Re: [chrony-users] why PPS one second ahead NTP?

[Rob Janssen <chrony-users@xxxxxxxxx>]

    This is because NMEA has a high uncertainty of time, it can
    sometimes be off by several hundred ms.
    That can then cause the clock to jump a second, which PPS cannot
    correct because it is modulo-1-second.
    Remember the time in most of the NMEA messages is "time of last
    fix", not "current time".  It can
    be lagging quite a lot.
    Your problem is caused by the "noselect" option on the pool line. 
    You'd better put that on the NMEA refclock instead.
    Aoso you have an offset 0.999 on your NMEA clock, not a good idea to
    do that.  The offset varies and such
    specs can bring you trouble.  Remove it or at least make sure it is
    smaller than ~400ms.
    
    Rob

On 2023-09-08 09:33, Petri Hyvarinen wrote:

Dear list readers,

I have been trying to figure out why PPS gets approx. one second ahead of NTP servers. Sometimes even PPS is ahead of NMEA approx. one second (perhaps depending of the offset set in configuration?).

 

Here is the situation how it ends up

MS Name/IP address         Stratum Poll Reach LastRx Last sample

===============================================================================

#* PPS                           0   4   377    14  +6587ns[+7168ns] +/- 5001ns

#- NMEA                          0   4   377    13   +109ms[ +109ms] +/-  200ms

^? 109-108-30-87.bb.dnainte>     2  10   377   854   +995ms[ +995ms] +/-   48ms

^? ntp3.dnainternet.fi           2  10   377   880  +1000ms[+1000ms] +/- 6362us

^? skg.vaxxi.net                 2  10   377  1031   +997ms[ +997ms] +/-   58ms

^? hell.ruselabs.com             2  10   377    55   +993ms[ +993ms] +/-   19ms

 

This is embedded system with u-blox module onboard where I have not all details available how HW is integrated between host and GNSS receiver module.

 

Below relevant lines (in my understanding) from chrony.conf

pool 0.fi.pool.ntp.org noselect

refclock PPS /dev/pps0 lock NMEA refid PPS precision 5e-6

refclock SHM 0 refid NMEA offset 0.9999 precision 2e-1 #offset 0.15 results approx. 41..44ms last sample

leapsectz right/UTC