[time-nuts] Leap Second quirk, Another hanging bridge

Chris Kuethe chris.kuethe at gmail.com
Tue Aug 12 15:03:47 EDT 2008


On Tue, Aug 12, 2008 at 11:36 AM, Hal Murray <hmurray at megapathdsl.net> wrote:
>
> I've got a couple of GPS units that use the SiRF chips feeding NTP.  I was
> looking for low cost units for time keeping.  They don't work very well.  The
> time offset of the NMEA message wanders/jitters by about 100 ms.  I can
> easily correct for a constant offset, but I can't dance around an offset that
> won't hold still.  (Oh, well.  I tried.)

Yep, that's to be expected, given the iterative solver they're using.
Try turning on the ZDA message, apparently the $ sign is aligned to
the start of the second ... if its implemented in your version of
firmware. Since you're running a BU-353 you could have all kinds of
broken things in your firmware (not impressed by Globalsat).

> But I still have a couple of them collecting data.

As an aside, how and what are you collecting? I'm starting to think
about setting up different loggers to catch the leap second...

> They seem to get confused by leap seconds.  At least that's the only thing I
> can think of that changed recently.  Here is a graph:
>  http://www.megapathdsl.net/~hmurray/ntp/leap-gps3.gif
>
> It started, midnight, July 31.  Does anybody know when the leap second
> announcement hit the satellites?

first notice of this in the archives is on the 28th... so probably
0000h UTC on the 29th.

> I assume it's a software bug.  Looks like it repeats on a weekly pattern.
>
> The red line on the top is a sane unit used as a reference to show that the
> time on the local system time isn't bouncing around.

I've used the LVC w/ PPS as the reference clock... not sure that I
trust the delay/jitter characteristics of USB enough to give me better
time than a wrist-watch.

> Here is a longer time span that shows a hanging bridge type pattern on the
> offset:
>  http://www.megapathdsl.net/~hmurray/ntp/leap-gps2.gif
> If you were (un)lucky, you could get fooled into thinking that the offset was
> reasonably stable.
>
> This graph includes a few more units:
>  http://www.megapathdsl.net/~hmurray/ntp/leap-gps.gif
> The spikes on the blue Garmin GPS 18 LVC usually happen when it is recovering
> from not-enough satellites.  (I haven't checked these particular samples.
> That's what I found the last time I investigated.)

it's  "nice" to see all 3 SiRF receivers failing in the same way. that
does make me think it's firmware rather than busted hardware.


-- 
GDB has a 'break' feature; why doesn't it have 'fix' too?



More information about the time-nuts mailing list