Every synth tutorial starts with the same block diagram — oscillator to filter to amplifier — and every beginner makes the same discovery: the diagram explains nothing until you can hear a knob turn. So here's the tour in the order that finally makes it click, with an FFT spectrum open the whole time.

The oscillator: your raw material

Four geometric waveforms plus noise is the whole palette. A sawtooth contains every harmonic, falling off gently — bright and brassy, the default for leads. A square wave has only odd harmonics — hollow and vocal, the chip-tune sound. A triangle is a softer version of the same odd-harmonic structure. And a sine is one pure partial. Critically, a square with adjustable pulse width (PWM) is a different instrument at every setting: narrow it and the timbre thins and nasalizes; sweep it and you get the shimmering hollow motion that defined 8-bit bass lines. If your synth's square is stuck at 50% duty, you're missing half the character.

ADSR: the shape of time

Attack, decay, sustain, release is just "how does loudness evolve" — but the four numbers cover an astonishing range. Long attack turns anything into a pad; zero attack with short decay makes a pluck; high sustain sustains (organs, strings); low sustain decays (guitars, mallets). The mental model that sticks: attack and release are transitions, decay and sustain describe what happens while you hold. Set a 900ms release on a short note and you've invented legato.

The filter: subtraction is the name

This is literally why the method is called subtractive synthesis — the oscillator is deliberately too bright, and the lowpass filter carves the final tone. Two controls matter. Cutoff sets the carving line: sweep it down on a sawtooth and hear (and see, on the FFT) the harmonics fall away in ranks. Resonance boosts a narrow band at the cutoff: a little adds focus and vocal character, a lot self-oscillates into the acid squelch. The reason filter sweeps feel musical is physical: formants in speech are exactly resonant peaks moving around.

The LFO: motion for free

An LFO is just a slow oscillator you can't hear, routed to something you can. To pitch it's vibrato; to cutoff it's wobble; to amplitude it's tremolo. The rates tell the story: sub-1 Hz reads as "evolving pad," 4-7 Hz reads as "electric piano tremolo," and once you pass ~20 Hz the pitch modulation stops being heard as motion and becomes sideband timbre — FM synthesis is born from exactly that threshold crossing.

Effects: space and dirt

Overdrive, delay and reverb are the performance end of the chain. Overdrive (a tanh soft-clip) trades peak level for harmonic richness — it's why guitars sound bigger distorted. Stereo delay with a little crossfeed between channels instantly widens a mono voice. Reverb is thousands of packed echoes; algorithmic ones fake it with feedback combs, and the decay knob is just "how long the room remembers." The mix rule that separates amateur from tasteful: effects are salt, not soup — under 40% wet unless the effect is the point.

Design a sound, ship a sound

The quiet superpower of browser-based synthesis is the export. Design your laser or powerup by ear, then download the WAV — or copy the generated Web Audio code and ship the synthesis itself, so the sound is 40 lines instead of a 400KB asset. Game audio has been moving this way for years: procedural audio means infinite variations, zero download cost, and parameters (pitch, intensity) that respond to gameplay. The slider you just understood is the API.

Turn one knob at a time, watch the FFT, and trust your ears over the diagram. Ten minutes of that beats a chapter of theory — and unlike a rented DAW, this laboratory is one browser tab.

Subtractive Synthesis in Ten Minutes: What Every Slider Actually Does