Gentest is a new system for the automatic generation of test sequences for sequential circuits. After a rather general survey of typical problems in test pattern generation, this paper outlines the Gentest architecture with an emphasis on its component STG2, the actual test sequence generator for a given fault. This description is held at a rather superficial level. For instance, the authors mention that “equivalent faults” will be “collapsed” without saying how fault equivalence is detected in their system (they define faults to be equivalent if they are detected by the same test sequences). As another example, they sketch initialization procedures for the case in which initial state information is missing, but do not address the question of how their test sequences will cope with faulty circuits which cannot be initialized.
The paper discusses the merits and drawbacks of several known algorithms for test generation, including the Podem algorithm, variants of the D algorithm, the back algorithm, and the split value model. The presentation of the analysis is fairly superficial, so that it is difficult to verify the assertions; there are no definitions or proofs.
The final section of the paper reports on experimental results concerning some of the standard benchmark circuits. These results are quite impressive.
Given the information provided in this paper, it is hard to judge the provable quality of the Gentest system. Some of the issues the paper claims to deal with are theoretically much harder than it makes apparent. On the other hand, the system seems to be very useful in practice.
For further reading on this subject, I recommend two references that address some of the theoretical and algorithmic issues in test generation in a rigorous mathematical fashion, a book by Graf and Gössel [1], and a paper by Brzozowski and me [2]. The emphasis of Graf and Gössel [1] is on fault-detection circuits. Some of the results are directly relevant for test generation as well, however. Reference [2] cites several additional relevant papers, most of them published before 1985 in European journals.