Aluminum-core PCB cost?

Phil Hobbs wrote in news: snipped-for-privacy@supernews.com:

OK. Use a 0.090" PCB and a LOT of COPPER.

The obvious approach is an aluminum-core circuit

Ummm... no. Do you think Al conducts heat that much better than Cu?

Get a custom copper heat pipe and use thermal epoxy to attach the LED PCB to that.

I've never heard of one.

Right...

snip.

Be aware.... a faulty aluminium LED board is a nightmare to rework. Deffo not a DIY job...

There are also extreme copper boards, from 4 oz up to 20 oz.

My current insane project has a 4-layer PCB bolted to a machined aluminum baseplate, with gap-pad and AlN insulators between. That approach might be more economical.

mandag den 23. april 2018 kl. 17.19.08 UTC+2 skrev John Larkin:

some time ago I saw this 20oz board on reddit:

the guy said it cost $2250

Looks silly, 20 oz copper and skinny traces.

Sorry, what is the benefit of Al over Cu here? Cu has higher thermal conductivity, 3x the density and is the standard material for PCBs.

Is the idea that Al will radiate heat away faster? Wouldn't a Cu PCB with Al heatsink be the better choice then?

The Al is not instead of the Cu, it is instead of (part of) the FR4. Of course it would be even better to use an equivalent thickness of copper, but it would be more expensive.

mandag den 23. april 2018 kl. 18.13.21 UTC+2 skrev John Larkin:

from the reddit thread,

?the side that you?re looking at has -450V / 0 / +450V tra ces. The other side has 14V @ 150A and 24V @ 80A.? The other side b eing this:

races. The other side has 14V @ 150A and 24V @ 80A.? The other side being this:

There is something strangely beautiful about that. Now I just need to come up with an excuse to design something like that here...

I'd have to go to a lot of layers to get that to work, because I need quite a lot of circuitry. It's a customer requirement, so I need to go see if it's possible and how much it would cost. I might laminate the Al on one side of the boards and use it as three sides of the box. (Yes, this probably means using two flexes to join the sides together. As I said, an *inhmmhhmm* customer requirement.) Hopefully their CM will explain that to them.

Cheers

Phil Hobbs

The benefit of Al over Cu is that it doesn't trash the litho resolution.

Cheers

Phil Hobbs

V traces. The other side has 14V @ 150A and 24V @ 80A.? The other s ide being this:

I don't know if fr4 bonded to alu is really that much much better a than regular pcb bolted to an alu plate

I don't know if fr4 bonded to alu is really that much much better a than regular pcb bolted to an alu plate

Basically the same thing JL's doing right now, put a Gap Pad under it and be done.

Upshot: lower thermal resistance than FR-4, Kapton or other bondings!

Tim

Yes, that's one approach I'm considering. The Cu-on-anodized-Al approach is an interesting one as well. I could just put the transistors and sense resistors on that, and let the low-power circuitry stay on a regular PCB.

Cheers

Phil Hobbs

Rob wrote in news: snipped-for-privacy@xs9.xsall.nl:

Then he should use an anodized Aluminum plate, which is NOT electrically conductive, and have his copper traces (or the entire PCB) bonded to that.

Problem solved. Use thermal epoxy to attach an anodized Al plate to the bottom of the PCB.

"Tim Williams" wrote in news:pbmag5$eth$ snipped-for-privacy@dont-email.me:

Screw the gap pad. Use a more themrally conductive bond.

not generally regarded as good engineering practice

What's wrong with multilayer, modest weight copper?

I've got an 8 layer 2 ounce board that's almost impossible to work on. It takes five minutes of hot air before anything even begins to budge. I think a D2PAK could dissipate 10W on minimal pads (with thermal vias, mind).

On the topic of exotica, have you considered ceramic PCB? There are a few shops doing it now, it doesn't look too expensive -- probably not more than what you proposed?

Tim