How to synchronise a pendulum clock

[peterbalch]

What is the best way to synchronise a pendulum clock with a quartz oscillator? It seems like a simple question but I've searched the web and can find no convincing answer.

I have 150 year-old "black forest shield clock" and I want it to be as convenient as a cheap quartz clock. I realise that's not a direction that many horologists would go in but I don't want something I have to fiddle with. I want to change the battery once a year and I want it to keep time to a few minutes a year - like a clock I can buy in a car-boot sale for £0.50. I've already built an auto-winder (An Auto-winder for a Weight-driven Clock).

The pendulum should have a period of exactly 3168 ticks per hour (i.e. 6336 beats per hour). The pendulum has a wire link as its pivot. Its adjustment is by turning the whole bob (a disk) round by +/- one turn - i.e. 0.5mm or 0.15%. So, on its own, the clock has never been more accurate than a couple of minutes a day.

I want to build a quartz oscillator that runs at 3168 ticks per hour (to a few ppm). I can do the electronics, that's no problem. I'm assuming there will be a magnet (neodymium?) attached to the pendulum and an air-core electromagnet that's switched on occasionally. That leaves a lot of questions unanswered.

Does the circuit blindly send pulses to the electromagnet or is there a sensor to measure the current pendulum period and somehow to adjust the pulse? Huygens-style sync depends on blindly applying the impulses. Instead one could try to adjust the average period without attempting to sync.

Where is the magnet? I'd like it up just under the clock so it's inconspicuous. Does the north pole face the wall or does the north pole face to the side?

Let's assume the north pole faces the wall and the electromagnet is centrally placed on the wall. Is the electromagnet energised to attract or to repel the magnet? Is it pulsed once per cycle or once per beat? How long is the pulse? How powerful? Should the impulse from the magnet be as big as the impulse from the escapement? Bigger? Smaller? How would I measure it?

Should the electromagnet be on continuously to attract or repel the magnet? That would be equivalent to increasing or decreasing the force of gravity to change the period. There would then be some way of measuring the pendulum to know whether to advance or retard it.

Or let's assume the north pole faces sideways and the electromagnet is to the side just beyond the furthest swing of the pendulum. Should the electromagnet attract or repel the magnet? Or perhaps do both: attract then repel? Should it sense the current period and, if it needs to be shorter, wait until the pendulum is approaching then turn on the electromagnet to speed it up. Does speeding it up shorten the period? Doesn't it just increases the amplitude without changing the period?

You would think that the pulses from the electromagnet would simply entrain the pendulum but will they?

What is the "capture period" or "lock-in frequency"? Most people who sync a pendulum build a precision pendulum then assume the quartz driver will add those final few ppm accuracy. So the two oscillators' resonant frequencies differ by less than 0.01%. I have such a rubbish pendulum I need a capture period of, say, 0.3%. Is that even possible?

To get a big capture period you need a bigger impulse from the electromagnet. The pendulum becomes a driven oscillator. That's a classical example of a chaotic system. With a small frequency difference and a small impulse, you get lock. With a bigger difference you lose the lock and when you try to compensate by increasing the impulse you get chaotic behaviour.

What about the circular error? Can that be used to control the period? Altering the swing by +/-20% changes the circular error by a few ppm - nowhere near the 0.3% control I need.

How have other people done it?

 

[thesnark17]

It seems obvious to me that any system is only going to work if the adjustments are applied to a known state, i.e. the system needs to know where the pendulum is to interfere at the right time. A Hall sensor would do the job nicely, though it would require installation of a permanent magnet on the pendulum. On the other hand, having a magnet on the pendulum would help your electromagnet do its work more efficiently, so on the whole it sounds like a good thing.

Any interference with a pendulum will change its rate. If you give the pendulum a push (impulse in the current direction of travel), the resulting period will be shorter and the clock will speed up slightly. A pull (impulse against the current direction of travel) will act in the opposite way, but comes with the danger of possibly stopping the clock (highly unlikely of course). I recommend setting your pendulum slightly slow and then giving it a push now and then with the electromagnet to speed it up.

The Shortt-Synchronome clock used this principle to keep an electric regulator clock synced to a master pendulum in a vacuum chamber. In that implementation, the two clock rates were compared every 30 seconds, and the slave clock was set up such that it would lose ~6 seconds a day without correction. (They used a spring system to apply correcting power to the slave pendulum, rather than an electromagnet directly, but I don't see why it would be an issue in your implementation. Their implementation needed to work using minimal processing power, while I assume that yours will have no such issue.)

