Proof-of-concept project - DIY PWM-controlled TEC on air

B

buzzbomb

Gawd
Joined
Sep 29, 2009
Messages
594
Forgive me if this would qualify as extreme cooling, if it belongs there feel free to move it.

So I was throwing some spare parts into a case the other day and had an idea: I've always wanted to play around with a TEC, plus everyone says they are a bad idea (challenge accepted!). I've tried googling to see if it has been done before and I couldn't find any info so maybe it's a dumb idea? Regardless, the parts are already ordered and on the way - no point in turning back now! I'm going to run a TEC through a PWM controller with a potentiometer on it as a way of directly controlling the rate of cooling, to fight condensation, plus the power will be fed through a relay so it'll only be cooling when the PC/PSU are powered and running. I'll be using a Scythe Mugen 3 on the hot side of the TEC (with at least one 120mm fan, possibly two in push/pull, we'll see), a 90w/138w max TEC, and a copper cold plate on a Phenom 9850BE/M2N32-SLI combo. From what I've read the temperature of the cold side of the TEC is directly effected by the amount of heat you can dissipate from the hot side; I figure the Mugen 3 will dissipate enough to match or beat the 125W TDP of the 9850BE at stock clocks with no problem, so if the TEC is PWM-controlled to throw even 200W of equivalent heat off the hot side, the cold side should be at least below ambient. The only thing I'm worried about is whether the TEC will hold up to a PWM pulsing the power - but if the concept works I might upgrade to a more powerful TEC/cooling combo. The whole thing is going to be well under $100 invested so if it fails miserably it was at least a fun project and I'll finally be able to say I got to play around with a TEC.
 
OK it looks like the PWM-controlled TEC idea itself has been done and proven to work, I guess the proof-of-concept lies in the DIY/inexpensive sourcing aspect of the project.
 
Remember to check the input voltage and current ratings. If its a 15v TEC, you NEED a 15VDC power supply that can put out the large current it will absolutely need.
 
You are essentially making a cooler master V10 or Ultra ChillTEC heat sink. You are probably going to get lack luster performance without water cooling the hot side of the peltier.
 
StatutoryApe said:
Remember to check the input voltage and current ratings. If its a 15v TEC, you NEED a 15VDC power supply that can put out the large current it will absolutely need.
Click to expand...

It's 12VDC nominal, 15.4VDC max. If the current draw proves to be too high for the PSU I'll be using I will adapt a 19VDC laptop power supply for the job.

Kardonxt said:
You are essentially making a cooler master V10 or Ultra ChillTEC heat sink. You are probably going to get lack luster performance without water cooling the hot side of the peltier.
Click to expand...

The V10 was actually what got me thinking about TECs again. Low cost is the aim of the project, I'm not going for subzero temps - sub-ambient will suffice. And this goes back to the first comment, low cost includes not having to necessarily upgrade the PSU (within reason, of course).
 
Update: got all the parts, just have to wire it up. The inventory:

mugen.png


Scythe Mugen 3 SCMG-3000 with 2 120mm fans in push/pull, $32 for heatsink and one fan, $10 for second fan from frozencpu.com

pwm.png


12-24V PWM DC Motor Speed Control 10A Pulse Width Modulator Controller Switch, $12.79

relay.png


12VDC 40amp accessory relay, $6.99

tec.png


TEC1-12709 Thermoelectric Cooler Peltier 90W 138.6W max, $8.99; 40mmx40mm copper cold plate from frozencpu.com, $5.49

Total so far: $76.26. Everything is going to be mounted in a 5.25" housing harvested from a dead DVD drive.
 
Fun pic, frosty copper block:

copper.JPG


Just a reminder - I know this isn't new territory per se, but it's new to me; if it's successful and I can help someone else do the same, that's just extra satisfaction. :)

Edit: got the PWM inline and working, a pic:

max.JPG


The number on the multimeter is not the temperature, it is the resistance of a TTF103 thermistor...the max I've seen so far is 40.3 which according to the datasheet of that thermistor equates to about 14°F. Well below ambient, just as I desired. We'll see if it can keep up with a 125W Phenom. :cool:
 
Last edited:
Historically, TEC to air just doesn't work very well. Good luck with your interesting project!
 
