PDA

View Full Version : R507 in an R134a compressor: any danger?



DaBit
17-03-2003, 05:45 PM
Hi again,

I have the possibility to obtain some R507, and a second hand Danfoss NL11F to do some experimenting. If I blow it, so be it.

Now, here is my theory: I can use R507 in a R134a compressor under the following conditions:
- The oil must be compatible with R507
- Low side and high side pressures must be within the original limits
- Discharge temperature must be low enough.

The oil should be compatible since the same oil is used by Danfoss for R404/R507 and R134a.

Allowable low side pressure for the NL11F is 0.84 bar to 2 bar (-30 °C evaporation .. -10 °C evaporation with R134a)
When these pressures are translated to R507 evaporation temperatures, it shows that the evaporation temperature should be within the -50 °C .. -30 °C range. In my application, this requirement is easy to fullfill.

High side pressure is a tougher story since the saturated condensing temperature must stay below 31 °C to fullfill the requirement. Tough, but a liquid-cooled condenser should be able to fullfill this requirement.

Since the compression ratio stays approximately the same and the vapour heat capacity of R134a and R507 does not differ significantly, I do not expect problems with very high discharge temperatures.

Thus, as far as I know there are no technical reasons why it should not work, but am I correct?

And what about safety? I expect no troubles with safety, but I am not 100% sure.

Gary
18-03-2003, 02:15 PM
What is it with this fixation on refrigerants? A different refrigerant doesn't magically yield better results. High pressure refrigerants can be advantageous for low temperature systems, but the system must be designed for it.

Without substantial alterations, you can kiss your compressor goodbye.

BTW Dabit, whatever happened to the liquid chiller system you were building, with the TEV and dual heat exchangers? I don't recall seeing the final numbers on that.

DaBit
18-03-2003, 04:54 PM
Originally posted by Gary
What is it with this fixation on refrigerants? A different refrigerant doesn't magically yield better results. High pressure refrigerants can be advantageous for low temperature systems, but the system must be designed for it.

It is proven that replacing the original R134a refrigerant in a Chip-Con Prometeia (http://www.chip-con.com/index.php?pageid=14) with R404a instantly decreases full-load CPU temperatures with 10-15 °C. This without adapting the capillary tube to R404a.

The compressor used in the Prometeia is the very same Danfoss NL11F I am using, which makes me believe that using R507 should be possible when the operating conditions are adjusted as well.


Without substantial alterations, you can kiss your compressor goodbye.

Yes, but substantial alterations are performed. Suction and condensing pressure are kept within the compressor's limits, and I do not expect problems with discharge temperature since the compression ratio stays the same, too. The things troubling me are the behaviour of the oil at such low evaporator temperatures, and safety (mechanical strength of the compressor).

About blowing the compressor: I will use an old second hand compressor ,which I can get for free, for this experiment, and for now I do not really care if it will hold or not. It is not that much work to coil 10 feet of 1/4" tubing and submerge it into water (condenser), and another coil to serve as evaporator. If the compressor does not hold, then I know I should not use R507 in my 'production chiller'. If it does, and it holds after extensive testing is performed, I might consider using R507 in my 'production chiller'.


BTW Dabit, whatever happened to the liquid chiller system you were building, with the TEV and dual heat exchangers? I don't recall seeing the final numbers on that.

No, you are right. I still have to write the text and process the pictures for my website. I just have to do it, but I have to scrape myself together. After work I prefer using a hacksaw over a keyboard. But until now the system is running smoothly.

A sort update: I ditched the TEV since it was not able to cope with the low load when the processor and GFX card is running idle. Instead, I mounted a capilary tube and adjusted it for full load. During low load situation I get liquid out of the evaporator (of course), which is trapped in an accumulator.

While rendering an animation using 3D Studio MAX (3D modelling/animation tool), the measured values are as follows:
- secondary coolant temp: -24 °C (-11F)
- R134a evaporating at -32 °C (-25F)
- Temperature in the middle of the condenser: 33 °C (91F)
- Liquid temperature exiting condenser: 27 °C (80F)

I am unable to measure the TD between secondary coolant entrance and exit since the difference is within the temperature meter's accuracy of +/- 0.5K

The estimated heat load is 150-200W. I am not sure of the heat load since the Armaflex insulation is not so good after all, so I cannot ignore losses though the insulation.

The latest image of the case containing completed system can be found here: *click* (http://www.icecoldcomputing.com/cgi-bin/vdocview.pl?section=phasechange&spath=case_2&language=english)

Gary
18-03-2003, 05:52 PM
It is proven that replacing the original R134a refrigerant in a Chip-Con Prometeia with R404a instantly decreases full-load CPU temperatures with 10-15 °C. This without adapting the capillary tube to R404a.


