseudo-random +1/-1 sequences. The correlation output is integrated overa fixed time interval and then digitized. Regarding IR-UWB signals, the symbolduration Ts is commonly adopted as the fixed integration interval. For every symbolduration, one measurement value ym is obtained after the correlate-integrate-and-dump.

Multiple CS measurements are required to successfully reconstruct the originalsignal. Let’s denote the number of measurements by M . From implementationpoint of view, the M CS measurements can be processed either in parallel or serial.

4.2. CS FOR IR-UWB 55

st T

t

( 1,2, )my m M

+1-1

( )x t

Sample Time m sT

Figure 4.2: The RMPI architecture.

To enable parallel processing, M branches of RMPI must be implemented in thereceiver front-end. On the other hand, serial processing avoids the increased costin the hardware implementation, but at the expense of M times repetition of pulsetransmission. Whether to adopt the parallel processing or serial processing shouldbe considered along with the application scenario. For instance, regarding theapplication scenario of RFID and wireless sensor networks, the parallel processingis preferable for the IR-UWB receiver. This is because the sensor nodes are usuallyenergy and memory constraint and thus only short packet transmission is supported.The reader, on the other hand, could be a powerful device.

Back-end Signal Reconstruction

In the CS based IR-UWB receiver, after the front-end CS measurement, the sam-pled measurement values are sent to the digital signal processing (DSP) back-endfor signal reconstruction and then demodulation/TOA estimation. Various recon-struction algorithms have been proposed in recent works, e.g., BP [102], OMP [104],StOMP [107], CoSaMP [108], and HHS [109]. In this work, we focus on the twomost widely used algorithms, the linear programming technique BP (BPDN for thenoisy signal) and the greedy algorithm OMP. For each of the two algorithms, byexploring its characteristic, we propose a practical strategy to fit into the recon-struction of noisy and multi-path IR-UWB signals.

In many cases, the IR-UWB signals at the receiver side may be buried in heavynoise and also rich in multi-path. The CS sampled and reconstructed signals arevery much likely to contain lots of false atoms either using BPDN or OMP al-gorithm. Fortunately, it can be observed that, there is a high probability thatthe most significant atoms of the signal can be successfully reconstructed. The BP(BPDN) is a sort of swap-down reconstruction method. It starts with a “full model”and then iteratively improves the solution by pruning [102]. Hence, for the signalreconstruction with BPDN, we perform a filtering process afterwards, keeping theseveral most significant atoms while small atoms are filtered out. The number ofsignificant atoms can be predefined according to the UWB channel model. On theother hand, OMP is a sort of build-up reconstruction method. It starts from an“empty model”, and at each iteration the most significant new atom is added to themodel [102]. Thus, regarding the OMP reconstruction, we can just directly limitthe iterations of the algorithm.


SENSING

4.3 Noise Driven Architectural Analysis

As mentioned in Section 4.2, in CS sampling of IR-UWB signals, multiple CS mea-surements are performed either in serial or parallel for one signal reconstruction.In Paper VII as well as the other existing works, e.g., [80, 81, 110], the receivedUWB signal is considered to be invariant during the multiple CS measurements ofeach signal reconstruction. However, this may not always the case in reality. Thereceived UWB signal is composed by the ideal signal part and noise components.Although the channel impulse response and thus the ideal signal part can be as-sumed to be invariant for the short time period, the random noise process can beindependent and uncorrelated for different CS measurements. We also find thatwhether the noise is correlated or uncorrelated is depending on the noise source(sky noise or amplifier noise) and the receiver architecture (serial or parallel). Andthe noise situation has a significant influence on the accuracy of CS reconstruction.In Paper VIII [35], we present this noise driven architectural analysis of the CSbased IR-UWB receivers.

4.3.1 Correlated Noise versus Uncorrelated NoiseA numerical study is carried out to explore how the accuracy of CS signal recon-struction will be affected by the noise situation (correlated or uncorrelated) in theCS measurement process1. Both the bit error rate (BER) for a communication sys-tem and the time-of-arrival (TOA) estimation for a ranging system are evaluated.

In the BER simulation, 3-5 GHz IR-UWB signals with PPM modulation and100 ns pulse repetition interval are used. And we simply consider the AWGNchannel. The sampling rate of the Matlab simulation environment is set to be 20GHz, while the CS measurements are performed at 10 GHz which is the Nyquistsampling rate of the incoming signals. Different compression ratio K = 0.3, 0.5and 0.7 are considered in the simulation. Here the compression ratio K is definedas K = M/N , where M is the number of measurements and N is the length of theoriginal signal as stated in Section 4.1. Figure 4.3 shows the simulation results of theBER performance with respect to different noise situations. It can be observed thatthe BER performance of the uncorrelated noise situation significantly outperformsthe correlated noise situation [35].

In the TOA simulation, we still use 3-5 GHz IR-UWB signals but without anymodulation scheme. The mean absolute error (MAE) averaged over 1000 TOAestimates is selected as the performance metric. Since this work is focused onthe performance comparison between the correlated noise situation and uncorre-lated noise situation, in the TOA estimation scheme, we simply choose the mostsignificant reconstructed path as the leading arrival path. Regarding the chan-nel environment, 200 different channel realizations of the IEEE 802.15.4a channel

1In all the simulations, we adopt the waveform dictionary for the sparse basis, the randommatrix composed by either +1 or -1 entry as the measurement matrix, and the greedy algorithmOMP as the reconstruction algorithm.

4.3. NOISE DRIVEN ARCHITECTURAL ANALYSIS 57

model CM1 are employed in the simulation. The pulse repetition interval is set tobe 200 ns, and the predefined TOA is uniformly distributed in [0, 100 ns] so thatall the multi-path components can fall within the observation interval [0, 200 ns).Figure 4.4 depicts the simulation results of the TOA estimation. As expected, theuncorrelated noise situation outperforms the correlated noise situation.

The result derived from the above numerical study is actually intuitive. Thereceived signal can be regarded as the sum of the ideal signal and the noise compo-nents. When CS sampling operates on the sum, the signal part keeps invariant forthe whole measurement process and can be reconstructed as expected. Ideally, wedo not want the noise components to be reconstructed at all. However, when thenoise components are invariant or correlated between every two CS measurements,the noise components will also be reconstructed. This also can be understood as thesparsity of the received signal is decreased. On the other side, in the uncorrelatednoise situation, the reconstructed noise components are largely reduced. In otherwords, there is an effective denoising for the uncorrelated noise situation and theaccuracy of the CS reconstruction is improved correspondingly.

