Yo David!
On Thu, 21 Jan 2021 10:44:02 +0000
David Taylor<gm8arv@yahoo.co.uk> wrote:
My simple NTP data is on my Web site, and the MRTG/Perl scripts I
use. For example:
https://www.satsignal.eu/mrtg/performance_ntp.php
https://www.satsignal.eu/ntp/NTPandMRTG.html
Your "simple" is my "insufficient". No temp data, no jitter, no
comparison data, no statistics, no fix data. Try ntpviz
Yes, I am trying to measure the RTC without disciplining. As
interrupts are not possible I will try polling, recording the time
when the second digit (or value) changes.
Or, just use "hwclock --update-drift" hen look in /etc/adjtime
The first problem with NTP reading the RTC may be that once NTP knows
about it it may try and discipline it, preventing a long-term
measurement,
Which is why I keep repeating to you: "noselect".
and the second that the RTC can only be read in seconds,
so even without the disciplining you might be waiting for many weeks
for the RTC to be a second off!
Yes, patience is required.
On making the RPi alone as accurate as possible, some RPI 4B show a
ppm offset of -9 -11, and -19, so far worse than the RTC.
I'm not exactly sure why yours is more than 10x worse than mine. I'm
guessing you need to let ntpd run longer to discipline the clock.
RGDS
GARY