Even though this article is entitled Mastering 101.1, it’s not an entry-level piece that gives you a grounds-up viewpoint.
Instead I wish to just be quite honest about
the tools of mastering and how to approach their use, to figure out what to use, and most importantly when not to use them.
Rules of audio
My dear friend David Collins, formerly of A&M Mastering, now with Steven Marcussen Mastering in Hollywood, often quotes that audio engineering is easy, you simply “move the knobs ’til it sounds right.” It’s not his quote, but he’s the one who’s said it the most to me. I’d like to amend the rule by adding at least one rule AFTER that one, which is simply this: stop.
Most people don’t do that. They move the knobs around until it sounds right, then keep moving them, as if they can find something better. This is a critical mistake. They forget to “get out of the way” of the material.
Therefore, my rules of audio engineering are as follows (with #1 not being applicable to mastering):
1. Move the mic, not the knob.
2. Move the knob until it sounds right.
Of course there can be more rules and sub-rules, such as “pick the right mic/preamp/ feng shui,” etc., but these will work for a top-level guide.
Same goes for mastering. “Have a light touch and get out of the way,” says Bob Ludwig of Gateway Mastering Studios. For me, mastering is the art of revealing the true sound that is hidden there, the sound that could be heard with the proper use of tools.
This is true even if you have to do a lot more fixing than finessing, although everyone prefers to master mixes that don’t need a lot of fixing. “Fix it in the mix” has been replaced by “fix it after the mix,” which is simply not a good idea.
Worse yet, too many mixes are “pre-mastered” by some preset on a mastering box, just to have a loud copy for the band to cruise around with. That’s perfectly okay to do as long as you keep an original copy that hasn’t been run through the box.
You might be thinking this is not such an original idea, but it’s surprising and sad to find out how many engineers don’t keep copies of original mixes at all. Now that many people mix straight to hard disk in their computer, they love to normalize, compress, eq, and maximize their mixes without working on a copy (have you done this?). All you have to do is save the original mix just as it is, make a copy, and do the “temp mastering” on the copy.
Three handy tricks
I don’t pretend to be giving away any secrets here. The very small industry of mastering does not have a code of ethics as the magicians do, merely because mastering is hardly a technical skill. Instead, it’s a skill of the ear. You can be told and shown how to do something, but to have the sensitivity and discretion to do it is another matter.
This is my disclaimer, just a way to say that you may indeed try these at home, but your mileage will vary no matter what happens. Mastering is largely a delta variable moving toward the ability to take a subjective experience and have an objectively based point of action. Of course, there is no objective point of view, just something that is close to it.
#1: de-esser + 2 eqs:
Believe it, there’s something beautiful about multiple eqs and de-essers. But beware, it’s a recipe for high-speed disaster if you enjoy changing values around a lot.
In this example, using an eq, then a de-esser (splitband type, such as the Waves C1), then another eq, gives you control of air and ess, whether you can handle it or not. The de-esser can not be a wideband device, obviously, so you need a “multiband” device that has only two bands!
It is not good to use a 4-band device when you really only need one band because of all the unneeded crossovers. Even with a phase-compensated device, there will be an increase in peak values that is reduced if you use a “directable band” device, which I termed a “parametric compander” in 1990. (If someone else came up with that term prior to 1990, I am unaware of it and apologize to the Archetypal Karma Center.)
Did I mention that your de-esser and any other splitband kind of stuff should be last in your chain, next to your final limiter (if you use one), and wordlength reduction processor (dithering gadget)? In other words, after any shaping eq or compression or stereo re-imaging, or whatever else you do that is broadband.
Here’s the signal flow, and of course everything is optional:
broadband stuff --> eq1 --> de-esser --> eq2 --> limiter --> wordlength-reducer
So take your eq, your Parametric Compander, and your other eq, line ’em up with unity gain throughout (I’m not going to tell you how to handle headroom, sorry), and try this series of steps:
• Adjust 1st eq to get the air of the overall song, and ess level you want of the soft stuff. Ignore the loud esses for the moment.
• Adjust the de-esser to control the loud esses and other sizzlers in the mix. May I suggest a ratio of something like 3:1, attack of 25 ms, and release of 5 ms, basing my numbers from the Waves C1c/sc (which is the sidechain with splitband).
• Tweak the 2nd eq to fix any little holes in the resulting response.
