[ntp:questions] Re: May ntp be off by a second in case of GPSsource?
bg at lysator.liu.se
Sun Jan 1 00:48:11 UTC 2006
Thomas Laus <lausts at acm.org> writes:
> On 2005-12-30, Bjorn Gabrielsson <bg at lysator.liu.se> wrote:
> > hmurray at suespammers.org (Hal Murray) writes:
> > PPS pulse yes. But still the serial port message is usually sent after
> > the pulse. On some receivers the NMEA message is unpredictable - the
> > jupiter receiver I have is either 1 or 2 seconds late in _NMEA_
> > mode. Binary mode for the jupiter is flawless.
> The same is true for the Motorola Oncore receivers. The usual word of
> advice is to get a GPS receiver optimized for timing if you plan to use
> if for timing and one optimized for positioning if the planned use is
> for finding ones position.
I have not seen any GPS receiver with worse spec on the PPS than
1us. For NTP use 1us is enough. If you are building your own GPSDO or
have your own CS etc, then the best timing recievers that get down
towards 10ns or better could make a difference.
> The binary mode for the NTP reference clock
> receivers will provide much better performance than using the same
> receiver in NMEA mode.
Thats usually true. But it also requires a refclock driver for the
binary mode, which is not always available. With reasonable care from
the manufacturers side, there should be nothing wrong in using a NMEA
mode, when a PPS is also available.
More information about the questions