The Three Boxes Every Synth Is Made Of

Open a Minimoog, a twenty-dollar phone app and a wall of modular gear, and the same three things are happening in the same order. Once you can see them, every synth you meet becomes a variation on something you already know, and the front panel stops being a wall of knobs.

Three boxes and a set of hands

The audio inside a synth passes through three stages, in this order:

1. A sound source — the oscillator. It makes a continuous tone with a particular harmonic content.
2. A tone shaper — the filter. It removes some of those harmonics.
3. A level shaper — the amplifier. It decides how loud the result is, moment to moment.

That is the whole audio path: oscillator into filter into amplifier. Three boxes.

Then there is a second group of components, and this is the part that is easy to get wrong: the modulators. Envelope generators and LFOs belong here, and the essential thing about them is that no audio passes through them. They sit off to the side and produce changing values that the three audio stages respond to.

It is worth being blunt about this, because the chain is often written as oscillator into envelope into filter into amplifier, and that is not what happens. An envelope generator has no audio input and no audio output. It produces a number that rises and falls, and something else uses that number. Put the envelope in the audio path in your head and half of what a synth does stops making sense; take it out and put it alongside, and the architecture falls into place.

So the accurate picture is two rows. Along the bottom, audio flows left to right through three stages. Above it, a handful of control sources reach down and turn the knobs of those stages for you, faster and more precisely than a hand could.

Why control and audio are the same stuff

There is an idea underneath all of this that pays off repeatedly, and it comes from how analog synthesizers were built.

Inside an analog synth, everything is voltage. The audio is a voltage that swings back and forth quickly. The control signals are voltages that change slowly. There is no fundamental difference between them — no separate kind of wire, no separate kind of circuit. That is why the classic component names all begin the same way: VCO for Voltage Controlled Oscillator, VCF for Voltage Controlled Filter, VCA for Voltage Controlled Amplifier. Every stage is a thing whose behaviour is set by a voltage.

Two consequences follow, and both are the sort of thing that makes synthesis click:

• An LFO is just a slow oscillator. Nothing else. It is the same circuit as the one making your sound, running below the range you can hear, and used to control instead of to be heard.
• Speed it up and modulation becomes timbre. Raise a modulating oscillator into the audible range and it stops producing a wobble and starts producing new frequencies. That is not an analogy — it is literally the same routing, and it is what FM synthesis is.

Digital instruments keep the architecture and drop the voltages. An oscillator whose pitch is set by a number rather than a voltage is a DCO, a Digitally Controlled Oscillator, and on a sampler the same slot in the chain is filled by a recording instead of a generated waveform. The names change; the three boxes do not.

What happens when you press a key

Worth walking once in time order, because it puts every component in its place and takes about fifteen seconds to read.

1. You press a key. Two things leave the keyboard: a value saying which note, and a gate saying a key is down. On some instruments a third, velocity, says how hard.
2. The note value sets the oscillator's pitch. It has been running all along; now it is running at the right frequency.
3. The gate starts the envelope generators. They begin producing rising values.
4. One envelope's value is added to the filter's cutoff, so the filter opens and then settles.
5. The other envelope's value sets the amplifier's gain. Until this happens the amplifier was closed and you heard nothing at all — this is the moment the note becomes audible.
6. Meanwhile the LFO has been cycling continuously, whether or not anything was playing, nudging whatever it is routed to.
7. You release the key. The gate closes. Both envelopes enter release, the amplifier closes, and the note ends — while the oscillator carries on producing a waveform nobody can hear.

Two things in that sequence surprise people the first time. The oscillator never stops — it is producing sound continuously and the amplifier is what decides whether any of it reaches you. And the amplifier is closed by default, which is why a patch with no amplitude envelope makes no sound at all rather than a continuous drone.

A voice, and why notes disappear

One complete chain — oscillators, filter, amplifier, and its own envelopes — is a voice, and one voice can produce one note at a time.

How many voices an instrument has is its polyphony, and it is a real physical limit on analog instruments because each voice is a separate set of circuits. This is the whole explanation for a class of behaviour that otherwise looks like a fault:

• Monophonic — one voice. Play a chord and you get one note, usually the highest or the most recently pressed depending on the instrument's note priority. Most classic bass and lead synths are monophonic, and being monophonic is part of why they play the way they do.
• Paraphonic — several oscillators that can be at different pitches, but sharing one filter and one amplifier. You can play a chord; you cannot articulate the notes independently, because a single envelope is governing all of them. Notes added to a held chord will jump to the middle of the existing envelope rather than starting their own.
• Polyphonic — several complete voices. Eight-voice and sixteen-voice instruments are common, and software is usually limited only by processing.
• Voice stealing — what happens when you ask for more notes than there are voices. The instrument reallocates one, normally the oldest or the quietest, and that note stops. A sustained pad losing its bottom note as you play a melody over it is not a bug.
• Unison — stacking every available voice onto a single note, usually detuned against each other. It trades polyphony for thickness, which is why the biggest-sounding leads are often monophonic by necessity.

A hundred years of the same three boxes

A short history, because it explains why the architecture is so consistent across instruments that otherwise share nothing.

Electronic instruments predate the synthesizer by decades — the theremin, from around 1920, is played without being touched, the performer's hands altering the fields around two antennas to control pitch and volume. It is an oscillator and an amplifier with a very unusual interface, and nothing else.

What arrived in the 1960s was not the idea of electronic sound but the idea of voltage control: a standard way for one module to govern another, so that oscillators, filters and amplifiers could be built as separate units and connected however a musician wanted. That is the modular synthesizer, and the reason its architecture became universal is that voltage control made the three-box chain the obvious thing to build.

