Measuring Clocks

[SillyOldDuffer]

On SillyOldDuffer Said:

On blowlamp Said:

What are we even measuring? 😉

[…] That led me to wonder if there is a minimum time.   Is time a smooth analogue flow, or, is it a series of tiny steps?  Physics suggests many physical properties are quanta, so time might be too.   I believe science has so far failed to answer the question.

What precision to count in is part of a bigger problem.  […]

Which is why ‘they’ digitised the Second

MichaelG.

.

Ref. https://www.model-engineer.co.uk/forums/topic/measuring-clocks/#post-821851

[SillyOldDufferModerator @sillyoldduffer]

Quick report.  I got the Q code to run on command.

• Accepts command – OK
• Stops impulsing – OK
• Per beat calculator call – OK
• Reports answer when amplitude reaches target – OK
• restarts clock – OK

Correct answer – NO!  Q=0, the calculation is wrong.  More work.

During testing noticed that my pendulum was taking a suspiciously long time to decay.  Another bug. I assumed setting impulse to zero would result in a zero length pulse.  Nope, the code turns the electromagnet ON, then starts a timer to automatically switch it off when the impulse is complete.  Unfortunately that emits a short pulse during the time it takes to start the timer, about 70uS.  A 70uS pulse isn’t enough to keep the pendulum going, but it sure keeps it swinging for a long time.

Cure: when testing Q, the code mustn’t turn the electromagnet on at all.  Pesky details.

Arduino aficionados will have spotted a 70uS delay means I must be using digitalWrite().  digitalWrite is notoriously slow because it does beginner friendly error checking.  I’ll change to one of the faster alternatives that trust the me to get everything right.  What could possibly go wrong…

Dave

 

 

[SillyOldDufferModerator @sillyoldduffer]

After applying my genius to the Q calculation, I’ve made it worse.  Previously got the answer zero, which has the merit of being a number, even if hopelessly wrong.   Now getting NaN, which is computer speak for Not A Number.

More usefully, the pendulum takes 826 beats to decay from roughly 4 to 1°    Which a manual sum says Q=2368.

Bad news, big Q is better:

• good grandfather clock, Q about 12000
• precision regulator, about 20000
• Burgess-B, about 40000
• Shortt-Synchronome, about 110000
• Fedchenko, about 150000
• 32768kHz wristwatch, 100000 to 200000
• Oven Controlled Crystal Oscillator (OCXO), 1000000 – 2500000
• Lab grade ultra stable OCXO, up to 3000000

A well-made quartz wristwatch is 20x to 30x better than the best ever pendulum clocks, and they have Mickey Mouse dials!

My pendulum is miles behind the opposition.  Confirmed faulty.  A similar pendulum in the Mk1 clock was Q approx 15000.  May be because the clock is running on my dining table, which is vibrated by the pendulum, and steals energy from it.  Noticed tonight that:

• sitting still with my legs about 25cm away from the table increases period stability by 5x,
• letting the vinyl table cloth touch my leg decreases period stability by about 3x, even if I keep very still.  Slightest movement upsets it.

Surprising that a bob weighing about 100g swinging slowly on a thin spring period 0.920s can move a heavy table, yet it is so.  About 1.3mJ, roughly 0.001 foot-pounds, most of it moving the bob, not the table.  The table is, ahem, inexpensive. I guess weight about 70kg, and the spindly legs don’t help.

Can’t see anything obvious.  Will try running it on the floor, teak on concrete, to see if that helps, before ripping it apart.

🙁

Dave

[blowlampParticipant @blowlamp]

Are you still using a razor blade for the suspension?

 

Martin.

[SillyOldDuffer]

On SillyOldDuffer Said:

After applying my genius to the Q calculation, I’ve made it worse.  Previously got the answer zero, which has the merit of being a number, even if hopelessly wrong.   Now getting NaN, which is computer speak for Not A Number.

More usefully, the pendulum takes 826 beats to decay from roughly 4 to 1°    Which a manual sum says Q=2368.

