Establishing realistic test tolerance limits for test program sets

A statistical analysis approach is taken to determine test tolerance limits for outputs of units under a test program set. The root-sum-of-squares (RSS) statistical analysis used is predicted to provide for a 99.7-percent yield during any type of performance testing; should the Monte Carlo approach be taken, it is expected to provide the same results for a sufficiently large number of trials. The RSS is contrasted with the worst-case and systematic analysis approaches in a computer analysis; a yield of 99.55 percent was obtained for RSS. An empirical verification of the computer analysis is also presented.