VX-303

The complete acid machine: the TB-303 voice (saw/square into the famous resonant 4-pole ladder, with accent and slide) with the sequencer built in — a 16-step pattern grid, a keyboard along the bottom, and an 8-pattern bank. Clock it three ways — from a cable, from its own tempo, or from the song timeline. Program a line, start it, and ride Cutoff and Resonance.

Programming a pattern

Each step in the grid is a vertical stack: the NOTE / step button on top (shows the step's note), then dedicated GATE, ACCENT and SLIDE lamps, and an OCT knob at the bottom. Press a lamp to flip it, or drag across a row to paint the same value down the row.

GATE — on plays the step, off is a rest.
ACCENT — louder and brighter: the accent works the filter envelope

harder, which is where the squelch lives.

SLIDE — ties this step into the next: the pitch glides there and the

envelope doesn't retrigger.

OCT — the step's octave (±2, the classic acid jump). Drag the knob to

set it, or click to cycle.

Set a step's pitch by clicking its NOTE button to select it, then:

Keyboard — the keys along the bottom set the selected step's note and

its octave in one press. Pressing a key programs, it does not play: it writes the step and makes no sound, so you can edit a line while it runs without hearing stray notes over it. The lamp row above the keys shows the selected step's note — it stays put rather than chasing the playhead, because it is the readout of what you are editing; the step being played is shown on the step strip. A rest lights no lamp. Use ◀ ▶ to move the keyboard's range — there is no HOLD, because nothing sounds to hold. C3 is the sequencer's octave 0; steps only reach ±2 octaves, so keys past that write the nearest one they can, and a step below the current range lights no lamp until you shift down to it.

LENGTH — steps per loop (1–16) for this pattern.
RANDOM — generates a fresh acid line in the current pattern (root-heavy

scale walk, octave jumps, beat-biased accents, off-beat slides). One undo step brings the old pattern back.

Patterns switch immediately — the moment you pick one, the next step comes from it. Swapping patterns mid-line is a playable gesture, and it's what lets Song mode stay exact: what plays depends only on what's selected now, never on which bar boundaries have gone by. Switching to a shorter pattern keeps running from where it was, wrapped to the new length; in Song mode the position comes from the song's position instead, so it lands wherever that bar falls in the shorter pattern.

Clock

Clock: Ext / Int / Song — who advances the sequencer.
Ext — one pulse at CLK = one step. The classic wiring: patch a clock

or divider in and the 303 runs at your patch's tempo. RST snaps it back to step 1.

Int — it free-runs at its own BPM while ▶ Run is on, no cable needed.
Song — the step is derived from the song timeline (one step per 16th), so

the line is locked to the arrangement by construction. Drop the playhead anywhere, in the Automation dock or on a Song Clock, and it plays the step that belongs at that point in the song — the same line every time, however you got there. Scrubbing repositions silently; stopping releases the gate, though the voice's own envelope rings out after it, so a long Decay keeps sounding for a moment.

⏮ ⏹ ▶ — the transport. ▶/⏹ run the internal clock, so they're live only

in Int; ⏮ works in every mode (in Song it rewinds the whole song).

BPM — the internal tempo, editable in Int. In Song it dims and shows

the timeline's tempo; in Ext it dims, since the cable is the clock.

For a pattern chain that survives seeking, drive the PAT input from an automation lane: pattern-per-time then depends only on the song position, so jumping to any bar gives you the pattern that belongs there.

Right-click the step grid for pattern operations — Randomize (same as the Random button), and Shift Left / Shift Right, which rotate the whole pattern (notes, gates, accents, slides, octaves) one step within its length. Each is one undo step.

Memories

The MEMORY block holds 16 slots, and each one remembers the entire configuration — every voice knob (Cutoff, Reso, Decay, Env Mod, Accent, Tune, Volume, Wave), the full 8-pattern bank, and the selected pattern. Click a slot to recall it; the whole machine changes at once.

MEM input — a CV (0–10 V → memory 1–16) that selects the memory while

patched, sample-accurately. When MEM is patched the slot buttons lock and the lit one follows the CV, so you can sequence memory changes from another module.

Right-click a memory for Copy / Paste / Clear. Copy grabs the whole

memory; Paste overwrites another slot with it (it even pastes across separate VX-303 instances); Clear resets a slot to factory-fresh. Each is one undo step.

Editing always targets the memory selected by the buttons — so under MEM CV you can keep editing one memory while a different one plays.

CV mod

Four of the inputs are modulation inputs — CUTOFF, RESO, DECAY and ACCENT — and the CV MOD knobs are their attenuverters, one per jack, in the same order. Patch a CV in, open its knob, and that control starts moving.

This is the answer to "I just want the filter to open over eight bars." Without it you would be building a memory per cutoff value; with it you patch one automation lane into CUTOFF and the knob you already set becomes the starting point.

It's an offset, not a takeover. The panel knob keeps working underneath

and stays editable — the CV moves the control from wherever you left it.

A cable alone does nothing. The attenuverters start closed, so patching

changes nothing until you open one. That's deliberate: you set the depth before you hear it, not after.

They go negative. Turn one below centre and the CV works in reverse — a

rising envelope closes the filter instead of opening it.

Cutoff moves in octaves (fully open sweeps about four of them), because

that is how this voice already modulates its filter — a linear sweep would crawl at the bottom and leap at the top. The other three are plain 0–1 controls, so a fully open knob at 10 V takes them to maximum.

The depths are not part of a memory. They stay put when the memory

changes, because they are wiring rather than sound design — which is the whole point of not having to snapshot them.

Controls

TUNE — transposes the whole pattern, ±12 st.
CUTOFF / RESO — the filter. This is the instrument; sweep it.
ENV MOD — how far the filter envelope opens the cutoff per note.
DECAY — envelope decay, 200 ms – 2 s. Short snaps, long sweeps.
ACCENT — how hard accented steps hit (level and filter sweep).
WAVE — saw or square oscillator.
VOLUME — output level.
CV MOD — four attenuverters, one per modulation input; centre is off,

either side sets how far and which way that CV moves its control.

Inputs & outputs

CLK / RST — the external clock (one step per pulse, in Ext) and

reset-to-step-1. Patch a clock or clock divider into CLK to run the 303 under your patch's master tempo; RST snaps it back to the top of the pattern (in Song mode it rewinds the song instead, since the step is derived).

CUTOFF / RESO / DECAY / ACCENT — four modulation inputs, one per control,

each with its own attenuverter in the CV MOD row. Unpatched they read 0 V and change nothing.

MEM — memory-select CV (0–10 V → memory 1–16); overrides the memory

buttons while patched.

PAT — pattern-select CV (0–10 V → pattern 1–8). While patched, the CV

decides which pattern plays, immediately; the Pattern menu keeps choosing which one you edit. Drive it from an automation lane for pattern chains that stay correct when you seek around the song.

OUT — the acid line, ±5 V.
V/OCT · GATE · ACC — the sequencer itself, as CV: the stepped pitch,

the gate (ties held through slides), and an accent gate — patch them to a different voice and the 303 sequences it instead (or as well).

Tips

Start with everything at noon, RESO at 3 o'clock, and sweep CUTOFF slowly

while a pattern runs — that's the sound.

SLIDE into an octave-up root with ACCENT on is the acid signature move.
Feed OUT through a distortion or overdrive module for the harder Detroit

flavor.