GA based diagnostic test pattern generation for transition faults

Accuracy of any diagnosis algorithm depends on the test set used. Test that is able to distinguish more fault pairs is better suited for aiding diagnosis. Standard detection test set is generated to detect faults using less number of test patterns. It is unable to distinguish many fault pairs. To distinguish pairs, more patterns are required that consumes ATE memory and time. In this work we devised an algorithm to generate tests from the detection test itself. The generated tests distinguish transition fault pairs undistinguished by the detection test. Our distinguishing tests are made from the detection test by rearrangement of pseudo-primary and primary input of the two-pattern tests. Additional tester memory is not required for our diagnostic tests.