A Novel Free Pendulum Clock

[Martin KyteParticipant @martinkyte99762]

Posted by SillyOldDuffer on 12/03/2021 17:43:04:

Posted by Martin Kyte on 12/03/2021 12:10:59:

I hadn't realised there might be a relationship between circular and impulse error. Needs thinking about – it's on the list now.

Dave

Not so much a relationship as which generates the biggest variation. For a portion of the swing the pendulum is not swinging free but is being controlled by the impulsing system or escapement if you like. If the driven portion of the swing varies it will affect the period. For example in a electromagnetic impulsing system if the pendulum swing is not exactly central to the field it will get a sideways kick which disturbs it's path. It is unlikely that the swing will repeat its path each period and there may be all manner of periodic eccentricities or quasi stable states. How big these errors are will be a matter of testing. If they are more of a problem than variation due to circular error then that will determine how often you want to impulse for best stability.

That is probably as clear as mud for which I appologise but I hope you catch my thinking.

Keep up the good work

regards Martin

[John HaineParticipant @johnhaine32865]

Tony, not sure how you are checking the clock but one way if you have an android device is to take a photo of the face. Android saves photos with the system timestamp as part of the filename. Also there's an app called clocksync that shows the error between ntp and system, on my phone it's usually less than a second.

[SillyOldDufferModerator @sillyoldduffer]

Posted by Martin Kyte on 12/03/2021 18:44:10:

Posted by SillyOldDuffer on 12/03/2021 17:43:04:

Posted by Martin Kyte on 12/03/2021 12:10:59:

I hadn't realised there might be a relationship between circular and impulse error. Needs thinking about – it's on the list now.

Dave

Not so much a relationship as which generates the biggest variation. For a portion of the swing the pendulum is not swinging free but is being controlled by the impulsing system or escapement if you like. If the driven portion of the swing varies it will affect the period. For example in a electromagnetic impulsing system if the pendulum swing is not exactly central to the field it will get a sideways kick which disturbs it's path. It is unlikely that the swing will repeat its path each period and there may be all manner of periodic eccentricities or quasi stable states. How big these errors are will be a matter of testing. If they are more of a problem than variation due to circular error then that will determine how often you want to impulse for best stability.

That is probably as clear as mud for which I appologise but I hope you catch my thinking.

Keep up the good work

regards Martin

All too clear I'm afraid! I think you're right. Some of my test runs with a stiff 'sprod' (the whole rod behaves like a spring) show a kind of sawtooth effect due to the impulse extending the period with a bump, and then speeding up as the amplitude decays.

Interesting stuff. I'm letting my imperfect clock do a long run while I'm bogged down redecorating and house-mending, but I'd like to get back to it and attempt a Mk2, plus something like Tony's clock for comparative purposes.

Currently typing with my screen tilted sideways at 30° to let red wine drain out, whilst two Arduinos dry out on a radiator. Knocked the glass over after the first sip. Aargh!

Dave

[duncan webster 1Participant @duncanwebster1]

John's link is very interesting, but it's a lot easier with a processor! According to my clocky contact (an FBHI no less), carbon fiber has taken over from Invar, and is a lot is a lot easier to get hold of

My heretical thoughts are that it doesn't matter what you do to a pendulum, as long as you do the same every swing. Impulsing every swing and keeping the amplitude constant means that almost every swing will not be the true period of the pendulum, but they will be the same. I miss an impulse out when the amplitude grows, but over a sensible number of swings it all averages out.

I used slotted opto switches incorporating Schmidt triggers to get fast edges. I don't have a direct sunlight issue, if it was a problem then I'd try modulating the opto switch LED supply and sending the photo transistor output to a Phase Locked Loop chip LM567 PLL. No doubt someone who knows what he is talking about will be along in a bit to tell me why this won't work! The flag on my pendulum has appreciable length so it triggers the mid point opto a bit before the actual mid point, the magnet is then energised for a time generated by the processor so that it switches off at or before the mid point. If anyone were to object to this processor derived timing then the capacitor discharge system in John's link could be adopted.