4.3.2 Architectural Analysis

The numerical study above suggests that the uncorrelated noise situation in the CSmeasurement process is preferred for better CS reconstruction accuracy. To clearup the cause of the two different noise situations, the noise sources and the receiverarchitectures are jointly investigated. In CS receivers, before the CS correlator,there are mainly two noise sources, sky noise and amplifier noise [111, 112]. Con-sidering the two noise sources, as well as the serial or parallel implementation ofthe CS measurement process, in Figure 4.5 we present several possible architecturesfor the CS based IR-UWB receiver. Figure 4.5a shows a parallel architecture withsingle antenna and single LNA. No matter sky noise dominates or amplifier noisedominates, this architecture leads to the correlated noise situation. In Figure 4.5b,a parallel architecture with single antenna and multiple LNAs is presented. Whensky noise dominates, we still get the correlated noise situation. On the other hand,when amplifier noise dominates, the uncorrelated noise situation is obtained dueto the space diversity. Figure 4.5c is a parallel architecture with multiple antennasand multiple LNAs. This architecture guarantees that the noise is uncorrelateddespite sky noise dominates or amplifier noise dominates. Figure 4.5d depicts theserial architecture. Due to the time diversity, no matter sky noise dominates oramplifier noise dominates, we get the uncorrelated situation. This analysis of noisesituation in the CS measurement process subject to the noise source and receiverarchitecture is briefly listed in Table 4.1. It can be served as a design guideline forthe CS based IR-UWB receiver regarding different application scenarios.


SENSING

Table 4.1: Noise Situation in CS Measurement Process Subject to Receiver Archi-tecture and Noise Source.

Sky Noise Dominates Amplifier Noise Dominates(a) correlated noise correlated noise(b) correlated noise uncorrelated noise (space diversity)(c) uncorrelated noise (space diversity) uncorrelated noise (space diversity)(d) uncorrelated noise (time diversity) uncorrelated noise (time diversity)

2 4 6 8 10 12 14 16 18 2010

−5

10−4

10−3

10−2

10−1

100

Eb/No (dB)

BE

R

Correlated noise, K=0.3Correlated noise, K=0.5Correlated noise, K=0.7Uncorrelated noise, K=0.3Uncorrelated noise, K=0.5Uncorrelated noise, K=0.7

Figure 4.3: BER comparison between correlated noise situation and uncorrelatednoise situation (AWGN channel and Ts = 100 ns). Figure adapted from Paper VIII.

4.4 Noise-Reducing Architecture for CS Ranging Receiver

The high time domain resolution is the most attractive characteristic of IR-UWBsignals for ranging and positioning applications. Yet the acquisition of such hightime domain resolution is very challenging. The compressed sensing theory pro-vides a sub-Nyquist sampling solution without sacrificing the high time domainresolution. Nevertheless, due to the sparse reconstruction principle, conventionalCS receiver as introduced in [80–82] is natively sensitive to noise, causing significantperformance degradation in heavy noisy situations. In order to improve the per-formance of CS receiver, in Paper IX, a novel two-path noise-reducing architecturefor the receiver RF front-end is proposed. Besides the noise suppression effect, theproposed architecture also relaxes the hardware implementation of the CS randomprojection as well as the back-end signal reconstruction.

4.4. NOISE-REDUCING ARCHITECTURE FOR CS RANGING RECEIVER 59

10 12 14 16 18 20 22 24 26 28 300

10

20

30

40

50

60

70

Eb/No (dB)

MA

E (

ns)

Correlated noise, K=0.3Correlated noise, K=0.5Correlated noise, K=0.7Uncorrelated noise, K=0.3Uncorrelated noise, K=0.5Uncorrelated noise, K=0.7

Figure 4.4: MAE comparison between correlated noise situation and uncorrelatednoise situation (CM1 channel and Ts = 200 ns). Figure adapted from Paper VIII.

4.4.1 Two-Path Noise-Reducing ArchitectureThe proposed architecture for CS based IR-UWB ranging receiver is shown inFigure 4.6. By adding an identical input path (antenna and gain stage) togetherwith a mixer, the noise in the received signal before feeding into the CS samplingblock is suppressed comparing with the conventional CS receiver. Assume thetransmitted signal is a sequence of repetitive pulses in a UWB ranging system, andlet x1(t), x2(t) denote the two inputs, then they can be formulated as

x1(t) =+∞∑i=−∞

L1∑l=1

αl1p (t− iTs − τl1 − τTOA1) + n1 (t) = s1(t) + n1(t) (4.8)

x2(t) =+∞∑i=−∞

L2∑l=1

αl2p (t− iTs − τl2 − τTOA2) + n2 (t) = s2(t) + n2(t) (4.9)

Since the two antennas are closely spaced together relative to the transmitter andreflecting objects, it can be assumed that the incoming UWB signal s1(t) and s2(t)are identical and correlated, written as s1(t) = s2(t) = s(t). On the other hand,the random noise processes n1(t) and n2(t) produced in the two independent pathsare uncorrelated. Thus, the output of the mixer y(t) can be expressed as

y (t) = x1 (t) · x2 (t) = (s(t) + n1(t)) · (s(t) + n2(t)) (4.10)


SENSING

LNA Sub-Nyquist ADC

1 ( )t

1y

1 ( )t , 2 ( )t ( )M t

(a)

M

LNA Sub-Nyquist ADC

DSPSignal Reconstruction and Detection at every symbol interval

1 ( )t

1y

sT

(b)

M

LNA Sub-Nyquist ADC

1 ( )t

1y

(c)

M

LNA Sub-Nyquist ADC

DSPSignal Reconstruction and Detection at every M symbol interval

st T

t

(d)

sMT

DSPSignal Reconstruction and Detection at every symbol interval sT

DSPSignal Reconstruction and Detection at every symbol interval sT

st T

t

st T

t

st T

t

Figure 4.5: Possible architectures of the CS based IR-UWB receiver: (a) parallelarchitecture with single antenna and single LNA; (b) parallel architecture withsingle antenna and multiple LNAs; (c) parallel architecture with multiple antennasand multiple LNAs; and (d) serial architecture. Figure adapted from Paper VIII.

Passing through the mixing stage, the signal part (common part) is magnified whilethe uncorrelated noise part is effectively suppressed, leading to an increased signal-to-noise ratio (SNR) for the subsequent CS sampling procedure. This approachis particularly beneficial for impulse radio signals where the peak instantaneousSNR is far beyond 0 dB. This noise suppression effect is visualized in Figure 4.7using a sequence of 3-5 GHz UWB Gaussian pulses with randomly allocated pulseposition. The noise suppression results in a higher probability of successful CSreconstruction. As the example presented in Figure 4.7, with the same compressionradio K = 0.5, the proposed architecture successfully reconstructed all the pulses,whereas only half of them were recovered at the exact position with the conventionalarchitecture.

In addition to the noise suppression effect, the proposed architecture also de-creases the implementation complexity of the CS sampling and reconstruction block.The power spectral density (PSD) of the signal x(t) (before mixer) and signal y(t)

4.4. NOISE-REDUCING ARCHITECTURE FOR CS RANGING RECEIVER 61

1 ( )t , 2 ( )t ( )M t

sub-Nyquist ADC

DSPsignal reconstruction & TOA estimation

st T

t

LNA

CS sampling & reconstruction

LNA

( )y t1( )x t

2 ( )x t

Figure 4.6: The proposed noise-reducing architecture for CS based IR-UWB rangingreceiver. Figure adapted from Paper IX.

