Electrical Parameters Prediction for Fab-to-Fab IC Product Migration

Abstract

In the case of production migration of a specific semiconductor device from one fabrication plant to another, yield estimation in the target fab can be accelerated by employing information from the source fab, assuming that the process parameter distributions in the two fabrication plants are similar, but not the same. In this paper, we employ a five-block yield estimation methodology with different methods of feature selection and prediction, which are based on the same operating scheme but in which the method of features selection and the algorithm used for efficient yield prediction of a device in the target fab, vary. Each of the five blocks adopts prior knowledge from the source fab in order to predict the yield in the target fab. The methods used for features selection are Pearson Correlation Coefficient, Distance Correlation, Maximal Information Coefficient (MIC) and Hilbert-Schmidt Independence Criterion Lasso (HSIC Lasso). Linear Regression, Polynomial Regression and Multivariate Adaptive Regression Splines (MARS) are the algorithms used to obtain quick and accurate yield estimates at the onset of production in the target fab.