Go Back   Pro/Forums > ProCooling Technical Discussions > General Liquid/Water Cooling Discussion
Password
Register FAQ Members List Calendar JavaChat Mark Forums Read

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.

Reply
Thread Tools
Unread 04-05-2002, 04:04 PM   #1
Roscal
Cooling Savant
 
Roscal's Avatar
 
Join Date: Apr 2002
Location: North of France
Posts: 198
Default Computational Fluid Dynamics for design

First, excuse my english..

I want to know if people here have experienced CFD for waterblock design? The next pictures show some studies I have realised to have a better look about quality of design..

This is a look of absolute speed for different mazes:







So with simulation we can find bad design, has anyone do some similar job or not here??? After temperature can be introduced to know temperature of water and waterbloc too...There is a lot of possibilities.


Last edited by Roscal; 04-05-2002 at 04:07 PM.
Roscal is offline   Reply With Quote
Unread 04-05-2002, 06:16 PM   #2
Brad
Thermophile
 
Brad's Avatar
 
Join Date: Oct 2001
Location: Nuu Zeeelin
Posts: 3,175
Default

whats interesting in the top on is that with all the channels, only the top one shows a lot of resistance, this is obviously due to much more water flowing through it. the next two show a bit as well, the middle ones have no flow going through them, and the final few have quite a bit of flow. this is exactly not what we want.

the second one is fairly even, which is good, although near the entry there seems to be a lot of resistance, presumable due to the turbulance of the water at this point.

the third one is much the same as the second one I think.

do you think it would be possible to do the same as for these designs I've attached? (central inlet, two outlets) Thanks in advance
Attached Images
File Type: gif wb3.gif (14.8 KB, 423 views)
__________________
2x P3 1100's at 1400, Abit VP6, 2x Corsair 256mb PC150 sticks, 20gb 'cuda ATA-III, 2x 40gb 'cuda ATA-IV in raid 0. 20" Trinitron. No fans

2x 2400+ at 2288mhz (16.0 x 143), Iwill MPX2, 2x Kingmax PC-3200 256mb sticks, 4x 20gb 60gxp in Raid 5 on a Promise SX6000. Asus Ti4200 320/630. Cooled by Water
Brad is offline   Reply With Quote
Unread 04-05-2002, 07:04 PM   #3
Fixittt
CNC Beyatch
 
Fixittt's Avatar
 
Join Date: Aug 2001
Location: Tulsa Spell it backwards
Posts: 721
Default

very interesting, I dont quite follow how you desicided on temp, and hot sopts, and so on. One one of your pics I see a hot spot by one of the inlets, so very far from the actual heat source. And what bease thickness are you thinking of, because the more base material, the more heat will go outwards.

or maybe I am just not understanding what you are tying to illistrate.
__________________
Creator of the Spir@l Block
Longest post ever
http://forums.procooling.com/vbb/showthread.php?s=&postid=43808#post43808
Fixittt is offline   Reply With Quote
Unread 04-05-2002, 07:06 PM   #4
Brad
Thermophile
 
Brad's Avatar
 
Join Date: Oct 2001
Location: Nuu Zeeelin
Posts: 3,175
Default

he is talking about resistance, not temperatures
__________________
2x P3 1100's at 1400, Abit VP6, 2x Corsair 256mb PC150 sticks, 20gb 'cuda ATA-III, 2x 40gb 'cuda ATA-IV in raid 0. 20" Trinitron. No fans

2x 2400+ at 2288mhz (16.0 x 143), Iwill MPX2, 2x Kingmax PC-3200 256mb sticks, 4x 20gb 60gxp in Raid 5 on a Promise SX6000. Asus Ti4200 320/630. Cooled by Water
Brad is offline   Reply With Quote
Unread 04-05-2002, 08:19 PM   #5
EMC2
Cooling Savant
 
EMC2's Avatar
 
Join Date: Feb 2002
Location: Dark Side of the Moon
Posts: 365
Default * I know nothing * :P

Roscal - your English is fine CFD is even better Which CFD program are you using?


Brad and Fixiit - those are pics of the water velocity. The top one is a good example of the type issue I was talking to DoTheDog about.

The scale shows what color represents what velocity (speed) the water at a particular point has. Darkest blue is 0 (no flow), red is equal or greater than the highest flow he has scaled.