#2: kick/bass remix
Fix it after the mix—that really is a most inappropriate way to do business, but unfortunately, at most ends of the recording industry, we don’t have the power to force people to re-mix the CD if it really really really sucks. As a matter of fact, the Top Mastering People can’t force re-mixes either, because some clients are just too big to handle that way. So be it.
However, fixing after mixing is needed, so here’s another tooltip. You can use a parametric compander to control the bass and kick drum mix, somewhat, so that your client grovels on your feet—or at least says “Oh My Gosh How Did You Do That,” which is usually preferable to foot grovelling in most cases.
We’ll solve a few scenarios, not just one. The setup is nearly the same. Plus I’ll break one of my suggested rules from the first example, just to illustrate how there are no rules, but there are sensible practices that conflict with other sensible practices.
Here’s the setup. Use a parametric compander (aforementioned C1) or at least band 1 of your multiband tool, such as TC Finalizer, dbx Quantum, Waves C4, or MasterX. I find it is rather important to use only the tool you need, not simply to keep the same stuff patched up all the time or (ugh) use a template for your setup. This example will assume you have the C1 or something like it.
Kick too loud
Let’s break my rule first; although I wrote to do the broadband stuff first, sometimes it’s not right to give your broadband compressor (IF it is needed) some mix that has a kick that is 6 dB too loud! If you need to ask why, please think a bit. Therefore, it might be best to put your De-Kicker as the first thing in line, before your broadband chain.
Supposing this is true, here’s the chain:
De-Kicker --> broadband stuff --> splitband stuff --> limiter --> ditherbox
If this is a little confusing, then you’re missing my point about the attempt to be objective about such a subjective thing as the ability to hear. Anyone can listen, few can really hear. Okay. I’m jumping off the soapbox for just a few paragraphs, but will hop up again; it’s just my style.
First, let’s suppose the kick is too loud and the bass guitar is just right. What is needed is to simply push the kick back into place, and although this will change the nature of the kick sound, it is much preferred. I’m using the settings based on longterm experience with the C1, which is a reasonable benchmark because it is a manual device without ‘smart’ time tools such as auto-release.
Try a ratio of 4:1, lowpass shelf with crossover at 110 Hz, attack time of 70 ms, release of 50 ms. Bring the threshold down until you start squooshing the kick down a few dB. If you have an input meter that shows actual energy after the crossover, then you can see the kick transient and the bass guitar sustaining parts clearly. If you don’t have such a meter, then ignore the previous sentence.
You might find that a bandpass (“bell”) centered around 90 Hz to be a good starting point, depending on the frequency distribution of the kick sound. What you want is to control the kick with minimum adverse affect to the bass guitar/synth/sample.
Then lower your attack time until you really just about kill it completely (such as 1 ms). This will be too far, for sure, but you’ll have an idea of what you can do just with the attack time. By balancing the threshold and the attack time, you can re-shape the amount of kick transient, kick punch, and kick foreground/background.
Of course these are just starting points; there are thousands of settings that won’t work, and it’s your job to learn to hear and avoid them. It’s impossible to teach someone to hear, but it’s definitely possible for you to teach yourself to hear.
Finally, adjust the makeup gain of the compressor so that you put the right amount of kick/bass into the mix. If the bass guitar was already okay, then chances are the makeup gain adjustment will be small.
Kick too soft
I hate when this happens, but it’s possible to get something that works better. Assuming that the bass guitar is still at a good level, then adding some new transient to the mix is okay. But it’s much trickier than the previous example. The worst scenario is that the bass guitar is way too loud and the kick is way too soft; that would be the time to try to get a remix—or price yourself out of the job.
Use the same setup, but flip the parametric compander to the expander mode (ratios less than 1:1) so that it will actually expand the transient of the kick drum (and the bass).
What is needed is just the perception that the kick is increased (perception is reality), without making the bass guitar jump around too much. Fortunately, most of us are acclimated to hearing a bass and kick tied together, either because the players are great or due to a linked gate to tighten up some loose pickers. The effect is similar because the kick will generate the new transient and take the bass guitar with it.
Try a ratio of 0.7:1, lowpass shelf with crossover at 110 Hz, attack time of 15 ms, release of 200 ms. Adjust the threshold so you get some upward expansion each time there is a kick. You’ll have to juggle the attack, release, and threshold carefully to get exactly what you need, IF it is possible.
