Heatsinking electrically 'hot' nodes, SMD

I was thinking just stuffing a few 1/2 watt resistors, or something else with beefy copper leads.

The difference in thermal conductivity to a via is ~6-to-1 for a .6mm (24mil) lead.

Cheers, James Arthur

How about a 0.025" square post header? Double-row, if you've got the mania.

9C/watt per pin, 0.062" pcb. That's the post and its 0.040" plated-thru via, not including any solder.

Cheers, James

Have you looked into Bergquist thermal pads? They have a ton of different materials with different thermal conductances. Frequently used to cool power switching devices. They also have gap filler mushy materials to conduct lower amounts of heat from multiple devices with slightly different heights above the board.

Jon

It looks likely those will be in the mix. It's a double-sided board, so my aim is to drag the heat to a flat spot cleared of components on the bottom layer, then couple that to the aluminum case.

A rectangular piece of aluminum bar is my preferred method of filling the gap, but we've got to see if that's feasible mechanically.

I calculate this 20x20x4mm pad's thermal resistance as 1.7C/W. Not awful, but contrary to my earlier posting to John, I'm realizing every 1.7C/W would halve the operating life of the product. That battle's worth fighting hard.

Cheers, James Arthur

Here's one corner of a mongrel test board that we'll fab soon. It has a bunch of goofy circuits.

There are three resistor arrays, with various amounts of copper-pour heat sinking. The lower array has pours on layers 1 and 4.

I included a test array for the Kelvin resistor connection that we discussed recently.

It's not too late to add something!

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

link:

Measure the tempco of an FR4 cap! I recall something like +900 PPM/K. I've had to take that into account on some LC oscillators.

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Woe be it for me it burden your board, but if I were doing one right now I'd make a square inch of copper on top and bottom, fitted with a couple watts' worth of resistors for heating on top, maybe thermistors on both sides, plus a 0.025 header here or there for the thermal circuit.

I might do that just for fun with ye olde Dremel and Mr. Carbide. (I ought to get some decent SMD thermistors--much easier than swinging my sole thermocouple from side-to-side.) (I've got a cheapo IR too but don't trust it. S'pose I could use the Kapton tape trick and test it while I'm at it.)

Cheers, James

I have one unused patch 0.7 x 0.9 inches; I could do something with that. It's in the corner of the board; I could leabve the L2 ground plane, or cut it away.

Yeah, it would be easy to just shear and Dremel some FR4.

We stock surface-mount platinum RTDs, which are real accurate. They look like 1206 or 0805 resistors. Some DVMs will read them directly as temperature.

A good (small FOV) IR thermometer might work to measure a square inch of copper, assuming you make it black in the thermal IR. The cheapies are good for measuring the temp of, say, a large wall. A proper imager will show you the hot spots.

This is 1 watt, 30 mil double-side copperclad, no vias (Dremeled):

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

I have made sandwiches of gap-pad and aluminum, to fill up big gaps. Pad, alum, pad makes good thermal contact top and bottom.

Incidentally, you can generally let thru-hole leads, like header pins, just poke into the gap-pad stuff, as long as they don't short to the case or something.

1.7 c is half life? The conventional wisdom is 10C, unless you are near some failure point. 17C would reduce life by 1024:1!

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

1.7C/W * ~6W = +10oC.

James

Oh, c/w not c. Sorree.

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

"John Larkin" wrote in message news: snipped-for-privacy@4ax.com...

I invented it last week. Do I win a patent? ;-)

1/8" thick copper plate, soldered into a net-size route in the board. Think of it as an extremely thickly plated "poor man's [solder-in] via".

Those are RF transistors soldered on the top; the top of the bar is obviously hard to see under the shorting straps and transistors, but that's what they are stuck to.

As an aside:

These transistors seem to be rated very roughly similar to 2N7002s (~2 ohm, 2A, 60V). But with roughly 18dB more power dissipation... and 30dB more price. Although the capacitances are lower, and being LDMOS and rated for a little wattage (20W), I imagine the die size is a bit larger. But not by 30 freaking dB... the profit margins on RF amps must be ludicrous. And these are just the baby silicon variety.

Tim

--
Seven Transistor Labs 
Electrical Engineering Consultation 
Website: http://seventransistorlabs.com

Den mandag den 9. marts 2015 kl. 23.41.05 UTC+1 skrev Tim Williams:

They look quite similar to some we use for a preamp in a PA, afair they were quite picky about soldering temperature, hand soldering without killing them was not easy.

Do they still work?

-Lasse

No worries. The customer wants to stack multiple boards in the box. That makes thermal design and minimum dissipation both critical.

(The thing is already close to 96% efficient through two cascaded switchers, after a few tweaks. Lots of Linear Tech parts.)

I might just use the 0.025" header trick, plus a few more. I'm thinking stacked copper spacers may be essential for moving heat out of the stack, if it's even possible.

Cheers, James Arthur

Fans!

Maybe you can transfer heat from the edge of the boards to the side of the box. You might even use conduction-cooling card guides, on one edge or both. They are available, either spring types or cam action.

Is this price sensitive? With LTC parts, I assume not! Cam guides or machined heat transfer widgets might get pricey.

Maybe an aluminum right-angle thing, bolted to each board and to the side of the case?

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

In a sealed box--that actually helps. :-)

Doesn't seem to be price sensitive.

The case is a hard limit--it has to dissipate the whole load. They've offered fins, an offer gladly accepted.

Cheers, James

I entered the wrong thickness. That fixed, the spreadsheet's quite good.

Cheers, James

Hey, one could embed a not-especially-accurate temperature sensor interior to a PCB, using the FR4 capacitance tempco.

I guess a copper trace RTD would be a better idea.

--

John Larkin         Highland Technology, Inc 
picosecond timing   laser drivers and controllers 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com