I am asked almost everyday if it's OK to limit or compress the final stereo mix BEFORE mastering. While there is no clear answer to this without actually listening... the screen shots below are excellent examples of the process and what's at stake. Thankfully, in this particular instance, the client provided me with both limited and unlimited versions of their tracks. They preferred the limiting but were concerned that it would get in the way of the mastering.
Track 1 (top) is the client's final mix down without any limiting. You can see that there is tons of dynamic range. While this is not necessarily an indication of the way the track sounds, I can testify to the fact that it was rich, warm and open sounding.
Track 2 is the the client's preferred mix and it's obviously brick-wall limited almost beyond recognition. Strangely, the output was set to -4 dbfs which is an indication to me that they mistakenly thought this would provide me with enough room to "take it on home." The truth of the matter is that the output of a limiter doesn't matter as much as what's occurring on the Threshold side of the equation. A quick reminder regarding limiting, it's basically compression with a ratio of 10:1 or higher (up to 100:1 with brick-wall). 10:1 means it would take an increase of 10 decibels to cause the output to increase by 1 db. That's a lot on a stereo mix, especially when you're driving well beyond peaks. And don't forget that just because you're not clipping doesn't mean the limiter's not causing distortion.
Back to this particular version. You can tell just by looking at the wave form that there's some nasty stuff going on. On the original, unlimited version, the volume discrepancy between the quiet and loud parts was fairly huge for pop music... close to 20 or 30 db, but all of that was stripped away by the limiter, which left me scratching my head as to why they even bothered mixing it with such dynamic range to begin with. What did they like about the limited version? Certainly the upper mids became more pronounced due to all of the distortion, and the "quiet" parts were now "in your face." But, in order to make the "quiet" parts sound like that, the limiter had to be driving at least 20 db deep into average level (not peaks) and the chorus' sounded like the audio equivalent of 'Henry, Portrait of a Serial Killer' (ie, a song nobody wants to listen to because it's just too brutal and gruesome to bear.) So why would I master from this?
Track 3 is the final master. Needless to say, I worked from the original, unlimited mix. Notice that the dynamics are still somewhat intact (especially for a dance/pop tune). I was able to get a substantial amount of volume WITHOUT flatlining the entire mix and destroying all of the wonderful low-end. And equally important, when this track is played next to other commercially released tracks of the same ilk, it holds up both sonically and volume-wise. I'm happy. Client's happy. No need to belabor the point.
In summary, I like compression and enjoy hearing the results. If it's done lightly, in stages starting with the individual tracks and working your way up to the final mix down, I say "yay." The more you control your peaks and valleys in the mix down, the better the master will be. I wish I could play you these examples without singling anybody out and making them feel bad, but the waveforms speak for themselves. Avoid brickwall limiting but if you just absolutely have to use it, only grab peaks.