Although I have suggested that the counters may be arranged in a 1-2-2'-4 configuration rather than 1-2-4-8, this has not been verified by more circuit tracing. But I don't have to do this because I don't intend to use the counters anyway.

So the question is where to interpose an extra circuit to drive the nixies from a WiFi MCU, as planned? Should it be before the latches, or after? After is easier, as the shift register GPIO will already come with a latch, and there is no need to power the MC778s to keep them working as latches. Instead of cutting traces, probably the easiest way to connect to the driver transistors is to desolder the MC778s from the board, using some combination of soldering iron, solder sucker, solder wick and heat gun. This also has the advantage that the pin holes exposed by the removal of the chips are concentrated in particular areas corresponding to the digits, so the flying leads to an auxiliary PCB will be neater.

As mentioned the driver scheme uses 6 inputs to encode 0-9. The even/odd input is actually 2 lines, so 7 lines are required in all. We could invert the even input to get the odd input, but a shift register GPIO will have lines to spare so we can avoid an inverter.

There is actually one more pass transistor below the even/odd transistors, which seems to be for display blanking. Easiest solution would be to simply bridge the C-E pins of that.

I haven't discussed what to do with the symbol nixie and the 0/1 nixie. At this point, the easiest solution is to just unplug those nixies, to get a 4 digit display, one spare digit nixie, and one symbol nixie to ponder what to do with.

Another goal was to substitute a more efficient power supply. Since the RTL chips don't have to be powered as the driver transistors only need base drive current, I no longer need the 3.6V logic supply. The 200V nixie supply can be replaced by a boost converter. Then I don't need to supply any power to the transformer and all the other circuits on the board won't draw current. The whole box and the WiFi MCU could be powered by a wall wart SMPS. That also solves the problem of the dangerous exposed mains terminals at the back. I could remove the RS232 socket to provide an entry hole for the interface cable.

My ideas on the shift register interface and the power supply will be in the next log.