A Nonlinear Compensation Scheme for High Baud Rate Assisted by Neural Network Based on First-Order and Second-Order Perturbation Theory

In order to effectively suppress the performance degradation caused by the cumulative nonlinear impairments due to the higher transmission rate and larger transmission capacity of optical communication systems, we propose a nonlinear compensation scheme assisted by CNN-BiLSTM neural network based on first-order and second-order perturbation theory. It is verified by the polarization multiplexing (PDM) single carrier (SC) and 5-channel wavelength division multiplexing (WDM) simulation systems with 128 GBaud 16 quadrature amplitude modulation signal and transmission distance of 1600 km, the Q-factor of our scheme can be enhanced by 0.48 dB and 0.26 dB, respectively, in SC system when compared with the linear compensation scheme and the digital backpropagation at 40 steps per span (DBP40StPs). Furthermore, in 5-channel WDM systems, our solution resulted in a 0.53 dB and 0.34 dB increase in Q-factor gain, respectively, with a 29% reduction in complexity, when compared with the aforementioned schemes.