MMSE/LSE ESTIMATION AND EQUALIZATION FOR BETTER SIGNAL QUALITY AND PACKET DETECTION IN ULTRA-WIDE BAND SYSTEMS JEBIN JACOB

Transcription

1 MMSE/LSE ESTIMATION AND EQUALIZATION FOR BETTER SIGNAL QUALITY AND PACKET DETECTION IN ULTRA-WIDE BAND SYSTEMS by JEBIN JACOB DALE W. CALLAHAN, COMMITTEE CHAIR GREGORY A. FRANKLIN THOMAS C. JANNETT A THESIS Submitted to the graduate faculty of The University of Alabama at Birmingham in partial fulfillment of the requirements for the degree of Master of Science BIRMINGHAM, ALABAMA 21

2 MMSE/LSE ESTIMATION AND EQUALIZATION FOR BETTER SIGNAL QUALITY AND PACKET DETECTION IN ULTRA-WIDE BAND SYSTEMS JEBIN JACOB MASTER OF SCIENCE IN ELECTRICAL ENGINEERING ABSTRACT Ultra-Wide Band (UWB) radio is a fast-emerging technology that operates at a huge bandwidth using low-power and ultra-short information-bearing pulses. Coupled with Orthogonal Frequency Division Multiplexing (OFDM) which is highly efficient in terms of bandwidth utilization and a robust multi-carrier modulation scheme, UWB can be an effective solution to the demand for low-cost, high-speed, wireless links for shortrange communication. Packet detection is defined as the process of detecting the presence of data packet symbols in the received signal. The Federal Communications Commission (FCC) has limited the maximum emission strength of all UWB signals to be very close to the noise floor, which is defined as the strength of the sum of all noise sources, such as thermal noise and other interfering signals present in a communication channel. This limitation on emission strength increases the chances of a receiver missing a packet symbol, and ultimately the whole system can go out of synchronization. This situation demands a very high signal quality at the UWB receiver so that the probability of packet detection is high and the system remains synchronized. This thesis explores Minimum Mean-Square Error (MMSE) and Least-Squares Error (LSE) methods for estimating the impulse response of the UWB channel. Digital signal processing is performed on the received signal using the estimated impulse ii

3 response to carry out a process called equalization. Equalization helps increase the signal quality by improving the signal peaks and lowering the noises during correlation. The processed output is called the equalized output. Extensive simulations were carried out to establish the effectiveness of schemes for improving the received signal quality in an UWB communication system. Based on the distance between the transmitter and receiver and their line-of-sight as defined by the IEEE P82.15 Working Group for Wireless Personal Area Networks (WPANs), four different channel scenarios were considered for the simulation. Results show that the MMSE and LSE equalizers improve the overall signal quality and make the equalized outputs more accurate and similar to the transmitted signal. iii

4 DEDICATION This thesis is dedicated to my family and friends, for the never-ending love, support, and encouragement they have given me. iv

5 ACKNOWLEDGMENT It is a pleasure to thank the many people who made this thesis possible. First, I thank my advisor, Dr. Dale W. Callahan, for his invaluable advice, kind assistance, and contribution towards the thesis; without him, this thesis would have been impossible. My sincere gratitude is extended to my committee members, Dr. Gregory A. Franklin and Dr. Thomas C. Jannett, from the Department of Electrical and Computer Engineering. I especially thank Dr. Jeffrey R. Foerster, for his kind help and technical assistance during my thesis. From the formative stages of this thesis to the final draft, I owe an immense debt of gratitude to my mentor, Dr. George P. Koomullil. I also thank Dr. Roy P. Koomullil for his guidance, love, and support, which were invaluable in the completion of this thesis. I am grateful to Ms. Sandra Muhammad for assisting me with all of the administrative formalities required to bring my thesis to completion. Last and most important I wish to thank my parents, Jacob P. Koomullil and Saly Jacob. They raised me, supported me, taught me, and loved me. Their support and confidence in me helped make the completion of my graduate work possible. v

6 TABLE OF CONTENTS Page ABSTRACT...ii DEDICATION... iv ACKNOWLEDGMENTS... v LIST OF TABLES... ix LIST OF FIGURES... x LIST OF ABBREVIATIONS...xiii CHAPTER I. INTRODUCTION... 1 A. Ultra-Wide Band Communication and its Importance... 1 B. Problem Definition... 3 C. Approach and Contribution... 4 D. Thesis Outline... 6 II. MULTIBAND OFDM UWB SYSTEM... 8 A. History of UWB... 8 B. Definition of UWB... 1 C. Types of UWB D. Regulatory Issues E. Introduction of Multiband OFDM System F. OFDM System Model ) Transmitter ) Receiver G. Mathematical Analysis of OFDM H. Overview of an UWB Model ) Bernoulli Binary ) Rate Encoder ) Interleaver vi

7 TABLE OF CONTENTS (CONTINUED) Page 4) QPSK Modulator ) OFDM Transmitter ) Frequency Hopping and Filtering ) UWB Channel ) Frequency Dehopping and Filtering ) OFDM Receiver ) QPSK Demodulator ) Synchronization ) Deinterleaver ) Viterbi Decoder III. UWB TIMING SYNCHRONIZATION A. Synchronization ) Frequency Synchronization ) Timing Synchronization B. Packet Detection C. Packet Detection Algorithms ) Received Signal Energy Detection ) Double Sliding Window Packet Detection ) Correlation Detection ) Delayed Correlation or Autocorrelation Detection IV. RESEARCH METHODOLOGY A. Introduction B. System Model C. Standard Test Data D. UWB Channel Model E. Cross-Correlation F. Autocorrelation G. MMSE and LSE Estimators H. Data Analysis I. UWB Operating SNRs... 5 J. MMSE/LSE Channel Estimation and Signal Equalization Block Diagram V. RESULTS VI. DISCUSSION vii

