[ntp:questions] Single GPS/PPS time source gets marked as a falseticker

Mike S mikes at flatsurface.com
Wed Jul 13 22:56:18 UTC 2011


At 04:38 PM 7/13/2011, Michael Eder wrote...
>We have looked at our GPS on a scope, the PPS it is dead on and the 
>NMEA
>(just one sentence) is also reliable with about a 680 ms latency and 
>10 ms
>jitter....
>ce000:~/current# ./ntpstatus | tee ntplog.txt
>      remote           refid      st t when poll 
> reach   delay   offset
>jitter
>============================================================================
>==
>xSHM(0)          .NMEA.           0 l   13   16  377    0.000  -74.683
>68.862
>*SHM(1)          .PPS.            0 l   15   16  377    0.000   -0.034
>0.061

...

>ce000:~/current# ./ntpstatus | tee ntplog.txt
>      remote           refid      st t when poll 
> reach   delay   offset
>jitter
>============================================================================
>==
>xSHM(0)          .NMEA.           0 l    1   16  377    0.000  -101.07
>65.086
>xSHM(1)          .PPS.            0 l    4   16  377    0.000   -0.036
>0.061

As I read this, ntp is seeing more like 70 ms of jitter, not 10, and 
sees a -75 to -100 ms offset, so the 680 in the config doesn't seem 
correct.

So maybe something like:

tos mindist 0.080
-and-
fudge 127.127.28.0 time1 0.593 refid NMEA
-or maybe-
fudge 127.127.28.0 time1 0.767 refid NMEA
(Trying to split the 75-100 offset difference to 87 - I can never keep 
the signs right, so use trial and error)

What I've seen with the Thunderbolt GPSDO is that the offset for the 
serial output slowly walks around, maybe +- 7 ms, centered around ~50 
ms offset. If your Garmin does anything similar, even though you saw 
680 ms on a scope, if you came back in an hour and measured again it 
might be different. You may be able to analyze the peerstats file to 
see how it changes on yours.

So, what I did was first widen the mindist, then tweaked the offset for 
the serial time so it was centered around the PPS (which if things are 
locked, shouldn't wander much at all). Finally, I reduced mindist to 
about 15 ms (0.015 - about twice the actual variation I see).

oPPS(0)          .gPPS.           0 
l    7   16  377    0.000   -0.016   0.016
*GPS_PALISADE(0) .GPS.            2 
l    5   16  377    0.000    4.092   0.604

oPPS(0)          .gPPS.           0 
l    3   16  377    0.000   -0.019   0.025
*GPS_PALISADE(0) .GPS.            2 
l    1   16  277    0.000    3.808   0.695





More information about the questions mailing list