An Atomic Pendulum Clock Accurate Enough For CERN

That big grandfather clock in the library might be an impressive piece of mechanical ingenuity, and an even better example of fine cabinetry, but we’d expect that the accuracy of a pendulum timepiece would be limited to a sizable fraction of a minute per day. Unless, of course, you work at CERN and built  “the most accurate pendulum clock on the planet.”

While we’re in no position to judge [Daniel Valuch]’s claim, we’re certainly inclined to believe him, mainly because the 1950s-era Czechoslovakian pendulum clock his project was based on, the Elektročas HH3, was built specifically as a master clock for labs, power plants, and broadcast use. The pendulum of this mid-century beauty is made of the alloy invar, selected for its exceptionally low coefficient of thermal expansion. This ensures the pendulum doesn’t change length with temperature, but it still only brings the clock into the 0.1 second/day range.

Clearly that’s not good enough for a clock at CERN, the European Laboratory for Nuclear Research, where [Daniel] works as an RF engineer. With access to a 10-MHz timebase from a cesium fountain atomic clock — no less a clock than the one that’s used to define the SI second, by the way — [Daniel] looked for ways to sync the clock up to it. Now, we know what you’re thinking — he must have used some kind of PLL to give an electromagnetic “kick” to the bob to trim the pendulum’s period. Good guess on the PLL, but the trimming method is a little cruder — [Daniel] uses a stepper motor attached to the clock’s frame to pay out or retract a length of fine chain into a cardboard dish attached to the pendulum’s rod. The change in mass changes the pendulum’s center of gravity, which changes its effective length, and allows the clock to be tuned a couple of seconds per day.

It seems like [Daniel] is claiming that his chain-corrected clock won’t drift more than a second from the cesium clock for 158 million years. Again, we’ll take his word for it, but it’s a wonderfully ad hoc approach to tuning the clock, and we appreciate its simplicity.

31 thoughts on “An Atomic Pendulum Clock Accurate Enough For CERN

  1. i don’t understand. it sounds like his clock is trained to the nuclear clock, not an alternative to it? i mean, ntp running on my home pc won’t ever get more than 1 second away from the stratum 0 time source either

    i guess it *is* a hack though :)

    1. That’s how I read it. It seems to me to be an output display peripheral for a remote atomic clock rather than a standalone one-second-in-zillion-years type clock. It is nice though.

      1. ^ this

        The adornment on the end of a pendulum within grandfather clocks was purposely designed to slide up and down its pole to specifically adjust the frequency and thus the time as faster/slower.

        1. I’m currently struggling with a Bavarian coocoo clock that is running slow (minutesper day). The ornament on the pendulum is attached by a spring clamp, making it difficult to determine the distance moved from a previous setting.

  2. That’s hilarious. I love it.
    Also, I had no idea before reading that, that you could effect ppm-scale rate changes in pendulum clocks by moving them up or down just a few floors. Had to do the arithmetic myself to believe it. Cool.

  3. Pendulum grandfather clocks are pretty good, nowhere near a minute drift per day. Better ones, regulators, can be seconds per year territory or better. For comparison: The modest chronometer on my wrist right now is COSC certified which is like -4 to +6 seconds a day and the operating environment of a wristwatch is much, much harsher than a pendulum clock will ever see.

      1. I do wear it to bed, but only because I’m lazy. It’s operating temperature is all over the map. I guess it is moderated a little by my wrist itself but the body and mechanism is certainly quite different in, say 115F heat vs like 35 degrees and that is just in air, not swimming or diving or anything. Also to be COSC compliant it is timed in different orientations as well. All of which are things a pendulum clock doesn’t have to deal with which is why they are pretty darn accurate. Basically I was questioning why a pendulum clock, devices that centuries ago held rates of seconds per year, would have an error rate “a sizeable fraction of a minute” per day when a cheaper device (my watch) operating under comparatively brutal conditions is within a second or two a day.

    1. I’ve got a mechanical movement clock that is accurate to a minute every two weeks. It’s better, or worse, depending upon the temperature, as the wooden pendulum stick moves. I was going to replace it with carbon fiber, which has a much lower temp coefficient, but find this accurate enough for me.
      I also have a stratum 1 NTP server in house in case I need to really know what time it is.

      1. I have a digital Timex wristwatch in the drawer, that (before its strap mounts broke) drifted around 15 seconds between the dates of switching to/from Daylight Savings.

  4. Enslaving a clock to another more accurate standard does not make the clock itself more accurate. And as long as that clock is enslaved, it won’t drift a second, not even in “158 million years” from that other clock.

    I also see some weirdness in the logged data. The first graph with “Pendulum period error [sec]” has some noise, which is understandable, but it also has some systematic deviation and those all have nearly the same amplitude. What is going on there?

