Arduino Pendulum Clock Design – Comments Welcome

[SillyOldDufferModerator @sillyoldduffer]

Posted by Glenn Turner on 29/03/2022 14:31:26:

… The plan is to see what causes pendulum variations and find the practical problems, then start to fix those problems. I imagine that will have to make a pendulum out of invar and seal the clock in a container to stop pressure changes. If things go well I would like to make a new clock. Ultimately the plan is to manufacturer a highly accurate mechanical pendulum clock that has a completely detachable electronic adjustment aid. But right now it is just a hobby.

…’m using the Arduino Due to measure the period of the pendulum and calculate the average period each minute along with the standard deviation.

GPS PPS is going to be used for longer measurements…

Ah ha, we might be twins. I'm the evil one! Our breadboard electronics are remarkably similar.

I started with a Pendulum Clock Analyser, my goal being a portable device that could be pointed at a conventional pendulum clock to accurately measure, log and analyse what the pendulum is doing – basically identifying irregularities and their causes.

Tiny time differences between each swing and finding the tick/tock ratio reveals problems due to the escapement, suspension and vibration. Likewise, analysing long term data reveals problems due to the weather, bent teeth etc, and the clock's rate. (quickly revealing how it drifts steadily over time, used to advance or retard the pendulum finely.) Also possible to measure Q and other stuff useful to anyone making and adjusting mechanical clocks.

Problem was I couldn't make the Analyser portable because it depended on reliable GPS or MSF reception, neither of which work everywhere in my house, even with improved antennae. Worked well on my dining table, not at all in the front hall way, and the whole point was doing the analysis without disturbing the mechanical clock.

Testing the analyser, I noticed a baby light bob pendulum suspended by carbon fibre from a bodged Meccano stand kept remarkably good time – similar to a quartz kitchen clock. It led me to wonder what could be done with a well-made version of the same idea, potentially considerably smaller than a conventional pendulum clock, and easily sealed in a vacuum.

Led me into a jungle of problem solving! The law of diminishing returns kicks in – improvements become progressively more difficult. Beware, Arduinos, pendulums and improving time-keeping are all addictive.

I wish you'd been about earlier: for reasons unknown I didn't look at the Due and now you've mentioned it, it looks like a good choice. It has a crystal oscillator and an internal Time Counter that can be clocked by a precision external source, potentially giving a resolution of about ±12 nanoseconds. The Leonardo also has a real crystal, but it's too slow for high resolution work.

After your Hermie is investigated be interesting to hear if lack of a temperature compensated pendulum is the main source of error. My guess is lack of rigidity and a simple escapement introduce bigger errors that will have to be fixed before Invar and vacuum running are worth doing. I could be wrong…

One thing worth doing whilst measuring individual swings is to repeatedly calibrate the Due's notion of 1 second against GPS in software. If PPS is used to trigger interrupts every second and the internal clock is used to measure how long the Due thinks GPS seconds are, the difference makes it easy to get a correction factor allowing the crystal's temperature and other errors to be reduced mathematically. The Arduino tracks GPS well because a new correction is available every second.

PPS ticks can also be counted to have the Arduino do HH:MMS more accurately than an RTC, which depends on how stable and accurate it's crystal is – not in the same league as GPS!

Hours of fun ahead!

Cheers,

Dave

[John HaineParticipant @johnhaine32865]

Ordinary steel is about 1e-5 per Deg C tempco which amounts to about half a second a day per degree. So temperature is probably the first thing to compensate. Carbon fibre is better than Invar, which is anyway hard to get and expensive. I think it's sensitivity to humidity is overestimated. It's cheap and easy to get in rod or tube form.

[Michael GilliganParticipant @michaelgilligan61133]

Posted by SillyOldDuffer on 29/03/2022 20:41:56:

[…]

My guess is lack of rigidity and a simple escapement introduce bigger errors that will have to be fixed before Invar and vacuum running are worth doing. I could be wrong…

.

With some regret, Dave … I very much doubt if you are wrong.

MichaelG.

[Glenn TurnerParticipant @glennturner80107]

Regarding the Due's notion of 1 second. Yes I already spotted that it is slightly off compared to GPS. But I should look at this further and do some more detailed tests particularly regarding temperature.

Now for a word that may get you really thinking…..of course you may already have thought about this.

The word in question is……Ephemeris

Yep. I talking about the moon and its impact on gravity and hence the pendulum. My maths suggests it should be detectable.

[Glenn TurnerParticipant @glennturner80107]

Posted by Michael Gilligan on 29/03/2022 21:52:00:

Posted by SillyOldDuffer on 29/03/2022 20:41:56:

[…]

My guess is lack of rigidity and a simple escapement introduce bigger errors that will have to be fixed before Invar and vacuum running are worth doing. I could be wrong…

.

With some regret, Dave … I very much doubt if you are wrong.

MichaelG.

A cheap clock was just a quick way to get started. I suspect the escapement maybe an issue quite quickly.

I managed to pick some invar relatively cheaply. Fused quartz looked interesting but noticed it needs to stretched. carbon fibre wasn't on my radar until now.

[Michael GilliganParticipant @michaelgilligan61133]

I suspect that the somewhat flexible ‘pendulum bracket’ might prove to be a problem, Glenn.

MichaelG.

.

P.S.__ Lots of good information, and some great photos of Shortt clocks here:

http://www.royalobservatorygreenwich.org/articles.php?article=1327

Edited By Michael Gilligan on 29/03/2022 22:57:47

[SillyOldDufferModerator @sillyoldduffer]

Posted by Glenn Turner on 29/03/2022 22:22:33:

Posted by Michael Gilligan on 29/03/2022 21:52:00:

Posted by SillyOldDuffer on 29/03/2022 20:41:56:

[…]

My guess is lack of rigidity and a simple escapement introduce bigger errors …

.

With some regret, Dave … I very much doubt if you are wrong.

MichaelG.

A cheap clock was just a quick way to get started. …

And a good place to start too! Applying an ordinarily programmed Arduino to an imperfect pendulum clock will detect erratic running and working out what causes detected errors is educational. If an ordinarily programmed Arduino is applied to a very good clock, then the Arduino's own timing errors start muddying the picture.

By 'ordinarily programmed', I mean code written using the standard libraries without regard to any timing issues they might introduce. For example, digitalRead() and digitalWrite() are both very slow compared with directly manipulating ports because they do lots of friendly behind the scenes checking: they're easy to use, but slow. The alternatives are fast, but harder to program.

Dave

[Author]

Posts