Fault secure property versus strongly code disjoint checkers

摘要

The final checker of a self-checking system is an embedded double-rail checker (the partial checkers have in general two outputs). The self-testing or the strongly code disjoint property of this embedded checker can be lost if it is not exercised by an appropriate set of inputs. If some partial checkers are strongly code disjoint, then some undetectable faults can modify the input/output mapping of these checkers. This can compromise the exercising of the final double-rail checker especially if this modification leads to the reduction of the partial checker's output code space. In that case it will be required that the partial strongly code disjoint checkers must also be strongly fault secure. In this work we show that the (two-output) strongly code disjoint checkers do not allow such reduction of their output code space and on this basis we show that the strongly fault secure property is not necessary. We also give some techniques that ensure exercising the final checker.