Statistical defective subpopulation model for risk analysis: usage case study to compare sensitivity to delamination defect, for automotive semiconductors

Abstract

In semiconductor industry for automotive, risk analysis required by customers when facing failures in field, are performed by field modeling: this modeling allows to estimate quantity of failures that customer may expect in field in the next months or years. Field model choice and modeling results can be validated by reliability tests performed in parallel, or by studying sensitivity to defect from design or application considerations. This paper presents a rare case study when a same inter-layer dielectric defect impacted several analog products and sensors: thus, it was possible to compare sensitivity to defect between the different products, and in particular the specific field model for this defect and its parameters according to each product. This case is all the more interesting as the best model for this delamination defect in term of fitting on observed field failures, was in defective subpopulation.