Re: refclock local

On Tue, Aug 08, 2023 at 14:33:10 +0200, Miroslav Lichvar wrote:
> On Tue, Aug 08, 2023 at 08:15:35AM -0400, Josef 'Jeff' Sipek wrote:
> > I understood this to mean:
> > 
> >  1. some of the sanity check are skipped (e.g., checking phase alignment
> >     with other sources)
> >  2. the source will be used for *frequency* calculations
> >  3. the source will not be used for *phase* calculations
> > 
> > So, based on these, I expected chrony to report some random but very stable
> > offset but still make (partial) use of the clock.
> > 
> > Is my understanding wrong?
> 
> I think it's good.

Ok, good :)

> The chronyc reports in your original post are as expected.
> 
> The reported phase offset is expected to drift slowly between -1 and 1
> seconds with occasional jumps when the maximum is reached.

Out of curiosity, why is the range (-1,+1) instead of [-0.5,+0.5]?  I caught
me off guard when I restarted chrony and the PPS "jumped" from +400 to
-600ms.

> The most useful report would be sourcestats.

sourcestats from same time I captured the sources output yesterday evening:

Name/IP Address            NP  NR  Span  Frequency  Freq Skew  Offset  Std Dev
==============================================================================
PPS0                       64  54  1010     +0.491      0.305   -590ms   201us
kerberos.mit.edu           29  17   98m     +0.035      0.083    +20us   200us
time.cloudflare.com         8   7   51m     +0.016      0.238    +14us   115us
...

> If the stabilization
> works well, the skew of other sources should become smaller and they
> should use more samples (NP).

Noted.  I restarted chrony this morning with the PPS configured but no
signal going in, then after ~4 hours (which I assume is enough for the
system to stabilize), I turned on the PPS signal.  I'm going to give it a
few more hours and then look at the logs to see how the frequency skew of
the sources changed throughout the day - and especially after I enabled the
PPS signal.

> With a refclock on USB (which adds a lot
> of jitter) I'd not expect it to perform very well. You can try longer
> polling intervals and more filtering.

My desktop doesn't have a real serial port and that's why I used a USB
adapter.  This is purely an experiment to make sure I understand how things
work before I make a simple circuit around an OCXO and connect it to a real
serial port on my server.

Thanks,

Jeff.

-- 
To unsubscribe email chrony-users-request-kWFZVVI9zxvPqho9SqqRMmD2FQJk+8+b@public.gmane.org 
with "unsubscribe" in the subject.
For help email chrony-users-request-kWFZVVI9zxvPqho9SqqRMmD2FQJk+8+b@public.gmane.org 
with "help" in the subject.
Trouble?  Email [email protected]