Toying with TECs

Frozennova

Frozennova

Frozennova

Man of Honour
Joined
13 Nov 2009
Posts
11,643
Location
Northampton
After finding out about TECs I became very curious about them working, and once I'd saw pnemonics build this only increased the curiosity leaving an itch that must be scratched. but firstly this leaves a few questions regarding heatdump and cooling power.

Firstly I'm looking at doing chilled loop as this will be most effective. I've looked at undervolting charts of xtremesystems and decided on 3 possibilities either the 226W danger den, or 320w and 437w elsewhere do the figures below look accurate as far as heat moved and heat dump im looking at.

226W Peltier

Vmax 12v Heatdump Single 368.18W
Qmax 140.9W Heatdump 1472.72W
Imax 18.94A

Number of Tecs 4
Total Cooling Power 563.6W
 
ive always liked the idea of pelts but ive never had enough time & resources to play. and the results ive read about havnt always been that impressive for the amount of investment. all the same id still like to toy around with 1
 
I don't suppose you have a link to those undervolting charts? I couldn't find them :(

How are you going to cool the hot sides? How are you going to power them?

If you manage to cool them properly you'll need to add antifreeze :)

I think you'd be better off with just 1 to start with and power it from 15v instead of 12, this should get you a sub ambient loop if it's cooled properly.
 
Do a google search for tec undervolting its the result that points to xtemesystems

it'll be powered by a standard psu with the 12 rails internally joined.
I plan on using around 4 of these to give a high cooling capacity.

The pelts will be cooked using a second water loop, and yes i'll be using anti freeze
 
Ultrasonic is the chap you want really, but he's not around particularly often. You don't need undervolting tables from xtreme systems as badly as you need to learn to read the performance graphs from the people you'd buy the tecs from. I'll dig out a couple of links for you shortly.

What current is your atx psu rated for? Peltiers will draw rather a lot, and a cheap atx psu will just drop dead at the sight of them.

To confirm, you're planning a loop with your processor in, cooled by peltiers. The hot side of the peltiers is cooled by a different loop, which has only radiators in. If so you can take advantage of radiator performance scaling near linearly with coolant temperature, i.e. if a 240 radiator will dissipate 250W with a ten degree water - air delta, then it'll dissipate 750W with a thirty degree air-water delta. Letting the hot loop run much hotter than you'd be willing to with components in it means the radiator efficiency improvement can be used to offset the peltier heat dump.

What temperature do you want the cold loop water to be at, what wattage will you have components dumping into it, and critically how much are you willing to spend to achieve this?

I'm budgeting around 500 for my prototype once I get started on it, but with luck will be spending less than half of this.

Handbook
The 12711-5M31-15CQ which I will probably use at 6V
The 19911-5L31-15CQ which Pneumonic is using at 12V

After some initial sketches to show what I thought Pneumonic was planning I realised that I'm hopelessly ignorant regarding heat exchanger design, so I'm learning the theory behind that at present (from textbooks after failing to find much online). Then I'll learn to use Ansys competently, and hopefully start prototyping after that.

I believe that with a sufficiently flexible budget, any wattage can be dealt with. Anything more than 30 degrees between hot and cold water wont happen in a single stage device, but if you can keep the hot loop at 5 above ambient, the cold at 15 below is nothing too shoddy. If you move to two stages you're looking at around 50 degrees delta between hot and cold, the main thing that changes is you need some more space about about 3 times as many peltiers. I believe there is enormous potential for peltier cooled above ambient m-atx systems, so that's what I'm looking into at present.
 
Last edited:
About 9 years ago I used a single peltier on a piece of copper which sat directly on the CPU, then used a standard loop on top of that, worked fine until the PSU died. Celeron 600 at 1.1GHz :)
Problem these days if the CPUs pump out more heat so the tecs you need a re a bit bigger, but I would have thought the direct method should still work?
 
Possibly. The highest rated 62mm tec that can be found through google will cope with a 200W load through a delta of about 10 degrees when running at 12V. Intels quad cores are seriously pushing the limits of this, and efficiency is woeful, but it is still viable. Makes insulation a lot easier than the chiller approach too. This is of definite interest with the i3 series of processors as the heat put out is much more reasonable than the i7.

