From: Mxsmanic on
JohnT writes:

> You would only be able to verify that claim if you measured the time errors
> over 3 million years. Not even you could do that.

You can measure the error over one day and extrapolate.

One second in three million years is an accuracy of
99.999999999998943708%, incidentally.

--
Transpose mxsmanic and gmail to reach me by e-mail.
From: Mxsmanic on
Jim Ley writes:

> How do you update it for leap-seconds?

The radio signals to which it synchronizes cover leap years, leap
seconds, and daylight saving time. No manual adjustment is ever
necessary.

> It's not a function of the phone, it's a function of the network -
> anything that is CDMA based or the majority of the 3G methods (AIUI)
> have clocks synchronised with the base-station, which are GPS sync'd.

Which _can_ be GPS synchronized.

GSM (the standard used everywhere else in the world) provides for time
synchronization, also, but most GSM phones don't use it, and I don't
even know if operators keep their time signals accurate.

NTP is extraordinarily accurate for computers, if you have high-speed
Internet access. It is about 20 times more accurate for short periods
than radio control (with daily sync).

All of these methods are limited over long periods by the accuracy of
cesium atomic clocks--the aforementioned 1 second in three million
years. Cesium fountains will be the next major improvement
(possibly).

--
Transpose mxsmanic and gmail to reach me by e-mail.
From: Mxsmanic on
Dave Frightens Me writes:

> What do you care if it's only 1 second in every year?

I'd have to reset it yearly.

--
Transpose mxsmanic and gmail to reach me by e-mail.
From: Mxsmanic on
Jim Ley writes:

> If they didn't set them accurately (well if they didn't set them all
> to be the same to a reference clock, the only available reference
> clock being the GPS clock system) then their network would stop
> working, accurate clocks are essential for the protocols.

Relatively accurate time measurement is required by the protocols I've
seen, but the absolute time of day need not be as accurately kept.

> Why? For what purpose?

Why do I need to justify it? It costs no more than wristwatches with
much lower accuracy. It has no disadvantages, so why not?

--
Transpose mxsmanic and gmail to reach me by e-mail.
From: Mxsmanic on
Miguel Cruz writes:

> I don't think the service would work without accurate clocks; extremely
> accurate timekeeping is a key element of the system used to select cells
> and hand off calls.

If it works like GSM, it has to be relatively accurate, but not
absolutely accurate with respect to time of day.

> Yes, I have used NTP to keep my computer's clock accurate for at least
> 10 years. Which is why its agreement with my cell phone is so
> confidence-inspiring for me.

I've written a program that beeps like the time signal from WWV. It
soothes me at night to leave it running. All the hands of all the
clocks in my apartment move in lock step, in perfect synchronization
with the beeps. Even my watches do that.

I can think of a practical use for extremely accurate time, by the
way: watching solar eclipses. Totality is often measured in seconds,
and if your watch is off, you miss the show. I used GPS timing for
the last one I observed.

> It's nice to hear, by the way, that Windows has finally caught up in
> this regard.

It required widespread, continuous access to the Internet, which is a
fairly recent phenomenon for home computers, if not office computers.

> You may want that, but I don't see why it's important or why 99.99% of
> people would care.

Since it has no disadvantages, why not use it?

People buy more quartz watches than mechanical watches, and greater
accuracy is part of their motivation.

--
Transpose mxsmanic and gmail to reach me by e-mail.