[ntp:questions] Two time sources with offset => fails to synchronize

Serge Bets serge.bets at NOSPAM.laposte.invalid
Tue Feb 12 15:20:06 UTC 2008

Hello Thierry,

 On Tuesday, February 12, 2008 at 10:39:21 +0100, Thierry MARTIN wrote:

> It seems that the Radio clock has 50ms offset compared to the GPS
> source, which leads a failure in synchronisation.

Radio clocks may need a fudge factor, typically around 20 ms, to report
the right time. That's due mostly to filter delays in the radio
receiver, a little bit to serial port, to interrupt latency, and
2 or 3 ms to waves propagation delay.

I have a DCF77 receiver, driven by Jonathan A. Buzzard's radioclkd 1.0,
and talking to ntpd via shared memory SHM() aka REFCLOCK(28):

| server	maxpoll 6 prefer
| fudge	refid "DCF"
| fudge	time1 0.020

The time1 fudge is the needed thing. Calibrate it at best possible on
the remote machine. Search for "Driver Calibration", "fudge", and
"enable calibrate" in the html docs.

If that's not enough, then check the calibration of the GPS refclock on
your own machine. If that's not yet enough, then you probably have an
assymetric path to the remote machine. Anyway even if all else gets
perfect, you still have the worst number of clocks problem. One is good;
two worst; four best.

Serge point Bets arobase laposte point net

More information about the questions mailing list