I've noticed that the clock rate changes regularly on an hourly cycle. It appeared to be synchronized to the minute hand. I've noticed this on 400 day clocks. The rate varies significantly depending on whether the hand is going up or down. The self winding clock has much more power available but also has a large steel minute hand. I decided to verify that the unbalanced hand was the cause. I logged the clock rate and then attached a temporary counterweight to balance the hand. With the unbalanced hand the rate varies by about 40 microseconds. With the hand balanced the change was about 20 microseconds. The 20 microsecond change is synched to the clock winding. The clock speeds up when the spring is wound and gradually slows until the next winding. I actually found two things that affect the rate. The unbalanced hand and the amount of power available. First is a plot of the clock with the unbalanced hand. Next is the same data without any filter applied. The large spikes are caused by the clock winding. Here is with the balanced hand. The software that I'm using auto-scales the plot to fit the data. Notice that the scale has changed on these plots. And finally the same data without noise filtering. This allows you to see when the clock winds. You can see that the abrupt change in rate occurs when the clock winds.