Measuring Clocks

[duncan webster 1]

On duncan webster 1 Said:

Just to make it even more complicated you could just have 2 pin connector for power and do everything else by wireless / Bluetooth. Might put delays into it, beyond my pay grade

Duncan identifies the problem – those pesky delays.   The wires are: 5V, Gnd, I2C (to the temp/pressure/humidity sensor), Magnet On/OFF and the photo-interrupter output.  The last two are time critical, so wifi no good.  Optical is a possibility.

Robert’s MIL-STD connector is tempting but how big are they?  Not much space inside the clock – an ⌀8mm hole. Good for the wires, but I also have to pass a ⌀4mm pipe out to the vacuum pump.   Though I’d like to avoid another hole in the base, it’s a maybe.

Needs more thought.

Not much done this evening – family problem intruded.

Dave

 

 

[John HaineParticipant @johnhaine32865]

Regarding stiffness, Duncan calculated:

I make the overall stiffness to be ~600N/mm. For comparison, a 75 OD 1mm wall tube would be ~5300 N/mm. I don’t think I can readily post a SMath document here, if anyone wants a copy send your email via pm.

Assumption 1 is reasonable, the others less so. I’d expect the real stiffness to be lower

The actual side load is (mass of bob) * g * tan(half angle of swing), so pretty low

I think the last sentence must be incorrect.  The side load on the pendulum support must be equal to and in the opposite direction to the restoring force, which is just m.g.sin(theta) where theta is the angle from vertical, or for small deflections m.g.theta.  If it was tan then for an angle of 90 degrees the force would be infinite but in fact it will be zero.

Before we get too hung up on the support deflection (pun intended) it’s worth considering what affect it has.  The sideways deflection of the support very slightly reduces the effective deflection of the pendulum.  There are two springs in series, the “gravity spring” which has spring constant mg/L where m is bob mass and L the length.  If you do the sums the effective spring constant of the two in series is:

Keff = Kpend/(1+[Kpend/Ksupport])

So one effect of the support compliance is to very slightly increase the effective length of the pendulum.  When I did the calculations for a much heavier bob and probably stiffer support [1] the change in clock rate produced would have been of the order of one ppm.  I’m not sure of the weight and length of Dave’s pendulum but worth doing the calculations, I suspect the answer will be around the same.

1. On the amplitude of the Synchronome oscillator, HSN 2017-5

This assumes that the support is essentially reactive, not absorbing energy.  In reality there may be energy loss as well reducing the Q.

This doesn’t tell us much about effects going the other way.  If one had a totally rigid support fixed to a huge mass, if that mass moves it will affect the pendulum.  Even if the mass was the earth itself, there are earthquakes.  Oxford Prof. E T Hall on his retirement dedicated himself to building the “Littlemore Clock” (he lived in this village which is really a suburb of Oxford).  The pendulum swung in a vacuum, had a Q approaching a million, and was mounted on a 6 ton concrete pillar cast into a large hole in his garden.  Despite his best efforts the clock was actually worse than the Shortt at detecting earth tides, and it is believed this was because of seismic noise.  Amongst other things there was a large tree quite close by and a railway line carrying freight to and from Cowley motor works through the village.  (Interestingly the line is being reopened for passenger traffic and a new station being built at Littlemore.)

[John HaineParticipant @johnhaine32865]

As for feed-thrus etc.  Provided the wall thickness isn’t too large one can feed several watts wirelessly between two coils using quite low cost “wireless charging” chips as used on mobile phones for example.  So how about putting everything inside the vacuum chamber that’s needed for maintaining and measuring the pendulum with a serial Bluetooth or WiFi link to extract the data?

[duncan webster 1Participant @duncanwebster1]

JH has spotted the deliberate? mistake, but for small angle it won’t make any difference. If I were really into nit picking I’d point out that m.g.sin(theta) is at right angles to the pendulum, not horizontal, so at the support, the horizontal load  is m.g.sin(theta).cos(theta), but for small angles cos(theta)=1.