(after mixer) is illustrated in Figure 4.8. It can be seen that the major part of thesignal frequency spectrum is shifted to the baseband up to 2 GHz band after themixing stage. Therefore, in the proposed receiver, the CS random projection of theincoming signal can be performed by correlating with the random rectangular wave-forms at 4 GHz, while 10 GHz correlating signals are required in the conventionalreceiver. Furthermore, the number of virtual samples N and the projection timesM are decreased by a factor of 2.5 which consequently brings down the operationcomplexity of the back-end signal reconstruction.

4.4.2 Performance EvaluationMatlab simulations are performed to evaluate the performance of the proposed CSreceiver2. The simulation is based on 3-5 GHz IR-UWB signals and the meanabsolute error (MAE) averaged over 10000 TOA estimates is selected as the perfor-mance metric. Regarding the TOA estimation scheme, we simply choose the mostsignificant recovered path as the leading arrival path. In AWGN channel situation,we set the symbol duration Ts = 100 ns, and the predefined time-of-arrival τTOAis uniformly distributed in [0, 98 ns] which ensures that the UWB Gaussian pulseswith pulse duration around 1.5 ns can always fall into the observation interval [0,100ns). In the dense multi-path channel situation, 200 different channel realizationsof the IEEE 802.15.4a channel model CM1 are adopted, and Ts = 200 ns, τToAis uniformly distributed in [0, 100 ns] so that all the multi-path components canfall within the observation interval [0, 200 ns). The simulation results for AWGNchannel is shown in Figure 4.9. As illustrated in the zoomed-in part, with respectto the same compression ratio K = 0.5, the Eb/N0 needed to provide a ranging

2In all the simulations, we adopt the waveform dictionary for the sparse basis, the randommatrix composed by either +1 or -1 entry as the measurement matrix, and the greedy algorithmOMP as the reconstruction algorithm.


SENSING

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x 10−6

−1

0

1

time, s

original UWB pulses

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x 10−6

−2

0

2

time, s

UWB signal x, Eb/No = 10 dB

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x 10−6

−2

0

2

time, s

reconstructed signal x with compression ratio 0.5

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x 10−6

−2

0

2

time, s

cross-mixed UWB signal y

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

x 10−6

0

1

2

time, s

reconstructed signal y with compression ratio 0.5

Figure 4.7: Signaling comparison between conventional and proposed architecture.Figure adapted from Paper IX.

accuracy of 30 cm (1 ns MAE) is reduced by more than 4 dB with the proposedCS receiver instead of the conventional one. Similar performance enhancement canbe observed for the CM1 channel as shown in Figure 4.10.

4.5 Summary

This chapter addresses the major issues of the IR-UWB receiver design based on thecompressed sensing theory. By exploring the main concepts of the CS theory andthe characteristics of IR-UWB signals, it was found that the CS theory is well suitedfor the acquisition of IR-UWB signals, and the most appropriate sparse basis, mea-surement basis, front-end architecture, and signal reconstruction algorithm were

4.5. SUMMARY 63

−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

x 1010

0

2

4

6x 10

−11

frequency, Hz

power spectral density of signal x

PS

D, V

2 /Hz

−1 −0.8 −0.6 −0.4 −0.2 0 0.2 0.4 0.6 0.8 1

x 1010

0

1

2

3

4x 10

−12

frequency, Hz

power spectral density of signal y

PS

D, V

2 /Hz

Figure 4.8: Power spectral density before and after the mixing stage. Figure adaptedfrom Paper IX.

5 10 15 200

5

10

15

20

25

30

35

Eb/No, dB

MA

E, n

s

proposed noise-reducing CS, K=0.3proposed noise-reducing CS, K=0.5proposed noise-reducing CS, K=0.7conventional CS, K=0.3conventional CS, K=0.5conventional CS, K=0.7

7 8 9 10 11 12 13 14 15 16

1

2

3

> 4 dB

Figure 4.9: MAE of TOA estimation in AWGN channel. Figure adapted from PaperIX.

introduced. Regarding the performance of the CS receiver in heavy noisy situation,a noise driven architectural analysis was provided. Furthermore, a noise-reducingarchitecture was proposed for the CS based IR-UWB receiver. Simulation resultsexhibited significant improvement of the ranging performance over the conventional


SENSING

10 15 20 250

10

20

30

40

50

60

70

Eb/No, dB

MA

E, n

s

proposed noise−reducing CS, K=0.3proposed noise−reducing CS, K=0.5proposed noise−reducing CS, K=0.7conventional CS, K=0.3conventional CS, K=0.5conventional CS, K=0.7

Figure 4.10: MAE of TOA estimation in CM1 channel. Figure adapted from PaperIX.

CS receivers.

Chapter 5

Conclusions

5.1 Thesis Summary

The vision of Internet-of-Things calls for short range wireless technologies featuredby low-power and low-cost, robust communication (∼ Mb/s data rate and tensof meters range) and high-precision positioning (sub-meter accuracy). Traditionalshort range wireless technologies, such as UHF RFID, ZigBee, Bluetooth and Wi-Fi, either suffering from limited operation range and data rate or lacking of powerefficiency and ranging accuracy, cannot satisfy the above requirements. IR-UWBtechnology relying on the sub-nanosecond pulse transmission occupying several GHzfrequency band with extremely low power spectral density emerges as a promisingcandidate. Yet several challenges must be confronted in order to take the fullbenefits of IR-UWB technology, the first of which is the receiver design.

In this dissertation, IR-UWB receiver design that fits to the IoT applications wasproposed and investigated. It began with a comprehensive investigation of the UWBtechnology in Chapter 2. FCC regulations, types of IR-UWB signal (carrier-basedor carrier-less), modulation schemes (BPSK, OOK, PPM), channel characteristic(sparse or dense multi-path) were introduced and their impact and requirements onthe receiver design were discussed. Various receiver architectures were reviewed andtheir trade-offs in terms of operation flexibility, performance, hardware complexityand power consumption were explored. Based on those investigations, we narrowedour focuses to the two most promising sub-Nyquist sampling candidates for IR-UWB systems regarding IoT applications: the energy detection receiver and thecompressed sensing receiver.