You can see the effects turns have on the flow (usually highest on the "inside" of the turn, slowest on the outside).

You can also see fairly clearly the turbulence in the middle picture near the walls, as well as the boundary layer (0 or near 0 velocity) right up against the walls in a lot of areas.


Fixiit - the only temperature you assign when doing the thermal side of the simulation is the inlet water temperature and the boundary temperatures (you specify say 1/2" from the block surface the ambient temperature is 30C). The other inputs are usually the inlet/outlet pressures or the input flow rate and the heat flux and area it is applied to (the size of the CPU die and the amount of power you are wanting to model).

The CFD software then calculates everything else (if it's good software). It starts with the water velocity, which determines the heat transfer coefficient (how well the heat is conducted to and through the water). It then uses these values, the thermal resistance of the block itself, and the input power to calculate all the temps.
EMC2 is offline   Reply With Quote
Unread 04-05-2002, 11:17 PM   #6
jaydee
Put up or Shut Up
 
jaydee's Avatar
 
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
Default

If I had that software and new how to use it.....
jaydee is offline   Reply With Quote
Unread 04-06-2002, 04:29 AM   #7
Roscal
Cooling Savant
 
Roscal's Avatar
 
Join Date: Apr 2002
Location: North of France
Posts: 198
Default

Brad> For the moment this is only pressure, turbulence, speed that are computed. I do not still have to realize the coupling between temperature and other parameters because it's much more complicated . But with speed design we can we can have an idea of the losses of loads in the design because where the water speed is weak means many losses of load. The water go rather where it is the least in charge.

Pictures below show examples of turbulence in a maze2, speed and trajectory of a particle into the maze. Flow rate is always 10L/min.

Turbulence:


Speed:


Trajectory:



In the first post picture 2 and 3 are made about the same block design here :



So the red color show us that water passes more above fins near inlet and outlet because less load..


EMC2 > I see that you know a little about CFD. Thank you for the more precise explanations for the others . I have some problems with technical terms (english isn't easy to learn )
For the moment I use COSMOS Designstar with Flow module. But I work about new studies with the best software of CFD which exists : I talk about Fluent 6 but it's much more complicated than COSMOS . Have you already done some similar studies???

++
Roscal is offline   Reply With Quote
Unread 04-06-2002, 05:42 AM   #8
Brad
Thermophile
 
Brad's Avatar
 
Join Date: Oct 2001
Location: Nuu Zeeelin
Posts: 3,175
Default

one thing we should note from all these designs is that every single one has a blue boundary layer of very slow moving water. Thats something we need to fix, maybe a series of ‾‾‾‾‾\/‾‾‾‾ might fix it, or something similar
__________________
2x P3 1100's at 1400, Abit VP6, 2x Corsair 256mb PC150 sticks, 20gb 'cuda ATA-III, 2x 40gb 'cuda ATA-IV in raid 0. 20" Trinitron. No fans

2x 2400+ at 2288mhz (16.0 x 143), Iwill MPX2, 2x Kingmax PC-3200 256mb sticks, 4x 20gb 60gxp in Raid 5 on a Promise SX6000. Asus Ti4200 320/630. Cooled by Water
Brad is offline   Reply With Quote
Unread 04-06-2002, 06:05 AM   #9
Roscal
Cooling Savant
 
Roscal's Avatar
 
Join Date: Apr 2002
Location: North of France
Posts: 198
Default

Every fluid which gets through a channeling has any facons a very small speed near walls and even no speed on the wall because of the viscosity of the water. This is the limit coat (limit layer I don't know the exact term) phenomenon and it always appear in any situations.

In order to improve turbulence and break this limit coat the wall of maze should be lined(crossed off) perpendicularly in the direction of drainage of the stream to favor the appearance of mini-vortex which improve the thermic exchange wall/fluid..

So when I see nice waterblocks with walls very shiny and flat, it is not good for the thermic exchange, it could be improved by damaging the state of surface of walls. But I think people who buy waterblocks want the most beautiful block very clean and very shiny . You have probably see this technics of "raw state of surface" on some aircooling systems to favor turbulence....

