Helium results!!! Window air con chill box build. - Page 12

Well I've replaced the chamber air with Helium and have been doing some tests.....the results aren't as good as I hoped and are not very easy to interpreter.

First off I noted that the cpu fan speed has gone up from 2400rpm to 3000rpm because the helium is less dense...I guess that increase also applies to the other fans but I'm not monitoring those.

It has taken 10C off the gpu temps across the board regardless of the test, furmark at clock speed 960mhz, furmark at stock 700mhz or heaven at 960mhz...all show a 10c reduction on both gpus.

The cpu results are quite baffling at max heat production with IBT at 5.4GHz the temps is just about 5c lower where as at lower heat production with IBT at 5.0GHz the results are identical to air....and if I drop the heat production again by just using the heaven benchmark then temps drop by 5c again.

The way I'm rationalizing the strange cpu temp behaviour is by considering the cooling limiting factors of the heat sink design and the heat gradient.

I'm thinking that at some point along the cooling fins the temperature will equal the chamber temp, the position of that point will vary depending on the amount of heat production, the more heat produced the further away from the block will be the balance point and visa versa for less heat production.

Obviously any part of the heatsink fins further away from the block than the balance point will not contribute to cooling as they are at the same temp as the chamber and therefore not transferring heat.

So lower heat loads will only use the part of the fins directly by the block,with my Zalman flower design heatsink the fins converge closely as they come to join the block.

Heat removal is governed by two processes, conductive through the metal and then removal by gas flow through the fins and can be limited by which ever is the weaker component.

I think at lower heat loads only the inner part of the heat sink is used and because the fins are closely packed in that region then removal of the heat by gas flow becomes the limiting factor.
That's why I see and improvement with the helium at lower heat loads as it is improving the weak link in the cooling process.

At intermediate heat loads I see no improvement....at this point the heat balance point has moved out into the main body of the fins, gas flow through the fins is good and this part of the heat transfer process is not limiting whether I use air or Helium, the heat transfer is being limited by the conductive part of the heat transfer process so I see the same temps.

At higher heat loads the balance point has actually moved beyond the extremities of the fins so the entire heatsink assembly is now above chamber temperature and the cooling is now again being limited by the gas cooling aspect of the heat transfer process and again I see an improvement using the helium.

I guess the reason I'm seeing a 10c improvement on the gpu's and only a 5c improvement on the cpu is because the gpu heatsinks are quite poor and so have more room for improvement.

All interesting stuff and not as simple as I first thought, improvement depends on the strengths and weaknesses of the individual cooling components.

I also did not notice any alteration in my minimum chamber temp without load or any change of the loaded chamber temp.....which I think is good as it actually shows that my temps are not being limited by transfer of heat from the air/gas to the evap and is solely limited by the capacity of the evap system to remove heat energy..... ....I'm going to have to leave it for some time now to see how long my chamber can hang onto that Helium.

Thanks for that info on the fans.

I did anticipate that helium escape would be a problem, I've left the system overnight and there doesn't appear to be any marked reduction in the size of the expansion sac...but its early days to draw any conclusions on that.

On a power efficiency note, well any active form of cooling is going to push you into the red in that respect, but if you are going the active cooling route what are the alternatives?....peltiers are notoriously inefficient.
Direct phase change cooling of the individual components uses smaller less power hungry units, but then without the cooled chamber you could have condensation problems....although I do acknowledge that using a combination of a direct phase change cooling unit and a sealed chamber containing dry bottled gas would solve the condensation problem...but then you would need someway of removing the the heat from the chamber produced by the non cooled parts...perhaps a dual radiator loop, one rad inside and one outside would help, but ultimately I think your other component temps would suffer.

My air con unit is rated at 12000btu and uses 2.6KW......it's probably a bit of overkill and you could get away with a 9000btu unit for a chill box design. Also although rated at 2.6KW it doesn't draw that all the time, it depends on how much heat the pc is pushing out, the compressor runs noticeably cooler under lower heat load than under higher heat load so I guess is consuming less power......although I don't have a socket power meter to be able to confirm that.

On a power/efficiency note I'm also aware than by insisting on following and air/gas cooled component approach I'm also leaving possible performance gains unused and on the table, redressing that would at least bring the balance back towards the side of efficiency......a water cooling loop would greatly improve my component temps and allow for further performance gains.

I'm hoping to redress that last issue shortly....not by water cooling but as I mentioned previously by using two gases in the chamber, one of which will liquify out so I can pump it to the components to cool them.

I'm happy that my chamber idle and load temps are not being restricted by heat transfer to the evap from the air/gas so will move away from using helium due to possible containment problems. I'm not concerned about the small improvement in gpu cooling it gave me as in the next stage air/gaseous cooling will only be a back up to the liquid gas/phase change cooling. All that is important is that the one "background" gas is inert as the second gas which will liquify out will be flammable.

