Laser Phase Noise Mitigation based on Autoencoder for End-to-end Learning of CO-OFDM Systems

This paper proposes an end-to-end learning approach for coherent optical orthogonal frequency-division multiplexing (CO-OFDM) fiber communication transmission to mitigate laser phase noise. The approach is based on the autoencoder (AE) concept, which is a type of deep neural network designed to learn how to reconstruct input data at its output. In the proposed approach, the encoder component of the autoencoder generates robust symbol sequence representations for incoming data, ensuring resilience to laser phase noise impairments. The proposed approach exhibits impressive tolerance to laser phase noise of low-cost distributed feedback (DFB) lasers (a linewidth of above 1 MHz), making it effective in compensating for the impact of inter-carrier interference (ICI) phase noise within an OFDM symbol.