Last edited by Roscal; 04-06-2002 at 06:13 AM.
Roscal is offline   Reply With Quote
Unread 04-06-2002, 08:35 AM   #10
Fixittt
CNC Beyatch
 
Fixittt's Avatar
 
Join Date: Aug 2001
Location: Tulsa Spell it backwards
Posts: 721
Default

Roscal,

would U mind doing a drawing of the Spir@l block, I would be very interested in seeing how that looks.
__________________
Creator of the Spir@l Block
Longest post ever
http://forums.procooling.com/vbb/showthread.php?s=&postid=43808#post43808
Fixittt is offline   Reply With Quote
Unread 04-06-2002, 09:38 AM   #11
jaydee
Put up or Shut Up
 
jaydee's Avatar
 
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
Default

Quote:
Originally posted by Roscal

So when I see nice waterblocks with walls very shiny and flat, it is not good for the thermic exchange, it could be improved by damaging the state of surface of walls.
I think I have found an endmill that not only will rought the walls but keep it looking good aswell. I had a theory about this same thing and you just proved it right. This end mill will go through and make mini channels in the walls of the main channels which may work.
jaydee is offline   Reply With Quote
Unread 04-06-2002, 10:13 AM   #12
EMC2
Cooling Savant
 
EMC2's Avatar
 
Join Date: Feb 2002
Location: Dark Side of the Moon
Posts: 365
Default CFD - what's that, Complex Flatulence Development :P

Roscal -
They say English is the hardest language to learn. But your English is waaay better than my French I use Flowtherm at work, but can't use it for waterblock analysis as my company would tend to frown on such endeavors Here at home, I'm limitted to doing things the old fashion way with equations and spreadsheets for fluid analysis as I couldn't afford commercial CFD software for personal use (also using my experience to look at general design problems, knowing what the flow will look like as a result of the block features)

BTW, the technical term for the "limit layer" in English is the boundary layer. Oh, and you get the "Understatement of the Year" award for your comment "the coupling between temperature and other parameters because it's much more complicated"

Regarding the perpendicular slots on the surface - in one project at work I found slots with an minor angle in the direction of flow worked better overall for that type method


Brad -
As Roscal described, you almost always have a boundary layer in fluid flow. However, you can greatly reduce its size by various techniques. The two major determining factors are the velocity of the fluid and the surface structure. In general higher velocity means a thinner boundary layer and more turbulent flow. The other way (surface structure and channel features) can get a bit more complicated, but as I've said before, there is a decent volume of research in that area.

A good common example of the effects of surface is a golf ball. The boundary layer and fluid flow characteristics are what help determine how far a golf ball travels through the air (air in this case is the fluid). That's why golf balls are dimpled - to help reduce the drag created by the boundary layer (although the goal in a golf ball is different than what you want in a wb - to use a similiar method in a wb you would want rounded protrusions rather than concave dimples).

In the case of a wb, we want increased turbulence near the surface which also causes an increase in the flow resistance. One example of a minor improvement to turbulence is Fixitt's sandblasting of the surface (note to Fixitt - shot peening would work better for that method).

You get to a trade off point in a wb though - if you increase it too much, your flow rate for a lower pressure pump starts to drop off, and you could end up with a net loss in the heat transfer coefficient do to the lower Reynold's number (which is an indicator of the turbulence level).

The surfacing on the floor of the newer Maze3 is an attempt to address this issue.

If you look at the middle pic of the Maze2, you'll notice that one of the thicker boundary layers it at one of the worst possible locations - the inner wall over the core area (just after the first 180 turn from the center inlet, the wall closest to the inlet). The other major not so great "weak spot" is the low velocity at the back wall of the inlet and the back half of the floor under the inlet. (the greatest temp delta is in the area just over the core, so that is where you want the best heat transfer).


Fixitt - if Roscal is so inclined, it would probably help him if you had a 3D model already done up in Autocad (you might have to export it as an IGES model - not sure how well Cosmos handles regular Autocad files) If you only have a 2D model of it, and Roscal is willing to look at it for you, I might be able to help (I could convert it to a 3D model with Rhino without too much effort - mainly depends upon what types of 3D model files his version of Cosmos can handle).


