We lay out the purpose and the role that those machines in charge of dynamic processing of audio signals, known as dynamics processors, can, or rather should, fulfill. Before going further, though, and in order to keep things on the right track, it’s worth clarifying what that term “dynamics” refers to. An expression that, more properly speaking, we should call musical dynamics; since that is what we are really referring to, and in many cases it is advisable to preserve it, meaning to keep it faithful to the original.
The musical dynamics (of a work or a piece of music) are specified in the familiar dBs; they refer to the difference in level between the loudest and the softest parts of the work in question; that is, the level ratio between the two. This is perfectly defined in musical notation, where the softest part is marked “ppp” (pianississimo) and the most vigorous “fff” (fortississimo), both extremes representing the maximum possible musical dynamics. That said, the use of dynamics processors, and in general any type of processing of audio signals, does not necessarily have to be related to music; an obvious case can be seen in the effects and dialogue tracks included in film soundtracks; or, to give another example, in advertising.
There are two basic electronic devices par excellence used for the task at hand. One is, undoubtedly, the compressor and the other the noise gate; we should also consider the limiter a third machine, although we don’t really count it as such; since, as we’ll see later, a limiter is really a compressor with an infinite ratio; so it’s more prudent to speak of a compressor with limiter action than of a limiter proper. That is, by the way, not very important; I only mention it to clarify the terms as well as possible and aid understanding of the subject; but keep in mind that compression and limiting are two different things.
As a clear warning, I’ll note something you can surely find in any treatise on the subject we’re covering here. More is less; which is to say that using whatever device we treat an audio signal with is one thing, and abusing it is quite another. Something that can lead us to what is called listening fatigue by the fast track.
And above all, you have to keep firmly in mind that musical dynamics are an integral part of the music itself, so altering them means altering the musical message itself; distorting it, to put it plainly. Especially when we’re talking about classical music. Don’t quite believe it? Wait until the conductor catches you, and you’ll learn a new use for his baton the moment you overstep a hair with his interpretation.
The Compressor
Used both live and in the studio, this tool is nowadays indispensable. Its primary function is to reduce the dynamic range of signals that exceed a certain level threshold, which we have previously set. This makes possible, for example, an automatic gain reduction in the presence of unwanted signal peaks, which in turn makes the signals below the threshold move closer to it; that is, it levels the signals. Hence their dynamic range is reduced.
As with a mix, the use of a compressor, and therefore of any sound-processing device, does not follow any rulebook. Whoever works in this field is certainly doing creative work. A painter mixes and applies colors according to his intuition, knowledge, purpose, etc., and this comes to the same thing; our palette of colors is that console and those processing racks. What I mean by this is that nobody can tell you with certainty how to act on the controls at a given moment. For example, in the case of compression for a bass drum (kick), we can’t apply the same compression settings to every kick we come across, since they’re not all alike in terms of head, beater, tuning, the drummer’s playing, and a long etcetera. And even in the very hypothetical case that they were all identical, it wouldn’t be advisable to create the same sound for all of them. As if that weren’t enough, add the dependence on the type or style of music we’re mixing, intermodulation between instruments, and a long list of additional circumstances.
Another matter is the definition of a compressor’s controls, and guidance on where to set them as a reference point for applying them to a given instrument, mix, submix, or mastering job. Here are those controls.
Threshold: Sets the signal amplitude threshold at which compression begins to act once it is exceeded, whereas if it is not exceeded the compressor will do nothing. It is a hugely important parameter, so setting it is critical.
Attack: This is a time control (attack time). It determines the compressor’s reaction time once the signal has exceeded the threshold, at which point gain reduction begins, usually displayed on an LED bar or an analog VU meter, labeled for this purpose as gain reduction. Its value ranges from one ms, or even less (very fast attack), to a few hundred milliseconds. The attack time setting tends to be shorter (slower, or longer in time) for low frequencies (bass) than for high ones (treble). Although in reality it depends heavily, among other things, on the transient of the signal itself. The transient of an audio signal represents its own attack (not to be confused with the attack control on a processor); percussive sounds usually have a very fast transient, so for these the compressor’s attack speed is usually also fast, but taking care not to cut the signal’s own attack, something that, by the way, has to be kept in mind with all signals, unless for some reason you want to deliberately cut the attack of a signal. In general, as stated: don’t cut the transient of any signal.
Hold: With this control we set how long compression is maintained once the signal drops below the set threshold. We will get smooth control over the sound if we set a long time, especially if we have signal peaks close together, as is usual in fast songs (high tempo), though without overdoing it, because otherwise compression would be held longer than necessary, compressing the program signal in places where it isn’t needed, which would produce inappropriate or unnatural level changes.
