The Best Measurement Capability (BMC) of a Small Calibration Laboratory

[sermet]

Hi everybody! This is my first post on the Cove. The Cove is great and it has been of much help in my work.
I am trying to assess the best measurement capability (BMC) of a small calibration laboratory in the dimensional area (calipers, micrometers, dial indicators). I would appreciate your comments on the following questions.

1. What is the exact meaning of the qualifier of “nearly ideal instrument” which is given in the definition of BMC? As in the document EA-4/02 “Expression of the Uncertainty of Measurement in Calibration” , “… inherent to the concept of being nearly ideal is thus that there should be no significant contribution to the uncertainty of measurement attributable to physical effects that can be ascribed to imperfections of the device to be calibrated. However it should be understood that such a device should be available…”
This is quite confusing to me.
Let’s consider for instance the uncertainty budget of a nearly ideal caliper. Shall I include contributions such as parallelism of the jaws? Or should I consider that contribution to be zero? Or should I estimate it based on the assumption that the nearly ideal caliper is within the tolerance values reported in a relevant standard? or should I rely on available information given by mayor producers? Or should I set up a method for determining the parallelism in the lab. and estimate the minimum value of the uncertainty of that method? (we don´t measure parallelism in our routine calibration of calipers)
We do measure the parallelism of the anvils in the case of a micrometer using an optical parallel. What is the parallelism contribution for the uncertainty budget of a nearly ideal micrometer? Should it be our minimum uncertainty value (say 1 red interference band) or the tolerance for parallelism reported by the standard (say 3 red bands)?

2. How should I estimate the type A uncertainty contribution for the BMC assessment? We do not have at hand any type A uncertainty studies on nearly ideal instruments or brand new calipers, micrometers, indicators, etc. Shall I look for repeatability values reported by the mayor producers? Does anyone know where could I find such data?
Thank you for your kind support.

 

[Hershal]

Welcome to the Cove!

The nearly ideal means essentially their measurement standards, and the best example of the type instrument they are calibrating.....understand this is not real world necessarily.....sometimes it is.....

For Type A, take your readings (10 preferred), get the standard deviation, divide that by square root of n (where: n equals the number of readings), and you have Type A.

Next, get the Type B.....

For gage blocks (or whatever standard) take the specified MU from the calibration certificate, divide by 2 to return to standard uncertainty.

Ringing is when two blocks are joined to create a value, that is typically assigned 0.00005 inch, and is rectangular distribution, so divide by square rrot of three (1.732).

Difference in temperature between the calibration of the measurement standard (typically at 20 C) and the demonstration temp is also rectangular.

Co-efficient of expansion for the calipers (in the example) and the gage blocks must both be taken into account. Steel gage blocks expand/contract at 6.75 microinches per inch per degree C, and is rectangular.

Resolution error is half the least digit, so for calipers with a resolution of 0.001 inch, the resolution error is 0.0005 inch, and is rectangular.

Paralax (for analog instruments) must be considered and is rectangular. Use the same value as resolution error.

Once you have all this, take the values, square them, add them, take the square root of the total, you now have standard uncertainty. Multiply times two for expanded uncertainty.

See how easy this is?

Hershal

 

[sermet]

Thanks Hershal for your advice.
What you suggest is to take a new instrument, say one of the best calipers in the market in my country, and perform a routine calibration on it; calculate the expanded uncertainty and this would be the laboratory best measurement capability for a caliper of that range and resolution. It looks fine.

You didn’t mention uncertainty contributions such as the parallelism and Abbe error in your budget. Actually they are important contributions in my budget for the caliper. For a 300 mm vernier type caliper with a resolution of 0.05mm contributions from parallelism and Abbe error are 0.01 and 0.02 mm respectively. I also consider a contribution from the zero setting, 0.013mm, a half the resolution error.
Am I overestimating my budget?

I have looked over a number of sources about uncertainty budgets, but not a single example is the same. I have found differences in the number of contributions to be considered, and even their size.
As a closing thought: nobody wants to overestimate the BMC values being too much conservative (but we are required to consider all the contributions), since, as I understand, BMC defines the laboratory quality. Don’t you think that there should be more specific line guides on this topic from the competent authorities in order to harmonize criteria?

