I've read and heard a significant number of opinions from mastering engineers on whether stem mastering is good or bad. Some debate that stem mastering is really mixing. Is it easier or harder to master stems, and is it worth it? First, let me elaborate a little on what stem mastering is vs. standard stereo mastering.
Stereo Mastering
Most mastering engineers work with stereo (Left and Right channel) mixes exported from the mix session. These are the same types of audio files you can listen to immediately as they are mixed. Stereo mixes are by far the most common type of mix a mastering engineer will start with.
Stem Mastering
Stereo mixes are essentially like a bouquet of flowers in a vase. Each flower is a separate set of instruments, like "guitars", or "drums". Imagine each flower has many pedals and each pedal is a single guitar drum, or vocal. That said, each flower also has a stem. You might have a vocal stem, guitar stem, drum stem, etc.
Stems can be in stereo or mono. Some stems can branch into multiple branches. For example, a stem might be "music" and another stem, "vocals". There's no real limit to exactly how summarized or individual a stem might be, but it's typically of a mixdown of a particular instrument or group of instruments. You might have a single stem that's "bass and drums", or you might have a stem that's a single lead vocal.
What's terrible about stems?
Certainly, there are engineers that believe a mix should only be the stereo mix (L+R) from the mix engineer. Anything outside of that means there's lingering disagreements or questionable qualities of the mix. This is not unusual, but to most mastering engineers, this is a crime--indicative of the mix engineer, producer, artist, etc. doesn't have a full idea of exactly how the mix should sound. It also leaves a lot of interpretation up to the mastering engineer, something that truly should never be done. Why should the mastering engineer have a say in the production that should be completed by the mix engineer? Is there a lack of competence before mastering? Is there lack of agreement? Is there doubt?
These are valid excuses, but the most frequent reason is to gain a second opinion, and just let the ME (Mastering Engineer) make tiny adjustments that could potentially help push the master just a little farther. At the time I'm writing this I'm working on a stem master. The reason isn't because there's a lack of competence or agreement. It's just that we're all a team, and we want to keep the mix open to adjustment at all stages. It's not common but it can be asked.
So, why would I condone a stem master?
While a mix engineer is listening to tiny details in a mix, there's always a chance the mix engineer might like to bring the ME to the table and have another pair of ears through which to offer sonic quality adjustments to the mix, because I'm listening to the audio spectrum, dynamics, punch, and vibe of the whole mix. Fresh ears also help.
More, and this can't be stressed enough, most believe that stem mastering gives far too much decision-making power to the mastering engineer. True. And...false! A stem doesn't automatically give the ME too much power. It affords the ME the ability to perform very necessary tweaks to each stem where needed, not free reign. If the ME is serious, and values the art as it was produced, nobody should worry that it will be destroyed. In fact, the mix should sound quite close to the original. See the illustration below. Here, you can see some automation performed in the mix for guitars. This is all done inside the stem (or bus) before I receive it. That isn't something tweaked by the ME. Those decisions are already made and I respect them always.
Stems are more like mix buses (busses). If I receive a stereo stem of all guitars, I couldn't directly modify the guitar mix on my own without mangling the rest of the guitars. I can only tweak the entire guitar ensemble. The mix engineer has gone to great lengths to mix the balance of all guitars, added all automation, and tweaked individual elements (EQ, compression, panning, effects, etc.) of the guitar mix before ever piping the guitars to a mixbus. For drums, all would be mixed and automation should be completed before creating a buss or stem of drums. All vocals and vocal effects and automation should all be done prior to exporting them all as one stereo stem. See what I'm getting at?
When I receive the stems they should line up and be at exactly the level they were in the stereo mixdown that the artist and producer are already listening to. I simply import the stems into a master session and start working on the stereo mix of these stems. Only if (!) I feel there's something that needs to be adjusted in the stereo master do I tweak a stereo mix. Same goes with stems. If the stereo master needs to be adjusted do I tweak the stems of the mix.
Can stems be exhaustive or require lots of tweaks? Sure. I've even received stem mixes that required a significant level of my input. Below is an example of that. However, this is an example of when it's not really mastering, but rather extending the mix session between multiple engineers. Rates are quoted for this type of collaboration, not standard.
Let's set a real-world example:
The mix engineer sends a stereo mix to the ME. The ME listens to the mix and suggests a few minor tweaks that could help the mastered result be even better. The ME suggests that the guitars are slightly dull but the vocals are sibilant than the mix engineer thought. To correct the sibilance might mean to dull the crash cymbals and hi-hat from the drums depending on how bad the sibilance is. The ME could request a mix revision before mastering. This isn't unusual---for the human ME, anyway. What's the difference between sending feedback to the mix engineer when the mastering engineer can quickly tweak them and move on?
A mastering engineer should work more closely to the mix engineer now than ever before.
I'm not suggesting that every mix be sent to mastering this way. If the stereo mix is perfect, then we prefer that mix. However, sometimes I might find a mix engineer is light on kick drums or snare, and my ears might feel they can be commercially more successful if brought out, slightly. While I do have some clever ways of fixing issues with weak snare or kick, I don't mind asking the mix engineer to supply a separate stem of the kick or snare. Again, this doesn't dictate I receive a raw kick or snare track, no. I would ask for the mixed stem of that. I want to use exactly what the mix engineer thought was best.
Even if I receive a separate kick or snare stem, I would only tweak it and mix it back into the master to the level where it makes just the right difference and no more. It's not about re-mixing the song. It's about addressing specific character of the mix in order to maximize the best mastered result, whatever that may ultimately be.
What you should know
I don't typically perform stem mastering. Like most ME's I prefer the stereo mix be done as best it can be. That said, there are times I feel it might be worth trying a few tweaks. Most of these tweaks go undetected, but the result is noticeable.
Should we engage in a stem master, I'll suggest what I believe will be best to "stem" out of the mix. Many times, I will request the stem be extra from the mix, not require the mix to be split up into stems. This is primarily why I say that stem mastering to some level is just fine. There's no harm in letting a mastering engineer, especially one who's got mix experience to adjust how a kick drum performs or possibly even to add the slightest hint of reverb to the drums or lead vocal. You'd be surprised at how subtle such tweaks can be, but might make an incredible master!