# Problem with a nixie clock

I bought this nixie clock kit and it is partially working. One of the issues I'm having is that the 6th hour does not light up so if the clock is at 06:00:00 to 06:59:59 the 6 will not light up for the duration of the hour. At one point after fiddling around with the positioning of the tubes, it is the 9th hour that has this problem. The 6 (and the 9 at one point) works in the minutes and seconds. No it is not the tubes.

The other issue is that the time starts lagging by a second every other day.

I've followed the instructions and built the kit but I'm not too well-versed in electrical components. What exactly is controlling the time functionalities? I am thinking it is software related and thought it was the MOSFETS IC2 and IC3 that controlled it but apparently all the MOSFETS do is regulate the voltage... I may have dropped the IC2 MOSFET at one point and thought that it may have been damaged which caused this abnormal behaviour but since it only controls the voltage, that can't be it...

I have contacted the seller but he is not offering much help. If any one have experience with nixie clocks or this kit in particular, any help/pointers/advice would be greatly appreciated.

• The time is controlled by the 32.768kHz crystal attached to the secondary oscillator of the PIC microcontroller. Hard to know why it's lagging without hands on investigation. With the tubes, you may have assembled the connector incorrectly for your tube (in the manual there are a few variants covered) or the 6 filament pin is not being driven (check soldering and continuity from pin to driver with power off, OR check voltage carefully with multimeter with power on when 6 should be lit) – Oli Glaser Aug 30 '12 at 7:38
• Nixies in that type of clock are multiplexed, so if it's not the tube, it could only be a connection problem on that socket. – Axeman Aug 30 '12 at 8:04
• 1 second every other day equates to about 3 minutes per year or about 6ppm frequency error. Not bad for an uncalibrated oscillator. – MikeJ-UK Aug 30 '12 at 12:29
• @MikeJ-UK - good point, wasn't thinking last night, a 5.8ppm error sounds about right. One way of improving things would be to use the GPS option to provide regular corrections (it looks as if they sell a compatible module) Unfortunately you can't adjust the firmware clock settings otherwise you could use e.g. a more accurate external oscillator (TCXO) instead of the crystal. – Oli Glaser Aug 30 '12 at 19:30
• From what I can see on the schematic, the tubes are directly driven from the 5812 (IC3) and the bipolar transistors Q8, Q9 and Q10 so the MOSFET should have nothing to do with it. If you measure whilst on, and the voltage is low or 0V, then it's either a bad connection or a firmware issue. Since it's multiplexed, it's possible it is a firmware issue as I would have thought you would see the same problem with the other tubes if it's a bad connection. Hard to know exactly without an oscilloscope to probe the signals. – Oli Glaser Aug 31 '12 at 10:00

The manual says that booster generates 170V DC, when datasheet for IC HV5812 specced for 90V absolute max. The driver IC is killed by design. There is no means in booster schematics to limit the voltage to any predictable value (no zener of voltage feedback).

Old russian equivalent driver K155id1 is BJT based and specced to 60V max, but has a voltage limiting zener per each output 0.5 mA max, so it could survive better. Possibly the replacement with contemporary device was done with mistake of not adding a single 80V zener to booster.

• There is some attempt to limit the voltage going to the HV5812 - Look at the Vpp pin - it's connected to a divider (390K||100K to ground). The actual HV rail only seems to be running the blanking grids in the nixies, and the AM/PM neon bulbs. The shift-register is run off the divided-down Vpp. – Connor Wolf Sep 1 '12 at 11:08
• Also, there is a feedback mechanism. Look at R3, VR1, and R4. It's a voltage divider, with the wiper of the pot connected to AN0 on the PIC. As such, there is feedback from the HV voltage rail. – Connor Wolf Sep 1 '12 at 11:09
• On the whole, it's a pretty horribly drawn schematic. I mean seriously, did the person who drew this's CAD package not even have ground symbols? – Connor Wolf Sep 1 '12 at 11:12
• Correction, Nixie-tubes don't have blanking grids (I was thinking about VFDs). Apparently the circuit is even jankier. Basically, the digit electrodes in the nixie tube are either at Gnd, or Vpp. When they're at Gnd, there is ~170V between the digit and the anode, and you get a standard neon discharge. When the digit is switched to the voltage at Vpp (~35V, assuming a 390K||100K divider, which gives only 135V across the tube), there is insufficient voltage between the digit cathode and the anode for the nixie to strike, to the numeral stays off. – Connor Wolf Sep 1 '12 at 11:20
• It seems that you need ~160V+ to strike a neon discharge, and by positively biasing the cathode by a smaller voltage (the 35V Vpp), the digit electrode stays off. It's actually a pretty clever way of using a low-voltage VFD controller to control HV nixie tubes, even if it is horribly documented. – Connor Wolf Sep 1 '12 at 11:24

For problem 1) check for solder opens or shorts For Problem 2) most 32.768 KHz resonantors are 7 or 15pF load rating which requires two 14 or two 30 pF caps.

Since 1 second in 2 days lag is approx -5ppm and these parts usually have an initial tolerance of +/- 20ppm, you are within spec.

To speed up the clock you need to reduce either C6 or C5. You can file off the top of the part to shave off capacitance on small ceramic parts that are not metallized layers. I would expect the board adds 5pF load so consider dropping C6 from 30 pF to 25 pF or using a dremmel sanding disk to remove 25% of the top of the part carefully. They are brittle.So you May have to buy parts or consider a trimcap instead so you can adjust from 15~25pF with a fixed 15 pf and a variable 10pF for example. With a frequency counter you can measure to 1ppm in a minute with other methods using difference Hz in a few seconds.