Geothermal cooling with a 30 gallon steel drum buried in my crawl space

D

Delexicus

n00b
Joined
Feb 11, 2013
Messages
21
Edit: updated Schematic to v2

geothermalcoolingloopv2.png


So I finally got fed up with all the noise and dust build-up of my air heatpipe coolers and decided to make my water cooling debut. But if I'm going to go water I might as well go extreme, right? I've always wanted to try geothermal cooling for low noise and overclocking potential but was put off by the sheer amount of manual labor plus copper pipe soldering that would be involved. Then I came across this sweet setup made by Sverre Sjøthun at DWPG.com. Definitely not for the faint of heart but I thought I could simplify the setup a bit to suit my needs.

I decided to bury one of these guys into my crawl space under my house. I would have gone for a 55 gallon drum but my crawl space opening only measures 22 inches at it's narrowest, and that's just a hair too narrow. It's pretty tall down there measuring about five feet (six if you stand between the beams) so it's not too inconvenient to work in.

The computer to be cooled is thankfully on the first floor so I won't need to consider running pipe work through the walls. The ground and air is always at around 55F plus or minus 2 degrees in the crawl space so I don't need to dig any deeper for steady temps.

The drum comes with two threaded openings so that simplifies connections a bit. A 2" and a 3/4". I'll have the 3/4" be the hot return and the 2" be the cold supply. I'd like to make the 3/4" hot return have a nozzle protruding down into the drum to direct the water towards the drum wall at an angle. This should help make a circular current to help reduce the boundary water layer, as well as increase water flow speed with the steel heat transfer surface. That should theoretically help transfer heat to earth faster. The cold supply will have an extension bit on it so it will always take water from the bottom of the drum. I'm not sure if I should also make a nozzle for the cold supply so the water is being taken at an angle towards the drum wall or if it should just make it pointing straight down. So far I've just been guessing at the fluid dynamics so if anyone here has experience or suggestions in this field I would appreciate any feedback you may have.

It also occurred to me to pack some nice fine grain top soil around the drum to seal it in nice and tight.

This will be a total system cooler and that means CPU, GPU, PSU, CPU Mosfets, chipsets, hard drives, SSD, and ram will all be liquid cooled in some way. There will be NO fans whatsoever. I'll be using the extra space in the case to either add sound dampening material to silence the hard drives, or add a radiator and pump for when I want to take this thing backpacking. :p

If everything works out well I may consider extending this system to other PCs in the house. That may increase the equilibrium temp in the loop but heck, I could always sink another drum down there in the future.

Everything so far has just been a lot of thought and mental experiments on what would would give me the the coolest temps with geothermal without either A) making it noisy, or B) making it ridiculously expensive. Any thoughts, suggestions, or comments would be greatly appreciated.

For those that are curious, my planned computer build will include:
CPU: Intel Core i7 3770K
GPU: either a Geforce 580 or 680*
Motherboard: ASRock Z75 Pro3
RAM: 4x8gb G.SKILL Ripjaws DDR3 1866
Case: AZZA Genesis 9000 Black
PSU: KINGWIN Lazer LZ-1000
Hard drives: six or seven inherited from previous builds
SSD: OCZ Vertex 4 VTX4-25SAT3-512G.M 2.5

*I already have aftermarket air cooling for my 580 so I'm hesitant to ditch it. Plus it's proven to be a sour overclocker.
 
Last edited:
So I had a bit of a eureka moment thinking about my project. I was driving to work at night and stopped at a red light. I was letting my mind wander while lazily looking up at a street lamp. After awhile my eyes hurt a bit so I squinted and I noticed the usual star pattern appear around the bright light. Then it just clicked. What if I welded some fins to the steel drum for added surface area? I did some digging and found out that it's actually not such a crazy idea after all since it was done in this passive cooler too. I'll have to work up some CAD drawings to get my thoughts in order. I'll make sure to post some pics when I have more to show.
 
You're going to have 30+ gallons of water to soak up the heat produced by a system that'll draw around <500 watts at full load. The fins aren't going to do anything.
 
Don't forget the antifreeze and a way to bleed air out of the top of the tank.
 
