homemade water cooling for everyone ;) - january, 2006.

Not often I comment on project logs anymore but this is looking very impressive, just love that radiator. Keep up the good work

 

Wow, the blocks look amazing. I can't imagine doing all of that work and then NOT showing it off! Cheers, and keep the updates coming!

 

Optimal work...

p.s.:Is this(click)a lighten scooter piston? Beautiful

 

Very nice, some questions and thoughts

I really like this 19th century steam powered look, and I'm planning to build myself a phase change cooler based on some experimentation I've done using LPG as a refrigerant. So I'm very interested in your copper work.

You don't know what pressure your water main is, around here our pressure is around 500kPa or 5 bar, if that helps. (It seems that your system is going to be operating mostly at ambient anyway) It is safer to test things with water than compressed air, as compressed air can be quite dangerous if something breaks suddenly. You might be able to use a hydraulic piston to test pressurise your system. A 1MPa pressure gauge with a 10mm screw thread cost me $20 at a plumbing shop.

You can get special drill bits that cut sideways for milling, you can use these in drill presses to cut a smooth channel - talk to your metal worker friend.

What actually motivated me to post was where you attached the power transistors and diodes to blocks in the PSU. In several of the PSUs I've fixed the heatsink itself can sit at say 400V, and even if there is an insulated screw and isolation washer on the transistor, it might be easy to get a short connecting the high voltage to your plumbing. Please check this out carefully before using it.

If you need help with thermal calculations feel free to email me, I've studied heat transfer engineering and know how to design heat exchangers. Soldering those copper fins onto the pipe will improve thermal efficiency considerably.

Keep up the good work and I hope your finger gets better.

 

Thx!
It could be quite a mess with all tubes inside

Yes, I've later found out that water pressure is about 5bar (max.) and it's definetly overengineering parts to hold that much pressure, but hey, if they won't leak at 5bar, it certainly won't at pressure of one aquarium pump

hm, didn't have play with those drill bits, but anyway blocks are completed so no more drilling is required
I have some sweet ideas about alternative blocks and how to do it, but this project won't see it, maybe in some future I'll do another water cooling to try a different approach

Hm, you gave me some thinking to do ... What you saying is pretty logical and possible, and I'll have to consult with my friend that knows alot more than me about electronics how to avoid possible hazard ... thx for advice !

thx, but this is all start like an idea with some research how others did it, and simple, I've just start building it that way - no calculations are done (cause don't know how to do it )
And the biggest challenge was, will that manifold provide same flow at all blocks, and yesterday I've discovered that it will works like a dream.

thx again, and finger is pretty much like new already

2 hellas91 : No, not from scooter, It's from Yugo 0.9

JaredC01 : That's what this forum is for

 

The calcs for things like water heat moving capacity are very easy. A gram of water a second warming 1C carries away 4.2W of heat. If your pump moves 5g/s (18l/hour), which is easy for even an aquarium pump, and you are happy for the water to rise 5C you can cool a 100W CPU.

You can also measure these things by connecting a suitable heat source to the blocks (say a bunch of 20W resistors) and measuring the temperature rise.

Calcs for a radiator are a little more tricky, we need to know things like maximum expected ambient (perhaps 45C), power dissipation rate (basically the same as the power coing in the plug - measure this with something like the sparometer) and air flow.

I think it is worth working these things out - firstly to make sure your system is sized appropriately, but also to work out whether you need the fans running, how fast to run the pump etc.

 

Seen that turpija?
http://www.bit-tech.net/hardware/2006/03/29/Alphacool_Starter_Kit/1.html
You got mentioned... and you deserve it.

 

He's right. The transistors aboulutely HAVE to be isolated from the heatsink. If one of the pipe nipples on the outlet were to brush the case, *POP* your Power supply would be dead, not to mention the shock hazard it presents.

What happened to the original mounting hardware? There should have been scerews with specialised plastic washers, that center the scerew in the transistor tab hole, and those little grey rubberey shims, this are specialised non-conductive heat transfer thingies.

In the end, i suggest measuring the resistance from each transistor tab to the waterblock before powering the supply on. Also, running a ground lead to each WB is probably a good idea.


incidentally, your "turner" is called a machinist.

 

2 njh : This is some interesting informations, but there would be probably some difference between calculations and actual performance. Anyway, this is practicaly finished so if it perform bad (which i doubt) I won't change it a bit
But it could be nice to do some calculations before actually started to build it.

2 aon`aTv.gsus666 : wow, sweet ... I'm fameous
thx for link

2 Wolfe : Hm, this is starting to worry me. But original mounting of transistors was with normal screws, screwed directly in aluminum heatsinks, only on one transistor was that plastic washer, but only thing between transistor and heatsink was some rubber (alike) material and I though it was like those on oem pocessor coolers, but it actually could be electric insulation, hm.
Here is picture, where you can see there is no direct connection between screws and transistor, except on this smaller one where you can see that plastic washer... so if you know what should i put in between transistors and waterblock to stop short circuiting and still allow enough thermal conductivity, please give me advice, cause I don't want the whole thing blow in white smoke

 

If the calcs disagree then you've forgotten something, and you need to update your model. That's called engineering If it performs badly perhaps your CPU or ram will cook and you will have wasted lots of money.

