Compressed Gas Cooling

[Shinei]

I would have put this in Cases & Cooling, but the question is a bit more highly technical than C&C could handle, I'd think.
Anyway, the question is this: Would it be possible to take compressed gas, like CO2, and run it in a heatpipe type system where the heat from the cooled components provides the energy to move the gas from block to reservoir? So the system would thus look something like:
Gas starts in reservoir, under pressure, cooled. It leaves the reservoir via an open pressure valve that leads to the separate blocks for the GPU, video card RAM, and CPU, and cools the heatblocks. The heat from the blocks in turn generates pressure from the expanding gas, forcing the hot gas to flood away from the blocks and into an overpressure reservoir that acts as the radiator as well as to reduce system pressure. Gas builds in the overpressure reservoir as the system accumulates more and more used gas, forcing a pressure valve to open, sending the gas back into the primary reservoir for the system to cycle again.

Throughout the process, there is no outside forces acting on the blocks or gas, just the heat from the GPU, CPU, and video card RAM. Would a system like this cool efficiently, or would I be better off using a compressor? Note that I'm not going for VapoChill level cold (-33C), I just want a silent cooling system that can keep my system at around 25-30C, possibly cooler if I can get colder gas. Also, I had a question about radiating heat: Would the overpressure reservoir require a fan to cool off, since it's going to be taking in three different sources of hot gas, or will ambient air be enough to cool the tank down?
If you think of anything I may have missed, feel free to tell me and explain why this is a good/bad/wtf idea.

 

[LurchFrinky]

I don't really feel like giving a long, detailed answer, so no, you cannot do that.

The two biggest flaws are:

1. You have no driving force. A typical heat pipe has fluid inside and uses gravity to transport energy. You have nothing.

2. You are greatly over-exaggerating the pressure differentials in your system and the amount of momentum present in the gas.

 

[NeoPTLD]

Heat pipe works by liquid vaporizing and hot vapor carrying the heat to other point. Since the vapor actually moves, it's much faster than conductive thermal movement.

To make it work, you need to fill the system with enough working fluid so that you have liquid gas in the system.

Don't use carbon dioxide. I am here at the University library and just looked up the info on CO2 using CRC handbook.
The vapor pressure of CO2 at 25°C is 66.3atm(970 PSI) and as you can tell, it's a serious rupture hazard.

 

[uart]

Effectively what you are trying to achieve there is to have both a Heat Engine (converting heat into the work required to re-compress the gas) and a refrigeration cycle built into the one "thermal machine". It wont work because you'll never get the combined efficiency of the two stages together to equal or exceed 100%.

It's rather like trying to take an Electric Motor and couple it to a suitable Electric Generator such that, once started, the generator can supply all the power to drive the motor thus giving perpetual motion, that wont work either.

PS: Note that the "efficiency" of a refigerator cycle (if defined as the ratio of input mechanical power to pumped thermal power) can actually exceed 100% without violating any physical laws, however the combined efficiency of the two stages cannot exceed 100% (or in practice even equal 100%) .

 

[Shalmanese]

Actually, its not refridgeration since he doesn't need it to get down below ambient. He just wants faster heat transfer.

 

[uart]

Gas starts in reservoir, under pressure, cooled. It leaves the reservoir via an open pressure valve that leads to the separate blocks for the GPU, video card RAM, and CPU

This sounds to me like he's talking about cooling via gas expansion, which is the basic principle of a refrigerator.

 

[rjain]

If he could find something that boils at low temperature (35°C or so), then he can run it through a heat exchanger to release the heat. However, transferring the heat from the gas will probably be difficult. The way these things usually work is to let the evaporated gas simply escape into the atmosphere, AFAIK.

 

[LurchFrinky]

I just recently built a couple of heat pipes for my cpus (unfortuanately, one of them had a leak ) for the purpose of silent cooling. The one that still works keeps the cpu at around 55 C at full load which is fine by me. The pipes are pretty close to 4' long and stick out of the top of my case like a couple of antennae (well one antenna anyway). They are too hot to touch for any length of time. Since I was going for quiet, I didn't want to add any extra fans to cool them off, and they were already heavy enough that I didn't want to risk adding heatsinks. The point is that with passive cooling you will not be able to keep your temps even close to ambient without a really large surface area. You can try for some middle ground by bringing the heat transfer outside of your case and adding some large, low-speed fans, but you still can't get it below ambient.

