A C wrote:
On 10/11/2011 13:52, David Woolley wrote:
A C wrote:
Let us assume that the pulse from the GPS receiver is off from the
master UTC clock tick by about 3 ms due to a variety of delays in the
overall system starting from the satellite and working its way down to
ntpd. (No arguments about how I get to that number, we're just
assuming this is the case).
What's your basis for assuming all the error is attributable to the PPS,
rather than the other sources contributing to ntpd's time solution.
The GPS receiver should have solved for the time of flight from the
satellite to better than 100ns for GPS to be able to achieve its stated
accuracy. If the time of flight solution was out by 3ms on just one
satellite, your position error would be about 500km. Even if was out by
the same fixed amount for all satellites, you'd have a position error of
around 30 metres.
Nearly all the GPS offset will be attributable to the last few feet -
most of it within the PC.
I'm asking based on an assumption that this is the delay end to end from
the satellite to ntpd's final acceptance of the PPS tick. This is why
I'm saying not to argue this point of the delay calculation. I want to
know what the expected behavior of ntpd is given the conditions outlined
where a time offset is set into time1 on the PPS refclock. What the
actual value of that is doesn't matter to my question other than to
provide a fixed number for the operation.
I'll gladly discuss the reasoning for a delay later but I first want to
have the operation question answered. Note that this is also only the
PPS refclock, not a NMEA refclock. NMEA is not being used in this case,
only PPS. The actual time is being derived from pool servers.
I've generated a PPS from a xtal and used
"127.127.22.1 fudge time1 0.nnnnnnn" to bring the PPS offset to
a few microseconds for a short period depending on changes in
temperature. The best I achieved was for offset to be within a
few milliseconds for several hours with temperature being
reasonably constant but when temperature was fluctuating several
deg C during the day the offset drifted out by more than a second
in just a few hours.
I had intended using an ovened xtal but gave up on that idea as
too expensive as the xtal still needs the fudge factor or some
method to sync with PPS from GPS.
David
_______________________________________________
questions mailing list
[email protected]
http://lists.ntp.org/listinfo/questions
