dataReal_to_ram_pingpong with the redpitaya/double_iq_pid_vco example?

[aewallin]

For gap-free logging I am trying the dataReal_to_ram_pingpong IP. It seems to work OK for me in the dataReal2ram_pingpong example with the sin.dat example (16384 samples) datafile provided. In gnuradio the plots are 3*16384 samples long, so we see two discontinuities as expected in the top plot, and no discontinuities in the lower plot:

Next I want to use the dataReal_to_ram_pingpong in the double_iq_pid_vco example for streaming the phase-data over ZMQ to a host computer.
This seems to work, but I get some kind of glitch for the first two values (out of the 8192 samples being read per ZMQ send() ). This screenshot shows a gnuradio plot and also the raw samples, where diff() applied to the samples clearly shows that the two first samples contain anomalous data.

• any ideas why the two first samples out of 8192 transmitted per ZMQ send() are different?
• the double_iq_pid_vco example averages (NB_ACCUM=8192) samples using meanReal - would it be useful to instead test with a decimating FIR-filter?
• is there a good solution for timing of the python-script that reads from /dev/dataReal_to_ram_slow and sends the samples over ZMQ? the examples have a time.sleep(0.05) in the while True loop, but does this ensure that no data is sent twice and that there are no gaps?