The problem is the tiny doesn't seem to be sinking current properly. For testing I switched to a 20 PU on the breadboard and 100 ohm load resistors. At 3v supply this should give approximately 30 mA per pin, well under the rated 40 mA per pin and 200 mA total. However with the loads connected only pin 6 sinks any significant current, and appears to be higher impedance than it should be (1.6v across the resistor for 16 mA). The other pins seem to be very high impedance, they drop to ~7 mV (70 microamps) when the loads are connected. With the load resistors disconnected they function normally (digitalWrite HIGH = supply voltage - junction drop).
I'm using pinMode [this was a problem] (OUTPUT, N) [/this was a problem, should be (N, OUTPUT)] to set up pin N, and digital writes for on/off. I've been able to drive transistors before using an exactly similar setup, although admittedly these have much lower current requirements.
Anybody know what's going on here?
EDIT: Turns out I had accidentally swapped the order of OUTPUT and N in the pin mode command. Not sure why the compiler had no problems with this, but I am at least getting useful outputs now.
I'm gonna leave this up out of curiosity/etc, because I'm still getting higher impedance than I'd have expected on the outputs. With the 100 ohm load resistor, I'm only getting about 16 mA (expect closer to 25). Supply voltage from the CR2032 is ~2.48v under load.
Based on the graph on the datasheet (figure 22-19, PG 182), at 25 c ambient (currently 20) I should expect a voltage drop from VCC to output somewhere on the lower end between 400 and 600 mV when supplying 15 mA, but my readings at 16 mA above represent a voltage drop of almost 900 mV across the Tiny. Additionally the LEDs drop 2v at 10 mA, for which you'd expect approximately 300 mV drop across the tiny according to the graph, but mine is dropping almost 500 mV. I'm not sure if this is something in my setup or if the characteristic curves on the datasheet simply aren't that accurate for production models.