2W from a SOT-223 ?

Take into account that each square of 35=B5m copper foil has 70k/W thermal resistance. If your copper track to the heat sink is 4mm wide and 6mm long this would make an additional 105k/W !!! Do you need electrical insulation? Can you drill a hole in the PCB and solder the collector directly to a copper heat sink? Frank

Yes, 50 C rise above the connection point. The thermal resistance of the [ SOT-223 tab -> copper strip -> clamps to large heatsink ] path is at issue, especially the 2oz PCB copper-strip portion.

Or does Adam really mean a mean old copper strip? That could work.

SOT-223 doesn't ring immediate bells with me.

I can't believe there's no other suitable substitute. How far did you look ?

Have you looked at D-pak for power dissipation ? IR have an app note on using pcb foil area for heatsinking.

Graham

You are correct in your assumptions. I have used a SOT223 linear regulator in a similar manner although I use thermal vias to conduct the heat to a baseplate below the PCB.

Graham Holloway

Yup. Two watts sounds feasible. If you can afford a couple of square inches of copper top and bottom, and if there's some air flow, you probably don't need the heatsink at all.

Keep in mind that 1 oz copper foil has a significant thermal sheet resistance, about 70 K/W per square. So you've got to be concerned with spreading the heat laterally. Put the package in the center of a copper island and use lots of fat flooded (not thermal) vias to another island on the flip side. Parallel that with inner layers if available.

But the 26 K/W theta of the device itself seems high; that's throwing away 52 degC of junction temp right there. You might consider something with a bigger die.

See pic in a.b.s.e.

John

It's not very "thermally conductive." The best thing, by far, is classic messy silicone grease, which will squeeze down below 100 microinches if the surfaces are flat.

John

The thermal resistance from junction to "soldering point" of a PZT2222/PZT2907 SOT-223 is 26 K/W. Say I were to soldering the SOT-223 tab to a copper strip, which in then clamps to a large heatsink. Assuming Tj rise of 50°C then;

Pd_max = 50/26 = 2 Watts

Anyone else heatsinked a SOT-223 this way ? I would use a device with a larger package but there isn't anything else around with similar specs and/or availability. The PZT2222/PZT2907, replace the old metal canned TO-39 2N2219/2N2905.

regulator

of the PCB?

often quite

Siol

I don't have any specific details for you. However, I would flood the area beneath the device with copper and then add the vias. Say, 28mil holes on a staggered rows/columns pitch of 30mil would be a good start. I used a thin Berquist insulator between the underside of the PCB and a raised section of a baseplate, but I've seen other people used the gapfiller quite effectively to get the heat across several mm. I think you might need to experiment. Also bear in mind the vias will fill with solder (extra conduction) unless you take specific steps to avoid it.

Graham Holloway WPS/Accuphon Electronics (Tel/Fax 0(044)1233 662599)

How effective is this method in terms of conducting heat to the underside of the PCB?

Nowadays SMD packages are taking over even for power devices and it is often quite difficult to heatsink them :(

Gapfiller = that white thermal conduction paste?

I have been using aluminum block under the pcb, lots of vias and lots of copper on both sides. The problem with filling vias with solder is that it creates uneven surface, which adds extra thermal resistance to the aluminum block underside the pcb. It would be best to have a copper block under the pcb, a hole in the pcb and solder the part directly to the block. But it might be unhealthy for the part, its hard to get that block hot enough to solder and it cools slowly.

underside

area

on a

thin

of

effectively

unless

No, the thermally conductive rubber that comes in various thicknesses, up to about 6mm.

copper

creates uneven

underside the pcb.

and

part, its hard to

the PCB?

Suggest you take a look at this.

I think IR may have another app note too but I didn't find it instantly using the site search.

Graham

You need Motorola's AN1040 or something !

Even better - it's on my bookshelf in rev 6 of the power devices manual.

Should be compulsory reading for anyone involved with power semis.

I looked in the book. It is indeed AN1040. Does that make me a power semis nerd ?

Here's the link. Everyone should read this.

Graham

the PCB?

the site search.

This one's a keeper, very usefull, thanks!

I think even silicone grease is not very conductive unless it is squeezed out almost entirely. If two surfaces are not reasonably flat, it doesn't help much.

You're *very* welcome.

There are some snippets of information out there that are absolutely invaluable and likely never ever to be bettered.

I was lucky to find AN1040 early in my working life ! Before the internet and stuff - when you had still had to ask for data manuals. How times have changed !

Graham

much.

nerd ?

Another keeper, definitely.

Thanks.

there was a nice paper published in the 1996 High Frequency Power Conversion conference proceedings about using multilayer PCBs as heatsinks, along with some detailed measurements. thermal performance is actually pretty good, and transient behaviour is excellent.

Cheers Terry

stingy bastards with databooks. If you were a

pay, sometimes $50 per book. I used this argument to

That was not even that long ago (or am I getting old).

Nowadays google is your friend. Can't imagine any engineering work without internet anymore. Even though I am saving many PDFs and have a large collection, searching google is often faster and more convenient than searching the hard drive.