[ntp:questions] best jitter on qnx always = 0.977

Dave Hart hart at ntp.org
Wed May 11 05:30:53 UTC 2011


2011/5/11 Николай Орехов <nowhere at mail.ru>:
> My BSP gives right nanoseconds from ClockCycles since last 1ms interrupt. So
> It's far more precise than 1ms.
> How can I convince ntpd that I have higher presision ?

I would argue you shouldn't try, as ntpd assumes all local timestamps
are no more precise than the system clock.  Instead, I would focus
your effort on porting that "ClockCycles" interpolation between 1ms
timer interrupts into the QNX system clock routines, so that ntpd sees
the same high-precision timestamps from the system as from your PPS
source.

If QNX hacking isn't feasible you could shim interpolation under the
portable ntpd code (replacing calls to adjtime(), gettimeofday(), etc
with versions you supply that work with an interpolated clock).

Cheers,
Dave Hart



More information about the questions mailing list