[ntp:questions] Re: May ntp be off by a second in case of GPS source?

Thomas Laus lausts at acm.org
Thu Dec 29 12:00:04 UTC 2005

Hash: SHA1

On 2005-12-28, Komka Péter <KomkaP at prolan.hu> wrote:
> (If anyone cares:)
> I guess that NTP's time, calculated from NMEA stamps, may be off by 1 or 2 
> seconds.
> As I know, 1 or 2 seconds have to be added to the stamp's time to obtain 
> current time. Because it took some (not negligable) time to
> - compute the time (in the GPS receiver) (known as "position fix calculation")
> - tranfer the stamp to me.
> (Millisecond part may be separately set via PPS.)
> I had done so calculation in my own software, and checked it against 
> independend time source. (Matched.)
> In the ntp source, so far I have not found such second-ajdustment.
> This is why I wonder if it could give the precious time.
> (I have not tested its time yet.) (No ntp-compiling experience (PC, Win).)
In general, any GPS receiver that it used in timekeeping needs to
provide the PPS signal for use by NTP.  It operates in a manner similar
to the radio voice time announcements:

"At the tone, the time will be xx:xx"

The GPS receiver sends the timestamp message for the next PPS pulse and
the NTP software uses the exact time of the pulse to synchronize the
local clock.  The "position fix calculation" message has nothing in
common with a GPS receiver used for timekeeping.  The NTP NMEA clock
type turns off all position messages and only enables the ones that are
related to timekeeping.


- -- 
Public Keys:
PGP KeyID = 0x5F22FDC1
GnuPG KeyID = 0x620836CF
Version: GnuPG v1.4.2 (FreeBSD)


More information about the questions mailing list