A QUICK RECAP ON WHY WE USE COMPRESSION
OK, so you record some vocals on your new song. Naturally, it has it’s louder moments and it’s quieter moments.
You haven’t compressed it yet. Maybe you have, but you’ve simply evened it out a touch while recording was taking place, using a compressor with a very low ratio setting and/or a very high threshold level to be safe and minimalist in your impact on the sound.
The vocal take is perfectly executed in terms of pitch, rhythm and emotion. This is great! There’s one little problem, though…
A few plays through the track soon reveal that there is no fader position that you can leave the vocal take at that works from start to finish. Every position results in noticeable problems with level somewhere during the song. It’ll take more than one or two changes in position to deal with this.
Either the vocal is way too loud at times, or it’s way too quiet.
This happens on a syllable by syllable basis, or even song section by song section.
No matter whether the issues are on the micro or macro level in your song, you can’t solve this problem by leaving the fader in a good place. You have to reduce the dynamic range to minimize these fluctuations.
You can try to manually whip the fader around to different positions during the mix on the fly, but you can already see that words may work OK but this isn’t going to be a realistic proposition at the syllable level, let alone the (even harder to treat) consonant level within syllables. It’s impractical.
a) enter automation data to control gain changes
b) use a compressor to take care of gain changes for you
Once these actions have made the sound’s level appear stable to you throughout the mix, or at least good enough to automate subtleties later, you can safely move on to worry about any other unstable faders that persist.
The fader instability is due to the level variations during the song on that track. Compression is a good candidate to solve this issue, since the dynamic range can be shrunk, reducing fader instability.
This is why compression normally involves two stages.
In a nutshell, you turn down the loud bits of your signal first, and then you can turn up the whole signal to return to the starting gain value.
Ideally, your (pre-compression) input level arriving at the compressor will match the (post-compression) output level leaving the compressor.
The end result is a much more stable signal level that, naturally, requires less attention to control moment by moment through the track.
The fader for that track is no longer so unstable during the mix. Some instability may remain, but you will have less work to do to contain the fluctuations with fader automation and other techniques.
In fact, any modern commercial recording will have seen quite a few hours spent on the lead vocal’s level throughout the song, syllable by syllable, typically using several dynamic processing techniques together with detailed fader automation. The lead vocal sound is typically quite different in the verses than in the choruses, for example, in most modern pop productions. The middle or bridge may well call for another sound again.
Typically, it is impossible to set your vocals so they stay at the perfect level from top to bottom of your mix unless you use compression for overall control and follow that with detailed fader automation moves for any necessary remaining tweaks. If you have radical changes in sound from section to section, then you will be likely to use different compression settings and/or compressors in each section.
COMPRESSION IS A TWO-STEP PROCESS
First, you set a threshold level that triggers compression only when you want it, thus reducing the distance between the loudest and quietest parts of a signal.
This has the intended effect of compressing its dynamic range down into a smaller range of values. Lowering the gain values of the peaks closes the gap between their values and those of the quieter bits of the signal, which remain unchanged.
Once the signal is compressed appropriately, the second part of the process occurs, wherein you turn up the signal back to its original pre-compression level, using a “make-up gain” control.
You are matching input and output levels deliberately so that the level sounds as loud to you when it’s leaving the plug-in as it was coming in. Input matches output.
This maintains a good, clean gain structure flowing through the DAW channel strip. You can then use your channel faders, which come later in the channel than any inserted plug-ins, to balance the now-compressed sound in your mix.
WHAT ARE ALL THESE COMPRESSION PARAMETERS?
Well, here’s a rundown on them, one at a time. They do interact to a degree, so you often have to revisit settings after changing other settings.
Threshold controls the level at which the signal is loud enough to trigger the onset of compression. Until the signal crosses the threshold level, the compression circuit (or algorithm) does not act upon the signal.
Set a very low threshold to make the compressor act on the signal more or less all the time, or a very high threshold to have it only act occasionally on the peaks. Threshold normally starts at 0dB and you can reduce it from there.
