WB-design for direct pelt cooling
 

http://users.pandora.be/computerwackos/WC_bovenkant.jpg



Could you plz giver some commnts on the design?
It is merely a rough schets but enough to giv you a general idea.

My goal is to built a pelt into an waterblock, so i have direct cooling, which is more efficient.
The whole inside area presses on the pelt, so normally i shouldn't have any problems with low pelt pressure or so if i use a coldplate of 3-4mm or am i wrong with that?

Greetz ZaVkE

I think it would be better to split the tube like and do this. but you can do something like ]this or the following
http://dosage.diablo2gear.com/trash/f.jpg

that won't work terribly well, just get rid of the channels and have an open block design

I can't get rid f all the channels because i need something to press the pelt against the bottom, otherwise i have a large descent in performance of the pelt.

My design was done because of the high flow rate of it, so the hot water will be out of there quicker.

Maybe an approach with a larger channel, but still my desing would be better?

Nice idea. I feel that you should have the chanel walls a little lower so when you put the pelt in and then the base would just cover up the whole inside. But take note of the allowance for the wires and seal the wire connections very well so that no water can get to it. Try to make the chanelwalls as thin as possible so that you can have more surface of pelt to water while still applying pressure on the pelt.
But again, there's a problem to this. Soldering the base which is the coldplate to the shell, which is the hotside, there will be quite a signifcant thermal leak from cold side to hotside.
Epoxy maybe better, but then again, how insulative can a thin layer of epoxy be?

One problem I have encountered with epoxy; it doesnt like thermal stress. Sure it works for a while (meaning anywhere from 2 - 60 days) but eventually the epoxy will crack, warp, or just loose its bonding properties. I have had it happen to me 3 times with my old res. Used epoxy to seal the inside and outside of a bottle but after a while of running my water throught it (the temps going up and down) the epoxy gave way and leaks happened.

I could be wrong in this situation however I should point out that there is a risk.

Epoxy won't be the best solution, any Pelt requires heavy clamping to work at its best, and this can only be done, ironically, with metal screws.

'tis best to clamp, then isolate.

That WB design (1st one) is best for cooling a core, as it receives cold coolant in the middle, in line with the CPU core.

The channel design is best for Pelts, because the heat is spread out more evenly.

You must work around the channels, by making a fat corner for the clamps to the cold plate. Where there's a will, there's a way.

I think you'll have to sacrifice some clamping pressure if you want to direct water cool the pelt. I think the reason why pelts work so much better with high clamping pressure is because the many many little metal things inside that transfers the heat across dont have such good thermal contact with the ceramic plates to begin with. And with so many of them spread across the whole pelt, applying pressure just at the perimeter would not have as good an effect as applying pressure throughout.
That said, you have some rather spreaded out pressure with that chanel design so it shouldn't be too bad. Perhaps like they said, epoxy isn't such a good idea, but some form of insulative barrier should be implemented.
Perhaps you should have the shell with the same height throughout so that when the coldplate comes on, it isn't in contact with the shell. Then you can use screws to tighten. As for preventing water from leaking out, maybe silicone or plumber's goop. Actually what I'm describing is much similar to Tom Leuken's block except with your chanel design + forgoing the rubber gasket, and a cold plate clamped on.
All IMHO.

Maybe i can just make the top block a bit bigger and the coldplate also. That way i can drill 4 or 8 holes in the edges of each and use screws to clamp them together (with or without isolating material between the block and the coldplate, the idea is the same) and then use the 4 outside holes to hld it on the MB.

How about that idea?

Yes, that was what I was talking about, just make sure you have spreaded clamping pressure over the peltier.

And what is i drop the channeldesign and go for a open design but supported with round vertical bars (eg the alpha pal aircoolingsinks, but far not as many) or will the channeldesign work better?

It just because i want to restrict the flow as less as possible because i heard that that can really be a problem with a channeldesign.

Thats a nice idea too, but bear in mind the more copper in contact with the pelt, the less direct watercooling you're doing. I would opt for pins spaced evenly and spaciously apart, since the main thing here is directly cooling the pelt with water.
With them spaced generously apart, there shouldn't really be much flow rate problem. And try to make the chanel depth shallow.

my idea exactly

For those still in doubt about clamping, read this:
http://www.procooling.com/articles/h...just_how.shtml