Quoting Miroslav Lichvar <mlich...@redhat.com>:
On Wed, Dec 28, 2016 at 05:57:15PM -0600, Dan Drown wrote:
That looks like this (green=Linux system's clock, purple=TCXO's measurement
of Linux system's clock).
https://dan.drown.org/odroid/run7/tcxo-acc-1024.png

I've adjusted the starting offset to be within 50ppb, but the TCXO drifts
off and the difference widens to over 300ppb after 3 days.

Would a high-pass filter help?

I suspect this TCXO drifting is actually a hardware problem, my TCXO is already out of spec at above 3.6ppm[1]. My current theory is I toasted it during assembly a soldering iron set too hot. I'm having a pair of boards assembled by a place that has a proper reflow oven to see if that helps.

[1] https://dan.drown.org/odroid/latest/chan2.png

Here's how tempcomp.log vs tracking.log looks:
https://dan.drown.org/odroid/latest/tempcomp.png

This looks awesome :).

I especially like this[2] graph, comparing frequency stability of tracking.log offsets: odroid (tempcomp enabled) vs BBB (normal). Odroid's initial point at 64s is 2e-9 (2ppb).

[2]https://dan.drown.org/odroid/latest/adev-bbb.png

Unrelated to tempcomp, I'm also trying out the new asymmetry code because
I'm on a recent git version.  I assume the offset value in statistics.log is
corrected for asymmetry while the offset value in measurements.log is not.

Yes.

I've added the two offsets to my graphs:
https://dan.drown.org/odroid/run9/remote-statistics.96.126.122.39.png
https://dan.drown.org/odroid/run9/remote-statistics.216.239.35.0.png
https://dan.drown.org/odroid/run9/remote-statistics.216.239.35.4.png

This looks pretty good to me. At this moment it nicely shows how
updating the minimum measured delay shifts the offset line.

The graphs don't show it anymore, but IIRC when I looked at it the
first time one of the sources had the statistics offset overlapping
the bottom offset-rtt/2 line and I was not sure how would that happen.
It looked like the minimum measured delay was zero or maybe even
negative, or it was shifted by the offset option. If you still have
the old data, I'd be interested to see a separate graph of the delay.

I forgot that measurements.log offset had the opposite sign from statistics.log. Correcting that in the graphs to show the same sign fixed that problem :)

--
To unsubscribe email chrony-dev-requ...@chrony.tuxfamily.org with "unsubscribe" 
in the subject.
For help email chrony-dev-requ...@chrony.tuxfamily.org with "help" in the 
subject.
Trouble?  Email listmas...@chrony.tuxfamily.org.

Reply via email to