Another WB Idea! Comments?

[DoTheDog]

I have recently found that I may have access to a CNC Mill. All I needed was a WB design to try out. After thinking about it, I came up with this.



Pardon the lame skillz:shrug: But I think you get the idea. There are eight 1/4" deep 1/4" wide channels flowing from the center. The outer channel gets wider as more water enters it. The outlet is 1/2". Do you think this would be worth calling in a favor to get milled?

Any and all comments/flames appreciated.

[redleader]

Most of the flow will go through 2 or 3 channels and the others will fill with stationary (warm) water. Add some more flow restrictions to the channels nearest the exiting barb.

[DoTheDog]

Thanks for the response redleader. I was thinking that if the water was moving in one direction around the outside that it would pull water through all the channels.

What kind of flow restriction were you thinking about? The way it is now, it should have very little flow restriction in that the 1/4" x 1/4" channels are about the same volume as a 1/2" ID tube, so there is 8x (8 channels) the volume coming out of the center as there is coming in and there are no 90 degree turns. Could you explain? I am a newbie at designing these things. Any help appreciated.

[DoTheDog]

Another question. Does the die hit the WB dead center or is it off-set a little. Anyone know where the die hits relative to the mounting holes?

[Butcher]

It's very slightly off centre.

[EMC2]

It's offset 45mils to one side. You can get a spec with the exact dimensions from AMD's sight.

Regarding the issue of even water flow - remember that the channels you have are rectangular and the hose is circular (hint - look at the formulas for the area of a circle versus a rectangle).
What Redleader was referring to regarding flow restriction is that water takes the path of least resistance. If the amount of flow through the 1/2" input can pass through say half your channels, then the other channels will have almost no, if any, flow.

To prevent this you need to balance the total flow of the not just the channels in comparison to the inlet, but balance the flow restriction seen between all 8 paths. The later not only involves the size of the channel, but also the length of the channel and the turns in them.

[Brad]

yeah, make the channels closer to the exit much narrower than the ones that have to go all the way around. really cool design by the way

[DoTheDog]

Thanks for the replies guys.

EMC, I did do the calculation and it looks like a 1/4" x 1/4" square has about the same area as a 1/2" circle. Is the idea to get the total of the channels = 1/2" diameter tube? If that is the case, I would have some either very narrow channels (e.g. 1/8" or less) or very shallow ones. Is there any way to calculate that or do you just need to make some educated guesses?

yeah, make the channels closer to the exit much narrower than the ones that have to go all the

Brad,would it help if I turned it over so the channels closest to the outlet were going up hill?

really cool design by the way

Thanks, I was thinking Nautilus shell.

[Marco]

You'll have to get fixittt's professional word on this, but I think that's going to be a bitch to machine. Might be wrong.

You might want to try progressively smaller paths. As in the radial path that terminates furthest from the outlet is the widest, and that nearest the narrowest. Might look cool too.

[DoTheDog]

Marco,
Thanks for the reply. The CNC machine I may be able to use is an industrial sized one. I will talk to the operator and see if the mill can handle this design though.

The decreasing channel size idea is a good one but I guess I need some help in understanding how to calculate how much smaller each channel has to be from the previous one. Anyone know how to do that?

DoTheDog

[DigitalChaos]

Originally posted by DoTheDog
Is there any way to calculate that or do you just need to make some educated guesses?

you want to find the ID of the inlet, not the hose.

then do this
R= radius (1/2 of the ID)

area of inlet = 3.14 x R x R

then take the area and divide it by the # of channels you are going to use. the resulting # is the area that each channel should be.

if you want perfectly square channels take the answer from above and find its square root. otherwise just estimate with rectangle sizes to get the same area.

[DigitalChaos]

just a thought though... even if all of your channels DO equal the inlet area.... wouldnt velocity of the water change things?

like the water may move faster on one side... and slower on the other.

something you may want to think about.... maybe make channels smaller as they get closer to the inlet, and bigger as the get farther away.

[Marco]

Originally posted by DoTheDog
The decreasing channel size idea is a good one but I guess I need some help in understanding how to calculate how much smaller each channel has to be from the previous one. Anyone know how to do that?

I was wondering about that myself. You probably need to look at the relation between the surface tension at the copper/water interface and the length of the channel in relation to that. I wish I knew something more about fluid flow mechanics but I don't. I bet it's not a very difficult problem though, just a question of balancing the overall resistance to flow for each channel to be the same.

[DoTheDog]

Wow, this is getting complicated!

