Aldaya Ivan, Giacoumidis Elias, Tsokanos Athanasios, Jarajreh Mutsam, Wen Yannuo, Wei Jinlong, Campuzano Gabriel, Abbade Marcelo L F, Barry Liam P
Opt Lett. 2020 Oct 15;45(20):5820-5823. doi: 10.1364/OL.403778.
A novel nonlinear equalizer based on a multiple-input multiple-output (MIMO) deep neural network (DNN) is proposed and experimentally demonstrated for compensation of inter-subcarrier nonlinearities in a 40 Gb/s coherent optical orthogonal frequency division multiplexing system. Experimental results reveal that MIMO-DNN can extend the power margin by 4 dB at 2000 km of standard single-mode fiber transmission when compared to linear compensation or conventional single-input single-output DNN. It is also found that MIMO-DNN outperforms digital back propagation by increasing up to 1 dB the effective-factor and reducing by a factor of three the computational cost.
提出了一种基于多输入多输出（MIMO）深度神经网络（DNN）的新型非线性均衡器，并通过实验证明其可用于补偿40 Gb/s相干光正交频分复用系统中的子载波间非线性。实验结果表明，与线性补偿或传统单输入单输出DNN相比，MIMO-DNN在2000 km标准单模光纤传输时可将功率裕度扩展4 dB。还发现，MIMO-DNN通过将有效因子提高多达1 dB并将计算成本降低三分之一，优于数字反向传播。
