Accurate alpha soft error rate evaluation in SRAM memories

Abstract

Radiation sensitivity of SRAM memories is of vital importance in applications demanding high reliability levels. Soft error rates (SER) are usually determined through accelerated tests where target devices are subjected to very high levels of radiation, in order to increase the number of induced events. Two main factors determine the accuracy of this technique, on one hand, when the number of induced events is low the result is subjected to statistical errors, on the other hand, if the number of induced events is high, the probability that each cell experiences more than one event is increased, as a result there is the risk that a fraction of the events will not be counted. In this paper we propose an accelerated test method to determine the soft error rate in SRAM memories from a model based on the evolution of the cell population being at logic zero (or logic one) during the irradiation experiment. This model has been contrasted with experimental results obtaining a very good correlation.