Topics

Effect of open case vs. closed case


Tom Van Baak
 

Bepi,

A few years ago I measured a Synchronome; case closed vs. case open. The change is quite immediate, dramatic, and easily measurable.

You mentioned Q. Yes, the state of the glass case door will affect Q. My goal back then was to actually compare how stable the rate was, that is, how good the timekeeping was, in both cases. For example, even though the calculated Q is higher with case open, it is possible that a clock actually keeps better time with a case closed. In other words calculated Q and actual performance are not absolutely linked.

Now, about the effect of the glass door. Rather than find that old data, I just re-ran the experiment for you. The three plots below cover the past 14 hours, about 3 PM to 5 AM.

1)

Clock rate is shown below. On the left is 5 hours of normal running, that is, closed case. On the right is 5 hours of open case. In the middle I opened and closed the door a few times, an hour apart.

The results are pretty clear. For this experiment, rate increases about 15 ppm when the case is open. That's 1.3 sec/day. I hope this roughly agrees with results others have found. If not, let me know.

Closer inspection of the plot shows that it takes a while (about an hour) for clock rate to fully stabilize when the case is opened. This is not unexpected. An open case is less resistance for the bob, and since the impulse is fixed, the energy (amplitude) will slowly increase, impulse by impulse, until a new equilibrium is reached. It appears to take less time for rate to stabilize when the case is once again closed.

2)

Speaking of amplitude. Whether using Microset of picPET I always monitor amplitude. This is derived from the measured width of the photogate pulse. The larger the amplitude, the more energy in the swing, the faster the bob travels at BDC, the narrower the pulse as the thin flag passes through the photogate sensor. So here is a plot of amplitude for the same 14 hour run.


Note: rather the measure amplitude in degrees or radians the y-axis scale is percent where 100% is the normal amplitude with case closed. Comparing the left 5 hours with the right 5 hours we see amplitude increases by 6% when the clock runs with case open.

Also notice that in both cases, going from case closed to case open, or case open to case closed, there is a nice exponential ramp between the lower and the higher amplitude. The time constant conveys how aggressively the clock reaches a new equilibrium.

Quick guess. If the clock were running normally (case closed) at 2.7 degrees semi-arc (+/- 5.0 mm displacement), a 6% change means the new amplitude is 2.86 degrees. Circular error changes from 143 ppm to 161 ppm for a net change of 17 ppm. So some of the rate change is due to differential circular error.

(3)

Finally, for completeness, a temperature plot for these 14 hours. Normally my eye-level wall-mounted Synchronome is fully enclosed with an oversized, 6 foot tall, 12 cu ft, sound isolated and thermally insulated, temperature controlled vertical "coffin". In order to perform this door closed/open experiment I had to remove the coffin and set it aside. This occurred at 8 PM, 5 hours into the run. This exposes the clock to normal room temperature.


The left part of the plot shows how well the enclosure works. The temperature of the clock is always about 79.2 F and often stays within 0.1 F.

The right side of the plot shows the clock running at room temperature, about 74 F. Also, it shows the effects of HVAC cycling, wiggles that are highly attenuated on the left. The middle hours of the plot show the Synchronome slowly shedding 5 degrees of heat. It takes about 3 or 4 hours.

I hope these three plots help answer your question and stimulate some thought. Let me know if you have any questions.

Thanks,
/tvb

Plots also available at: http://leapsecond.com/pend/synchronome/case.htm



Bepi
 

Tim, interesting what you show. Comparing it with what I measured roughly a month ago and was posted on this blog oct 21st, I see similarities and differences. The measuring technique is probably the same, in my case I am showing data averaged over the 30 s synchronome impulsing period.

I similarly notice a fast and a slow effect on the rate.

The fast one in my case has the same sign as yours but is a 40 ppm variation instead of a 25 ppm. Your case might be larger than mine to start with. Mine is an original Synchronome case, 225 x 110 x 1210 internal dimensions, the bob has only an 8 mm gap towards the inside wall. It would be nice to complement these measurements with rate noise spectra. I'll do it sooner or later, some independent evidence there.

The fast rise time of this effect, faster than the amplitude variation and the rod heating/cooling time constant, makes me willing to go back to the raw data to look at what happens on a single beat time scale.

The slower and smaller effect, the one which correlates with the amplitude trace, makes one think immediately to the circular error. The sign is wrong in my case though, although it's right in yours, and its amplitude seems to be wrong too. The correlation with the angle/energy is strong and it's incompatible with the heating time constant of the rod.

Still a lot to understand.

PS Your displacements are probably in cm not mm. It reminds me of when once, while visiting a UK university, they announced the switch from imperial to metric in the rulers distributed by the internal stationary store. I went immediately to get a metric ruler and the guy at the counter said: shure we just got them, do you want a 6 or a 12 inches long one?

--
Bepi


Tom Van Baak
 

Bepi,

My case is the same size as yours, 8.5" x 4" x 48", which agrees with your metric numbers. I think our signs agree; note that I plot rate (frequency) and you plot period so your down is my up.

The slow (on the order of an hour) rise and fall in amplitude is likely just the clock hunting for its new equilibrium based on the new "environment" that the door open / close creates. With a pendulum clock amplitude changes tend to be very gradual because a clock can only gain amplitude in tiny quanta every 30 seconds, or only lose amplitude only as slow as its Q allows.

The immediate change in rate is more puzzling, and its much greater in magnitude than amplitude / circular error alone. I suspect it's buoyancy effect(s). The topic is covered in books and horological papers. It's not quite as simple as air pressure; there are effective weight or gravity considerations as well as air mass effects.

I would be happy if the textbook calculations specifically addressed the open / close door case. But so far the calculations that I've seen are more abstract.

Some online reading for you:

"The Pendulum", Alan Emmerson
http://users.qldnet.com.au/~ajay/From%20an%20Engineer's%20Notebook/The%20Pendulum.pdf

"Exploring Buoyancy (HSN2007-2)", Robert Belleville
http://thirdandlark.com/posts/HSN_2007_2_buoyancy/hsn_buoyancy.pdf

"The Trinity Clock, 4.6.2 Buoyancy"
http://trin-hosts.trin.cam.ac.uk/clock/theory/pendulum.pdf

"Studies on astronomical time-keepers and time-preserving systems"
http://adsabs.harvard.edu/full/1921AOTok...5d...1S

Also, this topic is likely covered by Rawlings, "The science of clocks and watches" or by Matthys, "Accurate Clock Pendulums".

/tvb
(N.B. it's Tom, not Tim).


On 12/5/2019 1:24 PM, Bepi wrote:

Tim, interesting what you show. Comparing it with what I measured roughly a month ago and was posted on this blog oct 21st, I see similarities and differences. The measuring technique is probably the same, in my case I am showing data averaged over the 30 s synchronome impulsing period.

I similarly notice a fast and a slow effect on the rate.

The fast one in my case has the same sign as yours but is a 40 ppm variation instead of a 25 ppm. Your case might be larger than mine to start with. Mine is an original Synchronome case, 225 x 110 x 1210 internal dimensions, the bob has only an 8 mm gap towards the inside wall. It would be nice to complement these measurements with rate noise spectra. I'll do it sooner or later, some independent evidence there.

The fast rise time of this effect, faster than the amplitude variation and the rod heating/cooling time constant, makes me willing to go back to the raw data to look at what happens on a single beat time scale.

The slower and smaller effect, the one which correlates with the amplitude trace, makes one think immediately to the circular error. The sign is wrong in my case though, although it's right in yours, and its amplitude seems to be wrong too. The correlation with the angle/energy is strong and it's incompatible with the heating time constant of the rod.

Still a lot to understand.

PS Your displacements are probably in cm not mm. It reminds me of when once, while visiting a UK university, they announced the switch from imperial to metric in the rulers distributed by the internal stationary store. I went immediately to get a metric ruler and the guy at the counter said: shure we just got them, do you want a 6 or a 12 inches long one?

--
Bepi


Bepi
 
Edited

The sign difference between our clocks is in the slow rate variation, not in the fast. I recorded it again today with a modified clock, the escapement latch is now triggered by a tiny solenoid controlled by a delaied zero crossing photo detector. The larger Q pendulum is now impulsed once a minute, instead of every 30 s, and the standard deviation of the 60 s cycle rate is now down to < 5 μs. Time scale in hours.





I don't understand both, the fast and the slow, rate variations. The slow one tracks the energy/amplitude but has the wrong sign, in my case, for cyrcular error. The fast can't be buoyancy since the air pressure/density can't, and doesn't, change in my case. It could be a dynamic resistance asymmetry effect and we should be able to change its sign (and affect its amplitude) by changes in the bob oscilaltion induced air flow. I'll try to modify it introducing obstacles.
It's still all unexplained and all the more fun.
PS Do you by chance have a copy of Woodward's collected horology writings, apparently there is book other than my own right time which is not in print any more.
--
Bepi


Tom Van Baak
 

PS Do you by chance have a copy of Woodward's collected horology
writings,
apparently there is book other than my own right time which is not in
print any more.

My Own Right Time (MORT) was his first book; Woodward On Time (WOT) was the later collection you speak of. Both are out of print, I think. Doug knew him well. I've met both Doug and Philip on several occasions a decade ago.

Most of what Philip wrote is in AHS (Antiquarian Horological Society) or HJ (Horological Journal) or HSN (Horological Science Newsletter). I'm told you can get back-issues with a subscription. For me, HSN is the best. I highly recommend that you subscribe, and also buy the CD of back issues. And maybe consider posting some of your recent results there. It's more of an informal publication, not a fancy glossy highly edited magazine.

