[Discuss-gnuradio] multi-rtl phase synchronization

[jmfriedt]

I am tackling again the coherence issue with multi-RTL dongles for 
interferometric measurements.
I have two R820T2 based DVB-T dongles connected to the same USB bus (USB hub) 
and am feeding
both dongles with a high quality 100-MHz source (-34 dBm output of a 
Rohde&Schwarz SML03 synthesizer,
30 dB gain on the RF frontend of the dongles) -- setup shown at 
http://jmfriedt.sequanux.org/DSC06032.JPG. 

I am trying to reproduce the graph "IQ plot of the z_1/z_2 relative phase 
signal over ~6000 seconds at 
1 Hz sample rate" from 
http://kaira.sgo.fi/2013/09/16-dual-channel-coherent-digital.html
Actually I notice that this graph shows something in units in the tens of 
thousands: what is
"Z1/Z2 relative phase" ? How can I convert this axis to angle in degrees ?

I have manually applied the patch described at 
http://www.rtl-sdr.com/forum/viewtopic.php?f=1&t=531 to
turn off dithering (I have checked that the function is called when starting 
gr-osmosdr -- whether 
dithering is actually disabled is challenging to assess). The resulting phase 
measurement -- recording 
either with multi-rtl source or two osmosdr sources, multiplying one output by 
the complex conjugate 
of the other, complex to phase and saving the resulting phase at 200 Hz rate, 
is at
http://jmfriedt.sequanux.org/allan.pdf (left is time domain plot, right is 
Allan deviation). On the
short term we have sub-degree stability, on the long term several degree drift. 
I have checked that
thermal transfer between the two dongles is of utmost importance: if I cool one 
of the dongles and keep
the other at room temperature, I observe phase rotations of several tens of 
degrees. I believe such
long term phase instability is of little concern for passive radar measurements 
where coherence over
several ms only is needed, but I find conflicting conclusions on long term 
phase stability.

Can other experimenters on this topic comment and assess whether 
http://jmfriedt.sequanux.org/allan.pdf
can be improved, or if is in line with the expected behaviour of these dongles ?

Thanks, Jean-Michel

-- 
JM Friedt, FEMTO-ST Time & Frequency/SENSeOR, 26 rue de l'Epitaphe, 25000 
Besancon, France