Re: [chrony-users] 'fallbackdrift' and temporarily connected ref-clocks

[simon@xxxxxxxxxxxxx]

I might still be confused, it seems that there may be a mode of 
operation between 'tracking' and 'fallback'.
--
tracking --loss_of_clock--> ????? --drift_time_out--> freewheel
--
I do understand that I can stay longer there (in ?????), but what's 
actually happening, and why is it better if there are no other clocks to 
correct the system?


You asked about requirements:
For DIY LTC I would want a precision of < +-100us (when tracking clock) 
and a drift of < 1frame (+-15ms) in 8hrs


I did a longer test run and uploaded some zip-file here:
https://github.com/x42/ltc-tools/issues/17#issuecomment-662027246

With 6hr of training (previously deleted 'chrony.drift') and then 9hr of 
freewheel I achieve around -6ms drift. The drift looks like it could be 
further corrected for.
Setup was a 'DigiDesign SyncIO' generating/feeding LTC into a Dell 
laptop runing libltc and tools.

Cheers,
Simon.

On 2020-07-20 01:11, Miroslav Lichvar wrote:
> On Fri, Jul 17, 2020 at 10:57:44AM -0600, simon@xxxxxxxxxxxxx wrote:
> > I am confused about 'fallbackdrift'; if my understanding is correct 
> > the
> > above will cause the real time drift corrections from 'LTC' will be 
> > dropped
> > between 16s and 32s.
> >
> > Q. Is this understanding correct?
>
> The average frequency offset will be applied at 16s and 32s after the
> last update of the clock. You would probably want to use a much longer
> interval, one that covers daily temperature changes, etc.
>
> > Q. Is there a better way to handled a disappearing ref-clock?
>
> For a start, I'd recommend to disable the fallbackdrift, enable the
> tracking log, and let it run for a day with the refclock to see how
> the frequency is changing over time.
>
> > Q. Is there a (better?) way to confirm drift compensation source?
>
> The tracking log will show you the corrections as they are applied.
>
> > Some (considerable) time later, I can see that local time has drifted 
> > off.
>
> What are your requirements? Don't expect miracles with an ordinary
> computer clock.
>
> --
> Miroslav Lichvar

--
To unsubscribe email chrony-users-request@xxxxxxxxxxxxxxxxxxxx 
with "unsubscribe" in the subject.
For help email chrony-users-request@xxxxxxxxxxxxxxxxxxxx 
with "help" in the subject.
Trouble?  Email listmaster@xxxxxxxxxxxxxxxxxxxx.