Parametric and Threshold Studies of Single Event Sensitivity

There are a number of papers that examine single event effect (SEE) sensitivities. This paper examines published results and demonstrates how the SEE critical charge is related to various parameters. Depending on the type of circuit and the type of test, the critical charge may increase or decrease linearly, or increase quadratically or as the square root of the test parameter. Parametric studies of SEE sensitivities are a very important tool for understanding SEE phenomena; however, they are much more informative when the data is fit using the log normal distribution. We examine the effect of various threshold fits on predicted space upset rates by comparing fits with Weibull functions and log normal functions. As the threshold region makes a very small contribution to the total upset rate, variations in the shape of the curve used to describe the threshold make a very small change in the upset rate. The exception is for very hard devices, where the threshold region makes a larger contribution to the upset rate as the steeper slope of the high LET part of the LET spectrum gives greater emphasis to the onset region. The relative number of upsets in space is the only operationally reasonable measure of device sensitivity. Relative upset sensitivity remains the same for nearly all proton and heavy ion intensive orbits. Scatter plots of space-upset rates versus device thresholds show little correlation. Device onset thresholds, whether determined by the first bit to upset with an exposure of 10⁷ ions/cm² or by the Weibull threshold, are not an accurate measure of device sensitivity. If one uses the intuitively obvious method of requiring a particular onset threshold as a surrogate for a given upset level requirement, wrong decisions will be made approximately 30% of the time. If upsets are observed at LETS below 15 meV/mg/cm² , upset rates for protons in low earth orbit may be important. If upsets are observed below 120 meV/mg/cm² (200 for SOI), system survivability needs to be examined. It is not necessary to use the onset threshold for either of these decisions. The threshold determined by laser testing is not the same as the onset threshold and this measurement may be valuable for hardness assurance purposes. There is a strong correlation between the figure of merit (FOM) and the space upset rate so this approach can be used for quick estimates of system performance.

I. Introduction

There are several commonly used approaches to single event effect (SEE) data acquisition, plotting, and analysis that need to be carefully examined by researchers in this area. This paper will attempt to identify these and the associated problems and present alternate approaches. The problem areas are: 1) Plotting experimental data using logarithmic cross section versus linear energy transfer (LET) on a linear scale, 2) Not extending the experiment to high LET, 3) Fitting the data with the Weibull distribution, 4) Overemphasis on acquiring data in the threshold region, and 5) Using the onset threshold as a parameter for indicating relative SEE rates in space.