I did some measurements on the PPS performance of the ublox
LEO-6 GPS

 http://abelian.org/vlf/tmp/170701a.png

I'm using a 29296.875 Hz reference from a GPS disciplined
Trimble OCXO, into two independent M-Audio 192 PCI sound cards,
each sampling about 192k frames/sec.  Sample stream from the
two cards are resampled and timestamped by vttime using the PPS
from a LEO-6 GPS.

The top two traces shows the phase of the 29296.875 reference
from each sound card.  The phase variations are clearly in
step with each other.

That means either the 29296.875 reference is jumping around,
or we are seeing variations of the LEO-6 PPS timing.

The two sound cards have an antenna signal mixed in which
supplies some 60kHz MSF (a very strong signal here).  The
lower two traces show the phase of MSF seen through the two
cards.

The variations of MSF phase agree (allowing for the frequency
ratio 60000/29296.875) with the variations of the reference.

From this I conclude that it is the LEO-6 PPS which is causing
most of the timing/phase variation seen here.   A 1 degree
phase change of the 29296.875 reference corresponds to a timing
shift of about 95nS, which agrees with the approx 80nS mean
absolute difference of the raw PPS intervals reported by vttime.

Maybe the steps occur when the LEO-6 switches to using different
satellites?

I hope to get one of the NEO-7 GPS modules and repeat this
test to see if the newer model has better PPS stability.
At the same time we can check the phase stability of its
frequency output by running it at 29296.875 Hz and comparing
directly with the Trimble reference. I'm wondering if the NEO-7
frequency output is good enough for the local oscillator of
a 137 kHz coherent rx.

--
Paul Nicholson
--