Passive Immersion Cooling of Electronics

[Anthony]

Waaaaay of topic...

Wanting an interesting and silent way to cool a PC, I elected to build a sealed case and submerge the motherboard in oil...

However, one doing some more reading on t'internet, I read a few reports that capacitors (I'm guessing electrolytic) are sensitive to even changes in air humidity, and that immersing them will allow the coolant to migrate into the capacitor and destroy it.

Obviously this would be a bit of a bummer if it killed my new motherboard :(

I know we've got a few members who are very knowledgeable when it comes to electronics, and am wondering if they've got any input on submerging electronics in oil?

[Flake2m]

well I have seen a website where a guy submergered his pII 266 into oil. The heatsink fan even circulated the oil around the tub.
However he didn't have the HDD and CD drive in the oil as HDD need air intake and Oil in CD drive would have stuffed it up.
I wouldn't try immersion cooling with a brand new system. You could try a water cooling setup. The better water cooling setups require significant case modifaction as the radiator is often 2 big to fit in the case.

[Anthony]

Obviously the drives aren't meant to go in the oil :)

The idea is that with the mobo and PSU submerged, there is no need for fans, thus the PC is practically silent. A submersible pump in the oil would keep up a circulation to prevent hotspots. The tank would radiate a certain amount of heat, possibly enough to allow continuous operation. If not, the high thermal mass of the oil would allow operation up to a certain duty cycle (I don't tend to run my PC continuously).

I've seen the experiment you mention, along with a few others. However, they all say "yeah, here's a pic of what I did, do I not rule?", but neglect to state anything about reliability! Only one that did, mentioned that the PSU died after three days due to a blown capacitor...

There are more suitable coolants than oil. Mainly "Flourinert", a coolant made by 3M and used in applications such as the Cray supercomputers. Trouble is, it costs $500 a gallon!

I've considered traditional watercooling, but most systems just tend to move the fan from the CPU to the radiator. I suppose I could use a heatexchanger to boost the water tend and run it through an oversized radiator and use no fan.

Ideally, I'd run watercooling with the reservior buried a few feet underground outside of the house, no fans, no radiators, no heat exchangers. Not practical in the house I'm in though.

Also, immersion cooling would cool *everything*. There are plenty of components that get quite hot on mobos/expansion cards. Conventional watercooling does nothing for this, and air cooling doesn't do much.

[Tuatara]

A lot of components are capapable of absorbing small amounts of moisture, ceramic capacitors being one of the worst. And there will be a lot of ceramic caps on a PC mobo. Oil cooling should work fine, so long as you use the right sort of oil, ie NOT car engine oil. One of the silicone oils should work - Dow Corning do a lot of different silicones (not $500 a litre either!) get in touch with them and see if they can recommend something suitable. If you're concerned about component damage, try overcoating the PCB with a conformal lacquer first - available in spray form at most electronics hobbyist shops.

[vulture]

Ever thought of Argon gas? Protects your contacts from oxidation and has far better heat conduction (noble gas) than normal air.

Ofcourse, airtight seal would be required.

SF6 could be used too.

[Anthony]

Maybe try gases next time, Vulture ;) One step at a time...

Thanks Tuatara, that was exactly the kind of answer I was looking for :) Good call on the motor oil, a cheap bottle of that was going to be my first try.

I've had a look on Dow Corning's website (under "silicone fluid") and they do some nice thin ones e.g. 0.65 and 1 cts, but they're as volatile as acetone, with a flashpoint of -3*C! The thinest non-volitile one is 20cts, which should be good. I'd like it to be as thin as possible to aid pumping. From the few distributers I've found, price seems to be about £4.50/kg. Might have to revise the design to cut the volume down!

Some kind of spray laquer was also something I considered. I imagine it'd be a bitch if it got onto unshorted jumpers or into expansion sockets. Careful masking might be called for.

At least I dismissed parafin oil and deionised/distilled water from the outset. "pure water wouldn't be conductive", I thought. Yeah, untill it came into contact with anything metallic and became re-ionised! Not to mention corrosive.

[Tuatara]

Glad to be of assistance! Thinking about it, synthetic motor oil might be a starter - you'd want to try it on an old mobo first though. I discounted motor oil initially as you don't know whats in it - could be all sorts of crap capable of dissolving epoxy and other plastics.

If you're on a really tight budget, and not afraid of toxic nasties, I bet you could scrounge some old transformer oil for free (you're not afraid of poly chlorinated biphenols are you?).