8 TABLE OF CONTENTS (CONTINUED) Page VII. CONCLUSION AND FUTURE WORK A. Conclusion B. Future Work LIST OF REFERENCES viii

9 LIST OF TABLES Table Page 1 Summary of FCC Restrictions on UWB Operation Summary of the Four Channel Model Properties Channel Characteristics and Corresponding Model Parameters ix

10 LIST OF FIGURES Figure Page 1 Spatial capacity comparison between IEEE 82.11, Bluetooth, and UWB Comparison of the fractional bandwidth of a narrow band and UWB communication system Spectrum of an Impulse Ultra-Wide Band signal Spectrum of an OFDM-based MB-UWB signal FCC spectral mask for UWB systems FCC spectral mask for UWB systems An ideal model of an OFDM transmitter An ideal model of an OFDM receiver Top level model of a typical UWB system Application of the packet detection in the timing synchronization Packet detection using received signal energy detection method Packet detection using double sliding window packet detection method Packet detection using cross-correlation detection method Packet detection using delayed correlation detection method Base-band OFDM system Block diagram of test data constructed using PLCP preamble Impulse response realization for channel model Impulse response realization for channel model x

11 LIST OF FIGURES (CONTINUED) Figure Page 19 Impulse response realization for channel model Impulse response realization for channel model Cross-correlation between test series f(t) and PLCP preamble g(t) Autocorrelation plot of the test series f(t) using a single PLCP period g(t) Available SNR at the receiver as a function of distance between the UWB transmitter and the receiver Steps involved in the estimation and equalization process using the MMSE/LSE estimation and equalization method PLCP preamble cross-correlated with ideal channel output (a), non-equalized output (b), MMSE equalized output (c), and LSE equalized output (d) PLCP preamble cross-correlated with ideal channel output (a), non-equalized output (b), MMSE equalized output (c), and LSE equalized output (d) Expanded version of Fig Autocorrelation using ideal channel output (a), non-equalized output (b), MMSE equalized output (c), and LSE equalized output (d) Autocorrelation using ideal channel output (a), non-equalized output (b), MMSE equalized output (c), and LSE equalized output (d) Average autocorrelation power for ideal channel output, nonequalized output, MMSE equalized output, and LSE equalized output Average autocorrelation power for ideal channel output, nonequalized output, MMSE equalized output, and LSE equalized output xi

12 LIST OF FIGURES (CONTINUED) Figure Page 32 Average autocorrelation power for ideal channel output, nonequalized output, MMSE equalized output, and LSE equalized output Average autocorrelation power for ideal channel output, nonequalized output, MMSE equalized output, and LSE equalized output Average autocorrelation power for ideal channel output, nonequalized output, MMSE equalized output, and LSE equalized output Cross-correlation PAPR for ideal channel output, non-equalized output, MMSE equalized output, and LSE equalized output Cross-correlation PAPR for ideal channel output, non-equalized output, MMSE equalized output, and LSE equalized output Cross-correlation PAPR for ideal channel output, non-equalized output, MMSE equalized output, and LSE equalized output Cross-correlation PAPR for ideal channel output, non-equalized output, MMSE equalized output, and LSE equalized output Cross-correlation PAPR for ideal channel output, non-equalized output, MMSE equalized output, and LSE equalized output xii

13 LIST OF ABBREVIATIONS ADC AWGN DAC DARPA db dbm DSSS DS-UWB ECMA EIRP ESD FAA FCC FEC FFT GPS HDTV IEEE IFFT Analog to Digital Converter Additive White Gaussian Noise Digital to Analog Converter Defense Advanced Research Projects Agency Decibel Decibel-milliWatt Direct-Sequence Spread Spectrum Direct Sequence Ultra-Wide Band European Computer Manufacturers Association Effective isotropic radiated power Energy Spectral Density Federal Aviation Administration Federal Communications Commission Forward Error Correction Fast Fourier Transform Global Positioning System High Definition Television Institute of Electrical and Electronics Engineers Inverse Fast Fourier Transform xiii

14 LIST OF ABBREVIATIONS (CONTINUED) I-UWB LANL LLNL LSE MB MC MIR MMSE NTIA OFDM PAM PAPR PLCP PN PPM PSK QAM QPSK RF SNR TH-UWB USAF Impulse Ultra-Wide Band Los Alamos National Laboratory Lawrence Livermore National Laboratory Least-Squared Error Multi-Band Multi-Carrier Micro power Impulse Radar Minimum Mean-Squared Error National Telecommunications and Information Administration Orthogonal Frequency Division Multiplexing Pulse Amplitude Modulation Peak-to-Average Power Ratio Physical Layer Convergence Protocol Pseudo Noise Pulse Position Modulation Phase-Shift Keying Quadrature Amplitude Modulation Quadrature Phase-Shift Keying Radio Frequency Signal to Noise Ratio Time-Hopping Ultra-Wide Band United States Air Force xiv

15 LIST OF ABBREVIATIONS (CONTINUED) USB UWB WiFi WPAN Universal Serial Bus Ultra-Wide Band Wireless Fidelity Wireless Personal Area Network xv

