On Tue, Sep 29, 2020 at 11:29:13AM -0400, Matt Corallo wrote:
Right, ok, so should I presume this isn't a bug normally, but is caused by
whatever nonsense brokenness is going on in my u-blox receivers that they
mysteriously sync to different time despite identical setups?
The problem with refclocks is that there is a missing estimate on
their accuracy, which is needed in the source selection (in terms of
the root distance). Ideally it should be provided by the user for each
clock, but it's not always needed.
It's not ununusal for PPS signals from different GPS receivers to be
slightly off from each other, but I think a larger error usually comes
from the timestamping of the pulse. If there is an interrupt involved
(e.g. pps-ldisc or pps-gpio), it will add a few microseconds of delay,
that can change over time (CPU speed, utilization, etc). If you had an
I210 and tried it with a HW-timestamped PPS, you would see how this
interrupt delay is changing.
What PPS driver do you use? Are your PPS signals configured to trigger
at different times (avoiding two interrupts at the same time)?