Hi All,

ANSI/NCSL Z540-1-1994 para 10.2b) wrote:
the collective uncertainty of the measurement standards shall not exceed 25% of the acceptable tolerance (e.g. manufacturer's specification) for each characteristic of the measuring and test equipment being calibrated or verified. What does this really mean? What does acceptable tolerance mean?

If my measurement specification is 10 +/- 1 mm, what would be my required uncertainty? Is it 25% of the total tolerance (i.e. 2mm)?

If i have an accuracy requirement of 10% of the total tolerance, will this affect my measurement uncertainty requirement?

Appreciate all your help here. I'm really new to this and a complete lost soul now. Thanks. :confused:

I'm not sure what you mean by

"If i have an accuracy requirement of 10% of the total tolerance.."

The collective uncertainty of the measurement standards must be at least four times more accurate than the tolerance being calibrated. If the tolerance is +/-1 mm, then the collective uncertainty of the measurement standards must be no greater than +/-0.25 mm.

You'll have to explain the quote above. The requirement is in Z540, which prescribes calibration laboratory requirements (to put it simply).

If what you are referring to by the quote above is your measurement requirement, this is completely different than the Z540 requirement. The Z540 requirement is a calibration requirement, not a process measuring requirement.

You would need to determine if your instrument is adequate for your measurement. A common rule of thumb is that the instrument should be 10x more accurate than the limits of the measured parameter. Therefore, if you were making a process measurement (not calibration measurement) with a tolerance of +/-0.1 mm, then the measuring instrument would need to have a specified tolerance of +/-0.01 mm. Subsequently, the collective uncertainty of the standards used to calibrate that instrument must be 4x more accurate than the instrument (+/-0.0025 mm).

Please let me know if I've interpreted your quote correctly. If so, this may mean you need a more accurate instrument.

I recently found this site that helps discuss your post:

http://www.isa.org/InTechTemplate.cfm?Section=Control_Fundamentals1&template=/ContentManagement/ContentDisplay.cfm&ContentID=56927

Thanks Jerry and Bob.

I've another question.

If I have a process measurement with a tolerance of +/-0.1 mm, then the measuring instrument would need to have a specified tolerance of +/-0.01 mm. (this i understand).

If the measuring instrument has a specified tolerance of +/- 0.001mm (which is much better than +/- 0.01mm), is the collective uncertainty of the standards used to calibrate that instrument +/-0.0025mm or +/-0.00025mm? Should it be 4x of the actual tolerance of the measuring instrument or 4x of the expected tolerance?

PS: really appreciate the replies above. Thanks!

Hi,

Is there anyone who can help me with the following questions?

1. I've a process measurement of 10 +/-0.1mm, therefore, the measuring instrument should have a specified tolerance of +/-0.01mm. However, if the measuring instrument have a specified tolerance of +/-0.02mm, can i still use this measuring instrument by tightening the process measurement (i.e. to 10 +/-0.09mm), since the measuring instrument's specified tolerance is +/-0.01mm difference from the required specified tolerance of +/-0.01mm? If no, What should i do?

2. With the above process measurement of 10 +/-0.1mm, and the measuring instrument should have a specified tolerance of +/-0.01mm, and this was met with better results of +/-0.001mm, can i still accept that the uncertainty of the standards used to calibrate the instrument as +/-0.0025mm? Or, the collective uncertainty of the standards used to calibrate that instrument must be 4x more accurate than the instrument, that is +/-0.00025mm?

3. With a process measurement of 10 +/-0.1mm, and the measuring instrument's specifed tolerance is +/-0.01mm, however, the collective uncertainty of the standards used to calibrate the instrument is +/-0.0050mm. Can i still use this measuring instrument by tightening the process measurement (i.e. to 10 +/-0.0975mm), since the difference in the uncertainty values is +/-0.0025mm? If no, what should i do?

