Traceability of Thermoelement Batches

Benjamin Weber

Trusted Information Resource
We are currently using a set of 6 thermoelements only for temperature measurements during testing of medical devices. So far each thermoelement is calibrated once a year. Since the calibration of the thermoelements is quite expensive, we only have a limited number of these. The problem is, that the temperature tests we are doing usually has to be performed at more locations than we have thermoelements. We solve this by repeating the tests after the the thermoelements have been placed at the different locations.

I would like to change this by the following:
- Buy a set of thermoelements from one production batch (lets say 20) that come with a batch certificate from the manufacturer, stating compliance with the corresponding accuracy class. This would be an ISO certificate only, not an accredited calibration certificate.
- Calibrate a sample of this set by an accredited calibration lab (maybe retain this one as a reference).
- Add a safety margin to the uncertainty of the calibrated sample and apply this extended uncertainty to the whole set (that even already has the ISO certificate for compliance with +-0,5K, but better safe than sorry). Lets say I use type T thermoelements and the calibrated sample complies with +-0,5K (type T, class I), I could maybe add another +-0,5K resulting in an uncertainty of +-1,0K for the whole set.
- After a given use time (lets say a year), dispose the set of thermoelements and start at again with a new set.

I would like to not perform in-house calibration, because I fear that the requirements for this would outweigh the benefits.

This would lead to less calibration costs (1 vs. 6 times). The costs for an accredtited calibration of one single thermoelements are approximately the same as for buying 10 new ones. So we still safe money here. And we could use more thermoelements (20 vs. 6) in order to save time and efforts during testing. Addtionally it would also be OK to perform tests, were we lose a thermoelement (e.g. due to having to cement it to the device under test). Currently sacrifying a thermoelement is not an option.

My concern is, that the accreditation body will give me a non-conformity, since not each single one of the thermoelements is metrologically traceable (only the one that has been calibratied by an accredited calibration lab). I tried to get some information from the accr. body, but unfortunately they are a little reserved, saying they cannot really say if this is OK or not, unless they have seen this during an official audit :-(

What do you think?


Trusted Information Resource
It sounds like you are trying to do the right things here, but your concern about a non-conformity due to lack of traceability is well founded. You most probably would get one in an audit.

A possible solution might be for you to get the batch of thermoelements and send one out for the accredited calibration to become your reference standard. Then take advantage of 17025 allowing a test laboratory to calibrate their own standards without needing a calibration scope of accreditation. Compare your other thermoelements to the reference standard, keeping the conservative 1°K tolerance you have mentioned (that would keep your measurement uncertainty below the tolerance). Now you have traceable calibrations on your elements.

You will need to do the back end work to support these calibrations: generate a calibration procedure, calculate an uncertainty budget, record the results, etc. In the end you should have a basic calibration for your elements that will withstand review.

Benjamin Weber

Trusted Information Resource
Thank you very much for your opinion. In-house calibration (without being accredited as calibration lab) so far seems to require to much efforts.

Any other opinions on that? Do you have experiance with such in-house calibration?


Trusted Information Resource
I really can't help you there. You really have two options:
Calibrate the elements yourself.
Pay someone else to calibrate them.


Involved In Discussions
Ive, been a little on the fence replying to this one...

We are a test laboratory, accerdited as such, but we also perform in-house thermocouple calibration, the reason that im on the fence is that it is still unclear to me as to why this has not been a problem with any of the Accreditation bodies that we deal with, the national AB we use, lacks the technical knowledge to notice or inquire about it, and the 2 other international AB we deal with; do check 3 things:
1.- that our metrology well and PRT that we use as reference material is calibrated by an ILAC recognized lab
2.- that we have sampled and analized our calibration results to ensure compliance with the accuracy standard (forgot the name, and my file servers are down for maintenace, ill get you the standards name later), we basically run an excel file where we input the data form samples taken from the TC Spool and statistically validate error is within acceptable parameters
3.- that there is an internal procedure to perform these calibrations

it would be nice if some-one would banish my ignorance by pointing out where am i going wrong or right on this because altough the lab makes their best effort to ensure the validity of our TC readings i am usure as to by which means are we authorized by the standard to perform such calibration feats being a test lab.


Trusted Information Resource
I have experience with assessors for testing laboratories being short on knowledge regarding calibration requirements, so I can see your concerns.

17025:2005 was a little more specific in dealing with a testing lab that calibrated their own equipment, but that was only to state that a testing lab could perform their own calibrations (, and that they had to "estimate" their measurement uncertainties. In 17025:2017 (7.6.2) if you calibrate equipment you must "evaluate" your measurement uncertainty - do the math and show the results. In either case, a testing lab can calibrate their own standards, so you are OK there.

You are getting accredited calibrations with uncertainties for your well and reference PRT. Good, that is required for traceability.

You have a calibration procedure. Good, but since it is an "internal" procedure you are going to have to validate the procedure, and maintain records of the validation results. This is covered in 7.2.2 of 17025:2017. You will need to have a validation procedure for this as well, and a statement in it "detailing its fitness for use" (I have been written up with non-compliances for missing that statement!).

You have analyzed your results. Good, now you need to produce your measurement uncertainty budgets. Include all of the significant error contributors that affect your results, such as axial and radial uniformities, possible insertion depth issues, thermal transfer from the well to the TC elements, etc. Include the calibration uncertainties for your standards produced by your calibration providers.

You have a method, you have evaluated your uncertainty, and you have demonstrated measurement traceability. It sounds like you are ready.


Involved In Discussions
If i am verifying my results to comply with ASTM E 220 and as i understand that standard is a also a calibration method for thermocouples... what if i just use that method, then take care of my uncertainty budget, as i understand i would effectively save myself the trouble of method validation right?

i mean, i'd really rather convince my organization to spend the money on purchasing a copy of the standard (i'll even pay for it myself) if it saves me the trouble of doing method validation


Trusted Information Resource
Using a standard like ASTM E220 will take care of the validation issue.
Basing your calibration method on ASTM E220 is a different story. Unless you follow E220 completely you will need to justify (and validate) any changes you make to the procedure.
  • Like
Reactions: Gus
Top Bottom