There's no reason why you couldn't compare rates more frequently, or correct for a fairly large error, using this method.

 

[wisty]

Have a look at Bryan Mumford's website Horology | Horological topics (bmumford.com) .
He has written some interesting material about electronically driven pendulums that might be relevant.
I use one of his governors on a Eureka clock I own and it works well.

 

[peterbalch]

Thank you - that's given me some things to think about.

> It seems obvious to me that any system is only going to work if the adjustments are applied to a known state,

Yes and no. If you think of Huygens' "two clocks on a shelf", the tiny impulses were being applied blindly.

But I suspect you're right. It's certainly more efficient to apply the push when the magnet is close to the electromagnet.

> The Shortt-Synchronome clock used this principle

I've now read up on the Synchronome. A Synchronome slave has a falling arm to give it an extra push. That makes the period slightly shorter. There's a mechanism to ensure that extra push only happens "when it's needed". It's a push - not a pull - so (in my case) a side-coil would repel a magnet on the pendulum. The exact mechanism of "when it's needed" escapes me but it only happened if the phase of the slave was behind the phase of the master. One can imagine doing that with modern electronics. As you say, it requires sensing the phase. You can't just send the impulses blindly Huygens-style.

I hadn't thought of a Hall effect sensor. I've just looked at a few datasheets and all seem to require a few mA. I'd like it to be battery-powered which implies a supply current of around 100uA-200uA. I suppose one could momentarily switch on the sensor just when the slave pendulum ought to be at its leftmost - which is rather like what the Synchronome does (I think). The fake pendulums in quartz clocks run on microamps and use a coil as a sensor. According to (Free Pendulum Clock) a coil will produce around 1V and can trigger a micro to wake up and produce a pulse. So a sensor coil will use far less power. If I'm using a micro then I think I can use a single coil for both sensing and driving.

Anyhow, putting aside the question of power consumption, would a Synchronome-style system work for me?

A Synchronome slave is adjusted to run 6 seconds per day slow and then the extra push speeds it up as required. I suspect I can adjust my clock to no better than +/-2 min per day. If I set it to 2min/day slow, will the extra push be able to speed it up sufficiently? Any thoughts?

Reading about the Synchronome has given me a couple more ideas.

I've got a vague memory of someone putting a magnet under their pendulum to "increase gravity". They automatically adjusted the period by using a servo to raise/lower the magnet. It has the advantage that no power is used other than when you're doing the adjusting. Or maybe I could attach a magnet facing the wall then put a sheet of aluminium behind it: the eddy currents in the aluminium would slow the pendulum and I could automatically adjust the position of the aluminium.

Or I replicate the fake pendulum of a quartz clock. That must be giving an extra kick in phase as you suggest. I just turn it on or off as required.

> Have a look at Bryan Mumford's website

A fascinating site.

> I use one of his governors on a Eureka clock I own and it works well.

How long do the batteries last?

So the overall design is a "fake pendulum" driver and adjust its rate by changing the battery voltage. Does it attempt to do phase-lock with a quartz signal or it it adjusting the rate to be correct on average?

In Mumford's experiments, he manages to change the period by over 3% (but the amplitude of oscillation almost doubled!). I guess the adjustment range depends on the Q of the pendulum. Right? I don't understand clock-theory.

Peter

 

[thesnark17]

The Shortt-Synchronome makes up 6 seconds a day through adjustments applied (on average) once per minute. The >15lb pendulum beats with a 2 second period, so the adjustment is being applied approximately every 30 cycles. Given that your pendulum is much lighter (so can be pushed a little harder), and that you can apply corrections more frequently, I would think that 30-60 seconds of adjustment per day could be achievable. But it might not be necessary to adjust that much.

Very fine ATO electromagnetic clocks have a fine adjustment of rate with a magnet, using the artificial gravity principle you discuss. The magnet is mounted below the pendulum and moved up and down to fine-tune the rate without touching the pendulum assembly. This idea could be used to adjust your pendulum to an acceptable rate.

