A Low-complexity Nonlinear Compensation Scheme Assisted by Space-Time Adjacent Symbols for Wavelength-Division-Multiplexed Systems

In this paper, we propose a low-complexity nonlinear compensation scheme based on space-time adjacent symbols-assisted semi-connected feedforward neural network (SpTiAS-SFNN-NLC) for wavelength-division-multiplexed (WDM) systems. With the help of the adjacent symbols during the same moment on different wavelengths including the target wavelength, the scheme accumulates and averages self-phase modulation (SPM) and cross-phase modulation (XPM) of current symbols of target channel with a sliding window to equalize the nonlinear phase mismatch during the signals transmission, thereby reducing nonlinear in-band distortion. In order to verify the effectiveness of the proposed scheme, a 5-channel simulation system with 28 GBaud dual-polarization (DP) uniform 16/64 quadrature amplitude modulation (QAM) signals after 1200 km/400 km transmission is constructed. The results show that compared to the classical digital backpropagation with 20 steps per span (DBP-20StPs) under the same performance, the computational complexities of real-valued multiplication and real-valued addition for nonlinear compensation of 16/64QAM signals in this scheme are about 10.02% and 14.53%, 20.03% and 29.05% of that of DBP- 20StPs, respectively.