The listening room.

The call, filtered

The note is read dry for six seconds; the full-rate codec then takes the line. What arrives is the 1994 call — half-swallowed, faintly underwater, exactly as designed.

Eight generations down

Degrade is walked from nought to full — one encode generation stacking toward eight — and at twenty-four seconds Rate drops to Half. Distance accumulates.

A bad line

A bad line, assembled deliberately: Half rate, handset Noise, adjacent-channel Crosstalk, and DTX swapping each pause for comfort noise.

The handset, too close

The material plays clean for six seconds; then the buzz arrives at bench settings — Rate at 217 Hz, Pattern set to Voice — the cadence a handset leaves in nearby equipment.

Voice, Idle, SMS

Pattern is stepped through Voice, Idle, and SMS burst against a held pad — three cadences of the same interference — and Rate then climbs from 100 Hz to 400 Hz.

Buzz on the line

A call, interfered with deliberately: Drive at 90 % demodulates hard, Bleed crossfades from ring-modulated speech to the injected burst itself, Tone tilts the pulse dark and then bright, and Sync locks the frame rate briefly before it runs free again.

A weak cell

The note reads plainly for six seconds; then the concealment begins — Erasure at its bench 15 %, each lost frame patched with a repeat of the last good one, fading at 3 dB a frame.

More and more missing

Erasure is walked from nought to sixty percent — the repeats and freezes thickening as the frames go — and at twenty-four seconds Burst rises to ten frames, so each loss holds longer.

Dropped by hand

Mode set to Manual: each dropout is forced with Trigger, first sparingly, then in a closing run — Burst raised to twelve frames, Attenuation at 6 dB a frame, Hold at six, so each long dropout fades and then goes to nothing.

Three bars

The note reads dry for six seconds; then the codec takes the line at bench settings — Signal at 60 %, three bars, Mode held Manual — the rate ladder sitting mid-table.

Walking out of coverage

Signal is walked from five bars to none over the length of the note. The codec steps down its ladder as the bars fall, the voice coarsening rate by rate; near nothing, the frame losses begin.

The signal wanders

Mode set to Auto with Drift at 80 %: the bars wander on their own around a low Signal, the rate climbing and falling with them, and with Dropouts at 60 % the losses come and go where the signal thins.

A near-perfect silence

Dry for six seconds; then the unit takes the line. Where the voice stops, the detector declares silence and substitutes its own estimate of the room — and this room is next to nothing, so the substitution is next to perfect. Update refreshes the estimate every 480 ms.

The detector, advancing

Threshold is walked from −60 up to −20 dB: at first the detector cedes nothing; then the pauses go over to it; by the end it is taking the ends of words — handing back a frozen breath of each one in place of its decay. Noise Level stands at 80 %.

Words, cut short

Threshold raised to −20 dB with no Hangover: the detector calls the end of every phrase early and hands it to the noise generator — the last breath of each word frozen and held rather than allowed to die away. The machine decides, a little too early, that no one is speaking.

The local line

The note reads dry for six seconds; then the line takes it — 300 Hz to 3.4 kHz, the companding crunch at its bench 60 % — and the voice arrives as a call.

The band, closing

Low climbs from 200 to 600 Hz while High falls from 4 kHz to 2.4 kHz — the band closing on the voice from both ends — and Compand then brings the µ-law crunch from nought to full.

On hold

The mix is sent down the line: the Dial tone opens the call, Line Noise at 70 % and Hybrid Echo at 60 % colour the passage — the music of a phone left off the hook — and the Busy signal ends it.

The ring, plainly

A monophonic ringing phrase at bench settings — the Square wave through the piezo-buzzer model, Buzzer at 60 % — repeated, as the machine insists.

Three shapes and a slide

The same phrase through Square, then Pulse — Pulse Width swept from 5 to 95 % — then Triangle; Glide at 80 ms lets the final phrase slide between its notes.

The buzzer, in and out

Drive at 90 % clips the wave hard while Buzzer walks from nought to full — the plastic resonance arriving — then drops away and returns for the close. Attack at 10 ms, Decay at 800 ms.

Sixteen voices

A chord cycle in A minor with a melody above — the FM voices at bench settings, Algorithm 1, the small speaker doing what it can.

Four algorithms

Algorithm steps from 1 through 4 as FM Depth rises from 20 to 70 % — the same chords recast four ways — and at the close Ratio jumps to 5.0 and the voicing goes to bells.

Four voices, then thirty-two

Eight-note chords against a four-voice cap — the earlier notes stolen mid-ring — then Voices is raised to thirty-two and the same chords hold whole. Speaker at full.

The connection

Trigger starts the full sequence — the dialling, the answer tone, the training screech, then the carrier settling in — a V.90 negotiation heard end to end.

Three negotiations

The same handshake spoken three ways — Standard stepped through V.32, V.34, and V.90 at Speed 1.5 — each era of the line with its own accent.

Connected

The mix passes untouched for six seconds; then the line takes it — Scramble at 70 % imposing the modem's texture on the material while the Carrier level rises beneath it.

The voice, played

The note reads dry for six seconds; then the Internal carrier takes it — a held line sung by the filterbank, sixteen Bands wide — and the reading becomes the machine's.