The design issues of the sub-Nyquist sampling IR-UWB receiver based on EDprinciple were addressed in Chapter 3. A low-power ED receiver featured by flex-ibility and multi-mode operation to better fit into various IoT applications wasimplemented and fabricated in 90 nm CMOS. 16.3 mW power consumption and -79 dBm sensitivity at 10 Mb/s data rate corresponding to over 10 meters operationrange under the FCC regulation was achieved. As the key enabler for the receiver

65

66 CHAPTER 5. CONCLUSIONS

flexibility and multi-mode operation, the timing circuit was implemented using amixed-signal design flow to cope with the high frequency reference clock as well asthe high programmability. In addition, back-end signal processing to further op-timize the receiver performance was explored and developed. A practical and lowcomplexity threshold optimization scheme based on look-up table was suggested forOOK modulated signals. And adaptive synchronization and integration region op-timization based on the study of signal energy capture was proposed. Furthermore,robust burst packet detection based on sensing the preamble signal characteris-tic instead of the received signal strength thus adaptive to the variations of noisebackground and link distance was developed. The simulation or field test resultsdemonstrated the effectiveness of the above proposed signal processing algorithms.

The ED receiver is attractive mainly due to its low power and low complex-ity implementation, but it cannot exploit the high time domain resolution of theUWB pulse exhaustively. CS based IR-UWB receiver, on the other hand, allowssub-Nyquist sampling without sacrificing any time domain resolution. Chapter 4addressed the design issues of the CS receiver. The CS theory was comprehensivelystudied and appropriate sparse basis, sensing matrix and reconstruction algorithmswere suggested for the IR-UWB signals. And then, a noise driven architectural anal-ysis of the CS receiver was provided. We found that the noise situation (correlatedor uncorrelated) during the multiple CS measurements of each signal reconstruc-tion has a significant influence on the accuracy of CS reconstruction, and the noisesituation is depending on the noise source (sky noise or amplifier noise) and the re-ceiver architecture (serial or parallel CS measurement procedure). Furthermore, anovel two-path noise-reducing architecture was proposed for the CS receiver. Sim-ulation results exhibited significant improvement on the ranging performance overthe conventional CS receivers. Besides the performance improvement, the proposedarchitecture can also relax the hardware implementation of the CS measurementprocedure as well as the back-end signal reconstruction.

5.2 Future Work

Some future research suggestions are given as follows:

• Optimization of CS reconstruction algorithmThere exists a number of CS reconstruction algorithms to recover the sparse orcompressible signals (not limit to IR-UWB signals), for instance, BP, BPDN,OMP that are introduced in this dissertation as well as those emerged in re-cent publications (e.g., [113–116]). It could be interesting and well worthy offurther investigation on these reconstruction algorithms, in terms of effective-ness in recovering IR-UWB signals and hardware implementation complexity.Furthermore, optimization on the existing algorithms to better fit to the IR-UWB application is preferred.

• Hardware implementation of the CS-based IR-UWB receiver

5.2. FUTURE WORK 67

So far the research opportunities for the design of CS-based IR-UWB receiverare still wide open, especially in the hardware implementation perspective.Current research efforts are mainly focused on the system definition and sig-nal processing algorithms. There are already few works have discussed ap-propriate hardware architectures for the CS receiver, such as [80, 81] as wellas Section 4.3 and Section 4.4 in this dissertation. However, the real hard-ware implementation in ASIC is remain untouched. This is an important yetchallenging task for future research.

• IR-UWB radio for wearable and implantable biomedical devicesIn addition to WPAN, the wireless body area network (WBAN) is anotherimportant branch of future WSN towards the IoT. On-body/in-body wirelesscommunications are demanded to connect the wearable/implantable micro-devices to the Internet to achieve better personal health monitoring and care.IEEE 802.15.6 standard has defined the UWB as one of the physical lay-ers for the WBAN. Such kind of application scenario poses more stringentrequirement on the IR-UWB transceiver design, e.g., miniaturization in ge-ometry, high power efficiency. Besides, the channel conditions are much morecomplicated than common indoor channels, severe frequency-dependent at-tenuation may occur due to human skin, blood and tissues [117, 118]. Thisbrings new research topics worth further investigation on the IR-UWB radio,from channel modeling, low power circuit design to system integration.

Bibliography

[1] ITU Internet Reports 2005, “The Internet of Things,” November 2005.

[2] M. A. Feki, F. Kawsar, M. Boussard, and L. Trappeniers, “The Internet ofthings: the next technological revolution,” Computer, vol. 46, no. 2, pp. 24–25, Feb 2013.

[3] A. Al-Fuqaha, M. Guizani, M. Mohammadi, M. Aledhari, and M. Ayyash,“Internet of things: a survey on enabling technologies, protocols, and ap-plications,” Communications Surveys Tutorials, IEEE, vol. 17, no. 4, pp.2347–2376, 2015.

[4] F. Mattern and C. Floerkemeier, “From the Internet of computers to theInternet of things,” From Active Data Management to Event-Based Systemsand More, Springer, pp. 242–259, 2010.

[5] L.-R. Zheng, M. Nejad, Z. Zou, D. Mendoza, Z. Zhang, and H. Tenhunen,“Future RFID and wireless sensors for ubiquitous intelligence,” in Proceedingsof IEEE NORCHIP Conference, pp. 142–149, November 2008.

[6] [Online]. Available: www.dartmouth.edu/livinglearning/communities/riots.html

[7] R. Want, “An introduction to RFID technology,” IEEE Journal of PervasiveComputing, pp. 25–33, January 2006.

[8] K. Sohraby, D. Minoli, and T. Znati. Wireless sensor networks: technology,protocols, and applications. Wiley, May 2007.

[9] R. J. M. Vullers, R. V. Schaijk, H. J. Visser, J. Penders and C. V. Hoof,“Energy harvesting for autonomous wireless sensor networks,” IEEE Solid-State Circuit Magazine, vol. 2, no. 2, pp. 29–38, Spring 2010.

[10] ZigBee Alliance. [Online]. Available: www.zigbee.org

[11] Bluetooth. [Online]. Available: www.bluetooth.com

[12] Wi-Fi Alliance. [Online]. Available: http://www.wi-fi.org

69

70 BIBLIOGRAPHY

[13] J.-S. Lee, Y.-W. Su, and C.-C. Shen, “A comparative study of wireless pro-tocols: Bluetooth, UWB, ZigBee, and Wi-Fi,” in Proceedings of the 33rd An-nual Conference of the IEEE Industrial Electronics Society (IECON 2007),pp. 46–51, November 2007.

[14] A. Darif, R. Saadane, and D. Aboutajdine, “An efficient short range wire-less communication technology for wireless sensor network,” in Proceedingsof 2014 Third IEEE International Colloquium in Information Science andTechnology (CIST), pp. 396–401, October 2014.

[15] C. Park and T. Rappaport, “Short-range wireless communications for next-generation networks: UWB, 60 GHz millimeter-wave WPAN, and ZigBee,”IEEE Journal of Wireless Communications, vol. 14, no. 4, pp. 70–78, August2007.

[16] D. Porcino and W. Hirt, “Ultra-wideband radio technology: potential andchallenges ahead,” IEEE Communications Magazine, vol. 41, no. 7, pp. 66–74, July 2003.