16 I. INTRODUCTION A. Ultra-Wide Band Communication and its Importance One of the main features of any Ultra-Wide Band (UWB) communication system is its huge bandwidth. Usually, the instantaneous available bandwidth is much higher than what the communication system actually needs to deliver the data [1]. Shannon s channel capacity equation gives the maximum data rate that can be achieved for a given bandwidth and signal-to-noise ratio (SNR) in a data channel [2]. S C = B log (1) N In (1), C is the maximum channel capacity in bits/sec, B is the channel bandwidth in Hz, S is the signal power in watts, and N is the noise power in watts. The traditional narrow band technologies focus on improving the SNR to increase the data rate. However, the UWB uses a larger bandwidth to increase the total throughput through the channel. From (1), it is evident that the data capacity increases faster with an increase in bandwidth rather than an increase in the SNR. As a result, the UWB system can achieve higher data throughput than a traditional narrowband system. The spatial capacities of some of the wireless standards that are being developed by the Bluetooth special interest group and the Institute of Electrical and Electronics Engineers (IEEE) 82 working group are shown in Fig. 1. Spatial capacity is defined as the total data throughput of all the systems that can coexist in a non-interfering basis in an available spectrum and area and is calculated as the ratio of total throughput to the area 1

17 for a given wireless standard [3]. The spatial capacities of other narrowband wireless standards are nowhere comparable to that of UWB (Fig. 1). The limited capacities of other standards can be traced back to the restricted available bandwidth for these systems, since all of the systems are bound by the channel capacity theorem. UWB systems have a very high data throughput, since they usually have more than 2 GHz of available spectrum [3] Ultra Wideband 1 kbps/sq.m Projected Spatial Capacity (kbps) b 1 kbps/sq.m Bluetooth 1 3 kbps/sq.m Wireless Standards 82.11a 83 kbps/sq.m Fig. 1. Spatial capacity comparison between IEEE 82.11, Bluetooth, and UWB. Even though UWB can provide a very high channel capacity, this very high data throughput is available only at a limited range. The Federal Communications Commission (FCC) mandated the Effective Isotropic Radiated Power (EIRP) emission of all UWB signals to be very close to the noise floor. Noise floor is defined as the strength of the sum of all noise sources, such as thermal noise and other interfering signals present in a communication channel [4], [5]. This limitation in EIRP made UWB technology the 2

18 most effective in short-range (less than 1 meters) applications. The throughput decreases exponentially after this range [4]. B. Problem Definition One of the important steps to be performed at the receiver in a complex system, such as UWB, is data synchronization. FCC has mandated that EIRP emission of all UWB signals to be below -4 dbm [4], [5]. This limitation in the UWB signal power increases the chances of a receiver missing a data packet and, ultimately, the data received at the receiver become worthless. In addition, a UWB system functions at a very high data rate, and missing a synchronization packet can result in the loss of a huge amount of data. Therefore, the receiver must continuously scan for incoming data and perform rapid data synchronization. The most important step in data synchronization is the detection of the data packet at the receiver, which is the process of detecting the presence of the data packet in the received signal. Normally, the packet detection is performed by monitoring the energy of the received signal. A sudden change in the energy indicates the presence of the data packet. The packet detection is successful if the power of the signal goes above a preset threshold. This concept of modeling packet detection as a binary hypothesis is further discussed in Chapter 3. This scheme for packet detection is critical for UWB communication, since the maximum allowed EIRP is close to the noise floor. A sudden change in noise power can result in high energy in the received signal. This peak in noise energy can lead to a false detection of a packet and, sometimes, missing the packet detection completely. 3

19 The probability of successful packet detection depends on the quality of the received signal. If the quality of the received signal is sufficiently high, the receiver can successfully distinguish the energy peak at the instant of packet arrival and discard the noise between the peaks. This situation demands a scheme that can increase the quality of the received signal, leading to a high probability of detecting a data packet. This thesis focuses on increasing the received signal quality through the use of equalization to eliminate the channel-induced signal distortions by estimating the channel impulse response and performing digital signal processing on the received signal. C. Approach and Contribution As mentioned in the previous section, the central purpose of this thesis is to investigate methods to improve received signal quality. The improvement in signal quality is achieved by equalization, which is the process of convolving the inverse of the estimated impulse response of the channel with the received signal. The estimation and equalization process is performed in three stages: estimating the channel impulse response from the channel model, computing the inverse of the impulse response, and convolving the inverse impulse response with the output of the channel. Ideally, the equalized signal is completely devoid of channel-induced distortions and is similar to the transmitted signal. Two estimators based on well-known Minimum Mean-Squared Error (MMSE) and Least-Squared Error (LSE) methods are presented for the estimation of the channel impulse response. These two methods have been proven to be effective in Gaussian channels and other scenarios, such as in probability theory and linear prediction models 4

20 [6], [7]. However, MMSE and LSE methods have not been used in UWB systems for estimation of channel impulse responses and equalization. Extensive simulations were carried out using a UWB channel model defined by the IEEE P82.15 Working Group for Wireless Personal Area Networks (WPANs). Four different channel scenarios based on the distance between the receiver and transmitter and their line-of-sight were considered for establishing the effectiveness of the scheme for estimation and equalization. The output signals from the four channel models are without any signal processing and are called non-equalized outputs. These non-equalized outputs are equalized using the MMSE and LSE estimated channel impulse responses. The autocorrelation and cross-correlation results of MMSE/LSE estimated and equalized signals are compared to that of an ideal channel without any multipaths, having only Additive White Gaussian Noise (AWGN). Ideally, the equalized output should be similar to the ideal channel output, since the equalization process is trying to remove the effect of channel multipaths from the non-equalized output. The simulation results are also compared with the autocorrelation and cross-correlation results of a non-equalized signal. Cross-correlation graphs show that the MMSE and LSE estimators and equalization perform better in the presence of high SNRs. As the complexity of the channel impulse response grows, a gap develops between the ideal performance and the performance of MMSE/LSE equalization and estimation. The Peak Average Power Ratio (PAPR) of the cross-correlation is also used to analyze the results of the channel estimation and signal equalization schemes. The PAPR of a signal is defined as the ratio of the peak power of the signal to the total noise between the two peaks in the cross-correlation. Equalization improves the PAPR of the 5

