NTS Time Server ntp.bollar.com

Martin Boissonneault dev at ve2mrx.dyndns.info
Sat Mar 21 00:22:37 UTC 2020


Hi Rick,

I noticed that, except from the NTS part, our setups look similar (Pi 
3B+, Buster, NTPsec 1.1.8, u-Blox MAX-M8Q GNSS, you probably use GPSd 
3.20?).

I also noticed that you don't have quite as good precision and jitter as 
I do (but close), and that you seem to use multiple GNSS systems. I 
offer below many elements that helped me get my Pi to it's best, use 
what you need. I very recently went through all my notes and validated 
them to the current ways as there is a lot of obsolete stuff out there. 
This is from my refreshed notes for Buster.

Extract from my ntp.conf:

# kernel PPS
refclock pps unit 0 	refid kPPS flag2 0 flag3 0 minpoll 4 prefer 	# Best is minpoll 4!

# GPS PPS reference - PPS shm from GPSd - the serial shm is redundant as the PPS shm has all the info
refclock shm unit 1 	refid gPPS minpoll 4 	# Best is minpoll 4!

# *** NOTE: by default, the ARP cache flushes after 60 seconds. If so, use maxpoll 5 ***
# *** I modified the ARP flush on rPi to 65sec, so maxpoll 6  (64s) works ***
# sudo cat /etc/sysctl.d/ntp_arp.conf < net.ipv4.neigh.default.gc_stale_time = 65

# Take note: using iburst will make the ntpd first lock to the pool, instead of the PPS.

# LAN servers: <example>
server 192.168.10.11 	minpoll 4 maxpoll 6 		# !!! On LAN !!! # For ARP reasons, set maxpoll 6 -> 64s

# I use official government NTP servers to validate the quality of the pool
# Less an issue with a reliable GNSS timesource, but helps if you have an antenna problem and GNSS becomes unsynchronized
# Official Canadian gov NTP servers (Stratum 2)
server time.nrc.ca 	maxpoll 6 	# For ARP reasons, set maxpoll 6 -> 64s
server time.chu.nrc.ca 	maxpoll 6 	# For ARP reasons, set maxpoll 6 -> 64s

# Pool: use a geographically near pool - I set maxpoll to 10, but it is usually set to 6 by NTPsec
pool <local pool>.pool.ntp.org 	maxpoll 10

Also, if not already done, you should configure the Max-M8Q for time 
service. GPSd settings are too general purpose:

  * UBX-CFG-PRT set baudrate to 115200 but GPSd will need to be told
    about it in it's startup script, ubxtool can set the uBlox GNSS,
  * UBX-CFG-PRT set protocol out to 0-UBX, ubxtool can set the uBlox GNSS,
  * UBX-CFG-GNSS enable only ONE GNSS system (GPS or Galileo or GLONASS
    or BeiDou),
  * UBX-CFG-SBAS select the SBAS system covering your zone else disable,
    disable all SBAS features except integrity (activate it),
  * UBX-CFG-PMS set power to 0-Full power or
  * UBX-CFG-RXM set power to 0-Continous mode
  * UBX-CFG-NAV5 set dynamic model to 2-Stationary, Fix mode to 1-2D,
  * UBX-CFG-TMODE2 set Time mode to 2-Fixed mode, enter surveyed (fairly
    accurate) coordinates,
  * UBX-CFG-TP5 set antenna cable delay (from cable length and velocity,
    google it)
  * Last, UBX-CFG-CFG Save configuration to nvram

That should optimize your Max-M8Q for stationary time service. Sorry, 
I'm too lazy to write some script using ubxtool to do it, so I use 
uBlox's u-Center on Windows, save, and disconnect the RxD line of the 
M8Q to prevent GPSd from re-configuring it. I think there is a GPSd 
toggle (for "Broken" receivers, -u?) that prevents GPSd from writing to 
the GPS, but my home-made interface card has a jumper I remove.

Also, check if you need to disable the systemd time setting service:

# https://www.freedesktop.org/software/systemd/man/systemd-timesyncd.service.html
sudo timedatectl set-ntp false

 From your graphs, it looks like it's OK, but it is an SNTP time setting 
service that competes with NTPd and messes with precision. It shows up 
as a few time jumps per hour on ntpviz graphs.

If you want my scripts/notes/miscellaneous files, just ask. I'm open to 
corrections too, as I said, there is plenty of outdated information out 
there and I want to get rid of mine.

Have a nice day!
Martin

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.ntpsec.org/pipermail/users/attachments/20200320/df2c27ea/attachment.htm>


More information about the users mailing list