PC in aquarium full of mineral oil

The alternative to direct air cooling

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Post by drownmypc » Mon Dec 29, 2008 9:35 am

rill2 wrote:Even if the material doesn't matter, which I'll believe in an instant, can't you get a bigger surface area / volume ratio by finning the metal plates? It sounds a lot easier to get a finned metal plate than a finned glass plate.
Oh, fins! Yes, I certainly agree with you there - glass fins would be harder to make.

In fact, I think maybe the material actually matters with fins. Even setting aside the manufacturability question, I would guess that metal fins would be more effective than glass fins at cooling. That's just a feeling, though. You'd have to either do an experiment, or run some kind of thermal FEA (finite element analysis) simulation.

rill2
Posts: 10
Joined: Thu Dec 25, 2008 12:16 pm
Location: Netherlands

Post by rill2 » Mon Dec 29, 2008 11:16 am

Also, if thickness matters, you might be able to get away with a plate that is thinner than the aquarium glass. A metal plate can be a lot thinner and have the same strength, right? It seems like changing the back side into finned metal could benefit heat-transfer by a combination of effects.

I don't really fancy changing front or sides into metal, since I'd want to be able to admire the insides. I guess one side glass, one side metal is acceptable if it's really necessary. Bottom plate is possible from aesthetics point of view, but then you might have a problem with circulation. Top is possible, but oil to air directly probably works better than oil-plate-air.

I'm not sure what effort it could take or what it would cost to make an aquarium with the back replaced by a finned metal plate - and I'm not sure exactly how to calculate approximately how much heat you would lose extra if you did that.

How do you calculate how much heat you lose through a single one of those plates? What does it depend on? The material, thickness, surface area, shape, oil temperature, air temperature?

On the topic of lava lamps, you mentioned that it would be necessary to keep a stable temperature inside the aquarium. Is this a hard thing to do for any reason? Does the equilibrium oil temperature fluctuate a lot depending on CPU-load, or does it average out to keep the temperature comparatively stable? Does the surrounding room temperature influence the oil temperature significantly? Are there any random effects that would give unsuitable (for a lava-lamp) fluctuations in temperature?

Even if stable temperature is not a problem, designing a blob-composition that doesn't hurt your electronics, and that has just the right properties to produce the lavalamp effect inside mineral oil of a certain temperature is still a challenge of course.

Edit:

Maybe even impossible.

It might be easiest to just wall off a section of the aquarium and to use a formula that works under whatever the temperature of the mineral oil is. Of course, that would be much less awesome than free-floating blobs, but perhaps still fun. If you wall it off you also prevent the evaporating problem you mentioned might be an issue.

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Post by drownmypc » Mon Dec 29, 2008 6:23 pm

[Edit: oops, didn't mean to post this one. How do I delete a post?]
Last edited by drownmypc on Mon Dec 29, 2008 6:27 pm, edited 1 time in total.

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Post by drownmypc » Mon Dec 29, 2008 6:23 pm

rill2 wrote: How do you calculate how much heat you lose through a single one of those plates? What does it depend on? The material, thickness, surface area, shape, oil temperature, air temperature?
I posted something about thermal transfer in this post: viewtopic.php?p=395902#395902

To summarize: H = U * A * T

where U = 6.42 W / m^2 K, A = area in sq meters, T = temp difference in centigrade, and H = heat loss in watts (joules/sec).

rill2
Posts: 10
Joined: Thu Dec 25, 2008 12:16 pm
Location: Netherlands

Post by rill2 » Thu Jan 01, 2009 3:49 am

In your link to engineeringtoolbox.com (another post) I browsed around and eventually I found the following:

The overall heat transfer coefficient for a wall or heat exchanger can be calculated as:
1 / U A = 1 / h1 A1 + dxw / k A + 1 / h2 A2 (1)
where
U = the overall heat transfer coefficient (W/m2K)
A = the contact area for each fluid side (m2)
k = the thermal conductivity of the material (W/mK)
h = the individual convection heat transfer coefficient for each fluid (W/m2K)
dxw = the wall thickness (m)

I guess now I have to learn about conductivity and convection heat transfer coefficients.

