Adaptive state-pruning MLSE for high-efficiency FTN-16QAM symbol-rate sampling coherent interconnection

This paper presents a novel logic-partitioning-based adaptive pruning maximum-likelihood sequence estimation (LPAPMLSE) algorithm designed to address the high computational complexity of traditional MLSE in high-speed coherent optical interconnects. Leveraging quadrature duobinary (QDB) shaped faster-than-Nyquist (FTN) 16QAM signaling, our approach dynamically prunes improbable state transitions through intelligent signal space partitioning, reducing computational complexity by 56.25% while maintaining near-optimal detection performance (0.2dB penalty). Experimental validation in a 100Gbaud symbol-rate sampling digital signal processing (DSP) system demonstrates robust performance across 10/20/40km fiber transmissions, achieving comparable performance to full-state MLSE.