Quick and dirty micro channel experiment

[nicozeg]

Update: Did the nozzle, disapointing results.

I think my pump is too weak for the task. is rated at 1.4m head 800L/h; roughly similar to an eheim 1048

I made some flow tests timing the filling of a one gallon container:

-Pump with filter and 1m hose = 25 seconds.

-Full system without nozzle = 40.5 seconds.

-With 6.5mm ID nozzle = 48 seconds.

The temps stayed the same with the nozzle, maybe a bit higher. Its difficult to know with 1ºc monitoring resolution.

So looking at this results I have a question for cathar: Does the middle (nozzle) plate of white water make any performance dif whith such a pump? Maybe it just impact flowrate, but not temps.

[Cathar]

Originally posted by nicozeg
So looking at this results I have a question for cathar: Does the middle (nozzle) plate of white water make any performance dif whith such a pump? Maybe it just impact flowrate, but not temps.

The nozzle plate works just fine for an Eheim 1048 in my testing. It makes a real difference that you'd definitely notice with even a 1C resolution.

Mind you, I did experiment quite a bit to find the right configuration, and further the location and shape of the nozzle, the is just as important as it's size.

[nicozeg]

Maybe What I need to be sure is another mobo, its very dificult for me to get consistent results. If I intentionaly tweak aiflow arround socket, reported temps can vary as much as 8º.

End of Rant.

Flow wise, this block should be less restrictive than white water. Channel cross section area is about the same. Mine is 16 channels of 4 x 1.2mm, WW is 16 ch. of 5 x 1mm. The difference is in the nozzle. Mine is a 6.5mm diameter circle, equal to 33mm2. WW has 8 slits of about 1 x 3mm for a total of 24mm2.

So this case could be similar to what Volenti proved with his direct die: Several small jets are better than a bigger one.

[Cathar]

It should be noted that every single one of the WW measurements you noted are from the initial prototype. The "production" version differs in almost every respect (fin dimension, nozzle geometry, base thickness, except the number of channels) and is a higher performing solution as a result.

[freeloadingbum]

Have you thought of soldering a flange to the end of the input tube? It should keep the water flowing along the base a little longer instead of deflecting away from it. I posted this in another thread but I think it would work with your design as well.

[LiquidRulez]

I though a freeloadingbum was supposed to steal your ideas, not give them.

[nicozeg]

Originally posted by Cathar
It should be noted that every single one of the WW measurements you noted are from the initial prototype. The "production" version differs in almost every respect (fin dimension, nozzle geometry, base thickness, except the number of channels) and is a higher performing solution as a result.

Is not my idea to do a public reverse ingeneering of WW, I just realized that the multiple small jets are a clear advantage over a central one. This feature is a natural implementation on a parallel fins design, but very difficult to do in a radial one. Did you notice that Big Ben?

[bigben2k]

It is indeed difficult to do in a radial pattern.

In Radius, I'm limited to 4 "jets" (versus WW's 8), and I'm still not convinced that it'll reach the bottom of the center fins (nor in WW).

In any case, the purpose is to speed up the coolant, for more turbulence, which is "relatively" inefficient, because turbulators can do this better (true "jet inpingement" notwithstanding), where the former requires a more powerful pump.

In multiple jet direct die cooling, where true "jet inpingement" principle is applied, it'll work very nicely, but it should be remembered that it's a requirement because of the large surface area. A heavy pump is a must. Also, it's probably the most efficient way of doing it, but only because there are no other options: no surface dimpling, no fins, etc...

WW combines a couple of principles, so does Radius, so they should come very close to direct die results.

[nicozeg]

There's another problem with jets in this radial pattern. (Warning, this is completely from my guts, just how I think this thing work.)

In jet impringement, fast moving water is thrown perpendicularly against a flat surface. It goes to a direct collision, after which it has to choose the way to go, spreading radially.

In WW block, water is thrown in the middle of the channel, where it has two ways to go. This is not the ideal jet because it has some drag from the sides, but after the impact to the bottom it has to choose side.

When the jet is placed at the beginning of the channel, there’s only one way to go. My fear in this case is that water doesn’t wait for the impact at the bottom and just choose the shortest path, avoiding the corner where we want it to go.

All this are theories I’m making, trying to explain why I didn’t see an improvement with the nozzle. Does it make sense?

[Cathar]

Originally posted by freeloadingbum
Have you thought of soldering a flange to the end of the input tube? It should keep the water flowing along the base a little longer instead of deflecting away from it. I posted this in another thread but I think it would work with your design as well.

Then the water wouldn't be cooling the fins as effectively. Heat does flow up the fins and they need to be cooled just as much as the base. Well, that's my consideration of the applicability fo such a thing to the WW design. It may work well in a different design.

Originally posted by LiquidRulez
I though a freeloadingbum was supposed to steal your ideas, not give them.

What is your problem?

[Cathar]

Originally posted by nicozeg
In WW block, water is thrown in the middle of the channel, where it has two ways to go. This is not the ideal jet because it has some drag from the sides, but after the impact to the bottom it has to choose side.

When the jet is placed at the beginning of the channel, there’s only one way to go. My fear in this case is that water doesn’t wait for the impact at the bottom and just choose the shortest path, avoiding the corner where we want it to go.

All this are theories I’m making, trying to explain why I didn’t see an improvement with the nozzle. Does it make sense?

The WW "jet" is not an ideal jet, but that "drag" that you refer to as actually a highly turbulent convection layer. Swings and roundabouts sort of thing. What is lost in terms of impingement, is gained in terms of boosted convection from the fins.

I am however a little fuzzy on the exact jet impingement theory and how it pertains to striking surfaces with "furniture" about, but it would seem to me that what you're saying about the single path out makes sense. There would indeed be a tendency for the fluid to take the shortcut, however this does not mean that it won't be doing something useful, depending on the balance of fin height and how far up the fins (whether radial or parallel) the heat travels.