I wish I knew something more about fluid flow

Marco, I hear ya. I am not an engineer so this is starting to get a little above my head.

you want to find the ID of the inlet, not the hose.

then do this R= radius (1/2 of the ID) area of inlet = 3.14 x R x R then take the area and divide it by the # of channels you are going to use. the resulting # is the area that each channel should be.

DigitalChaos, I did the calculation for the channel cross section area. The inlet is 1/2" or 6.35mm^2 x PI is 126.7mm. You divide that by 8 channels and it is about 15.8mm channels. That is TINY! 5/32"^2 or thereabouts. So I figured if I doubled the area of the output area to 253mm then I could run 8 channels 3/16" wide by 1/4" deep.

I hear ya on the differential speed going through the channels but would that make a big difference if it was all moving?

[DigitalChaos]

Originally posted by DoTheDog

I hear ya on the differential speed going through the channels but would that make a big difference if it was all moving?

if a channel or 2 are moving fast enough, it could keep water from going to the other channels.

ideally you want each channel to move at the same speed that the inlet is moving, its just that the inlet is split up into 8 channels.

now lets say that 2 of the channels are moving 2x as fast, they would be taking the flow from 2 other channels. now this reduction should be reduced over all of the channels, but still. You dont have an even flow, and possible dead/slow heat spots.

you would have to "tune" the block, or find some way to calculate it. i have a feeling your calculations will have to deal with water flow rate too.




here is another idea, get rid of the outlet. and make the entire design somewhat 3D. You would want something perfectly symetrical. perhaps the chanels go out like you have them, then go up into a "second story" chamber for the outlet (using 2 blocks and 1 plate would probably be needed). you would only have to deal with the inlet blocking a bit of flow, and somehow design around that.

[jaydee]

Nice, and the machining would not be that hard actually. Making to code would be though.

[jaydee]

also another thing to consider is putting a divider across the middle instead of that center post. divide the block in half so equel parts of water go to the top 3 channels and equel amounts go to the bottom three. Not sure if that is scientifically sound, but something to think about.

[EMC2]

There are standard equations for rectangular vents for calculating the flow resistance they present. There are also equations for calculating the effect of the turns in the channels (and yes, those turns will have an effect).

Regarding making the channels significantly larger than 1/8th the opening port size, you run into 2 problems, both significant. First, because of the very large differences in channel length and turns, you would indeed end up with very low flow in the last 4 channels (farthest from the outlet). More than 80% of the flow would go through the first 4 channels if they are double sized.

The second problem is that you will be slowing the velocity of the water down to less than 50% of what it is at the inlet. You would end up with a large boundary layer of water against the block surfaces and poor heat flow into the water unless you used a high pressure, high flow pump. (you would still be losing velocity, but you could get enough to at least keep semi-decent thermal flux going)

Digital's description of an alternate way to solve a lot of those problems, with small radial channels and a "second tier output combiner" is probably an easier solution. There's a thread in here somewhere that he and I were discussing just such a concept that has a drawing I posted up that will give you an idea of what he means.

DC, you remember what thread that was in bro?

[DigitalChaos]

yea i remember that.. think i started it actually =) lemme look around.

[DigitalChaos]

here its: http://forums.procooling.com/vbb/sho...&threadid=2358

the designs EMC2 was showing was based off the loscircuits.com link.


just skip ahead and take a look at the final product of what we were building off:
http://www.lostcircuits.com/advice/radeonmods/5.shtml



now these are just Fins instead of channels... something else you might want to think about.

now im all inspired again, gonna bore out a copper block, add those super easy fins, then cap it up in a second chamber somehow. if i actually do this ill post some pics =) ive currently been melting my hands on copper water block while torching them =) so this should be pretty easy

[DoTheDog]

Thanks for the input guys.

First, because of the very large differences in channel length and turns, you would indeed end up with very low flow in the last 4 channels (farthest from the outlet). More than 80% of the flow would go through the first 4 channels if they are double sized.

I realized that the velocity would decrease if the channels out were greater in volume than the input. I was hoping that it wouldn't be too drastic and that it wouldn't hinder performance too much. Whaddya think? Would it be that problematic?

As for most of the flow going through the 4 channels nearest the outlet, I was counting on the water moving around the outside channel acting like a syphon and pulling water through the rest of the channels. Am I being simple minded here?

Also, I saw the 3d model that EMC posted in that thread. It looks very cool but I don't know if I could pull that one off. I want to keep it a simple construction.