30+ gallons of water may seem like a lot of cooling potential but the water will eventually heat up to a high equilibrium temp if it can't dump it's heat to ground quickly enough. I realize the fins will have little effect considering the temperature delta my computer will produce in the water. I was considering when I expand the loop to include more systems. I'm thinking about a three to five system loop here. I was hoping that the fins would wick heat away to earth faster than the earth would by itself. My reasoning is the thermal conductivity of clay and soil is very poor at about 0.15 to 2.5 while steel is rated at 43 and aluminum for 205. I used this handy chart for my numbers. I'm using the fins to basically make the drum reservoir into a more conventional heat sink design on the outside surface of the drum at least.
 
You'd need fins inside the tank too. I think the cons outweight the pros here. 30g of water is going to take quite a long time to heat up even with half a dozen systems on it.
 
Jorona said:
You'd need fins inside the tank too. I think the cons outweight the pros here. 30g of water is going to take quite a long time to heat up even with half a dozen systems on it.
Click to expand...

This...

You might want to see if you can find more info on water heating calculations, but here are some numbers to throw around just for food for thought.

2000BTU is about 586W of dissipation. That is already 5 heavily OC'ed CPUs. (Probably more then that.) A quick check of a cheap natural gas water heater shows it uses 35,000 BTU/hr to heat up water. (40Gal tank) That's 17x the amount of heat you could produce with those pcs.

I know it can easily take 15 minutes to get a cold water tank to heat up to operating temps (140F or 60C) and that's in an insulated tank that's trying to keep all of the heat in. I'm not sure if this would be accurate to say or not, but if we're talking in BTU/hrs I would think that 15 minutes of 35K is 8,750BTUs. To equal that at only 2,000BTU output you would need about 4 1/2 hours of constant heat. I would guess that the amount of surface area on the tank could dissipate that much heat given that amount of time.

I figured out the surface area on a 17" x 30" tank (~30gallons) to be about 171 Square feet. So if you have a metal tank there is basically no insulation. for 2,000 BTUs the temperature delta would be about 12 degrees. So your tank should be able to effectively dissipate 2,000BTUs/hr at 12 degrees above the nominal temperature of the area. (Assuming 1sq ft is dissipating 1BTU/hr in air)

It is true that adding more metal is going to conduct heat better, but in the end you are still dumping all of that energy back into the air. I think you already have more than enough surface area on that tank to where you would be better served by just making sure there is enough airflow. I figured a high end CPU cooler to have around 3 to 4 square feet of surface area. The tank has about 40 times that much.
 
This is not geothermal cooling. It is more along the lines of storage tank cooling. Geothermal cooling is very different.
 
You could just put a remote quad radiator with sandwich fans in your attic or basement and ran tubing to it and had 0 noise?
 
Thank you bman212121 for your thoughtful comments and useful calculations. I never was one to crunch the numbers. Though I should reiterate. I am completely burying this steel drum in my loose soil crawl space. There will be no part of the drum exposed to air. So yes, this is geothermal cooling. The biggest reason for adding the fins in my mind is to try to lower the temperature delta as much as possible.
 
Are you planning on digging a hole into the ground or setting the tank then mounding dirt around it? I think you'll be much better off digging a hole I think. Will there be a way to measure your water level or a way to top it off every once in a while?
 
I'll be digging a hole then filling it level with the surrounding ground. I'll have a top off reservoir in my computer room where I'll air bleed it and top it off when necessary. I have most of the loop planned out but I'm holding back on details until I can make up a nice CAD schematic.
 
I just posted a schematic for my cooling loop. That was a fun exercise I must say. I haven't used AutoCad since I was in high school.
 
Dok Ju5tice, it's nice to see you again my old nemesis. I'll give you bonus points if you can point out the major flaw in my schematic before anyone else does.
 
Delexicus said:
Dok Ju5tice, it's nice to see you again my old nemesis. I'll give you bonus points if you can point out the major flaw in my schematic before anyone else does.
Click to expand...

The pump. If it's pulling from the barrel, it's pumping cold water, not hot as the diagram illustrates. If it's pumping into the barrel, then your loop flow rate is gonna suck. :D
 
I like this concept :)

Take pictures!

I would think a automotive in tank fuel pump might work well here. Cheaper ones hit about 40psi, and newer ones can be up to 120psi. Of course this depends on some variables.

You could even filter the intake water.
 
Ozymandias said:
The pump. If it's pulling from the barrel, it's pumping cold water, not hot as the diagram illustrates. If it's pumping into the barrel, then your loop flow rate is gonna suck. :D
Click to expand...

