Next to reverb, compression is arguably the most important studio process and can enhance most recordings. Paul White explains how compressors work and gives some practical tips on using them.

Most people have a general idea of what a compressor does, but many users aren't confident about actually applying them. And this is a pity, because, next to reverb, compression is arguably the most important processor in the studio. Put simply, a compressor is a device which reduces the dynamic range of an audio signal; it reduces the difference between the loudest and quietest parts of a piece of music. If you like, think of it as a helping hand which automatically pulls down the fader for you when the level gets too high. Before going further, it might be useful to explain why we might want to do this.

Pop music tends to have a limited dynamic range — the levels involved are fairly even and consistent. If we were to add an uncompressed vocal track to a typical pop backing, we would be aware that certain loudly sung words and phrases were too obtrusive, while some of the quieter phrases might be almost completely buried beneath the backing. Aside from making the lyrics difficult to hear, this variation in level makes the recording sound very unprofessional. And it's not just vocals that suffer — any acoustic instrument is likely to suffer from excessive dynamics when used in a pop context. Even electronic or electric instruments such as electric guitars and basses benefit from the use of a compressor to help produce a smooth, even level. And because quieter sounds are brought up in level, compression increases the average signal level of a recording, which helps produce a tight, punchy sound.

PARAMETERS

• THRESHOLD: All standard compressors work on a threshold system; signals exceeding the threshold are processed (reduced in level) while those lower than the threshold are unchanged. A further stage of 'make-up' gain is added after processing, to allow the user to restore or 'make up' any gain lost by the compressor action.

• RATIO: Whenever the input signal equals or exceeds the threshold, gain reduction is applied. The actual amount of gain reduction depends on the 'Ratio' setting, and on most compressors, that ratio is variable. The higher the ratio, the more gain reduction is applied so the more pronounced the compression. Very high ratios effectively prevent the signal from ever exceeding the threshold, and this is known as limiting. Because most compressors can also be used as limiters, they tend to be called compressor limiters.

The best way to illustrate the concept of compressor ratio is by example. If a compression ratio of 5:1 is set, an input signal exceeding the threshold by 5dB will cause only a 1dB increase in level at the output. The relationship between threshold and ratio is shown in Figure 1.

• HARD KNEE: With a conventional compressor, if the input signal is just below the threshold, no compression takes place, but if it is just over, gain reduction is applied at whatever ratio is set up. This is sometimes known as hard-knee compression because the onset of compression occurs as soon as the signal reaches the threshold level.

• SOFT KNEE: Some types of compressor operate on a Soft Knee principle, where gain reduction doesn't happen suddenly when the threshold is reached but instead is brought in progressively over a range of 10dB or so. Because of this 'soft' action, these compressors are less positive than Hard Knee types, making them less suitable for tight, positive gain control. Their strength is that they are less obvious in use, and this feature makes them very suitable for processing complete mixes or other sounds that need some compression but at the same time shouldn't sound processed. Figure 2 shows the characteristics of a Soft Knee compressor.

• ATTACK: The attack time of a compressor is simply how long it takes to react once the input signal has reached or exceeded the threshold level. With a fast attack setting, the signal is brought under control almost immediately with very little overshoot, whereas a slower attack time will allow the start of a transient or percussive sound to pass through uncompressed before the compressor realises what's happening and takes action. Creating a deliberate overshoot by setting an attack time of several milliseconds is an effective way of emphasising the percussive nature of instruments such as guitars or drums. For most musical uses, an initial attack setting of under 1mS is typical.

• RELEASE: The Release control sets the time taken for the compressor's gain to return to normal once the input signal has fallen below the threshold. Too short a release time can result in an audible level pumping effect which is usually undesirable, but may be used creatively to make rock mixes sound powerful. Conversely, if the release time is too long, the gain may not have returned to normal by the time the next high level sound occurs. For typical music applications, a good starting value for release time is around half a second.

• AUTO ATTACK/RELEASE: Some compressors have an Auto function which sets attack and release characteristics to suit the dynamics of the music being processed. These can be very effective, especially in applications where the signal dynamics are unpredictable, as in the case of complex mixes or virtuoso bass performances. The auto setting is also often the best choice for handling vocals.

• PEAK & RMS: The part of a compressor that monitors the incoming signal level is known as the side-chain detector, and this may be peak detecting or RMS detecting, depending on the model of compressor. A peak detecting circuit, as its name suggests, will respond to peaks in the input signal regardless of how short they are, while the RMS detector averages the signal level over a short period of time.

"Too short a release time can result in an audible level of pumping effect which is usually undesirable, but may be used creatively to make mixes sound powerful."

The RMS detector works much like the human ear and tends to give a natural type of dynamic control, but it has the disadvantage that it can fail to stop short signal peaks, even though a fast attack time is set. The peak detector, on the other hand, will never miss a signal peak, no matter how brief, but the trade-off is that the style of level control may sound a little heavy handed and unnatural.

• HOLD TIME: Normally, a compressor's side-chain follows the envelope of the signal being fed into it, but if the attack and release times are set to their fastest positions, it is possible that the compressor will attempt to respond not to the envelope of the input signal but to individual cycles of the input waveform. This is particularly significant when the input signal is from a bass instrument, as the individual cycles are relatively long. If tracking of the individual waveform cycles is allowed to occur, very bad distortion is audible as the waveform becomes compressed 'out of shape'.