[SillyOldDufferModerator @sillyoldduffer]

Posted by duncan webster on 12/03/2021 20:06:41:….I don't have a direct sunlight issue, if it was a problem then I'd try modulating the opto switch LED supply and sending the photo transistor output to a Phase Locked Loop chip LM567 PLL. No doubt someone who knows what he is talking about will be along in a bit to tell me why this won't work! …

While we wait for someone who knows what he's talking about, whilst modulating the beam certainly helps with sunlight doesn't it mean the trigger precision is limited by the modulation frequency and how long it takes the PLL to respond to change. Dunno about the PLL, but if modulated at 15kHz, detection would be not better than ±67µS, which is rather slower than a micro-controller can do with interrupts or input capture mode, .

Thinking about how long it takes to detect the pendulum, Tony's clock uses Hall Effect sensors. I've no idea how Hall Effect and IR sensors compare speed wise?

More deep water – it's a shame I can't swim!

Dave

[Tony JeffreeParticipant @tonyjeffree56510]

Posted by John Haine on 12/03/2021 19:12:02:

Tony, not sure how you are checking the clock but one way if you have an android device is to take a photo of the face. Android saves photos with the system timestamp as part of the filename. Also there's an app called clocksync that shows the error between ntp and system, on my phone it's usually less than a second.

Thanks John – that's very useful.

[John HaineParticipant @johnhaine32865]

I have an HED on my "JH5" clock to sense the extreme of swing. It's an integrated device with a built-in schmidt trigger, and is described as "continuous time" which means that it doesn't use chopping to reduce drift. I don't have the data sheet to hand but my impression is that it's very fast – after all these things are used in tachometers and car gearboxes and stuff. My main concern is variability of the switching point with distance between the magnet and sensor – though I use a CF rod the pendulum support is mounted on 2 brass pillars with the sensor at the bottom end. As temp rises the pillars get longer, the sense magnet gets further above the sensor, and the pendulum has to swing further to activate it. So higher temp = larger amplitude = slower rate. On the other hand the sense magnet gets further from the coil so the impulse reduces and amplitude decreases. I haven't a clue how these interact!

[Martin KyteParticipant @martinkyte99762]

So try out this train of thought.

If you have a pendulum with an escapement capable of delivering a set force for a set time and arrange things such that exactly the same amount of energy is given to the pendulum at the same point in it's swing then with varying atmospheric conditions the amplitude will change according to frictional interactions with the air. Suppose now the atmospherics change so that the amplitude is smaller and that the impulse is made correspondingly larger to restore the swing to the previous amplitude by, for example, extending the duration of the impulse. Although the pendulum has the same amplitude in both cases it does not follow it will have the same period. The arc through which it is driven by the escapement is longer in the second case. The period in the second case should be shorter than in the first case on the basis that the pendulum is driven for a longer portion of it's swing.

In the first case if the amplitude was less the period would also be less due to circular error. Whether the shorter period in the second case exactly matches I don't know but at least they are in the same direction. I suspect that the two case match in terms of energy budget but not neccessarily in terms of time.

regards Martin

[Martin KyteParticipant @martinkyte99762]

So essentially I think I'm saying you cannot correct for circular error by amplitude regulation because all you do is move the circular error to the escapement error. I

regards Martin

[Tony JeffreeParticipant @tonyjeffree56510]

What do you guys use as a suspension for your pendulums? I have used a conventional spring suspension and am noticing that I can get some rotational oscillation (rotating about the axis of the rod) – some of that was due to me using steel screws to mount the sensors on the backboard, and replacing those with brass has improved matters considerably, but the possibility is still there. Makes me wonder if I would be better off using knife edges or bearing races.

[John HaineParticipant @johnhaine32865]

Knife edges are difficult to make well enough to take the loads of a bob without breaking, you need sapphire or something similar (carbide perhaps?). Ball races have been used but for low friction need to be very small and the limited angular excursion may lead to "brinelling" (see current thread on this subject). I used a single 16mm wide x 0.1mm CuBe strip on my first clock and rotation is an issue. On the second and also the current build I use a 2 spring version, it has 2 x 6mm wide strips spaced apart ~20mm sandwiched and soldered between brass strips.

You can just about see the details in this photo (in my JH5 album).

[John HaineParticipant @johnhaine32865]

Duplicate post

Edited By John Haine on 13/03/2021 12:55:50

[Martin KyteParticipant @martinkyte99762]

Posted by duncan webster on 12/03/2021 20:06:41:

My heretical thoughts are that it doesn't matter what you do to a pendulum, as long as you do the same every swing.

