Oops

formatting link

I woke up in the middle of the night and realized that I'd not plopped enough vias around voltage regulator U5. It's weird that my brain works that way, checking things in background. Unfortunately, the boards are already ordered.

It's a SOT223 and might dissipate three watts. Each via could be roughly 30K/W but they aren't strictly in parallel, and all land nearby on the layer 2 ground pour, which will be a hot spot. I should have made the topside patch bigger and had more vias. Or helped the regulator with some resistors or something.

I guess I could have production fill the vias with solder; that would cool the regulator a little.

Gotta go to Molly Stones now. They do a dynamite corned beef and cabbage on St Patricks Day. One bad thing about this town is its scarcity of decent corned beef.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin
Loading thread data ...

sot223 and 3 watts, shame on you!

--
 Thanks, 
    - Win
Reply to
Winfield Hill

With a good copper pour, a SOT89 is fine at 3 watts. There's the same silicon inside my SOT223 as is inside the DPAK version.

The TI data sheet claims that theta junction-to-board of the SOT223 is actually better than the TO-263, 17 to 19 K/W respectively. I just need to keep the tab cool.

I could kluge a zener, or a couple of resistors, in series with the input to share the dissipation. Ugly.

Maybe I can reduce the load a bit.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Too much iron in that combo; better be sure you don't end up looking like Popeye.

--
This message may be freely reproduced without limit or charge only via  
the Usenet protocol. Reproduction in whole or part through other  
 Click to see the full signature
Reply to
Cursitor Doom

(...)

"Understanding Thermal Characteristic of SOT-223 Package"

Got any vertical space? Mount the regulator vertically off the PCB. Hand solder terminals 1,2, and 3 to their corresponding pads. Make an L bracket from copper sheet. The lower end of the bracket is soldered to ground pad 4. The vertical end of the bracket is soldered to ground tab of the regulator.

--
Jeff Liebermann     jeffl@cruzio.com 
150 Felker St #D    http://www.LearnByDestroying.com 
 Click to see the full signature
Reply to
Jeff Liebermann

More like >"MISUnderstanding Thermal Characteristic of SOT-223 Package"

That would be a lot of work and might not reduce the temperature. I can kluge a big 5-volt TVS (which is actually about a 7 volt zener) in series with the input and slurp about a watt out of the regulator. It would not be very visible on this board.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

the DPAK version.

Same silicon dies maybe, but one wonders after seeing how they completely fail to get the heat out of the die into its metal tab. There are specially-designed heat sinks for sot-223 and DPak parts, but the former consistently fail to get you up to the 3W level. 1.5 to 2W, maybe.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

"Completely fail" is not literally true.

The TI data sheet for their LM317/SOT223 specs 16.9 K/W junction to PCB. So it would be OK at 3 watts if the tab were well heat sunk... which I didn't do.

Interestingly, the part is only 3.6 K/W junction-to-top. So junction-to-bottom must be significant.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

Well heat sunk means much more than a square inch of copper. Even space-consuming heat sinks mounted on top of sot-223 and DPak packages, e.g. Thermalloy 7106DG or 573100, fail to handle 3 watts. Fairchild's AN1028 measurements show 1in^2 copper failing to reach the 3-watt region, even without any safety margin.

--
 Thanks, 
    - Win
Reply to
Winfield Hill

We try to use a topside copper patch, but then via down to a plane pour, and sometimes to another bottom-side copper patch. The trick is to get close-in to the part with vias, and then spread the heat laterally, to avoid a local hot spot, and use multiple copper layers for spreading. 1 oz copper is about 70 K/W per square, so lateral heat spreading is critical. Assuming only one layer of copper available for cooling, theta approaches some asymptotic value as patch area increases, limited by spreading.

My topside patch here extends under the part, too. Some goo or epoxy under the part would help.

On the typical 1 oz PCB, a small part will make a local hot spot, top and bottom. On this particular board, it's mounted above a machined, water-cooled block with a small air gap; I have to option of adding patches of gap-pad between the bottom of the board and the baseplate, which will slurp heat away from hot spots. That potentially changes everything.

This is an analogous situation. Some topside heat on a fairly small patch

formatting link

formatting link

and a hot spot on the opposite side of the board.

formatting link

If the bottom were cooled, theta would be a lot less.

The real problem with most surface-mount parts is not the theta of the package itself, but the small contact patch area.

--
John Larkin         Highland Technology, Inc 

lunatic fringe electronics
 Click to see the full signature
Reply to
John Larkin

An LM317/SOT223, soldered down to a big slab of 2 oz copperclad FR4, about 20 square inches, has a theta from the top of the part (approx junction temp?) to the tab of 7 K/W.

It thermally shuts down at 5.8 watts, about 150C on top. At that point, the nearby PCB is around 50C. The effective theta, top to ambient, is about 22 K/W.

If I stick the bottom of the board to an aluminum sheet with a sticky gap-pad, the PCB drops to about ambient and that top-theta drops to

18.5. The heat still has to pass from top to bottom through epoxy-glass, no vias here.

So this part is good for about 5 watts *if* you can heat sink the tab well. The typical PCB footprint doesn't do that.

Needs more vias! My real PCB will be here tomorrow, so I can test that.

--
John Larkin         Highland Technology, Inc 
picosecond timing   precision measurement  
 Click to see the full signature
Reply to
John Larkin

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.