The way to get around the problem is either to increase the release time of the compressor or, if your model allows it, increase the compressor's hold time. Hold time is a simple time delay that prevents the compressor from going into its release cycle until a certain time has elapsed, and if the hold time is set longer than the single cycle time of the lowest audio frequency, the compressor will never be able to release quickly enough to corrupt individual cycles. A hold time of around 50mS should keep things safe down to 20Hz, though some manufacturers build a fixed hold time into their compressors to make them, in effect, undistortable. In my view, this is a good idea, as a 50mS hold time isn't going to adversely affect any other aspects of the compressor's operation.

• STEREO LINK: When compressing stereo signals, it is imperative that both channels of the compressor function together, otherwise the stereo image will appear to shift away from whichever side is loudest. This is obvious when you think about it; if a loud sound occurs only in the left channel, then the left channel gain will be reduced and everything else in the mix will appear to swing towards the right channel, which is not undergoing so much gain reduction. The Stereo Link switch of a dual-channel compressor simply forces both channels to work together, based either on an average of the two input signals or on whichever is the highest in level at any one time.

NOISE

For every dB of compression we apply, we bring up the level of quiet sounds by 1 dB, assuming that we set the make-up gain to keep the peak level the same. This means that any noise present in the input signal is also amplified, especially during quiet periods when it's most likely to cause trouble. Excessive noise can be gated out by patching a gate or expander before the compressor, and many compressors come with their own inbuilt expanders or gates. It is true to say, though, that every care should be taken to minimise the noise at source rather than trying to clean up after the event.

USING COMPRESSORS

A compressor is a processor and so should be used via an insert point or patched in-line with a line-level signal. Many engineers tend to add some compression while recording and then add more if necessary while mixing. Working in this way makes good use of the tape's dynamic range while helping to prevent signal peaks from overloading the tape machine. It is best to use rather less compression than might ultimately be needed while recording, so that a little more can be added at the mixing stage if required. If too much compression is added at the beginning, it is virtually impossible to reverse the effect. If the compressor has an integral gate, this might be better left switched off during recording and only employed while mixing. This prevents a good take from being ruined by an incorrect gate setting and also enables the gate to remove any tape hiss that might be present, along with other noise.

OTHER SIDE EFFECTS

In pop music, most of the sound energy occupies the bass end of the spectrum: the bass drum, bass guitar, deep snare drum and so on. Unfortunately, any high frequency sounds in the mix will also be brought down in level as the compressor reacts to loud bass sounds. A quiet hi-hat beat, for example, occurring at the same time as a bass drum beat will be reduced in level even though the hi-hat sound alone is not too loud. One way around this is to set a slightly longer attack time on the compressor to allow the attack of the hi-hat to get through before the gain reduction occurs. Even so, if heavy compression is applied to a full mix, the overall sound can become dull as the high frequency detail is pulled down by peaks at the bass end.

Some recent compressor designs use a degree of deliberate harmonic distortion or dynamic equalisation to provide a subtle exciter or treble boost effect related to the amount of gain reduction taking place. This can go some way towards counteracting the dulling of high frequency detail, though the solution is only partial. Alternatively, additional EQ or a harmonic exciter may be patched in after the compressor.

DUCKING

In addition to their more conventional applications, compressors may also be used to make one signal control the level of another. This is known as ducking and is frequently used to allow the level of background music to be controlled by the level of the voice-over. As the voiceover starts, the level of the background music drops, but whenever there is a pause in the speech, the background music is restored to its former level, at a rate set by the compressor's release control. (See box for some more applications of Ducking).

To try ducking, you'll need a compressor with a side-chain access socket. This allows an external signal to control the compressor action rather than the input signal. When an external signal is patched in, its dynamics will control the gain reduction process. If music is fed into the main compressor input and a voice-over is fed into the side-chain input, whenever the voice level exceeds the threshold, gain reduction will be applied to the music. The amount of gain reduction will depend on the compression ratio that has been set. Figure 3 shows how a ducker is used to enable a voiceover to control the level of background music. Certain gates also have ducking functions and my own preference is to use a Drawmer DS201 gate for ducking; it has a dedicated ducking function which is far more intuitive and precise to set up than trying to achieve the same thing with a compressor.

Sibilant vocal sounds are a particular problem which compressors can help address. Sibilance is simply a narrow band, high-pitched sound accompanying normal speech and is caused by air passing around and between the teeth. Viewed on an oscilloscope, it might appear as a high-pitched whistle or noise. If an equaliser is inserted into the side-chain signal path of a compressor, the equaliser can be made to selectively compress only the sibilant part of the audio spectrum; this is known as de-essing.

Most sibilance occurs in the 5 to 10kHz region of the audio spectrum, so if the equaliser is tuned to this frequency range and set to give around 10dB of boost, then in the selected frequency range compression will occur 10dB before it does in the rest of the audio spectrum. The equaliser should be set up by listening to the equaliser output and then tuning the frequency control until the sibilant part of the input signal is strongest. Either a graphic or a parametric equaliser can be used in this application, though a parametric offers more control. Figure 4 shows how a compressor and equaliser may be used as a de-esser.