Cold Junction Compensator - how to decide

[dl1640]

hi

i want to choose a device to generate/maintain a constant temperature to calibrate some temperature measuring and indicating instrument, with the help of some good quality thermocouple. as i know such device is a cold junction compensator and my target temperature range is 0~100C, accuracy +/-0.1C or less, short-term stability +/-0.05C or less, and need it to be compact as we are very small lab.

do not want to diy, just want to buy some on the market.
please share your opinion.

also please tell what kind of thermocouple is good quality, i use T type or K type, un-calibrated

thanks

[HKJ]

Thermocoupler is not the right way to keep a reference stable.
Some semiconductor sensor or a PT100/PT1000 is much better.

[Vgkid]

You could also look into bead/glass thermistors for this.

[dl1640]

i feel it is difficult to have a thermocouple calibrated.
we sent a piece of T type thermocouple to external calibration institute, calibrated error is 0.3C and measurement uncertainty is also 0.3C and i feel such calibration result is useless.

[HKJ]

i feel it is difficult to have a thermocouple calibrated.

That is one detail, the other detail is that thermocouplers always gives a temperature difference and you must have another sensor to get the absolute temperature. In a 0~100 range you can just as well use the other sensor for the full range and avoid the thermocoupler.

[David Hess]

Only the best self contained calibrators are that accurate; thermocouples are not even close.  Were you expecting to pay thousands of dollars?

https://www.omega.com/section/precision-calibration-equipment.html

[dl1640]

i feel it is difficult to have a thermocouple calibrated.

That is one detail, the other detail is that thermocouplers always gives a temperature difference and you must have another sensor to get the absolute temperature. In a 0~100 range you can just as well use the other sensor for the full range and avoid the thermocoupler.

Hi Hkj

In fact what i need is the temperature difference, the reason is the UUT the temperature measuring device measures temperature with the help of a thermocouple in real life, the only temperature sensor inside the temperature measuring device is called a RJC (reference junction compensator), usually it is mounted near the input terminals of the device, the abosulute temperature is calculated of thermocouple output + RJC, normally RJC is around room temperature.

So a standard thermocouple or a well calibrated one is preferred as i can compensate the errors caused by the thermocouple during the calibration of the temperature measuring device.

Then with the help of a known and stable temperature source, we can complete the calibration setup, i think.

I said target temperature range is 0~100C, but indeed just one point calibration is fine to verify the RJC, such as 0C or 20C, i think.
Range calibration can be done by turning off the RJC and calibrate by inputting dc voltage with a precision voltage source.

[dl1640]

UUT is such a device 

https://www.ebay.com/itm/Yokogawa-Hybird-Chart-Recorder-LR-8100-/271648646956

[dl1640]

Only the best self contained calibrators are that accurate; thermocouples are not even close.  Were you expecting to pay thousands of dollars?

https://www.omega.com/section/precision-calibration-equipment.html

Hi David

I figured out we need both ice point bath (for reference junction) and furnace (known and stable temperature), that way we can calibrate UUT’ internal RJC and thermocouple as well.

Thanks 

[David Hess]

I figured out we need both ice point bath (for reference junction) and furnace (known and stable temperature), that way we can calibrate UUT’ internal RJC and thermocouple as well.

If I was doing this on the cheap for roughly room temperature calibration, I would use an ice bath, boiling water bath, and a water triple point cell to provide three accurate reference temperatures.  All can be constructed by an individual with a minimum of resources and calibrated knowing only barometric pressure.

[dl1640]

Hi David

I too use a roughly a bottle filled with ice and pure water to get a 0C reference junction. However if our company want to pay for better instruments then why not:)
Besides Omega there is Fluke doing such bussiness as I know they will be more money...

[nfmax]

I think what you probably need is what is called a 'dry block calibrator' - basically an insulated, temperature stabilised lump of aluminium into which you can insert multiple thermometers for inter-comparison. They are quite compact & generally fit on a bench top. While the calibrator will keep the block temperature stable, the uncertainty in the stabilised temperature is generally quite large. Hence you should also invest in a 'semi-standard' PRT thermometer and accompanying readout, with individual calibration. This should get you to about 20 to 50 mK calibration uncertainty over your temperature range of interest. Depending on the performance (uniformity of temperature between the thermometer positions, temperature stability over time) of your calibrator, you should be able to maintain 0.1 K overall uncertainty. Thermocouples, unless individually calibrated, cannot meet these uncertainty levels (and even then are subject to drift, requiring frequent recalibration).