Obviously, the manufacturer is too stupid to simply switch refrigerants and thus improve their product.

Pardon me for interrupting. Let us know how it works out.


While rendering an animation using 3D Studio MAX (3D modelling/animation tool), the measured values are as follows:
- secondary coolant temp: -24 °C (-11F)
- R134a evaporating at -32 °C (-25F)
- Temperature in the middle of the condenser: 33 °C (91F)
- Liquid temperature exiting condenser: 27 °C (80F)


That's not what final numbers look like. This is what final numbers look like:

Low side:

Evaporator water in temp
Evaporator water out temp
Saturated suction temp (SST)
Suction line temp at coil outlet
Suction line temp near compressor inlet

High side:

Condenser air in temp
Condenser air out temp
Saturated condensing temp (SCT)
Liquid line temp near condenser outlet
Liquid line temp near metering device inlet

In fact these aren't the final numbers, either. These are the numbers by which we evaluate the system balance and performance, so that we can make the adjustments needed in order for it to work right.

Just because you slapped it together and it works doesn't mean it is achieving anywhere near what it is capable of doing.


A sort update: I ditched the TEV since it was not able to cope with the low load when the processor and GFX card is running idle. Instead, I mounted a capilary tube and adjusted it for full load. During low load situation I get liquid out of the evaporator (of course), which is trapped in an accumulator.


See above.

DaBit
19-03-2003, 10:12 AM
Originally posted by Gary
Obviously, the manufacturer is too stupid to simply switch refrigerants and thus improve their product.

It was not the manufacturer. People who bought such a system tried it.


Pardon me for interrupting. Let us know how it works out.

I am sorry if I made you angry somehow. I am just asking your opinion on the subject. I would really like your opinion since your opinion has proven to be very valueable in the past.

About the R507 in a R134a compressor: it is a bit unconventional, but I can always try. It does not cost me a dime, and if it does not work out, fine, and I will stick with R134a. Fact is: R134a is not the best imagineable refrigerant for < -30 &deg;C evaporation temperatures.


That's not what final numbers look like. This is what final numbers look like:

Should have lookup an old post of you, I agree :)

Low side:

Evaporator water in temp: -24 &deg;C/-11F

Evaporator water out temp: -24 &deg;C/-11F
(difference in temperature within accuracy limits of the temperature sensors)

Saturated suction temp (SST): cannot measure due to lack of pressure sensors. Assume -32 &deg;C/-25F

Suction line temp at coil outlet: -32 &deg;C/-25F, measured inside the suction line, 10cm (4") from the coil outlet.

Suction line temp near compressor inlet: 4 &deg;C / 39F

High side:

Condenser air in temp: 19 &deg;C/66F

Condenser air out temp: 23 &deg;C/74F

Saturated condensing temp (SCT): 33&deg;C/91F

Liquid line temp near condenser outlet: 27&deg;C/91F

Liquid line temp near metering device inlet: 27&deg;C/91F
(please note that condenser outlet and metering device are very close together)

The accuracy of the used K-type thermocouples and measuring device is +/- 0.5K


Just because you slapped it together and it works doesn't mean it is achieving anywhere near what it is capable of doing.

Currently it is performing very close to the capacities mentioned in the Danfoss NL11F datasheet. The only thing that disappoints me is the large TD between secondary coolant and evaporating refrigerant.

Gary
19-03-2003, 01:15 PM
Are these numbers with the TEV or the cap tube?

I would assume the SST to be lower than -32C, which means the heat transfer in the evaporator is even worse. We are severely limited by the lack of pressure reading. You might consider installing a permanent guage in the suction line.

What happened when the TEV was used?

Gary
19-03-2003, 02:47 PM
I've been looking over your website. Very nice. :)

Your first project had two major flaws. You would have gotten much better results by reversing the water connections for counterflow. And of course, the downfall was the elbow fittings inside the evaporator. These were not necessary.

I would point out also that copper pipe caps are readily available at a plumbing supply house.

Etching the cap tube is very creative and resourceful, but there is an easier way to cut it. Use a small file to cut a groove around the tube and then snap it. Use a straight pin, nail, or whatever to make sure the end is fully opened.

Oil return could be ensured by standing the evaporator on end with the suction line down. I would then bend the suction line in a U shape up to the top of the heat exchanger and then down to the compressor. This helps oil return and traps excess refrigerant in the off cycle.

As to the current system:

A TEV is far superior to a cap tube, especially for these experimental purposes. Within limits, it will adjust itself to match the compressor and condenser, leaving you with a single unknown variable, i.e. the evaporator.

The problems you encountered with it on your current system are most likely due to the lack of heat transfer in the evaporator. You should get the evaporator right before you move on to other experiments. One step at a time.

