Compression and Limiting
Studio Sound Techniques
The lowdown on Compressor-Limiters, and how to use them.
The correct use of compression and limiting is one of the fundamental techniques of modern recording and one of the hardest to come to terms with. Freelance recording engineer Dave Lockwood explains the theory behind such devices and below outlines some of their typical applications in today's recordings.
Dynamic range control, in the form of compression and limiting, is one of the most important factors in engineering a 'tight', commercial sound onto tape. The need for this sort of processing arises because many acoustic sources have a dynamic range that exceeds the dynamic capability of most recording systems, which means that when the recorder's sensitivity is set to capture the quietest signal component, the loudest signals will then cause overloading and distortion.
The simplest form of compression can be manual, achieved just by reducing the system gain control during the loud passages and increasing it during the quiet ones, indeed, when used skilfully this method, known as 'gain riding', can be quite successful with a relatively predictable source, such as an orchestral music signal, where the rate of level change tends to be fairly slow, and where the peak levels can be anticipated from the continuity of the music. But the close-miking techniques used on today's recordings tend to emphasise dynamic range, and trying to adjust the gain to compensate for the random peaks of a solo instrument's response is an impossible task, for the transients cannot be anticipated, and may be of such short duration that the hand cannot possibly react fast enough.
The compressor-limiter provides an electronic means of automatically controlling gain, which can react very rapidly to changing signal levels. The device operates around a predetermined 'threshold' level, set by the user, above which gain reduction, or compression will occur. The degree of compression is expressed as a ratio (sometimes referred to as a slope), eg. compression at 4:1 means that once the threshold level has been exceeded, a further increase in input of 4dB will result in an increase of only 1dB at the output. With the peak levels automatically restrained in this way, the system's sensitivity can then be safely increased to achieve a high recording level without fear of overload.
The term 'limiter' is normally applied to a unit with a fixed high compression setting (usually in excess of 10:1), referred to as a 'limit ratio' or 'infinite slope', where the output level cannot be made to rise beyond the threshold, remaining effectively constant regardless of any amount by which the input signal may increase. A 'compressor' tends to have a much softer slope and a more subtle action, but sophisticated units described as 'Compressor-Limiters' often feature a very wide range of compression settings, from a gentle 2:1 to a 'tight' 20:1 or more, up to infinity.
As you might expect, the audible effect of severe limiting is to produce a noticeably more flattened and constricted sound than subtle compression at 2:1, but the ability to alter the threshold allows the different ratios to have their own particular applications.
By setting a high threshold, well above the average signal level, a limiter can be used purely as a protection device to prevent overload in a recording system or PA; the general signal dynamics will remain completely unaffected as the majority of the sound content is below the threshold level, but potentially troublesome momentary peaks, well above the average level, will be attenuated and kept within the acceptable range. Although the limiting action is severe, it can be of such short duration that its less pleasant side-effects are not audible.
When softer slopes are used, gain reduction can be made more continuous by setting a threshold that is below the average signal level. Compression will then occur over most of the dynamic range of the source, but without producing an unnatural 'squashed' effect, as any increase in level at the input is always able to produce some increase in level at the output. The compressed signal thus retains the general dynamic characteristics of the input, but with the overall range reduced, producing a more constant and predictable level that is easier to record.
Some compressor and limiter designs do not feature an actual 'Threshold' control, but have a fixed internal threshold with an 'Input' or 'Sensitivity' control to vary the signal level going into the unit. Keeping the input control down will give a low level signal that rarely crosses the threshold and therefore produces little effect, whilst increasing the input control will result in a higher level signal that consistently exceeds the threshold, producing more continuous gain reduction. An 'Output' level control is also often featured, to make up the level lost through compression, but where none is provided, simply increasing the sensitivity of the following stage in the signal chain will achieve the same result.
The attack and release characteristics of a compressor govern the time taken for the device to initiate gain reduction in the presence of a signal that exceeds the threshold, and then to return to normal gain when the high level signal is removed.
From the user's point of view, compressors can be divided into two main types: those in which the attack and release is automatically determined by the unit according to the dynamics of the input signal, and those which allow the user to set the attack and release times via front panel controls. In general, the first type is rather easier to operate, as its characteristics are always optimised to produce the minimum of unwanted side-effects whatever the signal, but this is achieved at the expense of some flexibility, for settings that depart quite substantially from the optimum can produce some very usable deliberate effects.
