Chen Xinyu, Ming Hao, Li Chenjia, He Guangqiang, Zhang Fan
Opt Lett. 2021 Apr 1;46(7):1700-1703. doi: 10.1364/OL.422195.
We propose an artificial neural network (ANN)-based scheme to improve the performance of nonlinear frequency division multiplexing (NFDM) optical transmission systems in both time and frequency domain. Through two-stage ANN equalization at the receiver side, time-domain distortions between adjacent bursts and frequency-domain cross talk between neighboring subcarriers can be jointly mitigated. Burst ANN and Subcarrier ANN equalizers are characterized and validated by numerical simulations of a dual-polarization NFDM transmission system. Compared with the basic detection scheme, the proposed two-stage ANN achieves a Q-factor gain of 3.01 dB for an NFDM system with 256 Gb/s gross data rate transmitting over 960 km standard single-mode fiber (SSMF). The two-stage ANN approach offers an effective way to jointly equalize the signal in multiple dimensions.
我们提出了一种基于人工神经网络（ANN）的方案，以在时域和频域提高非线性频分复用（NFDM）光传输系统的性能。通过接收机端的两阶段ANN均衡，可以共同减轻相邻突发之间的时域失真和相邻子载波之间的频域串扰。通过双偏振NFDM传输系统的数值模拟，对突发ANN和子载波ANN均衡器进行了特性分析和验证。与基本检测方案相比，对于在960 km标准单模光纤（SSMF）上传输256 Gb/s总数据速率的NFDM系统，所提出的两阶段ANN实现了3.01 dB的Q因子增益。两阶段ANN方法提供了一种在多个维度上联合均衡信号的有效方法。
