Contributing Factors in a Supermicrometer Measurement Uncertainty Budget?

[Mchurch]

Hi everybody, Great forums! Just joined a bit ago when doing some searching on google.

My name is Mark, I'm a senior cal tech for a 3rd party company. I had formal metrology training in the US Navy and through University while overseas in England. I am having issues with my Supermicrometer budget being that I am primarily an electronic guru and stepped into the Phys world about a year ago.

First issue, My BMC will be my standard uncertainty used. It is being made wide enough so that it is achievable under virtually all conditions the dimen lab happens to be in.

My uncertainty budget so far consists of ... Well that would take a bit, See attached, and make recommendations for improvement. I feel pretty solid on my numbers and calculations, Just feel as if I am missing a source of uncertainty that I haven't even thought of.

[jfgunn]

Couple of thoughts:

1.) You use the acuracy of gage blocks in your calculation. This is correct. You should also have a factor equal to the calibration of the gage blocks. When you had your blocks calibrated, you were given an uncertainty at k=2. Take this number and use a divisor of 2.

2.) It would appear that the bench micromter you are to test has a resolution of 20 millionths. I would use 1/2 this value for the contributing factor, therefore 10 millionths.

3.) I would reccomend calculating the uncertinaty at 1, 5, and 10 inches. you could then have a statement of the uncertainty that has the form something like "(16 + 3L)uin where L is the length in inches".

4.) Is your bench micrometer the only one you will be testing? If so, you temperature stability numbers look OK. If not, you should allow for a larger swing of temperature for on-site work where the lab is not within 1C.

Hope this helps some.

[Mchurch]

Originally Posted by jfgunn

Couple of thoughts:

1.) You use the acuracy of gage blocks in your calculation. This is correct. You should also have a factor equal to the calibration of the gage blocks. When you had your blocks calibrated, you were given an uncertainty at k=2. Take this number and use a divisor of 2.

2.) It would appear that the bench micromter you are to test has a resolution of 20 millionths. I would use 1/2 this value for the contributing factor, therefore 10 millionths.

3.) I would reccomend calculating the uncertinaty at 1, 5, and 10 inches. you could then have a statement of the uncertainty that has the form something like "(16 + 3L)uin where L is the length in inches".

4.) Is your bench micrometer the only one you will be testing? If so, you temperature stability numbers look OK. If not, you should allow for a larger swing of temperature for on-site work where the lab is not within 1C.

Hope this helps some.

Thanks for the feedback

1.) I didn't use the accuracy of the gage blocks, I will be using actual data values given to me at thier last calibration, The uncertainty of measurement reported on the gage blocks was (4.2 + 1.9L) so the SU there was 23.2 uIn on my 10 inch block (worst case unc)

2. and 3.) Thanks, I will be making the recommended changes.

4.) Most likely scenario is it stay in the lab. If my customer has the demand and the money, My hands would be tied by my management forcing us into the field with it. That being said, Temperature correction values? This is what I've found so far:
-ASME/ANSI Y14.5M 1.4(k) states all temperature correction factors should be referenced to 20 C.
-ASME B89.1.9 states the CTE of gage blocks should be between 10.5 and 12.5. (No uncertainty associated with these values)
AISI 1050 steel has a stated of 11.5 between 15 and 25 degrees C.
AISI 52100 steel is stated at 10.5 between 19 and 21.

With all of that, What number do I assign to my blocks for expansion values and is and auditor/assessor going to nitpick the value I assign. I see some labs using 9.5 and some as high as 13.5.

I have mitutoyo grade 2 blocks, I contacted mitutoyo and they claimed all thier blocks manufactorerd during the period of time mine where conformed to AISI 1050 Steel within 90 percent of thier material. Do I take the 11.5 stated and add 10 ucertainty to the value?

Also using the Veriteq Spectrum as a lab temperature monitor, It has a stated accuracy of + or - 1 degree, Should I include this in my budget as a possible 2 degree swing in lab temp due to the unknown of my tempurate monitor? Or will the auditor not even bother with that much information?

