Adaptive RAKE Receiver Structures for Ultra Wide-Band Systems

Transcription

1 Adaptive RAE Reeiver Strutures for Ultra Wide-Band Systems A Thesis Submitted to the College of Graduate Studies and Researh n Partial Fulfillment of the Requirements For the Degree of Master of Siene n the Department of Eletrial Engineering University of Sasathewan Sasatoon By Quan Wan Copyright Quan Wan, Deember, 005. All rights reserved

2 PERMSSON TO USE n presenting this thesis in partial fulfillment of the requirement for a Postgraduate degree from University of Sasathewan, agree that the Libraries of this University may mae it freely available for inspetion. further agree that permission for opying of this thesis in any manner, in whole or in part, for sholarly purposes may be granted by the professor or professors who supervised my thesis wor or, in their absene, by the Head of the Department or the Dean of the College in whih my thesis wor was done. t is understood that any opying for publiation or use of this thesis or parts thereof for finanial gain shall not by allowed without my written permission. t is also understood that due reognition shall be given to me and to the University of Sasathewan in any sholarly use whih may be mae of any material in my thesis. Request for permission to opy or to mae other use of material in this thesis in whole or part should be addressed to: Head of the Department of Eletrial Engineering University of Sasathewan Sasatoon, Sasathewan, S7N 5A9

3 Abstrat Ultra wide band UWB) is an emerging tehnology that reently has gained regulatory approval. t is a suitable solution for high speed indoor wireless ommuniations due to its promising ability to provide high data rate at low ost and low power onsumption. Another benefit of UWB is its ability to resolve individual multi-path omponents. This feature motivates the use of RAE multipath ombining tehniques to provide diversity and to apture as muh energy as possible from the reeived signal. Potential future and rule limitation of UWB, lead to two important harateristis of the tehnology: high bit rate and low emitting power. Based on the power emission limit of UWB, the only hoie for implementation is the low level modulation tehnology. To obtain suh a high bit rate using low level modulation tehniques, signifiant inter-symbol interferene S) is unavoidable. Three N N means the numbers of fingers) fingers RAE reeiver strutures are proposed: the N-seletive maximal ratio ombiner MRC), the N-seletive MRC reeiver with least-mean-square LMS) adaptive equalizer and the N-seletive MRC reeiver with LMS adaptive ombiner. These three reeiver strutures were all simulated for N8, 6 and 3. Simulation results indiate that S is effetively suppressed. The 6-seletive MRC RAE reeiver with LMS adaptive ombiner demonstrates a good balane between performane, omputation omplexity and required length of the training sequene.

4 Due to the simpliity of the algorithm and a reasonable sampling rate, this struture is feasible for pratial VLS implementations.

5 Anowledgements This thesis ould not ome through without the help of many wonderful people around me. First of all, would lie to express my deepest sense of gratitude to my supervisor, Dr. Anh Dinh, for his patient guidane, valuable advie and tremendous tehnial and moral support. Woring with him has been a true privilege. His diligene, insights and enthusiasm demonstrate great sholars and dediated teahers. His ideas and inspirations have greatly helped the development of my thesis. really appreiate the ountless times that he spent on disussions and feedba on my wor. would lie to anowledge TRLabs for supporting this researh in the past two years. would also lie to anowledge my olleagues at TRLabs for maing my graduate studies and researh wor enoyable and produtive. The reation of the outstanding researh environment at TRLabs in Sasatoon is due to their tireless efforts. Finally, would lie to express my sinere gratitude to my father who always enouraged me to fae new hallenges and never hesitated to provide me with ease and omfort towards my eduations. V

6 Dediation To my dear parents To my wife, Ying Cui, who has always been there for me and filled my life with love and happiness To my dear son, YingJia Wan V

7 Table of Contents Abstrat... Anowledgements...V Dediation...V Table of Contents...V List of Tables...X List of Figures...X List of Aronyms...X Chapter ntrodution.... Historial baground.... Motivation Problems and obetives Thesis overview... 8 Chapter UWB Communiation Tehnology Definition of UWB Advantages of UWB....3 Existing UWB system shemes Channel haraterization of UWB... 7 Chapter 3 Reeiver Signal Proessing in the Presene of Multi-path RAE reeiver Combating S Optimum maximum-lielihood reeiver Linear equalization V

8 3..3 Deision-feedba equalization LMS adaptive equalization Chapter 4 Proposed Reeiver Strutures Basi requirements and onsiderations of reeivers A simplified RAE demodulation implementation Channel estimation for RAE demodulation N-seletive MRC RAE reeiver N-seletive MRC RAE reeiver with LMS adaptive equalizer N-seletive RAE reeiver with LMS adaptive ombiner Chapter 5 Simulation Set-up and Results Simulation set-up Signal generator UWB hannel Mathed filter sampling and hannel estimation N-seletive MRC RAE reeiver and LMS adaptive equalizer N-seletive RAE reeiver with LMS adaptive ombiner Simulation results Analysis and disussion Chapter 6 Conlusions and Future Wor Conlusions Future wor Bibliography Appendix A: Disrete-time model for a hannel with S V

9 Appendix B: Zero-foring equalizer Appendix C: Mean-square-error MSE) riterion linear equalizer Appendix D: Proedure to alulate a gradient vetor... 0 V

10 List of Tables Table -: Unliensed bands..... Table -: The EEE UWB hannel harateristis....0 Table 5-: Parameters of different reeiver strutures.88 X

