1. Home
  2. Electronics Design
  3. some thermals

some thermals

I have a board where I want to use four small mosfets in parallel, to switch 100 volts or so in 2 ns or less, into 50 ohms. One gets speed here by stressing small fets, sot23 or super-sot6 packages. The problem becomes how to get the heat out. So I was wondering how much heat we can dump into the PCB ground plane. Candidate numbers are

4 fets

4 layer board, 1 oz copper, L2 ground plane

12 mils of FR4 from top to ground plane

1 watt target dissipation, more if possible

Topside drain pour about 600 x 250 mils

So, what's the thermal situation? The pour area is so small that we can ignore convection; 0.15 square inches at maybe 150k/w per, is 1000 k/w.

So most of the heat travels through the FR4 to the ground plane. Using published values (conductivity of the epoxy-glass at 0.27 w/m-k) and doing a simplistic calculation, the theta between the drain pour and the ground plane is around 10 k/w. Not bad at all, and the capacitance is tolerable.

But then there's the thermal spreading resistance of the ground plane, not so easy to calculate. Time for copperclad and x-acto knives.

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_t4.jpg

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

The kapton tape is black in the thermal IR, allowing me to thermally image the thing. The copper is a nearly perfect mirror, so all I'd see is the room ambient reflected.

P = 1 watt total

As expected, theta through the board is right around 10 k/w. But the bottom of the board (t5 above) is about 10k or so above ambient, so net theta is around 20k/w. Tolerable.

If I want to decrease theta without adding capacitance, I think I could add pours to additional layers, specifically L4, and use lots of vias to thermally stitch to the L2 ground plane; that would reduce the thermal spreading resistance.

What would be nice would be a board material with high thermal conductivity and low dielectric constant.

FR4 0.27/4.6 = 0.058

BeO 300/6.7 = 44.8

John

Reply to
John Larkin
Loading thread data ...

[snip]

Diamond : dielectric constant around 5.5 and thermal conductivity of about

2000W/m-K

Someday...

Bob

--
== All google group posts are automatically deleted due to spam ==
Reply to
BobW

= 363. And that's the junky mixed-isotope stuff.

Yeah, that would open up possibilities.

John

Reply to
John Larkin

age

ut

As far as I known, you can go out and buy vapour-deposited diamond layers today.

formatting link

I don't know how much it costs and what the lead times are.

formatting link

but you could ask.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

Do they make Alumina PCB's? 18 W/(m*K), the dielectric constant is a bit high (9).

George H.

Reply to
George Herold

They certainly used to - they were called thick film hybrid circuits, and were usually restricted to doubled-sided.

-- Bill Sloman, Nijmegen

Reply to
Bill Sloman

Yes, but they're expensive and fragile. That dielectric constant wouldn't bother me... I'd just decrease the fet copper pour area and still be way better off thermally. There are some ceramic-filled epoxy-glass PC boards that are better thermally, but are again expensive.

It's looking like we can dissipate a watt, maybe two, on the desired footprint, if we're careful about the layout, especially spreading the hot-spot heat close to the fets.

John

Reply to
John Larkin

What about a piece of aluminum core PCB as a heat spreader? They seem to be available from some proto shops (eg. pcbpool) cheaply. The insulation is only something like 100um thick though, so capacitance might be an issue. Bergquist is the material source most of the offshore factories use, from what I've seen.

Reply to
Spehro Pefhany

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.