Measuring Clocks

[John HaineParticipant @johnhaine32865]

I’m not sure is the same as any of them, but my approach would be to –

• Have a separate amplitude detector (AD) photogate offset from BDC by the required amplitude
• When this is triggered, inhibit the next impulse
• Examine the average hit/miss sequence from the AD and adjust the impulse to drive this to hit/miss/hit etc.

I would apply constant length impulses, ideally symmetrical around BDC, and adjust their strength using a DAC.  This could either be a real DAC driving a variable power supply or possibly use PWM of the coil current.

[Joseph Noci 1Participant @josephnoci1]

…. but less energy makes my pendulum in air unreliable

I would suggest if that is the case, it will not be very stable or consistent at any amplitude then. Somewhere your losses a far to high then. Do you know what sort of Q the pendulum has?

[John Haine]

On John Haine Said:

I’m not sure is the same as any of them, but my approach would be to –

• Have a separate amplitude detector (AD) photogate offset from BDC by the required amplitude
• When this is triggered, inhibit the next impulse
• Examine the average hit/miss sequence from the AD and adjust the impulse to drive this to hit/miss/hit etc.

I would apply constant length impulses, ideally symmetrical around BDC, and adjust their strength using a DAC.  This could either be a real DAC driving a variable power supply or possibly use PWM of the coil current.

How do you cope with Tilt with this method? – Tilt will be seen as an amplitude change.

[Joseph Noci 1Participant @josephnoci1]

Re my post with all the Bold characters – I did not do that – just clicked QUOTE and added my bit, submit, and the forum bots did it all on their own..

Time to go away again…

[blowlampParticipant @blowlamp]

Anyone familiar with the Grand Seiko Spring Drive movement will see certain similarities with what Dave is trying to do.

Spring Drive is an entirely mechanically powered movement, which has a braking wheel in replacement of a conventional escapement. An electronic circuit determines the rotational speed of this wheel, which is referenced back to a quartz crystal and applies a braking force to maintain accuracy. The movement performs with a deviation that is usually less than +/- 1 second per day.

At the end of the day, the mechanical components aren’t determining the accuracy, they are simply following that of the quartz crystal, but not completely faithfully because of how the movement is ‘governed’.

 

Martin.

[John HaineParticipant @johnhaine32865]

Joe, this is essentially based on Bateman’s method with some additional control.  I’m not sure how significant “tilt” (i.e. the actual ground level moving) is for most locations.

[Joseph Noci 1Participant @josephnoci1]

I hope I am not unique in my pendulum’s tilt affinity…

You will have seen these in my posts on the other forum- How the odd/even beat delta increases with environmental Temp increase – This was verified as being due to tilt – with Doug’s help some time back.

Also not sure its only ground level moving that is the issue – a lot depends on the Pendulum mounting – the pivot is a meter above the photogate and that’s a long lever arm when the target is microseconds, or even less. Sitting on a average home floor, or bolted to its wall – there will be tilt..

[Joseph Noci 1]

On Joseph Noci 1 Said:

…………….. I’m detecting changes in amplitude by measuring transit time…….

Measuring elapsed time requires some sort of time base, window, clock, i.e. , an infiltrating clock..

The measurand remains a slave of that time…

Greetings Joe, good to have another big-hitter on the case!

Can anyone explain how measuring amplitude with a separate low accuracy clock infiltrations and how the pendulum becomes a slave to it?

I say there is no synchronisation because I can’t identify the mechanism.  So far neither have the critics, and, with respect, it’s not enough just to state it!

If I’m wrong, and the pendulum is somehow slaved to the microcontrollers crystal, why doesn’t the Shortt Synchronome suffer the same defect?  The Shortt-synchronome impulses the master pendulum by dropping a gravity arm timed by the slave clock.  Why isn’t that an infiltration.  Woodward, see quote above, calls it  “the servant’s kitchen clock”.  What’s good enough for Shortt, is good enough for me.  And bear in mind, we may be about to conclude the Shortt clock has a design flaw, when I don’t believe it has,

I cheerfully accept the suspicion, but where’s the proof?  I don’t mind being wrong, but I’m not giving in without a good explanation.

