A potential problem with soldering a transistor down is ensuring
there aren't any voids. Where transistors have a carrier and
flange soldered together it's normal to do it in a reflow oven
with an inert atmosphere, using preformed inserts.
Both Freescale and NXP discuss mounting the latest high power
devices in application notes. For example, see NXP #10800 para. 2.5.
It's interesting to note the difference between the regular white
heatsink goo we're all used to and some of the new stuff prompted
by CPU cooling demands. At 500W dissipation in one of the kW
devices there could be several 10s degrees (C) improvement.
I'm hoping to find time to experiment with these transistors - my
plan is to lap the transistor onto the spreader with ultra fine
grinding paste or jewellers rouge and then use Arctic Silver 5 or
Shim Etsu 7762 thermal grease. Both are readily available for
about 2% of the transistor cost.
> Your thought of soldering also needs to take the differential thermal
> expansion to be taken into account - solders vary widely in their ability to
> take repetitive thermal cycles .
> I would think the device manufacturer would supply you with the materials
> used in the package and their thermal expansion properties as well as
> recommended mounting techniques. It is certainly to their benefit to do so.
Amps mailing list