EEVblog Electronics Community Forum
Electronics => Manufacturing & Assembly => Topic started by: DerekG on November 30, 2017, 02:10:55 am
-
I'm hoping that someone may have some experience with soldering the heat sink of a MOSFET directly to a sheet of aluminium in an oven (either infrared or vapour phase)?
Can it be done? Is it reliable? What to watch out for? What flux to use? What solder paste to use?
I'm wanting to solder TO-220 / TO-3P / TO-247 through hole MOSFETs onto a 2 mm (or 3mm) thick aluminium plate. The MOSFETs would need to be clamped in place before soldering & the legs would hang over one edge of the plate so they can be soldered onto a regular PCB adjacent to the plate.
We require 4 MOSFETS to be soldered to the plate, so there would be 12 component legs to give some strength between the boards. Both boards would be mounted with 4 screws into our plastic enclosure.
The MOSFET heat sinks are all at the same potential, so we can safely connect them all together on the one aluminium plate.
This would be a production process with several hundred boards being run through the oven over the morning.
I presume the aluminium plate will require some preheating. The aluminium plate to be used is a standard commercial grade plate which is why I ask about special fluxes etc to clean it first (& get the solder to adhere to it).
This process would alleviate us having to manually screw all the MOSFETs to the aluminium plate & to do away with the manual process of applying heat transfer paste.
We will have the aluminium plate laser cut to shape to begin with & later on, punched as the volume grows.
Any thoughts appreciated. Especially from anyone who has actually tried it.
-
i'm no expert, but, wouldn't you'd have to heat the heatsink up so much, that by the time you got it hot enough to take the solder, it would fry the mosfet?
I guess you can heat the heatsink first, get it tinned with solder, wait for it to cool down, then quickly hit it with solder iron and see if you can get the mosfert soldered on without damaging it.
-
I have experience in trying. It is some kind of hard getting solder to stick to aluminum, at all.
I have heard it is possible by putting oil on the surface, scraping/grinding the surface away, then soldering it. All the while under a layer of oil to keep out any oxygen.
I have tried it. I didn't get it to work.
Seriously, try getting solder to stick to aluminum, at all. If you can do that, then you can move on to figuring out how to get the FET on there. Until then, I think you're dreaming that this is going to be cheaper than screws and paste. Take the cheapest products out of China and show me where they save money by soldering anything to aluminum! Their labor might be cheap, but they also have reflow ovens and solder and aluminum heatsinks. :-//
https://app.aws.org/wj/2004/02/046/
-
To solder to aluminium, coat the area in gel flux. Scrape the base material through the flux. Then begin soldering. With the thick oxide layer broken and the flux to further rip of the very thin layer that would have just formed you can get solid connection to the base metal.
I've done it this way for a number of old gauges with Ali housings. You get the solder to bite to the metal first, then come back and solder what you need by melting the blob you have attached.
-
Tips for Soldering Aluminum (https://app.aws.org/wj/2004/02/046/) might be of some use. That said, even if you manage it, I expect you'd have reliability issues.
You'd be better off using copper plate/sheet/bar IMHO, and 0.125"/3.175mm thick would work and is easy enough to source. You can even get it on eBay (example (https://www.ebay.com/itm/125-1-8-Copper-Sheet-Plate-2-x-5-/282062698821)).
-
Getting Aluminum to tin for a one-off is doable by abrading through a puddle of flux, solder or high temperature oil that excludes air. The surface must be freshly abraded then abraded through the puddle as even the thin layer of oxide that forms while heating will totally prevent success. It is in no way a production-ready process.
I'd also be concerned about the difficulty of cooling the heatsink fast enough not to be *way* outside the MOSFET's permissible soldering thermal profile, without thermally shocking it with a too rapid quench.
Consider using a specialist thermal adhesive. To get a reliable bond with thermal epoxies, the heatsink surface and device mounting surface will need to be freshly abraded and cleaned. Alternatively look at self-ashesive silpads, and mounting clips rather than screws.
-
...
Consider using a specialist thermal adhesive. To get a reliable bond with thermal epoxies, the heatsink surface and device mounting surface will need to be freshly abraded and cleaned. Alternatively look at self-ashesive silpads, and mounting clips rather than screws.
