Eye Diagram Analysis with Deep Neural Networks for Signal Integrity Applications

Abstract

To facilitate the design for signal integrity in interconnect networks, this study explores the application of a deep neural network called convolutional neural network in eye diagram recognition. A multi-module memory bus interconnect structure is built and simulated. The eye diagrams for different types of signal impairments are generated using the ADS circuit model and used as the training data for convolutional neural network. Three basic signal impairments and their combinations were studied in the experiment. The results validate that the CNN model developed in this work can accurately identify the types of signal impairments and even locate the position of the signal impairments. Machine learning can also improve the eye diagram metrics with the help of linear regression algorithm.