I think the problem with flexibility of support isn’t the change in effective stiffness, it is damping, which reduces Q and so increases the necessary drive

[John HaineParticipant @johnhaine32865]

Well that’s my point really – support compliance only reduces Q if it’s also lossy which it isn’t necessarily.  I still think the best way forward is to use 2 pendulums in antiphase so their support reactions cancel.  It turns out that for this to work really well you want a high compliance support with lots of damping!

[duncan webster 1Participant @duncanwebster1]

July 2017 BHI Journal, article on falling ball gravity escapement mentions a double pendulum clock running in vacuum. Made by a demigod called Jim Arnfield. I have a scanned copy, but it’s probably copyright

[SillyOldDufferModerator @sillyoldduffer]

Much interrupted today, and haven’t had a chance to absorb posts yet.

In the margins thought about stiffening the tower, and wonder if clamp on cheeks in between the fore-aft pillars would help.  Removable to allow access.  In place they inhibit fore-aft movement.  Could be few of them providing spaced out cross-bracing, or several could be used to wall most of the structure in creating a poor man’s box-girder.

 

 

The cheeks are grooved with an 8mm chord to grip the pillars, and could be glued with weak thread-locker. (Because they have to come off for access.)

Big problem is having so little space inside the vacuum chamber.

Is this idea worth the effort?

Won’t get much done tomorrow – the family are coming…

Dave

Ps the cat is persecuting me.  No idea what it wants.

[SillyOldDufferModerator @sillyoldduffer]

If anyone fancies a little light reading, the Mk2 design overview is now available as a PDF on Dropbox.  37 pages and rising. Comments welcome, no obligation.

Dave

 

[duncan webster 1Participant @duncanwebster1]

I think the first thing to do is find something solid to stand it on. It may well be that the clock frame is stiffer than the mount at present. It almost certainly is contributing less to damping.

Then some shear member connecting the top and bottom plates in a plane parallel to the plane of the pendulum  swing. Could be a bit of square bar bolted to the bottom plate between the uprights with a piece of sheet steel, say 1mm thick, 76 wide bolted to it and the top member. Friction should be enough to transmit the very low force.

Even strips bolted diagonally top to middle and middle to base, 2 X shapes one atop the other.

Sums to follow

[duncan webster 1]

On duncan webster 1 Said:

I think the first thing to do is find something solid to stand it on.

+1 for emphasis

MichaelG.

[duncan webster 1]

On duncan webster 1 Said:

I think the first thing to do is find something solid to stand it on. It may well be that the clock frame is stiffer than the mount at present. It almost certainly is contributing less to damping.

Agreed.

Then some shear member connecting the top and bottom plates in a plane parallel to the plane of the pendulum  swing. Could be a bit of square bar bolted to the bottom plate between the uprights with a piece of sheet steel, say 1mm thick, 76 wide bolted to it and the top member. Friction should be enough to transmit the very low force.

Even strips bolted diagonally top to middle and middle to base, 2 X shapes one atop the other.

On the to do list.  Requirements are:

• bracing must fit in the limited space available.   Adding cross-support would be a doddle if the pendulum wasn’t swinging inside a gantry and the gantry wasn’t a very close fit inside the vacuum chamber.
• bracing must be removable to allow access to the electromagnet, photo-interrupter, bob, and suspension.

Sums to follow

Many thanks!

Won’t get much done today – hosting a family event.

So I quickly moved the clock to a safer/firmer location and set it going.  Reluctant to start, probably because moving it disturbs the pendulum and level.   Have an hour to tweak the starter settings, and then I want to let it run to see how it behaves.  Stopped yesterday and I think the impulse may be a shade weak.  Mustn’t jump to conclusions because I bumped the clock at least twice;.  putting it on the floor next to a well used office chair with sticky out wheels wasn’t smart.

Too much going on.

