A pivoted weight that winds the mainspring from the wearer’s own movement, through a reduction train to the barrel.
A hand-wound watch must be wound daily, and each winding turns the crown, which turns the stem: the largest opening in the case and the main route by which dust, perspiration and water reach the movement. There was also the spring itself. A watch wound each morning runs at its highest torque in the hours afterwards and its lowest before the next winding, so its rate wanders across the day, and a watch forgotten for a day stops.
John Harwood patented the first workable self-winding wristwatch in 1923 and reached production through Fortis in 1926. His weight was a pivoted segment swinging through a restricted arc between two buffer springs, winding in one direction only, and his watch had no crown at all: the hands were set by turning the bezel. Rolex’s Oyster Perpetual of 1931 removed the buffers and let the weight turn through a full circle, which wastes no energy striking anything and harvests small movements of the wrist far better.
Early rotors still wound in one direction, the other simply free-wheeling. Bidirectional winding uses reverser wheels, each a wheel and pinion coupled by hooked pawls that lock one way and slip the other. Two of them sit in the train so that whichever way the rotor turns, one locks and drives while the other slips, and the drive delivered to the barrel is always in the winding direction.
The costs are height and wear. A rotor and its reduction train add roughly two millimetres, which is why automatic chronographs and automatic calendars arrived so much later than their hand-wound equivalents, and the rotor bearing is the hardest-worked bearing in the watch. Because the wearer cannot be stopped from winding indefinitely, the mainspring is not anchored to the barrel wall but held by a sliding bridle that slips when the spring is full.