|
A common problem in semiconductor manufacturing is the inability to accurately quantify the repeatability and reproducibility of gauges (GR&R) in the conduct of destructive tests. Typically, attempts are made to perform GR&R studies using calibration samples. However, this approach does not model the actual performance of the gauge in its true manufacturing environment. This paper demonstrated how to design and execute a GR&R study for destructive tests, using the post-wirebond ball shear test as vehicle. What the author did in this study is to have a homogeneous population of samples built, since many identical samples are needed to generate data for this destructive testing GR&R study. All known sources of variation among these samples must be blocked, so that variations in the destructive test data will primarily come from the testing itself, allowing a good estimate of the GR&R. To ensure homogeneity of the samples, all of them came from the same wafer and basically used the same raw materials, as well as underwent exactly the same assembly process. This means that all samples must go through a single equipment for each assembly step, with all assembly process parameters held constant as much as possible. Lastly, a single, calibrated/qualified ball shear tester was used in the study, with the ball shear testing performed by two certified operators in a randomized order without interruption. The above approach resulted in a Ball Shear Test GR&R value of less than 30%. This variation is below the minimum acceptable level of metrology system error, thus giving the author's manufacturing group the confidence to utilize this metrology system as a process optimization and control tool.
|
