Re: GPSd/NTPd timesync issue in JSON mode

[Jürgen Hausladen]

Hi Gary,

thanks for all your information and explanations, this seems to confirm some of the presumptions I made.

> > Not a surprise.  Tey gpsd in the standard configuration.  If that fails
> > then we could work on some patches. 

I started using the JSON mode with NTPd due to an issue we had back a couple of years ago, where it happened, that the system time (sometimes after days or even a week) suddenly started drifting away in shared memory mode even in the range of seconds. However, I haven't experienced this with 3.23.1 so far.

> > 10Hz mode is a lie anyway, little
> > need to use it.

I actually don't need the 10Hz update rate for the time sync, but when using the device also for location based event triggering in a vehicle.

What do you actually mean by 10Hz mode is a lie anyway? The GPS module I am using is at least from its specs capable of running with 10Hz.

Thanks,

Jürgen

Am Mo., 5. Sept. 2022 um 20:02 Uhr schrieb Gary E. Miller <gem@rellim.com>:

Yo Jürgen!

On Mon, 5 Sep 2022 14:10:24 +0200
Jürgen Hausladen <hausladen.j@gmail.com> wrote:

> I have a device that is running with a 1Hz GPS update rate, and NTPd
> configured in JSON mode. The device has a real PPS pulse towards it
> that GPSd makes use of.

Not a recommended mode.

The PPS code in gpsd is much better than that in ntpd.

> However, a problem occurs when the GPS update rate is set to 10Hz.
> Then the time starts to drift away and ends up floating somewhere
> between 100ms-200ms off to the reference devices (Usually the device
> is off around 1-2ms and 15ms at max.).

Not a surprise.  Tey gpsd in the standard configuration.  If that fails
then we could work on some patches.  10Hz mode is a lie anyway, little
need to use it.

> When looking at the respective PPS & TOFF messages in NTPd, one can
> see a strange jump in the real_nsec field of the TOFF message, when
> the time starts to jump to the ~100ms offset.

But since you have ntpd using PPS over JSON, the results will be poor anyway.

> Has anyone experienced this problem so far as well?

Almost no one uses 10H mode.  No one uses it for timing.

> Could this be a potential GPSd issue or might this be related to NTPd?

Yes.

> Running GPSd configured in shared memory mode with 10Hz seems to work
> fine so far but still subject to long-term testing.

Then nothing for us to do, if gpsd works in normal mode then the problem is
in ntpd.

RGDS
GARY
---------------------------------------------------------------------------
Gary E. Miller Rellim 109 NW Wilmington Ave., Suite E, Bend, OR 97703
        gem@rellim.com  Tel:+1 541 382 8588

            Veritas liberabit vos. -- Quid est veritas?
    "If you can't measure it, you can't improve it." - Lord Kelvin