Impulsing every swing and keeping the amplitude constant means that almost every swing will not be the true period of the pendulum, but they will be the same.

I miss an impulse out when the amplitude grows, but over a sensible number of swings it all averages out.

Point 1 I totally agree with.

Point 2 If you mean the amplitude of the impulse, yes. If you vary the impulse to regulate the amplitude of the pendulum you break your rule 1/point 1

Point 3 Missing an impulse again breaks rule 1.

regards Martin

[Tony JeffreeParticipant @tonyjeffree56510]

I'm using a single 1/2" wide spring steel strip. I think part of the problem is that I have made it longer than necessary – just made up a pair of larger cheeks that will reduce its effective length.

Agree with your comments about the other alternatives. If the larger cheeks don't improve matters, I may try a similar setup to yours, but with two 1/2" springs.

[duncan webster 1Participant @duncanwebster1]

Tony:

I made the same error at first, spring too long. My clocky man said no more than 10mm, so I reduced it. Some clever FEA man could no doubt tell us how short it could be without increasing the bending stress too much.

Martin:

I think you're right, but assuming higher humidity slows the pendulums down and more impulse speeds it up (both these assumptions might be wrong) they will tend to balance out. The only way to get rid of atmospheric effects is a vacuum enclosure as used on the best astronomical clocks. I'm not going there (yet), but it has occurred to me that just a hermetic enclosure might be beneficial, at least the atmospheric composition would not change, it's temperature would. The advantage is that you don't have in/out leakage, any interchange would be from small atmospheric pressure changes and diffusion.

SOD:

I need to have a think about possible delays to the impulse triggering. I'll be back

#2 son is here today, he has promised to show me how to post videos, bet you can't wait

[Martin KyteParticipant @martinkyte99762]

Lower amplitude shortens the period. Higher amplitude increases the period.

So if atmospheric conditions create more drag the amplitude will decrease and the period will be shorter.

Correspondingly if the impulse is increased to correct for the lower amplitude that will also shorten the period (I think).

I don't think they cancel.

regards Martin

[John HaineParticipant @johnhaine32865]

Humidity seems to have very little impact, it's air pressure and temperature that mainly affect the drag and hence amplitude. More pressure and lower temp = air denser. There is also a small viscosity affect, which gets higher with temperature.

Some very successful clocks have been made with aneroid compensators which work very well. Or you can take the Harrisonian approach and get a balance between escapement and circular deviation so changes in amplitude with pressure compensate for buoyancy changes.

The affect of changing impulse on period depends on where it is – if the centre of the impulse is after BDC a larger impulse slows the pendulum down; if before it speeds it up. Ideally the impulse is centred on BDC when it doesn't change timing.

If you did use a pulsed light source to avoid ambient effects, it would be better to use a "synchronous rectifier" to detect the signal rather than a PLL.

Just as a benchmark, Doug Bateman's electronically maintained clock uses an opto sensor to keep the amplitude stable to within a second of arc – described in Horological Journal Jan 1972. It is also fitted with an aneroid compensator and in a later article (October 75) he describes its performance with this which seems to reduce the error due to pressure variations by a factor of 10.

[John HaineParticipant @johnhaine32865]

Posted by Martin Kyte on 13/03/2021 13:44:04:

Lower amplitude shortens the period. Higher amplitude increases the period.

So if atmospheric conditions create more drag the amplitude will decrease and the period will be shorter.

Correspondingly if the impulse is increased to correct for the lower amplitude that will also shorten the period (I think).

I don't think they cancel.

regards Martin

The condition that create more drag is greater density, but this makes the bob lighter through buoyancy, which increases the period. The effect of impulse on period may make it longer or shorter depending where it's applied in the cycle.

For a perfectly central impulse it turns out that there is a nominal amplitude at which the buoyancy, drag and circular deviation effects cancel out and the pendulum rate is insensitive to barometric pressure. For a typical regulator pendulum this is about 2 degrees. However you have to let the amplitude change with pressure to actually get the compensation, so a system that regulates the amplitude can't be made to work in this way.

