Timing data for Synchronome # 2177


Bepi
 
Edited

The data I presented earlier on this blog, Chasing the Losses, about the energetics of my synchronome were obtained with a simple timer based on arduino uno whose clock is a ceramic oscillator. The quartz on the board is not the processor clock. Its precision was adequate for the task but not good enough for the period measurements. I am waiting for a new version of arduino, leonardo, which uses a quartz for communications and the processor, it should be stable enough in my relatively constant-temperature home.

The following picture shows a comparison between period measurements by the arduino (trace 1) and a timer controlled by an ordinary quartz (top trace) during a 22 hours stretch when the temperature changed by roughly 1°C (3rd trace) and the oscillation amplitude changed by a few hundreds of one degree (bottom trace).



The above shown period data have been acquired with an ordinary quartz timer designed by Luke Mester which I assembled in Venezia, in a couple of hours, on one of his circuit board. A very satisfying experience.

https://mesterhome.com/timer/index.html

It has now been upgraded with a TCXO which I tested at the shop of a local radio-amateur against a frequency counter with rubidium timebase.




 

The amplitude measurements are still arduino's. More recently the timer by Mester is also providing amplitude data in complete agreement with the arduino ones.

In future posts I'll show more measurements with reference to evidence of spurious pendulum oscillations, the effects on stability of pressure and temperature variations, circular error and the presence of the case. All comments and requests for more detail are very welcome on the blog as well as by personal email, b.-

--
Bepi


Bepi
 
Edited

One of the first things I have done with the new timer has been to optimize the pendulum suspension position with respect to the impulsing machinery.

The following graph shows that the positioning of the yoke with reference to the frame is pretty sensitive, tenths of a mm make a difference.


The most significant part is probably the period fluctuation standard deviation, AVperiodSD, the SD of the period averaged over the 15 period synchronome cycle, less noisy when the mechanism is the most efficient.

A smaller standard deviation doesn't necessarily mean more long term accuracy, a sufficient not a necessary condition, but it is an indication for the better if everything else stays the same. Unfortunately if one looks at the period fluctuations just after the arm strikes the pendulum the pattern can change completely from case to case.

Why is the max pendulum frequency at maximum efficiency? It's also a position where the pendulum rate is the most sensitive to changes.

In my case the zero position of the graph corresponds to the suspension at 17.8 mm from the frame as in the picture:


and the drawing arm position as in the following picture:


I wonder why they have designed the steel frame fork so asymmetric with respect to the optimal position of the suspension, or is it just mine? Maybe things would change some with a different drawing arm length.

--
Bepi


Eric Scace
 

Depending on how the frame has been attached to the case, and the case attached to the wall, the location of the yoke can be significantly different than apparent centerline.

There is not an easy way (nor reference marking) to get the frame inside the case to be strictly vertical — and if the frame is attached to the case a little bit off of vertical, to place the frame vertically will mean the case looks crooked on the wall.

— Eric

On 2019 Oct 20, at 15:12 , Bepi <pepicima@...> wrote:

One of the first things I have done with the new timer has been to optimize the pendulum suspension position with respect to the impulsing machinery.

The following graph shows that the positioning of the yoke with reference to the frame is pretty sensitive, tenths of a mm make a difference.
<YOKE POSITION.jpg>

The most significant part is probably the period fluctuation standard deviation, AVperiodSD, the SD of the period averaged over the 15 period synchronome cycle, less noisy when the mechanism is the most efficient.

A smaller standard deviation doesn't necessarily mean more long term accuracy, a sufficient but not necessary condition, but it is an indication for the better if everything else stays the same. Unfortunately if one looks at the period fluctuations just after the arm strikes the pendulum the pattern can change completely from case to case.

Why is the max pendulum frequency at maximum efficiency?

In my case the zero position of the graph corresponds to the suspension at 17.8 mm from the frame as in the picture:

<17.8mm.jpg>

and the drawing arm position as in the following picture:

<17.8 mm arm2.jpg>

--
Bepi


Chris
 

We take so much for granted that all Synchronomes are almost perfect and built to the same, high standard.

