MSA results on % Tolerance charts

Stuart Andrews

Involved - Posts
We conduct our MSA studies and compute the results on a % Tolerance chart that looks very similar to the % Total Variation charts.

However, I have now fully read and understood the MSA manual, second edition, and it doesn't mention a calcualtion based on tolerance.

Does anybody know what grounds I have for using this calculation because I do get more acceptable gauge results using this calculation.:confused:

Al Dyer

Should a gage be measured against the tolerance when the tolerance has no bearing on the gage?


Ryan Wilde

Here is where my ignorance becomes readily apparent...

I have no idea what an MSA study entails, except what I have gleaned here. I don't actually do any of the QS9000 stuff, I have strictly worked in the calibration arena for some 17 odd years.


From what I have learned (possibly mistakenly) an MSA is the study of a measurement system including short-term variation, drift, etc. In this case, the tolerance of the gauge has absolutely no bearing whatsoever. What does have bearing is the absolute error (bias ± the uncertainty of the gauge), its repeatability (short-term), its drift (those yearly calibrations, with data, on a spreadsheet gives a great baseline), the effect of your specific environment extremes on the gauge (I've been in shops with 39°C seasonal temperature swings and they have no idea why they have problems measuring aluminum), and variation between the petite guy that is oh-so-careful and Gunther-who-uses-a-micrometer-like-a-c-clamp.

General rule of thumb in metrology - measurement is measurement, and tolerance is analysis after the measurement process. Tolerance is not a gauge attribute, it is a (sometimes arbitrary) number a design engineer figured his/her gauge should be able to meet (and possibly even meet over time) when compared to the oh-so-elusive absolute "nominal".

So, on a good note, if your gauge variation as a percent looks just like your total process variation, you have a mighty fine process.

In other words - I use tolerance for 3 things. First, it gives me boundaries for computing the recall cycle of a gauge. Second, it allows me to take a gauge's reading as gospel if the tolerance is sufficiently small compared to what I am testing. Third, it tells me when to adjust/repair/replace a gauge, which means my company makes more money and can give me raises. ;)

I don't know Stuart's application, so it is hard to say if % tolerance gives any indication whatsoever. But, if total allowable deviation of a gauge is specified in the quality system as the same as the tolerance, it would be a very good indication of gauge condition and drift.

I hope I wasn't entirely off-base, as I've never needed to look at an MSA manual. I just have to do those uncertainty studies.


Atul Khandekar

My interpretation

MSA studies are conducted primarily to determine if a given instrument is suitable for measuring a certain part. Afterall measurements taken with this instrument will be used for SPC analysis. So MSA studies determine the suitability of an instrument for monitoring and controlling a process. You check the bias, discrimination, stability over time and R&R etc.

The first thing to do is to analyse Appraiser and Equipment variation and correct the problems if any in measurement method, fixturing etc.

There are occasions when your %R&R ( R&R error as a % of Study variation) cannot be brought below 10%. This can also happen if your Process Capability is high. In such cases, one alternative is to scrap the instrument and find a better (and probably more expensive) one. However, you can compare the measurement error with the total tolerance on the part (IMO, not the gage) you are measuring. (Error as % of tolerance) And if this % is low, the instrument may be 'Accepted Conditionally'. An instrument can be conditionally accepted if R&R is between 10% and 30%.

Similarly, you can also compare the bias with tolerance to decide if bias is significant. (This is probably not statistically correct method, may be you should do a control chart with bias study measurements)


Ryan Wilde

Thanks, Atul, that is the kind of straightforward explanation I've needed. Really, it doesn't sound too bad to perform and calculate.

So, from what I can gather, I should revise my statements. Since the MSA seems to be specific to a product, then the tolerance of the product not only needs to be taken into account, but has to be the key factor in decision making as to whether the MSA study has proven or disproven that a measurement system is adequate for a particular function. Therefore, I would have to say that % tolerance of the product should be the last step in the MSA study.

I went at it entirely from the gauge tolerance viewpoint, which was obviously the wrong stance.


Top Bottom