What would it take to build a cryogenic cooler?

[Shingoshi]

I did a search of this forum first before posting this. Ironically, there was nothing even given for the term cryogenic anywhere on this forum. So it seems like this may be new territory.

EDIT: It would have helped if I had spelled cryogenic correctly!

After seeing all the discussions of extreme phase-change cooling, I wondered if anyone here really ever considered building a system capable of cryogenic temperatures. We've all seen the accounts of builders on the web using LN2 or even Helium for short runs to set records for overclocking. But what about the possibility of actually building a cryogenic system that could run 24/7?

I'm certain such a system would require multiple stages, and probably in combination with autocascading techniques. And the thing is, I think some of the advanced builders here actually have the skill to do this. So what would it take to build a system to stop the fires of hell, so to speak. If anyone has looked into this, let us know what you've found out.

In the long run, I guess this thread will be primarily theoretical. Although it would be nice if an actual guide could be established from this.

Shingoshi

EDIT: Because of the repeated questions I've had about vortex tubes, please don't ask me any questions about "vortex tubes" without FIRST having looked up "vortex tube" on the internet. There's way TOO MUCH INFORMATION on the subject to have to keep covering the same issue.

EDIT: Throughout the course of asking this question, I have pursued many directions of attacking a solution to this. In some ways, my posts may seem rambling. I acknowledge that. But I think of greater importance are the various studies I've found be researchers probing similar goals. As a result I've made frequent use of patent searches.

The greatest objection that has been raised is the amount of energy that would have to be consumed to provide such a level of deep cooling. It is here that I've sought understanding of how various devices and technologies that I had become aware of have been shared here. Many of these investigations are beyond the scope of what is typically pursued by users of this site. But then again, if that weren't the case, many of them would already have cryogenically-cooled computers. So here are some of the links I've found related to this topic:

I express my gratitude for FreePatentsOnline - Free Patent Searching, where membership is required to access.
CRYOGENIC ABSORPTION CYCLES
Two-statge sorption type cryogenic refrigerator including heat regeneration system
Super critical helium refrigeration process and apparatus
Oxygen chemisorption cryogenic refrigerator
Cryogenic refrigerator

I am also very grateful for the work of the many who have contributed to Wikipedia.org. These are some of the links which serve as background material for the discussion here:
Regenerative heat exchanger

I will continue to update this list.

[trance565]

millions of dollars to start.

[yngndrw]

A large contract with BOC or another LN2 supplier. Probably cheaper than making your own.

[-EC-UnRockStar]

http://www.extremecooling.net/index/....php?47785.380

not finished and without ARCs

[[XC] gomeler]

Cryogenics technically starts at -150 Celsius. Biggest problem is most chips, the reason we build these things, won't run at -150 Celsius and below. Next up is cost. A two stage cascade runs around $1k in parts. Each additional stage runs ~$400 after that. Problem is after r1150 you jump to r14(expensive) and then methane and argon. Provide a builder with $6k to $8k and after blowing $2k on random problems I'm sure you'd see LN2 temps. Unrockstar's cascade is massive and I imagine the price is somewhere close to $6k.

It all boils down to money in the end. I know I'd like to hit -200C as an experiment but can't afford to.

[Postal Dude]

Also Shingoshi, most cryogenic coolers are designed to bring something down to temperature and then just keep it there, heavily insulated. For what we use, it becomes alot harder as were pumping a hell of a lot of energy into the system. A very auto cascade could run extremely low temperatures if it wasnt expected to cope with any sort of load. There is a thread somewhere on here (good luck finding it) either wrdzal or chilly1 showed what an auto-c could accomplish without any load holding capabilities, i think they got down to -140ish with one compressor.

[Shingoshi]

The reason why I posed this question is the result my buying four Rechi compressors on Ebay. http://cgi.ebay.com/ws/eBayISAPI.dll...m=160287043717 Granted, I only paid $20/each for them. But I figured if I'm going through all of this trouble, how much of a reward could I expect to receive from my efforts.

