Heat sink paste

Computer CPU analogies to LED flashlight applications are tricky. The contact areas can be vastly different, and thus make apples-apples comparisons difficult. If you have a large contact area, then thermal conductivity differences for various pastes/epoxies have that much less impact.
 
. . . On my computer I lapped both the CPU and heat sink and saw about a 15 degree reduction in temperature. . . . .

I didn't think it would make that much difference, considering how much money you spend on a CPU !

What process do you use for lapping ? I'm considering now finishing with a polishing paste, rather than just the 1200 grit paper I've been using.

When I lapped the back of my Luminus SSR-90 star, I was surprised how curved it was. I kept on lapping with 500 grit until I got 3/4 of the surface flat.

I'm getting motivated to do a controlled test of a 20mm aluminium star dissipating a controlled 10 watts onto a copper block and testing the effects of -
- no compound
- toothpaste
- white compound
- silver compound
- silver epoxy

Initial test with light pressure, then repeat for full pressure

Then lapping the two surfaces and repeating the tests
 
I didn't think it would make that much difference, considering how much money you spend on a CPU !

What process do you use for lapping ? I'm considering now finishing with a polishing paste, rather than just the 1200 grit paper I've been using.

When I lapped the back of my Luminus SSR-90 star, I was surprised how curved it was. I kept on lapping with 500 grit until I got 3/4 of the surface flat.

I'm getting motivated to do a controlled test of a 20mm aluminium star dissipating a controlled 10 watts onto a copper block and testing the effects of -
- no compound
- toothpaste
- white compound
- silver compound
- silver epoxy

Initial test with light pressure, then repeat for full pressure

Then lapping the two surfaces and repeating the tests

That would be awesome data to have!
:thumbsup:
 
My lapping procedure used 4 different grits. 400, 1000, 1500 then 2000. The sandpaper was taped to a thick piece of glass, about 1/8" thick, WD-40 was the lubricant. You do 10 back and forth strokes turn it 90 degrees and then 10 more strokes when the oil becomes clogged with metal I wipe it off and then reapply more WD-40
 
I'd also like to see that data.

This would probably add a fair amount to your workload, but I wouldn't mind seeing if there's a difference between cheap stars from DX, compared to "name" brands, e.g. the stuff Cutter or LuxeonStar stock. Am surprised that the star from Luminus was that curved, now I'm wondering if it's a general issue or not.
 
When you tighten the screws onto the heatsink, it will tend to pull the star flat - within limits.

But as the bottom moves, the top of the star will also move, putting stress on the LED or the insulator between the metal core and the LED !
 
I've been air, liquid, phase-change cooling and OC'ing for a very long time and what I've always done for a more secure bond for surfaces that I didn't want perm-bonded was; I mix a smaller ratio of 5min epoxy with with Arctic-Silver5 or with ICDiamond-7. This enables you to make good solid contact with the intended heat-sink, while not bonding it so rigidly that it will not separate with without applying an above avg force.

It depends on the 2 surfaces intended to being bonded; obviously the less epoxy-to-paste ratio increases C/W values and efficiency. Or for a much more flexible, yet still efficient temp conductance; use polyurethane glue to surround the object. PU-glue expands many times beyond it's original state, this expansion is gas-based and produces micro-size bubbles which will NOT conduct very well at all, so mixing it like epoxy isn't advised, but would still work better than just glue. Or you could use Goop with the same methodology as PU-glue.
_Overall I've found that general epoxy works great, if gluing plastics to metals, I like Devcon™ Plastic weld, since it actually dissolves into the plastic's surface for a much more secure bonding-layer etc...

I have tried using super-glue also, but I've found that the mix doesn't work all that well bc it dries so fast. Also, S-glue hardens so rigidly that it's not at all suited for shock/hard impacts; it will eventually develop cracks and flake off w/o warning and catastrophically fail. It work fast thought, but doesn't at all survive cold+shock.

JB-Weld and thermal paste works pretty well and has the strongest bond that I've found with non-industrial based resin and hardeners. Once you mix and bond/coat w/ JBW = it's near PERM and will not come off w/o using sharp prying and or grinding tools!

Thermal-tape is very convenient, semi-perm, but not at all comparable to properly applied and secured thermal-paste... Then again this is more for extremes where a few degrees separate functional operation and glitching or failure.

The prob with general epoxies, PU-glue, S-glue and lapped surfaces is adhesion is very low the more polished the surface. Since at the micro level the glue/resins cannot sink further into the valleys and pits in the metal to make an anchoring, velcro type bond. They will peal with increased heat or cold. I only polish a very tiny area where the 2 objects will bond, and roughen the surroundings. Any extra thermal paste is merely a waste and does not aid in conduction.

Seesh that's some nerdage! lol. :thumbsup:
 
Top