Units are now available which feature just a single front panel control, for the user to set the threshold, with the compression ratio and best attack and release times all being automatically determined by the internal circuitry, according to the input signal's characteristics, and with even the output automatically adjusted to maintain a consistent level. For many users, the slightly compromised facilities of this type of compressor are more than compensated for by the predictable performance and absolute simplicity of operation.
Recommended control settings for compressors and limiters are difficult to specify as the conditions of use, and the units themselves, can vary so much, but certain general principles are evident which can be of assistance in the utilising of this type of device for specific applications.
Vocals can benefit from the use of compression in recording, perhaps more than any other type of signal. A typical singing voice tends to contain frequent high level peaks that restrict the maximum recording level, while the average level still remains comparatively low, and this can make it practically impossible to place an uncompressed lead vocal at a satisfactory level in a mix, without it being either too dominant or becoming lost in the backing track.
Limiting can be used to restrict the higher levels, but in order to avoid producing an excessively 'flattened' effect, it is usually preferable to employ a softer slope, in the region of 4:1, operating over a wider area of the total range, as this will raise as much as possible of the low level content, and also retain some of the dynamics of the performance, whilst still providing the necessary control to allow the average level to be significantly improved.
A fast attack time, around 1 to 2 milliseconds, is usually needed for compressing vocals, for the transient rise-time can be very short. As the attack setting is lengthened, more of the transient is allowed to pass through before the onset of gain reduction, and with voices this can have the effect of emphasising sibilance (the sounds of the letters 's' and 'c'). In general any transient content much shorter than 1 millisecond will tend to be lost through tape saturation, so attack times any faster than this are usually unnecessary.
Release time, sometimes called recovery time, is a most important factor in determining the overall effect of compression. A fast release, which ceases gain reduction as soon as the high level signal is removed, will allow the maximum amount of low level information to come through unattenuated, producing the highest possible continuous level. However, the combination of a fast attack and release with a large amount of gain reduction can produce very noticeable side-effects, with the rapidly altering gain being evident in the form of 'pumping' and an impression of levels surging up and down. A longer release time, which causes the compressor to return to normal gain more smoothly, will make such modulation effects much less objectionable, but it will also prevent such a high average level being achieved due to the lengthened periods of attenuation.
An optimised release setting usually represents the best compromise between these two factors. Sounds that have very familiar tonal and dynamic characteristics, such as voices, will usually require a smooth release in the region of one second or more, in order to avoid unnatural effects, but with less well known, or electronic sounds, the ear will often tolerate much higher levels of dynamic corruption.
Electric guitar sounds, for example, can sometimes benefit from the effects of the most extreme limiting, using the fastest available attack and release times. A consistently high level can be maintained to give the guitar track enormous presence in the mix, with the modulation caused by the fast release actually having a desirable effect in altering the natural decay of the sound, causing notes to be held at full level when they would normally be dying away. A guitar footpedal 'Sustain' effect is in fact just a compressor with a fixed, very fast attack and release, which is fine for guitar use but which tends to rather restrict the effectiveness of this type of unit in other applications.
A very effective 'tight', clean rhythm guitar sound can be produced by using compression on a DI'ed (direct-injected) signal. The compressor not only assists in achieving the cleanest possible sound, but can also make the playing seem exceptionally well controlled. When a long release time is selected (2 seconds or more), heavy limiting can be used to hold a rhythm guitar part at a constant background level in a mix, where it can contribute effectively without needing to be prominent.
Bass guitar is invariably tightly compressed on commercial recordings, to the extent that the home recordist working without a suitable compression or limiting device may find it impossible to achieve anything comparable to the definition and articulation of the modern recorded bass sound. The compressor enables a constant high level to be recorded, and also smooths out the inherently uneven response of the different registers of this instrument.
The levelling action of tight compression is fundamental to the recording of the currently popular 'thumb and string pull' funk bass style, as it enables the deep, 'thumbed' bass note and the trebly, 'pulled' upper note to assume equal prominence. With more conventional styles the slight 'click' of the initial contact between the finger, or pick, and the string is emphasised, giving audible punch to the start of each note, but without a high level transient to cause overload problems.