You can just hook up a bottle of compressed nitrogen and let it flow through the blocks and into the atmosphere. You will get a constant hissing noise, but since the gas is expanding, it will be below ambient. You might have to worry about condensation in that case. You will also have to change the bottles occasionally.

 

[ZeroNine8]

Gas, compressed or not, is a horrible medium for heat transfer when compared to liquids or solids. If you want to compress a gas into liquid state, you will be talking some serious pressure and the equipment necessary to handle this isn't feasible for cpu cooling by any stretch of the imagination, not to mention being quite dangerous.

Unless you have some sort of mechanical pump or magic valves, you will not be able to get the gas from the low pressure state back to the high pressure state using the heat from your cpu alone. Without a mechanical means to repressurize the gas, the pressure throughout the system will equalize and you'll be left with nothing more than an overly complicated convection cooler, probably less efficient than a stock heat sink.

 

[FrankSchwab]

You guys have never seen a natural-gas powered refrigerator, have you?

Yes, I remember days when I had to go relight the pilot light on the refrigerator in our RV.

You can use a heat source (Natural Gas burner, or, presumably, P4 processor) as the driving force in a refrigeration cycle. Don't ask me how it works; it seems an awful lot like black magic to me.

He's not trying to build a perpetual-motion machine; he has a source of input energy (power given off by the proc/video card/etc). Seems kinda elegant to use that waste heat to drive the cycle needed to cool it. If the input energy isn't sufficient to run the cycle, he could always add more energy to the system (high-wattage resistor, anyone?)

This sounds like it would make an excellent second-year thermodynamics project.

/frank

 

[Shinei]

Thanks for the input, everyone, it's always enlightening reading these HT posts.

Frank: That's what I had figured, I had asked my physics prof about the idea before I even posted it up here, and he said the same thing you did. It should work, given enough pressure in the reservoir or enough heat output from the cooling blocks, even without a compressor; I had wanted to avoid using a compressor because that would require a temperature controller device and a self-built power supply to convert 120VAC into 12VDC, plus the noise and risk of overcooling would be less than desirable (I'm well aware of the need to coat just about every possible part of your CPU socket with anti-condensing compound to use a VapoChill setup).
Supposing I built the system, what kind of gas should I be considering? I had initially thought CO2 because it's so easy to get a hold of from a paintball facility, and because the gas can get rather cold when highly compressed, but according to some of the people here, it's too high-pressure for the purpose without blowing apart the system under heating.

 

[NeoPTLD]

Originally posted by: Shinei
Thanks for the input, everyone, it's always enlightening reading these HT posts.

Frank: That's what I had figured, I had asked my physics prof about the idea before I even posted it up here, and he said the same thing you did. It should work, given enough pressure in the reservoir or enough heat output from the cooling blocks, even without a compressor; I had wanted to avoid using a compressor because that would require a temperature controller device and a self-built power supply to convert 120VAC into 12VDC, plus the noise and risk of overcooling would be less than desirable (I'm well aware of the need to coat just about every possible part of your CPU socket with anti-condensing compound to use a VapoChill setup).
Supposing I built the system, what kind of gas should I be considering? I had initially thought CO2 because it's so easy to get a hold of from a paintball facility, and because the gas can get rather cold when highly compressed, but according to some of the people here, it's too high-pressure for the purpose without blowing apart the system under heating.

heat pipe

To work against gravity, you can't use plain pipe. Although, plain pipe should work if it will be a straight tube with the dissipation side higher than absorption side. I'm not sure where you'll get heat-pipe pipe with capillary structure built into it.

I'd use water or rubbing alcohol.

You need to evacuate the system before filling with fluid though and that's tough.

 

[Shinei]

The only reason I mentioned heatpipes was because the general concept of using heat from the source to cool the system is the same as the idea I had. I don't want to resort to liquids unless I absolutely have to, and then I'd be better off using a pump with water in it. Nifty link though, thanks for the information.

 

[Matthias99]

The problem is that you *are* trying to build a heatpipe, just going from a cold gas to a hot one rather than a liquid to a gas. And you're missing a few pieces.

I'm having difficulty visualizing how this will all work. Either I'm missing a step, or you're trying to violate the Second Law of Thermodynamics.

Gas starts in reservoir, under pressure, cooled.

Got it.

