Most recordings sound amateur without it - but the art of successful compression is to make it as 'transparent' as possible. Bob Dormon explains how uneven guitar-playing, singing, and mixing can be expressly suppressed by a compressor...

Last month I described a noise gate as a tool to help a dodgy recording sound decent. But what about a dodgy player? Well, this is where that ubiquitous studio accessory; the compressor, comes in. A compressor can help smooth out uneven playing by acting as an automatic volume control. When things get loud, they get turned down; when they're quiet, they're unaffected. The overall effect is a consistent sound without any vast differences in level.

Instruments can be loud or quiet depending on how they are played. The playing style will govern what is called the dynamics (or dynamic range) of the instrument. The dynamic range of any recording medium is limited, so capturing the dynamics of a particular instrument can adversely affect the recording of the instrument in one of two ways:

1. If the instrument has been recorded too quietly, you hear the inherent noise of the recording medium;
2. If the instrument has been recorded too loudly, the recording medium distorts.

The simple solution to this problem - besides practising your playing or having an experienced engineer to 'ride' the faders - is to use a compressor. A correctly-set compressor will allow you to get the hottest signal to tape, thus reducing the chances of noise while ironing out any transient peaks that could lead to distortion. One reason why, once they've invested in a compressor, many studio owners will only part with the treasured item when they want to buy a better model...

Compression is cool, and anyone interested in serious recording should get used to the idea that they are going to need it sooner or later. Admittedly, if you're totally MIDI'd up then you can probably avoid using compression while putting a backing track together, but what do you do when the vocalist turns up? The Italian for good singing is bel canto but can belto seems more appropriate to many of today's untrained singers. Rather than melt your tape heads, the sensible application of a compressor will allow you to capture careless whispers or the sound of the crowd all on the same track. And it's nothing to fret about; you'll only be following the recording methods employed on practically every contemporary recording made today.

Compression is everywhere. The secret of making a good recording, however, is that the listener doesn't notice you're using one. The compression effects should sound natural and reasonably dynamic. All you're doing is attempting to contain the dynamic range to something more appropriate - squashing the signal to match the limitations of the tape, or, to put it another way, compressing it.

The functions

Depending on the complexity of your compressor, you will get a number of controls to manipulate an incoming signal and alter the compressor's response to it. Typically, these are: Input Level, Threshold, Ratio, Attack, Release, Gain, and Bypass. Metering is also important, as you need to examine how the compressor is responding to the incoming signal while setting the thing up. On the popular Drawmer models there's just one meter with a switch to allow you to view either the VU output or the amount of gain reduction; dbx compressors usually have two meters to show this vital information simultaneously. But setting up a compressor is not just about looking at meters to get a textbook result. It is very much to do with using your ears to decide just how subtly or brutally the effect should be used.

If you over-compress a recording, reviving it can be a tough task. If, on the other hand, you don't use enough compression you can always add a bit more when mixing, although initially you run the risk of getting a distorted or noisy recording. This is where the Bypass switch comes into its own; if you're not sure what the compressor's actually doing for your track, then switch it out and draw your own conclusions. Consider what you like or don't like about the compressor's action, then try to adjust it to compensate.

Typically, novice users over-compress things so that they may look great on the meters but sound flat to the ears. The problem usually stems from the idea that the compressor should be doing something to the signal all the time. This is not always necessary; in the case of vocals, for example, the compressor may only need to work on the louder parts, leaving the rest of the performance virtually unprocessed. Consider the ability of the singer; many have exceptional mic technique, and can perform their own volume control by moving in and out of the mic. Your job could be easier than you think!

Input Level

Well, this may seem pretty obvious, but it does in fact have quite a significant effect on the response of the compressor. As you change the input level, you also alter the area of the signal's dynamic range where the threshold control functions, and in so doing you change the point at which the compressor begins working. In conjunction with the threshold control, the input level conveniently provides a way of choosing how hard you want to drive the compressor.

