Channel Aware •Strategies in Wireless Communications by Maged Elk'ashlan •B'.Sc, Arab Academy for-Science.and'Technology, 1997 M . S c , Arab Academy for- Science and Technology,. 2000 A THESIS S U B M I T T E D IN P A R T I A L F U L F I L M E N T O F T H E R E Q U I R E M E N T S : F O R T H E D E G R E E OF DOCTOR OF PHILOSOPHY in T H E F A C U L T Y OF G R A D U A T E STUDIES (Electrical and Computer Engineering) T H E U N I V E R S I T Y O F BRITISH C O L U M B I A August 2006 © Maged Elkashlan, 2006 11 Abstract Channel-aware resource allocation is widely considered to be crucial for realizing high data rates in wireless networks. This thesis considers the resource allocation problem in a multiuser wireless network using link adaptation. Link Adaptation (LA), which loosely refers to changing transmission parameters over a link, such as modulation, coding rate, power, etc., in response to changing channel conditions is considered to be a powerful means of achieving higher efficiency or throughput in wireless networks. The adaptation of the transmission parameters is performed according to the predicted future quality of the channel, also referred to as the channel state (CS). The objectives of this thesis are threefold: devise new methods for the improvement of multicarrier code division multiple access ( M C - C D M A ) as an effective signaling scheme in correlated fading channels, the development of new channel aware algorithms with adaptive subchannel allocation, and the investigation of the statistics of general order selection in correlated Nakagami fading channels. A novel adaptive subcarrier allocation algorithm is developed for M C - C D M A to improve the overall bit error rate (BER) performance. The proposed method is suitable for use in correlated fading channels. This algorithm assigns users to subcarrier groups that provide favorable fading characteristics, while simultaneously reducing the amount Abstract 111 of interference caused to other users. T h i s method examines the effect of equalizing the interference i n each subcarrier group while maintaining a reduced correlation among the subcarrier fading processes for a group. Consequently, subcarriers are separated into non-contiguous groups to maximize frequency diversity a n d m i n i m i z e multiple access interference (MAI). Previously proposed adaptive subcarrier allocation algorithms, are considered to be greedy algorithms. T h e y are greedy i n the sense that they consider every reservation request individually, a n d make the choice that looks best at the moment. Adaptive allocation using the simple greedy method gives an o p t i m a l solution for a single user system. A d a p t i v e b i t allocation using the greedy method cannot give an o p t i m a l solution for multiuser cases. It is possible that the sub-channels w i t h the largest channel gain for one user are also the largest for another user. W e present i n this thesis a new class of dynamic resource allocation schemes that are based on a l l the users' subband gains. Due to the time-varying nature of the wireless channel, dynamic resource allocation makes full use of multiuser diversity to achieve higher performance. In this thesis, we formulate the multiuser subcarrier allocation problem and propose an iterative algorithm to perform the subcarrier allocation. T h i s method involves the ordering of subband gains. T h u s , the analytical foundation is the theory of order statistics. T h e users are recursively assigned to that subcarrier which provides the highest possible signal-to-noise ratio ( S N R ) . W h i l e ensuring that no more t h a n one user is assigned to the same subband. O u r objective is to find low-complexity schemes which can improve system capacity and throughput, a n d simultaneously m i n i m i z i n g the M A I . Finally, the cumulative distribution Abstract iv function (cdf) (and hence outage probability) of the r-th order branch S N R in correlated Nakagami—m fading is studied. The accuracy of a simple exchangeable approximation to reduce the computational load is examined. The effects of correlation on the error performance of channel aware systems that involve the ordering of channel gains in a correlated Nakagami fading environment is investigated. A useful analytical formula for the B E R of the r-th order statistic of a set of arbitrarily correlated and hot necessarily exchangeable diversity branch gains is described and shown to be applicable in the performance analysis of various diversity systems. V Contents Abstract ii Contents v List of Figures ix List of A b b r e v i a t i o n s xii Acknowledgements xvi Dedication xviii 1 Introduction 1 1.1 Mobile Communications Systems: Past, Present, and Future 1 1.2 From Second- to Third-Generation Multiple Access Schemes 3 1.3 From Third- to Fourth-Generation Multiple Access Schemes 5 1.4 Motivation 8 1.5 Link Adaptation Fundamentals 10 1.6 Subchannel and Power Allocation 12 1.7 This Thesis 17 Contents 2 3 4 vi 1.7.1 Objectives 17 1.7.2 Contributions 17 1.7.3 Road Map 20 M a t c h e d M C - C D M A in Slowly Fading Dispersive Channels 22 2.1 Introduction 22 2.2 System Model 23 2.3 Numerical Results 25 F r e q u e n c y - H o p p i n g M u l t i c a r r i e r C D M A in R a y l e i g h F a d i n g 31 3.1 Introduction 31 3.2 F H - M C - C D M A System Model 33 3.3 Numerical Results and Discussions 38 F r e q u e n c y - H o p p i n g M u l t i c a r r i e r C D M A in Correlated R a y l e i g h F a d i n g 47 4.1 Introduction 47 4.2 System Model 48 4.2.1 FH-MC-CDMA 48 4.2.2 Channel Model 50 4.3 4.4 B E R Analysis of F H - M C - C D M A 51 4.3.1 Uniform F H - M C - C D M A 52 4.3.2 Random F H - M C - C D M A 53 4.3.3 Optimum F H - M C - C D M A 53 Numerical Results 54 Contents 5 6 A C h a n n e l A w a r e Frequency H o p p i n g M u l t i p l e Access Scheme . . . 63 5.1 Introduction 63 5.2 System Model 64 5.3 C A F H Scheme 65 5.4 B E R analysis 66 5.4.1 B E R analysis of C A F H for r = l 66 5.4.2 Numerical Results of C A F H for r=l 70 5.4.3 B E R Analysis of C A F H for r >1 72 5.4.4 Numerical Results of C A F H for r >1 74 O r d e r Statistics for Correlated Fading Channels 88 6.1 Introduction 88 6.2 Preliminaries 89 6.3 cdf of 7 91 6.4 7 vii r : n 6.3.1 Correlated Fading 91 6.3.2 Independent Fading 93 Numerical results 93 6.4.1 Constant Correlation Model 94 6.4.2 Linear Model 96 E r r o r A n a l y s i s of O r d e r Selection in Correlated F a d i n g C h a n n e l s . . 102 7.1 Introduction 102 7.2 Channel and System Model 103 Contents 8 viii 7.3 Error Performance for Order Selection B P S K 104 7.4 Numerical Results and Discussion 108 7.4.1 B E R evaluation for Order Selection B P S K 109 7.4.2 Application: B E R evaluation for C A F H 110 Conclusions a n d F u t u r e Research 118 8.1 Summary 118 8.2 Future Work 122 Bibliography 126 ix List of Figures 2.1 (a) Transmitter model (b) Receiver model 2.2 Matched M C - C D M A subcarrier pattern, channel transfer function T F , and integral \&(-) with : = 2/BT- The frequency corresponding to the I 27 T H subcarrier is f + FiB 28 2.3 B E R as a function of number of users for S N R = 9 dB 29 2.4 B E R as a function of S N R for M = 64 users 30 3.1 (a) Transmitter model (b) Receiver model 40 3.2 F H - M C - C D M A pattern 41 3.3 Theoretical and simulation B E R comparison for N = M 42 3.4 Theoretical and simulation B E R comparison for N < M 43 3.5 B E R comparison between random and coordinated F H - M C - C D M A for M c = 32 users 3.6 ch ' 44 B E R comparison between random and coordinated F H - M C - C D M A for M — 64 users 45 3.7 B E R as a function of S N R for M = 64 users 46 4.1 (a) Transmitter model; (b) Receiver model 57 List of Figures x 4.2 Frequency groups when N 4.3 B E R versus S N R for M = 32 MS's, and L = 32 59 4.4 B E R versus S N R for M = 32 MS's, and L = 8 60 4.5 B E R versus TV for L = 16, and S N R = 10 dB 61 4.6 B E R versus M for L = 16, and S N R = 10 dB 5.1 A C A F H flow chart describing the proposed subband assignment scheme. 77 5.2 B E R of conventional F H and C A F H with B P S K modulation, L = 32. . . 78 5.3 B E R of conventional F H and C A F H with N C F S K modulation, L = 32. 79 5.4 B E R of C A F H with B P S K for r = 1 and 2, L = 32. , 80 5.5 B E R of C A F H with N C F S K for r = 1 and 2, L = 32 81 5.6 B E R of conventional F H and C A F H with B P S K modulation. 70 = 2 dB Gj = N V j with /V — L = 4, and F = g = s/N. 58 . and L = 48 5.7 62 82 B E R of conventional F H and C A F H with B P S K modulation. 7 o = 8 dB and L = 48 83 5.8 B E R vs. average SNR, 70, of C A F H with B P S K for r = 1 and 2, L = 48. 84 5.9 B E R vs. average SNR, , of C A F H with N C F S K for r = 1 and 2, L = 48. 85 7 o 5.10 B E R vs. the number of rounds, r, of C A F H with B P S K . 7 o = 8 dB and L = 48 86 5.11 B E R versus number of MS's of random allocation, C A F H with B P S K for r = 1 and 2, and exhaustive search method. 70 = 4 dB and L — 8 6.1 Plot of cdf of 7i:3,72:3 and simulation (solid) and 73:3 with m — 1 and 7 87 = 1: theoretical (dotted) 98 List of Figures 6.2 Plot of cdf of 7i;3,72:3 and 7 3:3 xi with m = 1 and 7 = 1: independent (dotted) and correlated (solid) fading channels 99 6.3 Plot of cdf of 7 1:3 with 7 = 1 for different values of m 100 6.4 Plot of cdf of 7 r;5 with m = 1 and 7 = 1 for a linear model (solid) and its exchangeable approximation (dotted) 101 7.1 B E R vs. average SNR for B = 0.9 112 7.2 B E R vs. average SNR for B = 0.5 113 7.3 B E R vs. average S N R for B = 0.25. 114 7.4 C A F H B E R vs. number of mobile stations, for 7 = 2 and 8 dB, and n n n B = 0.9 115 n 7.5 C A F H B E R vs. number of MS's, for 7 = 2 and 8 dB, and B = 0.25. . . 116 7.6 C A F H B E R vs. number of MS's, for 7 - 2 and 10 dB, and B = 0.33. 117 n n . List of Abbreviations 2G 2nd Generation 3G 3rd Generation 4G 4th Generation ACE Active Constellation Extension AMC Adaptive Modulation and Coding AWGN Additive White Gaussian Noise BER Bit Error Rate BPSK Binary Phase-Shift Keying BS Base Station CAFH Channel Aware Frequency Hoppin cdf Cumulative Distribution Function CDMA Code Division Multiple Access CF Characteristic Function CP Cycle Prefix CS Channel Sate CSI Channel State Information CQI Channel Quality Information List of Abbreviations DAB Digital Audio Broadcasting D-AMPS Digital-Advanced Mobile Phone Service DFT Discrete Fourier Transform DMT Discrete Multitone DS Direct Sequence DS-CDMA Direct Sequence C D M A DSL Digital Subscriber Line DVB Digital Video Broadcasting EDGE Enhanced Data Rates for G S M Evolution EGC Equal Gain Combining FDMA Frequency Division Multiple Access FH Frequency Hopping FH-MC-CDMA Frequency Hopping M C - C D M A GPRS General Packet Radio System GSM Global System for Mobile Communications HDR High Data Rate H-S/MRC hybrid selection/maximal-ratio combining ICI Interchannel Interference IP Internet protocol ISI Intersymbol Interference ITU International Telecommunications Union LA Link Adaptation xm List of Abbreviations LAN Local Area Network MA Margin Adaptive MAI Multiple Access Interference MC Multicarrier MC-CDMA Multicarrier Code Division Multiple Access MC-SS Multicarrier Spread Spectrum MIMO ' Multiple-Input Multiple-Output MISO multi-input single-output MMSE Minimum Mean Square Error Estimation MRC Maximal Ratio Combining MS Mobile Station MU-OFDM Multiuser O F D M NCFSK Non-Coherent Frequency-Shift Keying OFDM Orthogonal Frequency Division Multiplexin PAPR Peak-to-Average Power Ratio PDC Personal Digital Cellular pdf Probability Density Function PER Packet Error Rate PF Proportional Fair PG Processing Gain PSAM Pilot Symbol Aided Modulation PSD Power Spectral Density xiv List of Abbreviations PTS Partial Transmit Sequence QoS Quality of Service RA Rate Adaptive RF Radio Frequency rv Random Variable SC Selection Combining SFH/CDMA Slow F H / C D M A SINR Signal to Interference plus Noise Ratio SISO Single-Input Single-Output SLM Selected Mapping SNR Signal-to-Noise Ratio SNRr Received Signal-to-Noise Ratio TDL Tapped Delay Line TDMA Time Division Multiple Access TI Tone Injection TR Tone Reservation UMTS Universal Mobile Telecommunications System WCDMA Wideband Code-Division Multiple Access WLAN Wireless Local Area Network WLL Wireless Local Loop xv XVI Acknowledgements First and foremost, I am obediently thankful to God, praise and glory be to him, for his countless blessings and mercy. Without his guidance and mercy I can do nothing. I would like to express my heartfelt gratitude to my supervisor, Dr. Cyril Leung, for his guidance, technical advice, invaluable feedback, understanding, generous support, and friendship. I have passed through many personal and difficult circumstances during the period of my Ph.D. program. Without Dr. Leung's genuine kindness, I do not think you would be reading these words today. I am very grateful to my co-supervisor, Dr. Robert Schober, for his valuable suggestions, and technical assistance. Although he is very busy, I always found him easy to approach and available for help. I would like to express my appreciation to my thesis committee for their valuable comments which enhanced the presentation of this thesis. I have been blessed to come in contact with many wonderful people throughout my study period at U B C . This page is definitely not enough to mention all of them, but, the least I can say to them is: thank you for making my stay at U B C such a pleasant experience. Words cannot describe my feelings toward my parents. I can only pray to God to Acknowledgements xvii reward them for their constant support, persistent encouragement, and most of a l l for their unwavering love. xvm Dedication To m y beloved parents. 1 Chapter 1 Introduction 1.1 Mobile Communications Systems: Past, Present, and Future The common feature of next generation wireless technologies will be the convergence of multimedia services such a speech, audio, video, image, and data. This implies that a future wireless terminal, by guaranteeing high-speed data, will be able to connect to different networks in order to support various services: switched traffic, IP data packets and broadband streaming services such as video. The development of wireless terminals with generic protocols and multiple-physical layers or software-defined radio interfaces is expected to allow users to seamlessly switch between networks based on different standards. The rapid increase in the number of wireless mobile terminal subscribers, which currently exceeds 1 billion users, highlights the importance of wireless communications in the new millennium. The increase has been sustained through the continuous evolution of standards and products. The adaptation of wireless technologies to the users' rapidly changing demands has been a hallmark of this rapid growth. It is expected that a variety Chapter 1. Introduction 2 of standards and systems will continue to exist. This plethora of wireless communication systems is not limited to cellular mobile telecommunication systems such as Global System for Mobile communications (GSM), IS-95, Digital Advanced Mobile Phone Service (D-AMPS), Personal Digital Cellular (PDC), Universal Mobile Telecommunications System (UMTS) or cdma2000, but also includes wireless local area networks (WLANs), e.g., H I P E R L A N / 2 , I E E E 8 0 2 . i l and Bluetooth, and wireless local loops (WLL), e.g., H I P E R M A N , H I P E R A C C E S S , and I E E E 802.16 as well as broadcast systems such as digital audio broadcasting (DAB) and digital video broadcasting (DVB). Currently, the research community is focusing its activity on beyond 3G, i.e. J^th generation (4G) systems, with more ambitious technological challenges. The primary goal of next-generation wireless systems (4G) will not only be the introduction of new technologies to cover the need for higher data rates and new services, but also the integration of existing technologies in a common platform. Hence, the selection of a generic air-interface for future generation wireless systems will be of great importance. Although the exact requirements for 4G have not yet been defined, its new air interface should fulfill at least the following requirements [1]: • generic architecture, enabling the integration of existing technologies, • high spectral efficiency, offering higher data rates in a given scarce spectrum, • high scalability with different cell configurations (hot spot, ad hoc), hence better coverage, • high adaptability and reconfigurability, supporting different standards and tech- 3 Chapter 1. Introduction nologies, • low cost, enabling a rapid market introduction, and • future proof, opening the door for new technologies. 1.2 From Second- to Third-Generation Multiple Access Schemes Second Generation (2G) wireless systems are mainly characterized by the transition from analog towards a fully digital technology and comprise the G S M , IS-95, P D C and DA M P S standards. Work on the pan-European digital cellular standard G S M started in 1982 [2, 3], and now accounts for two-thirds of the world mobile market. In 1989, the technical specifications of G S M were approved by the European Telecommunication Standard Institute (ETSI), and commercialization began in 1993. Although G S M is optimized for circuit-switched services such as voice, it offers low-rate data services up to 14.4 kbit/s. High speed data services up to 115.2 kbit/s are possible with the enhancement of the G S M standard General Packet Radio Service (GPRS) by using a larger number of time slots. G P R S uses the same modulation, frequency band, and frame structure as G S M . However, the Enhanced Data Rates for Global Evolution ( E D G E ) [4] system which further improves the data rate up to 384 kbit/s introduces a new modulation scheme. Parallel to G S M , the U.S. IS-95 standard [5] (recently renamed cdmaOne) was approved by the Telecommunication Industry Association (TIA) in 1993, and commercial- ization started in 1995. As in G S M , the first version of this standard (IS-95A) offered data services up to 14.4 kbit/s. In its second version, IS-95B, up to 64 kbit/s data ser- 4 Chapter 1. Introduction vices are possible. During the same time period, two other 2G mobile radio systems were introduced: D-AMPS/IS-136, called T D M A in the U S A and P D C in Japan [6]. Demands for more capacity for mobile receivers, new multimedia services, new frequencies, and new technologies are behind the introduction of 3G systems. A unique international standard was targeted: Universal/International Mobile Telecommunication System (UMTS-IMT-2000) which would enable a new generation of mobile personal communications services world-wide. The objectives of the 3G standards, namely U M T S [7] and cdma2000 [8] went far beyond those of 2G systems, especially with respect to [9]: • the wide range of multimedia services (speech, audio, image, video, data) and bit rates (up to 2 Mbit/s for indoor and hot spot applications), • the high quality of service requirements (better speech/image quality, lower bit error rate (BER), higher number of active users), • operation in mixed cell scenarios (macro, micro, pico), • operation in different environments (indoor/outdoor, business/domestic, cellular/cordless), • and finally flexibility in frequency (variable bandwidth), data rate (variable), and radio resource management (variable power/channel allocation). The commonly used multiple access schemes for 2G and 3G wireless mobile communication systems are based on either Time Division Division Multiple Multiple Access ( T D M A ) , Code Access ( C D M A ) or a combinations of T D M A / C D M A in conjunction with an additional Frequency Division Multiple Access ( F D M A ) component: Chapter 1. Introduction 5 • The G S M standard, employed in the 900 M H z and 1800 M H z bands, first divides the allocated bandwidth into 200 kHz F D M A sub-channels. Then, in each subchannel, up to 8 users share the 8 time slots in a T D M A manner [3]. • In the IS-95 standard, up to 64 users share the 1.25 M H z channel by C D M A [5]. The system is used in the 850 MHz and 1900 M H z bands. • The aim of D - A M P S ( T D M A IS-136) is to coexist with analog A M P S , and the 30kHz channel of A M P S is divided into three channels, allowing three users to share a single radio channel by allocating unique time slots to each user [10]. • The recent International Telecommunications Union (ITU) adopted standards for 3G ( U M T S and cdma2000) are both based on C D M A [7, 8]. For U M T S , the C D M A F D D mode, which is known as wideband C D M A , employs separate 5 M H z channels for both the uplink and downlink directions. Within the 5 M H z bandwidth, each user is separated by a specific code, resulting in an end-user data rate of up to 2 Mbit/s. 1.3 From Third- to Fourth-Generation Multiple Access Schemes Besides offering new services and applications, the success of 4G wireless systems will strongly depend on the choice of the concept and technology innovations in architecture, spectrum allocation, spectrum utilization, and exploitation [11, 12]. New highperformance physical layer and multiple access technologies are needed to provide high speed data rates with flexible bandwidth allocation. A low-cost generic radio interface, Chapter 1. Introduction 6 operational in in different environments with scalable bandwidth and data rates, is expected to have better acceptance. The technique of spread spectrum may allow the above requirements to be at least partially fulfilled. A multiple access scheme based on direct sequence code division multiple access (DS-CDMA) relies on spreading the data stream using an assigned spreading code for each user in the time domain [13]-[16]. The capability of minimizing multiple access interference (MAI) is given by the cross-correlation properties of the spreading codes. In the case of severe multipath propagation in mobile communications, the capability of distinguishing one component from others in the composite received signal is offered by the autocorrelation properties of the spreading codes [14]. The so-called rake receiver should contain multiple correlators, each matched to a different resolvable path in the received composite signal [13]. Therefore, the performance of a D S - C D M A system will strongly depend on the number of active users, the channel characteristics, and the number of branches employed in the rake. System capacity is typically limited by self-interference and M A I , which results from the imperfect auto- and cross-correlation properties of spreading codes. It will be difficult for a D S - C D M A receiver to make full use of the received signal energy scattered in the time domain and hence to handle full load conditions [13]. The technique of multicarrier transmission has recently been receiving wide spread interest, especially for high data-rate broadcast applications. The history of orthogonal multicarrier transmission dates back to the mid-1960s, when Chang published his paper on the synthesis of band-limited signals for multi-channel transmission [17]-[19]. The Chapter 1. 