I suggested the Hall effect sensor because it won't affect the pendulum and I had heard of it used on a clock before. Ironically, that clock is on the website mentioned above! A Seconds Beating Oak Clock | Horology Regrettably, I can't give more information than that. I know that there are other possibilities, but I don't know if they can be efficiently powered by batteries.

 

[wisty]

> I use one of his governors on a Eureka clock I own and it works well.
How long do the batteries last?

So the overall design is a "fake pendulum" driver and adjust its rate by changing the battery voltage. Does it attempt to do phase-lock with a quartz signal or it it adjusting the rate to be correct on average?

In Mumford's experiments, he manages to change the period by over 3% (but the amplitude of oscillation almost doubled!). I guess the adjustment range depends on the Q of the pendulum. Right? I don't understand clock-theory.

The batteries (3 rechargeable C size cells) last about 9 months. I have a similar (more modern - my BM governor is 20 years old!) device from Frank Roesky which uses 2 cells and runs for about a year.

Both clocks phase lock to a quartz derived signal. Both also use a small microprocessor to derive the drive signal from the quartz frequency. The software in the microprocessors is tweaked to compensate for the slight difference between the nominal and actual crystal frequency so both clocks keep time to seconds per year.

If the pendulum period varies sufficiently with amplitude (circular error), then the pendulum will inherently phase lock to the drive signal without the need to provide sensing. In precision clocks with long pendulums and small arcs - Fedorchenko is a prime example - the circular error is probably too small to create phase locking and sensing is needed.

 

[praezis]

Hi,

just a short post: what you intend is possible and imo you are on a good way. I developed something similar a couple of years ago - and during a couple of years. The clock in my picture is running for many years now synchronized to DCF77, the German radio time transmitter.



The small mobile box under the pendulum contains everything: 2 AAA cells, the radio receiver with antenna, and the regulator
The pendulum needs to carry a small weak ferrite magnet.
Its cells last for 3-4 years. Max error of the regulated clock is 0.1 seconds absolute. The adjusting range is several hundred seconds/day with that light hollow pendulum, less with a heavy pendulum.
It is true what was already mentioned: the device needs to know the clock's beat rate - unless the clock has a real seconds hand like mine.

Such device makes sense only with clocks running one week or more with one wind. But you added a motor winder to your clock, if I understood well.

If you give the pendulum a push (impulse in the current direction of travel), the resulting period will be shorter and the clock will speed up slightly. A pull (impulse against the current direction of travel) will act in the opposite way,

This is not quite true, just valid before the neutral position. After the opposite is true.

Frank

 

[peterbalch]

wisty:

I'm a beginner in this whole field but I think adjusting circular error is of no use to me. The changes due to circular error are too small.

The normal half-angle of my pendulum is 4 degrees - so its circular error is 0.03%. If I increase the swing by 50% the circular error increases by 0.04%. 50% is a huge increase - the clock might not even cope with that.

Adjusting circular error may work for a high quality pendulum but mine is rather poor. I need to be able to change the period by maybe 0.15%.

Your governor works with a Eureka clock (right?) and a Eureka clock has a hairspring, not a pendulum. So the "circular error" due to amplitude is _very_ much smaller. I presume the effect being used is that the increased supply voltage increases the electromagnet "restoring force" applied to the balance wheel. That's equivalent to increasing the force of gravity for a pendulum.


thesnark17:
> using the artificial gravity principle you discuss

I found the "gravity" auto-adjuster: Auto-Pendulum Tuner


praezis:

> 2 AAA cells ... last for 3-4 years.

Wonderful!

> The adjusting range is several hundred seconds/day

That's what I need. It's very reassuring that you've managed to do it. And with such a small current.

> the device needs to know the clock's beat rate

You mean it has a sensor? A coil or Hall effect? Does the coil attract or repel the magnet?

Have you published the project somewhere?

> you added a motor winder to your clock, if I understood well.

It's published here:

An Auto-winder for a Weight-driven Clock

An Auto-winder for a Weight-driven Clock: I have a clock that's around 150 years old. It's what's known as a "Black Forest Clock" or a "Shield Clock" and is driven by a weight. I can never remember to wind it so I thought I'd build an electric auto-winder. After several …

www.instructables.com

I don't know if the AA cells will last a year - that's just my calculation.

Peter

 

[praezis]

You mean it has a sensor? A coil or Hall effect? Does the coil attract or repel the magnet?