Handled with care, platinum resistance thermometers are very stable. While annual recalibration is preferable, because the most likely cause of error is shift in their ice point resistance due to mechanical stress or abuse, regularly measuring and recording the thermometer's resistance at ice point, or preferably at the water triple point, will warn you if something has changed.

Maintaining an ice point bath to within 20 mK is not as easy as it sounds, though, and water triple point cells are rather bulky, fragile, and require training and care in use. If it were my decision, I would use a third-party calibration service on a periodic schedule.

[Edit to add] I should emphasise that the job of the dry-block calibrator is just to keep the temperature of two or more thermometers constant and the same, it isn't itself particularly accurate. For that, you use a calibrated PRT as one of the thermometers being compared.

[2N3055]

Take a look at this..
It so happens that well  made distilled water ice bath can be used instead TPW for most of uses...  They achieved Expanded Uncertainty, 95% (k=2) as low as 3.5 mK..
Also there is an overview of RTD calibration...
For higher temps use a dry well...

[HalFET]

Been using an ice bath for ages now. Other than having to show up in the lab at weekends to replenish the ice once in a while it is easily the best method on the cheap.

[nfmax]

Been using an ice bath for ages now. Other than having to show up in the lab at weekends to replenish the ice once in a while it is easily the best method on the cheap.

Agreed - so long as you know what you are doing!

[HalFET]

Been using an ice bath for ages now. Other than having to show up in the lab at weekends to replenish the ice once in a while it is easily the best method on the cheap.

Agreed - so long as you know what you are doing!

My main advice if you wish to do this:

• Get a large double walled  thermos flask/can as vessel.
• Wrap the entire thing in glass wool insulation.
• Use DI water for the ice.
• Don't use tape and instead epoxy wires in place
• Stick an RTD in there with it.
• Monitor air pressure if you're measuring long term steady-state.
• Silica gel (or another drying agent) is your friend to prevent condensation in unwanted areas.
• Put the entire thing in a styrofoam box for added effect.

The better you insulate it and the cleaner everything is the better the result really. Ratio of ice to water is something you should figure out yourself, you can also go for other "reference solutions" if you want, like dry-ice combined with solvents.

[dl1640]

So many inputs 

The temperature measuring device, by its nature, is a Voltmeter + a RJC (a thermistor circuit or so).
It measures both TC output and RJC output, then calculate final result for abusolute temperature.
A PRT or SPRT of course is better than thermocouple, but it cannot calibrate the temperature measuring device under TC mode.
If use PRT, it only calibrate under RTD mode, I think. So TC mode will not be checked and calibrated (dcv + RJC).
I see NIST on their website introduce how to calibrate thermocouple, here is link, on the bottom of this website has some reference documents.
https://www.nist.gov/calibrations/thermocouples-materials-thermometer-indicators-calibrations#32010C

My idea is "stable source + ice point + precision dmm + good thermocouple".
To calibrate thermocouple, we need "stable source + thermocouple under test + ice point (reference junction) + precision dmm".
To calibrate temperature measuring device under its TC mode, we need "stable source + previously calibrated thermocouple".

So uncertainty of dmm and ice point need to keep low..

Maintaining an ice point bath to within 20 mK is not as easy as it sounds, though, and water triple point cells are rather bulky, fragile, and require training and care in use. If it were my decision, I would use a third-party calibration service on a periodic schedule.

This is one of the options to keep budget down and ease of use. But that means we have thermocouple calibrated externally so re-calibration is not convienient and thermocouple should be checked everytime before we conduct the calibration of temperature measuring device.

Need more study..thanks

[nfmax]

It isn't quite clear to me whether you are trying to calibrate the thermocouple, or the thermocouple indicating instrument (or both together).

If you are calibrating the instrument alone, you can always use a precision millivolt source in place of the thermocouple and a precision resistor in place of the reference junction sensing thermistor. Temperature need not come into the process (except for the intrinsic temperature sensitivity of the instrument itself).

If calibrating a thermocouple, you need a stable, known temperature (or possibly two if you want to calibrate your reference junction sensor at the same time). This can be measured using any type of thermometer, not necessarily a thermocouple: a PRT is more suitable.

You need to ensure your local standards are themselves traceably calibrated, ideally to national standards.

[dl1640]

Hi

i want to calibrate thermocouple indicating meter. but if i can also calibrate the thermocouple used in the calibration, it is very good. i do not know how to calibrate a built-in RJ sensor with a standard resistor, seems impossible because no place to hook up a resistor to a built-in sensor and read some data..?