Add logging in syncword sweep, fix unsigned underflow bug,
fixed symbol average computation for quantization.
Do syncword sweep whenever available.
TG-81
Add script to receive the demodulated data from Module17/radio, to use
it you need to add in meson.build:12 the ENABLE_DEMOD_LOG define.
Build and flash into the platform, build and run scripts/get_demod_log,
turn on the radio, remember to turn of radio before you terminate
get_demod_log, print the output with
`scripts/plot_m17_demod_csv.py serial_demod_log.csv`.
TG-81
DC removal filter had a bug that created weird positive spikes on
Module17, add a syncword hot re-sync to be able to correct small phase
errors that appear every five seconds more or less.
TG-81
Due to the RRC computation sometimes violating the timing constraints,
we switched the M17 demodulator to 5 samples per symbol (24KHz), thus
dividing the processing time by four ar the number of samples halves as
well as the number of taps of the RRC.
Minor adaptation on the code followed.
TG-81
Due to the RRC computation sometimes violating the timing constraints,
we switched the M17 demodulator to 5 samples per symbol (24KHz), thus
dividing the processing time by four ar the number of samples halves as
well as the number of taps of the RRC.
Minor adaptation on the code followed.
TG-81
Quantization algorithm is now based on separate rolling averages for
positive and negative samples.
Merged csv outputs from syncword search and quantization on Linux.
TG-81
Make quantization work also on DC offset signal
Add samples plot code
Fix bug in buffer wrap around
Fix oob memory access in M17 modulator
Update tests with DC offset signal
Switch to 48KHz sample rate for Module17 and MD380
Add DC biasing script
TG-81
RRC filter taps have been replaced with the ones generated by Wojciech,
a new baseband for the M17 demodulation testsuite has been generated,
using OpenRTX itself and some further minor changes.
TG-81
Added implementation of the M17 4FSK demodulator, including clock
recovery, phase detection, and quantization algorithms.
A testsuite is also included to do regression testing against a
reference baseband pre-generated and the corresponding reference
bitstream.
A 1% BER is still present due to a fault likely in the RRC filtering,
since the eye diagram of the filtered stream is bad.
TG-81
Silvano Seva
Niccolò Izzo