Characterizing the Uncertainty in Dosimetry Using Data From Multiple Tests

Recently, errors in dosimetry have become of renewed interest together with their effect on the accuracy of estimating on-orbit upset rates . In those studies, the focus was on zero mean errors in the dosimetry-the dosimetry for runs with different ions was assumed to be measured with error, but that these errors were zero mean. There is anecdotal evidence, however, that the errors in dosimetry may not be zero mean. From a single set of tests it is impossible to estimate the magnitude of this non zero-mean effect. From multiple tests of the same part, it is possible if it is assumed that the true cross-section of the part does not change, and that the fluctuations around the common non zero mean are independent. We show non zero-mean errors in dosimetry of between 10 and 30% in data sets that were not collected specifically for this study. The presence of non zero-mean errors can have a large effect on the uncertainty in the experimental determination of a part's cross-section. When estimating upset rates for a device, users of experimental test data must use their best judgement as to the magnitude of the non zero-mean error that may be present.