Are you sure you didn't clamp the peltier between two bits of copper, then mount this on the processor? I don't believe atx spec allows for the 250ish psi mounting pressure peltiers work best with.
 
I've looked back over things and decided 3 of the 226w pelts undervolted should provide enough heat transfer (440w).

As far as heat exhangers go I haven't a clue so sharing you knowledge would be appreciated. Im assuming yours will be CNC machined as from previous posts your in the engineering field. I havne't got the budget for such luxury so I'll either be buying them off the shelf or hand mill my own heat exchangers proving I can get hold of a family friend who owns an engineering company fairly local.

The tecs will be drawing ~ 19*3=57amps and I'm looking at using the 1000w corsair as it provideds 40amps per rail, I'm contemplating building a rad/psu box to seperatly house these as I want to build this is a mid tower rather than full tower case so internally mounting two loops, a heat exchanger, 2 psus and 2 240 rads is going to be a squeeze.

My initial project will be based around a 775 rig that im currently stockpiling components for as they pop up on MM or auction sites, then once the budget allows I'll move onto i5 or i7 probably tweaking initial errors or making improvements.

Budget has no been considered as I have only just begun researching and budgets are rarely followed.

I will be doing a good 6 months of research and design work before the actual build commenses so I can be sure of enough research and planning
 
CNC cut copper brazed together would be ideal. There's a distinct risk that I won't get time on the cnc and it'll end up painstakingly cut on a bridgeport then silver soldered, but I rather hope not. Engineering student here. Pneumonic bought some which look about as good as any could do, cost about 60 quid a plate.

While I'm sure you've got the numbers right, could you point me at a datasheet for the peltiers you're planning on using? The 226W is a value for Qmax, the wattage it'll move through a temperature difference of zero. If you've scaled that value down to account for undervolting it, then your 440W will also be through zero degrees, i.e. the maximum load wattage before the system fails.

Bear in mind that taking it below ambient temperatures introduces a parasitic load, heat flow from environment into your cold pipes. This is difficult to estimate, though far from impossible. Wrapping everything in insulation helps considerably. Essentially you have to overspecify relative to the wattage you want to move.

Based on skinee labs testing, a 240mm radiator should be good for a C/W value of about 0.04, or a combined value for both of yours of about 0.015.

Choosing which tecs to use is tricky. The one I'm looking at moves 20W at delta30 with a cop of 0.66, or moves 43W at delta20 with a cop of 1.3. Using six of these and hoping for d20 gives 260W moved and 460W to remove, so a single 240 radiator will keep the hot water at 20 degrees above ambient. D20 then gives the coolant at ambient temperature.
 
Last edited:
The only problem I had with peltiers was condensation, the chip would be covered in water droplets within an hour. if i left it on for several hours the motherboard around the socket was drenched in water.

I got through a few motherboards before I eventually gave up.

That was back in the celeron 300 and amd k6 cpu's. -15C got that celeron over 100% overclock :D
 
I looked at TECs and Pelts 5 or 6 years ago... then it was only really used to Cool GPUs. IIn the end I opted for a Vapochill and then a Mach II GT Prometeia. I have no idea why you'd choose a tec over a Vapochil LS etc.... these offer far easier cooling and far safer too. Not to forget a much bigger capacity for cooling.
 
Datasheet(226w 15.2v): http://www.customthermoelectric.com/tecs/pdf/12711-5M31-24CZ_spec_sht.pdf

The initial build will be using an Intel E1200 (65W TDP) that I picked up for £10 off the bay, I plan on buying most of the components used however PSU and Case will be new.

Could you expand on the parasitic load part, is this essentially as the pelt will be below ambient heat will be transferred from the atmosphere to the cold side?

I'm by no means considerably skilled within engineering I can use a lathe and a mill to an acceptable accuracy which I believe will be invaluable in a project like this.

I know assumptions are generally wrong but are the heat exhangers two larger copper waterblocks with the pelts clammed between them at a high pressure ~200psi?

EDIT:

I've just looked over the datasheet and would I be interpretting the graphs correctly by thinking the Qmax at a DeltaT of 30° would be ~105w giving a Qmax of 420w across 4 tecs

I've just looked at the 225w tecs that you linked and running 6 of these would require 6amps less to power than the 226w units and would yield a matching Qmax at a DeltaT 30°
 
Last edited:
If you're only dealing with a 65W processor then you've overspecified the peltiers by a factor of about 4 at the moment. Are there plans to cool more things than this one in the future?