Aremco make some printable high-temperature adhesives:
http://graphitestore.com/Graphite/High-temp-coatings-and-adhesives/custitem_gs_applications/Screen-Printable (http://graphitestore.com/Graphite/High-temp-coatings-and-adhesives/custitem_gs_applications/Screen-Printable)
e.g. Aremco-Bond 556-HT-SP for chip bonding.
The OP may be able to print the adhesive onto the heatsinks in a semi-automated way. I would ask board assemblers about this: what methods and materials would best suit their processes.
-
I've never successfully soldered to aluminum. Be prepared for a lot of baked semi's.
Instead, I would use D2PAK MOSFETS on aluminum-core PC board. These are common PCBs in the automotive industry and for LED lighting.
Then mount that to your final heatsink.
-
This is definitely a vapour phase job (since it's an option for you). The fluorocarbon liquid makes an inert atmosphere that should protect the aluminium from oxidation during soldering.
-
This is definitely a vapour phase job (since it's an option for you). The fluorocarbon liquid makes an inert atmosphere that should protect the aluminium from oxidation during soldering.
Ooh, that's actually a good idea. It's probably not quite good enough, though: aluminum will getter any remaining oxygen, so the vapor needs to be very pure indeed, not diluted with air, say. Anyway, you still need the skull-and-crossbones flux to get through the initial oxide layer. Would be cool if the surface could be prepared inside an inert gas glovebox, then soldered. Hm.
Anyway, guessing that's a lot more setup than OP is looking to use here. :-\
Tim
-
It may be worth looking at getting the aluminum sheet copper or tin plated so its solderable.
-
I'm hoping that someone may have some experience with soldering the heat sink of a MOSFET directly to a sheet of aluminium in an oven (either infrared or vapour phase)?
Can it be done? Is it reliable? What to watch out for? What flux to use? What solder paste to use?
To echo what others have said...
No. Not even no, Hell No... :-DD
I'm wanting to solder TO-220 / TO-3P / TO-247 through hole MOSFETs onto a 2 mm (or 3mm) thick aluminium plate. The MOSFETs would need to be clamped in place before soldering & the legs would hang over one edge of the plate so they can be soldered onto a regular PCB adjacent to the plate.
0. Those aren't SMT packages, for a reason. (There are TO-263 and TO-268's for that!) Many of those come with tin plating which is just fine and solderable, but many also come with nickel plating that's a nightmare to bother with. I've SMT'd TO-220's here and there for little projects, I know. :horse:
1. Use copper. Copper is solderable, also more conductive so you need less of it, perhaps even to the extent that you need one or two fewer transistors.
2. Use a PCB. Heavy copper PCBs are available most anywhere. There's even the kind with machined (I guess) inner "layers" of considerable thickness. Though you may end up paying as much for the PCBs as for trying to solder aluminum...
3. Use more transistors. You're looking at five or ten bucks of transistors, versus sky's-the-limit inert gas, breathe-the-fumes-and-you-die soldering. Transistors are fucking cheap. I haven't seen a single design where the transistors were anywhere near a significant fraction of the total product cost. The labor of inserting them and screwing them down easily overwhelms their cost (except maybe in China, but then, locally sourced, classic type, transistors are quite cheap, too).
They're not vacuum tubes. Pile them on! ;D
3a. While you're at it, give up on all that annoying mounting hardware! Spring clips are fantastic: they don't bend the device, they spread the load evenly, and you can use much bigger devices (MAX247 and such) that aren't hampered by pesky trivia like screw holes.
There are even heatsink-and-clip duos that are cam-lock or clip-in-place. No screwdriver at all! ;D
We require 4 MOSFETS to be soldered to the plate, so there would be 12 component legs to give some strength between the boards. Both boards would be mounted with 4 screws into our plastic enclosure.
Plastic?
So, no heatsink?
God man, just stretch the PCB a bit and use D2PAKs on 2oz copper! You won't notice a difference!
Tim
-
I used to work for a company (well known one) involved in 'power electronics' and they developed a method of soldering the THD type packages directly to the heatsink of one particular product, which gave them good advantage in terms of the product's performance..
