SINGLE EVENT EFFECTS RADIATION TESTING REQUIREMENTS: A BAYESIN AN APPROACH

Abstract

A key aspect of the safe operation of computing systems in environments with exposure to ionizing particles (such as ions, protons, and neutrons) is single-event effects (SEE) hardness assurance. Statistically reliable determination of the probability of failure-free operation of integrated circuits (ICs) based on radiation testing results poses several fundamental challenges. In the absence of observed failures during testing, it becomes nearly impossible to interpret the results conclusively without utilizing prior knowledge about the device and the nature of SEEs. Prior quantitative information about SEE behaviour of the devices of the same type can be used to justifiably reduce test requirements while the desired confidence level in the hardness assessment. This paper proposes a method for determining testing requirements based on Bayesian methodology. In this approach, SEE sensitivity parameters are treated as a random vector variable, and the prior probability distribution for this variable is constructed from available empirical data. Both parametric and non-parametric methods for constructing the prior distribution from empirical data are considered. The analysis reveals that the uncertainty in testing requirements estimation is predominantly influenced by the prior information and its representation, while accounting for sample-to-sample variability and particle fluence measurement errors only slightly increases test requirements. The proposed approach was also used for assessing the confidence level of hardness requirements compliance check based on prior information without conducting actual tests.

Keywords

• test standard, radiation hardness, prior information, probability of failure-free operation, bayesian probability.