Post Rad Chiller Concept

[NaeKuh]

Hey nol when you get some numbers out, can you make a new thread so we dont need to go searching for it?

[littleowl]

Hey nol when you get some numbers out, can you make a new thread so we dont need to go searching for it?

thanks NaeKuh

[Clue69Less]

Call it the backup band thread.

[n00b 0f l337]

Hmmm, working with some of my thermocouples here, and I have to say, I think the setup is a proof of concept, however it needs to be over a larger dissipation area, moving a few hundred watts of energy across something this "small" still makes a rather hot hot side.

[NaeKuh]

nol your telling me a MCR320 with a D5 isnt enough to cool the hotside?

[n00b 0f l337]

Its not the radiator, its getting the heat out of the pelts and hotside of the block more so. Spreading out the load over 12 pelts would be much easier per say

[Scarlet Infidel]

I see what you mean, running some quick estimates Id say the water block has a surface area for heat transfer of around 480cm^2. Its a nice block design, but that's not a huge number.

For reference, the Coolit products have a hot side surface area of 565cm^2 (except the Boreas which I think doubles this).

Also you may get some losses from laminar flow but I don't think that's a major problem in that design.

[NaeKuh]

I see what you mean, running some quick estimates Id say the water block has a surface area for heat transfer of around 480cm^2. Its a nice block design, but that's not a huge number.

For reference, the Coolit products have a hot side surface area of 565cm^2 (except the Boreas which I think doubles this).

Also you may get some losses from laminar flow but I don't think that's a major problem in that design.

You cant compare my unit with the coolit's. My hotside is watercooled, coolit's is aircooled. Theres a difference. Also coolit's uses ALU. That company name is taboo on XS.

So nol you think i should drop 12 little TECs on that instead?

You mind trying it out? How the hell are you going to even fit 12 little TEC's.

[n00b 0f l337]

Would need a bigger plate setup sadly, the idea is to dissipate the heat over a larger surface area, and with smaller TEC's to dolocalize heat.

[Scarlet Infidel]

Sorry, i meant that their cold side surface area was that. It was purely for reference to give the numbers some meaning.

You cant compare my unit with the coolit's... That company name is taboo on XS.

That's kind of my point. If their products are so lousy, you should be trying to beat their specifications.

[initialised]

Check out my project here.

10 basic air heatsinks on 10 40mm peltiers (rated as 137w). The idea is to run each peltier at no more than 5v (from the 5v line of my existing PSU, so no extra PSU required). Not aiming for ultra cold temperatures, but a little boost without using too much power. Also, I'm trying to do it as cheaply as possible.

I did something similar with 6x50W Qcmax Pelts sandwiched between a Koolance HDD block and rack mount radiator, it held ~19C chilling RAM (OCZ Flex), NB & SB (P5K).

[Uncle Jimbo]

There 90W TEC's if my memory serves me. And theres 5 of them.

The block is uber free flowing. Placed right along with DD MPC Universal.

Nol is doing all the testing needed. :T Thats why he has it. After he figures out the best way to use it, thats how i will adopt it.

Well I stand corrected on getting more heat transfer than heat in. I took a small foam cooler case and mounted the TEC (a 12730) between 2 melcor sinks in the top so that the cold side sink was in the water almost to the TEC, and the hot side was in air. I used a 110CFM fan on the hot side, giving me .06C/W thermal resistance. With 1 gallon of water initially at 25C and room ambient at 25C, I applied 5.1V from a 500W Ultra supply. Current draw settled at 5.8A after a few minutes. That represents 29.6 watts in.

After 15 minutes, the hot side measured inside the sink by the TEC was 30.1C, and the water was at 22.7C. So pulling 2.3C out of a gallon of water in 15 minutes, with a differential of about 7 degrees. That represents about 40W transfer. The total hot side load of about 70W squares with the 5.1 rise pretty well, so I believe the numbers.

