Ryan Brooks's 6502 Nixie Clock

So, I wanted a Nixie clock and I really wanted to design the hardware and program it myself.  Most of the kits out there are based around a PIC or some other new, easy to use microcontroller.  I thought it would be much more interesting to make a clock powered by an old school computer. UPDATED Oct-06

The result is a 6502-powered Nixie clock.  I started this project about two years ago when large tubes (Z5680M in this case) were still available.   Since then, any large tube is 5x the price it was -- if you can find them.


The design consists of three boards: 

6502 controller

    8k of SRAM and clock, Dallas DS1742 timekeeper RAM
    8k of EPROM: 2764
    I/O: 6522 VIA
    Misc logic: 74LS00, 74LS138, DS1233 reset...

Shift board and HV drivers

    Daisy chain of 24 bits of shift registers: 74LS595
    BCD drivers: Russian 74141-alikes

HV power supply

    Switch-mode: MC34063A - A design from the NeoNixie group

The display consists of four East German Z5680 (beautiful) tubes for hours and minutes, and two Russian IN-18 tubes for seconds.   Each tube anode is fed via a 10kohm resistor and a pot mounted on the back of the clock chassis.   The individual digit cathodes are then wired into the output of the 74141s that decode a BCD input from the 74LS595s.  The 595s (serial shift registers with latches) are fed from three output lines on the 6522, which I use as serial clock, data, and output latch.   This allows you to use three lines of I/O to control the 24 bits necessary for 6 BCD digits.

Two of the boards, the 6502 controller and the shift board, were PCBs that I designed in Protel 99.   You can get the entire design database for these here, as a zip file.  The HV board was made with point-to-point wiring on a protoboard.  With these schematics and circuit board layouts with Gerber files, you are welcome to duplicate my design.   You can find the 6502 tasm source and S-record object code here.

Pictures of the construction are here. 

Outputs - Port B

0,1,2 - Data, Clock and Strobe
3 - Alarm annucator   (future: is 4k or so enough for a small sample?)
4,5 -neat LEDs (scrapped, kept everything neon)
6 -HV modulation (unused currently)

Inputs - Port A

0,1 - Hour +/-
2,3 - Minute +/-
4 - Second Reset

All inputs are active low and are connected to momentary switches (DPST); No hardware debounce. 
DC Power supply

All three boards are driven from the same low-voltage DC power supply.   The 120VAC input goes through an IEC connector/filter to a fuse holder and SPST switch.  From there, it enters a 10VAC transformer (sized for 2 amps) and then a bridge rectifier.   The ripply DC is filtered with a large cap and then made available on solder lugs.   This raw ~10VDC goes to the HV board and to the two logic boards through a standard 7805 for the digital circuits.

What I learned

• Check each digit on your tubes when you buy them.  One of my tubes had two digits sort of twisted around each other -- hard to see without careful inspection.

• Don't use a similar looking probe for your oscilloscope and DVM.  I was using the DVM to test current on each tube, and then the computer stopped responding -- I thought I had crashed the computer.  I picked up what I thought was the osc probe to check the activity on the address bus, and spark!  170v!   Toasted almost everything on the 6502 board, including the decoupling caps.  Crap.

• Be very organized with your digit wiring; use lots of cable ties or something.  6 digits is 120 solder joints on 60 wires.  Argh.

• .100" plexi is too thin.

• Don't be afraid to make your own Nixie sockets.    I stood my tubes up on my flatbed scanner, arranged them with a ruler and then scanned.   Imported to Photoshop, did a levels to make the pins contrasty and printed the output 1:1.   Glued the paper onto my plexi, drilled all the holes, dropped in pin-sockets w/ CA glue and bam! Instant Nixie Tube Clock Socket!

• A 35mA draw on a  170v switchmode power supply makes quite a big magnetic field;  enough to crash the computer.  Use short wires and shield the inductor.

• The DS1742 is very picky about timing.   Set the read bit and then wait quite a long time before really trying to get data from the clock registers.  If you don't wait, you always get 0s -- not garbage like the data sheet implies.

• In tasm, ROR does not imply ROR the accumulator like it did in the old days.  I think it was doing a ROR $0 (zero page) instead, which is stupid.   You'd never do that.  Just "ROR" in the C64/Apple II days meant ROR the accumulator, but in tasm, you must say ROR A.  (I think this also goes for LSR).   Until I figured this out I was pulling my hair out since my code seemed fine.

• Those sexy 0.1uF polarized tantulum decoupling caps that people like will blow if you look at them wrong.   Something is to be said for just using ceramic disk caps, but the PCBs weren't designed for that.

Schematics





Thanks to everyone on Neonixie Yahoo! group and 6502.org!  Awesome people!   - Ryan K. Brooks