Help wanted designing a source of 0.5Hz pulses, please

[Nickthedentist]

First of all, I do hope that this topic is not considered too OT for this forum. My sincere apologies if it is. If, as I suspect, it is just borderline, I would be very grateful if it could be allowed to stay for a few days.

I have recently acquired a home-built 1930s electric clock. Not the usual mains-driven synchronous sort, but a battery-operated design based on the "Hipp Toggle". Essentially, when the pendulum's amplitude of swing becomes too small, a switch operates momentarily, putting a pulse of current through an solenoid on the pendulum's bob, which gives the pendulum a little "kick" to get it swinging more vigorously. This happens every half minute or so.

Sadly, both I and the previous owner/repairer have struggled to get things going the way they were designed. It seems impossible to get it to run accurately, in spite of many hours of tinkering. The problem is that the amplitude of swing varies too much. The Hipp contact only closes when the swing becomes too small to advance the train, and then the solenoid gives it such a boost that the first wheel is advanced twice for the subsequent few swings (it’s driven by a ratchet and pawl arrangement).

So I'd like to cheat, in such a way that the original design can be easily re-instated of course.

My idea is simply to pass a pulse of current through the pendulum’s solenoid every two seconds, to keep it swinging with constant amplitude, and leave the Hipp toggle's contacts completely out of circuit. The solenoid has a resistance of about 200 ohms, and momentarily passing about 25-50mA through it (i.e. with about 5-10V) seems enough to keep it swinging.So what I need is a source of 0.5Hz pulses, a few volts in amplitude at a few tens of mA.

I thought I could use the PCB from a modern battery-operated quartz movement as a convenient source of accurate 1Hz pulses (at about 0.5V) to trigger the circuit. So I suppose I need a divide-by two circuit and then some kind of “output stage”.

Any ideas, anyone?

Cheers,
Nick.

[Brian R Pateman]

Nick,

You need something very basic.

How about a small transformer 6 volt primary for example. Use a half wave rectifier and the result is 100Hz.

Then all you might need is a simple pulse shaping circuit, perhaps a Schmitt trigger or even a good old fashioned relay.

Regards,

[Nickthedentist]

Thanks for your reply, Brian!

But I have to admit that I'm confused - circuit design is not my forte

How do I get 0.5Hz I need from the mains-derived 100Hz?

Thanks,
Nick.

[Brian R Pateman]

Sorry Nick,

I was working in metric minutes (they have 50 seconds) the cold has addled my brain.

I'd still keep it simple. Use a simple CR circuit to operate the aforementioned relay.

I used to run several "Pulsynetic" clocks from a small motor with a cam operating a microswitch and that gave pretty good results.

Regards,

[GrimJosef]

The only thing you haven't said Nick, explicitly at least, is how long you want the output pulses to be. Are they meant to be "high" for 1 sec then "low" for 1 sec or do you just want a "short" pulse every couple of seconds ?

Depending on the answer you'll then need just a CMOS flip-flop (cheap as chips) followed by a transistor to handle the current and probably an input transistor to amplify the 0.5V trigger pulses. That'd give you 50% duty cycle. If you want shorter pulses you'll probably need to use the other flip-flop in the package to shorten the output.

Cheers,

GJ

[John M0GLN]

Won't the 555 timer IC do for an application like this? Just been looking in the Radio Shack booklet and they got a Metronome and a Pulse Generator circuit, besides various other oscillators.

John

[Nickthedentist]

Hi, GJ!

Originally Posted by GrimJosef

The only thing you haven't said Nick, explicitly at least, is how long you want the output pulses to be. Are they meant to be "high" for 1 sec then "low" for 1 sec or do you just want a "short" pulse every couple of seconds ?

Good question! I think a "short" pulse every 2s would be what I need (though a really accurate 0.5Hz sinewave would probably do as well).

Originally Posted by GrimJosef

Depending on the answer you'll then need just a CMOS flip-flop (cheap as chips) followed by a transistor to handle the current and probably an input transistor to amplify the 0.5V trigger pulses. That'd give you 50% duty cycle. If you want shorter pulses you'll probably need to use the other flip-flop in the package to shorten the output.

That makes perfect sense. Do you know which 4-series IC contains 2 flip flops? Then I could look up some appliaction circuits, maybe.

Thanks,
Nick.

[Nickthedentist]

Hi John,

Originally Posted by jcaines

Won't the 555 timer IC do...

Yes, I thought "555" to start with too, they're very versatile things.

But accuracy wouldn't be good enough for running a clock - unless I used one to provide the pulse, and manually triggered it from my 1Hz quartz signal via a divide-by-two arrangement.

Nick.

