Power-aided probabilistic shaping parameter optimization for long-haul optical fiber transmission

Abstract

A power-aided probabilistic shaping (PS) parameter optimization scheme is proposed to localize the optimal transmitting setup and maximize the transmission capacity in long-haul optical fiber transmission. Using a neural network and genetic algorithm (NNGA), the proposed scheme can select the optimal transmission parameter set, including the launching optical power (LOP) and three PS characteristics. We built a testbed using MATLAB and VPI to validate the proposed scheme. The proposed scheme was demonstrated in a dual-polarization coherent transmission system. The results show that the proposed scheme improves the generalized mutual information (GMI) by 0.3035 bits/symbol/pol and normalized GMI by 0.1022 in a 1000 km G.654E fiber transmission at 420-Gbit/s, compared to a traditional MB-based PS technique. The signal-to-noise ratio (SNR) achieves a gain of 1.003 dB, which validates the outperformance of the proposed scheme.