That is a single 62mm TEC, and with a different mounting arrangement I could probably get 4 of them under that Melcor sink, but the hot side would get a lot hotter with a 280W load. The TEC would be more efficient at the higher temp but also working across a bigger differential.

Just to see where it would go, I put 12.2V across the TEC and ran the same test. Current stabilized at 22A for a total heat in of about 270W. After 10 minutes, the hot side was at 50C, and the cold side just under 20C, representing about 130W transfer across a 30C differential.

That leads me to believe that 4 of these, run at 5V, would cool a 1GPM flow by .5C and require 120W to do it. That is clearly in the same league as phase change cooling for efficiency.

I'd like to try the same thing with WC on the hot side to see what it would do with a lower differential - my guess is it would move over 50W but use more current, so it might not be as efficient.

[Scarlet Infidel]

Well I stand corrected on getting more heat transfer than heat in...

Glad to see the theory working in practice. Thanks for trying that out, I hope that as more people see this effect and start believing it the more people will come up with innovative ways to exploit it.

I'm still itching to finish my chiller, but I'm so indescribably busy that there's no way I can spare a weekend to return home to do it.

Edit: also, I must try to spend a little time working on my peltier calculator program, it seemed to be giving startlingly accurate results with a few tests I ran.

[systemviper]

nobody has any pictures./////

[NaeKuh]

That leads me to believe that 4 of these, run at 5V, would cool a 1GPM flow by .5C and require 120W to do it. That is clearly in the same league as phase change cooling for efficiency.

I'd like to try the same thing with WC on the hot side to see what it would do with a lower differential - my guess is it would move over 50W but use more current, so it might not be as efficient.

I have 5 90W TEC's and nol is saying he's having a hard time pulling the heat off the hotside. :\

What you described above was also my theory, use them downvolted nicely, and try to pull the most efficiency out of it. :T

[n00b 0f l337]

Sorry mate, cam trouble and not alot of time, moving setup most likely a bit later to series as original intended

[Uncle Jimbo]

Glad to see the theory working in practice. Thanks for trying that out, I hope that as more people see this effect and start believing it the more people will come up with innovative ways to exploit it.

I'm still itching to finish my chiller, but I'm so indescribably busy that there's no way I can spare a weekend to return home to do it.

Edit: also, I must try to spend a little time working on my peltier calculator program, it seemed to be giving startlingly accurate results with a few tests I ran.

There has been some discussion here on different TEC products. For the most part, the product number is the number of elements, then the max amps i.e. NNNAA so a 12730 is 127 elements, 30 A max. Different doping of the elements and different geometries will produce slightly different performance.

For Bismuth telluride, the most common materials, each element has a nominal voltage of about .123 at 25C max. So TECs with 127 elements have nominal voltage of about 16V at 25C. The current capacity is determined by the size of each element and the voltage by total number of elements.

Peltier elements have a negative coefficient of resistance, which means the effective series resistance increases with heat. That means more voltage is needed to achieve the same current as the hot side gets hotter, or conversely, that lower current will result from the same voltage. Larger elements have a lower nominal resistance. The effective resistance (R=V/I in a running system) varies with both heat and heat transfer and is typically about twice the nominal resistance with a variance of 20% or more over the range of operation.

The heat transfer is determined by the amps and the temp differential. For a given amperage, the lower the differential, the more heat is transferred. Higher currents move the whole curve, so a larger differential can be sustained, or more heat can be moved for a given differential.

You should never exceed the max current for a TEC. High voltages can also destroy the elements. Because different formulations have different resistance, the voltage needed to sustain current varies somewhat, TEC by TEC, in addition to the hot side temp. 127 element units are designed for a max voltage of around 16V at 25C, which is about 18V at 50C. Some TECs have a design point at 75C or 100C or even 200C instead of 25C or 50C - the elements are the same, but higher temp solder is used to allow the higher operating points.