7 Introduction basic principle is to transmit data simultaneously through a band-limited channel without interference between sub-channels (without interchannel interference, ICI) and without interference between consecutive transmitted symbols (without mtersymbol interference, ISI) in the time domain. A major contribution to multicarrier transmission was made in [20] when Fourier transform for base-band processing was proposed in place of a bank of subcarrier oscillators. To combat ICI and ISI, a guard time was introduced between successive transmitted symbols. The main advantages of multicarrier transmission are its robustness in frequency selective fading channels and, in particular, the reduced signal processing complexity by equalization in the frequency domain. The basic principle of multicarrier modulation is to divide a high-rate data stream into several low-rate sub-streams. These sub-streams are modulated on to different subcarriers [21]-[23]. By using a large number of subcarriers, a high immunity against multipath dispersion can be provided since the useful symbol duration T on each sub-stream will typically be much larger than the channel dispers sion time. Hence, the effects of ISI will be minimized. Since the number of filters and oscillators necessary is considerable for a large number of subcarriers, an efficient digital implementation of a special form of multicarrier modulation, called O F D M , with rectangular pulse-shaping and guard time was proposed in [21]. O F D M can be easily realized by using the fast Fourier transform ( F F T ) . O F D M , having densely spaced subcarriers with overlapping spectra of the modulated signals, obviates the use of steep band-pass filters to detect each subcarrier as in F D M A schemes. Therefore, it offers high spectral efficiency. Chapter 1. Introduction 8 O F D M came into prominence in the 1990's as it was the modulation method chosen for A D S L in the U S A [24], and for the European D A B standard [25]. This success continued with the choice of O F D M for the European D V B - T standard [26] and for the W L A N standards H I P E R L A N / 2 and IEEE802.11a [27, 28] and recently for the interactive terrestrial return channel ( D V B - R C T ) [29]. It is also a candidate for wireless M A N standards, H I P E R M A N and IEEE802.16a [30, 31]. The advantages of multicarrier modulation on the one hand and the flexibility offered by spread spectrum techniques on the other hand have motivated many researchers to investigate the combination of both techniques, known as multicarrier spread spectrum (MC-SS). Such combinations [32]-[38], have resulted in new multiple access schemes called M C - C D M A and M C - D S - C D M A . They offer attractive features such as high flexibility, high spectral efficiency, simple and robust detection techniques, and narrow band interference rejection capacity. M C and MC-SS are today considered promising candidates to meet the requirements of next generation (4G) high-speed wireless multimedia communications systems, where spectral efficiency and flexibility will be the most important criteria for the choice of the air interface. 1.4 Motivation The wireless communications industry is growing rapidly for both fixed and mobile applications. The increasing demand for all types of wireless services (voice, data, and multimedia) is fueling the need for higher capacity and data rates. Although improved compression technologies have reduced the bit rate for voice calls, data traffic will re- Chapter 1. Introduction 9 quire much more bandwidth as new services come online. In this context, emerging technologies that improve wireless systems' spectrum efficiency are becoming a necessity, especially in broadband applications. Some examples include smart antennas, multipleinput multiple-output (MIMO) technology, coded multicarrier modulation, power control, channel allocation, link-level retransmission, and adaptive modulation and coding ( A M C ) techniques [39]. Popularized by cellular wireless standards such as E D G E , A M C techniques that adapt to the time-varying characteristics of wireless channels allow significantly higher data rates, reliability, and spectrum efficiency for future wireless data-centric networks. The set of algorithms and protocols governing A M C is often referred to as link adaptation (LA). While substantial progress has been made in this area to understand the theoretical aspects of time adaptation in L A protocols, more challenges surface as dynamic transmission techniques must take into account the additional signaling dimensions explored in future broadband wireless networks. More specifically, the growing popularity of both M I M O and multicarrier (MC) systems that take channel aware algorithms into account create the need for L A solutions that integrate temporal, spatial, and spectral components together. Currently a few existing and emerging wireless systems such as IS-95B, W C D M A , CDMA2000, H i p e r L A N / 2 and (E)GPRS of G S M have already introduced some kind of L A [40]. While it is desirable to adapt the transmission parameters according to the channel state information (CSI) to capture even small-scale variations, there are Chapter 1. Introduction 10 practical limitations to channel state prediction and link adaptation. Frequent adaptation increases the number of mode-change messages transmitted over the channel, consuming bandwidth, and time resources [41]. Moreover, predicting the future channel quality may also consume significant amount of system resources (e.g., time, bandwidth and power), since it may involve transmission of training-sequences, pilot tones, or feedback messages carrying the CSI. Naturally, the additional cost and complication of predicting CS and L A increase with the desired prediction accuracy. Common channel estimation methods documented in the literature are pilot symbol aided modulation (PSAM) and minimum mean square error (MMSE) estimation. A classical paper on pilot symbol aided modulation for Rayleigh fading was presented in [42] where the optimum Wiener interpolation filter was derived to minimize the variance of the estimation error and system performance for B P S K , Q P S K and 16-QAM was studied. Because the optimum Wiener filter requires a priori information of the channel and is computationally complex, several suboptimal interpolation filters have been proposed among which the sine interpolator is most widely used due to its simple implementation and close-to-optimum performance [43]. The bit error rate (BER) of pilot symbol aided M Q A M in Rayleigh fading using the sine interpolator has been analyzed in [44]. 1.5 Link Adaptation Fundamentals The basic idea behind L A is to adapt the transmission parameters to take advantage of the prevailing channel conditions. The fundamental parameters to be adapted include modulation and coding levels, but other quantities can also be adjusted such as power Chapter 1. Introduction 11 level (as in power control and power allocation), channel allocation, spreading factor, and signaling bandwidth. L A is now widely recognized as a key solution to increasing the spectral efficiency of wireless systems. L A techniques are incorporated in current proposals for third-generation C D M A wireless packet data services, such as cdma2000 and wideband CDMA ( W C D M A ) , G P R S and GPRS-136 [39]. In practical L A implementations, the values for the transmission parameters are quantized and grouped into what we refer to as a set of modes. A n example for a mode might be a pair of modulation level and coding rate. Since each mode has a different data rate (expressed in information bits per second) and robustness level (minimum S N R needed to activate the mode), they are optimal for use in different channel/link quality regions. When using packet-switching technology and multiple modulation and coding levels, the E D G E system employs a link-adaptation technique to adapt packet transmission to one of six coding levels [45] (a recent proposal has nine modulation and coding levels [46]), where the highest data rate can exceed 550 kb/s. When the channel condition is poor, a low modulation level (i.e., few information bits per symbol) and/or heavy coding should be used for packet transmission to enable correct signal detection. On the other hand, if the channel is in a good state, a high modulation level and/or light coding can be used to increase data rate. Due to the unreliable nature of radio links, the provision of quality of service (QoS) such as packet error rate (PER) in wireless networks is challenging. For real-time services such as Internet protocol (IP) voice, music and video, stringent delay requirements severely limit or even preclude the retransmission of lost packets. Thus, tight delay Chapter 1. Introduction 12 requirements often translate into stringent P E R requirements. T h e goal of an L A algorithm is to ensure that the most efficient mode is used for given channel conditions, based on a mode selection criterion ( m a x i m u m d a t a rate, m i n i m u m transmit power, m i n i m u m P E R , etc). M a k i n g modes available that enable communication even i n poor channel conditions increases system robustness; under good channel conditions, spectrally efficient modes can be used to increase throughput. In contrast, systems without L A are constrained to use a single mode that is often designed to maint a i n acceptable performance when the channel quality is poor to get m a x i m u m coverage. Such systems are designed for worst-case channel conditions and do not fully exploit the channel when good conditions prevail. In practice, performance can be degraded by problems such as imperfect synchronization or quantization effects. However i n such cases, these degradations can be reduced to manageable levels. To avoid unnecessary complication of the models, these effects are ignored i n this thesis. 1.6 Subchannel and Power Allocation W i t h the i n t r o d u c t i o n of new IP-based multi-rate, m u l t i - Q o S , future networks should be designed for economic packet d a t a transfers [47] w i t h decreasing cost per megabyte i n order to make t h e m more attractive to the users. These services are highly asymmetrical and require high transmission b a n d w i d t h . However, due to the limitations of the available frequency spectrum, its efficient use is crucial to the success of next generation wireless networks. Novel access methods coupled w i t h adaptive resource management techniques, Chapter 1. Introduction 13 for both uplink and downlink transmission, are required to improve spectrum efficiency. It is well established that dynamically allocating transmission rate and power are effective techniques to improve the performance of wireless networks. For example, [48]-[52] consider these problems in the context of the information theoretic capacity region of a multi-access fading channel under various assumptions. Multicarrier transmission schemes have been introduced into code-division multiple access ( C D M A ) systems to enable high data rate transmission. One of the methods is to transmit identical narrowband direct sequence (DS) waveforms in parallel over a number of subchannels using frequency diversity. In [53], a multicarrier C D M A system with an adaptive subchannel allocation method for forward links was proposed. In this system, instead of identical DS waveforms being transmitted over a number of subchannels in parallel, each user's DS waveform is transmitted over the user's preferred subchannel which has the largest fading amplitude among all the subchannels. In [54] subchannel allocation in forward links for M C - C D M A with random signature sequences is studied. A near optimal subchannel allocation policy is formulated by maximizing the total average signal-to-interference-and-noise ratio (SINR). A n iterative algorithm similar to the water filling algorithm was also proposed. In general, efficient subcarrier and power allocation have been shown to reduce performance degradation caused by channel dispersion [55][60]. Among the potential performance-enhancing technologies is orthogonal frequency division multiplexing ( O F D M ) . It is suitable for high speed downlink transmission as it can yield high spectral efficiency due to its robust performance over heavily impaired Chapter 1. Introduction 14 links. O F D M has been demonstrated as an efficient way to mitigate the adverse effects of frequency selective multi-path fading by transmitting signals over a number of flatfaded narrow-band channels. The inherent multi-carrier nature of O F D M also allows the use of adaptive modulation and power allocation. Furthermore, dynamic allocation techniques that efficiently use resources such as bandwidth, power and modulation, need to be devised to increase the system spectral efficiency. O F D M is a promising technique for the next generation of wireless communication systems [61], [62]. O F D M divides the available bandwidth into N orthogonal subchannels. By adding a cyclic prefix (CP) to each O F D M symbol, the channel appears to be circular if the C P length is longer than the channel length. Each subchannel can, thus, be modeled as a time-varying gain plus additive white Gaussian noise (AWGN). Besides the improved immunity to multipath fading [22] gained by the multicarrier property of O F D M systems, multiple access is also possible, because the subchannels are orthogonal to each other. In [59] [63]-[67], adaptive bit and power allocation are studied for O F D M . In these studies, optimal allocation of resources requires a centralized controller with knowledge of every user's channel state. Resource allocation for O F D M has been given a lot of attention. Different distributed resource allocation algorithms for circuit-switched traffic in an O F D M multi-cell environment are compared in [68]. For packet-switched traffic, a distributed resource allocation algorithm that employs channel segregation and interference sensing for efficient resource management is proposed in [69]. A non-frequency selective channel is considered in both papers. In [70], a distributed dynamic resource allocation scheme for a downlink O F D M Chapter 1. 15 Introduction packet-based cellular system that adaptively uses channel information and individual rate requirements to determine the subcarrier, bit and power allocation was proposed and its performance was evaluated for real-time multimedia traffic via simulation whereas for non real-time traffic, the performance was evaluated in [71]. In [72], an analytical model of this algorithm is developed to provide insight into the performance gain in terms of achievable system load for voice and data. • Two classes of resource allocation schemes exist, namely: 1) fixed resource allocation [65]; and 2) dynamic resource allocation [59] [63] [66] [67]. Fixed resource allocation schemes, such as T D M A and F D M A , assign an independent dimension, e.g., time slot or subchannel, to each user. A fixed resource allocation scheme may not be efficient since the scheme is fixed regardless of the current channel condition. On the other hand, in dynamic resource allocation, resources are adaptively assigned to users based on their channel gains. Due to the time-varying nature of the wireless channel, dynamic resource allocation makes full use of multiuser diversity to achieve improved performance. Two classes of optimization techniques have been proposed for multiuser O F D M , namely: 1) margin adaptive (MA) [63]; and 2) rate adaptive (RA) [66][67]. The M A objective is to achieve the minimum overall transmit power given the constraints on the users' bit error rate (BER). The R A objective is to maximize each user's error free capacity with a total transmit power constraint. These optimization problems are nonlinear and, hence, computationally intensive to solve. In [59], the nonlinear optimization problems were transformed into a linear optimization problem with integer variables. The optimal solution can be achieved by integer programming. However, even with integer Chapter 1. Introduction 16 programming, the complexity increases exponentially with the number of constraints and variables. In [66], it is shown that for R A the sum capacity is maximized when each subchannel is assigned to the user with the best subchannel gain and power is then distributed by the water-filling algorithm. However, fairness is not considered in [66]. When the path loss differences among users are large, it is possible that users with higher average channel gains will be allocated most of the resources, i.e., subchannels and power, for a significant portion of the time. The users with lower average channel gains may be starved. In [67], a max-min problem approach is used to ensure that all users achieve a similar data rate. However, the maxmin optimization problem can only provide maximum fairness among the users. In most wireless systems of interest, different users require different data rates, which may be accommodated by allowing users to subscribe to different levels of service. In [73], an adaptive subchannel allocation and bit loading scheme for the multiuser M I M O / O F D M A system is proposed through some modification of the approach for O F D M A systems in [74]. A distributed power control scheme involving iterative waterfilling is proposed [75] for the digital subscriber line (DSL) interference channel. A n iterative power control algorithm is proposed [76] for an M I M O interference system with feedback. The capacity of M I M O interference systems without feedback is treated in [77], and it is shown that putting all power into one antenna is optimal when the interference is sufficiently strong. Optimum signaling for M I M O interference systems with feedback for flat Rayleigh fading channels is treated in [78]. In [79], long-term proportional fairness resource allocation with dumb antennas is studied. It was pointed out that in multiuser Chapter 1. Introduction 17 systems, channel fading can be exploited as a source of randomness, i.e., multiuser d i versity. However, i n some scenarios, due to slow channel variation, the dynamic range of channel fluctuations i n the time scale of interest may be small. 1.7 This Thesis 1.7.1 Objectives T h i s thesis has the following m a i n objectives. T o develop new methods for M C - C D M A that involve efficient subcarrier allocation to improve system capacity for cellular systems. Subcarrier correlation is also considered. T o introduce and study a n adaptive subchannel allocation algorithm based on order statistics suitable for channel-aware systems. T o i n vestigate the effects of correlation on the error performance of order selection, a multiuser diversity method that selects the r - t h order channel gain for transmission. 