I'm designing this RTD/resistor simulator and it's convenient to use an LM8261 opamp in the output stage. The 8261 doesn't really current or thermal limit (that's part of its charm) but I can limit power dissipation by adding a resistor in series with the output. We have the SOT23-5 parts in stock.
But the SOT23 8261 is spec'd at 325 K/w, and I'd like to push it as much as possible. So I was thinking that the V- lead is probably the substrate, and there might be a copper paddle to that pin, not just another wire bond.
I soldered all 5 pins down to a chunk of FR4 and then started sanding the top with a Dremel. This is what I got:
What's interesting is that the paddle seems to be just under the epoxy surface, so the chip must be mounted upside down. It does look like pin 2, V-, is the paddle and has a pretty good thermal path from the chip to the pin. So we can add some topside copper pours and thermal vias to a big -12 volt internal pour, maybe some more copper on the bottom side, and probably reduce theta a bunch.