This time we will cover dynamics processors (compressors, limiters, expanders and noise gates), explaining their basic principles and many of the techniques commonly used with them. The first devices capable of controlling the dynamic range of a signal appeared in the field of telephony. They were used to keep the amplitude of the telephone signal from exceeding certain limits (limiters) or to hold the signal down at its highest points while raising it at its lowest (compressors), producing a signal with a higher apparent level.
Not long afterward, these devices began to be used in the music recording industry, and today practically every commercial record uses dynamics processors of some kind. Despite this widespread use, some engineers are against using these processors, especially compressors, while others consider them an indispensable tool in their daily work.
Whichever side we are on, the truth is that in recent years the war to "sound louder than everyone else" has driven up the use of limiters and compressors, even making the dreaded, somewhat mythical inter-sample peaking a reality on some records.
Basic Operation
All dynamics processors work on the same principle. An amplitude threshold is set, and the device reduces the signal's amplitude depending on where that signal sits relative to the threshold: compressors and limiters act when the signal exceeds the threshold, and gates and expanders act when it is below it.
To determine how much to reduce the level, a ratio of the form A:B is set, where A is the input level and B is the output level. For example, with a compressor set to a -6dB threshold and a 4:1 ratio, a signal with an input level of -7dB will pass through the device unaltered, since the device does not act when the threshold has not been exceeded. However, if the signal reaches -4dB at the input, the 2dB excess above the threshold will be reduced at a ratio of 4 to 1.
We can say that the only difference between a compressor and a limiter is the ratio. When ratios of 10:1 or higher are set, we usually consider that we are limiting the signal, and below those ratios we speak of compression.
Similarly, gates are used to remove the signal below the threshold, while expanders attenuate it.
Other Parameters
Makeup Gain
Using a compressor often reduces the overall level of the signal, making it seem weaker. Makeup gain is a gain stage at the compressor's output that lets you restore the signal to the apparent level it had before processing.
Attack and Release Time
Attack time, generally in milliseconds, is the time the processor takes to start attenuating the signal once the threshold has been exceeded, and release time is the period it remains active even after the signal has dropped back below the threshold.
These two parameters are very important, since they determine which section of the waveform envelope we apply the amplitude reduction to, and therefore define the processor's behavior. Short attack and release times act on the transients of the signal. Reducing these fast amplitude peaks lets you, for example, raise the level of the rest of the signal without risk of clipping.
If we use longer attack times, we let the transients pass through without being compressed, and act on the rest of the signal, increasing the difference between the peak level and the average level of the signal. This increase creates the subjective impression that our signal is stronger and more aggressive.
Common Uses of Compression
Evening Out Dynamics
A vocalist can't sing every word at exactly the same level. There are variations due to pronunciation, intonation and so on. Likewise, electric basses often have quite variable dynamics, with some notes sounding louder than others.
In these cases, a compressor can be used to even out the dynamics of the performance: by reducing the level of the loud parts we bring them closer to the level of the soft parts, getting a more uniform overall result. Ratios between 2:1 and 5:1 are typically used for this task, and the threshold is chosen to achieve reductions of up to 6dB.
If we also use the output gain to bring the signal back to its input peak level, we will get a signal with a higher apparent level, which also increases the instrument's natural sustain.
Be careful with the attack time: one that is too fast can make the highs disappear quickly or hurt vocal intelligibility. One way to find a suitable attack is to start with a slow time and shorten it until you hear the sound start to get darker.
As an Effect
When compression is used as an effect, the ratios are usually more aggressive, and it is common to see 10 or 12 dB of reduction, even going to 18-20 dB, or more, in some cases.
A typical example is applying heavy compression to the kick drum to give it more punch. Again, the attack time must be chosen carefully, since the goal is to let the kick transient pass cleanly through the compressor and act on the "tail."
If you're not sure what a fast attack time is on a kick drum, try zooming in on a single hit of this instrument. Select the whole hit with the cursor, and make sure (by listening to the selection) that you're not leaving out part of the sustain. It will probably last about 200-250 ms. If we now check how long the transient takes, we will see it is around 15-25 ms. With these numbers in mind you can get an idea of where the attack is and where the tail is, and start trying times on the compressor.
If we also set the release time so that the compressor returns to unity gain in time for the next kick drum hit, we will get the classic effect of the compressor "breathing" in time with the music.
Parallel Compression
When you want the benefits of compression without it being excessively obvious, a very useful resource is parallel compression.
This technique consists of sending a copy of the signal you want to process to a compressor and applying parameters that are generally somewhat aggressive. The result is blended with the original sound, so you hear the compression, but not as clearly as if we had applied it directly to the original sound.
