ACKNOWLEDGMENTS
My deep gratitude goes to Dr. Laurie Joiner, who has been supportive, knowledgeable, thoughtful, and considerate during this journey. I am so thankful with no limits, Dr. Joiner, I learned a lot from you and I have the pleasure knowing such a smart and a wise woman like you.
Thanks to every person in the ECE department at UAH. Special thanks are passed to Jackie Siniard and Linda Grubbs for their lovely smiles and the relieving chats.
I have to mention my wonderful friend, Aditi, without whom, my PhD years would not have been as joyful as they were. Thank you, Aditi, for the encouragement you provided me with and for the funny talks. I am really lucky getting to know such a positive and warm-hearted person like you.
I would also like to thank Yarmouk University in Jordan for funding my PhD study. It will be my honor to get back and teach in such a reputable institution.
Last but not least, I would like to pass my thanks and love to my mom and dad,
Fatimah and Ahmad, and to all my family members in Jordan. They have been always there: loving, keeping up with me, tolerating the full spectrum of my personality, and supporting me until the final mile of every pursuit. I cannot wait to get back to you guys and hug you all…
Asma Ahmad Alqudah
vi
Table of Contents
ABSTRACT……………………………………………………………………………. iv
ACKNOWLEDGMENTS……………………………………………………………… vi
List of Figures……………………………………………………………………………. x
List of Tables……………………………………………………………………………xiii
CHAPTER
1. Introduction…………………………………………………………………………… 1
1.1 Motivation for FTN signaling………………………………………………………….. 5
1.1.1 FTN signaling: background……………………………………………………… 6
1.2 FTN prior work………………………………………………………………………….. 7
1.3 A discussion on tree-based detection algorithms………………………………………. 10
1.4 Dissertation contributions…………………………………………………………..….. 12
2. Basic principles of linear modulation systems……………………………………… 14
2.1 Single carrier linear modulation systems………………………………………….…….. 14
2.1.1 Bit and block error rate definitions…………………………………….………... 17
2.1.2 Bandwidth characteristics……………………………………………………..… 20
2.1.3 The squared Euclidean distance definition……………………………….…..…. 24
2.1.4 T-orthogonal modulation pulses………………………………………………… 26
2.2 Introduction to maximum-likelihood sequence estimation………………………….….. 30
2.2.1 The recursively-structured MLSE…………………………………………….…. 33
2.2.2 The error performance of the MLSE………………………………………….…. 36
2.3 The M-algorithm……………………………………………………………………….... 38
2.4 Maximum a posteriori decoding……………………………………………………..……40
2.4.1 The BCJR algorithm…………………………………………………………… 43
vii
2.5 Faster-than-Nyquist signaling………………………………………………………… 46
2.5.1 The system model……………………………………………………………..…. 49
2.5.2 The capacity estimation of FTN signaling …………………………………..….. 52
2.6 Principles of turbo equalization………………………………………………………… 57
3. Employing higher order modulation in combination with nonbinary code alphabets in the context of faster-than-Nyquist signaling………………………………………………………....62
3.1 Problem statement……………………………………………………………………..….63
3.1.1 The selection of the FTN pulse shape………………………………………….….67
3.2 Equivalent discrete-time system models…………………………………………………68
3.2.1 An improved minimum phase model………………………………………….….69
3.3 The extension of the M-BCJR algorithm for nonbinary alphabets………………….….. 71
3.3.1 Performance of the M-BCJR in simple detection…………………………….…..74
3.3.2 Backup M-BCJR for nonbinary alphabets……………………………………..…76
3.4 Turbo equalization………………………………………………………………….……79
3.4.1 System model……………………………………………………………….….....80
3.4.2 Simulation results………………………………………………………….……. 95
3.5 Binary code-aided QPSK-based FTN………………………………………………….. 103
3.5.1 System model……………………………………………………………………103
3.5.2 Simulation results………………………………………………………………..108
4. Turbo equalization of the faster-than-Nyquist signaling
using the reduced-complexity Z-MAP algorithm………………………………… 111
4.1 Introduction……………………………………………………………………………..111
4.2 Error moments………………………………………………………………………….113
4.3 Z-MAP applied to turbo equalization of FTN signals………………………………….117
4.3.1 Simulation results………………………………………………………………..118
viii
5. Summary and future directions…………………………………………………… 128
REFERENCES…………………………………………………………………………132
ix
List of Figures
Figure Page
2.1 A basic device for transmitting information via carrier modulation……………….. 15