11 List of Figures Figure -: UWB emission limit for indoor systems Figure -: UWB emission limit for outdoor hand-held systems.. 0 Figure -3: Time divisions in THSS-UWB sheme... 5 Figure -4: One realization of UWB hannel model CM Figure -4: One realization of UWB hannel model CM.... Figure -6: One realization of UWB hannel model CM3... Figure -7: One realization of UWB hannel model CM4 Figure 3-: Tapped delay line model of frequeny-seletive hannel... 7 Figure 3-: Optimum demodulator for wideband binary signals... 9 Figure 3-3: Model of binary digital ommuniation system with L th -order diversity. 33 Figure 3-4: Reeiver struture for proessing signal orrupted by S.. 35 Figure 3-5: Linear transversal filter equalization struture Figure 3-6: Blo diagram of hannel with zero-foring equalizer Figure 3-7: Struture of deision feedba equalizer Figure 4-: UWB transmission system model Figure 4-: The traditional RAE demodulation implementation Figure 4-3: A simplified RAE demodulation implementation 53 Figure 4-4: The number of single-path signal orrelators in a UWB Rae reeiver as a funtion of perentage energy apture for reeived waveforms in an offie P upper plot) and H lower plot) representing typial high SNR and low SNR environment. n eah plot, 49 measurement waveforms are used 58 Figure 4-5: N- seletive MRC RAE reeiver struture... 6 X

12 Figure 4-6: N-seletive MRC RAE reeiver with LMS adaptive equalizer 67 Figure 4-7: N-seletive LMS RAE reeiver struture. 7 Figure 5-: Blo diagram of simulation Figure 5-: Simulin blo diagram of the signal generator.. 75 Figure 5-3: Blo diagram of UWB hannel. 76 Figure 5-4: Simulin blo diagram of N-seletive MRC RAE reeiver and LMS adaptive equalizer Figure 5-5: Simulin blo diagram of N-seletive RAE reeiver with LMS adaptive ombiner.. 80 Figure 5-6: The BER for different pilot sequene Figure 5-7: The BER for N-seletive MRC RAE reeiver strutures. 8 Figure 5-8: The BER for LMS equalizers with different stages Figure 5-9: The error signal for 8-seletive MRC RAE reeiver with LMS adaptive equalizer Figure 5-0: The error signal for 8-seletive RAE reeiver with LMS adaptive ombiner. 85 Figure 5-: The error signal for 6-seletive MRC RAE reeiver with LMS adaptive equalizer.. 86 Figure 5-: The error signal for 6-seletive RAE reeiver with LMS adaptive ombiner. 86 Figure 5-3: The error signal for 3-seletive MRC RAE reeiver with LMS adaptive equalizer.. 87 Figure 5-4: The error signal for 3-seletive RAE reeiver with LMS adaptive ombiner Figure 5-5: The BER for N-seletive MRC RAE reeiver with LMS adaptive equalizer. 89 Figure 5-6: The BER for N-seletive RAE reeiver with LMS adaptive ombiner Figure 5-7: The BER for three different reeiver strutures with N X

13 Figure 5-8: The BER for three different reeiver strutures with N6...9 Figure 5-9: The BER for three different reeiver strutures with N3.93 X

14 List of Aronyms AV BER CMOS DD DFE DSSS FCC FR F S LAN LMS LOS LT MB ML MRC MSE NLOS OFDM PAM audio and video bit error rate omplementary metal oxide semiondutor deision-direted deision-feedba equalizer diret-sequene spread-spetrum Federal Communiation Commission finite impulse response intermediate frequeny inter-symbol interferene loal area networ least-mean-square line of sight linear time-invariant multi-band maximum-lielihood maximal ratio ombiner mean-square-error non-line of sight orthogonal frequeny division multiplexing pulse amplitude modulation X

15 PAN PPM RF RMS SS THSS TR UWB WAN WPAN personal area networ pulse position modulation radio frequeny root-mean-square spread spetrum time-hopping spread-spetrum transmitted referene ultra wideband wide area networ wireless personal area networs XV

16 Chapter ntrodution n this hapter, a brief history of ultra wideband UWB) tehnology is desribed first and then motivations of the researh are given. Desription of the problems and obetives are also ontained in this hapter.. Historial baground The onept of UWB ommuniations originated in the early days of radio. n the 900 s, the Maroni spar gap transmitter the beginning of radio) ommuniated by spreading a signal over a very wide bandwidth. This use of spetrum did not allow for sharing. The ommuniation world abandoned wideband ommuniations in favour of narrowband in order to share the available bandwidth. The government of eah ountry governs spetrum alloation and provides guidelines for radiated power in the bandwidths of narrowband ommuniation systems and for inidental out of band radiated power. The origin of ultra wideband tehnology stems from wor in time-domain eletromagneti begun in 96 to fully desribe the transient behaviour of a ertain lass of mirowave networs through their harateristi impulse response [, ]. The onept was indeed quite simple. nstead of haraterizing a linear timeinvariant LT) system by the more onventional means of a swept frequeny response i.e., amplitude and phase measurements versus frequeny), an LT system ould alternatively be fully haraterized by its response to an impulsive exitation -- the so-alled impulse response ht). n partiular, the output yt) of suh a system

17 to any arbitrary input xt) ould be uniquely determined by the well-nown onvolution integral: y t) h u) x t u) du.) However, it was not until the advent of the sampling osillosope Hewlett- Paard 96) and the development of tehniques for sub-nanoseond pulse generation, to provide suitable approximations to an impulse exitation, that the impulse response of mirowave networs ould be diretly observed and measured. One impulse measurement tehniques were applied to the design of wideband, radiating antenna elements [, ], it quily beame obvious that short pulse radar and ommuniation systems ould be developed with the same set of tools. While at the Sperry Researh Center, part of the Sperry Rand Corporation, Ross applied these tehniques to various appliations in radar and ommuniations [, ]. The invention of a sensitive, short pulse reeiver to replae the umbersome time-domain sampling osillosope further aelerated system development. n 973, Sperry was awarded the first UWB ommuniations patent. Through the late 980's, this tehnology was alternately referred to as base-band, arrier-free or impulse ommuniations. The term "ultra wideband" was not applied until 989 by the U.S. Department of Defene. By that time, UWB theory, tehniques and many hardware approahes had experiened nearly 30 years of extensive development. By 989, for example, Sperry had been awarded over 50 patents in the field overing UWB pulse generation and reeption methods, for appliations suh as ommuniations, radar, automobile ollision avoidane, positioning systems, liquid level sensing and altimetry [].

