Precision pendulum techniques

Yes, apart from pseudo-gaussian. I think that's a form of curve, but exactly what and why?

I've thought of filtering the digital pulse to my electromagnet, but not tried it. The resulting magnetic field is already filtered by the magnet's inductance, but I've not measured it. Must put a Hall Sensor on my scope and have a look. Hadn't thought of a DAC, (how fast it would it need to be?), but I have a waveform generator that can shape pulses to test the idea.

I've attempted a soft impulse by having the electromagnet some distance from the bob, because the field weakens quite rapidly with distance. The impulse is fired soon after the bob passes dead centre, and stops before the bob has travelled far. (Not done the sums!) Then I adjust the pulse length down until the pendulum gets just enough energy to keep going. I've not felt the need to experiment with deliberately shaped pulses because statistical analysis of the log data doesn't show much sign of impulse disturbance. (That the pendulum is disturbed by more powerful pulses does show up in the stats.) I guess a low Q pendulum would benefit more from a shaped impulse more than a high Q job. Low Q needs more energy, making it likely that the pendulum has to be thumped. On the other hand, a high Q pendulum could be more sensitive...

The combinations feel endless!

Dave

I have very little knowledge of the hi-tech electronic devices that are being used here to test experimental pendulum systems.

I can see that the testing system must be more accurate than the tested system, and that the results look rather messy.

Have these hi- tech testers been applied to known high performance clocks (such as synchonomes ) in order to see how  messy the results look?

I thought Dave was trying for a high performance clock, not a world beater.

dave8

Edited By david bennett 8 on 11/03/2023 22:34:57

A true Gaussian (a "bell curve" ) has infinite tails, which you wouldn't want, so it would need to be a modified one that starts and stops with zero force. Perhaps a modified Gaussian-like profile without the infinite tails is not perfect for this, but it has got to be pretty close. You want it to start almost imperceptibly, grow smoothly to a peak, and then decline back to zero in the same way.

P.S., that darned insertion of a winking smiley when you type " followed by ) is so irritating!

Edited By S K on 11/03/2023 22:35:43

We could call it a raised cosine, as someone mentioned before, either on this thread or one of the others.

Ah, good. Thanks for that.

Robert Matthys and Ned Bigelow both experimented with sinusoidal drive. A moving magnet transducer attached to the pendulum connected in an oscillator circuit with an op amp. IIRC there's a chapter on it in Matthys' book.

Edited By John Haine on 12/03/2023 08:06:20

Isn't the drive entirely controlling the pendulum frequency at that point?

No, the pendulum is the resonator. Like the crystal in a crystal oscillator.

Yes, but not by me, in various ways, and as far as I know the data isn't available. I've thought of building a John Haine style clock - long rod, heavy pendulum, Helmholz coils, bolted to wall etc, simply so I can measure it for comparative purposes.

I'm certain from reading books that measuring a pendulum accurately reveals 'noise', i.e the period is almost never spot on, but beats scatter more-or-less evenly about an average, or move consistent with environmental change, notably temperature and air-pressure. Many other sources of noise. Vibration due to traffic or seismic activity, the house moving with temperature, and exotics, such as Invar being unstable at the molecular level, or micro-gravitational changes due to ground water moving with the tide.

Although I expect to find noise, I don't have a good measure of how good or bad my clock is relative to better made examples.

I'm aiming for a world beater in the sense I'm comparing my 'made with modern parts' clock with a Shortt-synchronome of 1921. Be good if my clock performed better than a Shortt, but I'm a long way off achieving that!

A more cheerful test result today. A few days ago I reported a laser beam bounced off by tripod detected movement. The tripod was on a dining table, so I repeated the test by putting the tripod on the floor, which is teak parquet on concrete. I also rearranged the beam to amplify movement more. This time I couldn't see any movement in the remote dot.

So, the tripod is stiffer than my dining table. Interesting that a 40g bob swinging about 15mm shakes a two-man lift wooden table! The reason is the table top is sat on four long and relatively wobbly legs.

Dave

I've got a triple fusee bracket clock which runs if you put it on a concrete floor, but not if it's up on furniture. It obviously needs some attention, but it shows the effect of lack of rigidity in the support.

Same thing with longcase clocks standing on carpet. My old mate Barrie got caught out with testing a known runner of a long case movement to sort the strike out and it just would not run. Eventually worked out it was the ancient clock horse standing on carpet to be the cause. The thing that alerted him was the clock weights had started swinging.

Now there is a thought. If you attach a second pendulum to the suspension frame of similar period to the clock pendulum would that provide a sensitive test for rigidity. You could perhaps measure the time constant for increase in amplitude. It would only really be a comparative measure but after all you just want to know if you are heading in the right direction. Sort of wobble integrator.

regards Martin

Edited By Martin Kyte on 12/03/2023 14:27:35

Edited By Martin Kyte on 12/03/2023 14:28:26

Or attach a second identical pendulum permanently to swing in antiphase and balance the clock...

