Good has many meanings . The web site http://www.dictionary.com has over 50 definitions for the word. The definition used in this post is number 5, well behaved. A good, or well behaved test is one that passes the parameter being measured when it is within limits and fails when the measurement is outside of the limits, all of the time.
Since it not possible to evaluate a test for the “all of the time” criteria it is necessary to turn to statistics to provide a measurement for the goodness of a test. And, since test is a process a good statistic to use is the capability index (Cp) and the centering capability index (Cpk). A good explanation of the definitions of Cp and Cpk can be found at the QIMacros website, http://www.qimacros.com/qiwizard/process-capability.html
For an complete evaluation of the test it would be necessary to test a number of different boards on different test systems. However, executing the test on even one Unit Under Test on one tester will reveal a good amount of information about the test. For instance, is the Cpk less than 1? In that case there will definitely be problems. Typical industry demands are for a Cpk of 2 or over.
No matter what, if the tests are not evaluated and the results recorded trouble looms ahead. Having the evaluation proves that at least one moment in time the test was well behaved.
- Rick Wagner