A General Method of Time Delay Estimation and Compensation Based on CNN in Optical Parallel Channels

A general time delay estimation and compensation method based on multi-path effect analyses and convolutional neural network (CNN) for synchronization of optical parallel channels is proposed. Experimental verification is conducted in a 16-channel 400G system, and the root mean squared error (RMSE) of predicted time delay is 4.81ps, which meets the requirement for no error bit.