On 7/12/2014 3:38 PM, Frank Stearns wrote:
Judging by some of the
specs of these supposedly 24 bit consoles, there really didn't seem to be 24 bits
inside, but something less (18? 20?), with the extra bits perhaps being used for the
results of larger multiplies. Is that the case? Or did they just bit-shift and let
the LSBs fall off?
How are you judging? By the S/N or dynamic range specifications? When
it comes to the real world, everything goes in as 21 or 22 real bits (a
24-bit A/D converter on a good day) and has to go out through a 24-bit
D/A converter (because that's all we have).
As Scott and I have mentioned, with 32 bits or more of space for words
to grow when being processed, and allows you to sum a whole bunch of
24-bit streams without truncation or overflow. For example, when I
reviewed a PreSonus StudioLive digital console, I put up 16 channels of
(just to be on the safe side) -0.1 dBFS channels, turned all the faders
all the way up, and the master meters were well off the scale and the
analog output was horribly clipped as you'd expect. But turning down the
master fader so the meters were back on scale and the THD at the analog
output was just a couple of thousandths of a percent greater than any
one channel playing alone. That's really impressive.
Of course you still have to worry about clipping the inputs, but once
you're there, you don't really need to worry about clipping an internal
mix bus. Of course you still need to watch out for limits on plug-ins.
If you sum all your drums and have peaks that (if it worked this way)
were +10 dBFS or so and try to run them through a compressor plug-in,
you'd need to drop that back so that you were feeding the compressor
with something under 0 dBFS peaks. So you still can't be sloppy when
things get more complicated than mixing.
Just guessing, but it seems that given two not so great choices, the one with the
smaller but untruncated word might be better sonically than fixed truncation steps,
especially if your input level was low (for whatever operator oversight -- perhaps
the reason for paying careful attention to gain structure).
I'm not sure what you're getting at here, but you really shouldn't
operate a digital mixer any differently than you'd operate an analog
mixer. In the real world, with real world signals, you can truncate a
32-bit audio word to 24 bits without applying dither and not have it
sound bad. Some mixers may use dither at the master fader. I know that
this was a big deal in Pro Tools a few years back when they figured it
out, but now people are saying that with the mixer in PT10 and 11,
selecting the dithered mixer adds things that you don't hear when you
use the just plain ol' mixer (that they seem to have done right).
But I dunno. I'm not one to listen for "artifacts" so I don't hear them.
--
For a good time, visit
http://mikeriversaudio.wordpress.com