Have you published the project somewhere?

First a tiny reed was used as sensor, later a hall sensor.

No, not yet.

Frank

 

[peterbalch]

I've been doing a little practical research. I made a simple "timegrapher" using an Arduino and a photo-interrupter. It's not calibrated. It uses the Arduino's resonator so it has little long-term stability. But it will tell me in a few minutes whether a change I've made has changed the pendulum's period.

The first thing that's obvious is that the period is far from regular. It varies by up to 2% from one swing to the next. Is that normal for this kind of clock? (Wood frame, steel arbors, brass bushes, weight 550g, steel pendulum shaft, light brass pendulum bob, low Q.)

I guess it might be a problem with my photo-interrupter - the vane that interrupts the beam may not produce a sharp cut-off. But I don't think that's the problem: I can also see short term drift from the draught if I walk near it.


I managed to convince myself that the best arrangement for me is a neodymium magnet on the pendulum facing an electromagnet coil. The coil is to one side of the pendulum and the magnet almost touches the coil on each swing. The coil resistance is 1600ohms and the current is 3mA. If I turn on the coil to repel the magnet, the period decreases by 2%. If it attracts the magnet, the period increases by 2%. 2% is 28min per day so by turning the coil on briefly and occasionally, I should be able to regulate the clock. But it looks to me like I won't be able to phase-lock the pendulum (because the period varies by more than 2% in the short term). I may have to regulate it "on average".

Peter

 

[Gary Heaven]

I've been doing a little practical research. I made a simple "timegrapher" using an Arduino and a photo-interrupter. It's not calibrated. It uses the Arduino's resonator so it has little long-term stability. But it will tell me in a few minutes whether a change I've made has changed the pendulum's period.

The first thing that's obvious is that the period is far from regular. It varies by up to 2% from one swing to the next. Is that normal for this kind of clock? (Wood frame, steel arbors, brass bushes, weight 550g, steel pendulum shaft, light brass pendulum bob, low Q.)

I guess it might be a problem with my photo-interrupter - the vane that interrupts the beam may not produce a sharp cut-off. But I don't think that's the problem: I can also see short term drift from the draught if I walk near it.


I managed to convince myself that the best arrangement for me is a neodymium magnet on the pendulum facing an electromagnet coil. The coil is to one side of the pendulum and the magnet almost touches the coil on each swing. The coil resistance is 1600ohms and the current is 3mA. If I turn on the coil to repel the magnet, the period decreases by 2%. If it attracts the magnet, the period increases by 2%. 2% is 28min per day so by turning the coil on briefly and occasionally, I should be able to regulate the clock. But it looks to me like I won't be able to phase-lock the pendulum (because the period varies by more than 2% in the short term). I may have to regulate it "on average".

Peter

View attachment 752045

That graph looks pretty much what I get when I monitor my hand cut wooden clocks. The shape of each tooth in each gear will fundamentally change the amplitude and therefore time of each pendulum swing. Also your measurement should measure each alternate beat as it is impossible to place a sensor of zero width in the center of the swing. This means you detect from a falling edge, in say the left direction to the next left falling edge. Hard to tell if you are already doing this.

The NTP protocol implementation on arduino is pretty loose so you may see huge (i.e. 30 second) updates as NTP corrects itself. Later versions of the ESP stack are a little better with their smooth update but you need proper hardware (an external GPS) to be sure to have a constant signal to synch to. The internal ESP clock seems remarkably stable however (code word for it took me a while to notice the NTP problems).

Until now I have been using a phototransistor for detection but being swamped by ambient light is an issue. Since I am building it is easy for me to hide a small magnet and use a Hall Effect sensor. I will post back on my success in this area.

Lastly, I have a magnetic impulsed clock and I have found it more repeatable to vary impulse width and repel than it is to vary current or attract. This is analogous I suppose to the feeling one feels in front of a vacuum cleaner vs that one feels standing in front of an air hose. Run your clock a little fast and speed up the pendulum with pulses. I dont think you will ever get total synchronisation though. In my case I vary the pulse width to overcome the inevitable movement of the gap between magnet and coil that humidity in wood creates for us. I measure the signal from the coil as the magnet approaches and then adjust the pulse width out based on it. Less signal = more gap so longer pulse.

