[lmm]

> This doesn't make sense to me. When I say "the parts of the music that would have been quieter", I'm mostly not speaking of softer passages vs. louder ones. I'm referring to the fact that even within the span of a single measure, the original signal will generally have had significant peaks and valleys.

I don't think that can be right - sound is a waveform so even the loudest notes will oscillate around zero. "Loudness" would usually be defined as something like root-mean-square amplitude, but if we zoom in to talk about a single note (sine wave) then the peak amplitude and the RMS amplitude are in a direct relationship with each other. On the scale of a whole piece, yes there's a difference between average and peak, but on the scale of one groove to the next - a couple of seconds - I don't think there's any practical difference.

The complaint I've heard about "loudness wars" e.g. Californication is precisely that there are no soft passages - every bar is pushed up to full volume. I'm not aware of anyone talking about the effects of compression on a smaller scope than the couple-of-seconds scale.

[ScottBurson]

> sound is a waveform so even the loudest notes will oscillate around zero

Sure. When I refer to "peaks and valleys" I'm talking about the envelope of the waveform. I think that's standard usage in audio engineering.

I think you should download Audacity or equivalent and look at some waveforms of both uncompressed and compressed music. I think you'll see that there are loudness variations on a much shorter timescale than you expect, and that compression reduces these considerably.