Abstract:

The predicted reliability of an electronic component is dependent on application conditions, part characteristics, and modeling constants. Information related to each of these factors must be found (or approximated) for a proper reliability estimation to be made. As each input can be treated as a distribution representing information uncertainty, the more information that can be found, the less uncertain the final time to failure distribution assessment will be. In this paper, we show three main input aspects in the context of methods of collecting information and quantifying the uncertainty in the information. We evaluated the information that is available for COTS and non-COTS semiconductor parts that can be used for completing reliability assessments. When such information is not available, we developed methods for quantifying information availability, consistency, and quality for parts, and finally evaluate the different sources of uncertainty in failure model inputs and their impact on reliability assessments.