Oscillators

Everything you hear from Phaseburn One starts with four sources: the Phase-Locked Loop (PLL) that gives the instrument its voice, a VECTOR phase-shaping oscillator, the MORPH oscillator with its twelve waveform families, and a sub. This chapter walks through each one — what it is for, how to play it, and every control it exposes.

Four sources, one voice

Pick the sources you need, balance them with four volume knobs, and the instrument plays them as a single line. All four live on the OSC tab — a section each for the PLL, VECTOR, MORPH and the sub — and every section ends in its own volume knob. Their outputs sum into one voice: Phaseburn One is monophonic by design, so the entire processing budget goes into making that one voice deep rather than eight shallow ones.

Each source has a distinct job. The PLL is the lead character — a feedback circuit whose struggle to stay in tune is the tone itself, covered in the next two sections. VECTOR is the sculptor: two knobs bend a wave from soft to snarling, with a formant mode for vowel-like resonances. MORPH is the conventionalist, holding every classic waveform and a few less classic ones on a single morphing knob. The sub is the anchor: low-end weight with shape and saturation controls, and routing that keeps it clean when the rest of the patch gets violent.

Blending is direct: raise the volume of what you want, pull down what you don’t. The volume knobs top out at unity, so the balance you set is the level the mix bus receives, and the same four volumes reappear as channel faders on the Mixer tab. There, each pitched source also gets a PAN knob — a constant-power stereo balance that is bit-transparent at centre and a modulation destination in its own right, so a source can sit off to one side or swing across the field under an LFO. Stereo character itself comes from the pitched sources — the PLL, VECTOR and MORPH each run independent left and right engines, so width is built at the point of generation instead of being pasted on afterwards. The sub is the exception: it is mono, feeding both channels equally with no pan, which keeps the low end centered where it belongs.

The tables at the end of each section in this chapter list every parameter with its range and default; the same data, compact and complete, lives in the full parameter index (Chapter 11 — Reference). Prose in this chapter deliberately avoids repeating numbers the tables already carry.

The Phase-Locked Loop

Make a note lock solid and weighty, then loosen the loop until the pitch has to fight its way to every note. The Phase-Locked Loop (PLL) is Phaseburn One’s main voice, and the fight is the sound.

A phase-locked loop comes from radio engineering. One oscillator listens to a reference signal, measures how far its phase has drifted, and corrects itself continuously. In every textbook application the loop exists to erase that error as fast and as invisibly as possible. Phaseburn One is built on the opposite idea: the error is the interesting part. An ordinary oscillator jumps to a new pitch instantly and perfectly. A loop has to get there — it notices it is wrong, accelerates, brakes, overshoots, corrects again — and that whole struggle happens at audio rate, in the waveform itself. Engineers spend careers suppressing this behavior. Here it has knobs.

Keep the loop tight and it locks: solid pitch with analog character, because the correction never fully rests — a live feedback process breathes under the tone. Loosen it and the chase becomes audible: the oscillator glides toward each note, overshoots, rings around the target before settling. Fast lines turn into sirens; held notes with the loop barely holding on turn into growls.

Three knobs are the core, and they map directly onto the physics:

TRACK sets how fast the oscillator chases the reference. High values snap to the note; low values drag, so every interval becomes a glide the loop has to earn. Push TRACK high enough and it overtracks — the loop overshoots in bursts.

DAMP sets how hard the correction brakes. Heavy damping eases into the target pitch. Light damping arrives too fast, overshoots, and rings — a pitch that wobbles around the note before it lands.

INFLUENCE sets how much the reference pulls at all. Back it off and the oscillator only half-believes the note you played; the loop drifts, leans, and finds its own way there.

None of these are macros over hidden parameters — they are the loop’s actual coefficients, scaled to musical ranges. The reference the loop chases is an oscillator of its own, and you can move the target: transpose it with the octave and tune controls, or detune it slightly with the fine control so the lock leans and slowly beats. OCTAVE reaches four octaves down from whatever you play, which on a low note takes the reference under hearing — territory with its own rules, described at the end of the next section.

Pushing the loop

Sirens, growls, interval chases and squelching acid lines all come out of the same handful of controls, because they are all the same circuit failing in different directions. No matter how hard you shove the loop, it always wants to come home — every excursion bends back toward the note you played. That gravity is what keeps the chaos playable.

