[time-nuts] Z3801A Leapsecond Output

M. Warner Losh imp at bsdimp.com
Sat Dec 31 22:31:10 EST 2005


In message: <43B73B9D.7060402 at febo.com>
            John Ackermann N8UR <jra at febo.com> writes:
: Looks like the Z3801A did just what it was supposed to, other than
: taking an extra second for the leapsecond bit to clear:
: 
: 53735.99993 T22005123123595530+003F
: 53735.99994 T22005123123595630+0040
: 53735.99995 T22005123123595730+0041
: 53735.99997 T22005123123595830+0042
: 53735.99998 T22005123123595930+0043
: 53735.99999 T22005123123596030+003B   <----
: 53735.99999 T22006010100000030+001E
: 53736.00000 T2200601010000013000024
: 53736.00001 T2200601010000023000025
: 53736.00002 T2200601010000033000026
: 53736.00003 T2200601010000043000027
: 53736.00005 T2200601010000053000028

We found in our testing of our leap second code that most GPS
receivers had some inconsistancy in their data.  Some were minor like
the above extra +, while others would keep reporting bogus offsets for
several minutes (until a new almanac was complete).  Others would do a
two step around the leap second, sometimes jumping forward a second or
backwards a second more than they were supposed to.  Most of those
issues were resolved with firmware upgrades from the manufacturer, but
many off by a second or current/future leap second counts are
anomalous near the leap second.  Usually the ones we thought were
stable and relaible, leading to bad effects in code that was ill
preprared to be lied to in the exact manner that it was lied to (but
well prepared to be lied to in other ways)...

We recorded the raw data streams from all the equiptment we had
against the future need to test changes to our software as well as
having multiple different streams of data to use to test rather than
the one 'golden' samples that we obtained when we sent our gear to the
simulators this past fall.

I'll see what of the data actually recorded (I wasn't there today, but
was on call) and what data we're allowed to share with others (we have
some sensitive hardware that has a bunch of strange rules surrounding
it).

Warner




More information about the time-nuts mailing list