SFF PHASE CHANGE

[AleksandarK]

I was wondering if there are any sff phase change coolers. It would be COOL, pun intended, if there is one so it could be implemented in a case.

 

[craigbru]

There are very few phase change coolers period, let alone sized for SFF. Most units I've seen are quite large. You could still build a relatively small system if you used ITX hardware for example, but I doubt you'd get much smaller than a standard ATX sized case. That would be small for phase change for sure, but certainly not what we'd consider SFF. Now, that said, I've long been a proponent of highly dense systems regardless of actual size designation, and I think it would be pretty awesome.

 

[jeshikat]

I'd like to see someone build a system inside the Little Devil PC-V2 box:

Image source

 

[craigbru]

I'd like to see someone build a system inside the Little Devil PC-V2 box:

Image source

Oh hell, there is a little more room in there than I thought...

 

[EdZ]

Technically, anything using heat-pipes or a vapour chamber is a phase change cooler!

Joking aside, the sticking point for a compact sub-ambient phase change cooler would be the compressor. I'm struggling to think of any situations where somebody would need to build a compact low-power refrigerant compressor, except maybe for very early thermal-guidance missiles (prior to the switch to solid-state cooling with Peltiers, or open-circuit expansion cooling with a gas bottle), so this would need to be a completely custom self-developed part.

You would be able to make an over-ambient phase-change system with a regular liquid-phase pump and a low-boiling-point Fluorocarbon working fluid like Fluorinert or Novec. The tricky part there is that most compact pumps (and pumps in general) no not like two-phase flow (fluids with gasses mixed into them). You'd need to set up your system so the pump is drawing from a 'sump' of settled condensed fluid. This imposes a volume penalty (need a large enough sump that it cannot 'run dry' at full-load operation, need expansion volume to prevent the loop overpressurising as vapour is generated), orientation penalty (can't turn it sideways or the sump will not fill up), and because the liquid condenses in the radiator then you end up with an inefficient - so large - radiator; the fluid vapour will probably only be 10°-20°C above ambient, and gas-to-gas heat exchangers need a larger surface area than liquid-gas exchangers like normal WC radiators. You are also limited in that if your ambient temperature ever reaches above the fluid's boiling point, the system no longer works; condensation never occurs so the entire working fluid turns into a gas and the system vents itself and overheats. On the upside, because these fluids are non-conductive you can omit the 'waterblock' entirely and just flood the whole board directly.

 

[jeshikat]

I'm struggling to think of any situations where somebody would need to build a compact low-power refrigerant compressor

There's the Stirling cryocoolers used in higher-end thermal imagers.

 

[|||]

I'm struggling to think of any situations where somebody would need to build a compact low-power refrigerant compressor, except maybe for very early thermal-guidance missiles (prior to the switch to solid-state cooling with Peltiers, or open-circuit expansion cooling with a gas bottle), so this would need to be a completely custom self-developed part.

There's the Stirling cryocoolers used in higher-end thermal imagers.

@Aibohphobia beat me to it: Long-wave Infrared cameras; but as mentioned, there are alternative methods to cool these components.

You would be able to make an over-ambient phase-change system with a regular liquid-phase pump and a low-boiling-point Fluorocarbon working fluid like Fluorinert or Novec. The tricky part there is that most compact pumps (and pumps in general) no not like two-phase flow (fluids with gasses mixed into them). You'd need to set up your system so the pump is drawing from a 'sump' of settled condensed fluid. This imposes a volume penalty (need a large enough sump that it cannot 'run dry' at full-load operation, need expansion volume to prevent the loop overpressurising as vapour is generated), orientation penalty (can't turn it sideways or the sump will not fill up), and because the liquid condenses in the radiator then you end up with an inefficient - so large - radiator; the fluid vapour will probably only be 10°-20°C above ambient, and gas-to-gas heat exchangers need a larger surface area than liquid-gas exchangers like normal WC radiators. You are also limited in that if your ambient temperature ever reaches above the fluid's boiling point, the system no longer works; condensation never occurs so the entire working fluid turns into a gas and the system vents itself and overheats. On the upside, because these fluids are non-conductive you can omit the 'waterblock' entirely and just flood the whole board directly.

You could just let the working fluid pump itself, like what Calyos does. That gets rid of the pump. There would still be the limitation of being thermodynamically limited to above ambient temperature without putting work into the thermodymic cycle.

 

[EdZ]

There's the Stirling cryocoolers used in higher-end thermal imagers.

I though those would all have moved over to solid-state cooling now. Though I guess older ones will still be in use (because who throws away a perfectly good LWIR imager) and somebody must make spare parts of them. Those compressors would probably cost quite a bundle though, and IIRC do not have the power to sink a lot of energy: they pull the FPA down to temp slowly and then just have to make up the insulation loss.

