New waterblock concepts

[Mr Evil]

I have started designing myself a new watercooling system, starting with the waterblock. I have designed several, and must decide which one to actually build. A picture of one of the designs is shown below:



More pictures and information can be seen on this page.

Any comments you may have would be very welcome.

[futRtrubL]

I always like to see new designs and inovations.

That should get good surface area contact fith good water speed however it looks like you have necked down the water flow alot. You might get very bad flow rates if so. How do you plan on making something like that?

Edward

[Mr Evil]

I would make it from sheets of brass or copper, cut, bent and soldered together, although I may have to alter the design if it proves too difficult to make.

[Brad]

you'd need to have a fairly thick base, and i'd definately make the channels out of copper in that design, I like X2 and X1 combined

ie the shape of X2, but with 5 or so tubes with the vortex concept going on

[Mr Evil]

Combining the two sounds like a good idea.

I would like to make everything out of copper, but it depends on availability.

[Volenti]

Here's something that should make the construction of your first block a lot easier, you can get box channel brass in various sizes from hobby/toy stores that sell model aircraft/trains ect.

I have a peice at hand, 305mm long, 9mm wide, 4.5mm high, with a wall thickness of about .5mm, should only cost a buck or 2.

[#Rotor]

That is a very valuable little tidbit of information....

those two blocks are well though out... the first one looks like a micro-channel concept.... where the idea is to divide the liquid up into as small a particle as possible, and then introducing the heat to it individually, thus getting maximum heat transfer per volume moved...

the second one looks like it's going for the brute force method... using velocity and surface shaping to generate a violent and turbulent flow-pattern , also very effective and of the two, probably the easiest one to make...

[EMC2]

X1 block -
You can use flow diversion walls (basically a fin) to help split the flow between the channels. At the inlet pipe put one in the center. Then halfway to the main block put another one splitting each of the two areas into 4. Also, if you can adjust the size of the channels, you can affect the flow, smaller channels in the middle, slightly wider on the outsides. Just keep in mind that the center is where most of the heat is, so you don't want to divert too much to the outside.

From the pic, the channels look really small. You might want to increase the height of the channels. Same goes for the base plate thickness.

To solve the smooth curve fabrication issue, you could use 45 degree angled turns and make the ends like mini-surge tanks (similiar to a radiator or heater core).

Of course, pure copper will be better than brass from a thermal standpoint.


X2 block -
Shallow threads aren't enough to get a swirling flow at the flow rates you want. Nor would the density difference be enough to create the mixing action from centrifugal force.

From the positive side though, the threaded walls would create a quite turbulent flow at the wall surface. The big issue would be keeping the threads very shallow with as coarse of a pitch as possible. If there is much depth to the threads, you can end up hurting your heat transfer do to almost no flow on the inside surface of the threads. You could solve that by adjusting the tap size versus drill/pipe size to get shallow threads. Or you could go back after tapping in deep threads and redrill the hole with the next size drill bit, which would both reduce the thread depth and create a flat surface at the peaks of the threads.


X1-X2 combined -

You could actually do the same thing with your X1 design by using tubular channels "embedded in a box" like Brad mentioned. Approximately 70% (depends on tube wall thickness) of the flow would be through the inside of the tubes with the remainder of the flow between the outside of the tubes and the inside of the box walls the tubes are embedded in. The one problem there is the size of the 'contact patch' with the bottom plate. A way around the problem would be to crossdrill the top of the base plate to create round grooves that the pipes rested in. A groove that was as deep as 1/4 the diameter of the pipe would give you slightly more surface contact than the pipe diameter.

An alternative way to get the same results would be to take a solid block and cross drill the round channels directly into it. The two issues to solve there are fabricating the "end caps" to create the chambers at the inlet/outlet sides (same as with the X1 block) and doing it this way would force you to use a larger size channel (the smaller the drill bit the shorter the channel can be).

[Brad]

I think that the idea of the micro tubes is good, like in Xj's block, but laminar flow is surely going to occur with a long thin tube, which could be stopped by doing something like the vortex thing, or a helicoil design.

Basically combining X1 and X2 still

[Volenti]

Speaking of Xj's block, was it ever put into operation? how did it perform?

On the laminar flow issue, how thick is the boundary layer? I suppose it would depend on the flowrate and surface, would there be a tube size to flowrate ratio that would reduce laminar flow to a minimal amount?

[EMC2]

Boundary layer thickness - is proportional to flow rate and is also affected by the surface. Both the flow regimine (laminar vs turbulent) and pipe size affect the heat transfer coefficient (film coefficient h).

For a purely laminar flow, the film coefficient h is inversely proportional to the hydraulic pipe diameter. Double this diameter and you halve the film coefficient. The larger the diameter, the worse off you are regarding heat transfer.

The greater the turbulence (and the less laminar) the flow, the higher the film coefficient h is, and thus the better the heat transfer. Turbulence greatly aids heat transfer

Based strictly on pipe/channel size (NO structural parametrics such as surface roughness to alter the flow regime one way or the other), yes there is a ratio. What you're really talking about here is the Reynolds number Re. The equation is:

Re = ( D * p * v ) / u

where D is the hydraulic pipe diameter, p is the fluid's density, v is the fluid's mean velocity, and u is the fluid's dynamic viscosity. (say that fast 3 times) Note that they hydraulic pipe diameter is NOT the diameter of the pipe. Flows with Reynolds numbers below ~2300 are considered to be laminar.

Also, turns in a pipe increase local turbulence (but increase pressure drop, although not at the same rate).

<edit: big fingers + speed = typos>

[Brad]

Joe is testing the block, he isn't letting out too many details, but he did say it was very very good

[Mr Evil]

Lots of good feedback so far, thanks everyone. I'm getting more ideas...

Volenti: I already used various brass bits and pieces from my local modelling shop, such as you can see in my GPU waterblock:



And the same one half constructed:

[Volenti]

Ahh, nice work very elegant.