1.7.2 Contributions T h e m a i n contributions of this thesis are: • A channel-matched multicarrier code division multiple access (matched M C - C D M A ) scheme is proposed for downlink m u l t i p a t h slowly fading channels. T h e design overcomes the performance degradation caused by the channel dispersion. In contrast to a conventional M C - C D M A scheme, subcarriers are chosen so as to be located i n frequency subbands characterized by favorable channel transmission gains. T w o diversity reception techniques, equal gam combining ( E G C ) a n d maximum ratio Chapter 1. Introduction combining 18 (MRC), are considered. The bit error rate performance improvement over conventional M C - C D M A is examined. The results have appeared in [80]. • A new transmission scheme, frequency hopping multicarrier code division multi- ple access ( F H - M C - C D M A ) , is proposed and investigated. This scheme is compatible with existing narrowband second-generation frequency hopping and thirdgeneration wideband C D M A systems. A n analysis of the downlink bit error rate with E G C and M R C in slow frequency-selective Rayleigh fading is provided. Two frequency hopping (FH) schemes, namely random and coordinated F H , are used. The performance improvement of F H - M C - C D M A over conventional frequency hopping is demonstrated, with a capacity gain approaching that of M C - C D M A but using a much lower number of subcarriers. The results have appeared in [81]. • The F H - M C - C D M A scheme is modified for operation in correlated fading channels. The bit error rate (BER) performance of the modified F H - M C - C D M A scheme is compared to those of conventional F H and M C - C D M A , respectively, assuming a tapped delay line (TDL) channel model in which the subcarriers undergo correlated fading. Numerical results indicate that the proposed F H - M C - C D M A scheme can yield a much lower B E R than conventional F H ; with a proper choice of parameter values, the modified F H - M C - C D M A also outperforms M C - C D M A . This work has been reported in [82]. • A channel aware multiple access scheme based on frequency hopping (CAFH) is proposed. In contrast to conventional F H , which uses a channel state independent Chapter 1. Introduction 19 hopping sequence, the transmitter in the channel aware scheme hops to the available frequency subband which currently has the largest transmission gain. It is shown that the proposed scheme can offer large performance gains over the conventional F H scheme [83] [84]. A n efficient method is presented to evaluate the performance of C A F H with r rounds. The resulting closed-form expressions are used to investigate the effect of the number, r, of rounds on the error probability for a cellular communication system. Numerical examples are presented to illustrate the application of the new analysis. This work has been reported in [85]. • A procedure for computing the cumulative distribution function (cdf) of the r-th order statistic of a set of arbitrarily correlated fading channel gains is studied. The method should prove useful in the study of various communication problems. For example, it can be applied to the performance analyses of diversity systems operating over correlated Nakagami-m fading channels. Numerical results are presented to illustrate the effect of fading correlation and the fading severity parameter as well as the accuracy of an "exchangeable" approximation. The results have been reported in [86]. • A systematic characteristic function (CF)—based methodology has been developed for computing the B E R of the r-th largest channel gain for correlated Nakagami fading channels for any given covariance matrix. The method should prove useful in the study of various communication problems. Numerical results are presented to point out the application of the new analysis. The expressions derived are then used to evaluate the error performance for C A F H . The effect of the fading correlation Chapter 1. Introduction 20 on the B E R performance is illustrated. 1.7.3 Road M a p In Chapter 2, matched M C - C D M A is proposed as an adaptive subchannel allocation scheme with-non-uniformly spaced subcarriers matched to the channel transfer function to improve B E R performance. A new scheme, Frequency-Hopping Multicarrier C D M A ( F H - M C - C D M A ) , is de- scribed in Chapter 3. The performance of F H - M C - C D M A used over a Rayleigh fading channel is analyzed. F H - M C - C D M A may be viewed as a combination of F H and M C - C D M A in which M C - C D M A is applied to portions or subbands of the available bandwidth. A new analytical model for correlated F H - M C - C D M A is presented in Chapter 4. The effect of equalizing the interference in each subcarrier group while maintaining a reduced correlation among the subcarrier fading processes for a group is studied. A novel channel aware frequency hopping (CAFH) multiple access scheme is described in Chapter 5. The B E R of C A F H is analyzed assuming independently Rayleigh faded subbands. It is shown that C A F H can offer significant performance improvements over conventional F H . The resulting channel assignment algorithm is iterative in nature. The performance analysis of C A F H with r rounds is also considered. Expressions for the cumulative distribution function (cdf) of the r-th order statistic, r = 1, 2 , . . . , n, for correlated Nakagami fading branches are obtained in Chapter 6. The B E R of a channel aware diversity system, hereafter referred to as order select-ion, is Chapter 1. Introduction 21 evaluated in Chapter 7 for correlated Nakagami fading channels, when the r-th order branch is selected for transmission. Chapter 8 concludes the thesis and outlines future research problems. 22 Chapter 2 Matched M C - C D M A in Slowly Fading Dispersive Channels 2.1 Introduction Frequency selective multipath fading is common in urban and indoor environments and can significantly degrade the performance of wideband, high rate mobile communication systems. M C - C D M A , a combination of O F D M and C D M A , has been proposed as a viable transmission technique [32, 33, 35, 38, 87]. W i t h M C - C D M A , the conventional serial transmission of a data stream is converted into a parallel transmission of data symbols using a large number of narrowband orthogonal subcarriers. In this chapter, a modified M C - C D M A technique, hereafter referred to as matched M C - C D M A , with non-uniformly spaced subcarriers matched to the channel transfer function is proposed to improve the B E R performance. Two receiver diversity combining techniques, E G C and M R C , are considered for combining subcarrier signals. °A version of this Chapter has been published [80], Chapter 2. Matched MC-CDMA in Slowly Fading Dispersive Channels 2.2 System Model The matched M C - C D M A transmitter spreads the original data stream over different subcarriers in the frequency domain using a given spreading code. Subcarriers are concentrated in the regions of the channel with low attenuation, thereby reducing the performance degradation caused by the frequency-selective channel. A n appropriate channel measurement technique is assumed. The channel is modeled using a two-path transfer function developed by Rummler of the form [88], [89] C(f) = a[l - Be-**V-M *], T f £ (/ , f + B ) c c (2.1) T where B? is the total available bandwidth, a and (3 are the overall attenuation parameter and shape parameter respectively, f is a fade minimum frequency, and T I is the relative 0 ( time delay between the two paths. deep fades at The effect of the multipath component is to create / = f + i/r^, % = 0, ± 1 , ± 2 , • • •. Because of the reduced M C - C D M A 0 symbol rate, a synchronous downlink channel is assumed [87]. The distribution of the matched frequency subcarrier locations is proportional to the square of the magnitude of the channel transfer function, | C ( / ) | , and is obtained from the cumulative probability 2 distribution function (cdf) of | C ( / ) | as in [55] 2 (2.2) elsewhere. A matched M C - C D M A signal is generated by replicating a single data bit into L 23 Chapter 2. Matched MC-CDMA in Slowly Fading Dispersive Channels copies, each multiplied by a different element (chip), Cj[l] € 24 — 1, 2, •• •, L of the spreading code assigned to user j and then binary phase-shift keying (BPSK) modulated onto a subcarrier. The transmitted matched M C - C D M A signal can be written as sj(t) = ^ J Z ^ - o o E f = i • cos ( 2 T T ( / + g ) i ) • p(t - kT ), b c FeZ (2-3) where aj[k] € {—1, +1} is the A;" data bit of user j, T is the bit duration, p(t) = rect(£/Tb), 1 b and {Fi} are the L subcarrier indices selected for channel matching. The total number of available orthogonal subcarriers, uniformly located within BT, is denoted by N. Each subcarrier has a bandwidth of B . CB Conventional M C - C D M A can be viewed as a special case of matched M C - C D M A with Fi = IF, F 6 Z. The received signal to noise ratio (SNR ) for the I th r subcarrier is where fi — fc + ^ and A = \Ai\e i = C(f + | f ) is the I j&A th t c subcarrier complex channel gain and E is the bit energy. Matched M C - C D M A is expected to provide an improved b performance over conventional M C - C D M A since L L \Al,match\ > ^2 \Al nv\ > 2 2 !CO (= 1 where A^ match and Ai 1 = 1,- ••,£/. (2-5) 1=1 are the complex channel gains corresponding to the I th iConv sub- carrier in the matched and conventional M C - C D M A schemes respectively. The received Chapter 2. Matched MC-CDMA in Slowly Fading Dispersive Channels signal with M active transmitters can then be written as r(t) = i Ef=l Et°°-oo KIU'M*] ' COS (2TT(/ c + g)t + 6) A • (t - kT ) + n{t) P b (2.6) where n(t) is a sample function of a white Gaussian noise process with zero mean and power spectral density N /2. 0 Assuming perfect phase recovery, the decision variable for the k symbol of user j is th D (kT ) - Ef=i {G i [E!=i J MC b it +n {kT )) l b (2.7) where Gj i is the gain factor for user j at subcarrier I and depends on the combining t scheme used, and {ni(kT ), I = 1,2, b • • -,L} are sample values of independent Gaussian random variables with mean zero and variance 2.3 Numerical Results The Walsh-Hadamard sequences with a processing gain (PG) of 64 are assigned to users as follows: the reference user (user 1) is assigned each of the 64 sequences in turn. Assume that user 1 is assigned sequence number j. If j = 1 or j > M, user i, i = 2, • • •, M is assigned sequence number i. If 2 < j < M, then user i is assigned sequence number i if i < j, and i +1 if i > j. Numerical results show that the proposed matched M C - C D M A scheme can provide a much lower B E R than conventional M C - C D M A . As shown in Fig. 2.2, the number, 25 Chapter 2. Matched MC-CDMA in Slowly Fading Dispersive Channels N, of available subcarrier channels is 128 from which L = 64 matched subcarriers are selected. The channel parameters are a — —1 dB, (3 = 0.5, f = 0.2B and 0 T = 2/BT- Fig. 2.3 shows the B E R as a function of the number of users for matched M C - C D M A and conventional M C - C D M A with F = 2, and S N R = E /N b 0 of 9 dB, where E is b the transmitted energy per bit. The figure shows that the proposed matched M C - C D M A scheme can yield up to a 10-fold reduction in B E R compared to conventional M C - C D M A . A B E R comparison as a function of S N R for M=64 users is illustrated in Fig. 2.4. For the same combining scheme, the S N R difference between matched and conventional M C C D M A required to achieve a given target B E R increases as this target B E R decreases. For 64 users and a B E R of I O , matched M C - C D M A with E G C provides a gain of - 2 about 4 dB compared to conventional M C - C D M A with E G C . It should be noted that for a small number of users (i.e., in a noise limited environment), M R C outperforms E G C . However, for a large number of users (i.e., in a interference limited environment), E G C has a superior performance. This is likely due to the fact that the orthogonality of the codes is affected to a greater extent by M R C . M R C is only optimal in an additive white Gaussian noise (AWGN) environment, which is the case for an uplink scenario. 26 Chapter 2. Matched MC-CDMA in Slowly Fading Dispersive Channels (b) Figure 2.1: (a) Transmitter model (b) Receiver model 27 28 Chapter 2. Matched MC-CDMA in Slowly Fading Dispersive Channels 21 1.75 1 1 r - Subcarrier number, / Figure 2.2: Matched M C - C D M A subcarrier pattern, channel transfer function T F , and integral \T/(-) with T — 2/B - The frequency corresponding to the I subcarrier is f + F B . th d c t ch T 29 Chapter 2. Matched MC-CDMA in Slowly Fading Dispersive Channels o 10 20 Matched M C - C D M A (EGC) Conventional M C - C D M A (EGC) Matched M C - C D M A (MRC) Conventional M C - C D M A (MRC) 30 40 Number of users, M 50 60 Figure 2.3: B E R as a function of number of users for S N R — 9 d B . 70 30 Chapter 2. Matched MC-CDMA in Slowly Fading Dispersive Channels io u SNR, dB Figure 2.4: B E R as a function of S N R for M = 64 users. 31 Chapter 3 Frequency-Hopping Multicarrier C D M A in Rayleigh Fading 3.1 Introduction Multicarrier modulation is an attractive scheme for high data rate digital transmission because it allows a reduction of the symbol rate, at the cost of increasing the number of subcarriers used to transmit the information. In C D M A , multiple users can transmit simultaneously within the same frequency band by using a number of different pseudo-noise sequences. M C - C D M A is a combination of these two techniques. Unlike D S - C D M A , in M C - C D M A the input data is duplicated and transmitted through a number of narrowband subcarriers. The signaling period can thus be much larger than the delay spread of the channel making the system less susceptible to ISI due to multipath propogation. On each subcarrier, the transmitted bit is binary antipodally modulated based on the spreading codes assigned to the user. At the receiver side, the received signals at the different subcarriers are weighted and summed to produce a decision variable. For the downlink system, orthogonal spreading codes can be used to reduce multiuser interference. How°A version of this Chapter has been published [81]. Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading 32 ever, since the channel gains for different subcarriers are different, strict orthogonality among users is lost and multiuser interference is introduced. The two principal types of C D M A are DS and F H . A major advantage of F H is that it can be implemented over a much larger frequency band than direct-sequence spreading, and the band can be noncontiguous [90]. A second advantage is that frequency hopping provides resistance to multiple-access interference while not requiring power control to deal with the near-far problem. In D S - C D M A systems, accurate power control is critical on the uplink [91]. A form of F H which allows simultaneous transmission in several frequency bins has also been investigated [92]. Because at each frequency hop, a frequency-hopping multicarrier CDMA ( F H - M C C D M A ) user transmits in only a small portion of the band at a given time, multiple transmissions may occur simultaneously. In addition, users are separated through the use of different spreading codes in the frequency domain thereby reducing the effects of hits. In a synchronous down-link mobile radio communication channel, we may use orthogonal Walsh Hadamard codes. The proposed F H - M C - C D M A may be viewed as a combination of F H and M C - C D M A in which M C - C D M A is applied to portions or subbands of the available bandwidth, to which users may hop to. M C - C D M A has disadvantages [93] and problems associated with nonlinear amplification [94] and high peak-to-average power ratios (PAPRs), which result from the fact that the M C - C D M A signal is composed of many subcarriers, are reduced, as F H - M C - C D M A generally uses a smaller number of subcarriers than M C - C D M A . In an F H - M C - C D M A receiver, at the different hops, the received signal is combined in the frequency domain, therefore, the receiver can employ Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading 33 all the received signal energy scattered in the frequency domain. 3.2 FH-MC-CDMA System Model Consider a system where M users are transmitting over the channel. Each user transmits one message bit per hop in one of g available frequency groups. A synchronous frequencyhopping system is assumed, with the hop intervals (of period T ), of all transmitters b aligned at the receiver. The users' hop sequences and message bits are assumed to be independent random sequences, uniformly distributed over the g frequency groups and the set {-1,1}, respectively. To evaluate the performance of the system, it suffices to consider one of the users, hereafter referred to as the reference user. The proposed transmitter and receiver are illustrated in Fig. 3.1. The I bit d [l] of the m th th m user is first mixed with a frequency-hopping tone of frequency fh* '™^• M C - C D M A processing 1 with a spreading code of length N is then performed. The transmitted signal for the I th bit of the m th user can be written as Sm(t) = J2ll d [l]Cm\i\ • COS ^ ( / ^ + f)t) m with the subcarrier tone frequency f — , IT„ < t < (I + l)T b (^) where F and i are integers. In F H - M C - C D M A , the total radio frequency (RF) bandwidth (BWRF) is divided into s subcarriers which are organized into g frequency groups. Each frequency group consists of N subcarriers (s — gN). The R F signal from a given transmitter is hopped from group to group by changing the carrier frequency as illustrated in Fig. 3.2. A n F H - M C - C D M A Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading system can be viewed at one extreme as a M C - C D M A system when 5 = 1, and at the other extreme as a conventional F H system when N = 1, for a fixed total bandwidth, i.e., a fixed value for s. A pseudo-FH system is obtained by choosing a large number of frequency groups, each with a small number of subcarriers. For example, with s — 128, we might choose g = 64 and N = 2. In this way, some of the advantages of both F H and M C - C D M A systems may be obtained in a single hybrid scheme with those of the F H dominating. On the other hand, a pseudo-MC-CDMA is obtained by choosing a large number of subcarriers and a small number of frequency groups, e.g., N = 64 and 5 = 2. Assuming that signals in different groups do not interfere with each other and that delay spread effects are negligible, the received signal corresponding to the transmission of the I message bit is given by th r(t) = £. (3.2) lT <t<(l b where M h is the number of users in a given group, p th user, 9^ m) b is the channel gain at the i th m<i subcarrier of the m + l)T is a random phase offset and n(t) is a sample function of a white Gaussian noise process with zero mean and power spectral density N /2. It is 0 assumed that {p i}m=i,...,M ,i=i,...,iv are outcomes of statistically independent Rayleigh mi /l distributed random variables and {fV' ^}m=i,...,M ,i=i,...,JV are statistically independent ,m h random variables, each of which is uniformly distributed in [0, 27r). The probability that M — 1 other transmitters (interfering users) share the same h 34 Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading frequency group as the reference user is given by ^ M- 1 ^ P{M (3.3) Pf?»