Sorry for the lengthy questions. Really hope someone can help me with this. I'm still confused... :confused:

I think I understand your questions. First, to make things a little easier, you might consider dividing this into two separate issues and treat them separately.

ISSUE 1: Adequate instrument for your process measurement

ISSUE 2: Adequate Uncertainty of the standard used to cal your instrument.

They are governed by separate requirements, and may become confusing to some if they become mixed together. If you must mix them together, this should be within the requirements of your quality system (create a documented procedure to specially calibrate for this special purpose).

I will leave ISSUE 1 alone, as it appears you are covered by an adequate instrument.

Regarding ISSUE 2, if you are less than 4:1 (cumulative uncertainty of the cal measurement 4 times more accurate than your tolerance) there are ways around that through documenting as an exception. Another option is to loosen the tolerance of the instrument to what ever you need to and remain 4:1 uncertainty ratio. If this still gives you 10:1 between the instrument and your process measurement, and your quality system permits it, then as long as you properly document the loosened specs, and properly label the instrument so the user knows it is at a looser spec, then that could work.

As for adding the extra uncertainty directly onto the process measurement, that may be confusing, and I don't believe would work.

I would definitely recommend keeping this as two separate issues for purposes of simplification and following of various requirements.

Those are my thoughts.

Hi Jerry,

Thanks for your reply. However, I'm still unclear of certain stuff. Hope you are able to help me with it.

Issue 1: Adequate instrument for your process measurement
- This is done by having the 10:1 ratio between the instrument and the process measurement.

- If I have a process measurement of 10 +/- 0.1mm, i know that my instrument would need to have a specified tolerance of +/- 0.01mm. Where do I find the instrument's tolerance? is it from the calibration record? Usually the calibration record indicates the deviation from nominal or the max instrumental error, is this the instrument's tolerance?

- Also, what should be done if the 10:1 ratio is not met?

Issue 2: Adequate uncertainty of the standard used to calibrate the instrument
- The ratio of 4:1 (cumulative uncertainty of the cal measurement 4 times more accurate than your tolerance), this tolerance refers to the actual tolerance of the instrument or the required tolerance that i derived from the process measurement? I think that the ratio 4:1 should be the actual tolerance of the instrument, however, if my actual tolerance of the instrument is 0, wouldn't the uncertainty be 0 and this would be rather difficult to attain?

- You mentioned that if it is less than 4:1, one option is to loosen the tolerance of the instrument to what ever you need to and remain 4:1 uncertainty ratio. i dont quite understand how this works? Is it possible to quote an example with its process measurement, tolerance, etc?

Thanks so much for your patience in reading yet another lengthy post, and my apologies if I've asked "silly" questions.

WHERE TO FIND THE INSTRUMENTS TOLERANCE. This is normally specified by the manufacturer. I don't know the specific manufacturer/model you are using. But if you refer to its manual, the measurement you are making should be specified in there. Also, if you have had it certified by a proper calibration lab, and had before/after data recorded, it would list the tolerance at the measurands they took. However, using the second method above, you mar/may not correctly ascertain the tolerance at your measurement.

WHAT TO DO IF THE 10:1 RATIO IS NOT MET. The ideal answer would be to use a different instrument that meets the 10:1. If you can not do so, this gets out of the realm of calibration and into determining what your company (or quality standard you must comply with) determines to be acceptable risk.

ADEQUACY OF THE 4:1 UNCERTAINTY RATIO. If you write your own procedure (RISKY - AND NOT RECOMMENDED), you can re-specify tolerances tighter than mfr tolerance. I do not recommend doing this, as without a lot of extra work, you can not prove it can actually meet the tighter tolerance; and you would likely have to calibrate more often.

WHAT TO DO TO LOOSEN TOLERANCE IF 4:1 CAN NOT BE MET. If you can not meet 4:1, you would have to adjust the tolerance on the instrument being calibrated to what ever the 4:1 computes to in comparison with measurement uncertainty of the standard. You would also have to label the instrument stating that it's tolerance is reduced (less accurate) to what ever that new calculated tolerance is. This is NOT recommended, and is only a makeshift method to cover special circumstances. The right thing to do is calibrate at at least 4:1 uncertainty ratio.

