Liquid cooling heatsink design

e.

like this

with chassis in place of the mica washer ?

Something like that, but if I'm going to use screws to directly fasten the sinks to the top of the enclosure (not sure how else I'm going to be able to make it secure to get a tight connection to apply thermal paste to)...I don't know the geometry of the hollow inside to know how much material there is to fasten the screw into without breaking thru into the hollow interior.

I may have to sacrifice one to find out. :(

9 sounds like a lot.

I'd expect those to leak if you drilled a hole on one.

One bigger classic cold plate might be better than a bunch of those:

You can safely drill/tap those.

Water has a huge specific heat, so a little water flow does a lot of cooling, something like 0.004 K/W at 1 GPM. So a small cold plate is a lot of cooling, provided you can transport the heat into it.

What's inside the enclosure? Does it have a good thermal path to the roof?

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

About ten or so years ago, I made some liquid cooled heat sinks using coppe r and solder. Your local roofer is likely to have some suitable copper. I just made a wood bit with the desired form and had at it with a small ba ll pein hammer. If you made your own , you could make one larger heat sink instead of using a bunch of bought heat sinks.

Dan

I'd like to cool an AMD x86 processor, max TDP around 65 watts, that has to live in a tiny easily-portable aluminum box with not much room for active cooling. It's going to be used for various multimedia tasks for live performances at art galleries/small exhibitions/etc. where extremely quiet bits are going to be part of the production so speedy fans whining is a no-no. There's very little height for a heatsink and fan already, a low-profile sink and cooling fan will probably fit, but I'm not optimistic about the ability of a low-profile sink and fan spinning slow enough to be quiet to cool such a CPU when running flat-out.

The thermal path to the roof would be pumped water from a sink on the processor up through the sink on the roof interfaced to the chassis, with fins on the outside if necessary. Is the back-of-napkin idea at least

I have enough time to experiment with whether a "custom" (i.e. hacked-together) liquid cooling system with big-ass heatsink will work, if not I'm going to give the sink + slow fan a shot and hope for the best

nk

sure.

99.9% you can't drill into them, and why would that be easier than a bar ac ross?

I already had most of the PC-related parts on hand for this one-off on hand already, if I can come up with a cromulent cooling solution for a tiny box that ideally doesn't involve fans I'm going to take a stab off it.

They make off-the-shelf-solutions for more money but the problem with that is then they have the money instead of me

Ok, I guess I'm not understanding from the diagram how the hollow aluminum block attaches to the underside of the roof of the enclosure solidly enough to make a good thermal connection. If I just paste it there with grease it's just going to fall down!

If you can get them mounted probably just simple convection could be used instead of the pump ?

sink

ed

om

losure.

be

e

r across?

a metal bar with screws into the heatsink clamps them to the inside of the enclosure at the same time clamping the heatsink to the outside

e.

Water cooling sucks. Heat pipes work better, and use the heat being dissipa ted to pump the coolant around.

I had my nosed rubbed in this back around 1994, when the water-cooling that I'd put in developed a tendency to stop working - when bubbles appeared in the circulating water after about six months in the field - and Doug Stewa rt replaced it with a heat pipe.

Sloman A.W., Buggs P., Molloy J., and Stewart D. ?A microcontroller

-based driver to stabilise the temperature of an optical stage to 1mK in th e range 4C to 38C, using a Peltier heat pump and a thermistor sensor? ? Measurement Science and Technology, 7 1653-64 (1996).

--
Bill Sloman, Sydney

If I can avoid having to find a pump that's quiet and will fit in this tiny form-factor I'm all ears. There may be an off-the-shelf low-profile heat pipe + fins thing that will work, I have a couple of these heat-pipe beasts sitting around which with a 120mm fan at about 500 RPM would probably be sufficient for near-silent operation

but it would require a hole in the roof of the case!

Here's some pics of the patient:

The box:

Internals:

Motherboard with socket at right, dual 120W SMPS boards with outputs paralleled thru a daughterboard at left, to be powered via either an external 12V adapter in parallel or 24V in series if I can come up with a scheme to equalize the input voltages. They're narrow enough to possibly both fit on standoffs mounted to the front of the enclosure by the power switch.

It's all going to be a very tight fit and run pretty warm but I think I can pull it off with the proper cooling arrangement. I can undervolt/underclock the processor but would prefer not to do that if not absolutely necessary

pated to pump the coolant around.

Water cooling can work well if it is designed well. The heat being dissi pated can be used to circulate the water. I made a water cooled heat sink and used it with the reservoir higher than the heat sink. It did not need a ny pump to circulate the water. I am not saying that heat pipes are not a good solution , but you do need more room in the computer case. And water cooling is easy to implement at low cost if you have a reasonable assortmen t of tools.

Dan

This cold plate:

I like a lot and looks like it would fit and mount easily. Around 1 GPM doesn't seem like an unreasonable figure for a small 12V 5W pump that would fit inside my little enclosure. Price could be a little better tho :|

I'm pondering mounting one of these dealies:

To the outside of the enclosure. Could put a little "No Fishing !" aquarium sign in there

sipated to pump the coolant around.

sipated can be used to circulate the water. I made a water cooled heat sin k and used it with the reservoir higher than the heat sink. It did not need any pump to circulate the water. I am not saying that heat pipes are not a good solution , but you do need more room in the computer case. And wate r cooling is easy to implement at low cost if you have a reasonable assortm ent of tools.

The problem we ran into was that air bubbles kept on showing up in the circ ulating water, after about six months of continuous operation. There may be a well-known solution to this, but the heat pipe was the easiest we found, and it let us get rid of the water pump. Convection driven systems can wor k, but you need enough vertical height for the mass of the column of thinne r, warmer water to generate enough pressure to drive enough circulation.

We were shoe-horning our solution into a carefully styled case which indust rial designers had wrapped rather closely around the proof-of-principle pro totype, making it tricky to add in the stuff that kept the gear running rou tinely on the customer's bench top for years at a time.

--
Bill Sloman, Sydney

I'm doing a gadget that will mount inside a laser. The customer already has a water cooling system, so we'll just nab a bit of his flow.

This is the baseplate to be cooled:

For testing in our lab, I was thinking that I could have two gallon jugs, one almost filled with water and one almost empty. The water would syphon between jugs, and I would just swap their height now and then.

If I disssipate 40 watts and want maybe 0.1 K/W at the cold plate, the flow rate needs to be about 1/25 GPM. The gadget has an overtemp shutdown, so if I forget to swap jugs that gets tested too.

Or I guess I could buy some sort of submersible pump and put it into a bucket of water. That would be quiet.

--

John Larkin         Highland Technology, Inc 

lunatic fringe electronics

Is there a good guide/website/app/spreadsheet for making calculations re: active cooling systems?

Is there some type of "thermal circuit" analogy that can be used when one knows or can guesstimate the thermal resistances of the processor->plate->coolant interface, the mass flow rate and type of coolant, and the resistance of the radiator plate->enclosure->ambient?