🙂
Dave

[Joseph Noci 1]

On Joseph Noci 1 Said:

…. but less energy makes my pendulum in air unreliable

I would suggest if that is the case, it will not be very stable or consistent at any amplitude then. Somewhere your losses a far to high then. Do you know what sort of Q the pendulum has?

Agreed,  it does seem stiff, and there is probable cause.  I broke the suspension spring when reassembling the mechanicals and replaced the spring with similar.  It’s made from a disposable razor-blade, which is thin springy stainless.  Possible the replacement is thicker or a different alloy.  The razor was a different make.   I’ll measure Q later – added to the TODO list.

Dave

[John Haine]

On John Haine Said:

… I’m not sure how significant “tilt” (i.e. the actual ground level moving) is for most locations.

Joe lives near the sea-shore, so the tide may be moving the ground.  Million of tons of seawater pressing on many square kilometres of squidgy ground. Wonder what the geology is underneath?   Might include a thick layer of water soaked sand…

Dave

[duncan webster 1Participant @duncanwebster1]

I’ll don my asbestos suit before I put in my two pennorth. If the factors which slow a pendulum don’t change, I dont think it matters what you do to keep the pendulum swinging as long as you do exactly the same on every swing. Unfortunately at least atmospheric conditions do change, unless you run in a controlled atmosphere (temperature, pressure and humidity), so the impulse has to be varied in some way. It seems sensible to use low expansion pendulum and reduced pressure to reduce atmosphere effects, and to impulse at mid swing which disturbs the pendulum least.

As I don’t have analogue measure of amplitude one approach could be to count the swings between detection of excess. If more than some value, reduce the impulse a tad, if less than some smaller value increase it. The latest clock has a very low Q for reasons I don’t understand, so thi will have to wait until I rebuild the electronics on Clock1

[John HaineParticipant @johnhaine32865]

Dave –

“The Shortt-synchronome impulses the master pendulum by dropping a gravity arm timed by the slave clock. ”

Well, not quite!  That places a “dead roll” surface formed on the gravity arm onto the impulse roller which is fitted to the pendulum, but the moment when the arm “rolls off” and the impulse is actually generated is determined by the arm/roller geometry. Likewise on my “Arduinome” the roller on the arm is placed onto a dead roll surface on the pallet, and the impulse is generated as a curved ramp on the pallet passes under the roller, so again the impulse timing is determined by the geometry.  (Though I have to say that when I looked at the pallet ramp surface this morning there was a slight wear mark suggesting that the roller might be being lifted a bit early – need to check,  It has been operating for nigh on 5 years.)

[SillyOldDuffer]

On SillyOldDuffer Said:

On Joseph Noci 1 Said:

…………….. I’m detecting changes in amplitude by measuring transit time…….

Measuring elapsed time requires some sort of time base, window, clock, i.e. , an infiltrating clock..

The measurand remains a slave of that time…

Greetings Joe, good to have another big-hitter on the case!

Can anyone explain how measuring amplitude with a separate low accuracy clock infiltrations and how the pendulum becomes a slave to it?

I say there is no synchronisation because I can’t identify the mechanism.  So far neither have the critics, and, with respect, it’s not enough just to state it!

If I’m wrong, and the pendulum is somehow slaved to the microcontrollers crystal, why doesn’t the Shortt Synchronome suffer the same defect?  The Shortt-synchronome impulses the master pendulum by dropping a gravity arm timed by the slave clock.  Why isn’t that an infiltration.  Woodward, see quote above, calls it  “the servant’s kitchen clock”.  What’s good enough for Shortt, is good enough for me.  And bear in mind, we may be about to conclude the Shortt clock has a design flaw, when I don’t believe it has,

I cheerfully accept the suspicion, but where’s the proof?  I don’t mind being wrong, but I’m not giving in without a good explanation.