Anything below ambient temperature will suffer a heat transfer into it from the environment. The cold side of the pelter, the cold plate, the waterblock, all the tubing. First off this leads to condensation, but more importantly it means wattage going into the cold loop that the peltiers have to get rid of. This is a parasitic load, as it's a load which you don't really want and comes along for free as it were, irrelevent if no temperatures are below ambient.

That's pretty much it as far as heat exchanger goes. The 240 radiators could be reasonably described as "forced convection air-water heat exchangers", or the water block as a "solid to water heat exchanger." The more specialised and carefully designed the waterblock is, the more efficient the heat exchanger. You don't really want a 5 degree temperature difference between the liquid and the face of the heat exchanger if you've only got a 30 degree delta to start with. I'm currently planning on adapting this design.

No to the last bit. For delta 30, the right hand graph gives about 18.5A at 12V, from the left hand one 18A at delta 30 gives a Q of 105. So it can be expected to move 105 watts under these conditions. It'll use 12*18.5=222W to do so, meaning you need to remove 222+105=327W per peltier from the hot side. This is a COP, or coefficient of performance, of 0.5 as it needs about 2 watts for every one watt moved. Three of them should be good for 300W moved through a delta of 30. The radiators have to cope with 980W, using a C/W value of 0.015 (1600rpm fans, assuming thick radiators) gives a hot water temperature of 15 degrees above ambient. Cold side would then optimstically be 15 below ambient. Electrically very expensive but impressive performance.

Qmax is 226W, as this is Q at maximum voltage (15.2) which gives maximum current (24A) and a delta t of 0. There are equations available which use Qmax, Imax etc as parameters and generate these curves, I haven't got around to comparing them with the graphs yet (need some parameters that I can't find) but I'm sure they're better. For one thing the graphs are at a constant hot side temperature, probably 25 centigrade, and so aren't really valid.

The psu is probably a bad choice by the way. Meanwells are much loved here. I'm going to work to the limitations of my 860W psu as I want to run the peltiers and the computer from the same power supply. An 8 pin pci-e cable is rated for 150W, so I'm looking to stay below 12 or 24 amps and look for higher cop. An 8 pin atx cable is good for 28A all by itself, but few power supplies come with a spare one.
 
Last edited:
JonJ678 said:
Are you sure you didn't clamp the peltier between two bits of copper, then mount this on the processor? I don't believe atx spec allows for the 250ish psi mounting pressure peltiers work best with.
Click to expand...

I didn't use that pressure, the 2 tabs on the socket would never have taken it :) I also used cheap and nasty paste, still got good temps. I seem to recall getting 25 under load, which was about ambient at the time. I think the tec was only 70W (40x40mm). Rad was 120x240 with 1 standard fan on it. 2 Gallon bucket for the res :D
 
Dimension said:
The only problem I had with peltiers was condensation, the chip would be covered in water droplets within an hour. if i left it on for several hours the motherboard around the socket was drenched in water.

I got through a few motherboards before I eventually gave up.

That was back in the celeron 300 and amd k6 cpu's. -15C got that celeron over 100% overclock :D
Click to expand...

Nice overclock :)
I was using mine on a slot1 setup (slotket adaptor and a celeron 600). Came home from work oneday to find it had crashed (W98....) couldn't see the CPU socket, too much ice! rebooted and went into the BIOS, -56 :D Then had to quickly find a dish to put on top of the gfx card to catch all the water that was running off....
 
The E1200 is just a test processor, depending on how long the build takes and cost I will then either move on to i5/i7 most probably tweaking along the way.

As far as PSUs I hoping for something that was variable voltage, and small enough to fit into a drive bay but I doubt anything capable enough can be found. Size will be an important issue as I want everything to go into a CM-690 II as I love the look of these cases and they provide excellent W/C options without breaking the bank, however means using thinnner rads which is why two will be used as I dont want any toasted pelts

Thanks for expanding on heat exchangers and effieciency, this should give me a firm basis to work upon
 
You must log in or register to reply here.
Back
Top Bottom