Compression is one of the most essential tools in mixing but often one of the most misunderstood. For many, compressors are a kind of black box but the operation and usefulness of this processor are actually quite simple to understand. In this article we break down the individual controls and parameters that make up a compressor using animated demonstrations of the concepts.
Try this exercise. Grab a drum loop and use a fader to take down the volume of every hit's attack, & I do mean just the attack – you need to snap the fader back to its usual level for the rest of the hit. Of course, this is difficult to do with any kind of accuracy or consistency – we’re talking about milliseconds here. One option is to abandon the impossible task of working on this in real time & instead manually shave off every drum's transient using automation.
This allows for detailed, granular control over shape & dynamics but it would take ages to completely go through & make subtle edits to a whole track. Is there some way to automate this process when an instrument is too "clicky" with attack to blend nicely in the mix? The answer is yes – compression. Compression has become a dirty word in the fallout from the “loudness war”, but when you understand it as an automatic version of the above process, it can be treated as a versatile utility as opposed to some dangerous effect that risks ruining your track. Think of compression as your personal mulit-armed android whose single task is to move your tracks' faders with superhuman speed, accuracy, patience, and foresight. The android's precision is incredible but its intelligence isn't - it has a limited, sometimes unintuitive vocabulary of instructions it can understand but will follow them blindly, even if they make no sense at all. How well this "android" works depends entirely on your understanding of the handful of instructions you have to communicate with.
threshold & ratio
Threshold is a critical control for compressors. In simplest terms, it determines the amount of volume a signal needs to activate the compressor. That is, it establishes a boundary that triggers the processor when crossed. Notice how the signal above the threshold turns red – this means that part of the amplitude is “fair game” for the compressor to work on.
It’s important to note that threshold does absolutely nothing on its own – it merely sets the point of “activation”. The amount of work or gain reduction a compressor applies is not determined by a fixed amount but instead a ratio. The ratio control determines what percentage of the gain that exceeds the threshold will be reduced at any given moment. Notice that the actual amplitude of the signal is irrelevant – it is only the amount that exceeds the threshold that matters.
In other words, any dB's above the threshold will be scaled according to the ratio. It’s easiest to just think of the ratio as a fraction – if the ratio is set to 4:1, ¼ of the dB over the threshold will be preserved (or ¾ will be reduced). A ratio of 5:1 leaves 1/5 of the signal uncompressed, a ratio of 1.5:1 leaves 1/1.5 of the signal uncompressed, & so on. A compressor with a ratio of inf:1 is called a limiter – it creates a “brickwall” ceiling so that no signal can pass the threshold.
That all being said, knowing the exact amount of gain reduction is somewhat unimportant since it is in constant flux because the sound feeding through the compressor is constantly changing volume. The most important takeaway is that the amount of gain reduction that takes place is primarily an interaction between these two parameters – threshold and ratio. "Has the signal crossed the line?", "if so, by how much?", & "what percentage of the offending audio should I take out?" are questions the compressor is constantly asking.
One thing to note about threshold and ratio (and compressors in general) is that they are entirely signal-dependent. If you’re used to scrolling through synth presets, you've probably developed a habit of looking for a pre-made patch as your starting point when loading in devices to your DAW. Compressor presets make you take a more active role. Say you have a "kick drum compression" preset. 4:1 may be a good ratio for most kick drums, but the designer has no idea how loud your specific kick is since it can literally be any level. Therefore, the threshold (& many of the other controls) are set somewhat arbitrarily & require you to make the adjustment yourself. I recommend resetting the threshold to 0dB when you load in a compressor to force yourself to set it intentionally every time.
makeup gain & knee
“So compression makes things quieter? I always thought it was supposed to make things louder!” is what you might be saying to yourself at this point. Technically, a compressor reduces the dynamic range of a signal, and it does so by turning the loudest points of the signal down. Imagine an uncut lawn – the shortest blade of grass is 2” and the tallest is 6”. The dynamic range is the difference between these, or 4”. When you mow the lawn down to 3”, any blades that were standing more than 3” tall will obviously be cut down to 3” but those that were shorter – like the 2” blade – will not be affected by the mower at all. In this case, the dynamic range of the yard has been reduced to just 1" (3" - 2"). When the grass grows by 3” (and assuming it all grows at the same rate), the tallest blades will all be 6” again but the shortest will only be 5” and the dynamic will still be 1” (a reduction of 3”). The lawn will appear taller now than it was before the cut, even though the tallest grass is still the same height.
The goal of compression is to make the quiet bits in a signal louder. It just so happens that we accomplish this in the two-step process of selectively turning down the loudest peaks before returning them to their original volume by raising everything all at once afterward. Just like the lawn, even if the loudest peak doesn't change its volume, the higher average volume will make the whole signal seem louder.
Many compressors have a control that does this automatically, called makeup gain. While this is handy for some workflows, I find it can make the compressor's function misleading. The volume increase may distract from the fact that I've taken away too much of the dynamic range. I prefer to add the gain back in manually so I can get a truer sense of what's happening.