MULT, DIV and DIV LOCK. The loop can track at a ratio of the reference instead of the note itself. MULT climbs in doublings, octave by octave; DIV divides. Together they set the ratio the loop chases — note × MULT ÷ DIV — and RATIO SLEW glides between settings over a musical length of time, so interval sweeps become something you perform rather than program.

Both ratio knobs carry an S badge, and both arrive with it lit. Snapped, they land only on the whole ratios above, one detent per drag step. Switch the badge off and the same knob sweeps everything in between — ×3, ×5.5, ÷2.75 — for ratios that beat against the note instead of consonating with it. Toggling the badge never moves the sound: it changes how the knob travels, not where the loop is, so you can find a ratio free-hand and then relight the badge to keep the dial on the detents.

Alone, MULT only ever reaches octaves, which the OCTAVE knob already does. DIV is what makes the ratio musical: it is the second divider of every hardware frequency synthesiser, and dividing by something that is not a power of two lands the loop on intervals no amount of MULT can reach. Divide by three and the oscillator drops a twelfth below the note; by five and by seven it descends the undertone series, the mirror image of the harmonic series and the reason organ builders bothered with subharmonics at all. Combine the two and the ratios turn into intervals: MULT 4 over DIV 3 is a perfect fourth above, 8 over 5 a minor sixth, 8 over 3 an octave and a fifth. Because the whole ratio is one number to the loop, DIV inherits everything MULT has — the glide, the modulation and the snap grid.

The DIV LOCK switch — a divider lock, and a different control from DIV — decides what “track at a ratio” means. Off, the loop chases the ratio the way it always has: the correction never finds a true resting point at the target interval, and the residue is a gritty shimmer of extra frequencies clustered around the note. On, a divider goes into the loop’s feedback path, and the oscillator genuinely locks to the ratio — MULT and DIV become an interval selector with the solid, glued quality of a true lock, a slow hunting wobble when disturbed, and an audible swoop as it captures each new note. Ratios that are not whole harmonics need that lock to hold their tuning; without it they are a pitched fight rather than a note, which is its own instrument.

Four detectors. The part of the loop that measures being wrong is the phase detector, and the detector selector picks its temperament. PD is a smooth analog-style comparator: it measures continuously and produces a rounded, elastic error — the misbehavior stays melodic. EDGE reacts only at signal edges, so the error arrives in abrupt decisions: locks are snappier, failures are harsher, and the same TRACK and DAMP settings turn noticeably more aggressive. XOR is the classic logic-gate comparator: it settles a quarter-cycle off-center and lets a fast ripple leak into the correction as a constant buzz, and because it hard-limits whatever reference it hears, it will happily lock onto odd harmonics instead of the note — a detector with opinions. S&H samples the loop’s position once per reference cycle and holds it, so the correction moves in steps: quiet and steady when the loop is relaxed, audibly hunting in staircase wobbles when TRACK and INFLUENCE push it — and with GRIT feeding the detector noise, the held value itself turns random, a pitch that stumbles around the note. A companion parameter, edge sensitivity, sets how small an edge still counts for the edge-triggered detectors — it has no panel control and is reachable only as a host automation parameter.

The reference below hearing. Wind OCTAVE far enough down on a low note and the reference falls under the bottom of the audio range — a few cycles per second, then less than one. The loop keeps working; it stops being about pitch. Its edges become rhythm, and the controls you were using to tune a note now shape a pulse. PD compares phase continuously, so it simply follows the reference down: the oscillator sits on it, and with OUT SHAPE toward pulse the voice turns into a click train you play from the keyboard, or a drone you can beat against another voice. EDGE has nothing to measure between one edge and the next, so it ratchets — long still stretches broken by a fast chirp each time the reference crosses. XOR breaks into bursts. S&H holds its last measurement until the next edge arrives, which down here is a very long hold; below a couple of cycles per second it wanders rather than tracks, exactly as hardware sample-and-hold loops do. Nothing in the signal path strips this material out: sub-audio output moves woofers and pumps a limiter at the reference rate, so the master high-pass is worth switching on when a patch lives down there.

Patches saved before the reference could go this low keep the behavior they were written against, where it stopped at the bottom of hearing and the lowest OCTAVE settings on a bass note all sounded the same. The switch that carries that choice, PLL Ref Floor, has no panel control and is reachable from a host’s parameter list; setting it to FULL opens an older patch up. It also clears a tuning error the old limit caused — with the reference pressed against the bottom the loop’s hunt was cut off on one side only, which dragged the pitch of low EDGE patches sharp by as much as a third of an octave.

