I posted recently on gearslutz.com about this subject and I don’t want to rewrite it all again, so I’m just going to paste it here for anyone that is interested:
I’ve spent a lot of time lately doing comparisons of super super loud masters (-4) to quieter ones (-14). Really good masters still sound good even at -4… so much so, that I don’t have a drastic preference listening to music either way. A great sounding master at -14 IS very slightly capable of greater detail… but it’s slight. The difference comes down to transient handling.
This comparison (loud masters to soft) is similar to comparing a GML preamp to a Neve 10xx. The GML retains a little more transient detail, while the Neve tends to smooth transients… but the Neve smooths them so well, it can even be perceived as euphonic - not displeasing. Likewise, a good master does the same thing if/when loudness is achieved.
Somebody mentioned density earlier in the discussion… Even if something is quiet, it can still sound really dense if it’s heavily compressed or if the arrangement is dense… but in master comparison, I think “transient density in the loudest range” is the biggest difference… not overall density. What I mean by that is density of quick immediate bursts that happen to occur in the loudest dynamic range of the master. In quiet masters, these little bursts are all over the place – in every instrument - even the loudest ones. The transients are subtle, but perceived as “detail”. In loud masters, the transients of the loudest elements in the mix are gone whenever the momentary loudness of that element is high… those transients get flattened. Quiet elements in a loud master can still have those transients though as they’re beneath the range where limiting/distortion would affect them… Anyway, I think the reason loud masters can still sound decent, is because our brains don’t easily hear or detect transients well, as they are super fast…
Just be aware though, when you go for that loud master you ARE losing something (transient detail), but if it’s well done what you lose can be very subtle, or even euphonic in the way it’s achieved, and the music can still be completely enjoyable for most people. This is why some of the most beloved songs in history can be either super loud masters OR super quiet masters. As engineers, we might complain, but the simple fact is that the public can tolerate either loud or quiet masters if the master is good enough and the song is good enough… and if loud and quiet masters are actually presented at equal volume so as to not bias the listener (so I AM for universal volume normalization without limiting/processing).
As far as how loud to master something, that depends. Do you have the equipment and knowledge to do a loud master really well? If not, go quieter as it’s safer… or better, hire a pro mastering place to do your master. If you do have the ability to make it both loud and good, you have a choice. Do you want it to have the transient detail of a GML preamp, or the thickness and warmth of a driven Neve preamp? Loud masters vs soft masters are a similar comparison. Try doing a master that way sometime… take your -14 master, and slam it through a Neve preamp cranked up… the Neve can reduce -14 to -10 very easily. It adds distortion, but you may notice it’s a cool sound, and you can still enjoy the music. If we aim for a system of universal volume normalization, then loudness becomes a irrelevant, and sound is all that matters… so you could print your master confidently at -14, or at -4 - whichever level has the “sound” you’re going for (the amount of dynamic range in your transients that are near to 0dbFS).