Stan, I was thinking along the lines of a simple RC circuit and a transistor or FET switch. I don't want to use a uP pin for this function, that seems a waste. Also, if I use a uP pin, I'd have to have track power monitoring in the uP to know when the track power dies, since the battery is continuing to power the uP, I wouldn't know the track power was gone without some sensing.
If you're considering a 20-pin version, why not a 28? Assuming you have these professionally assembled should be no penalty to put a 28-SS pkg vs. a 20-SO pkg. You'd have to fuss with the hand-built prototypes of course but should free up the debug interface. I think the conventional battery circuit is still better done with the uC (yes, requiring a track-voltage-presence input pin and a battery-enable output pin). The circuit I showed earlier suffers from a modest bias current which draws on the battery when track voltage is present. This would vanish if the uC only turned on the battery when it needs to. The incremental cost of a larger uC part is probably a wash with the components and real-estate to implement the discrete circuit "logic" to implement a zero bias current battery switch when track-voltage is available. It would also handle that no-man's land situation of the battery powering the entire system at low track voltage.
OTOH if you've come up with a low-cost RC circuit to switch the battery I'd like to see it!
