Flat Cold plate for TEC not optimal?

[jaydee]

It was said by someone in another thread:

P.s. Flatness of cold plates is one of the reasons why ther are not performning as well as they could

I asked for an explaination but because of previous bickering he declined. For the last 4 years of design and construction I have always been under the impression that a TEC was fairly flat on both sides. I was also under the impression that to get the optimal performance of the cold plate the flatter the better because it will give more contact area between the cold plate surface and the TEC surface. Being it is highly unlikely TEC's are going to change design then this statment leaves me confused.

So, if I am incorrect in this, then what?

[Jabo]

Oh, you are as always perfectly correct.
I just followed what is written under your nick, that's all.
I concluded you are not interested in explanation since theory and design are worthless waste of time.
Since you want to her explanation afterall I will be glad to provide it.

It is siml misunderstanding and me employing some shortcuts in expresing myslef thus not bing orecise and clear about what I wanted to say.

Surfaces of a cold plate in contact with TEC on one side and CPUs die on the other should be as flat as possible since thermal energy source is flat and planar.

I was thinking about SIDES of cold plates and 3D shape of them.
Apologies for inprecise thoughts expressing.

[chrisbmx68]

so your saying tha tlike a coldplate with rounded/chamfered sides would be more thermally effiecient than a rectuangular /square coldplate?

[jaydee]

Originally Posted by Jabo

Oh, you are as always perfectly correct.
I just followed what is written under your nick, that's all.
I concluded you are not interested in explanation since theory and design are worthless waste of time.
Since you want to her explanation afterall I will be glad to provide it.

It is siml misunderstanding and me employing some shortcuts in expresing myslef thus not bing orecise and clear about what I wanted to say.

Surfaces of a cold plate in contact with TEC on one side and CPUs die on the other should be as flat as possible since thermal energy source is flat and planar.

I was thinking about SIDES of cold plates and 3D shape of them.
Apologies for inprecise thoughts expressing.

Ah, I see. I am working on the cold plate of my TEC block and in fact it isn't just a regular rectangle plate like many use. I think there is something to what your getting at but I need to run some tests to make sure.

It seems to me the more extra copper in the cold plate the more area for secondary heat/cold loss. If you make your cold plate as small as possible and romove as much "fluff" copper then it would perform a little better by removing secondary losses (kinda like making a good die simulator). A standard TEC is 50mm square (or 40mm), so why would you want a cold plate much bigger than that if the CPU die area is less than that (other than mounting considerations of course)?

Also it isn't that I am not interested in design and theory, I am just not interested in theory that cannot be implemented and substaniated and turned into fact. All my design are based on theory just like everything else. Only thing that seperates my theory is I actually test it with a working prototype to make sure my theory is correct. I try to take my theory, make it happen, and then talk about it. Theory holds more water if you have something "actual" to base it on.

There is TONS of misinformation running around the water cooling forums because people spread their "theory" around with out anything to back it up. Causes considerable problems.

But I digress. Sorry for being a pain in the ass.

[chrisbmx68]

i just ordered a maze 4 and i am holding off on getting their coldplate but i dont really see how much it could be improved upon?

another question would it be benefficial to sanwich neoprene/or other insulation between the coldplate and the waterblock or will it just melt?

[jaydee]

Originally Posted by chrisbmx68

i just ordered a maze 4 and i am holding off on getting their coldplate but i dont really see how much it could be improved upon?

another question would it be benefficial to sanwich neoprene/or other insulation between the coldplate and the waterblock or will it just melt?

The Maze 4 TEC setup looks about as good as it will get I belive. Not much room for changes.

DD sells a neopreme kit for their TEC setups. You do want some insulation in there.

[Brians256]

Originally Posted by chrisbmx68

i just ordered a maze 4 and i am holding off on getting their coldplate but i dont really see how much it could be improved upon?

another question would it be benefficial to sanwich neoprene/or other insulation between the coldplate and the waterblock or will it just melt?

I'm not qualified to say what improvements should be made on the DangerDen coldplate.

However, putting neoprene between the waterblcok and coldplate is just ... I can't say. Are you sure you typed that right? Putting any insulation between the heat source and the method of cooling is counter-productive.

