PDA

View Full Version : Which A/C unit?



unixxx
10-10-2004, 08:19 PM
After thoroughly reading the threads I've decided that I want to get started on bulding a chiller system. I just want to know if there's anything I should look for in the victim A/C unit and whether any particular brand/model is more modable than another. I'm going to be cooling five processors plus some other miscallaneous chips with this system so I'm pretty sure I'll need a fairly high BTU, at least 9,000 or more. Just wanted to get other people opinions on what A/C unit I should choose before I shell out the money. Thanx

Redwolf
10-10-2004, 08:39 PM
What are you looking for loop temp (water temp) wise?
With a 9Kbtu unit I wouldn't expect more than around 10F maybe 0F with a 1500W load. Depending on piping length and head you'll prob need around 1/2 to 1gallon per/min pump @ 5ft head.
If you can find a 1/2 ton water cooled condensor or brazed plate heat exchanger for a good price... Buy it.

unixxx
10-10-2004, 09:11 PM
I'm looking for around -20C loop temp wise. BTW, is a 1/2 water cooled condensor really as large as it sounds and what's a brazed plate heat exchanged? Thanx

Juicy629
10-10-2004, 09:20 PM
....9k...uhh thats a alot colder then 0 bro...and since when does a computer generate 1500watts of heat...im lost

unixxx
10-10-2004, 09:36 PM
I'm in the process of making a heat emmision chart to get a total on how much heat I need to dissipate to remain at -20C. Basically, as stated above, I'll have five processors and some other miscelaneous stuff.

saratoga
10-10-2004, 10:33 PM
....9k...uhh thats a alot colder then 0 bro...and since when does a computer generate 1500watts of heat...im lost

You can't compare BTU/hr to degrees C because they're not related things. Basically all the 9000 BTU/hr thing tells you is that its probably a 1 HP compressor (or maybe a 3/4). It doesn't really tell you want final temp you'll get, that depends on the cap tube, gas, exchanger, etc.

unixxx
10-11-2004, 05:11 AM
Right, but I want to try to predict how many BTU will be required for me to absorb the heat my processors will output. If I pick an A/C unit with a lower BTU than what my processors put out than it will be impossible for me to get below freezing because the ambient air will end up doing some of the cooling as the A/C unit won't have enough capactiy. But yeah, I know what you mean, they aren't directly related and I'm not gonna lower my temp just by having more BTU. Anyone have any suggestions about what kind I should buy, or should I just figure out my heat output and match it up to a BTU rating so I get an A/C unit that can handle the heat. Thanx.

Juicy629
10-11-2004, 06:06 AM
yes and the fact that many people have made chillers that have been certain btus and typically they get the same temps...oh and i had no clue bout 5 cpus....ya id shoot for like 1500 btus..

Redwolf
10-11-2004, 06:17 AM
1 Ton of refrigeration = 12,000Btu/hr. At what ever your evap temp is going to be.
In A/C terms with R22.... 1Ton of cooling = 1Hp compressor. (rule of thumb)

A 3/4hp or 9,000btu/hr window shaker might do it. But pull down time would be slow and It might be iffy on those Hot VA summer days. I'd go with a 1hp or 12,000 Btu/hr unit. But your best bet would be a 1hp R404A EM (extended Med temp) or LT condensing unit.

A brazed plate heat exchanger is just what it says. www.flatplate.com
A tube in tube water cooled condensor does the same thing but is somewhat larger...However they are cheeper when purchased brand new. So I'd search around Ebay. I've seen them go for 20bucks.

unixxx
10-11-2004, 06:49 AM
Are you suggesting that I use the watercooled condensor as an interface between my water loop and my A/C loop (so it would be used as an evap) or as a way to cool my condensor?

Redwolf
10-11-2004, 01:02 PM
As an evap.. They can be used either way. Or you could just build your own.
If you buy one.. dont buy a used one. Trust me.

unixxx
10-11-2004, 02:51 PM
Even with a direct transfer of heat in the evaporator using pipe-in-pipe I'll still need a resevoir to prevent me from having one massive bottle neck at the evaporator. Since I'll be running coolant to at least four different systems I'll need at least four pumps and a resevoir to act as a buffer for the pump. The exception would be if I could get a very large diameter pipe in pipe (or make one). Do you think I'll increase my efficiency at all using a pipe-in-pipe evaporator if I have to put the coolant into a resevoir after anyway? Thanx

Redwolf
10-11-2004, 03:58 PM
The secondary coolant will flow no matter what. No need for a rez. This isn't straight room temp water cooling.
You have to think of it in different terms. The more secondary coolant you have to use the more of a load your going to put on the evaporator. End up with a longer pull down time and run time.

HawainPanda
10-11-2004, 05:05 PM
Also, its not necessary to have..4 pumps, that overkill, just get one strong one, that?s how many people work their chillers

unixxx
10-11-2004, 05:32 PM
It'll have to be one strong pump with a cross section area four times that of the smaller four lines (that go to each comp). I guess it should work under these circumstances. But if I had say a 3/4" pump going to four parallel 3/4" lines I'd only have 1/4 the pressure (unless the pump where 4 times as powerfull) and the pump would act as a major bottle neck (can you compensate for the smaller diameter pump with an increase in pressure???). But say if I had a 1-1/2" pump going to four 3/4" lines then I should be fine. For obvious reasons I don't want to run the water blocks in series. I also have to keep into account that the coolant system is going to be situated on the bottom of my tall case and the systems will be setup horizontally, so with a one pump coolant setup the pressure will decrease for each system as I go up. I guess I should put the hottest system at the bottom and the coolest one at the top. The other reason I was thinking about a resevoir is the downward pressure on the pump from water falling from above would put strain on it, if I used a resevoir the pump would only have to deal with pushing the water up, and not with water falling back down. Just some thoughts, I'm no expert, what you guys think given my more detailed description of my setup? Thanx.

HawainPanda
10-11-2004, 05:54 PM
Ah, okey?makes sense unix, I think I?ve actually seen someone do that..run everything in parallel, looked freakin amazing..but really daunting, lol, I can?t imagine if it leaked

Redwolf
10-11-2004, 06:01 PM
This is how I would do it... w/Taco 003 pump (3gallon per min @ 3ft head)

Build me two manifolds out of 1" copper with 1/2" lines (one w/ ball valve on each)
Have all the water blocks return into the manifold with valves. (This way I could control the flow to each block).
Have the supplies to each block on the other with out valves.
Pump will be on the return pumping into the chiller barrel.
Put in a bypass valve (3/4") from the outlet of the pump to the other side of the chiller. This will help give you a 10F TD across your chiller.
Build an expantion tank out of something... 2gallon tank maybe. After filling up the system with fluid I'd pressurize the tank with 5psig of air. But this would be after getting all the air out.

This almost exactly how we build boiler, chiller, cooling tower, and dry cooler loops.

unixxx
10-11-2004, 06:13 PM
Redwolf, excellent idea, I never thought about pressurizing it, that would greatly reduce the strain on the pump. A couple questions though: Should I have a one way valve on the pressurized resevoir the prevent the coolant from being pushed backwards by the air pressure? What exactly is a bypass valve in the context below? What's a 3ft head on a pump? Sorry if any of these sound basic. Thanx.

Redwolf
10-11-2004, 06:24 PM
You put 5psig of air pressure on the expantion tank... So no it wont mix with the water

Bypass valve does just that. Bypasses the chiller inorder to keep it from short cycling and to keep the TD (tempature difference) across it at around 10F

Pumps are rated in gallons per min and feet of head. So a pump rated for 3gallons per min @ 3ft of head. Will do just that. Pump 3 gallons per min when it has to pump up to a point 3ft above it.

Don't forget to plumb in some sort of drain.... So if for any reason you have to drain the system.

unixxx
10-11-2004, 06:31 PM
I know the purpose of an expansion tank in household plumbing is to keep pressure on the main lines so the pump doesn't need to turn on and off as much which would quickly burn it out. If the pump is constantly running what is the purpose of an expansion tank here? What exactly is short cycling?

Juicy629
10-11-2004, 06:45 PM
cooling 5 comps...damn...lotta pump..one massive res...like massive...at least a huge cooler...

Redwolf
10-11-2004, 06:46 PM
As the coolant temp rises and lowers its density will increase and decrease. Expantion tank prevents this.

Well I'm assuming your going to have the chiller on a thermostat which is set for around -21C Off and -19C On.
Short cycling is when the system turns of and off rapidly.. less than a min usually.

unixxx
10-11-2004, 06:51 PM
Ok, so if the chiller got too cold the bypass valve would be opened and if the chiller got too warm the bypass valve would be closed?

Juicy629
10-11-2004, 07:08 PM
umm..if hes going to be chilling 5 cpus...typical chiller is 5000 btus...thats a cpu and say a nb or gpu...now nb and gpus dont generate as much heat as a cpu..so if you figure 5 cpus...if you think a 9000btu or even a 13000 btu will be sufficient?...uhh at first he would never be able to turn the thing off...second he would need a large res to chill the mass amount of liquid running in five systems...not to mentino like a 3-4 pumps to get the liquid moving cuz after the 3 cpu that liquid is going to be freaking warm...which you will need a big ass res to cool it...correct me if im wrong but this is just from experience..

unixxx
10-11-2004, 07:12 PM
Correct, exactly what I was thinking. Anyway, two of the systems should emit much less heat than the other three. But I was originally thinking larger size resevoir and at least 10,000 BTUs, hopefully more, and a pump for each system. I want others opinions as well though.

HawainPanda
10-12-2004, 08:21 AM
OH!...5 systems?!...wow
lol, dang..your def gonnna need, well what temps are you going for?..how much total heat load per system? we gotta decide about that first before looking at how powerful of a compressor you need
man unixxx, this is gonna be pretty awsome, also, you know that your gonna have at least..tens of feet of tubing, you'll need super good insulation

but awsome neverthe less..i can't wait to see

unixxx
10-12-2004, 10:48 AM
I'm going for around -20C or less in each system. I still need to start recording the heat emission of all of the components, this information is pretty easy to get from the manufacturer's data sheet. I'm gonna have two Opteron 244's, one P4 2.4, one PIII 733, two PowerPC's, and a Hitachi SH4. Guess which systems I'll have :)

JSU
10-12-2004, 11:03 AM
um mind explainig why you are going to cool the powerpcs???? also I have no idea what a Hitachi SH4 is......

unixxx
10-12-2004, 02:58 PM
Mainly just to do it and because I don't want any fans in any of the actual systems.

HawainPanda
10-12-2004, 03:17 PM
Oh?..lol, neat!, although I don?t see the point in cooling macs?lol or the Hitachi, but whatever, its your project, I don?t think?I dunno, why not just go with plain water for the macs and Hitachi?

JSU
10-12-2004, 03:32 PM
b/c he already has a loop and buying a rad and pump and stuff would be more expensive.

unixxx
10-12-2004, 04:57 PM
Exactly, if I already have half of the system and these two processors barely produce any heat that I have nothing to lose :)

unixxx
10-12-2004, 06:09 PM
Ok, I'm looking through the AMD Opteron thermal design guide and I see the following value:

Thermal Design Power = 84.7W

As it's "thermal" design power would this be the heat radiation of the processor? I just want to double check because it also has "power" in there and "I/O" power it also listed in the vacinity and it has a value of 4.1W and I know you don't measure the heat output of I/O :) I can't find any other values that could be indicate the heat produced.

unixxx
10-12-2004, 06:19 PM
Sweet, so this is the right value, I found this on Google:

The thermal design power is the maximum amount of power the thermal solution is required to dissipate.

There's some more good info here: http://www.tomshardware.com/cpu/20001107/mobilecpu-01.html

unixxx
10-15-2004, 04:30 PM
Ok, I've bought all of the basic components for one system so I'm gonna multiply their heat output by five for an educated guess of the heat output of all five systems. Even though some of them have less heat output I also haven't factored in anything other than the barebones system into this figure so it should all even out.

ArkhAngelos
10-17-2004, 12:18 AM
Remember that your water temps as you go down the loop will get progressively warmer. Put the processors you wish to cool most first.

Redwolf
10-17-2004, 06:50 AM
Remember that your water temps as you go down the loop will get progressively warmer. Put the processors you wish to cool most first.
Only if you have a slow pump.

unixxx
10-18-2004, 02:53 PM
Each computer will have its own parallel loop, so I should be fine heat wise per system. However, I'm gonna make sure I place my CPUs first (again in parallel), then video and northbridge second (yet again in perallel) for each individual comp. Hopefully this should make sure the coolest water gets evenly distributed among the two processors and then the videocard and northbridge can share what's left. I've bought all of my system parts now, all that's left it to buy the A/C unit and build a case :) ("all", LOL)

Redwolf
10-18-2004, 06:02 PM
nice

gkiing
10-18-2004, 07:07 PM
Good luck insulating it all. If you want to get a few extra degrees, once its built let a little bit of gas out of the a;c unit as theya re overcharged for chilling purposes.