Injection, bursts, and the loop lab

Past the detectors and the lock, the loop takes passengers: signals fed into its correction path on purpose, a chaos drive that takes its stability away, and a rack of four ways to break the servo itself. This is where growls, whoops and glitches live.

BURST and INJECT. BURST is overtrack intensity: push TRACK above the threshold you set with BURST THR and the loop’s own overshoot feeds back into the correction as extra kick, scaled by BURST AMT — chirps, whoops and squeals that always resolve back to pitch. The threshold works in every color mode, so it is always on the panel. INJECT feeds the reference straight into the correction path, from a subtle undertow of growl at the bottom of its range to controlled chaos at the top. A switch beside it picks what the injection is: INJ FM modulates the pitch with the reference waveform for sideband growl, while INJ LOCK is true injection locking — the reference captures the oscillator and holds it within a range set by the amount, so detune the reference and the tone bends, fights, then breaks free at the edge of the capture range and snaps back when you return. (It is a different lock from DIV LOCK — one couples the injection, the other divides the feedback path.) RANGE narrows the main loop’s own lock bandwidth — the chase, not the injection — so low settings hunt slowly toward the note like a worn analog circuit.

CHAOS. The loop normally settles at a stable equilibrium; CHAOS takes it away. The knob — first in the advanced row — overdrives the detector’s error curve until competing lock points appear, raises the loop gain past where the servo can settle, and slows the error smoothing so every correction arrives a little too late — while a slight keytracked wander keeps the fight seeded. Low settings ring the loop into a vibrato-like growl. The middle mode-hops between rival equilibria in period-doubled warbles. At the top the correction scribbles chaotically around the note without ever abandoning it, because everything still happens inside the loop’s own limits — RANGE, the integrator’s saturation and the detector all keep their say, so each detector breaks down with its own accent.

The COLOR switch adds harmonics drawn from the loop’s core, with a mode menu picking the flavor. HARM blends in an upper partial that climbs with MULT — bright, glassy, sometimes gloriously rude — and SAT drives the output itself into saturation, from a gentle analog warmth at zero to a soft square rasp at full; in both, SATURATE sets how much. FOLD synthesizes a fold-back partial — the clang of a harmonic reflected off a virtual rate at RATIO times the note — with FOLD setting its level and RATIO choosing the inharmonic interval, which stays constant across the keyboard because the fold rate tracks the note. SATURATE, FOLD and RATIO all stay on the panel in every mode; the mode menu decides which of them the engine is listening to. Beside the mode menu sits the RAW / CLEAN switch: RAW, the default, lets the FM operator and HARM’s partial fold past the audio band into the rich high-register hash the factory sounds were voiced against; CLEAN tapers and clamps them for clean ringing at the top of the keyboard. An FM section modulates the output oscillator at a ratio of the pitch — the FM and FM RATIO knobs in the advanced row. An envelope-to-FM depth — for attacks that clang and then settle — exists as a host automation parameter only, with no panel control.

The loop lab. Four knobs break the servo itself — they follow CHAOS in the advanced row — each one a real defect of hardware loops pushed to musical extremes, each a modulation destination, and all of them at zero by default, where the loop is exactly the stock circuit. STEP samples-and-holds the whole correction at a servo rate: continuous at the bottom, and as it slows the pitch hunting starts moving in discrete steps — R2-D2 chirps, bubbling glitch arps, gated glissandi. GRIT feeds noise into the phase detector upstream of the loop smoothing, so TRACK sets its color: analog fizz at low drive, a drunk wobble in the middle, a full random stumble at the top — every detector gets the trick the old noise reference reserved for S&H. LAG inserts a transport delay into the correction path, up to thirty milliseconds, so the loop steers by where the note was: metallic servo flutter at short settings, standing-wave hunting and time-smeared chases at long ones. RELOCK opens and closes the loop in a cycle, from a slow breath up to twelve times a second: while open the corrections freeze and the pitch sags away; at close the loop wakes to everything it missed and fights its way home — the capture swoop, the best sound a PLL makes, on demand.

Locking the loop to something else

Everything above assumes the PLL is chasing its own internal reference. It does not have to. The patchbay has a PLL REF jack, and whatever you patch into it becomes what the loop tries to lock to instead — including audio arriving from the host, whether Phaseburn is inserted on an audio track or fed from another track’s send. The INPUT REF knob — on the PLL section’s main row, just before OUT SHAPE — blends between the two: at INT the loop is exactly as described above, and toward EXT it hands the reference over. It is a modulation destination like everything else on that row, and it glides rather than steps, so an LFO or an envelope can open the loop to the outside world and close it again mid-note.

