[BrianH2O]

It seems that the Swiftech MCW50-T, a GPU cooler with an 80w pelt, runs the water diectly over the pelt. If this is the case, then a new generation in WB design for peltiers may be coming.

A test over at kconfig.com shows excellent results as
Stock Heatsink (324Mhz - stock)
Room: 19-21°C
Idle: 41.5°C
Load: 49.5°C

MCW50-T Block (420Mhz - overclocked)
Room: 19-21°C
Idle: 5.8°C
Load: 10.5°C (using Unreal Tournament 2003 / 3 Hours Play)

Notice the 420mhz OC on the 9700pro running at 10.5C load!

There are may consideration for type of design, i.e., turbulence, flow speed, leak prevention, etc.

I would image the Rh would be quite low in Kryotherm.

Any thoughts?

[Paxmax]

Yeah, it seems to be very waterproof, unless you get a crack in it.

The tough case with this solution is low surface area. It can only be compensated with higher water speed across surface. Higher speed means higher restriction which leads to lower flow. Another matter with TECs is that the coldside needs to be strapped HARD to the coldplate. What happens when the TEC is only supported along the outer rim? My guess is that it will bend and trap paste while trying to squeeze it out. One might gain alot due to the lack of TIM on hotside, but loose a little by lower pressure on coldplate.
Still, I'd like to try direct die TEC cooling too. I think I will support the TEC here and there at the middle aswell. I don't wanna crack my pair of 226 W TECs.
Leufkens has been selling direct die TEC WBs for a while(one or two years maybe). I haven't seen any tests yet though. He clearly states something like "A HIGH flow pump needed, minimum 317 gph".

So PRO's: No TIM on hotside, big gain. Very short heatpath.

Con's: Low surface area compared to a WB, high flow pump or high pressure pump needed.
Lower pressure on coldside might create a bigger TIM C/W
Possible bending of TECs, lower life expectancy?
The sealing ring (or whatever) steals a bit of TEC surface?
If the pump stops you have to react quickly before water boils off.

An idea, or the next step would be to strip the TECs hotside plate. The white thingy, alu-oxide perhaps?
Then insulate half the height of the thermocouples (on the coldside part) inside TEC and submerge the other exposed half!!! Now we have stripped away another C/W stealing component. The portuding blocks would expose more surface area than what's in contact with hotside plate. Though this field of use would be very limited, so we will never see it. A thought though.

[BrianH2O]

Thanks for the response Paxmax

I have been working on designing a direct cooling system for the TEC's hot surface....

Here are my thoughts

The block would be 60x60mm and would be created in 2 parts.

The bottom part would be a standard cold plate say 15mm think. On the inside, a 50x50mm recessed area 3mm deep for the TEC. This is slightly less than the thickness of a regular TEC. Of course, channels for the TEC's wiring would also be provided. On the outside, cut away a recessed edge of 10mm all the way around. This would leave an area 50x50mm for the actually heat transfer from the CPU.

The top part would also be 60x60mm in size. The inside would be 45x45mm and 5 to 10mm deep. This would allow an overlap of 2.5mm over the TEC. This is where the gasket would fit being 60x60 outside with a 7.5mm width. The top of the block would have 1/2 barb holes at each corner.

Now here is the fun part. Tap screw holes for say 2.5mm hole thru to the top of the block. We would end up with around 20~25 of these holes. Insert flat bottomed, no head screws and tighten against the pelt. These copper screws would has 3 benefits

1) Help to clamp the TEC in an even and distributed pressure to the bottom cold plate with the edge also clamped
2) Help with the turbulance of the water
3) Help to radiate the heat up the screws giving more cooling area.

I do not think that this block would be that hard to build.

Any ideas or suggestion ie I am total nuts here (lol).

One of my first questions... Does the bottom have to be 15mm thick? Or could I go for 10 or even 5mm

Thanks

[Aswethinkweis]

your idea for the block seems pretty sound, i think i have a few ideas though, im gonna make some quick drawings, cause i think they might get across better than words

[Aswethinkweis]

well heres the drawings i whipped up, not the greatest, but i think its what you were getting at... keep in mind i drew these in about 20 minutes :P ill make an asembly drawing as well

heres the top "cover" type part of the block:



heres the retention ring that would fit between the top part, and the coldplate:



heres the cold plate, where the pelt would sit:



hope the drawings make a bit of sense, im sure the complete assembly drawing will help out when im done, but i got some stuff i wanna do before i do that drawing

[Since87]

Brian,

I believe around 6 - 7mm is optimal for the coldplate thickness.

