I was recently watching the Steinberg Internal Mixing DVD and, as a result, I have a question.
One of they points they make on the DVD is that, when one is sending pop and rock music off for mastering, the output of the mix at the loudest part should be no more than -14dB. So my question is:
does that -14dB apply to the entire mix before one mixes it down to a stereo track or after it is mixed down to a stereo track?
So, for example, I have my 16 tracks mixed down to a stereo file and I'm still at 32-bit float. Should the stereo file I'm sending to the mastering house be no louder than -14 dB? Or, can the 16 tracks before the stereo mix down peak at -14dB?
It seems to me that if all the tracks combined before the stereo mix must be no louder than -14dB, this can create some problems with gain staging i.e. no fader for an individual instrument should be set higher than the stereo bus fader.
First of all, there are no rules for sending something off to a mastering house, and that -14 dB number they came up with seems rather arbitrary. The general consensus, though, is that you don't want to slam the levels right up to digital zero, because then that doesn't really leave the mastering engineer any headroom if they are working entirely in the digital domain, or it will be driving their D/A converters well above the optimum operating range if they are sending the signal out to analog gear, possibly adding some distortion by driving those D/A converters too hard.
However, if you are simply going to turn down the master fader to get your levels reasonable, then there is no reason why the mastering engineer couldn't also simply attenuate things digitally before doing his thing... the results should be the same either way.
Also, most of the better professional mastering engineers have really great A/D and D/A converters with plenty of headroom that can handle hotter levels if needed... but, it's still best to keep things in the range they were designed for.
As discussed before, the main reason for keeping your signals at a reasonable level is to not overload the analog gear, such as your A/D converters, or your D/A converters, and anything else you are using on the analog side of things (compressors, pre-amps, equalizers, etc.). Most of the analog gear was optimized to be used at 0Vu, which is usually about -18dBFS for average RMS level. That would be the number I would try to keep your average RMS level at (NOT your digital peaks, which are different) instead of the -14 dBFS. Maybe they meant that -14dBFS should be about the highest RMS level you hit. Your digital peaks could go as high as -6 dBFS if you are keeping your RMS levels in check, thus giving you plenty of digital headroom still. I think those are good numbers to shoot for, which will allow any analog gear the mastering engineer uses to operate in its optimum range.
Yes, you can always just turn down your master stereo fader to get the levels right, but you should start out trying to keep your master fader at the 0 (unity) position, and bringing down all the channel faders as needed to put the levels right where you want from the start. This will also optimize the gain throughout all the channels and ensure you aren't driving any plug-ins on group channels or busses too hard. However, some channel faders can certainly end up being higher than the master channel in parts of the song if the audio on that channel is really low in level at some points and needs boosting to get it to the right level in the mix. There is nothing wrong with that at all. Don't worry about where the actual fader positions are as far as the numbers are concerned, just watch your master fader level, and start with all the faders low enough to give yourself plenty of room for some dynamics in your song, especially in the loudest parts, so that you aren't getting so hot that you need to pull the master fader down. But, then, just use your ears and put everything where it needs to be to get the musical balance you want.
Again, there are no rules, though, and I've seen plenty of people slamming all the channel levels super hot, and then pulling the master fader way down to keep from clipping the D/A converters. Most people would agree that's not the best practice, but if everything in the DAW is floating point, then, at least in theory, it shouldn't make a difference if you pull the master fader down, or reduce the level of all the individual faders.