[17] P. Nikitin and K. Rao, “Performance limitations of passive UHF RFID sys-tems,” in Proceedings of IEEE Antennas and Propagation Society Interna-tional Symposium, pp. 1011–1014, July 2006.

[18] U. Karthaus and M. Fischer, “Fully integrated passive UHF RFID transpon-der IC with 16.7-µWminimum RF input power,” IEEE Journal of Solid-StateCircuits, vol. 38, no. 10, pp. 1602–1608, October 2003.

[19] R. Barnett, G. Balachandran, S. Lazar, B. Kramer, G. Konnail, S. Ra-jasekhar, and V. Drobny, “A passive UHF RFID transponder for EPC Gen2 with -14dBm sensitivity in 0.13 µm CMOS,” in Proceedings of IEEE Inter-national Solid-State Circuits Conference (ISSCC 2007), Digest of TechnicalPapers, pp. 582–623, February 2007.

[20] J.-S. Lee, “Performance evaluation of IEEE 802.15.4 for low-rate wireless per-sonal area networks,” IEEE Transactions on Consumer Electronics, vol. 52,no. 3, pp. 742–749, August 2006.

[21] D. Dardari, R. D’Errico, C. Roblin, A. Sibille, and M. Win, “Ultrawide band-width RFID: the next generation?” in Proceedings of the IEEE, vol. 98, no. 9,pp. 1570–1582, September 2010.

[22] FCC, “Revision of part 15 of the commission’s rules regarding ultra-widebandtransmission systems, first report and order,” February 2002.

[23] W. Yao and Y. Wang, “Direct antenna modulation - a promise for ultra-wideband (UWB) transmitting,” IEEE Microwave Symposium Digest, vol. 2,pp. 1273–1276, June 2004.

BIBLIOGRAPHY 71

[24] Z. Zou, D. Mendoza, P. Wang, Q. Zhou, J. Mao, F. Jonsson, H. Tenhunen, andL.-R. Zheng, “A low-power and flexible energy detection IR-UWB receiverfor RFID and wireless sensor networks,” IEEE Transactions on Circuits andSystems I: Regular Papers, vol. 58, no. 7, pp. 1470–1482, July 2011.

[25] M. Baghaei-Nejad, D. Mendoza, Z. Zou, S. Radiom, G. Gielen, L.-R. Zheng,and H. Tenhunen, “A remote-powered RFID tag with 10Mb/s UWB uplinkand -18.5dBm sensitivity UHF downlink in 0.18 µm CMOS,” in Proceedingsof IEEE International Solid-State Circuits Conference - Digest of TechnicalPapers (ISSCC 2009), pp. 198–199,199a, February 2009.

[26] Z. Zou. Impulse radio UWB for the Internet-of-things: a study on UHF/UWBhybrid solution. PhD thesis, Royal Institute of Technology, December 2011.

[27] Q. Zhou, J. Mao, Z. Zou, F. Jonsson, and L.-R. Zheng, “A mixed-signaltiming circuit in 90nm CMOS for energy detection IR-UWB receivers,” inProceedings of 2010 IEEE International SOC Conference (SOCC), pp. 413–416, September 2010.

[28] J. Mao, D. Sarmiento M, Q. Zhou, J. Chen, P. Wang, Z. Zou, F. Jonsson, andL.-R. Zheng, “A 90nm CMOS UHF/UWB asymmetric transceiver for RFIDreaders,” in Proceedings of 2011 ESSCIRC, pp. 179–182, September 2011.

[29] Q. Zhou, Z. Zou, F. Jonsson, and L.-R. Zheng, “A flexible back-end withoptimum threshold estimation for OOK based energy detection IR-UWB re-ceivers,” in Proceedings of 2011 IEEE International Conference on Ultra-Wideband (ICUWB), pp. 130–134, September 2011.

[30] Q. Zhou, Z. Zou, H. Tenhunen, and L.-R. Zheng, “Adaptive synchronizationand integration region optimization for energy detection IR-UWB receivers,”in Proceedings of 2012 IEEE International Conference on Ultra-Wideband(ICUWB), pp. 62–66, September 2012.

[31] Q. Zhou, Z. Zou, Q. Chen, H. Tenhunen, and L.-R. Zheng, “Low complexityburst packet detection for wireless-powered UWB RFID systems,” in Proceed-ings of 2015 IEEE International Conference on Ubiquitous Wireless Broad-band (ICUWB), pp. 1–5, October 2015.

[32] Z. Zou, T. Deng, Q. Zhou, M. Sarmiento, F. Jonsson, and L.-R. Zheng, “En-ergy detection receiver with ToA estimation enabling positioning in passiveUWB-RFID system,” in Proceedings of 2010 IEEE International Conferenceon Ultra-Wideband (ICUWB), vol. 2, pp. 1–4, September 2010.

[33] M. Sarmiento, Z. Zou, Q. Zhou, J. Mao, P. Wang, F. Jonsson, and L.-R.Zheng, “Analog front-end RX design for UWB impulse radio in 90nm CMOS,”in Proceedings of 2011 IEEE International Symposium on Circuits and Sys-tems (ISCAS), pp. 1552–1555, May 2011.

72 BIBLIOGRAPHY

[34] Q. Zhou, Z. Zou, H. Tenhunen, and L.-R. Zheng, “Exploration and per-formance evaluation of a compressed sensing based IR-UWB receiver,”in Proceedings of 2013 IEEE International Conference on Ultra-Wideband(ICUWB), pp. 226–230, September 2013.

[35] Q. Zhou, Z. Zou, H. Tenhunen, and L.-R. Zheng, “Architectural analysis ofcompressed sensing based IR-UWB receiver for communication and ranging,”in Proceedings of International Conference on Ultra-WideBand (ICUWB), pp.222–227, September 2014.

[36] “IEEE standard for local and metropolitan area networks part 20: air in-terface for mobile broadband wireless access systems supporting vehicularmobility - physical and media access control layer specification,” IEEE Std802.20-2008, pp. 1–1039, Aug 2008.

[37] K. Kalliola, “Spectrum sharing and flexible spectrum use,” in FUTURAWorkshop, August 2004.

[38] A. Batra, J. Balakrishnan, and A. Dabak, “Multi-band OFDM: a new ap-proach for UWB,” in Proceedings of the 2004 International Symposium onCircuits and Systems (ISCAS), vol. 5, pp. V–365–V–368, May 2004.

[39] M. Win and R. Scholtz, “Impulse radio: how it works,” IEEE Communica-tions Letters, vol. 2, no. 2, pp. 36–38, February 1998.

[40] J. A. Lopez-Salcedo. Coherent and non-coherent ultra-wideband communica-tions. PhD thesis, Universitat Politecnica de Catalunya, March 2007.