-\\-P ) - \ M M h h M -l h where P — ^ is the probability of another transmitter hopping to the reference frequency h group. Two F H - M C - C D M A schemes are considered, namely random and coordinated F H - M C - C D M A . In random F H - M C - C D M A each user randomly selects a spreading code (out of the N possible Walsh Hadamard codes) and a frequency group (out of the g possible groups). In coordinated F H - M C - C D M A , the M users are assigned to the g frequency groups in such a way as to equalize as much as possible the number in each group. For each frequency group, each of the M h users is assigned a distinct code, so as to avoid multiple users sharing the same code and group simultaneously (i.e., hit). This is done by picking at random a set of Mh distinct codes from the N available codes. There are a total of such sets. \M ) h Assuming that the receiver is synchronized with the desired reference user {m — 1), we can express the decision variable for the I th Z = %I i n where m r(t) E i = i bit as follows • cos {2,{h ' [l l) + fi)t + * i ) dt, (3-4) U ) is the reference receiver gain factor (which depends on the diversity reception combining scheme) at subcarrier i, and df' ^ is the estimated phase at the i 1 of fh~ • Assuming perfect phase correction (of l,r> th subcarrier — O^ ' ^), the decision variable reduces 1 1 35 Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading to Z = M} 1 where 0J' ,m) T,i=l Pl,iCl[i]Ql,i = 9i {l,1) - ^ ' subcarrier is given as ZEGC = d [l} X Ef ( , m ) q\^ = = 1 Pl,i + Emi m[l] d 2 HiLl Pm,iCm[i]qi,i • COs(0f ) + m ) ( -5) 3 p, and p. is & noise term. With E G C , the gain factor at the i th C\[i], and the decision variable becomes + E ^ = 2 dm[l] Eill • COs(0<'' PrnjCmlilCllil m) ) + p, E G C . (3-6) The three terms in (3.6) may be expressed as I$2° = E ^ 2 Msec = i C 1)Tt drnll] E i = l / V i ^ W c x W »(*) Ef=i i W • c c o s ^ • COS(ff (2T(A ( M ) + /<)* + A. The mean and variance of the random variable for the interference term are given by a^oc = = where P m — N^f^- E[{l^f] (3-7) 2(M,-l)A(l-f), represents the mean power of the m th user. The mean and variance 36 37 Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading of the noise is given by \PBGC\ = E l c 0 = ^ E G C P E G C \ = %N. G E EG W i t h M R C , the gain factor at the i subcarrier is given as q\ = C\ \i\p\,i, and the decision th ti variable becomes Z RC M = di [I] Eili Pl,i + £„= m[l] E,=l Pm,iP\,iC [i)Cl [i] • COs(0J'' ) + d 2 M) m P RCM (3.8) The three terms in (3.8) may be expressed as < B C = di[/]E^ P?,i 1 = E^2 dm[l] Til % f l C = t C i m Pm.Pl^cM n(t) £ f • C08(fli'' ) m) Pi,iCi[i] • cos (27r(/ C- ) + f )t + 0f ) dt. l) 1 = 1 h t The mean and variance of the random variable for the interference term are given by o)u,c = E[I%r } = .2 ]\2] (M — l)N[E[pfj] - (E[pl \) (3-9) 2 int h ti Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading 38 The mean and variance of the noise are given by \PMRC\ = E l a MRC 0 - EI^MRCVMRC] — -Pi 2 No The B E R for a F H - M C - C D M A system can be calculated as 7ZMC BER FH-MC BER = Etf =l i h P M t ~ ) \MC-CDMA),EGC SM„=1 ( h ~ ) { MC-CDMA),MRC P M 1 CDMA)MRC CDMA)>EGC BEP (3.10) = and BER^_ where BER$£_ l BER t are the downlink B E R ' s of an M C - C D M A system employing E G C and M R C respectively. Approximating the interference, I , from other users by a Gaussian random variable results in an M C - C D M A B E R given int by [38] ~ lerfc {J^^ BER^_ CDMA)EGC ^ J ^ p ^ + M ,N ^ r_ I / PoTb (MC-CDMA),MRC ~ {\j ^P T J ' Nq ^ 1 7 h BEB 3.3 Numerical 2 Results and 0 b +N J ' 0 Discussions The B E R values for coordinated F H - M C - C D M A calculated from (3.10) and obtained from computer simulations for N — M, are plotted as a function of the average S N R in Fig. 3.3. The S N R is defined as PiT /N . The simulation results are represented by the b 0 ^ Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading 39 symbols and the theoretical results'by the dotted curves with symbols. Fig. 3.4 shows the theoretical and simulation results for coordinated F H - M C - C D M A with N < M. We find that simulation results are well approximated by the analytical results when the number of users is large. For a small number of users the Gaussian approximation of (3.10) is not as accurate, e.g., 32 users using E G C . The B E R performance of random F H - M C - C D M A and coordinated F H - M C - C D M A using E G C for different number, N, of subcarriers is.compared in Fig. 3.5, for M = 32 users. A B E R comparison of random and coordinated F H - M C - C D M A for M = 64 users for different.subcarriers and employing E G C is illustrated in Fig. 3.6. It is clear that coordinated F H - M C - C D M A provides a lower B E R than random F H - M C - C D M A . The B E R comparison between conventional F H employing 128 subbands, M C - C D M A having 128 subcarriers and coordinated F H - M C - C D M A employing different number, N, of subcarriers for M = 64 using E G C and M R C are shown in Fig. 3.7. As expected, for a given average SNR, the B E R decreases with N, due to the bigger diversity gain. The figures show that the proposed F H - M C - C D M A scheme can yield up to a 10-fold reduction in B E R compared to conventional F H when employing two subcarriers in the interferencelimited region. In the noise-limited region F H - M C - C D M A becomes comparable with M C - C D M A . We find that the performance difference between a F H - M C - C D M A scheme employing 64, 32, 16, 8 and 4 subcarriers and a 128 subcarrier M C - C D M A is small. The performance difference is even smaller at low S N R values. Chapter 3. Frequency-Hopping Multicarrier CDMA COS 2K -7f [ 1 ] 1 4sn (l.m) COS(2JC/ A Band Pass Filter COS 271 "= ' c [2] 40 in Rayleigh Fading )t Frequency Synthesizer COS 27C [N] NF t (a) Band Pass Filter COS(27r^ r(0 + 9 (l.m) ) ^ ; , 4jn Band Pass Filter (l.m) )t • cos(2nf Frequency Synthesizer t Integrator a™; COS(27t— f + 9,) Band Pass Filter j/, ; m cos(2n— t F +6 ) <7 fc (b) F i g u r e 3.1: (a) T r a n s m i t t e r m o d e l (b) Receiver m o d e l Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading 41 Frequency Time Hop interval Figure 3.2: F H - M C - C D M A pattern. 42 Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading -if " * * * & + * A • o k M R C - 64 users M R C - 32 users EGC - 64 users EGC - 32 users 10 average SNR, dB 15 Figure 3.3: Theoretical and simulation B E R comparison for N = M 20 Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading HT't 43 I. A + * A • 0 MRC MRC EGC EGC - 64 users, 32 subcarrier 32 users, 16 subcarriers 64 users, 32 subcarrier 32 users, 16 subcarriers 10 average SNR, dB 15 Figure 3.4: Theoretical and simulation B E R comparison for N < M. 20 Chapter 3. Frequency-Hopping Multicarrier CDMA io in Rayleigh Fading 44 u 4 ' AA A n D 10 10 -3 * -*— o -aA -A- 16 subcarrier coordinated F H - M C , EGC 16 subcarrier random F H - M C , EGC 8 subcarrier coordinated F H - M C , EGC 8 subcarrier random F H - M C , EGC 4 subcarrier coordinated F H - M C , EGC 4 subcarrier random F H - M C , EGC 10 average SNR, dB •* •.. " ' 15 • * 20 Figure 3.5: B E R comparison between random and coordinated F H - M C - C D M A for M 32 users. Chapter 3. Frequency-Hopping Multicarrier CDMA in Rayleigh Fading 10" A ••• *• O• 32 subcarrier random F H - M C , E G C 32 subcarrier coordinated F H - M C , E G C 16 subcarrier random F H - M C , E G C 16 subcarrier coordinated F H - M C , E G C 8 subcarrier random F H - M C , E G C 8 subcarrier coordinated F H - M C , E G C 4 subcarrier random F H - M C , E G C 4 subcarrier coordinated F H - M C , E G C 10" 10 average SNR, dB A •Q. • *• o• 15 20 Figure 3.6: B E R comparison between random and coordinated F H - M C - C D M A for M 64 users. 45 Chapter io 3. Frequency-Hopping Multicarrier CDMA in Rayleigh 46 Fading u 10 -11 04 W CQ '•A. 10" A- A • ••• • #• •q>. 128 subcarrier M C - C D M A , E G C 128 subband F H 64 subcarrier F H - M C , E G C 32 subcarrier F H - M C , E G C 16 subcarrier F H - M C , E G C 8 subcarrier F H - M C , E G C 4 subcarrier F H - M C , E G C 2 subcarrier F H - M C , E G C • • *• ••6> : 10 10 15 average SNR, dB F i g u r e 3.7: B E R as a function of S N R for M — 64 users. 20 47 Chapter 4 Frequency-Hopping Multicarrier C D M A in Correlated Rayleigh Fading 4.1 Introduction In M C - C D M A , the available system bandwidth is partitioned into s subbands, each associated with a subcarrier frequency. Each data bit of a mobile station (MS) is multiplied by a chip sequence of length s; each of the resulting s chips is then transmitted using a different subcarrier so as to exploit the diversity available in frequency-selective channels. In [81], F H - M C - C D M A was proposed for use in a frequency-selective channel in which different subbands fade independently. In F H - M C - C D M A , in each hop, a MS transmits simultaneously over a predefined set of frequency subcarriers, referred to as a group. In this chapter, we propose a modification of the F H - M C - C D M A scheme in [81] that is suitable for use in correlated fading channels. In particular, with the appropriate choice of the number of subcarriers in each group and subcarrier interleaving, M A I and °A version of this Chapter has been published [82]. Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 48 correlation among the subcarrier fading processes for a group are largely reduced. We compare the B E R performance of the modified F H - M C - C D M A scheme with those of conventional F H and M C - C D M A , respectively. The results show that F H - M C - C D M A with appropriate subcarrier interleaving and an appropriate number of subcarriers for each group can provide a significantly lower B E R than M C - C D M A . 4.2 System Model 4.2.1 FH-MC-CDMA Let the total number of subcarriers which can be accommodated in the available system bandwidth, W, be denoted by s. The subcarrier spacing can be chosen to be approximately 1/Tj,. Thus, for a system which fully utilizes the available bandwidth, W ~ s/T . In F H - M C - C D M A , the s subcarriers are partitioned into g frequency groups b Gi, G2, • • -,G , each consisting of Nc subcarriers so that j g In each hop of bit duration T , one bit from each MS is sent using one of the g groups. b The M G . MS's in a given group Gj transmit data bits simultaneously to the base station (BS). Each of the Mc MS's is assigned a distinct spreading code of length Nc i r The F H - M C - C D M A scheme in [81] was proposed for the case in which the fading gains in the s subbands are uncorrelated. For the case when these gains are correlated we propose a modified F H - M C - C D M A scheme in which the frequency separation between Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 49 adjacent subcarriers in a given group Gj is increased so as to exceed the coherence bandwidth, A / , by interleaving the subcarriers of group Gj with those of other groups. c This allows subcarriers assigned to a given MS to experience largely uncorrected fading. We note that an M C - C D M A system with a larger subcarrier spacing would require a larger system bandwidth when compared to that of the proposed F H - M C - C D M A scheme. Based on M S allocation and subcarrier assignment, three F H - M C - C D M A schemes are considered: 1) In uniform F H - M C - C D M A , the M MS's are assigned to the g groups in such a way as to equalize as far as possible the number of MS's in each group. Each group consists of Nn = N — - V j subcarriers. 3 9 2) For comparison purposes, we also consider random F H - M C - C D M A , where each MS randomly selects one frequency group (out of the g possible groups) for use. Each group Gj has N Cj = N = * V j subcarriers, and M Gj G {1, • • •, N} MS's. 3) The numerical results in Section 4.4 suggest that optimum error performance is achieved when choosing M Gi = 1 V j . In this case, hereafter referred to as optimum F H - M C - C D M A , we have g = M. The proposed system is illustrated in Fig. 4.1. For a given M S m, each component of the spreading code (c [0], c [ l ] , . . . ,c [N .-l)), where c [i\ € { - 1 , +1}, is multiplied by m m m G the k data bit, d [k] G {-1, +1}. The i th th m m product signal, where i = {0,1, • • •, N . - 1}, G is used to modulate a baseband subcarrier of frequency fi = TJT, where F is an integer that determines the subcarrier separation. The N Gj modulated baseband subcarriers are then summed to form the composite baseband signal. The R F signal is formed by frequency Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 50 shifting the composite baseband signal by an amount equal to the carrier frequency, f (k,m)_ Yox F H - M C - C D M A , the transmitted signal of MS m for bit k can be written as h s (t) = E £ O ~* d [k}c [i\ • cos (2n(f ^ m 4.2.2 m m h kT < t < (k + 1)T„. + /<)*) , b ( -2) 4 Channel M o d e l The channel is modeled as a tapped-delay line (TDL) with a baseband equivalent impulse response [95] ^-)()=x:^' ^-^) m) i I (4.3) (=o where S(-) is the unit impulse function, and the complex channel coefficients, h^' \ m for path / are independent, identically distributed zero-mean complex Gaussian random variables, and L is the total number of resolvable paths. The complex subband coefficient of channel subcarrier i, located at / = f min + jr, i — 0,1, • • •, s — 1, with f min representing the lowest subcarrier frequency in the available transmission bandwidth, W, is L-1 hf ,m) = ^2 e- '^wh<i ' - K j27Tf k m) j2vi e (4.4) (=0 where h\ ' ^ — p^ ' )e^ ' ', with p^ ' ^ representing the Rayleigh distributed channel k m k m ei(k m) k m amplitude gain and 9^ ' ^ representing the random phase offset. For F H - M C - C D M A , the k m signal received by the BS from group Gj, over a slowly time varying, frequency selective Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 51 Rayleigh fading channel can be expressed as r(t) = Em=! d [k] ES" 1 m Pm ' cm\i] • cos ^ ( / ^ + f)t + (k m) %) ' ) (fc m) + n(t), fcT < t < [k + l)T b b where = (fh~ ' ^ ~ k m fmin + • ^1, and n(t) is a sample function of a white Gaussian fi) noise process with power spectral density No/2. Assuming that L divides s, subcarrier indices separated by nf, n G {1, 2, • • •, L — 1}, experience uncorrected fading since = ^ E/TQ e? ™± = 0 1 2 (4.5) where A = E[\h[\ } and E[-] denotes expectation. 2 4.3 BER Analysis of F H - M C - C D M A Assuming that the BS is phase synchronized with the target MS, m, we can express the decision variable for the k bit as th z = where q i &T )Tb r W _ 1 ^ c o s ( 2 7 r (A ( f c , m ) + /*)* + 0m ' ) — p\ ^c [i] is the receiver weighting factor for subcarrier k m<l • m ratio combining (MRC) is used by the BS. {k m) > dt ( 4 when maximum 6 ) Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 52 4.3.1 Uniform F H - M C - C D M A In this case, each group consists of Nc — N V j subcarriers. Improved error performance 3 is achieved by setting N < L, F = g = jj, and f ' {k h m) - f G{0, Y >T >" '> f } E min b b shown in Fig. 4.2. This may be explained by the fact that a group Gj will experience larger M A I for larger values of N (i.e., larger values of M Gj in a given group Gj), and correlated fading among its subcarriers as N increases above L. Here, we choose j groups to contain % + 1 MS's and the remaining g — j groups to contain i MS's, i.e., M = j.(i + l) + (g-j).i (4.7) = i.-g + j, and j = M mod g; [x\ and "mod" denote the largest integer smaller where i = than x and the modulo-operator, respectively. The average B E R can then be calculated as p _ uniform where P ' ' M-(i+l)j M pi,N c (j+l)j M pi+1,. c (4.8) is the average B E R on the uplink of an M C - C D M A system using M R C , when having i MS's and N subcarriers [89], [38], [95] ^ N - 1+n ^ N < L, i = 1 n . 7 2 Q r J 0 N < L, i > 1 L z ~ eL L 1 (L-1)! Lz 1 b N > L, i > 1 (4.9) a s Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 53 where Q(z) = -j=e~ / dy, 7 = N ^ % y2 E PH No (SNR) of the m th 4.3.2 2 * ^ is the average signal-to-noise ratio M S , and p, = \/j^=, where 70 = ^ is the average S N R per channel. Random F H - M C - C D M A As in uniform F H - M C - C D M A , each group consists of N =-N V j subcarriers. In this Gj case, each M S randomly selects one frequency group for use (i.e., M Gi € {1,- • •, N}). The average B E R for random F H - M C - C D M A can be calculated as P D =T? P(i-i)P - random Z_-/2=l T , N 1 \ / o (- ) 4 10 ' where P(i — 1) is the probability that i — l interfering MS's are in the same group as the desired reference MS, and is given by ^ M- 1^ 1Y~* P(i - 1) = 4.3.3 1 (1--) M 1 . (4.11) Optimum F H - M C - C D M A The total number of MS's may be written as M = 2 + j, (4.12) l where j = {0,1,' • -,2 — 1} and N = |f. Our results in Section 4.4 suggest that l optimum error performance is achieved by choosing (2 — j) groups to have TV subl Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 54 carriers and the remaining M — (2 — j) groups to have y subcarriers. In this case, l f (fc.m) _ t p f n i 1 , , . E=l 2. 1 + J_ £-L-2_ ... s + E=l\ F — = Mr = 1 V j and g = M. Using this method, M A I is eliminated by having one MS in a group, while maximizing the diversity for a group. Reducing the number of subcarriers in a group below the proposed number will result in diversity reduction, while increasing the number of subcarriers in a group beyond the proposed number will result in more MS's in a group, thus, M A I is introduced. The average B E R can then be calculated as = M_ <_j) op tirnurn 4.4 ( 2 M l.g o (4.13) &-j) i.TT p M e N u m e r i c a l Results Numerical results are now presented to compare the performance of the proposed modified F H - M C - C D M A with that of M C - C D M A and conventional F H . Although not shown, simulation results were obtained and found to agree closely with the theoretical curves. In Fig. 4.3, the number of resolvable paths, the total number of subcarriers, and the total number of MS's are chosen as L — 32, s = 256, and M = 32, respectively. When operating at N = 8 subcarriers, and F = g = s/N set to 32, uniform F H - M C - C D M A reduces to optimum F H - M C - C D M A having a single MS (M Gj = 1 V j) in each group. In this case, F H - M C - C D M A may be viewed as an optimum access scheme with no M A I , thus, outperforming all other schemes. This result is expected due to the fact that a Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 55 group Gj will experience larger M A I if N > 8, and a diversity gain reduction if N < 8. When operating at N = 16, uniform F H - M C - C D M A has M Gj = 2 V j. As expected, the B E R performance degrades with an increase in N. The error performance rapidly worsens when N > L due to the correlated fading among the subcarriers of a group. M C - C D M A which is a special case of F H - M C - C D M A when N = s and g = 1, has the highest M A I and the largest correlation among its s = 256 subcarriers. The figure clearly shows that F H - M C - C D M A outperforms M C - C D M A . In particular, smaller values of M . G result in a lower B E R . To demonstrate the effect of lower diversity orders, L = 8 is chosen in Fig. 4.4. When operating at N = 8, uniform F H - M C - C D M A achieves independent fading among its N subcarriers with no M A I (i.e., M . = 1 V j). Comparing Figs. 3 and 4 for N — 8, the G error performance is the same for both cases (L — 8 and 32). For N > 8, it is clear that a smaller value of L results in a significant performance deterioration due to the increasing correlation among the subcarrier fading processes. Fig. 4.5 depicts the B E R of uniform F H - M C - C D M A as a function of N when L = 16 and S N R = 10 dB. For a given M, the figure clearly indicates that N = 1 (i.e., F H ) and N = s (i.e., M C - C D M A ) are not favorable choices for N. The B E R performance improves when N is varied between 1 and s. In particular, smaller values of M Gj result in a lower B E R . This is due to the fact that for a given M, a reduction in the number of MS's in each group results in smaller M A I . Optimum error performance is achieved when M . = 1 V j (i.e., optimum F H - M C - C D M A ) . G Finally, in Fig. 4.6, we show the B E R as a function of M when L — 16 and S N R = 10 Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 56 dB. Clearly, F H - M C - C D M A outperforms all other schemes over a wide range of active MS's. More specifically, optimum F H - M C - C D M A provides the lowest B E R . Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 57 c [0] - OF COS(ZK — t ) |i) \F COS(2K Y, I) Sfi) 'b d[k] ) (km) Band Pass Filter Transmitted signal for group Gj 2cos(2nf t ) Frequency Synthesizer (N -\)F cos(2n - „ t) T C c IN - l ] <" j H (a) Band Pass Filter 'h r(0 41*1 Band Pass Filter 2 ft"") — cos(2nf t) Frequency Synthesizer Integrator W 1 C Band Pass Filter 7i ' (b) Figure 4.1: (a) Transmitter model; (b) Receiver model Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 58 0 mm 1 J frequency 7VF 1 jj W • group 1 H group 2 • g r o u P^r Figure 4.2: Frequency groups when N Gj = N V j with N — L = 4, and F = g = s/N. Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 59 Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 60 Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 61 2 4 8 16 32 64 Number of subcarriers, /V 128 Figure 4.5: B E R versus N for L = 16, and S N R = 10 d B . 256 Chapter 4. Frequency-Hopping Multicarrier CDMA in Correlated Rayleigh Fading 62 63 Chapter A 5 Channel Aware Frequency Hopping Multiple Access Scheme 5.1 Introduction The B E R of F H systems has been studied in [96]. Although a great deal of emphasis has been placed on DS, recent developments in F H technology [97, 98, 99, 100] have resulted in commercial applications such as mobile cellular communications, personal communications and wireless local area networks (LANs). On the other hand, F H systems have been considered for a variety of military applications due to their frequency diversity and resistance to near-far problem [91, 101]. In an F H system, the total available bandwidth is divided into a number of subbands. F H allows many mobile stations (MS's) to share a common channel by employing user-specific hopping patterns which specify the subband occupied by each MS at a given time and allow the data of different MS's to be recovered at the base station (BS). Thus, the'transmitted signal appears as a data-modulated carrier that is hopping from one frequency to the next. A n advantage of F H over DS is that it can be implemented over a larger frequency band, and the band can be non°A version of this Chapter has been published [83, 84, 85]. Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme contiguous [90]. A second advantage is that F H provides resistance to multiple-access interference while not requiring power control to deal with the near-far problem. In [55], matched F H was described for a single-user system to provide performance improvement on a time-invariant frequency selective channel. The subbands were designated as either usable or unusable (i.e., very poor) and hops were only allowed to usable subbands. . In this chapter, a channel aware frequency hopping ( C A F H ) multiple access scheme is described. In this scheme, a suitable channel measurement technique is assumed. The utilization of one of the recently developed blind identification techniques [102] would be useful in this aspect. In C A F H , the BS monitors the channel to estimate the subband gains for each MS. The BS then uses this knowledge to assign to each M S the subband in which the M S has the highest gain, while ensuring that at most one MS is assigned to each subband. The B E R of C A F H is analyzed assuming independently Rayleigh faded subbands. It is shown that C A F H can offer significant performance improvements over conventional F H . 