This is not pitch tracking, and it is more interesting than pitch tracking. The note you play still decides where the loop sits; the reference pulls it around within the capture range. Every control you already know becomes a way of shaping how the oscillator reacts to the incoming signal — how hard it grabs, how far it can be dragged, how it recovers:

• Feed a drum loop and open RELOCK and BURST. The pitch lurches on every transient and resolves back, so you get a tuned follower locked to the groove without a clock anywhere in the patch.
• Feed a bass line or a vocal, switch DIV LOCK on and bring MULT up. The loop locks a harmonic of the input rather than the input itself.
• Play deliberately against the incoming pitch. The servo cannot decide, and you get beating, chirping and slow chaotic wandering — the sound of a loop failing, which is the reason to have it.
• Switch the detector to XOR. It multiplies the input and the oscillator square together, which is ring modulation bound to the source’s rhythm.

Level matters. The comparator only registers an edge once the reference crosses ±0.02 — about −34 dBFS — so a quiet input never triggers a lock at all. Drive AUDIO IN until its meter peaks near the top, which is roughly the amplitude of the internal pulse it is standing in for; the strip has +12 dB above unity for exactly that. The input is soft-clipped above −3 dBFS, so pushing hard costs character rather than stability.

The detector matters here. XOR, S&H and EDGE all read the incoming waveform directly. PD compares phase against a signal the input does not carry, so it keeps following the played note — useful when you want part of the loop anchored.

PLL REF also takes Phaseburn’s own oscillators. Patch MORPH, VECTOR or SUB into it and you have cross-modulation with no host routing at all: one oscillator steering the loop. Those taps are one sample behind — the PLL is computed before the other oscillators inside the voice — which is exactly how patching one analog oscillator into another’s reference input behaves.

Effects returns are deliberately not offered as reference sources. The effects run after the voice, so a reverb tail would arrive a whole block late, and a lock reference that stale smears every transient.

Coloration

Perfect stability is a digital habit; the coloration stage is where you break it. It applies its character to the PLL voice — the core of the instrument — rather than politely post-processing the mix, which is why its knobs — DRIFT, DRIFT RATE and TUBE — live in the PLL section of the Oscillators tab rather than next to the filter.

DRIFT lets the pitch wander. It applies a slow wobble to the PLL’s reference — the pitch the loop chases. Low amounts read as an aged oscillator that no longer quite trusts its tuning; higher amounts become an audible, seasick waver. DRIFT RATE sets how fast the wander moves. When the PLL is spread into stereo (raise TRACK Δ or DAMP Δ — see the oscillators chapter (Chapter 2 — Oscillators)), the right channel’s loop gets its own wander at a slightly different speed from a different starting point, so the two sides slip against each other and the stereo image slowly leans; on a mono PLL both channels share one wander. Because the wobble goes through the loop rather than around it, the PLL’s own tracking behavior — lag, damping, overshoot — shapes the drift on its way to the output, which is why it feels like a machine misbehaving rather than an LFO on pitch.

TUBE is the matching wear on the amplitude side: soft-clip saturation on the PLL voice’s output. Low settings round the peaks and thicken the tone; pushed hard it compresses the waveform’s edges into a denser, warmer growl, and the dial keeps going past 100 into outright crush. DRIFT and TUBE are both live modulation destinations, so an LFO or step modulator can make the mess itself move — see the modulation chapter (Chapter 4 — Modulation).

The VECTOR oscillator

Two knobs take one wave from soft and hollow to snarling — and the same two knobs can park a vowel-like resonance on any harmonic you choose. That is VECTOR — vector phase shaping, a technique published by Kleimola, Lazzarini, Timoney and Välimäki in 2011, implemented here in double precision.

Instead of filtering a bright wave or crossfading between stored shapes, VECTOR bends the phase that reads the waveform. Picture the wave’s single cycle being traced by a clock hand: X and Y grab that clock hand and bend its path through one inflection point. X places the bend horizontally — how early or late in the cycle the distortion strikes — and Y sets its depth, which is the main timbre sweep. Low X with moderate Y is smooth and rounded; high X with low Y turns aggressive. Because you are shaping phase rather than amplitude, the sweep has a vocal, resonant quality that a filter sweep does not.

