Well for me it came from running multiple copies of the same bass heavy tracks encoded in different formats through spectrum analysers. But I guess those lie?
>Regarding solo vocals, the history of the MP3 format says: "The song "Tom's Diner" by Suzanne Vega was the first song used by Karlheinz Brandenburg to develop the MP3. Brandenburg adopted the song for testing purposes, listening to it again and again each time refining the scheme, making sure it did not adversely affect the subtlety of Vega's voice".
Human voices come in a wide range of tones and frequencies. Optimizing something for one voice doesn't mean all voices will benefit from the same optimizations. The specific track I was referring to had a lot of variation in high and low notes. You can tell me all you want what I did and didn't hear.
Of course it's going to look different in a spectrum analyzer, the whole point of lossy compression is to discard parts of the audio to save space.
You can't evaluate the quality of a lossy codec by looking at spectrograms. They're designed to fool human ears, not measurement software.