Yes, that seems like the best approach. At 20Hz waves, ntp would probably be good enough. The real issue I can see is to get that time-data into the sound data.

Yes, that seems like the best approach. At 20Hz waves, ntp would probably be good enough. The real issue I can see is to get that time-data into the sound data. Maybe you could time-stamp the beginning of a recording, and start a new recording every minute (to compensate clock drift). That might work, biggest issue I see is potential variable latency between 'starting a recording' and the actual first recording happening.