21 signal, since the peak of the signal is higher and the noise level between two peaks is lower. Improved PAPR indicates a higher quality received signal and, thus, PAPR gives a measure of the quality of the signal. D. Thesis Outline This thesis is divided into two parts. The first part (Chapters 1 to 3) deals with the problem statement and covers UWB radio fundamentals, OFDM modulation, and spectral characteristics of UWB channels. The second part (Chapters 4 to 7) is dedicated to the discussion of the research methodology and the idea of using MMSE/LSE estimation and equalization for improving signal quality, followed by results and detailed discussion. The chapters are organized as follows: Chapter 2 introduces the core concepts of UWB radio communications, OFDM systems, and definitions used for the UWB radio signal. This chapter also includes a toplevel UWB model and various modules in the system. Chapter 3 introduces the concept of timing synchronization, packet detection, channel estimation, and signal equalization. This chapter also explains various problems associated with achieving a good synchronization. Chapter 4 explains the research methodology and approach used to improve the received signal quality by the use of better equalization methods. A new approach for channel estimation and equalization using MMSE and LSE methods is presented in this Chapter. Chapter 5 includes the results from the simulations using MMSE and LSE methods for equalization. 6

22 Chapter 6 is dedicated to a detailed discussion of the results and figures presented in Chapter 5. Chapter 7 includes a conclusion and a discussion of the results. Some thoughts on future work are also included. 7

23 II. MULTIBAND OFDM UWB SYSTEM A. History of UWB The progress made in the field of microwave networks led to the origin of UWB. These pioneering studies during the early 196s were lead by Ross and Robins at Sperry Rand Corporation, Harmurth at the Catholic University of America, and Van Ettan at the United States Air Force (USAF) Rome Air Development Centre, as well as engineers at Lawrence Livermore and Los Alamos National Laboratories (LLNL and LANL) [1]. These studies tried to explain the transient behavior of microwave networks using their impulse response. The arrival of sampling oscilloscopes and the development of techniques for the generation of sub-nanosecond pulses helped researchers directly observe and measure the impulse response [8]. Later in the 196s, it became obvious to the researchers at Sperry Rand Corporation that short-pulse radar and communications systems could be developed using the latest technologies. They later started using these radars widely in applications as radar and communications. The first patented design of a UWB communications system in 1972 was made possible at the Sperry Rand Corporation by the invention of a sensitive baseband pulse receiver as a replacement for the sampling oscilloscope [1]. By the early 197s, the commercial applications of UWB began to gain popularity. Morey at the Geophysical Survey Systems Corporations made the first ground-penetrating radar using UWB technology, and this radar was commercially 8

24 available in 1974 [1]. By 1975, researchers could build UWB systems from commercially available Tektronix parts [1]. After the 197s, the researchers started experimenting with UWB technology as a medium for RF communication and sensing using transient pulses in a way that did not interfere with other existing systems. Robert Scholtz at the University of Southern California, in 1993, described a multiple access technique for UWB communications in which each user/system is given a unique spreading code that determines specific intervals when the user allows the data transmission [9]. This publication was a landmark paper since Scholtz s technique can be used not only for UWB radar and point-to-point communications but also for UWB wireless networks. In 1994, McEwan at LLNL was the first one in history to develop a compact, inexpensive, low-power UWB system called Micro-Power Impulse Radar (MIR) [1]. Recently, many companies have entered the UWB market. This large-scale interest in UWB was mainly due to the FCC s decision to allocate a huge bandwidth for the operation of unlicensed UWB devices [11]. Some of the big players in the UWB technology are the FCC, the National Telecommunications and Information Administration (NTIA), the Federal Aviation Administration (FAA), and the Defense Advanced Research Projects Agency (DARPA). They spent many years investigating UWB technology and its effect on existing wireless systems. The results from their investigation helped in guiding the FCC on setting the UWB standards and mode of operations. 9

25 B. Definition of UWB As mentioned in the previous chapter, the main attribute of any UWB communication system is its huge bandwidth. For any UWB transmission system, the instantaneous spectral occupancy is more than 5 MHz, or the fractional bandwidth is more than 2% [4]. The fractional bandwidth of a system is defined as the ratio of energy bandwidth to the center frequency. The energy bandwidth concept is illustrated in Fig. 2. In this figure, f L is the lower limit and f H is the higher limit of the Energy Spectral Density (ESD). The energy bandwidth is identified by the frequencies f L and f H, which delimit the interval where most of the instantaneous energy of the waveform falls. The interval[ f L, f H ] is called the energy bandwidth, which is 1 db bandwidth, and the center frequency is defined as f c ( f ) H + f L =. 2 Narrowband Energy Spectral Density (db) f L -1 db f H Frequency (Hz) Fig. 2. Comparison of the fractional bandwidth of a narrow band and UWB communication system. 1