There's a similar amplitude for other impulse positions, which may be at larger or smaller amplitude. In the case of Burgess' Clock B at Greenwich it's about 6.25 degrees.

[Martin KyteParticipant @martinkyte99762]

Posted by John Haine on 13/03/2021 14:06:45:

Posted by Martin Kyte on 13/03/2021 13:44:04:

Lower amplitude shortens the period. Higher amplitude increases the period.

So if atmospheric conditions create more drag the amplitude will decrease and the period will be shorter.

Correspondingly if the impulse is increased to correct for the lower amplitude that will also shorten the period (I think).

I don't think they cancel.

regards Martin

The condition that create more drag is greater density, but this makes the bob lighter through buoyancy, which increases the period. The effect of impulse on period may make it longer or shorter depending where it's applied in the cycle.

For a perfectly central impulse it turns out that there is a nominal amplitude at which the buoyancy, drag and circular deviation effects cancel out and the pendulum rate is insensitive to barometric pressure. For a typical regulator pendulum this is about 2 degrees. However you have to let the amplitude change with pressure to actually get the compensation, so a system that regulates the amplitude can't be made to work in this way.

There's a similar amplitude for other impulse positions, which may be at larger or smaller amplitude. In the case of Burgess' Clock B at Greenwich it's about 6.25 degrees.

So my feeling is that for electronically impulsed clocks the objective is to create the best pendulum you can such as the above and use the electronics to generate the most stable impulse you can. Amplitude regulation causes too much interference with the pendulum. Essentially you want to provide enough energy to the oscillator to maintain it as near to it's natural frequency as possible in the most stable way you can.

regards Martin

[duncan webster 1Participant @duncanwebster1]

I hope John won't mind if I link to this which explains quite a bit Burgess clock. I've learned quite a lot from this thread, not least how little I knew before!

If I shift my electromagnet so that it is offset from the centre and only impulses in one direction that could give impulse both before and after mid point. Probably not equally, but would it be worthwhile?

[Martin KyteParticipant @martinkyte99762]

I hope you don't take my contributions as gospel without some conformation from elsewhere Duncan. I have some understanding but don't set up as an authority.

Like you I have found this and the other similar thread emmensy entertaining and informative.

regards Martin

[Tony JeffreeParticipant @tonyjeffree56510]

Reducing the effective spring length to ~5mm seems to have controlled the rotational oscillation pretty well, so that will hopefully improve matters. I've been playing around with the height of the electromagnet (the gap between it and the armature on the pendulum) and it is easy enough to increase the time between impulses – up to just short of 3 minutes at one point. However, given some of the discussion above, it probably makes sense to increase the gap (reduce the force of the impulse) as that will result in a smaller variation in amplitude.

Edited By Tony Jeffree on 13/03/2021 16:09:48

Edited By Tony Jeffree on 13/03/2021 16:09:58

[John HaineParticipant @johnhaine32865]

Posted by duncan webster on 13/03/2021 15:16:23:

I hope John won't mind if I link to this which explains quite a bit Burgess clock. I've learned quite a lot from this thread, not least how little I knew before!

If I shift my electromagnet so that it is offset from the centre and only impulses in one direction that could give impulse both before and after mid point. Probably not equally, but would it be worthwhile?

Not at all Duncan, a good find.

[Tony JeffreeParticipant @tonyjeffree56510]

Posted by John Haine on 12/03/2021 19:12:02:

Tony, not sure how you are checking the clock but one way if you have an android device is to take a photo of the face. Android saves photos with the system timestamp as part of the filename. Also there's an app called clocksync that shows the error between ntp and system, on my phone it's usually less than a second.

Couldn't find clocksync on the Android app store, but I did find an app called "Atomic Clock & Watch Accuracy Tool (with NTP time)" which gives NTP time and in the unpaid version will allow you to record measurements for a single timepiece (more if you pay the sub). Pretty neat – you tap the screen when you want to take a measurement and enter the time from the clock & it records the NTP time at the moment you tapped. Nice graphs etc. and calculates a day rate from your data when you give it enough measurements.

[John HaineParticipant @johnhaine32865]

Tony, that's weird! I just checked and you're right, ClockSync is no longer available. However there is one called System Clock Monitor which does the same and actually seems better. Taking a photo of the clock face is an easy way to do the recording.

[Author]

Posts