DaBit
19-03-2003, 02:58 PM
Originally posted by Gary
Are these numbers with the TEV or the cap tube?

Numbers are with the captube.


I would assume the SST to be lower than -32C

Why? There is not much pressure drop over suction line piping between the outlet of the evaporator and the compressor inlet. Mass flow is low, the 3/8" piping is wide enough to reduce gas speed in the suction line and therefore friction losses, and there is only 45cm / 1.5ft of piping between the evaporator outlet and compressor inlet.

The measurement of the gas temperature from the evaporator can be considered accurate since the small themocouple is mounted inside the receiver in a piece of 1/4" copper, where the suction gas passes over directly.


which means the heat transfer in the evaporator is even worse. We are severely limited by the lack of pressure reading. You might consider installing a permanent guage in the suction line.

I am planning on mounting a pressure gauge on the compressor's service port. I have more space over there to mount such a gauge.


What happened when the TEV was used?

During pulldown everything worked fine, and the TEV showed it's best side by providing a much quicker pulldown than the capillary tube. However, as soon as the temperature and load reached their lowest points, the TEV became instable and started oscillating by injecting not enough refrigerant , an increasing amount of refrigerant up to the point where liquid was returned to the compressor, and then reducing the amount of refrigerant again until the evaporator starves.

I tried mounting the bulb at several spots (before the SG<->LL HX, after the SG<->LL HX), removing the SG<->LL HX altogether, I played around with the charge in the system, and I played around with the superheat setting of the TEV.

The oscillating effect (which is called 'hunting', right?) could only be removed by reducing charge quite a bit or setting TEV superheat to 12K or more. In both cases the performance of the system was far less than optimal.

Therefore my conclusion was that the ability of the TEV to cope with low loads was insufficient. Which did not surprise me: nominal capacity of the used TEV was 0.35kW. For a lowest operating capacity of ~30% of rated capacity, this still yields more than 100W.

Using a capillary tube and a receiver which captures excess refrigerant at low loads works better for these very low load applications. The captured refrigerant seems to increase vapour content at the captube entrance and therefore limiting the mass flow.

The disadvantage of the capilalry tube: pulldown is not nearly as quick as with a TEV.

I think the best suitable setup for cooling various PC components with a classic refrigeration system is a flooded evaporator setup. Those systems pack more heat transfer per square inch of evaporator metal area, and they don't suffer from the load variations.

Gary
19-03-2003, 03:45 PM
Why? There is not much pressure drop over suction line piping between the outlet of the evaporator and the compressor inlet. Mass flow is low, the 3/8" piping is wide enough to reduce gas speed in the suction line and therefore friction losses, and there is only 45cm / 1.5ft of piping between the evaporator outlet and compressor inlet.

The measurement of the gas temperature from the evaporator can be considered accurate since the small themocouple is mounted inside the receiver in a piece of 1/4" copper, where the suction gas passes over directly.


SST is not gas temperature. It is the temperature at which the liquid is evaporating inside the evaporator. This can be determined by reading the pressure and consulting a pressure/temperature chart.


I am planning on mounting a pressure gauge on the compressor's service port. I have more space over there to mount such a gauge.


This would be very helpful. :)


During pulldown everything worked fine, and the TEV showed it's best side by providing a much quicker pulldown than the capillary tube. However, as soon as the temperature and load reached their lowest points, the TEV became instable and started oscillating by injecting not enough refrigerant , an increasing amount of refrigerant up to the point where liquid was returned to the compressor, and then reducing the amount of refrigerant again until the evaporator starves.


Caused by lack of heat load due to insufficient evaporator surface area. This is confirmed by the very wide difference between SST and water temp (high evap TD). To be sure, there is a point at which the TEV can become unstable, but that point would be at a much lower temperature, given ample heat transfer. This is one area where (within reasonable limits) more is better. That limit will be reached when the TD drops below about 3K. And there are ways to take it even lower. You need a lot more surface area.

DaBit
19-03-2003, 03:48 PM
Seems that we were typing a reply at the same time :)
(edit: and again...)


Originally posted by Gary
I've been looking over your website. Very nice. :)

Thanks :)


Your first project had two major flaws.

Then, I knew even less about refrigeration than I do now...


I would point out also that copper pipe caps are readily available at a plumbing supply house.

They are not easy to obtain here in The Netherlands. Most of that stuff is made from messing. Now I managed to get a few copper endcaps.


Etching the cap tube is very creative and resourceful, but there is an easier way to cut it. Use a small file to cut a groove around the tube and then snap it. Use a straight pin, nail, or whatever to make sure the end is fully opened.

I managed to do so, in exactly the same way you describe. File a groove and snap it. I open up the tube by drilling the last few mils of the tube up to 1mm diameter (captube I.D.: 0.8mm)


I would then bend the suction line in a U shape up to the top of the heat exchanger and then down to the compressor. This helps oil return and traps excess refrigerant in the off cycle.

Now I use a horizontal spiral (phase-change III) where the refrigerant enters on top, and exists at the bottom. Excess refrigerant flows into the compressor's hermetic shell. Is this bad? I don't think so since it happens in every captube system with the compressor at the lowest point.



As to the current system:

The problems you encountered with it on your current system are most likely due to the lack of heat transfer in the evaporator. You should get the evaporator right before you move on to other experiments. One step at a time.

Are you sure my TEV problems were not capacity related? The TEV worked fine during pulldown, when a higher load was imposed on the evaporator.

Another thing: why does my coaxial evaporator (http://www.icecoldcomputing.com/cgi-bin/vdocview.pl?section=phasechange&spath=pc3_2&language=english) perform so bad? As can be seen om my website, I am using extra fins to improve heat transfer, and about 3 meters (~10ft) of 3/8" tubing should be capable of transferring 200W without a too high temperature difference.

After building the heat exchanger I cleaned out the copper with sulphuric acid and acetone, so all residues, oxidation and grease were removed. After taking apart the system to switch from TEV to captube, I checked the inside of the evaporator tube, and it is open and clean. No oil residues etc.

Gary
19-03-2003, 04:17 PM
Are you sure my TEV problems were not capacity related? The TEV worked fine during pulldown, when a higher load was imposed on the evaporator.


Increasing surface area imposes a higher load at any given temperature, or the same load at a lower temperature. The ideal in this case is to impose sufficient load to avoid instability at the lowest temperature the system is capable of achieving.


Another thing: why does my coaxial evaporator perform so bad? As can be seen om my website, I am using extra fins to improve heat transfer, and about 3 meters (~10ft) of 3/8" tubing should be capable of transferring 200W without a too high temperature difference.


200W at what temperatures?

Gary
19-03-2003, 04:38 PM
Now I use a horizontal spiral (phase-change III) where the refrigerant enters on top, and exists at the bottom. Excess refrigerant flows into the compressor's hermetic shell. Is this bad? I don't think so since it happens in every captube system with the compressor at the lowest point.


Whether this causes problems or not, and how severe those problems are, depends on the amount of refrigerant in the compressor on startup. At the extreme it can destroy the compressor. Trapping it in the coil is a good idea, even though it may not be needed. Think of it as insurance.

Gary
19-03-2003, 04:52 PM
Another thing: why does my coaxial evaporator perform so bad?

I should point out that my area of expertise is not in designing systems, but rather in figuring out what is wrong with them once they are up and running.

The two areas are not un-related, but any design suggestions I might offer are based on theory rather than experience. Theory is good, but experience is better. :)

DaBit
19-03-2003, 05:36 PM
Originally posted by Gary
SST is not gas temperature. It is the temperature at which the liquid is evaporating inside the evaporator.

I know, but I have no way of knowing that value without a pressure sensor. The temperature of the gas measured as closely as possible is the best guess I can make.


You need a lot more surface area.

Do you have a wild guess on how much surface area I need? What problems can I expect when I use a small plate heat exchanger? Their construction guarantees an enormous surface area for the refrigerant to evaporate, and an enormous contact area to the secondary coolant.


200W at what temperatures?

Say 200W at -20 &deg;C. But I alway thought TD should not change with absolute temperature? (to put it differently: that TD is a result of various heat resistances and power input only)

Gary
19-03-2003, 08:04 PM
Do you have a wild guess on how much surface area I need?

My wild guess would be about 4 times the current surface area in order to achieve -40C.


What problems can I expect when I use a small plate heat exchanger?

None if sized correctly.


Their construction guarantees an enormous surface area for the refrigerant to evaporate, and an enormous contact area to the secondary coolant.

It's amazing how much surface area they manage to squeeze into such a compact package. Layers of criss crossed accordian pleated plates sandwiched together, with all of the tiny passes in parallel to avoid pressure drop.


Say 200W at -20 °C. But I alway thought TD should not change with absolute temperature? (to put it differently: that TD is a result of various heat resistances and power input only)

This stuff isn't as easy as it looks, is it?

DaBit
20-03-2003, 12:24 PM
Originally posted by Gary
My wild guess would be about 4 times the current surface area in order to achieve -40C.

Geez.
That means about 40ft/12m of 3/8" tubing, with a copper<->water surface area of 3591 cm<sup>2</sup> (0.36m<sup>2</sup>/3.2 sq.ft) and a refrigerant<->copper surface area of 2912 cm<sup>2</sup> (0.29 m<sup>2</sup> / 2.6 sq.ft).

Quite a lot for a lousy few Watts. Where is the bottleneck: the water->copper interface or the refrigerant->copper interface? Are there any numbers on heat transfer coefficients between boiling refrigerant and copper?

