I'm getting about 2 degrees standard deviation of phase
measurements of the 19531.25 Hz reference, taking one
measurement per second.  The 2 degrees equates to about 285
nS timing jitter.

The PPS intervals have a mean absolute deviation between about
190 nS and 300 nS with long term average 256 nS.

So the two measurements are commensurate.

For a sanity check on the reference I set it to 39062.5 and
allowed the VLF rx to sniff the signal with good S/N and
measured a standard deviation of phase of about 0.46 degree
on one second measurements, corresponding to 33 nS jitter.

That really points to the sampling clock of the Octo bouncing
around.  The sample clock comes from a 49.152 Mhz crystal
oscillator divided by 512.

By way of confirmation, I replaced the Octo audio interface with
a stereo Audioinjector (also from Flatmax Studios).  Leaving
everything else the same.  With this audio interface, the PPS
is showing between 20nS and 30nS jitter of the pulse centroid.

--
Paul Nicholson
--