Welcome to the Cove!

The nearly ideal means essentially their measurement standards, and the best example of the type instrument they are calibrating.....understand this is not real world necessarily.....sometimes it is.....

For Type A, take your readings (10 preferred), get the standard deviation, divide that by square root of n (where: n equals the number of readings), and you have Type A.

Next, get the Type B.....

For gage blocks (or whatever standard) take the specified MU from the calibration certificate, divide by 2 to return to standard uncertainty.

Ringing is when two blocks are joined to create a value, that is typically assigned 0.00005 inch, and is rectangular distribution, so divide by square rrot of three (1.732).

Difference in temperature between the calibration of the measurement standard (typically at 20 C) and the demonstration temp is also rectangular.

Co-efficient of expansion for the calipers (in the example) and the gage blocks must both be taken into account. Steel gage blocks expand/contract at 6.75 microinches per inch per degree C, and is rectangular.

Resolution error is half the least digit, so for calipers with a resolution of 0.001 inch, the resolution error is 0.0005 inch, and is rectangular.

Paralax (for analog instruments) must be considered and is rectangular. Use the same value as resolution error.

Once you have all this, take the values, square them, add them, take the square root of the total, you now have standard uncertainty. Multiply times two for expanded uncertainty.

See how easy this is?

Hershal

 

[crendfrey]

Welcome to the Cove!

See how easy this is?

Hershal

It is easy when explaned by you sir but then you can do this magic
I am still working on the profeciency testing results (scales).
Do you give lessons?

"Don’t you think that there should be more specific line guides on this topic from the competent authorities in order to harmonize criteria?"


Yes, I do. Everyone seems to be on their own. However, one size fits all just does not work with MU.

 

[Ruebenn]

Mr.Hershal,

I was browsing through some scope of accreditation and i hope that you can shed some light on the scope mentioned.
I was going through A2LA accreditated Agilent Local Calibration center in Florida.
There i see parameter range Sine wave using the HP3335A.
I see it is under Generate but when i see the same A2LA accreditated Simco Electronics.
I see the same 3335A under MEASURING instead of Generate and i find it very difficult to digest this as i know the 3335A is a generating unit and not a measuring unit.
And i see the BMC readings differ as well.
And i find it difficult to comprehend how Agilent BMC can be bigger that Simco Electronics taking into the consideration that they are the OEM for the 3335A unit as well.
Plus can you please walk me through the values stated in the BMC?
How can we arrive to such a value?

 

[jfgunn]

I have looked over a number of sources about uncertainty budgets, but not a single example is the same.

I think this quote really says everything you need to know about uncertainty. It shows that two very intelligent people can perform two very good calcualtions and end up with different, yet aceptable, answers.

My reccomendation for uncertainty would be to pick the factors that you think are reasonable. Document what you have chosen to include (or exclude) and why you have made these choices. Then move on with the actual business of calibration. If you want to add a few items into the budget that others don't typically add, go ahead and do it.

Hershal's example has many things in it that are contained in my budget, with a few exceptions:

1.) He says to take the uncertainty from the calibration cert for the gage blocks. This is true. I think you should also use the accuracy of the gage blocks (as a rectangluar distribution) unless you are using the exact reported value of each block while using them.

2.) Ringinging of blocks: Hershal shows 0.00005" . I think he might have missed a zero and he meant 0.000005". This would be 5 millionths of an inch instead of 50 millionths. I do not include this in my uncertainty budget (maybe I should).

3.) I might throw in somehting for alignment error. Of course if I throw this in aren't I possibly doubling it because the caliper I use for my Type A has some alignment error????

My though on uncertainty is that I have good reasons for using what I have used. If Hershal were to come in to audit me, and he said "your uncertainty should have ringing in it", I would gladly add it. It would not matter to the final answer anyway. The next auditor through might tell me to take the ringing out and add something else. Each auditor has things that they might pick on in terms of uncertainty. As long as you have documented your choices, you should be fine.