Threshold

This works in a similar way to the threshold control found on noise gates. It sets a point that incoming signals have to cross in order for the compressor to begin the process of gain reduction. If a signal falls below this point, it passes through the compressor unaffected, but when it exceeds the threshold setting the compressor will process the signal and reduce its gain by a factor set by the Ratio control.

Ratio

As a compressor is a device used for gain reduction, you may want to be more selective about this process. Cheaper compressors will often do this automatically as they sense how hard the compressor is being driven by examining the input level. On more expensive models, the choice is yours. Setting the ratio on a compressor determines by how much the incoming signal will be attenuated (or reduced in level/gain). It's a mathematical function that simply divides the amplitude of the signal when it exceeds the threshold. This produces a corresponding drop in level at the output..

So, if a signal exceeds the threshold by 12dB and you have set a compression ratio of 4:1, then the output level will be 3dB.

If the same 12dB input signal exceeded the threshold with a 2:1 ratio, then 6dB would appear at the output.

The higher the Ratio value, the more severe the compression will be on signals once they exceed the threshold. Using the Ratio and Threshold controls together, you can specify just how loud your recordings get. In doing so, you're containing the dynamic range of your music and tailoring it to fit the limitations of your recording media.

Talking of limitations, there is a special kind of compressor called a limiter. Limiters use very high compression ratios; their function is to ensure that no matter how loud the input level gets, the limiter's output will never exceed a certain level. It's like a brick wall for signals - guaranteeing that nothing too loud gets through. It's the audio 'bouncer' at work again, but this time no-one gets past the brute. I'll talk more about the use of limiters later on.

"If the attack is too fast, you can get some nasty blips as the compressor wrestles with high signal levels"

Attack

Whether you're using a noise gate, a compressor, a sampler, or just a farty old analogue synth, the Attack control performs a similar function: altering the response time of the device. With a compressor, the attack rate determines how quickly the gain reduction of the compressor takes place once the threshold has been exceeded. If the attack is too slow, then the compressor may well have missed the boat and the loud sound will slip through unprocessed, while the compressor kicks in on a quieter sound sometime later.

It's as though our audio bouncer has a hangover, and instead of apprehending the villain, takes it out on the next few unfortunate punters. If the attack is too fast, you can get some nasty blips as the compressor wrestles with high signal levels. Also, a fast attack can really stifle a sound.

With practice - and as always with recording, using your ears will help - you'll soon get a feel for the sounds that benefit from short or long attack times, but I've highlighted a few in the Tip Sheet side panel.

Release

The Release control also works over time. While attack times normally cover the range from 100µsecs to 100ms, release times are more sober, and typically span the 100ms to 10secs range (although manufacturers pride themselves on producing devices with ever faster attack and release times). Adjusting the release time alters how soon the compressor recovers after it has compressed a high-level signal and returns to normal when there is no signal exceeding the set threshold.

If your singer is belting it out for the chorus with the compressor merrily doing its duty, and then whispers a few lines, you'll want to make sure that the compressor is able to let the quiet material through without reducing the gain of them, too. This is when you need to set the release time to an appropriate speed, so that the compressor can accurately track the sudden changes in signal level and produce a relatively smooth but musical output.

Output Level

As the main function of a compressor is to actually perform the task of gain reduction and effectively turn down the level of loud sounds, the output signal will ultimately end up quieter than the original input signal. To compensate for this drop in level, compressors have an output gain control to raise the output volume and bring the compressed material up to a more usable level. The amount of adjustment needed depends largely on how hard you are driving the compressor, where you've set the threshold, and the compression ratio being applied.

Metering

While using your ears will figure largely in the recording process, there are times when meters speak louder than words. Generally, the metering on a compressor will display the output level and the amount of gain reduction that is being employed. Being able to see the changing levels of gain reduction allows you to determine where your threshold point should be, and any adjustment needed from the input level control.