Now, insulating the whole assembly by putting insulation on the outside of the stack (CPU, cold-plate and waterblock) can help prevent condensation+frost and it can help reduce temperatures by blocking parasitic secondary heat paths. Was that what you meant?

Or if the cold plate is smaller than the waterblock, you can put neoprene insulation around the edges of the cold plate (so the neoprene insulates the waterblock from the CPU that could be colder than it), then that also makes sense.

[jaydee]

Originally Posted by Brians256

Now, insulating the whole assembly by putting insulation on the outside of the stack (CPU, cold-plate and waterblock) can help prevent condensation+frost and it can help reduce temperatures by blocking parasitic secondary heat paths. Was that what you meant?

Or if the cold plate is smaller than the waterblock, you can put neoprene insulation around the edges of the cold plate (so the neoprene insulates the waterblock from the CPU that could be colder than it), then that also makes sense.

I just assumed that is what he ment. I don't think I am wrong. I hope.

[Butcher]

Well generally you do put an insulator there brian - air. He's talking about the gap around the TEC between waterblock and coldplate which is generally left empty.

I'm not sure if air is better or worse than a layer of neoprene off the top of my head though. I suspect the comparison is not clear-cut anyway due to the different states of the two materials.

[chrisbmx68]

im aware of how to insulate the socket etc im running a swiftec 4000-t atm. (modded to fit my amd board) Im stealign the tec out of that setup and using the maze 4 and cooldplate and i did mean the gap between the coldplate and the water block around the tec unit

[Jabo]

jaydee116 -> that's exactly what I meant, have only as much copper and in such size as to represent thermal energy 'paths' or better 3D map of temperature spread (isothermals) - I just wish I had some explaining talents...
and I wish I had time/equipment to be able to test each of my theories as you, lucky one, have .

[Brians256]

Originally Posted by chrisbmx68

im aware of how to insulate the socket etc im running a swiftec 4000-t atm. (modded to fit my amd board) Im stealign the tec out of that setup and using the maze 4 and cooldplate and i did mean the gap between the coldplate and the water block around the tec unit

Whew! OK, since you are definitely knowledgeable, how about using styrofoam instead of neoprene? It has a better R factor, and you can carve it close enough and then compress the last bit.

As for melting I would think not. If you are cooling the hotside well enough, it shouldn't get much warmer enough. Even if the water is about 40C (i.e. really hot) for a 250W pelt (I'm rounding for ease of computation), and the C/W of the waterblock and TIM joint is 0.2, that only gets up to 90C on the hotside. Boiling water is safe in styrofoam, and that's 100C.

Anyone know the melting point of styrofoam?

[chrisbmx68]

ok cool so would i see any advantage by doing that do other ppl do it? i may just try it with my current setup since im still looking for a pump

[Butcher]

TECs will fail if the hotside gets above 85C in general, so if you're getting a 90C hotside, you need to sort the TIM joint out.

[chrisbmx68]

im not getting temps that high lol even air cooling wiht with a 92mm nado half turned down

[Brians256]

I was just trying reductio ad absurdum (or whatever the latin phrase is) to show that even in a horribly worst case, the styrofoam material should be safe. If you've got 40C water, you have a horrible radiator/fan problem, too. Also, the hottest TEC most people can get is only 226W.

[chrisbmx68]

i got a 226 but im running it at around 11v so its not too bad (considering i cant keep the thing cool since the ambient temp has been liek 90f lately)

[bigben2k]

Ok, let's stir things up a bit...

I've come across some unsubstantiated information that states that a TEC does NOT spread heat, it carries it quite linearly.

I have to assume, for the sake of common sense and for safety, that the heat spreads would flow ******d at ~ at 45 deg angle, from the heat source (CPU die). Given a 1/4" cold plate, the two ceramic (?) layers that constitute a TEC, and a water block base plate, what's the resulting area that actually requires cooling? Is it less than a typical 40 mm by 40 mm TEC?


Here's another mystery: does the cold plate bend or flex, when clamped, and by how much? Does it impact performance?

