Here are some explanations on how I demodulated the slow-voice transmission from R2BM.
The "PSK mode" transmission is effectively amplitude-limited audio ("Jimi Hendrix" style), transmitted as dual-sideband with suppressed carrier with an XOR function. To avoid the need of carrier reconstruction, I listened only to the upper sideband, ignoring the lower one.
For the experiment, I reused one of the wav files which are normally sent to jt9.exe for FW-5 decodes (21:20 to 21:25 UT). These are periodically recorded with 136 kHz dial frequency and 12 kHz samplerate. The samplerate of the file was forged to 1200000 samples per second, using a hex editor (804F1200 at offset 18h and 009F2400 at 0Ch). It was then read by SpecLab ("Analyze and play file with DSP"), with "play in an endless loop" enabled, allowing convenient realtime playback and adjustments.
In this mode, SpecLab shows a short repeating spectrogram from 0 to 600 kHz, with the desired SSB signal above 158 kHz. In the first blackbox, the signal is processed in a bandpass filter (fc = 161 kHz, bw = 6 kHz, Butterworth) and frequency shifter (shift=-158000 Hz, bw=6000 Hz), producing audible voice at the DAC output. For the recording of L6, I activated "use different samplerate for output, 12000, low quality) in audio settings.
For future voice transmissions, you might consider sending single-sideband mode, which would be more efficient in terms of bandwidth and power. An amplitude-limited (phase-only) version can also be transmitted using a class-D nonlinear PA.
73, Markus (DF6NM)
