[quote]Originally posted by Cathar
[b]Again, I am struggling to find the exact link to the paper that I read, but basically it was showing that the effect of extra turbulence in the jet stream by standing it off from the base and having it shear against the water around it was actually more effective at stripping the boundary layers despite the force of the jet's effect being slightly reduced. I believe that I've been able to mimic and observe the paper's statements through experimentation myself. The paper was talking about simple submerged jet behavior and not really talking about jet-in-a-cup behavior which is something else that I had to explore.

[quote]

Yup, it has been proven that the collision between the the boundary layer of two side by side jets hiting the base will create turbulance and enhance heat transfer provided the jets are close enough and the boundary layer has not lost much of its momentum.

But this research is done using a flat base as the heat transfer wall. In the case of a base with cavities (cascade design), the boundary layer at the botom of the cavity created by the jet will hit the sides of the cavity creating a 2nd impingement effect. As the fluid rises out of the cavity, the fluid has not much energy in it due to the double impingement previously. Even if there is still enough momentum, the fluids momentum will be directed upwards to the centre plate.
Then only will it react with the boundary layer that came from the fluid of the neighbouring jet.

But im still not sure which of the cases above (flat base & dimpled base) gives beter heat transfer.

Double impingement with reduced crossflow effect

or

Single impingement with enhanced turbulance with slightly higher crossflow effect

which is beter ?

Let's call it "twin inpingement", in reference to two jets side-by-side, and leave "double inpingement" to describe what's in Cascade!

Interesting.

So instead of having a perfectly round dimple, we used an "oblong" shape, and use two jets instead of one, there might be a slight performance improvement? Got a link to this research?

The problem, I believe, is that it would significantly reduce the amount of fins (aka dimple walls), which are really critical in increasing the surface area of the baseplate. I guess it could be compensated a little bit, with a bit more dimple seperation. A flat plate isn't going to cut it, IMO.

Of course this assumes that the second inpingement in Cascade has a significant impact. A CFD model of the effect would be most interesting, at least to help visualize it.

Otherwise, the jet tubes, as I've already mentionned, are a critical part in allowing the coolant to exit. How it interferes with the jet I can only guess: a toroid "vortex"? Too small. I'd try tracing the flow paths in an experiment on a (much) larger scale, but I'm not versed enough in fluid dynamics to affirm that the larger scale would be, in any way related to the small sizes involved here. I'm also pretty sure that I'm never going to have the scaled pumping capability :p

The turbulance created by the collision of the boundary layer does not apply only to 2 side by side jets but every jet neighbouring it in the jet array.

The jet array arrangement also plays an important role in the creation of turbulance due to boundary layer collision.

There is also a paper I recently read that said an array of jets that has no particular arrangement works better.

Even arrays with varying jet diameter can also enhance the heat transfer

I understood it as you described Aleck. Sometimes analogies help explain somewhat complex ideas to people best ;) And unfortuantely the surface to surface contact between the materials does have a boundary layer of sorts, hence the need for thermal grease.
Sometimes I over generalize to make things simpler but yeah I got Ya :)

A nice flow diagram showing the actual water flow at the surface for a single jet would show people a lot I think. Flow is next to impossible to determine in your head, until it's actually plotted out.

Here's a simple diagram showing the flow of fluid as the blue arrows.

The fluid closes to the wall slows down due to the intermolecular forces. The area where the fluid velocity is lower than a free streem velocity is the boundary layer

In the case of a cascade, the diagram shown above is at the base of the cavity of baseplate where the jet impinges marked by the red elipes in the diagram below

as the fluid gets further away from the axis of the jet, it gets slower and the boundarylayer gets thicker

The effective water velocity remains pretty much unchanged at the edges of the "cone" of impingement. It's only after the water "bounces" that it slows down.

The trick to the Cascade was to guage the cup diameter to present the cup wall just after the point of "rebound" from the cup base. This of course is dependent upon the jet velocity but I believe that I managed to pick a jet/cup width ratio that applies to the general flow rates that people use. In general it was okay to have the cup wall being slightly too far away as opposed to being too close. This allowed me to design to higher flow rates and not suffer terribly at lower flow rates.

If I were designed a low-flow only block (<2LPM), I would have done quite a number of things differently.

Speaking of that...

Do you know what the relative durability of silver and copper would be?

So the diameter of the cavities are based on the geometry of the jet which in turn is based on the velocity of fluid and ratio of jet height to jet nozzle diameter

You have designed the cascade so that it does something like in the modified diagram below.,..?

Sorry..
left out the pic

Well, according to the Wolverine handbook on cooling, copper is able to withstand 6fps (~2m/s) velocity without any form of long-term erosion. The Wolverine book is talking about ultra-conservative long-life scenarios.

An Eheim 1048 plus heater-core will give about a 2.5m/s jet velocity.