You could just let the working fluid pump itself, like what Calyos does. That gets rid of the pump. There would still be the limitation of being thermodynamically limited to above ambient temperature without putting work into the thermodymic cycle.

Calyos' cooler is a heatpipe in it's more traditional separated-flow form: they still use low-pressure water as the working fluid, but instead of the fluid flowing through a wick on the outside of the pipe and the vapour flowing back through the centre, the liquid and vapour phases get their own separate pipes to flow through.

 

[Chrizz]

I'd like to see someone build a system inside the Little Devil PC-V2 box:

Image source

The dimensions are apparently 540 x 220x 275mm (with legs) so you could fit an itx motherboard upside down on the ceiling. It's also long enough to fit an itx GPU next to it which you could connect with a 90 degrees riser. At $1000, I doubt anyone will do this though.

 

[AleksandarK]

You guys are gineuses.
Stupid question. Is it possible to trim the lenght of the pipe it self, or do you need to ask the manufactures?

 

[Biowarejak]

Technically, anything using heat-pipes or a vapour chamber is a phase change cooler!

Joking aside, the sticking point for a compact sub-ambient phase change cooler would be the compressor. I'm struggling to think of any situations where somebody would need to build a compact low-power refrigerant compressor, except maybe for very early thermal-guidance missiles (prior to the switch to solid-state cooling with Peltiers, or open-circuit expansion cooling with a gas bottle), so this would need to be a completely custom self-developed part.

You would be able to make an over-ambient phase-change system with a regular liquid-phase pump and a low-boiling-point Fluorocarbon working fluid like Fluorinert or Novec. The tricky part there is that most compact pumps (and pumps in general) no not like two-phase flow (fluids with gasses mixed into them). You'd need to set up your system so the pump is drawing from a 'sump' of settled condensed fluid. This imposes a volume penalty (need a large enough sump that it cannot 'run dry' at full-load operation, need expansion volume to prevent the loop overpressurising as vapour is generated), orientation penalty (can't turn it sideways or the sump will not fill up), and because the liquid condenses in the radiator then you end up with an inefficient - so large - radiator; the fluid vapour will probably only be 10°-20°C above ambient, and gas-to-gas heat exchangers need a larger surface area than liquid-gas exchangers like normal WC radiators. You are also limited in that if your ambient temperature ever reaches above the fluid's boiling point, the system no longer works; condensation never occurs so the entire working fluid turns into a gas and the system vents itself and overheats. On the upside, because these fluids are non-conductive you can omit the 'waterblock' entirely and just flood the whole board directly.

Could you not run it without a pump, but instead rely on the pressure of the expanded gas? Not sure if that would incur more or less of a volume penalty.

 

[|||]

I though those would all have moved over to solid-state cooling now. Though I guess older ones will still be in use (because who throws away a perfectly good LWIR imager) and somebody must make spare parts of them. Those compressors would probably cost quite a bundle though, and IIRC do not have the power to sink a lot of energy: they pull the FPA down to temp slowly and then just have to make up the insulation loss.

I think some of the consumer grade LWIR sensors do not have active cooling. I haven't looked into it in too much detail, so not quite sure how they work, but they may be using lenses that have extremely low radiative properties.

Could you not run it without a pump, but instead rely on the pressure of the expanded gas? Not sure if that would incur more or less of a volume penalty.

Not so much the pressure, but you could rely on buoyancy (density) to facilitate flow. Gigabyte had a non-sealed server submerged in Novec at CES. Novec is heavier than air, so it collected and condensed above the pool of Novec to cycle around.

 

[EdZ]

I think some of the consumer grade LWIR sensors do not have active cooling. I haven't looked into it in too much detail, so not quite sure how they work, but they may be using lenses that have extremely low radiative properties.

IIRC the real basic stuff like FLIR's Lepton periodically slide a shutter over the sensor assembly (either an automatic shutter, or prompting the user to slide a manual shutter) and measure the 'dark' output from the sensor, and just minus that out of the captured image. Cheap, but the downside is you lose a big chunk of the potential sensor dynamic range depending on how warm it is, and the reading can drift depending on temperature change since last calibration.

 

[Biowarejak]

I think some of the consumer grade LWIR sensors do not have active cooling. I haven't looked into it in too much detail, so not quite sure how they work, but they may be using lenses that have extremely low radiative properties.



Not so much the pressure, but you could rely on buoyancy (density) to facilitate flow. Gigabyte had a non-sealed server submerged in Novec at CES. Novec is heavier than air, so it collected and condensed above the pool of Novec to cycle around.

That's a very interesting implementation.