[chrony-users] How to plot statistics to show tracking accuracy?

[ Thread Index | Date Index | More chrony.tuxfamily.org/chrony-users Archives ]

Hi, I'm getting very decent results with chrony and a usb gps. According to the statistics the std dev is around 1ms, which I might hope is slightly too high and that as low as 0.5ms is feasible? I think that this might indicate that we can do better measuring the arrival time of the $ in the NMEA sentence, but now I need to collect some evidence for that

So my theory is that gpsd is measuring the timestamp of the end of the ZDA sentence (and working that back to get the initial timestamp) and this gives an additional 0.5-1ms jitter that might not be there if we measured the timestamp of the initial $. If so then I might expect to see a jitter distribution which is positively skewed (every time the last character drops into a later 1ms read)

Can someone give some guidance on the correct statistics to plot to investigate this further? I'm a little unsure on how to interpret the stats in /var/log/statistics.log also /var/log/tracking.log seems interesting?

On a related note, did anyone ever try and improve the time accuracy of USB serial reading by noting that serial jitter is lower than USB jitter and trying to interpolate character arrival time vs USB timestamps to try and figure out a kind of subsample accuracy? ie every time you read from the USB, note the timestamp and how many characters read - given that the usb freq isn't the same as the serial freq we might be able to improve our estimate of the exact time the intial $ character was emitted?


Ed W

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.

Mail converted by MHonArc 2.6.19+ http://listengine.tuxfamily.org/