As to the scale of the erosion, my personal belief is that it is extremely low even at 10m/s velocities. So long as there is no gritty particulate matter being thrust against the metal I figure it should be pretty safe.

Hard to put an exact figure on it, but have been running this silver block with ~5.5m/s jet velocities with the Iwaki MD-30RZ for close 3 months now. About every month I pull it apart and inspect it. Last time I checked the fine-level machining marks left at the base of the cups, which numbers in the very small number of microns in scale, were still plainly visible. If there's any erosion happening, it certainly doesn't look like it'll be an issue within 10 years at the least, even with this fairly powerful pump and soft silver.

Something like that. Not sure I agree with the "scale" of various parts of that picture, but yeah, something like that.

As only the bottom part walls of the cavity is involved in the secondary impingement...so i was thingking if making helical groves on the walls will give added performance..?

The groves could be easily made by using a small inner threadmaker (taps)..

Theoriticaly, this will surely thin the boundary layer of the fluid flowing out of the cavity walls because of the turbulance created by the whirling effect

Cathar, I was thinking about how you said that your cup:jet ratio was such that the "rebounded" water would hit the wall just after rebound. For some reason I got to thinking about parabolas and hyperbolas and focusing the rebounded water to a certain area to acheive added cooling after rebound. Haven't gone any further with that thought because I also wondered, do you know if the rebounded water is rebounding straight back out of the cup or could the rebounded water also be going straight back into the jet and slowing it down? Obviously your block works great and you've spent much time designing it but I haven't seen anyone mention where the exiting water is going exactly. I hope you see where I'm going with this :rolleyes:
I was just wondering if exiting the rebounded water to somewhere else besides back out the cup might help the impingement effect some.

Yes, I had considered tapping the holes like you say. Aside from the difficulty of tapping into copper into such small holes and not breaking taps when doing so, I was concerned that the pitch of the thread would be such that no real swirling would develop as the vertical velocity is just too quick. Instead there may be a more dangerous scenario developing whereby the water inside the thread grooves remains fairly static and the water just rushes up the middle. I never directly experimented with it though.

Part of the rebounding is useful in that it creates extra turbulence as it shears against the force of the incoming jet. Remeber what I was saying above about how having some jet shear is actually very useful, and one can see improved performance as a result, and this is true up to a point of when the jet loses too much power for the extra shear turbulence to compensate for.

Guess I should have stated my thoughts more clearly as I was assuming the rebounded to be too much interference to the point of slowing the jet down.

I guess the clearer question about the jet shear would have been, have you done flow testing to chech this or are there computer models that will show what the flow of your cup/jet is?
You obviously know alot about fluid dynamics it seems ( I know basically nothing as if you couldn't tell) so was just curious if the amount of shear happening was a guess aor calculated. I'm assuming calculated from the amount of study you've put into this.

Most of my data is a mix of study of research, and experimental evidence. Went through quite a large number of variations/prototypes of the Cascade. I don't have access to the sort of computational fluid dynamics software that would be needed to simulate this sort of flow, or if even the models of such are good enough to do it properly? :shrug:

Nah, it's all gross design evaluations, nothing calculated. Cathar did all the experimenting, to find the optimal solution, and the rest is history.

For straight water, this design is going to be extremely difficult to beat. All from a simple concept: double inpingement.

As I've stated, you could easily build 30 variations of this block, to find an optimal solution: hole spacing, hole diameter, jet diameter, jet distance, fluid density, thermal conductivity of the baseplate material and coolant, and jet velocity. Those are the variables.

I believe that Cathar may have used a home made simulator, but I don't know how useful it was here.

(BTW Cathar, how many variations did you have to build/try?)

8 different plate/jet pairs for which each was modified 3-5 times, exploring cup depth, inter-cup geometry, cup-jet width ratio, jet height, cup/jet density effects, base-plate thickness, etc. By the time I finished each pair was basically useless.

Sounds like something that Im doing....

Only that im not using the protruding nozzles....

and im also doing some computational fluid dynamics simulation using CFD

In the simulation i will only model an array of 2 x 4 jet nozzles....
to minimize computing time.

I figured out that only a part of the block needs to be simulated in order to predict the characteristic of each specimen....

Cathar, thanks for your patience and your willingness to share your hard earned knowledge with us :)
This is and interesting subject and one I haven't been able to find much in depth coverage on. At least not that are readily available to the mass public.

Sorry if I'm a pest but this whole subject has my brain a buzzin :)

Have you done any research into what back pressure on the return flow side does to performance? Seems as if too much back pressure on the return side could cause hindrence of the jets velocity in the cups. In other words if the return flow flowed more freely, with less resistance than the pressure side, then the jet velocity would also be maximized?
Possibly larger ID on the return side dumping into a res that's not full.