My multi spiral concept.

[nicozeg]

Warning, the following are pure divagations that were floating in my mind! There's no real block to show.

Since cathar revolutionized Block design some months ago it seems difficult to think of a better performing one with a much different concept. That parallel mini fins are hard to beat as an overall concept in simplicity and efficiency. Maybe there’s still some tweaking available in the design to improve performance using different construction techniques, as morphling is doing now with his single outlet version. But my goal is to find another approach.

As I like to swim against the flow I still believe that a radial flow design can be better, or at least as good. But a radial pattern has some problems: it’s impossible to keep a constant flow speed at the same time that the fin/channel size relation. The only way of achieving both goals with a circular expansion is with spirals. As geometric challenges are my specialty I designed a multi spiral pattern that keep both channel and fin width at 1mm. Maybe this is going to be very difficult to mill, but at this stage my only interest is designing an ideal shape, without considering manufacturing costs.

[Neomoses]

that looks pretty cool. and it shouldn't be terribly difficult to mill; it needs to be done on a CNC anyway. It may take some time, however.

[SysCrusher]

I think that's exactly what you want. Keep the pin in the center and that will help guide the flow at the same time reducing the the stagnation point. Fit a barb or nozzle even and around the pin. It should perform great.

[jaydee]

Humm... If this conversion is correct "1 Millimeters = 0.03937 Inches" then this will probably not be possible to mill with a CNC mill in a standard shop. Even the $500,000 mills can have .03" of slop. That would be more than enough to collapse the 1mm fin. I am not saying it is impossible though. A shop like this http://www.mmsonline.com/articles/hsmdm/case4a.html maybe able to do it.

But with that aside I like your thoughs there. Nice drawing aswell. That has to be SolidWorks? So would you have a nozzel directing water at that center pin and then hope it goes into all the 1mm channels? And where does it go after that? How is the output setup?

I think there shold be a lot of turbulance right over the core with the inlet right over it. I am not so certain that channels of anytype are going to gain anything after the water hits that center spot on the core. The water hitting that center "should" cool the metal down enough to prevent the heat from spreading throughout the rest of the base providing the base thickness is optimal. So anything beyond the area of the CPU core is really un-needed and may be a waste of time and money on development.

But from the looks of your drawing those fins and channels may not be much bigger than the area over the core? hard to tell without a full view.

Keep it up!

[SysCrusher]

Originally posted by jaydee116
Humm... If this conversion is correct "1 Millimeters = 0.03937 Inches" then this will probably not be possible to mill with a CNC mill in a standard shop. Even the $500,000 mills can have .03" of slop. That would be more than enough to collapse the 1mm fin. I am not saying it is impossible though. A shop like this http://www.mmsonline.com/articles/hsmdm/case4a.html maybe able to do it.

That's what sucks. I like to try some cool ideas with sharp edges and corners and 1 mm walls but would be impossible to do with cnc or to expensive to even bother. Laser maybe?

[jaydee]

Originally posted by SysCrusher
Laser maybe?

I am not sure. It would take one hell of a laser to cut into copper. Our $50,000 laser at work can't even make a mark in any type of metal and is only good to 3/8" acrylic. I know there are lasers that can cut metal as www.designcomp.com claims to use one to cut the aluminum cases and fan grills with. But Copper is very dense and would be (I think) much much harder to laser than aluminum. But yes if there is a laser capable of this it would be a way to go and may not be much more expensive than milling.

I am not to familiar with how wire EDM works but that would probably be a better solution if it is capable of something like this. But still that is very expensive.

[Roscal]

Nice design, a lot of surface but a lot of pressure drop with a lot of small channel (friction at wall like capillaries)

EDM is great to do this but cost is $$$$ (graphite electrode to mill) and very very slow but high precision at final .

