Solid-State PSU Cooling: Beyond the Pail!
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee, Devonavar
Correct! Another advantage Coconut Oil has over Mineral Oil. It's solid at room temperature, so even if it did get pulled up the wires it would stop a few inches above oil level... thanks for pointing that out!
Today I bought more components: total cost is now $207.80 CAD!
But that's OK, because I have plenty more to buy the remaining items; first here's the new pile:
[img]http://isochroma.com/Testfiles/Misc/SPC ... ardware%20[2007-04-10]%20[sm].png[/img]
In the photo above, you can see the items:
1. Coconut Oil: 20kg
2. 16-Qt thin-walled stainless-steel stockpot with lid
3. Tremclad flat-black metal-compatible spray paint
4. Yellow gloves
5. Hacksaw
6. Sandpaper
7. 100' 12/2 Cable
8. Hooks x 6 (brown paper bag)
Tremclad is for painting the outside of the stockpot, so that it is the best radiator. I'll leave the inside as is, because conduction will be the primary heat transfer mode inside, and radiation outside.
Yellow gloves: for protection against lead contact when I solder the ATX & device power leads. (Motherboard: 20, Device: 12, ExtraSquareConnector: 0 [omitted; not needed yet!])
Hacksaw: to make a 3-mm wide cut in the stockpot lid, extending from the outer edge to the center, where it will expand into a round hole for the cables to enter.
Sandpaper: to sand the 3-mm cut so it doesn't cut wires when they are slid in/out from the edge. After sliding in, Duct Tape with quick-release foldbacks on the edges will be used to cover the 3-mm entrance and leftover spaces in the hole, to prevent dust from entering. It can quickly be peeled off as needed.
Cable: Anticipated length of cable required is 10'x20 for the motherboard connector, and additional 10'x12 for devices.
Hooks: Hold the cable up. Pairs will hold the ends up, for backup purposes.
Already Have
1. Eutectic electronic solder & iron
2. Duct Tape
Still Needed
1. Marettes: join cables to PSU
2. Heat-Shrink Tubing: Soldering will be used (due to limited space) to join the cable to male connect on mobo side. After connections are soldered, the HST segments will be slid over and blow-dried into place.
3. Black vauum-cleaner hose: Or something similar, wires will be pushed through it to look neat & tidy.
4. Zip Strips: those whitish plastic strips that cinch tight. One will loop through each hook and then around the cable hose. At the ends, a second pair will be tightly zipped around the cables themselves, to provide suspension.
I may just forget the suspension and use the pot's nice lid for suspension!
Today I bought more components: total cost is now $207.80 CAD!
But that's OK, because I have plenty more to buy the remaining items; first here's the new pile:
[img]http://isochroma.com/Testfiles/Misc/SPC ... ardware%20[2007-04-10]%20[sm].png[/img]
In the photo above, you can see the items:
1. Coconut Oil: 20kg
2. 16-Qt thin-walled stainless-steel stockpot with lid
3. Tremclad flat-black metal-compatible spray paint
4. Yellow gloves
5. Hacksaw
6. Sandpaper
7. 100' 12/2 Cable
8. Hooks x 6 (brown paper bag)
Tremclad is for painting the outside of the stockpot, so that it is the best radiator. I'll leave the inside as is, because conduction will be the primary heat transfer mode inside, and radiation outside.
Yellow gloves: for protection against lead contact when I solder the ATX & device power leads. (Motherboard: 20, Device: 12, ExtraSquareConnector: 0 [omitted; not needed yet!])
Hacksaw: to make a 3-mm wide cut in the stockpot lid, extending from the outer edge to the center, where it will expand into a round hole for the cables to enter.
Sandpaper: to sand the 3-mm cut so it doesn't cut wires when they are slid in/out from the edge. After sliding in, Duct Tape with quick-release foldbacks on the edges will be used to cover the 3-mm entrance and leftover spaces in the hole, to prevent dust from entering. It can quickly be peeled off as needed.
Cable: Anticipated length of cable required is 10'x20 for the motherboard connector, and additional 10'x12 for devices.
Hooks: Hold the cable up. Pairs will hold the ends up, for backup purposes.
Already Have
1. Eutectic electronic solder & iron
2. Duct Tape
Still Needed
1. Marettes: join cables to PSU
2. Heat-Shrink Tubing: Soldering will be used (due to limited space) to join the cable to male connect on mobo side. After connections are soldered, the HST segments will be slid over and blow-dried into place.
