Pelt Water Chiller

I'm Making an in line water chiller, that will use 2 110 watt pelts. The hot side of the pelts will be cooled by a copper heatsink I'm making, which will be similar to a really long SLK, minus the quality. The cold side of the pelts will chill the water via a custom water block I am also making from copper.

The pelts are both 4cm x 4cm, and the waterblock I am planning on making will be 5cm x 12cm x 1cm. In addition to this main waterblock, I am thinking of making a second larger waterblock to attach to the main one. I have a piece of 3/4" aluminum which would be ideal for this. The idea behind the second waterblock is for the water to spend more time inside the cooling aperatus. The water would come from the pump and go through the first aluminum waterblock to cool it a little, and then through the second copper waterblock for further cooling.

I have tried to attach a picture of some screen shots of what I am planning on making from SolidWorks. To give you an idea of the size, the fans are 80mm. If I were to use the second aluminum waterblock, it would be attached directly to the bottom of the copper waterblock.

Does anyone know if the second aluminum waterblock would just be redundant? If the copper waterclock can transfer heat from the water fast enough then there is no need for the aluminum one.

Also, does anyone have any tips, or reasons this is the dumbest thing you have ever seen in your life? Anyone wana Host pictures so you can all see what I am talking about either?

-edit-

Nevermind about the hosting, the site hosting looks like it will do.

I have 2 things that you may want to do a little different.

1) The water flow should go from R to L or L to R not turn back and cross the same TECs twice with a lower temp on pass 2.

2) The water deflectors should not be parallel to the water flow or should at least be staggered, so as to create turbulence.

yea, the idea behind the water turning back and passing the tec's twice was to try to get about the same load and hot side temps on both pelts. My CPU load is always at 100%, and everything else is constant, so the whole thing should be more or less at a steady state. This means I am more concerned with the maximum amount of heat I can suck from the water over time, and not so concerned with the temp drop from a single pass. (I know these are essentially the exact same thing, but you get where I'm going here)

The chunks sticking out arn't for diflecting water, they are for increasing the surface area, like the fins on a heatsink. I guess befor I make the thing I should figure out roughly what kind of heat transfer coefficent I will be getting from the water (Renolds numbers and the like) and the appropriate thickness for the copper fins sticking into the water. I didn't bother putting all the detail into the model, the fins in the water will be tapered, getting thinner as they get further from the pelts. The thing is going on a really low flow system as well, so the inside of the block will have to be pretty rough to get a turbulent flow. I'm thinking it will probably have about a .5 mm boundary layer. I'll see what I can come up with to mix the water a little more between fins. Most of the flow will be laminar, but it will be mixing with the turbulent flow and spending enough time in the waterblock that I should still get decent performance.

Deflectors was a poor way to say it. I understand that they also add more surface area, but should also create turbulence as well. Water has a high heat cap. and a low thermal conductivity therefore needs some turbulence for good heat exchange.

I know what you mean. I guess I should rearange the deflectors a little. The design I have in the picture was probably influenced a lot by the fact that I drew it in solidWorks which I have only been using for a few hours.

Due to the really low flow rate I won't be able to get fully turbulent flow across the entire waterblock. The best I can do is to create a thin turbulent layer against the fins, and then try to design them to promote mixing between fins,

Ok, here is a new design for the pathway through the waterblock. Solidworks wasn't getting along with me too well so I couldn't actually get the pathway put into the model. Maby I shouldn't have tried to do the whole pathway as one cut... oh well... here is an overhead view showing the new pathway. The ends arn't done... one end is totally missing and the other one is really messed up... i'll do something about those later. In the picture, white is the area cut out from the block where water will flow. I think I have designed it to provide quite a bit of swirling and mixing.

OK, now I'm just talking to myself, but after looking at that last design I decided it was more or less crap. I have decided to take Doc's advice and not make the water double back. I have also made the flow path much narrower and longer. The idea behind this is to increase the velocity of the water in the cooler, which will raise the Reynolds and Nusselt numbers to increase the heat transfer coefficent. Here Is a Pic of the new flow path. The grey boxes are roughly where the pelts are going to be sitting.

Does anyone have any idea if I will still need to use the radiator in my system? My CPU puts out 84 watts at it's current settings, but after this project and several other cooling projects I am probably going to be putting out somewhere around 100 Watts. Will two 110 watt pelts be able to disapate this much heat or will I have to use the radiator to help them out?

