[ntp:questions] PPS from an external OCXO source. Correcting drift is it possible?

Kat Schwar3Kat at gmail.com
Sun May 3 04:05:15 UTC 2009


Nero Imhard wrote:
   > Since you seem to build stuff yourself, you could also regulate
the
   > temperature of your computer's timing crystal. Should work rather
well.
Great minds think alike :) and this was my first thought too - I tried
this approach, but space constraints and less than ideal PC board
layout caused this experiment to fail miserably with only a slight
improvement over the unregulated xtal.

   > A much more reasonable approach would be to hook up the OCXO to
your
   > computer's hardware clock for better stability.
Unfortunately not as easy as it seems high frequency clocks are
notoriously difficult to construct and calibrate and a new custom
clock or clocks would be needed every time I changed server so I opted
to do something portable.

   > It seems that you are doing things the wrong way round. An OCXO
is not
   > synchronized to anything so it really can't be used as a time
source
   > (which is what a PPS input is supposed to be).
Time synchronisation is unnecessary as synchronisation is derived from
external NTP servers by the NTP algorithms. Only frequency needs to be
disciplined by the PPS.

I have now downloaded and had a look at the source for the type 22 PPS
clock driver (it's much simpler than I imagined).  It does indeed
discipline frequency and it looks as though it may be very easy to add
a fudge time2 for drift that modifies the pre-defined length of the
PPS interval.  (More complex automated drift control may also be
possible).

Now if only I could figure out how to get my modified code to compile
successfully on my PCBSD (FreeBSD with GUI) server.




More information about the questions mailing list