18 On February 4 th, 00, the Federal Communiation Commission FCC) in the US approved the use of this very ontroversial ultra-wideband tehnology for ommerial appliations [3]. The targeted appliations for UWB tehnology are those that traditionally suffered from the multi-path fading effets lie indoor highspeed ommuniations and positioning, ground penetrating radars, through-wall and medial imaging systems or other seurity systems. n issuing its rules for UWB, the FCC ommissioners said that they were taing extreme are to avoid any possible interferene and believe that after a trial period, the ommission will be able to broaden the UWB appliations permitted. UWB proponents believe that UWB pulses will not ause interferene with other narrowband appliations beause the pulses ontinuously hange frequenies and operate at extremely low power levels.. Motivation Aording to FCC s definition, ultra wideband radio is a ommuniation system whih utilizes a signal whose frational bandwidth is greater than 0. or whih oupies 500 MHz or more of the spetrum. Typial UWB radios ommuniate using sub-nanoseond pulses without a arrier. The reason why a ommuniation sheme using narrow pulse signals has been proposed is beause of their novel properties whih possess advantages over onventional narrow-band or wide-band signals. A UWB signal supplies that bandwidth at a lower enter frequeny, whih is advantageous for operation in heavy multi-path environments and for penetration of materials. Resolvable multi- 3

19 path and the penetration apability enable a vision of potential UWB radio appliations in omplex multi-path environments, inluding indoor wireless loal area networ LAN). Furthermore, the absene of a sinusoidal arrier may allow a simpler radio arhiteture beause no intermediate frequeny F) stage is neessary. The benefits of an inreasingly mobile lifestyle introdued by wireless tehnologies in ell phones and home PCs resulted in a greater demand for the same benefits in other onsumer devies. Consumers have enoyed the inreased onveniene of wireless onnetivity. They will soon demand it for video reording and storage devies, for real-time audio and video AV) streaming, interative gaming, and AV onferening servies as the need for digital media beomes more predominate in the home. Many tehnologies used in the digital home, suh as digital video and audio streaming, require high bandwidth onnetions to ommuniate. Considering the number of devies used throughout the digital home, the bandwidth demand for wireless onnetivity among these devies beomes very large indeed. The wireless networing tehnologies developed for wireless onneting PCs, suh as Wi-Fi EEE 80.a, b, g) and Bluetooth tehnology are not optimized for multiple high-bandwidth usage models of a digital home. Although data rates an reah 54 Mbps for Wi-Fi, for example, the tehnology has limitations in a onsumer eletronis environment, inluding power onsumption and bandwidth. When it omes to onneting multiple onsumer eletroni CE) devies in a short-range networ, or wireless personal area networs WPAN), a wireless tehnology is required to support multiple high data rate streams, onsume 4

20 very little power, and maintain low ost, while fitting into a very small physial paage, suh as a PDA or a ell phone. The emerging UWB wireless tehnology and silion developed for UWB appliations offer a ompelling solution..3 Problems and obetives UWB differs substantially from onventional narrowband radio frequeny RF) and spread spetrum tehnologies SS), suh as Bluetooth Tehnology and 80.a/g. UWB uses an extremely wide band of RF spetrum to transmit data. n doing so, UWB is able to transmit more data in a given period of time than traditional tehnologies. The potential data rate over a given RF lin is proportional to the bandwidth of the hannel and the logarithm of the signal-to-noise ratio as stated in the Hartley-Shannon law S C Blog ).) N where: C Maximum hannel apaity, in bits per seond. B Channel bandwidth, in Hertz. S Signal power, in watts. N Noise power, in watts. RF design engineers typially have little ontrol over bandwidth parameters as they are ditated by FCC regulations that stipulate the allowable bandwidth of the signal for a given radio type and appliation. Bluetooth Tehnology, 80.a/g Wi- Fi, ordless phones, and numerous other devies are relegated to the unliensed frequeny bands at 900 MHz,.4 GHz, and 5. GHz. 5

21 Eah radio hannel is onstrained to oupy only a narrow band of frequenies, relative to what is allowed for UWB. UWB is a unique and new usage of a reently legalized frequeny spetrum. UWB radios an use frequenies from 3. GHz to 0.6 GHz a band more than 7 GHz wide. Eah radio hannel an have a bandwidth of more than 500 MHz. To allow for suh a large signal bandwidth, the FCC put in plae severe broadast power restritions. By doing so, UWB devies an mae use of an extremely wide frequeny band while not emitting enough energy to be notied by nearby narrower band devies, suh as the 80.a/g radios. This spetrum sharing allows devies to obtain very high data throughput, but they must be within lose proximity. Strit power limits mean the radios themselves must be low power onsumers. Beause there is no need to emit a high power signal, it is feasible to develop ost-effetive CMOS implementations of UWB radios in plae of expensive high power omponents. With the harateristis of low power, low ost, and very high data rates at a limited range, UWB is positioned to address the maret for high-speed WPAN. Potential future and rule limitation of UWB, lead to two important harateristis of the tehnology: high bit rate and low emitting power. Based on the power emission limit of UWB, the only hoie for implementation is using low level modulation tehnology. To obtain suh a high bit rate using low level modulation tehniques, the required symbol period is very small. Aording to the UWB hannel model from EEE P80.5 [4], there is a root-mean-square RMS) delay spread of approximately 5ns for a 4-0 meter range with a non line-of-sight transmission. This spread indiates that a signifiant inter-symbol interferene S) 6

