Out of Tolerance Condition - Definition of *Significant* Out of Tolerance Condition

My company follows MIL-STD-45662. This standard requires that we provide our own definition of "significant out of tolerance condition" as applied to gage calibration. If a gage fails calibration and meets this definition, action is to be taken to correct possible nonconforming product. What is a good definition of significant out of tolerance condition? How does industry perceive this definition and what factors should be considered? Our current definition is to take the calibration tolerance of the gage and add a certain percentage. I would like some information that could be used as a "benchmark".

[Jerry Eldred]

Significant Out Of Tolerance, is the amount out of tolerance that has potential to impact product quality. Because it has always been so vague, I have done away with significant out of tolerance. "Significant" truly can't be defined by the calibration lab, but must be based on the user's application. There is the presumption that the user has the correct tool for the measurement it is used for. With that presumption, it can be further presumed that he needs the applicable tolerance. So with those presumptions, significant out of tolerance is any out of tolerance. With the use of Gage R&R/MSA's, it has become better defined what measurement tool accuracy/stability/repeatability/reproducibility is needed to do the job. Using those concepts, and perhaps some control charts, and by reporting all out of tolerance conditions to the user, there is these days a better concept of when a measurement tool has drifted far enough to impact the process measurement (out of control).

I believe back in the days when MIL-STD-45662A was implemented, in the defense contractor environment (which it was designed for), philosophies functioned differently. I am not sure how many people out there are still using a separate limit for significant out-of-tolerance. By all rights, if you use a separate limit, you have to define it based on test equipment applications, track it, define it, and defend it. By sticking simply with the instrument tolerances, it makes the reporting mechanism to the end user much simpler.

The user then makes the determination based on the particulars what if any corrective action is needed. The same out of tolerance condition on the same model test equipment could have two different implications in two different applications.

A bandwidth issue on a 400 MHz oscilloscope, let's say a Tek 2465B (good equipment, just an example). If measured bandwidth at the -3dB point moved from say 420 to 320 MHz over a cal cycle, and the scope is used at two different test stations. At one station, they measure 1 KHz sine wave amplitudes, and at the other station, they measure risetimes of let us say 0.50 nanoseconds.

That out of tolerance condition would not be significant to the test station that measures amplitudes of 1 KHz sine waves. But it would be very important to the other test station where they measure risetimes (which would have been effected by this OOT).

My two cents is that the wise thing is to either call all out of tolerance conditions significant (which is, I believe a pretty common practice these days as the more recent standards don't give the latitude for defining a significant out of tolerance condition), or a detailed review of the applications of the test equipment (very cumbersome), and set significant out of tolerance limits based on that information.

I'll also be interested to hear some of the other viewpoints and opinions on this. I haven't heard a good discussion of this topic in a number of years.

------------------

Hard gage tolerancing always goes toward the center of the prescribed part tolerance. A plug gage has a plus go tolerance and a negative nogo tolerance. An in-tolerance go gage can reject good parts if you're riding the limit but will not accept bad parts. The go gage is the one subject to wear. Should it wear to the point of affecting assembly, you have a significant condition. That's up to you and the customer to determine.
Mike