Re: For the believers and SoundBlaster haters

"Randy Yates" <yates@xxxxxxxx> wrote in message
news:m3k5m4qqiq.fsf@xxxxxxxx
"Ethan Winer" <ethanw at ethanwiner dot com> writes:

Don,

applying artificial dither at conversion to 16 bits
will be totally swamped.

No kidding, but try telling that to the believers.
Recently I was involved in three different simultaneous
dither discussions in forums, so I put up the test file
described here:

www.ethanwiner.com/dither.html

Of course, nobody could tell where the dither was
switched on and off, even though it happens right in the
middle of musical phrases. So instead the believers
attack the test as invalid. If dither is so incredibly
important, why can't anyone tell which parts are
dithered and which
are not? :->)

Ethan,

I wouldn't say the use of dither makes a difference in
all scenarios. At 16 bits, the quantization noise is
already so low that, in order to hear the effect of
dither, the level of the passage must be very very low,
perhaps no less than 80 dB below full-scale. I didn't
measure your clip, but I bet even the soft parts were
well above -80 dBFS. Try the test again with levels set
lower.

Time to call time on the grounds that in the real world, there are no
recorded audio signals that are 80 dB down unless someone screws up very
badly. Normally, the softest sound on a recorded track is about 65 dB down,
more or less. When you have a recorded audio signal this small, it is
usually about 80% or more broadband noise, perhaps with a reverb tail
sliding under it.


.