3. Black vauum-cleaner hose: Or something similar, wires will be pushed through it to look neat & tidy.
4. Zip Strips: those whitish plastic strips that cinch tight. One will loop through each hook and then around the cable hose. At the ends, a second pair will be tightly zipped around the cables themselves, to provide suspension.
I may just forget the suspension and use the pot's nice lid for suspension!
jaganath
aren't plant-based oils less damaging to the plastic than petroleum-based ones?
Probably, but they still are damaging. This guy reports insulation hardening after 5 months of soaking the PSU in sunflower oil. It's his info about oil climbing up the cables too.
aren't plant-based oils less damaging to the plastic than petroleum-based ones?
Probably, but they still are damaging. This guy reports insulation hardening after 5 months of soaking the PSU in sunflower oil. It's his info about oil climbing up the cables too.
Sunflower, Canola, Flax oils are highly polyunsaturated. They should never be used for cooling! Polyunsaturated oils react with oxygen and even without, causing oxidation of other materials in contact, and themselves hardening up (cross-polymerizing). Think paint, which is based on Flax oil (in this application it is called Linseed Oil).
Coconut oil is non-reactive, because it is totally saturated. There are no reactive carbon-carbon double bonds, like polyunsaturated fats contain. Thus no cross-polymerization or formation of reative radicals which corrode other organic compounds (insulation, etc.).
Coconut oil is non-reactive, because it is totally saturated. There are no reactive carbon-carbon double bonds, like polyunsaturated fats contain. Thus no cross-polymerization or formation of reative radicals which corrode other organic compounds (insulation, etc.).
@wwenze: hope you were joking about the HDD
Tomorrow I buy the last of the supplies, and hopefully start construction. The remaining items are:
1. cute little rubber feet to screw onto the bottom of the circuit board, so that it can rest on the pot's metal bottom without shorting out. They will also be placed on the sides by epoxy or other sticking agent, to prevent shorting to the pot's sides - but I doubt there are any conductive components on the sides, and the heatsink is in the middle of the board.
2. little pointy screws to screw thru the PSU-board's existing mounting holes, and into the cute little rubber feet.
I've decided to use Duct Tape to hold the cable-bundle to the underside of my desktop, because it won't damage the desk and is easy to remove/move.
-----------------------------
Finally, a Note about the method:
Pole Pig: Circuit in a Can of Oil
This is my inspiration! Those wonderful grey cans you see on poles by the billions, powering homes all over the world... what's inside them?
They are just steel cans, and inside is a coil immersed in mineral oil. It is passively cooled by oil convection, and is precisely the same design as my PSU-in-a-Pot project.
Passive oil-convection in a cylindrical metal case is the cheapest, most reliable electric cooling system in the world today - it is likely powering the computer you are reading this post on.
The surface area for heat transfer of the above transformer is about 5 times greater than my extra large stock pot. However, the above transformer handles at least 10,000 watts of power, and is probably rated at 15kW (likely more).
At minimum, it is moving 25 times more power than my 400W PSU at max, but has only 5 times the surface area (maybe a bit more, but not much).
Granted, the efficiency of my PSU is rated by the manufacturer at >70%, so we'll just assume 70%. The pole-pig does a bit better, probably around 95%, so at peak load waste energy is...
Pole-Pig @ 95%: 15kW: 750W
PSU @ 70%: 400W: 120W
6.25:1 waste heat (W)
5:1 surface area (approx).
So my design for the PSU-in-a-Pot is about right; it will be able to dissipate heat as well as the pole-pig can, and won't be exposed to the same temperature extremes (direct sun in particular). Remember though, my system doesn't pull 400W, and was previously running just fine on a 350. It is probably taking around 300W max, or 90W of waste energy.
Tomorrow I buy the last of the supplies, and hopefully start construction. The remaining items are:
1. cute little rubber feet to screw onto the bottom of the circuit board, so that it can rest on the pot's metal bottom without shorting out. They will also be placed on the sides by epoxy or other sticking agent, to prevent shorting to the pot's sides - but I doubt there are any conductive components on the sides, and the heatsink is in the middle of the board.
2. little pointy screws to screw thru the PSU-board's existing mounting holes, and into the cute little rubber feet.
I've decided to use Duct Tape to hold the cable-bundle to the underside of my desktop, because it won't damage the desk and is easy to remove/move.
-----------------------------
Finally, a Note about the method:
Pole Pig: Circuit in a Can of Oil
This is my inspiration! Those wonderful grey cans you see on poles by the billions, powering homes all over the world... what's inside them?
