More heatsinking issues...

Well, as you must already know i'm building a 2x15w based on the JLH'96. Now, i'm still battling with the heatsinks. Each channel dissipates a maximum of 90W, so that's atleast 0.3°C/W per channel, or 0.15°C/W total. I found some small 0.6°C/W sinks, but at 25$ each, 100$ IS painful :)

So, my dad runs a metal shop. I've been doing some calculations, and if i stack 20 15x15cm plates (bolted together with a separator in between, say three 5mm thick bars), i'd be in the vecinity of 0.1°C/W. Nice so far; and i really don't mind the metalworking. My issue is this: my dad works mainly with stainless steel; hence, i can have enough 2mm thick steel sheet to build the damned thing. Alumminium is not as readily available unfortunately. Am i mistaken, or should the steel work better dissipating heat than the Al? I thought the only reason aluminium was used for heatsinking was that it can be easily anodized. Any takers? It would be a bit more heavier, yes, but i can live with that if there's a thermal benefit.

Ah, and i plan to have them black painted (with black automotive paint); it's nowhere as thin as an anodizing, but's not thick either, so radiation should improve a lot. Is this ok?
 
First of all, steel might be better at disipating heat than aluminum (I don't know), but it's nowhere near as good at conducting heat (a completely different issue). For a heatsink to function properly, it must be able to conduct the heat away from the sources (transistors, MOSFETs, resistors, etc.), to the fins, where it is disipated to air. Steel can't conduct heat well enough for this task.

Where there are only a few heat sources, I've heard of copper bars being used (perpendicular to the fins) to spread the heat (attach heat sources to the copper, and the copper to the aluminum heatsink). This creates a 'line source', instead of a 'point source' of heat, and will often allow the heatsink to function better. This might be useful for you.

Also, the general design you describe can be quite effective (in aluminum). A good example of this type of heatsink can be found at 'http://home.earthlink.net/~lotusblossom/index.html'.

Finally, most automotive paint has a gloss finish. This is not good for a heatsink. A heatsink works better if it's got a flat finish.

For best results, use black anodized aluminum.

Good luck
 
Thoth said:
First of all, steel might be better at disipating heat than aluminum (I don't know), but it's nowhere near as good at conducting heat (a completely different issue). For a heatsink to function properly, it must be able to conduct the heat away from the sources (transistors, MOSFETs, resistors, etc.), to the fins, where it is disipated to air. Steel can't conduct heat well enough for this task.

Really? That's funny, i was almost convinced it was the other way arround... i'll have to stick to Al then; it's available for me, just not as easy. But again, buying sheets of Al will STILL be cheaper than paying for those 4 heatsinks.


Finally, most automotive paint has a gloss finish. This is not good for a heatsink. A heatsink works better if it's got a flat finish.

No no no, this is a matte black finish; it's a thin coat of paint, like those used in industrial equipment. Again, nothing as good as black anodizing, but i think worth the try.

Thanks!
 
peted said:
Lisandro, if you are up to a bit of metal bashing, check the stuff on my site re handmade sinks. Mail me direct if you want more.
[Edited by peted on 07-31-2001 at 02:14 PM]

Saw your pics the other day; great work btw! :D Basically, my idea is to do almost the same, only a tad smaller and box-shaped. BTW, could you brief me on how you painted yours and how it performs? Thanks.
 
Lisandro_P said:
[BTW, could you brief me on how you painted yours and how it performs? Thanks.

Haven't tried auto paint - maybe it's too thick? Either way you'd have to use =matt= black. I used black-board black - water based - used watered down a little, rub the Al with wire wool - it's too shiny for the paint to take otherwise, then degrease (I used washing up detergent in the sink) and paint with 2 coats. I found it best when watered down a little. Objective is a =thin= matt black layer - it looks OK and meets those objectives, though anodising is better. I tried some Al black from a gunsmith - it doesn't work so well, and is horrendously expensive and chemically unfriendly too!