Dave

[Michael GilliganParticipant @michaelgilligan61133]

May I suggest an experiment, Dave ?

Given the design weight of 7kg, and the bottom-heavy configuration, it would be a relatively simple matter to suspend the clock ‘free-free’ using bungee cords.

Observing its behaviour in that condition would help highlight any mechanical issues with the construction.

Alternatively … for a quicker/dirtier test … just stand it on a soft foam cushion.

MichaelG.

.

Edit: __ For anyone not familiar with ‘free-free’  suspension … see the bungees in use here:

[Michael Gilligan]

Interesti

On Michael Gilligan Said:

…

Given the design weight of 7kg, and the bottom-heavy configuration, it would be a relatively simple matter to suspend the clock ‘free-free’ using bungee cords.

Observing its behaviour in that condition would help highlight any mechanical issues with the construction.

…

Interesting idea, sort of naively done already by comparing Q on a table-top with Q on a solid-floor.  (My clock’s Q on a solid floor is almost double that on my dining table.  Shows there is a problem.)

But you mean something smarter I think?  Revealing how far the gantry moves and in what directions as an aid to minimising deflections inside the clock. How would you instrument the structure?  I have zero experience or understanding, and am pretty sure none of my engineering books cover it beyond saying it’s done.

Solid Edge can do vibration analysis so it might be possible to model it.

I can provide the model if you’re up for attempting the analysis.   Too much for me to take on, and I’d be starting from square one.  Ignorance is not bliss!

Dave

[John Haine]

On John Haine Said:

As for feed-thrus etc.  Provided the wall thickness isn’t too large one can feed several watts wirelessly between two coils using quite low cost “wireless charging” chips as used on mobile phones for example.  So how about putting everything inside the vacuum chamber that’s needed for maintaining and measuring the pendulum with a serial Bluetooth or WiFi link to extract the data?

Would work for the data connections and the power.   The electromagnet feed and photo-interrupter are time critical though – have to be wired, though optical could be made to work.  Passing power wirelessly through the cast-iron block wouldn’t work.  Except the charging and comms loop could be on top, with only a few millimeters of PVC in the way.  The evacuating pipe still needs a hole though.

I’ll add it to the possible improvements list.

Ta,

Dave

[John Haine]

On John Haine Said:

…  Oxford Prof. E T Hall on his retirement dedicated himself to building the “Littlemore Clock” (he lived in this village which is really a suburb of Oxford).  The pendulum swung in a vacuum, had a Q approaching a million, and was mounted on a 6 ton concrete pillar cast into a large hole in his garden.  Despite his best efforts the clock was actually worse than the Shortt at detecting earth tides, and it is believed this was because of seismic noise.  Amongst other things there was a large tree quite close by and a railway line carrying freight to and from Cowley motor works through the village. …

Wow!

🙂

Dave

[SillyOldDufferModerator @sillyoldduffer]

For comparison, this is the Livermore Clock’s block diagram:

And my Mk2:

Many similarities, except Professor Hall’s clock is much better made and isolated than mine. That he failed is worrying, I’m doomed!

🙂

Dave

[SillyOldDuffer]

On SillyOldDuffer Said:

Interesting idea, sort of naively done already by comparing Q on a table-top with Q on a solid-floor.  (My clock’s Q on a solid floor is almost double that on my dining table.  Shows there is a problem.)

[…]

Too much for me to take on, and I’d be starting from square one.  Ignorance is not bliss!

Dave

Sorry, Dave … I think I may have misled you by including that video

I was only using it to illustrate the method of free-free suspension, NOT actually suggesting that you attempt a modal analysis [which usually requires a great deal of preparation, and some expensive kit].

I do think, however, that suspending the clock and simply letting it ‘do its own thing’ would be informative. …

MichaelG.

[John HaineParticipant @johnhaine32865]

Littlemore please!

https://maps.app.goo.gl/gNtLBnREeJocbFrJA

I once managed to find what I think was Hall’s house:

https://maps.app.goo.gl/HmJb9WjfWbvQRe1i8

Shrink the scale and look for the railway line.

I don’t agree that the clock is well isolated.  Yes it sits on a massive concrete pillar set in the ground, but given that the density of concrete is about the same as the soil, and there’s only a layer of rigid foam separating them, why should it be any more isolated than just plonking the baseplate on the ground?  A couple of reasons have been advanced for why it didn’t perform better – this by Tom Van Baak:

http://leapsecond.com/hsn2006/pendulum-tides-ch6.pdf

The most likely source of random FM is ground vibration.

Berneard Cleyet somewhere also suggested that the period measuring system may contribute noise.

[John Haine]

On John Haine Said:

Littlemore please!

Doh! I blame my brain.

https://maps.app.goo.gl/gNtLBnREeJocbFrJA

I once managed to find what I think was Hall’s house:

https://maps.app.goo.gl/HmJb9WjfWbvQRe1i8

Shrink the scale and look for the railway line.

I don’t agree that the clock is well isolated.  Yes it sits on a massive concrete pillar set in the ground, but given that the density of concrete is about the same as the soil, and there’s only a layer of rigid foam separating them, why should it be any more isolated than just plonking the baseplate on the ground?  A couple of reasons have been advanced for why it didn’t perform better – this by Tom Van Baak:

Well, isolated by my standards.  My clock is less than 2m from my feet and only 0.8m away from the fans on the back of my computer.

I wonder if the pillar got down to bed-rock?  If not, the foundation is floating.

My home is built on a filled in quarry.  My end of the estate may be on solid ground just beyond the face.   Not clear from the deeds or old maps if the quarry extended all the way to the public road.   Houses beyond  the end of my garden are definitely built on concrete rafts to spread the weight on made ground.   Gawd knows what that would do to a pendulum.

http://leapsecond.com/hsn2006/pendulum-tides-ch6.pdf

The most likely source of random FM is ground vibration.

Berneard Cleyet somewhere also suggested that the period measuring system may contribute noise.

Thanks for the link to Tom’s paper. I ran a spectral analysis of my Mk1 clock and couldn’t see any tides.  Might have got it wrong! More likely the clock wasn’t good enough.   Tom’s paper is helpful – I can probably replicate what he did.   Unlike Woodward, I don’t have to write my own FFT.

Dave

[SillyOldDufferModerator @sillyoldduffer]

Results from last night’s run, which is testing reliability not accuracy.   Had some bother setting impulse power so that amplitude neither rises or falls, and noted some odd racing and stoppages.  Almost as if the pendulum is tightening and loosening.  Could also be a photo-interrupter or software problem.

There shouldn’t be any spikes in the left-hand graph.

The first spike is a double counted beat, backswing and downswing both counted.   One error in 50000.  Hmmm,

Next event shows amplitude dropping to zero (much too high), then decaying quite rapidly.  Then, for no apparent reason, amplitude slowly speeds up, and returns to normal.   Period isn’t badly affected by either decay or recovery.

First glitch: amplitude jumps from to 70492 to 291260, then back to 70492μS.   A misread!

Second glitch: amplitude slows to over 100000μS. then picks up speed, up to 99688 on the screen, thereafter gradually accelerating back to normal.

I think the impulser may be getting out of phase with the bob.  It’s supposed to impulse on the downswing only: if the electromagnet impulses on the backswing, it acts as a weak brake.  If it gets back in phase, the impulse accelerates the bob, and it speeds up again.  Matches the data, but why?.

Next step – see if it happens again. If it’s a software bug, there may be a pattern.  Absence of pattern suggests a mechanical or electronic problem.

Measuring is double edged.   Wouldn’t have noticed this problem had I been clock-watching manually!  And if I hadn’t seen the graph, I would be happily turning and milling metal, not typing this!

Dave

[blowlampParticipant @blowlamp]

I suspect some of this could be due to slight rocking.

Could you setup a DTI next to the clock (at the top) to see if it moves as it swings and if walking around the room has any influence?

 

Martin.

[blowlamp]

On blowlamp Said:

I suspect some of this could be due to slight rocking.

Could you setup a DTI next to the clock (at the top) to see if it moves as it swings and if walking around the room has any influence?

 

Martin.

Yes, though I doubt my 0.02mm dial is sensitive enough; maybe with a lever?  The pendulum is disturbed if I clump around the room.  The Mk1 version, with a similar pendulum, detected heavy traffic in the road, and the Moroccan 2023 earthquake.

I wish you’d posted before I moved the clock yesterday! I put it too close to the wall, so now I can’t bounce a laser pointer off a mirror on to a wall 8m away.  Amplifies the movement, and Mk1 caused the distant dot to wobble, less than a millimetre.  Not tried it on the Mk2, but the mechanics are almost identical.  Thanks for reminding me!

The glitches occurred at about 01:30 last night. Should have been quiet.  Maybe I was sleep-walking.   Or, this house moves as it cools overnight, various odd creaks.

Dave

 

[John HaineParticipant @johnhaine32865]

Doug Bateman’s clock has a “phase” sensor to remind the control circuit which side to impulse on every swing.  Relying on “machine memory” can be unreliable.

[John Haine]

On John Haine Said:

Doug Bateman’s clock has a “phase” sensor to remind the control circuit which side to impulse on every swing.  Relying on “machine memory” can be unreliable.

So it seems!  I’ve added a phase sensor to the TODO list.    Means another wire through the bulkhead though, and changing the electronics.

Went unhappy to bed last night:

• Clock glitching
• Realised that the clock has to echo commands back to the PC prefixed with a # (which the stats program ignores as comments.) Fixing it caused a new bug to do with CRLF line ends.
• Discovered my governor logic is too simple; hunts forever about 1 in 10 times.
• Still not cracked:
  • setting the clock accurately,
  • PVC vacuum chamber concerns,
  • the bulkhead penetrator
  • where to put the clock; my house is no good!
• Rigidity needs improvement.
• Depressed by the failure of the Littlemore Clock, despite it’s expert best practice build, no expense spared.   Cause appears to be noise, the clock being located near a railway line.  Anyone know more?
• Not keeping on top of the paperwork.  Losing the earlier notes was a disaster, plus new information is flooding in.  I restarted the project with a clear plan, now I have to re-prioritise a big TODO list.
• Found a paper on leapsecond.com in which Tom van Baak explores ideas I thought were original to me.  With one exception, they’re not, and Tom lists a few I hadn’t thought of.   I empathise with Scott, who, after an enormous effort, arrived at the South Pole to find a Norwegian flag flying on it!
• Despite feeling considerably better, experimental treatment, I’m still unwell, and had a relapse last night.  Can’t take good health for granted.

Time for a break from armchair engineering.  Far harder than working to someone else’s plans and I’m exhausted.  Off to the workshop for therapeutic metal work – the vacuum valve assembly.  Might test the vacuum pump too; can I get below 500mb.  Robert suggested sealing wax and I’m keen to try.

Dave

[John HaineParticipant @johnhaine32865]

Regarding Littlemore.  One of Tom’s papers looks at the power spectrum of the rate data and concludes that the clock was suffering from white noise modulation of its frequency.  A potential cause of that is a noisy gravity field with a white spectrum, as frequency is (to a first order) proportional to g.  This would imply vertical vibration (acceleration) of the clock.  Unfortunately a 2-pendulum clock is insensitive to horizontal accelerations but not vertical.

Precision pendulum clocks are essentially gravimeters, so g variations set the ultimate limit on accuracy.  Somewhere I read that the Russians at one stage were marketing the Fedchenko as a gravimeter.  The location of the clock is as important as its design.  “Suspending” the clock so it is isolated from vertical vibration is remarkably difficult.

[Author]

Posts