Slashdot Log In
Coating a Motherboard In Thermal Resin?
Posted by
timothy
on Wed Sep 03, 2008 03:58 PM
from the oh-did-you-want-to-use-it-afterwards-too? dept.
from the oh-did-you-want-to-use-it-afterwards-too? dept.
Bat Country writes "I've had an idea in the back of my head for some time (and I'm surely not the only one) that it would be a worthwhile project to coat a motherboard in thermally conductive electrically insulating resin — complete with all of its various components — for the purpose of immersion, shock resistance, whatever. I'm curious to find out if anyone's undertaken a similar project or if it's known to be a shockingly bad idea (due to shrinkage during the curing process) already. Thoughts?" If you've done anything similar (even an experiment that failed), how did you go about it?
Related Stories
[+]
Hardware: Full Immersion Cooling Comes To Desktop PCs 192 comments
mr_sifter writes "After three years of research and around £100,000 of R&D costs, UK-based Armari has unveiled its XCP prototype. It's a full immersion liquid cooled PC which supports standard ATX components. Unlike conventional liquid cooled PCs, the components are all easy to swap in and out as they're swimming in liquid, rather than under waterblocks. It also looks amazing, pumping around 70KG of electrically inert cooling fluid (salvaged from an old Cray) around its military grade perspex shell."
This discussion has been archived.
No new comments can be posted.
The Fine Print: The following comments are owned by whoever posted them. We are not responsible for them in any way.
Full
Abbreviated
Hidden
Loading... please wait.
Conformal Coating (Score:5, Informative)
Re:Conformal Coating (Score:5, Informative)
Conformal coat is generally a thin film applied over the board and components. I get the idea he is talking about something more like 'potting'.
Parent
Potting blocks air cooling, of course (Score:5, Informative)
Electronics has to be designed for potting, at least if it dissipates any significant power. You have to provide a heat path (usually a metal heat sink) out of the potted block. This is done routinely for DC-DC brick [coselusa.com] power supplies. But it's not going to work on a PC motherboard.
Parent
Re:Potting blocks air cooling, of course (Score:4, Informative)
Parent
Re:Potting blocks air cooling, of course (Score:4, Insightful)
If you use the low power VIA chipsets like Eden you only have to dissipate under 10W of power, which these thermal goos could handle. But potting is very insulating compared to open air.
Don't forget the SSD. Standard hard drives won't handle any of this.
It might be useful to rethink the process and figure out what it is you are trying to address. Just moisture - conformal coating. Just vibration - make a better box? If you are strapping this to the outside of a Mercury outboard motor I would go with a better box.
Parent
Re:Conformal Coating (Score:5, Informative)
Those of us who work in the electronics industry know that doing CC in-house is a bitch and inspecting an outsourced job is an even bigger bitch, especially when you're dealing with military parts.
Parent
Re:Conformal Coating (Score:5, Interesting)
Parent
Re:Conformal Coating (Score:5, Informative)
Right, it's conformal coating. People do that all the time. I've used Fine-L-Kote on boards.
Boards with connectors or jumpers have problems. If the CPU and RAM are soldered to the board (as they often are in industrial, consumer, mobile, and automotive devices), just mask off the connectors, jumpers, switches, battery contacts, etc. with masking tape and start spraying. Fine-L-Kote is transparent, but glows in UV, so you can use a UV lamp to check if you missed anything. There are much heavier coatings, ones that really encapsulate the board, and those are widely used for automotive, boat, and aircraft applications.
PC-type motherboards aren't a good choice for this, because of all those connectors. Those are a weak point for corrosion anyway, so protecting the soldered-in components may not be all that useful. But if, say, you're putting something like a single-board PC on your boat, it's quite reasonable.
Parent
Re:Conformal Coating (Score:5, Insightful)
For immersion, the only sensible means is to use a non-conductive coolant and not worry about having to protect your hardware from the coolant it will be immersed in. It might be more expensive than just cooling with water, but it will likely be cheaper than having to replace ALL of your hardware that you just fried.
Parent
Re:Conformal Coating (Score:5, Insightful)
That's what he wants to do?
pfft. Waste of time and hardware. A leak will kill it.
Use something cool, maybe some old Cray coolant. It's out there.