In the process of trying to find more data on these Rechi compressors, I came across Spudfiles.com where someone had posted a question about how to wire the capacitors. Come to find out, there's a guy there who works with cryogenic systems in his employment. He told me to pm him so that I could find out how to work with cryogenic systems. Of course, I emailed him right away. So I'm now waiting to see what comes of this. It would be nice to have people who actually work with cryogenic systems professionally.

Anyway, his initial suggestion was that I tie three of the compressors in a parallel circuit as my first stage, and then run that line into the fourth compressor. Doing that would up my capacity while also increasing my pressure. So I'll be looking into this just to see what actually can be done with multiple compressors.

Shingoshi
And for those of you who complained about me not posting pictures, I just got a digital camera today. But I still need to get the mini-USB cord for it.

[-EC-UnRockStar]

Unrockstar's cascade is massive and I imagine the price is somewhere close to $6k.

10k

[MeltedDuron]

10k

a duck!!! Let alone when you switch it on!!

[wdrzal]

10k

I bet 10K doesn't include Labor either.

[Duh]

10k

you should have your rank updated I guess.. from registered member to extreme crazy phase change of something like that

[[XC] gomeler]

I was factoring my costs scouring eBay. Impressive nonetheless

I bet 10K doesn't include Labor either.

Just parts. Can't charge true labor costs on these things or nobody could afford them.

[Shingoshi]

As a generalization, for all phase-change systems:
1.) What would happen if you had something like an insulated scuba tank (a dewar) in which you could store very cold liquefied refrigerant?
2.) How much insulation would you need?
3.) How long could you keep the refrigerant cold?
4.) How much might this reduce the load on the system, by having a reservoir?
5.) Would a much smaller high-pressure bottle like those used for paintball work as well?

EDIT: I realize that how long you could keep the refrigerant cold depends on how much insulation you have around the tank. So for short-term usage, you might not need a lot of insulation. But for 24/7 as I've suggested here, you'd need much more insulation. So if you wanted to keep the liquefied-refrigerant cold for more than 24 hours, what would you need? And again, I realize this will also depend on the refrigerant being used as well. But interestingly enough, the larger the volume of the reservoir, the easier it would be to keep the liquid cold. Because the liquid itself will begin to act as an insulator.

The greater the volume of the reservoir, the less surface area proportional to volume there will be enclosing it. So there will be less absorption of heat, proportional to mass.

Shingoshi

[Moc]

Big question: Why do you want to store liquid refrigerant??? A phase change unit is a circuit, you don't need to store cold refrigerant.

[Shingoshi]

Big question: Why do you want to store liquid refrigerant??? A phase change unit is a circuit, you don't need to store cold refrigerant.

I'm building a system for 24/7 uptime. And the system will have multiple boards in it. So for safety concerns, I want a means to allow for safe shutdown in the event of compressor failure. Given the amount of wattage I'm building for (@8 cpus), the processors would have no reserve to keep them cool in an emergency.

I was thinking that two benefits might exist here.
1.) Quicker startup times if any shutdown is required for maintenance.
2.) It might lower the overall compressor load. Like a tank on an air compressor.
EDIT: And this ties both of the above into a single concern. It would take a long time to bring the system up to capacity to handle the load of all the processors. It would take even longer if any graphic cards are included. Having the reservoir would eliminate that problem.

I would use a switch based on my system's activity led, which would turn the tank off only when the computer is actually down. That way if the compressor went offline for any reason, the system would still have a reserve of coolant.

Shingoshi

[SexyMF]

So you want to build an eight head multi-stage rig?

Sounds like you are better off to purchase a LN2 generator and pots.

[Shingoshi]

So you want to build an eight head multi-stage rig?

Sounds like you are better off to purchase a LN2 generator and pots.

The expense of the LN2 generator would likely be greater than what I'm considering. And given the fact I said this is for 24/7 uptime, please explain how that would even be a practical suggestion.

EDIT: Confirmed! http://www.xtremesystems.org/forums/...d.php?t=130368

Shingoshi

[Shingoshi]

So you want to build an eight head multi-stage rig?

Sounds like you are better off to purchase a LN2 generator and pots.

The expense of the LN2 generator would likely be greater than what I'm considering. And given the fact I said this is for 24/7 uptime, please explain how that would even be a practical suggestion.

EDIT: In looking over the Elan LN2 generator, I found the following statement:

The liquefier takes about an hour to cool down before liquefaction of nitrogen can begin.
This process is tracked by the Orange LED (“B”) and Blue LED (“A”) located on the left
of the Display Panel above the Green (“C”) LED. See Fig. 6. When the temperature falls
below -73°C (200K) the Orange LED comes on. This occurs about twenty-five minutes
after start-up. When the temperature reaches -173°C (100K) the Blue LED comes on and
liquefaction begins shortly thereafter. Liquefaction begins about an hour after start-up.

Now while I'm not building an LN2 system, this was precisely my point in having a liquefied-refrigerant reservoir. The start-up times for my system will be horrendous. Having a dewar to pull refrigerant from would shorten that time significantly.

Shingoshi

[SexyMF]

What you are asking to do is in the commercial league. I work in a lab. We have looked at LN2 gas generators and LN2 liquid units. They are expensive but exist all the same.

I can only infer from the application you desire that:

- Since you are using 8 CPUS that your budget is large. Cooling anything 24/7 means you run a large risk of condensation (creeping beyond the CPU zone).
- You can afford to break expensive hardware


Cascade systems handling hundreds or watts are more power hungry and hence more expensive to run 24/7 than a LN2 unit.

You must weigh up the capital expense vs ongoing costs.
Unrockstars rig is a money sink and has only one head.



So give us something more to work with here. What is the application? Is it commercial or simply for 'bling' factor?

[Shingoshi]

I can only infer from the application you desire that:
1.) - Since you are using 8 CPUS that your budget is large. Cooling anything 24/7 means you run a large risk of condensation (creeping beyond the CPU zone).
2.) - You can afford to break expensive hardware

Cascade systems handling hundreds or watts are more power hungry and hence more expensive to run 24/7 than a LN2 unit.

You must weigh up the capital expense vs ongoing costs.
Unrockstars rig is a money sink and has only one head.

3.) So give us something more to work with here. What is the application? Is it commercial or simply for 'bling' factor?

1.) My current system isn't intended to be cryogenic. I was simply asking the question here as to the possibility of cryogenics. The design that I'm working on is a chilled liquid-cooling system. I've already built a liquid-cooled four-processor system. I'm only now wanting to chill it as much as I possibly can. I'm attempting to find azeotropic solutions which can drop my freezing point down as close to ethanol's as practical.
2.) Hopefully won't break anything else.

I'm not following in Unrockstars footsteps here. I simply want an extremely chilled liquid-cooled system.

3.) My system is ultimately intended to be a software compilation server. A build engine if you will. I'm looking for something that can compile all of the packages needed in a Linux distribution (of my own making) in the shortest amount of time possible. Having experience with Gentoo Linux, I'm familiar with just how time intensive the "emerge system & emerge world" process is. I would like to have a machine that can complete that process in twenty-hours or less. As well as be used for community development, testing and research.

Shingoshi

[[XC] gomeler]

Then I would suggest building a cluster to distributed the load or get a quad-socket AMD system(supports modest overclocking) or a dual-socket Gainestown(no overclocking). Mixing sub-ambient cooling with computers will always result in headaches and will in the end kill your productivity as you deal with the myriad of issues associated with it. Get yourself beefy air heatsinks and stay with air. You'll thank me down the road

[Shingoshi]

Then I would suggest building a cluster to distributed the load or get a quad-socket AMD system(supports modest overclocking) or a dual-socket Gainestown(no overclocking). Mixing sub-ambient cooling with computers will always result in headaches and will in the end kill your productivity as you deal with the myriad of issues associated with it. Get yourself beefy air heatsinks and stay with air. You'll thank me down the road

With all of the information on this site and others, I don't know why people keep complaining about condensation. I've already built and operated my quad-socket Tyan S4980 motherboard with an anticondensation layer properly applied. I used Koolance CPU-330 cooling blocks and ran the system sub-ambient. Ran just fine. So why the constant refrain about condensation? Yes, it takes some effort to apply. But anyone who can build a phase-change system is expected to have the skill to defeat condensation.

