Statistical modeling for particle impact noise detection testing

Particle impact noise detection (PIND) testing is used to test electronic devices for the presence of conductive particles which can cause catastrophic failure. A statistical model based on the rate of particles induced by the test vibration, the escape rate, and the false alarm rate is presented. Based on data from a large number of PIND tests for a canned transistor, the model is shown to fit the observed results closely. Knowledge of the parameters for which this fit is made is important in evaluating the effectiveness of the PIND test procedure and for developing background judgment about the performance of the PIND test. By varying the input parameters to the model, the resulting yield, failure rate, and percent fallout can be examined and used to plan and implement PIND test programs.<>