Precision pendulum techniques

[John HaineParticipant @johnhaine32865]

When Harrison designed his gridiron pendulum he must have made or used expansion measurements for the two materials used. I'm not sure that's different in principle from applying a correction for accumulated temperature deviation as the clock runs? Likewise quartz movements sometimes use a digital correction for measured crystal frequency to eliminate a trimmer capacitor. Both my running clocks do something similar, correcting for the actual period error by occasionally swallowing seconds. I could make that adaptive and add terms for estimated temperature and pressure error. Potentially better to reduce the complexity of the pendulum assembly by eliminating extra joints adding temperature compensation and aneroid capsules etc.

[Michael GilliganParticipant @michaelgilligan61133]

Philosophically … I would say that the very concept of experimental clockmaking is futile.

Civilisation is in its third epoch of timekeeping

1. Solar Time
2. Sidereal Time
3. Defined Time

The first two lent themselves to experimental work, because mankind was trying to comprehend and measure something [rather fuzzy] provided by ‘the Heavens’ … and better clocks led to better understanding

The third is different … Man has defined ‘Time’ and prescribed how it will be measured.

So … returning to the Subject line of this thread : all we can hope to achieve is a better understanding of why our pendulum is incapable of measuring what we now call Time.

Plato would presumably be offended.

.

Note: it is worth looking here: **LINK**

https://faculty.washington.edu/smcohen/320/timaeus.htm

… under the heading Temporal

MichaelG

[Martin KyteParticipant @martinkyte99762]

Hi John

Im not objecting to Dave’s method of correcting for variations (what I would call external compensation) by altering the integration constant in as the clock runs based on a predetermined lookup table of rate against environmental parameters. The issue is the lookup table has been generated using atomic time keeping and builds that into the clock. If the look up table could be generated by other methods such as long runs in environmental chambers at set parameters then that would decouple the clock from atomic timekeeping. I know this is a slightly pedantic point but it’s enough to generate that little niggle of a slight cheat.

regards Martin

[Martin KyteParticipant @martinkyte99762]

Responding to Michael

I can’t really agree with you. All timekeeping depends on watching the changing state of something physical in the universe be it Sun Stars swinging pendulums or atoms. It’s all fuzzy just some less fuzzy than others.

regards Martin

[Tony JeffreeParticipant @tonyjeffree56510]

Posted by Martin Kyte on 16/07/2023 09:15:08:

Hi John

Im not objecting to Dave’s method of correcting for variations (what I would call external compensation) by altering the integration constant in as the clock runs based on a predetermined lookup table of rate against environmental parameters. The issue is the lookup table has been generated using atomic time keeping and builds that into the clock. If the look up table could be generated by other methods such as long runs in environmental chambers at set parameters then that would decouple the clock from atomic timekeeping. I know this is a slightly pedantic point but it’s enough to generate that little niggle of a slight cheat.

regards Martin

Hi Martin

At some point you have to use some kind of reference in order to determine what effect the parameters you are setting in your environmental chamber are having. I don't see the pedantic point you are making here.

Cheers,

Tony

[Michael GilliganParticipant @michaelgilligan61133]

Posted by Martin Kyte on 16/07/2023 09:21:17:

Responding to Michael

I can’t really agree with you. All timekeeping depends on watching the changing state of something physical in the universe be it Sun Stars swinging pendulums or atoms. It’s all fuzzy just some less fuzzy than others.

regards Martin

.

No problem with lack of agreement, Martin … it prompts thinking and discussion.

In response to your comment, though … the currently defined second may itself be fuzzy, but it’s spectacularly crisp-edged compared with the other two.

The problem [in my opinion] is … There is nothing that we amateurs can do it investigate that fuzziness !

MichaelG.

[Martin KyteParticipant @martinkyte99762]

With the environmental chamber method you a for instance set the temperature constant. Run the clock for a long time an reference it against conventional non atomic time keeping. The long period gives you the resolution you need by making the variation cumulatively large. In order to do this in real time with fast changes of temperature you need the resolution of atomic timekeeping. This means you cannot make your clock without using such technologies.

regards Martin

[Tony JeffreeParticipant @tonyjeffree56510]

Either way though you are still in some sense building the reference clock(s) into your clock.