It leaves the reservoir via an open pressure valve that leads to the separate blocks for the GPU, video card RAM, and CPU, and cools the heatblocks. The heat from the blocks in turn generates pressure from the expanding gas, forcing the hot gas to flood away from the blocks and into an overpressure reservoir that acts as the radiator as well as to reduce system pressure.

If you keep the pressure in the reservoir higher than the pressure in the "overpressure reservoir", this will work. Otherwise the gas would just get forced right back into the original reservoir instead. But OK, you've got cold gas going from a high pressure tank, over your heat blocks, and into a reservoir. Eventually the pressure equalizes and we close the valve.

Gas builds in the overpressure reservoir as the system accumulates more and more used gas, forcing a pressure valve to open, sending the gas back into the primary reservoir for the system to cycle again.

So, we close the first valve and wait for the increasing pressure to open a valve leading from the "overpressure reservoir" back into the original one, right? This won't work. The problem is that you're going from "cold" pressurized gas to "hot" unpressurized gas, and somehow then magically making it cold again while repressurizing it! The heat from the processor will make it hot and pressurized -- and if you push that gas back into the "cold" reservoir, the "cold" reservoir's temperature will gradually increase until it stabilizes somewhere above ambient (depending on how good your heatsinks are). Odds are a non-active cooling system will be unable to remove enough heat, and the whole thing will just heat up until it melts down. Unless I'm missing something here.

 

[Shinei]

The overpressure reservoir is designed to act as the radiator, so the heat collected disperses through the walls of the reservoir, stabilizing pressure and temperature, until the first reservoir is nearly depleted, forcing the pressure valve in the overpressure tank to open and flood the main reservoir with cooled gas. The piping I planned on using was copper for the "return" pipes, so I could vent off a few extra joules of energy as the gas moves from block to radiator, facilitating a lower need for the reservoir to dump off 250w of heat on its own.

Suppose the tank has to supply cold gas to three blocks: GPU, CPU, and VC-RAM. The tank vents three separate pipes to the three blocks, which in turn seal upon pressure equalization, collect pressure from heat transfer, and then open the pressure valve going from the block to the radiator tank through copper piping. The heat loss from being away from the heat source, combined with a low specific heat, and the copper heat conductance should bring the temperature down a bit as it reaches the radiator, where it will continue to gather pressure as the gas collects in the radiator, giving off heat. The radiator, now nearly filled to capacity with gas, forces another pressure valve to open and sends the air down a copper tube back to the main reservoir, where the gas continues the cycle.
I'm definitely not an expert in thermodynamics, but it seems like it should work, given enough gas pressure and a gas with a low specific heat (meaning faster heat conductance and faster gas transfers start to finish). Am I still missing something, or does the system need to be tweaked more?

 

[EmMayEx]

Originally posted by: FrankSchwab
You guys have never seen a natural-gas powered refrigerator, have you?

Yes, I remember days when I had to go relight the pilot light on the refrigerator in our RV.

You can use a heat source (Natural Gas burner, or, presumably, P4 processor) as the driving force in a refrigeration cycle. Don't ask me how it works; it seems an awful lot like black magic to me.

He's not trying to build a perpetual-motion machine; he has a source of input energy (power given off by the proc/video card/etc). Seems kinda elegant to use that waste heat to drive the cycle needed to cool it. If the input energy isn't sufficient to run the cycle, he could always add more energy to the system (high-wattage resistor, anyone?)

This sounds like it would make an excellent second-year thermodynamics project.

/frank

Certainly you could use the heat of a P4 or Athlon XP to run a refrigeration cycle. The natural gas refrigerators work with an ammonia water mixture. When you heat the mixture the ammonia and water are separated. You recombine them in a pipe inside the refrigerator and they cool off. The process of dissolving requires energy so when ammonia dissolves in water (or when just about anything dissolves in water) is consumes heat energy and the temperature of the system drops. You design the system so that when the ammonia and water recombine they produce a temperature lower than what you want so when you run the mixture through the fridge or freezer it can absorb heat from the refrigerated space. When the mixture is nearly the temperature you want in the refrigerator or freezer you pipe it back to the burner and separate the components then let them recombine in the cooled space again. In a closed cycle it works quite well.

The problem with your scheme is that you are using the heat of the chip as an energy source. As you cool the chip off you lose your source of energy. No temperature difference = no heat energy to be extracted. It's somewhat analogous to trying to lift youself by pulling up on your own belt. You can give yourself a wedgie but you won't levitate.

