SubSDJ

Vocal — a vocal tract per voice (a bank of 5 formants between the drive and the filter). Vowel sweeps u→o→a→e→i · Size shrinks/grows the tract (child ↔ giant) · Q: narrow = robot/vocoder, wide = soft · Life = the formants wander a little, like a real mouth · Bright = the «singer's formant» (F4-F5, ~3 kHz: what lets an opera voice carry over the orchestra) · Bite = a consonant at the start of the note: the mouth STARTS closed and opens into the vowel (a formant transition, which is what the ear uses to hear /k/) + the burst of the plosive — low = a soft «wa», high = a hard «ka»; scales with velocity; with a slow attack it gets lost (it is a transient). With the sequencer: «ka-ka-ka». The formants are FIXED in Hz (they don't follow the note): that is what sounds «like a voice». Breath = the Noise knob. Ideal source: an osc on «Wavetable» with the glottal set (Morph = vocal effort). In the matrix it is the «Vowel» destination: S&H→Vowel = one vowel per note; env→Vowel articulates the attack.

Resonator — a bank of six modal resonators per voice, tuned to the NOTE (f0 × ratios), between the Vocal and the filter. «Lite» modal synthesis: a vibrating body is its modes, and each resonator (a 2-pole biquad) is one mode with its frequency, its Q and its gain. Material goes from string or tube (harmonics 1, 2, 3…) to free bar or bowl (1 · 2.76 · 5.40 · 8.93 · 13.3 · 18.6), interpolating the logarithm of the ratios: in between, the inharmonicity of a piano or a bell. Decay = Q from 10 to 600 on a log scale: how long each mode takes to ring up and to die away (τ = Q/(π·f): at full and 130 Hz, a second and a half). Bright shifts the weight between low and high modes without changing the level. Excite it with Noise (Mixer module) for bowed bowls and wind over strings: noise enters the bank boosted ∝ √Q (the power a resonator lets through is ∝ f/Q; without that, at high Q the bowl wouldn't sound), and the tonal signal enters at 0 dB (a partial right on a mode passes as is: a saw doesn't blow up). With a saw: buzzing metal. Follows the note and the glide. At 0 it isn't there.

Per-voice filter. Cutoff with a logarithmic travel · Reso in dB. Type: LP↔HP (Low↔Hi blends low-pass and high-pass in parallel), BP (LP and HP in series: band-pass) or Notch (LP + HP in parallel at gain 1: an exact zero at the cutoff; the morph at 50 % already is that notch, at −6 dB). Keytrack: the cutoff follows the pitch of the note (1 = 1:1) · 24 dB = two stages in cascade, creamier (not for Notch). Type and slope are structure: they apply from the next note. Drive saturates BEFORE the filter, per voice, with two characters: soft (tanh: compresses and rounds off) or fold (West Coast synthesis: as Drive goes up the wave folds over itself and harmonics appear, rise and fall again — a sweep no filter can do; move it with an LFO or an envelope from the matrix, destination «Drive»). Asym shifts the working point of the curve: EVEN harmonics come out (the «bent» body of a tube or a kick drum); on fold it is the wavefolder's symmetry — at full, only even harmonics: you hear the octave. Silence in gives exact silence out, and the DC the bias creates is filtered out (5 Hz). Its LFOs are the Cutoff and Reso rows in the LFO bank; its envelope, «Filter Env».

Filter envelope. Amount in Hz over the cutoff (± = opens or closes). «Follow Amp Env» takes the times of the amplitude envelope; «own» uses the A/D/S/R faders here. The peak scales with velocity (MIDI). The «Amp Env» graph draws it dashed.

Amplitude envelope (and the filter's too if «Filter Env» follows it). The graph draws both envelopes. Faders on purpose: an envelope is time bars, it is easier to draw than to turn. Log travel: attack and decay up to 20 s, release up to 30 s (the scale of the cinematic pad), and the millisecond still has room; the bottom end of the attack is exactly 0 = a step. Curve: at 0 the ramps are linear (as always); as you raise it, attack and decay become RC-style exponentials, like an analog ADSR — the attack rises fast and slows as it arrives, the decay falls fast and settles slowly. Since loudness is logarithmic, an exponential decay sounds even while a linear one «hangs and then drops». The release was already exponential. Flam (hits · gap): with more than one hit, hits−1 short bursts sound before the ADSR, «gap» ms apart, and the full ADSR is the last hit — the pulse train of the 808 clap: a clap is not one burst of noise, it is three or four bursts ~10 ms apart and then the tail. The graph draws them.

Voice. Poly / Mono (one voice, last-note priority). Trigger (Mono only) decides what happens when a note arrives while another is still held: Legato slides the sounding voice without retriggering anything (trills, the 303 glide); Retrig always retriggers: envelopes, pitch env and FM4 operators start again, and the old note fades while the new one's attack rises (like an analog ADSR retriggering from its current level). Separate notes retrigger in both. With the arpeggiator or sequencer, «overlapping» is decided by Gate: below 1 every step is a new note; at 1 the steps touch and tie in Legato. Glide = portamento (up to 10 s: sliding drones) · Pitch env: the note starts Amount cents off and falls to pitch over Dec (kicks, toms, zaps; up to 10 s = risers) · ±Cents shifts the tuning of the WHOLE synth (MIDI pitch bend moves it too) · Amplitude = global gain before the effects (the one the tremolo modulates) · Master = output volume (in dB, up to +12), BEFORE the master limiter. The master leaves −9 dB of headroom so a four-note pad chord doesn't reach the limiter and stays clean; every factory preset comes with its Master measured to sound equally loud in its use (one note for a bass or a lead, a chord for a pad). Push it too far and the limiter clamps down: louder, but compressed (the volume pumps with the beating of the oscillators).

LFO bank: one row per LFO, like the strips of a mixer (same column = same control). Click to cycle the shape (sin/tri/saw/sqr/S&H/drift; S&H = random steps at the LFO rate, drift = smooth random). Fixed destinations: PWM → pulse width of oscillators on «Square» (the row dims if there are none) · Cutoff → cutoff in Hz (for exponential modulation, in octaves, use the matrix: «LFO cutoff» → «Cutoff») · Reso → dB over the Q of the last stage · Tremolo → gain (organs, vibraphones) · Vibrato → pitch · Pan → autopan per voice. Speeds are logarithmic from 0.005 Hz (one cycle every 200 s: at that scale the LFO no longer «oscillates», it drifts). In the matrix the LFOs come in RAW as sources (their Level doesn't apply). Fade: vibrato and PWM ramp in after each note, like a player would (0 = immediate; up to 20 s). Sync: «free» = Speed in Hz; a division (4 bars … 1/32; «.» = dotted, T = triplet) = the LFO runs IN TEMPO: its rate comes from the BPM and its cycle starts on the common clock grid (the one the arpeggiator, sequencer and conductor use), also when the BPM changes. Phase (applies per note): «global» = one LFO per engine, the chord breathes as a block; «per voice» = each voice starts its cutoff/reso/vibrato/PWM/pan LFOs at a different phase (a Weyl sequence k·φ: deterministic and well spread; the first voice sounds as always); «retrig» = each note starts them at phase 0 (the LFO tied to the note: every pluck moves the same); «once» = a single cycle from the note, then they stay at its end — an LFO acting as an envelope (S&H and drift have no end: they behave as retrig). The tremolo is always global. Drift = rate of the matrix source «Drift (voice)» (0.05 Hz = one sway every 20 s).

Distortion, global (tanh, dry/wet: at 0 it is transparent) · Flanger with fixed feedback (the classic jet).

Chorus: stereo, 2 lines · Ensemble = a triple 3-phase chorus in the style of the Solina (string machine). Speed drives the slow LFO; the fast one runs ×10.

Stereo. Width spreads the voices across the stereo field by pitch (mono at 0). The autopan (LFO → pan per voice) is the Pan row in the LFO bank.

Arpeggiator. Hold notes down and it sequences them at the BPM (knob in Delay). Swing (below) is GLOBAL, like the BPM: in each pair of eighths or sixteenths the second arrives late — 50 % straight, 67 % triplet, 75 % dotted — and it swings the arpeggiator, the sequencer, the Steps and the players all at once (quarter notes and triplets don't swing). Gate = length of each step: in Mono, below 1 every step retriggers; at 1 the steps touch and, with Trigger on Legato, they tie and slide. Steps (and their note-offs) land with audio-clock precision.

Step sequencer. The held key is the root: it transposes the phrase (semitones) · drag a step = pitch · short click = rest · notes (selector next to the phrases): SEMITONES above the key or DEGREES of a scale from the key — the same phrase in major, dorian or pentatonic; «harmony» follows the scene's scale (Parts) · articulations (as on a 303): Shift + click = accent (that step louder than the rest, yellow cap, and the matrix source «Accent (Seq)» is 1 on its note: the preset decides what it does — open the filter, more drive…) · Alt + click = tie (the note isn't cut: the next step carries it on without an attack; a dash) → slide (it also glides to the next step's pitch with Glide, or 60 ms if Glide is 0; a diagonal) → none · Ctrl + click = repeats: the step fires 2, 3 or 4 times (split cap). In the phrase text: «7!», «5~», «3/», «0x3», combinable («7!/»). · Latch: the phrase keeps going when you let go · Retrig: if the phrase is ALREADY playing (Latch or legato), a new key restarts it on the grid; with Off only the root changes. Starting from stopped always begins at step 1 · in mono with Gate 1 the steps tie (glide) if Trigger is on Legato.

Mod Env — a FREE envelope for the matrix: pick it as a source («Mod env», 0..1, doesn't scale with velocity) and send it wherever you like (wavetable, Drive, pan, a mixer level…) without touching the filter's. Mode: «normal» = ADSR; «loop» = attack and decay repeated for as long as the note lasts (from 0 to 1 and back to 0; sustain doesn't count): a timbre rhythm tied to each note, with no tempo — A 10 ms + D 90 ms = 10 times a second; on release, the release starts from wherever it is; «trigger» = ignores the note-off: full attack, decay and release even if the note lasts an instant (percussion, arpeggiator hits). Curve like the Amp Env's: 0 = straight ramps, 1 = RC (falls fast and settles).

Steps — four lanes of per-step values (−1…+1) that the matrix uses as sources: «Steps i» is a continuous signal (the step also moves a held note: a trance gate is one lane on Amplitude; a pad that changes timbre in steps, on Cutoff or Wavetable) and «Steps i · note» takes the value of the step where each note starts and holds it for the whole note (arpeggios and plucks: each note with its own color, no steps inside). Draw by dragging; len = how many steps (up to 32); with different lengths the lanes only line up again after many bars (16 · 12 · 5 → 15 bars): complex while staying rhythmic — the line above tells you how long. Div = how long each step lasts, on the common clock grid (the one the arpeggiator, sequencer and conductor use). Smooth = from step to ramp. Rot = start N steps later. Drift = one more step of rotation every N bars: Reich's phasing. Reset: on play (the first note after silence is step 0) · never · every N bars.