Semi-supervised Regression Model for Eye Diagram Estimation of High Bandwidth Memory (HBM) Silicon Interposer

This paper presents a semi-supervised regression model for accurate eye height and eye width estimation of high bandwidth memory (HBM) silicon interposer channels. The model combines a channel characteristic-based deep neural network with a co-trained convolution neural network. For verification, the proposed model was applied to the stripline channels of an HBM silicon interposer. The root-mean-square errors (RMSE) of the predicted results of eye width and eye height are 4.95 mV and 7.32 ps.