The impulse pallet is always taken as a standard. What about its exact length from the center of the rod?
How much effort went into making the impulse curve exactly the same as every other Synchronome?
How accurately were the pallets orientated around the rod? How was it even possible to set precisely with a clock already on the wall?
I used a small mirror, from below, but that assumes a great deal about exact perpendicularity of the wall, the case and the movement.

The Early Synchronomes were fixed into their cases, almost as an afterthought, with short, fat, wood screws.
It could probably be argued that they didn't dare bolt it into the supporting wall.
The painters might have wanted to move it without seeking advice first.
The maintenance chap might have needed to swap movements.
Or the wall might have been solid, polished granite. Chewed-up, slotted wood screws would look bad.
This was before the days of carbide tipped, masonry drills and every tradesman had an electric, hammer drill.

Chris.B

 

On 20/10/2019 22:55, Eric Scace wrote:
Depending on how the frame has been attached to the case, and the case attached to the wall, the location of the yoke can be significantly different than apparent centerline.

There is not an easy way (nor reference marking) to get the frame inside the case to be strictly vertical — and if the frame is attached to the case a little bit off of vertical, to place the frame vertically will mean the case looks crooked on the wall.

— Eric

On 2019 Oct 20, at 15:12 , Bepi <pepicima@...> wrote:

One of the first things I have done with the new timer has been to optimize the pendulum suspension position with respect to the impulsing machinery.

The following graph shows that the positioning of the yoke with reference to the frame is pretty sensitive, tenths of a mm make a difference.
<YOKE POSITION.jpg>

The most significant part is probably the period fluctuation standard deviation, AVperiodSD, the SD of the period averaged over the 15 period synchronome cycle, less noisy when the mechanism is the most efficient.

A smaller standard deviation doesn't necessarily mean more long term accuracy, a sufficient but not necessary condition, but it is an indication for the better if everything else stays the same. Unfortunately if one looks at the period fluctuations just after the arm strikes the pendulum the pattern can change completely from case to case.

Why is the max pendulum frequency at maximum efficiency?

In my case the zero position of the graph corresponds to the suspension at 17.8 mm from the frame as in the picture:

<17.8mm.jpg>

and the drawing arm position as in the following picture:

<17.8 mm arm2.jpg>

--
Bepi


John Haine
 

Hi Bepi, those are really interesting results, we've obviously been working on similar lines in finding the optimum lateral position for the pendulum.  I suspect that the minima for the two period measures are because when the "impulse" is on the exact centre of the swing, it changes only the amplitude (actually velocity) of the swing not the phase, so variations in impulse have minimum effect.  In this case the impulse is either the negative force of moving the count wheel or the actual positive impulse.

You say "Unfortunately if one looks at the period fluctuations just after the arm strikes the pendulum the pattern can change completely from case to case".  What do you mean by "case" here?  Is it oscillation to oscillation or different positions of the pendulum?  If the latter then one sees the period spiking either up or down depending on whether the impulse is lagging or leading.

I haven't measured the lateral position of my suspension but will try, but it certainly isn't central in the fork.  (By the way the frame is cast iron not steel - I know for certain because I have milled away two of the upstands!)

Of course, we are now making measurements on performance which would have been nigh on impossible when the clocks were originally being made or installed given we have the means to cheaply measure times in tenths of microseconds.  I'm sure the clocks were originally just made to run reliably with little thought to setting up the impulse position as only the gross long-running time keeping could be measured.


Ian Richardson
 

I have been following this thread, but not necessarily understanding it all, but one thing is clear to me.  It is a tribute the the design of the Synchronome that there is sufficient tolerance in all the "critical" parts to ensure reliability, long service between maintenance and sufficient accuracy for purpose.  A fine precision piece of kit requires precision setting up and accurate measurement etc. but if you bang a nail in the wall, hang a Synchronome on it and set it up as per the simple instructions - it will work!!

What I believe is called "intermediate technology"

Ian R
Macclesfield, UK
(with 6 Synchronomes hanging on nails on the wall - all running for years!)



