Kylotan
I disagree with the assessment that you leave the loudness until the very last step.

Of course there are many things you can do during the mixing process to make a louder overall two-track mix. I'm referring to the loudness of the two-track mixdown. At least given the way I master, this requires doing tweaks at the waveform/surgical level, which can done only when you have a final two-track mix with which you can work. You can't make these kinds of tweaks when mixing.
 
For example suppose four different tracks "collide" on one note to create a big peak on three or four consecutive half-cycles. Trying to fix that through automation of the four tracks will drive you crazy. Fixing it in the two-track mix is easy, so you can reduce the level of those peaks, open up headroom, and raise the overall level without having to do dynamics control. 
 
I think that generally the best option for optimizing levels is to do so during the mastering process, but as long as you don't master at the waveform level but simply use processing, you can certainly do that in the master bus (and that's another reason why you want to have rational levels at the master bus).
 
However I also think that your mix should not need a lot except for optimizing levels, although multiband dynamics processing will affect the spectral balance and that may require subtle EQ. When I do final tweaks in Wavelab, waveform editing aside, the processing I typically use is multiband dynamics and maybe some image widening. But if you use Wavelab's feature of A/B comparisons of processed and bypassed with level compensation, you really don't hear much difference. The balance is already where it needs to be.
 
One thing I find intriguing about LANDR is it seems it would make it very convenient to get a rough idea during the mixing process of how mastering will affect the song, especially because you can audition variations among light, medium, and heavy-handed approaches.

vladasyn
The reason we got to talk about how limiting works is because somebody said- they would mix to -3 and then use limiter to bring it to -0.1, which to me is not a good idea. I mix to -0.1, then limit peaks by master limiter and bring it slightly higher. I still do not see a reason to mix under -3.

Loop output levels can increase by a much as +6 dB (not common, but it happens) when changing tempo. Randomized modulation effects and chorusing can produce peaks that are several dB higher than average levels. So suppose you mix to -0.1. You think everything's fine. Then a modulation effect changes the level of a track you thought was okay, or a loop output level changes during a tempo increase, and now the master bus is over -0.1. (And you were probably getting inter-sample distortion anyway during D/A conversion when you were mixing to -0.1, but that's a whole other topic.) So you reduce the levels of the two tracks that are contributing to the over, but now you've upset the balance you worked on. If you had mixed to -6 on your master bus, you wouldn't have needed to change your balance.
 
Remember that the final signal leaving the two track master bus is what feeds hardware. Hardware does not have the near-infinite headroom of SONAR's summing bus and audio engine. You can go into the red all day on individual channels, assuming your plug-ins can handle it (which as mentioned before is not a certainty), but hardware does not tolerate going into the red.
 
You can do and think whatever you want, but there are valid technical reasons for mixing to a lower level than 0 on your master bus. You can choose to abide by or ignore those reasons, it won't affect my mixes .

I keep a "mastering" style chain on my "Premaster" bus (a bus just before my master that I routed everything into). This includes two stage compression, EQ, saturation, tape emu, etc. I only keep a limiter on the master bus and that only acts as ear protection if I accidentally crank something.
 
The premaster effects chain (mostly Prochannel stuff) mostly stays off. I just turn it on every so often and fiddle with it to see how the mix would respond to the various effects.
 
As the mix evolves the settings on these get less and less invasive and I'll slowly start excluding each effects because they are no longer adding anything useful. Ideally by the time I'm done the mix sounds as good or BETTER with the entire chain off.
 
This is based on advice I've received from Danny and others about not relying on any kind of mastering stage to fix stuff. The premise is if an effect on the master bus makes the mix sound significantly better then you can usually go into the mix itself and tweak something to do the same thing. It's just a matter of figuring out what that thing (or combination of things) is.
 
The result has been much more even mixes that I can't seem to make ANY better with my own mastering (but a real mastering engineer with proper tools could). So really my "mastering" projects tend to be just light limiting/raising the volume and little things like adding a bit of air or rolling off unwanted subsonic frequencies (which are mostly already gone by this point anyway).
 
