Subtractive Synthesis Explained

Start with too much, then remove

Subtractive synthesis is named for how it works: you begin with a sound that has more harmonic content than you want and take pieces away. A sawtooth wave contains every harmonic of its fundamental, which makes it buzzy and bright and — crucially — a good raw material. Filtering it down is how you arrive at everything from a soft pad to a punchy bass.

This is the opposite of additive synthesis, where you build a tone by stacking individual sine waves. Additive is more precise in principle and vastly more laborious in practice: reproducing a sawtooth additively means summing dozens of sines. Subtractive gets you most of the way in three controls, which is why it dominated hardware synthesis and still does.

The whole architecture is four stages in a line: oscillators → filter → amplifier → output, with envelopes and an LFO modulating the middle two. Every control on the synthesizer belongs to one of those stages, and once you can name the stage a control affects, the instrument stops being intimidating.

Oscillators: the raw material

The oscillator produces the initial waveform, and the shape you choose determines what harmonics are available to work with later. Four shapes cover almost everything.

The practical consequence is that you cannot filter in what is not there. Starting from a sine gives you nothing for the filter to remove, so a sine patch will always sound plain no matter what you do downstream. If a sound is too dull, the fix is usually a brighter oscillator rather than a higher cutoff.

A second oscillator is where thickness comes from. Detune it a few cents and the two drift slowly in and out of phase, producing a gentle beating the ear reads as warmth and movement — the reason a two-oscillator patch sounds so much richer than one. Larger detunings give chorus-like width, and 1200 cents is exactly an octave, which is how you get weight underneath a lead.

The filter is the instrument

If subtractive synthesis has one signature control, it is the low-pass filter cutoff. It sets a frequency above which harmonics are progressively removed, and sweeping it while a note sounds is the most recognisable gesture in electronic music.

Resonance boosts the frequencies right at the cutoff point, creating a peak that emphasises whatever sits there. A little adds character and focus. A lot produces the squelchy, almost vocal quality that defines acid basslines. Pushed to the extreme, a resonant filter will self-oscillate, generating a sine wave with no oscillator input at all — a happy accident of analogue design that became a feature.

Filters are usually described by slope, in decibels per octave. A 12 dB/octave filter rolls off gently and keeps some brightness above the cutoff; a 24 dB/octave filter cuts hard and sounds much darker at the same setting. A great deal of the difference between famous synths comes down to their filter designs, which is why people talk about a particular instrument's filter as if it were a personality.

A high-pass filter does the reverse, removing lows, and is most often used to thin a sound so it sits above a bass without competing. It is less dramatic than a low-pass and far more useful in a mix than beginners expect.

Envelopes: how a note behaves over time

A sound with a constant volume is a drone. What makes a note feel like a plucked string, a struck bell or a bowed cello is how its level changes over its life, and that shape is described by four numbers.

Attack is how long it takes to reach full level. Near zero gives a percussive start; a second or more gives a swell that fades in. Decay is how long it takes to fall from that peak to the sustain level. Sustain is not a time but a level — the volume held for as long as the key is down, which is the part beginners most often misread. Release is how long the sound takes to fade once you let go.

The step that transforms a patch is giving the filter its own envelope. Real instruments are brighter at the moment they are struck and darken as they ring out — a plucked string loses its high harmonics fastest. A filter envelope that opens quickly and closes as the note decays reproduces that, and it is the single biggest difference between a synth patch that feels alive and one that feels like a test tone.

The LFO

A low-frequency oscillator runs below hearing — typically 0.1 to 20 Hz — and rather than being heard, it modulates something else. Where you point it decides what it does.

Pointed at pitch, it gives vibrato. Around 5 Hz with a few cents of depth is what a singer or violinist does naturally, which is why it makes a synth line feel more human. Pointed at amplitude, it gives tremolo — the pulsing of a Rhodes or a surf guitar. Pointed at filter cutoff, it gives the rhythmic wobble that underpins a great deal of electronic music.

The most common mistake is too much depth. Vibrato at forty cents sounds seasick rather than expressive; four to eight cents is usually plenty. As with most modulation, the setting that sounds right in isolation is generally too strong in a mix.

Building a patch from nothing

A reliable order of operations, which works whether you are on a hardware synth or a browser one.

