[ntp:questions] Using Trimble TSIP under Linux

Kennedy, Paul P.Kennedy at fugro.com.au
Mon Oct 29 00:06:47 UTC 2012

Hi Dave,
read through this thread and there does not seem to be any reference to
the good work by the folks at:

There are several folks with the pi running as an ntp server. I have had
a unit in soak test over a month now without hitch.

some notes:

1. I did use not gpsd.  Instead I use direct PPS + Serial RS232 off the
GPIO bus.  I used the Sure gps as it has the 3.3v serial output which is
all the GPIO bus will accept. 5V from a Trimble will cook your pi (been
there already)
2. I have a USB GPS as secondary, but the serial has not missed a beat
3. I needed the patched kernel which handles the pps on GPIO pin16 -
this was the most curly part.
4. I had to recompile ntpd so it would display a few more decimal places
of precision (nice).  The biggest step out I have noted is a few
microseconds. Generally it is good +/- 2 microsecs.
5. I made a couple of web pages so I could keep an eye on it (and the
raw inputs) in a browser. http://secondthoughts.no-ip.org/ntp
6. You can see the raw data from both the serial (NMEA) and GPIO (PPS)
at: http://secondthoughts.no-ip.org/ntp/page-ntpqc.php
7. I also expose the live output of ntpq-p and my current ntp.conf file
at that link, so you can see for yourself the stability.

The thing that I really like about the pi is the low power and silence.
All of a sudden, I can run an ntp server, host web pages, and develop
using 2 watts of power.

I heard a rumour the ethernet port on the pi is serviced as part of the
USB port, and has a 1 millisecond interrupt interval.  I tried to read
up on this, and did see something to that effect, but it is still very
vague to me.  So although the pi's clock is good to a couple of
microsoconds (for more than a month) It really needs some testing to see
if the ethernet is the achilles heal.  Not much fun having a very stable
server if it cannot serve very accurately! (that is the important bit of
this mail ;-)


-----Original Message-----
From: questions-bounces+p.kennedy=fugro.com.au at lists.ntp.org
[mailto:questions-bounces+p.kennedy=fugro.com.au at lists.ntp.org] On
Behalf Of unruh
Sent: Monday, 29 October 2012 6:20 AM
To: questions at lists.ntp.org
Subject: Re: [ntp:questions] Using Trimble TSIP under Linux

On 2012-10-28, Richard B. Gilbert <rgilbert88 at comcast.net> wrote:
> Lose the "minpoll 4 maxpoll 4 "!

Well, no, standard advice for refclocks is exactly that. get the time
every 16 sec. 

> It may not be causing your problems but but I doubt that it's helping 
> matters any!
> The NTP software should pick the best clock(s) from those offered!

You know what the best clock is-- the pps intervals. 

> NTPD will adjust the polling intervals to suit the current conditions.

> The short poll interval will correct the large errors quickly.  The 
> longer poll intervals will be used to "fine tune".

Not really. The ntpd poll stretching is used to minimize the network
traffic on ntp servers. That is not a problem here. Longer poll
intervals are better (up to a point) for getting the rate of the local
clock better, but it leaves the local time worse off. With pps you are
better off getting the time right. A local refclock is not the same as a
networked ntp server.


questions mailing list
questions at lists.ntp.org

More information about the questions mailing list