Aluminium has thermal conductivity of about 250, Glass has 1.08, so it seems like aluminium would do a better job at conductive heat transfer than glass. The difference between copper and aluminium (about 400 vs about 250) isn't quite as large as the difference between glass and aluminium, so that may explain why the material doesn't matter as much if the choice is between copper and aluminium. It may matter when it's between aluminium and glass.

I haven't figured out yet how to calculate the convective heat transfer coefficients for mineral oil and air.

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Post by drownmypc » Sun Jan 11, 2009 8:43 pm

rill2 wrote:The difference between copper and aluminium (about 400 vs about 250) isn't quite as large as the difference between glass and aluminium, so that may explain why the material doesn't matter as much if the choice is between copper and aluminium. It may matter when it's between aluminium and glass.
I think the wall material is relatively unimportant because the air itself poses a heat transfer bottleneck. Air has poor heat capacity and poor thermal conductivity, hence it is often used for insulation.

An illustrative point: your goose down jacket is mostly using air as an insulator. The goose down itself is not the major barrier to the heat energy transfer. The air itself is the barrier to heat energy transfer. The goose down just keeps the air from mixing within the coat, thus preventing convective heat transfer.

At a microscopic level, a relatively small amount of heat energy transferred from glass to air causes the layer of air adjacent to the glass to quickly reach the same (or almost the same) temperature as the glass. That is due to the low heat capacity of air. When their temps are the same (or similar), the temp gradient is zero (or small), and heat transfer ceases (or slows).

Once that innermost layer of air heats up, it wants to pass its heat energy to the next outer layer of air (which is cooler). But because air has lousy thermal conductivity, that transfer takes place slowly.

Consequently, heat energy moves through air rather slowly, and air becomes a bottleneck.

But, that's all in the absence of mixing -- turbulence or air currents could cause the warm air to mix with cool air and transfer heat more quickly. Hence, the use of goose down in a winter coat to minimize the mixing of the air trapped in the coat's material.

If you use a fan to blow that layer of warm air away, to be constantly replaced by cool (ambient, room temp) air, you can keep cool air air next to the warm glass, thus keeping up the temp gradient, which keeps the heat transfer going. My double-hull tankis passively cooled, but if I point a fan at it, the temp drops by 15F.

My point is this: the thermal conductivity of the glass (or metal or whatever the tank wall is made of) is not the only consideration. You must also consider how readily the oil inside the tank will give up its heat to the tank wall, and how readily the air outside the tank will accept heat energy from the tank wall. I believe that in my case, the air is the bottleneck, unless perhaps a fan is used.
I haven't figured out yet how to calculate the convective heat transfer coefficients for mineral oil and air.
I haven't either! Seems like convective heat transfer would be avery complex subject. I don't know any of the engineering or scientific theory for that.

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Post by drownmypc » Tue Jan 13, 2009 8:15 pm

rill2 wrote: The overall heat transfer coefficient for a wall or heat exchanger can be calculated as:
1 / U A = 1 / h1 A1 + dxw / k A + 1 / h2 A2 (1)
where
U = the overall heat transfer coefficient (W/m2K)
A = the contact area for each fluid side (m2)
k = the thermal conductivity of the material (W/mK)
h = the individual convection heat transfer coefficient for each fluid (W/m2K)
dxw = the wall thickness (m)
Rill2, I apologize for not looking more closely at the formula you posted here. This equation resembles electrical resistance in parallel and also electrical conductance in series.

Two resistors in parallel, with resistances R1 and R2, have an aggregate resistance R, and are related thus:

1/R = 1/R1 + 1/R2

Alternative view is to consider conductance (inverse of resistance). Resistors can be said to have an electrical conductance, and two resistors in series, with conductances K1 and K2, will have an aggregate conductance

1/K = 1/K1 + 1/K2

Thermal conductance and electrical conductance have mathematical parallels. The aggregate thermal conductance of thermal "resistors" in series has the same formula as for electrical resistors.