Performance: each of the half circle sinks in the photos achieved 0.17C/W (free convection), and 0.11C/W with a slow (silent) fan. The thick Al plates are 300mm wide, the rest of the plates are dimensioned to be a 320mm diam circle. All the Al plates/fins are 250mm high. I wrote a little C prog to calculate areas and cutting data based on the formula for a circle. I guess the same could be written in Excel. Want the C source code?

Given the lack of design data I could find, I determined the figure for a single vertical sheet (free and forced) by experimentation, and worked from there. Turned out to be reasonably accurate, but have to say there was an amount of faith involved! Dunno how many devices are in JLH - I used the v.thick Al plate to carry the MOSFETs because only 2 in a Zen stage - very concentrated heat compared to being able to spread many around a sink. My MOSFETs run at 128 (gain)and 112 (cc source) watts. Get yourself one of these 3 terminal temp chips to use as a temp probe (with some thermal grease) - use with battery and a voltmeter. BTW the outer plates =do= run almost as warm as the others so the heat transfer spacers do their job well. Mine are 180mm long(ie 35mm free either end), 25mm wide, 9mm thick.

Also bear in mind temp derating for the device, junc <> case (0.5C/W for me), and case <> sink (0.2C/W for me) values. always worth going for as much extra h/s as you can.

Al Vs Stainless? Al wins hands down - remember how they put Copper inserts in the bottom of those expensive stainless saucepans to stop local burning? Also a tad easier to work!!

Cheers

Pete
 
peted said:
Haven't tried auto paint - maybe it's too thick? Either way you'd have to use =matt= black. I used black-board black - water based - used watered down a little, rub the Al with wire wool - it's too shiny for the paint to take otherwise, then degrease (I used washing up detergent in the sink) and paint with 2 coats. I found it best when watered down a little. Objective is a =thin= matt black layer - it looks OK and meets those objectives, though anodising is better. I tried some Al black from a gunsmith - it doesn't work so well, and is horrendously expensive and chemically unfriendly too![/B]

That was pretty much what i had in mind... the paint i'm refering to is that kind of "industrial" black matte paint used on cabinets, floors, equipment, etc. If that's no good, i'll go your way.

Performance: each of the half circle sinks in the photos achieved 0.17C/W (free convection), and 0.11C/W with a slow (silent) fan. The thick Al plates are 300mm wide, the rest of the plates are dimensioned to be a 320mm diam circle. All the Al plates/fins are 250mm high. I wrote a little C prog to calculate areas and cutting data based on the formula for a circle. I guess the same could be written in Excel. Want the C source code?
... Dunno how many devices are in JLH - I used the v.thick Al plate to carry the MOSFETs because only 2 in a Zen stage - very concentrated heat compared to being able to spread many around a sink. My MOSFETs run at 128 (gain)and 112 (cc source) watts. Get yourself one of these 3 terminal temp chips to use as a temp probe (with some thermal grease) - use with battery and a voltmeter. BTW the outer plates =do= run almost as warm as the others so the heat transfer spacers do their job well. Mine are 180mm long(ie 35mm free either end), 25mm wide, 9mm thick.

I'd like that program, send the source to [email protected] :) . Now, i've been working with Rod Elliot's Excel sheet, and i plan to go with 20 15x17cm plates, separated in the middle by a 2cm wide bar (5mm thick). Just as you did, two of these plates will be thick to allow heat to flow faster off the semiconductors. Ideally, this should set me in the vecinity of 0.1°C/W. The original JLH'96 uses two power BJT's per channel, but i'm putting two more in parallel so they run cooler and power derrating does not become a problem.

Al Vs Stainless? Al wins hands down - remember how they put Copper inserts in the bottom of those expensive stainless saucepans to stop local burning? Also a tad easier to work!!

Yah, i did some research yesterday and it's completely true. I don't know where i readed once that steel was a much better heat conductor... oh well :D Thanks for the feedback!