I may have got it wrong about Analog going out of the business. At least I don't think it is on purpose. They just don't have complete documentation and can't seem to answer questions about timing on the built in DACs.
Might want to consider one with very fast wake-up that can be placed into sleep mode a lot.
Yeah, use the 80C51 architecture since they won't go away :-))
But seriously, main things are, regardless of uC type:
a. Common and full ground plane. Full VCC plane if you have the luxury of a 4+ layer board.
b. Tuck the crystal really close to XIN and XOUT, mount 1M or whatever starter resistor farther away. Not the other way around. You wouldn't believe how often that is done wrong.
c. Bypass caps really close to VCC pins but to a guy like you that'll be obvious anyhow (but maybe not to other readers). I've often seen lengthy skinny traces to bypass caps, they just left the CAD at the 10mil default. So if someone else does the layout that's a point to watch.
d. Unused pins: Either as inputs with pull-down, or outputs with "low" written to their port register bit.
e. AC terminate longer traces that can switch during signal acquisition. Or bury them if multi-layer.
The NXP parts and a few others seem to be able to idle between interrupts and get going again fairy fast. This will save some power.
I started with a Silabs part. It doesn't quite have enough through put.
I will have internal layers so I can put a Vcc plain under the micro and a v-analog under the analog parts. I had already decided on this much.
I think I will be running on an external clock that is piped in on a connector. I may use a HC14 or the like locally to clean it up. Multiple cards will share one clock so that they all agree on exactly what 1Hz is.
I often put extra bypassed on the back too.
It was a false impression I got from the fact that they never answered a simple question about the part. I got a responding e-mail from the robot and then over a week of nothing and then a note asking me to resend the question because of a technical problem with the web site and then another long delay. I bugged them a gain and finally got a non-answer from a human. They appear not to be going out of the processor business on purpose. If they don't fix their problems with answering questions, however, the result may be the same.
All Billion of 'em? ;-) I had a rep bring yet another ARM wannabe company's FAE around today. "That's nice; get in line". These guys were selling low power. Ok, but the rest of the stuff takes a lot more than the DSP now.