The View from 2005: This article is just as relevant today as it was in 1994. We use peak meters a lot more today, to head off dreaded “overs” in digital audio. The principles of compression are almost exactly the same, and the ambiguity between subjective loudness and objective amplitude remains as perplexing as ever.

Look out! Timbre!

Faithful readers will recall that I’ve been harping about timbre as a function of equalization in several different articles over the past year. One of the things I mentioned in those articles was the use of a compressor along with parametric equalizers. As I was writing, it occurred to me that it would be really useful to discuss compressors by themselves, in terms of using them to modify and control timbre and tone quality, instead of the more traditional view of using them to control loudness, which is how they are normally thought of, if not used.

Ah! Compressors!

Now, compressors are really interesting toys. They are not simple, in either their operation or their musical impact, so it’s worth considering them in a little detail.

To begin with, compressors are gain controls. The active element in the device is a voltage-controlled amplifier that changes the level of the signal passing through it.

Changing level is pretty mundane stuff, you would think. Bring up the bass 2 dB, fade out the horns. Pretty straightforward. Not so with compressors, because what they do is change level as a function of the audio signal being passed through the compressor. What this means to you is that compressors can be used to shape the envelopes of the individual notes being played. Further, and this is not obvious at all, they can be used to change the overall spectrum (huh?) of the processed signal, its perceived loudness and a great deal of its fundamental musical character.

This all happens as a result of the sensing circuit that listens to the audio signal being sent to the compressor and which derives a control voltage from the amplitude of the audio signal that is used to operate the voltage-controlled amplifier (VCA).

	Figure 1. Generic compressor.

Let’s discuss how this sucker works for a second. The audio input is split and sent to both the VCA and to the sensing circuit. At the VCA, its level is changed as a function of the control voltage being applied to the VCA, which in turn is being derived from the sensing circuit (AKA Envelope Follower (EF), Transient Detector (TD), RMS Rectifier (RMSR), Amplitude-to-Voltage Converter (AVC), Hyper Yield Processing Eliminator (HYPE) and other such acronyms – each manufacturer has his/her own name for it, and you too can make up your own special term of endearment if you wish). Assuming that the VCA is a good one (and most current compressors use good ones, as they’re cheap and easy to get), the real character and quality of the compressor comes from the sensing circuit and the way it and its controls are configured. One of the most important aspects of this is how it senses changes in level over time. We’ll get to that.

Another compressor control that turns out to be really important is the little knob called ratio, which describes the degree of compression going on. The actual ratio referred to is the ratio (in dB) between input and output levels when the compressor is invoked (i.e. when the threshold level has been exceeded, or deceeded in the case of gating or downward expansion). A 1:1 ratio means that no compression is taking place: 10 dB above threshold at the input will result in 10 dB above threshold at the output. A 2:1 ratio means that 10 dB above threshold at the input will yield an output only 5 dB above threshold. A 10:1 ratio means that the output in this case will be only 1 dB above threshold, while an infinite (∞ ratio will result in no output ever exceeding the threshold, regardless of the input. Big ratios are usually thought of as limiting, while small ones are thought of as compression.

Now’s the time

Musical envelopes are time envelopes. They shape or describe the shape of dynamic processes that change in a recognizable pattern over time. The compressor’s sensing circuit modifies the loudness envelope of the audio signal, and it usually has a couple of controls to allow you to have some say about this. They are generally called “attack time” and the “release time.”

The audio signal’s amplitude is changing over time, as a matter of course. In a general way, we can say that each note or musical event has its own separate envelope of loudness, which includes starting from silence and getting loud, sustaining or slowly changing during the course of the note, and then dying away to silence. It is from this generalized notion that we get the analog synthesizer’s ADSR (Attack-Decay-Sustain-Release) Envelope Generator and its descendants, which are used to mimic this general music behavior.

In the case of the compressor, however, things are a little different. The compressor’s envelope follower is responding to changes in the audio envelope rather than creating an envelope on its own, and the time-based controls are determining how quickly the VCA will respond to the actual changes in the audio envelope level. If the compressor is set to be ideally, perfectly instantaneous in its response (there are some electronic design problems here that I am not going to get into), then as the audio signal changes, so does the VCA. In the case of compression, which counteracts loudness changes, a “perfectly” responding sensing circuit would cause the VCA to hold the audio signal, regardless of its original level, at a constant output level.

In such a case, for instance, the peak of the kick drum attack would be at the same level as the noise floor, and the transition from noise floor to peak amplitude (a change in level of approximately 90 dB!) would take place in about 50 microseconds. This would be a strange sound indeed: first you are hearing air conditioner rumble and mic preamp hiss and the distant leakage of traffic and then comes a sudden not-very-loud tick-whooomffffwhoomfffwhmffwffwffwff morphing back to air conditioner rumble and mic preamp hiss and the distant leakage of traffic over about 200 milliseconds, all at say 80 dB SPL, without the meter or loudness changing at all! Very interesting, but not funny, as Artie used to say. A sound that only an audio nerd’s mother could love!

This is why we want to have some control over how closely in time the action of the compressor counteracts the audio envelope level. The attack time allows us to control how much time of the audio attack we wish to allow to pass through the compressor before the compressor begins to turn down the level. The release time allows us to control how rapidly we want the compressor to turn the level back up after the audio envelope has died away. Naturally, there are no rights and wrongs to this. It all depends on what you are trying to accomplish (have you heard this before?).