26 Often, the term percent bandwidth is used instead of fractional bandwidth. Percent bandwidth is defined as the fractional bandwidth represented in percent units. For example, a signal with an energy bandwidth of 1 MHz and a center frequency of 2 MHz has a percent bandwidth of 5% and is a UWB signal, since its fractional bandwidth is.5, which is higher than the lower limit of.2. According to the FCC first report and order [11], UWB systems with f > 2. 5 GHz need to have a 1 db bandwidth of at c least 5 MHz, while UWB systems with f < 2. 5 GHz need to have a fractional c bandwidth at least.2. Due to these properties of UWB signals, the UWB radios using these signals have some unique advantages [4]: 1) UWB signals can penetrate through obstacles more efficiently. 2) UWB signals can be used for precision sensing and tracking even at the centimeter level. 3) UWB can be used for very high data rates even if the number of users/systems that coexist are huge. 4) UWB radios can be made smaller, with lower processing power requirements. C. Types of UWB There are two methods in which UWB signal transmissions are performed: the first method is based on sending very short interval pulses to convey information, and the second method is based on using several simultaneous carriers. The first method of UWB transmission is known as Impulse Ultra-Wide Band, and the second type of UWB 11

27 transmission is called as Multi-Carrier Ultra-Wide Band. These two methods of UWB transmissions are discussed in the following section. UWB signals are traditionally radio frequency (RF) pulses that are of very short duration. UWB transmissions that use these types of signals are called Impulse Ultra- Wide Band (I-UWB) and are the most commonly used method of transmission. The spectrum of an I-UWB signal is shown in Fig. 3. Typically, the information data symbols are modulated using Pulse Position Modulation (PPM) and Pulse Amplitude Modulation (PAM). The data symbols are encoded using pseudo-random or pseudo-noise (PN) codes in order to shape the spectrum of the generated signal according to the FCC mandated spectral mask for UWB communications. A time dither is introduced, usually to the data symbols, and such a signal is called Time-Hopping UWB (TH-UWB). The encoded data symbols are amplitude modulated by Direct Sequence Spread Spectrum (DSSS), and the resultant UWB signal is called Direct Sequence Ultra-Wide Band (DS-UWB). An alternative to DS-UWB is to divide the available bandwidth into sub-bands and then split orthogonal sub-carriers into a train of short pulses, send the pulses over a channel, and reassemble them at the receiver to recover each sub-carrier separately [1], [12]. This mode of operation is called Multi-Band (MB) Orthogonal Frequency Division Multiplexing (OFDM), and the UWB signal is called MB-OFDM UWB. Recent proposals regarding UWB in the United States and in the IEEE TG3a working groups seems to favor MB-OFDM UWB. This fact is evident by the final channel modeling sub-committee report released in April 25. A detailed discussion about MB- OFDM is presented later in this chapter. The spectrum of an OFDM-based MB-UWB signal is shown in Fig

28 -15 Relative Power (db) Frequency (GHz) Fig. 3. Spectrum of an Impulse Ultra-Wide Band signal. Relative Power (db) Frequency (GHz) Fig. 4. Spectrum of an OFDM-based MB-UWB signal. 13

29 The major difference in the spectral characteristics of I-UWB and MB-UWB signals are shown in Figs For an I-UWB, the frequencies lie in the whole spectrum for a single impulse. The power slowly increases, reaches a maximum at 3 GHz, and then decreases slowly. For a MB-UWB signal, there are many sub-carriers operating in a band of frequencies (Fig. 4). Since these carrier frequencies are orthogonal, they do not interfere with each other. D. Regulatory Issues In the late 199s, the FCC realized the importance of UWB technology and that it could be used for some very important applications, such as radars for high precision tracking, plotters for medical and through-wall imaging, sensors for remote sensing, and transreceiver for secure voice and data communications. This realization about the importance of UWB led the FCC to issue a Notice of Inquiry on September 1, 1998, for revising Part 15 rules allowing the use of bandwidth for UWB devices without any license. Usually, the FCC divides unused spectrum into smaller bands and allocates the bands to specific users or services. However, the FCC allowed UWB devices to function in all frequencies from 96 MHz to 31 GHz. The maximum EIRP was limited so that the UWB operation would not hurt the other wireless systems coexisting in the frequency band, such as Global Positioning Systems (GPS) and Wireless Fidelity (WiFi). UWB was allowed to interfere with the operation of these coexisting systems, but these systems would not experience any performance degradation since UWB operates at very low signal strength that is close to the noise floor. Other narrow band wireless systems 14

30 observe the UWB signals as a noise, and this noise is filtered out at the wireless system receiver. On February 14, 22, the FCC issued its First Report and Order regarding the unlicensed operation of UWB, based on more than 1 documents from 15 different organizations [11]. The report classified UWB operations into three separate categories: 1) Communication and measurement systems. 2) Vehicular radar systems. 3) Imaging systems, including ground penetrating radar, through-wall imaging and surveillance systems, and medical imaging. Each category was allocated a specific spectral mask (Figs. 5 6). The FCC limits EIRP levels of radio transmissions in the frequency spectrum. These limitations are known as the spectral mask, sometimes also referred to as the transmission mask. These restrictions on the spectral mask reduce the interference in other systems by limiting EIRP in specific bandwidths that are being shared by different wireless systems. For example, the maximum allowed EIRP at 6 GHz is below -4 dbm/mhz for an indoor commercial system. For a vehicular radar system, the maximum allowed EIRP at 6 GHz is less than -6 dbm/mhz (Fig. 5). Table 1 summarizes some of the UWB applications and their frequency band of operation, along with user restrictions that are imposed when operating in that particular application mode [1], [4], [11]. E. Introduction of Multiband OFDM system Orthogonal Frequency Division Multiplexing (OFDM) is a form of multi-carrier transmission. In this form of transmission, the sub-carriers are made to overlap in 15