/tvb


Bepi
 
Edited

Tom, repeating the open and closed door test adding a flat, 20 mm thick, piece of styrofoam to the glass door as shown in the picture ( the back wall is still closer to the bob than the styrofoam )



yielded the following result:



where the time scale is in hours and the period data are averaged over 30 periods, the periodicity of gravity arm impulses.
The amplitude still behaves in the intuitive fashion: nearer wall => more friction => smaller amplitude.

The rate wall effect seems to be linked to the distance both for the fast and the slow time scale. About the slow time scale we know that the phase can be affected by the amplitude through non-linearities, this time is not the circular error one but something else.

To quantify an effect from theory which so strongly depends on the turbulent airflow might be impossible but we can certainly see it empirically.

The regularity of the rate, what interests us most, is also affected. Judging from standard deviation measurements, and visually from the traces themselves, the "closed" condition seems to be the best, maybe, in my case, a sort of compromise between the more turbulent "closed + styrofoam" case and the "open" case where a butterfly in the room would certainly show up in the period trace.

 

Form this experience I would deduce that an enclosure is very important for regularity, but equally important is that its small dimensions are large enough to minimize air friction, certainly larger than in a standard Synchronome enclosure.

Paying attention to the aerodynamics of the bob would be probably just as useful. It would be fun to test an aerodynamic lead bob of same weight of the traditional larger one.

 

These results have been obtained with a new gravity arm trigger setup (probably very similar in its effects to John Haine's stepper motor trigger) and 30 skipped oscillations between impulses, a much less noisy setup than the traditional 15.  The inherently higher Q of this solution allows for the same amplitude as with the traditional mechanical trigger. It would be super easy to investigate the dependence of rate regularity from amplitude by changing the skipped periods between impulses, not quite the same thing as changing the strength of the impulse but a lot easier. John, did you try it already?

 

PS Thanks Tom, Woodward on Time is the book I am looking for. I'll be glad to purchase it in any kind of condition and form. I recently purchased all the back issues of HSN but never had the patience to look at them, since you recommend them I'll look at them more seriously now.

--
Bepi


John Haine
 

I've yet to experiment with the number of oscillations between impulses.  I started with 40 and that gives me about 1.8 degrees amplitude and as yet I haven't changed it as it seems to work.

There are quite a few used copies of WOT on Amazon right now.


John Haine
 

Oops!  Sorry, mistaken, that's MORT, no copies of WOT.  Would be good if the BHI could either reprint of make available as an e-book.


Bepi
 
Edited

About the door opening effects on the period I subdivided the measurements according to their separation from the impulse period to find out if all cycle phases were equally responsible for the changes. The answer is definitely not.

To remind you of what I am talking about on the top trace is the "open door" overall effect, an increase of the average period of the clock by 160 μs with the case door open, a fast and slow change, the slow change mimics the amplitude change, the fast change is instantaneous. Neither is consistent with a circular error effect. The plot time scale is in hours and what's shown is the averages over the clock impulsing period, in this case 30 periods, an average over 60 seconds.


The third and fourth traces show single period measurements, every dot is a period, 2 second intervals, at two different times during the same experimental run.

The third trace corresponds to door open, t = 3.06 h, the fourth to door closed, t = 12.58 h.

As reported here many times it's evident that at the time of impulse the period is more heavily perturbed, followed by some spurious oscillations and a quieter phase. This is true with different intensity at door open and closed.

if one now looks at the period variations due to the door opening at different delays with respect to the impulse discovers that the slow variation is all linked to the quite large variation of the period at impulse which contributes disproportionally to the average.

Here on the ms scale 



while the free running period variation is all of the instantaneous kind.

Here on a 300μs scale, the same as the above averages




--
Bepi


Bepi
 

It has been shown above that the biggest effect on the clock rate is not related to an energy change, which is also present, but is an instantaneous phase change, an instantaneous change in the forces structure, a "tangent rule" error. It's a free pendulum effect, it can develop completely when the clock is in its free pendulum phase between impulses, I find hard to call it "an escapement error".

Friction losses are often absent from theoretical analysis leading to the "tangent rule" because it's assumed that they have a phase even-function dependence on geometric symmetry grounds, i suspect it's not true in the synchronome. 

To estimate the asymmetry responsible for the "fast" close-open-close door effect I performed the following measurements.



In the picture one can see the effect on the period of opening the case door twice, for 10 minutes each time, for two power pulse delays. No change whatsoever in the power pulse for door open event N.1, between blue lines, and delaying by 80 ms the 67 ms power impulse, in door open event N.2, red lines.

80 ms correspond to a roughly 7 mm bob shift in physical space.

Same current, same pulse length, in principle no difference in the energy delivered by the escapement, just a modest difference of when.

 

I would deduce the need to investigate the clock case asymmetries, a few millimeters can matter, they could induce dissipation asymmetries in turn responsible to significant rate noise.

--
Bepi