yep's Guide to Compression

I'm going to do this a little backwards, so bear with me. I'm first going to describe the theory of compression, the whys and wherefores, and then describe how to actually use the compressor. So if it seems like I'm skipping over stuff, be patient. This is a big topic.

Bass, kick and snare almost always call for some degree of compression in conventional popular music production. These are the most critical elements of the mix, and they are also some of the most prone to wide dynamic fluctuations. Modern popular music tends to have a very fat, full, saturated sound, with electric guitars, organs, synthesizers, multi-layered vocals, and other elements that create a "wall of sound." Drums on the other hand have a very "spiky" dynamic sound that can sound wimpy behind these fat, flat, modern electronic instruments, unless you either turn all the other instruments way down or find some way to fatten and even out the drums. Similarly, bass can be difficult to control in a dense mix. The bass is notoriously prone to wide swings in level from one note to the next, in part simply because of the nature of the instrument, and in part for pshychoacoustical reasons that are a topic for another thread. This can lead to the infamous disappearing/reappearing bass line or the equally notorious bass note that would be king. Compression offers a solution to all of these problems, and also offers some opportunities for creative sound sculpting that can really make your music move and groove the way envision it.

There are two basic strategies for using compression that we'll take one at a time. the first and most basic use of compression is to transparently even out the dynamics so that the instrument sounds the same as it did pre-compression, but it maintains a more consistent overall level. "more consistent overall level" could mean a few different things. It could mean that each individual note or drum hit maintains its original dynamic profile, but that the level is more consistent from one note to the next, or it could mean that the individual notes themselves are modified for instance to reduce the peak level so you can bring up the "body" or sustain of the sound relative to the other instruments in the mix. Both of these goals can be achieved with great transparency using modern digital compressor plugins including the Sonitus compressor.

By "transparently," I mean that compression is applied in a way that does not perceptibly change the "sound" of the instrument. This means that using this kind of compression can be somewhat tricky for the novice, because unlike say reverb or distortion the compressor can be hard to hear if done properly, even with quite heavy compression. It doesn't "sound" like an effect. In fact, that's the point-- to make the sound of the compressor transparent. If you can hear the compressor as an effect, you're doing it wrong.

With practice, you learn to sort of "feel" the way the compressor affects the breathing and pulsing of the music. But when you're first starting out, it is critical to use your meters. You need to look at both the track's output meter (the cakewalk "track" meter, preferably set to RMS+peak), and the compressor's gain reduction meter. Make sure you are looking at the gain reduction meter on the compressor-- many have a meter that switches between showing different things. You will use these meters in conjunction to see what you're doing to the sound. I'll describe later how this actually works, but for now I just want to cover the theory.

The second type of compression is what I'm going to call "saturation"-style compression. This can sometimes be achieved without even using a compressor. When you overdrive certain high-quality analog audio equipment, there is a gradual compression and "soft clipping" that occurs before the ugly, distorted clipping sets in. This is especially true of tube equipment and analog tape. The effect is that peaks are limited, and a warm harmonic "fire" or "fullness" is added to the sound, along with a hyped-up sense of detail. An extreme example is of this phemonenon is guitar distortion, and you may actually be surprised at how effective a distortion pedal can be when used on drums in a dense mix. You can often apply surprisingly heavy amounts of distortion to drums before they start to sound noticably affected.

This type of saturation can be achieved with a variety of special "vintage" or "warming" style compressor plugins, and also with any number of dedicated effects or hardware components, including some effects that aren't really intended as compressors at all. This is an area to get creative and play around. Saturation effects can be a subtle, musical way to add new colors and textures to your recordings. Blockfish by digitalfishphones is a good free plugin that can get you started.

Note that whatever type of compression you use, you should make your compressor settings while listening to the whole mix, not to a soloed track. Compression that sounds extremely transparent and appropriate while the whole mix is playing can sound weird and unnatural when you solo the instrument. Make changes in small increments and use the bypass feature of your compressor as a gut-check. Be wary of the siren song of ever more compression. Overcompression can make everything sound slamming, hyped, and loud, but only for short bursts. After a few seconds, overcompressed music becomes fatiguing to listen to and your audience will turn down the volume, defeating the purpose of the compression in the first place!

