.. _time_20170725:

20190725 Ublox Z9P PPS and RTCMv3 corrections
=============================================

For unrelated experiments, I have "borrowed a cup of RTCMv3" corrections
from `GPSnet.dk <http://gpsnet.dk>`_ - one of the "real" RTK networks in
Denmark.

On `time-nuts` the consensus is that RTK corrections should not matter
for a position-locked receivers PPS output, and since I had the chance
I decided to sanity check this consensus.

The setup is as follows:

RTCMv3 corrections from GPSnet.dk to:

UBlox Z9P receiver, (Note: 'P', not 'T') GPS+GLONASS L1+L2

PPS output feeds START input on:

HP5370C (ie: a B with `JKS' improved CPU board <http://www.jks.com/5370/5370.html>`_)

HP5065A feeds STOP input with 5MHz.

The raw data looks like this:

.. image:: 20190725_fig1.png

After unfolding the 5MHz period and removing the HP5065's frequency
offset (-5.3e-11) it looks like this:

.. image:: 20190725_fig2.png

The RTCMv3 corrections were on from 5000 to 10000 and again from
35000 seconds onward.

I have asked UBlox support if the 20 ns shift is intentional
or a bug.  Will update here if I get a reply.

The Modified Allan Deviation indicates a small but not uninteresting
improvement:

.. image:: 20190725_fig3.png