Quote:
Originally posted by redleader
http://www-esg.lbl.gov/esg/personnel...nsoldering.pdf
Take a look at the indium-bismuth solder. The eutatic point occurs at 78C for a ~79% indium solder. Meaning we could use it to solder a waterblock directly to the CPU die without cooking the CPU (which is typically rated to at least 90C). It also lists a few more exotic solders with even lower melting points.
The pdf mentions typical difficulties in soldering silicon, however it also proposes a solution. Add a thin metallic coating to the silicon die and then solder that.
|
How about
eutectic indium-bismuth-tin with a melting point of 60 C ?
Did some searching around, found some data about Field's metal
here. It's listed as alloy #19 (it's also in the PDF above) Doesn't include thermal conductivity for some reason, but I used the electrical conductivity to calculate a thermal conductivity of 14.4 W/mK. However, this calculation can be wildly innacurate, if anyone remembers the old copper-silver alloy myth that used to float around. It's the right order of magnitude compared to the other alloys listed.
As for the problem soldering to silicon, I think that's a non-issue. That's a problem with
bonding to silicon, meaning the soldered joint has to sustain a load. We've got clips or bolts to do that. In fact, its an advantage because your HS/block will be removable without desoldering.
Field's metal appears to be the "low melting alloy" used in
this TIM. Someone managed to get a sample and describes it
[H]ere, although I'm pretty sure he's got the alloy contents wrong in his post. Thermagon gives the thermal conductivity as 18 W/mK, but I think that would be across the entire TIM, not just the Field's metal.
I Think the easiest way to mount it would be as a foil. like the Thermagon TIM except no extra layer of copper. Burn it in at 65 C and it solders itself. The main problem I see would be getting an consistent thickness in the foil if your home-brewing it.