Thanks for the tip about the flow rate. I was actually referring to the fact that air would get trapped in the tank. This would happen if it is flowing from the pump to the tank then being drawn from the bottom of the tank. I found a gadget that would fix that for me though. It's called an auto air vent and it works as follows.

Autoairvent.jpg
 
Last edited:
This is about the awesomest waste of money I've ever seen in the name of [H]ardcore overclocking. Props to you, if you go through with it. Also, I think you could run several servers off of that loop.
 
This idea doesn't seem worth the money. Biggest problem that I think you will encounter will be servicing it. I think it is much easier to work with a much smaller reservoir for maintaining the liquid quality.

I think a much more smaller loop with a chiller will be more effective over all for achieving nice cool temps and for servicing. Chiller doesn't always have to be turned on to achieve decent temps for water cooling. Well designed radiator with quiet fans works well. Adding the chiller to the loop would be for achieving sub-ambient temps and you would have to safety system for condensation.
 
You should think of getting something to monitor the water level, so you don't have to crawl around in the dirt and shine a flashlight into a barrel to check it.
 
I won't need to monitor the water level at the drum since I can monitor the water level from the PVC fill pipe. The steel drum won't need to be manually air bled since I'm installing an auto air vent that purges any extra air in the tank. I have added the auto vent in version 2 of the schematic.
 
I'm subscribing to this thread.

Just make sure to do something for condensation that water should get pretty cool.
 
I have central forced air heating and cooling in my house so the relative humidity stays quite low inside. The low humidity decreases the dew point quite a bit so condensation should not be a problem. My house indoor temperature reader says that for the last two years the max temp and humidity were 75f and 60% respectively. Of course I'll monitor it just to be on the safe side. Here is a dew point chart for your consideration.

dew_chart.jpg
 
Last edited:
So I'm still keen on adding fins to this drum. I know they probably won't do much to help with cooling but I always wanted a project to kick start my brazing learning experience. Problem is, it seems most mainstream hardware stores don't carry much, if any, brazing gear. I'm going to have to travel a bit to find a decent welding supply store it seems.
 
Nettwerk said:
Why don't you coil up a bunch of copper tubing and bury it instead?
Click to expand...

That actually sounds like a better, less complicated idea that OP's original, because you can get a hell of a lot more thermal conductivity, and surface area with coiled copper pipe, than with a steel drum.
 
The main reason I'm using a steel drum instead of copper coil or pipe is because I don't have the tools or equipment to solder or braze while I'm under my house. That plus I'm already aware of other systems that use copper pipe and coil for geothermal cooling and I wanted to see just how much of a difference a different system would have. This type of setup will also have significantly fewer points of failure compared to tons of coils run in parallel buried under my house. I've also already purchased most of the materials for this system that would be incompatible with a copper coil/ tube system, so inertia plays a role too.
 
Last edited:
Thanks to some brainstorming with my brother-in-law I have a much simpler method of adding fins to this drum. I'll use aluminum flashing and make three collars to clamp between the reinforcing ribs on the drum. After that I'll braze them in place to make them permanently attached. I'll use additional flashing to make the fins on the drum in a similar fashion. By taking an eight inch wide section of flashing and brazing one side flat to the drum, then bending the flashing perpendicular to the drum I'll have fins that I can deploy after I fit it in my crawl space. Here are some brazing videos that helped convince me this was the method that would work best for my project. With Flux and without flux. As you can see the added flux will be great for making a nice long bond that will transfer heat better. I should have all the parts I need after one more trip to the local welding store. I plan to start brazing today.
 
Last edited:
Toohotforflashing.jpg

steeldrumbracket.jpg


So it looks like I really do need some special brazing rod to connect such thin material. I had already ordered some HTS-2000 to do this project. But since I was already at the welding store I couldn't help but buy some standard aluminum brazing rod too. As I thought, the standard stuff melts at way too high a temp for such thin material. On the bright side, I was able to finish my pinch brackets that will hold the aluminum flashing tight when I finally do receive my HTS-2000 kit. I don't have time tonight to give a proper explanation of what I plan to do. Will update in the morning.
 
Cool idea, but watch out for corrosion. You have added steel and aluminium to your all copper loop with the drum.
 
BTU_Natas said:
Cool idea, but watch out for corrosion. You have added steel and aluminum to your all copper loop with the drum.
Click to expand...