Edited By Tony Jeffree on 16/07/2023 11:19:39

[Martin KyteParticipant @martinkyte99762]

Yes but with the chamber method you don’t need to invoke atomic technology. Indeed you could do it by astronomical observations and you don’t need a better clock than you are attempting to build. I would be interested in Dave’s observations.

regards Martin

[Tony JeffreeParticipant @tonyjeffree56510]

Why use a less accurate reference when access to a world class time standard is freely available?

I entirely "get" and agree with Dave's point that the pendulum (other resonators are available) should be independent – i.e., you aren't constructing some kind of phase lock between the reference clock and your own resonator, so the resonator and its control circuitry/mechanics are truly detached from the reference, but I'm having a hard time understanding what the philosophical difference is between (e.g.,) constructing/adjusting a mechanical temperature compensation mechanism by observation/measurement relative to a reference clock and achieving the same result by constructing tables in software and using that data in an algorithm to control the frequency of the oscillator (or to correct the indicated time).

Cheers,

Tony

[Martin KyteParticipant @martinkyte99762]

Because the original question was about the philosophy of building a pendulum clock. I said that my take was that no other external time reference should be used in the clock except the pendulum itself. Therefore the derivation of the lookup table employed by the clock for compensation should not use technologies in advance of the pendulum. As I say it’s my viewpoint. For testing or manufacture you can use what you like.

regards Martin

[SillyOldDufferModerator @sillyoldduffer]

I'm still thinking about the philosophical debate! Off the cuff, I say experimental results take us where the evidence leads. It's always necessary to confirm the evidence is valid, because it's always possible that the experiment was faulty. Likewise, conclusions drawn from the evidence also need careful consideration, and it's always possible new evidence might overturn a conclusion.

It's entirely possible my ideas are flawed. Martin's comments echo my own concerns, and I don't have a solid rebuff yet, and might never have. His comments, and others, have been extremely helpful. Whilst I would like to end up with a clock that everyone agrees outperforms a Shortt-Synchronome fair-and-square, could be I can't get it to work anything like that well, or, the approach is a step too far – cheating.

Meanwhile it's certainly interesting.

I said earlier that I'd see if Q varies over time. This graph, which also shows temperature, suggests it does:

I think I see a loose correlation between Q and temperature. Might be castles in the flames, but Q seems to fall with temperature after a delay. If the effect is mechanical, it fits with a suggestion made by John Haine, whereby the structure of the pendulum and supports take longer to warm and cool than the sensor. There's been some discussion about what to do about temperature lag, such as Duncan's dummy pendulum. Not proven though: at present the clock is running independently with GPS switched off. This means the Arduino crystal frequency isn't compensated, so the hint of a Q/temperature relationship may be measurement error. That wouldn't explain why Q ranges between 21000 and 28000 though. I suspect something mechanical, perhaps elasticity of the suspension spring changing. Any suggestions?

I've just finished a 17 day test run of my experimental pendulum, or rather a power-cut finished it for me, and haven't had time to look at the results. I'll report them later in my own thread.

Dave

[S KParticipant @sk20060]

On the "philosophy" question, it may be beneficial to distinguish between pendulums and clocks, and further sub-divide the categories according to their features. For example:

• A pendulum. The only pure pendulum is a completely free one, with no deliberate external energy input applied except to start it (the initial lifting force). These are not very useful for clocks since they must eventually stop, exhibit non-isochronous behavior as their motion decays, etc.
• A pendulum with external energy input to maintain its motion. The external energy becomes "entangled" with the pendulum and alters its performance in both desired and undesired ways. A lot of the discussion here has been on how to apply external energy without disturbing the pendulums natural motion – something that can never be achieved perfectly.
• A mechanically compensated pendulum. Normally attempted by applying opposing actions (e.g. to temperature changes), and again a goal that can never be achieved perfectly.
• A pendulum clock: A pendulum with energy restoration plus a rating device (mechanical or electronic) to translate the pendulum's motion into standard time units. Most discussion here currently centers on electronic rating.
• An electronically-compensated pendulum clock: A pendulum clock with informational input from electronic environmental measurements (temperature, air pressure, local gravity, etc.) that is computationally processed to correct the translation between the pendulum's motion and standard time units.

