'I was ready to fine-tune my compression algorithm,' Brandenburg recalls. 'Somewhere down the corridor a radio was playing [Suzanne Vega's song] 'Tom's Diner.' I was electrified. I knew it would be nearly impossible to compress this warm a capella voice.'... He wound up listening to the song thousands of times, and the result was a code that was heard around the world. When an MP3 player compresses music by anyone from Courtney Love to Kenny G, it is replicating the way that Brandenburg heard Suzanne Vega.
Wouldn't that only apply to MP3s encoded with the Fraunhofer psychoacoustic model?