Author Options:

Thermocouple calibration - How to Answered


I have a few questions about the thermocouple calibration.

I want to use three type R thermocouples to measure the temperature during a 3d printing process using an Arduino Uno and the max31856 amplifiers. Right now I am doing tests with K type thermocouples before we buy the R-types. It seems to work fine, it measures a room temperature of 22ºC more or less and reacts reasonably in my opinion but I would like to know if the measured temperature is the actual temperature.

1. Is a calibration needed? I know the amplifiers (max31856) do some corrections automatically, but maybe an extra calibration is needed.

2. If such a calibration was needed, how would I do it? For example, I will put copper in an oven and heat it and measure the temperature with my thermocouples and with a bolometer. If my thermocouples measure something different, how do I correct it? Theoretically my program transforms (via amplifiers) the measured tension difference into temperature, correcting it at the junction. How would I tell it now that it has to measure differently?

3. If no calibration is needed and, as I explained in 2, my thermocouples measure something different as the bolometer, which one is the accurate?

Thank you very much for your help.


Calibration of thermocouples (and other temperature sensors) is always needed. Calibration is what relates the electrical signal to the real world. The most fundamental calibration of temperature sensors is the Triple Point of Water (0.01 deg C) but that is difficult for a DIY'er to achieve and maintain. However the Ice Point of Water (0.0 deg C) is achievable and practical.

The internet has (of course) a multitude of sites showing how to make an ice point apparatus. Some of them even work. The things needed are: a vacuum insulated container such as a glass lined Thermos(R) container; distilled water; ice made from distilled water; an ice shaver (such as used to make flavored shaved ice treats in the summer); and a length of flexible tubing. Towels are also useful, both to clean up spills and to insulate equipment from drafts. It is helpful if the vacuum container can be kept in a refrigerator for several hours to pre-chill. The ice needs to be shaved - not crushed or in cubes.

The vacuum container needs to be filled with a mixture of the distilled water and shaved ice. One end of the flexible tubing goes down to the bottom of the container and the other end goes to a bowl or sink to drain excess water. (There should be an equal mix of ice and water, but the is must be floating.) Add shaved ice as needed to keep it full. After allowing time for a stable temperature to be achieved, the thermocouple to be calibrated can be placed, allowed to stabilize, and then record the value. (A better practice is to do it three times and take the average.) If you have a reference thermometer of known accuracy and uncertainty you can also use that as a reference. (Translation - the reference thermometer should be calibrated.)

The ice point gives you a single known point for the thermocouple. Getting a known reference for the other temperatures is more difficult. The best solutions are expensive - a calibrated platinum resistance thermometer, or thermometer calibration setup that can generate the temperature. (I used those and other things in calibration labs, but I'm retired from that now.)

The use of an infrared (IR) thermometer as a reference may be useful depending on your situation. They can be reasonably accurate provided they are well maintained, the lens is clean, all instructions are followed and the "emissivity" value is correctly set. The distance from the area being measured can be critical - including enough in the field of view while excluding areas that may be hotter or colder. An moderately expensive one may have accuracy of (+/- 1% of reading or 1 deg C) with repeatability of (+/- 0.5% of reading or 1 deg C). A cheap one may have the accuracy stated as simply (+/- 2 deg C). (Note: "or" means "whichever is higher".) It is up to you to determine if this is precise enough for your needs, keeping in mind that the inaccuracy of the IR thermometer adds to the inaccuracy of the thermocouple.

Wow thanks for the answer, I really really appreciate that you took the time.

I don't need an accuracy of 1K or so, a few would be ok too. So in the end, after asking in a few places and talking it with a collegue we have found a solution. It should work with our IR thermometer, having the difficulty to find the correct emissitivity value of our material. But it should be fine!

I really appreciate all the answers, it always amazes me how many people there is willing to help.


(Do I have to do anything to end the post? Sorry I'm a noob, I don't know anything)

There was a superb "amateur Scientist" column in Scientific American years ago showing how you can make your own triple point cell.

I suppose the other factor we need to know is what the OP considers "calibrated" Like you, I work in a world with uncertainties of 0.01 K. Maybe the OP thinks 1 K is good enough

A type R thermocouple appears to be fairly linear


Depending on the accuracy you require you should be able to do a check at 2 suitable points on the graph against a known accurate thermometer (IR or otherwise). You can then extrapolate the results up to the higher temperatures, it should be linear.

So, if I understood correctly, in my oven experiment I can measure two points with my thermocouple and the IR thermometer (why are they not to be trusted @steveastrouk?) and compare. Let's say my thermocouple always measures 50 ºC fewer than the IR thermometer. In the future I can say the real temperature is always the measured minus 50ºC?

@iceng I'm sorry, I don't really understand what you sent me. Is it another kind of thermocouple?

Yes essentially your right. Steve is also correct re IR thermometers, BUT a good one works OK if it works. Much depends on where your focused. At oven temps you should be able to get hold of a reasonable accurate digital thermometer.

Never trust an IR thermometer.