[time-nuts] Measuring the accurcy of a wrist watch
Tom Van Baak
tvb at LeapSecond.com
Thu Apr 17 03:16:28 EDT 2014
> I have a Timex watch that's probably seven or eight years old. It has an
> LCD readout. The buttons haven't worked in years. It looses about one second
> every three or four months. I have to take out the four microscopic screws
> in the back to get into it to set it. The only reason I hang onto it is that
> is the most accurate watch I've ever owned.
> Is there some way to detect its internal frequency inductively? In the
> past I have used old relay coils as inductive pickups with success, but
> never with a watch.
Hang on to that one; that's an unusually accurate watch.
I have been unsuccessful at picking up timing signals from a LCD watch. I did try a webcam once but the timing resolution of an LCD display is rather poor, in the 10 to 100 ms millisecond range. If you look a LCD closely you'll notice how slowly the contrast transitions occur. The acoustic, mechanical, electrical, and magnetic methods offer greater resolution, but none of those work on LCD watches.
I wonder if a fine (sub-mm) optical sensor that tracks contrast changes in a single segment would work. On a 7-segment display, the 'd' segment is the best one to aim for since it changes 8 times every 10 seconds.
If you open it you may see a number of traces from the IC to the LCD with pulse rates tied to the main oscillator. The problem here is that by opening the case and measuring these extremely low current signals you likely introduce a frequency offset.
When you say "set it" do you mean adjust the time or adjust the rate? One idea to reset the time by a few seconds without opening it: raise or lower the temperature by many tens of degrees for some number of hours or days.
More information about the time-nuts