Fairly Large Project: Oil+TEC Tank PC

[AkkerKid]

This'll be my first post. I'm half way through my project and just found this site today. I'll get you up to speed.
The Idea:
Put the entire PC (everything but the CD drive) into a custom built oil tank cooled by an outdoor radiator and Peltiers (TECs).
Basically I'm building everything from scratch. I've already built a tank, motherboard tray, radiator and blocks for the TECs. The basic idea is like this: The entire PC will be submerged into 5 gallons of the thinnest technical grade mineral oil I can find. The back of the mobo will be facing upwards. All connections will have short extentions going from the mobo to a top plate so that capilary action will not pull oil through the cables beyond that point. When the tank is filled, the warmest oil will flow towards the top where is will pour out of the side of the main tank into an attached reservoir. This oil will go first through the pump and into the hot side block of the TECs. It will be further heated by the TECs and flow outside through 3/8" pipe to my radiator. It will come back in and into the cold side of the TECs and directly into the tank and through the CPU block. From there it will either go through my GFX block or it will vent out to the rest of the case. Some stock fans will remain IE. northbridge, raid card.
The Tank:
The tank is 7 by 14 by 14 on the interior with a slot cut out of a side near the top where the top most oil will fall out into the reservoir on the side measuring 2 by 6 by 6 (all measurements in inches) The tank is completely made of lexan. The mobo tray slide into the case and will have a top plate where all of the external connections will be terminated above the oil. It may also have a handle so the entire PC can be pulled out of the oil whenever necessary.
The Loop:
When the oil heats it'll rise to the top where it pours into the side reservoir. The pump will be submerged in this oil pumping it out of the case and into the custom made blocks surrounding my TECs. (TEC aka Peltier Thermo Electric Cooler is a solid state heat pump. When powered with a fairly large direct current (DC) it will pull heat from one side (depending on the polarity of the current) and put it out onto the other.) I have a pair of 70 watt TECs between a pair of custom made aluminum oil blocks. The warmer oil passes through the hot side of the TECs and outside to the radiator. The radiator is what's left of a broken window air-conditioning unit. The coils in the condenser and the fan still worked so I removed it all and rebuilt it to have four oil paths and hung it from the overhang just outside my window. It has a three speed fan and plenty of other circuitry I can interface with to control various other parts of the system. The oil will cool to ambient outdoor temperature there and go back in the window to the cold side of the TECs and then to the CPU.
What's Done:
The tank, tubing, radiator, pump, TECs, fittings and bought the oil.
What's Left:
I must Seal the hard drives, build better TEC blocks, build the top plate with all of the connections, transfer the components, build circuitry to control the TECs and fan from the signal of the CPU fan header, mount everything. Get more/bigger power supplies for the TECs.

Tell me what you think!

[billbartuska]

Am I missing something? Aren't the hot sides and the cold sides of the tecs going to be in the same oil? Won't the overall effect of the tecs be just heating up the oil?

[AkkerKid]

After the hot-side of the TEC block, the oil goes to the radiator, bringing it to ambient temperature. Then it goes to the cold-side block for further cooling. I only have to make sure the radiator can cool the combined heat of the PC and the TECs.

[billbartuska]

Hmmmm...

I think you're trying to prove a corollary of the oft failed perpetual motion machine law.

If you just pumped the oil through the cold side and then through the hot side. and then set up another pump to pump the oil through the rad wouldn't you have the same thing you are designing? Why not eliminate the tec?

[AkkerKid]

Sorry to dissappoint, but I'm not going to break any laws of thermo dynamics or physics in general for that matter. Hopefully, this image will clear the idea up some.

[billbartuska]

Originally Posted by AkkerKid

Hopefully, this image will clear the idea up some.

Hmmmm.....

The reason people use TECs for cooling insn't because they are efficient at removing heat. Actually, they are very inefficient. In the example below, the TEC would actually be adding about 200 watts of heat.

The primary reason TECs are used for computer cooling is because they will operate at sub-ambient temperatures, not because they are better at removing heat than air or water cooling. Removing 100 watts of heat from 5 gallons of oil dissipates the same amount of energy whether the oil is at +50°c or -50°c. A TEC, a heat sink/fan or a water cooling system that will remove 100 watts of heat all remove tha same amount of heat, the TEC just does it at sub ambient temperatures. Both air and water cooling depend on the temperature differential between the hot side of what is being cooled and the ambient air temperature, and as such, can not cool to temperatures below the ambient air temperatures.

But, since the temperature of the cold side of a TEC depends on the amount of electrical power added to it (not the ambient air temperature), surface temperatures far below ambient are achievable. The penalty though, is that the added heat of the hot side of the TEC (in addition to the heat from whatever is being cooled) must also be dissipated.