Release: Once the signal’s amplitude level falls below the threshold and the compression hold time has ended, the release control comes into play, setting the speed at which compression finally lets go, so as not to have a possible (depending on the instrument in question, especially percussion) very abrupt re-attack during the course of the signal’s envelope. The same will happen if we enable a release time that is too fast (short), but it isn’t good to have it too slow (long) either, since the compressor should be completely inactive so that it can act on the next signal that exceeds the threshold. As we have seen, this is another time control; its range varies from a few milliseconds to a few seconds. Commonly, the release time is set longer (slower, or greater) than the attack. The release value is specified in dB of decay per second (dB/sec), which tells us the speed at which compression falls away, or the time that must pass before it restarts (a new attack action). Some well-built compressors also have a control to set this release slope. An excessive decay speed (many dB/sec) will cause a clearly audible abrupt change of level; conversely, too slow a speed will make the signal ignore the threshold by not reaching it when it should, especially at high musical tempos, likewise producing inadvisable level variations.
Compression ratio: The amount of gain reduction we want once the signal exceeds the threshold is what this parameter controls. If, for example, an input signal exceeds the threshold by 4 dB and we have set a ratio of 4:1, at the processor’s output we will get 1 dB above the threshold; if the ratio were 2:1, at the output we will have 2 dB over the threshold. We can clearly see that setting a low compression ratio will give us light compression, generally desirable, whereas with a high value we will get a heavily compressed output signal.
This control, together with threshold, are the two main parameters we must set, since the rest are directly (critically) dependent on how they are adjusted.
Output gain: While a compressor is working, it is normal for the level of the signal at its output to be lower than the one arriving at its input; restoring this level is the job of the output gain control. By comparing the unprocessed signal, by bypassing, with the processed one, we adjust this parameter until both have as nearly the same level as possible. And in general, we should also check the setting of all parameters by directly comparing the signal before/after (A/B comparison), whatever process is being applied to it; not just compression. It is also worth mentioning that during the compression process, besides a loss of level, we may also observe a change in the tonal character of the sound, which is why some of these machines also have tone (EQ) controls. Of course... indeed, controlling and/or correcting tonality in all its breadth is the job of an equalizer, a device we can conveniently chain at the compressor’s output; it is not wise to place it at its input, because then its correction curve could improperly trip the compressor’s threshold. But as I said, there’s no accounting for taste...
Stereo Link: The link switch (not always called link) is used to couple two mono processors for operation in stereo, or conversely, to convert stereo processing into dual mono (two independent mono channels). When working with a stereo signal, the compressors have to be linked. That way, when compression starts on one of them, the other is forced by it to operate as well. In a stereo coupling of compressors, it is common practice to have both compression channels with identical parameters; in fact, those parameters are normally set on only one channel, and the processor automatically copies them to the other. If we didn’t do this, when one channel began processing, the other would stay inactive waiting its turn, and the result would be a confusing sound image and a dreadful sense of autopanning between the two channels.
Limiter: By setting a compressor’s ratio control to its ∞:1 position, when the signal exceeds the set threshold the compressor operates as a limiter. A limiter acts on the signal that touches the threshold, making it impossible for it to be exceeded in any way; it cuts it off abruptly at the safety value we have set with this parameter, unlike the compressor, which exerts a smoother, more continuous control over gain reduction. Some compressors have a separate control for this purpose, with which what we are really setting is its working threshold, which should not be excessively low. Normally, the limiter section in a compressor, or in any device that includes one, has at least one LED that lights when dynamic limiting kicks in; well, in principle it is advisable for that LED to light briefly in the presence of hits or loud parts of the signal (the peaks), with the limiter remaining inactive for the rest of the program material; otherwise you risk wrecking or excessively clipping the dynamic range of an instrument or song, as the case may be. Not for nothing, this control can often be found under the name Peak Limiter.
Some compressor models don’t have attack, hold (few include it), or release controls; in that case, the machine itself determines these parameters automatically by analyzing the program signal fed into it. On other occasions, we may find on the control panel a manual/auto switch, which gives us the option of controlling all the parameters it includes manually, or doing so automatically as described above. In principle, it is advisable to use automatic control, especially in live applications, since this function spares us the complex adjustments of the aforementioned controls, and it usually works OK.
Hard Knee & Soft Knee
Depending on the mode or behavior in which a compressor treats signals that exceed its threshold, it is called hard-knee or soft-knee. As can be seen in the graphs, soft-knee mode will always be preferable because of its gentler action, with a more or less progressive response curve depending on the purpose pursued. Hard knee is mostly applied to individual instruments or tracks. Soft-knee behavior, meanwhile, is the most advisable for mixes and, above all, for preparing premasters. The graphs are quite explicit on this, so I won’t elaborate further. Note the abrupt response curve of the limiter mentioned earlier, compared with other ratios.
Something that, because of its interest, is worth mentioning about compressors is that there are models called multiband, with which we can process several bands separately. Their layout is basically a signal entering a crossover filter, with a compressor connected to each of its outputs. The output of these compressors is fed to a summing stage that rejoins the processed portions of the signal, thus restoring the full spectrum. A bit more complicated to handle because of the crossover frequencies, but super effective. Indispensable in mastering.