Eight bands to thirty-two

Bands step from eight — barely words — to sixteen, then thirty-two, the reading sharpening at each move: the resolution of the filterbank, heard arriving. Mode then switches to LPC, and the same voice is carried by the all-pole model instead.

Spoken through the pad

Carrier switched to Sidechain: the pad is made to speak. Sibilance rises to 80 % and the consonants return; Formant is pushed up seven semitones, then down; Attack and Release open at the close and the words smear.

The floor, held

Dry for six seconds; then the box takes over at bench settings — band-limited and boxy, the far-end feed ducking the near voice as the unit decides who may speak.

One voice at a time

Duplex walks from full — both parties audible — to hard half-duplex, one voice at a time; Switch then slows from 5 to 200 milliseconds and the handover becomes a pump.

The box itself

Room at 90 % and Band down to 2.5 kHz — the enclosure at its worst. Echo Cancel steps to full at ten seconds, scrubbing the reflection and pumping as it works, then falls away again.

The message, saved

The note reads dry for six seconds; then the machine answers — Microcassette at bench settings, the Band at 3.2 kHz, a little Wow, the familiar hiss floor — a message kept.

The tape, dying

Wow at 60 % keeps the pitch unsteady from the first word, Noise raised to 50 %; Wear then climbs from nought to full, and the dropouts take the message a piece at a time.

Solid state

Medium switched to Digital: the Greeting beep opens the take, Wear at 70 % sets the crush depth and Band at 2.4 kHz narrows the voice, Compand at 80 % crunching what remains; the Record tone closes it.

Eight bits, truncated

The phrase runs dry for five seconds, then the bench default: Bits at 8, Quantize set to Truncate, Dither off. The grain sits in the decays, where the level falls and the staircase shows.

Sixteen down to two

Bits steps from 16 down to 2 against calibrated tones, Quantize on Truncate, Dither off. The steps are audible by eight; by two the staircase is most of what remains.

What dither is for

Held at Bits 4 with Quantize set to Round. Dither steps from Off to TPDF, then Shaped-Hi; Dither Level rises to 2 LSB, and Bias tips the staircase off centre at the end. The grain trades for hiss, then the hiss moves up out of the way.

Eleven kilohertz, plainly

The phrase runs dry for five seconds, then the bench default: sample-and-hold to 11.025 kHz, anti-alias off. The top octave folds back into the material.

The rate walked down

Rate is walked from 26 kHz down to 1 kHz in one pass — sample-and-hold, no anti-alias filter. The foldover arrives on schedule; nothing here is misbehaving.

Two ways down

The same figure taken down two ways: Band-limited at 4 kHz with the anti-alias filter in, then Sample-and-Hold with it out. The Output Filter closes to 3.5 kHz at the end.

The calibrated wander

The pad runs dry for five seconds, then the bench calibration: Amount at 20 %, Random at 60, Periodic at 40, Rate at 400 Hz. A fine unsteadiness in the top of the chord — more felt than heard, which is the point of a reference.

Sidebands on schedule

Amount rises from 0 to 100 % over calibrated tones, then Rate walks from 1 Hz to 5 kHz. The sidebands step away from each tone as the periodic component climbs — exactly where the arithmetic says they should be.

Random against periodic

Amount held at 70 %. The first half is Random alone — a broadband smear across the chord; at twelve seconds Periodic takes over and the smear resolves into discrete sidebands at the Rate. Interp moves to Thiran for the close.

One twenty-eight, nominal

The mix runs dry for five seconds, then the codec takes it at Bitrate 128 kbps, Block Switch on Auto. What goes missing is what the masking model says you would not have heard anyway — mostly.

Down the ladder

Bitrate steps from 320 down to 32 kbps in six moves. The top of the band thins first, then the texture; by 32 kbps the model is rationing, and it shows.

The threshold, moved

At 96 kbps, Masking rises to +12 dB — the model discards more than it should — then falls to −12, where it keeps what it need not. Block Switch goes Off at twenty seconds and the hits smear forward; Joint Stereo closes the image at the end.

Twelve kilohertz, held

The mix runs dry for five seconds, then the converter model: Zero-Order Hold at a 12 kHz Rate. The images sit above the passband, unfiltered — here, the reconstruction is the instrument.

Forty-eight down to eight

Rate falls from 48 kHz to 8 kHz in one continuous move, Zero-Order Hold throughout. At the top the model is nearly transparent; by the bottom the images have walked down into the music.

Three reconstructions

The same figure at a 16 kHz Rate, three ways: Zero-Order Hold with its images, Band-limited clean, then Brickwall — Ringing raised to 100 % puts the pre-ring in front of every hit. Bandwidth closes to 8 kHz at the end.

A finishing tool

The reference pass at 48 kHz, 16-bit Word, TPDF Dither, Soft Converter — in at six seconds, out at sixteen, in again at twenty-two. The difference sits at the edge of notice. That is the instrument.

Long-play

Rate steps from 48 to 44.1 kHz — a move below notice — then to 32 kHz, the long-play speed, and the top octave is gone. Word drops to 14-bit at twenty seconds and the floor comes up to meet the tape.

Out of calibration

Jitter raised to 80 % blurs the top end; Converter switches to Brickwall at twelve seconds and rings at the band edge; