Because you're never quite sure when the DAW will trip you up. You can cheerfully freeze and flatten in Live, for example, and it will use 32bit files. No clipping. But, if you forget you've gone beyond 0dBFS and render stems to 16 or 24bit, you are suddenly in Clip City. The trouble, by the time you've reached this point, you then have to work out how to scale things back without affecting a more or less completed mix. Keep the levels lower, you don't have to go back.Syncretia wrote:If I were to explain it a student for example, what would I tell them? Why shouldn't they just turn the master down?
Similarly, recording into Aux channels on Logic and Pro Tools is generally at 24bit resolution, so may well clip even though the mix bus that provides the source is cheerfully running at 32bit. Logic seems to run channel inputs through the I/O code as recording is disabled (even on auxes fed by a bus) if you don't have an input selected.
This is why broad-brushed claims that "all DAWs use 32bit processing so everything's fine" don't really play out so well in the real world. They've largely used 32bit FP arithmetic because it runs fast on modern processors - but they've also made assumptions about how people use these tools, one of them being that people are treating 0dBFS as a limit and not deliberately going over it. So, they don't necessarily make 32bit processing available everywhere.
You also may have plug-ins that are programmed on the assumption that -18dBFS = 0dBu and so push anything above that into saturation. So, they won't clip as such, but things can get grainy real fast.
Running the channels so they are peaking at somewhere in the -12dBFS to -18dBFS range isn't going to hurt the audio ("oh noes, I'm wasting bits!!!!"). As it makes no overall difference to the sound assuming the monitoring levels are adjusted to compensate for the apparent reduction in impact and it's easier to remember to keep things quiet than to remember when a DAW will clip something, I can't see a good reason for running channels close to 0dB. If you want to deliberately saturate a channel, then just use a trim plugin there and then reduce the gain afterwards to fit the mix.
Those are my arguments. YMMV.
Seriously, I've adapted to three different recording formats and up until a few years ago, delivering signal well above the noise floor and just below the distortion point was critical. And I'm sure the day will come when you don't have to think about gain structure at all, but it ain't today.