The instrument that took it out of the studio was the portable, hard-wired, keyboard-equipped synth of the early 1970s. It fixed the routing — you could no longer patch anything to anything — and in exchange it could be carried to a gig and played immediately. Almost every synth since has been some point on that trade between flexibility and immediacy.

Then the constraints fell away one at a time. Digital control brought reliable tuning and, critically, storable patches: before that, saving a sound meant writing down or photographing the position of every knob. Digital sound generation brought methods that analog circuits could not produce. Software removed the voice limits and most of the cost.

What did not change is the chain. A modern software instrument with a hundred modulation destinations is still a source, a tone shaper and a level shaper with things moving them, because that division of labour turned out to be the right one.

Reading a panel you have never seen

Because the panel is effectively a diagram of the signal flow, the layout itself tells you most of what you need.

Look left and you will find the sources — oscillator sections, usually numbered, with waveform selectors, tuning controls, and a mixer for balancing them. Look in the middle and you will find the filter, with cutoff and resonance as its two large controls. Look right and you will find the amplifier and output. The modulators — envelopes and LFOs — are grouped somewhere nearby, often below or to the side, and their controls are the ones whose labels are times and rates rather than frequencies and levels.

The differences between instruments are then a short list of specific questions, and asking them deliberately is faster than exploring:

1. How many oscillators are there, and what waveforms do they offer?
2. What kind of filter is it, and how steep? This one question accounts for more of an instrument's reputation than any other.
3. How many envelopes are there? Two is the classic arrangement — one for the amplifier, one for the filter. More than two means envelopes are available for other things.
4. How is modulation routed? Fixed routings printed on the panel, or a free matrix where anything can control anything?
5. How many voices, and is there a unison mode?

A patch — or preset, or program — is simply the stored answer to every one of those controls at once, plus the modulation connections. It is a list of settings, not a recording, which is why a patch reproduces exactly on the instrument it was made for and does not transfer to a different one.

Where the chain bends

The three boxes are the rule and there are recognised departures, worth knowing so they do not look like exceptions to something you have just learned.

• Filter after the amplifier. A few designs put the amplifier first. It changes how the filter is driven, since the signal reaching it is now shaped by the amplitude envelope, and instruments that offer the choice sound noticeably different in each mode.
• Effects, which are not part of the voice. Reverb, delay, chorus and distortion almost always sit after all the voices have been summed. That means they are global — one reverb for the whole instrument, not one per note — which is why a synth reverb behaves differently from a per-note effect and why heavily effected patches can sound smeared on dense chords.
• More than one filter, in series or in parallel, covered in the filter lesson.
• Oscillators that are not oscillators. On a sampler the first box holds a recording; on a physically modelled instrument it holds a simulation; on a wavetable synth it holds a table being scanned. The remaining two boxes are unchanged, which is precisely why learning subtractive synthesis first transfers to everything else.

Start from init

One practical habit, and it is the fastest route into an unfamiliar instrument.

Every synth has an initialised patch — a blank state, usually one oscillator on a sawtooth, filter wide open, plain amplitude envelope, no modulation. It sounds dull and slightly unpleasant, which is exactly right: it is the sound of the architecture with nothing done to it.

Learning a new instrument by loading its presets teaches you what its designer likes. Learning it from init teaches you what it is. Start there, turn one control, listen, turn it back. An hour spent that way is worth a great deal more than an afternoon of scrolling through patches, and it is the only reliable way to find out what a particular instrument's filter actually does — which, as the previous section argued, is most of what you are buying.

Where the character actually lives

Worth knowing before you spend money on anything: take two synths with identical oscillators, identical envelopes and identical amplifiers, and give them different filters, and they will not sound remotely alike.

In subtractive synthesis the filter does most of the shaping, and the details of its design — how steeply it cuts, how it behaves as resonance is raised, what happens to it when a hot signal is driven into it — are what people are describing when they say an instrument has a particular sound. Everything else in the chain is comparatively interchangeable.

Which is also why the filter lesson in this course is the longest one, and why the practical advice for learning any new synth is to leave the oscillator on a sawtooth, set the amplifier envelope to something plain, and spend the first hour doing nothing but moving the filter.

What comes next

Each box gets its own lesson from here, in signal order:

• Oscillators and waveforms — what each classic waveform is actually made of, and why sawtooth is the default starting point for almost everything.
• Filters and resonance — the stage that decides what an instrument sounds like.
• Envelopes — the modulator that makes a note behave like a played note rather than a switch.
• LFOs and modulation — everything else that moves, including the routing systems that let anything control anything.
• Types of synthesis — the methods that replace the first box, or rearrange all three.

One habit is worth forming before any of that. When a patch will not do what you want, change one thing at a time and listen. Synths are dense with interacting controls, and most patches that refuse to behave turn out to have one modulation routed somewhere you have forgotten about. The three boxes give you somewhere systematic to look: is the problem the source, the tone shaping, the level shaping — or something moving one of them?

What to practice

• Open any synth and find the three audio-path stages on the panel — source, filter, amplifier. They are almost always laid out left to right in that order, because the panel is a diagram of the signal flow.
• Turn the filter cutoff fully open and the amp envelope to a plain organ shape. That is the raw oscillator, unshaped, and it is the sound everything else is built out of.
• Take a preset you like and work out which single stage is responsible for its character. Usually one stage is doing most of the work and the rest is trim.
• Count the voices on a synth you use by holding down more and more notes until one drops out. That number is its polyphony, and knowing it explains a whole class of dropped-note mystery.
• Find one modulation routing in a preset and disconnect it. What changes tells you what that routing was for, which is faster than reading about it.