Slimline Nixie Clocks

Everyone needs to build a Nixie clock at some point. It’s a fantastic learning opportunity; not only do you get to play around with high voltages and tooobs, but there’s also the joy of sourcing obsolete components and figuring out the mechanical side of electronic design as well. [wouterdevinck] recently took up the challenge of building a Nixie clock. Instead of building a clock with a huge base, garish RGB LEDs, and other unnecessary accouterments, [wouter] is building a minimalist clock. It’s slimline, and a work of art.

The circuit for this Nixie clock is more or less what you would expect for a neon display project designed in the last few years. The microcontroller is an ATMega328, with a Maxim DS3231 real time clock providing the time. The tubes are standard Russian IN-14 Nixies with two IN-3 neon bulbs for the colons. The drivers are two HV5622 high voltage shift registers, and the power supply is a standard, off-the-shelf DC to DC module that converts 5 V from a USB connector into the 170 V DC the tubes require.

The trick here is the design. The electronics for this clock were designed to fit in a thin base crafted out of sheets of bamboo plywood. The base is a stackup of three 3.2mm thick sheets of plywood and a single 1.6 mm piece that is machined on a small desktop CNC.

Discounting the wristwatch, this is one of the thinnest Nixie clocks we’ve ever seen and looks absolutely fantastic. You can check out the video of the clock in action below, or peruse the circuit design and code for the clock here.

26 thoughts on “Slimline Nixie Clocks

  1. Interesting how it skips a second at 0:13 (from 18:59:31 to 18:59:33). I guess he’s not updating the display often enough, e.g. maybe he updates it once a second and using an external clock crystal that’s not in sync with the CPU clock.

    1. Having looked at the code/circuit a bit, this is indeed the case, but what’s worse is that the updating happens every 1000ms, but it’s doing this timing in software, rather than using a timer and an interrupt, and worse, it’s adding 20ms between every loop, so you’ll see how this skipping will happen once a minute or so. And indeed, right after the slot cycle effect, it skips again from 19:00:05 to 19:00:07, and then again from 19:00:35 to 19:00:37. If you make a clock, make sure to get the clock essentials correct :)

      1. Maybe he intended to build a clock which is intended to conveys Plato’s Theory of Forms. Maybe the guy fumbled.

        That said, having these kinds of imperfections in a project based on old technology fits the theme and makes for a more interesting piece than a ‘perfect’ clock, especially since it does not affect total the accuracy of the actual time keeping.

        1. Well, unless these imperfections are intended, but in this case it’s just silly to insert 20ms into the loop for no good reason: this makes the issue much worse. And yes the nixies are old technology, but not the software. In fact, if you did use old discrete components to make this clock, it wouldn’t have this issue.

        2. If you go back 45 years, there wouldn’t be such “imperfections”. It would be a lot bigger, though.

          Back then, the clock would be a string of counters with decoders feeding the Nixie tubes. Either it would use a count derived from the 60/50Hz line, or a crystal oscillator divided down. Since many clock builders at the time were interested in accuracy, they probably used a crystal oscillator.

          Clock ICs arrived shortly after, making it all more compact, but so did LED 7 segment readouts, so probably few Nixie clocks were made with the clock ICs. But even then, the clock ICs were just a bunch of dividers and decoders, using one single clock input.

          One thing to consider, back then a digital clock was a big thing. But if someone built a Nixie clock, likely it was because other readouts hadn’t arrived, or Nixies were what someone found cheap as “surplus”. There wasn’t the same fetish over Nixie tubes that exists now.

          Of course it changed rapidly. The TTL changed to the more integrated clock ICs and then every hobby outlet sold some sort of clock kit, LED readouts, the IC, a few parts and maybe a circuit board and/or power transformer. That too morphed after a few years, National moving from clock ICs to clock modules, even less to do though at least the modules had various options so they were flexible. And a bit later, the craze was over, digital clocks in everything.


        3. If this were a project built the old way of doing it, you’d never, ever see skipped digits, and the accuracy would be either as good as the power company’s accuracy, or as good as a single crystal oscillator source divided down by sequential counters can be.

          So basically… No. The “theme” does not fit. Poor programming does not equate to any notion that old tech wasn’t as advanced, so it must be imperfect somehow. It’s just a bit of a fumble.

          1. Remarking that discrete logic clocks back in the day would not have done this is missing the point. The point was that older technology often is the best they could do or uses approximations or shortcuts and that this sometimes leads to peculiar solutions and quirks. Even though the skipped digit here is obviously an omission, it still has romantic merit.

            Besides, pretending that older technology wasn’t imperfect is as much romanticizing things as the reverse. In cases like vinyl records, tube amplifiers and the TB-303, those imperfections are a big part of the attraction. Rather than a perfect result every time, the permutations of what should be the same output holds people’s attention. Considering nixie clocks are conversation pieces to start with, having a small quirk – one that does not destroy the functionality of the project – makes it an interesting piece for a longer period of time than a perfect clock. Even though clocks are supposed to be as predictable as possible, our brains actually consider perfect predictability incredibly boring. Almost as boring as some of the commenters on Hackaday.

