My K-type thermocouple (grounded enclosure type) connected to MAX31855 adapter connected to Arduino Mega gives me faults every few readings. It's really, really bad. I've tried swapping thermocouples, adapters, and Arduino's in various permutations and reached the conclusion that this happens whenever the thermocouple sheath/shielding touches the metal frame I've got the rest of my contraption set up on. Also gives faults when I touch it. Works perfectly fine when floating or completely isolated from everything else.

I've tried connecting myself to ground, and I've tried connecting the frame to ground, but nothing works. What should I do? I thought this was the whole purpose of the shielding. :(

When I check continuity on the probe, I get 3 ohms between probe tip and one lead and 7 ohms between probe tip and other lead. This is exactly the same for another thermocouple I've tried. No surprise here, as it's a grounded type enclosure thermocouple.

Probe and module on rig

  • \$\begingroup\$ I've dealt with this type of issue in the past (ground loop, conducted pickup). The best, most-certain way to avoid this issue (especially when you may not have control of the rest of the equipment), is to buy ungrounded (and as a result, un-shielded), thermocouple sensors. They are readily available. You can also play games by grounding one of your grounded-thermocouple leads at the conditioner end (if the conditioner supports it), and hope for the best. \$\endgroup\$ Commented Apr 5, 2023 at 17:27
  • 3
    \$\begingroup\$ What does "gives me faults" mean?? Please be specific. \$\endgroup\$ Commented Apr 5, 2023 at 17:48
  • \$\begingroup\$ @ChrisKnudsen As I just found out "ungrounded" in TC jargon, is in fact a shielded type, but one where the shield isn't bonded to the junction. The unshielded type seems to be called "exposed". See here: temprel.com/learning/selecting-your-junction-type \$\endgroup\$
    – tobalt
    Commented Apr 5, 2023 at 21:09
  • \$\begingroup\$ related thread: Grounded thermocouple \$\endgroup\$ Commented Apr 5, 2023 at 22:10

1 Answer 1


Disclaimer: As Chris Knudsen has rightfully indicated, this answer is only applicable to shielded TCs without a junction-shield bond (i.e. the "ungrounded" type). However, the OP asks about the "grounded" type with a junction-shield bond, to which the discussion/suggestion below are not applicable.

The shield is only effective if is at a potential that the thermocouple wire pair is, too. At least, it should have a constant voltage to the wire pair. Else, the capacitance between the shield and wire pair will couple in common-mode noise currents. These are not perfectly rejected by the thermocouple readout instrumentation amplifier and a part of it will mimic a fake differential signal. AS the TC voltage is so small, a small differential noise voltage will make a large impact on the reading.

If your setup is powered from an isolated switching converter, that usually means that its common-mode voltage with respect to Earth is something like 50-100 VAC at mains frequency.

If you don't touch the shield anywhere and just co-float it with the rest of the setup, it crudely-speaking "disappears" and you obtain approximately the behavior of an unshielded wire pair.

However, when you touch the shield or when you attach it to PE, you force the shield onto a different potential than the wire pair. In this case, the aforementioned large noise can appear.

If you are looking for a robust and shielded layout of your setup, I suggest you attach the shield to your system power supply negative rail. Just to be clear: I am talking about the negative supply of - in particular - the MAX adapter, i.e. its "Ground", which might be also the "Ground" of the arduino.

  • 1
    \$\begingroup\$ "...when you attach it to PE, you force the shield onto a different potential than the wire pair." Except in a "grounded thermocouple" (which is what he has described) where one lead of the thermocouple is bonded to the thermocouple junction at the end of the sensor. All is fine, until the thermocouple itself contacts some other chassis bit - somewhere where nasty ground currents may be running (and their shady friends, The Potentials). This is a direct conducted interference - the shield is directly connected to the thermocouple junction at the end of the sensor. An unfortunate setup IMHO \$\endgroup\$ Commented Apr 5, 2023 at 20:56
  • \$\begingroup\$ @ChrisKnudsen You are right, I just googled and found out that there are indeed two different kinds of shielded TCs: Those with a regular shield, and ones with the shields bonded to the TC junction. My suggestion applies only to the former, but not the "grounded" type. I think I'll leave the answer up anyway for whomever it could help.. Btw, that also explains the varying degree of success I have had with this suggestion on visually "shielded" TCs lol. I can't see any reasonable usecase for the "grounded" TCs either \$\endgroup\$
    – tobalt
    Commented Apr 5, 2023 at 21:07

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.