Thermal Transfer Compounds


One I've been looking at but not yet tried is Apiezon N

The link to their data sheet is

http://www.apiezon.com/greasetable.htm

and they advertise (in a seperate circular) that results for
thermal and vacuum tests are available from NASSA Ames and Max Planck
('Thermal boundry resistance of mechanical contacts between solids at
sub-ambient temperatures' - E.Gmel, M. Asen-Palmer, et al.)

The normal white transfer compound used for power transistors is
about 3w/mK and the tube claims is good down to -100C
I have used this for peltier coolers to -30C.

Apeizon vac grease is used in vacuum CCD cryostats but always manages to end
up on the front surface by capillary action.

Indium foil is good but only works when you can put a lot of pressure on the
junction - it is good under heat straps.
Underneath detector chips it just creates another junction making the
transfer worse.

The best tip is to be careful to thermally anchor the signal wires, it is
possible to get more cooling into the device through the contacts than the
package.
Rockwell use this on their large format MCT arrays, there are 60 signal pins
and 200 cooling pins connected to the substrates.


Both Oxford instruments and Lakeshore have good descriptions of thermal
epoxys and transfer compounds on their websites and catalogues.

                  Martin Beckett

Hi Phil-

I had to deal with this too many years ago for chilled radiometers in early thermal
scanners.  We are doing it again today.  I guess the snake oil is just as rich now
as then.  We were trying to wring the last Deg-K out of our prototypes.  We learned
(probably re-discovered for the 497th time...) that the contact interface is
essentially the mashing together of two mountainous landscapes.  It was critical
that the surfaces be of the identical figure (flat is simplest) so that uniform
pressure is maintained across this interface without smashing your detectors, etc.

We also found that the finer those 'mountains' were the better the thermal
conductivity.  Period, no question.  Many a beer passed over the question of why
more, smaller mountains were better than fewer, larger ones.  A lot like today's
debates over final density of a Cantor dust.  Just smooth your surfaces as finely
as possible.  Remember, grind with water, wash in hydrocarbon (and vice versa) to
avoid grit embedding.  If you don't trapped abrasive will act as a stand-off and
the conductivity will go DOWN.

Final surface grinding is the key in my experience.  Even if you are going to
solder the surfaces.  I like the 600 and 1200 grit diamond files and 'stones' by
DMT.  Machinists use them for lapping carbide, ceramic, etc.  You can get them from
Enco supply at www.use-enco.com for about 25 bucks.  They work with water, so I use
naphtha followed by MEK for final cleaning to remove ALL grit, finger oils, etc.
from my pristine surfaces.  Lap the TEC's, heat sinks, intermediate plates, cold
finger ends, and don't forget the back of the chip package, too.

It's critical that your thermal goo truly wet the surfaces such that it can help
bridge the insulating air gaps between 'valleys' in your interface surfaces.  As
suggested elsewhere, wet the surfaces with your goo and wring them together to
displace all air.  The air has far worse thermal conductivity than the direct
contact between the mountains of your surfaces or the goo in the valleys.  We
found  it important to never reduce the wringing force.  If you do it wrong,
releasing the force will allow the joint to release and suck air.  Then you start
over.

Right now we load the joints with goo, pre-fit nylon bolts between the plastic cold
finger retainer saddle and the hot sink, start the parts together with a rocking
motion to wedge out the air, then tighten slooowly and evenly while heating the
whole stack with a hot air gun.  The detector package is similarly treated with its
retaining clips and nylon hold down screws.  Never re-use the ejected and scraped
off goo.  It somehow gets loaded with dust and other crud and it will now act as an
insulating spacer.  Murphy.

Today we can maintain about a 75 Deg-C difference with two TEC's in series (yeah, I
did it myself.  Next time, Melchor does it!).  The cool end is an aluminum finger
cooling a little TI TC-237 CCD.  I needed to insulate the devil out of my cool box
and my cold finger against  both radiative and convective heating.  Also, I use dry
air to pre-infiltrate my test camera chamber.  The chamber is aluminum insulated in
polyamide epoxy resin really, really filled with glass micro-balloons (model
aircraft hobby supplies).  Next time, we'll use PVC to get away from the aluminum.
A coat of the same epoxy mix  is on the cold finger up to the chip mounting
surface.

Plain old Zinc Oxide grease has worked well.  I tried silver loaded silicone
greases 40 years ago and we had a problem with (we thought) oxides and sulfides of
silver corrupting the soup and increasing insulating stand-off between the
surfaces.  I would like to hear any modern experience.  Right now we're testing a
sample of Bergquist's TC-7500.  It seems better; in my prototype camera I can hit
75 to 79 Deg-C DeltaT with ease versus 68 to (stretch) 73 C with Radio Shack zinc
grease.

We are using water cooling for the hot side sink from a room ambient of about 20
C.  The chip package at the cold finger interface is indicating about -47 C with
the hot side of the TEC stack reading +30 C.  Thanks Tom Droege for the hint to
thin the hot sink behind the TEC for water cooling.  Will do that tomorrow.
Probably mill a new one with good support and turbulent flow against the hot wall.

It has been a real long time since I fooled with TEC coolers and their relative
inefficiencies.  Analytical solution to the problem of extracting milliwatts of
heat from a CCD by pumping with tens of watts in the TEC stack is subtle.  The
folks at Melchor.com have ton of references that make fascinating reading.  They
also have a nifty - and free - tool at their site called AzTEC for calculating
power/ efficiency/heat flux/  Delta T in their TEC's.  They offer prebuilt TEC
stacks cheaply, too.  We will use them for our product, believe me!!

Jeff Thompson