TEC waterchiller

[Jorlen]

Ok here is a picture. Look carefully, water HAS to pass through the waterblocks.

[Sam__]

Ok here is a picture. Look carefully, water HAS to pass through the waterblocks.

ahhh. ok... sorry,i misinterpreted your first pic. looks like it should work nicely.

[Jorlen]

I like this:
http://www.overclockers.com/tips1240/

Another way to do it would be to make a box out of copper/aluminium instead of acrylics and positions the fan to create best possible circulation and let the mineral oil cool against the walls of the box. Then place it outside the chassi.

[Xeon th MG Pony]

ahh i see

perhaps you could create a circuit so that voltage generated by a peltier when the water-block temperature is higher than the heat-sink temperature would feed the "base" of a "transistor amplifier circuit" which would have a gradual effect on the amount of current that could flow through the "collector" and "emitter" - as water temperature drops to ambient the current shuts off in a nice analogue curve.
(i've drawn the "transistor amplifier circuit" with "base, collector and emitter" as symbols of transistor function - i dunno what the real circuit would look like)

then you'd still have the 'luxury' of a water-to-ambient-dT based power control - but not a digital on/off one.

you might also want a variable resistor that you could use to set a minimum amount of power that'd flow into the peltiers regardless of water temperature - this'd allow you to set a minimum temperature near or under ambient. without it the minimum temperature would be slightly over ambient. if full load temperature is over ambient i guess you'd leave it out.

though if the ambient temperature in your computer room is fairly steady this'd be a bit overkill

You could do this by feeding the pelts out put to a Voltage controlled Oscilatore then filter it through a schmit trigger for a more streamlined aproach.

[Xeon th MG Pony]

I like this:
http://www.overclockers.com/tips1240/

Another way to do it would be to make a box out of copper/aluminium instead of acrylics and positions the fan to create best possible circulation and let the mineral oil cool against the walls of the box. Then place it outside the chassi.

When that thing goes it's going to go good, not worth it IMO, fare more effective and safer ways to achieve the same effect.

[Jorlen]

Ok, I dug out the old PSU. Its called HiTRON model HSH250C-11. Its actually only 20A@12.5V, 250W.
The unloaded voltage range goes from 9.35V to 15.5V. I hope I have that range available under load too. I dont have any suitable resistance available to test it under load atm.
The coils (I suppose) started making some faint noise when voltage was tuned below 11.5V. Nothing to write home about, but I prefer it makes no noise in the final system.
Its completly silent above 11.5V which is nice. Hope it stays like that under load too.

[Sam__]

Ok, I dug out the old PSU. Its called HiTRON model HSH250C-11. Its actually only 20A@12.5V, 250W.
The unloaded voltage range goes from 9.35V to 15.5V. I hope I have that range available under load too. I dont have any suitable resistance available to test it under load atm.
The coils (I suppose) started making some faint noise when voltage was tuned below 11.5V. Nothing to write home about, but I prefer it makes no noise in the final system.
Its completly silent above 11.5V which is nice. Hope it stays like that under load too.

you may have to put some kind of load on the 5v rail to get a hgher/more stable 12v rail. some PSUs like a little on the 5v to get em going too.

[Jorlen]

Thanks for the advice Sam. But this PSU is not from a PC. It only has one output.
This is a picture of the HiTRON next a regular PC PSU.

[mike8913]

ah now I see the tubing more clearly. I didn't realize you had it configured more like a HX. Sorry for my mistake. Otherwise, it looks good so far. You may have said this, but i think you will need an enclosure with fans for the Heatsinks to be effective.

[hollo]

I like this:
http://www.overclockers.com/tips1240/

with a 'coarse' heatsink (wide fin spacing) on the water-block with the internal fan pointing at it it'd be excellent

Hollo, its completly possible to do as you describe. The drawbacks are that a power transistor works a variable resistor in the way that it increases its own resistance to control the current/voltage. Thats why you need massive heatsinks on these suckers. So although it would work, its my second choice. Dont wanna generate more heat then I absolutely have to, you know. Its a cooling system after all.

well in a way the power consumption of a transistor in such a circuit would be a moot point, because whatever voltage controlling mechanism you use it'll boil down to a resistor connected in series with the peltiers, and the resistor will need a heatsink and the voltage moderating circuit will be inefficient (unless its a binary on/off circuit, which you don't want).

however using 3 sets of 3 peltiers would be extremely efficient in its own right, because it would require the least damping of the voltage going into the peltiers during normal operation to avoid condensation. 5 sets of 2 peltiers would require a bit more resistance in series with it to moderate the peltier voltage enough to avoid condensation, and 10 peltiers in parallel would require the most resistance.

since 3 sets of 3 peltiers would be the most efficient use of the peltiers and PSU (so long as it wasn't underpowered) it would make sense to set it up like that .... the possible use of the extra water-block/peltier/heatsink for dT based temperature control was kindof a bonus

what were you planning - 5 sets of 2 peltiers each?

