Recently, I acquired a Karlsson nixie tube clock.
I just happen to design and build Nixie Tube clocks as a sideline hobby. So why the interest with one of those?  Perhaps a thorough examination of a competitors product was in order ;).

Surprisingly small, the base was quite heavy; due in part to the one piece cast, nickel-plated base design. It looked as though it would be right at home in a Pottery Barn or Restoration hardware store, sitting next to a reproduction of a 50's style phone.

I fumbled with the two toggle switches at the back. Not intuitive, and neither were marked to indicate their function. Being a normal guy, I refuse to look at the blurblet that accompanied it for directions. Besides, it was only a single paragraph of instructions in English, Spanish, German, and Ebonics.

After a bit of fiddling, I was able to set the time. Toggle A moved left sets the hours, toggle A to the right sets the minutes. Toggle B selects either 12 ot 24 hours display mode, with the center position turning the clock off !? Who'd want an on-off switch on their clock?  Besides, you have to reset the time whenever you shut it off by accident.

Reference B outlines a 1pps (pulse-per-second) oscillator assembly in a separate daughterboard, standing off of the main board. The 1pps signal is generated using a 32Khz crystal, and divided down to achieve the desired 1pps output. What the heck were they thinking? This has to be the "Rube Goldberg" winner of the nixie clock design contest!
I connected the 1pps signal to an event counter, and recorded the number of pulses over a 24 hour period. It should have recorded 86,400 pulses. The actual number was 86,394. It lost 6 seconds in a single day. Yuck. That's nearly a minute per week error when averaged to it's favor; considering normal room temperature changes.
My ideas as to why the engineer did this:  (1) the clock was originally intended to operate from the line frequency, but they changed it mid-design. (2) it was intended for operation using the Czeisler periodic day cycle, rather than a Circadian 24 hour day cycle ;). (3) They didn't know how to use the prescaler.  The really bad part about this particular 1pps generator:  That's a 32khz crystal they're using to generate the pulse!  It is subject to significant aging effect, and thermal influence.
A more intelligent approach MCU timekeeping:
In any typical MCU design that requires precise timekeeping ability, the programmer will select an MCU crystal frequency that can be divided down to the required period using the prescaler register (consider this a divide by n function). Then an interrupt is assigned to trigger an increment of the time routine based on the output of the prescaler. This interrupt will always occur at the same time, regardless of the program variations that change the number of instruction cycles in an average program loop.

Reference C outlines the cathode driver section. Ten HV NPN transistors are used to switch the paired cathodes to ground from ten individual MCU output pins.

Reference D outlines the Anode driver section. Six intermediate voltage transistors are used to switch another six HV PNP transistors to apply +165vdc at the anode of each nixie tube.

Reference E outlines a HV power supply to generate +165vdc for the anode drivers. A transformer, bridge rectifier, and two filter capacitors are visible in the image. The Karlsson clock uses an isolated HV supply based on the "back-to-back" transformer approach. Simply described; the line voltage is reduced to 9vac using a standard "wall wart" adapter, the 9vac is fed into another step down transformer wired in reverse to achieve an isolated 120vac output. The 120vac is full wave rectified and filtered, and provides a DC output voltage of +165v. This HV supply approach is robust and reliable, but requires transient suppression to avoid glitches due to surge and noise. A single MOV or primary side damping condenser would have been sufficient. For that reason, the Karlsson clock is very susceptible to glitches and lockup caused by transients from magnetic ballasts used in fluorescent lamps.

Reference F outlines the 5 volt power supply section. An input filter capacitor, three terminal voltage regulator, and output decoupling capacitor comprise the entire circuit.