IR and thermocouple measurement difference on large and high emissive objects

Question

First of all, the question is similar to this one.

I am working on a high-power, high-density power supply where the size and heat dissipation are quite limited. In the final product, the PCB will be enclosed in a metal chassis that has some "thermal" connections to the PCB (through gap pads) and a fan (driven by temperature-controlled driver logic).

During engineering tests I let the unit running with no cover/case/chassis whatsoever under room conditions. And I supplied a small fan temporarily to prevent any damage due to potentially excessive heat.

Now here's the interesting part:

During coverless tests I took temperature measurements of the transformer core and windings with a Fluke thermal cam. The results were quite high i.e. close to the safety limits. The core was beyond 100°C and the windings were close to 110°C.

Then I attached thermocouples to the core with epoxy (not to keep them in place but to isolate them from ambient to prevent false readings as epoxy is terrible in terms of heat conductivity) and I chucked another thermocouple into the windings and secured it. Then I put the entire "thing" into its original chassis having original fan blowing. The read results were lower: Core was less than 65°C, and windings were less than 90°C.

How could the measurements differ by huge amounts? How could a transformer have "lower (?)" surface temperature when it's fully covered?

NOTE:

• I'm sure the TCs are properly touching to the heat sources.
• I'm aware of the fact that the sensor itself helps to cool down the surface as the sensor is a metal ball so it may steal some of the heat. But since the surface is large enough, the cooldown effect of metal ball-tip sensor is negligible.
• The ambient temperature inside the chassis is, obviously, higher than the environment (around 60°C when the PSU is fully loaded).
• I can't put any schematics, photos, diagrams etc. as it's against the company rules.

Sorry for this long question. I tried to give the necessary details in "shortest" sentences.

Answer 1

Looks like I have made a grave mistake and ignored lots of external parameters.

So …

How could the measurements differ by huge amounts? How could a transformer have "lower (?)" surface temperature when it's fully covered?

The squishy gap pad which sits between the transformer core and the outer chassis provides really good thermal conductivity as well as isolation. So it "steals" some hot juice from transformer's core and transfers to the chassis. That's why the core is cooler.

Also, it appears that the forced air coming from the fan is more concentrated (and therefore, heavier) when the unit is fully covered, so it gets easier for the forced air to push the hot air around the hot components away. That's probably why the winding (and everything else's) temperature is lower when the unit is covered.

And, just for comparison, for the caseless run, the temperature measurements taken with thermal cam and thermocouples are close (differ by ±2 °C).