Digital VLSI circuits are tested by applying to their input pins a set of patterns that exercise some or all of the circuit's functions. During fabrication, some of these functions may develop faults that are sensitive to the test patterns. In most cases, however, a test pattern cannot, for practical reasons, test all faults since that would require use of exhaustive data patterns. Therefore, although a circuit may pass the test, there is no guarantee that the circuit is fault-free. Thus, it is important to determine how well a test can identify a faulty circuit.
The most common criterion for judging the quality of a test is the fault coverage obtained from a fault simulator. Typically, fault coverage refers to the percentage of single faults detected by the test. As a rule, test engineers attempt to provide as close as possible to 100% fault coverage. Due to the difficulty in developing such tests, however, in practice, a fault coverage that is too high is difficult to achieve.