    It also feels like sacrilege to put a paper basket and some chain on such a vintage clock, but it all looks reversible so no real harm done.

    When I read something about “one second in who knows how many years” of accuracy, I always have to think back to: Those clocks from the ’70-ties are being sold on Ebay for some time now, and they rarely work at all. It’s something with the cesium leaking out of the tube or some recombination, and when that happens, the only remedy is to replace the tube. (If they are even made anymore).

    I also remember having read about the most accurate pendulum clocks ever made. They have multiple pendulums swinging in vacuum, and they don’t get excited during most of the time. The “master pendulum” is just free swinging (maybe for half an hour or so) and when it looses amplitude and needs some boost, then it does receive it, but the phase difference is then compared with and corrected for with another pendulum, which is freely swinging at that time. I think that was the height of mechanical clock making before electronics took over.

    1. Obviously, a little bit of exaggeration. I guess reading through the article makes it clear. Hobby project, where one learns about PLL dynamics, system modelling, using noisy signals, design of control loops for non-linear systems, optimise the dynamics to profit the best from performance of each oscillator and many other. Leraning by doing.

      Nevertheless, locking two oscillators makes the locked one inherit some of the properties of the master one. In this case the aim was to transfer the long term stability. I might not be here in 158 million years to check if it had drifted by the mentioned one second. But I am sure it will be fine :)

      The clock was measured and observed free running for few months. The observed pulses of same amplitude up and down every few minutes is the kick, when the automatic winding mechanism kicks the weight up. It is there exactly one period and the very next period the same pulse in negative direction follows. The net sum is very close to zero. Observing the system shows you many interesting phenomena.

    2. we are talking different league.
      The Cs clocks used for the Keating experiment were the Caesium beam tube type. These are standard telecom equipment now, e.g.

      The one disciplining the Active hydrogen maser, which is already orders of magnitude better than Cs beam tube is a Caesium fountain type clock. The NPL link is already in the text, a video from NIST
      btw. few photos of this particular one

      Precision at the level of limits of nature is a very interesting field.

  5. “… we’d expect that the accuracy of a pendulum timepiece would be limited to a sizable fraction of a minute per day.”

    Erm, s/minute/second/

    From The Royal Society, 1911 “The measurement of time by pendulum clocks is one of the most accurate measures in daily use, but even with the best existing clocks it is not uncommon to find that the rate may vary from time to time by as much as half a second a day, roughly about 1 part in 150,000.”

  6. I use a plc with a GPS time base and a magnet on the pendulum and an electro magnet fixed in the clock case. The two come to a 5mm gap at each swing of the pendulum. By using the electro magnet, I can attract the pendulum magnet or repel it to make it swing faster or slower. I have accuracy of less than a second a day. It will correct every hour.

  7. We had an IBM pendulum clock at the South Pole Station (90 deg South) that was originally also for synchronizing other clocks. Indium pendulum also. It had relay outputs that were used to trigger the C3 Ionosonde for measuring the height of the ionosphere. The previous NBS (National Bureau of Standards) guy had plotted the timing inaccuracy and correlated it with the atmospheric pressure data he got from the Weather Bureau guys. So he experimented with putting whole, halves, and quarters of 6-32 brass nuts on the top of the bob using tweezers and was able to keep the clock within one second per day as known from the WWV radio signals we could usually get unless the reception was really bad.

    The clock’s cabinet was plain but beautifully made; I always wanted one like it. Years later, after I was married I looked them up on the internet and found you could get one for US $1000 in good condition. I haven’t looked since then and I didn’t have that kind of spare cash around then so I never got one.

  8. I take exception to the statement disparaging pendulum clock accuracy, too. (“Sizable fraction of a minute per day.”) I have a good quality but off-the-shelf grandfather clock I wind and set every week, and it is never off more than 10 seconds per week. It gains or loses depending on the ambient temperature, e.g., in summer it is a few seconds slow and in winter it is a few seconds fast (per week).

    As somewhat of a clock and watch fanatic, like one of the other posters I have a stratum one clock at home.
    Mine is a GPS-aided Rubidium atomic clock that I use for my ultimate home time reference. It is spec’d with a variance of 1msec per month when GPS is lost (which is never).

  9. Can someone suggest a circuit for syncing my grandfather clock to the one second pulses from a quartz clock? The circuits I’ve seen all seem to be self sustaining oscillators that use the pendulum to determine the time period. I’m looking for an external pulse to “correct” pendulum error.

    1. The company was/still is called Elektročas – Pragotron. It was an electric time keeping equipment manufacturer in the former Czechoslovakia, with products exported to the whole Eastern block.
      They still exist, but obviously changed the product portfolio a bit
      Search for the company name and you will find very nice pieces.

  10. This comment section is overflowing with debate about the accuracy of this and other time-keeping systems.

    I just want the stepper motor that lasts 158 million years :D

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.