demuxing from hackrf

[parity-error]

I have a (20MHZ ) bit stream of raw IQ data (form HackRf) coming in to my pipeline (its currently demodulating and recording all of the fm channels) and I want to take that same stream and individually shift it (like I do currently for the analog) then decimate and send it to nrsc5 for each digital channel that is present.

I have been playing around with trying this using CSDR (https://github.com/ha7ilm/csdr) something like this:

(in this example the sample is taken with the center at the frequency that I want to test so I don't need to 'slide' (shift_addition_cc) it.
cat 20Mhz.sample.fom.hackrf.iq | csdr convert_u8_f | csdr fir_decimate_cc 92 0.01 HAMMING | csdr convert_f_u8 | ./nrsc5/src/nrsc5 -r - 0

But I don't get any output from nrsc5. This seems theoretically possible but I am not sure if the problem is a slightly different format of the IQ data or if I am not getting it to the exact sampling rate that nrsc5 is expecting. Please advise. Any help on this would be greatly appreciated.

[icb-]

You want a 1488375 sample per second interleaved uint8 stream.

This GNU Radio flow graph generates a playable IQ file from my Pluto SDR.

[argilo]

Correct, the input sample rate needs to be exactly 1488375. A decimation factor of 92 will give you 217391 instead, which is way off. Also, I believe the HackRF produces signed 8-bit samples, whereas RTL-SDR tools (including nrsc5) use unsigned 8-bit.

[argilo]

In case it's any more convenient for your use case, nrsc5's API (accessible from C or Python) has an nrsc5_pipe_samples_cs16 function that accepts signed 16-bit complex samples at 1488375/2 samples per second (i.e. half the rate that the command line tool uses).

[argilo]

Also, the Python command-line tool (cli.py) has an --iq-input-format option which allows you to choose between cu8 and cs16 formats. The sample rates are as above.