The power supply, as you might guess, is nasty and complex. If I remember correctly, the deflection circuitry runs on about 300 volts and the CRT anode runs on about 800. This is pretty low anode voltage but I got decent focus and brightness. The X and Y D-to-A outputs are direct-coupled to the triode grids (!) so bias on the differential amplifiers is set by moving the entire logic power supply up and down with respect to the tube supplies. I think the software is intersting in itself - for instance I digitized my handwritten numbers to generate the vectors, so the clock is uniquely mine in that way too. To minimize screen burn, the image sweeps across the screen throughout the day. You don't notice it moving, but it does.
On the back of the clock is a center-off momentary switch for setting. When you push it one way, an underline moves under the number to be changed. When you push it the other way, the number increments. Since each digit is set individually, you have to click five, not fifty times, to set the time to xx:50. After setting, the screen clears and you are presented with an underlined 12 or 24 to choose the time display mode. The seconds start running only after the final click to exit set mode (when the underline disappears) to allow synchronizing the seconds. I haven't seen this scheme used for setting before, but it works surprisingly well and allows you to set the clock in just a few seconds, with no irritation. I think most clocks set badly (the FAST and SLOW buttons on commercial alarm clocks are the worst design ever.)
For handwired stuff I never draw out a schematic; I just build from my head. So other than sharing the general idea like this, I unfortunately don't have much else. I certainly don't have anything like a full schematic for this clock. I'm trying to get better about that, but old habits die hard.
Its neat display.
Its guts showing. Close quarters!
F8 microcontroller (lower right) with its piggyback ROM containing the program. This clock was built in '01 and has been running constantly since. The ultra-long-life (mercury added) tubes show no sign of darkening/aging. Since the WWVB signal provides only the day of the year (1-366), the clock has to calculate month/date. Since March 1 is sometimes day 60 and sometimes day 61 (leap years), the date algorithm has to know about leap years. The very complex code in this clock was not fully tested until I woke up and checked it on the morning of Mar 1 '04! Happily, it worked correctly.