Spectrally Efficient FDM Communication Signals and Transceivers: Design, Mathematical Modelling and System Optimization

This thesis addresses theoretical, mathematical modelling and design issues of Spectrally Efficient FDM (SEFDM) systems. SEFDM systems propose bandwidth savings when compared to Orthogonal FDM (OFDM) systems by multiplexing multiple non-orthogonal overlapping carriers. Nevertheless, the deliberate collapse of orthogonality poses significant challenges on the SEFDM system in terms of performance and complexity, both issues are addressed in this work. This thesis first investigates the mathematical properties of the SEFDM system and reveals the links between the system conditioning and its main parameters through closed form formulas derived for the Intercarrier Interference (ICI) and the system generating matrices. A rigorous and efficient mathematical framework, to represent non-orthogonal signals using Inverse Discrete Fourier Transform (IDFT) blocks, is proposed. This is subsequently used to design simple SEFDM transmitters and to realize a new Matched Filter (MF) based demodulator using the Discrete Fourier Transforms (DFT), thereby substantially simplifying the transmitter and demodulator design and localizing complexity at detection stage with no premium at performance. Operation is confirmed through the derivation and numerical verification of optimal detectors in the form of Maximum Likelihood (ML) and Sphere Decoder (SD). Moreover, two new linear detectors that address the ill conditioning of the system are proposed: the first based on the Truncated Singular Value Decomposition (TSVD) and the second accounts for selected ICI terms and termed Selective Equalization (SelE). Numerical investigations show that both detectors substantially outperform existing linear detection techniques. Furthermore, the use of the Fixed Complexity Sphere Decoder (FSD) is proposed to further improve performance and avoid the variable complexity of the SD. Ultimately, a newly designed combined FSD-TSVD detector is proposed and shown to provide near optimal error performance for bandwidth savings of 20% with reduced and fixed complexity. The thesis also addresses some practical considerations of the SEFDM systems. In particular, mathematical and numerical investigations have shown that the SEFDM signal is prone to high Peak to Average Power Ratio (PAPR) that can lead to significant performance degradations. Investigations of PAPR control lead to the proposal of a new technique, termed SLiding Window (SLW), utilizing the SEFDM signal structure which shows superior efficacy in PAPR control over conventional techniques with lower complexity. The thesis also addresses the performance of the SEFDM system in multipath fading channels confirming favourable performance and practicability of implementation. In particular, a new Partial Channel Estimator (PCE) that provides better estimation accuracy is proposed. Furthermore, several low complexity linear and iterative joint channel equalizers and symbol detectors are investigated in fading channels conditions with the FSD-TSVD joint equalization and detection with PCE obtained channel estimate facilitating near optimum error performance, close to that of OFDM for bandwidth savings of 25%. Finally, investigations of the precoding of the SEFDM signal demonstrate a potential for complexity reduction and performance improvement. Overall, this thesis provides the theoretical basis from which practical designs are derived to pave the way to the first practical realization of SEFDM systems.