Soft clipping is where peaks are brought down in level so that they don't cause distortion at the ceiling, they are not allowed to square off as in everyday clipping. Instead the peaks get rounded tops that do not draw attention to themselves through excessive high frequency energy. This is one of the best ways of adding softness to a mix in the digital domain because you can avoid the harshness that ordinary limiters cause when they are squaring off the peaks at the clipping level. Combining soft clipping with a limiter can be a way of getting the loudness you want and at the same time avoid much of the harshness that else would be produced, since you can limited much less in order to get the loudness you want.
I usually don't touch the output volume fader when I mix (I only use it temporarily to find out different things about the mix), I leave it at unity reference gain. I guess I am used to the analog domain where everything sounds the best near the unity gain marking. I don't know if this is the case in the digital domain as well. Just like you and many other mixing engineers you use the mix output volume fader, sometimes it can be a good indicator of how well a mix sits. You mentioned that you were tracking and boosted the faders and then you didn't have to adjust the mix output volume fader and you realised this was good sounding. To me this sounds like two things happened:
- You were applying less effects than you usually do and saved a lot of the signal this way, resulting in a cleaner final result
- You were tracking instruments that together produced balanced frequency energy, meaning the elements covered their own frequency range well (which is a best practise), the tracks were probably also recorded a little less dynamic or the mix included less dynamic instruments.
Especially the latter mentioned thing is important here, because with much less high peaks you get much less distortion in the end after limiting (through less squares at the ceiling and less distortion through better signal-noise ratio), especially in combination with soft clipping and this makes things louder too. I think that's mainly why you experienced a better final result.
Of course you can control loudness only through using the volume faders, in fact you should if you want a loud final result. I will sum it up. When you mix, some of the instruments are dominant, some are not depending on the kind of element they belong to and what instrument you feel you want to be dominant. If dominant instruments like for instance drums are recorded very dynamically, it will produce high peaks, but the actual RMS level might often be quite low. In order to mix this instrument louder than the rest (by ears) to make it dominant you will over compensate due to the high peaks, so that the volume level of the other instruments automatically are brought down. When the other instruments are brought down in level they get a worse signal noise ratio. During a such situation hard limiting on top of that will create an extremely harsh sounding final result since the transients are cut off really hard and the overall signal noise ratio is so bad. In other words, this is a very important aspect of recording..!!
Due to this phenomenon it's very popular to use compressors with soft clipping during the tracking process, especially when tracking drums, because drums are often one of the most dominant elements with very high recorded dynamic range. Some engineers say compressors are the best tools for getting high dynamic range and that's absolutely true even though it is completely against the nature of what a compressor really does. Above I described what happens in practise, now I will describe how the dynamic range can be improved by using compressors with soft clipping properties. When you track instruments that are very different in transient response you will easily loose signal noise ratio if the dynamic instruments take up a lot of dB on the mix and they are dominant on the mix. When you lose signal noise ratio you actually also lose dynamic range. The reason why compressors with soft clipping properties are so good at increasing the dynamic range is because they allow you to control the transient response so well during the tracking process that you are able to mix the song in such a way that you will not get a bad final overall signal noise ratio and the sound is still very natural sounding. This in combination with the correct track volume balance creates the perception of high dynamic range, because when limiting the overall signal noise ratio is still at max level which means you can more easily feel the dynamics in the mix when the mix is loud.
This is probably the best argument for limiting a song hard, in that loud mixes with good signal noise ratio can create the perception of a really clean sounding mix due to the low noise floor! (even though the actual dynamic range is very low)
This is one of the secrets behind good sounding mixes and I'm sure a lot of engineers would not like this to be discussed so much because this is probably one of the best ways of dramatically improving the sound quality of a mix and every professional engineer has his ways of doing this in the most efficient way. But discussing soft clipping in this context makes this secret really exploited!
This is exactly why the digital domain is seen as harsh sounding and why it's so easy to get a better sounding result in the analog domain. If you track to tape you can use the 0 -> +6 dB range for controlling the transient response and in that way you get a material that is much easier to mix with a good final result.
Generally I think the digital domain is a little raped today. Engineers that are used to recording in the analog domain just record, mix tracks with completely wrong volume level settings, pan wrong due to bad converters, apply a lot of effects on top of that with completely wrong wetness and finally they limit the song hard with something like the L2 to make it sound as good as an analog version. The result of that sounds like a song without energy and emotions, simply a dead sounding mix. Working in the digital domain demands an understanding of how to keep a signal clean, because it's just not like in the analog domain where analog clipping is the solution to evertyhing and applying effects on everything and with whatever wetness you like will create the result you want. Applying an effect wrong on a the wrong track can be devastating in the digital domain! It's not uncommon that engineers put reverbs too wet on the drums when the drums are dominant on the mix, because they are used to creating a large enough room simply by controlling the wet ratio on the reverb and/or delay effect.
Luckily there are still mastering engineers out there that can and also want to save digitally destroyed mixes somewhat. The first thing they do is to prepare their best analog recording medium. It's like: "Ok, this mix was a little digitally raped, I have to try saving the life of this one with the help of my analog partners!"