From memory it was around 3-5 mosfets per side (so 6 to 10 in total) on a long rectangular heat sink - the process was done on a modified for this purpose solder pot, all fets were soldered at the same time. Once the process was mastered there were no issues with damaging the mosfets while soldering. Having said this, if a mosfet was to fail in this module the whole assembly had to be thrown away as you could not rework it most of the time safely, or within reasonable time.
-
Altho, I am not 100% sure if the heatsink was from Aluminium (it looked like from memory)
-
This is definitely a vapour phase job (since it's an option for you). The fluorocarbon liquid makes an inert atmosphere that should protect the aluminium from oxidation during soldering.
Ooh, that's actually a good idea. It's probably not quite good enough, though: aluminum will getter any remaining oxygen, so the vapor needs to be very pure indeed, not diluted with air, say. Anyway, you still need the skull-and-crossbones flux to get through the initial oxide layer. Would be cool if the surface could be prepared inside an inert gas glovebox, then soldered. Hm.
Anyway, guessing that's a lot more setup than OP is looking to use here. :-\
Tim
Vacuum vapour phase is actually a thing. >:D
-
This process would alleviate us having to manually screw all the MOSFETs to the aluminium plate & to do away with the manual process of applying heat transfer paste.
Rivets instead of screws would be quicker and more secure.
Or SMD MOSFETS on an aluminium PCB.
Or clips.
Why mess about inventing a new process - the time taken will easily cover many thousands of rivet/screw operations.
-
I'd solder MOSFETs on copper (stainless steel with flux) pads with holes for screws and afterwards screw these onto aluminium heatsink. May be adding thermal paste between pads and heatsink. It will be much easier and more confident than trying to stick it to aluminium, which is solderable only in oxygen-free environment.
-
Thank you to everyone who has contributed to this thread.
Using copper plate instead of aluminium is of great interest. Something that I had not thought about before.
We are using the expensive Infineon 10.7 milliohm 200V mosfet IPP110N20NAAKSA1. Two push the power in one direction & the other 2 push the power to a second control board once the first board no longer requires it.
We currently use 2oz copper with a large heat sinking area around the mosfets (on both sides with solder filled plate through holes).
The design originally started out pushing 1.5KW & is currently pushing 3.4KW (bolting the mosfets to an aluminium plate heatsink using heat transfer paste).
We are wishing to automate the process as production numbers rise which is where my interest comes in.
We are also looking to push more power using the newly released expensive IXYS 250V 12 milliohm mosfet IXFQ120N25X3.
Production in Australia is not cheap, hence everything that can be automated needs to be automated.
We also want repeatability. We strive to ensure that every pcb is identical in every way.
Please keep the great suggestions coming. I'm taking them on board & will discuss them all with our other R&D & production engineers to pick the eyes out of it.
We don't want any internal fans as these usually fail at some time in the future & we don't wish to go above 2oz copper on the existing board as there is a lot of SMD with finer tracks already on it.
-
To ensure a uniform contact ( especially with temperature cycles) I wonder if some sort of spring-clamp detail would be best, to ensure consistent pressure.
Instead of paste, maybe a soft metal deformable shim ?
Have a look at what high-end PC coolers are using.
-
This process would alleviate us having to manually screw all the MOSFETs to the aluminium plate & to do away with the manual process of applying heat transfer paste.
Rivets instead of screws would be quicker and more secure.
Or SMD MOSFETS on an aluminium PCB.
Or clips.
Why mess about inventing a new process - the time taken will easily cover many thousands of rivet/screw operations.
I will agree that rivets are a potential option here. At my regular employer we rivet TO-220 packages to heat sinks, hundreds to thousands of boards each day.
Combined with self-adhesive sil-pads and you have a fairly quick process with consistent thermal contact.
Depending on volume you could potentially look at automatic rivet feeders.
Though for the wattages you're looking at, and the cost of the part, I'd of course say to do extensive testing on any potential new process.
My concern with soldering directly to the plate is that you're looking at different expansion rates between a PCB and an aluminum plate. Over time you risk fracturing solder joints due to the flex.