They are just steel cans, and inside is a coil immersed in mineral oil. It is passively cooled by oil convection, and is precisely the same design as my PSU-in-a-Pot project.
Passive oil-convection in a cylindrical metal case is the cheapest, most reliable electric cooling system in the world today - it is likely powering the computer you are reading this post on.
The surface area for heat transfer of the above transformer is about 5 times greater than my extra large stock pot. However, the above transformer handles at least 10,000 watts of power, and is probably rated at 15kW (likely more).
At minimum, it is moving 25 times more power than my 400W PSU at max, but has only 5 times the surface area (maybe a bit more, but not much).
Granted, the efficiency of my PSU is rated by the manufacturer at >70%, so we'll just assume 70%. The pole-pig does a bit better, probably around 95%, so at peak load waste energy is...
Pole-Pig @ 95%: 15kW: 750W
PSU @ 70%: 400W: 120W
6.25:1 waste heat (W)
5:1 surface area (approx).
So my design for the PSU-in-a-Pot is about right; it will be able to dissipate heat as well as the pole-pig can, and won't be exposed to the same temperature extremes (direct sun in particular). Remember though, my system doesn't pull 400W, and was previously running just fine on a 350. It is probably taking around 300W max, or 90W of waste energy.
You're either relying on convection or conduction for heat transfer; there's nothing in-between. Adding metal oxides to oil thickens it such that convection doesn't work. And bulk heat transfer by conduction through mineral-oil-metal oxide is very poor.
Heat-sink paste doesn't conduct heat very well - it is used primarily to fill microholes in the metal surfaces, but only a very thin layer. In this application, it has only to be better than air. The oil in thermal grease is to make it flow into all the microholes on the two surfaces. However, the price of using oil is very high, because it is a very good insulator. That's ok for heatsinks, because only a very thin layer is needed, so its hole-filling property has a much higher value/liability ratio than its bulk insulative property.
Oil works well when it is free to flow because convection can transfer heat, but it is a terrible liability if it cannot flow, due to obstructions or thickness.
For bulk heat transfer, metals, low-viscosity liquids, or tightly packed metal oxides are the game.
For example, the best thermal grease, Arctic Silver III material has a conductivity of 9.0 W/mK, but my next-best material for the PSU-in-a-Pot project, Aluminum Oxide, gets 30 W/mK.
See this article for more thermal grease numbers:
http://www.overclockers.com/articles662/
Heat-sink paste doesn't conduct heat very well - it is used primarily to fill microholes in the metal surfaces, but only a very thin layer. In this application, it has only to be better than air. The oil in thermal grease is to make it flow into all the microholes on the two surfaces. However, the price of using oil is very high, because it is a very good insulator. That's ok for heatsinks, because only a very thin layer is needed, so its hole-filling property has a much higher value/liability ratio than its bulk insulative property.
Oil works well when it is free to flow because convection can transfer heat, but it is a terrible liability if it cannot flow, due to obstructions or thickness.
For bulk heat transfer, metals, low-viscosity liquids, or tightly packed metal oxides are the game.
For example, the best thermal grease, Arctic Silver III material has a conductivity of 9.0 W/mK, but my next-best material for the PSU-in-a-Pot project, Aluminum Oxide, gets 30 W/mK.
See this article for more thermal grease numbers:
http://www.overclockers.com/articles662/
no problem, my knowledge of your language is probably even more insufficientSorry for my insufficient knowledge of English
http://en.wiktionary.org/wiki/live#Adjective
On some PSUs the heatsinks have very high voltage (300-400V?) going through them.(circuitry) Electrically charged or energized, usually indicating that the item may cause electrocution if touched.
example: Use caution when working near live wires.
Wires & other obstructions cause turbulence in the "laminar flow" of the oil, and make it difficult to remove the supply for maintenance. And yes, most PSU heatsinks are electrically "hot" with high voltage.
My most difficult problem yet to be solved, is how to bend 20 12-AWG solid wires to get them into the PC case.
The idea so far is to leave about 2 inches of flexible stranded wires on the mobo-end PSU ATX connector, allowing it to turn the nearly right angle needed for connection to the mobo socket.
My most difficult problem yet to be solved, is how to bend 20 12-AWG solid wires to get them into the PC case.
The idea so far is to leave about 2 inches of flexible stranded wires on the mobo-end PSU ATX connector, allowing it to turn the nearly right angle needed for connection to the mobo socket.