[Jorlen]

Hollo switching on and off at high speed (PWM) is very energy efficient. Losses in TEC performance is 1% roughly, and losses in the switching transistor I do not know exactly, but they are very small in comparison.
Yes I plan to run them 5x2. 2 in serial 5 in parallell.
Resistance of one peltier
15.4/8 = 1,925

Total array resistance = 1.925 * 2/5 = 0.77
at 3.5A per peltier (17.5A total) this requires 13.475V, gives 6.7V per TEC.
Total power draw 236W.
So thats pretty much my max without changing PSU.

I plan to cool GPU and NB too. So total cooling heat output might be in the range of 100W-150w.

38W per HS, at a passive C/W of around 0.8, means a temperature increase of, eeek, 30C degrees. but its easy to improve upon it, a quite fan will take me down to 15C degrees above ambient.

[hollo]

ahh, PWM
http://en.wikipedia.org/wiki/Pulse-width_modulation
i get it now

[Jorlen]

Got the peltier elements now. Whats missing now is the pump, more hose, insulation, structure that holds the HS and waterblock strucure a few inches or more above the top of my case.
The pump is a DDC 10W, which can be modded to 18 if I am lucky with the version number. I might build a box around it if I find it to loud for the rest of the system. Would be nice if someone could confirm that the DDC will be fine just being cooled by the coolant.
Pics:

[Jorlen]

Pump arrived. I tried to solder it to get 18W. But I screwed up and tore the targeted solder face from the PCB by mistake. I will need a finer soldering Iron if I wanna repair it. For now 10W will have to do.
Bought the OCLabs top cause it was the cheapest and capacity wasnt that far behind the XSPC top.

[Jorlen]

I managed to borrow an oscilloscope. And the PWM working frequency of my Gigabyte DS3 cpu-fan output is close to 23kHz. I changed to speed of the fan with Gigabytes Easytune, and the oscilloscope showed how the width of the pulse changed as one would expect.
To complete the TEC controller I now need a set of transistors that can pass a combined current of 20A and switch efficiently at 23kHz.

Didnt bother to take any picture of the measuring procedure. For those who want to know. I used an unused PWM fan from an Intel standard cooler and connected the oscilloscope to the black and blue wire. Then I counted on the screen the nr of cycles completed over the course of 1ms, which were 23.

[Norway]

Hope it will work fine for you, cause I'm thinking of doing something similar just with one 62mm 545W TEC ...
Do you think it will be cold enough and ?


Awaiting your results so get your build done asap

[Jorlen]

Update. I managed to break one of the heatsinks when mounting the TECs and peltiers. I am being a bit careless, I stress to get more work done. As a result, things go wrong. I dont recommend this way of working.
But things progress.
This is how far I have gotten. 9 out of 10 heatsinks+TECs+Waterblocks mounted in a frame.

[FrogBite]

looks sweet

[Jorlen]

@Hollo
Since I broke one heatsink I can only run 9 TECs. I am running the system outside the case atm running 9 TECs. I have no load on the system its just circulating water. The 9 TECs together draw only 5.1-5.4A from the supply.
Each TECs running at 4.1V and 1.7A. 63W of total energy draw to the TECs.
The water is getting a bit cooler. And the Heatsinks are getting a little hotter. Not much to speak of.I might run it like this on my CPU to see what the cooling capacity is. But first I need to fix a few problems.

I got two problems atm, reservoir is crap. Cant bleed worth a damn with it. Need to rebuild it. And the waterblock for the CPU is leaking. This probably wont be to hard to fix.

[rocket733]

That looks sweet, nice job so far

[Scarlet Infidel]

Pump arrived. I tried to solder it to get 18W. But I screwed up and tore the targeted solder face from the PCB by mistake. I will need a finer soldering Iron if I wanna repair it. For now 10W will have to do.
Bought the OCLabs top cause it was the cheapest and capacity wasnt that far behind the XSPC top.

I wouldn't worry about this. In a normal setup, a more powerful pump will put more heat into the water and hopefully this will be counteracted by the increased flow.

In a chiller setup, because of the peltiers, the increased heat in the water can become more or a negative factor than the increased flow is positive (I think...).

So, in my air cooled chiller project I am aiming to use a DB-1 pump as its only 6w despite the fact I own a more powerful 18w DDC. Then again my cold side loop is very unrestrictive which is a major factor. When I have finally finished the chiller I will test both pumps so try and prove this.

[Jorlen]

I can change voltage to the pump and see how the performance of the system changes with pump power. If the effect is noticable and favorable I might attempt the modification. But noise level is even more important for me. I havent heard the pump at 18W.
I have modified the reservoir and repaired the cpu-wb now. And I bought a new heatsink to replace the one I punctured.

[scifikg]

Do you have any thermal probes? Even those cheap ones like this.

[Scarlet Infidel]

I can change voltage to the pump and see how the performance of the system changes with pump power. If the effect is noticable and favorable I might attempt the modification. But noise level is even more important for me. I havent heard the pump at 18W.
I have modified the reservoir and repaired the cpu-wb now. And I bought a new heatsink to replace the one I punctured.

Good, I think you will find more or less the same temps with it at 18w but with more noise and wasted electricity. Personally I found my 18w pump too be too noisy as a fan of quiet computing.

[Jorlen]

I dont have any thermal probes. I will probably get one.