|
|
General Liquid/Water Cooling Discussion For discussion about Full Cooling System kits, or general cooling topics. Keep specific cooling items like pumps, radiators, etc... in their specific forums. |
Thread Tools |
05-27-2003, 10:56 PM | #1 |
Cooling Savant
Join Date: Nov 2002
Location: portugal
Posts: 635
|
Several random questions here
Ok, time to get deeper in this 'waterblock construction' thingy.
here are some questions, and i will ask more as i remenber them: How you do you calculate the perfect fin area? In cathar's ww i would just do 0.1 mm fins with 0.1 mm channels, but looks like that would be bad, why? what are the formulas/calculations/teories behind it? How do you calculate the area of the sprayers? i mean, in cathar's cascade block, he has 36 'tubes' with i dunno what diameter, making an X area of sprayers. How is that calculate, so that you calculate it without becaming low speed sprayers and without restricting too much the flow? The pump parameters enter here too right? How do you calculate base ticness? i saw somewhere it was related to the fins area?? :shrug: why turbulance is better? water makes more contact with the 'walls'? Cant we add something to the water to make it 'contact' more with the copper? how to calculate flow resistance in a block? what is the impingement efect? NOTE: english isn't my native language, normally i get quite confused, but in this treat if i do, i will just keep asking untill i understand it sorry |
05-27-2003, 11:23 PM | #2 |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Oh lordy!
Where does one begin without feeling like they're writing the first few chapters to a book on water-block design? Last edited by Cathar; 05-27-2003 at 11:30 PM. |
05-27-2003, 11:28 PM | #3 | |
Cooling Savant
Join Date: Dec 2002
Location: Dallas
Posts: 339
|
Quote:
What a great Idea A book on water block construction – I can imagine it now – “Water Blocks Manufacturing Basic” at Amazon.com only 14.99 – I’d buy it …… |
|
05-28-2003, 01:36 AM | #4 | |
Cooling Savant
Join Date: Jan 2002
Location: Ashland
Posts: 296
|
Quote:
Hell Yeah! thats a really good idea.
__________________
Air cooled my ass. |
|
05-28-2003, 08:31 AM | #5 | |
Responsible for 2%
of all the posts here. Join Date: May 2002
Location: Texas, U.S.A.
Posts: 8,302
|
Re: Several random questions here
Quote:
#1: You want to keep a channel-to-fin ratio between 1:1.5 and 1: 0.75. The actual width will be limited by the tooling you use. Btw, if you think you can make 0.1mm fins, think again, or let us know how! #2: The calcs are based on Cathar's simulation. #3: Trial and error. #4: Again, the tooling is a big limitation. Even at a low speed, the design will perform well. #5: As with many designs, the block will perform better with a higher flow rate. The trick is making a block that will allow you to use a reasonable pump, and reach a spot where the flow rate is at a point where increasing it further will yield little improvement. #6: Bp thickness is related to the topside design, yes. If your design isn't efficient, then you have to use a thicker baseplate, to allow the heat to spread, so that it can be dissipated efficiently. otherwise, the coolant flow won't be able to pick up the heat efficiently, and you'll end up with a higher temperature difference, between the CPU and the inside of the block, which will result in a higher CPU temp. #7: Water has certain thermal properties. Beyond that, it has properties as a fluid. By putting the water into turbulence, you're effectively reducing the resistance of the heat transfer to the water. Check out the OC article about it. Pins, fins, or whatever shape you see fit. Continous fins over the core area are extremely effective. #8: At the detriment of its thermal properties, maybe. #9: If you figure that one out, let us know! I've tried various formulaes, and applied them to each broken down individual components of a waterblock, but you also have to account for large variations in channel size differences (i.e. orifice plates), and that can be challenging to calculate, especially if a channel configuration doesn't fit a known model. It's often easier to just measure it, and an option to do that, if you are making the block from a CNC, is to make a plastic prototype, just for flow testing. #10: First seen in a document about how to cool a Korean nuclear reactor, in case of emergency, it's the action of shooting water straight up against the surface to be cooled. The idea is that the collision throws the coolant into turbulence. It's cheaper to achieve turbulence this way, because you wouldn't believe what kind of flow rates (and pump) are required to put flowing water into turbulence! I'll let you search for the documents I've mentionned. If you post a link to them, I'll answer more questions! |
|
05-28-2003, 09:04 AM | #6 |
Cooling Savant
Join Date: Jan 2003
Location: Sydney, Oz
Posts: 336
|
satanicoo - this forum and others like it (overclockers.com, overclockers.com.au, overclockers.com.nz) have a FEAST of information to get you in the right mindset for designing waterblocks.
You also have to keep up, as annoying network engineers from Melbourne have a nasty habit of turning up and changing the paradigm every once in a while. However, the design of waterblocks, not too suprisingly, is a graft of thermo-dynamics and hydro-dynamics. So, how heat moves through solids, how liquid moves through solids, plus the science behind the cooling effect of a liquid passing a solid. Lots of reading there, and hydro-dynamics is a black art in an age of science to the masses. In the end, even the greats take a basic design, and tweak it through trial and error. The washing-up-coffee-cup design is going through this process now. You make two, you compare and you "theorise" why differences in design effect differences in performance. You make a third to prove your theory. Rinse. Repeat. However, especially at the bleeding edge, everything impacts on everything, and then you might just get lucky with certain measurements. Especially when you have mere-mortal testing equipment such as a single pump and motherboard diode for temperatures....
__________________
Long Haired Git "Securing an environment of Windows platforms from abuse - external or internal - is akin to trying to install sprinklers in a fireworks factory where smoking on the job is permitted." (Prof. Gene Spafford) My Rig, in all its glory, can be seen best here AMD XP1600 @ 1530 Mhz | Soyo Dragon + | 256 Mb PC2700 DDRAM | 2 x 40 Gb 7200rpm in Raid-0 | Maze 2, eheim 1250, dual heater cores! | Full specifications (PCDB) |
05-28-2003, 09:06 AM | #7 | |
Cooling Savant
Join Date: Jan 2003
Location: Sydney, Oz
Posts: 336
|
Quote:
__________________
Long Haired Git "Securing an environment of Windows platforms from abuse - external or internal - is akin to trying to install sprinklers in a fireworks factory where smoking on the job is permitted." (Prof. Gene Spafford) My Rig, in all its glory, can be seen best here AMD XP1600 @ 1530 Mhz | Soyo Dragon + | 256 Mb PC2700 DDRAM | 2 x 40 Gb 7200rpm in Raid-0 | Maze 2, eheim 1250, dual heater cores! | Full specifications (PCDB) |
|
05-28-2003, 09:39 AM | #8 | |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Quote:
1:0.75 ranging to 1:1.5 came from a research paper I read, and seemed to match up with my anecdotal experience. It specifically applies to (mini)-channelled designs - being channels less than 2mm in size. With channels larger than that you can basically forget chasing leading edge performance. It does not apply to the impingement cup design. 0.1mm channels/fins would result in a waterblock who's pressure drop is WAY too high. The block would perform quite poorly unless you had a gear pump applying ~50PSI to it to get sufficient flow. If in a White Water style design, you'll also lose the breadth of the jet impingement region. Part of the reason for the width of the White Water's jet is because it dictates the resultant width of the impingement region. Making the block shorter would also alter the width of the impingement effect that is responsible for a fair portion of the block's extra "push" up the performance slope. The area of effect over which the sprayers operate is determined by working out how far the heat will spread for the largest CPU die you intend to cool. Currently that's the P4 with an IHS. With a thin base-plate, you need to focus on about a 16x16mm area. For the longer (~14mm) off-centre barton dies, the area also needs to be about 18mm wide, so a 16x18mm area is the focus here which works well for the CPU's current and forwards looking. The next step will have over 50 sprayers in a slightly smaller area. Hitting the limits of conventional machining here. The goal is to pack as many into the target area as possible. The total pressure drop for the sprayers is worked out on the total orifice area open for the pump to push through. Through experimentation I've determined what I believe to be the best trade-off of flow vs pressure-drop vs jet velocity, and the total orifice area which gives me that best trade-off point. The target total orifice area divided by the number os sprayers dictates the orifice area that each sprayer should