Weighted test generator in built-in self-test design based on genetic algorithm and cellular automata

Abstract

Weighted pattern generation is an effective method for cutting down the test length of pseudorandom test pattern set in a built-in self-test (BIST) design. For its natural weighting structure without additional hardware overhead, cellular automata (CA) was applied as test pattern generator of BIST in this paper. Furthermore, optimizing schemes based on genetic algorithm (GA) were also adopted so as to approach the desired weight of circuit under test (CUT) more efficaciously. Preparative programs consists of encoding the rules of a CA, constructing chromosome, calculating fitness of the chromosome, and selecting an individual for performing genetic operations, etc‥ Then, the characteristic of the individual is evaluated by judging whether the obtained weight is an approximate value to the desired weight or not. Finally, an optimized rule value set was searched and therefore an actual weight set and corresponding test set are also achieved. Experimental results based on some ISCAS'85 benchmark circuits show that this weighted pattern generation structure with CA based on GA is efficient in diagnosing some difficultly-detected faults and improving fault coverage.