Modeling for Nonlinear High-speed Links Based on Deep Learning Method

Abstract

With the nonlinearity degree increasing in high-speed links, the traditional fast time-domain system simulation techniques based on linear time-invariant hypothesis cannot accurately predict the system response. In this paper, a modeling method based on deep learning method is proposed to develop a time-domain model to accurately deal with the nonlinear factor in the high-speed links. Compared with the SBR-based fast time-domain simulation method, the simulated results show that the proposed modeling method based on deep learning can facilitate accurate and fast transient simulation.