Water cooled PSU
 

I'm going to make waterblocks for cooling my PSU ,now I need confirmation wheter cooling heatsinks in PSU is enough or do I have to cool anything else.

I've wondered that myself. I've got an old copper block not doing anything. The resistors in the supply will generate a good deal of heat as well, but I'm not really sure if that would be enough to cause ptoblems.

One way to find out would be to run an old supply without a fan for a little and touch the parts inside to see how warm they are. Of coutse if you didn't discharge it completely and touched the wrong thing you'd be in for a lot of pain.

Anyone ever taken thermal images of a power supply?

Watercooling the PSU

If you read the thread above you'll see where I'm at atm. As the current air cooled sinks are live It would require an isolation solution to prevent the coolant becoming live if the sinks were just replaced with blocks.

I outlined a solution I'll probably go with in the thread using a ceramic or glass isolation spacer, but I still want to know a definate answer as to what reason there is for making the sinks live?

Koolance also make a water-cooled PSU but I'm not sure how good it is.

Thanks BladeRunner for pointing some things, I totaly forgot to think about.I also read your thread on bbs. I went to friend of mine who know more of this stuff, and we open one PSU and took a look. You've got two heat sinks, one on high voltage AC part of the PSU and one on the low voltage DC part. If you saw the transistors on AC sink are all isolated with silicon, and the sink isn't live. And on DC side, two of the transistors are conncted to the sink and it act like a mass for them. So I would isolate all transistors on DC side and the reconnect two transistors with piece of wire and solder. Now the problem would be if we connect both sinks with water in the block, because if some sort of electrical surge would pass isolation we get 240V to DC part and then to whole computer.So some sort of insulator should be used, like you said ceramics, glass, I looked at thermal conduction coeficients and find that marble has 2.8W/m*K, it shouldn't be to difficult to find nice flat piece ,ceramics may be better, I couldn't find thermal characteristic for them , and I don't know where to search for flat piece of one. And for even better safety I'd use Alu. for water block and then annodize it for isolation , then ground all the blocks and water in my computer, and the last thing, I'd use surge suppression .

http://www.fortunecity.de/wolkenkrat...sh/english.htm

Here is one.

The only problem with cooling a psu with water cooling is your going to be introducing a very large ammount of heat into the h20 system, which means instead of having the heat blow out the back of the case as most psu do you are making that heat go into your water loop which would raise water temps all over aka the orginal koolance cases with water a cooled psu. Also I would assume that sence the power supply doesnt directily have anything to do with signal flow (other then providing power) that it would not allow one to have any gain in performance.

just my 2cents on it, interested in hearing results if you do end up doing it

inda

inda

It depends, I'm using an Enermax 550 watt which is overkill for my system, but the plus side is it has a walk in the park powering all my hardware so never gets more than warm, even under load.

My cooling "radiator" is well up to absorbing any additional heat. My reason for wanting to watercool it is to be able to remove the fans not to gain any more performance.

morphling1

The larger sink in my Enermax 550 watt is live 240v, but I still can't see why it needs to be, wish I could find a definate answer for this as it's holding the job up now :(

I used voltmeter on my 300W PSU and heatsink wasn't live, and it should't be. Check rubber pads on transistor.

I think it must depend on PSU or Type. Two others I have are also live and the sink on the Enermax is directly solder connected to the live part of the PCB circuit in two places. some other older ones, like yours are not live and have sinks mounted or connected to the earthed chasis. I'm pretty certian it's meant to be this way, just like I said It must depend on PSU or type. I just need to know why, and if it can be made non-live without effecting the operation, as this would be a far easier watercooling solution.

The heatsinks are connected (more or less) to the back part of regulators. Generally that (metallic) back part is either standalone - therefore needs to be connected to ground - or connected to the middle pin (all regulators have 3 pins).
You should consult databooks to find out your regulator references. The databook will tell you how is connected the back part.
I think most seminconductor companies have their databooks online, if you dont have access to any. If you're stuck with your references post the brand name and the exact reference here, maybe someone (including me) will be able to answer...

[ 09-28-2001: Message edited by: SureFoot ]

I know what I'm going to do. I find somebody that does hard coating, that's method of protecting Alu. with thin 30-200 mikro meter layer of ceramic like coating that is very hard 300-700MHV, and of course totaly electric uncounductive, so that way water in block will be insulated from any electric current. That should do the trick :)

Perhaps what I'm trying to explain here has got lost in the thread or the way I'm trying to explain it :). Whatever the connection of the transistors, (src's), it has no relation to the heatsink in my PSU. All I can tell you is the metal part of the sink is live mains, (240v), when the PSU is powered but the three pins and the metal back part of the transistors are not directly connected to the sink electrically. They might be connected via the PCB but they are completely isolated from the sink its self through contact. They are isolated via the membrane from the metal of the sink. It is therefore just the sink it's self that is live which is why It doesn't make much sense to me.

http://www.zytec.worldonline.co.uk/p...suwc1small.jpg http://www.zytec.worldonline.co.uk/p...suwc4small.jpg

I've been trying to find a definite answer to this in a few places, but although I come across a few suggestions no one has said. "it's almost definitely because of X"

http://discussions.hardwarecentral.c...ML/016106.html

http://www.lostcircuits.com/discus/m...html?997437264

I suppose I'm going to have to go with isolation via ceramic's like morphling1 as that would keep it all the same (live sink - live spreader plate - ceramic - (isolated) water-block).

I'm fully never happy when I don't understand the reasons for something, I guess it's just my nature, thanks for all the help folks ;)

Might be merely EMI. Did you try to ground it via a high-powered resistor ? Try it and read the current going through (should be AC too..).
If there's a lot of current then they're *really* connected to 230V. For instance if you put a 10K resistor you should read about 23mA. (That's 5.3W of power so pick up a 7W or higher resitor)
If not (=no current or very low current) then it's EMI for sure. Put a wire to ground them in that case.
Check back here with the results :p

[ 10-10-2001: Message edited by: SureFoot ]

My thoughts on all this:

Why not simply take an old CPU HSF, pull the fan off, and attach it to a spreader plate for the PSU heat sinks? You should be able to increase the radiation area enough to dissipate(sp) the heat without a fan.

But, I haven't had a chance to try it yet... I'll let y'all know when I do.

(First post! yay me. Longtime lurker, finally have an opinion worth posting about.... )

Congrats! Post more and let's fire this place up a bit :).