Fast and accurate deep neural network (DNN)-based eye-height and eye-width estimation method

Abstract

This paper propose a fast and accurate eye-height and eye-width estimation methods based on deep neural network. The proposed method estimate the eye-height and eye-width fast and accurately by simply reducing the complicated process from design parameters to eye-height and eye-width. Unlike previous methods, the proposed method does not require any additional simulation, so eye-height and eye-width estimation are performed in a few seconds. In addition, various methods were used to optimize the proposed model to lower the error rate and accurately estimate eye-height and eye-width. In this study,the proposed method was applied in PCB size channel and memory channels of HBM interposer. This is a fast and accurate way to estimate eye-height and eye-width of various high-speed channels. At the same time, it is easy to estimate eye-height and eye-width even on channels containing various noises such as crosstalk or ISI. This study was applied to the HBM interposer memory channel and the proposed eye-height and eye-width estimation method of this study is faster and more accurate than the previous eye-height and eye-width estimation methods.