Here's a description of my latest quiet PC cooling project. I was running a computer with a Zalman "flower" heatsink with a 120mm NMB fan blowing on it, but even with the fan slowed down, it was too loud when running at a CPU temperature under 50°C.

The system:

• ECS K7S5A motherboard
• Athlon 1.4 GHZ CPU
• 256 Mb DDR RAM
• 80 Gb Barracuda IV hard drive
• Plextor 12/10/32A burner
• Generic 48X CD ROM
• Generic floppy
• Aopen GeForce 2 video card
• Generic case
• 300 watt Codegen power supply-fan replaced with 80mm Pc Power and Cooling "Silencer"

I had water cooled once before using a homemade copper water block, Danner pump, and a Toyota Supra heater core as the radiator with a 120mm fan blowing through it, but the heater core fins were too closely spaced to get a good air flow. To keep the temperatures down meant running the fan at full speed. Not quiet. For this project, the waterblock and pump were used again.

Early on, I decided to try to make a fanless radiator. The inspiration came as I was looking at the radiators in our house. We have hot water heat with baseboard radiators. They are a copper tube with aluminum fins along the length of the tube. At one point, I was just going to buy one of these and mount it behind my desk. Without a fan, though I didn't feel it was the right solution. The answer was to mount a radiator in a chimney and use the difference in temperature at the ends of the chimney to generate an airflow.

The formula used to calculate airflow in a chimney is:
Q = 60 * Cd * A * sqrt(2 * g * (Ht - Hb) * ((Ti - To) / Ti))

• Q = flow rate in cfm,
• Cd = 0.65 (for unobstructed openings)
• A = opening area, square feet
• Ti = indoor temp (Rankine) (The temperature of the heated air measured at the top of chimney...Joe)
• To = outdoor temp (Rankine) (The temperature of the air entering the chimney...Joe)
• g = gravity.

Here's a website that explains stack effect and has a Java calculator to make calculations a lot easier:
http://chuck-wright.com/calculators/stack_effect.html

**Thanks to Rusty075 for the formula and website**

Plugging in the dimensions and measured temperatures for the completed radiator gives a theoretical airflow of 14.28 cfm. Not bad for no fan. An interesting note is that as the temperature of the cooling water rises, the airflow rises, too. More airflow in response to more cooling needed. Very nice...

The completed unit is made from two boxes of 1/2" particle board glued with yellow carpenter's glue. Carpeting is used in several places to quiet vibration from the pump. It's filled with about three quarts of distilled water. Click on any of the pictures to enlarge them.


The top box houses the radiator and is 8" wide x 8" deep x 31" high with two 4" x 6" holes for cool air to enter near the bottom. It has a layer of carpeting on the inside bottom. In addition, the boxes have a layer of carpeting between them. The rectangular clear box is the resevoir, made of 1/4" thick acrylic.


The radiator is made of four 3/4" copper tubes with seven 1-3/16" x 12" copper fins soldered to each tube. They are connected with soldered copper elbows and form a continous loop with the resevoir inline after the first tube.


The bottom box houses the pump and is 10" wide x 9" deep x 8" high, covered on the inside with carpeting. Foam rubber is used to seal any openings in the box. The black strip on top of the board is 1/16" thick black self-adhesive foam from WalMart.


The pump is mounted to a board that drops into the bottom box. The pump is suspended on small stretch cords to isolate it from the board. The vertical wood strips are just there to hold the board up for photographing. The black layer on top of the board is more of the WalMart foam.


The cooler is connected to the waterblock with 1/2" silicone tubing, reduced to 3/8" at the waterblock. The water flow goes from the pump to the radiator/resevoir to the waterblock and back to the pump.

So, how well does it work? As I'm writing this, the computer has been on for five hours and MBM5 reports the CPU temperature at 44°C with a 21°C ambient temperature. The only noise is a slight hum from the pump and a bit of noise from the power supply cooling fan, the only fan in the system. The CPU temperature rises as high as 47°C under stress.

I'm quite happy with this setup, but wondered how the cooler would perrform with a fan placed on top. I did some testing using the SiSoft Sandra Arithmetic Benchmark Burn-In running for 15 minutes to stress the CPU.

Three tests were done, first with no fan, then with a 120mm fan running at 8 volts-about the same noise level as the power supply fan, and last the 120mm fan running at 12v.


The test schedule:

• 0 min No fan Cpu at normal operating temperature
• 0 min Start burn
• 15 min Stop burn-in
• 30 min Fan at 8 volts
• 45 min Start burn-in
• 60 min Stop burn-in
• 75 min Fan at 12 volts
• 90 min Start burn-in
• 105 min Stop burn-in
• 120 min Stop test

The results were not as dramatic as I had imagined. There was only a 2°C differencu in CPU temperature from no fan to really noisy fan. The following graph shows the results. The temperatures are in increments of 1°C because that is the limit of accuracy for the motherboard sensors.