The formula you gave is

1 / U A = 1 / h1 A1 + dxw / k A + 1 / h2 A2

This resembles the aggregate conductance of three resistors in series,

1/K = 1/K1 + 1/K2 + 1/K3

where

K = U A = the overall conductance of the oil -> wall -> air pathway
K1 = h1 A1 = conductance of oil while losing heat
K2 = k A / dxw = conductance of tank wall for transferring heat
K3 = h2 A2 = conductance of air while gaining heat

Or, solving for K, we have

K = 1 / ( 1/K1 + 1/K2 + 1/K3 )

So, let's look at that. If K3 is small (the conductance of air), then K (the overall thermal conductance of the system) will be poor, even if the tank wall is very thin and very conductive (i.e. K2 is very big).

This is a kind of mathematical justification for saying that the wall/air interface could be a kind of thermal bottleneck, even if the wall itself is perfectly conductive (made of copper for instance).

But, like you said, we don't know the convective heat transfer coefficient (which is the effective conductance) for air. Or for oil. So we don't know whether those would be bottlenecks or not.

But my gut feeling is that K3 (= h2 A2) is rather large, and dominates the contribution by K2 (= k A / dxw).

jaganath
Posts: 5085
Joined: Tue Sep 20, 2005 6:55 am
Location: UK

Post by jaganath » Wed Jan 14, 2009 12:21 pm

drownmypc wrote:But, like you said, we don't know the convective heat transfer coefficient (which is the effective conductance) for air
it can be calculated if you have the Nusselt number:

http://www.cheresources.com/convection.shtml#natural

http://www.owlnet.rice.edu/~ceng402/ed1 ... index.html

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Post by drownmypc » Thu Jan 15, 2009 10:53 am

Wow. That blows me away.

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

add a case?

Post by cordis » Thu Jan 15, 2009 11:26 pm

Hey, thinking about this kind of mod a little bit, it seems like the hard part would be the motherboard mounting. Would it make sense to use something like the Antec Skeleton case, mount everything but the drives, and just drop it in? Admittedly, it would take a tank that's more square, but it could solve some customization headaches. Might look kind of cool. Having the big fan at the top would probably help with the circulation, although it probably depends on the type of radiator/loop thing you might have.

Come to think of it, if you're looking for something that dissipates more heat than glass, I'd think about something I've always wanted to make a case mod with, stone. I wonder how hard it would be to create a cube or something out of granite or marble tiles. You could let the stone heat up, and put a little waterfall on one face to create a water flow to carry the heat off. Now sure, it wouldn't be silent, but it would probably sound more pleasant than a case fan. Well, ok, I'm getting off track here, but there are definitely some possibilities. But turning a computer into a tabletop fountain would be pretty cool, don't you think?

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

cheapish metal case

Post by cordis » Sat Jan 17, 2009 5:19 pm

Thinking more about putting a system like this in a metal "case" to control the temperature better, one thing I thought of was to use a existing pot, like this:

http://www.kitchensupplydirect.com/114-6500.html

You could probably drop the motheboard in sideways, put the power supply in as well, and if you can find one that's deep enough, you could put the hard drives in the pot on something like a steaming rack held above the oil level. You could also try something like a more rectangular baking pan, if you could find one deep enough. Just a thought.

xan_user
*Lifetime Patron*
Posts: 2269
Joined: Sun May 21, 2006 9:09 am
Location: Northern California.

Re: cheapish metal case

Post by xan_user » Sat Jan 17, 2009 5:38 pm

cordis wrote:Thinking more about putting a system like this in a metal "case" to control the temperature better, one thing I thought of was to use a existing pot, like this:

http://www.kitchensupplydirect.com/114-6500.html

You could probably drop the motheboard in sideways, put the power supply in as well, and if you can find one that's deep enough, you could put the hard drives in the pot on something like a steaming rack held above the oil level. You could also try something like a more rectangular baking pan, if you could find one deep enough. Just a thought.
Image

rill2
Posts: 10
Joined: Thu Dec 25, 2008 12:16 pm
Location: Netherlands

Post by rill2 » Fri Jan 30, 2009 3:29 am

I got sucked in by life for a moment.