JD - have you ever seen a bit that you could create a knurled type surface on wb passages? Regarding the bit you found - an important piece of info : you do not want to create any surface channels that are parallel to the flow as this will contribute to a laminar flow along the walls.
EMC2 is offline   Reply With Quote
Unread 04-06-2002, 10:19 AM   #13
Psychotic
Cooling Neophyte
 
Psychotic's Avatar
 
Join Date: Mar 2002
Location: currently in hiding after a recent escape from the psychiatric ward of a local hospital
Posts: 69
Default

Wow, talk about an informative thread! I love reading this stuff and increasing my watercooling knowledge. Pro Forums keep getting better and better.
__________________
I don't suffer from insanity. I enjoy every minute of it.
Psychotic is offline   Reply With Quote
Unread 04-06-2002, 10:24 AM   #14
jaydee
Put up or Shut Up
 
jaydee's Avatar
 
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
Default Re: CFD - what's that, Complex Flatulence Development :P

Quote:
Originally posted by EMC2



JD - have you ever seen a bit that you could create a knurled type surface on wb passages? Regarding the bit you found - an important piece of info : you do not want to create any surface channels that are parallel to the flow as this will contribute to a laminar flow along the walls.
You are right and that was the bit I was thinking of as the channels would run in the same direction and that wouldn't help to much.

I also am going to try this one end mill that will make the botom of the channel groooved instead of flat. So instead of a 90 degree angle from the bottom of the channel to the side of the channel it will be U shapped. And then I was thinking about taking my 1/8" end mill and making the side walls bumpy. I will run the end mill into the side walls about every 1/4".
jaydee is offline   Reply With Quote
Unread 04-06-2002, 10:37 AM   #15
ondaedg
Cooling Savant
 
Join Date: Nov 2000
Posts: 130
Default

jaydee, I am not sure if what you said is what he meant, but I think I understand. He is saying that if you could find a bit that will score the side walls of the water passages vertically which is perpendicular to the waterflowing horizontally through them. If that is what you are saying, then disregard this =]
__________________
ondaedg@procooling.com
ondaedg is offline   Reply With Quote
Unread 04-06-2002, 10:41 AM   #16
jaydee
Put up or Shut Up
 
jaydee's Avatar
 
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
Default

Quote:
Originally posted by ondaedg
jaydee, I am not sure if what you said is what he meant, but I think I understand. He is saying that if you could find a bit that will score the side walls of the water passages vertically which is perpendicular to the waterflowing horizontally through them. If that is what you are saying, then disregard this =]
That is what I ment by running the 1/8" indmill into the sides of the channel wall about every 1/4". That would leave small vertical lumps in the channel walls about every 1/8" apart. I thinlk that would work pretty good. Wish my mill was operational as I would show you all.
jaydee is offline   Reply With Quote
Unread 04-06-2002, 10:54 AM   #17
EMC2
Cooling Savant
 
EMC2's Avatar
 
Join Date: Feb 2002
Location: Dark Side of the Moon
Posts: 365
Default

Here is a modified pic that shows the areas I was talking about in the Maze2:
Attached Images
File Type: jpg maze2m.jpg (41.7 KB, 336 views)
EMC2 is offline   Reply With Quote
Unread 04-06-2002, 10:58 AM   #18
Fixittt
CNC Beyatch
 
Fixittt's Avatar
 
Join Date: Aug 2001
Location: Tulsa Spell it backwards
Posts: 721
Default

jay the endmill U are talking about (for the U shaped bottom) is a ball nose endmill. I was not to thrilled by how these things work for what we want to du. My first Spir@ls were done this way, and it increases base thickness a bit to much I think.
__________________
Creator of the Spir@l Block
Longest post ever
http://forums.procooling.com/vbb/showthread.php?s=&postid=43808#post43808
Fixittt is offline   Reply With Quote
Unread 04-06-2002, 11:13 AM   #19
EMC2
Cooling Savant
 
EMC2's Avatar
 
Join Date: Feb 2002
Location: Dark Side of the Moon
Posts: 365
Default

JD - I tend to agree with Fixitt's assessment of using a ball nose. A fairly simple way you could increase the turbulence on the floor of the channel would be to put irregularly spaced small cresent like cuts in it with a regular end mill (not the ball nose) doing it something like this:
Attached Images
File Type: png anglecutfloor.png (20.0 KB, 324 views)
EMC2 is offline   Reply With Quote
Unread 04-06-2002, 11:18 AM   #20
TiTch
Cooling Neophyte
 