Those power semis in that last photo are all look low voltage (they are the diodes for the 5, 12, -12V lines I'd be guessing), I wouldn't worry about them, it's the other ones that I'd be scared of. To be honest I would take the copper out of the power supply and use the original cooling (or at least air cooling). How do you know you've got all of the hot devices?

How are you going to cool the transformer?

 

In the meantime I spoke to my friend and he knows all about that rubber/plastic insulation material under transistors so it wouldn't be a problem (or atleast it shouldn't).

I don't know if I got all hot devices, but on those transistors was originaly aluminum heatsinks, so I suppose they need some extra cooling.
I won't cool transformer, as there is not much more space left, and I will leave original fan inside, and hopefully as head will be extracted from transistors by water there would be much cooler inside for the transformer too, and other components also.

And no, I won't change my mind, copper stays in power supply unless something blew or melt

 

Finally it came the day to water test block, so after some time this is what I end with ...



And the most important question about whole project has been answered. Yes, that manifold will supply all waterblock with enough flow and pretty much all will get the same (except cpu block, as it has larger diameter tube), as you can see ...



But now, the bad news, at couple connections on block there was leaking occured, maybe because i was putting too much pressure from house water supply (and it's at 5bar max), and maybe it would never leak with normal pump. But it's better to leak now than later on my motherboard or something. So, all connections will be checked and refited with thicker layer of teflon tape ...

Time to move on those bare metal parts that need some peparation and primer. This weird yellow thing is aluminum primer ...



The dark caught me, but I had to cover it all before primer harden so it become useless. This is drying booth



The day after it's all dry and still weird looking



Next step was to put primer on plastic parts. But first I wanted to flatten dvd front mask and cover those useless logos ...



Plastic primer that I used is transparent and it's nothing special to see, but anyway, here is all parts drying on nice spring weather



All for now, next update will be in color

 

A question for you.

The spray paint have you use on the block is an anti oxide product?

 

first off, i woud like to say the radiator, resovoir and all the blocks look very good

However, IMO if you start trying to pump water into that many components your going to get a pressure drop

your trying to pump so much water, but now you have many more paths, so there will be less water going through each block

i would have thought 2 pumps may be better, so the pumps both suck from the resovoir, go their directions and then both join back into the rad, which then goes to the resovoir

P.S. http://forum.pcstats.com/showthread.php?t=33481&highlight=watercooling

this guy has kinda done somthing similar ...

 

yahooadam: That's why the system should be properly modelled. I claim that a single 5W pump is enough, you claim that 2 (unspecified) pumps are required. How can we decide?

 

im not saying it wont work, im not sure

but having 7 (right ..) waterblocking in your system is really gonna be spreading a small pump very thin

i was just saying

if it works, great, just a comment

 

But how thin? My little 5W pond pump will move 200L/hour at 0.5m head (which is probably more head that this system). If we are willing to let the water heat by 1C it will move 230W, by 5C, 1.1kW. More significant is the ease of conducting the heat into the water (with this design, probably fairly poor) and getting rid of the heat (again, probably fairly poor). But as turpija doesn't specify what the goal of the project is, I guess it's all fairly academic

 

I know it won't be easy for pump to handle all that blocks and make a decent flow, but only thing that bothers me was, will all blocks get flow at all, or some will and someone won't. And that was proved in last update that it will work great. Next thing is to test few pumps and see which one can do it properly, so I don't see big problem here ...

I'm aware that this is not the most performance WC sistem ever made but i think it could perform quite well. And real goal was to actually make a WC sistem from scratch, because it was interesting project for me, cause I never did it, and I learn alot during building it. And I think it just can't perform bad, cause there will be alot of water in sistem, and with that big radiator I think is properly overengineered.

 

with the test did u have all the waterblocks connected though ?

0.5m head is gonna be split between all 7 blocks though

so it will only go 0.0714m if you have all 7 blocks connected, however one of the pipes is bigger, so that will probably reduce it further

not forgetting, if you didnt have the blocks connected, thats extra resistance

assuming all the tubes are the same, each block will get 28.57L /hr

although that figure will be different, as the CPU block will get more, and the others less

 

Actually if you want to calculate anything there is 9 blocks.
And how you know how much flow will I get, what pump is that you calculating with. How you know which pump will I use before I do ?!

Anyway, when I test few of them then I'll know, this is all guessing.

and btw, yes, all blocks were connected during the flow test, as it's show on pictures in project log ...