I will have to look up how steel and copper react to each other but the aluminum is not going to touch and water. It's only an external heat spreader. I'll make sure that there is a nice coat of paint between the aluminum jacket and the steel drum. The hts-2000 brazing rod hasn't arrived yet so I have nothing else I can report on.
 
You may want to consider how your going to get the water to mix, don't want stagnant pockets of water sitting inside the drum.
 
Joseph F said:
That actually sounds like a better, less complicated idea that OP's original, because you can get a hell of a lot more thermal conductivity, and surface area with coiled copper pipe, than with a steel drum.
Click to expand...

This. And you're not mixing metals. As silly as it seems, if your house is properly grounded, you'll get a verrrrrrrrry slow galvanic reaction between the steel drum and copper parts.

Or make a copper coil and immerse that in a steel drum full of mercury :D.

Don't actually do that.
 
This is such an extreme build. I can't wait to see how it turns out. I also wonder what the temps will be.
 
bman212121 said:
This...

You might want to see if you can find more info on water heating calculations, but here are some numbers to throw around just for food for thought.

2000BTU is about 586W of dissipation. That is already 5 heavily OC'ed CPUs. (Probably more then that.) A quick check of a cheap natural gas water heater shows it uses 35,000 BTU/hr to heat up water. (40Gal tank) That's 17x the amount of heat you could produce with those pcs.

I know it can easily take 15 minutes to get a cold water tank to heat up to operating temps (140F or 60C) and that's in an insulated tank that's trying to keep all of the heat in. I'm not sure if this would be accurate to say or not, but if we're talking in BTU/hrs I would think that 15 minutes of 35K is 8,750BTUs. To equal that at only 2,000BTU output you would need about 4 1/2 hours of constant heat. I would guess that the amount of surface area on the tank could dissipate that much heat given that amount of time.

I figured out the surface area on a 17" x 30" tank (~30gallons) to be about 171 Square feet. So if you have a metal tank there is basically no insulation. for 2,000 BTUs the temperature delta would be about 12 degrees. So your tank should be able to effectively dissipate 2,000BTUs/hr at 12 degrees above the nominal temperature of the area. (Assuming 1sq ft is dissipating 1BTU/hr in air)

It is true that adding more metal is going to conduct heat better, but in the end you are still dumping all of that energy back into the air. I think you already have more than enough surface area on that tank to where you would be better served by just making sure there is enough airflow. I figured a high end CPU cooler to have around 3 to 4 square feet of surface area. The tank has about 40 times that much.
Click to expand...

Sorry I had to comment. You had a decent idea but you fudged the math so horribly wrong and guessed at so many things.

1) Just ignore BTU's calculate in the units we all are used to: Watts (Joules/sec).

2) 568W is *not* 5 heavily OC'd computers. A single heavily OC'd computer can put out that much heat if not more depending upon the number of video cards in the loop. Without knowing what kind of systems he is using, I'll guess 500W for the gaming PC and 250W for other PC's to make the math simple for a total of 1500W for 5 PC's.

3) ACTUAL MATH:

Specific Heat: Q (Energy) = m(mass) * c(specific heat) * deltaT(change in temperature)

Water has a specific heat of 4.186 J/g C. Which means it takes 4.186 joules of energy to heat up 1 gram of water by 1 degree C. If you assume that no heat is leaving the water; 1500J/s in / ( 4.186J/g C * 30 gallons * 3780 grams/gallon ) = deltaT. Do the mathy stuff and it works out to 0.003157895 degrees C rise per second, or about 53 hours to raise the water temperature by 10 degrees C.

Now in the actual system you have heat going in and heat going out. Delexicus your initial assumption of the water reaching an equilibrium is exactly correct. I do not know how much heat the drum will dissipate, but essentially if you are putting 1500W into the water, the water temperature will rise until the drum is taking 1500W out of the water. As the water gets hotter the drum will be able to draw more heat out (higher temperature delta makes for larger heat transfer).

So while it will take a very long time for the water to heat up, if you leave your PC's running eventually the water will heat up to an equilibrium. How high above ambient that equilibrium is will be determined by the efficiency of the drum at conducting heat from the water to the ground.

I think you have quite an ambitious plan, and for most peoples computer usage patterns you should be just fine with just the drum. Good luck, it will be interesting to see your project progress.
 
You must log in or register to reply here.
Back
Top