Ratio is a very important control. It’s a “more compression, please” knob. The ratio is the ratio of input signal to gain reduction at the output, before any “make-up gain” is added.
At a ratio of, let’s say, 4:1, any input signal that exceeds the threshold level by 4dB will be only 1dB louder at the output of the compressor.
If you selected an 8:1 ratio instead, then any input of 8dB greater than the threshold level will be reduced at the output to only an increase of 1dB above the threshold level.
If you set a ratio of 4:1 and the signal crosses the level of the threshold by 2dB, then that increase will be reduced to an increase of only 0.5dB at the output.
These examples make it obvious why you will use a make-up control to increase the output gain so that it matches the original input level.
This has the welcome side-effect of making your sonic judgments far more reliable, since you really can’t trust subjective decisions about two signals of unequal loudness. Once level-matched, you can hear more easily what effecct the compression is having on your signal.
Level-matching is a prerequisite to smart decision-making, due to the frequency response curves of human hearing.
The compression ratio you choose indicates how many dB of gain reduction will occur for every 1dB you go above the threshold level you have set with the threshold control.
The lower the threshold, the sooner (and more often) the compression works. Raise the threshold by a set amount, and the signal will have to be that amount louder in order to trigger the compression.
The higher ratios have a more pronounced effect, due to this rational inter-relationship between input and output levels.
Attack controls how quickly the compressor will start working on the input signal once the threshold level has been exceeded.
A quick (short) attack time is going to make the compressor react quickly after the input signal crosses the threshold level.
A slower (longer) attack time will mean the compressor responds more slowly, allowing transients through for a bit until the compression kicks in.
Using an attack time of less than about 16ms on a drum or percussion sound runs the risk of clamping down on the impact the drum has.
The sound of a drum can be thought of as being in two parts. There is an initial attack phase with a very short transient spike, a generally loud hit. This lasts but an instant, a handful of milliseconds. Then, we hear the second and typically longer part of the sound, the decay phase where the sound dies away to silence.
If you start to compress too quickly after that transient spike leaps across the input threshold level, the power of the drum hit will be damped down, and the drums will sound weedy and un-energetic, with clipped or clicky-sounding attacks.
This suggests a rule of thumb to stick to 16 milliseconds or longer with the attack time on your drum and percussion compressors if you want to avoid this terrible fate for your drum tracks. Keep them punchy and powerful by allowing the initial hit of the drum through and clamping down only on the boomier part of the sound that arrives slightly later.
Release sets how long the compression will take to stop acting on the signal once the signal has fallen below the threshold level.
When does the compression effect go away? When the release time is over, and not when the signal falls below the threshold. That’s just the cue for the chosen release time to begin.
Using the attack and the release can dramatically alter the rhythmic feeling the compression gives to the music. Choose musical times for the attack and release, ideally that breathe with the tempo.
Setting longer release times will force the compression to persist even though the signal has long since fallen below the threshold. This may or may not be what you want. Shorter release times will allow the compression to re-trigger appropriately because it will stop compressing very soon after the level falls back below the threshold.
As I said, release time only begins after your input level falls back below the threshold. Release is not found on every compressor, and some have a fixed release time, often called Auto-Release.
When the chosen release time has passed, the compressor action has ceased and the signal level is no longer being compressed. The input to output ratio will stay at 1:1 until the threshold level is exceeded once again.
On an 1176, it comes in the form of selector buttons, and you only have a few ratio values you can set.
On most compressors, it will be a knob that you turn to increase ratio. Sometimes, it is stepped to give a few discrete values instead, as with the ratio buttons on an 1176. More often, though, it’s a ratio knob that gives a continuously variable ratio from 1 upwards.
HARD KNEE AND SOFT KNEE
The knee decides the rate at which compression will start to happen after the input level has crossed the threshold.
The knee control governs the speed of the transition from “not compressing” to “compressing”, in other words.
This transition into compression can take place the instant the level crosses the threshold – this is a hard knee setting.