Gas will never be as good as liquid, and as for SF6 - well if you can afford that you can afford the fancy 3M shit!

[Bert]

Ever thought of Argon gas? Protects your contacts from oxidation and has far better heat conduction (noble gas) than normal air.

Ofcourse, airtight seal would be required.

SF6 could be used too.

If Argon is a superior heat conductor, why was I taught to use it to inflate my dry suit and told it had superior insulating value by my diveing instructors. Here's A link regarding Argon (http://www.decompression.org/maiken/Why_Argon.htm) Helium would be an excellent candidate as far as thermal properties go, but good luck trying to retain it in a home made enclosure!

[vulture]

Odd, normally all noble gasses are good heat conductors.

[NightStalker]

I was going to attach a small PDF on the physical properties of various CFC and HCFC's which might be useful for your liquid cooling project, anthony, pgp'd to your key, but attachments seems to be not available anymore. :confused: Anyways, find a copy on Rhodium's site.

I beelieve automotive A/C refrigerant, it's either HCFC 141b or 131, would bee useable, as both are liquids at room temp/non-flammable/cheap. :) Mainly cheap. Though they often have oils/anti-corrosion/other additives mixed in, so you may have to distill them to purify them before they'd be safe to use on a motherboard.

Oh, and if you have an old one, DEFINATELY test this on the old board beefore risking some fine new piece of gleaming techno-lust. ;)

With room temp fluorocarbons, you could set up a very long tube from your case up a wall or such to act as an air condensor and, assuming that your room temp never exceeeds the BP of the refrigerant, the coolant will condense and flow back into the case.

[Flake2m]

In one of the experiment I saw on te 'net the guy used ordinary cooking oil. The stuff that cost 1 euro a Litre. If you kept the heatsinks on the CPU the the fan will circulates to oil which woulds save you the trouble of having to buy a pump.

Also I wonder if there would be improvement in effiecentcy if oil was used instead of water for a water cooling setup? It would be also intresting to try methylated spirits and some other non-conventional cooling agents.

[Tuatara]

Water is particularly good because of its high specific heat, and low viscosity - both of which are essential to good heat transfer. High specific heat means lower flow rates can be used, and low viscosity means easier to pump, but more importantly a thinner boundary layer around the parts you are trying to cool. Read a fluid mechanics book if you want to know what a boundary layer is.

Its also non-toxic, and cheap! :D

[Anthony]

I should have added to my last post that I'm sure I have an old 486 *somewhere* which will be a guinea pig :)

Yeah, I've heard of the nasties that can form in transformer oil!

I was under the impression that fully synthetic motor oil was quite pricey, but having had a look, it seems to be £20 a gallon. Which isn't too bad - I wouldn't refuse to buy it for my car.

My initial thought was mineral oil. Which I think is non-synthetic motor oil (amongst a hundred and one other uses). Apparently cheaper to buy than dispose of, but I couldn't find anyone selling it as-is, save for tiny pharmaceutical quantities.

I think the first step will be to buy some synthetic motor oil and let rip on the 486. I can test the polycarbonate that I want to use for the case panels at the same time.

I planned to use a pump for two reasons: 1) I don't trust a PC fan to cope with the greater increased load posed by the oil 2) More importantly, I have an aquarium powerhead going spare. In addition to circulation, I'd also be able to attach a tube to it to drain the case if need be. Hopefully the impeller will be able to shift a fluid thicker than water.

Room temp refrigerants is something I will give thought to. Surely the long tube could replaced with a more compact radiator? Because there's no phase change involved, I assume the system could be driven by a normal pump, rather than a compressor? I don't think the inside temp in this country ever exceeds 30*C, and only then rarely.

[T_Pyro]

Regarding water cooling, you should see this (http://www.webx.dk/oz2cpu/pcmod/water.htm).

[Anthony]

Interesting page, never seen a watercooled PSU before.

I wonder though, surely there are more than FETs in a PSU that need cooling? With the airflow removed aren't any components not attached to the main heatsinks at risk of overheating?

Unfortunately I don't have the use of a milling machine. It'd probably be possible to do without one, but it'd be easier to buy a commercial waterblock. I'd source/make everything else myself though!

I was thinking about those room temp refrigerants...

Either there is no phase change involved, in which case, what's the advantage over water?

Or more likely, the idea is that the refrigerant doesn't boil at room temp, but does in contact with the relatively hot CPU, and will then recondense in the room temp heat exchanger. If so, how would the system be driven? Impeller pump on the liquid side?