🙂
Dave

Re the Shortt clock, I think John has summed it – It is not timed by something else that has an inherent time:

the impulse  generated is determined by the arm/roller geometry.

In your inferred timing instance, you determine amplitude as follows:

I’m detecting changes in amplitude by measuring transit time. Granularity ±1mS 

&

I’m using a “Kitchen Clock” too, but only to measure amplitude.

In my book, if you use the result of any measurement to control the clock – rate, amplitude, whatever,  – if the measurement required ANY external source of non-pendulum derived time,  then you are syncing, no matter how loosely, to an external time source. If you use a ‘clock’ to measure Amplitude, and you control to that value, you are synced to that clock, since your aim is to control the amplitude –

If I use a clock to feed a timer that counts clocks between the leading and trailing edge of the vane blocking and unblocking the photogate each beat at BDC, AND I use that value to determine amplitude, AND I use that information to correct and control the amplitude, I am synced to that external clock. That is a No-No..

If I measure amplitude, say using the Photogate at the end of the swing as suggested by John, and control amplitude to that, and then I use the photogate/timer value to MEASURE the amplitude to see how well I am controlling amplitude, then thats good…But use that value to control the amplitude and external time has infiltrated – you can wave whatever wand you may wish to kid yourself, but the clock is now part of the pendulum…

[blowlampParticipant @blowlamp]

In the Shortt, the slave clock is the detached escapement of the master and does nothing to it other than keep it moving in the least disruptive way through its consistent control of the almost invariable gravity arm.

 

Martin.

[blowlamp]

On blowlamp Said:

In the Shortt, the slave clock is the detached escapement of the master and does nothing to it other than keep it moving in the least disruptive way through its consistent control of the almost invariable gravity arm.

 

Martin.

In the Duffer clock, the microcontroller’s clock is detached from the master and does nothing to it other than keep it moving in the least disruptive way…

Specifically, the microcontroller establishes the pendulum’s amplitude (which could be done externally without a clock) and uses that to increase or decrease impulse power.  Unlike the Seiko it does not alter the oscillators frequency or phase.

What I need is an explanation as to how changing impulse power takes over the pendulum.  The impulse would have to alter the pendulum’s period, and it doesn’t.   What’s the mechanism?

Can I remind everyone that the governor doesn’t operate on every beat.  It starts by applying a series of  corrections that stabilise the amplitude and then disconnects.  Thereafter it does nothing for many beats unless the amplitude moves out of range, which might happen much later due to the bob gaining or losing energy due to the impulse being slightly too strong or weak.   After tweaking the impulse power for a better energy balance, the governor sleeps until the amplitude rises or falls too far again, if ever.

The process hunts for the impulse power that achieves a target amplitude, and having found it leaves well alone, unless a slow loss or gain of energy detects another power correction is needed.  It doesn’t play squash with the bob, the action is irregular, and it’s not a phase locked loop.  I see no way the pendulum can phase lock with the microcontrollers oscillator.

Might be wrong, but so far the case for the prosecution lacks evidence!

Dave

PS Joe’s post appeared whilst I was typing.  Bedtime here, so I’ll read it in the morning.

 

 

[blowlampParticipant @blowlamp]

Dave.

The Seiko Spring Drive doesn’t have an oscillator, unless you are talking about the quartz crystal, but obviously that frequency doesn’t change as it would render it useless.

If a change of impulse power doesn’t change the period of the pendulum, then why are you changing the impulse power?

The Shortt clock never seeks to change the impulse power and still manages to maintain an accuracy of +/- 1 second in every 1 to 12 years, depending on whose figures are used.

 

Martin.

[Joseph Noci 1Participant @josephnoci1]

What I need is an explanation as to how changing impulse power takes over the pendulum.  The impulse would have to alter the pendulum’s period, and it doesn’t. 

 

That is not really the point – the point is how you determine that you need to change impulse power – you decide this because you decided the amplitude had changed.

How did you determine that the amplitude has changed? If you used an external clock source to determine the amplitude, then you are synced to that clock, since the determination depends on that external clock.