How much refrigerant->copper->liquid surface area is used with those >20kW chillers you guys are used to work with? Those must be huge....


None if sized correctly.

Obvious question: what is 'sized correctly'? If I understand you correctly, larger is only better up to a certain point. But wouldn't the smallest available exchangers with only 6 plates or so be sufficient? I am passing only a very low amount of refrigerant through the exchanger (~4kg/hr).



This stuff isn't as easy as it looks, is it?

No, but that is why I post here :)
I always like to understand what I am doing and what is going on.

So, what factors influence the TD between refrigerant and coolant in a heat exchanger? I would say:
- Power input
- secondary coolant and refrigerant velocity
- secondary coolant properties
- the construction material used.

But where does absolute temperature fit in?

Gary, I also have another question on which you might have a good opinion. I am using a water/methanol mixture as the secondary coolant. Based on numbers it should perform very well. And in my system if outperforms propylene glycol based automotive coolant hands down.
See also my little research on the subject (http://www.icecoldcomputing.com/cgi-bin/vdocview.pl?section=phasechange&spath=pc2_2&language=english). But numbers are just that, numbers.

Do you have any experience with such a mixture? Why is this mixture not used in commercial systems? It is toxic, but so is propylene glycol.

Gary
20-03-2003, 01:49 PM
Geez.
That means about 40ft/12m of 3/8" tubing, with a copper<->water surface area of 3591 cm2 (0.36m2/3.2 sq.ft) and a refrigerant<->copper surface area of 2912 cm2 (0.29 m2 / 2.6 sq.ft).


How much copper would be needed if you used 1/4" tubing? Or 1/8"?


How much refrigerant->copper->liquid surface area is used with those >20kW chillers you guys are used to work with? Those must be huge....


We aren't trying to achieve ultra low temperatures in those chillers. If you wanted above freezing water temperature in your system, then your current evaporator could easily do the job. The colder you want to go, the more surface area needed.

As I recall, all of the heat flow formulas have electrical flow formula equivalents, with TD representing voltage and heat flow representing amperage, if that helps. In adding surface area, we are lowering resistance.


Gary, I also have another question on which you might have a good opinion. I am using a water/methanol mixture as the secondary coolant. Based on numbers it should perform very well. And in my system if outperforms propylene glycol based automotive coolant hands down.


I have little experience with glycol chillers. I know that throughout the refrigeration industry, using automotive anti-freeze is considered a VERY bad idea. Beyond that, you probably know more about the subject than I do.

An ultra low temp liquid chiller is quite unique. Usually the product is chilled directly, but then, this application is unique in that the product to be cooled (CPU) is extremely small.

Using a liquid chiller for this purpose limits the potential, but has it's advantages also.

Gary
29-03-2003, 09:22 PM
Here are some further thoughts concerning surface area:

You have already seen the high TD caused by insufficient surface area. Imagine how very high the TD must necessarily be at the CPU, given it's very tiny surface area.

While the temperature of the cooling fluid or refrigerant may be very low indeed, the actual temperature of the CPU is much higher.

This can be greatly improved by reducing the heatflow resistance of the material which separates the cooling fluid from the CPU, but this in itself causes problems.

The ideal would be a micro thin membrane made of highly heat conductive material. On the other hand, the membrane must be strong enough to prevent leakage and/or bursting under all circumstances.

If the cooling fluid is the system refrigerant, then normally the highest expected pressure would be when the system is shut down, and again the nightmare scenario would be if the computer is running without first starting the cooling system and allowing it to pull down to it's temperature, thus heating the refrigerant and greatly increasing it's pressure.

Can you say, "BOOM"?

Now consider the fact that using a much higher pressure refrigerant means much higher pressure under all conditions. While higher pressure refrigerant may allow you to achieve lower temperatures, it also necessitates a much stronger contact surface (membrane) and thus more resistance to heat transfer between the CPU and the refrigerant. It is also more likely to leak and/or burst.

Can you say, "KABOOOOOOOOOOOOM"?

Using a water block offers some advantages here. The pressure on the water side is constant and relatively low, which allows for a thinner membrane. Also, the water block, being professionally designed can reasonably be assumed to take both heat flow characteristics and pressures into account and is likely designed to achieve a reasonable and safe balance between the two.

herefishy
29-03-2003, 11:52 PM
I think that the CPU manufactuerers need to integrate a cooling circuit into their "mold" in order to accommodate cooling requirements. Preferrably in regard to heat dissipation vs. surface area in order to proliferate the sale and use of their product :/

Gary
30-03-2003, 01:18 AM
I think that the CPU manufactuerers need to integrate a cooling circuit into their "mold" in order to accommodate cooling requirements. Preferrably in regard to heat dissipation vs. surface area in order to proliferate the sale and use of their product

Yep. :)

Or possibly a CPU (on an extension cable?), which can be directly immersed in the cooling fluid and tightly sealed, thus eliminating the membrane completely.

