Glen,
As always it's "whatever works" but I look at the "glue" as more of a mixing decision, and then often in moderation much like general bus compression. Personally I just like to get the mix in order first, then apply processing that I feel is needed for the entire album. Kind of like broad strokes then filling in the details. For my money overall level is achieved by limiting (and occasional converter clipping) more so than compression. So it's something that I prefer to make a final decision on later in the chain. I don't limit more than once other than the potential combination of converter clobbering and final limiter.
There are even some MEs that "print" a version of the master without limiting so that that they can go back to it for revisions and limit further or less depending on how loud the client wants their CD. It saves them time in having to recall an analog chain or other processing.
I'm not sure that I see too much of an advantage in removing large transients first, unless it was a really spastic drummer or quirk in the audio for example clicks or pops. Hopefully large swings in volume are there for a reason otherwise they should be addressed at the track level.
As always a very reasoned and reasonable answer.
As I go back and look at this again, I think we're pretty much talking the same thing as far as limiting for final volume coming last.
The main difference I think is that I prefer not to leave it to the final gain limiter to tame runaway transients. By that I'm referring to peaks that stick up more than a dB or two over the average peak level for no really good sonic reason. Often times these peaks are a result of the summing process and not readibly noticable or best addressable on the individual track level.
For example, there may be only 6 or 8 separate beats within a 12 bar structure where the attack of the snare, the kick, the bass and the vocal are all fine on the track level, and sound fine when summed. Yet when summed they just happen to align in time a few ms finer than normal and/or have just a smal extra l amount more energy in each track to be unnoticable at the track level, but when combined they wind up producing a transient peak that's 3dB or 4dB above the rest of the similar wave crests.
I personally am a fan of manually knocking those down to the average crest level in the 2mix before sending the 2mix to compression or limiting. Personally, I find that to be a much more transparent way of recovering those few extra dBs of headroom before they even go to compression, plus it allows me to adjust each runaway peak by a specific dB amount instead of a single ratio (and therefore variable dB level) across the board.
Is that part of mixing or mastering? Good question. It occurs after mixdown, and is part of the final polishing procedure, yet it is something I'd often do before handing off to an ME. I guess if I'm self-mastering, I consider it part of the mastering phase, but if I do it before i give the final to another ME, then it's part of the mixing phase
.
Like you rightly say, whatever works, works
. I just prefer personally not to wait for final limiting to tame the runaways; it seems to me like those need to be taken care of before that, and that the final limiting should be reserved for final volume effect only. Tomayto , tomahto, probably.
G.