I'm on the side of electronic compensation leaning into the fever-dream of the spectrum. It results less in a "pendulum clock" and more in "an instrumented computer system incorporating an anachronistic oscillator." But that can be fun too.

 

Edited By S K on 17/07/2023 14:17:28

[S KParticipant @sk20060]

Posted by John Haine on 10/04/2023 22:57:57:

 

So I think my conclusion is that one will get an irreducible measurement error of the order of a few microseconds rms using the Sharp optos. To put that in perspective the expected time error over a year due to that would be about 14ms which is probably insignificant compared to other time errors for a pendulum clock.

 

It seems like this implies that one need not use a microcontroller much faster than an Arduino, too.

But of course faster opto systems are certainly possible. One time constant of the charge collection of a typical bare silicon sensor in the red to near-infra-red would be about 30ns, so you could probably drop the opto system's measurement error by a factor of 100 if you tried.

Because the absorption depth is so much less in silicon, blue or ultraviolet would buy you an additional factor of 10. But by then you would need some fancy electronics to go with it, and my guess is that this would not remain the largest source of timing errors anyway.

 

Edited By S K on 17/07/2023 16:23:16

[SillyOldDufferModerator @sillyoldduffer]

Posted by Martin Kyte on 16/07/2023 12:21:22:

Because the original question was about the philosophy of building a pendulum clock. I said that my take was that no other external time reference should be used in the clock except the pendulum itself. Therefore the derivation of the lookup table employed by the clock for compensation should not use technologies in advance of the pendulum. As I say it’s my viewpoint. For testing or manufacture you can use what you like.

regards Martin

Maybe worth repeating what's controversial about my clock. Not terribly clear from the various forum threads, which wander.

In short:

• There's a moderately well-engineered pendulum in a stiffish frame. Not done yet, but the pendulum is designed to run in a vacuum
• Impulsed by an electromagnet. A microcomputer can alter when an impulse occurs and how powerful the impulse is. I believe it's less intrusive and more tunable than any mechanical escapement because there's no physical contact and the computer is far faster than the pendulum
• The pendulum has no temperature or barometric compensation. Instead, it's fitted with electronic sensors allowing the microcontroller to compensate for them in the 'gear train'
• The gear train is electronic, and able to convert any period into DD/MM/YY HH:MMS, potentially with microsecond or higher precision
• Controversial:
  • Stage 1: the pendulum/clock runs and collects its notion of time, plus temperature and air pressure data, more the better.
  • Stage 2: the data is analysed. The period of each beat is compared with a high-accuracy clock, either GPS or NTP, giving the pendulum's actual period as modified at the time by temperature and pressure. The analysis allows the pendulum's period to be predicted at any combination of temperature and pressure with a simple formula
  • Stage 3: the clock's "gear train" is configured to use the formula. GPS and NTP are disconnected, and the clock operates independently. When the pendulum ticks, the clock, knowing what the temperature and pressure are, calculates what the pendulum period should be, and counts that.

Stages 1&2 create what Martin calls a look-up table. Actually not a table but conceptually similar. Although the statistical analysis is high-end, I don't think it's philosophically different from what conventional clockies do. Having built a clock, they're run for a known period of time, comparing them with a better clock: solar system, TIM, Radio pips, whatever. Then adjustments are made to improve the timekeeping. The clock might be left for days, weeks or months between manual checks, whilst I use a computer and better clock to check every beat. Doing so gives a fine-grained picture of what the pendulum is doing, but I think it's essentially the same process. Looking closely shows pendula are noisy in the signal sense, i.e varying in frequency. Many of these variations average out, so a pendulum clock can be a good time-keeper over months, even though its pendulum is noisy. If it were known, the same clock's short term performance could be distinctly wobbly.