It may seem subtle, but it’s there…

[blowlamp]

On blowlamp Said:

Dave.

The Seiko Spring Drive doesn’t have an oscillator, unless you are talking about the quartz crystal, but obviously that frequency doesn’t change as it would render it useless.

Yes, I mean the oscillator controlled by the quartz crystal.  I thought you introduced the Seiko because it has two clocks:, saying:

Spring Drive is an entirely mechanically powered movement, which has a braking wheel in replacement of a conventional escapement. An electronic circuit determines the rotational speed of this wheel, which is referenced back to a quartz crystal and applies a braking force to maintain accuracy.

If your description is correct, the Seiko’s escapement is controlled by the crystal oscillator.  Thus the watch doesn’t have an ‘entirely mechanically powered movement’.

Looks very much to me to be what Joe calls an infiltrating clock.  The mechanical movement is influenced by another clock which is electronic, period controlled accurately with a quartz crystal.   I wish Seiko were represented on the forum, because I think they’re guilty.

If a change of impulse power doesn’t change the period of the pendulum, then why are you changing the impulse power?

I’ll say it again.   Circular Error varies with Amplitude, therefore all pendulum (and balance wheel) clocks try to keep amplitude constant.  Circular Error is always present, but is small below about 5°, to the point that the pendulum’s period is nearly isochronous.  The simple escapements in basic pendulum clocks cause period error by rough impulsing.  Better clocks strive to reduce Circular Error by various methods, such as a gravity escapement.

The Shortt clock never seeks to change the impulse power and still manages to maintain an accuracy of +/- 1 second in every 1 to 12 years, depending on whose figures are used.

True, but consider how Shortt identified what the impulse power should be, and how it is delivered.  Delivery is by a gravity arm, which is nearly, bit not quite, mechanically independent of the pendulum.   It’s a good solution.

However, the magnitude impulse force delivered by the gravity arm was determined experimentally.  Shortt set a prototype clock going and measured period relative to astronomical time, resulting in many design improvements.  He also measured the pendulum’s amplitude, which enabled him to fine tune the geometry and operation of the gravity arm to deliver an optimised impulse.   I’m trying to do the same, the difference being I find relative amplitude by measuring how long the pendulum rod occludes the opto-sensor.   Shortt experimented manually, I’m using a microcontroller to do it.

 

Martin.

Interesting stuff eh, you’re making me think!  Thanks for contributing.

Dave

[NealebParticipant @nealeb]

Joe’s argument suggests to me that any electronic maintaining technology is automatically ruled out as an element of electronic timing is required and that means that pendulum timing is somehow linked to the external electronic timing.

Reasoning is that the pendulum swing is dependent on impulse power delivered. That power is a function of both amplitude – which can be controlled by purely electronic means – and pulse duration which must be determined by some kind of electronic timer. That might be a specific quartz oscillator, cheap ceramic resonator providing micro-controller clock, or something else, but time source there must be. So, your pendulum is no longer an independent entity in time-keeping terms.

I’m not a clock nut myself (although I find the current discussion fascinating to follow!) but would argue that you can measure the interference or level of synchronisation between pendulum and external time source by considering what happens if the external clock drifts. My feeling is that the external “clock” needs only reasonable short-term stability in order to allow the electronic governor to function reliably and any longer-term drift will alter slightly the power delivered per impulse but the governor logic will adapt to that such that average power delivered to the pendulum will remain constant even though individual pulse length will change. I am assuming that some kind of PID control is in there as P alone will not reduce amplitude error to zero.

Edit – and having read Dave’s last reply, I realise that he is dependent on external timing as he indirectly measures amplitude by measuring speed at a fixed point of swing, by measuring time taken to cover a fixed distance. Ah well, it was a good argument while it lasted…

[Joseph Noci 1]

On Joseph Noci 1 Said:

What I need is an explanation as to how changing impulse power takes over the pendulum.  The impulse would have to alter the pendulum’s period, and it doesn’t. 