Parent
Re:Conformal Coating (Score:5, Informative)
Conformal coating is typically a thin layer of silicone/urethane/acrylic used to keep moisture from getting at the parts on the board. It can not sustain immersion in liquid.
He's looking for epoxy potting, which we also do occassionly. The trouble with epoxy potting is getting the heat out of the board. You need to leave the thermally conductive parts outside of the potting so that you can remove the heat. The epoxy itself isn't thermally conductive enough to get processor heat out, even on a processor with passive cooling.
You can do this yourself if you have enough time and epoxy, though I'm not sure how much success you'll have. A failed attempt probably means a board that is no longer useable.
The very best ones get cured in a vacuum so that all of the air bubbles are pulled out. There are many other types you can use that don't require a vacuum.
Parent
Re:Conformal Coating (Score:5, Informative)
I wonder where he could readily get a vacuum chamber big enough to pull the air out. Getting the air bubbles out is going to be pretty vital to not baking the mobo, especially the cavities under components.
The most readily available vacuum pumps are smaller than a full motherboard + components, used either for basic science, paint or latex molding. Maybe call up Tap Plastics [tapplastics.com] or your local university and see if they have a big one you can borrow. Because you're going to need a real one. A vacuum cleaner on a Tupperware box or a concrete vibrator isn't going to get what you need.
But my main worry is about thermal expansion. When motherboards get hot they don't expand evenly. Locking everything in resin, not mater how thermally stable, will put a lot more physical stress on the components. And you won't be able to do a damn thing about it except chuck it in the bin and start over again. Though this would probably be lessened with a smaller form factor and lower energy components.
Parent
Re:Conformal Coating (Score:5, Funny)
best slashdot quote ever.
Parent
Re:Conformal Coating (Score:4, Informative)
The amount of vacuum needed for most of these kinds of epoxies is pretty minimal (usually a few inches of water).
Not that this is the recommended method, but we were able to successfully cure epoxy of this nature in a make-shift vacuum chamber. (A plywood box sealed with silicone and a shop-vac.) We even put a clear acrylic window on the front. Of course a real vacuum pump could probably implode our setup, but it was good enough for the epoxy and significantly cheaper than the alternatives.
Parent
Re:Conformal Coating (Score:5, Informative)
The trouble with epoxy potting is getting the heat out of the board. You need to leave the thermally conductive parts outside of the potting so that you can remove the heat. The epoxy itself isn't thermally conductive enough to get processor heat out, even on a processor with passive cooling.
Coming from a materials engineering background, this was my first thought about the question. A material which is very thermally conducting while also electrically insulating sounds like a pretty tall order to me. The reason is that one of the primary methods of heat transfer is through transferring excited electrons.
The two others are phonon transmission through the lattice and radiation. A lot of things which block electrons also block phonon movement (basically heat transfer through vibration) and I'm guessing radiant heat is not going to be sufficient. The one material I do know of which has high thermal conductivity and low electrical conductivity is diamond. It's kind of an oddball, though. I'd be interested to know if there are other, more common materials that would suit this application.
Then there's also the CTE mismatch issue mentioned later in the thread. Whether you design the specs for the high-heat state while it's in use or the cool state when it's off, the other state is going to put a lot of thermal strain on your circuit board components. I'd imagine you could solve some of these problems, but I doubt you're going to be able to take care of them all.
Parent
Re:Conformal Coating (Score:5, Informative)
I think the OP's best bet is to coat the mobo with a CVD diamond film and hope that the liquid cooling keeps the temperature down low enough the film doesn't crack.
Parent
Cray blood (Score:5, Insightful)
Probably a lot easier to source yourself a few liters of Cray blood (or some similar non-conductive coolant) to submerge the board in instead.
Cheers,
Re:Cray blood (Score:4, Informative)
didn't someone put up an article awhile ago here on Fluorinert [wikipedia.org]?
Parent
Re:Cray blood (Score:4, Insightful)
2) The stuff evaporates really fast
Some does, some doesn't. I've worked with the HFE-7500 stuff, which DOES evaporate really fast (much, much faster than water) -- as a result, you can also smell it a little bit, which can be annoying after a while. It also has a pretty low viscosity, which means that it tends to leak through any seals.
