Shortening Burn-In Test: Application of a Novel Approach in optimizing Burn-In Time using Weibull Statistical Analysis with HVST

Abstract

Burn-in and stress testing are becoming increasingly important, a sine qua non in the electronics industry as customers become increasingly sensitive to failures occurring in the useful life of a product or system. Burn-in subjects the product to expected field extremes by exposing the product to accelerated temperature and voltages stress to screen infant mortalities (latent failures). In the past, burn in duration studies use constant failure rate statistics to model the classical bathtub curve describing early-life failure behavior of the product throughout its operating lifetime. Thus, FIT (Failure in Time) rate calculations were greatly inflated by including failures occurring long after the time of interest. This manifests in a requirement for a very low failure rate after the conservative burn-in stresses causes the need to sample larger number of units in order to differentiate between the passing and failing criteria. Furthermore, it makes the Burn-In study more complicated and reduces the chance of success.