5.2 System Model We adopt a slowly time-varying, frequency-selective Rayleigh fading channel model which is commonly used for wideband systems [89, 103]. The instantaneous SNR, F, on each subband of the channel is distributed according to an exponential distribution given by 1 _JL Prh) = — e •«>, 7 > 0 7o (5.1) 64 Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 65 where 70 is the average value of T. The slowly varying nature of the channel implies that it can be treated as time-invariant over the duration of a few symbols. Consider the uplink of a system in which K MS's transmit over L subbands to the BS. The signal of M S k in signaling interval i is s {t) = ^Y~d kti kti cos(27r(A + f )t), (i-l)T<t< %T c M (5.2) where T is the bit duration, f is the carrier frequency, fh is the frequency offset, E is c ki the bit energy and <4,i € {—1, +1} depends on the information bit to be sent. The signal received during the i th symbol interval at the BS is K 9k,iSk {t - T ) + n(t) r(t) = tl (5.3) FC k=l where r is the k M S time delay, n(t) represents additive white Gaussian noise (AWGN) th k with mean zero and two-sided power spectral density (PSD) N /2 and g ,i is the k MS th 0 subband gain during the i th signaling interval. The subband gain, g i, k> k depends on the scheme that is used to assign subbands to MS's. 5.3 C A F H Scheme In the C A F H scheme with r rounds, the BS assigns subbands to active MS's as follows. The round number is initially set to 1. In round j , j = 1,2,- • -,r, for each M S , k, which has not yet received its subband assignment, the subband f G {1, • • -,L}, with k Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme the j th best subband gain, g^\, out of L subbands is determined. If a given subband has the j th best gain for exactly one M S , the subband is assigned to that M S and the scheme enters round j +1. If an available subband has the j th best gain for several MS's, the subband is assigned to the MS with the largest gain and the scheme enters round j + Once a subband is assigned to a MS, it is not available in subsequent rounds. The subband assignment procedure continues until all K MS's have been assigned subbands or round r + 1 is reached. In the latter event, i.e., there are still unassigned MS's after round r, the BS chooses each such M S in turn in a random order and assigns to it the unoccupied subband for which its gain is highest. A flow chart describing the C A F H scheme is provided in Fig. 5.1. It might be noted that the number of rounds required is at most min{r, K}. 5.4 5.4.1 B E R analysis B E R analysis of C A F H for r = l For simplicity of the analysis, we consider a synchronous system, i.e., T\ = T = . . . rK = 0 [104], 2 with r = 1 round. The B E R , PCAFH, for the C A F H scheme can be written as PCAFH = A + B , (5.4) 66 Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 67 where K l ( ) ^L = ll A Z K (5.5) P fc=l represents the B E R for the case when a subband is assigned to the target MS in round 1, and K - l \ / ^\k-l z(k) / T -, \ K-k (5.6) L represents the probability that k — 1 other MS's occupy the subband of the target MS in round 1. For the case in which a subband is not assigned to the target MS in round 1 (i.e., the highest gain among all the L subband gains of the target MS was smaller than that of some other MS that shared a common highest gain subband as the target MS), the B E R can be approximated by 1 K k=2 where K Z = Y\z{k) (5.8) fc=i is the probability that the target M S is selected in the first round. Let G^ random variable corresponding to the j th P ,j,h e be the largest gain out of h gain samples. The term represents the B E R when the transmitted signal from an MS experiences a subband gain G^ and is contaminated by A W G N with two-sided P S D N /2 and is given by 0 P e T K = / Jo Q(V^l)Pr(dl)dl (5.9) Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 68 for B P S K modulation, where Q(z) = -^e^^dy. For non-coherent FSK (NCFSK) modulation PS? *' = [ = [G^] E/N where is the j 2 th 0 \e- vMl)dl- 3 3 (5.10) l/2 largest S N R out of h samples. For independent, identically distributed Rayleigh gain samples, the probability density function, of p^il) can be obtained using a standard result in ordered statistics [105] as Pro»(7) - WMI^y. ( - ~ °) ~ l e lh h •e-^-^e- 3 7 7 7 0 ' ( ' 5 n From (5.9) and (5.11), the single-user B E R when using B P S K may be written as P ^ S h K = (^jnyr JT <?(V27)- (! ~ e^) ~ h " i ^ ' j l h ° d ^ ( 5 ' 1 2 ) h-3 0 V 1 J (5.13) Integration by parts as pBPSK _ ft! v' 1 - ^ —I ! s / °° Q U ^ ) • | l - ^ ( ^ ) | d , 0 e 7 (5.14) ) Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 69 yields h-3 pBPSK e,j,h r _ h\ (h-j)!(j-l)\ sr ~i Z^i=0 (5.15) h i+j Using lno-1 withCT = 2{lJli+j) Jo (5.16) 2 in (5.14), we have 2 h\ S^h-j (-1)' 2 ( / i — — 1 ) ! Z^i=0 i + j pBPSK e,j,h 1 - . / lo+i+j -S (5.17) For the best S N R case (i.e., j = 1), pBPSK h Y^/I-I (-i) 70 j 70+J+l (5.18) From (5.10) and (5.11), the single-user B E R when using N C F S K may be computed as follows pNCFSK e,j,h ST^h-j, Z^i=0 V _ h\ — (h-j)\{j-l)\ r 1 L \i ) 70+2(i+j) • V * (5.19) J (5.19) can be simplified to NCFSK e,J,h. P. _ (5.20) h\(h-j + l)< + 2 nt J 0 '(^+i+j)' Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 70 For j — 1, we have the result given in [106], i.e. pNCFSK V = _ ( 5 2 1 ) 2^t (?+^) 1 For comparison purposes, we note that the B E R , P c o m ) , for conventional F H in Rayleigh fading on the uplink is given by [107] Pconv,BPSK = \{ 1 JT+YO) ^ ~ ^ + \ P H ( 5 ' 2 2 ) for B P S K modulation. For N C F S K modulation it is given by Pconv,NCFSK (1 ~ = Ph) + TT-P/i- 2 + 70 where P = 1 — ( l — 1 h 5.4.2 (5.23) 2 is the probability of a hit. Numerical Results of C A F H for r=l The B E R curves on an uplink channel for C A F H with r — 1 are compared with conventional F H in Fig. 5.2 assuming B P S K modulation. The curves are plotted using (5.4) assuming that the L = 32 subbands undergo independent Rayleigh fading. The results show that C A F H / B P S K can provide a much lower B E R than conventional F H / B P S K over a wide range of number of MS's. The B E R of C A F H / B P S K decreases rapidly with SNR. A t an average SNR, 7 o , value of 2 dB, C A F H / B P S K provides roughly a 100-fold Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 71 reduction in B E R compared to conventional F H / B P S K . The B E R difference between C A F H / B P S K and conventional F H / B P S K increases with 70. Fig. 5.3 shows the B E R curves of C A F H and conventional F H for N C F S K modulation. Comparing Figs. 5.2 and 5.3, it can be seen that C A F H / B P S K outperforms C A F H / N C F S K . The B E R of C A F H / B P S K is one to three orders of magnitude smaller than that of C A F H / N C F S K for a given SNR and a small number of MS's. With a large number of MS's, C A F H / B P S K provides roughly an order of magnitude reduction in B E R . C A F H / B P S K and C A F H / N C F S K performance differences tend to increase with SNR. Computer simulation results were obtained in order to ascertain the accuracy of the analysis and are shown as dotted curves in Figs. 5.2 and 5.3. It can be seen that the simulation curves agree closely with the (approximate) analytic results. It might be noted that the analytic B E R results are closer to the simulation results when the network is lightly loaded. The performance difference increases with the number of MS's. Fig. 5.4 shows the B E R of C A F H / B P S K for r = 1 and 2. A larger value of r results in a substantial decrease in B E R as the number of MS's increases. The performance difference between r = 1 and 2 increases with SNR. For example when K = 8 MS's are transmitting, the performance difference is roughly 3-fold, 4-fold, 10-fold and 40-fold when operating at an average SNR of 70 = 2,4,6 and 8 dB, respectively. The B E R curves of C A F H / N C F S K for r = 1 and 2 are shown in Fig. 5.5 and are qualitatively similar to the C A F H / B P S K curves. Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 5.4.3 72 B E R Analysis of C A F H for r >1 The C A F H B E R with r rounds can be obtained as PCAFH = Pi + (1 - Z ){P 1 + (1 - Z ){P 2 2 + (1 - Z ){ 3 3 • ' ' {Pr-1 + (1 ~ Z -i){P r r + (1 — Z )Prandom}} ' ' ' }}}, r (5.24) where Zj, Pj, 1 < j < r, and P ndom denote the probability that the target MS is selected ra in round j, the B E R when a subband is assigned to the target MS in round j, and the B E R when a subband is assigned to the target MS after round r, respectively. The terms Zj, Pj, and Prandom can be expressed as k=l K-j+l E (5.25) K~k I E HKn)P^\n)T^{n), j>l fc=l n=j—1 K P =I J Ei*(fe,o)p ,i,fct. fc=1 i A:-J+I e K-k E I E fc=l n—j—1 (5.26) ^(^njpw^rwwPe,! ,/s(L-n)) J' > 1 and random l ^ (5.27) respectively. For the sake of mathematical tractability, we assume in (5.27) that when a subband is not assigned to the target MS in round j < r, one of the subbands with the {r + 1, r + 2, • • -,K} largest gains is assigned at random to the target MS. This procedure Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 73 is only an approximation of the C A F H assignment scheme described in Section 5.3, of course. However, as our results will show, (5.27) is a very accurate approximation of the true error rate. In (5.25) and (5.26), T^\n) is the probability that the subband with the jth i g a r e s r j gain is available when n> j — 1 subbands have been occupied by MS's in the previous j — 1 rounds. T^\n) is given by Furthermore, the probability, $(fc,n), that k — 1 other MS's occupy the subband of the target MS in round j when n subbands have been assigned in the previous j — 1 rounds is given by ( K - T7. - 1 \ ®(k,n) In obtaining (5.29), it is assumed that the j th largest gain of each of the K — n — 1 other remaining MS's is randomly located in one of the L — n remaining subbands. In (6.25) and (5.26), P^\n) denotes the probability that n subbands have been assigned in the previous j — 1 rounds. P^\n) may be obtained recursively as n-l pU)(n) = P ~ \i)q{L-i,n-i), {i l j > 2, (5.30) i=j-2 with the initial condition P ( n ) = q(L,n), (2) > (5.31) Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 74 where q(l, m) is the probability that m subbands are assigned in the previous round (i.e., round j —1) when I subbands were available. The term q(l,m) can be expressed as follows m-2 M - 2 - £ ij M-m M-m+l—ii M—m+2—ii—i2 3 q{l,m) = =1 (*) E E »2 = 1 il=l M-l V M-l-ix J ••• E 13 = 1 M - 1- i - i x E !m-l=l \ ( 2 m-2 \ j=i V «2 ^m— 1 (5.32) where M = K — (L — I) is the number of unassigned MS's at the beginning of round j — 1. In (5.32), the summation indices ii,i2,' • ^hn-i represent the number of MS's that could be assigned to each of the first m — 1 subbands and the combinatorial terms represent the number of ways of selecting the MS's in each subband from the available set of unassigned MS's. Let be the random variable corresponding to the j largest gain out of h gain th samples. In (5.26) and (5.27), P ,j,h denotes the B E R of a M S that experiences a subband e gain G and is impaired by A W G N with two-sided P S D N /2. 5.4.4 Numerical Results of C A F H for r >1 ij) 0 The B E R curves for the uplink of a system using C A F H with r = 1 and 2 are compared to conventional F H in Fig. 5.6 assuming B P S K modulation and an average S N R of 7o = 2 dB. The results show that C A F H / B P S K can provide a much lower B E R than conventional F H / B P S K over a wide range of number of MS's. A t 70 = 2 dB, it is observed Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme that C A F H / B P S K can yield over a 100-fold reduction in B E R compared to conventional F H / B P S K . The curves are plotted using (5.24) assuming that the L — 48 subbands undergo independent Rayleigh fading. Computer simulation results were obtained in order to ascertain the accuracy of the analysis and are shown as dotted curves. It can be seen that the simulation curves agree closely with the analytic results. It is evident from Fig. 5.6 that for a fixed target B E R , the number, K, of MS's that can be supported by a r = 2 C A F H can be significantly larger than in the r = 1 case. At 70 = 2 dB, and a fixed target B E R of 10~ , it is found that the system capacity is 18 3 MS's for r = 2 compared to only 12 MS's for r — 1. Moreover, the B E R ratio between r = 1 and r — 2 increases with K, for K < 15. Beyond K = 15, the B E R ratio is fairly constant. The curves for 70 = 8 dB are plotted in Fig. 5.7 and are qualitatively similar to the 70 — 2 dB results. The B E R ratio between conventional F H / B P S K and C A F H / B P S K increases with 70. Fig. 5.8 illustrates the B E R versus average S N R for C A F H / B P S K with L = 48 subbands. Clearly, the performance difference between the different numbers of MS's {K = 1,10,20) increases with 7 o . The curves for C A F H / N C F S K are plotted in Fig. 5.9. It is evident that as 70 increases, so does the improvement offered by r = 2 compared to r = 1. Fig. 5.10 shows the B E R versus the number, r, of rounds for C A F H / B P S K with 70 = 8 dB and L — 48 subbands. As expected, the performance difference between the different rounds gradually decreases with r as the C A F H scheme reaches saturation at a B E R of 10~ and 2 x 1 0 8 - 6 for 10 and 20 MS's, respectively. 75 Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 76 To compare the performance of C A F H with an optimal allocation scheme, an exhaustive search method that searches over all the possible L(L—l)(L — 2)---(L—K+l) channel allocation permutations is used. The exhaustive search tries all the different channel allocation possibilities and selects the one that gives the lowest B E R , on which the current information bit is transmitted. It can be seen from the results in Fig. 5.11 that the difference in performance between C A F H and the optimal exhaustive search method is very small. However, the channel utilization of C A F H is much higher than that of the exhaustive search method. This can be explained by noticing that C A F H requires the knowledge of the channel gains at each information bit transmission, which requires a single channel measurement session per information bit, regardless of the number of MS's and number of channels. On the other hand, the exhaustive search method requires the knowledge of the B E R for each of the different channel allocation permutations. This requires L(L — 1)(L — 2) • • • (L — K +1) channel measurement sessions per information bit. Since the exhaustive search calculates the B E R , a sufficient number of information bits is required for transmission. Moreover, system delay should be considered. The time taken to measure the channel, and to transmit the information bits for B E R calculation must be less than the channel coherence time, i.e., constant channel behavior. This becomes extremely difficult to achieve for the exhaustive search method. The exhaustive search is a hypothetical approach used only for comparison purposes. Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 77 Chapter 10 5. A Channel Aware Frequency Hopping Multiple Access Scheme I 1 i 1 1 1 1 ' r Number of MS's Figure 5.2: B E R of conventional F H and C A F H w i t h B P S K modulation, L = 32. 78 Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 79 10 V- • -X • • • x • • * x•• - X 10" 10" 04 W CQ 10 10 3 -4 FH (Y = 6 dB) FH (yo - 8 dB) CAFH (Yo - 2 dB) CAFH (Yo = 4 dB) CAFH (Yo = 6 dB) CAFH (Yo = 8 dB) 0 10" 10 6 8 10 Number of MS's 12 14 16 18 Figure 5.3: B E R of conventional F H and C A F H with N C F S K modulation, L = 32. Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme • X- • r • •+ • r • Q • r r -*- r —t— r - B - r —•—r 6 8 10 Number of MS's 12 = = = = = = = = 1 ( =2dB) 1 (Y =4dB) l(y =6dB) 1 (Yo=8dB) 2(y =2dB) 2(Yo=4dB) 2(Yo=6dB) 2(Yo=8dB) 80 Y o 0 0 - 0 14 16 Figure 5.4: B E R of C A F H with B P S K for r = 1 and 2, L = 32. 18 Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 81 10 10 % 10" Pi w 10 -4 r r D r r -*- r r r r • • X' • • •+ 10 10' • 1 6 8 10 Number of MS's 12 = 1 (To= = KYJ = = Hl6 = = 1C» = 2(7, = - 2 ^ = = 2(16 = = 2(7) = i 14 2 dB) 4 dB) 6 dB) 8 dB) 2 dB) 4 dB) 6 dB) 8 dB) i 16 Figure 5.5: B E R of C A F H w i t h N C F S K for r = 1 and 2, L = 32. . 18 Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 82 10 —*— -eo — Conventional F H C A F H <>=1) Simulation C A F H (r = C A F H ( r = 2) Simulation C A F H (r = 2) Number of MS's Figure 5.6: B E R of conventional F H and C A F H with B P S K modulation. -y = 2 dB and L = 48. 0 Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 83 10 Number of MS's Figure 5.7: B E R of conventional F H and C A F H with B P S K modulation. 70 = 8 dB and L = 48. Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 84 Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 85 Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 86 Number of rounds, r Figure 5.10: B E R vs. the number of rounds, r, of C A F H w i t h B P S K . 70 = 8 d B and L = 48. Chapter 5. A Channel Aware Frequency Hopping Multiple Access Scheme 87 Number of MS's Figure 5.11: B E R versus number of MS's of random allocation, C A F H with B P S K for r = 1 and 2, and exhaustive search method. 70 — 4 dB and L = 8. 88 Chapter 6 O r d e r Statistics for C o r r e l a t e d Fading Channels 6.1 Introduction In the design of wireless communication systems, much research effort has been devoted to the study of diversity schemes for mitigating the effects of channel fading and improving the received SNR. The theory of order statistics underpins the performance analysis of many diversity reception techniques. Selection combining (SC) [108] is a simple diversity scheme in which the signal from only one of the diversity branches (the branch with the largest SNR) is processed. In signal detection and estimation problems, the cdf of the r-th order statistic is often required. For example, it is useful in calculating the performance loss when the receiver makes an error in selecting the largest S N R branch. It is also applicable to the calculation of the average symbol error rates for various modulation schemes. Most studies on the ordering of channel gains assume that the fading processes on the different branches are uncorrelated. This is partly due to the lack of a simple procedure °A version of this Chapter has been published [86]. Chapter 6. Order Statistics for Correlated Fading Channels 89 for analyzing order statistics for fading envelopes with arbitrary cross-correlations. For correlated fading, expressions for the pdf of the highest S N R for two and three branch diversity systems are available [109, 110, 111, 112]. For a larger number of diversity branches, [113] and [108] present an expression for the correlated SC output pdf. In [114], the performance of a hybrid selection/maximal-ratio combining ( H - S / M R C ) diversity scheme over an exchangeable correlated Nakagami fading channel is analyzed. 1 . Expressions for the cdf of the r-th order statistic, r = 1, 2 , . . . , n, for correlated Nakagami fading branches are not available in the literature except for r = n (i.e. the largest order statistic). In this chapter, we apply a result from the theory of order statistics [115] for evaluating the cdf of the r-th order rv out of a set of n arbitrarily correlated rv's to the case of correlated Nakagami-m fading. Specifically, numerical results are obtained which illustrate the effects of the fading parameter m and the branch correlation matrix. For a large number of branches, the numerical procedure can be quite computationally intensive. Thus, the accuracy of using a simplifying exchangeable approximation is also investigated. 6.2 Preliminaries Let gi,- • -,g n be n arbitrarily correlated rv's which represent the branch gains in a diversity communication system. If g • • -,g 1: n are arranged in increasing order of their magnitudes and written as 9\:n X nn < 9n:n, (6-1) A set of random variables A\, • • •, A are said to be exchangeable if the joint distribution of A , • • does not depend on the permutation ir. n •,A <••• Wl Chapter 6. Order Statistics for Correlated Fading Channels 90 we refer to g , as the r-th order branch gain. In this chapter, g\,---,g are assumed to be r:n n statistically dependent and not necessarily exchangeable branch gains. The corresponding instantaneous branch SNR's are denoted 71,72, • • •, 7n = [ ] E/N ,i 2 7 l gi 0 with = l,2,...,n (6.2) where E is the transmitted bit energy and No is the one-sided noise PSD. The r-th order branch SNR is l r : n = {9r:n] E /7V . (6.3) 2 0 The joint cdf of , • • - ,j n 7 l can be expressed as [116, p. 140] where (p{t\, • • -,t ) is the joint characteristic function (CF) of 71, • • - , 7 . Hence, the cdf n n of the highest order statistic 7 n : n = max{7i, • • - , 7 „ } can be written as (6.5) (6.6) = Pr{all 7i < 1- 1 (27r)" j_ e - j t k v dt\ • • • dt , r fc=i (6.7) Chapter 6. Order Statistics for Correlated Fading Channels 91 6.3 6.3.1 cdf of 7 r : n Correlated Fading In this section, we derive an expression for the cdf, F 7 r ; n , of the r-th order branch SNR, 7 , in terms of {F .