31 -4 UWB EIRP Emission Level (dbm/mhz) Mask 1-75 Mask 2 Mask Frequency GHz Fig. 5. FCC spectral mask for UWB systems. Mask 1 represents the mask for indoor UWB communication systems. Mask 2 represents the mask for outdoor UWB communication systems. Mask 3 represents the mask for UWB vehicular radar systems. -4 UWB EIRP Emission Level (dbm/mhz) data1 data2 data Frequency (GHz) Fig. 6. FCC spectral mask for UWB systems. Mask 1 represents the mask for UWB low frequency imaging. Mask 2 represents the mask for UWB mid frequency imaging. Mask 3 represents the mask for UWB high frequency imaging. 16

32 TABLE 1 SUMMARY OF FCC RESTRICTIONS ON UWB OPERATION Application Communication and measurement systems (Sensors) Frequency Band for Operation GHz (different emission limits for indoor and outdoor systems) User Restriction None Vehicular radar for collision avoidance, and suspension system control GHz None Ground-penetrating radar to visualize or spot buried objects Wall imaging systems to visualize objects contained in walls Through-wall imaging systems to spot location or movement of objects placed on the other side of a barrier GHz and below 96 MHz GHz and below 96 MHz GHz and below 96 MHz Law enforcement, fire and rescue, research institutions, mining and constructions Law enforcement, fire and rescue, research institutions, mining and constructions Law enforcement, fire and rescue Medical systems for visualizing in the interior of people and animals Surveillance systems for intrusion detection GHz Medical personnel GHz Law enforcement, fire and rescue, public utilities, and industry 17

33 frequency and avoid mutual interference, resulting in better spectral efficiency than any other scheme. Multiple users can be supported in the same spectrum by allocating each user a group of sub-carriers. OFDM-UWB is mainly intended for data transfer in the physical layer for high bit-rate, short-range (1 2m) communications networks. The OFDM-UWB transmitter splits orthogonal sub-carriers into a train of short pulses, sends the pulses over a channel, and reassembles them at the receiver to recover each subcarrier separately [1], [12]. F. OFDM System Model 1) Transmitter: A number of orthogonal sub-carriers constitute an OFDM carrier signal. Each sub-carrier carries a bit of baseband data, and these data are independently modulated using a commonly used modulating method, such as Quadrature Amplitude Modulation (QAM) or Phase-Shift Keying (PSK). An ideal model of an OFDM transmitter is shown in Fig. 7. At the transmitter, x, x 1 x n are the symbols modulated by using a constellation map. A constellation map gives the relationship between the discrete inputs and the real space to which these discrete data will be converted during modulation. These input symbols are used to compute an Inverse Fast Fourier Transform (IFFT) and results in a set of OFDM symbols. The real and imaginary components are then separated. These separated components are converted to the analog domain using digital-to-analog converters (DACs). A different analog carrier signal having a frequency, f c, is used at the modulation. This carrier wave s cosine and sine waves are modulated by the analog 18

34 signal at the baseband from the DACs. The final transmission signal s(t) is obtained by adding these two analog signals. x x 1... Inverse FFT Real Imag. DAC f c 9 deg. s(t) x n DAC Fig. 7. An ideal model of an OFDM transmitter. 2) Receiver: An ideal model of an OFDM receiver is shown in Fig. 8. The receiver picks up the transmitted signal r(t). The cosine and sine waves at the carrier frequency are then quadrature mixed with the received signal, and the result is the baseband signal. A low-pass filter is used to filter out the components above 2fc that are created during the mixture. The resultant baseband analog signals are converted to a digital form by sampling and digitizing using analog-to-digital converters (ADCs). The Fast Fourier Transform (FFT) is computed to convert the OFDM symbols to data symbols. The result symbols at the OFDM receiver are y, y 1 y n (Fig. 8). G. Mathematical Analysis of OFDM In an OFDM system, a number of sub-carriers are used for symbol transmission. If k represents the number of sub-carriers used, and M represents alternative symbols 19

35 used for modulation of each sub-carrier, then the total OFDM word consists of M K symbols. The OFDM signal for these symbols can be expressed as r(t) 9 deg. f c ADC Real Imag FFT y y 1... ADC y n Fig. 8. An ideal model of an OFDM receiver. V ( t) N = 1 I k= k e i2πkt / T, t T. (2) In (2), { I k } are the data symbols, N is the number of sub-carriers, and T is the OFDM symbol time. The sub-carrier spacing of T 1 Hz maintains the orthogonality. This property is expressed as T T i2πk / * 2 / 1 1t T i πk2t T i2π ( k2 k ) t / T ( e ) ( e ) dt = ( e ) 1 1 T In (3), * denotes the complex conjugate operator. T dt 1, =, k k 1 1 = k k 2 2. (3) To avoid inter-symbol interference, a guard interval, t, is introduced in multi-path fading channels. The guard period is inserted before the transmission of the OFDM symbols, and a cyclic prefix is transmitted during this guard period interval. The T g 2

