An Alternative Approach to Traditional Lifetime Testing Based on Comparative Studies and Survival Analysis

Abstract:
Modern electronic devices and assemblies feature incredibly long lifetimes when designed properly. Experimental verification is still vital to warrant consistent quality. However, conventional accelerated lifetime testing requires numerous experimental resources, as testing has to be repeated at different stress levels and can last for thousands of hours. This work suggests comparative studies as a viable alternative. In this type of study a sample is tested at just one stress level and the results are compared to a baseline data set. The statistical models and methods for this analysis are presented and the trade-offs over more traditional methods are discussed. Their application is illustrated using a data set from a comparative study of high-end and low-cost offline LED drivers. Through the use of the Cox Proportional Hazard model the higher reliability of the high-end drivers can be demonstrated, albeit with a minor statistical significance.