semi-passive psu idea

kentc · Post by **kentc** » Thu Mar 09, 2006 3:16 pm

hey all,
i need your thoughts on this crazy idea i got from reading about you guys makeing your own passive psus.
at first i thought of just takeing the psu out of it shell and just let it be cooled by the general airflow in the case, fueled by the 120mm exhaust fan. but then i thought of makeing it semi-passive, and by that i mean putting the hot-running components of the psu on a big external heatsink but still retaining the original heatsink aswell, cooling the components passively and actively. the reason is i don't trust the external heatsink to lower the temperature enough - afraid it would raise ambient case temperatures too much but it would help expel atleast some of the heat so the fan in the case wouldn't have to do all the cooling. also it would be a failsafe against the external heatsink running to hot.

sorry excuses for illustrations, whiped up in xpaint:

second picture ofcourse with 120mm fan in it's proposed position, sucking air out of the case via psu components.

questions: will leaving the original heatsink on the components direct to much heat into the case or would the fact that the external hs is up mean heat wants to rise? does heat in metal care about up or down?
will the external heatsink work effectievly just sitting on top of the case or should they always be mounted sideways?

comments? i bet...

Mr Evil · Post by **Mr Evil** » Thu Mar 09, 2006 3:55 pm

I wouldn't run a PSU out of the case if I were you. Components inside may reach hundreds of Volts and, depending on how the insulation is arranged, the heatsinks may also be at a lethal potential.

You might make a complete custom case for one, as long as all touchable parts are securely earthed.

Captain John · Post by **Captain John** » Thu Mar 09, 2006 5:44 pm

OK- a couple of thoughts:

The heatsink on the PSU may be live... Meaning, it actually is carrying and is meant to carry current. Connecting this to the grounded case is just looking for trouble.

Have you attempted to remove the (ICs/mosfets/VRMs) from a power supply? It's a PITA. It's many times easier to replace the fan with one which barely turns and push a couple of CFM through the PSU.

-John

kentc · Post by **kentc** » Fri Mar 10, 2006 8:09 am

thank you for your thoughts.
ok, to clarify: i am ofcourse aware of the consequences of running the psu without it's casing and i will ofcourse isolate the components from earth. but there just won't be enough room for the casing inside the computer case and i do not want to run an external powerbrick, so this is the best solution i've come up with so far.

kent.

ps. ran a bbs for about 5 years back in the day, psu with fan removed open in case - only got shocked once!

Captain John · Post by **Captain John** » Fri Mar 10, 2006 8:58 am

I hate to be a stickler, but _how_ are you going to have an external top heatsink connected to an internal heatsink, attached to VRMs which possibly are passing current to the expected internal heatsink, and have it isolated from ground?

I can imagine a thermal interface material which passes heat but not electricity, but know of none. Silicone thermal paste typically has graphite in it. Even water will pass electricity to some degree, more when other chemicals are in it...

I'm extremely curious as I actually have tried to remove VRMs and heatsinks from a PSU, wasted probably an hour or more of my time, and had solder fly in my face in the process of removing the VRMs (I'm not mangled, don't worry). I think you might have better luck going down a different avenue. Perhaps removing the PCB from the PSU and directly attaching larger heatsinks to the existing heatsinks?

How much power does your system require? A low-power system (tualitin & HDD) could run as low as ~50w under load, and... even at 70% efficiency, you're only disappating 15w of heat. I see no reason why 15w cannot be dealt with sans fan, as long as the PSU is not required to remove extra heat from the case itself (65w likely would require at least a fan).

I can run my pentium d 820 and have the PSU put out air which is not detectably warmer than ambient. The trick is that the PSU is not in the path of heat from the proc/system.

-John

jaganath · Post by **jaganath** » Fri Mar 10, 2006 9:32 am

What you need is a material that is not electrically conductive but is very thermally conductive. Ceramic comes to mind.

Mr Evil · Post by **Mr Evil** » Fri Mar 10, 2006 9:46 am

Captain John wrote:I can imagine a thermal interface material which passes heat but not electricity, but know of none. Silicone thermal paste typically has graphite in it. Even water will pass electricity to some degree, more when other chemicals are in it...-John

Typically, thin sheets of mica are used in conjunction with thermal paste. A variety of other materials are sometimes used in place of mica, e.g. aluminium oxide. Silicone pads can be used without thermal paste, which is clean and simple, but performance is lower. Hard anodized heatsinks can be used without insulation (bolts must still be insulated though, of course), since the anodized layer is itself an insulator. Some component packages are already insulated and so need nothing extra between them and the heatsink.

TomZ · Post by **TomZ** » Fri Mar 10, 2006 9:56 am

Captain John wrote:I can imagine a thermal interface material which passes heat but not electricity, but know of none.

Actually, this is quite common. Here is one manufacturer of such materials (gap pad):

http://www.bergquistcompany.com/thermal_materials.cfm

Typically, thin sheets of mica are used in conjunction with thermal paste.

Mica is commonly used for only small area, e.g., a single transistor/FET. Products like gap pad easily outperform mica in nearly all characteristics.

kentc · Post by **kentc** » Fri Mar 10, 2006 11:33 am

Captain John wrote:I hate to be a stickler, but _how_ are you going to have an external top heatsink connected to an internal heatsink, attached to VRMs which possibly are passing current to the expected internal heatsink, and have it isolated from ground?

you're not a stickler (being a non-native english speaker i don't even know what that means litterally

), i'm very keen to have your input. especially since you have had led spit in your face from trying

this guy has done pretty much exactly what i wanna: http://astro.temple.edu/~xioti/fanlesspsu/
another couple of good articles: http://homepages.paradise.net.nz/max.wi ... assive.htm, http://www.silentpcreview.com/article80-page1.html.

the system is rojected to use something like 125-150w at peak, sure would like to expel some of that redundant heat passively.

best regards, kent.

Chris Chan · Post by **Chris Chan** » Fri Mar 10, 2006 1:40 pm

Interesting, but i would say the long wires would increase inductance/capacitance/resistance within the PSU, so it's iffy. I myself wouldnt do it, I'd just do a fan-mod.