NavList:

Gary:

A fascinating experiment and experimental confirmation of my long-held belief that even the cheapest digital watch makes a superb chronometer.

Now a challenge for anyone wanting to repeat the experiment or extend the results:

1.   Take the "watch board" and simulate motion.   Okay, maybe we can't take it to sea but maybe drive it around town with us?

2.   Vary the temperature.   Maybe put the board outside in direct sunlight (and rain and maybe even snow).

Remember, Harrison's biggest challenge was not making an accurate chronometer (that had been done already) but rather making one that remained accurate despite the motion and temperature changes experienced at sea.

Whoops, as I write this I'm looking at my digital watch on my wrist and watching it bounce around.   Maybe experiment #1 isn't required.   In fact, the very nature of digital watches should make them motion-insensitive (excluding relativistic effects :-P ).

Lu Abel

Gary LaPook wrote:

Based on our discussion, I became curious about the accuracy of digital
watches and their suitability for use as chronometers so I went to my
local TARGET store and purchased three identical watches for $17.00
each, the cheapest that they had. I set them and let them run for a few
days and, as I expected, they each had different rates. Based on this I
labeled them "A", "B", and "C" in the order of their rates starting with
the slowest. I then reset them to UTC at 0121 Z on May 28, 2009. I
checked them against UTC from WWV eleven days later on June 8th and
found that they were all running fast by 2, 4 and 7 seconds respectively
and I worked out their daily rates as .1818, .3636, and .6363 seconds
per day, respectively.

On July 11th, 44 days after starting the test, the watches were fast by
9, 17 and 28 seconds. Using the rates determined in the first 11 days
the predicted errors would have been 8, 16 and 28 amounting to errors in
prediction of 1, 1, and 0 seconds. If using these three watches for a
chronometer we could average the three errors and end up with only a .66
second error in the UTC determined by applying the daily rates to the
three displayed times after 33 days from the last check against WWV
which took place on June 8th.

I determined new rates now based on the longer 44 day period of .2045,
.3864 and .6363 seconds per day, respectively.

On September 15th at 0800 Z (per WWV), 110 days after starting the
test,  I took a photo of the watches which I have attached. The photo
shows the watches fast by 21, 41 and 69 seconds but by carefully
comparing them individually with the ticks from WWV the estimated actual
errors are 21.5, 41.8 and 69.0 seconds. Using the 44 day rates, the
predicted errors are 22.5, 42.5, and 70 seconds giving the errors in the
predictions of 1.0, 0.7 and 1.0 seconds which, if averaged, would have
caused a 0.9 second error in the computed UTC after 66 days from the
last check against WWV on July 11th.

If, instead, I used the 11 day rates then the predicted errors would
have been 20.0, 40.0, and 70.0 seconds which would result in errors of
prediction of -1.5, -1.8, and 1.0 which, if averaged, would cause and
error in the computed UTC of -0.6 seconds after 99 days from the last
check against WWV which would have been on June 8th in this example.

 From this experiment it appears that fifty one dollars worth of cheap
watches would give you a perfectly adequate chronometer.

gl


  

---

--~--~---------~--~----~------------~-------~--~----~
NavList message boards: www.fer3.com/arc
Or post by email to: NavList@fer3.com
To , email NavList-@fer3.com
-~----------~----~----~----~------~----~------~--~---