2.2 A system model of a communications system when transmitted over an AWGN channel………………………………………………………….. 16
2.3 The frequency content of the bandpass signal ………………………………22 ( ) 2.4 Root RC pulses at three different values of the excess bandwidth factors …………………………………………………………………………….30 2.5 A straightforward way to produce the sequence from the received signal ………………………………………………………………………….. 33 ( ) 2.6 A 4-state binary trellis example…………………………………………………….. 34
2.7 A block diagram of a serial concatenation communication system with encoding and ISI. denotes an interleaver………………………………………49 ∏ 2.8 A serial concatenation communications system employing iterative turbo equalization at the receiver…………………………………………………………..58
3.1 Turbo equalization structure……………………………………………………..….64
3.2 Model for converting the continuous FTN into discrete time……………………….68
3.3 BER vs. for simple ISI detection BPSK-based FTN………………………...74 ⁄ 3.4 BER vs. for simple ISI detection QPSK-based FTN……………………..….75 ⁄ 3.5 Backup M-BCJR procedure for and . Illustrating and recursions, hard decision path, and =backup 3 re cursion………………………………..78 = 2
x
3.6 Nonbinary turbo equalization receiver………………………………………..…….80
3.7 Turbo equalizer BER vs. for binary code-aided BPSK-based FTN signaling at ………………………………………………………………..96 ⁄ τ = 1⁄ 2 3.8 Turbo equalizer BER vs. for quaternary code-aided QPSK-based FTN signaling at ………………………………………………………………..96 ⁄ τ = 1⁄ 2 3.9 Turbo equalizer BER vs. for binary code-aided BPSK-based FTN signaling at ………………………………………………………………..98 ⁄ τ = 0.35 3.10 Turbo equalizer BER vs. for quaternary code-aided QPSK-based FTN signaling at ……………………………………………………………....98 ⁄ τ = 0.35 3.11 Turbo equalizer BER vs. for binary code-aided BPSK-based FTN signaling at ………………………………………………………………99 ⁄ τ = 0.25 3.12 Turbo equalizer BER vs. for quaternary code-aided QPSK-based FTN signaling at …………………………………………………………..…100 ⁄ τ = 0.25 3.13 Turbo equalizer BER vs. for quaternary code-aided QPSK-based FTN signaling for and for ⁄ different number of iterarions…………………….100 τ = 0.5 3.14 Turbo equalizer BER vs. for quaternary code-aided QPSK-based FTN signaling at ………………………………………………………..108 ⁄ τ = 1⁄ 2 3.15 Turbo equalizer BER vs. for binary code-aided QPSK-based FTN signaling at …………………………………………………………..…..109 ⁄ τ = 1⁄ 2 4.1 The most probable state in the correct instant…………………………………… 113
4.2 Presence of a concurrent at the error instant……………………………………..…114
4.3 A trellis with 4 states……………………………………………………………..…115
4.4 The Z-MAP principle [42]………………………………………………………….117
xi
4.5 Average number of states of the Z-MAP simple detection of the FTN binary signals at …………………………………………………………….…..119 τ = 1⁄ 2 … 4.6 BER comparison between M-BCJR and Z-MAP turbo decoding for binary FTN signaling at ……………………………………………………………..…120 τ = 1⁄ 2 4.7 Turbo equalizer BER vs. for binary code-aided BPSK-based FTN signaling at applyingE ⁄N the Z-MAP…………………………………………121 τ = 1⁄ 2 4.8 Average number of states of the Z-MAP turbo system of figure 4.7………………122
4.9 BER comparison between M-BCJR and Z-MAP turbo decoding at the 4 th iteration for binary code-aided BPSK-based FTN signaling at …………. 124 τ = 1⁄ 2 4.10 Turbo equalizer BER vs. for quaternary code-aided QPSK-based FTN signaling at applyingE ⁄N the Z-MAP…………………………………….… 125 τ = 1⁄ 2 4.11 Average number of states of the Z-MAP turbo system of Figure 4.10………….. 125
xii
List of Tables
Table Page
I Algorithm for computing the posterior probabilities for a memory-2 ISI channel………………………………………………………Pr( | ) …….84
II Algorithm for computing the posterior probabilities for the memory-2 quaternary convolutional code definedPr( in= this | ( )) section………….89
III Algorithm for computing the posterior probabilities for the memory-2 binary convolutional code definedP( in this = section……………. | ( )) 106
xiii
Dedication
To Abdullah…
CHAPTER 1
Introduction
Tremendous progress has been witnessed in wireless communications over the last two decades. Mobile telephony, which was primarily meant for voice-based services, has evolved to the extent that non-voice-based services now predominate. In addition, the explosive growth computer networking has gone through has laid the foundation for the largest global medium for information exchange, the Internet. Therefore, there is an ever increasing demand to make better use of the available resources in order to sustain this growth.
