[time-nuts] What is UTC after all?

[Tom Van Baak]

> Looking at the Thunderbolt manual, I came across this:
>
> <quote>
> It is important to remember that any real-time UTC is actually a
> prediction of UTC.
> The official UTC time is published approximately one month after the fact.
> </quote>
>
> It makes sense that real-time "UTC" is a prediction, since UTC is the
> average of a number of clocks around the world, we can't know ahead of
> time what they are going to do, we can only guess.

Didier,

Remember it's all a matter of statistics. If you want 1 us
or even 100 ns accuracy, it's much closer to "certainty"
than a "guess".

Any two clocks will drift apart, so time is always a sort
of a guess. The key point is not if clocks agree, but how
close they will be over some elapsed interval. Hence, the
use of statistics like ADEV, or MTIE, etc.

In a sense, every clock is a "prediction" and all of timing
is a guess. The past and present phase and rate combine
to create the future time, to some level of accuracy, with
some probability.

The measure of how well the past behavior of a clock is a
prediction of future time is exactly what the Allan deviation
measures.

> However, it is amazing to me that it takes a month to have the official
> time, which I guess is published in the form of an error from the
> predicted number.

Like a GPSDO, there is a crossover point where too
frequent a correction is either not good or technically
not even possible (due to inter-lab measurement noise
or clock synthesizer granularity). I'm not sure where
that point it; probably less than one month, to be sure.
But don't think that updating the rate of the worlds
clocks, like, every hour would be a good thing either.

/tvb