 

That is not really the point – the point is how you determine that you need to change impulse power – you decide this because you decided the amplitude had changed.

With respect it is.  Not enough in my book to assert the mere presence of a second clock proves infiltration.  Smacks of guilt by association, and surely depends on how the second clock is used.

Though I recognise infiltration might happen. I don’t see, in my clock, how it does.  Anyone who can explain the mechanism by which syncing can occur will convince me.

How did you determine that the amplitude has changed? If you used an external clock source to determine the amplitude, then you are synced to that clock, since the determination depends on that external clock.

It may seem subtle, but it’s there…

Too subtle for me.   I’m not convinced yet that infiltration occurs, and, even it does, that the effect is significant.  The pendulum’s amplitude is stabilised, not varied, and the impulse does not alter period or phase in the long run.

What do the team think of this compromise.  These steps:

1. The pendulum in this clock takes up to a few hours to settle after a cold-start.  It is not a timepiece yet.
2. After settling, I could switch the governor on to find the impulse power needed to maintain amplitude at the chosen target, say 4°.  Infiltration doesn’t matter because the clock still isn’t operational.
3. Impulse power is set to the fixed value found by the governor, and the governor turned off.   Now the microcontroller’s clock is disconnected entirely, so it cannot infiltrate.   The clock still isn’t operational.
4. Final step is to set the clock accurately to GPS UTC.  This is a forthcoming attraction because I’m not sure how best to do it.  Watch this space.
5. The clock becomes operational after it’s been set to UTC.  Only now can it’s performance be assessed.

I assuming there’s no objection to using the governor to initialise the clock provided it’s switched off before the clock is operational?

Many thanks for the contributions, all from people whose opinions and knowledge I thoroughly respect.

Dave

[SillyOldDuffer]

On SillyOldDuffer Said:

On Joseph Noci 1 Said:

What I need is an explanation as to how changing impulse power takes over the pendulum.  The impulse would have to alter the pendulum’s period, and it doesn’t. 

 

That is not really the point – the point is how you determine that you need to change impulse power – you decide this because you decided the amplitude had changed.

With respect it is.  Not enough in my book to assert the mere presence of a second clock proves infiltration.  Smacks of guilt by association, and surely depends on how the second clock is used.

Though I recognise infiltration might happen. I don’t see, in my clock, how it does.  Anyone who can explain the mechanism by which syncing can occur will convince me.

How did you determine that the amplitude has changed? If you used an external clock source to determine the amplitude, then you are synced to that clock, since the determination depends on that external clock.

It may seem subtle, but it’s there…

 

Too subtle for me.   I’m not convinced yet that infiltration occurs, and, even it does, that the effect is significant.  The pendulum’s amplitude is stabilised, not varied, and the impulse does not alter period or phase in the long run.

 

What do the team think of this compromise.  These steps:

1. The pendulum in this clock takes up to a few hours to settle after a cold-start.  It is not a timepiece yet.
2. After settling, I could switch the governor on to find the impulse power needed to maintain amplitude at the chosen target, say 4°.  Infiltration doesn’t matter because the clock still isn’t operational.
3. Impulse power is set to the fixed value found by the governor, and the governor turned off.   Now the microcontroller’s clock is disconnected entirely, so it cannot infiltrate.   The pendulum clock still isn’t operational.
4. Final step is to set the clock accurately to GPS UTC.  This is a forthcoming attraction because I’m not sure how best to do it.  Watch this space.
5. The clock becomes operational after it’s been set to UTC.  Only now can it’s performance be assessed, and there is no second clock

Is there any objection to using the governor to initialise the clock when it’s switched off before the clock is operational?

Many thanks for the contributions, all from people whose opinions and knowledge I thoroughly respect.

Dave

[John HaineParticipant @johnhaine32865]

I’m bemused by this discussion, and the use of the term “infiltrate”. The latter is overloaded with negative connotations and could mean anything.

• If you want to counter the effect of circular deviation then you need to control amplitude.
• Using a separate (reference) clock which is asynchronous to the pendulum will give you a measure of amplitude (or actually velocity) based on photogate time in terms of the number of cycles of the reference which fluctuates by +/- 1 cycle of the reference, which can be averaged out.
• There is no way that then using that average to adjust the impulse to keep the amplitude constant can “synchronise” the pendulum.
• If by some lucky fluke the reference was synchronised then the situation is no different.

The only difference I would have with Dave is that I think his control scheme is too slow, better to have a much tighter loop.

[Michael GilliganParticipant @michaelgilligan61133]

Could you please elaborate on [5] Dave ?

As I read it, you are now proposing to only use your clever electronics for an initial set-up and will then set the pendulum free to do its thing.

[ surely this must be misinterpretation on my part ]

MichaelG.

 

[Nealeb]

On Nealeb Said:

…Edit – and having read Dave’s last reply, I realise that he is dependent on external timing as he indirectly measures amplitude by measuring speed at a fixed point of swing, by measuring time taken to cover a fixed distance. Ah well, it was a good argument while it lasted…

Welcome Brian!

Your post on top of Joe’s, and the earlier comments, has got me worried!

Though I’m somewhat confident the governor mechanism that triggered this conversation isn’t a meaningful infiltration, there might be another that is!

The pendulum clock’s electronics are implemented with a digital microcontroller which is enormously convenient and flexible compared with a mechanical or analogue equivalent.

The problem is that a digital computer executes instructions at a rate determined by a crystal oscillator.  The computer ticks, so everything it does is time dependent and phased with the crystal oscillator.

My clock works thus:

• A function called loop() repeats endlessly.   It logs results to the PC, receives commands from the PC, and receives data from an optional GPS module, all these via Serial UART.  Also, interrogates the pressure, temperature and humidity sensor to determine environmental corrections, mainly temperature compensation.   It also manages control settings.  None of it is time critical.
• Time critical events are handled by interrupt functions.  When an event occurs, loop is stopped, the computer saves state, and transfers control to the interrupt function.  They are:
  • GPS second event, received from an optional module, and only used to set the clock and measure period accurately when the clock is in learning mode, not time-keeping.  Not normally active when the clock is operational,
  • Single Shot Timer.  A hardware counter timer is used to deliver accurately metered pulses.   The counter/timer is clocked by the crystal oscillator and started by the pendulum event interrupt function.
  • Pendulum events are triggered by the pendulum breaking the opto-sensor beam.  The function triggers a single shot pulse, and also measures the pulse characteristics for diagnostic and set-up purposes.  Apart from amplitude as discussed, the measurement values do not influence the pendulum.

My concern is this: a digital computer operates in steps of a time quanta is determined by the computer’s crystal oscillator.  So there is a delay of ‘n’ quanta between the event and the computer reacting to it.  Many stages, but all executed in a time quanta derived from a crystal.  When a pendulum event occurs, loop() stops and saves in ‘n’ time quanta.  Then control passes to the interrupt function, also taking ‘n’ time quanta to do so.  The interrupt routine executes it’s logic  in another ‘n’ quanta steps, before returning and restarting loop() from where it was suspended.   Also in crystal related time.

Question is, does the infiltration matter?   It shouldn’t artificially improve the pendulum’s time-keeping because that’s designed to be more accurate than the microcontroller’s crystal, but I fear using a microcontroller reduces pendulum accuracy…

Eeek.

I should write up how the clock works properly and send it out for peer review.  Order a gallon of red-ink!

🙁

Dave

PS I shall carry on regardless because I’m getting loads of interest out of this, theory and practical.   Fortunately it’s not about me being right or wrong.

[blowlampParticipant @blowlamp]

Dave.

A ‘free pendulum’ is set going and left alone. There is no on the fly adjustment to its impulse. It gets the gentlest of nudges as infrequently as possible to keep it going and nothing else.

Just two ingredients – gravity and its effect on a swinging mass. That is what makes it the reference.

I suppose a pendulum could be seen as ‘measuring’ gravity itself.

 

Martin.

[Author]

Posts