[AngryAlpaca]

Whew! OK, since you are definitely knowledgeable, how about using styrofoam instead of neoprene? It has a better R factor, and you can carve it close enough and then compress the last bit.

Styrofoam won't get in as close as neoprene would.

I've come across some unsubstantiated information that states that a TEC does NOT spread heat, it carries it quite linearly.

Makes sense to me... Hundreds of tiny thermocouples spread across, with only a 1-2mm spreading area, so it will probably be fairly linear. If it spread well, then what would be the point of the coldplate?

what's the resulting area that actually requires cooling? Is it less than a typical 40 mm by 40 mm TEC?

Why would it be less? Yes, most of the heat would spread to the coolest point of a waterblock (typically the middle) but there's still a bit on the outside, and cooling a large area removes any guesswork. Hell, get an MCW6000, then you aren't sacrificing anything (except appearance, supposedly.)

Here's another mystery: does the cold plate bend or flex, when clamped, and by how much? Does it impact performance?

I'll say no. The coldplate is clamped against the TEC (flat) which is clamped against the waterblock (it damned well should be flat.) Using one flat surface to pull another flat surface straight towards itself will cause both flat surfaces to remain flat. If the waterblock isn't flat, however, then both the TEC and the coldplate will bend a little bit, hurting the thermal transfer.

[jlrii]

Originally Posted by bigben2k

Ok, let's stir things up a bit...

I've come across some unsubstantiated information that states that a TEC does NOT spread heat, it carries it quite linearly.

I have to assume, for the sake of common sense and for safety, that the heat spreads would flow ******d at ~ at 45 deg angle, from the heat source (CPU die). Given a 1/4" cold plate, the two ceramic (?) layers that constitute a TEC, and a water block base plate, what's the resulting area that actually requires cooling? Is it less than a typical 40 mm by 40 mm TEC?


Here's another mystery: does the cold plate bend or flex, when clamped, and by how much? Does it impact performance?

A TEC as far as I know transfers heat in a linearly because it cannot be thought of as a single device.

A quote "A thermoelectric has analogous parts. At the cold junction, energy (heat) is absorbed by electrons as they pass from a low energy level in the p-type semiconductor element, to a higher energy level in the n-type semiconductor element. The power supply provides the energy to move the electrons through the system. At the hot junction, energy is expelled to a heat sink as electrons move from a high energy level element (n-type) to a lower energy level element (p-type)."

What that quote does not mention is that each TEC will have many of these couples (n>p) and that he heat will only move directly throught each pair and cannot spread to other sets. So a proper heat spreader is required on both the hot and cold sides for the device to function well.

As far as mounting goes there is flex involved and manufactures standards referring to same. Here is a quote on assembly

"Procedure for Assembling (Type L) Lapped Modules to Heat Exchangers

IMPORTANT: When two or more thermoelectric devices are mounted between a common plate, the thermoelectric devices thicknesses should vary no more than 0.0015-inches. Contact our Engineering Department for more information on close tolerance lapped thermoelectric devices.
Step 1. Prepare cold plate and heat sink surfaces as follows:

A) Grind or lap flat within +/- 0.001" in module area.
B) Locate bolt holes as close as possible to opposite edges of module (1/8" clearance recommended, 1/2" maximum), in the same plane line as the heat exchanger fins. This orientation utilizes the additional structural strength of the fins to prevent bowing. Drill clearance holes on one surface and drill and tap opposite surface accordingly (see sketch). If a spacer block is used to increase distance between surfaces, performance is greater if the spacer block is on cold side of system.
C) Remove all burrs, chips and foreign matter in thermoelectric module area.

