Soldering Different Metals (Al to Cu), Face to Face

APynckel · Jan 16, 2017

So I am putting together a small heatsink for a hobby (active cooling a raspberry pi 3) and I want to solder a copper base to an aluminum heatsink to help aid in pulling heat away from the processor. I have already gone and lapped down the anodize from the base of the aluminum and ensured a clean flat contact layer (took it up to 2000 grit on a flat granite countertop). I want to solder the Cu base to the Al fins to ensure as good a heat transfer as possible and to make one solid item.

A) What kind of solder should I use?
B) What's the proper method of soldering two faces together (30mm x 30mm Surface Area)?

Is there anything else I should be concerned with? Should I be considering a different method of joining the two metals?

Thank you in advance!

ewlsey · Jan 16, 2017

As far as I know it's not possible to solder aluminum.

What's wrong with the heat sink paste and some screws?

APynckel · Jan 16, 2017

Hmm.

It's too small for fasteners (and there are no points on the board near the CPU to mount anything), but I could probably use thermal paste. Just give it some time to "set" and hope I don't get too much shear across the interface during install.

PeteM · Jan 16, 2017

I've got some solder for aluminum, but it uses a corrosive flux and requires fairly high heat. Even if you managed to solder things without damage -- and they didn't get galvanic corrosion sooner than later -- the combo would probably pass less heat than a proper silver-based thermal paste.

Probably some way to hold it down? I'm pretty sure there are also thermal glues (filled epoxy); though I haven't personally used any.

Monarchist · Jan 16, 2017

ewlsey said:
As far as I know it's not possible to solder aluminum.

Radio and Audiophiles have been doing it since Coolidge was Prez. Ordinary 60/40 rosin core, too.

Razor-blade the oxide away, apply a drop of high-temp-capable oil, scratch-tin the Al through the oil drop, apply copper wire or tinned tab, reheat. Done.

That said, fool's errand for THIS project.

What's wrong with the heat sink paste and some screws?

Nothing that CPU will notice.

Still - far easier to JF use a solid-copper heat sink from the outset. Or Silver (tiny CPU). Or a heat-pipe.

Bill

Monarchist · Jan 16, 2017

PeteM said:
I'm pretty sure there are also thermal glues (filled epoxy); though I haven't personally used any.

I have. 50+ years and counting. Go-to WAS Wakefield Semi's "Delta" and "Deltacast" series. Filled epoxies from lots of folks since.

Bill

9100 · Jan 16, 2017

It is possible to solder to some aluminum alloys. One technique is simply to flow a puddle of hot solder on the part and scratch through it with a wire brush while the solder is molten. Ordinary lead tin solder will adhere to aluminum when the surface is clean and not had time to oxidize.

Manufacturers of flashlight bulbs solder to aluminum bases all the time.

Bill

Monarchist · Jan 16, 2017

APynckel said:
Hmm.

It's too small for fasteners (and there are no points on the board near the CPU to mount anything), but I could probably use thermal paste. Just give it some time to "set" and hope I don't get too much shear across the interface during install.

Ordinary thermal paste dasn't much 'set'. Until you do not want it to do. Look for Silver-doped thermal ADHESIVE in a tiny push-ram tube. As with Brylcreem, 'A little dab'l do yah'.

Meanwhile.. how easy for you to cut-down a solid-copper HS to fit if one fell out of the sky? IIRC they were for fanless C3 CPU before we switched to fanless C6. Copper isn't a lot of fun to cut or file, but ..patience..

Aluminium SUCKS for HS BTW. Only virtue vs Copper is massive cheaperness. Silver blows 'em both away, and then there are heat-pipes. DIY'able, too, those are. If you want to badly enough.

Bill

L Vanice · Jan 16, 2017

Aluminum can be soldered with pure tin, but it requires cleaning the hot aluminum surface with a clean stainless steel brush while melting the tin rod against the aluminum. The aluminum does not melt at first, but, once the tin joins to the aluminum, some of the aluminum makes a solution with the tin and the solution of tin and aluminum will melt. You have to use care to keep the tin layer thin. No flux is used and the heating is done with a propane or air-acetylene torch. The tin will not flow into a gap, so it is a different process from soldering copper to copper. Once the aluminum has a layer of tin melted into it, the copper can be soldered to the tin with ordinary electrical solder and flux.

I bought my tin rods decades ago and have not run out yet. The labeling ("Mad Dog Weld") does not say it is tin. I had some analyzed in a lab, and they determined it is pure tin. I found a product on Amazon that looks like the same stuff I have, but is, of course a different brand.
Amazon.com: BAC Industries AP-21 Alumi-Pro Welding Kit: Home Improvement

Larry

APynckel · Jan 16, 2017

Monarchist said:
Radio and Audiophiles have been doing it since Coolidge was Prez. Ordinary 60/40 rosin core, too.

Razor-blade the oxide away, apply a drop of high-temp-capable oil, scratch-tin the Al through the oil drop, apply copper wire or tinned tab, reheat. Done.

That said, fool's errand for THIS project.

Nothing that CPU will notice.

Still - far easier to JF use a solid-copper heat sink from the outset. Or Silver (tiny CPU). Or a heat-pipe.

Bill

It's a tiny CPU, seems like .500"x.500" or smaller. I cannot find exact specs on its geometry. Will measure tomorrow when it shows up.

Monarchist said:
Ordinary thermal paste dasn't much 'set'. Until you do not want it to do.

How easy for you to cut-down a solid-copper HS to fit if one fell out of the sky? It isn't a lot of fun to work with.

Bill

I just had to order 2 pieces of .250" copper plate (1.125in x 1.125in). I'm doing this completely as a hobbiest, as I lost my job in November.

What makes it not fun to work with? I wouldn't mind playing with it. I just don't have the material. If I were to do the entire thing from scratch, I'd like 30mm x 30mm x 30mm cube.

Monarchist · Jan 16, 2017

APynckel said:
It's a tiny CPU, seems like .500"x.500" or smaller. I cannot find exact specs on its geometry. Will measure tomorrow when it shows up.

Don't make a problem out of a solution. Chip dasn't NEED much.

1/2" square CPU, you use nothing more complex than a bit of sheet copper 'flashing', ELSE a bespoke 'winged' HS of the sort once used for support IC's.

PM me a ship-to so I can clear some wasted storage space.

On edit.. no that would be further wasting yer time:

hardware - Does the Raspberry Pi need a cooling system? - Raspberry Pi Stack Exchange

Cease "diving into the Pi", because such things are a bottonless pit when you are down. DAMHIKT.

Go find a replacement for that job you just lost instead. Then you can afford a ready-to-use three-dollah custom copper HS in a nice box.

Bill

JST · Jan 16, 2017

ewlsey said:
....

What's wrong with the heat sink paste and some screws?

What's wrong with it is that it is not very good....! At least not the silver stuff, or any of the thick paste varieties. Silver sounds great, but really is not. Thick paste is bad.

The best stuff I have ever found is this stuff from Aavid. It actually comes from a division of Henckel. It is used for high power IGBTs and similar applications, and works extremely well. I have NEVER seen a failure due to problems with the material, or the application. The key with it is that it will always be the very thinnest coating possible, with any mounting that involves some spring-type compliance. It will fill all the voids and give the best coverage and the thinnest film.

Phase Change Materials - heat sink interface material

APynckel · Jan 16, 2017

Monarchist said:
Don't make a problem out of a solution. Chip dasn't NEED much.

1/2" square CPU, you use nothing more complex than a bit of sheet copper 'flashing', ELSE a bespoke 'winged' HS of the sort once used for support IC's.

PM me a ship-to so I can clear some wasted storage space.

Bill

It still pumps out a LOT of heat. Gets close to 85*C during normal operation. I've pulled a 28mm x 28mm anodized aluminum heatsink from am "engineering sample" heatsink box that I've since lapped down flat, and found a 30mm x 30mm 5v fan that I was going to mount to it.

Monarchist · Jan 16, 2017

APynckel said:
It still pumps out a LOT of heat. Get's close to 85*C during normal operation. I've pulled a 28mm x 28mm anodized aluminum heatsink from am "engineering sample" heatsink box that I've since lapped down flat, and found a 30mm x 30mm 5v fan that I was going to mount to it.