(v), j = r , r + 1, • • • , n}. Recall that the probability, p , , of the lj: r:n r n occurence of at least r out of n events {Ai, • • -,A } is given by [117] [11, pp. 109] n Pr,n = 2(-1)J'"rCll)^'r = 1'2'"- ^ - (6 8) where Sj= ]T P l '{-'' ^--- A, }. 1 : i (6.9) 1 <ii <---<i-j<n In (6.9), the summation is over all different ways of selecting j out of n rv's. Letting Ai,i = 1, • • -,n denote the event {% < v}, (6.9) can be re-written as l<ii<--<ij<n Pr{max(7 ---,7i -)<«} £ i l ) J ( - ) 6 n l<ii<---<i -<n 3 _ Yl 1 < i i <•• i M - x r " ' ^ ( ; ) ( 6 - 1 2 ) <ij<n Kii<-<ij<n In (6.13), the sum is over all possible ways of selecting j out of n rv's and the superscript notation in F^f'' ^ 1 indicates that only j iit • • ,7^. are included in the selection. It is Chapter 6. Order Statistics for Correlated Fading Channels 92 assumed that the sum is equal to F {y) lrvn using (6.7). From the definition of A i} for j = n. The term F-£- we can write Pr,n = Pr{7l:n <V,-- - , 7 = Pr{ = F (f) is obtained 7 r : n 7 r : r:n < v} (6-14) < v} (6.15) ». (6.16) Following the result in [115, p. 99], and substituting (6.13) and (6.16) into (6.8), we can obtain the cdf of the r-th order statistic KJv) = t \<i\<---<ij <n ly.j (6.17) When n is large, the evaluation of (6.17) may be time-consuming. However, if the 7J'S are exchangeable, (6.17) can be easily evaluated as it reduces to KJV) = B-ir ( : r J r J) (6.i8) Chapter 6. Order Statistics for Correlated Fading Channels 93 6.3.2 Independent Fading Suppose that gi,---,g are n independent, b u t not necessarily identically distributed rv's, n (6.7) simplifies to Fjn:n(v) = ^ {ll , ' ' ', In < V} = Pr{ 7 l <v}--- Pr{ 7 n (6.19) < v} k=l B y substituting (6.19) i n (6.17) we get E (-iy (6.20) j-i r- 1 E l<ij+l<---<in<n UF% '- M +I M k=\ If the channel gains are now assumed to be independent, identically distributed (i.i.d) rv's, using the property of exchangeability, (6.20) reduces to * V : » = (6.21) i > i : 3=r 6.4 N u m e r i c a l results To illustrate the application of the results, we evaluate the cdf of the r - t h order statistic for correlated N a k a g a m i - m fading branches. I n this case, the joint C F is given by [118, p. 359] t) = |I-jTS|n m (6.22) Chapter 6. Order Statistics for Correlated Fading Channels 94 where I is the identity matrix of size n x n, | • | denotes the determinant, T = diag{t t , • • 1; 2 •,t }, m € [0.5,oo), and S is a symmetric matrix with elements n (6.23) In (6.23), Rk,e = Cov(7fc,7^) is the element in row k and column I of the correlation matrix R. Of particular interest for wireless communication systems are the values m = ~, m = 1, and m > 1. The Nakagami distribution is a one-sided Gaussian distribution (severe fading) for m = |, a Rayleigh distribution for rn — 1 , and a good approximation to a Rician distribution when m > 1. 6.4.1 Constant Correlation M o d e l The constant correlation model, discussed in [119, 120, 121], may be used either to approximate closely spaced antennas or to perform a first-order analysis using the average value of the correlation coefficients for all the off-diagonal entries of the correlation matrix [121]. Suppose we have a diversity receiver with an array of n = 3 antennas placed at the vertices of an equilateral triangle. Then the branch gains are exchangeable Nakagami-m rv's and the cdf can be obtained using (6.18). As an example, we use the correlation Chapter 6. Order Statistics for Correlated Fading Channels 95 matrix from [121] 1.0 0.6 0.6 0.6 1.0 0.6 0.6 0.6 1.0 (6.24) where 7 denotes the average SNR. Fig. 6.1 shows the cdf of the r-th order statistic in a correlated Nakagami-m fading environment with the correlation matrix in (6.24), m — \ and 7 = 1. Computer simulation results (shown as dotted curves) were used to verify the analysis results (solid curves). As to be expected, it can be seen that a lower order statistic, e.g. 71:3, has a steeper curve than a higher order statistic, e.g. 7 3:3 . In a SC system, this provides the performance loss when the receiver incorrectly selects a branch with a lower gain instead of the highest gain branch. Fig. 6.2 shows the cdf of the r-th order statistic in a correlated Nakagami-m fading environment (solid curves) with the correlation matrix in (6.24), m — 1 and 7 = 1. The cdf curves for independent fading (dotted curves) are obtained using (6.21) and are shown for comparison. When the highest SNR branch is selected (i.e. 73:3), the performance is better when the branches are independently fading; when the lowest SNR branch is selected (i.e. 7 i ) , correlated fading yields better performance. :3 The cdf of 7^3 is plotted in Fig. 6.3 for different values of m in correlated fading using the correlation matrix in (6.24). It can be seen that the performance degrades as fading severity increases (i.e. m decreases). Calculating the cdf of the r-rth order statistic can be applied to evaluating the average Chapter 6. Order Statistics for Correlated Fading Channels 96 error performance of various modulation schemes in correlated fading channels. In diversity applications, this provides the performance loss when the receiver incorrectly selects a channel with a lower gain. In multiuser diversity applications, this is a good measure for the channel gain loss from one user to another, which implies a total throughput improvement when selecting only the user(s) with stronger instantaneous SNR. 6.4.2 Linear M o d e l This type of correlation model applies to linear antenna arrays. We consider the correlation matrix example for a five-element array in [122] R = l\ 1.000 0.795 0.605 0.375 0.795 1.000 0.283 0.795 0.605 0.375 0.605 0.795 1.000 0.795 0.605 | - 0.375 0.605 0.795 1.000 0.795 0.283 0.375 0.605 0.795 1.000 (6.25) The correlation coefficient value decreases as the distance between array elements increases. In this case, the channel SNR rv's are not exchangeable and the cdf has to be obtained using (6.17). Using the correlation matrix in (6.25), Fig. 6.4 shows the cdf for 7i:5,72:5, • • • ,75:5 with m — 1 and 7 = 1 (solid curves). Approximate cdf curves (dotted curves), obtained using (6.18) and a 5 x 5 exchangeable matrix in which an average value of 0.514 is used for all off-diagonal entries, are also plotted. It can be seen that there is a significant difference between the exact and exchangeable approximation cdf curves for Chapter 6. Order Statistics for Correlated Fading Channels 7i:5 and 7 5:5 . 97 Chapter 6. Order Statistics for Correlated Fading Channels Figure 6.1: P l o t of cdf of 71:3,72:3 and 7 and simulation (solid). 3 : 3 w i t h m = 1 and 7 = 1 : 98 theoretical (dotted) Figure 6.2: Plot of cdf of 71:3,72:3 and 7 with m = 1 and and correlated (solid) fading channels. 3 : 3 7 = 1: independent (dotted) Chapter 6. Order Statistics for Correlated Fading Channels 100 v Figure 6.3: Plot of cdf of 7i 3 with 7 = 1 for different values of m. : Chapter 6. Order Statistics for Correlated Fading Channels 101 Figure 6.4: Plot of cdf of j with m = 1 and 7 = 1 for a linear model (solid) and its exchangeable approximation (dotted).. r : 5 102 Chapter 7 E r r o r Analysis of O r d e r Selection in Correlated Fading Channels 7.1 Introduction In order to increase system's capacity and to improve the offered QoS in wireless communication systems, several techniques have been proposed to mitigate the effects of channel fading and to improve the received SNR. Such techniques are diversity reception, dynamic channel allocation, and power control. The theory of order statistics underpins the performance analysis of many diversity reception techniques that involve efficient channel allocation and signal processing algorithms for signal detection and estimation. Expressions for the cdf of the r-th order statistic, r = 1,2, . . . , L , for correlated Nakagami fading branches have appeared in [86]. In this chapter, we apply a general result from the theory of order statistics [86] and [115] to evaluate the B E R of a channel aware diversity system, hereafter referred to as order selection, when the r-th order branch is selected for transmission when using B P S K modulation. The expressions derived are then used to evaluate the error performance for C A F H . The C A F H evaluation is carried out assuming correlated Nakagami-m fading when the coherence bandwidth notably exceeds Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 103 the transmission bandwidth, which is the real scenario in practical C A F H systems. 7.2 Channel and System Model Let <7i,<72, ' , 9 i be L arbitrarily correlated rv's which represent the branch gains in - - a diversity communication system. 7i>72,' • • ,7i» w i t The corresponding branch SNR's are denoted by h ^{ } E/No,i = l,2,...,L 2 7i gi (7.1) where E is the transmitted bit energy and N is the one-sided noise P S D . Consider an 0 L-branch diversity receiver of a cellular system in which K MS's transmit over L, L > K, branches to the BS. The received base-band signal on the 7r -th branch of the /c-th M S fc at the BS is ^ (t) = ^ s ( t ) e - ^ + n ( t ) , fc (7.2) f c where Sk(t) is the transmitted signal of the /c-th MS and n(t) is independently identically distributed A W G N with zero mean and one-sided spectral density N . The random phase 0 ' ipn is uniformly distributed over the range [0, 27r) and g„ is the random magnitude of the k k 7Tfc-th diversity branch gain. Here we model the fife's as correlated rv's with a marginal Nakagami pdf given by [123] J9n \ k ) r(m„ ) V "fc/ k = l,2,--;K fi fc where the fading parameter m Wk j- 3^ 7 is assumed to be a positive integer, T(-) is the Gamma Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 104 function defined by T(p)= J °° v ~ e~ dv, and Cl = E[(g ) }, E[-] being the expectation p l v 2 Q Vk nk factor. Let j „ k denote the instantaneous signal-to-noise ratio (SNR) on the ^ - t h branch, expressed as 7x = ^r{9, )\ fc N,o (7.4) k where E/N is the ratio of the bit energy to the Gaussian noise spectral density. It is 0 well known that the marginal pdf of 7 ^ follows the Gamma distribution f () v = 211^." -m* (vfrV e k v > (7.5) 0 where the average received S N R on the 7Tfc-th branch is given by 7 ^ =E[7 J=^|fi .. ?R 7.3 7r( Error Performance for Order Selection B P S K If the rv's <7i, • • •, are arranged in increasing order of their magnitudes and written as 9l:L <•••< 9L:L, (7.6) we refer to the r-th order rv, g , as the r-th order statistic (r = 1, • • •, L). In this section, r:L gi, • • •, gL are assumed to be statistically dependent and not necessarily exchangeable branch gains. Accordingly, the SNR of the r-th order statistic is then 7r:L = [g : } E/N . 2 r L 0 (7.7) Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 105 The average error probability when transmitting on the r-th order branch out of L branches for a B P S K modulation over an A W G N channel with two-sided P S D N /2 is 0 obtained by averaging the conditional error probability over the pdf of j -i, i.e., r poo Pl = L (7.8) \ P(e\v)f Jv)dv, Jo lT where P(e\v) is the conditional error probability. B y differentiating (6.17), we can write the pdf of 7 £ as r: U M = dv L = E •i^Cll) i=r L = E (-i^C-i) ^ (7.9) dv (fl,•••,*•<)/ U- E l<7Tl<-<7rj<i, where the pdf of the largest order statistic, U M E l<7Tl<-<7Ti<L ^L-.L, = j^yz JZo • • • JZo^i, out of L rv's is given by [108] • • ;t )h{y,t)dti L • • • dt , (7.10) L with h(v,t) T-T / \-e-3H v dv hut (7.11) E(-i) n=l n+1 j(b t +-+b t ) exp(jv\biti + ---+b t ]) E bH 1 1 L L L L \-b =n L Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 106 Inserting (7.9) into (7.8) yields p ; : L = E ( _ 1 ) i - ^ - l ) E l<7ri<---<7Ti<i x H ' '• • •,U )w^-^KU ,- I-oo^ '-^^1 i 1 • ;UAdh • • • dt ni (7.12) with 1 P(e\v)h '-'^ {v,t)dv. {7I1 (7.13) ) (27T)* Jo In (7.12) the integrand is decomposed into two factors. T h e first factor 4>(t) is the C F , dependent solely on the channel characteristics. T h e second is the weighting function w(t); it depends solely o n the m o d u l a t i o n scheme. T h e conditional error probability for B P S K is given by [89] P(e| ) = Q(v^7)- (7-14) 7 It follows from inserting (7.14) into (7.13) that POO w 7Tl ) ' ' ') t-JTi ) / Q(V2v)h " {v,.t)dv. Jo {nw (2TT)< (7.15) ,nt) Integration by parts as (Tl,-,*i)(t w •• • t ) 1 x E(-i) e=i n+l ^! '" S: j /o QW*J )• E favrj H 00 1-6*4 =n iri + exp(jv[b t ni n +b,ri H ds VK^]) (7.16) „dv. Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 107 yields UUU) w x E(-i) n + 1 n=l bit, H 1 Eh 6 . =n w 2 V5F 1 P O O - « ( + J ' l ' i r i 'irj "I 1 Ji) l e H>7rj'irj]) (7.17) L e t t i n g x — y/v, we o b t a i n T w \ 7Tl u , J 7T] / 1 (2TT)' -1 UUU) x E(-i) n + 1 n=l L J"°° -a; (i+jK^i+ -+^^1)^. i bjrjH 2 e H%j=n (7.18) Using (7.19) in (7.18) w i t h u 2 = 2 ( l + j [ 6 ^ + - - - + f c „ t . ] ) » we have Tl 1 i 1 (2^ w T uuu) x E(-i) n + 1 n=l • E birjH I ^/l+jffcir^TTjH 1-6^=" hbirjtTTj] (7.20) The error performance can then be obtained by inserting i f / " ' ' " ' " " ^ ^ , , • • 71 (7.20) and <j>^<"'^(£ , W1 • • -,t ) Wi 1 shown i n for a given operational environment into (7.12). Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 108 7.4 Numerical Results and Discussion In this section, numerical examples are provided for several correlation models, well known from practical diversity systems, i n order to provide the applicability and check the usefulness of the proposed analysis. W e consider the uplink of a cellular system i n which a M S transmits over L — 8 subbands to the B S . For a given M S , we assume that the various subbands are subject to correlated fading, where the amount of correlation depends on, among other things, the subband frequency separation. T h e correlation at any instant of time between the fade envelopes, gi and gj of the i th and j th subbands, respectively, is taken to be [124] 1 E{gi,gj} l + where B c ( ^ ) (7.21) 2 is the coherence b a n d w i d t h of the channel. Therefore, the L x L covariance m a t i x R is given by 1 Pl,2 Pl,L P2,l 1 P2,L Pl,l PL,2 1 R = 7 Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 109 where Pi,i = 1, 1 < i < L, and R is symmetric and positive definite [106]. For correlated Nakagami-m fading channels, the joint C F is given by (6.22). We shall assume that each subband experiences the same extent of fading, i.e., me — m for £ — 1, 2, • • •, L. 7.4.1 B E R evaluation for Order Selection B P S K Substituting the component correlation coefficients into the joint C F expression (6.22) and using B E R expression (7.12), we will see how those parameters affect B E R performance in a correlated Nakagami-m fading channel. To illustrate the application of the presented mathematical model, we evaluate the B E R , Pg , L when transmitting on the r-th order subband out of L available subbands for a B P S K modulation in m — 1 correlated fading environment. In Fig. 7.1 we select B n » — B /B s c = 0.9, where B n is the subband bandwidth nor- malized by the coherence bandwidth of the channel, and the bandwidth of each subband is given by B . s Fig. 7.1 shows the B E R , P subbands . In Fig. 7.1, P r L e r : i e , of B P S K when r = 8, 7, 4,1 for L = 8 is plotted versus the S N R in correlated fading environment (solid line). In the same figure, P r h e is depicted for an uncorrelated fading environment (dotted line). For comparison, the B E R of conventional B P S K without ordering is plotted (dashed line). When the highest S N R subband is selected (i.e. 78:s)> the performance is better when the subbands are independently fading. On the other hand, if the lowest S N R subband Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 110 is selected (i.e. 7i s), correlated fading yields better performance. It can also be seen : that the performance difference between correlated and uncorrelated fading subbands increases with r. A t a S N R of 10 dB, operating in an uncorrelated fading environment provides roughly a 20-fold, 7-fold and 2-fold reduction in B E R compared to a correlated case when operating at r = 8,7 and 4, respectively. A t r = 1 the performance difference is quite negligible. To demonstrate the impact of the fading correlation to the error performance, B = n 0.5 is chosen in Fig. 7.2. It is evident that the effect of the correlation to the B E R performance is greater for larger correlation coefficients. As expected, the B E R worsens for smaller B values when the correlation coefficient increases. For example when transn mitting at a S N R of 10 dB, the performance difference is roughly 100-fold, 30-fold and 4-fold between uncorrelated and correlated fading channels at r = 8, 7 and 4, respectively. At r = 1 the performance difference is more noticeable. Following the same trend, the B E R for B = 0.25 with even a larger performance difference between the corrrelated n and uncorrelated cases is depicted in Fig. 7.3. 7.4.2 Application: B E R evaluation for C A F H The presented C A F H analysis is carried out assuming correlated Nakagami-m fading when the coherence bandwidth notably exceeds the transmission bandwidth, which is the real scenario in practical C A F H systems. The B E R for C A F H with one round can Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 111 be calculated as [83] L-1 1 PCAFH )K-l — E (7-22) i=L-K where the first term in (7.22) represents the B E R for the case when a subband is assigned to the target M S in round 1. Accordingly, when the target M S was not considered in round 1 (i.e., a MS with a larger gain occupied the same subband as target MS), the target MS is chosen by the BS in a random order and is assigned to an unoccupied subband, this constitutes the second term in (7.22). And fc-i K-k (7.23) represents the probability that k — 1 other MS's occupy the subband of the target M S in round 1. To illustrate the sensitivity of C A F H to the fade envelopes correlation, Figs. 7.4 and 7.5 plot the B E R performance for C A F H versus number of MS's for B n and B n — 0.25, respectively when L = 8. performance for B n = 0.9 For smaller 'L, Fig. 7.6 plots the B E R = 0.33 and L = 3. It is observed that the performance difference between correlated and uncorrelated fading channels increases with 7. For a fixed 7, the performance difference decreases with the number of MS's. Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 112 BPSK with ordering (uncorrelated fading channels) I 0 i 2 i i i 4 6 8 Average S N R (dB),Y Figure 7.1: B E R vs. average SNR for B = 0.9. n I 10 Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 113 1(T Pe' 10 - I t - - - - - 10" ,4:8 Pi H 10 -3 PQ 10" Conventional BPSK without ordering 10" BPSK with ordering (correlated fading channels) BPSK with ordering (uncorrelated fading channels) 10 -6 0 2 4 6 8 Average SNR (dB), y" Figure 7.2: B E R vs. average SNR for B n = 0.5. 10 Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 10 U 10" 10" Pi w 10 „7:8: -3 ,7:8 CQ p4:8i 10" Conventional BPSK without ordering 10" BPSK with ordering (correlated fading channels) BPSK with ordering (uncorrected fading channels) 10" 0 2 4 6 8 Average S N R (dB), J Figure 7.3: B E R vs. average SNR for B n - 0.25. 10 11 Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 115 10 u I I I I I 1 10" 10 w CQ 10 10 -2 7= 2 dB 7= 8 dB -3 ^ -4 7=8dB 10 -5 correlated fading channels uncorrelated fading channels • 10" 1 1 1 • i 3 4 5 6 i Number of mobile stations Figure 7.4: C A F H B E R vs. number of mobile stations, for 7 = 2 and 8 d B , and B n = 0.9. Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 116 10 u I l" 1- ' y= 1 1 1 1 1 2 dB ; 10" pi 8 CQ -3 10 7= 8 dB 10" 7=8dB correlated fading channels 10"' uncorrelated fading channels 10" 1 1 I 3 ' 4 i I 5 Number of mobile stations • i 6 Figure 7.5: C A F H B E R vs. number of MS's, for 7 = 2 and 8 dB, and B = 0.25. n Chapter 7. Error Analysis of Order Selection in Correlated Fading Channels 117 10 I I I . 7= 2 dB y=10dB 7=10dB correlated fading channels uncorrelated fading channels i N u m b e r o f m o b i l e stations Figure 7.6: C A F H B E R vs. number of MS's, for 7 - 2 and 10 dB, and B n = 0.33. 