Mitigation of Nonlinear Phase Noise in Single-channel Coherent 16-QAM Systems Employing Logistic Regression

Abstract We propose and analyze a classier based on logistic regression (LR) to mitigate the impact of nonlinear phase noise (NPN) caused by Kerr induced self-phase-modulation in digital coherent systems with single-channel unrepeated links. Simulation results reveal that the proposed approach reduces the bit error ratio (BER) in a 100-km-long 16 quadrature amplitude modulation (16-QAM) system operating at 56-Gbps. Thus, the BER is reduced from 6.88·10 -4 when using maximum likelihood to 4.27·10 -4 after applying the LR-based classification, representing an increase of 0.36 dB in the effective Q-factor. This performance enhancement is achieved with only 624 operations per symbol, which can be easily parallelized into 16 lines of 39 operations.