Monkey God said:
Historically, TEC to air just doesn't work very well. Good luck with your interesting project!
Click to expand...

Thanks! I'm definitely going to keep this log updated. It would appear that the TEC is not overpowering the Mugen 3 but it doesn't have a real thermal load under it either. I also don't have the other fan mounted - the response from Scythe support about another set of fan clips was not positive...he suggested zip ties which I had really hoped to stay away from.

Sooo...anyone got a spare set of Mugen 3 fan clips I can get? :D
 
Remember that the TEC will generate a large amount of heat to move the thermal power from the hot to cold side.

Your 90w TEC will move 90w of heat from hot to cold @ a nominal 14v input.
90/14 = 6.5a

And it will use 6.5a in the process. This means the actual TEC will be pulling 90w of heat out of the cold plate, and producing another 90w of heat in the process, increasing the thermal load of the hot side to 180w or so.

I sincerely think this won't keep your processor anywhere near ambient temps @ stock clocks, and it will do so with the added expense of 100w consumption.
 
Cool project idea, best of luck!

But yeah, echoing others, I've dabbled with a TEC in the past and you just gotta watercool to make it worth it. And you seriously need a 200W one these days to really see the benefits over regular watercooling in terms of real-world overclocking potential.

The trick for you will be making the hot exhaust air not a problem for the rest of the case. Focus on getting the hot air away as efficiently as possible as this will be the limiting factor. TECs sound like they'll magically get freezing cold on one side and not care about the hot side however hot it gets, but they don't; it's all relative.

Get that heat away! :cool:
 
Last edited:
I had a whole update typed up yesterday and Chrome wigged out on me...meh.

Anyway, I did some CPU load testing over the weekend and had some triumphs and some defeats, as expected and advised - but no matter, I will soldier on. With the mentioned TEC, I first tested on an old 2.0GHz Northwood P4 and at 100% on the PWM I would have frozen it solid; I loaded the CPU up with benchmarks on a Lubuntu LiveCD/USB and the TEC kept it well below ambient and falling. I may make an attempt at volt-modding that board and going after the HWbot.org max overclock at some point, but that's a different project; as it was, I got the old Intel iron to POST over 3GHz with no additional mods beyond cranking the CPU voltage in the BIOS. With that minor victory (the P4 being a ~50W CPU and all) I moved on to an Athlon X2 4000+, which the setup could keep at ambient even under the same CPU benchmark load (65W TDP proves to be the best the TEC can keep up with very shortly). I tore it all down and threw the 9850BE quad in there and this is where the TEC really met it's match; with the PWM at 100% I was getting idle and load temps similar to a capable air cooler alone. This can only mean one thing - a more powerful TEC is necessary! I've got two more TECs coming (I'm really glad they're so cheap) - a 154W 10A unit and a 180W 15A monster - along with a pretty little adapter to simultaneously power up two ATX PSUs, one for the system and one for the TEC alone.

Hey, I've never said this project's aim was efficiency or even rationality. :)
 
Some things I've learned in the last couple weeks: it takes forever for stuff to get here from China if you don't pay for expedited shipping (and then it might be the wrong stuff anyway!), Shin Etsu G751 is junk despite what many reviews say (I'll stick to my tried-and-true MX-2), and I've had a lot of fun with this project. It's probably not going to work like I had hoped, but I'm going to keep trying some ideas and if I run across a cheap H80/H100 I might pick it up and see how bad that heatsoaks. This isn't to say that the Mugen 3 is being overpowered by the TEC, it seems more of a problem of heat transfer from the TEC to the heatsink.
 
Last edited:
Thanks for sharing the update. ;)

Yeah, TEC's are only really effective when their wattage exceeds the source, but you can stack them for ultra craziness.

If you get into TEC stacking though, you will DEFINITELY need to watercool! And have a meaty dedicated single rail PSU to power them. :cool:
 
You must log in or register to reply here.
Back
Top