Hacking Grandfather Clock Accuracy While It’s Still Ticking

grandfather-clock-tweaking

[Keith] got his hands on a few grandfather clocks. Apparently the price tag is greatly reduced if you are able to get them second-hand. The mechanical timepieces require weekly winding, which is a good thing since you’ll also need to correct the time at least that often. But this drift got [Keith] thinking about improving the accuracy of these clocks. He figured out a high-tech way to adjust the timepiece while it’s ticking.

The first thing he needed was a source of super-accurate time. He could have used a temperature compensated RTC chip, but instead went the more traditional route of using the frequency of mains power as a reference. The next part of the puzzle is to figure out how to both monitor the grandfather clock and make small tweaks to its pendulum.

The answer is magnets. By adding a magnet to the bottom of the pendulum, and adjusting the proximity of a metal plate positioned below it, he can speed up or slow down the ticking. The addition of a hall effect sensor lets the Arduino measure the rate of each swing and calculate the accuracy compared to the high voltage frequency reference.

36 thoughts on “Hacking Grandfather Clock Accuracy While It’s Still Ticking

      1. Well yeah. Plus on modern operating systems you can sync the system clock to an NTP server somewhere and always be super accurate. Right now the cable box, phone and computer all have the exact same time. How cool is that.

    1. Have you ever seen a 60Hz output from your PSU? Your PC doesn’t care if it’s powered by US 60Hz, EU 50Hz, 90V DC (most switching PSUs can use that) or 12V lead acid battery. That and proper RTC provides way better resolution.

        1. Yeah, but it’s compensated over any longer length of time to provide proper time, because huge amounts of our infrastructure rely on it to keep time. At least it is here in the US.

  1. Not that it matters to me,but doesn’t this take some of the character of using a old style clock? WTSHTF with no grid we are going have to make do with dad watching the Sundial & yelling to mom when the shadow with the hour mark to adjust the clock. Surely some shortwave time stations will remain on th air. Now is the time to build fixed tuned crystal radio receivers to be prepared. ;)

    1. Interesting. I didn’t consider an eddy current braking effect. But would damping make the pendulum slower? I think it might just change the Q of the system. But in any case, the pendulum seems to go faster when the plate is up, and slower when the plate is down (at least that’s how the program works) so braking wouldn’t seem to be the dominant effect. The nice thing about a feedback system is you don’t really need to understand the full function.

  2. Did you have a modern “clone” of a grandfather clock? Either there was something wrong with it, or whoever made it knows nothing about timekeeping. *Real* grandfather clocks are amazingly accurate. I have one in my dining room, and the twice-a-year daylight savings time correction is all it ever needs. Master clockmakers were geniuses at including corrections for temperature, altitude, air pressure, age, wear, and every other factor they could think of.

    A quote from the Scientific American article: “Without controls the clock kept time to within a few seconds per day for many years in spite of variations in temperature and barometric pressure and the degradation of the oil in the bearings of the clock.”

    The AC power line frequency has short-term variations, but is continually corrected to NBS standards. Clocks using it as a time reference aren’t off by more than a second a year, unless the clock itself is flawed.

    Now… what would be interesting is to take an inherently *inaccurate* clock, and use the micro to force it to be accurate anyway! A little research will reveal dozens of clock design strategies that just didn’t make it due to accuracy problems. Imagine a sand-filled hourglass that actually keeps accurate time! :-)

    1. It may be I just didn’t spend enough time adjusting the pendulum length–I hate to stop the pendulum and turn that little nut. I agree that at least the nicer of my two clock is very stable. It is making a correction for one second errors only a couple times a day and that may be the control system “hunting”. BTW, my original concept was to turn this thing off once the clock was tuned. The plate will stay put, and the tuning is done. I will try that in a few days and see how far the clock drifts when it is fully tuned and post it on my site. I am still guessing a couple of seconds a day, which is nothing to sniff at for something with no electronics..

  3. Using a magnet sensor would NOT adjust the time of the swing. The swing time is dependant upon the pendulums LENGTH. Should the pendulum be kicked by a pulse at regular intervals, that COULD change the timing. take a look at http://www.hvtesla.com/masters/po36_intro.html for a full accurate detail on how a pendulum works for a clock, how to maintain the inertia of the pendulum and how, to regulate the swing of the pendulum. The hack is a neat solution, but not a good solution.

    KS

    1. Wow. Thanks for all the comments guys/gals. You are right about the length being one way to adjust the pendulum period. And of course pendulum weight doesn’t change anything (within reason). Gravity is another term in the pendulum equations, however, and the magnet hacks the gravity term. Maybe a better way of thinking about it is that the magnet increases the downward force on the pendulum bob without increasing inertial mass. Of course, maybe I am wrong about the theory, but it is working, somehow,

      Keith.

    2. I’m struggling to understand what you’re saying; are you implying that the project doesn’t actually work? The pendulum swing time is depending on many factors other than the length. Gravitational acceleration, air resistance, bearing friction, etc. Exerting drag on the pendulum at some part of the swing would certainly modify the period.

    3. You missed a point here – it depends on the pendulum length only if the end of pendulum is treated as a point mass (like a metal ball hanging on a rope). In real case the axle has some weight and by moving some of the weight away from the pivot, the “appeared” length would be increased even if the total length of the pendulum remains unchanged.