In the meantime, I keep thinking the overclockers will eventually re-invent the MOAB.

DaBit
31-03-2003, 09:55 AM
Originally posted by Gary

You have already seen the high TD caused by insufficient surface area. Imagine how very high the TD must necessarily be at the CPU, given it's very tiny surface area.

I manage to keep the TD between the die (piece of silicon forming the processor) and secondary coolant between 10-15K. This is good, if you take all the heat resistances (coarse subdivision: die->package, package->waterblock, waterblock->coolant) and generated heat into account.

Also, keep in mind that the energy density at the core->cooler interface is almost 1MW/m<sup>2</sup> (~100mm<sup>2</sup>, 100W generated heat)


While the temperature of the cooling fluid or refrigerant may be very low indeed, the actual temperature of the CPU is much higher.

Indeed, but there is very little I can do to improve internal heat transfer. I can only start working from the outside of the chip on. This starts by flattening and polishing the mating surfaces and applying silver-based thermal grease.


The ideal would be a micro thin membrane made of highly heat conductive material. On the other hand, the membrane must be strong enough to prevent leakage and/or bursting under all circumstances.

Yes, and there are several possibilities:
- CuSil, an alloy of copper and silver with an extremely good heat conduction. But unaffordable.
- Silver. Very good, but expensive
- Copper. Only somewhat less suited than silver, and way cheaper
- Aluminium. Hardly useable since the commercial grade aluminium used in construction is an alloy with a quite bad heat conduction. And it is not that much cheaper than copper in small amounts.

There are also carbon composites which exhibit an extremely high heat conduction, but it takes a plane manufacturing plant to work with the stuff.

After all, copper is the best suitable material for the do-it-yourselver.

Making the copper base too thin is not so good either since the main problem is the copper->coolant interface layer. Thus, you need the copper to spread the heat so more effective coolant->copper surface area is available.

But is is, as always, a compromise. Make the base too thick and a too large TD occurs over the copper. Make it too thin, and the spreading is insufficient, so the main bottleneck becomes the copper->coolant interface.

A friend of mine ran a few simulations using professional fluid dynamics software. We started it half a year ago, and some simulations are still running. The results are very helpful in understanding what exactly is going on.

The thread on a Dutch forum can be found here (http://gathering.tweakers.net/forum/list_messages/478776/)

A few pictures:

Waterblock design (http://www.hyperreal.org/~geert/test/koel018.gif)

unconventional waterblock, temp gradients and heat flux vectors of the metal (http://www.hyperreal.org/~geert/test/koel018_01a.gif)

Coolant heat distribution (http://www.hyperreal.org/~geert/test/koel018_01b.gif)


I tried writing a simulator based on the Navier-Stokes equations which would be able to simulate the highly turbulent boiling refrigerant and it's interactions with the environment, but the memory and computation power required by such a simulation are way out of reach. No way a normal PC can handle this within reasonable time. A supercomputer with several terabytes of RAM and >1000 processors would be more appropriate.


without first starting the cooling system and allowing it to pull down to it's temperature, thus heating the refrigerant and greatly increasing it's pressure.

Can you say, "BOOM"?

It seems more disastrous than in reality. First of all: temps do not rise above 60-80 &deg;C before the internal protection cuts in. Even when that fails, the system starts to act as a heat pipe. Refrigerant evaporates in the evaporator and condenses at cooler spots, such as the compressor. Thus, pressure cannot reach dangerous levels.


Now consider the fact that using a much higher pressure refrigerant means much higher pressure under all conditions. While higher pressure refrigerant may allow you to achieve lower temperatures, it also necessitates a much stronger contact surface (membrane) and thus more resistance to heat transfer between the CPU and the refrigerant. It is also more likely to leak and/or burst.

Currently my evaporator is a 3/8" pipe, and such a pipe is able to withstand a tremendous amount of pressure before it bursts.

When building a direct-die evaporator (thus the evaporator put on the CPU itself, and not going through an intermediate coolant), the construction strength is also high by design. To spread heat, a base of at least 5mm (~3/16") is required, and adding the required pins, loops and barriers to increase surface area also increases strength. Compare this with a strip of aluminium, and a 'T' piece of aluminium. The 'T' is much stronger.

