When I first started recording, I got an inaccurate understanding of compression stuck deep in my mind. I only really started to exorcise it a few years ago, and I still think the way compression is often explained is confusing. (Maybe I’ll post a follow-up on The Compression Conspiracy, or maybe it will come out here.) I’ll attempt to explain compression clearly and simply, touching on limiting and clipping along the way.
Compression is automatic volume reduction, or “variable attenuation” if you prefer. Compression is not soft clipping or peak reduction*.
* Peaks will be reduced in the course of automatic volume reduction processing, but since the whole signal is being reduced it’s confusing to call that result “peak reduction”
Compression is like having a gremlin enthralled to your instructions, whose hand is on a volume knob. You can tell the gremlin how fast to turn the volume down when your input signal gets loud (compressor “attack” time), how fast to turn the volume back up when the input signal gets quieter again (compressor “release” time), how loud is “loud” to you (compressor “threshold”), and whether you want your gremlin to turn down the volume a little, or a lot (compressor “ratio”).
The gremlin’s name is the “control circuit” or “sidechain,” a version of the input signal that is processed to detect volume, and therefore when to turn the signal volume down and back up.
|Automatic volume reduction|
or “variable attenuation”
|Base version||Compressor||Soft clipper|
(Theoretically a compressor
with an infinite ratio. In practice,
a compressor with a ratio of
about 10:1 or greater)
|Change in wave|
|Change in wave shape (timbre)|
and peak level
|Reaction time||Varies; usually a few milliseconds|
to a few hundred milliseconds
The results of compressors are often visualized incorrectly, so let’s start with an article that does it right; “Understanding Compressor Attack And Release Times” by Pro Tools Expert. They created the excellent image below, which I’ve scribbled on.
Notice how the attack time begins when the input signal initially crosses the threshold, and the first cycle of the signal passes through the compressor with no volume reduction (red circles). Also notice how the output signal retains its wave shape (nice and pointy at the peak); only the amplitude (i.e. volume, i.e. distance from the gray center line to the peak in the image) is reduced.
What has happened here, musically? The compressor has emphasized the initial transient (red-shaded “attack time” section) relative to the sustain or “body” of the sound (orange-shaded section). We might describe the result as having more “punch.” We might perceive the compressed sound as quieter than the input (notice the decreased amplitude of the wave in the orange-shaded section) and decide to turn up the volume of the signal across the board, after the compressor in the signal path, to compensate.
Here is the incorrect – but shockingly common – way to visualize the effect of compression. What is shown here is actually soft clipping. (If you go to the trouble of image-searching this you’ll see the author acknowledges such later in the article; kudos to them for that. Many people don’t!)
Why is this misleading? If you compare this to the first image, you’ll see that this image doesn’t show anything about attack or release time, and it changes the shape of the signal peak, which compressors don’t do! Extremely unhelpful. This is how compression was first explained to me, resulting in many years of confusion and frustration. If this post can help one person avoid that, I’ll be happy. (HI CORY)
Here’s a slightly better version (scribble in orange), though it still doesn’t explain anything about attack and release.
Getting from compression to limiting is a matter of ratios. Think about our gremlin again. For gentle compression we might tell them “For every two clicks louder than the threshold the signal gets, turn the volume down one click.” That’s a 2:1 compression ratio. Or we could tell them four clicks in, one click out; a 4:1 ratio.
Now imagine we tell our gremlin “Whatever happens, DO NOT UNDER ANY CIRCUMSTANCES let the sound get louder than the threshold!” That would be an infinite ratio; no matter how loud the input signal, our gremlin turns down the volume knob as much as necessary to make sure the output does not exceed the threshold. (It still takes the gremlin a few milliseconds to actually turn the knob down and up, so attack and release times are still in play, and the peak levels of the output signal might exceed the threshold in those milliseconds while the gremlin is turning the volume down.)
In practice, once we get to a ratio of 10:1 or 20:1 it’s roughly equivalent to an infinite ratio to human ears. We call compressors with high ratios and fast attack and release times “limiters,” but it’s all automatic (gremlin-controlled) volume reduction and there’s not a sharp line between what gets called a compressor and what gets called a limiter.
Clipping is changing the peak shape of a signal. (Image credit.)
As you can see the peak level is also changed as a result of the shape change – they’re two sides of the same coin – but it’s the change in shape that changes the timbre (i.e. tone, or quality) of the sound. Clipping is distortion of the input signal in a literal sense, and it’s also the sound of distortion in a subjective sense. The sound you imagine when I wrote “distorted guitar,” that’s the guitar’s signal being clipped.
Clipping is instantaneous; just like you see in the image above, it occurs on a cycle-by-cycle basis.
I want to stress the differences between clipping and compression, since they’re often mixed up. But one similarity between soft clipping and lower-ratio compression is that the output signal remains correlated with the input signal above the threshold level. Likewise, hard clipping is similar to infinite ratio limiting in that above the threshold, increases in input signal do not result in corresponding increases in output (with the previously noted exception for limiters’ attack times).
Another difference between clipping and compression is where the “threshold” sits in the processing. In clipping the threshold is in the signal path; it changes the sound signal itself. In compression and limiting the threshold is in the control signal path (i.e. it is a rule given to the volume knob gremlin). A compressor’s threshold setting does not change the signal itself directly, but rather controls the automatic volume reduction of the audio signal in combination with the attack and release times and compression ratio.
Long story short:
- Compression and clipping are very different audio processing effects. Compression is like a gremlin-controlled volume knob. Clipping is distortion (which can range from subtle or planet-destroying)
- Limiting is a type of compression (not a type of clipping)
- There is a lot of confusing information about compression online; think critically!