Numerical approach to predict power device reliability

Abstract

The scope of this paper is to work out a predictive method to estimate the power device reliability under active cycle tests. The proposed method is able to predict, through a numerical model, the local maximum temperature during test. The results validation has been pursued correlating the numerical thermal maps results with the experimental temperature distribution obtained from an infra-red camera. Front metal ratcheting has been recognized as the main root cause of contact resistance degradation during the considered reliability test (Repetitive Avalanche). This failure mode is dependent on the temperature variation for cycle, by which it is possible to predict the device lifetime according to the Coffin-Manson fatigue model.