Bad news, big Q is better:

• good grandfather clock, Q about 12000
• precision regulator, about 20000
• Burgess-B, about 40000
• Shortt-Synchronome, about 110000
• Fedchenko, about 150000
• 32768kHz wristwatch, 100000 to 200000
• Oven Controlled Crystal Oscillator (OCXO), 1000000 – 2500000

<p style=”text-align: left;”>Lab grade ultra stable OCXO, up to 3000000</p>

• Bateman regulator 12000
• Burgess B around 4000 (so why is it so good?)
• Synchronome pendulum 12000?
• Shortt pendulum circa 100000
• I don’t think I’ve seen a figure for Fedchenko but would guess at about 250000 as it runs in a high vacuum

The Clock B example shows that high Q is not the only requirement.  The Littlemore clock had a Q of approaching a million but wasn’t as stable as the Shortt, nobody knows why but probably ground vibration.

[blowlamp]

On blowlamp Said:

Are you still using a razor blade for the suspension?

 

Martin.

Yes, and it’s chief suspect.  Earlier I used a modified disposable razor blade successfully until I broke it.  The replacement blade came from a different brand, so the steel might be unsuitable.  Both stainless, but apart from that, properties unknown.   Possibly the replacement is only hardened on the edge.  I didn’t think to test it.

Dave

[John HaineParticipant @johnhaine32865]

Can you spare a feeler gauge finger? A razor blade’s stiffness may vary across its width which might induce twist.

[John Haine]

On John Haine Said:

On SillyOldDuffer Said:

After applying my genius to the Q calculation, I’ve made it worse.  Previously got the answer zero, which has the merit of being a number, even if hopelessly wrong.   Now getting NaN, which is computer speak for Not A Number.

More usefully, the pendulum takes 826 beats to decay from roughly 4 to 1°    Which a manual sum says Q=2368.

Bad news, big Q is better:

• good grandfather clock, Q about 12000
• precision regulator, about 20000
• Burgess-B, about 40000
• Shortt-Synchronome, about 110000
• Fedchenko, about 150000
• 32768kHz wristwatch, 100000 to 200000
• Oven Controlled Crystal Oscillator (OCXO), 1000000 – 2500000

<p style=”text-align: left;”>Lab grade ultra stable OCXO, up to 3000000</p>

• Bateman regulator 12000
• Burgess B around 4000 (so why is it so good?)
• Synchronome pendulum 12000?
• Shortt pendulum circa 100000
• I don’t think I’ve seen a figure for Fedchenko but would guess at about 250000 as it runs in a high vacuum

The Clock B example shows that high Q is not the only requirement.  The Littlemore clock had a Q of approaching a million but wasn’t as stable as the Shortt, nobody knows why but probably ground vibration.

Q figures are always a bit suspect.  Different values for the same clock.  My source says 40000 for Burgess, so maybe 4000 is a typo, or the other way round.  And the various ways of calculating it produce slightly different answers.

I’m not too worried about the exact value other than Q should be high rather low.  Putting it another way, a pendulum that takes a long time to stop is more likely to be a stable resonator than one that quickly grinds to a halt.

My 2368 result is cause for concern.

I’m worried about vibration.  My clock is obviously vulnerable.  I can well believe it’s scuppered others too!  Livermore have my sympathy.   Although my design is free-standing, might be possible to bolt it to a wall. Fair amount of work though, and I don’t have a good wall.

Dave

 

[duncan webster 1Participant @duncanwebster1]

I once had  a gents pulsynetic stood on the carpeted concrete floor in my kitchen, it wouldn’t keep going. When it got to its final destination bolted to a solid wall if ran fine.

Changing tack slightly,  had anyone tried a hermetically sealed enclosure full of helium at atmospheric pressure, much lower density than air, and possibly easier than vacuum as any leakage would only be by diffusion, no big pressure difference.

[blowlampParticipant @blowlamp]

Combining a 0.05mm feeler gauge for the suspension with a much heavier pendulum would help I’m sure. Mount it on a massive, rigid chassis and I wouldn’t be surprised if performance shot up.

 

Martin.

[Michael GilliganParticipant @michaelgilligan61133]

Recommended Reading :

https://www.tandfonline.com/doi/full/10.1080/00207179.2019.1590736#abstract

With open access

AND

This [which is chattier]

https://physicsworld.com/a/the-secret-of-the-synchronized-pendulums/

MichaelG.

[John HaineParticipant @johnhaine32865]

Dave, I sent you a PM

[John HaineParticipant @johnhaine32865]

Clock B is now installed in the NAWCC museum in Columbia Pensylvania.  During this summer permanent measuring equipment was installed and amongst other things a run down test done, without the escapement fitted.  This was mainly to investigate the circular error compensation but is also gave a Q of 4800 at the running amplitude of 6.25 degrees.  This is in line with a number of previous measurements.

The clock is an existence proof that high Q is not essential for precision timekeeping.

[SillyOldDufferModerator @sillyoldduffer]

Here’s a photo of a similar blade.  Came from a triple-bladed razor, and the holes and notches punched in them are different.

I ground the edge off with a Dremel, taking it back so the spring is symmetric about the holes, though I didn’t do a good job.   The blade is 0.1mm thick.

In the clock, the blade is fitted such that it flexes across one of the square holes, so the amount of metal bent is rather small, 2x 0.94mm!

Passes the twang test, so springy.  Bending back with pliers, the example returned to straight up to about 270° before snapping cleanly.   Hard and brittle – on the face of it, an OK spring.

There’s a thin hair caught on the bob, plus detrital dust.  Might be that.

Dave

 

 

 

[blowlamp]

On blowlamp Said:

Combining a 0.05mm feeler gauge for the suspension with a much heavier pendulum would help I’m sure. Mount it on a massive, rigid chassis and I wouldn’t be surprised if performance shot up.

 

Martin.

Yes, except the much heavier bob.  The clock is designed to swing the pendulum in a vacuum, which constrains it’s size and weight.   I’m hoping a small pendulum in a vacuum will out perform a big heavy pendulum in air.

My lathe can face the end of an 18″ long 4½” soil pipe without too much bother, bigger is problematic.  Ditto grooving the 180 x 150mm cast-iron block.  The support tower has to fit inside the pipe, leaving little space for a bigger bob.  Denser metal maybe but the extra mass may be too much for the existing electromagnet.  It’s all interrelated.

Mild-steel – 100g
Lead        –  140g
Uranium  –  238g
Tungsten –  240g
Gold        –  250g

Not rushing to change! Note John’s comments about the tenuous link between accuracy and Q!  Also, this is an experimental clock, intended to test a number of novel hypotheses, not a rush to get everything right first go.  As I’m more interested in new alternatives than golden oldies, I’m keen to milk all the value I can from the Mk2 hardware before changing it. If the new ideas work to advantage, they replace previous best practice.

So, although the suggestions are good, they’ll have to wait for the Mk3!  Keep them coming please, the design document has a todo section so I don’t forget.

Many thanks.

Dave

 

 

 

[blowlampParticipant @blowlamp]

I’m wondering how problematic it’s going to be in keeping the soil pipe from imploding once under vacuum. It’ll probably have over a thousand pounds distributed over its area and If it doesn’t retain a good circular cross section, then it might well collapse.

 

Martin.

[Michael GilliganParticipant @michaelgilligan61133]

As you are still inviting suggestions, Dave … I have to say that I think the design of your support structure is a weakness [my references to Huygens were not entirely coincidental] … If only I had the skills and the software, I would love to do  FEM on that ^^^ but meanwhile I have rely on gut feeling developed years ago, from work in a vibration test facility.

In summary … the suspension point of the pendulum should be ‘grounded’ better.

MichaelG.

.

Ref.

[blowlamp]

On blowlamp Said:

I’m wondering how problematic it’s going to be in keeping the soil pipe from imploding once under vacuum. It’ll probably have over a thousand pounds distributed over its area and If it doesn’t retain a good circular cross section, then it might well collapse.

 

Martin.

Me too!  The base is locks into a groove, and the top will be reinforced with a 5mm thick Aluminium disc.  On the plus side, I’ve tested the prototype enclosure to below 600mb, and it didn’t bend noticeably.  Pipe dia 72mm, height 300mm, wall 2.25mm thick.

The maths is beyond me.  AI suggests “The cylinder will buckle (crush) at approximately 56 kN or 144 atm external overpressure. Use a safety factor of at least 3 for vacuum applications.”

However, running the numbers for the larger pipe is bad news:  a standard 3.2mm thick soil pipe will crush at about 83% of full vacuum.  Might get away with it because my pump isn’t that good! Otherwise Plan B is to run the clock at reduced pressure, 800mB, 20% of full vacuum, maybe 40%.

A 3.5mm thick pipe (heavy duty) should survive a full vacuum, though the safety factor is low.

The standard for these pipes specifies ‘at least’ wall thicknesses.  Mine is 3.72mm thick, so has some extra margin over 3.5mm nominal.

Don’t know how low my pump will go if all leaks are sealed and it’s left running for ages.   Very difficult to pump a hard vacuum, and I expect the PVC to gas long before I get near.  Holes might pull through the plastic long before I get there.   Using PVC because it’s cheap and available, not because it’s the best material.

Yup, I’m expecting trouble!

Dave

[Michael Gilligan]

On Michael Gilligan Said:

As you are still inviting suggestions, Dave … I have to say that I think the design of your support structure is a weakness [my references to Huygens were not entirely coincidental] … If only I had the skills and the software, I would love to do  FEM on that ^^^

I wish you would!  I’ve tinkered with FEM doing stress/starin and it’s not difficult to get results.  Except I don’t understand how to set the model up, or to interpret the answers. Colour codes the structure convincingly,  but I don’t understand the numbers.  Thermal and vibration not investigated at all.

In summary … the suspension point of the pendulum should be ‘grounded’ better.

Can you suggest how that might be done?   The prototype tower had 3 pillars (see above), discussed on the forum, and I think Duncan persuaded me to use 4 pillars. (I thought 3 for kinematic reasons)

The Shortt-synchronome uses the metal vacuum tube, not a tower, it bolts to the wall.  Hard to make.

 

 

MichaelG.

 

 

 

Dave

[Michael GilliganParticipant @michaelgilligan61133]

Do you have a photo, or a sketch, of the upper part of your pendulum, Dave ?

… I would like to give this some thought.

Using four relatively slender pillars with single-point fixings doesn’t really inspire confidence.

MichaelG.

[Michael GilliganParticipant @michaelgilligan61133]

One small step: This video only discusses torsion in a single bar of round section … but the thinking Duffer will soon extrapolate to imagine [roughly] what happens when four bars are connecting two plates.

At least two other modes of distortion/vibration exist due to ‘lozenging’ under side-loads.

MichaelG.

.

[Michael GilliganParticipant @michaelgilligan61133]

and if you can spare another 20:30 … this is very informative:

MichaelG.

 

[Michael Gilligan]

On Michael Gilligan Said:

Do you have a photo, or a sketch, of the upper part of your pendulum, Dave ?

… I would like to give this some thought.

Using four relatively slender pillars with single-point fixings doesn’t really inspire confidence.

MichaelG.

Yes if you don’t mind a short wait.  I’m upgrading the Design Overview which will include drawings.  They’re in Fusion360 on a Windows partition, so might take some finding.  I’m planning to redraw the clock in Solid Edge;  might do that instead.

I’ll be back!

Dave

[duncan webster 1Participant @duncanwebster1]

If you assume the top middle and bottom rings are rigid it wouldn’t be difficult to work out the sideways stiffness of the structure by traditional methods. It would probably overestimate it, but better than nothing. Send me a dimensioned sketch

[SillyOldDufferModerator @sillyoldduffer]

Clock moved from table to floor, about 40mm from me on chair, and cover on, so shielded from ambient light and draughts:

• starts reliably
• period variation reduced by at least 5x (not actually measured it)
• period varies if I clump about – likely due to vibration through the concrete floor and teak layer.
• Q up from 2168 to 4106.  Even though I forgot to clean the bob and remove the hair.

The levelling detect upgrade made setting up easy. Plonked on the floor, right/left levelled by bubble, switched on, told clock to report the photo-interrupter ratio. Set it to 0.988 with the adjuster screw.  No need to connect my oscilloscope.  Adjuster screw not so obviously too coarse as it was on the table.

Not massively different but my floor and table top aren’t parallel.

Surprise design failure.  The clock is hard to move – needs handles.

Next, I’ll model the base and stand in SE rather than mess with Fusion, which I abandoned over two years ago.  Bound to upgrade itself, and I’ve probably forgotten how to drive it.

In other news, the elastic in my underpants has failed!  Most uncomfortable.

Dave

 

 

[Author]

Posts