[time-nuts] Time Interval Counter(?) for high-precision watch measurement

[Tom Van Baak]

List -- There's a very nice review of modern high-precision wrist 
watches here:

https://www.esquire.com/uk/watches/a33819282/high-frequency-accurate-quartz-watches/

Even though time nuts is not a "watch and clock" mailing list, I 
recommend a quick read of the article. The current best result is the 
Citizen 0100, which was announced last year with an accuracy of 1 second 
per year. Obviously these are all quartz watches (not mechanical). 
Several review sites:

https://www.ablogtowatch.com/citizen-caliber-0100-introducing-the-worlds-most-accurate-autonomous-watch/

https://www.ablogtowatch.com/citizen-caliber-0100-worlds-most-accurate-watch-review/

https://www.hodinkee.com/articles/citizen-eco-drive-caliber-0100-introducing

https://www.hodinkee.com/articles/citizen-eco-drive-caliber-0100-review

And the Citizen site itself:

https://www.citizenwatch-global.com/s/news/2019/20190320-1/index.html


Raven -- A rule of thumb is that 1 s per day is about 1e-5 and 1 s per 
year is about 3e-8. To me it looks like your 1e-10 number is off by a 
factor of 300.

Some comments on measurement.

1) For long-term tests, yes, using GPS is the way to go. Note that in 
most cases you need only a GPS/1PPS source; a full GPSDO is not required.

2) Magnetic induction pickup is a popular choice. Note it can be highly 
sensitive to pickup location, angle, distance. When you are making 
millisecond or microsecond measurements you need absolute rigidity in 
your test fixture. There may also be thermal effects. All of this should 
be tested independently so that you don't start a long-term measurement 
only to discover that you merely measuring the instability of your test 
fixture instead of the wristwatch. Spend a lot of time with the 'scope. 
Spend time checking external perturbations on your test fixture. See if 
you need magnetic shielding.

3) It's good that you measured the rise time. 10 to 40 us is reasonable. 
Also check the jitter as that will impact your ability to make watch 
accuracy measurements. Some of these watches use variable pulse dropping 
techniques so a stepper motor (induction) measurement is *not* the same 
as a quartz oscillator measurement. You will see this in phase or 
frequency plots of your data.

4) The SR620 is a laboratory counter. See also the TAPR/TICC. But both 
are nearly a million times overkill for your application. For each tick 
you need only microsecond level precision, not nanosecond, nor 
picosecond. If you are into DIY, you can make your own timer. Other 
alternatives include:

https://www.bmumford.com/mset/

https://mesterhome.com/timer/index.html

http://leapsecond.com/pic/picpet.htm
http://leapsecond.com/pic/pp06.htm

My personal preference for the class of measurements you plan to make 
are USB-powered instruments. I would not measure a precision wristwatch 
with a big, hot, noisy, breezy, vibrating SR620.

5) You probably don't want a frequency or period counter for this 
project. Traditionally a time interval counter is used: you measure from 
your watch tick to a GPS tick. All you need to know is derived from that 
time series. You may find, no matter how expensive your time interval 
counter, that as the watch tick collides with the GPS tick your sample 
rate becomes unstable. The solution to this is to use a timestamping 
counter instead of a time interval counter. Both the TAPR/TICC and the 
picPET are based on this approach. I time all my mechanical and quartz 
clocks & watches using picPET's.

6) Never start a long-term measurement until you have analyzed an hour 
or day of your data. An ADEV plot will help verify if your measurement 
system is measuring your watch or if your watch is measuring your 
measurement system. Learn to use TimeLab. Live phase / frequency / ADEV 
plots of your data are a life saver.

/tvb


On 9/7/2020 3:23 PM, Raven L wrote:
> Hello time nuts and greetings from the 10^-10 world of high precision
> wristwatches.
>
> I'm trying to set up a lab for automated watch measurement. I have a basic
> GPSDO with a PPS and a 10MHz output. I use an inductive sensor to pick up
> the signal from the motor inside the watch.
>
> I've been using a basic digital oscilloscope to measure the interval
> between the PPS and the watch signal.  The signal from the watch has a
> total rise time of about 10us to 40us, varying by watch, and rises a few
> volts above a noise floor of about 50mV. The oscilloscope does a
> serviceable job but doesn't allow automated measurements and can't be
> driven by an external timebase. My goal is to make measurements with a
> precision of about 10us, with a goal of ultimately pinning down a rate to
> better than 0.001 seconds per year (initial testing shows this is what I
> need to resolve the effect of tipover on AT-cut MHz-range quartz crystals).
>
> Are there specific time interval counters or frequency counters that would
> make this easier?  I've been looking at the SR620 as a candidate, which
> appears to be available used for around 2k USD, but wouldn't be surprised
> to learn if there was something that was a meaningfully better fit. Thanks!
> _______________________________________________
> time-nuts mailing list -- time-nuts at lists.febo.com
> To unsubscribe, go to http://lists.febo.com/mailman/listinfo/time-nuts_lists.febo.com
> and follow the instructions there.
>