Almost any channel adaptive digital encoding (ie. where the sender gets feedback from the receiver about how much data is successfully arriving) will end up sending more useful data than any analogue scheme in any conditions.

Super weak voice channel where you can barely make out any words? That same channel, with the same transmit power, will get a decent telephone quality digital voice down, as long as you use modern low bitrate audio encoding. And it would send pages of text per second if you can handle typing not speaking.