On the other hand, the FC-73 stuff (which I've also worked with, though not as much) doesn't evaporate NEARLY so fast as water, and is more viscous, so it doesn't leak very quickly. It also doesn't attack silicone seals nearly so much as the HFE. For home hobby stuff, I'd recommend FC-73 over HFE.
Parent
Re:Cray blood (Score:4, Informative)
Parent
Re:Cray blood (Score:4, Informative)
FC77 and 73 also are longterm environmental hazards and attack the ozone layer. Which is why the HFE family was invented. Like most things of this ilk, the older stuff is better for the task (R12 Vs 134a anyone?)
I use FC77 and HFE7100 as thermal transfer fluids for ultra-cold applications (-60C and lower) in semiconductor testing. Wicked cool stuff.
IF you do manage to buy/acquire enough HFE7x00 remember it eats plasticizers for lunch (gloves are almost useless, better just to be careful).
As such I strongly recommend glass enclosures with as few seams as possible (bend two horseshoes of glass fit them together very tightly then seal with a torch). Obviously a hole for the MB to slide in would be helpful. A tight fitting lid will help to reduce evaporation.
In our professionally built system, a Temptronic Atlas series chiller and thermochuck (http://www.temptronic.com/Products/ThermoChuck_Overview.htm [temptronic.com]) we lost about a pint a month of HFE7100 under heavy load.
-nB
Parent
Re:Cray blood (Score:4, Informative)
Looks like I should read my MSDS more often...
FC-77:
ATMOSPHERIC FATE:
Perfluoro compounds (PFCs) are photochemically stable and expected to persist in the atmosphere for more than 1000 years. PFCs
have high global warming potentials (GWP), exceeding 5000 (100-yr-ITH). The Ozone Depletion Potential (ODP) is Zero.
HFE-7100:
ATMOSPHERIC FATE: Zero Ozone Depletion Potential (ODP). Atmospheric Lifetime: approximately 4.1 yrs. Global Warming
Potential (GWP): 280 (100 year ITH, IPCC1995 method). Global Warming Potential (GWP): 320 (100 yr ITH, IPCC2001 method).
Atmospheric degradation products are expected to include: for methyl nonafluoroisobutyl ether: predominantly isoperfluorobutyric
acid, CO2, HF, and perhaps also CF3COOH; for methyl nonafluorobutyl ether: n-perfluorobutyric acid, CO2, and HF.
The HFE was developed to address the environmental persistence of FC by photo-decomposition. The byproducts, however, still look to be a BadThing(tm)
-nB
Parent
Re:Cray blood (Score:4, Funny)
Either you're new here, or you've misspelt "ethernet."
Parent
Not sure (Score:5, Insightful)
Re:Not sure (Score:5, Informative)
That's why you don't use water, you use something non-conductive. Mineral oil is a relatively cheap and widely available option (just go to your vet and ask for a few gallons of horse laxative) if you don't want to spend the money on commercial grade cooling fluid.
Parent
Re:Not sure (Score:5, Funny)
Parent
Re:Not sure (Score:5, Funny)
You: Hello Doctor! 5 gallons of your finest horse laxative please!
Vet: Oh, got a sick horse then? Are you it needs a laxative?
You: No, it's for me, eh I mean, it's for my computer!
Vet: Ah, the computer! I see. Say no more. *wink wink*
Parent
Re:Not sure (Score:5, Funny)
Vet related too (I do the IT for a group of clinics) but not that relevant:
Customer: "Are gherkin (pickle) slices bad for dogs?"
Receptionist: "I'm not sure - I'll ask the vet...well, the vet says in small quantities they should not be harmful...has your dog swallowed one?"
Customer: "Oh, no, we just wondered whether we should remove it when we buy our dog a burger at the drive-thru"
Parent
Re:Not sure (Score:5, Informative)
It is important to keep in mind that light mineral oil like that, while not as bad as other choices, will leech plasticizers out of insulators. The power supply wiring on my machine very quickly became stiff and brittle and it dissolved the soft rubber that was holding the fan assembly to the processor's heat sink. Not sure if it will have any long term effect on the plugs of the electrolytic caps on the board but I wouldn't be surprised.
If you can afford to split the difference between mineral oil and florinert (perfluorocarbon), you might consider a low viscosity silicon oil. That should bu much nicer to natural rubber compounds and plastic insulators.