I have found the diameter of the magnet in this sort of thing makes a huge difference to the efficiency of the impulse circuit. The magnet diameter needs to be at least as large as the diameter of the impulse coil. Little Neo magnets don't seem perform as well as a large ferrite. The time the pendulum spends in the field of impulse is a factor of how much energy you can impart.

 

[peterbalch]

That graph looks pretty much what I get when I monitor my hand cut wooden clocks.

Thanks. You're the first person who has offered an opinion of any sort. Do you see 1% or 2% change from one pendulum period to the next?

I'd expect hand cut wooden gears to be more uneven than brass gears that have been "mass produced" - even in a cottage industry. So maybe my clock should be doing better.

One post said I should run the "timegrapher" for longer. The pendulum period may change with each rotation of one particular wheel.

> your measurement should measure each alternate beat as it is impossible to place a sensor of zero width in the center of the swing.
> Hard to tell if you are already doing this.

The detector and coil sit on either side (or the same side) of the pendulum just up under the clock so they're hidden in the shadows. Of course, mechanically, that's the worst place to put them. At the bottom end of the pendulum would be much, much better. But it's a wag-on-the-wall clock with an exposed pendulum and to put the coil/sensor at the bottom end of the pendulum would mean screwing it to the wall.

There's a flag on the pendulum (a piece of black tape). So I can only measure the whole period - not the two "beats" as horologists call them.

The coil is at the side. It could be on the same or opposite side to the sensor.

I tried putting the coil behind the pendulum in the middle of the swing but that didn't work. The pendulum was pulled towards the wall so much it would actually hit the coil. Most people put the coil under the centre of the bottom of the swing pointing upwards - so that isn't a problem for them.

> The NTP protocol ...

I'm not at that stage. I'd be happy if the clock were as good as a cheap quartz clock so a simple crystal is good enough for me.

> Until now I have been using a phototransistor for detection but being swamped by ambient light is an issue.

Putting it in the shadows under the clock helps. Plus, rather than using it as an on/off signal, you can use an ADC to measure the light. Measure it with and without turning on the LED and look at the difference.

> it is easy for me to hide a small magnet and use a Hall Effect sensor.

Why can't you use the "main" magnet? Why do you need a second one?

> I have found it more repeatable to vary impulse width and repel than it is to vary current or attract.

Interesting. I experimented with a short, stong pulse vs a long, weak one. Long and weak seemed better to me.

I want it all to run on AA cells - I don't want trailing wires. So I worry a lot about power consumption - under 200uA would be good. I was trying to find the best combination of

• pulse length vs current
• pulse timing: approaching, at closest, retreating
• attract vs repel or both
• pulse blindly or use sensor

Does it work if you send the pulse blindly? Yes but it needs a huge current - an average of milliAmps. It ought to work. Think of Huygens' "two clocks on a shelf". But his were precision clocks. Mine isn't. That 2% variation in period is a killer.

So maybe a sensor is needed. But a sensor takes current. You'd normally put 10mA through the LED. I can't afford 10mA. Can I switch on the LED for a few microseconds when I think the pendulum should be there? If the pendulum isn't there yet it needs speeding up; if the pendulum is already there it needs slowing down.

So I changed to a Hall sensor. I want the processor to spend most of its time asleep. It would be nice to have a sensor that wakes up the processor with an interrupt. So the sensor is always on. More current!

It's all getting a bit complicated and I couldn't get the total current low enough.

A reed switch takes no power but a reed switch contains soft iron and the magnet would just clamp onto it.

> I vary the pulse width to overcome the inevitable movement of the gap between magnet and coil that humidity in wood creates for us.

Magnets are complicated. You do them at school and you think you understand them but ...

I read in (I think) a JPL webpage that a good first approximation is that the attractive force falls off with distance cubed. It falls off really fast.

I've got an additional problem: the maximum angle varies a lot so the "gap" at maximum swing varies.

> I have found the diameter of the magnet in this sort of thing makes a huge difference

Now that's interesting.

Magnets are complicated. You do them at school and you think you understand them but ...

Say the N pole is being attracted to your coil and the S pole is being repelled. If it's a disk magnet then they're both close to the coil. Does that mean their effects cancel each other out? With a long magnet they don't cancel.

There's a Wikipedia page. I might go through the math and see what it says about a disk magnet vs a long magnet.

