Yes, it keeps being a bit hackery… The waterfall had a lot of noise and a white band at the higher frequencies. Playing with it gave a lot of failed observations. So I started with the basic satnog_fm.py flowgraph again and increased the audio sample rate from 48k to 480kHz. The waterfall was very good. But the audio ogg file was to big to upload. I have now removed the audio generation from the flowgraph. I got:
See: Observation 4484958
The wide bandwidth is good for seeing both peaks, and maybe for finding other sats. But the doppler deviation is hardly visible. I also increased the amount of fft channels from 1024 to 2048, but the picture did not change.