On 05/06/2022 06:59, jdow wrote:

The delay is not as important as the two way path difference. The delay is
automatically canceled out based on the calculated one way delay. So it should
be single digit ms spot on. That's the point of my quibble. A known or
calculated delay can be removed.

Yes, it's the difference between send (20 Mbps) and receive (200 Mbps) and that
asymmetry which introduces an unknown path delay difference which isn't
accounted for, and hence the offset I noted. As there will be a difference,
all that NTP does is to use the average.

For low interrupt latency from the serial port control lines - the source of
the PPS signal - even the Raspberry Pi 1 is significantly better under Linux
than a typical Windows PC, at least using standard device drivers.

Cheers,
David
--
SatSignal Software - Quality software for you
Web: https://www.satsignal.eu
Email: david-taylor@...
Twitter: @gm8arv