-----Original Message-----
From: Chris <chris.b@...>
To: synchronome1 <synchronome1@groups.io>
Sent: Mon, 21 Oct 2019 7:58
Subject: Re: [synchronomeelectricclock] Timing data for Synchronome # 2177

We take so much for granted that all Synchronomes are almost perfect and built to the same, high standard.
The impulse pallet is always taken as a standard. What about its exact length from the center of the rod?
How much effort went into making the impulse curve exactly the same as every other Synchronome?
How accurately were the pallets orientated around the rod? How was it even possible to set precisely with a clock already on the wall?
I used a small mirror, from below, but that assumes a great deal about exact perpendicularity of the wall, the case and the movement.
The Early Synchronomes were fixed into their cases, almost as an afterthought, with short, fat, wood screws.
It could probably be argued that they didn't dare bolt it into the supporting wall.
The painters might have wanted to move it without seeking advice first.
The maintenance chap might have needed to swap movements.
Or the wall might have been solid, polished granite. Chewed-up, slotted wood screws would look bad.
This was before the days of carbide tipped, masonry drills and every tradesman had an electric, hammer drill.
Chris.B
 
On 20/10/2019 22:55, Eric Scace wrote:
Depending on how the frame has been attached to the case, and the case attached to the wall, the location of the yoke can be significantly different than apparent centerline.

There is not an easy way (nor reference marking) to get the frame inside the case to be strictly vertical — and if the frame is attached to the case a little bit off of vertical, to place the frame vertically will mean the case looks crooked on the wall.

— Eric

On 2019 Oct 20, at 15:12 , Bepi <pepicima@...> wrote:

One of the first things I have done with the new timer has been to optimize the pendulum suspension position with respect to the impulsing machinery.
The following graph shows that the positioning of the yoke with reference to the frame is pretty sensitive, tenths of a mm make a difference.
<YOKE POSITION.jpg>
The most significant part is probably the period fluctuation standard deviation, AVperiodSD, the SD of the period averaged over the 15 period synchronome cycle, less noisy when the mechanism is the most efficient.
A smaller standard deviation doesn't necessarily mean more long term accuracy, a sufficient but not necessary condition, but it is an indication for the better if everything else stays the same. Unfortunately if one looks at the period fluctuations just after the arm strikes the pendulum the pattern can change completely from case to case.
Why is the max pendulum frequency at maximum efficiency?
In my case the zero position of the graph corresponds to the suspension at 17.8 mm from the frame as in the picture:
<17.8mm.jpg>
and the drawing arm position as in the following picture:

<17.8 mm arm2.jpg>
--
Bepi


Bepi
 
Edited

Thanks John, interesting remarks.

I checked how vertical my frame is, better than 1/2 deg, too small and with the wrong sign to account for the asymmetry, it's either design or another reason.

TO answer your question by "case" i mean different position but there are a couple of repeating patterns which make me think of undersampling interference.

About the period minimum as a function of the suspension position are you referring to Airy's tangent rule? A tangent is an odd function though I would have expected the "error" to change sign on either side of the "center", if 0 in my previous post graph is the pendulum trajectory center, like it appears to be from energy considerations.

I keep running in situations which one hopes would have simple classic explanations, but they don't.

Here is another example, the cabinet effect in my arrangement, the glass door gets opened and closed at the two obvious discontinuities in all traces:



The usual data taken in two different days against time in min and hours:  top is the period, in both cases a roughly 80 μs drop, middle frame is temperature measured inside the case, more or less in the middle, on the right side, roughly 1/3 of a deg C, and oscillation amplitude in the bottom graph, still rising after an hour but leveling. All data are averaged over the synchronome 15 period cycle.

I was surprised by the size of the energy change and one would have thought of a corresponding circular error effect in the period but the sign is wrong, the size of the effect is wrong and the time behavior is wrong. The energy is a nice exponential as expected but the period reflects the squarish timing of the door opening and closing. The small temperature change is inadequate to explain the period change as a contraction of the invar rod. If I remember well I tested the period temperature sensitivity to be less than 6 μs/C. Pressure is obviously not changing in this short time.

One can, barely, see in the data a period shift proportional to the amplitude change, wrong sign and amplitude again if interpreted as circular error.

