Unusual steampunk-themed clock with orange glowing cold cathode display and automatic brightness control based on ARM microcontroller

Amateurs around the world are building electronic clock using “cold cathode display” or ”nixie” tubes for displaying digits. Such clock are usually called nixie clock. If you search for “nixie clock” pictures you’ll see how popular they are. Interesting nixie clock feature is that they’re using obsolete indicators (for aesthetic purpose of course) but modern chips. I think this trend is charming and I’m happy to participate in it.

I have six IN12-A nixie tubes. You may find their properties at some reference.

Important radio valves’ requirement is high anode voltage. My IN12-A tubes are shining reliably at 170 V. Of course that’s far from thousands of volts for some CRT, but such voltage can kill typical digital component several times at once.

If you’re building such clock first part of your troubles will be produced by digital circuit (flashing your MCU for example) and second part by cranky high voltage generator. Such power source is responsible for stepping up voltage from e.g. 12 V to desirable 170 V. You should be ready to:

1. find somewhere coil with specified inductivity (hundreds of microhenry) and saturation current (approximately 1 A),
2. buy expensive field-effect transistors which will tend to burn out,
3. look for redundant and missing contacts between the parts carefully, the latter are risky in such schemes too,
4. debug the scheme bewaring of high voltage strokes continuously.

The following Sunny Clock scheme (ru) fragment based on MC34063AD chip was the starter for my discoveries. It takes 12 V at P12V node and generates high anode voltage at HV node. At first I’ve tried to reproduce this module as is but it turned out that transistor Q1 is pretty hot without any rational reason. It looks like most of amateurs are OK with such behavior because this heating isn’t significant but I wasn’t satisfied with this compromise.

Thanks to some article I’ve discovered that MC34063AD is able to only pull up Q1 transistor’s gate to +12 V but it can’t pull the gate down. FETs have significant gate’s capacity so you need resistor R12 to discharge and “close” it. Such solution isn’t perfect at all because during transition the transistor works in linear mode and heats up. The same article gave me good solution for this trouble. Additional PNP transistor (Q2 on the scheme below) closes the FET very good without any heating.

To manipulate terrifying hundreds of volts using weak microcontroller you’re able to use a bunch of high voltage transistors or specialized K155ID1 chip as I did. It also contains binary decoder which will keep some MCU’s ports for another tasks.

The photo shows decoder probing process. The breadboard is powered with +180 V for the tube and +5 V for the chip. To specify desired binary code you need to pull down “zero” decoder’s pins. TTL logic implies that all unconnected pins will be pulled up by themselves.

You may see modified Sunny Clock scheme below. I didn’t test it.

This schematic has significant flaws.

1. You are unable to see that using only the scheme but I assumed to distribute three anode optocouplers this way: 12:31:23—such distribution means that you need to turn high voltage on all cathodes of minutes (for example) to turn them off. The site I mentioned above says that turning tubes off this way isn’t appropriate:

   Do not blank, by turning all cathodes off. Especially, when using cathode drivers like a 74141 (or its Russian equivalent). If you attempt this with these devices, they will see over ~100V, and will conduct thru more than one cathode at a time. This will look similar to ghosting. These devices are leaky, so at least one cathode should be ON at all times.

   To be able to turn off desired pair of the digits, one must distribute the optocouplers this way: 11:22:33—that means you can blank minutes just by turning anode voltage with optocoupler number 2. That may be useful for setup mode to blink with configurable number: hours, minutes or seconds.

2. The scheme uses obsolete ATMega8 MCU. I want to change it to STM32L152RBT6 because I’m studying ARM microcontrollers.

3. MCU could take the responsibility of MC34063AD chip. Every microcontroller has wide abilities of generating PWM signals and ADC for feedback loop for high voltage. This makes possible not only reducing of chips’ count but also change anode voltage wisely depending on lighting conditions.

I hope that I’ll extend the article with prototype’s photos in the future.