Comparative Analysis of Upset-Multiplicity Occurrences due to Flash X-rays and Pulsed Neutrons in Commercial SRAMs

In this paper we propose an approach to distinguish between upset bursts induced by flash X-rays and pulsed neutrons, by analyzing upset-multiplicity occurrences. Test results of ISSI 65 nm SRAMs on a flash X-ray machine and a pulsed reactor are presented. We adopted binomial distributions to fit the obtained upset-multiplicity data. The pulsed-neutron results exhibit precise and consistent accordance with the binomial distributions. In contrast, the flash X-ray data sharply diverge from the binomial distributions as the dose rate exceeds the upset threshold. However, the discrepancy reduces significantly after the bit-upset proportion saturates at higher dose rates. Interestingly, the binomial distributions fit well with the flash X-ray data at all dose rates if combinations of 2 or 3 different binomial distributions are applied. To explain the observed phenomena, the underlying mechanisms of dose rate upsets and neutron-induced single event upsets are further discussed. Above all, the presented approach of analyzing upset-multiplicity occurrences of upset bursts provides a novel and straightforward means to differentiate localized and global effects, and is especially useful for dose-rate-upset analysis when the correspondence between physical and logical addresses is unavailable.