You can mill that with a 1mm milling bit on a high speed CNC machine, we will use some 0.5mm milling bit on a 5 axis machine to realize our mould foundry (cost-> 0$ ) to do our new bloc but milling firm which can do that are few because methods are expensive

[hara]

That could be cut using a waterjet? It looks like what my friend wanted to cut, squirrel.

http://forums.procooling.com/vbb/sho...&threadid=5296

[Roscal]

Not waterjet because jet must pass through the plate...

You can't stop the jet at the depth you want

[nicozeg]

My plan is getting more ambitious, so at this stage I’m just dreaming.

As the shape is going to be very complex, it could be milled an original in a soft material like plaster, then cast cheap copies. Before someone says that’s impossible with copper, I agree. I’m thinking in silver. Jewelers cast all the time complex minute details with simple equipment.

Jaydee, maybe you missed a decimal in your slope calculations, it should be more like .003

The water goes to a perimetral channel of a growing section and exit in only one barb, just like good centrifugal pumps. All the channels and housing of the block will be in casted resin, so the only metal required is the silver disk. Actually the disk is 28mm diameter and 2.3cc volume. As silver density is 10.5 I will need about 25gr of silver. That shouldn’t be expensive, anyone knows?

This is using just the crappy modeling of Autocad 2k2

[nicozeg]

Bah, missed the attachment.

Roscal, thats what I want to do, How are you goung to cast it?
U could post more details of your design, with that nice simulations.

[nicozeg]

Damn!, it don't want to go!

Third try:

[nicozeg]

Originally posted by hara
It looks like what my friend wanted to cut, squirrel.

I know, I had this spiral design before that, only in 2d drawing, thats why I sugested him that the spiral geometry could be tweaked a lot.

[g.l.amour]

i do'nt see the point letting it spiral outside the cpu die area. wouldn't that just be obstructing flow without a reason? it could be that u r thinking pelts though.

i think bigben's radius would be a little little bit more effective


beautiful design anyways.

[Roscal]

Nicozeg> No details of the evolution of our bloc for the moment because it's secret (not yet calculations all done, we change design every day with better ideas to reduce all costs and improve performance )

Our proto is like that, but the final version is completly different (using same concept) :


Milled at 1mm bit and EDM used to finish it (cost ->0$ too. Thx to my friend ). We have done flow rate test only for the moment and we have a very high speed for water and a very low pressure drop (what we want!).

PS : Our mould is a metallic one with copper injected under high pressure to have the block (best method to have tiny things in the bloc, 0.5mm lenght for elements authorized, fins are very thin and tall if we want...) and we hope to do about 100-200 blocks (20-30 pre-command at this time) , mould is almost "unbreakeable" so no problem for us, we can do blocks when we want !!

[nicozeg]

G.l.: Yes, the idea of the spiral is scalability, being able to use the same design over a large or small area. I can easily modify the geometry so after some radius flow speed is reduced. I dont think it's so restrictive, currently it has 12 channels with 26.5 mm lenght.

Roscal:

I'm dont know how EDM work, but is impressive! Really nice machines you have available.

It can't be so secret your development if theres already 30 orders! Where's that? in french forums?

[Roscal]

When I say 30 orders I mean 30 persons are very interesting to buy one (proto for the moment but final version will be ), they know me very well and they have confidence in me and my friend . And yes on french forums but no pics of evolution available

EDM is a milling method without contact between metal and tool (electrode graphite). All is in a tank with fluid and we put a lot of current (electricity) between the 2 electrodes (bloc and graphite) and under sparks on whole bloc metal is vaporized in the fluid and we have a small hole. You can mill all metals, even the strongest on earth very easily with EDM and precision is great (0.02mm during completion)

EDM is used to do sharp corner (impossible with a CNC) and complex geometry

[nicozeg]

It sounds like a very expensive process.

I was thinking also in rapid prototyping using 3d printers, That can't be very expensive in small parts like this.

[jaydee]

Originally posted by nicozeg


Jaydee, maybe you missed a decimal in your slope calculations, it should be more like .003

I used an online converter so I don't think it is off. http://www.onlineconversion.com/length.htm

Double checked. Copy and pasted: "1 Millimeters = 0.03937 Inches".

