Re: [chrony-users] Accurately measuring clock drift

[Rob Janssen <chrony-users@xxxxxxxxx>]

    Joe,
    
    Well, it is not merely a timeserver that I am after...  those are
    easy to find (also as surplus).
    And here at home I have a LeoNTP.  Nice timeserver, unfortunately
    IPv4 only despite lots of promises.
    
    We do operate a number of transmitter stations in a simulcast
    repeater network.
    For that to work, the time on the server itself must be accurate to
    a few microseconds.
    The original poster of this topic (that I sort of hi-jacked) seems
    to be doing a similar thing.
    
    For time on the server to be that accurate, sync to another server
    using NTP, even over the local
    network, is not good enough.  That is why we sync to 1PPS pulses
    from a suitable GPSDO that
    also outputs the 10 MHz reference frequency for the transmitter.
    
    We use the "1PPS to RS232 DCD input" method, with "ldattach" to make
    the serial port into
    a /dev/pps0 and lock chrony to that.  That works reasonably well,
    but with many GPSDO units
    that we use we need to stretch the pulse, and also the pulses are
    usually filtered on the
    RS232 board so I think it would be better to have a dedicated board
    for this purpose that
    solves all issues.  Maybe an Intel I210 card is closer to what we
    need than an RS232 port.
    (still requiring the pulse stretcher / level conversion)
    
    Rob

On 6/21/23 14:22, Joe Smith wrote:

It's been a while since I've messed with this but in a past life I created a dedicated time server using a BeagleBone/BeagleBoard SBC with a Adafruit GPS Breakout board. Runs Ubuntu and and other Linux flavors. The BeagleBone is about the size of a credit card. Put a small enclosure around it and voila! a compact time server running chrony. You can ssh into it like any other network connected Linux host to configure or maintain it. We were able to get very good time sync from this.

In one instance, my prototype was modified by someone else to use a different GPS board that had a very narrow pulse, too narrow for the BeagleBone to detect. Pulse width on the board was adjustable via cmd-line over serial cable so we just adjusted it to the same width as the Adafriut (1ms I think?).

Perhaps something along these lines might work for you. BeagleBone is about $50 and a bit more for breakout board, cables, etc. Probably around $150 or so total parts investment. I've seen people do it with Raspberry PI as well so that may be an option for you. Takes a bit of time and patience to put together but a very affordable solution. I don't have notes on the setup anymore but there was an abundance of info online and the chrony-users group is extremely helpful getting past some of the rough spots.

Hope this helps