[time-nuts] Hanging bridge question

Poul-Henning Kamp phk at phk.freebsd.dk
Tue Mar 25 07:27:49 EDT 2014

In message <6B362A4D-834A-4733-BED8-FCFEC0CCBA6C at rtty.us>, Bob Camp writes:

I should add here, that you _can_ do a little bit better than the
sawtooth correction.

We know, or at least assume, that the GPS's internal clock is step-less
and slowly changing, so if you put a predictive filter on this stuff,
it can actually do a reasonable job at estimating which way the rounding
of the sawtooth correction went (since it is integral ns).

This reduces the random rounding error on the sawtooth correction
from +/- 0.5 ns to something like +/- 0.3 ns.

Totally not worth it, but a cool and educational project :-)

Poul-Henning Kamp       | UNIX since Zilog Zeus 3.20
phk at FreeBSD.ORG         | TCP/IP since RFC 956
FreeBSD committer       | BSD since 4.3-tahoe    
Never attribute to malice what can adequately be explained by incompetence.

More information about the time-nuts mailing list