Pass-fail windows

On a production line, the devices under test (DUTs) have a set of specifications from R&D or the customer that they must comply with. This can be viewed as a window on a graph.

During testing, if the response falls within the border of the window, it passes, but if it falls outside, it fails. This is a simple process, until uncertainty is incorporated. Uncertainty can come from a number of areas, but a primary area is tolerances in sensitivity or response curves.

Tolerance, false passes, and false failures

All microphones have a plus/minus to their measurement accuracy, but some are more precise than others. Most production line microphones have a variation between microphone responses of around ±2 dB between microphones, in the range of 50 Hz and 10 kHz. 

Using a simplified example, this means that if the DUT needs to output a signal between 36 and 40 dB at a given frequency to pass, and the actual signal is 39.5 dB, the spread of results for (for example 100 microphones) will be between 37.5 and 41.5 dB, depending on which microphone in the line is used for the test, that DUT may result in a false failure. This means that the good unit will need to be scrapped or recycled. Similarly, if a DUT has an actual output at 40.5 dB, the spread of results will be between 42.5 and 38.5 dB. This means that a significant number of test stations would report false passes for that unit. 

Reducing uncertainty

The only way to reduce the number of false passes and failures from production tolerance is to reduce the uncertainty. EQset microphones have a tolerance of ±0.5 dB between individual microphones. This means that from the example given above, there would have been zero false passes or failures.

 Want to know more about "Optimizing production line testing" ?
Download our new application note where we explore what EQset is and includes content on TEDS, calibration, pass-fail windows, and a comparison of testing with and without EQset.