36 cyclic prefix transmitted is the last portion of the OFDM symbol having a length equal to the guard period. The OFDM signal with cyclic prefix can be represented as V ( t) N = 1 I k= k e i2πkt / T, T g t T. (4) This OFDM signal can be either real or complex-valued. Usually, the baseband data is transmitted in real-valued equivalent signals; however, for wireless applications, the signal is transmitted typically in complex-valued. A carrier frequency, fc, is used to up-convert the baseband signal and is represented as [1], [12] For wireless applications, V(t) becomes V ( t) = { e } i 2π f c V ( t) t 1 s( t) = R. (5) 2 [ 2 ( f + k / T) t+ arg( )] N 1 I k k cos c I k = π. (6) H. Overview of an UWB Model The purpose of this section is to illustrate the details of a multi-band orthogonal frequency division multiplexing (MB-OFDM) UWB system based on a draft proposed to the IEEE a standards group in September 23. This system also is the basis for wireless USB, which is the UWB common platform of the WiMedia Alliance. The essential technology has not changed in the later proposals. The proposal supports seven data rates in the range of Mbps. The highest mandatory rate is 2 Mbps, using a frequency hopping (multi-band) scheme for OFDM signals transmission. The functioning of a UWB system is explained as follows with the help of the top-level simulation model in Fig

37 1) Bernoulli Binary: The Bernoulli Binary Generator block generates random binary numbers using Bernoulli distribution function. The Bernoulli distribution generates zero with probability p and one with probability 1-p. The mean value of the Bernoulli distribution is 1-p and variance is p(1-p) [13]. 2) Rate Encoder: The rate encoder changes the rate of the bit stream using a convolutional encoder. Rate encoding is an important process in many digital communication systems involving Forward Error Correction (FEC) coding. Time diversity is provided by the encoded symbols to prevent localized alteration or burst errors in the symbols [13]. 3) Interleaver: The interleaver adds redundant bits to the rate encoded data to protect the data transmission from burst errors. Interleaving enables the receiver to successfully retrieve data from the received transmission if burst errors corrupt a large number of bits in a row [12]. 4) QPSK Modulator: The Quadrature Phase-Shift Keying (QPSK) Modulator modulates the input signal from the interleaver using the quaternary phase-shift keying method. The QPSK modulator accepts binary data and converts the binary to a complex form according to the QPSK constellation map [13]. 5) OFDM Transmitter: The OFDM transmitter converts a set of QPSK symbols from the QPSK modulator into OFDM symbols (165 samples each). These OFDM symbols are 22

38 then sent to the transmitter s front-end for transmission. Various additional bits are added in this section for timing synchronization and to prevent inter-frame interference [12]. Interleaver Rate Encoder Binary Data QPSK Modulator OFDM Transmitter Frequency Hopping and Filtering UWB Channel QPSK Demodulator OFDM Receiver Frequency Dehopping and Filtering Synchronization Deinterleaver Viterbi Decoder Data Fig. 9. Top-level model of a typical UWB system. 6) Frequency Hopping and Filtering: Frequency hopping is a modulation technique used for the output from the transmitter in spread spectrum signal transmission. During a transmission, the signal is repeatedly switched between frequencies. This frequency switching is implemented to minimize the unauthorized interception or jamming of the transmitted signal and to prevent data loss due to frequency selective fading [14]. In this UWB scheme, the signal is hopped between three frequencies. 23

39 7) UWB Channel: The UWB channel model simulates an indoor UWB channel programmed by Intel and used by the IEEE a group. The channel model simulates the channel response as the time increases. The current proposal defines four possible UWB channel models [15], which are discussed further in Chapter 4. 8) Frequency Dehopping and Filtering: Frequency dehopping takes inputs, which are being switched between three different frequency spectrums of the UWB channel, and retrieves the OFDM signal. During this process, the offset frequency due to the UWB channel is also calculated and synchronized [14]. 9) OFDM Receiver: The OFDM receiver decodes the OFDM signal and retrieves the data from the received signal. All of the redundant bits are removed and necessary compensations applied to recover the original signal. 1) QPSK Demodulator: The QPSK demodulator demodulates the signal using the quaternary phase shift keying method. QPSK demodulator accepts the complex-valued QPSK symbols and converts these symbols into binary data according to the QPSK constellation map [12]. 11) Synchronization: The synchronization block deals with the timing synchronization of the signal. This block ensures the correct identification of the data packets. The synchronization block makes sure that the start of the data packets is correctly identified by making use of the Physical Layer Convergence Protocol (PLCP) 24

40 preambles. PLCP preambles are redundant data bits that are added in front of actual data bits and are defined in the UWB standard [16]. 12) De-interleaver: The de-interleaver removes the redundant bits added during interleaving. Interleaving enables the receiver to successfully retrieve data from the received transmission if burst errors corrupt a large number of bits in a row [12]. 13) Viterbi decoder: The Viterbi decoder uses the Viterbi algorithm for decoding a bit stream that has been rate encoded using forward error correction based on a convolution code. This thesis focuses on improving the signal quality of UWB signals using equalization to improve cross-correlation PAPR. The improvement in cross-correlation PAPR will enable better signal quality, which can lead to better packet detection due to better signal peaks and less noise levels. 25

41 III. UWB TIMING SYNCHRONIZATION A. Synchronization In any wireless system, one of the important steps to ensure that the received signal is properly synchronized so that the output data is meaningful. Since the UWB systems usually deal with very high data throughput, a rapid and accurate synchronization of the incoming data at the UWB receiver is very critical. A missed synchronization can result in losing several packets of data and, ultimately, in system failure. There are two main types of synchronizations in an OFDM UWB system that are necessary for a successful communication between transmitter and receiver: frequency and timing synchronizations [12]. Discussions of these two kinds of synchronizations are included in the following section. 1) Frequency Synchronization: Frequency synchronization is defined as the process of correcting the differences between the carrier frequencies in the receiver and the transmitter. This frequency offset can occur for many reasons, such as instability in the receiver or transmitter oscillators. There is also a Doppler effect if the systems are in motion, and the frequencies can change when the transmitted signal reaches the receiver [12]. Due to errors in frequency synchronization, two phenomena can occur. First, there will be a reduction in signal amplitude, since the signal will not be sampled at peak points. Instead, the signals will be sampled elsewhere, resulting in a degradation of signal 26

