Abstract

In this thesis we consider Orthogonal Frequency Division Multiplexing (OFDM) technique by taking into account in the receiver design the fundamental and unique characteristics of Underwater Acoustic (UWA) channels in the context of Relay-Assisted (RA) systems. In particular, OFDM technique is used to combat the problem of Intersymbol Interference (ISI), while to handle the Intercarrier Interference (ICI), a pre-processing unit is used prior to the Minimum
Mean Squared Error (MMSE) frequency-domain equalization called Multiple Resampling (MR), which minimizes the effect of time variation. This pre-processor consists of multiple
branches, each corresponds to a Doppler scaling factor of a path/user/cluster, and performs of frequency shifting, resampling, and Fast Fourier Transform (FFT) operation. As a suboptimal alternative to MR pre-processing, Single Resampling (SR) pre-processing is also used to reduce the effect of ICI in the system, and it consists of only one branch that performs frequency shifting, resampling, and FFT operation, which corresponds to one approximated resampling factor, that is a function of one or more of the actual Doppler scaling factors. The problem of bandwidth scarcity is considered in the context of Two Way Relaying (TWR) systems in an attempt to increase the bandwidth efficiency of the system, while the problem of fading is
considered in the context of Distributed Space-Time Block Coding (D-STBC) to boost the system reliability. Also, joint TWR-D-STBC system is proposed to extract the advantages of both schemes simultaneously.

Second, motivated by the fact that OFDM is extremely sensitive to time variation, which destroys the orthogonality between the subcarriers, we consider another candidate to UWA channels and competitor to OFDM scheme, namely, block-based Single Carrier (SC) modulation with Frequency Domain Equalization (FDE). We start by the Point-to-Point (P2P) systems with path-specific Doppler model and Multiple Access Channel (MAC) system with user-specific Doppler model. The Maximum Likelihood (ML) receiver in each case is derived, and it is shown that a MR pre-processing stage is necessary to handle the effect of time variation, as it is the case in OFDM. Different from OFDM, however, the structure of this pre-processing stage. Specifically, it consists of multiple branches and each branch corresponds to a Doppler scaling factor per path or per user, and performs frequency shifting, resampling, and followed by and integration. FFT operation is not a part of the pre-processor. The goal of this pre-processing stage
is to minimize the level of time variation in the time domain. So, the output of the pre-processor will still be time-varying contaminated by ISI, and hence an equalization stage is required. To avoid the complexity of the optimum Maximum Likelihood Sequence Detector (MLSD), we propose the use of MMSE FDE, where the samples are transformed to the frequency domain by means of FFT operation, and after the FDE transformed back to the time domain, where symbol-by-symbol detection becomes feasible. Also, the channels are approximated such that all paths or all users have the same Doppler scaling factor, and the pre-processing stage in this case consists of only one branch and it is called SR. Having the basic structure of SC-FDE scheme, we then consider the corresponding schemes that are considered for OFDM systems, namely: TWR, D-STBC, and TWR-D-STBC schemes.

A complete complexity analysis, bandwidth efficiency, and extensive Average Bit Error Rate (ABER) simulation results are given. It is shown that MR schemes outperforms its SR
counterparts within a given signaling scheme (i.e., OFDM or SC-FDE). However, this superiority in performance comes at the expense of more hardware complexity. Also, for uncoded
systems, MR-SC-FDE outperforms its OFDM counterpart with less hardware complexity, because in SC-FDE systems, FFT operation is not part of the MR pre-processor, but rather a part of the equalizer. Finally, under total power constraint, it is shown that TWR-D-STBC scheme serves as a good compromise between bandwidth efficiency and reliability, where it has better bandwidth efficiency with some performance loss compared to D-STBC, while it has better performance and the same bandwidth efficiency compared to TWR.