Peter

 

[TQ60]

Maybe re-think this.

Place the sensor someplace else like a gear and make it sharp acting, something like the trip on a striking clock.

Next, determine the rate the pendulum needs to be, program pulse generator for the rate and place coil where it needs to be to drive it.

 

[peterbalch]

Place the sensor someplace else like a gear

To measure the average speed over the last few minutes?

> program pulse generator for the rate

Then send the pulses blindly - not knowing where the pendulum is?

I know what frequency the pendulum ought to be going at - it's 3168 cycles per hour. How does measuring the current average speed help me?

I tried sending the pulses blindly, adjusting the frequency by hand. It only worked with a huge current.

thesnark17 said some time ago "any system is only going to work if the adjustments are applied to a known state". That's not true. You can apply it blindly - in an unknown state - and it will work. But it takes far less current if the pendulum is near the electromagnet. That's because the magnetic force falls off very quickly with distance.

Most people wait until the pendulum period drifts close to the fixed pulse period then hope it will synchronise and stay close to it. For them it works because they've built a good quality pendulum.

But it doesn't work for me because the pendulum period varies so much - 1% or 2% from one swing to the next. So the pendulum is often several millimeters away when the pulse comes. Increasing the current fixes that but then my battery would only last a couple of weeks.

So I need to know where the pendulum is - I need a sensor. But sensors take current that I can't afford.

Peter

 

[peterbalch]

Gary

There's a Wikipedia page. I might go through the math and see what it says about a disk magnet vs a long magnet.

It turns out I don't need to do the math. On that page there's a graph called "Exact force between two coaxial cylindrical bar magnets for several aspect ratios".

It shows that the force due to a a disk magnet (L=R/2) falls off faster than a long magnet (L=4R). At a distance of R/2, the force of the disk magnet has fallen to 25%, whereas the force of the long magnet has fallen to 40%.

Peter

 

[Gary Heaven]

Gary



It turns out I don't need to do the math. On that page there's a graph called "Exact force between two coaxial cylindrical bar magnets for several aspect ratios".

It shows that the force due to a a disk magnet (L=R/2) falls off faster than a long magnet (L=4R). At a distance of R/2, the force of the disk magnet has fallen to 25%, whereas the force of the long magnet has fallen to 40%.

Peter

In my case I am using a open air coil to pulse a disk magnet. This forms a magnetic circuit (i.e. the coil is an electromagnet of certain dimensions interacting with a fixed magnet of certain dimensions) and the combination of the two needs to be considered. The more efficient this circuit is the less electrical input is needed for the coil.

 

[Gary Heaven]

To measure the average speed over the last few minutes?

> program pulse generator for the rate

Then send the pulses blindly - not knowing where the pendulum is?

I know what frequency the pendulum ought to be going at - it's 3168 cycles per hour. How does measuring the current average speed help me?

I tried sending the pulses blindly, adjusting the frequency by hand. It only worked with a huge current.

thesnark17 said some time ago "any system is only going to work if the adjustments are applied to a known state". That's not true. You can apply it blindly - in an unknown state - and it will work. But it takes far less current if the pendulum is near the electromagnet. That's because the magnetic force falls off very quickly with distance.

Most people wait until the pendulum period drifts close to the fixed pulse period then hope it will synchronise and stay close to it. For them it works because they've built a good quality pendulum.

But it doesn't work for me because the pendulum period varies so much - 1% or 2% from one swing to the next. So the pendulum is often several millimeters away when the pulse comes. Increasing the current fixes that but then my battery would only last a couple of weeks.

So I need to know where the pendulum is - I need a sensor. But sensors take current that I can't afford.

Peter

Sensing the time of arrival of a pendulum near the end of its swing is always going to be a bit erratic as it is moving much slower. Much better to measure mid swing to get a more vertical change in sensor input.

 

[TQ60]

To measure the average speed over the last few minutes?

> program pulse generator for the rate

Then send the pulses blindly - not knowing where the pendulum is?

I know what frequency the pendulum ought to be going at - it's 3168 cycles per hour. How does measuring the current average speed help me?

I tried sending the pulses blindly, adjusting the frequency by hand. It only worked with a huge current.

thesnark17 said some time ago "any system is only going to work if the adjustments are applied to a known state". That's not true. You can apply it blindly - in an unknown state - and it will work. But it takes far less current if the pendulum is near the electromagnet. That's because the magnetic force falls off very quickly with distance.

