Local calibration labs

[Cerebus]

It would be impossible to know the uncertainty of a measurement before the measurement is made, so stating that one would receive a specific uncertainty doesn't seem very forthright.

Eh? If you don't know the uncertainties inherent in a making a particular type of measurement with a particular method and particular instruments before making the measurement how will you suddenly know them after making the measurement. I think you're conflating absolute error and uncertainty.

You can have an incomplete picture of the uncertainty before a measurement, but there are components of the total uncertainty that are unknown until the measurement occurs. Repeatability is probably the most common. Sometimes this component completely swamps the other, referenced uncertainty contributors.

That just doesn't make any sense. It, frankly, just sounds like handwaving. What are these "components of the total uncertainty that are unknown until the measurement occurs"? How are they different from any other components of the total uncertainty that can be determined either from prior knowledge of systematic errors or prior statistical observations of the measurement process?

[Moon Winx]

That just doesn't make any sense. It, frankly, just sounds like handwaving. What are these "components of the total uncertainty that are unknown until the measurement occurs"? How are they different from any other components of the total uncertainty that can be determined either from prior knowledge of systematic errors or prior statistical observations of the measurement process?

For a quick example, say you have a magical measurement system that that makes perfect resistance measurements with absolutely zero uncertainty. You bring in a bunch of artifacts to measure, different models of resistors, and even multiples of the same model. Now, due to variations in the manufacturing, you would have different levels of defectiveness (inadequate shielding, oxidation of components, overloaded internal resistors, etc.) in the groups of resistors that produce errors during the measurement. These show up as a stability component, short term and long term drift and offsets from a previously known history of that item. All of these things affect the measurement uncertainty, so even using a perfect measurement system you would have components of the measurement uncertainty that come from the individual test units. We have a couple of systems in our lab where 99% of measurement uncertainty comes from test unit noise. The Fluke 792A is an example. Depending on when it was manufactured, we end up with significantly different measurement uncertainties even though the they all share the same model.

[CalMachine]

It would be impossible to know the uncertainty of a measurement before the measurement is made, so stating that one would receive a specific uncertainty doesn't seem very forthright.

Eh? If you don't know the uncertainties inherent in a making a particular type of measurement with a particular method and particular instruments before making the measurement how will you suddenly know them after making the measurement. I think you're conflating absolute error and uncertainty.

You can have an incomplete picture of the uncertainty before a measurement, but there are components of the total uncertainty that are unknown until the measurement occurs. Repeatability is probably the most common. Sometimes this component completely swamps the other, referenced uncertainty contributors.

That just doesn't make any sense. It, frankly, just sounds like handwaving. What are these "components of the total uncertainty that are unknown until the measurement occurs"? How are they different from any other components of the total uncertainty that can be determined either from prior knowledge of systematic errors or prior statistical observations of the measurement process?

You're failing to take into account the repeatability of the DUT. 

[Cerebus]

That just doesn't make any sense. It, frankly, just sounds like handwaving. What are these "components of the total uncertainty that are unknown until the measurement occurs"? How are they different from any other components of the total uncertainty that can be determined either from prior knowledge of systematic errors or prior statistical observations of the measurement process?

For a quick example, say you have a magical measurement system that that makes perfect resistance measurements with absolutely zero uncertainty. You bring in a bunch of artifacts to measure, different models of resistors, and even multiples of the same model. Now, due to variations in the manufacturing, you would have different levels of defectiveness (inadequate shielding, oxidation of components, overloaded internal resistors, etc.) in the groups of resistors that produce errors during the measurement. These show up as a stability component, short term and long term drift and offsets from a previously known history of that item. All of these things affect the measurement uncertainty, so even using a perfect measurement system you would have components of the measurement uncertainty that come from the individual test units. We have a couple of systems in our lab where 99% of measurement uncertainty comes from test unit noise. The Fluke 792A is an example. Depending on when it was manufactured, we end up with significantly different measurement uncertainties even though the they all share the same model.

