I'm developing a measurement system for optical sensors. The alignment error of the sensor in the test fixture causes a component of gage error that is proportional to the true value of the sensor response when alignment is perfect. For example, if the true value of one sensor response is 100 counts and alignment error adds another +/-10 counts, then another sensor of true value 200 counts will have an alignment error of +/-20 counts. This kind of measurement error doesn't typically happen in measurment of mechanical components. All of the discussions of gage R&R that I have seen assumes that gage variability is a constant whether you are measuring small parts or large parts. For my application, the largest contributor to gage variability is a function of the underlying values I am trying to measure. I can influence my gage R&R results by selecting only weak or only strong response sensors. Is there a way to characterize the variablity in my gage that that avoids this problem?
Thanks,
Mark

It sounds like you do have a constant based on your example. Its 10%. I say that based on GRR studies we did on our optical CMM's. When we varied measurement parameters such as point quantities, focal thresholds, light contrasts and intensities, etc., when testing on a calibration standard, there was a definite correlation to the error generated and the variation of the measurement parameters and it was accurately expressable as percentage. I learned long ago to avoid comparing mechanical to optical measurement analysis, obviously because optics introduce so many variables that do not exist mechanically.
A GRR was "mechanically" consistent though when optical measurements were taken manually by operators, but the automated measurements resulted in a reliable percentage product.
I dont know if what I just explained applies to your sensors, but if it helps, then great.

Steve Truchon