22 is unavoidable. The traditional RAE reeiver struture to ollet multi-path energy in UWB systems does not ombat S very well. Some published results on UWB have negleted this problem as most performane analyses employ a RAE reeiver under the assumption that hannel delay spreads are muh less than system symbol time [5, 4, 7, 9]. Still for a 4-0 meter range with a non line-of-sight transmission in the UWB hannel model from EEE P80.5, the average number of signifiant paths apturing greater than 85% energy is more than sixty. How ould the multi-path signal s energy be aptured effetively for a reasonable ost? How ould the dense multi-path hannel parameters be measured in the high data rate appliation environment? The first obetive is to find a method to estimate the hannel parameters and gather multi-path energy with the low omputation omplexity. The seond obetive is to find suitable equalization tehnologies to suppress the signifiant inter-symbol interferene. The overall obetive of this researh is to propose low ost UWB reeiver strutures with low omplexity and a low sampling rate that an ahieve satisfatory performane under a signifiant S. n this thesis, the sample and effetive sliding orrelation algorithm is applied for UWB hannel estimation. An analog and digital hybrid implementation of a RAE reeiver struture is proposed. Two shemes for suppressing S, LMS equalization and LMS ombining, are developed. Simulations are performed using the popular simulation tool, MATLAB Simulin. Aording to the simulation results, the obetives are ahieved. 7

23 .4 Thesis overview This hapter provides a brief historial baground on UWB ommuniation systems. Motivation, problems and obetives are also disussed. The rest of this thesis is organized as follows. Chapter provides a definition of UWB signal. Advantages of this tehnology are also disussed. Existing UWB system shemes are briefly reviewed and hannel haraterization of UWB is introdued. Chapter 3 deals with signal proessing tehnologies in the presene of multi-path, RAE multi-path ombining tehniques and different equalization tehniques. Chapter 4 desribes the proposed reeiver strutures. Chapter 5 fouses on simulation set-up. This hapter also provides analysis and disussion of the simulation results. Finally, onlusions on the aomplished researh obetives are summarized and researh topis for future wor are suggested in Chapter 6. 8

24 Chapter UWB Communiation Tehnology n this hapter, the offiial definition and advantages of UWB are given. Basi onepts of several urrent main stream UWB system shemes are introdued subsequently. Finally UWB hannel modes are desribed.. Definition of UWB The UWB tehnology was often referred to as base-band, arrier-free or short impulse. But there was not a lear definition of UWB until the Federal Communiations Commission s Report and Order [3], issued on Feb 00, gives an offiial definition for UWB. Aording to this definition, an UWB signal is any signal whose frational bandwidth is greater than 0.0 or oupies 500 MHz or more of the spetrum. The formula proposed by the FCC for alulating frational bandwidth is H L H L f f ) / f f ), where f is the upper frequeny of the 0 db emission point and f is the lower H frequeny of the 0 db emission point. The enter frequeny of the transmission was defined as the average of the upper and lower 0 db points, i.e., f f ) / L. Meanwhile UWB signals must meet the spetrum mas shown in H Fig. - and Fig. -. Aording to the FCC s rules, there is 7.5 GHz bandwidth 3. GHz 0.6 GHz) available for UWB ommuniations and measurement systems. The allowed power emission level is -4.3 dbm/mhz. The equipment must be designed to L 9

25 GPS Band Figure -: UWB emission limit for indoor systems GPS Band Figure -: UWB emission limit for outdoor hand-held systems ensure that the operation an only our indoors or it must onsist of hand held devies that may be employed for suh ativities as peer-to-peer operation. A omparison with the other unliensed bands urrently available in the US is shown in Table. 0

26 Table -: Unliensed bands Unliensed bands Frequeny of operation Bandwidth SM at.4ghz GHz 83.5MHz U-N at 5GHz GHz GHz 300MHz UWB GHz 7,500MHz Given the reent spetral alloation and the new definition of UWB adopted by FCC, UWB is not only ust onsidered as a tehnology anymore, but also available spetrum for unliensed use. This means that any transmission signal that meets the FCC requirements for UWB spetrum is aeptable. This, of ourse, is not ust restrited to impulse radios or high speed spread spetrum radios pioneered by ompanies so far, but opened to any tehnology that utilizes more than 500MHz spetrum in the allowed spetral mas.. Advantages of UWB Beause UWB waveforms are of suh short time duration, they have some unique properties. n ommuniations, for example, UWB pulses an be used to provide extremely high data rate performane in multi-user networ appliations. For radar appliations, these same pulses an provide a very fine range resolution and a preision distane and positioning measurement apability. These short duration waveforms are relatively immune to multi-path effets ompared to normal narrow band systems as observed in mobile and in-building

27 environments. As a result, UWB systems are partiularly well suited for high-speed wireless appliations. As bandwidth is inversely related to pulse duration, the spetral extent of these waveforms an be made quite large. With proper engineering design, resultant energy densities i.e., transmitted Watts of power per unit Hertz of bandwidth) an be quite low. This low energy density is translated into a low probability of detetion RF signature. The low probability of detetion signature is of partiular interest for military appliations; meanwhile, it also produes minimal interferene to proximity systems and minimal RF health hazards, whih is a signifiant benefit for both military and ommerial appliations. Among the most important advantages of UWB tehnology, however, are those of low system omplexity and low ost. UWB systems an be made nearly "alldigital," with minimal RF or mirowave eletronis. Beause of the inherent RF simpliity of UWB designs, these systems are highly frequeny adaptive, enabling them to be positioned anywhere within the RF spetrum. This feature avoids interferene to existing servies, while fully using the available spetrum. n summary, UWB presents a ompelling solution to many of the hallenges faing today's wireless industry and appliations. These inlude the following [30]: Low radiated power: UWB is limited by regulation to power levels that are a tiny fration of other radio tehnologies, with possible health benefits and adaptation to sensitive environments, suh as hospitals and airports.