[paulsherwin]

A simple R-C controlled multivibrator can be quite accurate, though it needs to be adjusted to run at the right speed just like a mechanical clock. Lots of battery clocks used this arrangement before the widespread adoption of crystal control.

If you build one with two transistors rather than using a 555, you can put the solenoid in one of the collector lines and can dispense with a driver transistor. It would be very easy to knock something together and experiment.

http://www.hobbyprojects.com/multivi...s/astable.html

Paul

[KeithsTV]

Trying to adjust a low frequency RC oscillator can be a very frustrating and time consuming exercise. A better solution might be to use a divider such as the CMOS 4060 which has a 14 stage binary divider. This has a section which can be used as an oscillator which can be run at 8.192kHz. The output of the final divider would be 0.5Hz. This can be used to trigger a 555 configured as a monostable to give a short pulse for the solenoid.

Keith

[ppppenguin]

A 4060 and a standard 32768Hz watch crystal will give you 2Hz. Another 4060, partially used will give 0.5Hz square wave. Alternatively the 2 halves of a 4013, part of a 4040 etc etc will do the divide by 4. This will run on anything from 3V to 15V. Then a single transistor circuit with a couple of Rs and a C to differentiate the square wave to get a pulse. Or maybe a 4098 monostable.

I can probably give you a selection of these chips from stock. Data on 4000 series CMOS should be readily available on line.

[Top Anode]

I'm so out of practice that I may be missing the point entirely, but isn't the aim to get the pendulum to do the timing?

I'm assuming that it is operating as a true pendulum.

The problem is that the amplitude of swing varies too much.

This issue would be made worse, surely?

I know you mentioned the ability to change back to the original design, but even trying to 'push the time into the pendulum' in the short term, seems counter (Groan) to the aims of a restorer...furthermore, wouldn't there be an issue of the energy being applied not having any feedback?

Frivolous mode: On (But it says what I'm trying to get at.)

Ah, I have it! You could fit micro switches onto the sides of the case that would cut the power for a few seconds if the pendulum's stride got too great.


I'm full of bright ideas at one in the morning.

[Station X]

It will also be necessary to make sure that the pulse occurs at the correct time ie when the pendulum comes close to the electromagnet.

[Mike Phelan]

Hi Nick
I've only just spotted this thread - a few thoughts about the various replies and a few solutions - especially the last paragraph:

Originally Posted by Top Anode

I'm so out of practice that I may be missing the point entirely, but isn't the aim to get the pendulum to do the timing?

Correct - that is (was) its function.

I'm assuming that it is operating as a true pendulum.

As opposed to what?

The problem is that the amplitude of swing varies too much.

True, that's circular error, but a Hipp toggle ought to work reasonably well with a constant voltage source - any variations will be because of mechanical problems.

Ah, I have it! You could fit micro switches onto the sides of the case that would cut the power for a few seconds if the pendulum's stride got too great.

A bodge that would not work at all!

A non-crystal controlled like a 555 would not be accurate at all.
A 32768 Hz Xtal could be divided down with bistables from the CMOS or TTL chips, or even transistors. As in the frequency dividers on an electronic organ.

However, you don;t really need any of that. Far too complicated.

You only need an impulse every 30 seconds or less to keep the pendulum maintained, not every half-second. This could be done by an optocoupler and the seconds hand on a cheap quartz clock. The impulse needs to be very short - exactly how short depends on the mechanical construction.

Finally, and this is the real problem that no-one has spotted yet:

The Hipp contact only closes when the swing becomes too small to advance the train, and then the solenoid gives it such a boost that the first wheel is advanced twice for the subsequent few swings (it’s driven by a ratchet and pawl arrangement).

That is why the timekeeping is no good; it should not miss a tooth or count two teeth. The pawl should count one tooth on all occasions.


There are two possible reasons for this, neither of them electronic.

• The Hipp toggle is only tripping when the pendulum arc is too small.
• The design or adjustment of how the pawl operates is wrong.

Finally, what sort of accuracy are you expecting, in, say, minutes per week?

HTH

[Antlong]

Hello Nick,

On the face of it, the solenoid is overpowerful - would it be adequate to reduce the current through the solenoid? Lower the supply voltage or insert a series resistance?

Regards Ant

[PJL]

I think I would stick to trying to solve it mechanically. My suggestions are:

Dirty contacts in the switch means too much hysterisis resulting in the extra long pulse when it does finally make. I guess you have tried cleaning the switch, can you fit something else?

Modern alkaline batteries have low resistance, try a series resistor to reduce the current?

Put a stop in to take out the excess momentum in the pendulum, perhaps a bit of spring steel?

[Top Anode]

Okay, no jokes. Mike says:

Finally, and this is the real problem that no-one has spotted yet:

and goes on to describe accurately the issues with the original design. But surely, the fixes that are being suggested are trying to modify the prime function of a pendulum.