unixxx
10-21-2004, 12:26 PM
Yeah, I saw Gary mentioned multipe times how much cooler you can get just by properly adjusting the refrigerant temperature. He seemed disapointed that no one did :) I'll do that eventually as soon as I get everything else setup. Right now I've gotta design the case, I've got a good idea of what it'll be like, just gotta get it drawn up and order the plexi. So would everyone agree that I need around 12,000BTU+? Thanx.

gkiing
10-21-2004, 01:24 PM
Yeah, I saw Gary mentioned multipe times how much cooler you can get just by properly adjusting the refrigerant temperature. He seemed disapointed that no one did :) I'll do that eventually as soon as I get everything else setup. Right now I've gotta design the case, I've got a good idea of what it'll be like, just gotta get it drawn up and order the plexi. So would everyone agree that I need around 12,000BTU+? Thanx.

Adjusting the refrigerant charge (amount) not temperature. 5000btu is eneough for a -30c chiller, -40c if you insulate it well.

unixxx
10-21-2004, 02:48 PM
Right, the two terms crossed paths in my mind :) I'm gonna have 5 systems on this so I was thinking 12,000, unless you think 5,000BTU will work fine.

unixxx
10-30-2004, 12:15 PM
I found an "Opened Box" 12,000 BTU A/C windows unit at Best Buy a few days ago for around $250 after discount. Do you think this is a good deal or do you think I could do better?

unixxx
11-21-2004, 07:24 PM
Just purchased a 12,000BTU Panasonic A/C unit today for $129 on eBay plus $85 shipping for a total of $214. As most of the units in the 12,000 range run for $400 I thought this was a pretty good deal, even with shipping. It should arrive next week at which point I'll disassemble it and try to post some pics. I've just gotta buy some plexi now. I'm thinking my case is gonna be like a plexiglass rack with plexi glass rack-mount cases so I can swap out systems in it. I ran my Opterons for a while today and the highest I saw them hit under load (installation of Fedora Core 3) was 50 degrees Celsius. The cool thing about chilled liquid is that even if your power goes out you don't need to power the A/C unit anymore as the water will be pretty cool already, you just need to keep it flowing through. So you can get those few extra minutes to save your work with your UPS.

unixxx
12-01-2004, 05:58 PM
Well, I opened up my new A/C unit and it appears to be quality made and has a huge a** evaporator and condensor :toast: along with a compressor with the following specs:

LG Electronics
1 PH 60Hz 115V~ LRA49
Part # 252OUKQC2AA
Model # QK156CAA

Datasheet: http://www.lge.com/products/component/compressor/aircon/download/r_QKseries/pdf/QK156C.pdf
Home Page: http://www.lge.com/products/component/compressor/aircon/r_QKseries.jsp

Its specifications indicate that it uses R22 (Freon) and "SUNISO 4GSI" oil. I can't find a HP rating on the datasheet or website anywhere, does anyone know how I could find this? Thanx.

unixxx
12-03-2004, 02:49 PM
Now for some pics :toast: Unfortunately the image tags don't seem to like ftp so you'll just have to click the links if you want to see them :rolleyes:

The front of the A/C unit:

ftp://68.100.68.78:1090/H2A/ac_unit/PC030001.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030002.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030003.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030004.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030005.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030006.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030007.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030008.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030009.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030010.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030011.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC030012.JPG

unixxx
12-04-2004, 01:00 PM
Ok, I successfully seperated the evaporator, condensor, and compressor without rupturing any refrigerant lines. Here are some pics:

ftp://68.100.68.78:1090/H2A/ac_unit/PC040013.JPG
ftp://68.100.68.78:1090/H2A/ac_unit/PC040015.JPG

And here's the left over crap (and fan):

ftp://68.100.68.78:1090/H2A/ac_unit/PC040014.JPG

Now I'm gonna attempt to remove all of the useless (so maybe it isn't) air routing material and evaporator fan and reinstall the evaporator, condensor, and compressor. The reason I removed them in the first place was to get at all of the useless stuff to be removed under them.

I said before that I wanted to put several systems in this setup and to my advantage the sizing will work out perfectly. A standard rack mount enclosure is 19" x 21", and the drip pan on the A/C unit it just a little bigger than that. It'll make a perfect floor for the rack. I'm gonna get back to work :)

Briggs
12-04-2004, 08:01 PM
That thing is monster! Good luck with the project!