I decided to get down to basics and prototype some of the power sections of an exploratory project (hand-held embedded Linux-based diagnostic fob), rather than take on too many sections at one time. The imminent death of the Intel Edison (goddamn, don't get me started...) means I'm probably going to work with the Pi Compute Module. I already built an extremely crude prototype from a Pi and an Adafruit button/display hat quite awhile back, which was successfully used in the field. So staying with the Pi is in some ways less work.
On the other hand, bye-bye easy battery management (grrrr....).
The end result is this:
Most of the circuit is a buck/boost converter, based on a suggested design and layout from TI's Webench tools. I spec'ed 5V at 1.5A, which is why I believe so much of the board is occupied by thermals (and perhaps accommodating high currents, so I ran with it in other sections of the board where space permitted).
On the lower right is a standard Lipo charger circuit. On the bottom is a latched-dual-MOSFET soft-power-button circuit. Press for power, press-and-hold for 3 seconds to release power.
Combine all three and I'm one step closer to providing a sane way to power the Pi on battery.
This is the first board where I used riveted vias. Per my other post, I do the riveting first, then let the ink provide the conductivity. That worked pretty well here. I tapped each side of the rivet a few extra times to make them smooth and even inset them a bit. I kept the nozzle height at a safe .14mm, which given I was only using TSSOP-level features wasn't a big drawback. I only had to go back over a few traces to make sure the ink was bridging the rivet properly.
I can't emphasize enough how nice it is to not have to hand-solder all the vias. Yes, my luck with hand-soldering these ink traces sucks, but it's also one less manual thing to deal with.