Using Compressors (Part 2)
Paul White discusses the more advanced aspects of compression, and includes a table of useful settings.
Last month, we covered the basic principles of compression and examined the main controls on a typical compressor. This month, Paul White looks at the remaining facilities and explains the uses of compressors in the studio.
As I explained last month, most compressors fall into the category either of Hard Knee (conventional Ratio type) or Soft Knee, though many new models are switchable between both. This might lead you to believe that all compressors will produce the same result but, surprisingly enough, there's almost as much mystique attached to the sound of certain types of compressor as there is to guitar amplifiers or studio microphones.
One variable between different compressors is the way in which the compressor listens' to the signal; some act on the average envelope of the sound, while others react to the highest peak level, no matter how short it might be. These two modes of side-chain operation are known as RMS (average) and Peak sensing, and though RMS offers the closest approximation to the way in which our ears respond to sound, many American engineers prefer to work with Peak sensing. The differences are not always obvious, especially if you haven't done a lot of work with different compressors, but if you have a model that offers both modes of operation, it's worth switching between them to check which gives the best-sounding result. You'll probably find that the differences show up most on music that contains percussive sounds, as the Peak type of compressor will track the peak levels of the individual drum beats.
More subtle is the way in which a compressor distorts the signal it is processing. You might think the best compressor is the one that doesn't distort at all, and modern designs can come very close to this ideal, but experience has shown that most engineers like the 'warm' sound of the older valve designs which, on paper, perform quite poorly. In fact, subtle distortions can have a profound effect on the way in which we perceive sound — a dash of even-harmonic distortion can tighten up bass sounds and make the top end seem brighter and cleaner. The reasons are quite complex and fall under the general heading of psychoacoustics, but it pays to keep in mind that in all areas of audio, what measures best doesn't necessarily sound the best!
To capitalise on these phenomena, some present-day designers have produced valve designs with a degree of distortion built in, while others have used FET (Field Effect Transistor) circuitry to mimic the behaviour of valve circuits. With the advent of digital tape recorders and more technically perfect mixers and effects processors, the interest in such compressors is on the increase, in just the same way as the rise in popularity of valve microphones.
As explained last month, compressors are really automatic level controllers — if a sound gets too loud, they turn it down. However, if the loud sounds are going to be turned down, it stands to reason that in order to keep the peak level the same as before, we're going to have to turn the overall level up — which is why most compressors have an output gain control. This is sometimes known as 'make-up' gain because it makes up for level lost during compression. When a lot of compression is used, background noise can become a problem during pauses in the input signal, because when the signal is quietest, the compressor's gain is at maximum. A simple rule worth remembering is that every dB of compression applied (as read off the gain reduction meter) reduces the signal-to-noise ratio by 1 dB. On a vocal track, this can lead to an increase in hiss, breathing noises and general background noise between phrases or during pauses, which can be a real problem if the singer needs a lot of compression.
To counter this undesirable side-effect, many compressors are fitted with expanders or gates, and though there is a subtle difference between the workings of expanders and gates, the way in which they are used in this context is identical, as is the subjective result. Essentially, an expander works like a compressor in reverse — the user sets a threshold, and when the signal falls below that threshold, the gain is reduced. Gates are similar, except that when the signal falls below the threshold, the gain isn't just reduced, the signal is shut off completely. However, a gate that closes too rapidly can chop off the ends of wanted sounds that have a long decay, especially those with long reverbs, so most gates (and expanders) of this type are designed to shut down slowly.
In use, the gate or expander threshold is set just above the background noise floor so that as soon as the singer (or other sound source) makes a sound, the gate opens instantly. When the singer finishes a phrase, the gate will close down over a period of a second or two and mute the background noise completely. Figure 1 shows the gate action in the presence of background noise and signal. It is very important to ensure that the gate threshold is set as low as possible to avoid adversely affecting the wanted signal.
Unlike dedicated studio gates, which have attack and release time controls and may also have side-chain filtering, the expander/gates fitted to compressors are invariably very simple, often with just a single threshold control and a single LED to indicate whether the gate is open or closed.
Many studio compressors come as dual channel devices, and though the channels can be used independently, it is usual to be able to run them as a stereo pair by means of a stereo linking switch. This is desirable when compressing a stereo mix, as it is important that the same gain reduction is applied to both channels at all times. The reason behind this might not be immediately obvious — you might think that if one channel contains a loud sound, then that channel should be compressed the most.
