what does a lossy compression scheme "lose"?

[the_bertrum]

Sure, you've listened to audio recordings that were compressed by various schemes and if you're like me you've made appropriate noises about how good or bad you thought they sounded.

But have you ever wondered what, literally, has been lost in a specific recording?

I remember reading about the development of MP3 (see, for example, MP3). Yes, the lossy scheme is based on psycho-acoustic theory (say that 3 times fast) but it contains a tweakable parameters. I was struck by this passage from the Wikipedia article concerning a Dr. Brandenburg who did a lot of the practical development

An acapella version of the song "Tom's Diner" by Suzanne Vega was the first song used by Brandenburg to develop the MP3 format. It was used as a benchmark to see how well MP3's compression algorithm handled the human voice. Brandenburg adopted the song for testing purposes, listening to it again and again each time he refined the compression algorithm, making sure it did not adversely affect the reproduction of Vega's voice.[45] Accordingly, he dubbed Vega the "Mother of MP3".[46] Instrumental music had been easier to compress, but Vega's voice sounded unnatural in early versions of the format. Brandenburg eventually met Vega and heard Tom's Diner performed live.

This quote came from the "Development" section of the article but and there's a bit more in the "Standardization" section.

As it happens, last night I came across a reference to an old VOX.com article about this very song: "Listen to what gets lost when..." The clever bit is that you can listen to both the song and the "remnants" of the song, e.g, the difference between the uncompressed and compressed versions of the song.

I know of no similar comparison for other lossy schemes but if they exist I'd be grateful for some references.

Regards,
Kent

Gah! that Vox article opens with this:

Part of the reason people love vinyl records so much (even though they don’t necessarily sound “better” than digital formats), is that they are closer in format to the way you would actually hear a song if you were standing in the room with Rihanna instead of listening to “FourFiveSeconds” on your phone.

So not true.  There's all sorts of magic going on to the way the sound is stored then reproduced from Vinyl and it is only if the chap doing the mastering was good that it sounds anything like "being in the room".  In fact, I contend that the best mastering and recording engineers recognise that how it sounded in the room is often not the best way for it to sound, even if there was an "in the room" version what with overdubs and multitrack recordings.  Also, what Vox presents the "lost" information is actually a new composition based upon that lost information.  The actual lost information is much less.

But, I do love these sorts of things.  I can't reliably tell between MP3 and FLAC, and hearing what it lost in Tom's Diner explains to me exactly why, especially in music with a small dynamic range like most popular music.  All the lost stuff is overpowered by the stuff that's kept and my ears are not golden enough to hear it.

Thanks Kent, entertaining reads