[41] J.-Y. Lee and S. Choi, “Through-material propagation characteristic and timeresolution of UWB signal,” in Proceedings of 2004 International Workshop onUltra Wideband Systems. Joint with Conference on Ultrawideband Systemsand Technologies (Joint UWBST IWUWBS), pp. 71–75, May 2004.

[42] “IEEE 802 part 15.3: wireless medium access control (MAC) and physicallayer (PHY) specifications for higher rate wireless personal area networks(WPAN),” 2003.

[43] “IEEE standard for information technology - telecommunications and infor-mation exchange between systems - local and metropolitan area networks- specific requirement part 15.4: wireless medium access control (MAC)and physical layer (PHY) specifications for low-rate wireless personal areanetworks (WPANs),” IEEE Std 802.15.4a-2007 (Amendment to IEEE Std802.15.4-2006), pp. 1–203, 2007.

[44] Decawave. [Online]. Available: http://www.decawave.com/

[45] Time Domain. [Online]. Available: http://www.timedomain.com/

BIBLIOGRAPHY 73

[46] Ubisense. [Online]. Available: http://ubisense.net/en

[47] Zebra Technologies Corporation. [Online]. Available:https://www.zebra.com/gb/en.html

[48] Bespoon. [Online]. Available: http://bespoon.com/

[49] V. Venkatasubramanian, H. Leung, and X. Liu, “Chaos UWB radar forthrough-the-wall imaging,” IEEE Transactions on Image Processing, vol. 18,no. 6, pp. 1255–1265, June 2009.

[50] R. Chavez-Santiago, A. Khaleghi, I. Balasingham, and T. Ramstad, “Archi-tecture of an ultra wideband wireless body area network for medical appli-cations,” in Proceedings of the 2nd International Symposium on Applied Sci-ences in Biomedical and Communication Technologies (ISABEL 2009), pp.1–6, November 2009.

[51] A. Reisenzahn, T. Buchegger, D. Scherrer, S. Matzinger, S. Hantscher, andC. Diskus, “A ground penetrating UWB radar system,” in Proceedings ofthe Third International Conference on Ultrawideband and Ultrashort ImpulseSignals, pp. 116–118, September 2006.

[52] L. Sakkila, C. Tatkeu, F. Boukour, Y. El Hillali, A. Rivenq, and J.-M. Rou-vean, “UWB radar system for road anti-collision application,” in Proceed-ings of the 3rd International Conference on Information and CommunicationTechnologies: From Theory to Applications (ICTTA 2008), pp. 1–6, April2008.

[53] Prithviraj, “UWB application for precision automobile parking system,” inProceedings of International Conference on Recent Advances in MicrowaveTheory and Applications (MICROWAVE 2008), pp. 326–326, November 2008.

[54] M. Z. Win, D. Dardari, A. F. Molisch, W. Wiesbeck, and J. Zhang, “Historyand applications of UWB,” in Proceedings of the IEEE, vol. 97, no. 2, pp.198–204, February 2009.

[55] H. Nikookar and R. Prasad. Introduction to ultra wideband for wireless com-municaitons. Springer, 2009.

[56] S. Iida, K. Tanaka, H. Suzuki, N. Yoshikawa, N. Shoji, B. Griffiths, D. Mellor,F. Hayden, I. Butler, and J. Chatwin, “A 3.1 to 5 GHz CMOS DSSS UWBtransceiver for WPANs,” in Proceedings of 2005 IEEE International Solid-State Circuits Conference (ISSCC). Digest of Technical Papers, vol. 1, pp.214–594, February 2005.

[57] A. Azakkour, M. Regis, F. Pourchet, and G. Alquie, “A new integrated mono-cycle generator and transmitter for ultra-wideband (UWB) communications,”in Proceedings of 2005 IEEE Radio Frequency Integrated Circuits (RFIC)Symposium. Digest of Papers, pp. 79–82, June 2005.

74 BIBLIOGRAPHY

[58] Y. Zheng, Y. Tong, C. W. Ang, Y.-P. Xu, W. G. Yeoh, F. Lin, and R. Singh,“A CMOS carrier-less UWB transceiver for WPAN applications,” in Proceed-ings of IEEE International Solid-State Circuits Conference (ISSCC 2006).Digest of Technical Papers, pp. 378–387, February 2006.

[59] L. Smaini, C. Tinella, D. Helal, C. Stoecklin, L. Chabert, C. Devaucelle,R. Cattenoz, N. Rinaldi, and D. Belot, “Single-chip CMOS pulse generatorfor UWB systems,” IEEE Journal of Solid-State Circuits, vol. 41, no. 7, pp.1551–1561, July 2006.

[60] D. D. Wentzloff and A. P. Chandrakasan, “Gaussian pulse generators forsubbanded ultra-wideband transmitters,” IEEE Transactions on MicrowaveTheory and Techniques, vol. 54, no. 4, pp. 1647–1655, June 2006.

[61] J. Ryckaert, M. Badaroglu, C. Desset, V. D. Heyn, G. ven der Plas,P. Wambacq, B. van Poucke, and S. Donnay, “Carrier-based UWB impulse ra-dio: simplicity, flexibility, and pulser implementation in 0.18-micron CMOS,”in Proceedings of 2005 IEEE International Conference on Ultra-Wideband,pp. 432–437, September 2005.

[62] S. Bourdel, Y. Bachelet, J. Gaubert, R. Vauche, O. Fourquin, N. Dehaese, andH. Barthelemy, “A 9-pJ/pulse 1.42-Vpp OOK CMOS UWB pulse generatorfor the 3.1-10.6-GHz FCC band,” IEEE Transactions on Microwave Theoryand Techniques, vol. 58, no. 1, pp. 65–73, January 2010.

[63] M. Benedetto and G. Giancola. Understanding ultra wide band radio funda-mentals. Prentice Hall communications engineering and emerging technologiesseries, 2004.

[64] J. Foerster, “Channel modeling sub-committee report final,” IEEE DocumentIEEE P802.15-02/490r1-SG3a, 2003.

[65] A. F. Molisch, K. Balakrishnan, D. Cassioli, C.-C. Chong, S. Emami, A. Fort,J. Karedal, J. Kunisch, H. Schantz, U. Schuster, K. Siwiak, “IEEE 802.15.4achannel model - final report,” IEEE Document IEEE 802.15-04-0662-02-004a, 2005.

[66] I. O’Donnell and R. Brodersen, “An ultra-wideband transceiver architecturefor low power, low rate, wireless systems,” IEEE Transactions on VehicularTechnology, vol. 54, no. 5, pp. 1623–1631, September 2005.

[67] R. Blazquez, P. Newaskar, F. Lee, and A. Chandrakasan, “A baseband proces-sor for impulse ultra-wideband communications,” IEEE Journal of Solid-StateCircuits, vol. 40, no. 9, pp. 1821–1828, September 2005.

