Re: [chrony-users] Commercial unit running chrony, benefits of 'xleave' ?

[Matthew Eshleman <matthew@xxxxxxxxxxxxxxxxxxxxxxxx>]

FYI. I received a LeoNTP today. Much better results, thank you for the suggestion. 

Results from the point of view of the target embedded Linux device:

# chronyc sources -v
MS Name/IP address         Stratum Poll Reach LastRx Last sample              
===============================================================================
^* LeoNTP-33f                    1   1   377     0  +3296ns[+5960ns] +/-   55us
^- TM2000B                       1   1   377     1    -13us[  -11us] +/- 1466us

# chronyc ntpdata  
Remote address  : 192.168.201.56 (C0A8C938)
Remote port     : 123
Local address   : 192.168.201.211 (C0A8C9D3)
Leap status     : Normal
Version         : 4
Mode            : Server
Stratum         : 1
Poll interval   : 6 (64 seconds)
Precision       : -25 (0.000000030 seconds)
Root delay      : 0.000000 seconds
Root dispersion : 0.000000 seconds
Reference ID    : 47505300 (GPS)
Reference time  : Mon Feb 07 21:43:44 2022
Offset          : -0.000003241 seconds
Peer delay      : 0.000108612 seconds
Peer dispersion : 0.000001047 seconds
Response time   : 0.000004814 seconds
Jitter asymmetry: +0.50
NTP tests       : 111 111 1111
Interleaved     : No
Authenticated   : No
TX timestamping : Kernel
RX timestamping : Kernel
Total TX        : 1188
Total RX        : 1168
Total valid RX  : 1168

Thank you very much for the suggestion. This should certainly get us moving in the right direction.

Thank you again,

Matthew

On Thu, Feb 3, 2022 at 9:03 AM Matthew Eshleman <matthew@xxxxxxxxxxxxxxxxxxxxxxxx> wrote:

Thank you very much, 

The ntpdata for this TM2000B server, as reported from our target embedded linux device:

> chronyc add server 192.168.201.20 minpoll 0 maxpoll 1 xleave trust prefer iburst

> ntpdata:

Remote address  : 192.168.201.20 (C0A8C914)
Remote port     : 123
Local address   : 192.168.201.211 (C0A8C9D3)
Leap status     : Normal
Version         : 4
Mode            : Server
Stratum         : 1
Poll interval   : 1 (2 seconds)
Precision       : -10 (0.000976562 seconds)
Root delay      : 0.000000 seconds
Root dispersion : 0.000000 seconds
Reference ID    : 47505300 (GPS)
Reference time  : Thu Feb 03 14:41:21 2022
Offset          : -0.000002094 seconds
Peer delay      : 0.000977595 seconds
Peer dispersion : 0.000977595 seconds
Response time   : 0..000320911 seconds
Jitter asymmetry: +0.00
NTP tests       : 111 111 1111
Interleaved     : No
Authenticated   : No
TX timestamping : Kernel
RX timestamping : Kernel
Total TX        : 3478
Total RX        : 3478
Total valid RX  : 3478

This server is probably good enough for initial use cases, but long term use cases likely require less error.

That LeoNTP looks interesting/good. I'll probably grab it too. Thank you!

Thank you very much,

Matthew

On Thu, Feb 3, 2022 at 8:04 AM Miroslav Lichvar <mlichvar@xxxxxxxxxx> wrote:

On Thu, Feb 03, 2022 at 07:09:27AM -0600, Matthew Eshleman wrote:
> For example, I acquired:
> https://timemachinescorp.com/product/gps-ntpptp-network-time-server-tm2000/
>
> which works fine so far, however on a local LAN we are showing about +/-
> 1.5 ms of estimated error.

What does chronyc ntpdata print for the server?

> Fundamentally, I'm curious how much benefit we would see if the NTP server
> supports the "xleave" option, and then I'm curious if there is a known NTP
> server, preferably small and portable, that uses chrony.

I don't know any. It is a business opportunity nobody tried to make
money on yet. :)

An appliance that I know works well is LeoNTP sold by Uputronics.
It doesn't use hardware timestamps, but it has an embedded NTP
implementation with timestamps accurate to few hundred nanoseconds,
without relying on the xleave mode. The main I issue I have with it is
that the firmware is not opensource, so you can't easily fix bugs.
There are few mostly minor issues I reported years ago, which are
still not fixed.

If you would prefer something DIY running chrony, a good option would
be the PC engines apus. They have onboard I211 NICs which support
hardware timestamping and a serial port, to which could be connected a
GPS receiver. There are pads on the board for the I211 SDPs which can
be used for highly-accurate PPS timestamping, but soldering wires to
those tiny pads might not be for everyone. See this great post from
Dan Drown:

https://blog.dan.drown.org/apu2-ntp-server-2/

A promising option seems to the Raspberry Pi CM4, which supports
hardware timestamping, but unfortunately there is no support in the
driver yet:

https://github.com/raspberrypi/linux/issues/4151

--
Miroslav Lichvar


--
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.