118 Chapter 8 Conclusions and Future Research We conclude this thesis with a summary of our contributions and suggestions for future work. 8.1 Summary The main objectives of this thesis were: • The proposal of new methods for the improvement of M C - C D M A as an effective signaling scheme. • The development of new channel aware systems with adaptive subchannel allocation. • The investigation of statistics of general order selection in correlated Nakagami fading channels. The following is a summary of Chapters 2-7. In Chapter 2, we have presented a new M C - C D M A scheme, matched M C - C D M A . In matched M C - C D M A , preference is given to the use of subcarriers with low channel attenuation. The B E R as a function of the number of users and the SNR for the proposed scheme and conventional M C - C D M A Chapter 8. Conclusions and Future Research 119 was investigated. It was found that matched M C - C D M A provides a lower B E R . The performance difference between matched and conventional M C - C D M A increases as the target B E R decreases. The results in this chapter also have appeared in [80]. The performance of a new F H - M C - C D M A scheme for transmission over a Rayleigh fading channel has been analyzed in Chapter 3. The proposed system can combine efficiently the techniques of slow F H , O F D M , and M C - C D M A . Both F H and M C - C D M A can be viewed as special cases of the proposed scheme. The performance of F H - M C C D M A using E G C and M R C was evaluated and compared with those of corresponding F H and M C - C D M A systems. The results show that F H - M C - C D M A is an attractive wireless multiple access candidate, which can be integrated into existing 2G and 3G C D M A systems. F H - M C - C D M A provides a lower B E R than conventional F H and a comparable B E R to M C - C D M A while employing a lower number of subcarriers. This can avoid some of the disadvantages of M C - C D M A . Results of this work have been published in [81]. Chapter 4 has presented a detailed architecture for a modified F H - M C - C D M A system as a new transmission scheme for wireless communications. We propose a modification of the F H - M C - C D M A scheme in [81] that is suitable for use in correlated fading channels. Both F H and M C - C D M A can be viewed as special cases of the proposed generalized F H - M C - C D M A when the number of subcarrier in a group j , N j G respectively. It is found that N j = 1 and N G GJ = 1, and NQ — s, 0 = s are not favorable choices for NQJ- Optimum error performance is achieved by varying N j between 1 and s while keeping G the number of MS's in a group j, M Gj = 1, V j . The B E R performance of F H - M C - Chapter 8. Conclusions and Future Research 120 C D M A using M R C on an uplink correlated Rayleigh fading channel has been compared with that of F H and M C - C D M A , respectively. The results show that F H - M C - C D M A can achieve a much lower B E R than F H and M C - C D M A whilst employing a smaller number of subcarriers in a group. This work has been published in [82]. In Chapter 5, we have presented an effective access scheme for slowly time-varying, frequency selective channels. A C A F H multiple access scheme was proposed for independently Rayleigh faded subbands. C A F H involves the ordering of subband gains of the MS's. In C A F H , the channel status is monitored at the base station (BS). The BS then determines the subband gains for each M S and uses this knowledge to assign to each MS the subband that enjoys the highest gain,'while ensuring that no more than one M S is assigned to the same subband. It was shown that C A F H can provide a much lower B E R than conventional F H over a wide range of S N R values and number of MS's. This is achieved by exploiting information about the MS's subband gains. The performance difference between conventional and C A F H increases with SNR. For a fixed target B E R , the number of MS's which can be supported by C A F H is much larger than that with conventional F H . This work has also appeared in [83] and [84]. The performance of C A F H with r rounds over slowly time-varying, frequency selective channels is analyzed. In [83], the analysis was limited to just one round. In this chapter, closed-form expressions that are recursively-based are derived to evaluate the B E R for C A F H with r rounds and its performance in a cellular communication system is studied. The B E R improvement achieved when using r > 1 is depicted. Numerical results show that the improvement obtained using r > 1 is considerable compared to the r = 1 case. Moreover, Chapter 8. Conclusions and Future Research 121 it is found that C A F H reaches saturation at larger r values; the dependence of the system performance on r is less significant for large r. This work has been reported in [85]. In chapter 6, the cdf (and hence outage probability) of the r-th order branch SNR in correlated Nakagami—m fading has been studied. The accuracy of a simple exchangeable approximation to reduce the computational load has been examined. Results from this work has been reported in [86] and [125]. Chapter 7 investigates the effects of correlation on the error performance of channel aware systems that involve the ordering of channel gains in correlative Nakagami fading environments. A useful analytical formula for the B E R of the r-th order statistic of a set of arbitrarily correlated and not necessarily exchangeable diversity branch gains is described and may be efficiently applied in the performance analysis of various diversity systems. In the case of coherent B P S K , a closed-form expression is extracted for the B E R when transmitting on the r-th order branch. This result is then employed for the B E R calculation of C A F H . Depending on the level of correlation, it can be observed from the results in Figs. 7.1, 7.2, and 7.3 that the performance over the r-th order branch is significantly degraded under correlated fading conditions when compared to the uncorrelated fading case, as a consequence, the average B E R performance of C A F H is noticeably affected. This work has been submitted for publication [126]. In conclusion, among the main contributions of this thesis, we proposed: • Matched C D M A and F H - M C - C D M A for wireless communications. • Channel aware frequency hopping ( C A F H ) as an effective access scheme for correlated and independently faded subbands. Chapter 8. Conclusions and Future Research 122 • B E R of order selection when the r-th order branch is selected for transmission. 8.2 Future Work We present below a list of research problems that can be investigated as possible extensions for research work reported in this thesis. 1. The use of multiple antennas to provide antenna diversity over a point-to-point wireless link has been studied extensively since the early 1960s. It is still a popular research topic for improving energy efficiency through the use of both transmit and receive diversity. Recently, with the increasing demand for high-speed wireless data services, antenna diversity in a multiuser data network has received particular attention [127] [128], as it requires a broader multilink perspective for the spatial diversity gain in the system [79]. Recent progress in information theory [49] [129] has shown that in a data network with multiple active users requesting services simultaneously, multiuser diversity can be obtained by a packet scheduler, which always allocates the common radio resource to the user having the best channel quality, on the condition that the instantaneous user channel quality information (CQI) in terms of S N R is available to the scheduler. Multiple transmit antennas can be used either to obtain transmit diversity or to form M I M O channels. We may refer to this multiuser scheduling scheme as greedy scheduling. Adaptive loading (or adaptive modulation) technique is an efficient technique to achieve power and rate optimization based on the sub-channel gains. Adaptive bit allocation using the simple Greedy method gives an optimal solution for a single user system. Adaptive Chapter 8. Conclusions and Future Research 123 bit allocation using the Greedy method cannot give an optimal solution for multiuser cases. It is probable that the sub-channels with the largest channel gain for one user also be the largest for another user. A practical channel-aware scheduling algorithm, proportional fair (PF) scheduling [79] [130], has been used in the downlink of cellular packet data system IS-856 [131] [also known as l x E V - D O or high data rate (HDR)], which trades some total system throughput for resource fairness among the active users. When users cannot share the same sub-channel at the same time, it is a difficult problem to find the best way to assign the sub-channels to the users. It would be interesting to study the improvement in average throughput due to multiuser diversity in M I M O cellular systems by assigning the transmit antennas to the users using the channel aware algorithm described in Chapter 5. 2. Multiuser diversity helps on two accounts. First, one can pick the user(s) that is (are) in a favorable channel condition, thereby increasing the network throughput for the forward link. Second, one can exploit the multiuser diversity to decrease feedback requirements in the reverse link. This novelty has been introduced in [79], where the transmitter pre-codes in a pseudo random fashion and that user is picked whose ratio of instantaneous data rate to its own average data rate is the largest. This idea can be extended to include users with different ratios. Hence, studying the effect of choosing users with different S N R levels on the network throughput. Using the concepts covered in Chaper 7, expressions for the B E R and throughput may be obtained. 3. High P A P R of the transmit signal is a major drawback of multicarrier transmis- Chapter 8. Conclusions and Future Research 124 sion such as O F D M , M C - C D M A and discrete multitone (DMT). It is possible to avoid high P A P R signals by employing a technique named clustered O F D M [132] [133] [134]. In this technique the subcarriers are clustered into several smaller blocks and transmitted over separate antennas. The P A P R is reduced since there are fewer subcarriers per transmitter. F H - M C - C D M A is another method that may be applied to P A P R reduction in multicarrier systems, as F H - M C - C D M A generally uses a smaller number of subcarriers. Since F H - M C - C D M A requires a single antenna for transmission, it may provide an effective method for reducing both cost and complexity when compared to clustered O F D M . 4. Multiple transmit and receive antennas can be used to improve the performance and increase the capacity of wireless communications systems. It is shown that when multiple transmit and receive antennas are used to form a M I M O system, the system capacity can be improved by a factor proportional to the minimum of the number of transmit and receive antennas compared to a single-input single- output (SISO)' system over a flat narrowband Rayleigh fading channel [135] [136]. However, for wideband channels O F D M has to be used with M I M O techniques for intersymbol interference mitigation and capacity improvement. This M I M O - O F D M system is investigated in [137] [138] [139]. A n alternative to O F D M is F H - M C - C D M A that can be applied to M I M O technologies. In principle, F H - M C - C D M A and M I M O can be synergistically integrated to offer the benefits of both system simplicity and high performance. M I M O - F H - M C - C D M A could be effective in reducing ICI as well as in obtaining a diversity gain even for highly-correlated fast fading channels. Since Chapter 8. Conclusions and Future Research 125 M I M O - O F D M systems are based on O F D M , they also suffer from the problem of inherent P A P R . M I M O - F H - C D M A can reduce such drawbacks. T o deal w i t h the increasing number of system users a n d fading correlation, a subcarrier allocation scheme m a y be applied t o M I M O systems w i t h channel state information (CSI) at the transmitter or receiver. C A F H for M I M O channels ( M I M O C A F H ) can be considered for wideband transmission to mitigate intersymbol i n terference a n d enhance system capacity. Studying the capacity of M I M O enabled C A F H system under a generic multiuser multicarrier framework m a y reduce I C I caused by high-speed mobiles i n cellular environments. M u l t i p l e transmit a n d receive antennas can be used w i t h C A F H to further improve system performance. Using M I M O - C A F H , P A P R related problems can be d r a m a t i c a l l y reduced. 126 Bibliography [1] S. Hara and R. Prasad. Multicarrier Techniques for 4G Mobile Communications. Artech House, Norwood, 2003. [2] ETSI GSM Recommendations, 05 series. Sophia Antipolis, France. [3] M . Mouly and M . - B . Paulet. The GSM system for Mobile Communications Palaiseau. Published by authors, France, 1992. [4] Q. Bi, G. I. Zysman, and H. Menkes. Wireless mobile communications at the start of the 21 century. IEEE Communications Magazine, 39:110-116, January 2001. [5] TIA/EIA/IS-95. Mobile base-station compatability standard for dual mode wideba spread spectrum cellular system. July 1993. • [6] D. J. Goodman. Trends in cellular and cordless communications. IEEE Commu- nications Magazine, 29:31-40, June 1991. [7] ETSI UMTS (TR-101 112), V 3.2.0. Sophia Antipolis, France, April 1998. [8] TIA/EIA/IS-cdma2000. Physical layer standard for cdma2000 spread spectrum systems. August 1999. [9] J. Korhonen. Introduction to 3G Mobile Communications. Artech House, Norwood, 2nd ed., 2003. [10] D. J. Goodman. Second generation wireless information network. IEEE Trans. Veh. Techn., 40(2):366-374, May 1991. [11] J. M . Pereira. Beyond third generation. In Proceedings of International Sympo- sium on Wireless Personal Multimedia Communications (WPMC'99), Amsterdam, Netherlands, August 1999. Bibliography 127 [12] J. M . Pereira. Fourth generation: now it is personal. In Proceedings of Interna- tional Symposium on Personal, Indoor and Mobile Radio Communications (PIMR 2000), pages 1009-1016, London, U K , September 1999. [13] D. L . Schilling R. L. Pickholtz and L. B . Milstein. Theory of spread-spectrum communications - a tutorial. IEEE Trans. Commun. Techn., 30:855-884, May 1982. [14] G . L. Turin. Introduction to spread-spectrum anti-multi-path techniques and their application to urban digital radio. Proceedings of the IEEE, 68:328-353, March 1980. [15] A . J. Viterbi. Spread-spectrum communications-myths and realities. IEEE Communications Magazine, 17:11-18, May 1979. [16] A . J. Viterbi. CDMA: Principles of Spread Spectrum Communication. AddisonWesley, 1995. [17] R. W . Chang. Synthesis of band-limited orthogonal signals for multi-channel data transmission. Bell Labs Technical Journal, (45):1775-1796, December 1966. [18] R. W . Chang and R. A . Gibby. A theoretical study of performance of an orthogonal multiplexing data transmission scheme. IEEE Trans. Commun. Techn., 16:529-540, August 1968. [19] B . R. Saltzberg. Performance of an efficient parallel data transmission system. IEEE Trans. Commun. Techn., 15:805-811, December 1967. [20] S. B . Weinstein and P. M . Ebert. Data transmission by frequency-division multiplexing using the discrete Fourier transform. IEEE Trans. Commun. Techn., 19:628-634, October 1971. [21] M . Alard and R. Lassalle. Principles of modulation and channel coding for digital broadcasting for mobile receivers. European Broadcast Union Review, (224):47-69, August 1987. [22] J. A . C. Bingham. Multicarrier modulation for data transmission: A n idea whose time has come. IEEE Communications Magazine, 28(5):5-14, May 1990. Bibliography 128 [23] L. J. Cimini. Analysis and simulation of a digital mobile channel using orthogonal frequency division multiplexing. IEEE Trans. Commun., 33:665-675, July 1985. [24] J.-C. Tu J. S. Chow and J. M . CiofR. A discrete multitone transceiver system for HDSL applications. IEEE Journal on Selected Areas in Communications (JSAC), 9:895-908, August 1991. [25] ETSI D A B (EN 300 401 V 1.3.1). Radio broadcasting systems; digital audio broad- casting (DAB) to mobile, portable and fixed receivers. Sophia Antipolis, France, April 2000. [26] ETSI D A B - T (EN 300 744 V 1.2.1). Digital video broadcasting (DVB); framing structure, channel coding and modulation for digital terrestial television. Sop Antipolis, France, July 1999. [27] ETSI H I P E R L A N / 2 (TS 101 475). Broadband radio access networks HIPERLAN type 2 functional specification - Part 1: Physical layer. Sophia Antipolis, France, September 1999. [28] IEEE-802.11 (P802.11a/D6.0). LAN/MAN specific requirements - part 2: wireless MAC and PHY specifications - high speed physical layer in the 5Ghz band. I E E E 802.11, May 1999. [29] ETSI D A B R C T (EN 301 958 V 1.1.1). Interaction channel for digital terristrial television (RCT) incorporating multiple access OFDM. Sophia Antipolis, France, March 2001. [30] ETSI H I P E R M A N (TR 101 856). High performance metropolitan area network, requirements MAC and physical layer below 11 GHz band. Sophia Antipolis, France, December 2002. [31] Draft I E E E 802.16ab 01/01. Air interface forfixedbroadband wireless access system - Part A: Systems between 2 and 11 GHz. I E E E 802.16, June 2000. [32] A. Chouly, A. Brajal, and S. Jourdan. Orthogonal multicarrier techniques applied to direct sequence spread spectrum C D M A systems. In Proceedings of IEEE Global Telecommunications Conference- Globecom'93, pages 1723-1728, Houston, USA, November 1993. Bibliography 129 [33] V . M . DaSilva and E . S. Sousa. Multicarrier orthogonal C D M A signals for quasisynchronous communication systems. IEEE Journal of Selected Areas of Communications (JSAC), 12(5):842-852, June 1994. [34] K . Fazel. Performance of C D M A / O F D M for mobile communications system. In Proceedings of IEEE ICUPC'93, pages 975-979, Ottawa, Canada, October 1993. [35] K . Fazel and L . Papke. On the performance C D M A / O F D M for mobile communication system. of convolutionally-coded In Proceedings of IEEE PIMRC'93, pages 468-472, Yokohama, Japan, September 1993. [36] S. Kondo and L . B . Milstein. On the use of multicarrier direct sequence spread spectrum systems. In Proceedings of IEEE MILCOM'93, pages 52-56, Boston, USA, October 1993. [37] L . Vandendorpe. Multitone direct sequence C D M A system in an indoor wireless environment. In Proceedings of IEEE First Symposium of Communications and Vehicular Technology, pages 4.1.1-4.1.8, Delft, Netherlands, October 1993. [38] N . Yee, J.-P. Linnartz, and G . Fettweis. Multi-carrier C D M A in indoor wireless radio networks. In Proceedings of IEEE PIMRC'93, pages 109-113, Yokohama, Japan, September 1993. [39] S. Nada, K . Balachandran, and S. Kumar. Adaptation techniques in wireless packet data services. IEEE Commun. Mag., pages 54-64, January 2000. [40] S. Nanda, K . Balachandran, and S. Kumar. Adaptation techniques in wireless packet data services. IEEE Communications Magazine, 38(l):54-64, January 2000. [41] S. Catreux, V . Erceg, D . Gesbert, and R. W . Heath. Adaptive modulation and M I M O coding for broadband wireless data networks. IEEE Communications Magazine, 40:108-115, June 2002. [42] J . K . Cavers. A n analysis of pilot symbol assisted modulation for Rayleigh fading channels. IEEE Trans. Veh. Techn., 40:686-693, November 1991. [43] Y . S. K i m , C. J . K i m , G. Y . Jeong, Y . J . Bang, H . K . Park, and S. S. Choi. New Rayleigh fading channel estimator based on P S A M channel sounding technique. In Proceedings of IEEE International Conference on Communications-ICC'97, pages 1518-1520, June 1997. 130 Bibliography [44] X . Tang, M . S. Alouini, and A . Goldsmith. Effect of channel estimation error on M - Q A M B E R performance'in Rayleigh fading. IEEE Trans. Commun., pages 1856-1864, December 1999. [45] E D G E - E v a l u a t i o n of 8-PSK. ETSI, Finland, ETSI SMG2, W P B - 4 , 1998. [46] E D G E : Concept Proposal for Enhanced G P R S . ETSI, Tucson, A Z , ETSI SMG2, W P A / W P B - 4 , 1999. [47] W . Mohr and W . Konhauser. Access network evolution beyond third generation mobile communications. IEEE Communications Magazine, 38:122-133, December 2000. [48] D. Tse and S. Hanly. Multi-access fading channels - part I: Polymatroid structure, optimal resource allocation and throughput capacities. IEEE Trans, on Information Theory, 44(7):54-64, November 1998. [49] R. Knopp and P. A . Humblet. Information capacity and power control in single-cell multiuser communications. In Proceedings of IEEE ICC'95, pages 331-335, Seattle, June 1995. [50] R. Cheng and S. Verdu. Gaussian multiaccess channels with ISI: Capacity region and multiuser water-filling. IEEE Trans, on Information Theory, 39(3):773-785, May 1993. [51] W . Y u , W . Rhee, S. Boyd, and J. CiofH. Iterative water-filling for gaussian vector multiple-access channels. IEEE Trans, on Information Theory, 50:145-152, January 2004. [52] P. Viswanath, D . N . C. Tse., and V . Anantharam. Asymptotically optimal waterfilling in vector multiple-access channels. IEEE Trans, on Information Theory, 47:241-267, January 2001. [53] Y . H . K i m , I. Song, S. Yoon, and S. R. Park. A multicarrier C D M A system with adaptive subchannel allocation for forward links. IEEE Trans. Veh. Techn., 48(5): 1428-1436, September 1999. [54] P. Fan, J. L i , and Z. Cao. A near optimal subchannel allocation policy in forward links for multicarrier C D M A system. In Proceedings of IEEE Veh. Techn. Conf, pages 896-900, May 2001. 