The Containment Set Approach to Upsets in Digital Systems

Fault analysis of digital systems is highly dependent upon the fault model employed. Much previous work utilizes fault models known to contain inaccuracies in order to permit mathematically tractable analysis. In this correspondence a new approach is taken which combines faults, hardware, and software together into one overall model. This new model is shown to be useful for the consideration of intermittent/transient faults. It supports a new method, based on the novel concept of a containment set, for realizing transient fault tolerance without massive redundancy. It also allows for a new approach to system fault tolerance evaluation and validation which uses a transition matrix which is defined in terms of the containment set.