Parent
Re:Not sure (Score:5, Informative)
I thought about this a little more and I think I should put this into perspective a little bit.
The first machine I submerged was done in a plastic tub and covered with 5 gallons of Tractor Oil (no, really) that I bought at the wal-mart for $20. It smelled, it was fairly viscous and it definitely leeched plasticizers and crawled up the interior of cables. I ended up with a small puddle of oil under my optical mouse. Wireless worked okay through it though. I tried submerging hard drives after covering the breather holes - I figured if I was keeping them in a fairly narrow temp range they shouldn't need to equalize. Unfortunately none were sealed sufficiently and all eventually died, though one did make it a month before it flooded.
The next machine was a recreation of the Puget Systems [pugetsystems.com] submerged machine. I went with light mineral oil from the local feed and tack store but was seriously considering silicon oil in it's place. Light mineral oil was going for about $17 a gallon, was less viscous than the first oil I used, didn't stink and hasn't effected the components as quickly. I can't find it now but I recall silicon oil being somewhere around $100 a gallon. If I had the money to spare I would have gone that route. Looking online the best price I can find for appropriate Fluorinert is around $1,000 a gallon.
So for 5 gallons of submerged cooling you're looking at $20-$5,000. When you consider the effect it'll have on components, it's probably worth it to use the $500 fill of silicon oil over the $100 fill of mineral oil.
Parent
Re:Not sure (Score:4, Informative)
From my experience with submersion cooling in mineral oil, if your connectors are submerged they will wick oil up the interior through capillary action. If you build a system similar to that made by Puget Systems [pugetsystems.com] it probably won't be a problem, provided you leave some head space at the top of the case. When I built a system similar to theirs I made a short socket extension for the power cord so it didn't have to go under the fluid. everything else was able to stay above the surface of the oil.
If you submerge your video cards and intend to use the fans as impellers, make sure they can start turning against the resistance of the fluid. All my fans worked when submerged except the video cards.
Parent
A Bad Thing (tm) (Score:5, Insightful)
Wouldn't that conduct the heat from the CPU over to the other components?
Technical Term (Score:5, Informative)
The technical term you're looking for is Potting [wikipedia.org].
Re: (Score:3, Funny)
Re:Technical Term (Score:4, Informative)
Then give him brownies!
-Taylor
Parent
No that's different (Score:5, Informative)
Potting is used to keep the components from moving (usually in high-G environments. Sometimes you use it to keep close conductors from shorting (like solder-cup connector), but again the risk there is mostly movement of the conductors, not the environment. Potting materials usually do not have good thermal dissipation properties, and aren't really the best thing for environmental protection (humidity, liquid immersion etc) either. Conformal coating is what you want for the latter.
Parent
SUN used to do it. (Score:5, Interesting)
When they offered the SUN Crypto Accelerator cards for offloading SSL computations, almost the entire PCI card was coated in resin to prevent tampering. I don't think they're still available for purchase from SUN but I'm sure we've still got a few in storage at work somewhere.
Re:SUN used to do it. (Score:5, Interesting)
I saw a documentary on card cheating devices, and one of the early card-counting computers was dipped in something to prevent people from backwards engineering it. It included a failsafe, as well, a thin filament wire designed to be pulled off if the stuff protecting the computer was scraped away, and without that wire in place it would malfunction.
Parent
You should see the IBM version (Score:5, Interesting)
The IBM crypto processors had the module containing the key wrapped in wires (which, if broken, or changed in length, would erase the key) and internal to the module were thermal and x-ray sensors to prevent sniffing the contents of the module that way.
SirWired
Parent
Re:SUN used to do it. (Score:5, Interesting)
Parent
Oil PC going for $140 a barrel (Score:3, Informative)
Tomshardware had a computer in a fishtank full of mineral oil a bit ago. Works well but what a mess.
http://www.pugetsystems.com/submerged.php
Re:Oil PC going for $140 a barrel (Score:5, Funny)
You're supposed to take the fish out of the tank first.
Parent
I've done it. (Score:5, Funny)
Re:I've done it. (Score:4, Funny)
I had an onion on my belt, which was the the style at the time.
Brett
Parent
Look into Fluorinert (Score:5, Informative)
It is electrically insulating and is commonly used for cooling electronics (think Cray supercomputers).