I can calculate the whole thing for oil using data I found online, the only problem is still convective heat transfer coefficient for air. "If you know the nusselt number," well, I don't.

Putting the whole thing in aluminium is good for cooling perhaps but it takes a lot of the wow factor out for me. I want to decorate my aquarium and people being able to look inside. I'm only willing to put back, bottom, and sides as metal.

I was hoping just the back side would be enough, so before I start thinking about materials and construction, I would like to calculate how much cooling I get by making the back wall out of, say, a finned plate of aluminium.

The problem is that none of us knows how to calculate it, although we've made progress.

Olle P
Posts: 711
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Post by Olle P » Mon Feb 02, 2009 1:35 pm

rill2 wrote:I would like to calculate how much cooling I get by making the back wall out of, say, a finned plate of aluminium.
Another option is to fold it in a zig-zag pattern to increase the surfaces facing both oil and air.

Start out with a thin plate (or possibly a couple, adding up to) 50cm high and 400cm wide, to get a 2.0 square metres surface. Then fold it sideways back and forth to get your desired case dimensions.

Another very "cool" way to solve the issue would be to find an old housing for a scrapped high power/voltage transformer and use that as a starter. They have cooling fins on the outside and transformer oil on the inside.
Open up one side and cover it with glass if desired, add connectors for all peripherals at the top, and you're ready to roll!

Cheers
Olle

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Post by drownmypc » Mon Mar 02, 2009 8:39 am

(edited to add pictures and more detailed description)
Olle P wrote:Another option is to fold it in a zig-zag pattern to increase the surfaces facing both oil and air. Start out with a thin plate (or possibly a couple, adding up to) 50cm high and 400cm wide, to get a 2.0 square metres surface. Then fold it sideways back and forth to get your desired case dimensions.
Olle, you read my mind!

I've been looking into exactly this strategy these past few months. In fact, I have a rough design for it on google sketchup (a very interesting CADish app). Here's a rendering of it:

Image

The idea (as you said) is to create an accordion out of the sheet metal, whose folds allow a lot of surface area between the hot oil and cool air. Furthermore, the manifolds (the "accordions") are surrounded by other surfaces to guide air and oil so that they move in parallel counter-flow (opposite directions), which optimizes the heat transfer.

The tank interior holds most of the oil. It's floor has holes cut for two ordinary low speed fans which draw oil from below, pushing oil upward into the tank interior (red arrows). The oil flows upward and over the lip of the interior tank's walls, down into the vanes of the manifold, and all the way down into the chamber below the inner tank, where it gets sucked upward by the fans again.

Meanwhile, there is an exterior framework which moves air past the oil, through the spaces between the vanes, in the opposite direction (parallel counter-flow). Below the lower oil chamber there is an air chamber. A fan pulls the air upward through the air chamber's floor and pushes it upward into the bottom of the vanes. Outer plates (not shown in illustration) keep the air between the vanes as it travels upward, until it exits the channel out the top of the vanes.

Image

The design has a vane (sort of a hollow fin) every 1/2". The vanes themselves are 1/4" thick (but hollow, for oil to flow within), and spaced 1/4" apart (air flows through these spaces between vanes). I might need to play with that -- maybe give the oil more space and the air less space, to account for their very different viscosities. A quarter inch isn't much space -- I expect the oil and air will be traveling fairly slowly.

I don't know squat about metal work, so I'm planning to take classes at our local shop.

I've also been looking around for aluminum and steel sheeting supply, but it is hard to find thin plates or sheets that are tall enough. Also, they are rather pricey. I've been looking at http://www.mcmaster.com to get a sense of availability and pricing.

We can get 10 ft lengths of 12" wide steel strips for about $40 from McMasters. With 4" vanes, we consume 17 inches of material per inch of manifold length. The 10 feet will fold to about 7 inches of manifold. So, we will need two of these for each 12" manifold. Total cost, about $160.

There is steel at 20 ft lengths and 12-3/8" wide. Spring steel. 0.01 or 0.012 inch thickness. About $100 each. Might be a bit too thin, though. Not sure.

Steel. At 10 foot lengths, 12-3/8" wide, we can get thicker (0.018") wide, about $100 each.