Start with the oscillator. Choose a waveform with enough harmonic content for the sound you want — sawtooth for anything bright, triangle or sine for something soft. Add the second oscillator and detune it slightly for thickness.

Open the filter fully, then close it. Sweep the cutoff down until the sound is too dark, then back off. This finds the tonal centre far faster than guessing at a number.

Shape the amplitude envelope. Decide first whether the sound is percussive or sustained, since that single decision sets attack and sustain together.

Then add the filter envelope. This is the step that brings it to life. Even a small amount of opening at the start of each note makes an enormous difference.

Add movement last. A touch of vibrato, and a second look at detune. Both are seasoning, and both are easy to overdo.

Polyphony, voices and why old synths were monophonic

Every note you hold needs its own complete chain — oscillators, filter, amplifier, envelopes. On an analogue instrument that meant physically duplicating the circuitry, so a synth capable of eight simultaneous notes contained eight of everything. That is why early instruments were monophonic: one voice was all the hardware anyone could afford.

It is also why monophonic synths developed features polyphonic ones do not need. Glide, or portamento, slides the pitch from one note to the next rather than jumping, which only makes sense when there is a single voice moving between pitches. Note priority — whether the highest, lowest or most recent key wins when you press two — is a real design decision on a mono synth and meaningless on a polyphonic one.

Digital removed the cost, and a browser synth like this one is limited only by processor time. Each held note allocates a fresh set of nodes and frees them once its release tail has finished, which is why the release setting matters for more than musical reasons: a very long release with many notes means many voices still running.

Retriggering is the detail that catches implementations out. Pressing a key that is already sounding must release the old voice before starting the new one, or you end up with two oscillators on the same pitch stacking up — twice the volume and a phasing artefact that sounds like a fault.

Where these ideas came from

The architecture is not one person's invention, but Robert Moog's instruments of the 1960s standardised it — voltage-controlled oscillators, filters and amplifiers, connected in exactly the order described above. The decision that made it a musical instrument rather than a laboratory device was using a keyboard to generate the control voltage, so pitch became something you played rather than something you dialled in.

The Minimoog of 1970 packaged the whole thing into something portable with a fixed signal path and no patch cables, and that fixed path is essentially the layout every subtractive synth has used since. When a plugin today shows you two oscillators, a filter, and two envelopes, it is following a template that is over fifty years old.

Digital changed the implementation and almost nothing about the concepts. The oscillators are now computed rather than analogue circuits, the filters are digital approximations, and polyphony is limited by processor rather than by hardware voices. But attack, decay, sustain, release, cutoff and resonance mean exactly what they meant in 1970, which is why learning them once pays off across every instrument you will ever touch.

Mistakes that make a patch sound amateur

Everything at maximum. Full resonance, full detune and full filter envelope on the same patch fight each other, and the result is harsh rather than powerful. Most good sounds have one dominant characteristic and everything else in support.

No filter movement. A static cutoff is the fastest way to sound synthetic in the bad sense. Even a small filter envelope gives a note a beginning, which is what the ear listens for.

Zero attack on everything. An instant attack is right for a pluck and wrong for almost everything else. A few milliseconds softens the click at the start of a note and is often the difference between a sound that stabs and one that speaks.

Ignoring release. An abrupt cut-off when you lift a key sounds unnatural, since nothing in the physical world stops instantly. Even 100 ms of release makes playing feel considerably more musical.

Designing in isolation. A patch that sounds enormous alone frequently disappears in a mix, because the frequencies that made it impressive are the ones everything else occupies too. If a sound is destined for an arrangement, build it against the arrangement.

Learning by taking things apart

The fastest way to improve at sound design is to reverse-engineer presets. Load one you like and change a single control at a time, listening to what each does. Set the filter envelope amount to zero and hear how much life it was providing. Take the detune to zero and hear the thickness disappear. Set sustain to full on a pluck and hear it become an organ.

The second habit is to build from scratch even when a preset would be quicker. Choosing every value yourself, badly, teaches more in ten minutes than an hour of browsing presets, because you find out which controls matter by having got them wrong.

Open the synthesizer and try exactly that — load the Pluck preset, set its amplitude sustain to 1, and listen to what happens. Then set the filter envelope amount to zero and hear the difference again. If you want to play what you build over a steady pulse, the metronome is there, the beat maker gives you drums to play against, and the chord finder will tell you which notes go together.