Step 2. Thoroughly clean and degrease thermoelectric module, heat exchanger and cold surface.
Step 3. Apply a thin continuous film of thermal grease (Wakefield Engineering Type 120 or Dow Type 340) to module hot side surface and to module area on heat exchanger.
Step 4. Locate module on heat exchanger, hot side down.
Step 5. Gently oscillate module back and forth, exerting uniform downward pressure, noting efflux of thermal compound around edges of module. Continue motion until resistance is felt.
Step 6. Repeat Step #3 for cold side surface and cold plate.
Step 7. Position cold plate on module.
Step 8. Repeat Step #5, sliding cold plate instead of module. Be particularly careful to maintain uniform pressure. Keep the module centered between the screws, or uneven compression will result.
Step 9. Before bolting, best results are obtained by preloading in compression the cold plate/heat exchanger/module assembly, applying a light load in line with center of module, using clamp or weights. For two module assemblies, use 3 screws located on module center line, with middle screw located between modules. To preload, torque middle screw first. Bolt carefully, by applying torque in small increments, alternating between screws. Use a torque limiting screw driver. The recommended compression for a TEC assembly is 150 to 300 pounds per square inch of module surface area. Using the following equation you can solve for torque per screw:

T = (C x D x F x in²) / (# of screws)

T = torque per screw (inch-pounds)
C = torque coefficient (0.20 as received, 0.15 lubricated)
D = nominal screw size (4/40 = 0.112, 6/32 = 0.138, 8/32 = 0.164)
F = Force (lbs / in²)
in² = Module surface area (length x width)

Check torque after one hour and retighten if necessary. Use Stainless Steel Screws, fiber insulating shoulder washers, and steel spring (Belleville or split lock type) washers (see sketch in Assembly Tips).
CAUTION!
1. To assure good thermal grease interfaces there should be no bowing of either surface due to torquing. To prevent bowing, apply less torque if one or both surfaces are less than 1/8 inch thick copper or 1/4 inch thick aluminum.
2. Lead wires are soldered to module tabs with bismuth/tin solder (136°C). If lead wire replacement is necessary, use bismuth/tin solder.
DO NOT use lead / tin solder (180°C)."


If you are interested in Peltiers I reccomend going to Melcors site and downloading their Aztec software which includes all kinds of info and a program to select what you need. They are fun to think about and play with but are not very eficient. If you want to cool a 100w chip to 0 deg c it can be done but expect to heat a 400 sf room in the middle of winter with your rig and the resultant rize in your electric bill.

[Butcher]

A TEC won't flex much as the surfaces are ceramic which will shatter or crack if you try to bend it too much.

[jlrii]

Originally Posted by Butcher

A TEC won't flex much as the surfaces are ceramic which will shatter or crack if you try to bend it too much.

That they will.

Another cause of failure is bowing of the hot/ cold plates. If a unit is not properly clamped the majority of the pressure exerted on the TEC will be around the edges somtimes with no pressure at all in the middle. If some of the p-junctions have nowhere to transfer heat to because of little or no clamping pressure they will not perform well and may self destruct.

Yet another cause of failure is stress fracturing. This is more likely to occur in larger TECs that are cycled. The clamping requirements of TECs are much higher than CPUs etc. Typical mounts are designed to exert about 300 psi on the TEC. A copper mount will not expand and contract at the same rate as the device and under 300 psi the TEC cannot move on it's own. This is why many maufacturers do not produce the mondo 220w + TECs used a lot on computers, as a larger TEC is repeatedly heated and cooled the MTBF from stress is lower than with 2 smaller TECs with the same combined output.

[bigben2k]

Nice info jlrii! Got a source for it, or is it on Melcor's site too?

[jlrii]

Here is a link to their software page. The program is free and includes a little "thermo electric handbook." Don't worry about the info requested as long as you fill anything in (be correct if you wish or simply put a letter in each box) it will let you download. My main complaint with their program is that it uses only their model #s, but after playing with it for a while you get a real good idea of what TECs can and can't do. When using the Thermoelectric Cooling Selection tab set up your processor amps , ambient etc and it'll tell you what TEC(s) to use and the cop etc...BUT make sure from that page you go to the analysis worksheet and play with the "Set Thermal Operating Point" options. Here you can set hot and cold side c/ws. See what you find on the cold side and how efficiency responds.

Edit: some info c/p'd from their manual, some from memory ....did a lot of reading about this stuff a few months back, played around with a couple of experiments., and reached some conclusions. See what you find

[jlrii]

Just looking throug some old links this might be of interest to you, lets you know what that prog is figuring out. Lots of info out there.