It's not normal mastering, it's not mixing...?
Stem mastering is a different, expanded and more involved version of the classic, stereo mastering. The main difference is, of course, in the stems. Stems are subgroups of the mix, bounced separately to stereo tracks.
They should, if left untouched, sum together to form the mix as constructed by the mixing engineer. They allow the mastering engineer to access individual layers of the mix and make less intrusive adjustments to make it translate better.
Just as with stereo mastering, it's all about making the mix and its storytelling translate as good as possible to the myriad of different reproduction systems. Between the 1,99€ earbuds and a 10kW club sound system with subs that make your trousers flap on kick hits, if there are weak spots in the mix, there is a system to expose them! Mastering is to find and minimize those weak spots.
Those subgroups usually include instruments with a similar place and purpose in the arrangement. Drums and other percussive instruments usually go together. Bass-heavy instruments form another stem and midrange melodic instruments such as synths, guitars, pianos go into one or two stems.
Of course, that depends also on the complexity and density of the arrangement. Solo instruments would likely get their own stem, as will effect returns, backing vocals and the lead vocal.
This should produce somewhere between 4 and 8 stereo files that are to be sent to your mastering engineer. If you're doing your own mastering, importing the stems to a new project to separate your workflow from the mixing project might be helpful to invite a different mindset and focus.
So why go to all that trouble?
An example
If you have a room mode (resonance) that nulls out frequencies around 200Hz in your listening position, there's high chance you will add too much snare fundamental frequency with an EQ to make it nice and fat.
But when you were recording your guitars, you were wearing your headphones! You were not affected by the room nodes, so your mic placement and amp settings were spot on for the low end. You know you nailed that guitar tone and really like it, so you don't change it while mixing and end up with a great guitar tone and a boomy snare.
Classic stereo mastering would have to make a compromise. Cutting a bit of 200Hz on the whole track will make the snare sound a bit more presentable and not cause the whole track to duck once it hits the stereo compressor and limiter, but you will loose some of the guitar body that gives the weight and energy to the track.
Stems on the other hand will allow the mastering engineer to apply the cut on the drum group only, leaving the guitars untouched! Nice.
How to prepare for it?
1. Decide upon your groups (stem outputs)
This is a nice excercise even if you aren't planning to send the track to mastering in stems! This will help you acknowledge the relations in the song's structure and frequency spectrum, so you can make better EQ decisions to fit everything together without "spectrum clogging".
Try to identify the layers of your mix, frequency-wise and content-wise, and group them into a stem!
A drum kit or a rhythm-machine is pretty obvious, but what about the percussive additions like tambourines or shakers? Do they follow the primary groove and fit into the spectrum between the drums and the cymbals? Or do they follow the keys or guitars?
Another simmilar question would be: "Do the synth pads and the piano play the same theme in different parts of the spectrum? Or does the pad follow the bass line, while the piano balances the guitar part on the other side of the stereo image?"
Try to identify as many of those relations as possible and put similar content together in groups. Next, choose a way to export those groups separately in a way that best fits your workflow and your DAW.
You can either solo them and do a normal stereo bounce/export, or you could route them to group/aux/folder* channels for a faster, simultaneous export.
*The naming of this type of summing track depends on your DAW, as does the exact export procedure.
2. Stereo bus processing?
It's common practice to use a master bus compressor or EQ (or both) while mixing. Those processors usually have a big impact on your mix and if activated early on, inform all of your subsequent mixing decisions!
While EQs can have enough headroom to not work any differently on a single stem than on the whole mix, that most certainly can't be said for compressors!
So for the stems to sum perfectly into a stereo mix, the gain reduction has to stay the same regardless of the material passing through the master channel to be bounced as a stem.
In The Box (ITB) Stem Mastering
For this purpose, some clever routing is required. In the digital/"in-the-box" workflow, one way is to make a stereo bus/aux that all the tracks feed into, independantly of the routing to the master bus.
The sends to this bus should be set to "AFL - after fader level" to always contain the whole mix with the appropriate levels and all the automation. Finally, use this bus as a sidechain input to the master bus compressor.
This way, it will always react as if the whole mix is passing through it!
Analog/hybrid Stem Mastering
Following this technique in the analogue domain is techically the same, but can be a bit more challenging. Not all analogue stereo compressors have sidechain inputs and even those that do aren't necessarily in stereo, which can mean additional trouble.
A free aux send on a mixing desk can usually be found, but summing boxes rarely include them. All that considered, it's sometimes easier to inform the mastering engineer about the choice and the settings of you master bus compressor, so he can try and replicate it with his own setup.
3. Exporting
With all the routing figured out, all you have to do is bounce the grouped tracks to stems! I would recommend exporting in .wav format, original sample rate of the project and 24- or 32-bit sample depth, but it can't hurt to ask your mastering engineer for preference.
I may be stating the obvious to many of you, but still - proper naming of the tracks is almost invariably going to translate into better service! The mastering engineer importing your files is going to nod in approval with a smile instead of muttering curses into his beard, and a happy person always does a better job!
Conclusion
Stem mastering can be a life-saver, especially if you're struggling with a dense and complicated mix on a less-than-perfect monitoring system. True, it usually comes with a higher price tag and it's still not a solution to a bad mix, but if in the right hands, it can do miracles!
Stem processing is also where mix:analog can be very useful. Using the quick access slots for fast switching between stems and imbuing your tracks with the analogue magic can be a most satisfying experience that yields great results fast and effortlessly.
If you haven't already, please give it a try- the free signup offer should suffice for a 30min session with some of the coolest gear around!