The "Cascade" - mini-cup directed jet-impingement block design (56K warn)

[Blackeagle]

nicozig,

A very fine idea indeed. Would help make the block more "idiot proof".

[Cathar]

Originally posted by nicozeg
Cathar, with that amazing precision parts your machinists can make; having to align the final assembly by eye seems a missuse of your resources. You could design some type of aligning feature on the lower face of the middle plate, that precicely match another one carved on the copper base. This could add some machinning time to your current design, but it will ensure a perfect centering of the jets every time. Once you switch to molded parts, this features comes almost for free.

Yes, very good idea. I had indeed thought about this when molding. A little harder to do with the present machining as I would need to remove extra material to create the bumps, but for a mold it's exactly what I intend to do.

I'll also get them keyed a certain way so there's no confusion as to which direction the plates go, since with the top two plates it is possible to turn them around such that the top plate blocks off the middle plate's exit.

[nicozeg]

A little harder to do with the present machining as I would need to remove extra material to create the bumps, but for a mold it's exactly what I intend to do.

But you're already removing material from the base to make the jets longer. You can forget about feature tapering to avoid a tool change, and you only need to leave some small spots of the base with the original thickness.

It's a good warning that of the srinkage, and that can be easily solved making linear shapes disposed in an axial orientation, alowing for srinking, and even termal expansion during use, without loosing alignment. The groove on the base just need to be a bit longer than the shape on the middle plate.

I marked with blue in this pic a suitable placement that can work both on the milled and molded versions. It just need a little cleareance from the o'ring, but the space available seems enough.

[bigben2k]

Originally posted by Cathar
Cheaper than molding eh? Like what are you suggesting?

You could simply use a good adhesive, to glue polycarb tubes to the top, and let the machinist drill the precision holes.

As for the alignment, I suggested (earlier) that you include a groove around the block's screw holes: there's no need to create new dimples/grooves, when you already have them .

[justbsn2]

Hello All!
Is it possible to pre-order this block as of now? I was just about to order the Whitewater block from cooltechnica.com

But......

I really like the clear top on this one over the whitewater block.Now,im thinking of waiting on ordering for a P4 CPU.

Where can i get on a waiting list and pre-order?

Thanks in advance

[Cathar]

Send an email to slf@connect.com.au and I'll get back to you with details.

[utabintarbo]

Originally posted by Since87
...

It might be difficult to get a precise fit between these features. When designing molds you have to take into account how the material will shrink when it cools. However, the machining of the copper could be modified after a sample molded piece was produced, to ensure that the alignment was good.

It's often easier to mold a piece, and make the rest of the design work around the molded piece, than it is to get the molded piece to match precise dimensions.

The precise shrinkage factor is probably well-known, and can be accounted for in the design of the mold. For prototype parts molded for the automotive industry, a positional tolerance of <.25mm (.010") is commonly maintained. I don't think (but I don't know) that such a small deviation will affect matters greatly. Only cathar knows for sure.

Also, a couple of round "knobs" might be all that is needed for locating the pieces together. They can be made assymetrical to ensure proper fit (see pic).

Bob

[bigben2k]

Ya'll...


I just got a call from Cathar (Hi m8!), and we got to "shoot the breeze" about this design.

Maybe it's too early, but this design is going to be very difficult to beat, because anything else that could come close would "require some expensive equipment", as Cathar said, and personally, I believe that the double inpingement effect that this design uses really makes maximum use of the coolant flow.

In any case, I still intend to try it with 2 Johnson pumps, and I might throw in my Little Giant as well, to see if I can throw the coolant in turbulence, even before it hits the microtubes, if that's possible (probably not). Actually, I think I was aiming for turbulence to occur within the microtubes, now that I think about it, before it's jetted.

The problem is that the speed involved is (not "near sonic", I was confused with Radius' inlet) relatively low as it is now (~60 km/h, or 40 mph, from Cathar), and I certainly don't have any hope of increasing that much higher.

I just ran the calc, taking 4 gpm, dividing it by 52 (52 tubes), and assuming that the ID of the tube is 0.033 inch (as linked, above), the flow speed would be 28.9 fps (feet per second), or 19.7 mph (miles per hour).

If I can increase the flow rate above 4 gpm, to say 10 gpm (which would be quite a feat!), then the flow speed through an individual tube increases to 72 fps, or 49 mph.

But I suspect that more than likely, what will actually happen, is that I'm going to hit a wall, where the pumping action will throw more heat into the loop, to the point where the resulting performance would actually decrease.

Send me the block Cathar, and I'll throw lots of pumping action at it!


In the mean time, I'll see if I can make some kind of calculation on the Reynolds number.

Turbulence is determined by a dimensionless value called a "Reynolds number." The exact range between turbulent and laminar flow is disagreed upon in some circles.

According to McCabe, Smith, and Harriott, turbulent flow will occur at Reynolds numbers above Re~24,000. Turbulence can be forced by obstructions in the flow as long as the Reynolds number is well above Re=2,100. Below that, there's no way to avoid laminar flow.

Sieder and Tate define turbulent flow for significant heat transfer to be abover Re=100,000.

Reynolds number is equal to linear velocity multiplied by channel diameter multiplied by density, all divided by viscosity. So to get turbulent flow by increasing flow only, you'd have to increase flow rate a hundred times.

Re = V * D * rho / mu


(posted by Alchemy, in the Radius thread)

[jaydee]

Originally posted by bigben2k


Maybe it's too early, but this design is going to be very difficult to beat, because anything else that could come close would "require some expensive equipment", as Cathar said, and personally, I believe that the double inpingement effect that this design uses really makes maximum use of the coolant flow.