Thanks again Jfgunn, And if anybody else can provide insight with the temp correction factors, Please educate me.

[jfgunn]

Your questions about CTE's show that you have thought this through quite well.

There are two items to include in the budget for temperature:

1.) Difference in CTE between master and UUT. In theory, both items would be steel and they would have identical CTE's. Even if your lab was at 30C, it would not matter because both the UUT and master would grow the same and the measurments would be the same. But, since CTE's vary by as much as 10%, the master and the UUT could have CTE's that are 20% apart (1 could have CTE 10% low and 1 could be 10% high). Pick a nominal CTE value and compute the difference if one were 10% low and 1 were 10% high. This CTE difference would be multiplied by each of the following: (1-temperature control in lab of 1 degree, 2-accuracy of thermometer, 3-uncertainty from calibration of thermometer). This would be a rectangular distribution.

2.) Part vs. Master temperature difference. Even if the part and the master are next to one another in the lab, they are not at the same temperature. If the lab can vary by up to 1 degree, the standard could be 2 degrees different than the UUT. This 2 degree temperature difference would be multiplied by the assumed CTE. This would be a rectangular distribution if you have a simple on/off air conditioning system. If you have a system that controlls temperature better (ie an air conditioning system that constantly pumps in a/c and pulses in heat to control temperature) then a triangular distribution might be more appropriate.

I use 6.4uin/in/F for my assumed CTE of Steel.

A2LA has a great document on their webiste that helps explain this better than I have done above. It is their Guide to Dimensional Uncertainty or something like that.

[Mchurch]

Resolution Uncertainty,

Anybody's thoughts are appreciated but am replying to JFGunn stating:

"2.) It would appear that the bench micromter you are to test has a resolution of 20 millionths. I would use 1/2 this value for the contributing factor, therefore 10 millionths. "

I'm reading a document titled "A Comprehensive Comparison of Uncertainty Analysis Tools" from the 2004 Measurement Science Conference, Author Suzanne Castrup.

Under 'Resolution uncertainty' it states;

"For analog displays, the resolution error is assumed to have a normal distribution. The resolution uncertainty is computed by setting the containment limits equal to a smallest increment of resolution and applying a containment probability that readings can be discerned within these limites."

It contiunes to state use ± 1/2 smallest resolution for digital readouts.

I'm using an analog display with 20uIn being my minor divisions resolution. Should I use 20uIn or 10uIn for my resolution uncertainty?

And with that resolution uncertainty, I would assume a divisor of 1.732. Am I wrong?

I have seen similar model units, use full resolution (20uIn)with a divisor of 3. Are they simply squaring the divisor to make up for the doubling of resoultion uncertainty?

Any thoughts are appreciated!

[jfgunn]

You ask good questions and I am not sure I know a definite answer.

I saw something once that talked about digital resolutions and the fact you should sometimes not use 1/2 because some electronics round at half and some round at 90%........

I suppose I might do the following in your case:

1.) Compute the total uncertainty either way.
2.) Choose the solution that gives you the larger unvertainty.
3.) Run with it.

The fact is that if your average customer will not care about an uncertainty of 55uin or 60uin (or wahtever the numbers are).

When you make notes in your budget record the fact that you might be able to use the other method. Make a note that you were unsure (or uncertain) of which way to go so you chose the worst case.

In my opion, uncertainty budgets do not have to be perfect, they have to be defendable.

[Mchurch]

Ok, Final version is attached. Came up with 17.1 uIn + 8 uIn/In. Am I in the ballpark for what other labs are claiming?

I would appreciate feedback positive or negative. I am relatively new at this and will be working on many more in the PD lab in the future.

Thanks for all the help JF. OCScal has a great scope in the pd lab, You do the work for all of that?

Mark.

[jfgunn]

Final version looks good to me.

Unfortunately, yes. I have done all of the work for our entire scope. It took a fair amount of time and changes a little each year.

Ever time we get audited, I just wait for an auditor to pick something to correct/revise. It keeps me on my toes!

I am glad I was able to help.