use, and simple geometry then tells me the diameter of each sprayer hole to meet that requirement. The pump parameters are important, but I don't tend to work in those terms directly. I work more by guaging what seems to be working for real actual pumps that people actually use, and finding the best balance point for them. Best base thickness can determined by trial and error fairly easily. It is directed tied to the convectional efficiency of your design. Lower convectional efficiencies require thicker bases. Using the results of established research, it's possible to estimate your convectional efficiency, and establish a good starting point to try with the base-plate thickness to suit. Turbulent water helps to disrupt the boundary layer that forms along the metal surface. The boundary layer is a thin slow-moving insulative layer of water that hugs the wall and reduces thermal transfer efficiency. Jet impingement is extremely effective at stripping away the boundary layer by blasting a narrow focussed jet of water at a surface. Where the jet impingement in most effective there is effectively no boundary layer occuring. This is the region of extremely high thermal transfer efficiency, far higher than simply trying to brute force the water into turbulence by using a bigger pump to make the water go faster. Well, that's my take on some of the answers to the questions. |
|
05-28-2003, 01:05 PM | #9 |
Put up or Shut Up
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
|
Well I am a die hard trial and error dude. I do not know much about scientific stuff. I design a block I think might work by whatever experience I have gathered in the past and make it. If it doesn't work well then I make another with moddified charicteristics that should make a difference such as base plate thickness or what ever uniqe feature you added. For example my pyrmid pins on my REV. 4.0, I started out with them being TO small (REV. 2 and REV. 3), and then worked my way up untill I hit the jackpot with REV. 4 (severly modified from the others REV'S and still has room for improvement). I never made any scientific calculations on anything. I showed the REV. 4 to a design engineer that designs heat exchangers and he was like "Wow, you came up with that?". I went to high school with the guy (same class in fact) so he knows I am not to bright when it comes to calculations. Basic fundementals will get you a long way without calculations. But you do listen to the people that know the calculations and the science as thats where you pick up the basic fundementals as with experience.
Browse around the forums and find blocks people made that seem to work well and read what they say is their theory behind why it works. Good chance they are right with their theory although not always. Ask questions if it isn't to clear. You first need to UNDERSTAND why it works before you make your own version though. There are plenty of people that see a design that works well and not understand why it works well and then makeup a some rediculous thing they call a block and think it will work because it looks similar to one that does work. |
05-28-2003, 05:17 PM | #10 | ||||||
Cooling Savant
Join Date: Nov 2002
Location: portugal
Posts: 635
|
Quote:
Quote:
But: Quote:
Quote:
Quote:
Quote:
"North korea back to Nuclear research" and stuff like that. Not a good googler unfortunately :shrug: . I loved Cathar's explanations on turbulence and impingement. Now it sounds natural . I know there are lots of info all over the web, but its normally quite dificult to understand, i really never understood why turbulence and impingement were such a great thing, but now i understand why. Hi Jaydee I'm glad you have the tools to do it. I am also making some tests. you will heard about them later. Wait for me . Thanks all, will post more questions as i remember. |
||||||
05-28-2003, 06:09 PM | #11 |
Responsible for 2%
of all the posts here. Join Date: May 2002
Location: Texas, U.S.A.
Posts: 8,302
|
The total flow resistance is the sum of all the parts. In this case, any way you add it, it's very hard to pass water through a slim opening.