        4. No, in my opinion it does not fit the theme. In the time of this old technology there were no microcontrollers with software glitches. There were only counters made of flipflops. The display would for sure update synchronous to the primary timing source, be it a crystal or the mains frequency. Nobody would have invested the resources (tubes/transistors) to make to sets of registers. :-)

    2. You can see the omission on his schematic. All of these external RTCs provide an interrupt to trigger a synchronous update of whatever is connected to it, but it’s not connected. Chances are he’s free-running his microcontroller and polling once a second. Not a good idea if you’re displaying seconds, but you can easily get away with it if you’re displaying only the minutes.

      1. The polling is possible, but then poll more often than once every 1020ms (e.g. at least twice a second), and use a hardware timer so that if it does skip over, the next time this will happen will take a very long time (more than a day given the precision of the average crystals).

    3. Maybe there’s a bug in the nixie tube? He should swap tubes… j
      (Just kidding of course). This is just a great example of poor programming. You should update the display when it’s supposed to change. eg:
      if (LastDisplay != Display)
      LastDisplay = Display;

  2. If you think your basic un-ovened un-aged crystal is more accurate for timing than the 60hz line (suitably filtered) then you need to learn a lot more about how it was done back in the day, why SSB took so long to catch on after WWII, and so on. Crystals weren’t magic by a long shot. The watch crystals of the right cut to keep decent time when “ovened” on your wrist did not yet exist.

    60 hz tends to be really good on average due to the requirement to sync the world together. There are tiny variations over short terms, but long term it’s linked to what is now NIST.
    You’re not going to beat that at home.

    1. Interestingly enough I’ve heard that when utilities use power line carrier it can sometimes interfere with clocks and other devices which use the 60 Hz for reference. Basically, these devices detect the zero crossing to keep time and power line carrier causes a dip in the voltage to represent a ‘1’ bit. If the dip is low enough, a zero crossing is erroneously detected and the device sees too many zero crossings.

    2. I’m saying that when digital clock making started as a hobbyist thing, they were putting consideration into the timebase. People who were already interested in a very accurate frequency reference, remember it was also a period when frequency counters became within the price range of hobbyists. I remember more than one article where the author built a really good frequency standard and used it as a reference for both a clock and frequency counter.

      They weren’t using low frequency crystals. They’d start with at least 1MHz if not 5 or 10, and at least some of them knew how to get things right. Even in the clock IC age, a time base, at least for backup used a color subcarrier crystal, no not a perfect choice, and that National IC that divided it down to a useful frequency.

      At least in the ham magazines, there were plenty of articles about getting precise frequency and keeping it there. It was a sub-hobby for some, and at least back then out in the ham magazines along with the simple projects. There were lots of articles about crystal ovens, and even burying reference oscillators to keep it at a constant.

      SSB took time for other reasons, in part because it was different, and also because it was perceived as more complicated. But one thing that propelled it was cheap crystals left over from WWII (in FT-243 holders, that were screwed together), and stable VFOs made from WWII surplus, either a Command set transmitter or a frequency meter, both having really good variable capacitors and reduction drives. Crystals didn’t fit into the equation, at least not stability wise.

      There was a lot more about crystal oscillators out there back then than you seem to think. Not everyone was interested, but one didn’t have to have access to professional publications.


      1. Funny, my collection of IRE journals from that time says emphatically in the SSB issue that the FT 243 xtals weren’t good enough for SSB, and experience confirms it (I’m 63…).
        Of course, there were other problems in the military – even with prop planes the doppler would mess up SSB, which is why Costas proposed DSB no carrier and a bunch of workaround with pilot signals were proposed (but none ever widely adopted).
        Some of this is in Fredrick Terman’s Radio Engineer’s handbook of WWII vintage.

        Yes, I have the ARRL handbooks back to then and the magazines too. Stable VFO’s and
        “pullable” xtal oscillators were used for SSB then. No one could just dial in a frequency and go.

        You seem to be dating hobby clock building from your own beginnings…it goes back a bit further.

        At the time of nixies, those good xtal oscillators weren’t around unless you did the oven, aging and calibrate to NBS thing – and you had to take into account that the skip time from CO might vary with time somewhat…
        Be careful what you assume about what someone else knows…or how long they’ve been at this game.

        Yes, a drop in the line (for whatever reason) can mess up a cycle counter. My first bread machine in my off-grid solar setup here didn’t like the inverter waveform either (it keeps fair time but no where near as good as the grid AC). I wound up using a tuned circuit to filter the 60 hz (which would ring for awhile if it dipped) and counting the output of that, which I why I mentioned “suitably filtered” above.

  3. It isn’t a work of art because it looks generic, or is being inspired by a hatred of RGB LEDs a STEAM thing this week? As far as clock hacks go it is rather lame because of buzzword insufficiency., where is the IoT part? ESP* + NTP? MQTT YouTube counter? Weather report feed? Gesture sensing? You can’t even use it as a headphone amplifier or detect radiation with it. It could be doing all of those things, without changing the way it looked at all. Now that is what I’s call a real STEAM’r.

  4. I have a German-made Chronotronix nixie tube clock. Mine has a feature which can display the “date” periodically at each 30second interval. I am wondering if this was a ‘place holder” for that purpose in this design, which somehow got forgotten about…
    The more useful [IMHO] display item at home would be a nixie thermometer (w/a barometer + indoor/outdoor temp).
    Some crafty guy (named Yuri) out of Ukraine is selling a non-kit versions (fully assembled but no case) for $75USD on ebay.

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.