I have a little hobby project where I record an FM radio music station using a SDR and then remove all the non-music portions for offline listening. I like the

I have a little hobby project where I record an FM radio music station using a SDR and then remove all the non-music portions for offline listening. I like the music selections the DJs pick, but I prefer not to listen to the DJ commentary and the advertisements. I evaluated three methods of recording: analog capture from a standalone FM receiver, using this nrsc5 library to record the "HD" radio stream, and using an AirSpy SDR with this library: https://github.com/jj1bdx/airspy-fmradion https://github.com/jj1bdx/airspy-fmradion Recording the "HD" (what a misnomer) radio was nice in that there was no hiss or multipath effects, but in comparison to the other methods the digital compression artifacts became impossible to un-hear. It seems to top out at about 96 kbps The airspy-fmradion library has some nice stuff in it to address multipath, resulting in the best audio quality of the three methods I tested. I use https://github.com/ina-foss/inaSpeechSegmenter https://github.com/ina-foss/inaSpeechSegmenter to identify which segments of the recordings are speech vs. music.