A soft knee allows a slower onset of compression. The slower the onset, the softer the knee. Softer knee settings are more gentle than the sudden clampdown of a hard knee setting.
You might get a knee switch with soft/hard positions, or you might get a control that lets you vary the knee continuously.
You will also find, due to the nature of the relationship between soft and hard knee settings, that the louder the input signal gets, the less marked the difference becomes in using the various knee settings.
If the signal is really a lot higher than the threshold a lot of the imte, then the soft and hard knee will give you about the same output level prior to make-up gain.
When input levels are hovering around the threshold, then there’s more of an audible difference between different knee settings.
The term knee derives from plotting compression graphically, where it takes the form of a knee joint, visually speaking, as this screenshot of the popular DigiRack Compressor/Limiter plug-in shows. This plug-in is bundled free with all ProTools software. Quite a few software compressors give a graphic indication of the compression characteristics you have set up.
The Digidesign DigiRack compressor plug-in has a very easy to use knee adjustment control, the lowest left knob marked KNEE. The knee shape is clearly visible in the graph window, which, as I say, will change in real time as you change parameters.
It is a great learning tool if you have any ProTools software, because it gives a visual indication of your settings as you change them.
It responds to changes in parameters as you make them by redrawing the graph on the fly in real time, which is very educational. Various dynamics plug-ins do this, but this one happens to be free with all current and many recent versions of ProTools software.
INPUT and OUTPUT are common controls on many compressors. A good example is the Universal Audio 1176LN. This is one of the simplest compressors to operate and they sound great, but an 1176 makes it all too easy to over-do the compression, since it has few controls. It does, however, have the serious advantage that you can easily dial in and hear the effect, thus getting used to what you are listening for a lot quicker.
If you’re a beginner at compression, it really helps to do this and get familiar with what compression sounds like. The fact the 1176 has less parameters also helps, but in some respects it may confuse you, so do try other compressors.
Almost all compressors offer a make-up gain, and in some cases it is simply the control labeled OUTPUT.
You already know why we need this control.
The compressor will determine if the input signal crossed the threshold using either Peak (Instantaneous) levels, or using RMS (Average) levels.
The difference between Peak and RMS levels in a signal is called the Crest Factor, and getting musically appropriate crest factors while still having the loudest peaks possible is a major goal of mastering engineers. They want the RMS level to be the right distance from the Peak level, so that the music has dynamic range, but is also consistently loud enough.
This is often a battle of incremental improvements to the audio painstakingly won by the mastering engineer over a fairly large number of plays.
In using tracking or mix compression, you will sometimes have a choice in how the compressor determines the threshold has been crossed.
You may have no say in the method. If you do, what are the differences to consider?
When you get a choice of Peak or RMS, you will find that Peak works best for sudden loud sounds like drums and percussion, whereas RMS may work better for sustained sounds like keyboard pad sounds or a vocal with a relatively static level.
The main downside with using Peak Detection is that it might make the compressor leap about in level dramatically, going in and out of compression too often and too easily due to a signal with a lot of sudden transient events that die away quickly (like drums).
Compression will be smoother based on an RMS detection circuit, but may be too slow to catch all the transients in the signal with “jumpier” sounds like a funky clavinet solo.
SOME COMPRESSION RATIO SUGGESTIONS
Very low ratios such as 1.2 to 1 are common in mastering applications, but in home recording you will use 2:1 or 3:1 for subtle compression. For more obvious compression, you’ll use 4:1 and above.
For a bass guitar played with fingers or a pick, 8:1 may be great. For slap bass, if you really must party like it’s 1985, you’ll be happier with a 20:1 limiting ratio, so the signal is the same level all the time no matter how loud the input gets, pretty much.
The higher the ratio is set, the more the compression is piled on. When the ratio is set to 1, there is no gain reduction because no compression is applied. You can leave the ratio the same, and add more effect by lowering the threshold. Alternatively, you can leave the threshold alone and raise the ratio.
