Single-Event Data Analysis

Data needs to be carefully examined and sometimes corrected before used to predict upset rates in space. The data needs to be examined for illegitimate, systematic and random errors. Blunders and systematic errors are very likely during single-event data accumulation as repeated changes of device and ion or energy correspond to performing a large number of different experiments in a very short time. The most likely types of systematic error are due to dosimetry. They may occur occasionally, repeatedly if the dosimetry is not controlled, or with large standard deviations with controlled dosimetry. The dosimetry can be checked using golden chips, monitoring internal data consistency, occasional repetition of a measurement or calculation of observed versus expected values. The calculated linear energy transfer (LET) may need to be corrected if the device exhibits the funnel effect. The depth of the device leads to discontinuities in the data as ions are changed. The correction for this effect enables the depth of the device to be determined and places more values on the cross section curve. The data is often fitted using the Weibull function. This can be misleading as a single set of data can be fit with a variety of width and cross section values. It is often valuable to fit data with the log normal function, especially when studying the changes in the cross section curve with changes of some parameter. What appears to be a change of cross section is usually a change in the critical LET. The dependence of the critical LET on parameter can lead to a better understanding of the device.