Exploring diagnostic capabilities of software-based self-tests for production and in-field applications

Self-test and diagnosis is an inherent part of fault tolerance approaches, which are realized by active hardware redundancy. The test must be available in-field and has to provide a diagnostic resolution, which matches the repair function of the fault tolerance approach. Continuous development lead to ever fine grained repair functions, which require a growing diagnostic resolution. Typically, fault dictionary based methods are not adequate here, because the required in-field memory exceeds the available resources, or a mapping from a structural fault to a functional misbehavior is not possible. In this paper we survey different possibilities for diagnostic software-based self-test (SBST) approaches. Finally, we show that a functional diagnosis by software is possible in-field. The proposed method reduces the size of the fault dictionary for the examined working example from ~8,9MB down to only 42 bytes.