  4. If a long case/grandfather clock needs adjusting weekly, it hasn’t been set up right. The pendulum length (screw adjusted) determines the time keeping in co-operation with gravity. My clock keeps awesome time, a spectacular feat of precision engineering!

  5. The time correction of the 60 cycles is being abandoned and will no longer provide reliable daily time. For more details on its previous accuracy, see:
    http://www.bmumford.com/clocks/60cycle

    I made my own “pendulum governor” in 2002. You can read about it here:
    http://www.bmumford.com/mset/PendGov.pdf
    It will run on four “AA” batteries for about 6 months.

    This is a perennial project, though a good mechanical clock needs no such help. Still, if you have a nice antique that was never a great timekeeper, it’s nice to bring it up to modern standards.

  6. i wonder how hard it would be to make a device small enough to be hidden on the backside of the pendulum, with batteries and all, that adjusts the screw that moves the pendulum weight up and down. I have an old clock, that just needs ½ a turn of the screw to atjust it a few secs per day, so maybe a precise servo could turn it back and forth until it was precise. the period of the pendulum could probably be measured by a accelerometer.
    If something like this could be made, you wouldn’t need the big visible parts in thos project

    1. That’s a good idea. If the adjustment nut turns easily, you could have a solenoid behind it that moves a rubber flap out to brush the nut as the pendulum swings past. Each time it brushes, it rotates the nut a fraction of a turn. By timing the solenoid right, it can brush the nut as the pendulum swings by left (slower), or right (faster), or not at all (no change).

      Another possibility: A Vetinari clock. Set the clock to run fast. Use a solenoid to hold the pendulum momentarily at the peak of its swing. Do this just often enough to make it keep perfect time. It would also “bug” people who were listening to it. Tick, tock, tick… tock, tick tock. :-)

  7. Nice work Keith!!

    For the other “arm-chair” hackers, why don’t you put your theory of better options to the test and do a hack like this and right it up for others to comment on.

    1. Now that sounds like an interesting idea. I have a web page or two sitting idle so it could be fun, and I can take the armchair critiques away from this page to a page where they no longer detract from the fine work made here.

  8. I work on old clocks as a hobby & have been for over 30 years. And the assumption that a grandfather clock will drift by as much at 5% per week just to come up with a project is pure BS. Grandfather clocks, because of their long pendulum, are / were some of the most accurate clocks made, because of their long pendulum.

    And so you know, as a pendulum type clock is moved twords or away from the equator the pendulum swing changes, requiring adjustment to make the clock accurate again. If his grandfather clock was fast or slow he only needed to adjust the swing weight on the pendulum to compensate for the move from the clock’s previous location.

    1. The underlying reason for the required adjustment isn’t the distance from the equator but the variation in acceleration due to gravity. This affects the period of the pendulum swing. Earth’s gravitational pull isn’t uniform across the world; as a rough guide, the more rock underneath you the higher the gravitational pull.

  9. 1. This method of time adjustment has been done for hundreds of years. “Master Clocks” using this method existed AFAIK in the 19th century, maybe even earlier.

    2. Mains frequency is notoriously unreliable when it comes to timekeeping. The mains frequency shifts quite a bit with load. In-fact the the power companies watch the shifting frequency to determine when load increases and decreases. Ideally the power companies are supposed to increase and decrease the line frequency over time to cancel out the error, but rarely does this work well enough to render a mains frequency reliable enough for daily human timekeeping applications. Where I live (and in most of the World), the power company is totally irresponsible with maintaining the line frequency – so it is totally useless for keeping time. I’m much better off using a 50ppm crystal reference.

  10. Any mechanical clock that is monitored and rate adjusted by computer is effectively an electric clock.
    To an observer, the peceived accuracy of a grandfather clock is a relative thing….especially if it doesn’t have a seconds indicating dial.
    Prior to quartz clocks, pendulum clocks reigned supreme until the advent of electronics – roughly mid 20th century – and every issue refered to above, directly and indirectly, in the above comments, some of which are misunderstandings, are covered in the excellent book “Electrical timekeeping” by Frank Hope-Jones, the designer and builder of the Synchronome clocks that ruled the roost from 1900 until they were made absolete by electronics.

  11. Just found this interesting thread. I agree with Hugozair, a pendulum that is regulated by an electronic device is essentially an electronic clock.
    I built a “free pendulum” clock with electronic readout, but the timekeeping is solely regulated by the natural rate of the pendulum. It has one moving part—the pendulum. To keep it swinging, it’s given a tiny nudge every two seconds by a small electromagnet. The timing of this nudge is triggered by the pendulum itself as it swings past the electromagnet. A microprocessor tallies the swings and keeps track of the time. It also drives an analog clock movement that is synchronized by PLL to the pendulum.
    The microprocessor’s oscillator (quartz crystal controlled) is used as a reference to measure the accuracy and stability of the pendulum. The pendulum has remained very stable at approximately +/-0.1 second per day for the past six months. The error is logged in the micro every 10 minutes. You can see it in action in this short video on youtube… https://youtu.be/UtUK1bYc5V8
    A comment on using AC line frequency for timing: As I understand, line frequency is less accurately regulated than in earlier years, in part due to the complexities of renewable energy contribution. I regularly see variation of +/-10 seconds over weeks. They pull its accuracy back toward zero error over time but without plotting the variation, you don’t know where you stand. Accuracy may be better or worse in your area.

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.