All of the test Charliewatch boards fab'ed and assembled by Aisler work perfectly (other than things that I broke, like not noticing that this MSP430 requires load caps for the crystal). For the next rev I want to squeeze the LEDs closer together to shrink the watch face a bit, which might require violating the ERC.
I've started porting the older Charliewatch animations to the bi-color version. I should rethink some of them now that we can distinguish the different hands.
Now with a 3D printed case and 22mm nato watch strap. Need to get a 35mm sapphire and maybe a wider band.
Somehow I had overlooked this totally unobvious disclaimer buried on page 316 of TI's 676 page data sheet for this chip.
Turns out the secret is that the gpio config on the msp430fr2433 is unchangeable except for port1 pins 4-7 until you set the magic LOCKLPM5 bit in the PM5CTL0 register. Now all the LEDs work!
Seems that the cpu browns out when some of the charlieplexed pins are driven, which makes me suspect there might be solder bridges under the QFN package.
Something is wrong: only three or four io lines seem to be functional, although they all pass the multimeter beep test