Tell Time Like It’s 1960 With This All-Transistor Digital Clock

When you’ve got time on your hands, doing something the hard way can be therapeutic. Not that the present situation and the abundance of free time that many are experiencing has anything to do with [Leo Fernekes] all-transistor digital clock build, which he started a year ago with his students. But if you’ve got time to burn, this might be a good way to do it.

[Leo] says one of his design goals with this clock was to do it with the technology commercially available in 1960, which means relying completely on discrete components. And he and his students managed to do just that, with the exception of the seven-segment displays, which were built from the LED filaments from some modern light bulbs. Everything else, though, is as old school as it gets, and really underscores all the complexity that gets abstracted away from timekeeping with modern chips. The video below covers each module in detail, from the Schmitt trigger that cleans up the 50-Hz line frequency to the ring counters and diode matrices used to drive the display. We found the analog stair step dividers used to bring the line frequency down to a more usable pulse train particularly interesting. That clever bit of engineering saved 10 transistors over what would be required for traditional flip-flop dividers.

There’s a lot to learn from this design, and the execution is great too – we’re suckers for Manhattan-style builds, of course. Hats off to [Leo] and his lucky students on a great build.

44 thoughts on “Tell Time Like It’s 1960 With This All-Transistor Digital Clock

  1. Would they not have had crystal oscillators (thesedays bought as 2 pin metal shielded oval shaped packages) back in the 1960s? Seems better than relying on 50Hz mains which could suffer a powercut.

    1. Very nice clock, one thing about using the mains is that in comparison to modern clocks it is not that accurate as the frequency varies over the day due to load on the network. That said it was quite acceptable back in the day and they do try to keep the long term average approx 50hz where ever possible. Another thing is that a 50hz timebase requires a shorter divider chain than 32khz a big thing if you are making it from transistors !

      1. Today the frequency is not universally reliable anymore because renewable power requires the utilities to maintain relaxed frequency limits. The phase differences in an AC synchronous grid define the direction of power flow. Areas that have intermittent power production maintain higher frequency to force the power to be transmitted elsewhere, and these other areas then have to maintain relaxed frequency limits to stop the grid from going into cascade failure every time there’s a surge of power coming their way.

        This causes intermittent frequency deviations, and in weaker grids the difference can be all the way to 47…53Hz which is a significant error up to 1½ hours over 24 hours. Six hours of over-production from e.g. wind turbines can put your clocks ahead by up to 20 minutes. These errors are so large that they take weeks and months to correct, and the next power surge is going to come before the error goes away, so the grid synchronous clocks are basically running fast all the time.

        1. At least where I am mains frequency is pretty darn precise. I have an old 1930s Telechron electric shop clock in my garage and never gains a minute between daylight savings adjustments.

  2. This was a great video, really enjoyed the deep understanding of a basic clock that we would all take for granted otherwise. And the digits! They’re excellent.

  3. Those dividing circuits are so cool. Imagine dividing a periodic signal by 10 with only a couple of transistors! This is the type of project I search for when skimming through HAD posts!

    One question: How can only 1 of the minute and / or hour shift-counters-registers be assured active at any given moment. I missed that if it was part of the design.

        1. Err @RW ver 0.0.1, @karl enter and I are talking about any one of the given 3 shift register that drive the individual 7 segment digits. (Not the divide by 50 counter circuit.) We do not see a way to assure there is one and only one SCR active at any given moment. Specifically, we do not see where, upon power up, that only one SCR is on and all other SCRs are off. Maybe there is some trickery in picking the resistors and capacitors for only one of the SCR circuits. But I don’t think @Leo Fernekes talked about it or showed it in the video. Which we both though was, er, an exception to a really well though design.

          1. I’m still a bit hazy about where you think there’s a problem, yes, the 7 segment drivers are still counters, and pulled low by Q11 to reset. (18:00 approx)

            Theoretically during the minute, deciminute or hour interval you could introduce a logic high to one of the latches and have the display wrong until the next Q11 reset cycle.

          2. I had this concern the first time I encountered ring counters, many years ago. I THINK the way this works, is that when an SCR switches on, it drops the voltage feeding all of the SCRs through their 120 ohm resistors. In fact, on power up, none of the SCRs will be on, but when the first pulse arrives, all of them will start to turn on, but one will inevitably turn on faster, reducing the drive to all of the others. In some ring counters, I’ve seen one resistor that was a different value for one of the switching devices, so that it is assured to be the first one to trigger. I don’t see that here, but in any case, the circuit is inherently stable only with one SCR on.

          3. Also, even if by some quirk, the circuit does come on with two or more SCRs on at the same time, with every incoming pulse, the SCR that turns on quickest will tend to prevent others from turning on, so in the process of setting the clock by applying extra pulses to the minutes, decaminutes, or hours counters, with every pulse you reduce the possibility of more than one SCR being active.

            I also may be wrong about none of the SCRs being on at power-on. I’m not sure about the biasing, but it could be that when it powers up, one will trigger spontaneously. But as I described above, as one turns on, it turns all of the others off.

