Originally Posted by CharlesJ
You mean that that recording session had greater than 96 dB of actual music dynamic range? Rally?
No, but dynamic range isn't the issue; the issue is the peak. When engineers record digitally, they adjust levels so the input comes as close to the red line as it can without going over. For the most part, digital clipping is harsh and unlike older analog gear, it pretty much never sounds good. They could be recording a whisper of 10 dB, but they're still trying to get as strong of a signal as they can, so they'll pump it either by turning the inputs up and/or compressing the vocals(these days it's almost always more of the latter and less of the former).
For recording, mixing and mastering, the name of the game is toeing that red line. So if a track is mastered in 24-bit, the loudest parts of the track are exceeding 96dB, peaking out at the limit of the equipment. I think the average real-world headroom of today's 24-Bit gear is around 122dB. The process may be changing a bit, but as recently as the last time I converted mastered 24-bit audio tracks to redbook audio, the resultant cd would have clipping at the peaks, and it was harsh.
I heard recently(on a recent HT Geeks podcast, I think) that a lot of today's pop engineers(especially on rap/R&B releases) do this on purpose. The mastering equipment is good enough that they can actually make the master clip by 2-3dB over the limit, which is just enough to boost the apparent loudness of the cd without the presses spitting the disc out as an error. The filters are apparently good enough to clip without it sounding like crap to the average listener.
Anyway, converting a file that peaks at 122dB to one that peaks at 90dB creates challenges and usually involves either lossy dithering or other means. I believe when jsalsburg originally said compression, he was talking about data compression, but either case the signal ends up undergoing negative digital processing to fit onto a cd.
I also think what jsalsburg said about playing it back on a big system
is key. The closer an audio system can come to cleanly reproducing the dynamics made possible by 24-bit files, the more likely a person is to being able to hear the difference.