Filtering Down
Because the side‑chain signal that feeds the detector circuit is not actually in the audio path, processing can be applied to the side‑chain signal without directly affecting the main signal. This possibility is often exploited to apply equalisation or filtering to the side‑chain signal, weighting the response of the detector towards one area of the frequency spectrum. Probably the simplest and the most common application of this idea is to incorporate an optional high‑pass filter into the side‑chain.
The effect of this is to make the detector less sensitive to low frequencies, which often helps to align its action more closely with what we are hearing. Our ears are much more sensitive in the midrange than in the bass; so there are times when we want a compressor to even out the midrange, but it ‘hears’ the low end as being dominant and responds to that instead. That’s especially the case on complex material such as full mixes.
Warm Audio’s Bus Comp is based on the classic SSL design, and incorporates a variable high‑pass filter for situations where you want the processing to respond to midrange rather than low‑frequency energy.
High‑pass filters are found in many compressors. Some also feature additional side‑chain EQ controls, such as the Thrust options on some API compressors. One common type of de‑esser is simply a compressor with an exaggerated treble boost in the side-chain, emphasising the bursts of high‑frequency noise that make up sibilance and triggering compression. Taking this further, some hardware compressors have an insert point that allows you to patch in the EQ of your choice.
Taking Sides
Most DAWs don’t allow one plug‑in to host another as an insert, so you can’t actually patch a software EQ into a software compressor’s side‑chain. However, many plug‑in compressors do have an external side‑chain input or key input which allows a signal of your choice to be fed to the side‑chain and used to trigger compression. If you want to experiment with side‑chain EQ, you can duplicate the source to a second track, equalise it and route it to this input. But it’s also possible to use the side‑chain input for other purposes. In particular, you can route a completely different signal to the side‑chain input, and this is the basis of numerous compression techniques.
Softube’s FET Compressor is one of many plug‑in compressors that can accept an external side‑chain signal.
This type of side‑chain compression can be used to create an effect that has been a staple of electronic music ever since French house became a thing. It’s achieved by placing a compressor across the master bus, with the side‑chain input fed from the kick drum track. Every time the kick drum hits, the entire mix is compressed, creating a ‘pumping’ motion. Other well‑known applications for side‑chain compression include ducking, whereby instrumental parts or reverbs and delays are ‘pushed’ into the background by a vocal, then allowed to swell again in the gaps.
Up, Up And Away
You’ll sometimes see compression described as a means of making things louder. On the face of it, this seems confusing: as we’ve seen, the fundamental job of a compressor is to turn things down when they get too loud! In fact, there’s no contradiction. What most compression settings do is to reduce the crest factor of the signal: by turning down the loudest peaks, they reduce the ratio between the peak level and the average level. If we then turn the compressed signal back up again so that it peaks at the same level as the uncompressed signal, the average level will now be greater, and hence it will sound louder. All compressors have the means of doing this built in, usually through a control labelled make‑up gain.
Knees Up
As obscurely named controls in music technology go, soft knee and hard knee are right up there. They take their name from a common graphical representation of compression, and relate to the behaviour of signals around the threshold level. In a hard‑knee compressor, the transition from doing nothing to full‑ratio compression takes place as soon as a peak reaches threshold level. In a soft‑knee design, the process is more gradual: compression at a low ratio begins to take place as the signal approaches threshold level, and the ratio steadily increases as it goes further above the threshold. (Dbx coined the alternative term ‘over easy’ to describe soft‑knee compression.)
In these two shots of the Compressor plug‑in bundled with PreSonus Studio One, the effect of different Knee settings is visualised. With a hard knee setting, compression occurs immediately at full ratio once the threshold is reached, but not before...
In most cases, the action of hard‑knee compression is more assertive and less subtle than that of a soft‑knee model; soft‑knee compression is often preferable when we want to retain the illusion that no processing is taking place.
...with a softer knee, the onset of compression is more gradual and begins before the threshold is reached.
Parallel Lines
Newer compressor designs often now feature a wet/dry mix control. This allows you to implement a technique known as parallel or New York‑style compression, whereby a heavily compressed version of the source is blended with an uncompressed version. This modifies the sound in complex ways, with results that are audibly different from straightforward compression. For more detail, I’d suggest consulting Hugh Robjohns’ in‑depth article on the subject: www.soundonsound.com/techniques/parallel-compression. Alternatively, dive in and experiment! Be aware, though, that there is no universal standard for how the ‘wet’ and ‘dry’ sides of the signal should be balanced, so you may find that a setting which works for one compressor is completely different on another.
