I'm performing an estimate of uncertainty for accelerometers in the coming months. The last time I did this was for a different company and my uncertainties were based on the tolerances of the standard accelerometer and the controller.
My uncertainties came out substantially higher than most of the published uncertainties I saw from other labs.
I'm thinking the other outfits were using sampling to get their uncertainties and that's how they got their numbers down.
Has anyone here used sampling instead of instrument tolerance to perform their estimates of uncertainty?