ESSENTIAL: an efficient self-learning test pattern generationalgorithm for sequential circuits

摘要

The authors present ESSENTIAL, deterministic automatic testpattern generation algorithm for sequential circuits. By combiningreverse time processing over time frames and forward processing withintime frames, ESSENTIAL avoids the detrimental a priori determination ofa topological path to be sensitized or of a primary output, to which thefault effects have to be propagated. Moreover, the proposed testgeneration approach fully exploits the beneficial techniques that havesuccessfully been used for combinational circuits by the automatic testpattern generation system SOCRATES. In particular, the authors discuss alearning procedure for global implications not only over reconvergentfanout, but also over time frames as well as static and dynamic uniquesensitization techniques. After introducing a couple of intelligentheuristics employed for guiding and supporting the decision-makingprocess, the authors report some preliminary but encouragingexperimental results