And, like mentioned, you need to either chemically or mechanically remove that outer oxide layer on the aluminum.
If you need to keep with the thermal paste instead of pads, you'd be surprised what an automated system dispensing system is capable of handling.
Depending on the volume that you're using, sometimes just the ability to use larger containers of liquids can be enough of a cost savings to justify the equipment purchase.
Another thought is to automate the existing screw process. Depending on how far you want to go, there's equipment to handle it. Even just going to a hand operated driver with an integrated screw feeder can be a huge time saver. I'd guess you don't do the volume for a full robotic system, but they are out there if you ARE doing that kind of volume.
-
...
Another thought is to automate the existing screw process. Depending on how far you want to go, there's equipment to handle it. Even just going to a hand operated driver with an integrated screw feeder can be a huge time saver...
Or make the existing screw process more efficient. One company I worked for used duo-taptite screws to fasten motherboard housings. These screws roll a thread instead of cutting it, so there is no debris when the thread gets formed.
The metalwork came with formed/extruded holes of the correct diameter. Factory staff would drive the taptite screws in with an air-gun. One short burst and the screw would be driven fully home, forming the thead as it went, and fastening two of the pieces. The process of assembling the metalwork pieces into a chassis was very quick.
-
Not a good idea. A taptite screw will deform the heatsink surface round the hole leading to poor thermal contact. One can alleviate it to some extent in a thick enough heatsink by countersinking the hole, but the countersink depth (and thus diameter) has to be chosen extremely carefully as if it is too small the deformation due to the taptite will leave an annular raised ring round the hole, lifting most of the device mounting surface off the heatsink and if it is too large the device tab will dish, tilting the device, also resulting in poor thermal contact.
The use of pop or impact rivets is also strongly dis-encouraged - the shock of the disposable mandrel separating in pop rivets, or of impact riveting can distort the tab or even crack the die.
See OnSemi's AN1040/D Mounting Considerations For Power Semiconductors (https://www.onsemi.com/pub/Collateral/AN1040-D.PDF)
As the O.P's using thin plate, countersinking other than just to break the edge of any burr left by drilling and tapping is NOT an option. As the heatsink is to be laser cut and later punched, the extra cost of rectangular cutouts for mounting clips will be small or even negligible.
-
While not technically being soldering, another form of joining metals is welding: Spot welding
https://en.wikipedia.org/wiki/Spot_welding
I do not know if it will fry your semiconductors during the process. I know it will easily join aluminum and other materials like copper just fine.
In any case, it seems like your construction is a one-time deal only. The devices are mounted, and never leave for replacement/service.
Best of luck!
-
The Al and Sn can form alloys, but the problem with soldering on Al is that an oxide Al layer is formed almost instantly when heating the solder joint. The oxide layer keeps the Sn and Al apart, so no melting together, no soldering. But, any oxide layer requires oxygen in order to be formed, so if we prevent the Al to oxidase, then we might solder on Al.
One trick to keep the Oxygen away from the solder point is to solder under oil.
You can try it for yourself by
- putting a big drop of oil on the solder point (even the cooking oil will do it),
- scratch the oxide layer (just a few scratches to break the already existing Al oxide layer, but do the scratch under oil so it won't immediately oxidize back),
- then solder under the oil drop.
I tried that by soldering normal wires on food grade Aluminum wrapping foil, and it worked unexpectedly well. The solder drop spread and wet on the Aluminum foil like it would have been on a Copper foil.
For soldering transistors on a big radiator, I guess you will need to preheat the whole radiator before soldering, not a simple thing to do, but totally doable.
NOTE
If you decide to try soldering under oil, use a sacrificial soldering tip.
My solder tip was very hard to wet with solder again after using it under cooking oil, so you may want to cover the soldering tip itself in protective Al foil to avoid oil contamination.
-
maybe not so hard, see my pictures :-DD IC controller directly on the aluminium board, assembled by makerfabs.