Not only hard to beat but I think the "limit" is getting pretty close on how low you can pysically get with just air as the temp control and the other minor limitations such as the TIM joint ect. It will take some serious design work to come up with something better at the same or even close to the same price. But I thought that about the White Water to.

[bigben2k]

One idea that came up in the Cascade thread over at OCAU, is the possibility of drilling the baseplate straight through, for something like a direct die cooling approach.

Otherwise, if Cathar's calculations are correct, this block is about 2.5 deg C from the maximum that is possible for any waterblock.

Interestingly, the TIM joint accounts for most of the temp difference between the CPU and ambient air: go figure!:shrug:

[leejsmith]

lets say the cups went all the way through the base plate and we are using a p4 cpu . what could be used that would be a good TIM and also stop any water escaping.

[bigben2k]

Clamping pressure and an RTV silicone seal arount the perimeter of the IHS.

[leejsmith]

i will add it to my list of things to try once i get my diy version of the cascade block optimised.

[jaydee]

Originally posted by bigben2k
One idea that came up in the Cascade thread over at OCAU, is the possibility of drilling the baseplate straight through, for something like a direct die cooling approach.

I have been thinking about this very thing for quite sometime but have failed to come up with a way to properly seal it. It would not be easy to get a decent silicone seal that wouldn't leak around the edges of the die. You really need it sealed ontop of the die itself but you will loose needed contact between the base on the die.

The closest thing I have come to an acceptable solution (IMO) is to actually mill out a spot on the base of the block to where the die fits into it and then you could use some silicone to seal it off from the sides. Would still be tough to get the seal right. All for what? a possible 2.5C gain? Thats where I am having difficulties justifiying trying it. I am not really to interested in a couple C that will not provide any measurable gains in anything.

[MadDogMe]

All for what? a possible 2.5C gain? Thats where I am having difficulties justifiying trying it. I am not really to interested in a couple C that will not provide any measurable gains in anything.

Then why buy a CNC mill to make blocks?. Would of been easier to slap a Maze3 on, and cheaper!. It may perform like a lemon but it'll get you an overclock. Being able to make the worlds Uber performing block justifies it in itself, specialy when you have a comercial audience/interest in mind...

Vorn sprock tekspeeks as they say down under ...

[jaydee]

Originally posted by MadDogMe

Whatever....

[Cathar]

Could we leave the bickering outside?

Been playing with approximating some form a venturi taper into the jets. Been using a regular 118° drill bit to chamfer the jet intakes, and then following up with a 45° cutting tool to create a smoother transition into the tubes. The result has been a small, but measurable, drop in flow resistance.

The following drawing is pretty much to scale on what's going on:

[hara]

Does this add considerable machining time to the whole block or is it uber fast on the CNC ?

[jaydee]

Originally posted by hara
Does this add considerable machining time to the whole block or is it uber fast on the CNC ?

Looks to me like just one more pass with the 45degree tool. Shouldn't add any significant time I wouldn't think. Just dropping down, lifting up, moving to the next hole, dropping down, lifting up, moving to the next hole ect... Should be pretty quick on the mill if my thinking is actualy what he has in mind.

[Since87]

Any news on the feasibility of molding?

Some thoughts on a cast in eyelet...

[Cathar]

Originally posted by Since87
Any news on the feasibility of molding?

Some thoughts on a cast in eyelet...

Totally infeasible to do the molding reliably with the cast-in eyelets. They are simply too small. I do have a sample of eyelets and we're investigating tapered press-fitting with them.

The above diagram represents what on the blocks currently moving out the door. The double-chamfer thing happens reasonably quickly.

[MadDogMe]

Thanks for editing your abusive post Jayuhdee. How about doing the rest on this site (if you've got the time?) ...


Sorry for the OT but I'm sick of this BS...

PS, you're PM's taking a very long time to arrive...

[jaydee]

Originally posted by MadDogMe


Sorry for the OT but I'm sick of this BS...

Then STFU. You got PM.

EDIT: PM taking a long time eh? BS!

[t00lb0x]

Don't fight its evil. People come here for help and flaming each other isn't helpful. Oh a bit late to say this but Cathar I love your block and you're an inspiration to us all.

[GTA]

tbh, i'm sick of the arguing, its not advancing anything to do with waterblock design, which is why I assume we're all here.

Leave it alone, both of you. I don't care who's right or wrong, just bicker in PMs, and stop forcing all of us to wade through a pile of your crap to read what should be a wonderful thread.

Don't reply to this, just PM me if you have a point to make. Stop posting in this thread. Its very clear no-one wants to read it.


Cathar - Your attention to detail never fails to astound me. I would never have thought that the campfering you're doing would be worthwhile, but it obviously is. Gives me SERIOUS food for thought on my own block.

The only way forward from here is by METICULOUS attention to very gradual improvements.

However, we said that about the WW, which everyone said was "perfect" and could not be beaten. Proved them, and me wrong mate.

The lesson to be learnt from this design, as far as I can tell, is that by paying attention to the little things, you can really puch things forward. The general design for this block has been around for 6 months or more, but it is the attention to details that many would consider irrelevant that pushes performance from average to extreme levels.

Lovely block, the only place I can see any improvment coming from here is the placement of the outlet. Its fairly obvious that its not optimised. I'd like to see a 1" outlet placed around the inlet personally. This would not suit many systems, but it seems odd to me to have one outlet where it is.

my 2p anyway.