When setting up a microphone for a vocalist, it's no use just asking them to speak. Play them some of the track so you can get an idea of how loud they'll be singing and make your adjustments that way. You'll have to explain that their headphone vocal level may change as you set things up. (Singers tend to go quiet very quickly when there's something weird going on in their cans.)

The gain-reduction metering will help you find a suitable threshold for the louder and quieter passages, while the output level will assist in preventing distortion (from too much output gain) or noise (from too little) as well as ensuring that there's a reasonable output signal going to tape. The gain-reduction metering also displays the response times of both the attack and release controls; observing the rate of change can help in finding an appropriate speed that suits the dynamics of the source material that's being compressed.

Linking

Quite a few of today's compressors come in pairs (one unit housing two identical compressors) or even in sets of four. While you can always use the modules independently, switching in the 'Link' function will enable, say, two of the compressors to act as one. This is ideal for compressing stereo sound sources - a keyboard, a stereo sample, or a whole mix. As one side of the stereo image may sound louder than the other at times, it is important that both sides are reduced in level (compressed) when this occurs. If the compressors were to work independently, then the actual stereo image would shift, as one side would get quieter while the other remained unchanged. To prevent this so-called 'centre shifting', the level-sensing circuits in both compressor modules act in tandem, causing identical gain reduction in both channels regardless of which side of the stereo image is causing the compression to be applied.

Sidechains & de-essing

The sidechain of a compressor is the part of the circuitry that actually detects the changes in level and then - with the aid of the controls above - produces the compression that features on the final output signal. This sidechain is often available for external control. As with a key-trigger input on a noise gate, you can externally dictate the action of the compressor on a signal fed into it as normal.

A typical application of this is de-essing. De-essing is a method of reducing the sibilance that is frequently a side-effect of singing. Lots of esses soon start things sizzling. (Try saying that out loud and you'll soon realise what I mean.) A way of overcoming this is to selectively compress those high frequencies. First, you split the vocal into two signals and feed one into the compressor as normal. You send the other signal to an equaliser and boost the high frequencies and cut out all the lows, so that the sibilant sounds dominate the signal. You then send this equalised signal into the sidechain input and the compressor will respond to those exaggerated high frequencies. You won't actually hear the sidechain input, but you will hear its action on the other sibilant vocal that's been fed into the compressor as usual. The effect is that the compressor will suppress the high frequency sibilant sounds only when they occur and leave the rest of the vocal unchanged.

You will have to adjust the compressor's response to the sidechain; to this end, some compressors have a switch called 'sidechain listen', which enables you to hear the signal that is present at the sidechain input. You may also have to switch in the sidechain control, though more compressors these days have a break-jack arrangement that detects when the sidechain input is being utilised. De-essing like this can be time-consuming, but while there are de-esser units available that specifically deal with this task, using a compressor's sidechain is still the obvious route for those with restricted rack space and/or budgets.

"Sensible application of a compressor will allow you to capture careless whispers and the sound of the crowd all on the same track"

Limiters

When compression ratios are in double figures, the compressor itself takes on the role of a limiter. A limiter will have a ratio of 10:1, 20:1, or maybe even 100:1 - sometimes written as ∞:1. A limiter will be used to guarantee that the threshold, when exceeded, will only produce a fractional increase in output or none at all. Limiters have extremely short attack and release times so that the listener doesn't detect their brief interaction.

Their use is usually to hover above the program material and act as a goalkeeper (or bouncer) should any transient peaks appear that would distort the equipment. Limiters find favour in live recording situations, particularly where digital equipment is being used. Unlike analogue recorders that produce a gradual increase in distortion when overloaded, digital recorders either drop out or create a completely unmusical fuzz. Using a limiter ensures that the levels don't get so hot that this occurs.