Most people wait until the pendulum period drifts close to the fixed pulse period then hope it will synchronise and stay close to it. For them it works because they've built a good quality pendulum.

But it doesn't work for me because the pendulum period varies so much - 1% or 2% from one swing to the next. So the pendulum is often several millimeters away when the pulse comes. Increasing the current fixes that but then my battery would only last a couple of weeks.

So I need to know where the pendulum is - I need a sensor. But sensors take current that I can't afford.

Peter

Not to be rude, build a better pendulum.

The pendulum is the "standard" that determines the accuracy of the time.

Differences in the time train may cause it to swing different angular amounts resulting in slight differences in rate, and it seems this is what you are looking t controlling.

Having a sensor looking at the pendulum or anything else is wagging the dog.

Having a tightly controlled device firing at the specific time the pendulum "should" be at some point would give it a push is it was behind, or pull it if it was ahead.

If you started the pendulum completely out of phase, not in correct place to receive pulse, it would naturally slow down due to lack of power.

At some point it will get close enough to the power pulse that it will get in sync and go.

Many electric clocks use some pulse to drive the pendulum, then the pendulum simply strikes a ratchet gear that pushes the motion works.

The pendulum still controls the rate, external forces apply pulse to pendulum to keep it going.

ATO comes to mind.

 

[peterbalch]

Not to be rude, build a better pendulum.

That would be sensible but I'm trying to get a regulate an existing clock. You can see pictures of the clock on this page.

I assumed that the way to do it was to use and electromagnet to synchronize the pendulum - that's what the majority of people do. For the reasons given above, I don't think that's going to work.

I'm currently thinking that maybe I can regulate it by automatically adjusting the tension in the chain as explained on this page (scroll down to find an animation).

I want it to run for a year on AA cells. There are other possibilities that take very little power: slowing the pendulum down using eddy currents or speeding it up using a permanent magnet. I haven't investigated them yet.

Peter

 

[TQ60]

Did not read much of the linked page, but I see where you could suspect variable rate.

We are working on a self winding clock that has 2 offset weights that get "fired" by a solenoid, both are active except for the fraction of a second it takes to fire the weight.

A suggestion for your winding...

We have another clock that rides along a rack gear, the clock is the weight, to wind, you lift the clock.

Shift gears, your existing gearbox is still used, but it gets configured so it rides the chain and becomes the weight.

Chain passes through the unit so you need to arrange things to balance it.

Your motor can be controlled by MAGNETIC REED switch.

A magnet attached to the wall behind the chain, when the motor-weight aligns with the magnet, the switch closes causing the motor to run, raising the unit until it gets away from the magnet.

With the RPM being slow, the tension on the chain should be very constant with tiny "blip" when motor starts.

Pendulum can still be improved for better consistency.

The system we mentioned before would keep it in sync if properly built.

If you really want to get crazy...

GPS has an output called 1PPS, one puse per second.

This is what is used for syncing most everything.

If your pendulum rate is such that it can divide or multiply to 1 pps. Second then it may be possible to fire the coil at 1 pps and sync the pendulum every how ever many strokes it has in a second.

As long as it is whole number of 1/2 swings in each second, it should be at the same position in each exact second.

 

[peterbalch]

We are working on a self winding clock that has 2 offset weights that get "fired" by a solenoid, both are active except for the fraction of a second it takes to fire the weight.

Do you have any photos or drawings you could show here?

I think that solenoids are very inefficient. I guess it's OK if it's mains powered. Personally, I don't want any trailing wires.

> Shift gears, your existing gearbox is still used, but it gets configured so it rides the chain and becomes the weight.
> Chain passes through the unit so you need to arrange things to balance it.

Like this arrangement? It's a "monkey climbing a rope". The monkey starts climbing when the weight touched the bottom of the case and stops climbing when it touches the clock. He calls it a "sloth".

I started off with various designs like that and simplified it repeatedly to get what I have now. My view is that the fewer parts there are, the fewer parts will go wrong.

> Your motor can be controlled by MAGNETIC REED switch.

It is. As I discuss in my Instructable, reed switches are very much more reliable than the tilt switches I started with.

> A magnet attached to the wall behind the chain, when the motor-weight aligns with the magnet, the switch closes causing the motor to run, raising the unit until it gets away from the magnet.