Jaydee, thanks for the ideas. Glad to know that it wouldn't be too hard to mill. The drawing is still in no shape to give to a CadCam guy to code for me yet so I have yet to show it to him.

I have actually tried to draw it again with different channel sizes (largest furthest from the outlet getting gradually smaller as they approach the outlet. It looks pretty cool. but I left it on a comp at work and forgot to FTP it up so I could post it. I will add it tomorrow.

Any help is appreciated. This is my first block design and I want it to not only look nice but also perform.

[Brad]

how about having a central inlet, with a southwest exit, like shown, and a northeast exit, with two sets of channels. it's probably spread out the water flow a bit more

[EMC2]

After doing some calcs, it's actually worse than I thought. Based on even sized channels for the "arms" radiating from the center and a smooth increase in size from start to end for the outside channel, here are the relative flow numbers in percentages, with Arm1 being closest to the outlet:

Arm1 51%
Arm2 27%
Arm3 13%
Arm4 6%
Arm5 2%
Arm6 1%
Arm7 & 8 < 1%

Couple of other pieces of info:

To get anywhere close to balanced you have to do some major changes.

If you have a 1/2" center "hole" for the inlet, the maximum width of the arm opening at that point is 0.196" without overlap. A 1/2" hole has a perimeter of 1.5708". Divide that by 8

If you make this in a 2"x2" block area, between the 1/2" inlet area and the 1/2" outside channel at it's widest point, you don't have a lot of room left for the arms. Your best bet will be to use the depth of the channels to get the flow up.

Last thing - the pic I did was to go on top of a block. If you didn't already, look at the link to the original sight that Digital referenced at earlier in that thread. In your case, all the top would be is a rectangular "box" with 8 holes drilled in the bottom to feed the output of the arms up and 1 in the center for the inlet pipe to go through, then the "top" has 2 holes, one for the inlet, one for the outlet. Not really a hard piece to make at all. Just a box with holes in it After looking at the ratios needed for balance based on your original block design, this alternate way would actually end up being much easier.

[DoTheDog]

Thanks for the input guys.

how about having a central inlet, with a southwest exit, like shown, and a northeast exit, with two sets of channels. it's probably spread out the water flow a bit more

Brad, you really like the 2 outlet idea I take it (I've seen your post with your 2 outlet designs). Would you put 1/2" in and 2 x 3/8 out? If you did that would it make sense to take each outlet to a different rad? Veeeery Interesting...

After doing some calcs, it's actually worse than I thought. Based on even sized channels for the "arms" radiating from the center and a smooth increase in size from start to end for the outside channel, here are the relative flow numbers in percentages, with Arm1 being closest to the outlet:

EMC,
Thanks for running the calc for me. How did you do it? Is it really complex? Given your calculations, what if I went to a 1/4" thick block with 1/8" channels? Would that be too thin? That way I could keep the volume of the channels much lower. Another idea would be to reduce the number of channels from 8 to 6. Also, would doing what redleader suggested going to larger channels farther from the outlet and making them progressively smaller as the channels approached the outlet make a difference? I.e. Channel 1 would be 3mm and channel 8 would be say 5mm wide?

Last thing - the pic I did was to go on top of a block. If you didn't already, look at the link to the original sight that Digital referenced at earlier in that thread. In your case, all the top would be is a rectangular "box" with 8 holes drilled in the bottom to feed the output of the arms up and 1 in the center for the inlet pipe to go through, then the "top" has 2 holes, one for the inlet, one for the outlet. Not really a hard piece to make at all. Just a box with holes in it After looking at the ratios needed for balance based on your original block design, this alternate way would actually end up being much easier.

The reason I thought it would be harder is you would end up having to mill 3 pieces of copper. One for the top with the inlet and outlet, one for the "box" with 8 holes in it, and 1 for the channels. I guess milling them wouldn't be that big of a deal but getting them all put together again would be. My soldering skillz are worse than my Illustrator skillz!!! BTW, if I did go this route how thick would the "box" have to be to allow optimal flow through to the outlet?

I appreciate you guys helping me think this through. This has been a great learning experience. Hopefully by the end of this I will have a great performing custom WB!!!

[DigitalChaos]

if you went the alternate route... hrmm im not sure on the OPTIMAL channel size. basically you would want to get something that is not too restrictive on flow, but you wouldnt want the channels too deep. you want the water to be closer to the die, so i guess find the thinest channels that offer the least flow restriction.
perhaps you could make the channels wider so you would basically be making FINS in there.??