I’ve now setup a new station ( https://network.satnogs.org/stations/85/ ), the setup has worked quite well (although I had to install the software “manually” using ansible, instead of using the Pi image, because the 3B+ isn’t supported yet.).
I’m getting this output in my command line: https://screenshot.tbspace.de/reahdpvmtof.png
the O-output seems to be GnuRadio buffer overflows. That of course isn’t too good, right?
Is the Pi 3 lacking the power required to process the AirSpy R2 signal? Is the AirSpy running in 2.5MSPS or 10MSPS mode?
If the Pi isn’t powerful enough: recommendations for hardware with enough power to process the R2’s signal?
The default configuration of gr-satnogs is to run the airspy at 10MSps. You can adjust this in gr-satnogs/python/hw_settings.py by changing the airspy samp_rate key/value pair in the hw_rx_settings dictionary to 2.5e6. Look for hw_settings.py in /usr, as that’s the installed version (so you don’t have to redo sudo make install.
I am using an up-board with intel atom cpu. It runs both my SatNOGS station at 10 Msps and my Airspy HF+ spyserver at the same time. It is running standard Ubuntu server 16.04 and hasn’t had any downtime since I have set it up.
There are two challenges with the Airspy R0/R2:
USB transfer gets close to max usable bandwidth of USB2
Driver library converts the samples to IQ at full sample rate
I have not yet found an ARM board that could reliably transfer 10 Msps from the Airspy. I suspect it is either an architecture or chip support issue. Enabling bit packing helps at the expense of a little more CPU. I could get this to work on the Odroid XU4, just for a while. At some point the IQ conversion thread got thrown over to one of the slow cores ruining everything. I could adjust the thread affinity to force the IQ conversion to run on a specific core, but that is not very portable.
So for now I will stick to intel-based boards for the Airspy R0 and R2.