Can you buy copper screws in the size you want? (copper allthread?) When they were at the right height for supporting the TEC would you solder the screws to the top? Or were you going to seal the threads some other way.

Sounds like you are on the way to a fairly good design.

Another possibility, is to look into the microchannel parts that Neomoses is selling in the classifieds. (You'd probably want a different fin to channel width ratio for this. But he says he'll do custom.) Put the microchannel section into the block "upside down" to provide the mechanical support, and give even, high velocity waterflow over the hotside. If you lined things up right, you could get the copper fins right over the TEC's "couples". (Those cube like things you can see inside an unpotted TEC.) That would be best mechanically. Thermally, I'm not sure.

I haven't thought this through very well. Your idea may well be better for a variety of reasons. Just thought I'd toss the idea out there. A key issue is to get high velocity water flow over as much of the TEC's surface as possible, and have a high pressure contact with copper everywhere else.

[McWarren]

How about this rough idea.




A simple idea and pretty easy to construct (easy=I can build it) Use the regular bolting method to apply the required pressure without having to worry about leaks. In fact, a higher pressure would make a safer seal.

[BrianH2O]

I was thinking of this design attached below...

As for the screws, I see tham as counter sunk into the top and then filled with a "removable" filled such as Plimber's Goop.

I am not sure if they make copper screws. I have yet to see them, but then, I have not really looked yet. How hard would them be to make?

The upside-down microchannel may also work. The up side (pardon the punk), it can be made right into the block without any screw holes to seal. The down side is getting it flat with the other edge... may be a better idea after all...


How do I attach an image to this forum?

[Since87]

Originally posted by BrianH2O
I am not sure if they make copper screws. I have yet to see them, but then, I have not really looked yet. How hard would them be to make?

I checked McMaster-Carr and they don't seem to have any threaded copper. I don't have any idea how hard they would be to make. I do know that tapping holes in copper is a real pita. I've got a Maze2 here I bought used, that has a tap broken off in one of the holes. I'm guessing DD was intending it to be a Maze2-2 and gave up on that idea when the tap broke.

You could use brass screws, but you only get about 30% of copper's thermal conductivity.

Originally posted by BrianH2O
The upside-down microchannel may also work. The up side (pardon the punk), it can be made right into the block without any screw holes to seal. The down side is getting it flat with the other edge... may be a better idea after all...

Yeah, keeping the two surfaces flat was the drawback that jumped out at me too.

Originally posted by BrianH2O
How do I attach an image to this forum?

Huh. Just noticed the attachment option is gone. Maybe OC has run out of disk space? Maybe someone with web space will offer to host an image

[Aswethinkweis]

email me the image and ill host it for you shikeieq@hotmail.com

[BrianH2O]

Many thanks to Aswethinkweis for hosting my pic!

Here is my pic/thoughts... remember not to scale on any thing



The clamping screws wiuld come thru the top and connect to the Pelt surface so that is can be screwed down tight.

We could use a brass screw with a copper "boot". I would tend to avoid alum. because of the"battery effect".


Another possiblity is to use a 3 piece setup as described by Aswethinkweis. Then we could have a bottom as I described, a ring as Aswethinkweis described and a top panel that could be machined like the inside of the base plate of the swiftech MCW5000 (.... scroll down for the base plate). Of course, the lining up of the tips with the top of the pelt would also be hard to achieve.

So we would have
Top plate
Gasket
Spacer ring
Gasket
Bottom plate

Any thoughts again?

[Since87]

One issue that occurred to me involves the recessed area in the coldplate for the TEC.

How do you get the recessed area flat. You can't lap it. Without lapping how can you assure flatness?

Another issue is; how do you you insure that the waterflow is spread evenly over the pelt surface? Why would water ever travel to the corners without barbs? (I certainly don't know the answers. Just bringing it up as something to think about.) I imagine the right arrangement of pelt pressure screws could give you fairly even flow distribution, but figuring out what the ideal locations are is way outside my area of expertise.

