Computationally efficient fixed-state MLSD for 100G C-band dispersion-uncompensated IM/DD links

Maximum likelihood sequence detection (MLSD) has been introduced in high-speed intensity modulation and direct detection (IM/DD) systems due to its ability to effectively mitigate severe inter-symbol interference (ISI). However, the use of digital noise-whitening post filter (PF) to enhance MLSD performance brings significant computational complexity, as the number of states grows exponentially with PF order and modulation level. To address this challenge, we propose a complexity-efficient MLSD (CE-MLSD) scheme that that selectively retains a small number of dominant states and symbol levels, while increasing PF taps to enhance frequency-domain compensation performance. Additionally, we provide a thorough comparison of the computational complexity and bit-error rate (BER) performance of different MLSD schemes. Compared to full-state MLSD, the CE-MLSD reduces the number of multiplications by 91.67%, 99.79%, and 99.68% for 100-Gb/s OOK, 120-Gb/s PAM-4, and 120-Gb/s PAM-6 signals over 80-km, 40-km, and 20-km standard single-mode fibers (SSMF), respectively, while maintaining similar BER performance. Furthermore, for PAM-4 and PAM-6 formats, the CE-MLSD offers additional complexity reduction over previously proposed schemes such as Trellis path-limitation MLSD (TL-MLSD) and fixed-state MLSD (FS-MLSD), achieving up to 99.63%, 43.75%, and 11.76%, 36.36% reductions in multiplication complexity, respectively. CE-MLSD provides a potential solution for low-cost, high-performance C-band IM/DD optical transmission systems with extended reach.