As my recent tests show, the heat transfer is pretty linear with amps down to some pretty low values, so with a low differential and low voltage you can achieve good efficiency but need a lot of big elements to move the total heat load. The conclusion is that for best efficiency use the biggest elements you can get, as indicated by amp capacity, and run low in the curve. I was looking at an 03140 - 31 elements, 40A max, and 3.5V nominal. They have bigger elements than the 12730, and therefore should move more heat. 4 of them in series would effectively be a 12440. I would expect that to have about 30% better heat capability than a single 12730 at the same voltage.

[Scarlet Infidel]

Some good info, seems to match with what Ive got in the way of equations but its nicer to have it in words.

That peltier you mentioned (03140) seems a little obscure. Do you know where to get them from?

When I chose the peltiers I did (TEC12709) I was aiming for:

Highest current,
Smallest physical size (40mm - this was to minimise copper cold plate costs),
Being dirt cheap ($35 for 10 inc. world-wide postage, packaging and paypal fees).

[Uncle Jimbo]

Some good info, seems to match with what Ive got in the way of equations but its nicer to have it in words.

That peltier you mentioned (03140) seems a little obscure. Do you know where to get them from?

When I chose the peltiers I did (TEC12709) I was aiming for:

Highest current,
Smallest physical size (40mm - this was to minimise copper cold plate costs),
Being dirt cheap ($35 for 10 inc. world-wide postage, packaging and paypal fees).

I don't think those TECs are cheap - oddball voltage is usually pretty high price. One example is 03111-5M31-40CQ at http://customthermoelectric.com/tecs...FRaQggodl1GdFA - That is a 03140 part for about $50, but for what you are doing, they are probably not a good choice. I was only looking at them to test the edge case for better efficiency.

The 12709s are 40mm, right? If your setup can take a 62mm, your low cost choice is a 12730 which is the highest current 127 element I have seen commonly available. I have also used the '437W' TEC from Frozencpu http://www.frozencpu.com/products/24...x_Peltier.html which is a 19933, also in a 62mm size. At $50 it is pretty expensive and undervolt performance is only a little better than the 12730.

The biggest amp unit I have seen is a 03178, one example is at http://www.inbthermoelectric.com/highperformance.html listed as inb 31-5.0-1.5 - that is a 55mm unit which can pass 78A. It is a little less efficient at full power than the 12730. I have not used it, but the data suggest 4 of those in series would move 750W at full power (1200W in!). But at 5V across 4 in series, it should move 140W or so with maybe 65W in. That might be interesting to check out the question is what it would draw at 5V, given that is only 1.25V per module.

[Clue69Less]

I have not used it, but the data suggest 4 of those in series would move 750W at full power (1200W in!).

Good Gawd, y'all!!!

[Uncle Jimbo]

Good Gawd, y'all!!!

doubles as a large room space heater for those frosty mornings!

[NaeKuh]

guys some update.

Nol took my concept wrong. LOL

He linked the hotside and coldside in one loop. Personally i dont see how he got any gains from this method as its using something simular as a perpetual pendulium.

Anyhow Nol finally understood what i ment, and he's gonna have to reloop it as well as bring out another pump.


Nol you must of gone though hell, and probably wondered wtf i was thinkn.

[n00b 0f l337]

No! Thats a possible way for it to work as well, some TEC chillers actually have that If the flow rates just right (often low) the radiator can remove all the heat from the pelts (+ there load) and the system, however the cooled object will have some gains from the cold flow of liquid (not perpetual motion as you have to keep pumping power in )

[Clue69Less]

guys some update.

Nol took my concept wrong. LOL

He linked the hotside and coldside in one loop. Personally i dont see how he got any gains from this method as its using something simular as a perpetual pendulium.

Anyhow Nol finally understood what i ment, and he's gonna have to reloop it as well as bring out another pump.


Nol you must of gone though hell, and probably wondered wtf i was thinkn.

What the fcuk ??? Looping the hot side to the cool side with TECs means the temps will increase, period.

[n00b 0f l337]

No, you can create a liquid delta
Its done but normally needs much larger systems, I've gotten cold side down a few degrees actually, seems to help temps when the rad can take the heat