AN EXAMPLE OF REDUCED TOLERANCE (Simplified). If you have an instrument mesauring 1 inch +/-0.01 inches and the cumulative uncertainty of the standard used is +/-0.003 inches, you have an uncertainty ratio of 3.333:1. In this example, you would reduce the tolerance of the instrument being calibrated to +/-0.012 inches (which is now 4 times less accurate than the standard ( [0.003] X [4] = [0.012] ). Or for better confidence, you could increase the uncertainty ratio even further.

A POSSIBLE ALTERNATIVE. If you can develop a check standard (some sort of well characterized device to be measured that you prove to be very stable - much more stable than your measuring instrument - you could call that device your source of accuracy. Especially if it is the same dimension as your product. I'm not able to adequately describe everything involved. So only consider that route if you can do all the lengthy homework to properly implement it (and if you can come up with one that will adequately do the job).

Hi,

I've another question to this. What happens if my process specification has a one-sided tolerance? i.e. (10 + 0.1 mm)

Previously, when we have the process specification as (10 +/- 0.1mm), our required accuracy of the measuring equipment is +/- 0.01mm.

For (10 + 0.1 mm), should the accuracy be +0.01mm, or +/- 0.005mm?

Also, one other question not directly relating to the above.
When a designer indicates the specification as (10 +/- 0.1 mm), is there a difference as indicating as (9.9 + 0.2 mm)?

Hope my questions are clear.

Maybe someone will correct me on this, but I would not consider 10 mm +0.1 mm to be a one-sided tolerance. I would normally call a one-sided tolerance something (for example) like <10 mm (only an example). In the case as you mention above, I would call that 10 mm +0.1 mm / -0 mm). So it would be a two-sided tolerance where the lower limit is zero.

Personally, I would use the same rule of 10:1 (if that is your company's rule) for +0.1 /0.0 mm as I would use for a normal two sided tolerance.

You ask whether for 10 +0.1 mm whether the accuracy for the standard should +/-0.01 or +/-0.005 mm. I'm not sure if I understand that.

Regarding the difference between 10 +/-0.1 and 9.9 +0.2 (same as 9.9 +0.2/-0.0). There is in some respects no difference, because the tolerance range is the same. However, in another respect, there IS a difference. That difference is that you try to ideally strive to produce the part at your nominal value. So ideally, in the first case, the best parts would be 10.0; and in the second case, the best parts would be 9.9. So you are theoretically trying to make two different sized parts. Also, since you are trying to make a part to a +0.2/-0 spec, I believe you may discard more parts that way (because if you make some a little too small, they would be discarded (if you understand my meaning).

WHAT TO DO IF THE 10:1 RATIO IS NOT MET. The ideal answer would be to use a different instrument that meets the 10:1. If you can not do so, this gets out of the realm of calibration and into determining what your company (or quality standard you must comply with) determines to be acceptable risk.


If you a control charting, you should be at 10:1 to the control limits. :cool:

Hi, Thanks for the reply. However, I've more questions to ask to clarify.

How do you apply the same rule of 10:1 for +0.1 /0.0 mm as you would use for a normal two sided tolerance?

Will you change the specification from (10 + 0.1/ 0.0) mm to (9.95 +/- 0.05) mm and apply the 10:1 rule? That is the accuracy required will be +/- 0.005 mm.

Or from (10 + 0.1/ 0.0) mm, the accuracy required will be just + 0.01 mm. There is no negative part of it?

Tolerance is really the zone between the specifications. So 10:1 applies to the zone without respect of the target. The idea is to get resolution of some value. So, the spec of +0.1 /0.0 is a tolerance of 0.1, applying 10:1 is .01 resolution required, or +/- .005. :cool: