On Sat, 12 Dec 2015 15:51:15 -0500, Phil Hobbs
<
pcdhSpamM...@electrooptical.net> wrote:
>The problem with grease is that it tends to pump itself out from under
>the hot spots under temperature cycling. If one could get the diamond
>particles to embed themselves in both sides of the metal, they'd stay
>put and probably improve the theta.
In a past life, I was responsible for getting the heat out of RF power
transistors. I posted my results, which didn't quite agree with the
internet published orthodoxy, about 2 years ago in this newsgroup.
Now, I can't find it. Argh.
Very roughly, I got best results when I polished both the aluminum
heat sink and the base of the power xsistor. The idea was to get as
much metal to metal contact as possible. What worked best was no
grease, no powders, and no mixes of grease and powders. Just metal to
metal. However, that requires that both the heat sink and xsistor be
flat and parallel, not twisted, bent, or torqued. Measuring the
flatness of the heat sink was fairly easy. I can't say the same for
the xsistor. So, I just tested for stiction against a known flat
gauge block. Good enough.
During the testing, I tried various powdered metals without any greasy
binder. I didn't have a wide selection of powders available, but I
managed to try silicon carbide, graphite, metallic silver, alumina,
and copper. I would say that powdered silver worked about the same as
Arctic Silver heat sink goo. The best of the bunch was silver.
W/m*K
Diamond 1000
h-BN 600
c-BN 740
Silver 406
Copper 385
Gold 314
Aluminum 205
Graphite 200
Carbon 150
SiC 120
Brass 109
The idea behind the powder was to force the powdered metal particles
into the aluminum heat sink and copper xsistor base, filling in the
cracks and crevasses in the metal. In order to make that work, I had
to clean/degrease the parts, smear on the powder, wipe off the excess,
attach the transistor, and compress the mounting flange. If I had
left a large quantity of powder on the heat sink, it would make things
much worse. Inspecting the surface with a microscope showed that most
(not all) of the particles were imbedded in the aluminum.
However, the copper xsistor base was another story. Most of the
particles would just stick to the surface and not fill the cracks. I
could beat on the xsistor to force the particles into the xsistor
base, but that would cause the base to loose its flat surface. I
could have left the gold plating on the base, but then it wouldn't be
flat.
One thing I didn't try was gold leaf. Working with power transistor
size pieces of gold leaf is difficult. It sticks to literally
anything and will not let go. It also likes to wrinkle creating
lumps. However, with a fairly high thermal conductivity and the
ability to fill in cracks, it's still tempted to fight the problems.
The bottom line is that powder only made sense if both metals were
soft, polished, flat and parallel. Controlling the amount of powder
was difficult because the amount needed depended on the type and depth
of the surface features, cracks and roughness.
In the end, we used a polished heat sink and xistor for the really
high power stuff, and conventional heat sink goo for most everything
else. It was possible to get an improvement by using some of the
aforementioned methods, but it just wasn't worth the time and effort.