My statement: I'm assuming that it is operating as a true pendulum.

As opposed to what?

Well, I can't answer this,

EDIT: Perhaps I should. My interpretation of the first post is that it's wished to get the clock going, abandoning the "Hipp Toggle", but allowing the old tech method to be recoverable if so desired. This would leave the question of the pendulum. If it's still going, then by my argument, it would not be a true pendulum, more an interesting visual feature.

I was simply trying to make quite sure that the pendulum was the fundamental source of the timing. I did this because I was confused. How could firing a pulse at a timed interval, be integrated with the natural physical laws of a pendulum? Accurate enough, yes, but it would have to stay in phase for days. Very tall order, unless there is a feedback from a pendulum position sensor. The thinnest wisp of wire would trigger a 555 for instance. But this all seems to be spoiling what is a rather interesting machine. This toggle I see was in use for 'over one-hundred years'...please keep it original.

If there is to be no feedback, then as I see it, the pendulum simply has to be given energy to overcome aerodynamic and mechanical friction, and also the more esoteric disturbances of the Earth's gravitational anomalies and perturbations. As far as I can see, this energy input has to be applied in a manner that will not alter the pendulum's timing, or the charm of the timepiece will be lost.

[Mike Phelan]

Originally Posted by PJL

I think I would stick to trying to solve it mechanically.

Absolutely!

Dirty contacts in the switch means too much hysteresis resulting in the extra long pulse when it does finally make. I guess you have tried cleaning the switch, can you fit something else?

The Hipp device will always ensure that the time that the contacts are closed will always be the same - it depends on the arc of the pendulum, which will be the same to all intents and purposes.

Put a stop in to take out the excess momentum in the pendulum, perhaps a bit of spring steel?

That will wreck the timekeeping altogether.

I still maintain the poor timekeeping is purely due to the fact that the count wheel is sometimes missing a tooth, sometimes catching two teeth. That cannot work properly!

Nick - have you watched how often the contacts operate (how many swings) over a period?

I made a similar Hipp-derived clock that runs on three HP2 batteries; with new batteries it takes about 8 swings to impulse, then for the next six months settles down to 6. As the batteries go down to less than a volt, it will stop after a few weeks whan the impulses get to every other swing.

Timekeeping is a minute or two a week.

[Hermit6345]

This problem sounds familiar!

I have a similar problem with a partially built clock, which seems to work in a similar way to Nicks. It has a one second pendulum and the clock mechanism relay needs a pulse every 30 seconds. I did maintain the pendulum momentum initially by just using the relay contacts in a sort of mechanical bi-stable. This meant that the pendulum was energised at every swing. Not how it's supposed to work, but the pendulum kept going for the 24 hours that I trialled it. My next idea was to use opto techniques to sense the pendulum swings, count these up and every thirty seconds, send a pulse to the clock mechanism and also energise the bit that gives the pendulum a kick so that the pendulum momentum is retained and also resets the mechanism.The pendulum weight is a huge lump of lead encased in brass. There are two bits to the pendulum control, one relay energises and drops the weight onto the brass bit on the pendulum and the other relay resets it all. I don't see any reason why the clock bit cannot be used being driven from a modern xtal controlled osc with some arrangement to operate the pendulum control relays. I have pictured my clock bits below. The pendulum is not shown because the threads holding the weight onto the end of the pendulum shaft require renewing.

[GrimJosef]

Originally Posted by Top Anode

I was simply trying to make quite sure that the pendulum was the fundamental source of the timing. How could firing a pulse at a timed interval, be integrated with the natural physical laws of a pendulum? Accurate enough, yes, but it would have to stay in phase for days. Very tall order, unless there is a feedback from a pendulum position sensor.

I think the effect of Nick's scheme, if it can be made to work, will be to make the crystal oscillator the fundamental source of the timing. The pendulum will then simply become a "driven resonator". As with any resonator the Q will be finite, so it will have some range of frequencies over which it can be driven perfectly effectively. Since the crystal is a lower noise source than the pendulum was and since, presumably, they're both tuned to the same nominal centre frequency (0.5000000... Hz or a harmonic thereof) there will be no problem with phase drift. If the pendulum should start to go out of phase the kicks from the solenoid should quickly (a few cycles ?) drag it back to where it should be.

I agree with Mike. In principle if this thing ever worked then it can be made to work again. However this may involve re-making worn mechanical parts. If it's just a few, if they don't need specialist tools to make them and set them up and if you're "good with your hands" (unlike me) then this would be the way to go. But if any of these is not the case then you may do more harm than good and the electronic fix should at least be completely reversible when someone with the necessary tools/time/expertise to fix the mechanics comes along.

Cheers,

GJ