It not uncommon to have the need to control the volume (dynamics) of a signal in an automated way.
We may be trying to avoid too high a level that will clip an amplifier or deafen the audience or send a speaker cone excursing to hyperspace. Or we may just want to regain control on the voice of a singer that will alternate shouting and whispering. Sometimes we will want to avoid background noise when no signal is present.
To perform all those functions, dynamics processors come to our aid. These are commonly used in live sound reinforcement as well as multi-track recording, while they are not used as often for pre-recorded sound, which is assumed to have canned controlled dynamics. These days more and more mixing is performed with digital mixers, which often integrate dynamics processing, so there is more dynamics processing taking place routinely than ever before.
1.2. Defining dynamics processing
The concept of a dynamics processor is really not that different from that of a person changing the volume by moving a mixer fader. For instance, if we have a singer that starts singing too loud or getting too close to the microphone, we will reduce the volume on that voice's channel. In that case we are proving compression. When the singer is not singing, we may move the fader all the way down to avoid background noise leaking into the main outputs, thereby acting as a noise gate. There is basically a process whereby someone is listening to the volume changes of a signal and taking the decision of whether the volume needs to be changed or not. The graph below illustrates the process : the auditory system detects the volume changes, and the brain commands the hand to bring the fader up or down as a function of the signal volume.
This human dynamics processor has its limitations. It can only control one channel, it is slow, and its actions are not repeatable. We might want to use a robot with an articulated arm that would ride a mixer's faders. In practice, though, we choose an electronic device that performs an equivalent function. The electronic version is not very different from a philosophic point of view, though it does away with its limitations.
As shown below, the input to a dynamics processor is split into two. One of these signal copies will be processed by a gain changing element, which will typically be a voltage controlled amplifier (VCA) or a digital equivalent. The other copy goes to a detection circuit that rides the VCA. For the volume changes to be smooth, an envelope generator is used to ramp volume changes and thus avoid abrupt audible changes. The slope and shape of the ramp can be modified, as we will see later. Often we can choose to feed the detector with the input signal or, alternatively, an external signal which is referred to as the "Side Chain" or "Key" signal.
One of the side effects of using analog volume control elements (VCAs) to process the signal is noise. The quietest VCAs tend to be expensive, and therefore only included in highly professional equipment. As well as a good VCA, a quality dynamics processor also needs a good detection stage, which is by no means easy to design. Which would explain why only a handful of brands on both side of the Atlantic enjoy a reputation for quality dynamics processors and are used for serious sound reinforcement. A good dynamics processor should make it easy to control dynamics transparently, avoiding any kind of audible undesired "pumping" and "breathing" effects. Newer digital units, often built into digital mixers, do not suffer from noise problems, though quality dynamics processing algorithms are not easy to come by, and therefore it will probably be unrealistic to expect quality compression or gating from inexpensive digital products.
The more commonly used dynamics processors are :
A compressor / limiter attenuates or limits signals exceeding a pre-defined signal level. There exists an special version of a compressor/limiter called "de-esser", which tames excessive levels of the portion of the frequency spectrum where sibilance occurs.
A limiter is only a form of compressor.
A noise gate mutes or attenuates signals below a pre-defined signal level. If it allows the selection of the attenuation level (as opposed to just providing total attenuation, i.e. muting), it is referred to as a "downward expander".
There also exists the "true expander", though in practice there are no commercial devices that perform true expansion, which would entail amplifying signals above a specified level and attenuating those below it, therefore expanding the dynamics of a signal.
We could also speak of digital (those that process a digitized signal) and analog devices. In reality, (good) digital devices can work like their analog counterparts, though normally one uses their processing power to increase manipulation possibilities. For instance, for recording and other non-real time applications, we could compress a signal using a "look-ahead" buffer to make compression/limiting decisions based on what is still to come, so that, for instance, we could start compressing a peak before it exceeds the threshold, avoiding the transient overshoot that would occur on an analog compressor and doing so in an inaudible way.
Different dynamics processors will provide differing sets of controls and indicators. In general, the controls we will find on dedicated units (built-in ones may obviate some of the controls) are :
1.5. Meters and indicators
The most common visual indicators provided on dynamics processors are given below. You may not always find them, or you may get additional ones:
Related reading »» 7a. Compressors, 7b. Noise gates