Mirror-polish heatsink contact surface?

[hank]

This is beyond me at the moment, but thought it worth pointing to here.
Comment is from a computer modders' site, suggesting putting a mirror polish on the contact surface of a heatsink for maximum heat transfer.

Makes sense to me.

Link:
http://www.overclockersclub.com/guides/heatsinklappingguide.php

 

[greenLED]

I've been lapping the surface of HS's for a while - after tvodrd's recommendation.

 

[will]

Lapping the surface will make it flat, eliminating bumps, which in turn will give a better contact. Is there any heat paste involved? also - is the component mirror polished as well?

 

[Sawtooth]

Lapping the surface will make it flat, eliminating bumps, which in turn will give a better contact. Is there any heat paste involved? also - is the component mirror polished as well?

Perhaps using the component itself as the "lapping tool" would make for the best contact?

 

[Sable]

Lapping is a time-honoured way to improve thermal transfer on CPU heat sinks. Generally only the sink itself is lapped, and there is the tiniest possible amount of thermal grease used. It's only there to fill in the microscopic gaps where there is no metal-metal contact.

ETA: DO NOT LAP THE CPU ITSELF, if you're doing this in a computer. Bad idea...

 

[tvodrd]

I've lapped my LS slugs for years with 2000 grit.

Larry

 

[Anglepoise]

When mounting bare emitters, I lap both surfaces with 1200 wet and dry on a flat plate. Glass or Granite. Cleaned with alcohol , this gives a good 'key' for the AA to set up and really hold with hopefully maximum heat transmission.. So far I have 100% success with isolation of the positive base on the P4 with just this one layer....but still wonder if one layer is satisfactory.

 

[65535]

After lots of talk about lapping on bit-tech.net I have decided that due to the characterstics of heat transfer a 800 grit lapping using a glass plate will allow both good thermal goop adhesion, and it will allow a good amount of surface area with minimal gap for the heat to trsnfer through. The best heatsink is actually one with pyramids on the surface that radiate heat at angles off of the piece, rather than a fin that bounces heat off of the other fins, the more surface area the more energy that can be disipated.

 

[OddOne]

It does help with thermal transfer, BTW...

Now THAT is a mirror finish.

From 400 grit all the way to Brasso, on 1.25" diameter solid copper bar. It was rough on the ends from bandsaw cutting when I started.


Side note: I used to be a contributing author to Overclockers.com and helped run an OCing site for a while. (Also been interviewed by cNet about overclocking.) I've lapped MANY a heatsink in my time...

oO

 

[Mel_PL]

It does help with thermal transfer, BTW...
<snip>

Good polishing job...
From my physics lessons over 30 years ago (I hope the basics have not changed since then) I remember that black surfaces dissipate heat better than mirrors. Now, if you want to rely only on intermolecular interaction between the surfaces then mirror polish makes sense.
But when I compare the workload and the net thermal effect ... well, IMVHO simple silicon paste and rough surface is more than enough. But it does not shine so marvellous

-- M.

 

[SKYWLKR]

I lap all my heat sinks AND my CPU's and LED dies.

I get them to a point to where the humidity in the air is enough stick them together with surface tension then I apply the thinest layer of Artic silver I can to both parts.

I end with 2000 and use a 1/2 glass pane

 

[TOOCOOL]

Do not use Brasso or any metal polish on a heatsink that will use thermal paste Artic silver etc the polish will go into the microscopic holes where the Artic silver is supposed to be I belive you can check this at AS web site

 

[Anglepoise]

All very interesting. I have never considered using a mirror finish ,due to perceived problems with adhesion. I have always thought that one needs two flat surfaces with a little 'key' to give something for the AA to really grip to.

Thus my choice of 1200 wet and dry.

In my tests to see how much AA I needed to stick down one emitter, I used clear Plexiglas so I could measure the amount of spread at different pressures used while curing.

The thin wafer of cured AA could be easily separated from the ultra smooth Plexiglas with an Exacto knife.The wafer popped right off. Just an observation.

 

[OddOne]

Good polishing job...
From my physics lessons over 30 years ago (I hope the basics have not changed since then) I remember that black surfaces dissipate heat better than mirrors. Now, if you want to rely only on intermolecular interaction between the surfaces then mirror polish makes sense.
But when I compare the workload and the net thermal effect ... well, IMVHO simple silicon paste and rough surface is more than enough. But it does not shine so marvellous

-- M.