[68] C.-H. Yang, K.-H. Chen, and T.-D. Chiueh, “A 1.2V 6.7mW impulse-radioUWB baseband transceiver,” in Proceedings of 2005 IEEE International

BIBLIOGRAPHY 75

Solid-State Circuits Conference (ISSCC). Digest of Technical Papers, vol. 1,pp. 442–608, February 2005.

[69] T. Terada, S. Yoshizumi, M. Muqsith, Y. Sanada, and T. Kuroda, “A CMOSultra-wideband impulse radio transceiver for 1-Mb/s data communicationsand ±2.5-cm range finding,” IEEE Journal of Solid-State Circuits, vol. 41,no. 4, pp. 891–898, April 2006.

[70] N. Van Helleputte, M. Verhelst, W. Dehaene, and G. Gielen, “A reconfig-urable, 130 nm CMOS 108 pJ/pulse, fully integrated IR-UWB receiver forcommunication and precise ranging,” IEEE Journal of Solid-State Circuits,vol. 45, no. 1, pp. 69–83, January 2010.

[71] J. Ryckaert, M. Badaroglu, V. De Heyn, G. Van der Plas, P. Nuzzo,A. Baschirotto, S. D’Amico, C. Desset, H. Suys, M. Libois, B. Van Poucke,P. Wambacq, and B. Gyselinckx, “A 16mA UWB 3-to-5GHz 20Mpulses/squadrature analog correlation receiver in 0.18µm CMOS,” in Proceedings ofIEEE International Solid-State Circuits Conference (ISSCC 2006). Digest ofTechnical Papers, pp. 368–377, February 2006.

[72] M. Verhelst and W. Dehaene, “Analysis of the QAC IR-UWB receiver for lowenergy, low data-rate communication,” IEEE Transactions on Circuits andSystems I: Regular Papers, vol. 55, no. 8, pp. 2423–2432, September 2008.

[73] S. Bagga, L. Zhang, W. Serdijin, J. Long, and E. Busking, “A quantizedanalog delay for an IR-UWB quadrature downconversion autocorrelationreceiver,” in Proceedings of 2005 IEEE International Conference on Ultra-Wideband, pp. 328–332, September 2005.

[74] S. Franz and U. Mitra, “Generalized UWB transmitted reference systems,”IEEE Journal on Selected Areas in Communications, vol. 24, no. 4, pp. 780–786, April 2006.

[75] F. Lee and A. Chandrakasan, “A 2.5nJ/b 0.65V 3-to-5GHz subbanded UWBreceiver in 90nm CMOS,” in Proceedings of IEEE International Solid-StateCircuits Conference (ISSCC 2007). Digest of Technical Papers, pp. 116–590,February 2007.

[76] L. Stoica, A. Rabbachin, and I. Oppermann, “A low-complexity noncoherentIR-UWB transceiver architecture with TOA estimation,” IEEE Transactionson Microwave Theory and Techniques, vol. 54, no. 4, pp. 1637–1646, June2006.

[77] A. Gerosa, S. Solda, A. Bevilacqua, D. Vogrig, and A. Neviani, “An energy-detector for noncoherent impulse-radio UWB receivers,” IEEE Transactionson Circuits and Systems I: Regular Papers, vol. 56, no. 5, pp. 1030–1040, May2009.

76 BIBLIOGRAPHY

[78] D. Daly, P. Mercier, M. Bhardwaj, A. Stone, Z. Aldworth, T. Daniel, J. Vold-man, J. Hildebrand, and A. Chandrakasan, “A pulsed UWB receiver SoC forinsect motion control,” IEEE Journal of Solid-State Circuits, vol. 45, no. 1,pp. 153–166, January 2010.

[79] J. Paredes, G. Arce, and Z. Wang, “Compressed sensing for ultrawideband im-pulse radio,” in Proceedings of IEEE International Conference on Acoustics,Speech and Signal Processing (ICASSP 2007), vol. 3, pp. III–553–III–556,April 2007.

[80] O. Khan, S.-Y. Chen, D. Wentzloff, and W. Stark, “Impact of compressedsensing with quantization on UWB receivers with multipath channel esti-mation,” IEEE Journal on Emerging and Selected Topics in Circuits andSystems, vol. 2, no. 3, pp. 460–469, September 2012.

[81] T. Thiasiriphet, M. Ibrahim, and J. Lindner, “Compressed sensing for UWBmedical radar applications,” in Proceedings of 2012 IEEE International Con-ference on Ultra-Wideband (ICUWB), pp. 106–110, September 2012.

[82] A. Oka and L. Lampe, “A compressed sensing receiver for UWB impulseradio in bursty applications like wireless sensor networks,” Elsevier Phys.Commun., vol. 2, no. 4, pp. 248–264, 2009.

[83] H. Arslan, Z.-N. Chen, and M.-G. Di Benedetto. Ultra wideband wirelesscommunication. Wiley, October 2006.

[84] T. Q. S. Quek and M. Z. Win, “Analysis of UWB transmitted-reference com-munication systems in dense multipath channels,” IEEE Journal on SelectedAreas in Communications, vol. 23, no. 9, pp. 1863–1874, September 2005.

[85] K. Witrisal, G. Leus, G. J. M. Janssen, M. Pausini, F. Troesch, T. Zasowski,and J. Romme, “Noncoherent ultra-wideband systems,” IEEE Signal Pro-cessing Magazine, vol. 26, no. 4, pp. 48–66, July 2009.

[86] E. Candes, J. Romberg, and T. Tao, “Robust uncertainty principles: exactsignal reconstruction from highly incomplete frequency information,” IEEETransactions on Information Theory, vol. 52, no. 2, pp. 489–509, February2006.

[87] D. Donoho, “Compressed sensing,” IEEE Transactions on Information The-ory, vol. 52, no. 4, pp. 1289–1306, April 2006.

[88] G. Shi, J. Lin, X. Chen, F. Qi, D. Liu, and L. Zhang, “UWB echo signaldetection with ultra-low rate sampling based on compressed sensing,” IEEETransactions on Circuits and Systems II: Express Briefs, vol. 55, no. 4, pp.379–383, April 2008.

BIBLIOGRAPHY 77

[89] N. Van Helleputte and G. Gielen, “A 70 pJ/pulse analog front-end in 130nm CMOS for UWB impulse radio receivers,” IEEE Journal of Solid-StateCircuits, vol. 44, no. 7, pp. 1862–1871, July 2009.

[90] Z. Zou, Y. Ruan, L.-R. Zheng, and H. Tenhunen, “Impulse UWB energydetection receiver with energy offset synchronization scheme,” in Proceedingsof IEEE International Conference on Ultra-Wideband (ICUWB 2009), pp.540–544, September 2009.

[91] M. Sahin, I. Guvenc, and H. Arslan, “Joint parameter estimation for UWBenergy detectors using OOK,” Springer Journal of Wireless Personal Com-munications, vol. 40, no. 4, pp. 579–591, 2007.