False. There is conduction everywhere except in a vacuum. You can think of conduction as the transfer of energy by collisions of atomic and subatomic particles.Isochroma wrote:You're either relying on convection or conduction for heat transfer; there's nothing in-between.
Convection exists whenever you have mass transport and a temperature difference - moving mass means moving heat. You've also got radiation heat transfer, from the PSU to the oil, PSU directly to the pail, from one oil molecule to another, etc. Radiation flows both ways, but the net effect is of course from hotter surfaces to colder ones.
However, you're correct that conduction in gasses and oils is very weak. Don't discount it completely, though. It has some effect on your temperatures. More so in the case of oil than air.
Convection in viscous fluids is much slower than in thinner fluids. It's better in denser fluids. Anyway, oil beats air there, too.
Oh, and turbulence: Don't worry about turbulence. A viscous fluid like that isn't going to go turbulent due to convection on the small scale of a PSU. Besides, turbulence near a surface means higher heat transfer. Worry instead about the fact that flow restrictions such as wires slow the flow.
Sorry if I'm being too pedantic.
Sorry if I'm being too pedantic.
The project is complete! Took three days of work/downtime. I learned a few things, and there were only two items which were not expected:
1. The pot's handles are riveted on; the rivets are gasketless, so oil will leak through them if its level reaches the rivets.
2. The stainless steel lid is uncuttable: I ruined a brand-new hacksaw blade and two drill bits trying. So in the end I bought some wide black cloth tape and taped the top over.
The machine runs almost totally silent now, and the oil runs about body temperature at circuit level, and warmer at the top. Everything works reliably. There is a faint smell of coconut oil in my room, but after 24 hours it is already half what it was and should be gone in a week or two.
There is no wicking of oil up the cables, and about 6 hours after the power is turned off, the PSU-in-a-Pot is fully solidified and ready for transport, without worry of spillage!
Below are images of various stages in the construction. The total cost is $290.59 CDN. This includes about $30 worth of re-usables, partially consumed consumables and tools that were purchased for the occasion, and does not include solder cost or soldering iron (already in my posession).
1. The pot's handles are riveted on; the rivets are gasketless, so oil will leak through them if its level reaches the rivets.
2. The stainless steel lid is uncuttable: I ruined a brand-new hacksaw blade and two drill bits trying. So in the end I bought some wide black cloth tape and taped the top over.
The machine runs almost totally silent now, and the oil runs about body temperature at circuit level, and warmer at the top. Everything works reliably. There is a faint smell of coconut oil in my room, but after 24 hours it is already half what it was and should be gone in a week or two.
There is no wicking of oil up the cables, and about 6 hours after the power is turned off, the PSU-in-a-Pot is fully solidified and ready for transport, without worry of spillage!
Below are images of various stages in the construction. The total cost is $290.59 CDN. This includes about $30 worth of re-usables, partially consumed consumables and tools that were purchased for the occasion, and does not include solder cost or soldering iron (already in my posession).
Last edited by Isochroma on Sat Apr 28, 2007 10:57 am, edited 1 time in total.
Re: Solid-State PSU Cooling: Beyond the Pail!
[quote="Isochroma"]After much investigation, I've decided to try solid-state cooling!/quote]
For info: Article in "New Scientist" April 10, Chilled oil bath...
http://www.newscientisttech.com/article ... sors-.html
For info: Article in "New Scientist" April 10, Chilled oil bath...
http://www.newscientisttech.com/article ... sors-.html
Isochroma, I believe you've done it.
I've been on forums the net over, during the last 3 years discussing submerged cooling. There have been some very smart people researching all sorts of exotic oil properties, particularly mineral oils. All in search of the perfect replacement for 3M's horrendously expensive range of Fluorinert's. (Which have pour points ranging down to around -70dC [from memory], great for chilling, but the evaporation pressure at room temp is like watching money fly out of your wallet.)
As mentioned in this thread, typical food oils are unsaturated, making them reactive and prone to going rancid. I remember one poster chasing some wild extra-saturated variation of canola oil, but it turned out not to be made in commercial qualities. Very stable mineral oils are fantastic in sealed cases, but air tight sealing of the cables (with an allowance for heat expansion) is tricky, and you don't really want to be breathing the vapour if it leaks.
Don't be surprised if in 5+ year’s time, coconut server solutions are being offered commercially.
Thanks Mr Coconut!
PS The danger of oil moving up USB/etc. cables via the capillary effect is overcome if they are hung around 2.5 feet vertically first. I agree that coconut oil would be less likely to exhibit this issue however, due to it solidifying.
