Characterization of the time-dependent reliability fallout as a function of yield for a 130nm SRAM device and application to optimize production burn-in

A model is presented to optimize burn-in time based on product yield. A relationship between yield and reliability is developed and demonstrated with tests on a 130nm SRAM test device. Voltage and thermal acceleration are added to this model to appropriately scale predictions to product use conditions. High-voltage test screening is found to have a similar effect as burn-in in reducing infant mortality and thus can also be incorporated into this model. A conditional reliability model is needed to model the effects of both burn-in and high-voltage screening; these effects cannot be modeled as if they simply reduce the defectivity of the population. This approach can be used to predict the amount of high-voltage test screening and burn-in needed in order to qualify a production device built with the same technology.