If you want to get below ambient you have to supply energy in some form. This energy may not be obvious. If you allow water to evaporate off the chip surface you can get to below ambient temperature (assuming the relative humidity is below 100%) without any electric motors or compressors or heaters. In essence you are just taking advantage of the fact that water absorbs heat as it evaporates and the fact that at less than 100% humidity water has a tendacy to evaporate on it's own. The sun and wind have dried the air for you so you can use that residual energy in the dry air to cool something.

If you are happy with close to ambient then some sort of liquid cooling (a water block) should work well. The best scheme I've heard about is using silicon rubber and PVC parts to build a water block around the chip die itself. The die is coated with silicon nitride for protection but you'll want to seal off any conductive features near the die (like the L1 bridges on Athlons). If you can impinge a jet of water directly on the die then you will nearly eliminate the heat resistance between the die and the water because the water will carry away the heat before it has a chance to heat up very much. You'll need a good sized radiator to cool of the water down to near ambient (quite large if you don't use any sort of fan) or you could build a small spray tower and let some of the warm water evaporate when it contacts ambient air (this is how power plants and chemical plants often cool things) but then you need to add makeup water so your system doesn't run dry and deal with all the scale that builds up when you evaporate large amounts of tap water.

If you used something like rubbing alcohol or ether you could get a phase change (evaporation) inside your "water block" but you'd need active cooling or compression to reliquify the cooling fluid (this is essentially what a refrigerator does but they use CFC's or HCFC's or other gases instead of highly flammable liquids. Ether as a refrigerant is a recipe for an explosion.

Max L.

 

[uart]

He's not trying to build a perpetual-motion machine; he has a source of input energy (power given off by the proc/video card/etc). Seems kinda elegant to use that waste heat to drive the cycle needed to cool it.

Yes I admit that the situation is not identical to the pertpetual motion (motor generator) example and that there is a source of power. The example was "alike" in the sence that one system violated the laws of conservation of energy and the other violates the laws of thermodynamics.

I know the situations are different and that's why I made the point about refrigeration being able to be more than 100% effecient without violating any physical laws. Taken in it's entiretly however, the proposed sytem does violate the laws of thermodymanics and therefore cannot work. Or at the very least it cannot provide better than passive cooling would do using the same sized heat exchanger at the same ambient temperature .

 

[LurchFrinky]

Well, your still missing something, and you can tweak it all you want, it still won't work. While I don't know exactly how a natural gas refrigerator works, I can come up with some guesses and some comparisons to what you are trying to do.

A refrigerator has a compressor which performs work. Nowadays, that work is provided by electical energy, but you could easily get it from an internal combustion engine or a turbine or something else. What you are trying to do is use the heat from your processors as a source of energy to cool themselves off. This would be analagous to placing your refrigerator's compressor inside of the cold part of the refrigerator and hope that you can keep the compressor cold. Not going to happen.

If you want to use the heat from your processors as an energy source, then you are going to need a temperature differential. Presumably, you will use ambient as the cool side and the processor as the hot side. If you plan on using this energy to pump the gas in your system, then the flow of the gas would decrease as your processor got cooler. Essentially, the efficiency of your cooling system is a function of how hot your processors are (similar to heatsinks, what a coincidence!), which means it is impossible to approach ambient temperatures without constructing a perpetual motion machine. All of this if you actually have the means of converting the temperature differential into work. You could do this by placing peltiers in your system and wiring the leads to a really small gas pump.

Also, as you have it described, you have valves which open and close. Either your system is going to have some sort of pulsatile flow, or you are going to be using magical valves which are both open and closed at the same time. Since I don't believe in magic, I will assume the former. If the period of your pulsatile flow is in the neighborhood of a few seconds long, you can kiss your computer goodbye the first time the flow stops. If you expect to keep the flow stoppage shorter, then you will need to either have a very rapid heat transfer at all important points, or you will have to have valves which are extremely sensitive to pressure variations. More than likely, you will have to have both just to get any semblance of flow in your system. It is possible (I suppose, I haven't checked) to find suitable valves in a process and control catalog somewhere, but I highly doubt you could get even close on your heat transfer without artificially altering your temperature differential.

If you notice, nobody here thinks it is a good idea, and several of us have pointed out that it is impossible. Trust us, it isn't going to work.

 

[Shinei]

Was worth a try, at least... And this is why I decided to ask around before I started building it; $200 down the drain, plus the cost of the device itself, is not my idea of a good experiment.
Supposing I did include a compressor or some sort of forced airflow device, would this type of device* work? I hadn't wanted to use a compressor, but failing a passive conductance, I can always fall back to a single blower device instead of a ton of fans...

