[Discuss-gnuradio] Python synchronisation of 2xUSRP N200 with SBX in one UHD container.

[John Shields]

Hi,

    One has O/B GPSDO, the other connected by MIMO cable on Ubuntu  
with UHD_003.010.002.000

    From the /manual/page_sync.html page there is information on how to 
align the CORDICs and LOs of SBX.

    I have modified the python generated by a GRC FG and here are the 
relevant changes I made:


        self.uhd_usrp_source_0.set_clock_source('gpsdo', 0)
        self.uhd_usrp_source_0.set_time_source('gpsdo', 0)
        self.uhd_usrp_source_0.set_clock_source('mimo', 1)
        self.uhd_usrp_source_0.set_time_source('mimo', 1)
        self.uhd_usrp_source_0.set_samp_rate(samp_rate)
self.uhd_usrp_source_0.set_time_unknown_pps(uhd.time_spec())
    # inserted code to align CORDICs ?


    # inserted code to synch LO
self.tune_time=self.uhd_usrp_source_0.get_time_now(0)+uhd.time_spec_t(0.1) 
# tune in 100 msecs from 'now'
        self.uhd_usrp_source_0.set_command_time(self.tune_time)
        self.uhd_usrp_source_0.set_center_freq(freq, 0)
        self.uhd_usrp_source_0.set_center_freq(freq, 1)
        self.uhd_usrp_source_0.clear_command_time()


        self.uhd_usrp_source_0.set_gain(rf_gain, 0)
        self.uhd_usrp_source_0.set_antenna('RX2', 0)

        self.uhd_usrp_source_0.set_gain(rf_gain, 1)
        self.uhd_usrp_source_0.set_antenna('RX2', 1)


While there is a discussion on the web re: the tune_time which seems to 
suggest, I could just put in the, uhd.time_spec_t(0.1), there is a 
comment from Josh saying to make sure the time is in the future so that 
is why I added the time now.


In that same thread there were comments to the effect that stream_cmds 
have not been swig-ed as the control of streaming is part of the 
scheduler so:

questions:

    1) have I got the correct code to synch the LO i.e. by tuning both 
devices at the same exact time?

    2) based on the comment about stream synch being controlled by the 
scheduler, what do I need to do re: synching the CORDICs?

    3) I have noticed that sometimes when I print out the real_seconds 
of time_now, it has a low number (likely the number of seconds the USRPs 
have been running.

    -- Setting references to the internal GPSDO
    --     1) catch time transition at pps edge
    --     2) set times next pps (synchronously)

        I presume that I am reading the time_now before it is synched 
at the next pps where the value of time goes to 1505510864.00000 . If 
the answer to 1) is that my code is correct, how do I make sure that the 
pps edge has occurred before I get_time_now ?

    4) the sync document also implies that the phase offset will drift 
(due to thermal and other characteristics) so phase-coherent 
applications will need re-calibration. is this done by somehow stopping 
the stream(s), doing a set_freq and whatever the answer to 2) above is 
and then restarting. I am trying to collect complex visibilities over a 
long period of time ( i.e. fringe stopping) so wonder if the drift is 
slow enough whether I need to do the "calibration" mentioned in the sync 
doc?


                          Kind Regards,


                                       John