Try sending the kick and snare to an aux channel, apply a good dose of compression, and then boost the compressed sound around 180 Hz and around 10 kHz. When you blend it with the original tracks, the sound will have gained body, brightness and power. Careful, it can be addictive.
Multiband Compressors
These processors are made up of several compressors, each working on a different frequency range.
Depending on how much we compress or expand each band, we will be applying a certain amount of EQ, but also controlling the dynamics of the signal.
Multiband compressors are a very powerful tool. They are typically used at the mastering stage, although it is not unusual to use them on an instrument during mixing, when EQ alone doesn't give the desired tonal balance.
Common Uses of Gates and Expanders
Gates
Noise gates are used to get rid of unwanted noise in the signal.
A typical example is applying a gate to the signal coming from a guitar amp, to remove electrical noise in the passages where you're not playing. Another very common use is on the tom mics of an acoustic drum kit. When the toms aren't being played, the signal from these microphones tends to muddy the overall sound of the kit. A noise gate on each mic lets us get rid of that unwanted signal, letting sound through only when the toms are struck and the signal exceeds the threshold.
These processors can also be used for creative effects. As we will see in the next issue, gated reverb is an effect that was hugely fashionable in the 1980s, giving the drums of that era their characteristic sound.
Expanders
Sometimes gates work very abruptly, and for a more natural result it is preferable to lower the level of the noise rather than eliminate it completely.
If we want to get rid of breaths and small lip noises in vocals, an expander is ideal. It lets you reduce these noises by several dB without removing them entirely, which would sound too artificial.
Sidechain Processing
So far we have fed these processors with the same signal we were going to compress or expand, but what about compressing one signal based on another? This is called sidechain processing, and the signal that drives the processor is called the key signal.
When we enter the realm of sidechain processing, the creative uses of dynamics processors skyrocket. For example, one way to improve the groove of the drums is to apply an expander triggered by the snare to the microphones above the kit (overheads). Using a small amount of reduction (1 or 2 dB), we get the overheads to increase their level slightly on every snare hit, boosting the snare's punch and adding a new rhythmic element to the performance.
In a similar way we can improve the interaction between the kick and the bass. If some of the bass notes have drifted a little and don't sound exactly in unison with the kick, we can put an expander on the bass track and trigger it with the kick. Again, we use a small amount of reduction (1 or 2 dB) so that every time the kick hits, the bass rises slightly in level, giving the impression that they are working perfectly in unison.
Sometimes the key signal itself is processed. If in the previous example the kick track also contained snare hits, we could filter the signal sent to the expander below 200 Hz so that it responds only to kick hits.
In fact, a de-esser is nothing more than a compressor triggered by the voice itself, but with the key signal filtered beforehand so that the processor responds only to the sibilance frequencies (around 6-10 kHz). Following the thread of sidechain processing, we move on to the audio example. We have selected a technique called frequency-dependent ducking.
Audio Example
We all hear an example of ducking every day on the radio: the music level drops automatically when the announcer starts talking, to give the voice prominence and improve intelligibility.
This technique can be applied more subtly to create space around the voice in a mix. The idea is to make any instrument that interferes with the vocal performance drop in level when the voice is present, but only at certain frequencies. This is called frequency-dependent ducking.
In our example we are going to implement the ducking on the guitars. First we must create a copy of each track we want the ducking to affect. The level and position in the stereo field must be exactly the same as the original tracks.
Next we insert a noise gate triggered by the voice on each copied track, and invert the phase of the gates. At this point, every time the voice shows up it opens the gates and, as the original guitar tracks mix with an exact copy that has inverted phase, the sound on those tracks will cancel out completely.
Obviously, this is very drastic. By lowering the level of the copied tracks we could make the cancellation partial and therefore less obvious, but our goal is for only certain frequencies to be canceled.
By inserting an equalizer after the noise gates, we can control which frequencies get phase-inverted. So what we need to do is set that EQ so that it lets through only the frequencies we want to cancel.
In our case we looked for the frequencies that benefited the voice most, and concluded that it gained body around 180 Hz and improved intelligibility around 2900 Hz. The EQ applied consisted of a high-pass filter at 149 Hz, a low-pass filter at 3198 Hz and two band-reject filters at 433 and 917 Hz, respectively.
The net result of the EQ lets through only the frequencies around 200 and 2900 Hz, approximately. This way only these frequencies pass when the noise gate opens, and they are therefore the ones that cancel against the original tracks, leaving the rest of the spectrum unaltered.
To refine the ducking behavior, we can vary the level of the inverted tracks to increase or decrease the degree of cancellation, and adjust the gates' attack and release times so that they open and close smoothly.
