Low-complexity approaches for faster-than-Nyquist signaling

A promising avenue to achieve higher data rates is through the adoption of a non-orthogonal physical layer transmission approach known as Faster-than-Nyquist (FTN) signaling, which holds the potential to transmit more data with identical bit energy and bandwidth utilization compared to Nyquist signaling. However, FTN signaling introduces complexity to the system due to interference. This thesis initiates an exploration into the impact of channel coding, including non-binary and binary low-density parity-check codes, as well as polar codes, on a low-complexity FTN detector. The simulation results reveal that channel coding effectively narrows the performance gap between the FTN detector and the optimal BCJR detector. Consequently, there is no imperative need for a high-complexity FTN detector, particularly when dealing with high values of τ, referred to as the acceleration or packaging parameter. Additionally, the thesis introduces a coordinate interleaved FTN signaling technique, upon which a low-complexity FTN detector is formulated and compared with existing literature. This innovative approach yields a 10% increase in spectral efficiency (SE) at a relatively modest cost of 0.6 dB. In essence, this detector strikes a favorable balance between performance and complexity. Furthermore, the thesis addresses FTN signaling detection as a semidefinite problem and proposes a low-complexity soft-output semidefinite relaxation (SDR) based detector for M-PAM, supported by lookup tables (LUTs). Simulation results demonstrate that this detector, with polynomial computational complexity and adaptability to higher modulation schemes, closely approximates the coded BCJR performance within 0.8 dB for 4-PAM. Moreover, it achieves a 0.3 dB improvement in performance when SNR-based LUTs are employed. Given its low complexity nature, this suggests that the soft-output SDR detector can provide good performance for high modulation in terms of the produced log-likelihood ratios (LLRs).