eddy currents in metal core PCB's

---

Isn't an aluminum core board going to be expensive? How thick would the aluminum be?

Can you make plated-through electrical contacts to the aluminum? It would be a shame to not use the layer as ground.

Will the aluminum core help that much thermally? It will reduce thermal spreading resistance some, but you still have to get the heat from the parts into the aluminum, and then get rid of the heat overall.

Eddy currents induced into the aluminum probably won't do any harm.

I've done high-power 3 layer boards, but a fourth layer costs about the same. You could do two 2-oz or so internal copper layers, which would add up to about 6 mils of copper, thermally about the same as 15 mils of aluminum, depending on the alloy. Use all that extra copper to reduce resistive losses maybe.

John

e

nt.net ---

John has it right on the copper layers. Put in two, 4 oz CU layers with thermal vias filled with conductive epoxy (CB-100). Better thermally and lower cost than AL. Most PCB fab houses treat it like a normal PCB. Cheers, Harry

The I-core inductors, or inductors with gap in the core, are indeed affected by solid metal layers nearby. Whether those layers are copper or aluminum, there could be significant degradation of Q with the corresponding increase of the losses. It could be a problem, depending on the particular design.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

e

nt.net ---

you can get alu pcb prototypes from pcb-pool, only single sided though

1.5mm thickness, 100um isolation, doesn't look like you can connect to the alu directly but for a price they'll machine the alu with holes, threads etc.

not sure how the thermal specs compare to fr4:

l.pdf

isn't cheap but not too bad

-Lasse

snipped-for-privacy@netfront.net ---

Hi,

I don't know what the cost would be for the aluminum PCB, I figured that the extra cost of the aluminum may be offset by the better cooling allowing higher operating powers given the same components.

Some aluminum thickness specs I've seen are 0.8mm~3.0mm(0.02"~0.12")

I think the way the aluminum can be connected to the circuit is by laminating a copper layer directly to the aluminum at first, then connecting the circuit to this copper layer. This seems like a good way to do it but the last board house I talked to said they aluminum is left floating.

I thought the aluminum being between the ground return and the circuit could cause a problem by blocking the ground return path. Would this be a problem with AC current circuits? I'm just thinking of the analogy of a magnet being dropped down an aluminum pipe :)

cheers, Jamie

--- Posted via news://freenews.netfront.net/ - Complaints to snipped-for-privacy@netfront.net

snipped-for-privacy@netfront.net ---

Hi,

At full power the board should dissipate at most 50watts, and its dimensions are 5.25" by 8". I guess using 4 layer standard FR4 with heavy copper may do it. Do the conductive epoxy filled vias work with vias in pad as well?

Here's a picture of the board so far:

The two switcher circuits (a buck and a boost) in the middle of the board generate nearly all of the heat in the board.

cheers, Jamie

--- Posted via news://freenews.netfront.net/ - Complaints to snipped-for-privacy@netfront.net

er

make

em

front.net ---

wboard.png

t.net ---- Hide quoted text -

Jamie, conductive epoxy works well with "via in pads" with no solder sucking. Thermal layers are about spreading the heat as soon as possible and conducting it to a heat sink, normally the PCB edge bottom, then to a thermal plane connected to the chassis. Your PCB edges are full of connectors with no room for heat flow. The center of your board can be used as the pilot lite. Cheers, Harry

snipped-for-privacy@netfront.net ---

Yikes, 50 watts is very serious on a PCB, especially a small one like this. Is there going to be air flow? The numbers don't look very good here.

You will get a hot spot in the middle of the board, so I'd spread out the heat generating parts if possible.

If you use a lot of fairly big vias, you probably don't need to fill them. Their thetas will be pretty low.

Personally, I'd hack a mockup, carved from FR4 with an x-acto maybe, dummy resistors to make heat in the right places, and measure heat rise in your actual enclosure. Heck, try a piece of solid aluminum your size and check its theta.

Here's a simple hacked mockup, 2-sided 30 mil thick FR4, dumping one watt:

ftp://jjlarkin.lmi.net/T750_t1.JPG

ftp://jjlarkin.lmi.net/T750_t2.JPG

ftp://jjlarkin.lmi.net/T750_t3.jpg

ftp://jjlarkin.lmi.net/T750_t5.jpg

That last pic is the back side of the board. The back is solid copper, with a strip of kapton tape to increase emmisivity so the thermal imager works better. Note the local hot spot, directly opposite the resistors. Lateral heat spreading tends to be poor on PCBs.

John

snipped-for-privacy@netfront.net ---

OK, serious aluminum thickness! You could just bolt and/or epoxy a PCB to a slab of aluminum! By the time you do all that, you may as well use a regular heat sink.

Your 50 watts is scary.

John

Just my 2 cents: I would never do that. I'd screw down some real heat sinks and mount TO220 onto those. You likely still need forced air cooling. But if you'd just sink it all into the board you'd also need the number for the next fire station and an evacuation plan :-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

snipped-for-privacy@netfront.net ---

How much did you estimate? I came up with about 60C rise.

John

snipped-for-privacy@netfront.net ---

Very, very roughly, 80 square inches of isothermal surface (both sides), in still air, might be 2K/w. So 50 watts gets you 100K rise. But no pcb will be isothermal, so the central hot spot will be a lot worse.

With good heat spreading and a lot of air flow, this would probably be feasible. Personally, I'd use heat sinks.

John

We're brainstorming a pockels cell driver that would dissipate 200 watts or so. The catch is that we need to keep it really small (for low inductance) and cool, which means low capacitance to the heat sinks. Aluminum nitride slabs over copper heat sinks, maybe.

Water would work, as a last resort. Messy.

John

Oil works really well. But put a (large enough) drip pan underneath. I had one of those contraptions stored in the garage and luckily it had a drip pan when it decided to take a dump.

With Eimac 8873 tubes they use beryllium oxide for a low-C heatsink. But be careful, this stuff is dangerous. Just a speck of dust accidentally scraped off can send folks into the grave. Still shows up at surplus dealers though:

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

snipped-for-privacy@netfront.net ---

I've got a feeling you're closer than I am. Thanks for the numbers.

John

If this goes onto one of Jamie's rockets maybe it falls into the categroy "but it doesn't have to last more than 60 seconds" :-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

A good reference I keep in mind is that the highest-powered x86-type CPUs dissipated around 150W from a package with something in the ballpark of a square inch or two of heat spreader going to a large heatsink with a coiling fan. (And hey, that was only with a heat sink on one side! Perhaps with one on both sides you could get upwards of double the dissipation...)

These days they (Intel/AMD) seemed to have backed off on that a bit -- most of the newer high-end CPUs are more like ~125W (and many are now

al

ower

e

o make

seem

t 15

to

s

th

water inside a heat pipe? I'm think I've read about soldering heat pipes directly to components using low temp (~140C) solder

-Lasse

It was also the case that the stock Intel CPU heatsinks of that era were complete dogs with terrible whining banshee fan bearing noise and lousy dust trapping fins. Swapping them for a well designed decent third party gamers heatsink with a bigger slower fan and usually a silly name was a really smart move if you valued your sanity.

Regards, Martin Brown