It looks like I'll be using the cooler for quite a while. It is quiet enough and cools well enough for me. In the future I'd like to make a smaller resevoir and veneer the outside to make it look a little better.

Very interesting DIY project. Somebody finally did it! Worth a write-up to turn it into a full-fledged SPCR article. Email me if you're interested.

what are teh dimensions?

JoeG11,

Color me seriously impressed.

You have got to do a write up on this for Mike.

Thanks, all.

I'll get in touch with you in a day or two, Mike. I'd like to borrow a better
camera to take some new pictures.

DryFire, the top box is 8" wide x 8" deep x 31" high. The bottom box is
10" wide x 9" deep x 8" high.

It's something I've never seen done before. It took about a week
and US$35 to make. About half of that was for the silicone tubing.

Fascinating project. I wonder what would happen if you used an 6-8 foot piece of pvc pipe for the chimney instead of a 31inch box. Perhaps since the air at the top of a room tends to be warmer than at the bottom, there wouldn't be enough temperature differential for significantly improved airflow.

How/where did you get that formula?
I'm not sure how to interpret it. Are those temperatures defined to be measured inside or outside the chimney? I would have guessed that if you want to work out the flow through a chimney (with hot air going in at the bottom) using only two temperatures, those would be the inside and outside temperatures.
You say Tt=68.4 and Tb=74.1, so (Tt-Tb) is negative so, since H is positive, H*(Tt-Tb) will be negative, which is a problem since you need the square root of it. Did you get the temperatures the wrong way round or should it be Tb-Tt or is the confusion from something else?

Anyway, congratulations on building the thing. It's very encouraging to see that it works so well: the experiment with the fan makes me think that the fanless flow rate must be impressively high.

It would be interesting to see water temperatures too, if you can find a way to measure them. Perhaps an ordinary old-fashioned mercury-in-glass thermometer in the reservoir? (Well, probably not mercury but that alcohol with red stuff in it, or whatever they use.)

grnarrow, that's a good idea if you could adapt the radiator to fit the tube. My first try at this had a chimney and radiator 12" longer than the total height of the finished design with the pump mounted in the same box as the radiator. It was just too large for the room and needed more sound deadening as it had a nasty hum. So I cut it down and added the second box for the pump. The pictures don't show it very well, but the radiator is made of four sections of copper tube/fins soldered together. They form a close to square cross section, that's how the box became square.

Bat, you're right. I mixed up the two numbers. I'll edit the original post to fix that. I measured the temperature of the air going into the rectangular hole on the bottom and the temperature of the air coming out of the top. I believe the formula has been around for a while. The formula is about half way down on this page:
http://www.ece.villanova.edu/~nick/usenet/00001730

It takes about half an hour to come up to normal operating temperature. When it gets there I'll measure the temperature of the water in the resevoir and post it.

very well done

Thanks, johnoh.

Bat, here's your temperature. MBM5 reports a CPU temp of 43°C and the water in the resevoir is at 29°C. Maybe a better waterblock would help the CPU. I'm working on a new waterblock design with a .051" thick copper bottom plate and a 3/4" thick clear acrylic top chamber. The area where the water flows looks like this from the side ^^. I'll post more information on it when it's done.

Very nicely done. Nice to see yet another custom case bodger in the ranks.


I do question your math for the stack effect though.



Using that formula and your numbers I get 14.28CFM

But having said that, Who Cares? as long as it works that's all that matters. I gree that the radiator design is probably the weakest link. What about replacing it with a couple of automotive heater cores placed diagonally inside the chimney?

Oh, almost forgot, here's a site with a good explanation of the stack effect, and a little java calculator to do the math for you:

http://chuck-wright.com/calculators/stack_effect.html

Thanks for that web site, Rusty. It looks like I have used a formula meant for some other use. I'll change the original post to use your formula. 14.28 cfm seems more like it. When the cooler is heated up, I feel a slight but noticeable flow coming out the top.

I think the heater core idea would work as long as the fins did not obstruct the airflow too much. When I calculated the area of the chimney, I subtracted the area of the radiator from the area of the chimney. You would probably need a fairly tight fit between the inside of the chimney and the heater cores. The efficiency of your radiators may overcome the loss of airflow. Sounds like a good project.

My first water cooler used a heater core from a Toyota Supra with fins that were very closely spaced. The 120mm fan attached to it barely moved any air through it.

I think you'd want a thin radiator/heater. A motorcycle radiator might be a good choice. Car radiators tend to have a lot of thin flat tubes with fins around them. For something like this, the same idea but with far fewer fins or none at all might be better: poorer performance for a given air flow rate, but a much higher flow for a given air pressure difference across the radiator.

Thankyou for the water temperature. (Water temperature in degrees Celsius to the nearest degree, air temperature in degrees Fahrenheit to the nearest tenth!) I worked out the conversions: the water is about 10C warmer than the chimney intake air. Certainly respectable for something silent.

