Killer Single Stage - Turns Autocascade 19300 Btu rotary build

[Xeon th MG Pony]

good to see more industrial level engineers on here! lots of excelend info Kevin!

[bartx]

Unit seems to be ready Used parts and components:

80 cm long HX before phase separator (6mm pipe inside 12mm pipe on CO2 return)

20 plate heat exchanger

2.7m 0.9mm captube on 1st stage

3.0m 0.9mm captube on 2nd stage

I was trying to charge unit today, but it's not as easy as I thought The biggest problem so far is really high static pressure (even around 17 barg when everything has room temperature). Tomorrow for sure I'm going to braze expansion tank into the unit. I saw some nice numbers (around -80C idle), but I haven't tried load testing.

Couple of questions:

one side of plate HX should be frozen, another side should be warmer than 0C, right?
frozen CO2 captube means proper charge or not?

I noticed, that this unit needs so much refrigerant to get some acceptable temps (-70C). That's why I'm getting really high static pressure. I'm charging with propane/butane mix (cooking gas which goes through the filter/drier) and CO2. I was charging to 3-4 bar static with this mix and then CO2, but it doesn't seem to be a proper proportions. I have also R507, but I'm not sure if it's worth to try.

[Kevin Hotton]

Hi bartx,

I see a couple of problems with your build. The first is that your auxilary condenser HX is much too short. It would need to be much longer (at least 3 meters long). Also for the needed heat-transfer area it should have three small tubes within a single larger tube. If you do need to use this short (small area) auxilary condenser, I would charge with n-butane and ethane and aim for a -50C evaporator.

The next issue is the orientation of your captubes. You need to elevate the phase-separarator higher. Notice that the inlet into your phase-separator is below the liquid outlet tube going into the top of the bullet strainer. The too low phase-separator will not allow the phase-separator to empty over this high tubing bend until the liquid level in the phase-separator is above the phase-separator inlet tube. The second captube shares the same problem - the liquid flow that should enter the captube inlet is forced to flow upward (this is a bad design, and will cause vapor to form at the captube inlets - greatly increasing back-pressure). This bad positioning of components will also increase the needed refrigerant charge that you have experienced. To correct your piping issues, you can move the plate HX. Mount the plate HX with the ports down directly over you phase separator (with the vapor outlet going directly into the plate HX. The liquid port out of the bottom of the phase-separator can flow horizontally into the bullet strainer and then into the captube. Once into the captube, the flow can go upward as high as needed without trouble.

Also, with an autocascade of this design, the first captube should flow about 1.5-times the second. So I think you need to shorten your 1st captube to 2m.

Glad you will be adding a fade-out vessel. Without one, your static charge will be high (over 20-bar) and with a fade-out vessel connected with a captube will allow a slow controlled chill-down.

With these changes, you can try again with your cooking gas and CO2. I think it will work better.

Kevin

[Moc]

Hey Bartx,

I would also advice to change some more things:
1) Suction gas coming from the evaporator should also pass the plate hx!
2) The suction gas coming from the evaporator should now go through the auxillary condenser (which I also think is way to small... 3m 16mm/2x6mm should be alright) and then into the compressor!
3) Co2 is not a good refrigerant for this application as it will freeze and block your evaporator/captube.... you'll find tons of posts about that if you search.

Great fabrication btw !!!

[Kevin Hotton]

One more thing bartx,

I didn't catch this earlier, but I believe your captubes are too small in diameter. I would recommend the next size larger (in the range of 1.2mm ID).

[bartx]

Today, I've brazed expansion tank to he unit, and lowered both filters (fortunately, no brazing was needed in this case, only some bending). Static pressure still seems to be high, but 13 barg looks much better than 17. And pull down discharge pressure is much lower too.

I'm not touching the auxiliary HX so far, because I don't have pipes and only half on this HX gets covered by frost- I believe, that on well charged unit it should be frozen up to compressor suction, which affect on phase separation.

I tried some load testing, unit was holding 260W @ -59C for 10-15min with 0.5barg suction and 17barg discharge (ambient temp around 18C). Together with high static pressure, I think that both captubes are way too long - I don't have different size of captube to change, but both will be shortened tomorrow. I'm going to try 2.2m on first stage and 2.8m on the second (0.9mm).

I know that CO2 is not the best, but gases like R23 or R1150 are expensive and the most important - hard to buy. I'd be happy with -65C loaded and I think that it's possible with this mix. On the other side, using R1150 is really tempting

[runmc]

Great project so far. I have always been interested in Autocascade, so this is a real treat.

[Kevin Hotton]

Bartx,

If not upsizing the captubes, you definitely need to shorten them. 0.5-bar suction and 17-bar discharge will result in compressor overheating. A properly designed autocascade does not need an excessively high pressure ratio (you have 18-bar abs. / 1.5-bar abs. = 12). You need to get that down to a pressure-ratio of under 6. For compressor cooling raising the suction pressure is best.

If you use CO2 with the cooking gas, I cannot help with charge and fine tuning of your design. With CO2 when you get down to 5-bar and below -57C you get dry-ice (solid CO2). If have read that blending hydrocarbons with CO2 (like you are doing) can be made to work below -57C, but results are unpredictable and I suspect compressor overheating issues as well.

Kevin

[bartx]

Yesterday, my friend ryba gave me a hint about SLHX. Today I bought some fittings and cut all small pipes, which I found in my workshop to make a SLHX.

SLHX is 30cm long, outer pipe is 18/16mm. Inside there are eleven captubes - five 3,5/2,2, five 3,0/1,8 and one 2,0/0,8. From my calculations it's equal to 1,8m of 6mm pipe and the most important thing internal cross-sectional area is almost two times bigger than 6,0/4,2mm pipe. It's also worth to mention that it's much smaller and easier to insulate. This copper rod between the captubes is placed there to make brazing easier.

I ran out of oxygen, tommorrow afternoon I'm going to collect my tank, braze whole SLHX and replace the older one in the unit. Last question - where exactly I have to mount this SLHX:

- on CO2 return (current SLHX is now mounted there)
- on propane return
- on combined CO/propane return

[samsarulz]

Great work! Let?s see how it performs when you finish it

[Kevin Hotton]

bartx,

Study the attached diagram to see how the copper tubing should go. The table shows the temperature and pressures for the -70C autocascade design I presented months back. The refrigerant blend for this one was 50% n-butane, 50% ethylene by mass. The cooling duty was predicted at 317W using a much smaller 10.2 cc/rev compressor (yours is 19.1 cc/rev). In the diagram Q-Cond1 is the air-cooled condenser, Q-Csc/Q-Hsh is the auxilary condenser, Q-Cond2/Q-Evap2 is the cascade condenser, and Q-Evap is the one cooling your computer's CPU. Notice the cooling duty are: Cond1 of 951W, for the auxilary condenser it is 605W, the cascade condenser at 528W, and finally the CPU cooling evaporator at 317W. This is the normal progression of heat-transfer duties (Large to small going from the air-cooled condenser to the final coldest evaporator). With this in mind, I believe it would be best to use your one plate HX as the auxilary condenser, and the "tube-in-shell" copper HX you are building as the cascade condenser.

Also, because your compressor is so large for your desired 300W. I would make the first captube (the one using the liquid out of the phase-separator) significantly shorter than the other captube feeding the evaporator. That is by-pass the majority of the refrigerant flow and charge predominately with the cooking gas with just the minimal amount of CO2 to reach you temperature goal (-65C loaded). This is because I feel your HX's are too small for the size of your compressor and a high by-pass rate should allow you a higher suction pressure and better compressor cooling while still reaching your performance goal.

One error in the piping diagram, I don't show the fade-out (expansion) tank. It would pipe in (using a captube) just before the suction accumulator on your compressor.

Kevin

[bartx]

This HX on photos above turned out to be complete crap - it was impossible to braze it. I put all those captubes into bigger pipe (35mm) and it seems to be working better.

Autocascade was rebuilded a bit, both filter-driers are now placed correctly, I've installed new SLHX and shortened both captubes - 1st stage now has 2.05m of 0.9mm, 2nd stage has 2.45m of 0.9mm. Discharge pressure is now around 13 barg, static is around 11, so there is big improvement. I'm considering shortening them even more, I'll see tomorrow how it will perform on longer load tests and proper charge (I've learned so far how to charge autocascade ). HX is also insulated, because it has big impact on condensing pressures.

[bartx]

Some pics

I made longer load testing and so far I've got -57C @ 260W heat load, what is good, but I believe that it can perform much better. Discharge pressure still concerns me a bit, because it's near 15.5barg, when suction is 0.3-0.4 barg. And unit behaves like it needs more CO2 charge, which means to me that captubes are still too long. I know that it isn't fully insulated, but I think it that full insulation will have big impact on pressures and temps. Idle is around -75C.

There is big improvement in comparison to my first runs, suction pipes and suction accumulator gets much colder, which means better cooling for the compressor.

I'll keep you updated. I'm going to try 1.8m and 2.25m of captubes and we'll see. You were right Kevin, that both captubes are much too restrictive. On the other hand, I've never seen any autocascade on XS which was tuned for 260W heat load. In older threads there are couple of autocascades, which are tuned for 150W but it's way easier to make unit to hold only 150W.

It's hard to make all modifications (vent all the gas, then vacuum, purge with co2 etc.), but I'm having a lot of fun with tuning this unit. Thank you guys for all your help My goal is -65C @ 260W and I hope that it will be achieved soon

[Moc]

Nice results. You will get even better results if you pipe your unit like its shown in Kevins flowchart....

[Kevin Hotton]

Hi bartx,

Glad things are going better for you now. I like the looks of your homemade HX. You are free to do as you see best, but this is what I believe will help the most.

1) I would swap the location of the two HX's (use the plate HX as the auxilary condenser after the air-cooled condenser, and the one you made to be the cascade condenser).

2) I would only shorten the first captube to 1.2-meter length

3) charge predominately with your "cooking gas" propane / butane mix. with a small amount of CO2 only.

This will accomplish the following. You are currently getting high discharge pressure and low suction pressure because both captubes have vapor at their inlets. By doing 1,2,3 above, you will have much better cooling to create liquid at both captube inlets. This will lower discharge pressure as well as give you better cooling duty. And by leaving the second captube unshortened, a colder evaporator temperature.

regards, Kevin

[n00b 0f l337]

I have to agree. The suction from evap needs to head back thru HX then back thru SLHX pre-separator. Otherwise, it looks great.

[bartx]

I've made another changes today. Shortened 1st stage captube to 1.60m and connected propane return before SLHX, but after interstage HX (before propane return was connected near the compressor). Charged as Kevin adviced and finally I have -61C @ 260W. I insulated most of the pipes, but it was only temporary - they were all wet under the insulation. Now I started to make a final insulation, but it's a lot of work I hope to get couple degrees lower temps after insulating. Pressures are 14.5barg discharge and 0.4 barg suction - they should be a bit lower after insulation, I'm also considering to install 120mm fan to make additional cooling to compressor. Idle is around -68/70C.

I will made some tests on AMD cpu soon

[Kevin Hotton]

Hi bartx,

Small but positive improvement to your autocascade. I have to ask, "did you swap the position of your HX's?" I feel it is very important that you use the plate HX as your auxilary condenser and the one you made with copper tubes as the cascade condenser. You need the larger heat-transfer area of the plate HX (in the auxilary condenser position) to make the autocascade work properly.

Kevin

[bartx]

No, I didn't swap them - I have cut half of the pipes and replace them

Now, at the output of SLHX I'm getting frost, with insulated HX and pipes. Previously I had couple degrees of Celsius above 0.

Loaded temps (pipe coming from evaporator and HX):

[bartx]

Tested today on Phenom II X6 CPU. 4800MHz @ 1.8V wPrime 1024m - -61/62C under load. I don't think that there is cpu with more power draw. I have to finish insulation (only small part is uninsulated), but I couldn't resist and I had to run some benches

[tiborrr]

Mine did roughly 250W on -60?C (LPG/CO-2) with 9600btu/h rotary. But great effort and excellent result!

[bartx]

Mine did roughly 250W on -60?C (LPG/CO-2) with 9600btu/h rotary. But great effort and excellent result!

On 250W load tester or on CPU? I think that Thuban on this vcore has more than 260W...

[tiborrr]

If you want to really stress it out then you should try FX-8150 or 8350 overclocked to about 6GHz (it should do it at -50?C) and 1.7V+ and then try Wprime That should be in exceed of 350W.

[piotres]

very clean autocascade and good results bart !

[samsarulz]

Great, looks insane jaja. Great job- Just to know, every time I see a rotary compressor most of the people uses R507, there must be a reason .