[92] K. Furusawa, M. Sasaki, J. Hioki, and M. Itami, “Schemes of optimizationof energy detection receivers for UWB-IR communication systems under dif-ferent channel model,” in Proceedings of IEEE International Conference onUltra-Wideband (ICUWB 2008), vol. 1, pp. 157–160, September 2008.

[93] B. Miscopein and J. Schwoerer, “Low complexity synchronization algorithmfor non-coherent UWB-IR receivers,” in Proceedings of IEEE 65th VehicularTechnology Conference (VTC2007-Spring), pp. 2344–2348, April 2007.

[94] N. He and C. Tepedelenlioglu, “Adaptive synchronization for non-coherentUWB receivers,” in Proceedings of IEEE International Conference on Acous-tics, Speech, and Signal Processing (ICASSP), vol. 4, pp. iv–517–iv–520, May2004.

[95] M. Flury, R. Merz, and J. Y. L. Boudec, “Synchronization for impulse-radio UWB with energy-detection and multi-user interference: algorithmsand application to IEEE 802.15.4a,” IEEE Transactions on Signal Process-ing, vol. 59, no. 11, pp. 5458–5472, November 2011.

[96] S. R. Aedudodla, S. Vijayakumaran, and T. F. Wong, “Timing acquisitionin ultra-wideband communication systems,” IEEE Transactions on VehicularTechnology, vol. 54, no. 5, pp. 1570–1583, September 2005.

[97] C. Li, Y. Pei, and N. Ge, “Synchronization acquisition threshold based onpeak-to-average ratio of correlation energy for UWB communications,” inProceedings of 2011 International Conference on Wireless Communicationsand Signal Processing (WCSP), pp. 1–4, November 2011.

[98] C. E. Shannon, “Communication in the presence of noise,” Proceedings of theIRE, vol. 37, no. 1, pp. 10–21, January 1949.

[99] E. Candes and M. Wakin, “An introduction to compressive sampling,” IEEESignal Processing Magazine, vol. 25, no. 2, pp. 21–30, March 2008.

78 BIBLIOGRAPHY

[100] E. Candes and T. Tao, “Near-optimal signal recovery from random projec-tions: universal encoding strategies?” IEEE Transactions on InformationTheory, vol. 52, no. 12, pp. 5406–5425, December 2006.

[101] E. Candes and J. Romberg, “Sparsity and incoherence in compressive sam-pling,” Inverse Problems, vol. 23, no. 3, pp. 969–985, June 2007.

[102] S. S. Chen, D. L. Donoho, and M. A. Saunders, “Atomic decomposition bybasis pursuit,” SIAM Journal on Scientific Computing, vol. 20, no. 1, pp.33–61, August 1998.

[103] S. Qaisar, R. Bilal, W. Iqbal, M. Naureen, and S. Lee, “Compressive sens-ing: from theory to applications, a survey,” Journal of Communications andNetworks, vol. 15, no. 5, pp. 443–456, October 2013.

[104] J. Tropp and A. Gilbert, “Signal recovery from random measurements viaorthogonal matching pursuit,” IEEE Transactions on Information Theory,vol. 53, no. 12, pp. 4655–4666, December 2007.

[105] W. Weidong, Y. Jun-an, Y. Haibo, and W. Shehui, “Reconstruction methodfor pulse position modulation-ultra wideband communication signal based oncompressed sensing,” IET Communications, vol. 8, no. 5, pp. 707–713, March2014.

[106] T. Le, J. Kim, and Y. Shin, “An improved ToA estimation in compressedsensing-based UWB systems,” in Proceedings of 2010 IEEE InternationalConference on Communication Systems (ICCS), pp. 249–253, November2010.

[107] D. Donoho, Y. Tsaig, I. Drori, and J.-L. Starck, “Sparse solution of under-determined systems of linear equations by stagewise orthogonal matchingpursuit,” IEEE Transactions on Information Theory, vol. 58, no. 2, pp. 1094–1121, February 2012.

[108] D. Needell and J. A. Tropp, “CoSaMP: iterative signal recovery from incom-plete and inaccurate samples,” Applied and Computational Harmonic Analy-sis, vol. 26, no. 3, pp. 301–321, 2009.

[109] A. C. Gilbert, M. J. Strauss, J. A. Tropp, and R. Vershynin, “One sketchfor all: fast algorithms for compressed sensing,” in Proceedings of 39th ACMSymposium on Theory of Computing, pp. 237–246, 2007.

[110] J. Paredes, G. Arce, and Z. Wang, “Ultra-wideband compressed sensing:channel estimation,” IEEE Journal of Selected Topics in Signal Processing,vol. 1, no. 3, pp. 383–395, October 2007.

BIBLIOGRAPHY 79

[111] M. J. Gans, “Channel capacity between antenna arrays-part I: sky noise domi-nates,” IEEE Transactions on Communications, vol. 54, no. 9, pp. 1586–1592,September 2006.

[112] M. J. Gans, “Channel capacity between antenna arrays-part II: amplifiernoise dominates,” IEEE Transactions on Communications, vol. 54, no. 11,pp. 1983–1992, November 2006.

[113] S. Sparrer and R. F. H. Fischer, “MMSE-based version of OMP for recovery ofdiscrete-valued sparse signals,” Electronics Letters, vol. 52, no. 1, pp. 75–77,2016.

[114] H. Zayyani, M. Korki, and F. Marvasti, “Dictionary learning for blind onebit compressed sensing,” IEEE Signal Processing Letters, vol. 23, no. 2, pp.187–191, February 2016.

[115] H. Zhao, Y. Wang, X. Peng, and Z. Qiao, “Gradient-based compressive sens-ing for noise image and video reconstruction,” IET Communications, vol. 9,no. 7, pp. 940–946, 2015.

[116] D. Craven, B. McGinley, L. Kilmartin, M. Glavin, and E. Jones, “Adap-tive dictionary reconstruction for compressed sensing of ECG signals,” IEEEJournal of Biomedical and Health Informatics, vol. PP, no. 99, pp. 1–10, 2016.

[117] S. Nagaraj and F. Rassam, “Improved non-coherent UWB receiver for im-plantable biomedical devices,” IEEE Transactions on Biomedical Engineer-ing, vol. PP, no. 99, pp. 1–6, 2015.

[118] H. Liu, J. Sarrazin, T. Mavridis, L. Petrillo, Z. Liu, P. D. Doncker, andA. Benlarbi-delai, “Performance assessment of IR-UWB body area network(BAN) based on IEEE 802.15.6 standard,” IEEE Antennas and WirelessPropagation Letters, vol. PP, no. 99, pp. 1–4, 2016.

Sub-NyquistSamplingImpulseRadioUWBReceivers fortheInternet …1045523/... · 2016. 11. 9. · end...

Documents

Transcript of Sub-NyquistSamplingImpulseRadioUWBReceivers fortheInternet …1045523/... · 2016. 11. 9. · end...