TiTch's Avatar
 
Join Date: Mar 2002
Location: London, UK.
Posts: 64
Default

Out of interest, will the type of material have an inpact upon the effect of the water "sticking" For example will the fashion for using acrylics actually negatively effect the operation of blocks due to there very smooth and shiny surfaces.
TiTch is offline   Reply With Quote
Unread 04-06-2002, 11:36 AM   #21
jaydee
Put up or Shut Up
 
jaydee's Avatar
 
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
Default

Quote:
Originally posted by EMC2
JD - I tend to agree with Fixitt's assessment of using a ball nose. A fairly simple way you could increase the turbulence on the floor of the channel would be to put irregularly spaced small cresent like cuts in it with a regular end mill (not the ball nose) doing it something like this:
Unfortunatly I am limited to verticle and horizontal only as I do not have the right equipment to do anything angled. As for the ball endmill I am not talking about one pass. I want a small andgle so the water dosn't have a dead spot at the 90degree angle. Instead I wnat to make it slightly sloped to where it is not a perfect U but more like I_I. That way the base thickness is still pretty much the same and it should keep the water going at that point.

I will try to make up a drawing of what I am saying.
jaydee is offline   Reply With Quote
Unread 04-06-2002, 11:37 AM   #22
Roscal
Cooling Savant
 
Roscal's Avatar
 
Join Date: Apr 2002
Location: North of France
Posts: 198
Default

EMC2 > In the modified pic you posted you have reason because this is the place where we need a very good thermic transfer but there is an "unsticking" here because the flow arrived in a corner with maximum speed.

I think spiral is the best way to avoid this bad situation because there is no 90° corner consequently no "unsticking".

If someone has a CAD file (.IGES please) with a spiral maze and only the volumic maze with inlet and outlet as my pics, he could mail me for a simulation of flow.Thx
Roscal is offline   Reply With Quote
Unread 04-06-2002, 11:43 AM   #23
EMC2
Cooling Savant
 
EMC2's Avatar
 
Join Date: Feb 2002
Location: Dark Side of the Moon
Posts: 365
Default

TiTich - the surface texture of the Acrylic top has no significant affect, since their is almost no heat flux through the Acrylic (very poor thermal conductivity). It also would have minimal impact on the flow on the bottom of the channel and the lower 2/3rds of the side walls.

That being said - a metal top is better from a thermal standpoint, amount depending on the thickness and height of the channel walls and how good of a thermal interface their is between the channel walls and the top { the top helps to conduct heat across the block and helps to conduct heat to the water }. However, if the walls are too thin or too tall, then you get to the point of diminishing returns since their wouldn't be very much thermal energy conducted to the top of the walls anyway.
EMC2 is offline   Reply With Quote
Unread 04-06-2002, 11:47 AM   #24
Fixittt
CNC Beyatch
 
Fixittt's Avatar
 
Join Date: Aug 2001
Location: Tulsa Spell it backwards
Posts: 721
Default

like in the Spir@l, I can pretty much garantee that there is no heat making it to the top of the block. So the clear top is not a problem or a hamper in preformance what so ever.
__________________
Creator of the Spir@l Block
Longest post ever
http://forums.procooling.com/vbb/showthread.php?s=&postid=43808#post43808
Fixittt is offline   Reply With Quote
Unread 04-06-2002, 11:57 AM   #25
jaydee
Put up or Shut Up
 
jaydee's Avatar
 
Join Date: Dec 2001
Location: Spokane WA
Posts: 6,506
Default

Quote:
Originally posted by Fixittt
like in the Spir@l, I can pretty much garantee that there is no heat making it to the top of the block. So the clear top is not a problem or a hamper in preformance what so ever.
The tops of my blocks don't seem to get hot either. I don't think there is much to worry about there. Once I get time i am going to make a couple of the socket A blocks Brad posted above. I alreadyh started on it, but I need my endmill holder so I can run the 3/8" endmill needed for the channel.



I am going to make a series of different tops. 1 single in/out metal top. One singlee in, dual out metal top and two of the same in acrylic. I will test them all and see what happens. Then I will add the turbulance tothe channel walls and see if that helps any.
jaydee is offline   Reply With Quote
Reply


Currently Active Users Viewing This Thread: 1 (0 members and 1 guests)
 
Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Forum Jump


All times are GMT -5. The time now is 05:09 AM.


Powered by vBulletin® Version 3.7.4
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
(C) 2005 ProCooling.com
If we in some way offend you, insult you or your people, screw your mom, beat up your dad, or poop on your porch... we're sorry... we were probably really drunk...
Oh and dont steal our content bitches! Don't give us a reason to pee in your open car window this summer...