What about internal pics of the storm?

[jaydee]

Originally Posted by Cathar

You're on the right track jaydee. In jet impingement on a flat surface, what's occurring at the absolute dead center of where the jet strikes the base?

BTW - I no longer really consider the Cascade to be a true impingement design. It's really more of a super-turbulent mash design, along the lines of its origin, being that of a faucet at full blast inside the mouth of a tall deep cup.

The problems I faced with jet impingement (JI) after a lot of experimentation was that the Cascade-style mashing worked out best, mostly due to the sheer density of jets/cups that is achievable with the Cascade approach. Making the cups wider and giving enough room to allow for true impingement made things worse, and peak overclock and temperatures dropped away as the convectional gradient is less consistent across the base of the cups.

What you're seeing in the picture is an attempt to solve the inherent drawbacks to jet impingement across a large flat surface.

Cool, I was hoping this sinus infection med's were not making me delirious.

[bigben2k]

Yep, it's a stagnation point. A pin there could add a tiny bit of performance, and since it's a really sensitive area, it just might be measurable.

Nice work, and nicely integrated too!

[threeputt]

What about the middle plate? We've seen the top and bottom. Show us the middle plate as well, unless it's just a spacer to route the water...

Oh, and what about the G-levels? I think this one is G4? What changes as G-level changes? The pin size?

[pHaestus]

He did show you the middle plate (with the jets)

[threeputt]

Oh... that is the middle plate... For some reason I kept thinking that was the top plate... oops...

[pHaestus]

Top plate is just the threads for the barbs

[threeputt]

Yeah... I realized it as soon as I look at the picture on the first page again... need more coffee ^-^

[Kobuchi]

Originally Posted by Cathar

In jet impingement on a flat surface, what's occurring at the absolute dead center of where the jet strikes the base?

A: stagnation

if the jets must approach the flat surface at a perfect right angle.

With this jet arrangement the stalagmites are a nice touch.

[Brians256]

Dang, but it's nice to see discussion of the block theory in this thread instead of flames.

Cathar, props for working to push the boundary of waterblock design. The more I learn, the more ignorant I become.

[SysCrusher]

Originally Posted by Cathar

OK, since you asked so nicely I'll let you chew on this for a bit.

If I'm seeing correct, I'm impressed with that. I was wondering how you would attack that problem. Custom made bit must of done the trick. I came close with a ball endmill but not entirely perfect. I couldn't imagine doing 52 tight circles with a 1/16" ball mill. The machine time would be to costly.

[murray13]

Can't remember when or where it was discussed, but I remember talking about optimizing the cups a while ago. I even posted the pic below. Can't find the thread though. Cathar said he was already thinking in that direction.

[snowwie]

i remember going through a tour of cape canaveral in FL at the space center.

they were talking about how the space shuttle's engines are so powerful that they had to build something like that to divert the jets from the engines ******d, otherwise the engines' jets would bounce back up (stagnate?) and destroy themselves. thought it was interesting at the time....relation to current topic? hell, i don't know.

edit: why is "o u t w a r" censored?

[JFettig]

ok now we got somewhere with this design, if I remember right I have seen drills like that but never that small. either that or its just milled like that.

Those are pins inside the cups right? I toyed with the idea long enough to try drawing it and realize its more work than I felt like putting forth copying someone elses work

Looks good.

Jon

[Cathar]

Top plate also houses the jet intake plenum chamber, which is just a circular recess in the plate.

Basically I had spent some time studying this basic picture and thinking about JI:



Jet diameter/nozzle separation (z/d) seems to create peak Nusselt numbers at a ratio of 4 < z/d < 6 for a number of studies, due to the shearing of the jet against the surrounding liquid. The jet has a "potential core" which is roughly cone shaped and is also typically around 4-6d in length depending on the initial jet velocity from my understanding of the theory. The problem with the shearing is that it also saps jet power/velocity fairly rapidly.

Also, immediately under the jet the radial velocity of the water is zero. I was looking for a way to get the way moving radially somewhat before it even struck the base-plate, much akin to the angled impingement models some have attempted where it was theorised that a 10° angle was potentially of more benefit (can't find the link to it now). I was also looking for a way to disrupt the potential core of the jet before it reached the base-plate, and also use that disruptor as additional surface area. As such, the disruptor should not be smooth edged as we want to promote turbulence.

I was also partially inspired by this image:



However I wasn't too sure about the apparant dimensions of the pedestal used in their simulation. I tried doing just that, but it didn't make any real difference for me, and my suspicion was that the pedestal was too large. In that image my reasoning was that it was not really turbulating the core, and the base of the pedestal, as seen in the simulated thermal graph, was actually getting too hot. Basically the results I saw with that approach was about the same as without.

So I tried for even smaller, where it's no longer really a pedestal that the jet bounces off, but more of a short pin that juts up into the core of the impingement jet, attempting to disrupt it, but also not to deflect the flow so much that bulk of the jet flow is pushed away significantly. It was also apparant to me that in that image that there was a fair amount of potential for entrainment (recirculation of already heated water) around the base of the pedestal.

So basically I experimentally hunted for the better ratios of nozzle separation, pin width and pin height, along with an optimal range of cup diameters across a range of pumping scenarios. This is where it all got rather muddy trying to pick a balance.

