Horizontal diversity in test generation for high fault coverage

Abstract

Determination of the most appropriate test set is critical for high fault coverage in testing of digital integratedcircuits. Among black-box approaches, random testing is popular due to its simplicity and cost effectiveness. An extensionto random testing is antirandom that improves fault detection by maximizing the distance of every subsequent test patternfrom the set of previously applied test patterns. Antirandom testing uses total Hamming distance and total cartesiandistance as distance metrics to maximize diversity in the testing sequence. However, the algorithm for the antirandomtest set generation has two major issues. Firstly, there is no selection criteria defined when more than one test patterncandidates have the same maximum total Hamming distance and total cartesian distance. Secondly, determinationof total Hamming distance and total Cartesian distance is computational intensive as it is a summation of individualHamming distances and cartesian distances with all the previously selected test patterns. In this paper, two-dimensionalHamming distance is proposed to address the first issue. A novel concept of horizontal Hamming distance is introduced,which acts as a third criterion for test pattern selection. Fault simulations on ISCAS’85 and ISCAS’89 benchmarkcircuits have shown that employing horizontal Hamming distance improves the effectiveness of pure antirandom in termsof fault coverage. Additionally, an alternative method for total Hamming distance calculations is proposed to reduce thecomputational intensity. The proposed method avoids summation of individual Hamming distances by keeping track ofnumber of 0s and 1s applied at each inputs. As a result, up to 90% of the computations are reduced.