F'cks sake. Ton of them out there at anywhere from $1.17 to under $3.00.

If you treat the hours of your life as 'worthless', they have a b****y strong self-reinforcing tendency to become so.

Bill

APynckel · Jan 16, 2017

Monarchist said:
F'cks sake. Ton of them out there at anywhere from $1.17 to under $3.00.

If you treat the hours of your life as 'worthless', they have a b****y strong self-reinforcing tendency to become so.

Bill

Not quite sure how to take that last comment....

I'm just doing what I can to make this thing ready to take an overclock for what my intents are with the device.

Trust me, every hour of every day I'm beating myself up trying to find work.... I NEED an escape.

Monarchist · Jan 16, 2017

APynckel said:
I'm just doing what I can to make this thing ready to take an overclock for what my intents are with the device.

Trust me, every hour of every day I'm beating myself up trying to find work.... I NEED an escape.

Isopropanol cooling took my first iAPX386-16 Mhz "Double Sigma" to 50 Mhz. The third-party so-called faster-than-Intel (it was..) 387 could only manage 40 Mhz though, so I backed it down.

Then realized greater gain for less risk by adopting faster memory as larger cache for the primitive HDD of the era.

Problem with chasing this 'overclocking' stuff is that there are by this late stage in the game mebbe 100,000 folks working on ARMishness and close competitors at any given hour of the day.

Those who MATTER are DROPPING clock rates in favor of adding more cores and the ability to park those not needed. Thermal and power thing - "batteries included", that is.

Some of them have mega-R&D budgets and mainframe-class - if not 'supercomputing' - simulations going as they help define what goes into the next three generations of CPU, too.

So "Escape" yah.

But don't let it suck you in so deeply you are hand-carving alternatives to heat sinks 'good enough' at a buck or two, ready-made. Best case, you can barely measure any gain.

Bill

APynckel · Jan 16, 2017

Not hand carving, just exercising the engineering degree. Proving something to myself, more or less. I've not been able to find any heatsink / fan combo that is compact enough to run on this thing, so I'm just messing about to get the core temps down. I just like to tinker.

Monarchist · Jan 16, 2017

APynckel said:
Not hand carving, just exercising the engineering degree. Proving something to myself, more or less. I've not been able to find any heatsink / fan combo that is compact enough to run on this thing, so I'm just messing about to get the core temps down. I just like to tinker.

Tinsnips degree for that CPU. Or even 'good grips' Oxo kitchen scissors.

Copper flashing.

Cut a square triple the size of the CPU package.

Cut slices about 1/8" wide outer edge to scribed package size boundary.

Give each several twists, bend up in a curve. Fan can be 'elsewhere' in/on the device - no need of it being CPU-only.

Turbulence is crucial to good transfer, metal to air. 'Pin' type heat sinks and wavy thin fin ones have it. So will this 'dreadlocks' gadget.

Thick straight rib heat sinks fare rather poorly at transferring heat. They make up for their lack of thermal utility by being impressively larger and anodized purty colours for an eye-candy effect.

Do yer research and see.

Bill

APynckel · Jan 16, 2017

So this is what I had found (just to inform where my brain was heading). Small 28mm x 28mm hard anodized Al heatsinks in my engineering "sample box" that I was gonna put to work. A 30mm x 30mm 5v fan from Frys that the fasteners just happen to form a very nice interference fit into the sink with. I can bend some of the sink fins to create turbulent flow.

I had already lapped the sinks on progressive grit sandpaper to ensure a nice flat uniform mating surface.

Monarchist · Jan 17, 2017

APynckel said:
So this is what I had found

Apply the 80/20 rule. You are chasing diminishing returns.

20 percent of the complexity will secure you 80 percent of the gain and there'll be no CPU fan to fail.

A better fan, remoted, can outlive that fragile POS by ten years and cool as well - even BETTER.

Bill

Soldering Different Metals (Al to Cu), Face to Face

Plastic

Diamond

Plastic

Diamond

Diamond

Diamond

Diamond

Diamond

Diamond

Plastic

Diamond

Diamond

Plastic

Diamond

Plastic

Diamond

Plastic

Diamond

Plastic

Diamond

Similar threads