Well yes but I was thinking of a simple way for Dave to see if he’s heading in the right direction. Not sure he has room for two pendulums.

regards Martin

So Matthys explored a continuous drive approach using clipped sine wave (apparently first tried in 1960). The notion is to continuously restore the lost energy. So at the extremes of the swing, the restoring force should be zero, and at BDC, when the speed and hence air resistance is at its maximum, the restoring force should be at its peak.

I don't understand the original motivation for clipping, though - it seems plainly contrary to what is desired, as Matthys eventually points out. Also, while continuous drive seems to make sense at some level, it's long been known that the ideal time to impulse (at least for a brief one) is at BDC. Continuous drive would also have to be quite precise about its phase relationship, or near the extremities it may start pushing when it should still be pulling and vice versa.

Matthys eventually comes to the conclusion that continuous drive is not ideal, and that a brief impulse at BDC is likely superior. He also expresses concern about the stability of the drive, which goes back a long way in the earliest attempts at electromagnetic drive. He did mention that a continuous drive could at least be more gentle over its profile than a shorter one at BDC.

My thought is that a raised-cosine drive or similar profile concentrated at BDC is likely to be superior to a "digital" one, as that "twanging" - the shock of a sudden sharp impulse that some here have observed - would likely be lessened.

I don't! This is the latest design:

Scale: about 300mm tall, the bob is 25mm diameter, the base 140 x 140mm.

Dave

As I'm not intending to try for a "world beating" pendulum-based system, and because I'm motivated to learn more about machining, I've been musing about developing some sort of "improved" mechanical impulsing technique.

I happen to have a small servo that purportedly can be set to apply no more than a preset maximum torque. That's not what I want, though: I'd like to apply a varying torque, e.g. in a raised-cosine. If the baud rate is high enough, perhaps a continuously-varying torque could be approximated just using the servo, though.

I've also seen cam-based systems that can apply varying torque, but they are not adjustable without cutting a new cam. In combination with a programmable-torque servo driving the cam, perhaps they could be made adjustable, but it feels like a fixed cam would inevitably have to be redesigned several times and perhaps never be quite right without some insane math to help figure out the profile.

The nice thing about a servo is that they are so easy to control. But another problem may by how to reduce the torque, as a servo may be too strong compared to the tiny amount of force needed. I guess a 1 out of n swing approach could mitigate that.

I was also thinking about a lever made of thin spring steel that would soften the torque application by flexing (I saw something like this in an ancient clock patent). Perhaps an appropriately-designed spring like this could, in conjunction with a fixed-torque servo and a feedback loop (strain gauge, capacitive sensor...) could work.

A voice-coil is another possibility, but it would require a more complicated drive circuit than a servo would. And at that point, you might as well make an electromagnetic drive anyway.

(Also, because I like quirky, amusing things that might make someone else smile too, I was even thinking of making a little hand or finger that applied the impulse to the pendulum - hehe.)

Unless you are completely satisfied with whatever residual horizontal movement is in the hinge area, why not add some cross-bracing in the direction of the swing?

I probably will, the pillars just need some small holes drilled in them to take wires I can tourniquet tight to cross-brace the verticals.

On another tack, I've attempted to capture the shape of the magnetic pulse emitted by my electromagnet, which I expect is typical. Not brilliant because my Hall sensor (A1302) is deaf as a post but:

The noisy yellow line is the strength of the magnetic field. The blue line is voltage across the magnet, which for the test is fed through a 220ohm resistor, so the blue curve is proportional to current.

There's a sharp current spike when the pulse starts, and it drops rapidly as current flow is resisted by the magnet coil's inductance. When the pulse ends, the magnetic field collapses causing a negative going kickback, which, as can be seen, is clipped at about -0.6v by a protective diode across the coil.

Field strength is proportional to the current flowing in the coil and how far away the bob is, and both change during the impulse. My maths is terrible, but should be possible to graph how the strength of the magnet field varies as the bob flies through it. My feeling is the two interact to soften the jolt, but I could be wrong!

Dave

There had to be some way to limit amplitude and the most direct way given the very low frequency was to allow the OP amp to clip. One could also nowadays do it digitally by sampling the waveform, rectifying the waveform in a processor, and use a multiplying DAC to adjust the loop gain. Matthys and Bigelow disputed the significance of the finite gain bandwidth product of the OP amp, I suspect it was negligible. I think the main problem with this method is that making the transducers is tricky.

Some form of low-pass filtering could help with those sharp edges, which can't be that great for the pendulum's performance.

I'm far from an analog guru, but I suspect John would be able to provide some guidance, perhaps even a step towards that proposed raised-cosine profile.