Mechanically, it's hard to design something that climbs a chain reliably. You need extra guides and jockey wheels - just compare the complexity of the "sloth" with my design.

I'd prefer not to have things attached to the wall. I considered a jockey wheel that sensed when the "monkey" reached the bottom of the chain-loop. But once again, that adds a lot of extra complexity.

> Pendulum can still be improved for better consistency.

How? It's an old clock and I don't want to "modernise" it.

> The system we mentioned before would keep it in sync if properly built.
> it may be possible to fire the coil at 1 pps and sync the pendulum every how ever many strokes it has in a second.

How - while running it on AA cells that last a year?

> it should be at the same position in each exact second.

It isn't. The period is not constant enough.

I'd love to see your design.

Peter

 

[TQ60]

The GPS part would need to be separate from your battery requirement.

The self winding part should be simple.

Imagine a tall triangle shaped object.

It Contains your motor and batteries with the heavy stuff at the bottom.

You locate the chain drive above center of gravity.

Above that in the point of the triangle is where the chain enters.

Tubing can be used as a chain guide.

The unit is fully self contained, the Reed switch(s) go on the backside.

You can have a few in parallel just for redundancy.

Get the magnets from inside an old hard drive as they are very strong.

You can make something decorative to hold it on the wall behind the motor unit.

If you place the Reed switch near the top, it should always be covered.

The motor should have some flywheel effect, meaning the motor should spin up when started and still coast after shut off so it should climb the chain a bit after the switch turns off.

The photo is our battery we will be using in our 1899 clock.

It WAS a communications battery rated for 7.5 VDC, 2 Lithium batteries in series.

We needed 3 volts so now parallel.

Maybe size of 4 AA batteries, but 3.4 VDC at 6.8 Amp hours.

The bright flashlights use a good size cell and have a charger to charge them, that is another option for power.

You could use different battery to extend run time.

 

[Gary Heaven]

Peter

You asked if I see a 1% or 2% change in time from one pendulum swing to the next and the quick answer to that is no. If I see changes of that order I start looking for sticky bearings or teeth that are not meshed and go and find them. I get fluctuation but it is an order of magnitude less - 400 us on a 0.5 second pendulum.

This is for a wooden clock that is like the "Magica" clock but with a carbon pendulum over approximately 24 hours. I am attributing the short term fluctuation to a couple of 8 tooth pinions and the main impulse wheel not being as balanced as it could be.. I am happy with the over all performance of this clock, runs to within about 5 minutes over 3-4 months when it needs a battery change.

Having read your other post you mentioned about the clock you are trying to synchronise, I would suggest that the clock needs a good service before applying any theory. Friction and other losses will counter any amount of theory at this point. I concur with the others that you probably should not be trying to make a 1960 Morris Minor go as fast and efficiently as a 2023 Lexus Hybrid.

 

[peterbalch]

I get fluctuation but it is an order of magnitude less - 400 us on a 0.5 second pendulum.

Gary

Thanks. That's interesting. But I wonder how relevant it is to me.

You could be right that my clock requires a good service but I'd be happier if I knew that it was behaving particularly badly. Maybe that's normal behaviour. The general feeling seems to be "two minutes a day is about right" and that's what I see. Maybe 1%-2% is normal.

I presume that the 1% variation is due to "escapement error" - a vague term. Our suspensions and escapements are very different. Your clock has a ball bearing suspension (??) whereas mine has a trapeze. Your clock has solenoid driver (??) whereas mine has a recoil-strip.

I think that the 1% "escapement error" may be due to non-linear - i.e. chaotic - effects in the escapement. Your solenoid could well have much smaller non-linear effects. My trapeze suspension makes the pendulum into a "double pendulum" which is about as chaotic as you can get.

Fig 3 of this paper shows a graph with 1%-2% variability. The clock is made of Lego! Their Q is 1500, mine is 200; maybe that alone explains what I'm seeing. They talk a lot about the relationship between Q and noise.

> trying to make a 1960 Morris Minor go as fast and efficiently as a 2023 Lexus Hybrid.

More like trying to fit cruise control. I'm not allowed to change the Morris Minor, just automate the accelerator pedal.

Everything you say could be true. I keep coming back to the same question: what is "normal"?

Peter