Watercooled Powersupplies

[Per Hansson]

I've been looking into this a bit, there seem to exist only two "retail" watercooled PSUs, made by Silentmaxx:

http://www.quietpcusa.com/acb/showde...ID=243&CATID=2

http://www.aqua-computer-systeme.de/...dd81f195ea3f59

There are some other, like the Enermax used in this amazing build..., however that is a regular fancooled PSU that a shop (http://www.heattrap.de/) has modded, i.e. I do not call that a "retail" PSU, since it has not been built by the manufacturer...

If you just look around on google there are several guys that have built their own watercooled PSUs, most unfortunantley on Deer designed PSUs

http://www.unixmonkey.net/articles/waterpsu/
http://www.webx.dk/oz2cpu/pcmod/water-psu.htm

I was thinking of making a copper waterblock for my Seasonic 430w PSU, though on second though what will happen to the coils etc that do get very hot in a PSU? Or are they just warmed by the Mosfets?

I.e. it it ok to make a waterblock like in the last url above with only the mosfets connected to it, or will the rest of the PSU get overheated then?

[tiresias]

Re: Watercooled Powersupplies

I.e. it it ok to make a waterblock like in the last url above with only the mosfets connected to it, or will the rest of the PSU get overheated then?

Hmm... well, the hottest components in a PSU are normally the double-diodes (which is why they have the biggest heatsink). The MOSFETs usually are next, with a smaller heatsink - of course the transformer isn't really sinked at all, although it does have a relatively large volume and surface area in which (and respectively with which) to disperse the heat.

Still, I wonder whether a totally fanless watercooled approach is really feasible, since with no airflow heat would gradually accumulate from other components, including coils, smaller transformers, even electrolytic capacitors, to the point that it may compromise the unit's functioning.

Perhaps a good idea would be to install watercooling on the three hottest component/component banks (rectifiers>mosfets>T1), and use a very quiet, weak fan to avoid heat buildup.

[Galvanized]

Re: Watercooled Powersupplies

Did read your post earlyer today and with a quick look, couldn't find the home built watercooled PSU at overclockers.com or SPCR. I do think air flow of some degree is nesessary. The Antec Phantom 500W has a low cfm 80mm and with The Etasis/Silverstone 300W passive it is stongly suggested to have proper case air flow.

I think in the future we will see heat pipe tech in PSU cooling, something like the MGE 500W but with better execution/design on a quality PSU. Note the lack of TIM in the pics.
http://www.xbitlabs.com/articles/oth...oundup_10.html

The sink in your second link would be much better done had he formed a U out of 1/4" copper tubing, then silver soldered it into the two 90degree bends. The contact area would be four times greater and silver is a much better transmitter of heat than lead based solder, imho.

I have never tackeled a WCed PC and doubt I ever will, but do have a thought about it based on my considerable automotive background.
The continous single loop most all use when water cooling a PC has faults (imho) because it directs coolent to everything with the same flow rate. I would think manifolding the pump discharge to each componet on a seperate line with all returns to it's own manifold. Where the return lines meet the manifold, each would have it's own shut-off valve. These valves would be used to regulate coolent flow as needed. Full stress testing while monitoring temps would have to be done inorder to adjust the restiction the valves would offer to controlling flow rates.

In doing the above, a water block with poor performance can recieve more coolent. While a very good performing block could do with less.

[Rainbow]

Re: Watercooled Powersupplies

http://www.unixmonkey.net/articles/waterpsu/

This is BAD. Never do this with a PSU. The long leads can (and will) cause problems.

http://www.webx.dk/oz2cpu/pcmod/water-psu.htm

This looks much better. Except for the PSU itself, which looks like a weak 200W one.

[japlytic]

Re: Watercooled Powersupplies

I've seen a picture of a power supply being placed into a container filled with castor oil (with the mainboard) in a computer magazine. This was entirely safe(!)

The link from Denmark (.dk) shows a Deer/L&C crap. Why bother modifying a crap supply?

[tiresias]

Re: Watercooled Powersupplies

This is BAD. Never do this with a PSU. The long leads can (and will) cause problems.

That particular one is not only bad, it's actually, from a computer's point of view, hazardous (and I'm by no means a safety freak, trust me!). If you look carefully, you'll see he/she has attached both the MOSFETs from the primary 300Vdc side and the diodes from the secondary, low voltage side to the same copper plate!

The only thing preventing the two making contact and sending all attached components to the "other side" are two thin layers of TIM, which have been disturbed (removed and reattached) at least once already.

Kind of defeats the point of having isolating transformers, right?

PS. Creative engineering and technological abuse indeed!

[tiresias]

Re: Watercooled Powersupplies

I've seen a picture of a power supply being placed into a container filled with castor oil (with the mainboard) in a computer magazine. This was entirely safe(!)

Apart from the risk of acute exposure to the oil giving one explosive diarrhea - it should be safe, it's a non-conductive ester after all.

Does sound a little messy though.

[Rainbow]

Re: Watercooled Powersupplies

That particular one is not only bad, it's actually, from a computer's point of view, hazardous (and I'm by no means a safety freak, trust me!). If you look carefully, you'll see he/she has attached both the MOSFETs from the primary 300Vdc side and the diodes from the secondary, low voltage side to the same copper plate!

The only thing preventing the two making contact and sending all attached components to the "other side" are two thin layers of TIM, which have been disturbed (removed and reattached) at least once already.

Kind of defeats the point of having isolating transformers, right?

