Single event effects radiation testing requirements: a minimax approach

Abstract

One of the key challenges of single-event effects (SEE) testing of integrated circuits (ICs) and semiconductor devices is determining the "test requirements". This refers to the level of exposure at the test facility that allows for conclusions about the device’s compliance with specified requirements to be drawn with a given confidence level, based on experimental results. Statistically reliable determination of the probability of failure-free operation (PFO) from the test results with small sample sizes is a difficult problem, especially for newly developed ICs. This challenge becomes even more complex when no failures are observed during the experiment. In the context of SEEs in ICs caused by cosmic particles, the issue lies in the combination of high specified fluence values and the required PFO. Analysis shows that even (hypothetically) a highly detailed reproduction of the special factors (SF) during testing is generally not sufficient by itself to ensure the desired confidence level in the results. A justified reduction of the test requirements can be achieved by incorporating prior quantitative information on the SEE behaviour in the devices being tested, combined with model-based representations of SEE mechanisms. This paper discusses a minimax approach to determining test requirements, where prior information is introduced in the form of "hard" constraints on the values of the problem’s parameters.

Keywords

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