[BrianH2O]

figuring out what the ideal locations are is way outside my area of expertise.

and mine as well....

Maybe the screws with a copper boots could be positioned to "directly" the waterflow a bit to the corner's(???).

Who laps?? Fill the area with AS3 to fill the gaps. Worked (almost) for ATI's 9700pro....

Just kidding. Perhaps the bottom should be flat (or even raised) with a square ring to raise the outer edge to just under the peltier's thichness. No gasket would be required between the bottom plate and the ring....

Also, have I made the block to small? I wanted it to be as small as possible for weight and to be able to insulate it well.

So new propsal is
Bottom plate
Square ring
Top block with pressure screws with cooper boots/ hopefully able to direct waterflow to the corners.

[Aswethinkweis]

you would be able to get the recessed area VERY flat, there might be a few imperfections, but you could use a finishing end mill, which would leave virtually no machine marks, it would be just about if not smoother than most stock heatsinks, right now im also working on another design, this idea really caught my attention, and as soon as i get the new one finished (which might take some time, im kinda teaching myself this program as i go along) ill post it here

EDIT: about the copper screws, if you really wanted them to be pure copper, and had about 20 minutes per screw to spare, you could make them

EDIT2: heres a copper alloy screw from mcmaster, ill look around for a thermal propertie sheet on the alloy *runs and grabs his machinerys handbook 26th edition*

[BrianH2O]

Thanks again Aswethinkweis

I found alot of info Here including thermal characteristics. Over my head I think.

But, it should be ok with a copper boot to conduct heat.

I am really starting to like the lower part ring. It could be made to different heights depending on the thickness of the peltier for a better seal.

about the copper screws, if you really wanted them to be pure copper, and had about 20 minutes per screw to spare, you could make them

I do not have any tools to make this stuff. I want a design and then I will visit a machine shop for estimates.

[Aswethinkweis]

i have minimalistic capabilities for something like this, i might be able to make some of it in my house, and the stuff i couldnt make i can run down to my uncles company and machine the parts over there (although thats kinda sketchy, hes very pressed for spare time, etc) I really want to get a [seemingly] good design, and actually make it happen to get some results

[BrianH2O]

OK I have a more accurate pic of the block. I have
1) reduced the number of screws thru the surface to 7. This shiuld make it easier to waterproof
2) Added copper boots to all the screws and taper some of they to encourage flow to the corners
3) Added the ring along the bottom

Hope this looks good...

[JFettig]

if you guys are gonna make one of these, remember, you want the water to go in and out like kaboom kaboom, no sitting around, make these as small as possible without loozing flow

[BrianH2O]

JFettig

The internal volume of the block is
4.5x4.5x.75 = 15cm3

minus all the internal blocks ... 6 of .5*1.0*7.5 = 2.25cm3

Therefore internal volume should be around 12.75cm3 which is .7778 in3. Is a in3 of water == to an oz. of water? if so, then...

Is a 1/2 ID going thru a barb powered by a ViaAqua 1300 rated at 395gph enough for this?

The pump is 395gph == 6.58gpm == .11gps == 3.5 oz/sec.
Therefore 3.5/.7778 = 4.5 changes of the internal volume of the block every second.

Of course, this is all theory (and it may be bad...). The real number would be less because of restriction to the flow. What is a reasonally number to use and what is required?

I could also reduce the height from 7.5mm to 5 mm if it would help?

With 5mm new internal volume is 8.6cm3 ... around .53 in3. Therefore 3.5/.53 = 6.6 exchanges per second.... humm ... thinking now...

[mulletron]

Without the mad mechanical skills you gentlemen are proposing-I got a Leufken block awhile back, and love it. Currently on XP1800 (2.1 v), 1/4" plate and 226 TEC. Using 1/2 ID tubing and Via Aqua 1300. I get 12*C idle and 19*C load, with ambient of 28*C. The flow is high, only restricted by 2 h/c. Thru numerous disassemblies-noticed no warp or inconsistent thermal grease patterns between plate and TEC. Im using a Lexan clamp, steel bolts and my best guess as to all the pressure it can stand! I think the concept works, good luck.