Abstract
The primary purpose of digital circuit manufacture testing is to detect defective parts so that they will not be sold to customers. Predicting the defective part level, which results after a set of test patterns has been applied, is not a simple problem, but such a prediction is necessary if the effectiveness of different test sets is to be analyzed without doing extensive surrogate simulation. The final defective part level depends upon the test vectors applied, the circuit structure, and the nature of the defects which are present. In order to detect a defect, a test pattern must both excite the defect and propagate the resulting incorrect logic value through the circuit to a primary output. This research introduces a model which accurately predicts the defective part level based upon the number of times each circuit site has been observed and the probability of exciting an undetected defect given that that site is observed. It is shown conclusively that the probability of excitation decreases exponentially as the number of observations at a circuit site increases. This information is then inserted into the MPG-D defective part level model which can be used to accurately predict the final defective part level that results from the application of different test pattern sets.
Dworak, Jennifer Lynn (2000). Modeling the probability of excitation and the defective part level as testing progresses. Master's thesis, Texas A&M University. Available electronically from
https : / /hdl .handle .net /1969 .1 /ETD -TAMU -2000 -THESIS -D86.