Escapement Papers ISSUE 01
ISSUE 01the bench publishes numbered issues

Photograph the watch against a reference clock, wait a known interval, and photograph it again. The difference divided by the interval is the rate. It gives rate only, and only to about a second per day.

How to measure watch rate without a timing machine

issue ISSUE 01checked 2026-09-16responsible Escapement Papers (a named bench, not a person)

watchreferenceTWO PHOTOS, ONE INTERVALthe gap between the two lines divided by the days is the rateoffsetdays
the gap between the two lines divided by the days is the rate

What you need A camera or phone; a reference clock you trust; a fixed position; a note of the interval; patience for at least 24 hours.

tape card: what the method can and cannot give

QuantityAvailableMethod
RateYes, about 1 s/dTwo photos over a known interval
AmplitudeNoNeeds a timing machine
Beat errorNoNeeds a timing machine
Position errorYes, roughlyRepeat in each position

Steps

  1. Set the watch to the reference clock and photograph both together. The photo is the record of the starting offset.
  2. Wait at least 24 hours, in one position, undisturbed. Longer is better; a week halves the error in the answer.
  3. Photograph both again, in the same framing, at the same time of day.
  4. Read the offset in each photo and subtract. The difference, divided by the number of days, is the rate.
  5. Repeat in a second position. The difference between the two is the position error, roughly.

What the method cannot tell you

ReadingWhy the camera cannot get itConsequence
AmplitudeIt requires timing a fraction of one beatA tired movement looks fine
Beat errorIt requires resolving two beatsA fork fault is invisible
Short-term stabilityThe interval is too longYou see the mean, not the spread
Rate to better than 1 s/dReading error in the photosSmall errors hide in the noise

Why the mean rate is still worth having

A camera measurement over a week gives a mean rate that is more representative of how the watch is actually worn than a ten-second reading on a machine, because it includes the real positions and the real temperature. It is a worse instrument and a better experiment, and the two answer different questions.

The bench holds that a watch owner who records a camera rate over a week before sending a watch for service will usually get a better conversation with the watchmaker, because the complaint comes with a number attached.

References

  1. Ref 1Time and Frequency Division, NISTsupports the reference timekeeping the comparison depends on
  2. Ref 2Frequency standards and clocks: a tutorial introduction, NBS Technical Note 616, NISTsupports the measurement framework for clock and oscillator stability
  3. Ref 3NIST Time and Frequency Metrology Groupsupports the standards work that reference timekeeping rests on

One link, one document, titled as it titles itself. No home pages, no search results.

Set this page down Set this page down when you have a rate from two photos and a known interval.