I'll see if I can dig up that nuclear reactor paper. |
05-28-2003, 06:50 PM | #12 | |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Quote:
With the mini/micro-channel blocks, fin height is optimally between 4-5x of the channel width. Heat won't travel up the fins in any significant fashion past 4x high, except at really low (<2lpm) flow rates where at 5x high is possibly a better balance. Making the channels and fins 10x thinner, but keeping them at the same height as in the White Water, would result in 90% of the water flowing through the block not doing anything. Making the fins a balanced height would result in a block that's 10x more restrictive than the White Water, and offer about 6mm^2 of orifice area for the water to flow through. This added exit restriction would also stop the impingement region forming properly (if at all), so you can see that there really is a balance going on here. As you keep making the channels smaller and smaller, you're actually interfering with one of the aspects that makes the White Water design work well. In essence you're making a strict micro-channel block, but now have to work super hard with the pump to make use of it. So the White Water design really does have a lower bound on how far you can go with reducing the channels before it effectively stops being a "White Water" design and merely becomes a micro-channel design with a central inlet. I do have a very good idea of what is truly optimal for the design, but the current design is very close to that point anyway, and the performance differences really would be minor (~0.5C on a hot CPU) for a large increase in machining time and cost, and that is the other main factor in water-block design - how cheap can it be done? All well and good making a one-off that takes many hours to make up, but it'd cost hundreds of US dollars if you ever wanted to produce it. |
|
05-28-2003, 06:58 PM | #13 | |
Put up or Shut Up
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
|
Quote:
|
|
05-28-2003, 08:10 PM | #14 |
Cooling Savant
Join Date: Nov 2002
Location: portugal
Posts: 635
|
another thing: jet impingement is best at 90 degrees to the copper right?
|
05-28-2003, 08:47 PM | #15 |
Responsible for 2%
of all the posts here. Join Date: May 2002
Location: Texas, U.S.A.
Posts: 8,302
|
Yes.
I can't find that Korean nuclear reactor emergency cooling jet inpingement doc. Anyone else? |
05-28-2003, 09:22 PM | #16 |
Cooling Neophyte
Join Date: Apr 2003
Location: USA
Posts: 96
|
you now...a little off topic....but I think I have found the perfect pump at work. Well, actually it is 2 pumps...2 100HP pumps that run a waterjet cutter. Nice pressure, something like 80,000PSI. containing this bugger would be a challenge, since odds are it would pierce through (well, its designed to ) any baseplate, then the processor, then the mobo, the case, the wall......But say you could come up with a diamond base plate...talk about impingement......
|
06-01-2003, 07:38 PM | #17 | |
Cooling Savant
Join Date: Nov 2002
Location: portugal
Posts: 635
|
Quote:
Got to apreciate yours and all answers you ppl are giving me. They are giving me some lights . And another thing, i readed that continuous fins are way better than pins or stripped fins. So, WW won with the impigement, but lost about continuous fins, since they are not connect. I must ask: with the fins connected and the midle plate, the one who makes the jet, wouldnt it be better, or the walls kind of destroy the impingement efect? And if they do, what about this: /edit/ off-topic: BigBen2k told me that, on air, the perfect fin/channel ratio is 1:8 to 1:10, and is it? Sorry to question this, but the best koolers like slk900, they have what ratio? 1:3? |
|
06-02-2003, 09:11 AM | #18 |
Responsible for 2%
of all the posts here. Join Date: May 2002
Location: Texas, U.S.A.
Posts: 8,302
|
No, WW does have continous fins: you've misunderstood the meaning of "continous".
|
06-02-2003, 03:12 PM | #19 |
Cooling Savant
Join Date: Nov 2002
Location: portugal
Posts: 635
|
there are 2 sets of 7 continuous fins, but they arent connected between them. If they crossed the zone right above the core, heat would pass to them easier right?
|
06-02-2003, 05:48 PM | #20 |
Responsible for 2%
of all the posts here. Join Date: May 2002
Location: Texas, U.S.A.
Posts: 8,302
|
In theory, yes, but it requires a special flow geometry, hence my proposal, and Cathar's subsequent "Cascade" block.
|
06-02-2003, 07:43 PM | #21 | |
Thermophile
Join Date: Sep 2002
Location: Melbourne, Australia
Posts: 2,538
|
Quote:
|
|
Currently Active Users Viewing This Thread: 1 (0 members and 1 guests) | |
|
|