One of the big worries about the new software-mux design is supplying enough current to the LEDs. The previous design was going to send 12 mA by using pins on the controller through series resistance directly to the anodes. The big worry with that is that it wouldn't be bright enough. When multiplexing, you have to take into account that the duty cycle of each digit is quite low - 1/8 - 12.5% in this case. Because of that, it's actually kosher to send quite a bit more current during each time slice. But the XMega I/O pins are only rated for 15 mA. I knew there were driver chips that could be used to switch higher currents, but my searches failed me.

Today I came across this tweet. And there was much rejoicing.

The TBD62783AFNG can be used to supply the anodes with positive voltage under software control, and the TBD62083AFNG can be used to switch the cathodes.

Unlike the schematic in the tweet, I'm going to see if I can live without series resistors. Instead, I'm going to use an LDO to supply a reduced regulated voltage to the Vcc pin of the anode driver. As long as the voltage is regulated, then there should be no brightness variations with more or fewer segments lit.

The new software mux design replaces the ~$7 MAX6951 with about $3 worth of parts - the two driver chips and an LDO. It's not the absolute slam dunk cost savings I had thought, but still pretty good. The big question is whether or not the increased BOM would kill my on assembly costs and whether the 300x reduced refresh rate will matter.