[bigben2k]

Here here.

Jet inpingement, in its pure form, isn't exactly what's been applied here. Instead, the flow has been speeded (sped?) up, to attempt to recreate a similar effect.

Wether or not this is actually achieved, is a mystery, but the results are there, so something is working.

[freeloadingbum]

By Cathar:
Then the water wouldn't be cooling the fins as effectively. Heat does flow up the fins and they need to be cooled just as much as the base. Well, that's my consideration of the applicability fo such a thing to the WW design. It may work well in a different design.

I don't think you quite understand what I mean. The idea of the flange is to force the water through the fins just as the plate on top of your block forces the water through the micro channels.

Because nicozeg has a large open area above the base (and fins), the water is deflected throughout the open area after it hits the center. By adding a small "Top" a short distance above the base (and fins), the flow is focused where the heat is. The distance the "Top" is from the base can be determined through testing, same thing with the diameter.

An added benefit of the flange is that the jet holes can be drilled into the center to speed up flow (like the slot in the middle plate of your WW)

[Cathar]

freeloadingbum, I now see what you mean. Yes, I totally agree. A flange like you describe would improve nicozeg's implementation. nicozeg's cap mechnaism doesn't force the water to stay down close to the heat.

[Jessfm]

Well I have still problems with my small unit, especialy getting the correct "Jet" design.

I still have to consider the fact that not everyones pump will be equal, and still require a certain flow.

Also other considerations and machine problems hamper me.

2months and counting so far, and still not a unit that would be remotely cost effective to manufacture.

its more then Swings & roundabouts, its compromises , all the time. Have to give something to gain. Exactly how the Atlantis ended up at its final design , hardly where I intialy planned on going, rather more where the path led.

Still eventualy I will get there

BTW Cathar, good look for your business venture m8

[nicozeg]

Originally posted by Cathar
freeloadingbum, I now see what you mean. Yes, I totally agree. A flange like you describe would improve nicozeg's implementation. nicozeg's cap mechnaism doesn't force the water to stay down close to the heat.

That could be useful, but it will require longer channels, given the cutting method I used.

In this drawing Orange is the base, Red the nozzle and Blue the palomino die size. It can be seen that flow is channeled OK over most of the crucial area.

Jessfm: What block are you talking about? Any thread or webpage where details can be seen?

[Jessfm]

Nico, you asking about Atlantis or deveopment unit ?

[gergregg]

I've built something very similar over the past few months. I called mine snowflake because that is what the base resembled after my drilling and dremeling. What I did with my latest block was start with a piece of 3/8" stock for the base, which was drilled and dremeled to leave a 1/8” thick base plate. Then I cut out a 1 3/4" doughnut out of another piece of stock with a hole saw. I threaded the inside of the doughnut I just created to accept 3/8 MPT. After that I took a 2" copper cap and cut it to a 1/2" height and drilled and threaded a 3/8" hole to line up with the center threaded hole of the doughnut. I then drilled a 1/4" exit hole at the edge of the cap to thread in another barb. The exit barb had the threads cut so they barely made it through the copper cap. Now the nozzle could be constructed from a 1/2" by 3/8" MPT hose barb. I pressed a 1/4' brass plug into the inner diameter of the 3/8" barb and soldered it together. I then proceeded to drill 7 small holes in the plug to create a multi jet nozzle. I was finally ready for the final construction of the block. I took the doughnut and soldered it over the center of the base. I then took the copper cap with only the single outlet barb on it and proceeded to attach it to the base with solder and the barb to the cap with the same method. I used a scrap 3/8" fitting threaded through the hole in the cap and into the soldered doughnut to hold everything in position while doing the previous step. So I ended up with a radial type rotor design with an interchangeable nozzle.

My results were quite good. I was using a maze2 (which had been mounted and remounted numerous times) previously and my new block degreased my temps by almost 10 degrees C without changing anything else in my setup (setup consists of 1/2" fittings, chevette heater core w/ shroud and 63 cfm fans in a push/pull config, custom block, submerged maxijet 1200). My block currently sits about 10-11 degrees C above room temp at full load (xp1600+ @ 1833 default voltage). My temps are measured by the on die diode on my Epox, so they are probably decent.

[nicozeg]

Nico, you asking about Atlantis or deveopment unit ?

Dont know, are’nt the same?

Gergregg: Not sure if I understand it completely, but seems a lot of work hidden under that cap. A pic is worth a thousand words...

[bigben2k]

It looks like a combination of a cross-drilled block (baseplate), with a cap on top. Similar concept, different construction.

[m3_arun]

maybe instead of the donut-shaped nozzle, you should have solid round one in the center.

[freeloadingbum]

Gergregg, your design is very similar to what I was thinking about for myself, except for the replacable nozzle.

What kind of flow are you getting? You said you drilled a 1/4" hole for the output barb but you mention 1/2" tubing. What size barb are you using for the output?

[gergregg]

Freeloadingbum

I used a 1/4" MPT to 1/2" tubing barb. The inside diameter is the same for the 1/2" barb as it is for the 1/4" MPT. As for flow, the pump with nothing more than about 18" of 1/2" tubing would flow about 8.4L per minute. With the block in line and another piece of tubing about the same length I got about 6L per minute.

[Gulp35]

I was inpired by this post and desided to design my own. the only difference it that I would use a pin grid like most HSFs instead of the radial cut design. do you think that would help performance and also what type of pump are you useing and what gph do you get with the waterblock attached?

[bigben2k]

If you want to go that way, you might want to take a look at Hoot's pin-fin block, over at OC.

Construction details

Results here