Thermal paste did not result in Power semiconductors runnign that much cooler

Hi,
We ran two thermal tests (with K type thermocouples) on our 60W offline Flyback LED driver prototype.
This is just a test unit for the software engineer to work on, so the heatsink on the TO220 power FET and the TO220 power Diode are just small ones as follows (Farnell 161145)
TO220 heatsink:
https://uk.farnell.com/wakefield-so...y=https:en-GB/Element14_United_Kingdom/search

Anyway, in both thermal tests, the internal ambient inside the LED driver unit was the same at 73 degrees.

However, the test with thermal paste added to the heatsinks resulted in the FET/diode running 2 degrees hotter than without the thermal paste.

This seems strange, you would expect thermal paste to result in the power semiconductors running cooler.

Anyway, this could in some way be accounted for by the fact that the test where the thermal paste was used was done at 65W power level, whereas the previous test, without thermal paste, was done at 61W.
However, even then, I would have expected the thermal paste to result in cooler FET and Diode.
Another point is the possibility of some reading error…since the Mains Rectifier (which doesn’t have a heatsink, and so no thermal paste in either test case) was measured as running 4 degrees hotter in the test where the thermal paste was used for the FET and Diode…
Anyway, even accepting the slightly increased power level, and that there may be a few degrees of measurement error, I would still expect the prescence of thermal paste to have resulted in cooler running temperatures recorded for the FET and Diode. Do you think that the extremely small size of the heatsinks used are the reason why the thermal paste didn’t give much of a difference in temperatures to the FET and Diode?

Hi,

Thermal paste just reduces the thermal resistance between FET_pad and heatsink.

But there are a lot of other parameters that influence the FET junction temperature.
* inside FET thermal resustance r_th_jc
* power dissipation
* ambient temperature
* heatsink thermal parameter

For verifying thermal paste function:
* use equal dissipated power
* measure / calculate the difference of FET_case_temperature to heatsink_temperature.

Klaus

Sounds about right to me. The paste doesn't reduce the heat dissipation, it just helps to spread it around a bit better by filling in the bumps on the heatsink and transistor tab. A perfect transistor and heatsink surface would result in poorer conduction if you added paste! With 28C/W increase the rise intuitively seems to be what I would expect.

Small heat sinks are usually applied without thermal paste, there's no relevant reduction of K/W, simply because the heatsink thermal resistance is already large. The effect of thermal paste is to fill small voids in the interface, e.g. due to surface roughness. If applied correctly, it surely doesn't increase the thermal resistance. Not considering the unrealistic case that you have perfectly flat surfaces at both sides. Anodized aluminium has always a certain roughness and will take advantage from thermal paste.

Different results may be due not fully reproducible conditions.