Guardbanding - How would one go about guardbanding a specification?

[Charles Wathen]

Hi everyone,
How would one go about guardbanding a specification? If my TUR's are less than 4:1, it's recommend that I use a guardband of 20% for all 1.5 >TUR<4. Can someone give me an example say of a micrometer or caliper? Any web links?

 

[Jim Wynne]

"Guardbanding" is the practice of adjusting specification limits (pass/fail criteria) to account for uncertainty in the measurement system. I don't think there are examples for specific devices such as calipers and micrometers; the object is to reduce the chance of accepting nonconforming material because of measurement error, regardless of the device(s) being used. It usually boils down to economics. If I alter the specification to guard against accepting bad product, how does the change affect the chance that I'll reject good product, and how do I balance the two? In other words, is it better to reject parts at the limits of the specifications than to ship nonconforming parts? It's a question that must be answered on a case-by-case basis.

 

[Govind]

Charles,
Take a look at the thread.
(broken link removed)
Regards,
Govind.

 

[Charles Wathen]

Thanks for the info. The Fluke paper in that link should do the trick. In my Goggle searches, I found hardly anything on the subject. The reason I picked a micrometer or caliper for an example to make it easy to explain and understand.

 

[Charles Wathen]

OK just to make sure I understand what is going on with guardbanding. I'll use the example in the Fluke PDF.

Subject Unit = 0.1
Standard = 0.04
TUR = 2.5:1

If I use method #2:
my answer is 0.6. This would mean a subject unit with a tolerance of ±0.06 (0.6*0.1)

If I use method #3:
my answer is 0.85. This would mean a subject unit with a tolerance of ±0.085 (0.85*0.1)

If I use method #4:
my answer is 0.916515. This would mean a subject unit with a tolerance of ±0.91651514 (0.916515*0.1)

It looks like method 3 or 4 is probably the best, but are those calculations correct?

 

[Hershal]

The Fluke reference is good, except it has one flaw.....the writer refers to a TUR for the 10:1 and 4:1 rules.

Under ANSI/NCSL Z540-1-1994 and the previous MIL-STD-45662A, that was acctually a TAR, becuase measurement uncertainty (MU) was virtually unknown for practical application outside of a NMI at the time the standards were written.

What does that mean? Use the TAR to find out how accurate your standard is compared to the unit being calibrated. Then the MU must still be calculated, regardless of the accuracy ratio.

Guardbanding cannot be successful without knowing the MU, and remember that while the accuracy ration feeds into MU, it is only one influence.

Hershal

 

[Charles Wathen]

Thanks Hershal for that clarification.