OK, if I get you right your "there are components of the total uncertainty that are unknown until the measurement occurs" are all attributable to the DUT. Fine, we all understand that, and we'd have got there a lot faster if you'd said something like "those figures don't include the inherent uncertainties of the DUT itself" instead of going around the houses.

Your implication in

"It would be impossible to know the uncertainty of a measurement before the measurement is made, so stating that one would receive a specific uncertainty doesn't seem very forthright."

is that there's something misleading in the accreditation of cal labs to "measure X to y%" or "source x to y%". There isn't. This doesn't amount to a claim to something like "after supplying 10V to your DVM it will be calibrated to 0.3ppm uncertainty" it's "we will supply 10V to an uncertainty of 0.3ppm at the end of the leads we plug into your DVM".

If you still think this is not "very forthright", what quantification of a cal lab's capabilities would you deem reasonable to make? How is a national accreditation scheme going to account for every random bit of test gear that an end-user's might turn up with at the cal lab being accredited?

[Cerebus]

You're failing to take into account the repeatability of the DUT.

That's ***** easy for you to say after we've been around the houses. 

It would have been helpful if Moon Winx had been just a tad clearer in what he was saying. Like perhaps saying something like "Ah, but that doesn't account for DUT variability that the cal lab can't know in advance". It's a little difficult to second guess that when it's cal lab's accredited capabilities that's the subject in hand.

We've got some very clever, knowledgable, and experienced people on here, but sometimes I do wish that perhaps some of them had taken an extra writing  class or two.

[CalMachine]

You're failing to take into account the repeatability of the DUT.

That's ***** easy for you to say after we've been around the houses. 

It would have been helpful if Moon Winx had been just a tad clearer in what he was saying. Like perhaps saying something like "Ah, but that doesn't account for DUT variability that the cal lab can't know in advance". It's a little difficult to second guess that when it's cal lab's accredited capabilities that's the subject in hand.

We've got some very clever, knowledgable, and experienced people on here, but sometimes I do wish that perhaps some of them had taken an extra writing  class or two.

Hahaha I wish I had seen the comments earlier   and I totally agree, but sometimes I guess things get lost in translation..  This is an international forum and English might not be everyones' native tongue.

[Cerebus]

Hahaha I wish I had seen the comments earlier   and I totally agree, but sometimes I guess things get lost in translation..  This is an international forum and English might not be everyones' native tongue.

Exactly why I said 'writing class' for what in school we called 'English'.

Do other nationalities use the same idiom, do they call their writing, literature and comprehension classes by the name of the native language?

[Moon Winx]

OK, if I get you right your "there are components of the total uncertainty that are unknown until the measurement occurs" are all attributable to the DUT. Fine, we all understand that, and we'd have got there a lot faster if you'd said something like "those figures don't include the inherent uncertainties of the DUT itself" instead of going around the houses.

Your implication in

"It would be impossible to know the uncertainty of a measurement before the measurement is made, so stating that one would receive a specific uncertainty doesn't seem very forthright."

is that there's something misleading in the accreditation of cal labs to "measure X to y%" or "source x to y%". There isn't. This doesn't amount to a claim to something like "after supplying 10V to your DVM it will be calibrated to 0.3ppm uncertainty" it's "we will supply 10V to an uncertainty of 0.3ppm at the end of the leads we plug into your DVM".

If you still think this is not "very forthright", what quantification of a cal lab's capabilities would you deem reasonable to make? How is a national accreditation scheme going to account for every random bit of test gear that an end-user's might turn up with at the cal lab being accredited?

I don't think the accreditation bodies are interested in that. In my experience, they care if you have your process and procedures documented and your process of determining measurement uncertainty is compliant. You show your auditor how you've calculated what you claim your best measurement uncertainty is and maybe demonstrate the measurement to prove it if needed. I've never witnessed an auditor giving grief about calculating a measurement uncertainty at the time of measurement and they have always allowed an "estimated" Type A to be included in a typical budget. This is probably a holdover from the days before automated measurement systems. Having said that, in order to do a proper measurement uncertainty you must calculate it at time of measurement because you fail to capture variations in the test units (AND how the system and test unit interact (which is why I didn't specifically state the variation in uncertainty at time of measurement is due solely to the test unit)) if you don't. I believe this is going to be addressed in the newer 17025, and is mostly addressed in the ANSI Z540.3 which requires proper TUR to be determined and reported. The TUR (test:uncertainty ratio) replaced the TAR (test:accuracy ratio) for the very reason we are having this discussion: it's not about the accuracy of your measurement system but of the combined measurement process of the system and the test unit.