*http://www.galleria-e.com/cgi-bin/Colemans.storefront/3fd562ae019fc592273fd81e6c0206b4/Product/View/196701

Edit: Apparently the link didn't come out like I expected it to, the link to the device is above this line.

 

[Matthias99]

With a blower in there, you essentially have a water-cooling system filled with gas instead of water. Most gases will have a lower (much lower) specific heat than water, but they may absorb heat energy more readily (higher thermal conductivity, maybe? It's been a while since I did any thermodynamics). I can't say whether or not it would be more effective than standard water cooling, but it at least has a shot at working.

 

[sao123]

A gas refrigerator uses ammonia as the coolant, and it uses water, ammonia and hydrogen gas to create a continuous cycle for the ammonia. The refrigerator has five main parts:

Generator - generates ammonia gas
Separator - separates ammonia gas from water
Condenser - where hot ammonia gas is cooled and condensed to create liquid ammonia
Evaporator - where liquid ammonia evaporates to create cold temperatures inside the refrigerator
Absorber - absorbs the ammonia gas in water

The cycle works like this:
1. Heat is applied to the generator. The heat comes from burning something like gas, propane or kerosene.
In the generator is a solution of ammonia and water. The heat raises the temperature of the solution to the boiling point of the ammonia.
2. The boiling solution flows to the separator. In the separator, the water separates from the ammonia gas.
3. The ammonia gas flows upward to the condenser. The condenser is composed of metal coils and fins that allow the ammonia gas to dissipate its heat and condense into a liquid.
4. The liquid ammonia makes its way to the evaporator, where it mixes with hydrogen gas and evaporates, producing cold temperatures inside the refrigerator.
5. The ammonia and hydrogen gases flow to the absorber. Here, the water that has collected in the separator is mixed with the ammonia and hydrogen gases.
6. The ammonia forms a solution with the water and releases the hydrogen gas, which flows back to the evaporator. 7. The ammonia-and-water solution flows toward the generator to repeat the cycle.


While the idea presented may not work exactly as described, with a different primary heat source, such as heat disipated from the power supply, then using the refridgerant to cool all your cpus, I think it has promise.

 

[Shinei]

Matthias99: From what I can recall of basic thermodynamics from my chem classes, low specific heat elements will heat up faster, as well as discharge their heat faster; they may have a lower thermal conductance (as the case is with aluminum) than other elements, but in this case the heat source is static, but the transfer medium is dynamic. I'd imagine that with enough surface area (read: gas-immersed heatsink) and a high enough airflow, it should produce a fairly effective cooling solution without hooking up a ton of fans. I may be (and often am) wrong, though.

 

[NeoPTLD]

Originally posted by: sao123
A gas refrigerator uses ammonia as the coolant, and it uses water, ammonia and hydrogen gas to create a continuous cycle for the ammonia. The refrigerator has five main parts:

Generator - generates ammonia gas
Separator - separates ammonia gas from water
Condenser - where hot ammonia gas is cooled and condensed to create liquid ammonia
Evaporator - where liquid ammonia evaporates to create cold temperatures inside the refrigerator
Absorber - absorbs the ammonia gas in water

The cycle works like this:
1. Heat is applied to the generator. The heat comes from burning something like gas, propane or kerosene.
In the generator is a solution of ammonia and water. The heat raises the temperature of the solution to the boiling point of the ammonia.
2. The boiling solution flows to the separator. In the separator, the water separates from the ammonia gas.
3. The ammonia gas flows upward to the condenser. The condenser is composed of metal coils and fins that allow the ammonia gas to dissipate its heat and condense into a liquid.
4. The liquid ammonia makes its way to the evaporator, where it mixes with hydrogen gas and evaporates, producing cold temperatures inside the refrigerator.
5. The ammonia and hydrogen gases flow to the absorber. Here, the water that has collected in the separator is mixed with the ammonia and hydrogen gases.
6. The ammonia forms a solution with the water and releases the hydrogen gas, which flows back to the evaporator. 7. The ammonia-and-water solution flows toward the generator to repeat the cycle.


While the idea presented may not work exactly as described, with a different primary heat source, such as heat disipated from the power supply, then using the refridgerant to cool all your cpus, I think it has promise.

They've a diagram
here