It’s all about deciding where and when the signal needs compressing during the song, and setting the threshold and ratio accordingly so that the compressor action is only invoked where it is needed or wanted, and only as much as necessary for a good sound. If that’s all the time, then so be it.
Breathing a little is more musical, however, so watch out for wringing all emotion away from the track by leaving it lifelessly squished. You can use the attack and release times to get the sense of musicality going, from gentle smoothing to full-on, “pumping” effects. In tempo, these can be very effective in making the kit sound alive and exciting.
A ratio of 10:1 and above is basically limiting. In fact, 20:1 is an even more common limiting ratio.
Limiting is a means of ensuring the level can’t get any louder than a certain gain, no matter how loud the input level gets. It’s handy for catching sudden peaks, transients that would otherwise clip during recording. It often gets placed after a compressor on a vocal signal chain.
Mastering limiters are very high-quality ones that do minimal harm to the signal even when driven pretty hard.
The compressor evens things out with a few dB of gain reduction on the louder passages, and the limiter catches the sudden bursts of energy before they clip anywhere. This lets you turn the vocal up a lot, making it more “in your face” than without compression.
EXPANSION RATIOS AND INVERSE RATIOS
When the ratio is less than 1 to 1, that is called expansion, a similar but different type of dynamics processing that emphasizes the sounds above the threshold instead of compressing them. This is quite popular for turning the subtle details in a song down – you make the quieter bits quieter with expansion, rather than making loud bits quieter as in compression.
Expanders are often paired with gates. We’ll look at gating tomorrow in the next post in this series.
There are even some exotic negative ratios (e.g. -2: 1) which are pretty much useful only for outlandish effects. These ratios are called inverse ratios. You aren’t likely to need an inverse ratio too often, if at all, but they can be found if you’re interested, for example in the elysia mpressor. They can definitely sound pretty intense on drums, and are worth a look for that reason alone.
A QUICK WAY TO DIAL IN COMPRESSION
1 Lower the threshold to -20dB or so, and raise the input level to the compressor if necessary so it can cross the threshold level easily
2 Raise the ratio to 4:1 or so
3 Set a fairly fast attack time, and a moderate release time
4 See what you get and adjust to taste, usually by reducing the effect. Back it off to a lower ratio, and/or raise the threshold, and/or soften the knee.
You’re better off with less compression rather than more, because it’s virtually impossible to undo compression successfully. Put it on, and you’re stuck with it.
Raise the threshold, and/or lower the ratio, to back off the effect. It’s a subtle effect and it may be a lot more noticeable than you realize, if you are listening when tired or when your ears are fatigued from a long or overly loud session. Be careful.
Lowering the ratio is usually a more dramatic influence on the degree of audible compression.
The compressor should now be compressing the signal most of the time, if not all of the time.
This lets you hear what you are doing, decide what is going too far for your tastes, and adjust accordingly. Beward of squashing the life out of your track.
Using too much compression will rob your music of it’s high frequencies and emotional energy, leaving a dull, lifeless impersonation of music in it’s place.
Tomorrow, we will spend a little time looking at expanders and noise gates as well as the incredibly useful “New York” compression method, a technique known nowadays as parallel compression and very widely used. It fattens sound up no end, and you will need to know how to do it yourself.
So we’ll look at using dynamics processing in series and in parallel, and we’ll consider the noise gate and the expander.
The following day will be the final blogpost on dynamics processing, in which we will look briefly at the topic of multi-band processing, a fantastically powerful method of compressing only specific frequencies. This awesome tool lets you make the bass sound very pressured and intense, while leaving the midrange and the highs unchanged, for example. There is a problem, though.
It’s a very dangerous tool, misused, and will ruin a mix in no time. It’s also hard to use properly.
Used wisely, it lets you really rock the house with the bass end amongst various other uses. It is mostly used in mastering applications, due to it’s complexity and the experience needed to apply it safely.
See you tomorrow for expanders, gates, sidechains, key inputs and frequency-conscious compression (e.g. de-essers).