[ntp:questions] SIRF time output wobble, GPS or NTP fault?
Ron Frazier (NTP)
timekeepingntplist at c3energy.com
Tue Mar 6 16:07:09 UTC 2012
I've been seeing some very strange behavior from my USB BU-353 GPS based
on readings from the Meinberg Server Monitor. The unit is based on the
SIRF Star III e/LP chipset. At first, I thought it was the GPS at
fault. But, now, I'm not so sure that it's not either NTP or the
Meinberg monitor program. Let me explain.
Let me say that my GPS consistently provides me with + / - 6 ms offsets
and close agreement with internet servers when everything is working.
That's fine for the moment, with the equipment I have. I have the
system polling the GPS at 8 sec - 64 sec intervals and polling the
internet servers, including some NIST, at 4 min - 2 hour intervals. For
whatever reason, those seem to never go above 4 min.
Days ago, when I set up the system, I set the system fudge factor for
the GPS so that, when the clock is synced to the GPS within a couple of
ms, it was also synced to the NY NIST server within a couple of ms.
Here is a fictitious example of what my offsets might have looked like:
GPS: - 2.3
Internet server A: - 1.5
Internet server B: - 2.7
Internet server C: + 4.2
Internet server D: + 7.1
The point is, I'm locked to the GPS, but am in very close agreement with
the internet servers. I have more internet servers being polled than
shown here. When I set the system up and tweaked the fudge factor, I
had almost all the internet servers showing single digit offsets when I
was locked into the time from my GPS.
Now, a few days later, I might see something like this. I'm still
locked into GPS time.
GPS: + 1.7
Internet server A: - 41.5
Internet server B: - 52.7
Internet server C: - 44.2
Internet server D: - 67.1
Assuming all the internet servers have not drifted backwards 50 - 60 ms,
then it would appear that my GPS has drifted forwards 50 - 60 ms.
Then, a few days later, I might see something like this. I'm still
locked into GPS time.
GPS: - 3.2
Internet server A: + 31.5
Internet server B: + 42.7
Internet server C: + 34.2
Internet server D: + 47.1
Again, assuming all the internet servers haven't drifted, it appears
that the GPS has now drifted about 80 ms backwards.
I used to believe that the GPS was doing exactly that. I'm now no
This morning, I saw numbers such as this on the Meinburg monitor:
GPS: - 2.3
Internet server A: - 991.5
Internet server B: - 982.7
Internet server C: - 974.2
Internet server D: - 967.1
Again, all these numbers are fictitious, but representative. 900 ms
off! Give me a break. Then, the most amazing thing happened. I did
one and only one thing. I restarted NTPD from the Meinburg control panel.
Now, here are the ACTUAL numbers I'm seeing at this very moment.
GPS: + 6.5
Internet server A: - 64.2
Internet server B: - 44.3
Internet server C: - 47.1
Internet server D: - 63.9
At all times when I've looked at these numbers, the PC has been locked
to the GPS time. The point is not that my GPS may now be 50 ms ahead of
the internet servers. The point is that, before restarting NTPD, it
said it was 900 ms ahead of the internet servers; and, seconds later,
after restarting NTPD, it says it's 50 ms ahead of the internet
servers! The GPS output would not have substantially changed during
this time. The variance to the internet servers being reported changed
dramatically, just because of an NTPD restart. Note that my PC clock is
STILL locked into GPS time within a few ms.
There is no way that the time the GPS is actually reporting changed by
850 ms. I now believe that the odd behavior I'm seeing is not the fault
of the GPS, but I think it is the fault of NTP. I believe there is a
bug in the algorithm which compares the internet servers to the
reference clock, or a bug in the Meinberg Server monitor, or in NTPQ, or
something. The only thing that changed in this little experiment, was
the behavior of NTPD when I restarted, not the behavior of the GPS.
(PS - If you email me and don't get a quick response, don't be concerned.
I get about 300 emails per day from alternate energy mailing lists and
such. I don't always see new messages very quickly. If you need a
reply and have not heard from me in 1 - 2 weeks, send your message again.)
timekeepingdude AT c3energy.com
More information about the questions