Machine Learning Enabled Nonlinear Phase Noise Mitigation for Coherent Optical Systems

Abstract: In the current development of coherent optical communication systems, nonlinear noise is considered to be the ultimate bottleneck when extending the transmission length. In this report we suggest a nonparamter digital signal processing scheme to extend the transmission length of a fiber link. The processing scheme was enabled by machine learning, implemented to compensate nonlinear noise in a communication system without needing any information about the physical state of the transmission line. It was shown that in the case of nonlinear phase noise in a long-haul fiber system, the proposed processing scheme could extend the transmission length of the fiber link. However, for interchannel noise a clear benefit could not be determined due to limitations in the simulation. It was concluded that for a long-haul fiber link, knowledge of the system could be replaced with learning through an optimal statistical algorithm.