-
Some time ago, a colleague tried something similar:
using a aluminium core PCB (as they are quite popular with LED lighting, see zeqing's post), single side and single layer copper. He reflowed a total of 6 or 8 TO220 Mosfets on this PCB, worked fine. There's one major issue: you need to fix the MOSFETs while reflow in some way, but still allow them to reflow, so not too fixed, just kept in position by some special mechanical guidance. Otherwise their pins won't fit anymore into your other PCB because they float a bit around while reflowing.
Using this kind of PCB it doesn't matter if the tabs are all connected together or electrically separated, since you can etch the PCB before assembly.
Otherise, IMO using a sheet of copper instead of aluminium looks like a good idea for your problem.
Edit: if this is for volume production, check with the MOSFETs manufacturer for the reflow soldering of TO220 or similar packages. Usually these packages are not specified for reflowing the tab and may cause issues.
-
Is this on an aluminium substrate pcb?
Can you please share some more information about this process?
Thanks in advance.
-
Yes, it's called IMS:
https://www.pcb-pool.com/ppuk/info_pcbpool_alupanel.html (https://www.pcb-pool.com/ppuk/info_pcbpool_alupanel.html)
Don't know much about the process itself. What my colleague did was some kind of pre-production testing, and we came to the result "yes, it's a genius idea" but no one wanted to actually mass produce the assembly - reflow solderability of the packages and final pin alignment were the issues.
-
Yes, it's called IMS:
https://www.pcb-pool.com/ppuk/info_pcbpool_alupanel.html (https://www.pcb-pool.com/ppuk/info_pcbpool_alupanel.html)
Don't know much about the process itself. What my colleague did was some kind of pre-production testing, and we came to the result "yes, it's a genius idea" but no one wanted to actually mass produce the assembly - reflow solderability of the packages and final pin alignment were the issues.
I'm aware of a company that uses the same tech for their power electronics, they ended up doing the whole assembly from P&P to re-flow in-house. Wouldn't surprise me if that was the reason... they probibly use jigs for alignment.
-
Just a quick thought, What about spot welding. Low voltage so no problem with damaging the MOSFET and the current is isolated around the pins of the spot welder. More so maybe a spot welder with multiple 2 pin terminal that can weld multiple point on the case metal.
Only an idea, never tried it and would be interested in another project I have for myself.
-
Unlikely to benefit, given the warping and thermal shock associated with spot welding, and the copper or other high conductivity substrates involved.
You could perhaps add a filler layer: a foil of solder, or anything lower melting than the base material, to get a larger bond area for less stress.
Tim
-
Spot welding, by definition, welds a single spot - not ideal for heat transfer purposes.
-
Thanks again to everyone who has contributed to this thread.
After the excellent suggestion to use copper plate instead of aluminium, we have decided this is the best way forward.
We will make up a jig to clamp the 4 x mosfets into the correct position on the copper plate after applying a layer of "thick" solder paste (approx 0.5mm) via a specially made stencil. We will then pass them through a vapour phase oven & then xray the finished plate.
If everything looks good, this is the production path we will follow.
-
I would second the jig and an aluminium core PCB, as this can easily and cheaply be done to precisely fit the device pinout and even have slots and other things, like mounting tabs, milled in for essentially free during production, plus you have quite good thermal resistance from the copper to the base, and can even integrate other parts in there simply, like a heatsink thermal sensor, power devices like resistors that need heatsinks, or even just test points. Then you only need to have a few holes to mount to the main heatsink with a thermal pad to lessen the tolerances needed ( thin pad is a lot cheaper than tight control of surface finish on 2 parts) and spring clips to hold each device down to the heatsink. That way rework is possible, and sevicing or repair is a lot easier as you can remove the individual parts easier. just use a thick copper on the aluminium core PCB, and make the pads larger with a solder mask to keep the solder under the TO packages, and in later versions you can use a SMT version ( lower cost immediately there) and simply have some surface mounting pins, to keep the footprint for the leads to the main board the same, to save a respin.
-
One thing I've not seen mentioned - phosphoric acid. Regular old (acid strength, not beer strength) phosphoric acid. It makes soldering to steel, stainless, and aluminum no different from soldering to a regular PCB. Much less hassle than oil and scrubbing and such.
There are probably even more effective chemicals, I'd look into aluminum etchants at that point, but phosphoric acid has worked great for me.