On techniques for handling soft errors in digital circuits

Abstract

Dealing with soft errors due to particle strikes is the next major challenge in implementing digital systems. This study thoroughly investigates the effect of device size on circuit soft error rate and identifies methods to reduce soft error rate in combinational circuits. In particular, we propose three novel methods that upsize only selected gates and /or transistor networks. In order to obtain the most appropriate technique for soft error rate reduction in small technology node circuits, we conduct experiments and compare the results for several upsizing techniques including all gates, selected gates and transistor networks based on their fault sensitivities, and parallel networks with soft error rate saturation consideration. Consequently, it is discovered that some upsizing scenarios perform large improvement whereas others do not or even increase the soft error rate. The use of fault sensitivity analysis approach with parallel transistor network upsizing based on the contribution of each sensitive gate can reasonably reduce overall circuit sensitivity. Experimental results show an average reduction in soft error rate about 20% with a very small area overhead of 2% for benchmark circuits using our technique.