A well compressed bass can be used at a very high level in a mix, giving the desired depth and solidity to the sound, but without swamping the track with the sea of uncontrolled, boomy bottom-end that seems to afflict so many 'demo' recordings. Release times in excess of 500 milliseconds may often have to be used, for in many compressors a very fast attack and release time will cause audible distortion at low frequencies. But given a unit of sufficient quality, a very short 100ms release can be most interesting. This modifies the normal decay of the instrument, and produces an effect that is particularly pleasing with a fretless bass, where the naturally more rounded envelope (attack and decay characteristics of the note) causes longer notes to seem to 'swell' or gain in level while sustaining, making the characteristic slides of the fretless player's technique even more effective.
Compression and limiting produces interesting effects for the recording of drums and percussion, where the aim is not usually just the controlling of levels. The side-effect of 'pumping', caused by a fast release, can be employed to give a 'larger than life' tom-tom sound, with an emphasised resonance caused by the extended decay. When a bass drum is heavily compressed, with a short attack (1ms) and a lengthy release (1 to 2 seconds), the sound becomes much tighter and less boomy, with an enhanced midrange and a more evident 'slap' to the start of each note, which all assists in enabling the currently favoured, very high bass drum levels to be used, without masking other bass instruments or overpowering the rest of the mix.
When a whole drum kit balance is limited, the attack time can be used to reduce or emphasise the most dynamic elements of the sound, such as the 'crack' of a heavily struck snare drum. By selecting a longer attack (25 to 50ms), the unit will be made to respond mainly to the average level, allowing the shorter transients to remain unaffected and therefore at a relatively higher level. However, dynamic range should never be restricted indiscriminately, and unless some sort of deliberate compressor effect is actually wanted, percussion instruments can very often be recorded quite satisfactorily without compression or limiting.
The breadth of tone and dynamic capabilities of a large acoustic piano can sometimes give problems in recording. However, the use of some gentle 2:1 or 4:1 compression, with a smooth release, will emphasise the attack of the hammers and enhance the richness of the tone by extending the decay, and whilst the resulting sound might not please the purist, it offers all the presence and clarity needed in a commercial mix. Similarly, the warm but rather ill-defined basic sound of many Fender Rhodes electric pianos can be assisted by the application of compression, to achieve what many people actually think of as a 'typical Rhodes sound', with the percussive contact of the hammers on the tynes clearly audible, and that beautiful bell-like, ringing sustain.
Compression and limiting are often used on mixed programme material to achieve maximum impact and continuous 'loudness'. If you have ever wondered how TV commercials manage to sound so much louder than the programmes, it is because their soundtracks are often so heavily compressed that the contents are practically all at peak level!
One of the difficulties encountered in compressing a mix is that the most dominant instrument will tend to control gain reduction, and therefore modulate the rest of the mix, but this problem can be alleviated to some extent by the use of a softer compression slope over a wider range of input level, or by a longer release time which tends to make any modulation less obvious. When processing stereo signals, it is important that a dedicated stereo unit, or a pair of mono units with a stereo link facility, should be used, for it is necessary to have equal amounts of gain reduction occurring simultaneously in both channels, otherwise unwanted shifting of the stereo image can occur.
Compressor manufacturers' advertisements sometimes make reference to a 'soft knee' characteristic. This simply means that full compression does not occur as soon as the threshold is exceeded, but that a more gentle action takes place, in which the severity of control is progressively increased with rising level, making the onset of gain reduction less audible and maintaining a more natural sound around the threshold point.
Inevitably, the use of compression and limiting has its less desirable aspects. When a compressor is used to bring up low level information, it is, of course, unable to distinguish between wanted and unwanted signal components; consequently noise, ambience, studio crosstalk and sound spill are all also amplified. In fact, every dB of compression is an extra dB of source noise, so it is obviously important that dynamic processing should always take place at the optimum point in the signal chain. The treatment of individual signals should not be left until the mixing stage, for compression 'off-tape' has the effect of exaggerating tape hiss, whereas the superior noise performance of the mixer's input channels will more acceptably withstand this sort of degradation at the original recording stage.