[Cerebus]

Much better, now you're making some sort of sense.

[Moon Winx]

Much better, now you're making some sort of sense.

If you don't understand something, just ask questions. I'd be more than happy to answer what I can. There's no need to act like an ass.

[Cerebus]

Much better, now you're making some sort of sense.

If you don't understand something, just ask questions. I'd be more than happy to answer what I can. There's no need to act like an ass.

I'm not having a go, and I did ask questions. It's just that you were expressing yourself so badly that you were incomprehensible. You had something very simple and non-controversial  to say (and could probably have been put clearly into one sentence) but you managed to say it in a way that was impenetrable and initially sounded quite close to nonsense. Note, I was not the only one to think so, as you can see above.

If you think you're having problems with me, wait until you hit the non-native English speakers on here, some of whom who shoot first and ask questions after.

[Moon Winx]

I'm not having a go, and I did ask questions. It's just that you were expressing yourself so badly that you were incomprehensible. You had something very simple and non-controversial  to say (and could probably have been put clearly into one sentence) but you managed to say it in a way that was impenetrable and initially sounded quite close to nonsense. Note, I was not the only one to think so, as you can see above.

If you think you're having problems with me, wait until you hit the non-native English speakers on here, some of whom who shoot first and ask questions after.

Obviously I'm biased, but what's so hard to understand about this:

"It would be impossible to know the uncertainty of a measurement before the measurement is made, so stating that one would receive a specific uncertainty doesn't seem very forthright. I know that in practice the uncertainty contributors that can only be determined at the time of measurement are usually swamped by other sources of uncertainty, but I also know of many cases where that is definitely not true."

Read carefully. I'm stating that a lab that tells the customer that they will receive a certain measurement uncertainty before the measurement is made is not being forthright. And that's true. The Scope of Accreditation lists the lab's capabilities, not the uncertainty you will see on a test report as a customer.

You finding this incomprehensible says more about you than it says about me.

[Cerebus]

You finding this incomprehensible says more about you than it says about me.

Now who's being insulting?

The man you're talking down to to used to be a section editor on a national magazine and writer for several others. I know how to write and I know how to edit other people's writing so that it makes sense to a target readership. I've had to take writing from specialist contributors who can't write for toffee and get it to make sense.

It also means that I've been edited by others and learned how to not get in a huff if someone says "You're not being clear here". If I took your attitude I'd have blamed the readers instead of learning how to write clearly.

I'll leave others to judge whether all that qualifies me to talk about writing clarity.

As to what says what about whom, I haven't stooped to insults - pause and think about the implications of that.

[Moon Winx]

I'm not having a go, and I did ask questions. It's just that you were expressing yourself so badly that you were incomprehensible. You had something very simple and non-controversial  to say (and could probably have been put clearly into one sentence) but you managed to say it in a way that was impenetrable and initially sounded quite close to nonsense. Note, I was not the only one to think so, as you can see above.

If you think you're having problems with me, wait until you hit the non-native English speakers on here, some of whom who shoot first and ask questions after.

...to me, this is a little insulting and I hope you did better work in your editing job than you've displayed here. I would hope that you would offer more feedback to a writer instead of "whah... I don't understand so this must be incomprehensible and you are expressing yourself badly!". This is a message board, not an academic journal, so I expect the conversations to be in a casual tone and again, if you find something confusing there are more polite ways to ask for clarification rather than insulting the poster's writing style.

But I'm tired of this back and forth that brings no value to the thread, so this will be my last post in here.

[Cerebus]

But I'm tired of this back and forth that brings no value to the thread, so this will be my last post in here.

Good!