So I'm thinking I'll get a cylinder of Nitrogen (oxygen free) as the background gas and will use Isobutane as the coolant liquid gas.

Isobutane has a boiling point of -11c so will be in liquid form at chamber operating temps. It's heat of vaporization is 218 joules/cubic centimeter of liquid so a flow rate 1cc per second would be sufficient to cool 218watts of heat production......so I would only need a very small flow rate 0.137 liters/min to cool 500 watts.

I'm thinking initially I will simply pump the liquid and have it pour over the air cooling heatsink so I have normal air cooling as a backup in case it doesn't work as planned....also much less expense for me if it turns out to be a non starter.

If it works I will either exchange the air heatsinks for water blocks or simply have nothing but a jet of liquid isobutane pumped directly onto the cpu/gpu heat spreader.
Edited by technogiant - 10/28/12 at 12:31am

The outer case of my chamber is acrylic not styrofoam.....the only foam in contact with the atmosphere is the edge of the gasket on the lid seal and that is a dense rubberized material.

I've measured gas loss, the air in my system is always at a very slight positive pressure, enough to inflate but not stretch the bag, under those conditions it looses in the vicinity of 1 liter of gas a day.
I understand with Helium that rate will be accelerated, but if I use bottled nitrogen as I plan the rate will be identical to air and quite within acceptable limits.....a 10 liter bottle of nitrogen at 300 bar will, I think, contain 3000 liters of Nitrogen at normal atmospheric pressure.......or over 8 years worth!!!.......actually as the gas mix will be in the region of 80% isobutane and 20% nitrogen the nitrogen will last 5X that or 40 years!!!

I'm more concerned about the loss of isobutane, not because of the cost of replacement, a 960 gram canister would last about 1.5 years and costs under £5.....but because of the safety aspect.
I'd be losing about 800cc per day.....I think that is very minimal and would easily be dispersed by natural ventilation.....but I'm also aware it is denser than air and will tend to pool at low points.....I'll probably put a battery operated gas detector under the floorboards in the vicinity of the unit to monitor that.

I'm not too concerned about the flammable nature of isobutane inside the chamber......it's only flammable in air at concentrations of 2 - 10% so at 80% it would not be flammable even if I just used it and air in the chamber.

But I'm concerned that as it is denser than air it would settle to the bottom of the chamber and the air to the top and at some point as you progress from the bottom to the top you would have a point where concentrations were flammable.

Also if I used air as a background gas then as the isobutane was liquifing as the chamber temp dropped again there may be a point where gaseous isobutane entered the flammable concentration region.

So it is essential that I use an inert background gas which bottled nitrogen (oxygen free) will be.

PS an external rad and direct water cooling of the components would be a good idea in cooler climates for winter use and even in the UK may give temps close to what I'm getting atm with my air cooling.....but you would still have to remove heat from the chamber produced by other non water cooled components and so would need two radiators one inside the chamber and one outside......but if all my plans come to naught and I end up using an internal water loop I could well run a loop to an additional outside rad to allow for that, could just use valves to isolate the external rad when I want to the air con unit.

But ultimately I'm hoping I will get mega low temps using my liquid gas idea...even better than using a water loop.....ideally I'd like to use a gas other than isobutane....a non flammable one, and if it had a lower boiling point, closer to the to the operating temp of the chamber.....perhaps -15c to -20c that would be better again....but I can't find one.....so if anyone has any thoughts on that I'd be grateful.
Edited by technogiant - 10/28/12 at 2:10am

Thanks Nol.....absolutely right......the chamber has lost the same volume of helium in one week that it normally looses in 5 weeks when using air.

Things have progressed a little.....I've been able to source a non-flammable gas to use as the liquified cooling gas and so there is no need to be concerned about an inert background gas anymore....I'll be using HFC-227ea....it's a refrigerant gas commonly used in fire suppression as a substitute for Halon.

It has a boiling point of -16c...it's a large molecule (heptafluoropentane) with a molecular weight of 170 g/mol so loses from the chamber should be minimal. I'm just waiting on a price for a 13.5kg cyclinder and hoping its not silly money.

I've had further thoughts on the design for this liquid gas cooling......I can't just use the chamber to contain it as the interior has a dense foam which will absorb hydrophobic materials like this gas/liquid is....so I'm going to have to contain it in a closed loop.

In the longer term I'll make a chamber with a hard non-absorbent inner material so the liquified gas can pool at the bottom. To provide enough volume for pumping I'd also have an amount of mineral oil or suitable carrier oil there which the liquified gas would mix with and pump that to the cooling blocks where the gas would phase change out of solution. Either that or just have the mobo submerged in the mix of liquid gas/mineral/carrier oil which would make for a really simple build.
Edited by technogiant - 11/5/12 at 11:01pm