Another control you might see on a compressor is the knee. The knee can be thought of as a sort of gradation of the threshold point. When a knee is set to 0, the threshold is binary. If the signal's volume is shy of the threshold – even if it is by .001 dB - no compression is applied. The moment the volume exceeds the threshold, the full ratio is applied. With a soft knee, this line becomes blurred. As the signal approaches the threshold, some amount of compression will be applied. If you have a 10:1 ratio, for example, the compressor will apply a 2:1 ratio as the signal approaches the line. When it crosses the threshold, more compression (but not the full amount) is applied to the signal, say 5:1. Only once the signal has gone well above the threshold point is the full ratio – 10:1 – applied.
The knee eases the compressor in, resulting in a more natural softening of the signal as opposed to the black & white hard knee. A good rule of thumb for compression is to not allow it to make audible artifacts in your audio. When the gain reduction is noticeably jagged, you may ease off the threshold or ratio, but it might make more sense to reach for the knee control.
attack & release
Not the Black Keys album. These settings allow reaction-based control over timing & largely determine the “sound” of the compressor. The attack is less of a delay & more of a gradual introduction of the gain reduction. Basically, it asks “how long should the signal be over the threshold until I’m sure I should fully clamp down”. It’s like if you had your car’s sunroof open and it starts to rain. You react as soon as you notice the droplets but rain is able to fall through while the window closes. The time it takes the sunroof to go from open to shut is the "attack" time.
Having little or no attack time on the compressor will remove short peaks in a signal, smoothing out the volume & removing the “clickiness” of dynamic signals like drums. A longer attack value will allow more of the sharp peak through the signal, allowing a significant burst of audio to pop through. Both values can offer valuable mixing applications - a slow attack will allow an instrument to stand out by "announcing" each note with a defined percussive pop while a very sharp attack will result in an instrument being more transparent & background.
When the parameters are all set right, a dynamic signal would be frequently crossing the threshold back and forth. If attack is a measure of how long it takes for the compressor to fully respond to a signal crossing upwards, release is like the opposite – how long does it take for the compressor to “shut off” after the signal is no longer above the threshold point. Many compressors have an “auto” release mode which determines the release time based on the content of the signal (how much does it exceed the threshold by, presence or absence of low frequencies, etc.) & this may prove a good starting point. It is also fairly standard to set the release time to be no longer than the most common interval between transients in the signal you are processing so the compressor fully shuts off by the time the next hit triggers it. This allows for a more musical effect, sort of equivalent to sync’ing a delay time.
Attack and release have a major influence on the resulting dynamics of compression & are quite versatile. As such, it's key to take special care & use intention when setting these controls. I like to start with the attack set to its slowest setting and the release set to its fastest setting. This allows all the transients through and shuts off the compressor immediately when the signal crosses back down over the threshold. Just like starting with a threshold of 0dB and a ratio of 1:1, this forces me to be deliberate with these settings.
Sidechain is a popular technique, especially in dance music & it’s really not as complicated as it sometimes may feel. Just think of the compressor as having two functions – 1) the analysis of a signal and 2) the application of those settings in the form of gain reduction. All sidechaining does is allow the compressor to accept a separate input for each of these. In other words, the "sidechained" source is analyzed by the compressor & the resulting gain reduction is applied to a different track.
The most common sidechain source is a muted four-on-the-floor kick drum. You’ll want the attack to be about as quick as it will go and the release to be about the length of the interval between drum hits (in this case a quarter note). Applying the gain reduction caused by these settings to just about any other track in the song will make it feel more danceable & groovy.
peak vs rms
The way we’ve discussed compression so far is in an instantaneous, momentary world of peaks. RMS (Root Mean Square) is an alternate means of volume evaluation that quickly averages a signal’s amplitude at regular intervals. RMS is a popular method of visual monitoring because it more closely simulates how we actually perceive volume (we don’t really hear transients). It comes as no surprise, then, that RMS detection is a useful way for compressor to analyze loudness. Practically, this means that a transient, highly dynamic signal (like a closed hi-hat) will not be read as loud as a more sustained sound (open hi-hat), even if they have equal peak amplitude.
One feature that is particularly useful for digital compressors that use peak detection is lookahead. Lookahead delays the signal slightly so that the compressor may "peer into the future". When you think about how a compressor functions, even the best ones involve some degree of latency since it must first analyze incoming audio before reacting it. For 1ms - 2ms transients, this may allow them to slip through even if there is no attack time set. Lookahead negates this by allowing the compressor to analyze the signal before it reaches the circuitry that actually performs the gain reduction.
In a lot of cases, it can be useful to use two compressors in series so that each one has to do less work (& therefore are much more transparent). In this case, you may choose to use a peak compressor to shave off loud bursts and transients & follow it with a compressor that uses RMS detection to more broadly even out the dynamic range.