Some pictures of the 2-layer evaporator a guy naming himself BOWMAN1964 is using:

evaporator design, pic 1 (http://members.fortunecity.com/ajr011/bowmanscoolingextreeme1/Scan10003.JPG)

evap design, pic 2. (http://members.fortunecity.com/ajr011/bowmanscoolingextreeme1/cnc.jpg)

Machined parts (http://members.fortunecity.com/ajr011/artical/cpuevapblock.jpg)

Parts put together, not soldered yet (http://members.fortunecity.com/ajr011/artical/cpublock.jpg)

As you can see, these kinds of design guarantee high strength.


Can you say, "KABOOOOOOOOOOOOM"?

Yes, I can, but I doubt pressure will ever rise above 30 bars, and even if it does this is still far below the burst pressure of the weakest component (which is probably the old and rusty condenser. If I ever switch to R507, this will certainly be replaced by a watercooled condenser to keep head pressure within R134a limits).


Also, the water block, being professionally designed

Haha. This area is still only starting to be explored. Most 'professional' waterblocks you talk about are mady by people like me who made a business out of it. There is absolutely no theory nor reseach behind them. At least nothing more than the reasoning I put behind my waterblock (http://www.icecoldcomputing.com/cgi-bin/vdocview.pl?section=watercooling&spath=wb_construction_2&language=english)

Nowadays it is even getting worse. Waterblocks should not perform well, but they should look good. Many people switch to a watercooled PC only for the 'coolness' of it.

My homemade waterblock can withstand comparison with almost every 'professionally' produced waterblock. Even though I made it with only the simplest tools possible: a butane torch, a hacksaw and a hand drill.


Or possibly a CPU (on an extension cable?)

Extension cables are not possible. Signal speeds are already so high that the transmission speed over the PCB traces is a bottleneck. I won't go into details, but a few inches is already quite a distance.


, which can be directly immersed in the cooling fluid and tightly sealed, thus eliminating the membrane completely.

Then you face the die->coolant thermal resistance, which is pretty high. Nothing would be gained from it.


In the meantime, I keep thinking the overclockers will eventually re-invent the MOAB.

What is the MOAB?

Gary
31-03-2003, 06:08 PM
What is the MOAB?

This is America's latest and largest conventional weapon. I don't recall offhand what the initials officially stand for, something like Military Ordinance Artillery Bomb. I am probably wrong on this.

It's widespread unofficial nickname is the Mother Of All Bombs.

Gary
31-03-2003, 06:26 PM
Dabit, I would say that you display a far greater understanding of refrigeration than the vast majority of overclockers I have encountered.

I would not recommend using a higher pressure refrigerant in your current system, however once you have worked out the design problems and optimized it's performance, it would be a relatively easy step to transform it into a two stage cascade system, using higher pressure refrigerant in the low stage and achieving extremely low temperatures.

Gary
31-03-2003, 06:38 PM
As I pointed out in an earlier post, using refrigerant tubing in the evaporator which has half the inside diameter would require 4 times the tubing to match the inside area. If this tubing is placed in series (4 times the length) it will almost certainly result in excessive pressure drop. This can be overcome by cutting the tubing into 4 equal lengths and running the refrigerant through in 4 parallel circuits. This will result in considerably more surface area with equal or less pressure drop.

And as you discovered in your first system, all brazed connections should be made outside the evaporator.

Gary
31-03-2003, 07:04 PM
On your website, you show an illustration for a basic evaporator. I would use this design, except that I would run many small refrigerant tubes in parallel circuits, joining them together with a manifold at their outlets and a distributor at their inlets. Also the illustration shows the water connections reversed from what they should be. These should be counterflow to the refrigerant.

DaBit
01-04-2003, 09:54 AM
Originally posted by Gary
It's widespread unofficial nickname is the Mother Of All Bombs.

That name somehow reminds me to a game my brother was playing a while ago (He is a gamer, I am not). It also had the Mother Of All Bombs in it, but using that one was not even half the fun of using the World's Most Interesting Bomb. An interesting concept: making a bomb looking so good that everybody comes to look for it, and then it goes BOOM.

I try not to turn my refrigerating systems into a real-world Most Interesting Bomb. I draw more than enough attention whenever my PC is 'in the public', and I hate 'critical failures', to put in in a politically correct setting.

More offtopic: I liked the name 'Daisy cutter' more. The guy who invented that name must have had a tremendous sense for understatement.


Dabit, I would say that you display a far greater understanding of refrigeration than the vast majority of overclockers I have encountered.

I always try to do my homework, though experimentation is just as useful. You need to get the feeling somehow, and you cannot learn that from books. I am pretty sure that my first steps in the world of phase-change (with the K75 cold spray as refrigerant) turns a smile on your face.


I would not recommend using a higher pressure refrigerant in your current system, however once you have worked out the design problems and optimized it's performance, it would be a relatively easy step to transform it into a two stage cascade system, using higher pressure refrigerant in the low stage and achieving extremely low temperatures.