[nicozeg]

Originally posted by jaydee116
I used an online converter so I don't think it is off. http://www.onlineconversion.com/length.htm

Double checked. Copy and pasted: "1 Millimeters = 0.03937 Inches".

I know the conversion is ok. I mean .03 inches is too much slope, that's about 1/32, it should be more like .003"

[jaydee]

Originally posted by nicozeg
I know the conversion is ok. I mean .03 inches is too much slope, that's about 1/32, it should be more like .003"

Oh, my bad. But still I belive it is .03. Especially on a mill that has any amount of hours on it. The more it is used the more slop it gets. I well kept up mill shouldn't loose to much but still it wears. My mill is about 1/64th right now and going fast. Thats why I stopped using copper. The software is set to compensate. But that is mainly for strait cuts. Curved cuts is a whole different set of problems as it is putting pressure in more than one direction. it wouldn't be so bad if you were not trying to cut fins and channels in 1mm encraments, but trying to keep that many channels that precise is not an easy challenge. I know I couldn't do on my mill even if I used aluminum. Well a 1mm endmill needs to spin around 16,000 RPM anyway which mine only tops out at 3,000rpm. This will take expensive mills that are very tight which are only going to come from very expensive shops.

Also if you go to a machine shop make sure you have a drawing that they can use to make the G code with or they will charge you around $500-$2000 for a drawing fee. And they will want to make at least 25-50+ at a time or they will not do it unless you got the bucks to cover their time which is usually $50-$200 an hour. It will take a good hour or two to get it setup just to run it. And of course they are going to add in a few hours of prep and clean up. So if you just want one done expect around $800+ if you have a usable drawing, unless of course you know someone in a good machine shop that is willing to try and do it for you.

You maybe lucky and find a shop that is low on orders and cut you a hell of a deal though.

[bigben2k]

Originally posted by g.l.amour
i do'nt see the point letting it spiral outside the cpu die area. wouldn't that just be obstructing flow without a reason? it could be that u r thinking pelts though.

i think bigben's radius would be a little little bit more effective


beautiful design anyways.

Very eye-catching indeed.

Spiraling may not be completely useless. If you look at the flow graphs for Radius, you'll see a fast flow at the bottom of the channel; in a spiraling design, this flow might rotate along the bottom of the channel.

If it does rotate, it will have to flow extremely fast in order for it to make a performance difference though. I don't believe that there's a pump out there that can do this (within reason). The channel flow must reach Reynolds 2'000 to 4'000 (turbulent flow), at which point the rotating effect might be dissipated...dunno.

[jaydee]

Originally posted by bigben2k
The channel flow must reach Reynolds 2'000 to 4'000 (turbulent flow), at which point the rotating effect might be dissipated...dunno.

And will most likely add more heat than the extra performance that might be gained by the block will remove. It's a tough battle.

[MadDogMe]

That is the best concept I've seen for a while nico!...

[nicozeg]

Originally posted by jaydee116

Also if you go to a machine shop make sure you have a drawing that they can use to make the G code with or they will charge you around $500-$2000 for a drawing fee. And they will want to make at least 25-50+ at a time or they will not do it unless you got the bucks to cover their time which is usually $50-$200 an hour. It will take a good hour or two to get it setup just to run it. And of course they are going to add in a few hours of prep and clean up. So if you just want one done expect around $800+ if you have a usable drawing, unless of course you know someone in a good machine shop that is willing to try and do it for you.

You maybe lucky and find a shop that is low on orders and cut you a hell of a deal though.

Nahh, forget about milling, that's old age technology

I uploaded an STL version of the drawing to www.xpress3d.com for an instant online quote. It cost U$45, and you get $25 discount on the first order. That's delivered in two days anywhere in the US; only problem is I'm in south america , so shipping can be more expensive than the part.

Also those models are based on a hardened plaster powder so have grainy texture that need to be polished away. But I've done more complicated things than that..

MDM: Thanks..