[time-nuts] accurate portable time source (Ronald Held)

Tom Van Baak tvb at LeapSecond.com
Fri Dec 28 18:16:29 EST 2007

> To Tom:
>   I will look at those products. Isn't there a problem if the
> crystals are not preaged, for the accuracies I am looking for?  Since
> I am circuit impaired, besides the power supply, what other inputs are
> needed and what is the output(1 PPS, a Frequency)?
>                                                       Ronald

As with car's position, velocity, and acceleration, at elapsed
time, t, a clock will show an time error of:
    T0 + F0 * t + 1/2 FD * t^2
where T0 is the initial time setting inaccuracy, F0 is the initial
frequency error, and FD is the [constant] frequency drift rate.

So a 10811-class oscillator with a frequency drift rate (ageing)
of 5e-10/day will have a net time error of 20 ms at the end of
a month, or about 3 seconds at the end of a year. In practice
it is highly unlikely that the drift rate will be constant, be that
large to begin with, or even always have the same sign so the
actual time error would likely be much less. But consider this
a worst case calculation.

it is not unusual to find well-aged laboratory quality OCXO on
the surplus market with ageing rates near 1e-11/day. Put that
into the equation and you get just 400 us at the end of a month
and 60 ms at the end of a year!

Working backwards, if you assume no initial time or frequency
offset and a constant drift rate and want a time error of less
than one second at the end of a year, the drift rate needs to be
under 2e-10/day. Someone should check my math on all this.

Correct, pre-ageing is always a good idea.

Not sure what to suggest about being circuit impaired. Perhaps
this project of yours is a good excuse to learn?

(p.s. please trim stray digest messages in your replies to the list)


More information about the time-nuts mailing list