All of this reminds me of my neighbors French poodle. She never wants to go on the grass and get her feet wet! I can only laugh at her as she considers whether a snack on the grass is worth the effort (discomfort). And this sounds no different. Rose cracks me up every time!

Shingoshi
This is cool enough to be shared here: http://www.xtremesystems.org/forums/...0&postcount=39

EDIT: I've asked a question on another forum which I often used for scientific questions. It pertains to vortex tubes, which is one of my fascinations. I'm trying to find out there whether a vortex tube can condense a refrigerant without needing to use a condenser. It really comes down to how cold I can push the lowest temperature of the tube. If the temperature of the outlet drops proportionately to the inlet temperature, it should be easy to either liquefy or condense any refrigerants in the range of the temperatures reached. If increasing the pressure of the inlet increases the speed of the vortex rotation (as I suspect), an even greater amount of cooling should be obtained. Doing those things could result in a supercritical liquid refrigerant. Meaning the temperature would be colder than it's freezing point, while still being a liquid.
http://www.sciencemadness.org/talk/v...2678#pid160080

[[XC] gomeler]

My concerns about condensation are due to the sheer amount of hardware I have killed with sub-zero benchmarking. I live in Atlanta and I bench sub-zero typically on cascades and LN2. Take a peek at my hwbot profile in my sig, I probably have a few thousand hours experience gathered over ~4 years and I still kill hardware due to condensation. This is why when I say your system would be best on air, I really mean that. Guess I'm going to sit on the sidelines now and watch

[Pointhore]

With all of the information on this site and others, I don't know why people keep complaining about condensation. I've already built and operated my quad-socket Tyan S4980 motherboard with an anticondensation layer properly applied. I used Koolance CPU-330 cooling blocks and ran the system sub-ambient. Ran just fine. So why the constant refrain about condensation?

How much lower then ambient are you talking about when you say sub-ambient? 5, 10, 20, 70 degree's?


These guy's aren't telling you about condensation just to give you a hard time.
One small overlooked missing piece of insulation can kill a component or even a system in a heartbeat. 24/7 use on 8 systems would be tough to make sure that this type of problem would not happen at some point in time.

[sjg0]

With all of the information on this site and others, I don't know why people keep complaining about condensation. I've already built and operated my quad-socket Tyan S4980 motherboard with an anticondensation layer properly applied. I used Koolance CPU-330 cooling blocks and ran the system sub-ambient. Ran just fine. So why the constant refrain about condensation? Yes, it takes some effort to apply. But anyone who can build a phase-change system is expected to have the skill to defeat condensation.

All of this reminds me of my neighbors French poodle. She never wants to go on the grass and get her feet wet! I can only laugh at her as she considers whether a snack on the grass is worth the effort (discomfort). And this sounds no different. Rose cracks me up every time!

Shingoshi
This is cool enough to be shared here: http://www.xtremesystems.org/forums/...0&postcount=39

EDIT: I've asked a question on another forum which I often used for scientific questions. It pertains to vortex tubes, which is one of my fascinations. I'm trying to find out there whether a vortex tube can condense a refrigerant without needing to use a condenser. It really comes down to how cold I can push the lowest temperature of the tube. If the temperature of the outlet drops proportionately to the inlet temperature, it should be easy to either liquefy or condense any refrigerants in the range of the temperatures reached. If increasing the pressure of the inlet increases the speed of the vortex rotation (as I suspect), an even greater amount of cooling should be obtained. Doing those things could result in a supercritical liquid refrigerant. Meaning the temperature would be colder than it's freezing point, while still being a liquid.
http://www.sciencemadness.org/talk/v...2678#pid160080

How can a vortex tube work properly if the refrigerant is condensing inside the tube? Surely that would screw up the mechanics of the vortex, no?