Thicker versions are available in shorter lengths:
.022" in 8' lengths:
.035" in 7' lengths:

Other notes on the design:

The tank interior is 16"x8"x10.5" high.
Volume is 5.8 gallons.

Front manifold (on long side) is 12" (30cm) high, 16" wide, 4" deep.

A vane is an outward-thrusting fold of the tank wall, 1/4 inch thick and extending 4 inches. The vanes are as high as the tank itself, which is 12 inches.

Volume of fluid held by each vane will be 1/4 inch * 4 inches * 12 inches = 12 cubic inches = 0.052 gallons.
There is one vane every 1/2 inch, so along a tank wall 16" wide, there are 32 vanes. All 32 vanes will hold 1.66 gallons.

The long side of the tank is 16" and the short side is 8", for a total of 48" of perimeter, allowing 96 vanes, holding about 5 gallons of mineral oil.

The tank interior requires 5.8 gallons, and the vanes will require 5 gallons, so 10.8 gallons total. That's a lot of oil. Good luck trying to move this thing.

The tank interior footprint is 16"x8", but the 4" vanes all around extends the footprint to 24"x16". That's pretty big.

Normally a 24"x16"x12" tank would have 4 sq ft of surface area on the sides (not counting top and bottom faces). With the 4" vanes at 1/2" spacing, this tank has 8.5x that area, which is 34 sq feet.

-stefan
Last edited by drownmypc on Mon Mar 02, 2009 5:39 pm, edited 2 times in total.

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Evaporative cooling!

Post by drownmypc » Mon Mar 02, 2009 8:44 am

Hmm, this is interesting.

I draped a wet cloth over my oil aquarium and pointed a room fan at it.

With the computer on standby (drawing 4 watts, effectively off) the tank oil temp got down to 65.6F / 18.6C.
That's below the room's ambient temp of 70F / 21.1C.

Evaporative cooling!

Nifty.

Olle P
Posts: 711
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Post by Olle P » Tue Mar 03, 2009 7:19 am

drownmypc wrote:... the manifolds are surrounded by other surfaces to guide air and oil so that they move in parallel counter-flow (opposite directions), which optimizes the heat transfer.
I'd say you don't need to guide the air, but your solution for oil flow management seems great!
drownmypc wrote:The design has a vane (sort of a hollow fin) every 1/2". The vanes themselves are 1/4" thick (but hollow, for oil to flow within), and spaced 1/4" apart (air flows through these spaces between vanes). ...
I don't know squat about metal work, so I'm planning to take classes at our local shop.
Your flat vanes have a couple of disadvantages:
1) They're probably too thin to get an even oil flow distribution over the surfaces.
2) They're pretty difficult to manufacture.

The solution to both problems is to have the vanes in a simple zig-zag pattern, with at least 1" base width. That's easier to manufacture and the oil flow should be better, reaching all the way out to the edges.
My suggestion is to do a nice blueprint of the design, and let some professional metal shop do the physical work.
drownmypc wrote:Spring steel. ...
Spring steel is definitely the wrong quality. Too hard to work with!
drownmypc wrote:Steel. .022" in 8' lengths: .035" in 7' lengths:
These are the thicknesses to go for. .022 should suffice, .035 is very good.

... But again: Do see what the professionals say! They have the know-how, the skills, the tools and the material needed.
Personally I think it's too difficult for me as a novice to get the construction sealed at the bottom and at every junction.

Cheers
Olle

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Re: Evaporative cooling!

Post by drownmypc » Sat Mar 07, 2009 5:37 pm

drownmypc wrote:Evaporative cooling!
A little followup on this evaporative cooling concept --

I had a 3D game paused for a while, came back and noticed that the tank's oil temp had risen to 55C (131F). Even with a couple small (and quiet-ish) fans pointed at the tank. That was a bit hot for my comfort.

