UPDATE 1/18/12: The graphic and article have been updated with a more accurate explanation of limiting.
Last night, I was giving one of our mastering clients some mixing advice regarding vocal compression. I thought I’d post the email here for those of you who might still be struggling with the concept (I know I did for years) or to refer your friends to. Also note that this article is about sound compression or even more specifically “dynamic range compression” and should not be confused with data compression (mp3, zip, rar, etc.) This isn’t a comprehensive explanation of compression, you can find that elsewhere. This is just the bare bones essentials.
I used drums instead of vocals in the diagram below because I felt that would be the easiest to understand visually. Click for full size:
What’s the point of compression anyway?
Back in the old days, vocal compression was used for one purpose:
Compression makes the volume of a vocal more consistent overall. In fact it was originally called “Automatic Level Control.” So if you’re singing or rapping some words louder than others, compression makes for a less drastic volume difference between the loud and quiet parts.
So above all else, vocal compression makes the lyrics easier to understand and helps to keep the words from being drowned out by the music. Obviously you could just turn up the vocal track to solve this problem, but then you’d have a new issue: when the singer hits a loud note it will now be way too loud compared to everything else, startling the listener and causing them to turn it off because they don’t like being startled. Or more likely they just turn it down. And you want people to blast your music right?
How compression works.
It’s pretty simple: Every time the compressor hears a sound that goes over a certain volume level (that’s your “threshold”), it turns down the volume for that moment.
After compressing a vocal track, it will be quieter than what you started with. To make up for this lost volume, most compressors have an output volume knob or “makeup gain” knob. (For our purposes, gain and volume are the same thing.)
So although technically compression turns down the loud parts with the quiet parts unaffected, if you raise your makeup gain enough you’ll effectively be turning UP the QUIET parts instead.
As I said before, your threshold is setting the cutoff point. Any word that’s quieter than the threshold will remain untouched. Everything above it will be turned down somewhat. For rap, you probably want to set the threshold so you’re compressing everything but the absolute quietest words. Until you get to the point where you can actually hear compression working (this can take years), you’ll want to use your gain reduction meter for this purpose.
How you set your threshold will be relative to how loudly or softly your vocal was recorded. That’s a big part of why presets are only meant to be used as a starting point. The truly important value to watch is gain reduction, because a gain reduction meter tells you how much your vocal is actually being turned down at any given time.
If your compressor doesn’t have a meter, it’s hard to say where you should set your threshold. As I said before, compression is very hard to hear starting out. And your listeners will only hear it subconsciously. Or they might notice when it’s not there because it sounds like a demo and it’s not as smooth/consistent as a professional mix. But they won’t know why it sounds that way.
After using compression, the loudest notes will still be louder than the threshold, but not as much over it as before (see the diagram above). So post-compression, the difference in loudness between the loud and quiet words will be smaller. That difference is called “dynamic range” and the ratio determines how drastic that difference is. High ratios (8:1 for example) make for a smaller range. Small ratios (2:1 or 3:1) will allow a bigger range.
Compression Today and the Effects of Attack & Release.
Today, compression is still used to make lyrics more understandable, but back in the day it pretty much only had one knob. More compression, or less. Today you’ve got more controls, so you can have different flavors of compression using attack and release. At the risk of oversimplifying, here’s the effect attack and release settings have on vocals:
Shorter attack and release times (smaller numbers) will make vocals sound more “energetic,” louder, and will also bring out the breaths between words.
Longer attack and release times (higher numbers) will make the sound “punchier” which obviously isn’t as important for vocals as it is for other instruments like kick and snare.
For vocals, generally you’re going to want the attack to be quicker than the release. It’s not worth getting into why for now. I will occasionally set the release time to its absolute shortest value if I’m having trouble getting a vocal to cut through or going for an extreme effect.
What a Compressor’s Attack & Release Knobs Technically Do (Versus Synthesizers and Drum Machines).
If you’re used to working with synthesizers and drum machines, you may be confused about attack and release settings for compressors. Unlike on a synth, the attack and release settings on a compressor do not tell you the attack and release times of the instrument the compressor is affect. You’re setting the attack and release of the compression itself.
Compression doesn’t turn on and off instantly. So attack is the length of time it takes your compressor to start compressing at full force. (How long it takes to go from zero to sixty, so to speak.) Release is the amount of time it takes for your compressor to stop compressing.
So what does all this mean in layman’s terms? Let’s say you’re compressing a snare drum and you’ve set a relatively slow attack time–80 miliseconds. And let’s also say you’ve applied this compression permanently to your sound wave and you’re looking at the waveform before and after. Visually, your snare drum probably already had a big volume spike at the beginning which trailed off, but now it will cut off even more abruptly since it was grabbed by the compressor after 80 miliseconds and taken way down in volume. You may hear this resulting sound as a punchier snare drum.
Now let’s say you set the attack to an EXTREMELY slow speed, like 1 second (1000 ms). You’re not going to be compressing the big pop of the snare drum much at all now because that pop lasts less than a second. We’re talking fractions of a second here. So by the time the compression kicks in you’ve already missed that spike.
If you want to control the attack and release of a recorded sound directly (like you would with a synth or drum sample) what you want is an “envelope follower” or “transient modulator,” such as Waves Trans-X, Envelope (built into Logic), or Oxford Transient Modulator. The most popular hardware version of this is called the SPL Transient Designer–for some engineers it was the secret to the 90s drum sound.
Limiting is the final effect in every mastering engineer’s chain but is also commonly used on vocals (most limiters are better suited to certain applications over others).
A limiter is a type of compressor which has an “infinite ratio”* which means that everything that goes above the threshold gets set to exactly the threshold’s level.
The controls on a limiter are also slightly differently from a compressor. Instead of using your threshold to push down the peaks and then compensating afterwards with makeup gain, for limiters you set the input gain, which pushes up the volume of everything at once, while the threshold is fixed at 0. (Go here to get unconfused about decibels.)
The end result is that although everything has been boosted by the same amount in theory, the volume of the quiet parts is effectively increased much more than for the loud parts, because the loud parts were closer to the threshold to begin with, and as I said before, a limiter doesn’t allow anything to get louder than the threshold.
If you’re working with vocals, you’ll probably want to use the limiter’s output gain control (if it has one) to turn the final result down.
A NOTE ABOUT LIMITING PLUGINS: Very confusingly, some of the most popular limiter plugins (including the Waves L1, L2, & L3) call their input gain control a “threshold” and use negative instead of positive values. In fact, the true threshold cannot be controlled at all–it’s fixed. Your L1 threshold is actually an input gain. The reason they’ve set it up this way is because it’s easier to understand visually. Unlike with most analog limiters, many plugin limiters have a vertical volume meter and use sliders instead of knobs. So the slider is positioned beside the volume meter, and you can drag the slider down as you watch the meter bounce in order to visually choose which peaks are limited and which are not. This allows you to use your eyes instead of your ears to decide how aggressively you want to limit something–do you want to flatten out the volume completely or only the loudest peaks? Or somewhere in between?
Feel free to leave questions in the comments.
*Some engineers consider anything with a ratio of over 10:1 to be limiting.