· By Buffy
The Moonphase
A 59-tooth disc pretending to be the moon: how moonphase displays work, their real error rate, and the high-precision versions that chase a century of accuracy.
The moonphase display is the most romantic useless thing on a watch: a little disc with two moons peeking through a shaped aperture, tracking the synodic month — new moon, first quarter, full, last quarter — in metal.
The mechanism
The lunar month (new moon to new moon) averages 29.53059 days. The classic display is a disc with 59 teeth and two moons, advanced one tooth every 24 hours by a pin on the 24-hour wheel. Two moons × 29.5 days = 59 days per revolution.
But 2 × 29.5 = 59.0, while the real month is 29.53059 — so the classic moonphase gains about one day every 2.7 years (the error accumulates: 59/2 = 29.5 vs 29.53059). Owner’s reality: reset it whenever you notice; no moonphase displays the actual lunar phase with astronomical accuracy anyway (the moon’s face as seen from Earth varies with latitude and orbit eccentrics).
The precision chase
Watchmakers can’t leave a rounding error alone:
- 135-tooth discs (and similar high-count trains) shrink error to roughly one day per 122 years or better.
- Astronomical moonphases (e.g. Patek Philippe’s with 135 teeth, Arnold & Son’s perpetual moon, Ochs und Jahre’s modular moon) are driven by precise ratios geared off the date train.
- Ulysse Nardin’s moonphase modules track the moon with startling accuracy using differential gearing.
A “one-day-in-122-years” moonphase is, mechanically, a very slow argument between 29.5 and 29.53059 — the same fight the perpetual calendar has with 2100.
Why bother
Because it is the only complication that points out of the watch. Everything else on the dial is a human convention — seconds, dates, zones. The moonphase tracks an actual object in the sky, imperfectly, forever. Set it to the last new moon and it will be slightly wrong for decades while still telling a truth about the machine inside.
See What Counts as a Complication for the family it belongs to.