28 Speed - The same UWB devie an sale from speeds far in exess of urrent ommuniation networs, to very low speed and low power) appliations, suh as meter reading. Multiple hannels - UWB an support hundreds of simultaneous hannels, ompared to three for 80.b, or ten for 80.a. Simultaneous networing - This tehnology an funtion as a personal area networ PAN), a loal area networ LAN), and a wide area networ WAN), simultaneously. t is the equivalent of Bluetooth, 80., and 3G onverging, but in a single networ, with a single devie. Lower ost and omplexity - Devies using RF spetrum require a real radio reeiver and so are more omplex in terms of omponents, higher ost, and onsume signifiantly more power than UWB whih operates at lower power, and requires fewer omponents. Greater seurity - The inherent digital nature of UWB transmission, oupled with its operation in the lower power level, maes UWB perhaps the most seure means of wireless transmission available. Co-existene - Beause UWB signals an o-exist with onventional RF arriers, the tehnology will open up vast new ommuniation possibilities by reating a new ommuniation medium that peaefully oexists with existing tehnologies. UWB is an RF wireless tehnology and, as suh, is still subet to the same laws of physis as other RF tehnologies. Thus, there are obvious tradeoffs to be made in 3

29 signal-to-noise ratio versus bandwidth, range versus speed, and average power levels, and so on..3 Existing UWB system shemes There are four popular UWB system shemes: time-hopping spread-spetrum ultra wideband THSS-UWB); diret-sequene spread-spetrum ultra wideband DSSS- UWB); multi-band orthogonal frequeny division multiplexing ultra wideband MB-OFDM-UWB) and transmitted referene ultra wideband TR-UWB). Time-hopping spread-spetrum ultra wideband n THSS-UWB, the transmitted signal antipodal modulation an be defined as [6] S TH t) for one user using binary where w tr S t) w t T T ) D ).) TH tr f NS t) denotes the transmitted pulse form that has a maximum amplitude of one, a duration of T and is transmitted with a repetition period T. The position of a transmitted pulse within eah repetition period is determined by a pseudorandom ode whih selets one of the N slots, eah having a duration of T. The f pseudorandom ode taes integer value between 0 and N and it is assumed that NT. Moreover, {0, } T f D is a data stream and x denotes the integer part of x. A new bit starts with 0 mod N S. Eah information bit is transmitted with N S pulses and has a duration of T N S S T f. Figure -3 shows the time divisions. 4

30 Figure -3: Time divisions in THSS-UWB sheme Diret-sequene spread-spetrum ultra wideband A DSSS-UWB system is basially idential to an ordinary DSSS system exept that the bandwidth spreading effet is ahieved by pulse shaping. Similar to THSS- UWB above, DSSS-UWB is defined as S t) w t T ) n D ).) DS tr m N DS where n is a pseudorandom ode that taes values {±}. Eah information bit onsists of N DS pulses and has a duration of T N T. b DS m The flexibility provided by the FCC ruling greatly expands the design options for UWB ommuniation systems. Designers are free to use a ombination of subbands within the spetrum to optimize system performane, power onsumption and design omplexity. UWB systems an still maintain the same low transmit power as if they were using the entire bandwidth by interleaving the symbols aross these sub-bands [7]. Multi-band orthogonal frequeny division multiplexing ultra wideband 5

31 The MB-OFDM-UWB system transmits data simultaneously over multiple arriers spaed apart at preise frequenies by mean of OFDM modulation tehniques. Benefiial attributes of MB-OFDM inlude high spetral flexibility and resilieny to RF interferene and multi-path effets. Regardless of present or future spetral alloations and emissions restritions in various regions of the world, MB-OFDM is apable of omplying with loal regulations by dynamially turning off ertain tones or hannels using software. This flexibility, not demonstrated by other system shemes, enables worldwide adoption of UWB systems. Transmitted referene ultra wideband To irumvent the drawbas of RAE reeivers, e.g. hannel estimation or finding a suitable template pulse form for orrelation, TR-UWB shemes are well suited. A TR-UWB system transmits a doublet every T S seonds. The first pulse of eah doublet is information free, and the seond delayed pulse that is modulated by pulse amplitude modulation PAM), or pulse position modulation PPM) and delayed by T d seonds arries the user s information. Denote the pulse by t) w tr with duration T and the binary PAM symbol by D {0, }. The transmitted signal p an be desribed by [8] S TR t) [ wtr t TS ) D ) wtr t TS Td )].3) Reasonably assume T d > T p and T T < T, meaning the first and seond pulses d p S do not interfere eah other before propagating through a hannel. However, large pulse spaing inevitably sarifies data rate for good performane, espeially when 6

32 the hannel spread is very large [4]. Meanwhile, the first pulse may severely interfere with the seond pulse and ause inter-pulse interferene..4 Channel haraterization of UWB The EEE UWB hannel model is based on the Saleh Valenzuela model where multi-path omponents arrive in lusters [4, 9]. n this model, a lognormal distribution was used rather than a Rayleigh distribution for the multi-path gain magnitude. n addition, independent fading is assumed for eah luster as well as eah ray within the luster. Therefore, the multi-path model onsists of the following disrete time impulse response: L i i i hi t) X α, lδ t Tl τ, l ).4) i l 0 0 i where α } are the multi-path gain oeffiients, { i } is the delay of the l th luster, {, l T l { τ i, l } is the delay of the th multi-path omponent relative to the l th luster arrival time { T i },{ X } represents the log-normal shadowing, and i refers to the i l i realization of the hannel. Finally, the model uses the following definitions: T l the arrival time of the first path of the l-th luster. τ,l the delay of the -th path within the l-th luster relative to the first path arrival time, T l. Λ luster arrival rate. λ ray arrival rate, i.e., the arrival rate of the paths within eah luster. th 7