Looking good! This should work better than your first design. 2 110W TECs should be good to go for your heat load, unless you are trying to get to some extreme like -50C.

nope.... i'm only trying to get my processor to ambient at most. I don't mind insulating the waterblock (the one on the CPU) and the water lines, but I'd rather not have to seal and insulate the soccet. If It does drop below ambient, it's just gona get more voltage and speed to bring it up again

Well.. I have about 650 holes to drill now....better get too it.

The second design is doing to make you much happier. Also it is good to see someone actually doing the math and thinking for themselves instead of just doing what looks good. You will have to keep us posted on your progress and I am definatly interested in your temps. If they do not go to far below ambient, I may follow your footsteps. OH and I like your way of keeping it warmer.... Keep overclocking untill it is back to ambient , now that is a perfect example of a true OCer thinking process. Keep up the modding and keep us posted. Oh and PICS!

-JS-

Very nice design. It is actually one of the methods i am considering for my cooling. How can i figure out how many watts my cpu puts off and do you think an 80W pelt with the radiator would get some nice temps?

I have a program called "cpu power 0.21", you tell it whay cpu you have, and what vcore and clock it is running, and it figures out the thermal output.

You might get OK temps with the 80 watt pelt and a radiator. What you do is run the water from the cpu to the radiator, which will bring the water close to ambient temperature, then run the water through the pelt chiller, which will hopelfully lower the water temp to below ambient. The water then goes to the cpu, and back to the rad.

oh, the 3200+ barton has a thermal output of 76.8 Watts at stock settings.

AWESOME! and thanks for the help. that sounds about right with what i'm thinking of doing, although i'd have cpu > gpu > nb > rad > pelt. i REALLY NEED to cool at least those first two because i'm in texas and it gets hot a hell here even in the winters! also, i'm looking for quite a large radiator as i would like to build my own case to put this current box in with an extra compartment up top for rad, fans, pump as someone in the mods section is doing. thanks again for the help and would you suggest i use a 110W instead of 80W? by the way, keep me updated on how your chiller turns out and is coming along.

Also, would both the 80W and 110W pelts be cooled sufficiently by a HS/F for which ever i decide to get?

a good copper heatsink and fan should be more than enough to cool a 110 watt pelt. The bigger the pelt you use the better performance you will get. I am using 2 110 watt pelts mainly for the reason that I had 2 crappy 300 watt power supplies with about a 12 amp max on the 12V line, so they will work perfectly to power the pelts.

I spent about 5 hours drilling holes for the chiller... only got about half of them done .... this is gona take a while. I'll se if I can find a digital camera somewhere so I can show you my progress.

yeah i want to see it,
are you just using a drillpress or something similar
or are you going balls to the walls and free handing it


~Magick_Man~

well... mostly with a drill press... but I use the term very loosly, my drill press was $30 canadian at Canadian Tire

It's being quite the b*tch to do since all the holes are intersecting, which causes the bit to wander over to the hole beside it, and I don't have any clamps, and with this many holes clamps would take too long anyway. Anyway, that's enough whining from me, time to get back to work.

I don't have a digital camera right now, but I do have a crappy WebCam!!! all will marvel at the quality of the picture below. It has to be at least a few dozzen pixels

I think you need more beer while your drilling those holes. Should have lots of turbulence, can't wait to see how good it works. Keep us posted.

That, is a sweet idea. Already filled my head with them

~t0m

any more beer and the holes will be in my hand......well.... more holes in my hand....

It doesn't show too well in the picture, but the path is more than a little turbulent.... I'm thinking I might need a better pump to get the water to go through the damn thing....

Been having some interesting problems with the "drill press". The bits were seizing, despit everything being liberally coated with 15W-30, so I changed the belt, give it less speed, more torque. This results in the bit still seizing, but the chuck not stopping. Obviously this isn't soo good.... crazy amount of torsion in the drill bit....bit pretty much explodes....one chunk of it shoots into my hand so hard that it punctured my palm, and then blood becomes my lubricant of choice. 15W-30 working well in my circularoty system though. Good times all around, I'm about 80% done now.

Anyone know anything about soldering copper? I have done tonnes of soldering for electrical things, but never structural. I'm assuming there are different solders for it. All I need to do for the waterblock is the back plate thing and the pipe connectors. Should be good practice befor I build my "thermalwrong SKL-2000"