In a weird closing of the circle, I now store the internal sounds backup of my vintage Juno 60 synthesizer as a WAV file recorded from that tape backup output.
So the digital info of the internal synthesizers gets converted to analog audio in the synth, then passed as audio to my modern computer’s audio interface, which converts it to a digital representation of the analog audio.
And vice versa to restore the backup into the synthesizer’s memory.
Incidentally those backups are more reliable now than when using analog tape decks, since one doesn’t encounter physical tape degradation or a cassette deck “eating” the tape.
I haven’t done any testing with compressed audio formats, but I would expect even lossy formats to perform well, if one keeps the lossiness within certain bounds, so that the highest frequencies in the audio file are preserved.
MIDI as a compression format for that kind of audio data would be a lossy way to encode such an audio stream, and it certainly would perform well, so yes, such lossy formats do exist.
Most research in audio compression has been done on compressors that exploit the limits of human perception, though, so of the shelf lossy compressors may not do very well.
Yeah you're transmitting information about the source but to call it lossy is an understatement.
Modern synths still do this. The Korg Volca has a library for converting audio into white noise that reprograms/adds more samples.
This site describes the format, which was basically a header tone, a sync tone, data bits, and then a checksum (not described there but other sites say it was just an XOR). When we got a Disk ][ (5-1/4" floppy drive) all those issues went away.
http://www.applevault.com/hardware/apple/apple2/apple2casset...
Certain things had really distinctive sounds, like loading screens. I could recognise Manic Miner loading from just about any ten second chunk.