Different peltier idea?

dis · Post by **dis** » Sun Jul 04, 2004 5:44 am

Has anyone seen or got any insight on building a voltage regulator for a peltier device? I'm hoping i can rig something up to ensure that a TEC won't cool the cold side to more than a few degrees below ambiet tempretures.

The motivation is that my radiator is currently the main noise maker, due i assume to the number of closely placed fins and air resistance caused by them. I figure that if i remove the rad and replace it with a second water block, with a TEC on top then an open fin heatsink (maybe one of those nice heatpipe based coolers) i can free up some space in my rig and reduce the airflow noise considerably.

The only part i'm worried about is regulating the pelt and also supplying it with power. I've seen dedicated TEC power supplies on the net and i assume a 230Watt pelt will need its own supply, or wired to a BIG main PSU, 550Watt etc.

Any thoughts?

Nowhere_man · Post by **Nowhere_man** » Sun Jul 04, 2004 6:37 am

That's an interesting idea. I dunno if waterblock/pelt would have enough surface area (even with a heatpipe/heatsink) to properly cool since the water volume and surface area of the waterblock is much less than a radiator and the water passes through it quicker than a radiator.

dis · Post by **dis** » Sun Jul 04, 2004 6:46 am

Yeah, i had given some thought to that and its pretty interesting. At the moment my water temps could be ambiet +10 so the rad needs a lot of airlow to cool back down to room temps. If the delta was 40 degrees i'd be able to disapate much more heat with the same airflow but the water temps would rise accordningly.

Usign the TEC (assuming a 230Watt one can cope) i'll have water temps of -5 ambiet and the heatsink should becomme very hot, up to the point at which i can shift the heat very rapidly. The question is when this settles down and what rating of TEC would be needed since the larger the detla between cold and hot sides the less effective it becommes. I don't see it being a problem though.

I'll probably give it a try when i get a 64 bit chip, that'll need a new waterblock and the old one can be used for mounting the TEC. Before that though i still need to work out the voltage regulator, otherwise it'll mean frozen water and dead PC bits. I'm surprised none of the commercial watercooling products take this approach since it makes it more compact, cooler and quieter.

Nowhere_man · Post by **Nowhere_man** » Sun Jul 04, 2004 8:19 am

Something else you might consider, the Reserator is down to around $200 (US) online and that would be less expensive than the pelt, powersupply, and related items required to implement your idea.

Of course it wouldn't be as much fun as sorting out your solution.

dis · Post by **dis** » Sun Jul 04, 2004 8:39 am

I did think about it but my hearts set on an internal solution, that what i've got at the mo and preferably one i might even be able to shrink down for a SFF pc.

Having said that, theres something to be said for a big blue tower - it does look kinda cool.

Nowhere_man · Post by **Nowhere_man** » Sun Jul 04, 2004 9:25 am

You know dis, this link might be helpful to your quest, specifically The MCW-CHILL 452™

Post by **Rusty075** » Sun Jul 04, 2004 12:36 pm

Um, guys...

Did anyone read Review: ActiveCool AC4G Thermoelectric cooler ? it does exactly what you're talking about: adjust the pelt's voltage to keep it from dropping below ambient.

jinu117 · Post by **jinu117** » Sun Jul 04, 2004 5:28 pm

IMHO, Peltier = expensive.
Unless you want the extreme temperature, waterchilling = waste of time/money/effort.
If you are looking for above freezing point temps, stick to direct die peltier setup. Swiftech makes nice blocks for them. I used to churn out 3.7ghz out of old 2.4b c1. Cost of whole setup back than with dual loop, two pumps, 2 peltier blocks, few heatercores was about $4-500 I think. (You do need spare PSU for the pelts

)
Now, it was rather quiet due to immense size of radiators I had (2 large heatercore with 120mm shroud + 1 BIX) and rather "quiet" pump 1048+1250. But the heat it generated drove me nutz. I really didn't have to heat up the room in middle of night during winter time with this system running. Something worth considering for any peltier or active cooling project.
If you still want to go ahead, I suggest getting some basic stamp, code it and use the io function with some sort of votage controller to control the power output of peltier PSU. Most of peltier PSU will have adjustable pot which you can find out the tolerance and make circuit to emulate the pot with basic stamp. Shouldn't be too hard.

Nowhere_man · Post by **Nowhere_man** » Mon Jul 05, 2004 5:26 am

Rusty075 wrote:Um, guys...

Did anyone read Review: ActiveCool AC4G Thermoelectric cooler ? it does exactly what you're talking about: adjust the pelt's voltage to keep it from dropping below ambient.

And from the article to which you refer:

but its noise performance is not yet up to SPCR standards.

dis · Post by **dis** » Mon Jul 05, 2004 6:08 am

Yup, i read that review a while ago. Seems like a good idea thats been let down by a bad implementation.

My current concern is A) the extra PSU and B) that adding a 230Watt pelt means i have 230Watts more head to disapate. Emptying the heat into my room isn't ideal but its more a worry about how capable the heatsink will have to be.

Anyone know what rating of TEC i'd need to cool a 64 bit AMD chip? I'm guessing it'll be a 230 one and that might just ruin my plans. I'm not wanting an extreme overclock just something reasonable, at the mo i can take my 2500 Barton up to 2350 from 1866 without many problems.

The only other way to keep my overclocking and pursuit of a quiet PC in harmony looks like building a custom radiator out of copper pipe and fins with a very open design. Funnily enough it'll probably work out cheaper than the pelt idea but won't be near as compact.

Post by **Rusty075** » Mon Jul 05, 2004 11:13 am

Nowhere_man, yeah, I was refering to the products' concept, not necessarily it implementation. The concept has some potential, but needs some serious work on the implementation.

jpsa · Post by **jpsa** » Thu Jul 15, 2004 3:33 am

That would be worst than just slapping the heatsink on the cpu...

You'll have cpu heat transfered to the second waterblock, and this 2nd waterblock is cooled by the pelt and the pelt cooled by the heatsink right? This heatsink will be dissipating CPU heat + water pump heat + pelt self heat... you can try it, but I'm almost sure that this very same heatsink alone directly on the cpu (without pelt) will work better.