The examples you've cited look to be mastered (or at least sourced) differently. True loudness normalization should not affect the short term dynamic range of

The examples you've cited look to be mastered (or at least sourced) differently. True loudness normalization should not affect the short term dynamic range of the material. It simply looks at the average level and lowers or raises it to some preset point. YT could be doing inter-song normalization, i.e. adjusting parts within the song to the same overall level - but this is not a good thing as it would affect properly mastered tracks too. Your example seems to show a completely clipped sample (in the CD) and a more reasonably mastered copy (in the YT clip) with greater dynamic range. Once the dynamic range has been destroyed through compression (not digital compression, I'm speaking of gain compression http://en.wikipedia.org/wiki/Gain_compression http://en.wikipedia.org/wiki/Gain_compression) it would difficult or nearly impossible to restore without affecting the quality. My guess - in this instance, I think the CD and YT clips come from different masters.