Defect Detection for the TSV Transmission Channel Using Machine Learning Approach

Through silicon via (TSV) is a key enabler for 3D integration technology to provide high-density and high-speed transmission channels. However, the serious defects including open circuit and short circuit will result in functional failure. In this paper, a method to identify the open and short defect and detect the defect position is developed based on machine learning approach. Taking S parameters as training dataset, the predictive model is trained and evaluated, the performance of models based on bagging method and boosting method are compared. The results demonstrate the predictive model could correctly classify the defect type and predict the defect position accurately.