Puzzling clock offset spikes

Hal Murray hmurray at megapathdsl.net
Sat Jun 30 20:25:12 UTC 2018

> refclock shm unit 1 refid PPS minpoll 0 maxpoll 0 prefer
> I think I'm good above in the ntp.conf, right?

I think 0 is a no-op so you get the default.  What does ntpq -p say?

> Hmm - wasn't even familiar with logconfig until I looked it up just now.  I
> think I'm okay, the manpage says that everything is on by default.

I thought the default was off.  Do you have lots of stuff in your log file?  
Things like:

27 Jun 12:49:11 ntpd[619]: PROTO: 8024 84 reachable
27 Jun 12:55:41 ntpd[619]: PROTO: d41a 8a sys_peer
28 Jun 21:03:31 ntpd[619]: PROTO: 8014 84 reachable
29 Jun 05:23:00 ntpd[619]: PROTO: 931d 8d popcorn 0.000063 s
29 Jun 12:29:08 ntpd[619]: PROTO: 901d 8d popcorn 0.000117 s

> Another area of confusion for me here. The datasheet for the MT3339
> specifically mentions 'reference oscillator' - 

That's internal to the GPS unit.  The system can keep time without any help 
from GPS.  (It may not be as good as with GPS.)  The usual approach is to use 
some form of cycle counter which is driven off the main CPU clock which is 
derived via a PLL from an external clock/crystal.

Old systems (pre cycle counter) used the scheduler interrupt for timekeeping.  
There was an intermediate stage where the cycle counter (if available) was 
used to interpolate between scheduler ticks.

On PCs, there is a calibration step at boot time to figure out the cycle 
counter frequency.  I'm not sure about the details on a Pi.  There has to be 
either a calibration or a compiled in value.

These are my opinions.  I hate spam.

More information about the users mailing list