Compressors, limiters, and noise gates provide various ways of controlling the amplitude of a signal.
To do their job, these algorithms have some method of sensing the amplitude of the incoming signal — either an envelope follower (see the section on filters, above) or the equivalent. The information coming from the envelope follower is used to adjust, in one way or another, the gain of the signal being passed through the effect.
To explain how the devices work, we’ll start by talking about compressor/limiters (sometimes called comp/limiters), because noise gates turn the process on its head. Trying to explain them both at once would make for confusing reading.
A comp/limiter has a threshold level parameter, which is usually defined in dB. Signals that fall below the threshold level — that is, signals whose highest dynamic peaks are quieter than the level defined by the threshold parameter — pass through the effect without any change. For instance, if the hottest peaks in the signal are at -17dB and the threshold is set to -15dB, the comp/limiter has no effect on the signal passing through it.
When the incoming signal rises above the threshold, the comp/limiter reduces (attenuates) the level of its output. The amount of attenuation is controlled by the compression ratio parameter. Typical compression ratios range from 2:1 to 20:1.
I’m not even sure how to explain what this ratio means in a single concise sentence, so I’m going to give you an example. Let’s say the threshold is at -15dB, as before, and the incoming signal rises to -7dB.
That’s 8dB above the threshold. If the compression ratio is 2:1, the compressor will turn that 8dB increase into a 4dB increase (compressing the added gain by a 2:1 ratio). Thus the output signal will rise to a level of -11dB (-15 plus 4), as shown in Figure 9-4. The amount of attenuation is half of the amount by which the incoming signal exceeds the threshold, because the ratio is 2:1. If the ratio is increased to 8:1, the same 8dB rise above the threshold would be compressed to 1dB above the threshold, resulting in an attenuation of 7dB and an output of -14dB.
If that still seems a little abstract, don’t worry about it. The point is, a higher compression ratio will cause the comp/limiter to “squash” loud peaks more forcibly, and a lower threshold will increase the effect by causing the comp/limiter to look at lower-level signals as “loud.”
Figure 9-4. A compressor reduces the amplitude of signals that are higher than its threshold value, while letting lower-level signals pass through intact.
If the compression ratio is infinite, the comp/limiter will prevent the signal from ever rising past the threshold. This is called limiting.
Many comp/limiters have attack and release parameters. The attack parameter controls how quickly the dynamic level is reduced — how quickly the compressor starts to do its job — when the incoming signal rises above the threshold, and the release parameter controls how quickly the uncompressed dynamic level is restored after the signal drops below the threshold. A relatively slow attack setting (more than a few milliseconds) will allow the percussive attack transients of sounds to pass through without being compressed. This can add “snap” to compressed drum sounds.
Compression is often used to give pop music recordings a uniformly high dynamic level, which is felt to make the recordings sound better on the radio. Too much compression, especially when it’s applied to an entire mix, can squash the life out of a performance. Applied to individual drum samples, however, large amounts of compression can produce a startling and colorful effect. The compressor raises the low-level room ambience at the end of the sample to a point where it’s clearly audible.
In a noise gate, which is a type of downward expander, the amplitude of the signal is unchanged as long as it exceeds the threshold. The gate reduces the amplitude of its output (or shuts off the signal entirely) when the signal falls below the threshold, as shown in Figure 9-5. If the threshold is set to a low value, the gate can clean up a recording by removing background noise without audibly affecting the timbre of the music. When the threshold is set high, a noise gate can chop a signal apart and produce unusual stuttering effects.
Vocoder
Vocoders have been around (in analog form) since the ’60s. They’re used most often to make a novelty
“singing synthesizer” effect, but they’re capable of other musical effects as well. They’re a potent resource for processing drum loops, for example. A vocoder requires two inputs to function: a speech signal and a carrier signal. What it does is impose the moment-to-moment frequency spectrum (the overtone structure) of the speech signal onto the carrier signal.
Figure 9-5. A noise gate reduces the output level when the signal being processed falls below the threshold.
Figure 9-6. A block diagram of a vocoder. The speech signal is split apart by a bank of bandpass filters.
Envelope followers attached to these filters track the amplitudes of the frequency bands from moment to moment. The carrier signal passes through a similar bank of bandpass filters, whose outputs pass through amplifiers. The control signals from the envelope followers are used to open and close the amplifiers.
A vocoder consists of two banks of bandpass filters, a bank of envelope followers, and a bank of amplifiers. A block diagram of how it operates is shown in Figure 9-6. The two signals are split into narrow frequency bands by banks of bandpass filters. In ordinary operation, these filters will be set to matching frequencies, but this is not a requirement, and in fact some synth vocoders allow the bandpass frequencies to be offset or reassigned in various ways.
The outputs of the bands coming from the speech signal are tracked by a bank of envelope followers, which register the amplitude of the overtones in a given band at a given moment. The control signals from the envelope followers are then used to raise and lower the levels of the signals coming from the carrier’s bandpass bank. The carrier signals are then mixed and sent to the output.
In order for a vocoder to work effectively, the carrier must have a rich overtone structure. Sawtooth waves and wind (white noise) are good carriers. If the carrier lacks overtones in a given band, the vocoder won’t create them; all it does is raise and lower the levels of the carrier’s existing overtones. In many synths, the vocoder effect uses the synth’s own voices to produce a carrier signal. Generally, the speech signal will come from an external audio source.
Human speech contains some prominent high-frequency components, which are created by consonants such as f and s. If these are missing from the vocoder’s output, the speech will be more difficult to understand. Since the carrier may not have enough energy in the high-frequency range to reproduce these consonants effectively, some vocoders have a pass-through for the highs: They mix the top end of the speech signal with the carrier before sending it to the output.
Vocoders shouldn’t be confused with speech synthesis. Speech synthesis (robotic-sounding computer-generated speech) is an emerging field that may have some interesting musical applications in the future.