Quote Originally Posted by Massive Master View Post
...in 24-bit, it really doesn't matter.
In floating point, it really actually doesn't matter. -80, +80, given that the dynamic range of the mix itself is ok - there's not unwanted noise or distortion in the actual render - the absolute level matters not at all. There are practical reasons why you probably should be mixing with your master peaking around 0, but in theory it really doesn't matter. A lot of times there's a point in the mix process where I slap something like my typical mastering chain - mostly just some "glue" compression and saturation - on the master bus in order to see how it will hold up and adjust some things to make it work better. I sometimes end up pushing things up a bit going into those things, so that when I bypass them to render a "pre master" mix, it peaks sometimes pretty significantly above 0dbFS. And I don't even care. Render to 32 bit FP and worry about it at the next stage.

All of which kind of contradicts the advice I tend to give people which is "leave nothing for the ME to do". Mix it until it actually sounds the way you want and don't expect anybody to come behind you and fix it or finish it or make it better. I guess I'd say don't worry too much about the final absolute levels at all. Dynamic range and "loudness" are things we should consider during the mix simply because the best fixes for issues there are almost always at the mix stage, but that is also part of what the ME should be doing, so if you know there's that next step, you maybe don't get obsessive about, and try to err toward a little more DR than you might really want in the finished product.

That's what people actually mean (if they know what they're talking about and not just parroting things they read on the internet) when they say to leave "headroom" for the ME. It has nothing to do with absolute dbFS numbers, but the relative level, the dynamic range or crest factor, the difference between the average (RMS/LUFS) loudness of the mix and the loudest peaks. That's where you need to leave some room. If everything is good but just a couple of samples along the way are peaking a couple decibles above the rest so that the overall absolute loudness can't be as high as you'd want, the ME can handle that pretty gracefully. Like, if you want -14 LUFS integrated, but you're at like -18. That's "easy". But if you go too far the other way - you want -14, but you've squashed it down to -9 - that's a LOT more difficult to manage.