In your application, having the surface of the cold side of the TEC at a sub ambient temperature serves no purpose. The surface of whatever you are cooling (the CPU I assume) can not be any cooler than the temperature of the oil, and the temperature of the oil can't be any lower than the ambient air temperature. If the oil was cooler than the ambient air temperature, the radiator would heat it up to ambient. If your goal is to increase the delta T (oil to air) across the radiator with a TEC, and thereby gain added efficiency, I think the net/net effect would be negative because, as I said, the TEC will add quite a bit more heat than it will be removing, and whatever gains you get from radiator efficiency will be more than offset by the added heat from the TEC that needs to be dissipated.

By the way, you have some amazing construction skills there.

[AkkerKid]

What I'm hoping will happen is the TECs will cool the oil going to the CPU colder than the CPU can bring back to ambient. That way even after the oil has passed through the CPU block it will still be cool and keep the overall temperature of the rest of the PC fairly low.
On another note, I was thinking about your option C (the one with two separate fluid loops) But I remembered that when a TEC goes bad (and, trust me, they do) they become very good insulators. In that case I'd be depending on one TEC to transfer all the heat the entire computer makes to the radiator side... What I'm thinking is this... The computer uses anywhere between 300 and 400 watts. Just about all of the is going to heat in the case after everything's said and done. I don't want to depend on the TECs to remove all of the heat, so rather than have two seperate loops on either side of the TECs, I wanted to only use the TECs as a small supplemental temp drop right before the CPU. If they turn out to be worth it and actually drop the temp a few degrees below ambient, I might add more TECs to the loop.
Just a side note: the radiator is made for the hot side of an 800 watt A/C so it should get rid of many more watts of heat then my pc could put into it.

[billbartuska]

Originally Posted by AkkerKid

What I'm hoping will happen is the TECs will cool the oil going to the CPU colder than the CPU can bring back to ambient. That way even after the oil has passed through the CPU block it will still be cool and keep the overall temperature of the rest of the PC fairly low.

Yes, the TECs may cool the the CPU loop to below ambient (depends on the efficiency for the "large" rad), but overall the TECs are going to be adding heat to the loop. All the electricity that goes into the TECs will be converted to heat in the loop. It has nowhere else to go, unless you use a second loop or air cooling to dissipate it.

Interesting hypothesis though.

You may find this thread interesting.

[ben333]

Any updates?

[AkkerKid]

I've been able to scrounge up a larger pump. I haven't any idea of the spec but it can shoot a column of water 0.5 inches wide about a foot up at full power. It was used in a medical device a relative needed that pumped ice water through a pad that strapped around the shoulder. I'm thinking it'll be much better suited to push the thicker oil through the many feet of tubing I have than the Swiftech MCP350 I'm testing now. I've since cut down all of it's useless mounting hardware so it will fit within the reservoir properly. I'm not exactly sure how long it will last either. It is "medical grade" but to me it looks like the dc motor from a power-wheels car in a pretty white case. I've got to figure out what I need to adapt it down to the sized of tubing I'm using. Maybe, with the higher flow rate it's got, I won't have to make new blocks for the TECs.

[ben333]

Well if you end up needing more power, try looking at a PetSmart for some marked down pumps. I picked up a 750 GPH pump for only $38 (marked down from $95) thats working very well on my main rig with the radiator over in the window. I also picked up a pump that normally costs $60 or so for $17 that is working quietly and with great flow for my HTPC.

[AkkerKid]

I'll be trying out this new pump soon. I didn't think of a pet shop for pumps. If this thing doesn't work out (I doubt that'll be the case) I'll look further into other sources for pumps. I do believe this pump would add a lot of noise to any power system I'd connect it to, so I won't be connecting it to the PSU for power. It' only seems to want to take a little over an amp at 12Vdc so it would be alright if I powered it with the same supplies that I'll be powering the TECs with. I might even have to PWM it to keep the flow rate and pressure within tolerances.

[billbartuska]

Originally Posted by AkkerKid

I didn't think of a pet shop for pumps.

Tropical fish.

[ben333]

Well in response to the thread Bill linked to above, the people there seem a little bit off on the idea. The poster of that thread is talking about cooling a radiator with another larger radiator, when he really should just be using the large radiator directly in the first loop. But here, where its easier to pump water outside to a large radiator and back than expensive mineral oil, it's worth the possible performance loss to use oil just where its needed to be non conductive and water in the rest.

Akkerkid, when you finish this project you should do a write up with pictures of building and testing your cooler. If you're interested, maybe Joe would publish it here.

[ben333]

Updates?

[ben333]

Any progress? Intereted to see how this will work