In article ,
Bob Olhsson wrote:
Unfortunately a lot of people confuse the signal processing appropriate
for low-bit computer audio with that appropriate for low-bit lossy
coding. A pristine recording from a mike feed with no further signal
processing can sound stunningly good in an MP-3 encoding. As the
quality of the source drops, the quality of lossy-coded playback drops
faster.

I've also found that to be true- and interestingly, have found
advantages to using high-res sources or fancy noise-shaped dithered 16
bit- though really that will work better with encoders which are
discarding information near Nyquist, because again it's a tradeoff-
unless the encoder disregards stuff over 20K (which seems common), it'll
waste information trying to encode the zone where error information is
dumped, namely very high-frequency noise.

In a way this is an argument for not doing any processing, which
would give the low-amplitude MFSL version the edge.

I think what's really going on here is this: Myke has distinct
preferences for a particular RMS/peak ratio. This is not unusual at all
as far as I'm concerned- I like about 15 db of peaks myself, Myke wants
only 10. There's a distinct sound associated with both extremes- well
mixed audio with very high crest factor is liable to sound particularly
open and stark (digital mixing can be prone to this effect if you're not
using a lot of compression) and very low crest factor sounds congested
and dense.

Super-high crest factor can sound almost magical and unreal, not even
physically there, more like an imagined fantasy of what music might be
like, easily disrupted by unwanted ambient noise.

Super-low crest factor can sound as solid and dense as a brick,
dominating over any other sound, crude and unvarying.

Dark Side happens to be very high on the crest factor, even higher
than I like it, so it's understandable that Myke doesn't like it the way
it was recorded, and wants to change it.

In a way this is a glimpse of the future- one day, you'll pick out a
recording that you think sounds nice (say, "Low Spark of High Heeled
Boys", with its high crest factor and extended top-end) and your trusty
iMusicBox computerized gizmo will, on the fly, tinker with every single
track on your random playlist to 'map' roughly the same frequency
balance and dynamic behavior onto everything you hear. The hard part is
putting crest factor back into stuff that's been peak limited, but this
isn't impossible, it's just impossible to do well. Something like Dark
Side of the Moon in your playlist, being a very well recorded piece with
excessive fidelity and crest factor, will turn out perfectly suited to
this new world, because it will be malleable into any possible desired
'sound' and will sound its best whatever you do to it.

Which is of course what 'mastering' is about, even if people don't
always use it for that purpose anymore. Done properly, it's the craft of
delivering recorded music in such a way that it's able to adapt to
whatever playing conditions you impose on it- and we have yet to see the
full extent of what that will mean. Boomboxes with EQs or fake reverb
presets are barely scratching the surface of what you can do.


Chris Johnson