Totally agree with you about the unprecedented ability we have nowadays to check things with fantastic accuracy, and the exceptional help we get from inexpensive, simple to use, data acquisition. All of this was unavailable just a few years ago, mostly the result of cheap and very dense digital electronics.
Ian what is the equivalent of a synchronome in todays world? I understand you are interested in technology history, can you tell me why I, as an italian, would identify the synchronome so much more of a british piece of equipment rather than a german or french one? citing John Cleese I would also add "but please don't mention the war".

--
Bepi


Bepi
 

John I read your last post to Reinventing the Synchronome and now I understand what you are saying, very interesting. I retract what I was saying above about the tangent rule, quite the opposite what I am observing, a standard deviation minimum of the period averaged over the synchronome impulsing cycle at the center of the pendulum oscillation, would be consistent with what you observe. No surprise if the rate irregularities get minimized together with the size of the disturbance. Looking forward to repeat your measurements, there still is the possibility that the impulse disturbance is more complicated in my case I see pretty large oscillations after the strike which dissipate slowly, so maybe the centrality of the impulsing has to be intended in an average sense over the oscillations train. Do I remember well that your rod is a carbon fiber tube? It might be a lot more rigid than my invar rod with consequences in this respect.


Bepi


Chris
 

The standard rod on a Synchronome is very prone to visible vibration at impulse.
Presumably this reduces with improved set-up but may still be lurking just below visibility.

Adding mass to the center of the rod might push the frequency down and out of the damaging bandwidth.
Though it would also require the bob be lowered to compensate.

The rod could be stiffened without increased mass by have a central cross spare restrained by tensioned wires.
Which would be triangulated to the top and bottom of the rod. Somewhat like yacht rigging.

Chris.B

On 22/10/2019 09:55, Bepi wrote:

John I read your last post to Reinventing the Synchronome and now I understand what you are saying, very interesting. I retract what I was saying above about the tangent rule, quite the opposite what I am observing, a standard deviation minimum of the period averaged over the synchronome impulsing cycle at the center of the pendulum oscillation, would be consistent with what you observe. No surprise if the rate irregularities get minimized together with the size of the disturbance. Looking forward to repeat your measurements, there still is the possibility that the impulse disturbance is more complicated in my case I see pretty large oscillations after the strike which dissipate slowly, so maybe the centrality of the impulsing has to be intended in an average sense over the oscillations train. Do I remember well that your rod is a carbon fiber tube? It might be a lot more rigid than my invar rod with consequences in this respect.


Bepi


Darren Conway
 

Hi

A better option would be to replace the rod with a flat strip, but that would increase the risk of torsional oscillation. 

I am going to stick with the rod on my clock.


Regards

Darren Conway


On 22.10.19 9:20 pm, Chris wrote:

The standard rod on a Synchronome is very prone to visible vibration at impulse.
Presumably this reduces with improved set-up but may still be lurking just below visibility.

Adding mass to the center of the rod might push the frequency down and out of the damaging bandwidth.
Though it would also require the bob be lowered to compensate.

The rod could be stiffened without increased mass by have a central cross spare restrained by tensioned wires.
Which would be triangulated to the top and bottom of the rod. Somewhat like yacht rigging.

Chris.B

On 22/10/2019 09:55, Bepi wrote:

John I read your last post to Reinventing the Synchronome and now I understand what you are saying, very interesting. I retract what I was saying above about the tangent rule, quite the opposite what I am observing, a standard deviation minimum of the period averaged over the synchronome impulsing cycle at the center of the pendulum oscillation, would be consistent with what you observe. No surprise if the rate irregularities get minimized together with the size of the disturbance. Looking forward to repeat your measurements, there still is the possibility that the impulse disturbance is more complicated in my case I see pretty large oscillations after the strike which dissipate slowly, so maybe the centrality of the impulsing has to be intended in an average sense over the oscillations train. Do I remember well that your rod is a carbon fiber tube? It might be a lot more rigid than my invar rod with consequences in this respect.


Bepi

Virus-free. www.avast.com


John Haine
 

Yes, my rod is actually CF tube.  I suspect that it is no more rigid than invar.  Actually I suspect that the vibration is not the rod bending, but rotating around the CG of the bob against the suspension spring - as Tom pointed out a while back it would be better to impulse through the centre of percussion of the pendulum to avoid this.  I do see a bit of this movement especially when starting up but I don't think it's a big problem, the vibration frequency is well above 0.5 Hz.

I agree that the 'Nome is a very robust and reliable design, a great bit of "intermediate tech".  That's not to say that it can't be improved with some modern techniques though.


Harvey Moseley
 

What are the dynamics when the gravity arm falls?  How should it really work?  1)  how far above the pallet is the roller when it drops?  2) does it drop onto a flat spot so the initial fall does not generate an  amplitude impulse?  3) the amplitude impulse occurs when it rolls into the curved section of the pallet?  Where should that be in the pendulum cycle?
I am thinking that we would want to avoid any impulsive interaction between the initial fall of the gravity arm and the pendulum amplitude, making the clearance between the gravity arm and the top of the pallet as small as possible?  Then all the push comes fairly adiabatically as the roller goes down the ramp.  Is this the right way to think of this?   Also, the shape of the pallet on which the roller works is a possible variable. I never understood the circular-looking profile.  Never seemed right to me.  We should decide what force profile we want and then machine that into the pallet.
As I have mentioned, I have a synchronome I built back in 1969-1970 when I was a 20 year old and played with it quite a lot then.  You  guys are spurring me into setting it up again.  I never had a clock dial with it - I did have a photocell to count cycles though.
Best,
Harvey


On Oct 22, 2019, at 6:35 AM, John Haine <john.haine@...> wrote:

Yes, my rod is actually CF tube.  I suspect that it is no more rigid than invar.  Actually I suspect that the vibration is not the rod bending, but rotating around the CG of the bob against the suspension spring - as Tom pointed out a while back it would be better to impulse through the centre of percussion of the pendulum to avoid this.  I do see a bit of this movement especially when starting up but I don't think it's a big problem, the vibration frequency is well above 0.5 Hz.

I agree that the 'Nome is a very robust and reliable design, a great bit of "intermediate tech".  That's not to say that it can't be improved with some modern techniques though.


Neville Michie
 

Vibration of the pendulum rod should not be a problem if it has completely
died out by the time the pendulum is next impulsed.
A pendulum has one resonant frequency, the rate at which it beats.
You could apply energy at any other frequency, and unless there was a
non-linear process that could modulate the main resonance, the frequency of
the pendulum will not be affected.
However, interference at the same frequency, like another clock mounted on
the same wall, can add or remove energy from the pendulum.
The energy absorbed from a short impulse has many frequency components,
but only the frequency of the beat can be taken up by the pendulum.
If the pendulum is still shaking when next impulsed then the phase
of the impulse could be modified, this being a non-linear process there
could be interference with the pendulums frequency.
cheers,
Neville Michie

On 22 Oct 2019, at 21:35, John Haine <john.haine@ieee.org> wrote:

Yes, my rod is actually CF tube. I suspect that it is no more rigid than invar. Actually I suspect that the vibration is not the rod bending, but rotating around the CG of the bob against the suspension spring - as Tom pointed out a while back it would be better to impulse through the centre of percussion of the pendulum to avoid this. I do see a bit of this movement especially when starting up but I don't think it's a big problem, the vibration frequency is well above 0.5 Hz.

I agree that the 'Nome is a very robust and reliable design, a great bit of "intermediate tech". That's not to say that it can't be improved with some modern techniques though.


Chris
 

A vibrating pendulum rod is also oscillating the bob vertically.
Any bend shortens the rod. Which is the time standard now?
The oscillating shortened rod? Somewhere in between? Or the relaxed one?
Imagine "bending" a string on a guitar....

Chris.B

On 22/10/2019 23:56, Neville Michie via Groups.Io wrote:
Vibration of the pendulum rod should not be a problem if it has completely
died out by the time the pendulum is next impulsed.
A pendulum has one resonant frequency, the rate at which it beats.
You could apply energy at any other frequency, and unless there was a
non-linear process that could modulate the main resonance, the frequency of
the pendulum will not be affected.
However, interference at the same frequency, like another clock mounted on
the same wall, can add or remove energy from the pendulum.
The energy absorbed from a short impulse has many frequency components,
but only the frequency of the beat can be taken up by the pendulum.
If the pendulum is still shaking when next impulsed then the phase
of the impulse could be modified, this being a non-linear process there
could be interference with the pendulums frequency.
cheers,
Neville Michie

On 22 Oct 2019, at 21:35, John Haine <john.haine@ieee.org> wrote:

Yes, my rod is actually CF tube. I suspect that it is no more rigid than invar. Actually I suspect that the vibration is not the rod bending, but rotating around the CG of the bob against the suspension spring - as Tom pointed out a while back it would be better to impulse through the centre of percussion of the pendulum to avoid this. I do see a bit of this movement especially when starting up but I don't think it's a big problem, the vibration frequency is well above 0.5 Hz.

I agree that the 'Nome is a very robust and reliable design, a great bit of "intermediate tech". That's not to say that it can't be improved with some modern techniques though.


Harvey Moseley
 

Yes, these are second order effects, but when one is making precision measurement, they need to be considered and their effects quantified.
Harvey


On Oct 23, 2019, at 1:12 AM, Chris <chris.b@...> wrote:

A vibrating pendulum rod is also oscillating the bob vertically.
Any bend shortens the rod. Which is the time standard now?
The oscillating shortened rod? Somewhere in between? Or the relaxed one?
Imagine "bending" a string on a guitar....

Chris.B


On 22/10/2019 23:56, Neville Michie via Groups.Io wrote:
Vibration of the pendulum rod should not be a problem if it has completely
died out by the time the pendulum is next impulsed.
A pendulum has one resonant frequency, the rate at which it beats.
You could apply energy at any other frequency, and unless there was a
non-linear process that could modulate the main resonance, the frequency of
the pendulum will not be affected.
However, interference at the same frequency, like another clock mounted on
the same wall, can add or remove energy from the pendulum.
The energy absorbed from a short impulse has many frequency components,
but only the frequency of the beat can be taken up by the pendulum.
If the pendulum is still shaking when next impulsed then the phase
of the impulse could be modified, this being a non-linear process there
could be interference with the pendulums frequency.
cheers,
Neville Michie

On 22 Oct 2019, at 21:35, John Haine <john.haine@...> wrote:

Yes, my rod is actually CF tube.  I suspect that it is no more rigid than invar.  Actually I suspect that the vibration is not the rod bending, but rotating around the CG of the bob against the suspension spring - as Tom pointed out a while back it would be better to impulse through the centre of percussion of the pendulum to avoid this.  I do see a bit of this movement especially when starting up but I don't think it's a big problem, the vibration frequency is well above 0.5 Hz.

I agree that the 'Nome is a very robust and reliable design, a great bit of "intermediate tech".  That's not to say that it can't be improved with some modern techniques though.







John Haine
 
Edited

Good questions Harvey.
  • There was an article by F H-J mentioning a "theoretically optimum" pallet profile derived by Shortt.  This has a profile that is supposed to provide a raised-cosine force waveform.  Before and after the "ramp" the pallet surface should be radial to the pendulum suspension (a "dead roll") so there is no net sideways force on the pendulum except from the ramp.  The raised cosine is not necessarily optimum, it just happens to be smooth and symmetrical.
  • F H-J says, and everyone agrees I think, that the phase centre of the impulse should be at the centre of swing to minimise escapement error, and as Bepi's measurements seem to show also small timing variations due to fluctuations in the impulse.
  • In the Synchronome book it says the clearance between the roller and the dead roll surface should be 0.25 mm.  Seems to me that smaller is better, less shock when the roller lands.
  • I have machined such a pallet for my clock - you can imagine milling the desired raised sine profile with a cutter the same diameter as the roller, so the curve it cuts will be the right shape.  When you do that it actually looks rather like the circular profile!
  • I have verified experimentally that the dead roll hypothesis does actually work, but I'm beginning to doubt other aspects of this theory!
Just to add that in my "reinventing" album under photos there's a couple of pictures of my pallet.


Ian Richardson
 

It is worth noting that Gent used an inclined flat, straight impulse face for their pallet, right from the start.  It would be interesting to compare the performance of that and the more elaborate F H-J design (which I thought was Bowell's design anyway).

Ian R
Macclesfield, UK



-----Original Message-----
From: John Haine <john.haine@...>
To: synchronome1 <synchronome1@groups.io>
Sent: Thu, 24 Oct 2019 12:24
Subject: Re: [synchronomeelectricclock] Timing data for Synchronome # 2177

Good questions Harvey.
  • There was an article by F H-J mentioning a "theoretically optimum" pallet profile derived by Shortt.  This has a profile that is supposed to provide a raised-cosine force waveform.  Before and after the "ramp" the pallet surface should be radial to the pendulum suspension (a "dead roll") so there is no net sideways force on the pendulum except from the ramp.  The raised cosine is not necessarily optimum, it just happens to be smooth and symmetrical.
  • F H-J says, and everyone agrees I think, that the phase centre of the impulse should be at the centre of swing to minimise escapement error, and as Bepi's measurements seem to show also small timing variations due to fluctuations in the impulse.
  • In the Synchronome book it says the clearance between the roller and the dead roll surface should be 0.25 mm.  Seems to me that smaller is better, less shock when the roller lands.
  • I have machined such a pallet for my clock - you can imagine milling the desired raised sine profile with a cutter the same diameter as the roller, so the curve it cuts will be the right shape.  When you do that it actually looks rather like the circular profile!
  • I have verified experimentally that the dead roll hypothesis does actually work, but I'm beginning to doubt other aspects of this theory!


Harvey Moseley
 

Hi John,
Thanks for the excellent comments. Does the roller fall off the pallet before the electrical contact is made to lift the gravity arm, or while the roller is on the pallet?  It seems that this is an area of possible trouble as well.
Thanks,
H


On Oct 24, 2019, at 6:24 AM, John Haine <john.haine@...> wrote:

[Edited Message Follows]

Good questions Harvey.
  • There was an article by F H-J mentioning a "theoretically optimum" pallet profile derived by Shortt.  This has a profile that is supposed to provide a raised-cosine force waveform.  Before and after the "ramp" the pallet surface should be radial to the pendulum suspension (a "dead roll") so there is no net sideways force on the pendulum except from the ramp.  The raised cosine is not necessarily optimum, it just happens to be smooth and symmetrical.
  • F H-J says, and everyone agrees I think, that the phase centre of the impulse should be at the centre of swing to minimise escapement error, and as Bepi's measurements seem to show also small timing variations due to fluctuations in the impulse.
  • In the Synchronome book it says the clearance between the roller and the dead roll surface should be 0.25 mm.  Seems to me that smaller is better, less shock when the roller lands.
  • I have machined such a pallet for my clock - you can imagine milling the desired raised sine profile with a cutter the same diameter as the roller, so the curve it cuts will be the right shape.  When you do that it actually looks rather like the circular profile!
  • I have verified experimentally that the dead roll hypothesis does actually work, but I'm beginning to doubt other aspects of this theory!
Just to add that in my "reinventing" album under photos there's a couple of pictures of my pallet.


Eric Scace
 

   The roller is intended to roll off the pallet. That brings the full weight of the gravity arm into the contact closure, in order to have a more reliable contact closure.

On 2019 Oct 24, at 09:25 , Harvey Moseley <harvey.moseley1@...> wrote:

Hi John,
Thanks for the excellent comments. Does the roller fall off the pallet before the electrical contact is made to lift the gravity arm, or while the roller is on the pallet?  It seems that this is an area of possible trouble as well.
Thanks,
H


Harvey Moseley
 

It would seem to me that as it rolls off the edge of the pallet, it would provide a lateral impulse that would be undesirable.
Thanks,
Harvey


On Oct 24, 2019, at 10:06 AM, Eric Scace <eric@...> wrote:

   The roller is intended to roll off the pallet. That brings the full weight of the gravity arm into the contact closure, in order to have a more reliable contact closure.

On 2019 Oct 24, at 09:25 , Harvey Moseley <harvey.moseley1@...> wrote:

Hi John,
Thanks for the excellent comments. Does the roller fall off the pallet before the electrical contact is made to lift the gravity arm, or while the roller is on the pallet?  It seems that this is an area of possible trouble as well.
Thanks,
H