Thermal Pad Tests

Excellent report and good work. Time to go and order some Al2O3 insulators.

Regarding the mounting, there is several appnotes from various manufacturers showing that clamping reduces thermal resistance between body and heatsink of the TO-varietys due to less physical deformation as compared to when using mounting screws. Especially for the mounting tab types.
 
That depends on the TO variety - its not the case for TO3 for instance! Single mounting screw versions its certainly an issue - especially if you over-torque the mounting screw and bend the tab - since the tab is soft copper this is an easy mistake to make.

The MT200 package design with two mounting screws on either side as partly I think to address this issue.
 
Mark, you are absolutely right, thanks for the clarification.

I actually successfully mounted TO-264s on a predrilled heatsink (TO-3 /TO-204) for reuse in a variable power supply. I used a bar and screw+nuts to press them against the sink. Work just fine.

(I actually used MJL21194s that was reused from a blown power supply, ie blown as transformer meltdown)
 
I've done a less scientific test using a Flir One Pro and screw mounting of devices, I will concur that thermal pads leave something to be desired overall. That said, I got 0.55 degrees C/W with a Wakefield CD-02-05-264 and a Toshiba 2SA1943-O dissipating 30W after its phase change and 0.8 C/W before. With the phase change type pads, the device and the heatsink need to be >65C for 30 minutes or so the get their performance at its peak. With a 4180G on a TO-264 device, I measured 0.5C/W and 0.63C/W with the Berquist SP2000-0.015-00-104. (The TO-247 pads just do cover the metal backing of a TO-264.)
I've found the Bergquist Silpad 2000 and Laird Tgard 5000 to be the best thermal pads available from common suppliers. These still are better after the device had been mounted for 24 hours as it "conforms" to the surface irregularities. As an example with the Laird Tgard 5000 and a Sanken TO-3P screw mounted, it started at 0.83 C/W and eventually settled at 0.7 after 24+ hours. This is also what I got with 3 mil mica+compound. I've never tried Keratherm. The 4180G and GC electronics heatsink compound on a Sanken TO-3P device was down to 0.55C/W. With an LM350T as the heat source for TO-220 testing here's what I got, again, with screw mounting (I can't use the thermal camera with a clamp, of course) Dry mounting 2.5C/W direct, with compound 1C/W, Aavid 4170G 1.3C/W, ~3 mil mica+compound 1.6 C/W, Thermasil III 2.34C/W. As far as TO-220 devices go, the Vishay IRFxxxx MOSFETs do well, and using a Karefonte 0.635mm thick AlO2 type I got from LCSC, 0.75C/W on a IRFZ44 with the same GC electronics heatsink compound.
Old Motorola AN-1040 has some good information, too. I hope this isn't too much rambling and somewhat helpful!