[Jacks Complete]

You could sidestep the whole issue in one.

How fast a MoBo do you need? If it is going to be a nice lean one, for everyday use as a video/DVD/stereo/web browser, try www.mini-itx.com (Check out the one in the ammo can here (http://www.mini-itx.com/projects/ammobox/?page=3)!) for a range of either very quiet (25dbA @ 1m for a 1Ghz) or totally fanless version at 666Mhz. You can even get solidstate PSUs that run off 12V for them.

Lots of devious ideas for those babies! Feds at your door? Close the can, throw it in the garden pond, and it is safe till you fish it out!

[NightStalker]

The entire board is vertically mounted in a liquid-tight casing, with the CPU at the bottom.

As the fluid is heated by the CPU, it becomes warmer (less dense) and rises to the top of the casing, where it overflows into the radiator.

As it flows through the passive air-cooled radiator, it gives up its heat, becoming cooler (more dense), where it is thusly lead back into the case.

Be proper adjustment of inlet/outlet positions, relative to the CPU, and accounting for flow impedence from video boards and whatnot, it should be totally passive with no need for pumps or fans. :)

As the liquid is in constant circulation, it should never reach boiling, and if it does, it can't exceed it's BP as long as there's fluid covering the CPU which, being at the bottom of the tank, would be the last thing to "fry" if it leaked coolant, hopefully giving you time to turn it off.

In fact, you might want to install a float switch (like in a toilet tank) in the casing so that, if it does leak, the power will be turned off beefore the CPU goes dry as the coolant level drops, thus saving your MoBo. ;)

For the watercooled CPU blocks, you might try using lost wax casting to make a hollow block from scrap aluminium. Or use copper tubing embedded in a cast aluminum block.

[Anthony]

I see, the system would run on a convection current. Although, doesn't warm water sink? Or is this only at low temperatures, as I know it does in ponds during winter?

I've looked at mini-itx before. They're neat for cramming into small spaces (I was considering one for a carputer). The upgradability/expansivity isn't very good though.

Also seen the ammo box thing, been there, done that though :) Admitedly it was a bigger ammo box (fuzes actually). But then it does contain an ATX mobo, a powersupply, drives and a 500VA UPS (hour runtime with monitor).

On the lid is a standard UK plug socket as PTO. I usually plug my monitor into it, but can run pretty much anything within the power rating of the UPS. I ran a jigsaw from it once.

Sad, but when I moved house, I made sure the cable was installed before moving in. I didn't turn the computer off when I moved it, hooked up the cable modem first thing and carried on, without suffering downtime on the computer.

[T_Pyro]

The anomalous behaviour of water is only between 0o and 4oC.
Do you swap HDD's often? If so, complete immersion of the motherboard would be quite cumbersome! Hence, I prefer the watercooled CPU blocks. On the page that I'd mentioned, that guy took care of all the essential heat generating components: CPU, GPU, chipset, HDD, SMPS. Apart from those components, there are hardly any other components that generate any considerable heat (soundcard, ethernet card, etc...) Hence, it doesn't really matter if there's no airflow in the cabinet.
I was thinking that similar cooling blocks could be constructed by welding cut pieces of a copper sheet. Might be simpler than the lost wax process.
Additionally, if you're not too worried about some noise, you could use a radiator+fan setup to cool the water.

[TheHitMan]

Off topic but Anthony is "t'internet" a Peter Kay referance?

[Anthony]

Indeed it is :D

I don't swap HDDs often. The plan is to have the mobo inserted through the open top of the case with the ports upwards. It'd be like having a vertical removable motherboard tray.

It was the PSU components that I was worried about suffering from lack of airflow. There must be transistors, resistors, ferrites etc putting out a decent amount of heat and working close to the limits of their capacities.

[T_Pyro]

By "PSU", you mean the SMPS (Switched Mode Power Supply, or Power Supply Unit), right? In the SMPS, the transformer and the series pass transistors are the only components which generate large amounts of heat. The rest of the components are there to sample the o/p voltage, vary the pulse width accordingly, and feed the clock pulse to the series pass element, to put it simply. The capacitors generally have a voltage rating higher than they'd be be used for, so there's no problem with them.

I've never had trouble with these elements, though the transformer and the series pass transistor have acted up once or twice in some power supplies (although not the regular PC SMPS). But then again, I've never had to use the SMPS to the full rated capacity, either. If you're driving the SMPS close to its limit, I guess it would be wise to have some extra cooling.