Equalization of Short-Reach IMDD Transmission With Enhanced Reservoir Computing
Yixian Dong, Yiqian Shi, Jiacheng Feng, Zhinan Sun, Lin Jiang, Xihua Zou, Wei Pan, Lianshan Yan
Abstract
By adding output feedback and a sliding window, a software-based enhanced reservoir computing (RC) is proposed and verified in intensity modulation direct detection (IMDD)-based short-reach optical transmission systems. Both simulation and experiment confirm the enhanced RC is effective for dispersion and nonlinearity compensation for short-reach IMDD systems. Moreover, compared with the original RC without output feedback and sliding window, the enhanced RC can reduce the required received optical power (ROP) by 2 dB at 7% FEC limit, and decrease reservoir layer neuron number from 1800 to 300 without jeopardizing the performance within 25 km transmission. In addition, traditional DSP equalizers, including feed-forward equalizer (FFE), decision feedback equalizer (DFE), Volterra equalizer and neuron network equalizers, such as feedforward neural network (FNN), long-short-term memory network (LSTM) and bidirectional LSTM (BiLSTM), are conducted for a thorough comparison with the proposed RC equalizer. The corresponding experimental results show that the enhanced RC demonstrates lower BER and longer transmission distance with a smaller sliding window size.