Did I mention that part of mastering is the ability to STOP using a tool when it won’t work? Oh yes, this is a critical part of audio engineering: the ability to quit when it’s right AND the ability to quit when it’s wrong.
Bass needed, insert comp
Dead strings, poor players, and inept engineers all seem to conspire on the same project. The kick and bass are fine, but the bass guitar dies out too fast, doesn’t have that nice firm earthiness you want.
It’s the same type of setup as the previous situations, but more subtle. Using the multiband device or parametric compander, the attack will be set not to affect the kick too much, but the release time will help pull the sustain out a bit.
A bass line, no matter what instrument plays it, usually has a range from about 41 Hz to 164 Hz, insofar as the fundamentals are concerned. Therefore it’s the best area in which to play that will avoid mucking up the guitars, vocals, and other things in the mix. You can use Waves MaxxBass to play with those frequencies and compress them (there is an upward compressor in the original Waves version), or you can use the parametric compander mentioned throughout this article.
Try a ratio of 2:1, lowpass shelf with crossover at 160 Hz; attack time of 200 ms, and release of 50 ms. This has a big window to allow the kick to go through, and a quick release time so that when the bass guitar mutes a string, the compressor recovers quickly and is ready for the next transient.
Pull the threshold down until you see steady compression on the beginning of the bass note (and NOT on the kick drum), say at least 4 dB of gain reduction. Adjust the attack time so that you are really letting the kick drum come through enough, and set the makeup gain to something around +3 dB so that you are bringing the entire low end up in level.
The compression pushes the bass down at the start of the note, and as the sound dies away, the compressor opens up and lets the sound sustain a bit better—in this example, about 4 dB better. Of course, attack and release times that are too fast in these low frequencies will cause distortion, but you knew that already.
#3: upward compression
Sometimes I’ve heard this achieved by something referred to as sidechain compression, but that’s confusing. ’Tis better to call it side-bus compression or something like that, because it takes a dual path to do it properly, or a processor that has this as part of the design.
Here’s how it’s done in a traditional mixing technique. Take the kick and snare, send them to the LR output (as usual), but also send them to a bus that has an insert comp on it. Smash the living daylights out of the snare and kick, then return the output to the console and “mix in just a little” of this smashed signal with the straight mix. (Did I say this is very very easy with all-analog studios and a complete pain to do in digital, due to plug-in processing delay?)
What’s happening is that the signal is made very dense, then some of the density is mixed into the signal, raising the low level material upwards without affecting the transients, i.e. Upward Compression, i.e. compressing the low stuff upwards, etc. This is somewhat like taking a cupful of sauce off the heat, making it seriously thicker with arrowroot, then stirring it back into the pot until the desired thickness is reached. Better control and more accuracy is possible, if you can cook.
Again I’ll use the Waves tool as an example because it has a very clean implementation of precisely this concept. Use a C1 gate (aha! not a compressor) in wideband mode (at least to start with).
Set the C1 gate into the Expander mode, set both thresholds (gate open and gate close) to somewhere around -20 dBFS (assuming reasonable levels of average pop material; of course it is source-dependent). Try an attack time of 5 ms, and a release of 70 ms. (These will have huge control over how this behaves, so this is just a starting point).
Now the fun part: raise the Floor control up to about +5 dB. Tweak the thresholds so that you see a gain increase—yellow upwards meter movement—in the soft parts. You’ll notice that when peaks do occur, the yellow disappears to unity gain. Voilŕ! Increased average loudness without any limiting, without any compression of the transients! But you’ve also raised your noise floor, caveat emptor. Que sera et cetera.
Don’t want the bass parts to affect this neat trick as much? Use a splitband expander with a positive floor setting, while the sidechain eq is set to a highpass at around 200 Hz. I realize this may sound a bit confusing, but just suffice to say that the expander functions as an upward compressor when the floor setting is positive.
For more of my subjective writing about the attempt to be objective, check my stuff at www.prosoundweb.com, and my mastering course available from www.coolschool.com.
Seva is a composer, a team member of audio tool designers for more than a decade, the associate founder of Waves, a mastering engineer (www.soundcurrent.com), a pianist, owns a museum full of equipment and instruments, teaches Sound Art at the University of Tennessee, and is only 887 years old.