View Full Version : Evaporator design basics?
Ok guys, gotta make myself an evaporator block, need to know what things I should be considering.
I understand the pressure drop concept, but what is the ideal channel dimensions or area?
How far out of the capillary tube does evaporation take place and is it worth directing follow onto the base plate?
How much channel length is ideal and do 180degree corners assist heat transfer?
any pointers and thoughts would be helpful, as I'm looking at making the block in this case to fit the GPU but also with possible CPU applications, ideal dimension will be around 45mm square 10 - 15mm high with centre side mounted pipes. It's the insides that I need help with.
Thanks
Nico
bowman1964
10-23-2002, 07:20 AM
Originally posted by Nico
Ok guys, gotta make myself an evaporator block, need to know what things I should be considering.
I understand the pressure drop concept, but what is the ideal channel dimensions or area?
How far out of the capillary tube does evaporation take place and is it worth directing follow onto the base plate?
How much channel length is ideal and do 180degree corners assist heat transfer?
any pointers and thoughts would be helpful, as I'm looking at making the block in this case to fit the GPU but also with possible CPU applications, ideal dimension will be around 45mm square 10 - 15mm high with centre side mounted pipes. It's the insides that I need help with.
Thanks
Nico
well i will try.since i do have a few running and making another one now.first check out this artical will give you alot of ideas on what i did.
http://www.phase-change.com/index.php?action=Articles_page&cat=Guides&id=23&page=1
now alot depends on if you are going to mount the motherboard vertical or horizontal.i need to know that first.to give you what design to use.i bet not one person even knows why i need to know that.but it can make a differance of proberly 5c on the block.
on if to put the capillary tube spraying on the die plate ...yes makes a differance.and evaperation starts as soon as it exits the tube.will be the coldest point.you need all the channel lenght you can come up with.i have 60" of channel and it still needs more, to get a few degrees colder.alot depends on capillary tube size and lenght and comp size.now if 90 bends make a differance yes..but hard to say if alot of improvement.I have 52, 90 dregree bends in my channel in my block.one thing that differs my block from all the others.i do not get a spike in temps because the block is so large the cpu cannot affect it with suuden heat changes.at most i get a 5c degree variance in temps when going from full load to idle.unless i run full load for a hour then i get about a 6 to 7c degree drop.i even tested mine by turning off the compressor .....took almost 5 minutes for the cpu to reach 20c at idle and i dont have halt enabled in the chip.
Bowman,
I guess in this application I'm looking to run it upside down (cool face pointing up) horizontally on the GPU. I would like to be able to use it vertically on a CPU, but I'd prefer to optimize it for the GPU. As far as the rest of the unit goes, its an old slot type vapochill.
Was also thinking of regasing with R290, or a mix of R134a and R290. Do you think the BD35F will be upto it. Also would it be possible to use other gases like R404 etc.
bowman1964
10-23-2002, 05:42 PM
I guess in this application I'm looking to run it upside down (cool face pointing up) horizontally on the GPU well that block will have to be one layer and any design will do fine for a video gpu(not much heat)just use small narrow channels.say 1/4"wide by 1/4" deep.since on a video and you want to run this nomaly in the pc.run the capillary tube into the side and suck the gas out beside it.sorta like a old danger den 1 maze.in and out on the same side but with small channels.
I would like to be able to use it vertically on a CPU, well vertical is differant.i have found that if you put the capillary in the bottom and work your maze upward.this will keep the liquid in the maze.and hold it a little longer on the cpu.this is how my newest design is being made.sorta.....those specs i will let everyone see them after i have it back and i am testing.
Was also thinking of regasing with R290, or a mix of R134a and R290. Do you think the BD35F will be upto it. Also would it be possible to use other gases like R404 etc
and i think i would use the r290 and r134a mixture since your compresor already had r134a in it.it will cool pretty damm good.[
Originally posted by bowman1964
[B]well that block will have to be one layer and any design will do fine for a video gpu(not much heat)just use small narrow channels.say 1/4"wide by 1/4" deep.since on a video and you want to run this nomaly in the pc.run the capillary tube into the side and suck the gas out beside it.sorta like a old danger den 1 maze.in and out on the same side but with small channels.
I might be reading into this?, but should the channel are be pretty close to the suction pipe area. Also in order to extend the channel length can I reduce the width to 2-3mm and increase the height to say 12-15mm,keeping wall thicknesses to 1.5mm?
and i think i would use the r290 and r134a mixture since your compresor already had r134a in it.it will cool pretty damm good.
I think I'll try the mix as it should at least improve temps over straight R134a.
bowman1964
10-24-2002, 03:45 AM
I might be reading into this?, but should the channel are be pretty close to the suction pipe area. Also in order to extend the channel length can I reduce the width to 2-3mm and increase the height to say 12-15mm,keeping wall thicknesses to 1.5mm?
that is just fine.just remember you will need 3 to 4 mm or thicker outside wall thickness.to give you more sealing area for the top to solder to.if possible machine a lip into the top to give you more of a rigid block.if you dont it will leak in a matter of time.you might not need the lip if you run around -30c or so but mine runs -80f unloaded and it will leak after a couple of days unless i put a lip into the block to keep the top on tight.
pic is showing my block and how i put a lip into it.
How important is the lenght of the channels vs a central chamber surounded by a maze. The diefference in channel length would be about 2", but would a better evaporation chamber over the core, control load variations better?
Come on guys, I see plently of evaporator pics around but none showing the internal channeling (except for Bowman's of course ;) )
bowman1964
10-24-2002, 04:31 AM
How important is the lenght of the channels vs a central chamber surounded by a maze
wont see that much differance cooling the gpu.but on the cpu yes ........i believe makes a differance.but to keep down the heat and control load variations you need mass.that is as much channel lenght as you can up with ,and weight of block.you see i never get these load variations every one else gets.the reason why????because i use almost 60" of channel lenght in my block and mass of block.this asures my entier block is,say down to -50c.
with a small block that is at at ther same temp,if you apply say 40 watts of extra heat load you will see a sudden temp increase.but with mine being such a large mass it can absorb most of the extra heat load.and only show a small heat change giving the refridgerant time to do its job and remove the heat properly.
Bowman,
Any preference to the grade of copper to be used? c110 seems to have good thermal characteristics, better than Berilym or Nickel alloys anyway.
Also I was planning on using a 1.5mm base thickness with an evaporation chamber and capillary aimed at the base plate. Should allow for good heat transfer but will it still control higher heat loads.
Lastly you said a multilayer design was the go but a two layer would certainly add channel length and mass. Whats the reason behind not using it for horizontal applications?
bowman1964
10-25-2002, 03:45 AM
Any preference to the grade of copper to be used? c110 seems to have good thermal characteristics, better than Berilym or Nickel alloys anyway.
yes c110 is just fine.
Also I was planning on using a 1.5mm base thickness with an evaporation chamber and capillary aimed at the base plate. Should allow for good heat transfer but will it still control higher heat loads.
no not thick enough.needs to be between 4mm to 5mm.
Lastly you said a multilayer design was the go but a two layer would certainly add channel length and mass. Whats the reason behind not using it for horizontal applications?
well the problem is you want to use the horizontal in a upside down configeration to cool the gpu.now you need to keep the refridgerant on the plate next to the chip.on a mutillayer design the refridgerant will try to stay at the lowest point in the block.and this will end up being the lowest layer most of the time until it boils off cooling the gpu.that is not very effecient.so that is why i dont recondmend using mulilayer blocks in a upside down mulitlayer design.
and you proberly already realized i am the first person to ever utilize multi layers in block design.this i guess is because of the complexity involved in making one.but it would even help in water cooling if done right.
Penguin4x4
10-27-2002, 01:18 PM
Originally posted by bowman1964
i bet not one person even knows why i need to know that.but it can make a differance of proberly 5c on the block.
Well no, but I am all ears on the explanation. :)
aenigma
10-27-2002, 02:24 PM
Guess I am one of the only ones that knows why ;)
Because the liquid will puddle up, if your board is laying flat it will pool up over your cpu.For example if you were to use a hollow block, your temps would be very bad because the liquid would be boiling away on the side of the block, lay that motherboard down and it will work fine :)
bowman1964
10-27-2002, 04:29 PM
aenigma i forgot my young friend you learn fast......very good. refrigerant works differant than water.to use it properly you must understand how it acts.
aenigma
10-27-2002, 08:05 PM
Originally posted by bowman1964
aenigma i forgot my young friend you learn fast......very good. refrigerant works differant than water.to use it properly you must understand how it acts.
I have known that for a long time now ;)
Guess we havent talked in awhile....
Originally posted by aenigma
Guess I am one of the only ones that knows why ;)
Because the liquid will puddle up, if your board is laying flat it will pool up over your cpu.For example if you were to use a hollow block, your temps would be very bad because the liquid would be boiling away on the side of the block, lay that motherboard down and it will work fine :)
So if a block mounts upside down as it would on a GPU, then the refrigerant will pool on the back of the block. Ideally then you would want thin channel walls to conduct the heat down a short block to the liquid. Also a thick base would be even more important as it will have to control the heat even better.
So much to learn, so little time!:D
Ok it's all finished and working, temps aren't as low as I would have liked, but I went with R134a and it is probably a little overcharged. With a dodgy k-type thermocouple, I read (no load) -15c off evaporator, -8c on the back of card, (in system) -8c off evaporator, 20c on back of card. From earlier cards with hardware monitoring I remember readings around 52 - 54c, so I'm guessing its working pretty well?
Ti4600 went from 327 to 351 and I vmod shortly to see how much more I can get.
bowman1964
12-14-2002, 05:33 AM
great job :banana: :banana:
keep us posted:smileysex
TheDude
12-14-2002, 07:27 AM
Great thread guys!
Bowman,
Could you please post a link to your first article on phase change cooling basics 101?
Thanks :D
nyTroX
12-14-2002, 12:00 PM
Good job,Nico!
This is some of the info about the evaporators I've looking for.
Thanks guy's!, and now a link to some piccy's ;)
Dual Vapo Li (Gpu Vapo mod) (http://pcdb.overclockers.com.au/view.php?name=Nico&page=pics)
bowman1964
12-14-2002, 11:05 PM
Originally posted by TheDude
Great thread guys!
Bowman,
Could you please post a link to your first article on phase change cooling basics 101?
Thanks :D
here you go
http://www.phase-change.com/index.php?action=Articles_page&cat=Articles&id=21&page=1
:p :p
TheDude
12-15-2002, 08:02 AM
Thanks for pics Nico!
And Bowman for the Link to 101 article!
Great job Bowman!
Good stuff here!:banana:
baker18
12-15-2002, 07:46 PM
Nico,
Looks very good. Keep it up.
baker18
seversphere
12-21-2002, 02:16 AM
great job nico, what are you able to overclock?
Bowman, may I suggest condensation proofing as a topic for future article. I finally understand the basics 100% after reading your two articles at phase-change.com :toast:
Just did some vmodding and can get 371 @ 1.88Vgpu.
Also let some gas out and temps came down a degree or two. The system was originally static charged to 100g. that being a little much as it should be around 85g or less. So far I've let out two 5 sec burst's (a quick 5 second's at that).
Can anbody tell me roughly how much they think I've let out? I'm keen to try some more, but I did notice it takes a little longer to cool down now, so I don't know if I'm allready going into undercharge?
After reading some info on the Vapochill forums, I decided to continue releasing refrigerant. I've let another three 5sec bursts go and temps continue to drop. However it takes longer each time to to get down to temp.
Currently it maintains an idle temp of -12.4 on the evaporator and drops to -11.7 in 3DMark, this is with the GPU running at 371 and 1.88 Vgpu mod. I think I'll check unloaded running next to see what the effective no load temp is on the evaporator. Should be nearing -17 to -19, although I would like to lower it to -21 (vapochill specs) I don't think I'll quite make it.
A little update.
Yesterday I checked the noload temp which came in at -20.7, sounds as if the charge is pretty close.
Today I remembered a little trick of tilting the compressor, Idle temps dropped to -17 in windows and 3dmark brought it down to -16.4. The good part is it now does 378 at the same 1.88Vgpu, with no artifacts.
The only possible problem I can see is that the compressor gets pretty warm, at least warmer than I remember it being. I think a little more testing is in order, but it seems to be do it's thing now which is good.
Anybody got any theories on the tilt and increased compressor temps? could it also possibly be a little under charged now?
bowman1964
12-25-2002, 02:24 AM
Originally posted by Nico
A little update.
Yesterday I checked the noload temp which came in at -20.7, sounds as if the charge is pretty close.
Today I remembered a little trick of tilting the compressor, Idle temps dropped to -17 in windows and 3dmark brought it down to -16.4. The good part is it now does 378 at the same 1.88Vgpu, with no artifacts.
The only possible problem I can see is that the compressor gets pretty warm, at least warmer than I remember it being. I think a little more testing is in order, but it seems to be do it's thing now which is good.
Anybody got any theories on the tilt and increased compressor temps? could it also possibly be a little under charged now?
sure i can tell you excatly what is happening.for one the increased temps in the compressor is just the lowercharge you gave it resulting in a lower low side vacuum.which in turn lowers the boiling point of the r134a.so lower boiling point = increased heat removal......thus more heat removed, the more heat the compresor will see before it gets to the condensor to cool.
on tilting it well this is going to be hard to explain.
if your compressor is like my vapo the capillary tube and filter/drier is mounted as a unit.so when you tilt the compressor you are really tilting the drier/filer also.this is what is helping you.if i am wrong about the compressor and filter being mounted together this wont apply to yours but.if it is,and if you could see inside the filter/drier you would see the capillary doesnt just pull refridgerant from the bottom.it depends on how far the tube was brased up into the drier.it they pushed it lets say 1/2 to 3/4" up in the filter the capillary is proberly getting a trickle of refridgerant into the capillary from the overflow into the stuck up tube end.(a mixture of liquid and gas )now when you tilt the unit you may have covered the tube with a better flow of refridgerant resulting in more refridgerant less gas.this in turn drops the temps .
i hope i typed this so it is understood.
it is simple to me but it sometimes isnt to easy to get typed out so everyelse can understand.
Bowman,
Thanks for the info. From what you say, it sounds like the heat may not be to much of a concern. I might try to get a faster fan for the condenser though.
The unit is the very first version of Vapochill, for slot1/A CPU's and was renowned for the tilt trick giving it up to 10 degree's improvement. The review's I checked up on to day indicated it was a rightside tilt of a couple cm's and is due to inefficiencies in the condenser (fixed in the socket versions). If anything the "filter/ dryer" becomes more level horizontally with the tilt, which could allow higher pressure once the refrigerant fills up high enough to enter the capillary.
Dissolved
12-27-2002, 08:14 AM
awesome job with the pics.
Im also thinking oh phase change cooling my gpu. But i wanted to make a block that cooled the memory as well. Does that seem possible?
i know i couldnt cool both sides of the gf4 card, so id figure that if the gpu/mem side got so cool the back of the card would be cool also. Maybe get a Large heasink for the mem on the back, and over it in Neoprene.
Well heck i guess i could run 2 compressors (1 for cpu and 1 for gpu) Since i have Free r22/414B id only have the buy the parts. So im in need of someone to help me design my blocks (im gonna atempt to mill/drill them myself) and help me a bit on the over all design of the phase unit. And im ready to go!
im gonna make a prommy look like a lil girls doll :p :banana: :banana: :banana:
Dissolved,
I think you hit on the point why it's hard to do the mem as well. Cooling both sides of the board would be possible, but very messy. You could make up block to mount onto a GPU side cold plate, and have the return suction line wrapping around with flexible pipe and either going thru another block attached to mem cold plate on the other side or even just silver soldered straight to the cold plate zig zagging back and forth to cover the plate. Insulation becomes alot harder but not impossible.
Check out Sysfailur's cold plates and you'll get a better Idea of what I mean. ;)
Dissolved
12-28-2002, 11:26 AM
yea, it sounds hard but its worth a try. By the time i get my setup working on the G FX will be out. So id only be setting records will older hardware if i ever get this running :)
Make sure you allow mounting options for both Nvidia and ATI cards, that way you can upgrade ;) .I mount mine using the four holes around the Nvidia core and a Swiftech MCW40 kit, but I have also added holes for GF2/3/ ATI. Now I can upgrade to a 9700 Pro or more likely a GF FX when the come out. As long as Nvidia keeps using those 4 holes in future products i'm right.