Neural Network Techniques for High-Speed Electronic Component Modeling

Abstract

This paper provides an overview of neural network applications for high-speed electronic component modeling. Two neural network techniques, an electromagnetic (EM)-based neural modeling approach and a state-space dynamic neural network method, are presented. Accurate neural models are obtained to represent behaviors of EM component and nonlinear circuit. The neural models are used in signal integrity design of high-speed integrated circuits, helping to improve efficiency in circuit simulation and optimization.