33 By definition, τ 0. The distribution of luster arrival time and the ray arrival time are given by 0, l p Tl Tl ) Λ exp[ Λ Tl Tl )], l 0.5) > p τ τ ) λ exp[ λ τ τ )], 0.6), l ), l, l ), l > The hannel oeffiients are defined as follow: α,, l p, lξl β, l 0 log0 ξ l β, l ) Normal μ, l, σ σ ), or ξ β l, l 0 μ, l n n ) / 0 where n Normal0, ) and n Normal0, ) are independent and σ σ orrespond to the fading on eah luster and ray respetively. E ξlβ, l Ω 0 e T / Γ l e τ,l / γ.7) where T l is the exess delay of bin l and Ω 0 is the mean energy of the first path of the first luster, and p, l is equiprobable /- to aount for signal inversion due to refletions. The μ,l is given by μ, l 0ln Ω0 ) 0Tl / Γ 0τ, l / γ σ σ )ln0) ln0) 0.8) n the above equations, ξ l reflets the fading assoiated with the l th luster, and β,l orresponds to the fading assoiated with the th ray of the l th luster. Note that, a omplex tap model is not adopted here. The omplex base-band model is a natural fit for narrowband systems to apture hannel behavior independently of arrier frequeny, but this motivation breas down for UWB systems where a realvalued simulation at RF may be more natural. 8

34 Finally, sine the log-normal shadowing of the total multi-path energy is i aptured by the term,, the total energy ontained in the terms { } is normalized to unity for eah realization. This shadowing term is haraterized by the following: 0 log0 X i X i ) Normal0, σ ). x As shown above, there are 7 ey parameters that define the model: Λ luster arrival rate. λ ray arrival rate, i.e., the arrival rate of path within eah luster. Γ luster deay fator. γ ray deay fator. σ standard deviation of luster lognormal fading term db). σ standard deviation of ray lognormal fading term db). σ x standard deviation of lognormal shadowing term for total multi-path realization db). These parameters are found by trying to math important harateristis of the hannel. Sine it is diffiult to math all possible hannel harateristis, the main harateristis of the hannel that are used to derive the above model parameters are hosen to be the following: Mean exess delay RMS delay spread Number of multi-path omponents whih is defined as the number of multipath arrivals that are within 0 db of the pea multi-path arrival α,l 9

35 There are four distinguished models: CM: This model is based on line of sight LOS) 0-4m) hannel measurements. CM: This model is based on non-line of sight NLOS) 0-4m) hannel measurements. CM3: This model is based on non-line of sight NLOS) 4-0m) hannel measurements. CM4: This model is generated to fit a 5 nse RMS delay spread to represent an extreme no-line of sight NLOS) multi-path hannel. The following table lists some initial model parameters for a ouple of different hannel harateristis that were found through measurement data. Table -: The EEE UWB hannel harateristis [4] Model Parameters CM CM CM 3 CM 4 Λ /nse) λ /nse) Γ γ σ db) σ db) σ db) x Model Charateristis Mean exess delay nse) τ m ) RMS delay nse) τ rms ) NP 0dB NP 85%) Channel energy mean db) Channel energy std db)

36 Normalized signal amplitude Figure -4: One realization of UWB hannel model CM Normalized signal amplitude Figure -5: One realization of UWB hannel model CM

37 Normalized signal amplitude Figure -6: One realization of UWB hannel model CM3 Normalized signal amplitude Figure -7: One realization of UWB hannel model CM4

38 One realization of eah hannel model impulse response is shown below ontinuous-time model). From Figure -3 the multi-path delay spread for CM is about 90 ns. For CM in Figure -4 the multi-path delay spread inrease to about 0 ns. For CM3 in Figure -5 this value is about 00ns and for CM4 in Figure -6 it is almost 330ns. 3

39 Chapter 3 Reeiver Signal Proessing in the Presene of Multi-path n this hapter, introdution of the RAE reeiver is given at the beginning. Then a desription of S and some of the most used methods to ombat S are introdued subsequently. 3. RAE reeiver The extremely large bandwidth of UWB signals means that the UWB hannel is highly frequeny seletive. Aording to the introdution of UWB hannel model in Chapter, a large number of multi-path omponents arrive at the reeiver with different time delays. This means that large losses in the reeiver s performane will our if a method of ombating this frequeny seletive dense multi-path is not employed. However, the extremely narrow pulses used for UWB transmission lead to inherent path diversity i.e., independent fading of different multi path omponents). This implies that the reeived UWB signal ontains a signifiant number of resolvable multi path omponents whih suggests a RAE type reeiver to oherently ombine them [5, 0]. This signifiantly redues the fading effets and the resulting redution of fading margins in lin power budgets leads to redued transmission power requirements. The RAE reeiver was invented by Prie and Green in 958 []. The reeiver is used to ahieve multi-path diversity by olleting signal energy from eah reeived path using a delay line struture []. n spread spetrum ommuniations, 4

40 the bandwidth of the spread signal is usually larger than the hannel oherene bandwidth. f the bandwidth of the transmitted signal is large enough, it is possible to resolve multi-path omponents into separate signals. n suh a ase, the atual hannel an be modeled as a tapped delay line with time varying tap oeffiients [3]. Now suppose that W is the bandwidth oupied by real band-pass signal, then the band oupany of the equivalent low-pass signal t) s l is f W. Sine t) s l is band-limited to expressed as f W, aording to the sampling theorem s l t) an be [ πw t n / W )] n sin sl t) sl ) n W πw t n / W ) 3.) The Fourier transform of t) s l is S ) W l f l n s n / W ) e 0 πfn / W f f W ) > W ) 3.) The noiseless reeived signal through a frequeny-seletive hannel is expressed in the form πft r t) C f ; t) S f ) e df 3.3) l l where C f ; t) is the time-variant transfer funtion. Substitution for f ) from 3.) into 3.3) yields r t) l πf tn / W ) sl n / W ) C f ; t) e df W n S l 5

