ASIC Implementation of Time-Domain Digital Backpropagation with Deep-Learned Chromatic Dispersion Filters

TL;DR

This paper presents a novel ASIC implementation of time-domain digital backpropagation that employs deep-learned chromatic dispersion filters, achieving better error rates and significant power savings in CMOS technology.

Contribution

It introduces a jointly optimized, quantized deep-learning approach for chromatic dispersion filters in digital backpropagation, enhancing performance and efficiency.

Findings

Improved BER performance over baseline methods

Over 40% power dissipation reduction in 28-nm CMOS

Effective joint optimization of filters using machine learning

Abstract

We consider time-domain digital backpropagation with chromatic dispersion filters jointly optimized and quantized using machine-learning techniques. Compared to the baseline implementations, we show improved BER performance and >40% power dissipation reductions in 28-nm CMOS.