So, I wrapped the tank in a wet cloth. The tank temp dropped to 40C (104F). It took a few hours, of course (after all, it is 5 gallons of oil).

drownmypc
Posts: 56
Joined: Mon Feb 11, 2008 2:14 pm
Location: chapel hill, nc

Post by drownmypc » Sat Mar 07, 2009 6:53 pm

Olle P wrote:I'd say you don't need to guide the air, but your solution for oil flow management seems great!
Maybe I'll be able to design the tank to have removable guides, so we'll be able to test that hypothesis. That would be fun. I'm guessing that forced convention will make a significant difference. I predict, um, 10C lower oil temp with forced air versus passive. That's a total SWAG, of course. It's not like I know what I'm doing, or anything. ;-)
Olle P wrote:Your flat vanes have a couple of disadvantages:
1) They're probably too thin to get an even oil flow distribution over the surfaces.
2) They're pretty difficult to manufacture.
Sounds like you know more about metal work than I do, so I can't really dispute your second point.

About your first point -- I originally considered a zig-zag pattern. The evolution of my thinking, from start to finish, went like this:

1 - A 12" high, 16" wide flat tank wall has 1.3 sq ft of surface area. Ok.

2 - What about adding fins? A 1" vertical fin spaced every 1/2" along the wall will triple the surface area of the air-tank interface to 4 sq ft! Woot!

3 - Oh, but wait, if the interior of the tank is just a flat wall, it will have just 1.3 sq ft for oil-tank interface. That may be the bottleneck for heat transfer from oil to wall to air.

4 - Fine, so I will put fins on the inside too.

5 - Ick. That's a lot of fins. A lot of welding. Or gluing? They might be breaking off and stuff.

6 - Hmm, fins on the inside and fins on the outside rely on the metal to conduct a lot of heat along it's thin cross section. And if the fin is "wide" - that is, sticks out a long way, like 4 inches instead of just 1 inch, then we need even MORE heat flow through that thin cross section of metal. What metals are heat conductive? Copper rocks. But too expensive. Aluminum is half as conductive as copper. Not so expensive. But hard to weld (so I'm told). Steel has about 1/5th the conductivity of copper. And cheap. And workable (so I hear). Hm. But is it conductive enough for super wide fins? This sounds like a manufacturing nightmare.

7 - Hmm, thicker metal means more heat conductance. So what if I make the fins super thick? Like, maybe 1/8" thick? Actually, what if I make them SUPER thick - like 1" thick and 4" wide? Or go with metal wedges as fins, 1/2" at base, 4" wide, packed along the side of the tank. That's certainly enough metal to conduct the heat along their length. Hmm, but expensive. And heavy. Ridiculously heavy.

8 - Oooh, what if I make the fins HOLLOW! Let the oil flow INSIDE the fins, to supply them with heat. Made how? Hollow out blocks of steel. No, of course not. Buy rectangular tubing and join them together? Hmm, maybe. Then the heat flows transversely THROUGH the hollow fin's metal wall, instead of ALONG the metal. In which case thin metal is GOOD, not BAD. Jeez, how will I avoid leaks after joining all these hollow things together?

9 - Oh, of course, let's just take a big sheet of metal and fold it like an accordion. No joints, no welds, no chances of leaks. The more extreme the folding, the more surface area we get. But two questions: what shape for the folds/vanes, and how closely spaced (how thin), and how wide (how much does a vane "stick out" away from the tank) ?

10 - First, what amount of surface area are we aiming for here? I want extreme surface area -- like about 10x what you'd get with an ordinary tank. Increasing surface area requires either using more closely spaced vanes (and therefore more of them) or using wider vanes (so each vane has more surface area). So, let's consider wide, closely spaced vanes, for lots of surface area.

11 - Now, rectangle or triangle? My feeling is that a triangular cross section with a narrow base will cause oil to flow poorly in the pointy end -- that would be a fair bit of surface area not being given heat. My feeling is that a rectangle, even one with the same "average" thickness as the triangle (that is, half the base width) would do better. That's just intuition though. I can't back that up with real science.

So, I'm trying rectangles, as shown in the design illustration I gave earlier. Unless I cannot manufacture it. I don't know how to do it yet.
Olle P wrote: My suggestion is to do a nice blueprint of the design, and let some professional metal shop do the physical work.
I'm doing myself. That's part of the purpose of the project - I do it myself. I WILL learn how to work with metal.
Olle P wrote: Spring steel is definitely the wrong quality. Too hard to work with!