If you're just starting out, I would recommend tweaking the compressor so that the gain reduction meter "bounces" only two or three dB on most notes, tracking the tempo of the music fairly closely. Then turn up the makeup gain by about the same (2 or 3 dB). Note that this may sound like nothing has happened--it takes practice to hear mild compression. Then stop playback, rest your ears for a few seconds, and BYPASS the compressor before you hit play. Now listen to the uncompressed track and then un-byapss the compressor and see if it's an improvement. The track should feel a little punchier and a little more "present." If it needs more, try and get the gain reduction to bounce 5 or 6 dB the same way, and evaluate your work the same way. Take baby steps and go carefully and it will pay off quickly.

Now the nitty-gritty. Below is something I posted on an earlier thread about how to actually use a compressor and what the different controls are for. Hopefully it helps:


You really need to read up on this and practice a little until you understand how compression works. Every setting on a compressor affects every other setting, and they are inter-dependent. depending on how your other settings are, adjusting, say, the attack setting up or down can have opposite effects.

There are tons of explanations all over the web and in books of how compressors work, but none of them are very clear until you already have a fairly good handle on certain things.

First piece of advice: avoid using your hardware compressor before recording unless you really feel 110% confident in what you're doing. If you record at 24 bit, there is no advantage to compressing before recording, and software compressors will tend to be better than your hardware compressor. Moreover, if you get it wrong at the tracking stage, there is no way to fix it or undo it.

Second piece of advice: Understand the following before you start using a compressor. Depending on how you use it, a compressor can actually INCREASE the dynamic swings in your program material. Use of a compressor DOES NOT guarantee flatter overall levels. This is very important to understand.

Here's how a conventioanal four-knob compressor works (there are also two-knob compressors, but that's another story). You have four basic controls: attack, release, threshold, and ratio (there is also usually a "makeup gain" knob, but that's seperate from the action of the compressor, it's just a gain control that comes after the compression circuit. We'll deal with it later).

Before we start to talk about what those four knobs do, it is necessary to understand how an audio waveform develops. When you pluck a guitar string, or hit a piano key, or strike a drum head, several things happen to produce a complex waveform. I'm going to use an acoustic guitar as an example, but the following applies to almost any instrument, in principle.

If you look at the waveform of a plucked guitar note, there are typically three (maybe four) distinct parts of that individual note that look like different sections in a visual representation (in sonar's audio clips, for instance).

First there is the inital transient (the "attack" of pick on string). This typically looks like a sharp, almost instantaneous "spike" in level, and is typically the loudest part of the note.

Second, there is the main "body" of the note, the soundboard of the guitar resonating with the vibration of the string. This tends to have a fairly even level, and looks on the computer screen kind of like a brick of sound energy that comes after the initial transient spike. The duration tends to depend mostly on the guitar construction and playing technique. This is the "steady state" portion of the sound, the primary "sustain" of the note.

Third, there is the "decay" or "tail" of the note. This typically looks like a rapidly narrowing triangle attached to the end of the "body" of the sound. It is the section where the string starts to lose energy and the soundboard is just vibrating from inertia. it overlaps slightly into the time where the guitar has stopped vibrating altogether, and the final tail of the sound is simply the soundwaves still bouncing around in the room.

The phantom "fourth" section is "silence." Unless you record in outer space, there is no such thing as actual silence, but the quality of your "silence" definitely matters for purposes of compression. We define silence as the point where random noise and room mode resonance takes over and is as loud or louder than the sound of the actual note.

Got all that? good. Here's how it matters to your compressor. Your compressor has a little gremlin in it that turns down a volume knob. That's all. Really. A compressor is just a little guy that changes a gain control really fast. Here are how you give him his instructions:

Threshold-- this is like an alarm clock that wakes up your little gremlin. If you set your threshold at -6, and your incoming audio signal never goes above -6, then your little gremlin never does anything and sleeps through the whole thing. Once you threshold goes above -6, the gremlin wakes up and springs into action.