  4. “[Leo] says one of his design goals with this clock was to do it with the technology commercially available in 1960”. 1960 or 1980? If 1960 then that would be germanium, not silicon, transistors. Resistors would be carbon-composition. Capacitors would be ceramic disc, bumble-bee paper-oil, dry electrolytic etc (no metal-film polyester). Also, the LED did not exist until 1962. Cold-cathode, Neon, or incandescent bulbs would be the display technology common to the era. 1960’s-era parts are obtainable today, but getting rare and expensive. Still you see the occasional nixie-tube project here from time to time.

    1. First commercial silicon was along in 1958 https://en.wikipedia.org/wiki/2N696 but they really exploded onto the scene with the likes of the 2N2222 in 1962. Bear in mind a lot of hobbyists and some pros took quite a while to get familiar with new components and were designing with old faithfuls for a decade after they were replaced. Thus you get what are thought of as 1980s micros using mostly 70s chips, 1990s micros, 80s chips. Also many tended to create with what was plentiful in surplus of the day, which was of course not the latest and greatest but someone’s old and superfluous stock. Mullard actually switched to BCxxx and ACxxx naming in 1960 so all the “60s” OC44, OC71 germanium parts you’re remembering were 50s parts.

  5. I think the shift counters could be corrupted to have multiple states on…
    But this should not happen if it starts with just one state on…
    So a corollary to your question is I was wondering how the Hour and minute shift counter get started from cold start with just one active state…

  6. I had to work on a circuit that was built in 1962, that used similar stair-step circuits to divide down the frequency of the master clock for a RADAR system, to provide 10- and 50-mile range marks for the display ‘scopes. The problem was that the circuit was not adjustable, but depended not only on precision components, but also the typical C-B capacitance of the transistor, so when I had to replace one of these transistors, there was no modern (at the time, ~1980) transistor that had a high enough capacitance, resulting in a countdown of 8 when it should have been 10, or something like that. I added some capacitance to make up for this, and got it to work, but was told that this was not an acceptable solution. Oh well!

  7. “All-Transistor”??

    Why do I see LEDs, plain diodes, resistors and capacitors amongst other components then? ;-)

    Definitely a nicely built all-discrete clock though.

  8. Mains frequency is an *exceptionally* accurate time source, being regulated by the network far more tightly than any crystal oscillator. Mains powered clocks almost always use it. Also, it’s very simple to extract and use.

    Also, how is a crystal going to help when the power is shut off? There’s no battery backup to maintain the time anyway!

    1. You are wrong, mains frequency varies +-0.2Hz, at least that’s variation i see regularly.
      50ppm crystal gives ~800 times better stability.
      That’s why mains syncronized clocks are. 1-2minutes off every month.

      1. The long-term accuracy of the mains frequency is far better than a typical crystal, because they actually tweak it to keep clocks on time long term. Short-term, yes, the crystal is better.

      2. Short term they can be off by more than 0.2Hz, but the industry has agreements with government to hold the long term stability to very high tolerance. Mechanical mains clocks (i.e. those with AC synchronous motors) can’t track the changes that closely, but electronic clocks can. Most modern electronic clocks do however use a quartz crystal for the short term stability advantage.

      3. Tell that to my nixie clock that I’ve been running for 15+ years on two different power grids (US, then EU). I typically do not reset the clock except when the time changes twice a year for daylight savings. I rarely notice a time difference of more than a minute, with the exception of the EU power generation row that happened a few years ago – there the time difference was noticeable, and could be calculated back to the day when generation was slowed down.

        After all, it’s a clock that sits on my desk. A few minutes doesn’t bother me..

      4. The number of periods of the mains voltage in a year will be roughly the same every year, making the long term stability extremely good.
        Besides that, all mains sync’d clocks will have exactly the same variation, which can make the influence of the variation less intrusive.

      5. Even THIS seems like ancient history now, but until NTP became a thing, most PCs I encountered had clocks that were off by many minutes, due to cumulative error and, let’s face it, really cheap 32.768 kHz crystals in uncalibrated circuits.

      1. That was a really cool day.. I left work the night before and glanced at my clock, which is *never* slow.. I thought, “wtf?”. Decided I’d look at it the next day.

        Then I read the HaD article the next day and was amazed that I could back-calculate the time difference..

  9. I have to admit, I’m a bit dismayed by the negative comments on this post.

    This is hands down one of the best videos I’ve seen in years, chock full of in-depth knowledge and know-how, and not a microcontroller in sight.

    1. While some critics can apply, and we can even learn with that, most of what I read today are people trying to build a reputation based on criticism, and in the faith that negative words and unrelated subjects will prove to the community that they are smarter than the authors of the hacks.
      If you allow me to give an advice, ignore the show-offers, specially when a hack of yours is the subject of the article.

    2. That’s a comment section for you. It’s depressing, ain’t it?

      By the way, does anyone else have a more specific term for this build style than “Manhattan?” I’ve seen plenty of Manhattan-style circuits but nothing quite structured this way. Just lots of scored copper-clad, but not bothering to divvy it up and glue down pads with super glue or anything. Just leaving it scored but whole and slathering it in solder. Any more detail on this particular kind of construction? It’s fascinating.

Leave a Reply

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