Also, keep in mind that when an item has a relatively corse resolution, the BMC is 58-60% of that resolution. All of the other factors become insignificant compared to the resolution.

 

[ggdjr]

For Type A, take your readings (10 preferred), get the standard deviation, divide that by square root of n (where: n equals the number of readings), and you have Type A.

Next, get the Type B.....

For gage blocks (or whatever standard) take the specified MU from the calibration certificate, divide by 2 to return to standard uncertainty.

Ringing is when two blocks are joined to create a value, that is typically assigned 0.00005 inch, and is rectangular distribution, so divide by square rrot of three (1.732).

Difference in temperature between the calibration of the measurement standard (typically at 20 C) and the demonstration temp is also rectangular.

Co-efficient of expansion for the calipers (in the example) and the gage blocks must both be taken into account. Steel gage blocks expand/contract at 6.75 microinches per inch per degree C, and is rectangular.

Resolution error is half the least digit, so for calipers with a resolution of 0.001 inch, the resolution error is 0.0005 inch, and is rectangular.

Paralax (for analog instruments) must be considered and is rectangular. Use the same value as resolution error.

Once you have all this, take the values, square them, add them, take the square root of the total, you now have standard uncertainty. Multiply times two for expanded uncertainty.

See how easy this is?

Hershal

that is very well done

can you do the same for a device that take a t/c input,i need a little hand holding for now

thanks,

GD

 

[Hershal]

Thanks Hershal for your advice.
What you suggest is to take a new instrument, say one of the best calipers in the market in my country, and perform a routine calibration on it; calculate the expanded uncertainty and this would be the laboratory best measurement capability for a caliper of that range and resolution. It looks fine.

You didn’t mention uncertainty contributions such as the parallelism and Abbe error in your budget. Actually they are important contributions in my budget for the caliper. For a 300 mm vernier type caliper with a resolution of 0.05mm contributions from parallelism and Abbe error are 0.01 and 0.02 mm respectively. I also consider a contribution from the zero setting, 0.013mm, a half the resolution error.
Am I overestimating my budget?

I have looked over a number of sources about uncertainty budgets, but not a single example is the same. I have found differences in the number of contributions to be considered, and even their size.
As a closing thought: nobody wants to overestimate the BMC values being too much conservative (but we are required to consider all the contributions), since, as I understand, BMC defines the laboratory quality. Don’t you think that there should be more specific line guides on this topic from the competent authorities in order to harmonize criteria?

You are correct, those are in fact important issues in caliper BMC; however, I was leaning towards the most common influences.....of course, we could include guardbanding and Monte Carlo also, but for most calibrations these are a bit beyond the realm of normalcy.....still, the numbers you present - in my estimation - should be evaluated generally as rectangular distribution and divided by square root three.....except the Abbe influence may actually be triangular distribution and divided by square root six.....a bit more investigation needed perhaps.....

With respect to BMCs.....just my opinion, this is just me, only Hershal, no one else, this is not an official position by any known AB.....if I have enough disclaimers in place.....BMC is an almost purely fantasy number.....don't rely on it.....

 

[Hershal]

It is easy when explaned by you sir but then you can do this magic
I am still working on the profeciency testing results (scales).
Do you give lessons?
QUOTE]

We have two uncertainty courses coming up, I will be present - though not usually presenting - at both.....still, that means you can - and should - pepper me and the instructor with REAL WORLD questions.....

Please see our website.....the flyers for the training are there.....

https://www.iasonline.org
(broken link removed)
(broken link removed)

If you have checked uncertainty course prices and look at ours afterwards.....and wonder at the difference.....we are a non-profit organization, and as such PREFER to keep our costs only a little above what we need to break even, so we can continue to offer training.....

Oh, we are also offering CCT prep, reply if that interest you......

Hope this helps.

 

[Hershal]

It appears that my reply in quotes was not as effective as it should be....

I apologize and ask that everyone read as intended, if the notations provide sufficient reference for that.....