Formant mode. The FORMANT switch extends Y’s reach into true formant synthesis: the resonant peak locks onto a harmonic of the note, and sweeping Y walks it up the harmonic series. Peaks are strongest when they sit exactly on a harmonic, so slow Y sweeps produce the stepped, vowel-like character of a voice changing shape. Melodies keep their formants in tune because the peaks are multiples of the fundamental.

Stereo divergence. X Δ and Y Δ spread the bend coordinates symmetrically around the panel values: the left channel shifts down by the offset, the right channel up. Both channels run at identical frequency, but because their phase paths differ, they gradually drift apart and the stereo image evolves over the length of a note — width generated by the shaping itself, not by delay or detune. This is deliberate, and for pads it is the VECTOR section’s quietest strength.

Phase modes. A phase-mode parameter — host automation only, with no panel switch — chooses how each note begins. FREE starts each note at a slightly randomized phase, so repeated notes vary. SYNC resets the phase to zero at every note-on for a consistent, hard-sync-like attack on each note.

A sine-based wavefolder (FOLD, with a range switch that folds far deeper) and a waveshaper (SHAPE, with Soft, Hard and Fold curves) stack on top for saturation and buzz. At extreme X and Y settings, raise the oversampling — a host-automatable engine setting with no panel control, fixed at 2x by default (see Engine settings (Chapter 11 — Reference)) — and the engine applies strong alias suppression, giving its cleanest results with headroom to work in.

Like MORPH, the VECTOR section carries TUNE and FINE: the interval in semitones and the cents either side of it, which is what you reach for when stacking VECTOR against another source and you want beating rather than a chord.

MORPH: twelve families on one knob

When a patch needs a saw that becomes a square, a vowel, a rising sync lead or a wall of detuned width, MORPH is where you go. It is the conventional-oscillator section — with the convention stretched: one knob, and a mode selector that decides which of twelve waveform families that knob travels through.

The six crossfade families blend their waveform pairs with equal-power crossfades, so perceived loudness stays even along the travel, and switching families mid-performance crossfades in a few milliseconds — no clicks, even with the knob under heavy modulation. The knob itself is a modulation destination, which is where this section gets fun.

The everyday crossfades. Saw to square, sine to saw, and triangle to square cover most bread-and-butter needs — pick the pair, set the knob, done. Saw-to-pulse narrows the pulse as it fades, turning increasingly nasal and resonant.

The sweeps. FULL runs five waveforms in a single travel — sine, triangle, half-rectified sine, saw, square — the complete gentle-to-bright range on one control, made for a slow LFO. RECT morphs sine through half- to full-rectified shapes: buzzy and organ-like, doubling in pitch at the top. HARM builds a saw additively, one harmonic at a time, and is inherently band-limited.

The performers. SYNC turns the knob into a hard-sync ratio — the classic rising harmonic sweep for leads, corrected at the sync point so it stays clean. SWRM (Swarm) is a detuned unison saw: seven voices per channel, the knob setting the detune spread, with the STEREO Δ control taking it from a true mono stack to full stereo width. STRS quantizes a sine into coarse steps for digital, retro grit.

The voice. FMNT is real formant synthesis: a bank of vocal-tract resonances shapes the oscillator the way a mouth shapes the vocal folds, and those resonances stay fixed in frequency no matter what note you play — so a vowel keeps its identity up and down the keyboard, and vibrato or glide sweeps the harmonics through the standing formants exactly the way a singer’s does. The knob sweeps the vowel — A, E, I, O, U — for vocal pads and leads; with the PITCH switch on, the played note also shifts the vowel position — higher notes land on later vowels — and the knob offsets from there, stopping at the last vowel. Two controls appear only in this mode: BREATH mixes in aspiration noise shaped by the same resonances — from a hint of air to a whispered pad — and SIZE scales the whole vocal tract, from big and dark to small and bright. Both are modulation destinations, so an LFO on SIZE talks and a slow ramp on BREATH exhales.

The wavetable. SPCL walks a spectral table from plain harmonics through tilted, combed and resonant spectra.

Around the morphing core: FOLD and SHAPE are the same wavefolder and waveshaper tools as the VECTOR section. TUNE and FINE set the interval and the cents — FINE is the control for the small detunes that thicken a stacked patch without pulling it out of key. STEREO Δ outside of Swarm detunes the left channel down and the right channel up for immediate width and slow chorus-like motion.