The block's jet restriction and jet intake pressure drop was also heavily experimented with to provide what I thought was the best tradeoff of pressure drop for jet velocity over the target cooling patch area with pumps from an Eheim 1046 up to my Iwaki MD30-RZ, and while still keeping volumetric flow rates acceptable. The block's pressure drop, while very high, makes as much "use" of that pressure drop as possible to get the jet velocity typically within 85-95% of what the pump can sustain. Trying to get the jet velocity up any higher just results in absolutely killing volumetric flow rates for very minimal jet velocity gains, so overall I was pretty happy with the balance point I found here. How restrictive is it? Not overly so (IMO). Typically you'll see flow rates of around 75% of that which you'd see when using a White Water, and around 85% of what you'd see when using a Cascade.

Really I struggled with finding much of any existing theoretical/research material immediately relating to what I was attempting, and wholly admit to mostly flying blind with just experimental results to indicate if I was heading in the right direction with my thoughts.

Some References I used which were of some help:

http://widget.ecn.purdue.edu/~jmurthy/me605/wu.pdf

http://www.cape.canterbury.ac.nz/web...ata/839rev.pdf

http://elecpress.monash.edu.au/ijfd/...ijfd_v7_a1.pdf

http://www.electronics-cooling.com/h...01_may_a2.html

http://www.pressure-drop.com/

http://www.coolingzone.com/Content/D...as/fcalc10.htm

[SysCrusher]

Have you tried this one Cathar for a pressure drop calc?

http://www.winsite.com/bin/Info?500000006272

[Cathar]

Originally Posted by SysCrusher

Have you tried this one Cathar for a pressure drop calc?

http://www.winsite.com/bin/Info?500000006272

Hmmm - have you tried the one that I linked to? Quite a deal more powerful.

[SysCrusher]

I took a quick look over, nothing in detail yet.

Edit: I like it much better now that I got into it.

[miladiou]

Thanks Cathar, what an explaination, i didn't dare to hope so much.
Congratulations for your work.

[peepingdan]

Originally Posted by jaydee116

The knowledge I have gained by unregistered and Cathar over the last 3 years is simply priceless.

Amen to that. Everything I have learned from CNC machining I have learned here. It has allowed me to make the base of my waterblock that I wish I had the tools to finish. Just reading the end of this thread alone has given me so much.

[Incoherent]

Originally Posted by Cathar

So basically I experimentally hunted for the better ratios of nozzle separation, pin width and pin height, along with an optimal range of cup diameters across a range of pumping scenarios. This is where it all got rather muddy trying to pick a balance.

Assuming/expecting tests do show an improvement, what has given you the most gain do you think? My calculated guess would have to be the flow rate vs water velocity optimisation. There is a peaking curve in that.
I am curious about your new jet vs cup diameter ratio with regards to the jet impingment / mashing transition. Cup walls make the scenario lean heavily towards mashing. They are wider, are your cups shallower? Is the 4-6d z really optimum with the increased shear from exiting rather than sitting water ... these are somewhat rhetorical Friday night questions... better go to bed...

[Etacovda]

He mentioned 'tall, deep cups' in regard to the cascade earlier, so you'd have to imagine these are somewhat shallower.

[Kobuchi]

Shallow they are. I'm squinting, trying to decide if this could be better photo-etched, like a PCB. Here is an example of photo etched gears for a 1/12th scale model motorcycle. These were etched twice. Note the 1cm staples for reference. Not to slight the equally fine work Cathar had done by milling.

Etching may be cheaper than CNC time. I believe high school students still make their circuitboard projects by etching, not CNC.

If the pins must be proud of the base plate then nevermind.

[Les]

Cathar
Do you have a link to the image?


Or a reference to the work? Please.

[Cathar]

Originally Posted by Incoherent

Assuming/expecting tests do show an improvement, what has given you the most gain do you think? My calculated guess would have to be the flow rate vs water velocity optimisation. There is a peaking curve in that.

Yes, and no. The flow/velocity optimisation improved things, true, but they were all "lost" again by moving to a truer impingement effect. I have disruptor-less bases here and they perform about the same as the Cascade does, but a whole lot more restrictive. Therefore this just re-affirmed to me that the Cascade's design approach was/is still very valid.

The disruptor pins are key here. Without them, the block is no better than a Cascade, although the block would certainly be cheaper to make than a Cascade without the pins, just a lot more restrictive than a Cascade, but then again with the disrupter pins the G3 also pips the Cascade for performance

Originally Posted by Incoherent

I am curious about your new jet vs cup diameter ratio with regards to the jet impingment / mashing transition. Cup walls make the scenario lean heavily towards mashing. They are wider, are your cups shallower?

The cups are marginally shallower than the Cascade, but this is done mostly to increase bp thickness to assist in a more even cooling effect across the base of the block. The Storm's bp is about twice as thick as that on the Cascade, which still makes it very thin. The base-plate thickness is something that I really focused on with respect to achievable CPU overclock, rather than relying on the actual temperatures being reported. With a thinner bp it's possible to get lower reported temperatures with the Storm block, at the expense of peak overclock. This was something that I also spent a lot of time optimising as really this block was more about me getting as much overclock as possible out of my CPU.

Originally Posted by Incoherent

Is the 4-6d z really optimum with the increased shear from exiting rather than sitting water ... these are somewhat rhetorical Friday night questions... better go to bed...

4-6 z/d is apparantly optimal for maximised Nusselt numbers for submersed jets. I don't use a z/d of 4-6 in the block though. This was something else that I worked with for optimising with the disruptor pin.

[Cathar]

Originally Posted by Les

Cathar
Do you have a link to the image?

Or a reference to the work? Please.

Sorry Les, I don't. Someone forwarded me the image a long time ago (~2yrs) and I never knew where it came from, other than the accompanying comment that it was meant to be better than normal impingement.

I'm in the process of attempting to track down its source for you, and will post here when I find it.