It all comes back to the concept which I have explained a few times already and that is peak and rms values of anything. By the time we get to a pre mastered mix we really should be talking what the average rms value of the music is. Remember the peaks take care of themselves when you think about rms.
 
I might mix to an average value of say -14 dB rms and the peaks will just rise up to where ever they go and yes they might get up to -3 dB say or even the odd one higher but the real issue is the fact that the pre mastered mix is at a certain rms level.
 
If I premaster to say -14 and I know my client wants to get to say -6 then I know I have got to find 8 dB of extra rms gain. If I premaster to -20 then I know I have got to add 14 db of rms gain and so on.
 
The rms values are way more important now towards the end of the project. They are really responsible for the overall loudness of the finished track. The levels that peak values rise above rms values determine how transient the music is going to be and that is important as well.
 
VU meters come into their own around the mastering phase and that is why all decent mastering houses have them either real or virtual. Real still having the edge.
 
So I agree with Mike's observation in post #25. peaks are of little importance, rms values are more important by the time we are at the pre mastering stage. The peaks will be arbitrary and random so as long as they are not crashing into 0dB FS then it does not matter too much where they are. A good mastering engineer is going to level them off and even them out in their processing stages and they should be concentrating on rms values instead.
 
For example if you want to know how loud a commercial CD might be that you are trying to reach then it is the rms values you are measuring and trying to reach not peak information at all.

peaks at -3 db won't increase the mastering 'Headroom' as it doesn't add dynamic range or change the crest. That 2.8 db will be raised without any limiting. That being said, My peaks of mixes are around -3 - -0.2. I normalize them to -0.2 to start the mastering process. No limiting involved with normalize. If my mix peak is at -3 and the RMS is -21, thats a difference of 18. the same mix normalized or volume raised at fader to peak at -0.2 would then have a RMS of -18.2 Still the same difference of 18. Only after that point you have to limit or face distortion. 5 dbs of limiting would have it up to -13 RMS a place where Dynamics still can live and breathe. We should all support dynamic range day! Louder mixes reduce the crest which is ok with dance/modern music/rock, but still my ears and my clients prefer the dynamics. Pre-mastered I like em -20 ish. Not big on music mastered to RMS -6. Usually distortion, sounds too compact, and lacks Depth.

I agree with not having to go as high as a mastered rms level of -6 or -7. That would be nice. BUT it can be done though. Some just dont know how to do it and that is the reality.
 
Firstly if I am going from -14 (or -20) to -6 you don't do it in any one stage, that is where people go wrong. I start by editing in a quality editing program. I limit the peaks of the music down to a certain value maybe -4dB or so then add 3 dB to the whole mix. Right there in the editor you can get 3 dB of rms gain cleanly.
 
Often the EQ stage can add a db or so, the compressor once it starts to perform gain reduction you can also add 2 dB or so there in makeup gain. By the time the mix hits the limiter the limiter only needs to add a max of 4dB of rms gain.
 
Limiters are not all created equal. PSP Xenon (costs $250 though!) can add 4db of rms power cleanly with NO distortion. Not only that it has amazing control of how it sounds and the sound of the transients. I can still get snappy and punchy transients happening, dont ask me how, that is something I think only PSP knows.
 
So in the end although I am getting close to -6 my mixes still sound clean, not compact at all and have lots of depth. Can be done like I said you just have to know how that is all, and the limiter is pretty important in the equation.
 
But having said all that I would love to leave the mastered versions around -12 or so which would be nice of course. Some clients are cool with that and in those situations all the processes I have just mentioned can be relaxed quite a bit. But other clients would not accept it and then you have to be clever about producing loud masters that still sound fantastic. It takes longer too but it can be done.
 
We are getting a bit off topic here as we are talking about final mastering while the OP was about preparing a pre mastered mix for mastering.
 
 

What is the end game? We want our stuff to be aired, even though most won't be, and we don't consider the end game. Broadcasters have their own processing that happens and it seems to me that should be considered more than ones own needs.
 
I find it interesting that when I listen to music from the 60s and 70s that it sounds much warmer than the music of the 80s. Yet, the 80s is still pretty good. The 90s stuff starts to get louder and less warm but come the turn of the century all bets are off. My wife likes to listen to the new stuff and there's this song where the bass drum thumps along and the synths, as well as everything else, audibly duck under it. I'll never understand the purpose, even though I do, of side chaining. It just sounds bad.
 
That's why I brought up the new LUFS law. It's not hit the music industry yet but it should. There is talk that all this distortion inserted into tracks to make them louder is actually damaging our ears. I guess that's the future?

Guitarpima
 
That's why I brought up the new LUFS law. It's not hit the music industry yet but it should. There is talk that all this distortion inserted into tracks to make them louder is actually damaging our ears. I guess that's the future?

Is that related to the i-tunes auto-loudness thingy?   Just read about that recently (though I don't use i-tunes).
 

The LUFS law was made so about a year ago. There was a grace period of a year and it was a response to commercials being so loud as opposed to the programs on television. I don't know that if applies to radio broadcasting. There was mention of I-tunes and MP3s and the distortion inherent to that medium not to mention the distortion added to mixes by the industry.
 
Actually, I'm hoping someone knows what it all means and can explain it as I don't quite understand it myself. There is a thread, in the software forum I believe, with links so presentations on the subject. Apparently, they are trying to end the loudness wars. Amen to that I say.

+1
 
And about time...
 

My understanding regarding the iTunes thing is that iTunes radio will have its loudness correction turned on by default. Bob Katz wrote an article hoping that this could help to end the loudness wars since it will make squashed stuff sound worse once it is automatically adjusted to the same perceived volume as the non-squashed stuff..

Yup... that's the one I saw.  

I bought some contemporary music a few days ago from Amazon and it is clear that they were mastered with too much loudness. Even a classic jazz album had been remastered with too much loudness.  I checked them in Sound Forge 11 with its new loudness meter and it stayed at the top and showed clipping. They do not sound good even though the music is great. I want to slap the "engineer". 
 
I'm not happy about this at all.