Surely there is the risk that conductive paste of any type might creep from the device heatsink to the fixing screw which would result in a short to the amp main heatsink?
Ian
A specific CFP configuration that I am considering, actually has the collectors of all of the devices directly connected to the speaker output rail. So the idea is, to screw all the relevant transistors (driver and power transistors) to a copper thermal bus, which at the same time serves the purpose of a speaker output rail. Indeed, the whole rail would then be carrying the speaker output signal.
It would be this whole (big surface) copper rail that is actually screwed on to the heat sink proper, by means of a silicon galvanic insulation pad. Such silicon pads can be purchased as whole "sheets", from which I could cut out the necessary 38 cm x 5 cm insulation pad.
As the transistors are basically "screwed on" directly to the "speaker terminal rail" - there is no need for their galvanic separation from this rail. That is the reason why I am considering how to increase K/W, with no worries about electrical insulation.
Actually, I also was thinking about those thermal pastes that are available for mounting heat-sinks on high performance CPU's. But these "hyper-performance" pastes are somewhat controversial in terms of the price per 1 gram being offered, which annoys me.
Below - a concept of such circlotron output stage, with CFP devices that are mounted electrically direct to the speaker output rails. Not sure about the stability (as in: no oscillations) of such a CFP stage, though.
As for the thermal stability - a non issue. Overall positive thermal coefficient. I hope.