Author Topic: Cu Heat Spreader to Al Heat Sink - Idea for Lowest Possible Thermal Resistance  (Read 366 times)

0 Members and 1 Guest are viewing this topic.

Offline PlasmateurTopic starter

  • Regular Contributor
  • *
  • Posts: 204
  • Country: us
I was wondering if anyone had tried to attempt soldering a copper heat spead to an aluminum heat sink in order to minimize thermal resistance for applications requiring heat sinks. Here's how I would approach this.


1.) First I would apply this method (https://youtu.be/Z14aU-PaioA) to plate copper to the top of the Aluminum Heat Sink
2.) Apply solder to one side of the heat spreader
3.) Apply solder to the plated copper on the aluminum heat sink
4.) Join the two in an oven.

Has anyone tried something like this?
« Last Edit: April 23, 2021, 05:23:01 am by Plasmateur »
 

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 21567
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Does this have to be done in the shop with meager equipment, or is an industrial solution acceptable?

And what kind of heatsink is this?  Because if you're niggling over 0.1 K/W resistances when your air-cooled heatsink is 10 K/W... come on, get outta here.  Or if it's water cooled, resistance of that magnitude does matter, definitely, but you'll probably be using an aluminum or copper plate with short fins/pins immersed in water (as the auto folks do it) -- again saving a thermal joint.

So, for the remaining cases --

0. If the spreader can be simply as big as necessary, do it! Cover the whole face of the heatsink for all that matters, and slip a gap pad in there.  Or grease or whatever.  Or nothing, just use the aluminum bare, if it doesn't need to be isolated.

This can be effective for power converters where relatively high dissipation is needed, but the devices don't have to be close together, so a fair amount of area can be dedicated to their heat spreaders.  You can push over 100W from a TO-247 this way, for example.  Again, the heatsink needs to be pretty damn big to keep the RthHA down enough to matter.

The catch is, if the heat spreader is electrically live, then it adds capacitance to the switching node (when applicable, i.e. low-side switching devices).  This needs to be accounted for.  Probably the capacitance by itself isn't a functional problem (~100pF versus Coss ~1nF?), so that leaves the common mode noise problem.  If the heatsink is grounded, then the circuit DC+/- bus is driven by full amplitude of the switching node through that capacitor; if the circuit is grounded and the heatsink is floating, the heatsink will need to be well shielded.  The two can be tied together, directly if the heatsink is floating, or with a Y1 capacitor if grounded, and all the other attention that needs to be paid to the related RLC equivalent network (damping, filtering).

1. Mechanical fit.  Mill out a pocket for the heat spreader, just slightly undersized (interference fit), then press in the copper slug.  Numerous CPU heatsinks have been made this way, so apparently it works; most of the thermal transfer is through the sides though, so a tall fit is preferred.

It's still not all that great, because, depending on surface finish, the gap between metals may be microns or less: short enough that, not only is the air filling that gap basically still (no convection at play, only diffusion -- it's like a monolayer of fiberglass insulation), the gap may be below the mean free path of the air molecules themselves, meaning it's even a vacuum for practical purposes(!).  Actual metal-metal contact points will be sparse, yet contribute most of the heat transfer.

Clearly the better solution requires some kind of filler in the gap.  I suppose these things could have a wide variance just from the factory, already, because of incidental oils on the parts before assembly -- cleanliness actually works against you here.  A liquid film is much denser (more conductive) than an air film, even if it's not loaded with conductive materials (like a proper heatsink grease is -- and what do you think is in the microscopic spaces between those particles?).  Those huge IBM CPU modules (see recent-ish Dave vids) handled this by filling the cavity with mineral oil, using a very close (but not tight, still sliding) fit between the copper slug and aluminum backplate.

With a heat spreader pressed into the heatsink, and backfilled with thermal grease (or loaded first, then allowed to squeeze it out during assembly), you get pretty much any ordinary facing joint with grease, plus the tight (hopefully oiled) sides.

Honestly I don't think it's worth it, bulk aluminum is merely half as conductive as copper, and copper is pretty damn conductive.  All these shenanigans leave me doubtful that it's going to do better than a solid aluminum block anyway.


2. Solder it directly.  Aluminum is easy to solder, though it takes a special flux or solder to do so.  You may want to involve a specialist for this.  (The fluxes contain free fluoride, so are very toxic.)  Solders include tin and zinc alloys.  The latter, can wet under the aluminum oxide layer, by itself -- they're often sold as "miracle rod" because they're supposed to be easy to use, just scratch the material against the surface to get initial adhesion; but as is usually the case with miracles, they tend to fall apart once the magic is over.  You can't fill a joint the way you wick solder into a fluxed joint, there's nothing to disturb the oxide layers deep inside the joint unless you make a point of "buttering" them all first, then assembling when hot.

So, with careful prep and inspection, you could prepare a "buttered" surface on both objects, then slide them together and hope you didn't trap air or oxides in the joint.  Probably not too awful, but fiddly, and open to a lot of variance.

2a. You can "butter" the surface with zinc, then tin it with actual tin (and regular rosin flux).  Probably want to mill or grind down the rough zinc surface first.  This is probably the easiest way to do it in a lightly equipped shop.

3. Who needs solder when you can weld?

Well, it's more of a brazed joint as the copper is concerned, but, let's see here.  You'll need a flux again as usual, possibly a milder one, but probably still fluoride based.  Filler can be aluminum alloy, silicon eutectic for example, or a few others.  Heating must be done carefully, so as not to melt the base metal -- a higher melting alloy must be used.

Downside, several Cu-Al intermetallics are quite hard, and others are quite brittle.  It'll stick together with great thermal conductivity, but if it ever cracks off -- and that's more likely the bigger it is -- well, yeah.

4. And, sure, plating I guess.  See, the problem with the linked method is, it's just...trash?  Copper doesn't stick to aluminum, it just sloughs off.  At least when done like that.  There's too much of a mess of oxides and trapped air in that joint, it'll perform as poor thermally as it does mechanically.

Good metallurgical bonds can be had, but they require different metals, and I think, typically toxic chemicals at that.  Nickel is a popular one.  Nickel also serves as a diffusion barrier, blocking dissolution of aluminum and formation of intermetallics (though NiAl itself is a rather brittle intermetallic, so don't let it cook forever, I guess).  Nickel is somewhat hard to tin, but substantially easier than aluminum, needless to say; a soldered joint would do fine here to finish it up.

So, I don't remember what steps are needed to make that effective, but they're probably not the kinds of things you'd want in a small shop, so I would suggest the other ways.

BTW, if isolation and low capacitance are required -- consider AlN plates.  Pricey, but they have all the thermal conductivity of copper, with none of the electrical!

Tim
Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!
 


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf