Electromagnetic pendulum clock motor

[CarverJerry]

This is a cool project. Hope to build it some day. Have any idea what the kit will cost?

[DickB]

Not yet. It will depend in part on how many component parts I am willing to order in advance, to keep shipping costs in check. But I would definitely like to keep it affordable.

[DickB]

While working on build photos and instructions for this project, I've also been experimenting and working on alternative drive electronics. I've learned some things about this type of pendulum and drive in the process, with a very promising development.

If a pendulum's angular displacement is small, the pendulum's period is dependent primarily on the effective length of the pendulum. Most of the wood gear clocks and plans that you see use long, two-second period pendulums with low angular displacement. So this rule largely applies. The pendulum that I (and others typically) use is a shorter, 1-second period with relatively high angular displacement in order to drive the clock works.

For a free pendulum, as the angle of displacement increases, the period of the pendulum also increases. However, for a driven pendulum, at least one driven as I am doing, the pendulum period decreases as the angle increases. I've been varying the pulse with of current applied to the driving coil, and measuring the period to fractions of a second using a microcontroller. I found that can alter the period of the pendulum sufficiently to speed up or slow down the clock without changing the pendulum.

If you look at my first post, you’ll see that I intended to run this clock under control of a microcontroller all along, to regulate the pendulum for accuracy. At the time, I wasn't sure how to do this or if it were possible at all. Turns out it appears to be relatively straightforward: Kick it harder and it moves faster.

The microcontroller is normally in a low-power sleep mode. When the pendulum swings by, it wakes up, and sends a current pulse through the coil of a fixed time period, say 15 mS. Using a watch crystal to run the microcontroller, a very accurate time base is available. By timing each swing of the pendulum, we can determine if the clock is running fast or slow. Then we adjust the pulse width accordingly. This should keep the clock running with very good accuracy.

I’m actually well into the implementation of this design, and hope to have it working very shortly. I think the wait will be worth it. A drawback of many of these wooden gear clocks is that they can be finicky to get running accurately. This approach should make it very easy to get an accurate clock without much if any manual tuning. I have not come across another example of use of a microcontroller to regulate pendulum period like this – I think this is a unique approach.

[DickB]

I have a clock running for several days now with the new microcontroller drive and it is working very well. No pendulum adjustment is needed; just start it running and the controller compensates. This greatly simplifies tuning - there isn't any, really - and makes for an accurate clock - which is the whole point of a clock in the first place.

This breadboard circuit is actually running the clock. The cable at left was connected to a programming/debug device and is no longer being used. The actual circuit is not that complex. I have circuit boards on order.

[DickB]

Even though I have been working almost every day on some part of this project, it seems like it is taking forever to reach the finish line. Patience is a virtue which I don't easily practice.

I've built three clocks now and have run them for several weeks, making observations and minor tweaks to the design to make it easier to build a reliable clock. You should be able to spot one change by comparing the clocks in the photo. I got lucky with the first application of the microcontroller drive circuit, in that no pendulum adjustment was needed. The pendulum just happened to be set about where it needed to be. So actually, in most cases some adjustment is necessary. But I've added an LED indicator that helps quickly set the pendulum in the ballpark, and the circuit then does a good job of keeping time and letting you know if further adjustment is needed. I don't think the batteries will last a full year as I initially expected - I think 6 months is going to be the best I can do with this clock.

I've made numerous refinements to the software. The latest lets you determine if the magnet and coil are of the correct polarity, which is important for proper operation.

Getting close now.

[gregsolano]

Still looking forward to making one!

[CarverJerry]

You and me both Greg.

[jaustin]

I also can't wait to make one.

[dltccf]

still watching this thread with interest. I will be early in line to get the kit when you make it available.

[b.sumner47]

I would like one as well. Great Work.


Capt Barry