Part of the problem with conformal coat is that it makes it hard to service the electronics after it is cured. It also may or may not be uniformly distributed and thus may not pass muster in a tank of conductive liquid.
There are conductive epoxies like Stycast, but they're not particularly good conductors. The only reason to do immersion cooling is for good thermal contact to all components. A thick epoxy layer between your components and your liquid will quickly destroy that advantage.
Also, if you have connectors to the circuit board (like PCI connectors), then you cannot fill the pins. Last time I checked, most PCI connectors are just slots and have no bottom fill. Water will certainly get in under the coating through the slot.
Try Mineral Oil (Score:5, Funny)
People have been running PC electronics submerged in mineral oil for decades.
Advantages:
1. Not too hard to do
2. If push comes to shove, you can can probably burn the PC in your fireplace or other suitable container to keep warm. Or just because you are pissed at it.
Problems:
1. It's messy.
2. The oil tends to creep up any wires to the outside world (capillary action?) and eventually show up at the other end.
3. I'm not sure if non-gas tight connectors are used in modern PCs, but if they are, they may be a problem.
4. It's messy.
Did I mention that it is messy?
cockroaches (Score:5, Interesting)
Years ago I worked at a company which had had problems with some telecom equipment in the field and no one could ever find any smoking gun. Random problems pointed to several different places on one particular board. One technician must have been working late, because apparently the CO filled with cockroaches once the sun went down. One of the theories was that bugs crawling across the board caused random short circuits. The customer was getting pissed, so management opted for a shotgun approach. Half a dozen shot-in-the-dark fixes were made, including adding an insulating coating. No one knows which one (or combination) of the fixes did the trick, but the random outages went away. That was engineering at its finest.
Industrial Electronics (Score:4, Informative)
In my former life I worked as an industrial electronics technician. My job was, in a nutshell, to modernize a manufacturing plant from its 1950s style, analog (pneumatic) technology, to digital electronic distributed control systems.
The environments these devices need to work in are quite harsh, with extreme temperatures and often corrosive atmospheres. The pneumatic control systems were quite robust in those environments... electronic devices need a lot of beefing up to survive these conditions.
One aspect of this was to treat all circuit boards with a conformal resin coating. The trick is to make sure the thermal coefficient of expansion of the resin, matches the expansion of the circuit board material. I am not a chemist, but I do know such coatings are available.
Another consideration which has been mentioned is how to treat connectors. The usual method is to apply a rubber like sealing compound after a connector is fitted and tested.
For less extreme environments, a much less expensive, but quite effective alternative, is to apply a cheap acrylic coating, using readily available sprays such as Krylon 1301. The procedure is...
Assemble the device (uncoated) and test thoroughly.
Disassemble the device.
Apply tape and / or petroleum jelly to connectors and contacts, to prevent damage from the spray.
Apply the spray to each component.
Assemble and re-test.
Hope this lights a bulb for you.
Yes it works (Score:5, Interesting)
When I was doing a start up called Nisvara 2002 (now dead) we were building Silent computers and server rooms that didn't require air conditioning. Something like 50% power savings!
I was able to Pot or coat, power supply's, hard drive and motherboards in various materials.
The key is thermal conductivity. Yes some one here mention diamond, but that is expensive in unrealistic although diamond dust power was available from GE at a much lower cost then I expected. Carbon Fiber and other carbons are great except they are electrically conductive so they are ruled out (except diamond that is).
What worked great was epoxy with silicon carbide which is dirt cheap and sold as sand basting powder. Also boron nitride works great too, but this is a messy white powder and expensive.
Also a thin layer of silicon carbide or boron nitride epoxy could be applied then a layer of cheaper carbon black or chopped carbon fiber mixed epoxy could be use for making a thicker layer if needed. Non-metallic heat sinks work great using these materials.
We were able to take a Antec 450 Watt Power supply and run it at full load with no fans or heat sink fins as just one big white sold block of epoxy with boron nitride.
Re:Old news (Score:4, Informative)
Power supplies for the C-64 were 'potted' as were many power supplies of the day.
The original C64 psu was renowed for its poor reilability, which was caused for the poor heat dissipation due to that very epoxy potting. They used big TO-3 transistors which got quite warm during normal operation.
Parent