Actually, color isn't the issue so much as what the surface consists of in terms of oxidation byproducts (Aluminum Oxide is a great conductor, Copper Oxide isn't as much) and surface texture (smoother is better).

There's been a LOT of renewed interest in thermal junctions over the last ten or so years thanks to increasingly tiny processors generating much higher heat loads than has been the issue in traditional power semiconductors. The findings were that a mirror finish with a TINY amount of thermal goop to make up the difference was a better junction than HA aluminum (which is far rougher) with the same type of goop.

I lap all my heat sinks AND my CPU's and LED dies.

I get them to a point to where the humidity in the air is enough stick them together with surface tension then I apply the thinest layer of Artic silver I can to both parts.

I end with 2000 and use a 1/2 glass pane

That's the ideal way to do it. I bet your processor temps dropped a good 3-5 C once you got your lapping procedures down.

Do not use Brasso or any metal polish on a heatsink that will use thermal paste Artic silver etc the polish will go into the microscopic holes where the Artic silver is supposed to be I belive you can check this at AS web site

I didn't mention it but I always clean contact surfaces with automotive brake cleaner on a lintfree rag before applying thermal compound, so as to remove the Brasso film, skin oils, etc. from the contact surfaces. Skin oil alone can adversely affect junction resistances when you're being this precise. Why brake cleaner? It's relatively plastic-safe, leaves no residue, and cuts most contaminants (including old thermal compounds). Using a lintfree cloth as the wiper removes the risks of spraying something into a comp case, etc. (And no, you don't need to breathe the stuff - ventillation and intelligent use required!)

Like I said, this isn't my first time down this road.

oO

 

[IsaacHayes]

If you can get 2 surfaces super flat, they will stick to each other like a suction cup. With good pressure this is the best thermal contact as it maximizes surface area. Now some thermal compound can fill the microscopic holes, but you need an extremely thin layer. If you had the 2 materials sticking bare, I'd probably only rub in some thermal grease and not even do a thin layer. If you do choose to put a thin layer, I like to put tons of pressure and sometimes twist back and forth until it oozes out and the parts bite.

Sometimes this doesn't always apply to LEDs (can't press super hard), but the smoother the surfaces and less thermal compound the better.

The thermal interface of the first heatsink to the LED slug is the most critical one. It is closest to the heat, and has the least surface area and mass (the slug).

 

[OddOne]

Another thing to watch for when lapping is to avoid creating a curve, which is why glass plate (which is generally flat to within a few thousandths of an inch) is a popular surface for underneath the sandpaper.

A meniscus (slight curve only at the edges) is not generally a problem though, so long as the contact area with the thermal source itself is maximally flat.

oO

 

[IsaacHayes]

Speaking of concave, those that have CPU's with heatspreaders, sometimes lap those. I've found MANY are VERY concave and the edges stick way up higher than the center.... That can be worse than a rough surface if you have a big gap filled with thermal compound in the center.

EDIT: then again maybe they do that on purpose to releive stress on the cpu die right under neath? Not sure how much it flexes when clamped to a HS.

 

[hank]

> black surfaces dissipate heat

Yep, the radiator fins need to be black -- that's radiation, more efficient; the 'mirror polish' is for close physical contact (conduction).

I expect a small gap between mirror surfaces would be, well, a Thermos Bottle insulator (grin)

 

[tebore]

The real reason why CPU heatsinks needed to be lapped back in the day was there were REALLY REALLY cheap aluminum cut outs that were not flat at all (we're talking convex to concave bottoms) which is why most old P3 Flipchips had thick pink thermal paste.

Fast forward to when overclocking starts going main stream. We get nice flat copper sinks (Thanks Thermalright) but they were rough. Which is why using AS3 could drop 10degrees just by bridging the gap. People noticed this and started to lap until it was a reasonablly smooth surface. These sinks were so good that after lapping you might get max 5degrees off with the CPUs were loaded to 100%. The difference between a reasonably flat surface and mirror shine is very small, sometimes negated completely just by using AS5. Stoping at 800grit might give you 0.5degree difference compared to say power buffing with a rubbing compound. Some say stay away from the actual polishing because it will work against AS5 and actually give you higher temps. I haven't got to that extreme yet so I'm not sure. But I do know that the difference between an 800grit finish and a 1200grit finish is almost nothing.