High-accuracy requires frequency wobble (noise) to be reduced. Conventionally done by smart mechanisms implemented with the best materials and workmanship. In contrast, my clock reduces noise by applying a sort of digital filter. When the pendulum ticks, the microcontroller ignores the actual time taken, which being noisy is always slightly wrong, and instead counts the time the pendulum would have taken had it behaved perfectly. Perfect was identified statistically during Stage 2.

The approach corrects for temperature, air-pressure, and the maths also filters out pendulum noise. So rather than needing to build a really good high-end pendulum, I'm hoping to do better with a cruder set-up. My pendulum can be noisy provided it reliably beats within its previously determined statistical range. I've not thought about how the filter effects Flicker and the other exotic noise phenomena that upset clocks of all types.

The clock depends on a mass of previous high-accuracy observations. Further, in independent mode, the pendulum's actual period isn't measured or used by the clock, which is unusual! Instead the pendulum generates events, from which period is inferred. The time quantity in the calculation comes highly polished from the statistical analysis, and the current temperature and pressure, not direct from the pendulum. Hmmm!

However, maybe filtering isn't controversial in that conventional clocks do it too. I argue that coincidental mechanical averaging is also a filter. A clockmaker might assume that his carefully set-up and compensated pendulum really does beat exactly 1 second every time, but they never do. I suggest the apparent accuracy of most pendulum clocks results from averaging of which the builder is unaware.

Accuracy issues only become apparent when enough precision is added to the display, or the clock is compared with a much better timekeeper. Quite serious timekeeping errors aren't obvious on a minute hand, though they often appear on the seconds. Trouble really kicks off displaying deciseconds and below, and millisecond accuracy is very challenging.

Dave

Dave

[Martin KyteParticipant @martinkyte99762]

Very clearly put Dave.

As you inferred I used the phrase “lookup table” as a simple way to describe your complex and elegant compensation technique.

My only real issue is the need for ‘atomic clock’ technology in order to produce stage 1 and 2. The results of stages 1 and 2 are then used as a component of the timekeeper. You are perfectly correct in saying that other clocks have been checked and adjusted by astronomical observations but these are existing events and not new technologies.
I cannot see a way of carrying out stages 1 and 2 with the resolution required to characterise the clock beat by beat without invoking new technologies. If your stated aim was to push pendulum technology as far as it will go but then use atomic clock technology as a component albeit the data set you may as well have gone with reading the time from your sat nav.
I am playing devils advocate on this but I think it’s possibly a question that you need to be able to defend. As I said before what you have produced thus far is astonishing and so interesting so please don’t think I’m in any way criticising.

best regards Martin

[Tony JeffreeParticipant @tonyjeffree56510]

" You are perfectly correct in saying that other clocks have been checked and adjusted by astronomical observations but these are existing events and not new technologies."

Existing events/phenomena, maybe, but in order to observe those phenomena, new technologies were required (such as optical telescopes,…). One could argue successfully that the frequency of oscillation of an atom is a similarly existing phenomenon; we just need the modern equivalent of the optical telescope to observe its frequency and make use of it.

I have little doubt that if you time travelled back to Harrison's time with an atomic clock in your back pocket and let him borrow it for a few months he would have had very few qualms making use of the opportunity, just as he did with the other techniques he invented in order to improve the way his clocks performed, none of which existed before his time.

[S KParticipant @sk20060]

Posted by Tony Jeffree on 17/07/2023 17:39:35:

 

I have little doubt that if you time travelled back to Harrison's time with an atomic clock in your back pocket and let him borrow it for a few months he would have had very few qualms making use of the opportunity, just as he did with the other techniques he invented in order to improve the way his clocks performed, none of which existed before his time.

Interesting. But (way, way off topic) why would he trust it? It wouldn't be anything more than a mysterious and perhaps even demonic box to him. If you proclaimed "it's as accurate or even more so than observing the stars!," he might say "let's look, then" and find out that there are deviations all over the place. He might then congratulate you that your box at least comes close to matching the stars, but which would he ultimately trust more? As the old saying goes, "a man with two watches never knows what time it is," and aside from convenience likely the stars will be trusted more in the end.