My current system is running smoothly, and I like to keep it that way unless I am sure I can improve it. But I think playing with R507 and a second compressor does not form a serious health risk. 1/4" copper pipe can handle an enormous pressure, my brazed/soldered joints are of a good quality, and the compressor's burst pressure is 5 times the maximum allowable pressure of 28 bars, thus 140 bars.

A cascade is indeed a step further, and I have been thinking about the possibility of using R134a or R22 for the first stage, and CO2 (R744?) for the second stage. This should bring me close to the -80 &deg;C (-112F), and then the bottom is reached. At lower temperatures the CPU won't run anymore.

But that is another story.....


If this tubing is placed in series (4 times the length) it will almost certainly result in excessive pressure drop.

Something I noticed when I used 8 meters (~24ft) of 1/4" tubing.


This can be overcome by cutting the tubing into 4 equal lengths and running the refrigerant through in 4 parallel circuits. This will result in considerably more surface area with equal or less pressure drop.

You read my mind. Using multiple circuits with 1/4" tubing instead of the single piece of 3/8" I am using now could improve performance.

Problem: distributing the refrigerant over these multiple pipes. Massflow per pipe is so low...


And as you discovered in your first system, all brazed connections should be made outside the evaporator.

Now I now.. Experiment, fail, and learn.


On your website, you show an illustration for a basic evaporator. I would use this design, except that I would run many small refrigerant tubes in parallel circuits, joining them together with a manifold at their outlets and a distributor at their inlets.

Would such an evaporator work better than a plate heat exchanger? I still have the idea that the plate heatex is one of the most suitable exchangers for this purpose.


Also the illustration shows the water connections reversed from what they should be. These should be counterflow to the refrigerant.

I know, and I had that comment many times. I will change the picture since a lot of overclockers use my site as a source of information, and they think I am always right. Which is not true.

Dan
03-04-2003, 03:19 AM
Massive Ordinance Airblast Bomb. We use it against much smaller Weapons of Mass Destruction. Regardless, this is a great thread, full of good thought and questions.

I love it when conventional wisdom is challenged. :)

flyinkiwi
13-12-2009, 12:42 AM
Given the size restrictions of the heatsink in contact with the CPU, it would make more sense (to my thinking at least!) to use a refrigerant with higher pressure and vapour density - such as R507 - to make the most of the small contact area available.

S.

flyinkiwi
13-12-2009, 12:47 AM
Given the size restrictions of the heatsink in contact with the CPU, it would make more sense (to my thinking at least!) to use a refrigerant with higher pressure and vapour density - such as R507 - to make the most of the small contact area available.

Furthermore, cryogenic temps = ice/condensation + electronics = BAAAAAD.

S.

DaBit
14-12-2009, 01:52 PM
Wow, it has been a long time since I visited this forum!

The R507 togeether with a Danfoss NL11F compressor was a success. Using a larger condenser kept high side pressures low, and the low evaporation temperature kept low side pressures within the compressors' limit also.

Later on I moved to an R507/R1150 cascade which was able to maintain a stable -110C (-166F) at 200W load on the evaporator. This system is still running, and is now used at CERN in Switserland by a person who uses it to test Hall-sensors for oil drilling purposes.

I'm not doing much overclocking anymore. Bought a house, went living together with my girlfriend, got a kid, etc. This all soaks up money (and time) so not much is left for buying new computer stuff on a regular basis.


But still, it itches sometimes and a week not in the shed is an unhappy week.

Currently I'm working out some details about a solar system using a refrigerant. The problem: PV panels are quite expensive, but provide their energy as electricity, which is by far the most versatile form of energy, and the most expensive to buy.

Solar water heating is cheaper, but this provides a lot of energy when I don't need it (summer time), and not that much when I do need it (winter, heating).

A fluid loop which boils off a working fluid using those modern vacuum+heatpipe tubes, passes the vapour through a turbine or reciprocating engine (which drives a PMDC brushless motor used as generator), condenses it in a regular condenser and pumps it back to the solar tubes could be much more interesting.

Such a system works on a temperature differential, and no absolute value of the temperature is needed. With the condensers at -10C and the working fluid at 0C it still works, although efficiency would be low. But there is still output. For hot water to be useable it must be at least about 35C or so, and preferrably 55-60C.

Such a system is also quite cheap to build. Well, at least for me it is because I have a pile of junk of which I can use stuff.

OK, this is not refrigeration, more the reverse of refrigeration. But nonetheless an interesting topic with the same set of rules.

BUMEMBOLUBZEM
21-12-2009, 01:44 AM
I have an NL Pajero and am in the process of fitting my compressor under the drivers seat. Plenty of room and ARB reckon their pumps will operate on any angle. Those compartments get pretty warm and I would think they are too small and where would you put the jack. Use them for spares.