Compressors used in recording normally operate at line level (0.775V, or 0.335V if using the Japanese -10dBV format), and should therefore be connected via the mixer's 'insert points', but where there are none available, the device can be patched in-line between the mixer and the tape machine, provided that due attention is paid to the monitor status to ensure that the compressed signal is actually heard!
It is always preferable to compress after any equalisation, rather than before, for the reverse arrangement inevitably results in a greater deterioration in the signal-to-noise ratio, and compression after the EQ stage also permits a wider scope of equaliser settings, with larger amounts of signal boost available before distortion.
Many recording studio compressors feature some form of metering to show the amount of gain reduction occurring at any given moment; reverse driven VU meters are encountered, although LED columns are now more popular. Whilst LEDs are faster and more compact, a large VU meter offers superior resolution and I feel, gives a more graphic indication of the release characteristics. An In/Out switch for comparing processed and unprocessed signals, is another useful facility, although not always provided for it enables the operator to check relative noise levels and signal deterioration when setting up the unit; it is also advantageous for the metering to remain operative when the device is switched out, as this allows preliminary control settings to be made visually, before switching the processor into the signal path.
Modern compressor-limiters often feature sidechain patching facilities which give access to the level detection circuitry. By inserting an equaliser into the sidechain it is possible to cause 'selective limiting' to occur, where some frequencies are limited more than others. The frequencies to be treated should be boosted by the equaliser, as this makes the limiter more sensitive in that area, causing an earlier onset of gain reduction at those frequencies.
One of the most common uses of selective limiting is for the control of sibilance in vocals and speech, indeed, dedicated 'De-Esser' units are available which perform just this specific function. It is also possible to arrange for one signal to control the level of another, by patching the controlling signal into the 'detector' or 'control' input of the sidechain, with the main signal connected via the audio circuitry. Gain reduction will then occur when the control signal exceeds the selected threshold, but causing the main signal to be attenuated.
This effect, known as 'ducking', is widely used in broadcasting to allow a DJ's microphone to automatically attenuate music when he's talking. However, careful setting of the release time is needed to arrive at a compromise that avoids both rapid level changes between words or phrases, and a long recovery gap after the speaker has finished. This process can also be used to provide a degree of automatic control in mixing, by attenuating certain backing instruments in the presence of the lead vocal, for example. The effect must be used with some subtlety, however, for the listener should not be aware of the technique, just the end result.
Hard limiting can be used very successfully to improve the performance of spring reverb units, which have a tendency to 'flutter' when confronted with a heavy transient signal such as a snare drum, spoiling the smoothness of the decay and severely restricting the amount of effect that can be used. By placing a limiter with a fast attack and release setting before the spring, short term transients can be kept below the level where problems occur, while all other signals remain completely unaffected.
In choosing a compressor-limiter I feel that it is important to consider what type of usage you envisage: If you are a musician recording mainly by multitracking, it can be advantageous to try to obtain the best quality single channel unit that you can afford, rather than necessarily considering a stereo unit, for with only one source to be recorded at a time, a single processor is all that is needed, and with the appropriate use of compression on individual signals there is often no need for processing the final stereo mix. Provided that your single channel unit has a stereo link facility, it is always possible to upgrade to stereo at a later date, with the purchase of a second unit of the same type.
Compressors, like microphones, seem able to excite widely differing opinions, for although the most 'state of the art' units should produce the cleanest, most transparent sound, particular models are often appreciated precisely for the characteristic colourations and distortions that they impart; indeed, old valve limiters are among the most highly prized possessions of some American studios! Engineers and producers often seem to become attuned to the performance of a particular model, being prepared to overlook any limitations in return for an indefinable 'character' - I find that I continually return to using a Urei LA4, which I have yet to hear surpassed as a compressor for vocals, although I know other engineers who rarely use this model from choice. It is therefore obviously important that the potential purchaser should evaluate a compressor-limiter by hearing it in use, for specifications alone will reveal nothing about the differing subjective impressions given by apparently similar units.
The correct use of compression and limiting is one of the fundamental techniques of modern recording, and for the home recordist, only the acquisition of a decent reverb might be considered more important. Overload protection is naturally of the greatest importance when working with the restricted headroom of simple cassette-based systems, but the discovery of the immense range of creative applications is of great value in all fields of recording.