41 sl n / W ) t n / W; t) W n 3.4) where τ ; t) is the time-variant impulse response. t is observed that equation 3.4) has the form of a onvolution sum. Hene, it an be expressed in an alternative form as: rl t) sl t n / W ) n / W; t) 3.5) W n t is onvenient to define a set of time-variable hannel oeffiients as n n t) ; t) 3.6) W W Then Equation 3.5) expressed in terms of these hannel oeffiients beome n n r t) t) s t n / W ) 3.7) l l The form for the reeived signal in Equation 3.7) implies that the time-variant frequeny-seletive hannel an be modeled or represented as a tapped delay line with tap spaing /W and tap weight oeffiients { t)}. n fat, it is dedued from Equation 3.7) that the low-pass impulse response for the hannel is n n n τ ; t) t) δ τ n / W ) 3.8) and the orresponding time-variant transfer funtion is C f ; t) n n t) e πfn / W Thus, with an equivalent low-pass-signal having a bandwidth W 3.9), one ahieves a resolution of /W in the multi-path delay profile. Sine the total multi- 6

42 path spread is T m, for all pratial purposes, the tapped delay line model for the hannel an be trunated at T L m W taps. Then the noiseless reeived signal an be expressed in the form L r t) t) s t n / W ) 3.0) l n n l The trunated tapped delay line model is shown in Figure 3-. The time-variant tap weights { t)} n are omplex-valued stationary random proesses. Sine { t)} n represent the tap weights orresponding to the L different delays τ n /W, n,,, L, the unorrelated sattering assumption implies that { t)} are n mutually unorrelated. When statistially independent. { t)} n are Gaussian random proesses, they are t) s l /W /W /W /W ) ) ) t) t t 3 t L Additive noise zt) L r t) t) sl t ) W l z t) Figure 3-: Tapped delay line model of frequeny-seletive hannel 7

43 Now onsider the problem of digital signaling over a frequeny-seletive hannel that is modeled by a tapped delay line with statistially independent timevariant tap weights { t)}. n Binary signaling over the hannel is onsidered. There are two equal-energy signals s l ) and s l t), whih are either antipodal or orthogonal. Their time t duration T is seleted to satisfy the ondition T >> T m. Thus, one may neglet any inter-symbol interferene due to multi-path. Sine the bandwidth of the signal exeeds the oherent bandwidth of the hannel, the reeived signal is expressed as L rl t) t) sli t / W ) z t) v i t) z t), 0 t T i, 3.) where zt) is a omplex-valued zero-mean white Gaussian noise proess. Assume for the moment that the hannel tap weights are nown, then the optimum demodulator onsists of two filters mathed to v ) and v t). The demodulator output is sampled at the symbol rate and the samples are passed to a deision iruit that selets the signal orresponding to the largest output. An equivalent optimum demodulator employs ross orrelation instead of mathed filtering. n either ase, the deision variables for oherent detetion of the binary signals an be expressed as t U m Re T 0 r t) v l m t) dt Re L T r ) ) / ), 0 l t t slm t W dt m, 3.) 8

44 Figure 3- illustrates the operations involved in the omputation of the deision variables. n this realization of an optimum reeiver, the two referene signals are delayed and orrelated with the reeived signal r l t). s l t ) W W W t ) t ) L t) t) r l Summer and integrator Summer and integrator U Re ) U Re ) t ) t ) L t) s l t ) W W W Figure 3-: Optimum demodulator for wideband binary signals Under the ondition that the fading is suffiiently slow, within any one signaling interval, t) is treated as a onstant and denoted as. f the binary signals are antipodal, then sl t) sl t). Thus the deision variables in 3.) may be expressed in the form 9

45 L T l l dt W t s t r U 0 ) / ) Re 3.3) Suppose the transmitted signal is, then the reeived signal is ) s l t ) ) / ) t z W t s t r L l l t T 0 3.4) Substitution of Eq.3.4) into Eq.3.3) yields L T l l L n n dt W t s W n t s U 0 ) / ) / Re 3.5) L T l dt W t s t z 0 ) / ) Re f inter-pulse interferene is negleted, the resulting signals have the property 0, ) / ) / 0 T l l dt W t s W n t s n 3.6) Then Eq.3.5 an be simplified to T l l L dt W t s W t s U 0 ) / ) / Re L T l dt W t s t z 0 ) / ) Re L L N Re α α ε 3.7) where and e φ α T l l dt W t s W t s 0 ) / ) / ε 3.8) T l dt W t s t z e N 0 ) / ) φ 3.9) 30

46 n effet, the tapped delay line demodulator attempts to ollet the signal energy from all reeived signal paths that fall within the span of the delay line and arry the same information. ts ation is somewhat analogous to an ordinary garden rae and, onsequently, the name RAE demodulator has been oined for this demodulator struture by Prie and Green []. t is apparent that the tapped delay line model with statistially independent tap weights provides L replias of the same transmitted signal at the reeiver. Hene, a reeiver that proesses the reeived signal in an optimum manner will ahieve the performane of an equivalent L th -order diversity ommuniation. The L replias of the transmitted signal at the reeiver an be onsidered as arrying the same information-bearing signal passed through L diversity hannels. Eah hannel is assumed to be frequeny-nonseletive and slowly fading. The fading proesses among the L diversity hannels are assumed to be mutually statistially independent. The signal in eah hannel is orrupted by an additive zero-mean white Gaussian noise proess. The noise proesses in the L hannels are assumed to be mutually statistially independent. Thus the equivalent low-pass reeived signals for the L hannels an be expressed in the form φ r t) α e s t) z t),,,, L, m, 3.0) l where { } m αe φ represent the attenuation fators and phase shifts for the L hannels, t) s m denotes the m th signal transmitted on the th hannel, and t) z denotes the additive white Gaussian noise on the th hannel. All signals in the set { s m t)} have the same energy. 3

47 The optimum demodulator for the signal reeived from the th hannel onsists of two mathed filters, one having the impulse response h t) s T ) 3.) t and the other having the impulse response s h t) s T ) 3.) t Of ourse, if BPS is the modulation method used to transmit information, then t) s ). Consequently, only a single mathed filter is required for BPS. t Following the mathed filters is a ombiner that forms the two deision variables. The ombiner that ahieves the best performane is one in whih eah mathed filter output is multiplied by the orresponding omplex-valued onugate) hannel gain α e φ. The effet of this multipliation is to ompensate for the phase shift in the hannel and to weight the signal by a fator that is proportional to the signal strength. Thus, a strong signal arries a larger weight than a wea signal. After the omplex-valued weighting operation is performed, two sums are formed. One onsists of the transmitted 0. The seond onsists of the real part of the outputs from the mathed filters orresponding to a transmitted. This optimum ombiner is alled a maximal ratio ombiner MRC) by Brennan [9]. Of ourse, the realization of this optimum ombiner is based on the assumption that the hannel a } { } attenuations { and the phase shifts φ are nown. A blo diagram illustrating the model for the binary digital ommuniation system desribed above is shown in Figure 3-3. f the modulation is BPS, the output of the maximal ratio ombiner an be expressed as a single deision variable in the form 3

48 U L L Reε α α N 3.3) where ε dt 3.4) T s t) s t) 0 N e φ T 0 z t) s t) dt 3.5) t is notied that Eq.3.3 is idential to the deision variable given in Eq.3.7, whih orresponds to the output of the RAE demodulation. Consequently, the RAE demodulator with perfet estimates of the hannel tap weights is equivalent to a maximal ratio ombiner in a system with L th -order diversity. s s t) t) Channel φ α e Reeiver s s t) t) Channel φ α e z ) t Reeiver Combiner Output Deision Variables s t) s l l t)..... Channel L α Le φ z t) Reeiver L t) z L Figure 3-3: Model of binary digital ommuniation system with L th - order diversity. 33

49 3. Combating S The RAE demodulator desribed above is an optimum demodulator based on the ondition that the bit interval T >> b T m, i.e., there is negligible S. When the ondition is not satisfied, the RAE demodulator output is orrupted. f S is left unompensated, high error rate will our. The detailed information about the equivalent disrete-time model for a hannel with S is in Appendix A. The solution to the S problem is to design a reeiver that employs a means for ompensating or reduing the S in the reeived signal. The ompensator for the S is alled an equalizer. From the potential future and rule limitation of UWB, it is well nown that there are two important harateristis of UWB: high bit rate and low emitting power. Based on the power emission limit of UWB one an only hoose low level modulation tehnology. To obtain suh high bit rate with low level modulation tehniques, the result is very small symbol period. Aording to the UWB hannel models from EEE P80.5 [4], there is a RMS delay spread of approximately 5ns for a 4-0 meter range with a non line-of-sight transmission. This spread indiates that a signifiant inter-symbol interferene is unavoidable. n suh a ase, an equalizer is required to suppress S. At the reeiver, after the signal is demodulated to base-band, it may be proessed by the RAE, followed by an equalizer to suppress the S. The RAE output is sampled at bit rate, and these samples are passed to the equalizer. This struture is shown in Figure 3-4. There are several types of equalization methods being extensively used. One is based on the maximum-lielihood ML) sequene detetion riterion, whih is optimum from a probability of error viewpoint. A seond equalization method is based on the use of 34

50 a linear filter with adustable oeffiients. A third equalization method that is desribed exploits the use of previously deteted symbols to suppress S in the present symbol being deteted; this is alled deision-feedba equalization. All these equalization methods are now desribed in detail. Reeived base-band signal RAE demodulator Sampler Equalizer Output Symbol-rate lo Figure 3-4: Reeiver struture for proessing signal orrupted by S 3.. Optimum maximum-lielihood reeiver The reeived base-band signal an be expressed as { } r t) h t nt ) z t) 3.6) l n n where is the information sequene, ht) represents the response of the hannel n to the input signal pulse g t) and zt) represents the additive white Gaussian noise. The maximum-lielihood estimates of the symbols those that maximize this quantity: p [,,, ] Ι are CM Ι p ) Re n yn nmxnm 3.7) n n m where p 35

51 y n y nt ) r t) h t nt ) dt 3.8) l x n x nt ) h t) h t nt ) dt 3.9) n any pratial system, it is reasonable to assume that S affets a finite number of symbols. Consequently, S observed at the output of the demodulator may be viewed as the output of a finite state mahine. This implies that the hannel output with S an be represented by a trellis diagram, and the maximum-lielihood estimate of the information sequene [,,, ] Ι is simply the most p probable path through the trellis given the reeived demodulator output sequene { } y n. Clearly, the Viterbi algorithm provides an effiient means for performing the trellis searh. 3.. Linear equalization The maximum lielihood signal estimation for a hannel with S has a omputational omplexity that grows exponentially with the length of the hannel time dispersion. f the size of the symbol alphabet is M and the number of p interfering symbols ontributing to S is L, the Viterbi algorithm omputes M L metris for eah new reeived symbol. n most hannels of pratial interest, suh a large omputational omplexity is prohibitively expensive to implement. One suboptimum hannel equalization approah is the linear equalization that employs a linear transversal filter. This filter struture has a omputational omplexity that is a linear funtion of the hannel dispersion length L. This linear transversal filter equalization struture is shown in Figure 3-4. ts input is the 36