What Is an Envelope? How a Sound Starts, Holds and Stops

Every sound has a loudness profile over time, and that profile is conventionally divided into four phases. The attack is the rise from silence to maximum, which is very fast for a plucked or struck note and slow for a bowed or blown one. The decay is the fall from that peak to a sustained level. The sustain is the level held while the note continues, which is zero for instruments that cannot hold a note. The release is the fall to silence after the note is stopped. Those four describe most instruments adequately and are the standard model in synthesis, where they are directly adjustable. The shape differs enormously between instruments and is a substantial part of what distinguishes them, independently of the frequencies present.

Experiments removing parts of a recording demonstrate how much the beginning carries:

• •Removing the attack from a piano note makes it very difficult to identify as a piano
• •The same treatment makes several instruments sound like an organ, since what remains is a steady tone
• •Playing a recording backwards, which reverses the envelope, renders familiar instruments unrecognisable
• •The attack contains transients, brief noisy components that do not persist and that carry the mechanical signature of how the sound was produced
• •Listeners identify instruments from very short excerpts including the attack, and struggle with longer excerpts that exclude it
• •Synthesis without accurate attack modelling produces sounds that are recognisably artificial

The profile follows from the physics of how the sound is made. A struck or plucked string receives all its energy at once and then loses it, which gives an immediate attack and a long decay with no sustain, so a piano note begins to die the moment it starts. A bowed string receives energy continuously, which permits a sustained level and a slower attack determined by how the bow engages. A wind instrument depends on an air column being established, which takes a few cycles and produces a brief unstable period audible as a chiff. Percussion varies from an almost instantaneous attack and rapid decay to a long ring. Electronic sources have no inherent shape at all, which is why synthesis requires it to be specified and why early synthesisers sounded artificial when it was applied crudely.

The loudness profile is one of several things that change over a note, and the others matter for the same reasons. The balance of frequencies shifts, generally with higher components decaying faster than lower ones, so a piano note becomes purer as it dies and a bell changes character continuously. Pitch itself moves slightly on many instruments, with a struck string sharp immediately after the strike as the large initial amplitude stiffens it, and with wind instruments settling to pitch over the first cycles. Noise components present at the start disappear. Vibrato may be introduced partway through rather than from the beginning, which is a stylistic choice differing sharply between periods and traditions. A synthesiser that holds all of these constant while shaping only the loudness produces something that is recognisably not an instrument, which is why realistic synthesis proved so much harder than expected.

Control over the profile is a substantial part of both performance and production. Players alter it directly, with a string player choosing bow speed and pressure at the start of a note, a wind player choosing how to begin it with the tongue, and a pianist controlling only velocity and duration, which is why piano touch is a narrower parameter than it is sometimes described as. In recording, compression alters the profile by reducing the difference between loud and quiet parts, which changes attack and sustain and is why heavily compressed recordings sound different rather than merely louder. Gating removes quiet portions, shortening the release. Reverse envelopes, applied by playing a sound backwards or by synthesis, produce the swelling effect familiar from film scoring. And sampling instruments must capture the attack accurately or the result is unconvincing regardless of everything else.

Attack, decay, sustain and release describe how a note rises, settles, holds and stops, and the shape follows from how the energy is supplied. Removing the attack makes a piano note hard to identify and makes several instruments sound like an organ. The attack carries brief noisy components recording the mechanics of production, which is why synthesis without them sounds artificial.