My present project is a ~6kW switching converter in a sealed stainless steel box. The box is grounded well. When I put a thermocouple on the surface of the box, the reading is noisy. Even if I put some tape between the box and the thermocouple, I get the same result. All thermocouples in the box give reasonable results, but the one on the actual box surface gives me so much noise as to be useless. My PC temp logger box rejects all the results as outliers and never updates. A calibrated multimeter gives me readings 30C higher than three IR guns (pointed at black tape on the box) and a PN junction temp meter give me. This problem only seems to occur when the unit is running. Grounding or not doesn't seem to matter. Two brands of thermocouples give the same results.
Why would I be getting such bad thermocouple readings in that particular circumstance? Are thermocouples just that untrustworthy and I'm getting unlucky? Or am I missing something?
Best Answer
I think you need to use an instrumentation amplifier like the AD8221. It does sound like common mode noise. My company fit plenty of thermocouples on aircraft engines directly to blades and other hot and large pieces of metal and we still get noise but it's livable with when using an IA.