A CNN-based One-shot Blind RX-side-only Equalization Scheme for High-speed SerDes links

This paper proposes a Serializer/Deserializer (SerDes) equalization parameter optimization technique based on circular convolutional neural network (CNN). This method facilitates rapid and precise acquisition of optimal receiver (RX) side equalization parameters for various high-speed channel types by iterative learning processes. Upon completing neural network training, the parameter configurations are directly determined according to the current channel without iterative computation. Comparing with other equalization algorithms, it is able to optimize the equalization parameters of continuous time linear equalization (CTLE) and decision feedback equalization (DFE). Owing to the absence of iterations and RX-aligned training sequences, this approach accelerates the design and simulation of the equalizer, significantly enhancing SerDes equalization optimization efficiency.