Limiters are also abundant in broadcasting. If Noel Edmonds' broadcast console were to perform a national service and self-destruct while he's on air, then if any loud erroneous noises were produced from this supreme sacrifice, a limiter would kick into action to prevent the transmitters from overloading and from damaging thousands of televisions the length and breadth of the land.

By and large, limiters are employed for safety reasons in order to prevent damage to sensitive/expensive equipment, or to avoid distorting a recording medium.

Types of compressor

Compressors have been around for some time, so it should come as no surprise that the first designs were built around valves. The characteristics of valve compression are highly revered, and rightly so. Valve compressors tend to have a 'soft' response to signals exceeding their set threshold, plus harmonic distortions that are considered excessive by today's standards, yet have a characteristic sound that has retained its appeal. You may have seen 'soft knee' as an option on some solid-state compressors. This is an attempt to mimic the gentle threshold/ratio response of valve models.

Another quite simple device for compression is the lightbulb. As the signal going through the bulb gets 'hotter', so too does the lightbulb, until eventually it illuminates and dissipates the excess signal energy. A friend of mine once had a guitar amp that was too powerful for the cabinet, so he inserted a lightbulb between the amp and the cabinet. It gave new meaning to the phrase 'a bright guitar sound'!

Employing a similar method is the opto-electrical compressor. Today this type couples an LED and an LDR (light-dependent resistor). The LED illuminates and the LDR tracks its response, performing the necessary gain reduction on the sidechain circuitry it acts on. As the response to the increase in light intensity is not linear, the way this circuit behaves has characteristics of its own that are also considered desirable.

Finally, there are the modern-day FET (field effect transistor) models. While these are less likely to exude warmth, their qualities include rapid attack and release times, plus low noise and relatively low cost. The FET varies in resistance in proportion to a voltage derived from the input signal fed into it. This changing resistance is used to vary the output signal, working basically as a VCA (voltage controlled amplifier) fader.

Expanders

An expander performs pretty much the opposite action to a compressor. When things get loud, a compressor turns them down. By contrast, an expander makes the loud bits louder and the quiet bits quieter. Where the loud/quiet crossover takes place is determined by the threshold. Setting an expander up is slightly easier than a compressor, as the results are far more obvious - yet the controls are much the same. While dedicated expanders are available, many compressors now provide an expander function.

Compression and expansion (or compansion for short) is a process employed by various noise-reduction systems. Simply put, material is compressed when recorded and expanded when played back. The compression allows a good signal level to go down to tape, while the expansion process reduces the level of background noise and tape hiss when it makes the quieter material even quieter on playback.

"If you over-compress a recording, reviving it can be a tough task. If you don't use enough compression, you can always add a bit more when mixing"

Summary

One thing about compression: there's a lot of it about. Dear old Simes (Simon Bates, formerly of Radio 1, now residing on LBC) has probably the most compressed voice on radio. It doesn't matter whether he's reading a romantic listener's letter or the UK Top 40, his voice drills into the ear like a lawnmower on a Sunday afternoon. Even the records he plays are compressed. Not only at the recording, mixing, and cutting stages, but also by the broadcasting engineers so that they can pump out a loud radio transmission that's had all the peaks ironed out.

The same goes for TV audio, too. Alas, some films still come across quieter than the ads. It's not supposed to happen that you get blasted out of your chair after the soothing sounds of Inspector Morse - but, needless to say, the advertisers love it.

When it comes to using compression yourself, just remember in the words of the old song: 'Don't fear the meter...' Use your compressor to help you get a healthy recording level. You may have limited resources, but you could be pleasantly surprised by the performance of many low-budget devices. Phil Collins was well impressed with the 50-quid Allen & Heath compressor he bought. He used it at home to record Face Value and has stuck with it ever since. It's the compressor he still uses in the studio for his vocals...

Hear the dulcet tones of Bob Dormon guide you through the mysteries of compression on this month's ReiMix CD. Topics covered include vocal, guitar, bass, and drum-kit compression.

- Compression Tutorial 1
- Compression Tutorial 2