One primary resource in wireless communications is the frequency band of operation. The frequency bands are controlled and allocated by regulatory bodies such as the Body of European Regulator for Electronic Communications (BEREC) [1], the
Federal Communications Commission (FCC) [2], and the Telecom Regulatory Authority of India (TRAI) [3]. In this context, a wireless system mainly refers to a mobile phone or a handheld device communicating with other wireless devices or base stations. Since mobile phones have evolved from simple communicating devices to portable computers, there is a requirement of more efficient transmission systems that accommodate the increasing amount of wireless traffic. Even though the amount of available bandwidth has increased some, the great demand for wireless access has triggered a strong competition
1
among the wireless systems operators, who are paying a high premium to own spectrum allocations. For example, the 4G wireless mobile deployment in Spain, also known as
Long-Term Evolution (LTE) or Evolved UTRA (EUTRA), brought EUR 1.5 billion from the bidding of a total of 310 MHz in different frequency bands [4]. Even though the
LTE/LTE-Advanced is currently meeting the wireless user access demands, the requirements to increase the system capacity is growing dramatically. It is expected that wireless traffic will increase beyond 500-fold in 2020 as compared to the traffic in 2010
[5]. Taking into account the plethora of smart phones and tablets, the need for developing new wireless radio access technologies is essential to enhance the system capacity as well as the user data rates for future operations beyond LTE-Advanced. In the meanwhile, developments in the semiconductor technology are enabling the small-sized microprocessors to handle more complex operations.
Modern applications demand more bits be carried over the wireless channels. The straightforward ways to do so are either to send longer or faster data streams, i.e., to consume more time or more bandwidth. However, both time and frequency are scarce and costly resources. How can more information be carried per hertz and second?
Modulation theory, since the pioneering work of Nyquist [6], has mainly considered the memoryless transmission of data, which greatly simplifies receiver design and theoretical analysis. In the context of memoryless transmission, the symbols transmitted in different time epochs are considered independent. Thus, the different signals are received assuming there is no intersymbol interference (ISI) among them, and a simple symbol-by-symbol receiver can be used to detect the sequence reliably.
2
Shannon in 1948 and 1949 [7, 8] developed a fundamental concept to communications with his work in information theory. He proved that highly reliable communication can be achieved if the transmitted symbols are made in groups. He also verified that this construction is possible if the time signals are generated using sinc pulses. Based on this work, most communication technologies maintained the memoryless modulation assumption (see Figure 2.2). Even this assumption is only optimal analytically; it incurs some capacity losses practically due to the non-ideal system components.
Nyquist criterion states that in order to transmit data at a rate symbols per second with no ISI, pulses with a minimum bandwidth of are required. This data /2 Hz rate requires the peaks of the pulses to be spaced in time by seconds. 1/ Most data transmission scenarios employ linear modulation, which is formed by adding up a sequence of data pulses shifted from each other by an integer multiple of the symbol time duration , with the form