Abstract

Volterra equalizations (VE) have delivered significant performance improvement on optical signals, but their extremely high demands on computation complexity severely impede the possibility of large-scale deployment in resource-constrained optical interconnects. In this paper, we propose and experimentally demonstrate a novel sparsity tuned elastic net-pruned Volterra equalization (SENVE) technique to remove the redundancy in the structure of conventional VE (baseline VE) to save complexity without degrading system performance. With a three-stage configuration: pretraining, pruning, and retraining, our SENVE scheme can: (1) realize the compressed structure from a larger VE to reduce computation complexity and (2) accomplish budget-conscious and hardware- friendly structured sparsity of VE to efficiently accelerate the VE's evaluation. We experientially demonstrate the proposed scheme on an 80-Gbps four-level pulse-amplitude modulation (PAM4) signal in O-band for 40-km single-mode fiber (SMF) transmission with a 30-GHz externally modulated laser (EML). Under optimal pruning but without retraining scenarios, the experimental results show that, compared with baseline VE at the KP4-FEC limit, the proposed SENVE can achieve up to 68.7% complexity reduction without degrading performance over 40 km transmission at an received optical power (ROP) of −5 dBm. Moreover, we also find our proposed SENVE with retraining can achieve up to 90.8% and 35.8% complexity reduction, with respect to baseline VE and L1-regularization VE, without sacrificing system performance under the same over-pruning scenarios.

Reducing computation complexity by using elastic net regularization based pruned Volterra equalization in a 80 Gbps PAM-4 signal for inter-data center interconnects

Govind sharan Yadav, Chun-Yen Chuang, Kai-Ming Feng, Jhih-Heng Yan, Jyehong Chen, and Young-Kai Chen
Opt. Express 28(26) 38539-38552 (2020)

Computation efficient sparse DNN nonlinear equalization for IM/DD 112 Gbps PAM4 inter-data center optical interconnects

Govind Sharan Yadav, Chun-Yen Chuang, Kai-Ming Feng, Jyehong Chen, and Young-Kai Chen
Opt. Lett. 46(9) 1999-2002 (2021)

Nonlinear equalization based on pruned artificial neural networks for 112-Gb/s SSB-PAM4 transmission over 80-km SSMF

Zhiquan Wan, Jianqiang Li, Liang Shu, Ming Luo, Xiang Li, Songnian Fu, and Kun Xu
Opt. Express 26(8) 10631-10642 (2018)

C-band 100-GBaud/λ PAM-4 transmission enabled by hardware-efficient nonlinear equalizer with weight-sharing pruning

Wenzhuo Cheng, Meng Xiang, Hailin Yang, Xuancheng Huo, Yunhe Ma, Jianping Li, Songnian Fu, and Yuwen Qin
Opt. Lett. 48(2) 351-354 (2023)

Amplifier-free 4×96 Gb/s PAM8 transmission enabled by modified Volterra equalizer for short-reach applications using directly modulated lasers

Di Li, Lei Deng, Yao Ye, Yucheng Zhang, Haiping Song, Mengfan Cheng, Songnian Fu, Ming Tang, and Deming Liu
Opt. Express 27(13) 17927-17939 (2019)