[ntp:questions] Odd results with Oncore UT+ ref clock
Richard B. Gilbert
rgilbert88 at comcast.net
Tue Feb 1 18:11:17 UTC 2011
On 2/1/2011 12:44 AM, Chris Albertson wrote:
> I just set up a new Linux based stratum 1 server using a Motorola
> Oncore UT ref. clock. Seems to mostly be working except for a
> periodic anomaly Look at this output from ntpq:
>
> chris at atom:~$ ntpq -p
> remote refid st t when poll reach delay offset jitter
> ==============================================================================
> oGPS_ONCORE(0) .GPS. 0 l 6 16 177 0.000 1.401 377.675
> +dp.cx 128.227.205.3 2 u 39 64 377 86.133 -4.062 0.423
> +64.73.32.134 129.6.15.28 2 u 50 64 377 75.212 1.608 0.804
Did you, by any chance, start NTPD a few seconds before issuing the
above command?
>
> The I do nothing but wait and try again and get this
>
> ==============================================================================
> *GPS_ONCORE(0) .GPS. 0 l 64 16 360 0.000 0.816 0.115
> +dp.cx 128.227.205.3 2 u 16 64 377 85.835 -3.523 0.642
> -64.73.32.134 129.6.15.28 2 u 24 64 377 73.538 1.635 0.869
>
>
> Some times the offset and jitter will be as low as 0.004 and 0.002
>
> In the clockstats file it looks OK I see many lines that, I think say
> the GPS sees 5 to 8 satellites with good signal
>
> 55593 20317.217 127.127.30.0 3505527516.999033699 2011 32 5 38 37 36
> rstat 08 dop 0.0 nsat 12,7 traim 1,0,0 sigma 44 neg-sawtooth -30
> sat 88088878
> 55593 20318.217 127.127.30.0 3505527517.999037365 2011 32 5 38 38 37
> rstat 08 dop 0.0 nsat 12,7 traim 1,0,0 sigma 44 neg-sawtooth -41
> sat 88088878
>
>
> What I've hoping is that before I spend many hours on this someone
> else has seen it and knows
>
> =====
> Chris Albertson
> Redondo Beach, California
It looks as if you issued that ntpq -p command a minute or so after
starting NTPD.
NTPD can take up to thirty minutes to get a halfway decent approximation
to the correct time and up to ten hours to achieve the best possible
results. It was meant to be run twenty-four hours a day and seven days
a week.
If you want to run "nine to five" it's a poor choice. If that's what
you want, consider a program called "chrony" which can give you a decent
approximation in a few minutes.
More information about the questions
mailing list