Intelligent 2-Dimensional Soft Decision Enabled by K-Means Clustering for VCSEL-Based 112-Gbps PAM-4 and PAM-8 Optical Interconnection

TLDR: The results indicate significant improvement of the proposed k-means 2D SD without training using prior-known sequences, which can be realized by additional consideration of nonlinearity, level-dependent noise and ISI.

Abstract: In this work, we proposed an intelligent 2-dimensional soft decision (2D SD) enabled by k-means clustering, for vertical-cavity surface-emitting laser (VCSEL) based 112-Gbps PAM-4 and PAM-8 optical interconnection. At high modulation speed, VCSEL based link suffers from severe level nonlinearity, level-dependent noise and inter-symbol interference (ISI). For characterizing the above-mentioned three distortions, 2D signaling is performed through time-slotted mapping of PAM. Without extra requirement of Monte Carlo approach, channel conditional probability density function (PDF) can be intelligently estimated using inline data, thanks to 2D k-means machine learning. Thus, improved precision of log likelihood ratio (LLR) can be realized by additional consideration of nonlinearity, level-dependent noise and ISI. Both simulations and experiments have been carried out for proof-of-concept investigations on VCSEL and multimode fiber (MMF) links. 112-Gbps PAM-4 and PAM-8 signaling have been experimentally realized using a commercial-product-level VCSEL with 100-m MMF transmission. The results indicate significant improvement of the proposed k-means 2D SD without training using prior-known sequences.