These are the thicknesses to go for. .022 should suffice, .035 is very good.

... But again: Do see what the professionals say! They have the know-how, the skills, the tools and the material needed.
Personally I think it's too difficult for me as a novice to get the construction sealed at the bottom and at every junction.
Good to know, and good advice. Thank you!

theycallmebruce
Posts: 292
Joined: Sat Jul 14, 2007 10:11 am
Location: Perth, Western Australia

Post by theycallmebruce » Sun Mar 08, 2009 2:16 am

Keep us updated drownmypc! I'm looking forward to seeing your work.

Olle P
Posts: 711
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Post by Olle P » Sun Mar 08, 2009 4:32 am

drownmypc wrote:I'm guessing that forced convention will make a significant difference. I predict, um, 10C lower oil temp with forced air versus passive.
I agree with that, but natural convection should be totally sufficient (and silent). Plus easier to build.
drownmypc wrote:
Olle P wrote:Your flat vanes have a couple of disadvantages:
2) They're pretty difficult to manufacture.
Sounds like you know more about metal work than I do, so I can't really dispute your second point.
Try folding stiff cardboard into the two shapes, using a ruler pressed against a table to guide the fold.
drownmypc wrote:9 - Oh, of course, let's just take a big sheet of metal and fold it like an accordion. No joints, no welds, no chances of leaks.
You still have to attach it securely and leakage free to the bottom and front plates.
drownmypc wrote:11 - Now, rectangle or triangle? My feeling is that a triangular cross section with a narrow base will cause oil to flow poorly in the pointy end -- My feeling is that a rectangle, even one with the same "average" thickness as the triangle (that is, half the base width) would do better.
What will happen (no matter what shape) is that the oil in contact with the metal surface will sort of stick to it, leading to an increasing in flow the further away from the surface you get.
- Thin rectangular vanes will have a relatively large flow near the interior of the tank, where the pressure is "high", and very low flow at the outer edges.
- Somewhat thicker triangular vanes will have their high flow at the interior of the base, with "spill" further outwards and sideways.
My bet is that this design will actually provide a better flow at the outer edge and more even heat distribution across the surface.

Since you're determined to build it yourself your main design goals should be:
1. Sufficient cooling. (Any extra performance is considered a bonus.)
2. Simple build. (To get it finished at all!) Every separate plate, cut, bend and joint will increase the complexity and build time.

Some more advice:
- The type of oil to use is called "transformer oil" (normally used in high power transformers). It's available from all major oil companies.
- For joining this type of steel plates I would try to hot soldering with silver. Welding is overkill and too difficult. Then I would seal the joints with some suitable silicone sealant.

Cheers
Olle

eric_sc
Posts: 1
Joined: Fri Mar 27, 2009 10:53 pm
Location: USA

Post by eric_sc » Fri Mar 27, 2009 11:05 pm

While this looks like fun, do you think it's a good idea to immerse a computer in a large volume of flammable liquid? The MSDS gives it a Flammability rating of only 1, but then so has soybean oil. I've known of at least one kitchen fire caused by cooking oil. Haven't heard of kitchen fire caused by mineral oil.

Then again, I suppose I have a can of gasoline in the garage.

Olle P
Posts: 711
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Post by Olle P » Mon Apr 06, 2009 12:38 am

Fire shouldn't be a problem since it takes some serious temperatures to get it starting.

You don't want the computer sitting next to your fire place anyway, and probably don't place lit candles on top of it.

Transformer oil is used for cooling high power, (very) high voltage (10-500 kV) transformers all over the world, and accidental fires are extremely rare.

Cheers
Olle

~El~Jefe~
Friend of SPCR
Posts: 2887
Joined: Mon Feb 28, 2005 4:21 pm
Location: New York City zzzz
Contact:

Post by ~El~Jefe~ » Mon Apr 06, 2009 8:27 am

i dont think that mineral oil is the answer.

theres lots of other stuff to fill up with.

rill2
Posts: 10
Joined: Thu Dec 25, 2008 12:16 pm
Location: Netherlands

Post by rill2 » Thu Apr 23, 2009 4:39 pm

I see that you have actually designed something rather than waiting until one of us figures out how to calculate it.