The sub

Anchor the low end with a source whose only job is weight. The sub tracks the played note at a whole-octave offset — the OCTAVE control chooses how far down, with its top position sitting at the played pitch itself. Like the other sources it rides the mix bus, so by default the main filter shapes it along with everything else; when a screaming resonance sweep must not hollow out the foundation, the routing choices below keep the floor solid.

SHAPE morphs the sub’s waveform continuously through five stops: sine, triangle, saw, square, and a narrowing pulse. Leave it at sine for pure, invisible weight that a mix feels rather than hears; triangle and saw add just enough edge for the sub to read as a note, and square is the classic hollow foundation. Past square the pulse narrows continuously, down to a tenth of the cycle — reedy and nasal, with the energy moving up out of the fundamental. That last stretch is level-compensated, so the sub keeps its weight as the pulse thins instead of disappearing the moment you leave the square. SATURATE adds harmonics on top of whatever shape you chose — the practical effect is that the bass becomes audible on laptop speakers and earbuds, where the fundamental itself does not survive.

Routing is where the sub differs from the other three sources. On the patchbay a dedicated SUB → FILTER cable adds a direct send into the main filter — useful once you have pulled the sub off the mix bus but still want it filtered. And an FX BYPASS switch in the SUB section’s own header — the same control as the FX BYPASS button on the mixer’s SUB strip and the patchbay’s SUB → OUT cable, all three mirrored — routes the raw sub around the tone-shaping chain entirely when engaged, injecting it near the output: heavy coloration, distortion and character stages on top, untouched fundamental underneath. Because the direct injection happens after the master high-pass, the bypass is also how a sub survives aggressive low-end cleanup. It is the difference between a bass patch that gets dirty and a bass patch that gets dirty while the floor stays solid.

Glide and legato

Make lines slur, slide and breathe instead of stepping. Phaseburn One has two independent glide mechanisms, and knowing which one fires when is the whole trick.

GLIDE lives in the PLL section and is always-on portamento: every note transition slides, taking the time you set to reach the new pitch. At zero, pitch changes are instant. Because the PLL is a loop chasing a reference, glide interacts musically with TRACK and DAMP — a slow glide with light damping arrives at the note, overshoots it, and rings, which no ordinary portamento does.

Legato is conditional: it fires only when notes overlap — a new note-on before the previous note-off. Raise the Legato slider off its bottom Off detent — it lives in the GLOBAL module below the beat grid on the Groove page — and an overlapping note does not retrigger the envelope; the sustain carries through and the pitch glides over the time you set, independent of GLIDE. A velocity-lock option, in the PERFORM module on the Groove page’s Notes tab, keeps the held note’s velocity through the transition, so a phrase keeps one dynamic level instead of jumping with every new finger.

The sequencer participates fully. When a step’s note length pushes past its beat division, consecutive sequenced notes overlap — and with legato mode on, they take the legato path automatically: no envelope retrigger, smooth glide. Shorter steps articulate normally. This means one sequence can mix tied, sliding passages and punchy re-attacks purely through the note-length lane.

When several MIDI notes are held at once, the note-priority setting decides which one sounds: the most recent, the lowest, or the highest. It has no panel control — it is exposed to the host rather than drawn on the window — and low priority with legato gives the classic bass-player behavior: hold the root, hammer notes above it, and the line always falls back home when you release.

Oscillator independence

Pull pseudo-polyphony out of a monophonic instrument. Each of the four sources — PLL, VECTOR, MORPH and the sub — can independently sit out a note or play a different pitch than the one the voice was given. Two knobs in the OSC INDEPENDENCE module on the Groove page’s Notes tab control it.

SKIP is a per-note probability that a source stays silent. When a note fires, every targeted oscillator rolls the dice separately: at moderate settings roughly half the sources sound on any given note, and the timbre thins and thickens from note to note. A four-source patch stops being one fused tone and becomes a small ensemble that does not always show up in full.

DRIFT is a per-note probability that a source plays a different note — chosen from the same note pool, with the same chance-and-strength weighting, that drives the sequencer’s own note selection. At low values, an occasional source leans away from the line and adds interval color. At high values, each oscillator effectively walks its own melody: moving harmony out of one voice.

Four target switches decide which sources participate. Anything untargeted always plays the base note at full presence, so the usual move is to protect the anchor — leave the sub untargeted so the low end never gambles — and let the upper sources skip and drift above it. Both knobs are modulation destinations, so an LFO or step modulator can push a sequence from disciplined to scattered and back across a phrase.

Skip and drift are saved with the preset and roll fresh on every note, so no two passes through a sequence land identically — which is the point.