[time-nuts] My Casio G-Shock watch and other fun stuff

[Palfreyman, Jim L]

I reported the unusual accuracy of my Casio G-Schock radio controlled
watch (with the radio controlled feature turned off) a month or so back.

Well to back up my rough observations I decided to measure it properly.

First I connected a GPS disciplined 1kHz signal to my frequency counter
in "total" mode. Next I started the stopwatch feature on my watch. Twice
a day (generally morning and night but whenever else I could) for nearly
a month I photographed both the watch and the counter.

Now there where some error in each reading because a) the frequency
counter wasn't updating it's LEDs a thousand times a second and b) the
stopwatch wasn't updating it's LCD a hundred times a second. However
over time these would average out.

What showed up was a sinusoidal pattern with an amplitude of about 100ms
and a period of about 12 days. The straight line of best bet showed a
slope of around 3 seconds per year gain. However this line changes slope
(positive and negative) according to which part of the curve you are on
and so more data is needed to confirm the general trend. One thing I can
confirm is that I set this watch's time 6 weeks ago on the second and at
some points between then and now it is fractionally fast (say 1/10th of
a second) but now there is no visual difference between the UTC pip and
the tick of the second.

The frequency counter method was a relative system - i.e. one power
failure and the whole experiment was dead (although I did work out its
epoch) and so I decided to build a unit to display UTC time accurate to
1 ms. I bought a three digit LED "counter" kit (K1 from
www.ozitronics.com) and supplied it's "count" pulse with my phase locked
1kHz and the "reset" with the 1 PPS. I also changed it's strobing
capacitor so it strobed at 3500 Hz - which means I can capture the time
using 1/1000th of second shutter speed on my camera.

I've restarted the experiment photographing the stopwatch against
absolute time and it's been going a week. Currently the watch is 0.005
of a second slow and still following the sine pattern. I'll just have to
keep running this experiment and get more data.

And I know you cynics out there are thinking the radio controlled
feature is on - I can guarantee you it's not. For a start the watch is
configured to "SYD"ney - so the feature is off AND I have not been able
to receive ANY longwave signal in Hobart from this watch (although I
have had success with another cheaper unit). It has the time of the last
successful "GET" and that was 13 September last year - before I owned
the watch.

Now along these lines I've thought of pulsing a cheap strobe light I
have with UTC seconds (or tenths or hundreths or thousandths?) and then
using my camera with a long exposure to capture events highly
accurately.

Can you people think of any cool experiments that could be done with
this set up? Measuring the velocity of a golf ball just popped into my
mind - but that doesn't require an accurate absolute time. There must be
some cool stuff to do.

Any thoughts?

Regards,


Jim Palfreyman