[LinuxPPS] Why is my ATOM clock a falseticker?

[Cirilo Bernardo]

On 2/2/09, Hal V. Engel <hvengel at astound.net> wrote:
> On Sunday 01 February 2009 05:03:17 pm Cirilo Bernardo wrote:
>  > On 2/2/09, Paul Simons <paul at thesimonet.org> wrote:
>  > > Surprizingly, it is CONFIG_HZ=250.  Is that the default?  Should I change
>  > > it to 100?
>  > >
>  > > On 01/31/2009, "Cirilo Bernardo" stated:
>  >
>  > [snip]
>  >
>  > >  > What is your setting for CONFIG_HZ in the kernel options?
>  > >  >
>  > >  > - Cirilo
>  >
>  > There was some discussion a few months ago about the Linux CONFIG_HZ
>  > and problems associated with it.  I can't recall the specific issues
>  > and can't find the discussions in my email.  You might try rebuilding
>  > with CONFIG_HZ_100=y and see if that fixes the problem; if it does
>  > then we definitely have a kernel problem.
>  >
>  > - Cirilo
>
>
> My experience is that 100Hz works better but the OPs clock was off by amost
>  0.5 seconds and the difference between 100Hz and 250Hz was on the order of
>  perhaps 50 microseconds for me.  So I don't think this is the issue with the
>  large offsets.
>
>
>  Hal
>
>

The offset for the PPS is bizarre.  Wasn't there a kernel+NTPD NANO
issue?  All my kernels are >1yr old and I haven't played with the
kernel NANO timer option yet.

Just to make sure that PPS is not missing ticks, (re)start ntpd, check
the event counter along with time,  and check the event counter and
time again the next day.  Check that the number of events elapsed is
consistent with the recorded times.

- Cirilo