PPS Now that solid state h/drives are finally beginning to come of age, it might be time to build some 0db high performance systems. This 32Gb puppy has a claimed sequential read of 57Mb/sec, a sequential write speed of 38Mb/sec, and is < US$410. Due to their IO speeds, these would kill a Raptor for many applications. I'm sure they'd submerge very nicely too
.
I've been on forums the net over, during the last 3 years discussing submerged cooling. There have been some very smart people researching all sorts of exotic oil properties, particularly mineral oils. All in search of the perfect replacement for 3M's horrendously expensive range of Fluorinert's. (Which have pour points ranging down to around -70dC [from memory], great for chilling, but the evaporation pressure at room temp is like watching money fly out of your wallet.)
As mentioned in this thread, typical food oils are unsaturated, making them reactive and prone to going rancid. I remember one poster chasing some wild extra-saturated variation of canola oil, but it turned out not to be made in commercial qualities. Very stable mineral oils are fantastic in sealed cases, but air tight sealing of the cables (with an allowance for heat expansion) is tricky, and you don't really want to be breathing the vapour if it leaks.
Don't be surprised if in 5+ year’s time, coconut server solutions are being offered commercially.
Thanks Mr Coconut!
PS The danger of oil moving up USB/etc. cables via the capillary effect is overcome if they are hung around 2.5 feet vertically first. I agree that coconut oil would be less likely to exhibit this issue however, due to it solidifying.
PPS Now that solid state h/drives are finally beginning to come of age, it might be time to build some 0db high performance systems. This 32Gb puppy has a claimed sequential read of 57Mb/sec, a sequential write speed of 38Mb/sec, and is < US$410. Due to their IO speeds, these would kill a Raptor for many applications. I'm sure they'd submerge very nicely too
.
Indeed... my PSU is still running nicely, it's been up 24/7 so far. No rancid smells or polymerization yet. The faint coconut odour lasted about 1 month and is now gone.
Solid-state drives... the last 2x1GB DDR sticks should be arriving for my iRAM SSD within a week. Already have 1 iRAM running 1GB for the pagefile, this one will run 4GB for the entire C: drive Insta-boot, insta-launch, lasts forever, no corruption, no noise, low power.
Solid-state drives... the last 2x1GB DDR sticks should be arriving for my iRAM SSD within a week. Already have 1 iRAM running 1GB for the pagefile, this one will run 4GB for the entire C: drive Insta-boot, insta-launch, lasts forever, no corruption, no noise, low power.
PSU is still working nicely, but lately I detect a slight high-frequency whine.
My next PSU-in-a-Pot project will use the same pot, but silicon carbide fill instead of oil: http://www.reade.com/Particle_Briefings ... amics.html
My next PSU-in-a-Pot project will use the same pot, but silicon carbide fill instead of oil: http://www.reade.com/Particle_Briefings ... amics.html
-
- Posts: 76
- Joined: Mon Oct 29, 2007 2:02 pm
Over time the coconut oil will develop polar compounds that are electrically conductive. You might look into fractionated coconut oil along with a non conductive stabilizer. I don't know the exact properties of particular stabilizers but there are a number of them. Outrageously expensive silicone liquids, TBHQ and rosemary oil come to mind. If you can find an oil plant that makes fractionated coconut oil(try the one that made the oil you bought) then they may send a 5 gallon sample with stabilizer for free or a small fee if they perceive a market opportunity.
Stainless steel will cut with a reciprocating saw. I have used abrasive and bimetal blades. The bimetal blades are faster and make tighter turns but the abrasive blades make a smoother cut. For drilling stainless look up cobalt bits. Aluminum is cheaper and probably better for what you are doing.
Have you ever seen concrete poured where workmen have vibrating shafts that they use to compact the concrete? You may be able to mimic this with a pager or cellphone vibrator available at many surplus outlets attached to a thin metal rod.
Finally, where I live it is obligatory to say "Hold my beer and watch this!" before flipping the switch on something like this.
Stainless steel will cut with a reciprocating saw. I have used abrasive and bimetal blades. The bimetal blades are faster and make tighter turns but the abrasive blades make a smoother cut. For drilling stainless look up cobalt bits. Aluminum is cheaper and probably better for what you are doing.
Have you ever seen concrete poured where workmen have vibrating shafts that they use to compact the concrete? You may be able to mimic this with a pager or cellphone vibrator available at many surplus outlets attached to a thin metal rod.
Finally, where I live it is obligatory to say "Hold my beer and watch this!" before flipping the switch on something like this.