42 strength. Second, an offset in frequency results in the sampling of two adjacent carriers, which can lead to inter-carrier interference (ICI) [12]. For very small frequency errors, the degradation in SNR of the signal is given by SNR loss 1 2 Es = ( π Tf ) db. (7) 3ln1 N o In (7), f is the frequency error, T is the sampling period, s and E is the symbol power, N o is the noise power [17]. There are many algorithms to correct the frequency error. A detailed discussion of those algorithms is beyond the scope of this thesis, and an interested reader can refer to [12] for more details. 2) Timing Synchronization: Timing synchronization is defined as the process of detecting the data packet in a received signal so that receiver can sample the signal properly to retrieve a meaningful data. The timing synchronization can be considered a two-step process. The first step is to successfully detect the presence of data packets when they arrive at the receiver. The second step is to align the receiver to start reading at the instant when the presence of data packet is detected. These two processes are termed as data packet detection and symbol synchronization [12]. This thesis deals with ways to improve received signal quality, which results in better packet detection in OFDM UWB systems. B. Packet Detection Packet detection is an important step to be performed during the timing synchronization, since the rest of the synchronization process is dependent on good 27

43 packet detection. The UWB uses the IEEE Medium Access Control (MAC) protocol, and the UWB receiver does not have a prior knowledge about the time of arrival of a packet. Any algorithm to be used in UWB receivers needs to conduct the packet detection without any prior knowledge, and that makes the synchronization of UWB packets very difficult. The application of the packet detection in the timing synchronization is illustrated in Fig. 1. There are two types of UWB communications: high data rate and low data rate. In high data rate UWB communication, two of the major task groups are IEEE a and wireless Universal Serial Bus (USB). In any type of UWB communication, synchronization between transmitter and receiver is important. As mentioned in the previous section, there are two types of synchronizations: frequency synchronization and timing synchronization. Packet detection is one of the most important steps in timing synchronization. Improving the received signal quality, which can result in better packet detection, is the focus of this thesis. The packet detection can be modeled as a binary hypothesis test, and the two outcomes of the test are packet is present and packet is absent. Assuming B is the parameter of interest, v is the decision variable, and Th is the predefined threshold, then the actual test can be represented as follows. B : v < Th Packet not present B 1 : v Th Packet present 28

44 UWB Low Data Rate IEEE a High Data Rate IEEE a IEEE a Wireless USB Synchronization / Acquisition Transmitter / Receiver Frequency Synchronization Timing Synchronization Packet Detection (MMSE/LSE Estimation and Equalization) Signal Quality Symbol Synchronization Fig. 1. Application of the packet detection in the timing synchronization. Two probabilities are used to evaluate the performance of any packet detection algorithm: probability of detection, P D, and probability of false alarm, P FA. The probability of detection is defined as the probability that the packet detection algorithm will correctly identify the presence of a data packet, while the probability of false alarm is defined as the probability that the algorithm will misinterpret the decision variable and falsely identify the presence of a data packet when there are no packets present [12]. 29

45 Ideally, the probability of detection should be high, and the probability of false alarm should be low. C. Packet Detection Algorithms The commonly used packet detection algorithms are described as follows. 1) Received Signal Energy Detection: The simplest algorithm used for packet detection is the received signal energy detection approach. The received energy of the signal is continuously monitored for changes in energy. Until the arrival of the packet, only the noise is present and the energy is low and remains constant. When the packet arrives, there is a sudden change in energy of the signal that is taken as the criteria for detecting the data packet [18]. In order to avoid false detection due to high magnitude noises, a window filter known as a sliding window is introduced. The energy that falls inside this window is summed. The summation increases gradually, reaches a peak, and then decreases. A packet detection is confirmed when the decision variable goes over a predefined threshold. In this case, the decision variable is the energy summation of the signal. Assuming the length of the window is L, and the received signal is r n, the accumulated energy in the window is given by [12] L 1 2 n = r n k k= E. (8) The concept of received signal energy detection is illustrated in Fig. 11. A packet having 1 bits is detected using a window with a length of 1 bits. The peak energy is reached at 5, so that the threshold can be set at 25. 2) Double Sliding Window Packet Detection: The double sliding window packet detection algorithm is similar to the received signal energy detection algorithm discussed 3

46 earlier. Instead of one window, there are two sliding windows. In this case, the decision variable is the ratio of the total energy accumulated in the two windows. Packet 5 E n Energy 25 Th 1 2 Sample Number Fig. 11. Packet detection using received signal energy detection method. A packet having 1 bits is detected using a window having a 1 bit-length. Threshold is 25. Peak energy is 5. The concept of received signal energy detection is illustrated in Fig. 12. Initially, when only the noise is present, the decision variable is flat. When the packet arrives, the window A starts accumulating energy. The decision variable slowly starts rising until the packet reaches the window B. This is the peak of the decision variable. When window B starts accumulating energy, the decision variable slowly starts falling. A packet detection is confirmed when the decision variable goes over a predefined threshold. Assuming the length of the window A is M, the length of the window B is L, and the receiving signal is r n, the decision variable is given by [12] a n E n =. (9) bn 31