Perpetual calendar

A simple date mechanism advances by one every twenty-four hours and knows nothing of the calendar, so it has to be corrected by hand at the end of every month shorter than thirty-one days: five times a year, and by three days after a common February. For anyone who dated documents for a living the annoyance was constant. Thomas Mudge built a perpetual calendar into a pocket watch in London in 1762.

The memory of the mechanism lives in one component. The programme wheel, usually called the forty-eight-month cam, turns once every four years and carries forty-eight steps around its edge, one for each month of the leap cycle, each cut to a depth matching that month’s length. Once a day a large pivoted lever is released and falls; one arm feels the step presented to it, and the distance it travels decides how far it drives the date ring. Other arms of the same lever index the day star, the month star and the leap-year indication, which is simply a reading taken off the cam itself.

It is fragile by nature. A great deal of stored energy is released in an instant around midnight, and the calendar train is under load for two hours or more either side of the change, during which a correction forced through the pushers can break a lever.

The mechanism fails in 2100 because a four-year cam can only express the Julian rule that every fourth year is a leap year. It has no way of knowing the Gregorian exception that centurial years are common years unless divisible by four hundred. Every perpetual calendar built on a plain forty-eight-month cam will need to be advanced by a day on 1 March 2100, and again in 2200 and 2300, but not in 2400.