Again, there is no such thing as "time." Only unified space-time, and so any attempt to fix or isolate "time" fails since it is always relative and therefore ephemeral. Our best efforts to define and measure it "accurately" are mostly for the sake of commerce, and at the human level it's really little more than a terrifying mutual hallucination that we grudgingly agree to believe in. 😉

To me, that relativity is what allows clock-making hobbyists to happily take pleasure in their pastime without needing justification.

 

Edited By S K on 17/07/2023 18:41:17

[Martin KyteParticipant @martinkyte99762]

I’m sure John Harrison’s response would be “we should use that then” rather than use it to make his clocks better. He had no qualms about ditching the sea clocks and going for the watch which is kind of the point I’m trying to make.

regards Martin

[Michael GilliganParticipant @michaelgilligan61133]

Posted by S K on 17/07/2023 16:13:58:

Posted by John Haine on 10/04/2023 22:57:57:

So I think my conclusion is that one will get an irreducible measurement error of the order of a few microseconds rms …

But of course faster opto systems are certainly possible. One time constant of the charge collection of a typical bare silicon sensor in the red to near-infra-red would be about 30ns, so you could probably drop the opto system's measurement error by a factor of 100 if you tried.

[…]

.

Which, incidentally is why I posted this on Friday morning:

.

My brain is feeble this morning, so I am just linking this page in case anyone can make use of it: **LINK**

Ballistic Chronograph MK2 #DIY

The salient point being that his ‘light gates’ have a very fast response.

MichaelG.

[S KParticipant @sk20060]

Sure. But with a ~1us propagation delay, those LM339's aren't particularly fast.

[Michael GilliganParticipant @michaelgilligan61133]

I defer to your wisdom … but I would be grateful if you could therefore explain how this bold claim comes to be untrue:

[quote] The sensor's net response time is ~100 nanoseconds. A .270 Winchester round takes about 25 microseconds to pass through the beam and another 600 microseconds to travel 500 mm! [/quote]

Naive as I am, I assumed he was speaking the truth.

MichaelG.

[SillyOldDufferModerator @sillyoldduffer]

Posted by Martin Kyte on 17/07/2023 17:11:32:

…

I am playing devils advocate on this but I think it’s possibly a question that you need to be able to defend. As I said before what you have produced thus far is astonishing and so interesting so please don’t think I’m in any way criticising.

best regards Martin

I agree. And I'm very grateful for awkward questions because peer review is essential to what I'm doing. Left to my own devices I'm perfectly capable of believing in all manner of deluded nonsense! Much safer to let the team comment and tackle any issues raised honestly.

Thought experiment: would it be acceptable if I replaced GPS in Stage 1 by comparing my clock to a century old Shortt Synchronome?

The Shortt clock's rate would be determined from star transits, and accuracy improved by resynchronising it to star time every night. (We shall have to move to Hawaii to ensure clear skies. A terrible imposition I know, but science demands sacrifices! I'm sure Morton's will pay the bills.)

Not as accurate or convenient as GPS, but I think a Shortt-Synchronome and a 24×7 team recording temperatures and pressures for a few months would get enough data to do the statistics. The necessary sums could be done with pencil and paper, though another team would be needed to do the mass of horrible arithmetic.

Dave

[S KParticipant @sk20060]

Michael: All I did was add to the conversation by pointing out the raw potential of silicon photodiodes. Of order 30ns should be possible in visual wavelengths and, when using UV light, the response time could potentially be sub-nanosecond.

Until it gets down to 1ns or less, I'm not interested in that slow-ass project. 😉

Edit: Those randomly-inserted emojis. 🙄

 

Edited By S K on 17/07/2023 21:51:17

[Michael GilliganParticipant @michaelgilligan61133]

Posted by S K on 17/07/2023 21:49:21:

Michael: All I did was add to the conversation […]

.

That’s absolutely fine

I wasn’t challenging you to provide an explanation, I was simply hoping that you would

… you are obviously more familiar with these things than I am.

MichaelG.

[Author]

Posts