Amplitude-Aware Computational Sharing: A Joint State-Space Optimization Framework for Baud-Rate MLSE and BCJR Decoders

This paper presents a novel computational sharing framework in faster than Nyquist (FTN) based baud-rate sampling (BRS) systems, addressing the critical challenge of decoder complexity in next-generation data center interconnects. The proposed scheme introduces memory-efficient MLSE with pre-decision-driven pruning and metric sharing (MEPS-MLSE) and joint state sharing and truncation BCJR (JSST-BCJR). By enabling cross-symbol branch metric (BM) and state transition metric (STM) reuse in temporal and spatial dimensions, the framework achieves 94.66% (MLSE) and 91% (BCJR) complexity reduction while maintaining near-optimal BER and NGMI performance.