Thermocouple Ensemble Comparison

Although I don’t have a good way to put thermocouples in a known temperature environment (ie, yes, I can freeze and boil water, but I doubt the trustworthiness of any measurements made therein), I can compare the results from several different thermocouples held at the same (unknown) temperature: if they all agree to within a reasonable tolerance, I assume that they’re all reporting the correct temperature. Even better, if some of the temperatures come from different thermocouple amplifiers, then the electronics cross-check each other.

With that in mind, I attached a 4 Ω 25 W aluminum-body power resistor to the back of the same isothermal block I built for the Thing-O-Matic extruder tests, atop a dab of heatsink compound for good measure:

Isothermal block with 25 W power resistor
Isothermal block with 25 W power resistor

Then I tucked four sensors into the drilled holes:

Sensors in heated aluminum block
Sensors in heated aluminum block

Clockwise from the front left corner:

  • T1: Fluke meter, eBay TC (black sheath)
  • T2: Fluke meter, Fluke TC (brown sheath)
  • TM5: TC4+Arduino, eBay 100 kΩ thermistor (invisibly fine bare wire leads)
  • TC1: TC4+Arduino, eBay TC (black sheath)

The TC4 Shield handles both thermocouple and thermistor sensors, so I added a thermistor from my collection just to see how it worked. The datasheet gives these parameters:

  • 100 kΩ resistance at 25 °C
  • beta = 3950 (from 25 °C to 50 °C)
  • beta = 3990 (from 25 °C to 85 °C)

Unfortunately, there’s no way to include the eBay thermocouple epoxied to the nozzle in this test, but it’s from the same lot as the two in this test, so I assume it’ll produce the same results. The consistency in earlier tests suggests they’re all really Type K thermocouples and produce the same results as the Fluke thermocouple and meter that I assume produces accurate readings.

The closed-cell foam insulating the block from the vice jaws seemed like a good idea at the time.

I connected the resistor to the bench power supply, channeled the true spirit of DIY 3D printing (ie, ignored the power derating curves), and fired it up:

Multiple Sensor Calibration - vs time
Multiple Sensor Calibration – vs time

The successive steps correspond to power levels of 0 W (the ambient temperature), 1 W, 2 W, 3W, 4W, and 6 W. The last point established that the foam melts at slightly over 100 °C, whereupon the test terminated. Eyeballometrically, the time constant of the resistor + block is on the order of 20 minutes, so these few points represent a rather tedious Basement Laboratory session.

Plotting the temperatures against the reading for T2, the Fluke thermocouple connected to the Fluke meter, and thinning the lines makes the results a bit more obvious:

Multiple Sensor Calibration - vs T2
Multiple Sensor Calibration – vs T2

The three thermocouples and Fluke meter / TC4 Shield pretty much overlay each other, with the thermistor reporting a somewhat higher temperature. Given that TM5 is an eBay thermistor, I’ll let you judge whether the beta value I got from its listing matches the beta of the actual thermistor.

In any event, I’d say the thermocouples report a temperature within at most a degree or two of the actual temperature. Plus, I didn’t get a steam burn in the process…

7 thoughts on “Thermocouple Ensemble Comparison

  1. As soon as I saw the picture of the resistor in the foam, I thought “he’s going to melt that foam!”.

    1. Like the horror films where the audience starts chanting “Don’t open the door! Don’t open the door!”

      I did not dress in my skimpiest underwear, so it wasn’t as ugly as it could have been…

  2. Re steam burns: In a memo to self, I learned that it’s not a good idea to debug a hot coffee maker (flaky auto-stop valve) before you’ve had a cuppa. There was a little water that went back to the heater, and my finger was in the way when it came back out as steam/hot water. Seems it takes about 4 weeks for the second degree burn scab to finish healing.

    I haven’t yet decided if the auto-stop on coffee makers is a Good Thing or a Tool Of The Devil. I think it depends on my blood-caffeine level.

    1. the second degree burn scab

      Until you’ve had that happen, you really don’t know why steam burns are so terrible. Just reading about guys who get full-body burns is a queasy experience; it converts the abstract concept of Enthalpy of Condensation into visceral knowledge.

      Done it a few times, on a very small scale, and that’s entirely enough for me.

  3. I took a heat transfer course 50 years ago. Essentially I learned two things. The once maybe useful thing is why the insulation on steam pipes is shaped and sized the way it is. The other was that almost all of the formulas had about as many empirical constants as variables.

    If I understand the physics correctly, thermocouples depend only on the composition of the wires so two of the same type should work the same although I suppose impurities might produce a variance. I do wonder though about EMI from the heater driver PWM getting into the thermocouple electronics.

    You might consider a rapidly circulating oil bath to speed up stabilization in your calibration experiments.

    Don’t give anything I say two much weight particularly if it doesn’t make sense. I have been known to say more than I know but I do try to avoid it.

    1. EMI from the heater driver PWM

      That, at least, seems to not be a problem, as the firmware averages (sorta, kinda) 16 successive temperature measurements. I ran the tests with a DC supply and the results match up with measurements on the extruder, so they’re pretty close.

      thermocouples depend only on the composition of the wires

      For a while, I was expecting to discover, ah, shall we say, melamine contamination, but all of the thermocouple beads nuggets agreed with each other.

      Definitely a puzzle…

Comments are closed.