Last year, I found a 31 day pendulum wall clock (dissambled) in a box of parts at a swap meet and decided to try and put it together and regulate it with a quartz crystal oscillator. The escapement part that rocks back and forth and drives the pendulum was missing and had to be made from a couple razor blade pieces and heavy copper wire. The razor blade escapement worked well but only allowed the movement to advance as the pendulum swung, and would not sustain the pendulum motion by itself. But this wasn't a problem since the quartz crystal divider circuit provides energy to the pendulum with an electromagnet to keep it swinging with only a 5 second error per day. The 31 day mainspring was included, but I wanted to use a weight and drive wheel in place of the spring.
The clock movement is made in Korea by "Sau Jin LTD" "Dae Woo CO LTD" and has no jewels. It measures 4.5 inches diameter by 1.5 inches deep plus 1.5 inches for the hands and drive wheel shafts, so the clock face and hands are about 3 inches from the wall. The pendulum period is close to 53.4 complete swings per minute. Can't figure out why that particular period was used. I have seen similar movements on ebay.
The clock pendulum was made using a 10 inch, 3/16" wooden dowel with a strong magnet attached to the bottom and a small weight near the top to adjust the period close to 53 beats per minute. The circuit board and electromagnet to drive the pendulum are located on a small shelf (not shown) and positioned so the pendulum magnet swings close to the stationary electromagnet and receives a small pulse on each swing to sustain oscillation. The pulse duration is about 5% of the pendulum period and a LED is used to indicate the pulse output. The clock starts fairly easily by releasing the pendulum near the magnet when the LED flash is observed.
The synchronizing circuit that produces a short magnetic pulse to keep the pendulum in near perfect time was made using a crystal oscillator and binary counters to generate a 60mS pulse at the required rate of 53.4241 PPM. I didn't know the exact rate, but it appeared to be close to 53.5 cycles per minute by just adjusting the pendulum length and monitoring the error over a several hour period. The oscillator circuit uses an old 20KHz quartz crystal, but other low frequency crystals can be used. A standard watch crystal of 32.768 KHz is probably the best and easy to obtain. The idea is allow the counters to count to the desired number and then reset the counters, generate the desired pulse, and repeat the cycle. In this case, I needed a time of 60/53.4241, or 1.1231 seconds. At a frequency of 20KHz, this is about 20000 * 1.1231 or 22462. To detect this number, a multiple input NAND gate (CD4068) is used with inputs connected to the appropriate counter stages. Each counter stage has a value of twice the one before (divide by 2), so the values are 1,2,4,8,16,32 and so forth. So the problem was to select the counter stages that add up to 22462, which works out to:
So, the error is about 2 counts out of 20,000 or maybe 4 seconds per day. Further adjustment can be made by fine tuning the oscillator frequency by adjusting the value of the 1000pF cap at pin 1 of the inverter.
circuit from http://www.bowdenshobbycircuits.info/