I am very happy about this development. Now you can try it out and determine by experiment if it works or not !

If it works, I can make my own : )

Flatlander
Posts: 2
Joined: Sun May 03, 2009 8:16 am
Location: Canada

Post by Flatlander » Sun May 03, 2009 9:13 am

...greetings guys...my first post in Silentpreview so be kind huh?

I've been following this thread for a while now and getting "fully immersed" in this topic so to speak. Seems from reading through the posts heat (ISN'T it always) keeps rearing its ugly head. Have given it a lot of thought and have some ideas of my own that I'd like you all to kick around.

THOUGHTS:
---------------
1) Pumps ARE NOISY ....not necessarily......Diaphram pumps YES.... but what about centifugals ? I provide a link here as an example...
---------------
--oops can't post links yet...so do search on ebay for centifugal pumps & sort price low-->high they are $15 - $25
---------------
....and there are others too...
inlines (fuel transfer)
pony pumps,
bilge pumps etc all working off 12v

2) Using an automotive "transmission cooler" in a circulation loop as a heatsink. They're realatively cheap...small...readily available.

3) Using a Peltier(s) to cool the fluid outide the tank perhaps in combo with 1 & 2. Again these are readily available off ebay and can be passively cooled.

Thanks guys, interesting topic that I'm really keeping my eye on

cheers..

Olle P
Posts: 711
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Post by Olle P » Wed May 06, 2009 4:53 am

Flatlander wrote:...greetings guys...my first post in Silentpreview so be kind huh?
I'll try... ;)
Flatlander wrote:1) Pumps ARE NOISY ... but what about centifugals ?
Centrifugal pumps are what's generally used. Even those can be considered noisy if they're not perfectly balanced (which they're not).
Often they're less noisy than a high power fan though.
Flatlander wrote:2) Using an automotive "transmission cooler" in a circulation loop as a heatsink.
I'm not familiar with those. Probably too small to make a difference though.
Flatlander wrote:3) Using a Peltier(s) to cool the fluid ... and can be passively cooled.
Huge misconception. What peltiers do is to add lots of heat while creating a temperature difference between the two sides.
Assume we have a computer that use an average of 200W off the grid. Then the peltier needs to be able to transfer 200W of heat, which is more than most CPU-peltiers are capable of. To have it actually do that transfer we need to feed it 40-50A @ 12V, about 500W, from an external source. The result is that instead of cooling 200W off some 55C oil we need to cool 700W off a 100C peltier surface.
I wouldn't try doing that passively!
(If we add the peltier's PSU to the oil for cooling, it will add some additional 100W at 80% efficiency level, which in turn increase the demands on the peltier.)

Cheers
Olle

mark19891989
Posts: 187
Joined: Tue May 26, 2009 10:37 am
Location: UK

Post by mark19891989 » Tue Jun 09, 2009 10:31 am

This looks cool, would look better if you could "split" the tank in to 2 sections, front half with water, and fish in it, and the rear half with the pc hard ware in it, so it looks like the fish and the computer ate in the same place .

Olle P
Posts: 711
Joined: Tue Nov 04, 2008 6:03 am
Location: Sweden

Post by Olle P » Tue Jun 09, 2009 10:45 pm

mark19891989 wrote:... "split" the tank in to 2 sections, front half with water, and fish in it, and the rear half with the pc hard ware in it,...
Poor fish! One has to pick very carefully to find fish that like water that warm...

Cheers
Olle

mrttt
Posts: 1
Joined: Mon Apr 26, 2010 7:27 pm
Location: SF

Post by mrttt » Mon Apr 26, 2010 7:28 pm

Can someone help me ponder the idea of using a synthetic motor oil? Would this be an alternative to mineral oil with its health risks?

Would it possibly be better at dealing with heat?


Did anyone find another solution to counteract the mineral oil health risks? Is there any other kind of oil aside from veggie oil?

Post Reply