Blind nonlinear equalizer using artificial neural networks for PAM-4 signal transmissions with DMLs

Abstract

A novel blind equalization scheme based on a machine learning algorithm called artificial neural network (ANN) is proposed and experimentally demonstrated to compensate for the dynamic nonlinear behavior of a 10-G class 1310-nm directly modulated laser (DML) in pulse amplitude modulated (PAM)-4 signal transmissions. The dynamic nonlinearity is often seen as the major limiting factor against the transmission capacity beyond 10 Gbit/s in an intensity modulation and direct-detection (IM/DD) optical access network. The bit error rate (BER) results for 12.5-GBaud PAM-4 signal transmissions over a 25-km single-mode fiber (SMF) verify that the proposed blind equalization scheme can mitigate the dynamic system nonlinearity without requiring any pre-determined data sequences and also can achieve comparable performance without increasing the complexity in comparison with the supervised learning technique.