So I don't mix INTO a mastering fx chain but have one setup anyway to a/b the mix with so it actually exposes problems... not cover them up.
 
Essentially I'm trying to get the mixes to be releasable as is and keep them like that until maybe one day I can either afford to get a pro to master a whack of the tunes in one go or learn enough about proper mastering (and own the tools) to do it myself.

Actually, thinking about it, that's exactly how my projects are set up.
 
The master bus is always bypassed during my final exports.

Bristol_Jonesey
Actually, thinking about it, that's exactly how my projects are set up.
 
The master bus is always bypassed during my final exports.

It's quite possible you are one of those who contributed to me developing this workflow... so thanks for that.
 
It takes a village, aye?
 
;-)

John T
You can mix into 2-bus processing, for sure. Lots of mixers swear by it. But that is in no sense "mastering".

Hence the phrase "mixing into" and not "mastering into" I guess!

Beepster

Bristol_Jonesey
Actually, thinking about it, that's exactly how my projects are set up.
 
The master bus is always bypassed during my final exports.

It's quite possible you are one of those who contributed to me developing this workflow... so thanks for that.
 
It takes a village, aye?
 
;-)

Aye laddie!!

Anderton
For example suppose four different tracks "collide" on one note to create a big peak on three or four consecutive half-cycles. Trying to fix that through automation of the four tracks will drive you crazy. Fixing it in the two-track mix is easy, so you can reduce the level of those peaks, open up headroom, and raise the overall level without having to do dynamics control.

 
Sure, sometimes a bunch of reasonable track levels sum up to an excessive bus level. But sometimes a single track, or several tracks, just have too high a dynamic range, and trying to fix that in the two-track could mean you have to make a far bigger adjustment than if each of those tracks had been limited by a couple of dB in the first place. For this reason I'll always put a soft clipper on drum tracks and the percussion bus, and audition it to make sure it's not noticeably damaging the transient, and that way I know my final peaks are not going to be too high relative to the general song level.

Kylotan

Anderton
For example suppose four different tracks "collide" on one note to create a big peak on three or four consecutive half-cycles. Trying to fix that through automation of the four tracks will drive you crazy. Fixing it in the two-track mix is easy, so you can reduce the level of those peaks, open up headroom, and raise the overall level without having to do dynamics control.

 
Sure, sometimes a bunch of reasonable track levels sum up to an excessive bus level. But sometimes a single track, or several tracks, just have too high a dynamic range, and trying to fix that in the two-track could mean you have to make a far bigger adjustment than if each of those tracks had been limited by a couple of dB in the first place. For this reason I'll always put a soft clipper on drum tracks and the percussion bus, and audition it to make sure it's not noticeably damaging the transient, and that way I know my final peaks are not going to be too high relative to the general song level.

I'm talking about a much more specific situation than that...for example where multiple notes hit at the exact same spot. It won't matter if they're limited or not at the track level, it's the combination that causes the problem. The only difference if the transients are really squashed is that the problem will occur over a wider time range, and thus be easier to fix in other ways.
 
This is a particular problem in classical music. Although I know you're not supposed to touch classical music, I do it anyway. If I can bring up the entire piece by a couple dB just by doing surgery on a couple peaks, I think it's worth it. Just don't tell anyone I do this...I'll be arrested by the mastering police 
 
 
 
 

Kylotan
Sure, sometimes a bunch of reasonable track levels sum up to an excessive bus level. But sometimes a single track, or several tracks, just have too high a dynamic range, and trying to fix that in the two-track could mean you have to make a far bigger adjustment than if each of those tracks had been limited by a couple of dB in the first place. For this reason I'll always put a soft clipper on drum tracks and the percussion bus, and audition it to make sure it's not noticeably damaging the transient, and that way I know my final peaks are not going to be too high relative to the general song level.

 
I know it's not for everyone, but I love Sausage Fattener for boosting loudness and clipping peaks on the drum bus. It sounds great set to about 8-10%. Whack it up even higher for added crunch.