Its called "stack effect", and all this time I have been searching using "gopher den convection cooling"

I'm wondering if you will get accelerated flow with a converging duct (largest at bottom).

A note on temperature measurement. All readings except CPU temperature were done with a cheap Radio Shack indoor/outdoor digital thermometer that reads degrees F and C to a tenth of a degree. I'm not sure of its accuracy, but within the narrow range of readings, I trusted it to give a relatively good accuracy.

fmah, I'm thinking you could look at a project like this as an art object. Once you've got the basic operation designed, add some decoration or change the shape to give it some character. Just be sure to keep the SAF in mind when designing. Spouse Approval Factor. Those of you with signifigant others in your life will know what I mean.

Got a little off the subject there, lol. A converging duct would look cool, not sure of how it would function.

I just remembered that allot of people back at hardforums build those kind of things but usually with pvc. Although it's not really the same. it houses a submerged pump and acts as a resevoir and has attached radiators.

Tower of Cooling Power Update

Everything has been working fine for four weeks now, but I have an itch to improve on the original design. The top box transmits a bit of vibration from the bottom pump enclosure. I'd like to mount the pump and radiator in separate boxes again, but connected by tubing instead of stacked.

I also have a feeling I could improve the performance of the radiator by having more surface area. I'm thinking of building a flat copper radiator with cooling fins mounted to it, similar to a large heatsink which could be mounted into a narrower tower.

Time to get the torch out again...

Hey, great project.

Hmm, if you get more then a few degrees difference between water temperature and CPU temperature, something is going on with your waterblock, or the water isn't moving. I got 20°(C) difference when I didn't apply enough pressure. More normal difference is 2-3°C (I think, i875 is too new for lm-sensors, so I can only check in bios). I use a Danger Den Maze 3.

If you make your own waterblock, don't make the bottom plate too thin or you wont be able to apply pressure to it.

Also, To is probably not the air temperature at the enter of the chimney, but the air temperature at the top outside the chimney. So Ti-To is the temperature difference at the top of the chimney.

ive planned todo something similer with some rads/heatercores. i will get it done soon . itll go underneath/behind my desk. all gonna be big things . oh wouldnt a top thats larger work better?

Hi, ping and Mirar...

ping, please post some pics of your system when it's done. Sounds like a cool project. You're right about a bigger chimney. I ended up with the final design after some trial and error. It is a size small enough to look good in my room and still give acceptable performance.

Mirar, I've always felt the cooling system could be more efficient. The waterblock is an old design I made for 3/8" ID hose with internal dimensions to match, but the tower cooler is 1/2" ID hose fittings and is reduced at the waterblock. I'm working on a new waterblock with larger fittings and a different water passage. .051" thick copper bottom plate and 1" thick acrylic top plate. I'll post the concept drawings when it gets farther along.

I'm still wondering about the effect of head in this thing. We debated it here:

http://forums.silentpcreview.com/viewtopic.php?t=4306&highlight=

The pump has a 30" head right now. I don't have a way to measure flow rate, so it's still an unknown. I'm also working on a new radiator that will drop the head to about 12".

Here's a picture of the latest change to the original tower. It has new legs to completely decouple it from the pump box. It is now very quiet. Click the picture for a larger view.

ill picture what i can, but itll be behind my desk which is against the walls and my bed
and itll be constructed behind there i rekon, well the tower of cooling anyways
ill find urls for pics of my rads ive uploaded in the past

pics
http://www.emotiveexcellence.com/pingu/ ... age_01.htm
the white thing with wires is a 40x40mm tec i think
the monkey is the future world leader

http://www.emotiveexcellence.com/pingu/water/pumpres/ <uber pump 10,000 a lil noisy so ill stick to dualy eheims to start with

http://www.emotiveexcellence.com/pingu/ ... age_01.htm uber pump testing. 1250 it is against...

there

The head is irrelevant. That's why siphons work: all that matters is the difference in height between one end and the other, and for you there aren't any ends: it's a closed system.

head does matter, there is resistance in the system

That's not the same thing at all.

joesgarage11 was concerned about height (head) differences.

Resistance (from friction) matters of course. It causes back-pressure in response to flow, and this is sometimes called "head loss" because pressure is sometimes measured in terms of the head that would give that much static pressure (at zero flow).

The point is that if you have a U-shaped piece of tubing full of water, and both ends are at the same height, then the resistance will be the same regardless of whether the U is right way up, upside down or lying flat. The same goes for pumping water around a closed loop: the relative heights of the different parts don't matter.

That's some nice pics, ping. Looks like a mighty powerful cooling solution. Will you be using fans on the radiator?

I got an email from a farmer that saw your pump in action...He wants to use it to irrigate his corn field .

Just kidding about the farmer.

lol yeah the pump is uber, sucks 550watts tho :O. im hoping not to use any fans at all, but i will probably have some on a fan control thing, i think i had a small thermal runaway when i tried with the tec, sorta like a milk float police chase
pump sucks dry that 3gallon tank in around 10sec btw :]