230 VLSI Test Principles and Architectures values greater than or equal to one-half of the best fitness from the previous GA run, and (3) random individuals to fill the entire population if needed. Instead of targeting individual faults, the GA tries to detect as many faults as possible by any individual. Because the target is to generate a sequence that can detect as many faults as possible, parallel-fault simulation (on 31 faults) is used during fitness evaluation; 31 faults are used instead of 32 due to the nature of the embedded fault simulator. A set of 31 undetected faults in the fault list are selected as target faults. All individuals in the population would then target the same group of 31 faults. For successive GA runs, faults are chosen cleverly so that efforts can be reduced. For instance, if the best sequence added to the test set detected a total of 20 faults, it may have also excited and propagated some faults to one or more flip-flops at the end of that sequence. These activated faults should be placed in the next targeted
This is the end of the preview.
access the rest of the document.