PS. Creative engineering and technological abuse indeed!

This is true for the second PSU too. To increase safety, you need two separate water blocks - and the one on the high-voltage side must be insulated from the water. To be really safe, don't use any water in PSU.

[Chris1992]

Re: Watercooled Powersupplies

>To be really safe, don't use any water in PSU.
Absolutely right. =D

[Per Hansson]

Re: Watercooled Powersupplies

Hmm, you are right about that tiresias, however that is true with all watercooled PSU, since even if the blocks where separate current can still very easily be led throught he water instead...

If I built it myself it would be one big U block milled out, with water canals in it... Think of a CPU block bent twice 90°, of course much larger LOL ;-)

To be really safe, don't use any water in PSU.

And now where is the fun in that :P

[tiresias]

Re: Watercooled Powersupplies

Hmm, you are right about that tiresias, however that is true with all watercooled PSU, since even if the blocks where separate current can still very easily be led throught he water instead...

True. Two solutions that come to mind could be:

1. Either use a non-conductive, non-flammable (and generally non-unpleasant) fluid with an acceptably high specific heat capacity.

2. Use something other than metal to attach the FET or diode to. Some ceramics can be very good heat conductors - one could use this in place of the standard, thin TIM layer.

[Per Hansson]

Re: Watercooled Powersupplies

2. Use something other than metal to attach the FET or diode to. Some ceramics can be very good heat conductors - one could use this in place of the standard, thin TIM layer.

I know of no other material that I can easily obtain and still be able to cut out in a mill...

[Galvanized]

Re: Watercooled Powersupplies

Per, someone on another board PMed me this link. Thought you and others might like to see what it has to offer. So far, no pics with the hood up.

http://www.xtremesystems.org/forums/...ad.php?t=98507

[jonnyGURU]

Re: Watercooled Powersupplies

Hmm... Looks like I've got a new project to tackle!

I'm still going to use a fan, though. Just a really slow one. Almost all of the components inside of the PSU are going to generate heat and even the water blocks are going to radiate some heat not absorbed by the water. I figure a single 80MM spinning at about 2000RPM or so should suffice.

[Galvanized]

Re: Watercooled Powersupplies

Just brain stormin' Per. I love mod projects. I've amazed myself with what's been done by me as well as others with limited tooling.

Distilled water is non-conductive, many additives are also, iirc, Water Wetter is one. Much has been posted about coolent condutivity, additives to use and the use of dissimilare metals in the cooling loop. Extremesystems forums would be but one source.

For a non-conductive block some type of plastic with good thermal expansion properties could be machined to accept two copper plates. The coolent chambers could be joined by cross drilling. The plates could be bonded inplace with silicone adhesive. The stuff I've used only rates to a high of 150F but there are higher rated types. Once the plates are inplace, the block would need lapping.
For mounting the block two T shaped bits of extruded aluminum or copper would be used. Inorder to have a dead flat mounting surface, these would have to be joined with some type of non-conductive brace prior to being installed. OTOH, The blocks could be seperate and the coolent fed to the other by a short hose.
For clamps, consider the type offered by http://www.abaofamerica.com/

It should be noted that the WCed PSUs pictured in the above links only have a single inlet/outlet. I want to see them opened up. :

The MGE I linked above does not have the two sinks joined. So tiresias has a very strong point about that not being done.

[jonnyGURU]

Re: Watercooled Powersupplies

I'm thinking of just machining the whole block and replacing the stock sink with the block. I would extend the block all of the way to the inside edge of the power supply and then have the fitting go through the housing. No hose.

That way the block doesn't leak and if it leaks at the fitting it drips down the side of the PSU and not inside the PSU.

The only problem is getting the water from one block to the other without each block having it's own in and out. I wouldn't want to have a PSU with four lines going to it. Two is enough. I'd rather not use a hose, but if you make it one big water block (U shaped so it goes in and across one set of FET's and then comes back out running across the other set) then you give yourself limited space to work with.

An interesting problem.... Let's see what happens.

[Per Hansson]

Re: Watercooled Powersupplies

It is better to have two separate blocks, because then you do not risk sending 230v AC to your computers 12v DC supply as easily, because the current must go through the mostly non-conducting destilled water to reach the other block...

But anyway I've realized myself that I do not want a watercooled powersupply, because the gains are so small compared to the risks...

[jonnyGURU]

Re: Watercooled Powersupplies

That's a thought... but the FET's should never ever ground to the sink no matter what. At least if they did, protection should kick in.

What's worse? AC to DC short or water dripping on the PCB? :o Hmmm.....

[jonnyGURU]

Re: Watercooled Powersupplies

Ok... my friend the PSU engineer told me that if an AC FET or DC FET shorted and grounded to a sink and it then in turn shorted a sink that a DC or AC FET was attached to, you'd pretty much destroy everything in your computer.

Sounds like two separate sinks then.

[tiresias]

Re: Watercooled Powersupplies

Two sinks will certainly be the safest option in that regard - particularly since attaching the devices to the sinks will invariably involve some 'fiddling around' which could further disturb/damage the already thin and small isolating TIM layer.

DC or AC FET

What do you mean by DC or AC FET? Usually a PSU only has one kind of heatsink-mounted FET - on primary-side heatsink that is, and these are part of the inverter which 'turns' 300Vdc into square wave high frequency AC.

I've never seen any FETs mounted on the secondary side heatsink - usually these are just rectifiers. (double-diodes, as it were)