A wet/dry control isn’t essential for implementing parallel compression, however. You can achieve the same effect by placing a conventional compressor on an auxiliary channel and sending to it from the dry source channel. The relative levels of the faders on the source and auxiliary channel will then determine the balance of dry and compressed signal.
Compulsive Compression
Like other common processors such as reverb or equalisation, compressors sometimes offer an endless variety of additional controls on top of their core parameters. Sometimes you’ll also find familiar knobs bearing unfamiliar names. Most of this, however, is icing on the cake. Once you understand threshold, ratio, attack and release, and how they interact, you’ll be a long way towards being able to use compression to get the results you want.
So, my advice to anyone starting out is to focus on these core controls first — or rather, second. The first thing should always be to ask yourself why you’re using a compressor at all, and what result you’re aiming for. When we have unlimited plug‑in power at our disposal, it’s easy to get into the habit of applying compression purely for the sake of it. But, like any process, compression is a means to an end, not an end in itself. It’s much easier to choose the right settings when you know what you’re trying to achieve with them!
Check out this list of software and hardware compressors SOS has reviewed
What About Stereo?
By definition, the two channels of a stereo recording contain different signals. These signals won’t necessarily peak at the same time or the same level; so what should happen when we place a compressor across a stereo channel?
Most stereo compressors default to acting identically on both channels. Gain reduction is applied equally to both sides whenever a peak occurs in either. This is the safest option, because applying gain reduction unequally risks an audible sideways shift in the stereo image. It also makes the design of the compressor simpler, because the side‑chain signal need only be a mono sum or average of the left and right channels.
The much‑loved AMS Neve 33609/N is designed as a stereo compressor, but allows the two channels to be unlinked. In this mode they can be used as independent mono compressors for different sources, or on ‘stereo’ material where preservation of image isn’t an issue.
However, some compressors do have a true stereo side‑chain, and give you the option to partially or wholly unlink the two sides. Fully unlinking the sides makes a stereo compressor behave as two separate mono devices that happen to share the same settings, and can be useful on group channels such as backing vocal auxes which aren’t carrying true stereo signals, just lots of related mono sources. (The same can be achieved in some DAWs by using a dual mono rather than a stereo plug‑in instance.)
Quite a few compressors also have the option to operate in Mid‑Sides mode, whereby the left and right channels are matrixed into Mid and Sides channels, compression is applied, and then the matrixing process is repeated to restore left‑right stereo. In this case, unlinking the two channels in a stereo side‑chain causes the stereo width to vary, which is much less objectionable than left‑right shifts in stereo image, and it can in fact be very useful, especially at mastering. For example, if we’re sent a mix to master where the vocal is too loud, we can often improve matters by compressing only the Mid component.
Meanwhile, Back In The Real World...
In this article, I’ve described threshold, ratio, attack and release as being the most fundamental compression parameters. But there are some classic hardware compressors — and plug‑in emulations of them — that don’t have even these controls. This is because analogue electronics can only ever approximate the behaviour of an ideal compressor, and some older designs are very limited in comparison.
Based on the legendary Urei 1176, Black Lion Audio’s Seventeen compressor is an example of a ‘fixed‑threshold’ design.
Real‑world electrical components can react to input signal levels in complex ways. The classic Teletronix LA‑2A, for example, runs its side‑chain signal into an illuminating panel which, in turn, triggers a light‑sensitive component that applies more gain reduction as the light gets brighter. This pair of components exhibits very complex behaviour, with fixed but relatively long attack and release times, a very soft knee and a highly variable ratio. Another classic design, the Urei 1176, does have time‑constant and ratio settings, but it too lacks a threshold control.
Both are examples of fixed‑threshold designs. With no user control over the threshold value, we vary the amount of compression by raising or lowering the input level. Instead of adjusting the threshold level relative to the input signal, these are designed so that we adjust the input level relative to a fixed threshold. A higher input level equates to a lower threshold and therefore more compression (and vice versa).