How important is it to apply a distributed compressive-load over a TEC module?

[L33T]

In the course of every design there comes a time when it is back to the drawing board because you think there might be a better way.

That is the case right now.

I'm using an MCW5002. The hotplate has been modded to allow TEC modules to fit. Specifically, I plan on using the 437W TEC.

Here are my options:

OPTION 1. Use only the compression from the retention mechanism used to hold the entire water block against the CPU. Cold plate doesn't have to be very thick; 0.25" or 0.375" will do.

o PROS:
- Cold plate will only be as big as the hot plate.
- Cold plate won't need any additional milling (see CON of option 2).
- Less Cu needed than OPTION 2, so cheaper than OPTION 2.
- Easier to insulate than OPTION 2, since the cold plate and the hotplate will be flush.
- Allows for a thinner cold plate, since there will be no bending moment, as is in the case of OPTION 2.

o CON:
- Possibly not enough compression.

OPTION 2. Use a custom-designed cold plate that is as big as the retention mechanism, with cut-aways to fit the SM capacitors. Cold plate must be at least 1/2" thick (see CONS).

o PROS:
- Excellent compression.

o CONS:
- More difficult to fabricate than OPTION 1.
- More expensive in terms of materials needed, albeit the difference is not substantial and should not be used as a consideration.
- Bending moment situation dictates that the cold plate must be thick enough to minimize yielding -- hence, 1/2" thick. The actual material sustaining the moment will be much thinner than 1/2" thick... it'll be probably half of that.. perhaps 1/4" thick.
- More difficult than OPTION 1 to insulate, since the hotplate and cold plate will not be of the same dimensions.



Pic of OPTION 2:
In the design, I've included an extra hotplate that is sandwiched between the water block's plate and the TEC module. In the same orientation as the drawing, the layers of the entire object looks like this:

[ Cold plate * ]
[ TEC 62mm x 62mm ]
[ Hotplate ]
[ Water block surface ]
[ Water block housing ]
[ Water block retention plate ]

Notes:
* Must be milled to look like a two-layered pyramid; without this, it will not fit, due to the capacitors around the CPU. Yes, I have measured.

As you can see, the outermost dimensions of the cold plate are the same as those of the retention plate.

[STEvil]

You may be able to drill holes in the mounting plate and skip using the hotplate if the tec fits the base of the waterblock.

Also, think of the situation regarding the compressive load: If you do not torque the bolts evenly you will effectively only receive a fraction of the cooling power on the side which is "loose" and may cause the opposite side to fracture thus destroying the tec if it is overtight or angled.

[L33T]

You may be able to drill holes in the mounting plate and skip using the hotplate if the tec fits the base of the waterblock.

Yes I almost forgot about that issue. There are "holes" for screws to mount the base of the water block to the housing. I suppose applying AS5 so that it fills these holes up will certainly help, but I am not sure that is best, seeing as the 437W TEC is quite hot on the hot side. Now even if the TEC will be covering several of these screw holes, will the hot plate be necessary, if AS5 is used copiously?


Also, think of the situation regarding the compressive load: If you do not torque the bolts evenly you will effectively only receive a fraction of the cooling power on the side which is "loose" and may cause the opposite side to fracture thus destroying the tec if it is overtight or angled.

Yes I know that's quite a risk, but I'll try to prevent something like that from happening by turning them each a little at a time.... though I'll have to sort of keep track of what angle I've rotated the screw, so that I know how much compression is applied to the TEC.

Now I recall an article written by somebody regarding how much of a compressive load is optimal, and how many turns of a screw that translates to, given the threading, &c. For some reason, I think it might have been on [H]|F, but then I again, its quite unlikely that [H]|F would think of something that XS hasn't. In any case, its quite easy of a calculation to just figure it out myself.

[STEvil]

The holes might not matter given the size of the diodes inside the tec. Surface temperatures of the ceramic may only rise a few degrees celcius a most as each diode only handles about 2 watts each and then must transfer it through the ceramic (which isnt a very good conductor of thermal energy anyways).

Be sure to lap all surfaces and apply only the most minimal amount of thermal interface material. The mounting surfaces will affect the efficiency of the system more than mechanical mounting pressure if you apply the correct amount of TIM.

[L33T]

I've read this same post for a few days so it would sink in ^__^

Okay: when you say "lap all surfaces", do you mean the TEC module surfaces as well? I've read that some people do lap them, but if unnecessary, I'd rather not risk breaking off the leads.

Basically, are you recommending that I simply "float" a cold plate, so to speak?

[Holst]

I dont see why you would need to lap the TEC, if itsnot flat something is very wrong with it.

Have you ever tried to drill or file a TEC, that seramic is very tough!

I think its very important to clamp the TEC, with it clamped propperly you will definately get better performance.

However I appreciate that with these big TEC its very awquard to fit everythting in. Getting the coldplate to fit is a struggle... and then you have to insulate the damned thing.

Personally I would probably try without clamping and see what the performance is like.
With an IHS on your CPU you should be able to crank down a reasonable ammount of pressure onto everything. And make sure you only use a very very very thin layer of TIM.
What I might do myself is to force the block, TEC and coldplate together with a heavy weight (or some other method) before you fit them to the CPU. Suction should keep the assembly together pretty good if your carefull.

Let us know how you get on and good luck.

[n00b 0f l337]

Yeah no need to lap the TEC.
Holst! Good to see you back!

[STEvil]

I would give the tec a slight roughing on very used high grit paper just to be sure.

I had an 86w tec that looked like a checkerboard after rubbing it across a couple times... havent checked my 320w yet.

[L33T]

Hmmm glad to hear the advice... ok then I guess I will briefly pass the TEC over high grit paper, but not lap it... Then I'll use the "suction" method. I'd rather not use the clamping design... it is a lot of work to make the insulation "fit".