131 Bibliography [55] S. E . El-Khamy. Matched frequency hopping (MFH) signals for slowly fading dispersive channels. IEEE Trans. Veh. Techn., 47(l):365-369, February 1998. [56] Y . Zhu and E . Gunawan. Performance of M C - C D M A system using controlled M R C with power control in Rayleigh fading channel. IEEE Elect. Commun. letters, 36(8):752-753, April 2000. [57] T. Lok and T. Wong. Transmitter and receiver optimization in multicarrier C D M A systems. IEEE Trans. Commun., 48(7):1197 - 1207, July 2000. [58] M . Ditzel and W . Serdijn. Optimal energy assignment for frequency selective fading channels, in Proc. IEEE International Symposium on Personal, Indoor and Mobile Radio Communications, 1:D104 - D108, Sept. 2001. [59] I. K i m , H . L . Lee, B . K i m , and Y . H . Lee. Increasing in capacity of multiuser O F D M system using dynamic subchannel allocation. In Proceedings of IEEE Veh. Techn. Conf, pages 1085-1089, Tokyo, Japan, May 2000. [60] Q. Chen, E . Sousa, and S. Pasupathy. Multicarrier C D M A with adaptive fre- quency hopping for mobile radio systems. IEEE Journ. Select. Areas Commun., 14(9):1852-1858, Dec. 1996. • [61] H . Sampath, S. Talwar, J. Tellado, V . Erceg, and A . Paulraj. A fourth generation M I M O - O F D M broadband wireless system: Design, performance, and field trial results. IEEE Communications Magazine, 40(9):143-149, September 2002. [62] T. S. Rappaport, A . Annamalai, R. M . Buehrer, and W . H . Tranter. Wireless communications: Past events and a future perspective. IEEE Communications Magazine, 40(5):148-161, May 2002. [63] C. Y . Wong, R. S. Cheng, K . B . Letaief, and R. D . Murch. Multiuser O F D M with adaptive subcarrier, bit, and power allocation. IEEE Journal on Selected Areas in Communications (JSAC), 17:1747-1758, October 1999. [64] L . Hoo, B . Haider, J. Tellado, and J. CiofH. Multiuser transmit optimization for multicarrier broadcast channels: Asymptotic F D M A capacity region and algorithms. IEEE Trans. Commun., 52(6):922-930, June 2004. 132 Bibliography [65] E . Lawrey. Multiuser O F D M . In Proceedings of Int. Symp. Signal Processing and Its Applications, pages 761-764, Brisbane, Australia, 1999. [66] J. Jang and K . B . Lee: Transmit power adaptation for multiuser O F D M systems. IEEE Journal on Selected Areas in Communications (JSAC), 21 (2): 171-178, February 2003. [67] W . Rhee and J. M . Cioffi. Increasing in capacity of multiuser O F D M system using dynamic subchannel allocation. In Proceedings of IEEE Veh. Techn. Conf., pages 1085-1089, Tokyo, Japan, May 2000. [68] M . Wahlqvist, H . Olofosson, M . Ericson, C. Ostberg, and R. Larsson. Capacity comparison of an O F D M based multiple access system using different dynamic resource allocation. In Proceedings of IEEE Veh. Techn. Conf. 1997, pages 16641668, May 1997. [69] J. C. Chuang and N . R. Sollenberger. Spectrum resource allocation for wireless packet access with application to advanced cellular internet service. IEEE Journal on Selected Areas in Communications (JSAC), pages 820-829, August 1998. [70] N . Damji and T. Le-Ngoc. Adaptive downlink multi-carrier resource allocation for real-time multimedia traffic in cellular systems. In Proceedings of IEEE ICC'04, June 2004. [71] T. Le-Ngoc, N . Damji, and Y . X u . Dynamic resource allocation for multimedia services over O F D M downlink in cellular systems. In Proceedings of IEEE Veh. Techn. Conf. 2004, May 2004. [72] N . Damji and T. Le-Ngoc. Dynamic downlink O F D M resource allocation for broadband multimedia services in wireless cellular systems. In Proceedings of First International Conference on Broadband Networks (BROADNETS04), pages 589-598, June 2004. [73] I-S. Park, Y - H . Jung, I. K i m , and Y . H . Lee. Dynamic subchannel and bit allocation in multiuser M I M O / O F D M A systems. In Proceedings of IEEE Vehicular Technology Conference - VTC'2004, pages 884-888, May 2004. 133 Bibliography [74] I. K i m , H . Lee, B . K i m . , and Y . H . Lee. On the use of linear programming for subcarrier and bit allocation in multiuser O F D M systems, in Proc. IEEE Glocal Telecommunications Conference GLOBECOM'01, 6:3648-3652, Nov. 2001. [75] W . Y u , G . Ginis, and J. Cioffi. subscriber lines. Distributed multiuser power control for digital IEEE Journal on Selected Areas in Communications (JSAC), 20:1105-1115, June 2002. [76] M . F. Demirkol and M . A . Ingram. Power-controlled capacity for interfering M I M O links. In Proceedings of IEEE Veh. Techn. Conf, pages 187-191, October 2001. [77] R. S. Blum. M I M O capacity with interference. IEEE Journal on Selected Areas in Communications (JSAC), 21(5):793-801, June 2003. [78] S. Ye and R. S. Blum. Optimized signaling for M I M O interference systems with feedback. IEEE Trans. Signal Processing, 51(ll):2839-2848, November 2003. [79] P. Viswanath, D . N . C. Tse., and R. Laroia. Opportunistic beamforming using dumb antennas. IEEE Trans, on Information Theory, 48(6):1277-1294, June 2002. [80] M . Elkashlan and C. Leung. Performance of matched M C - C D M A in slowly fading frequency-selective channels. In Proceedings of IEEE International Symposium on Advances in Wireless Communications ISWC'02, pages 105-106, Victoria, B C , Canada, September 2002. [81] M . Elkashlan and C. Leung. Performance of frequency-hopping multicarrier C D M A in Rayleigh fading. In Proceedings of IEEE Vehicular Technology Conference - VTC'02, pages 341-345, Vancouver, B C , Canada, September 2002. [82] M . Elkashlan and C. Leung. Performance of frequency-hopping multicarrier C D M A on an uplink with correlated Rayleigh fading. In Proceedings of IEEE Global Telecommunications Conference - Globecom'03, pages 3407-3411, San Francisco, C A , U S A , December 2003. [83] M . Elkashlan and C. Leung. Channel aware frequency hopping multiple access scheme. IEE Electron. Lett, 39:1854-1855, December 2003. [84] M . Elkashlan and C. Leung. Channel aware frequency hopping multiple access scheme. In Proceedings of IEEE International Conference on Communications ICC 2004, Pages 390-394, Paris, France, June 2004. Bibliography- ISA [85] M . Elkashlan, C. Leung, and R. Schober. Performance analysis of channel aware frequency hopping. To appear in the IEE Proceedings - Communications. [86] M . Elkashlan, T. Khattab, C. Leung, and R. Schober. Order statistics for correlated fading channels. In Proceedings of IEEE Pacific Rim Conference on Communications, Computers and Signal Processing, pages 356-359, Victoria, B C , Vancouver, August 2005. [87] S. Hara and R. Prasad. Design and performance of multicarrier C D M A system in frequency-selective Rayleigh fading channels. IEEE Trans. Veh. Techn., pages 1584-1595, September 1999. [88] W . D . Rummler. A new selective fading model: Application to propagation data. Bell Syst. Tech. J., 59(5):1037-1071, May-June 1979. [89] J. G. Proakis. Digital Communications, 4th ed. New York:McGraw Hill, 2000. [90] D. J. Torrieri. Mobile frequency-hopping C D M A systems. IEEE Trans. Commun., 48(8):1318-1327, August 2000. [91] D. J. Torrieri. Future army mobile multiple-access communications. In Proceedings of IEEE Military Communications Conference, pages 650-654, Monterey, C A , USA, November 1997. [92] C. T. Ong and C. Leung. Code diversity transmission in a slow-frequency-hopped spread spectrum multiple-access communication system. IEEE Trans. Commun., 43(12):2897-2899, December 1995. [93] W . X u and L . Milstein. Performance of multicarrier DS C D M A systems in the presence of correlated fading. In Proceedings of IEEE 47th Vehicular Technology Conference VTC'97, pages 2050-2054, May 1997. [94] A . Wittneben. Performance of multicarrier DS C D M A systems in the pres- ence of correlated fading. In Proceedings of IEEE International Conference on Communications-ICC'93, pages 1630-1634, May 1993. [95] W . Y . Kuo and M . P. Fitz. Design and analysis of transmitter diversity using intentional frequency offset for wireless communications. IEEE Trans. Vehicular ' Technology, 46:871-881, November 1997. 135 Bibliography [96] E . A . Geraniotis and M . B . Pursley. Error probability for slow-frequency-hopping spread spectrum multiple-access communications over fading channels. IEEE Trans. Commun., 30:996-1009, May 1982. [97] Y . R. Tsai and J . F. Chang. Using frequency hopping spread spectrum technique to combat multipath interference in a multiaccessing environment. IEEE Trans. Veh. TechnoL, 43:211-222, May 1994. [98] L - L . Yang and L . Hanzo. Software-defmed-radio-assisted adaptive broadband frequency hopping multicarrier D S - C D M A . IEEE Communications Magazine, 40(3):174-183, March 2002. [99] C. Jiang and J . Wang. A novel receiver for F H M A systems. IEEE Trans. Veh. TechnoL, 51(5):1128-1137, September 2002. [100] A . Al-Dweik and F. Xiong. Frequency-hopped multiple-access communications with noncoherent M-ary O F D M - A S K . IEEE Trans. Commun., 51:33-36, January 2003. [101] D. B . Brick and F. W . Ellersick. Future air force tactical communications. IEEE Trans. Commun., 28:1551-1572, September 1980. [102] P. Duhamel F. Alberge and M . Nikolova. Adaptive solution for blind identification/equalization using deterministic maximum likelihood. IEEE Trans. Signal Processing, 50 (4): 923-936, April 2002. [103] B . Sklar. Rayleigh fading channels in mobile digital communication systems part II: mitigation.. IEEE Communications Magazine, pages 102-109, July 1997. [104] H . E l Gamal and E . Geraniotis. Comparing the capacities of F H / S S M A and D S / C D M A networks. In Proceedings of IEEE 5th International Symposium on Spread Spectrum Techniques and Application, pages 973-977, Boston, M A , U S A , 1998. [105] R. J . Larsen and M . L . Marx. An Introduction to Mathematical Statistics and Its Applications. Prentice Hall, 2000. [106] M . Schwartz, W . Bennett, and S. Stein. Communication Systems and Techniques. New York: McGraw-Hill, 1966. Bibliography 136 [107] T. S. Rappaport. Wireless Communications, Principles, and Practice. Upper Saddle River, N J : Prentice-Hall, 1996. [108] Q. T. Zhang and H . G . L u . A general analytical approach to multi-branch selection combining over various spatially correlated fading channels. IEEE Trans. Commun., 50(7):1066-1073, July 2002. [109] M . K . Simon and M . S. Alouini. A unified performance analysis of digital communications with dual selective combining diversity over correlated Rayleigh and Nakagami-m fading channels. IEEE Trans. Commun., 47:33-43, January 1999. [110] U . Mengali G . M . Vitetta and D. P. Taylor. A n error probability formula for noncoherent orthogonal binary F S K with dual diversity on correlated Rician channels. IEE Electron. Lett, 3:43-45, February 1999. [Ill] Jr. K . Dietze, C. B . Dietrich and W . L. Stutzman. Analysis of a two-branch maximal ratio and selection diversity system with unequal SNRs and correlated inputs for a Rayleigh fading channel. IEEE Trans. Wireless Commun., 1:274-281, April 2000. [112] D. Zogas G . Karagiannidis and S. Kotsopoulos. Performance analysis of triple selection diversity over exponentially correlated Nakagami-m fading channels. IEEE Trans. Commun., 51(8):1245-1248, August 2003. [113] O . . C . Ugweje and V . A . Aalo. Performance of selection diversity system in correlated Nakagami fading. In Proceedings of IEEE VTC'97, pages 1488-1492, Phoenix, A Z , USA, May 1997. [114] R. K . Mallik and Z. M . Win. Analysis of hybrid selection/maximal-ratio combining in correlated Nakagami fading. IEEE Trans. Commun., 50(8): 1372-1383, August 2002. [115] H . A . David and H . N . Nagaraja. Order Statistics. 3rd ed., J. Wiley, Hoboken, N.J., 2003. [116] A . Stuart and J. K . Ord. Kendall's Advanced Theory of Statistics, vol. 1, 6th ed., Eward Arnold, London, U K , 1994. [117] W . Feller. An Introduction to Probability Theory and its Applications. Wiley, New York, 3rd ed., 1957. Bibliography 137 [118] M . D. Springer. The Algebra of Random Variables. Wiley, New York, 1979. [119]- M . K . Simon and M.-S. Alouini. Digital Communication over Fading Channels. Wiley, New York, 2nd ed., 2005. [120] V . A . Aalo. Performance of maximal-ratio diversity systems in correlated Nakagami-fading environmentm distribution using Green's matrix approximation. IEEE Trans. Commun., 43:2360-2369, August 1995. [121] D. Zogas G . Karagiannidis and S. Kotsopoulos. A n efficient approach to multivariate Nakagami-m distribution using Green's matrix approximation. IEEE Trans. Wireless Commun., 2(5):883-889, September 2003. [122] Q. T. Zhang. Maximal-ratio combining over Nakagami fading channels with arbitrary branch covariance matrix. IEEE Trans. Veh. Techn., 48(4):1141-1150, September 1999. [123] M . Nakagami. The m-distribution — a general formula of intensity distribution of rapid fading. In Statistical Methods in Radio Wave Propagation, pages 3-36, W . G. Hoffman, Ed. Oxford, U K : Pergamon Press, 1960. [124] W . C. Y . Lee. Mobile Communications Engineering. New York:McGraw Hill, 1982. [125] M . Elkashlan, T. Khattab, C. Leung, and R. Schober. Statistics of general order selection in correlated Nakagami fading channels. Submitted for publication. [126] M . Elkashlan, C. Leung, and R. Schober. Performance analysis of general order selection in correlated fading channels. Submitted for publication. [127] H . Huang, H . Viswanathan, A . Blanksby, and M . A . Haleem. Multiple antenna enhancements for a high rate C D M A packet data system. J. VLSI Signal Processing, (30):55-69, January-March 2002. [128] R. Gozali, R. M . Buehrer, and B . D. Woerner. The impact of multiuser diversity on spacetime block coding. IEEE Trans. Commun., 7:213-215, May 2003. [129] D. N . C. Tse. Optimal power allocation over parallel gaussian broadcast channels. In Proceedings of IEEE Int. Symp. Information Theory, page 27, Ulm, Germany, June 1997. Bibliography 138 [130] D . N . C. Tse. Data throughput of C D M A - H D R : A high efficiency-high data rate personal communication wireless system. In Proceedings of IEEE 50th Vehicular Technology Conf, pages 1854-1858, Tokyo, Japan, May 2000. [131] CDMA2000. High Rate Packet Data Air Interface Specification. T I A / E I A / 3 G P P 2 Standard IS-856/3GPP2 C.S0024, v.3.0, December 2001. [132] L . J. Cimini,Jr., B . Daneshrad, and N . R. Sollenberger. Clustered O F D M with transmitter diversity and coding. In Proceedings of IEEE Global Telecommunications Conference - Globecom'96, pages 703-707, London, U K , November 1996. [133] L . J. Cimini,Jr. and N . R. Sollenberger. O F D M with diversity and coding for advanced cellular internet services. In Proceedings of IEEE Global Telecommunications Conference - Globecom'97, pages 305-309, Phoenix, Arizona, November 1997. [134] R. Dinis, P. Montezuma, and A . Gusmao. Performance trade-offs with quasi- linear ly amplified O F D M through a twobranch combining technique. In Proceedings of IEEE Vehicular Technology Conference - VTC'96, pages 899-903, Atlanta, Georgia, April-May 1996. [135] G . J. Foschini. Layered spacetime architecture for wireless communication in a fading environment when using multi-element antennas. Bell Labs Tech., 1(2):4159, Autumn 1996. [136] G . J. Foschini and M . J. Gans. On limits of wireless communications in fading when using multiple antennas. Wireless Pers. Commun., 6(3):311-335, March 1998. [137] Y . L i , J. H . Winters, and N . R. Sollenberger. M I M O - O F D M for wireless communications: Signal detection with enhanced channel estimation. IEEE Trans. Commun., 50(9): 1471-1477, September 2002. [138] A . J. Paulraj et al. A n overview of M I M O communications - a key to gigabit wireless. Proc. IEEE, 92(2):198-218, February 2004. [139] G . L . Stuber et al. Broadband M I M O - O F D M wireless communications. Proc. IEEE, 92(2):271-294, February 2004.
- Library Home /
- Search Collections /
- Open Collections /
- Browse Collections /
- UBC Theses and Dissertations /
- Channel aware strategies in wireless communications
Open Collections
UBC Theses and Dissertations
Featured Collection
UBC Theses and Dissertations
Channel aware strategies in wireless communications Elkashlan, Maged 2006
pdf
Page Metadata
Item Metadata
Title | Channel aware strategies in wireless communications |
Creator |
Elkashlan, Maged |
Date Issued | 2006 |
Description | Channel-aware resource allocation is widely considered to be crucial for realizing high data rates in wireless networks. This thesis considers the resource allocation problem in a multiuser wireless network using link adaptation. Link Adaptation (LA), which loosely refers to changing transmission parameters over a link, such as modulation, coding rate, power, etc., in response to changing channel conditions is considered to be a powerful means of achieving higher efficiency or throughput in wireless networks. The adaptation of the transmission parameters is performed according to the predicted future quality of the channel, also referred to as the channel state (CS). The objectives of this thesis are threefold: devise new methods for the improvement of multicarrier code division multiple access (MC-CDMA) as an effective signaling scheme in correlated fading channels, the development of new channel aware algorithms with adaptive subchannel allocation, and the investigation of the statistics of general order selection in correlated Nakagami fading channels. A novel adaptive subcarrier allocation algorithm is developed for MC-CDMA to improve the overall bit error rate (BER) performance. The proposed method is suitable for use in correlated fading channels. This algorithm assigns users to subcarrier groups that provide favorable fading characteristics, while simultaneously reducing the amount of interference caused to other users. This method examines the effect of equalizing the interference in each subcarrier group while maintaining a reduced correlation among the subcarrier fading processes for a group. Consequently, subcarriers are separated into non-contiguous groups to maximize frequency diversity and minimize multiple access interference (MAI). Previously proposed adaptive subcarrier allocation algorithms, are considered to be greedy algorithms. They are greedy in the sense that they consider every reservation request individually, and make the choice that looks best at the moment. Adaptive allocation using the simple greedy method gives an optimal solution for a single user system. Adaptive bit allocation using the greedy method cannot give an optimal solution for multiuser cases. It is possible that the sub-channels with the largest channel gain for one user are also the largest for another user. We present in this thesis a new class of dynamic resource allocation schemes that are based on all the users’ subband gains. Due to the time-varying nature of the wireless channel, dynamic resource allocation makes full use of multiuser diversity to achieve higher performance. In this thesis, we formulate the multiuser subcarrier allocation problem and propose an iterative algorithm to perform the subcarrier allocation. This method involves the ordering of subband gains. Thus, the analytical foundation is the theory of order statistics. The users are recursively assigned to that subcarrier which provides the highest possible signal-to-noise ratio (SNR). While ensuring that no more than one user is assigned to the same subband. Our objective is to find low-complexity schemes which can improve system capacity and throughput, and simultaneously minimizing the MAI. Finally, the cumulative distribution function (cdf) (and hence outage probability) of the r -th order branch SNR in correlated Nakagami- m fading is studied. The accuracy of a simple exchangeable approximation to reduce the computational load is examined. The effects of correlation on the error performance of channel aware systems that involve the ordering of channelgains in a correlated Nakagami fading environment is investigated. A useful analytical formula for the BER of the r-th order statistic of a set of arbitrarily correlated and not necessarily exchangeable diversity branch gains is described and shown to be applicable in the performance analysis of various diversity systems. |
Genre |
Thesis/Dissertation |
Type |
Text |
Language | eng |
Date Available | 2010-01-16 |
Provider | Vancouver : University of British Columbia Library |
Rights | For non-commercial purposes only, such as research, private study and education. Additional conditions apply, see Terms of Use https://open.library.ubc.ca/terms_of_use. |
DOI | 10.14288/1.0065609 |
URI | http://hdl.handle.net/2429/18464 |
Degree |
Doctor of Philosophy - PhD |
Program |
Electrical and Computer Engineering |
Affiliation |
Applied Science, Faculty of Electrical and Computer Engineering, Department of |
Degree Grantor | University of British Columbia |
Graduation Date | 2006-11 |
Campus |
UBCV |
Scholarly Level | Graduate |
Aggregated Source Repository | DSpace |
Download
- Media
- 831-ubc_2006-199701.pdf [ 4.72MB ]
- Metadata
- JSON: 831-1.0065609.json
- JSON-LD: 831-1.0065609-ld.json
- RDF/XML (Pretty): 831-1.0065609-rdf.xml
- RDF/JSON: 831-1.0065609-rdf.json
- Turtle: 831-1.0065609-turtle.txt
- N-Triples: 831-1.0065609-rdf-ntriples.txt
- Original Record: 831-1.0065609-source.json
- Full Text
- 831-1.0065609-fulltext.txt
- Citation
- 831-1.0065609.ris
Full Text
Cite
Citation Scheme:
Usage Statistics
Share
Embed
Customize your widget with the following options, then copy and paste the code below into the HTML
of your page to embed this item in your website.
<div id="ubcOpenCollectionsWidgetDisplay">
<script id="ubcOpenCollectionsWidget"
src="{[{embed.src}]}"
data-item="{[{embed.item}]}"
data-collection="{[{embed.collection}]}"
data-metadata="{[{embed.showMetadata}]}"
data-width="{[{embed.width}]}"
async >
</script>
</div>
Our image viewer uses the IIIF 2.0 standard.
To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.831.1-0065609/manifest