Attack-- If the threshold is the alarm clock, "attack" is the gremlin's "snooze" button. The "attack" is a delay between the time the signal goes above the threshold, and when the gremlin actually starts doing his job. For instance, if your threshold is set at -6, and your attack is set at 20ms, then what happens is this-- when your audio signal goes above -6dB, the gremlin groans, rubs his eyes and hits a snooze button that lasts for 20ms. After 20ms, the gremlin actually wakes up and starts doing his job. This setting allows 20ms of "attack" (aka "transient," aka "peak") to sneak through before the signal gets compressed. So if you wanted to flatten the whole signal, you'd set the attack to 0ms. If you want to keep the "impact" or the pluck of the guitar string intact before you start squishing the sound, then you simply tweak the "attack" setting to taste.

Ratio-- this determines how much the gremlin turns down the volume. It works in relation to the "threshold" setting. So if you set the threshold to -6, and the ratio to 2:1, then the gremlin will turn down the volume by 1db for every 2db above -6 the input signal went. So if the input signal was -3dB, the gremlin will turn down the signal by 1.5dB, for an output level of -4.5. (Ratios above 10:1 or so can basically be consiered a hard limiter)

Release-- this is the gremlin's quitting time. If you set the release time to 0ms, the gremlin quits and goes back to sleep as soon as the incoming signal drops below -6db (or whatever the threshold is set to). If te release is set to, say, 50ms, then the gremlin keeps working for 50ms AFTER the input signal drops below -6, which can lead to smoother-sounding tails.

So here are some examples:

If the threshold is set ABOVE the level of the BODY of the sound, but BELOW the level of the TRANSIENT attack of the sound, and the attack/realease are set very fast, then the compressor will basically work as a peak limiter, only compressing the initial transient attack.

If the threshold is set below the level of the BODY of the sound, and the attack and release are set somewhat slower, then the slow attack time will allow the transient sound to pass right through and will then compress the body and "tail/decay" of the note. This will create MORE initial impact when you apply makeup gain, and the slow release time will create a longer, more gradual sustain. This is the most "dramatic" way to set your compressor.

If you set a fast attack and a slow release with a low threshold, you can actually create a sort of "inverse" waveform, where the attack is compressed and the "body" of the sound sort of "swells" after the initial sharp compression releases.

If you have very aggressive compression settings with a too-fast release time, the compressor will "let go" too soon, and the decay and silence will pop through as amplified background noise between notes, causing "pumping" or "sucking" artifacts. On certain kinds of dance/club/electronic music, this can be a cool "throbbing" effect on bass or kick or synth pads. On most kinds of acoustic music, this sounds awful.

If you set the release time slow enough, the compressor will overlap onto the next note, and reduce your transients even if you set the attack time to let the transients through, again creating a "pumping" effect, but with a slightly time-dragging feel.

Depending on how you set the compressor, you can turn a sequence of steady quarter notes into short, choppy, staccatto hits, or into a smooth, pulsating pad, or make the notes sound like they're pushing or dragging the beat. You can make a track sound punchy and slamming, or mellow and smooth.

Listen to how the compressor affects the way the music breathes and pulses. Listen for the changes in performance "feel." Don't try to use the compressor as an automatic volume control-- use your faders for that. Don't use a compressor to limit stray digital overs on the incoming signal-- use a limiter for that, or better yet, just turn down your record levels and deal with your levels in the software.

Finally, and THIS IS THE MOST IMPORTANT THING TO KNOW, never rely on "presets" or "recipes" for compressor settings, unless you're simply using the compressor as a hard limiter. Correct compression settings depend of the tempo and dynamic profile of the incoming signal, and on the desired dynamic profile. THERE IS NO WAY any preset-writer could have known this about your material. Presets may be useful for reverb, saturation effects, even eq sometimes, but almost never for compression. Anybody who tells you that the right compression settings for, say, electric bass are such-and-such is WRONG, unless they always record the same player and the same instrument at the same tempo every time.


Cheers.

Reprinted with permission February 2007; Author: yep