论文信息 - A Low-Complexity Recurrent Neural Network Based Joint Equalization and Decoding Method for Trellis Coded Modulation Link in Data Center

A Low-Complexity Recurrent Neural Network Based Joint Equalization and Decoding Method for Trellis Coded Modulation Link in Data Center

RNN-based joint equalization and decoding methods are proposed for data center with TCM signals. Numerical results show compared with traditional DSP algorithms, our method improves power sensitivity by 2.2dB@BER=3.8×10-3 with 96% reduced complexity.

[1] Roberto Llorente,et al. Performance Comparison Between 4D Trellis Coded Modulation and PAM-4 for Low-Cost 400 Gbps WDM Optical Networks , 2016, Journal of Lightwave Technology.

[2] Jeremie Renaudier,et al. Multi-dimension coded PAM4 signaling for 100Gb/s short-reach transceivers , 2016, 2016 Optical Fiber Communications Conference and Exhibition (OFC).

[3] Weisheng Hu,et al. Spectral Efficiency Improved 2D-PAM8 Trellis Coded Modulation for Short Reach Optical System , 2017, IEEE Photonics Journal.

[4] Liang Zhang,et al. Advanced modulation formats for 400-Gbps short-reach optical inter-connection. , 2015, Optics express.

[5] Talha Rahman,et al. 4D PAM-7 Trellis Coded Modulation for Data Centers , 2019, IEEE Photonics Technology Letters.

[6] Zhangyuan Chen,et al. Toward Single Lane 200G Optical Interconnects With Silicon Photonic Modulator , 2020, Journal of Lightwave Technology.

[7] Chen Xu,et al. Low-Complexity Bi-Directional Recurrent Neural Network Equalizer for Short-Range Optical Interconnect Links , 2019, 2019 Asia Communications and Photonics Conference (ACP).