The best way to expose the flaw in this reasoning is by example. Let's assume that we have a stereo mix which comprises a simple string pad sound at a modest level, and at some time, a loud orchestral stab occurs only in the left channel. If the two channels of the compressor were allowed to act independently, the gain in the left channel would drop as soon as the stab occurred, while the right channel would be unaffected. As soon as the gain dropped, the level of the string pad would go down too, which means that the string pad would be momentarily louder in the right channel than in the left. This would make the string pad appear to move across the stereo mix towards the right-hand channel for a moment, which sounds most unnatural.
When linked for stereo operation, both channels of the compressor react to a mix of the sound passing through the right and left channels, so both channels react in the same way, regardless of which channel is the loudest. This produces a more natural sound and avoids the spurious image shifts that occur when both channels are allowed to do their own thing.
Of course both channels must be set up in exactly the same way for this to work properly, but that's taken care of for us by the considerate designers; when the two channels are switched to stereo, one set of controls usually become master for both channels, though some manufacturers opt for averaging the two channel's control settings or for reacting to whichever channel's controls are set to the highest value. Personally, I like the 'one channel becomes master' option best, as it leaves less room for pilot error.
Usually, compressors simply control the level of whatever signal is passing through them, but they can also be used for creative purposes by passing one signal through the compressor and feeding a different signal into the side-chain. (This is only possible on compressors that have side-chain access jacks on the back panel, but many do.) Why should we want to do this? Well, as an example let's assume that a piece of background music is being played through the normal compressor input, but that the side-chain input is being fed with a voice signal from a mixer or wherever. Figure 2 shows how this is set up in practice. Now the compressor is no longer listening to the signal passing through it but to the new voice signal. When the voice exceeds the threshold set by the user, the compressor will apply gain reduction — to the music signal passing through it. And when the voice finishes, the gain will return to normal at whatever rate is set by the release control.
Does this sound in any way familiar? The technique is called ducking, and is used all the time in broadcast to allow a commentator's voice to be heard over the music background. If the threshold is set so that gain reduction occurs as soon as the voice appears, the compressor will obligingly turn down the music, allowing the voice to be heard. Exactly how much the music will be turned down depends on the threshold and ratio settings; some experimentation will be necessary. The attack time should normally be set fairly fast, but the release time should be long enough to stop the music surging back in too abruptly. A release time of one or two seconds is a good starting point. (In my experience, ducking is easier to achieve using a gate equipped with a dedicated ducking facility; this will be covered more fully when we look at gates.) Ducking isn't by any means restricted to voice-overs, though. It can be used subtly but effectively to control the level of certain instruments in a mix, to allow solos to cut through. For example, during a guitar solo, a compressor could be used to pull down the level of the rhythm guitar or keyboard pad part by a few dBs. Even ducking a sound by just 2 or 3dB will make a noticeable difference to the balance and you'll be surprised at how effectively it works on rock or other high energy music. However, don't duck the drums or the bass guitar — it'll just sound as though someone is messing with the level of the whole mix. In an ever-changing world, the rhythm section, at least, should remain constant!
Another very common use of the side chain is to patch in an equaliser so that the compressor can be used as a de-esser. As its name implies, a de-esser is used to tame sibilant S and T sounds which, with certain singers, can be disturbingly loud once the singer is miked up. Figure 3 shows a compressor set up as a de-esser — a parametric equaliser offers the maximum flexibility, but a graphic can be used just as easily.
The first step is to find out at what frequency the sibilance is most annoying; this is achieved by listening to the output of the equaliser and then pushing up the sliders, one at a time, until the sibilance becomes most obvious. Depending on the singer, the effects may occur anywhere between 3kHz and 8kHz. If you boost only these frequencies using the equaliser, the compressor will respond to them more than to the rest of the audio spectrum, which has not been boosted. For example, 10dB of gain at the sibilant frequency means that the compressor will react to those frequencies at a signal level 10dB lower than it would react to other sounds.
This solution to sibilance is not ideal, because whenever a sibilant sound occurs, the gain of the whole sound is pulled down for an instant, but if the compressor threshold and ratio are set so that just enough gain reduction is applied, the results needn't be too obvious.
Finally, while there are no hard and fast rules when setting up compressors, it is inadvisable to set both the attack and release times to their minimum settings, as your compressor might just be fast enough to track the envelope of individual cycles of the input waveform. If this happens, it causes audible distortion, as the act of compression actually reshapes individual cycles of the input signal; though some manufacturers build in a deliberate delay time to prevent this happening, you can't count on it.
The accompanying table shows suggested compressor settings for different instruments and sounds, but don't stick to them rigidly, as different models react differently, and my idea of a nicely compressed sound might not be the same as yours. Not all compressors have an Auto feature, but where this column is checked, using the Auto facility is likely to give good results.
Feature by Paul White
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!