It all started in January 2002. I had been looking at the web photos of Jeff Thomas’ Nixie watch, amazed that a complete four-digit nixie clock could be made small enough to be strapped to one’s wrist and powered for months by some dinky little batteries. This was too good to be true!
Then the engineer in me got to thinking about how to make it small enough to actually wear every day, not just to parties as a novelty item. I got out my defunct Fluke 8100B nixie DVM and took out the B5870 tubes. (Don’t worry, I put them back in!) They seemed just right for making into a very small clock.
I found that I could fit two tubes and a camera battery in a pretty small space, about 1-1/2 inches square. Four tubes would make a pocket watch sized clock. A microcontroller like the 68HC705J1A that I used for my scope clock looked like a good fit.
I discussed this in email with Jeff, and he thought it was a fine idea. He provided some information about what his customers wanted to see in a nixie watch, and I referred to that list when doing my design. I seem to have gotten everything except the year-long battery lifetime accomplished, although it took me a few years to squeeze it all in there.
In March I made a prototype PC board nixie clock with four tubes, a little LT1308A switching power supply, the HC705 and a 74141 Nixie driver chip. It worked, but the CPU wouldn’t run at 32 KHz like I wanted for power savings (I found later that only one special unobtanium version of their HC705 CPU runs at 32 KHz). I ended up switching to a PIC16F872 chip, which had a good low-power 32 KHz operating mode. I also switched to TD62083 high-voltage cathode driver chips. Both these parts were available in tiny TSSOP packages suitable for a wristwatch.
I built a breadboard two-digit watch circuit using DIP parts and the power supply cannibalized from the four-digit clock prototype. I spent a fair amount of time developing the software and the power control hardware. The power supply ended up being current-regulated rather than voltage regulated, which saves power since no anode voltage-dropping resistors are needed. Some software tricks were used to get the multiplexed display to be flicker-free with a measly 8192 instructions per second.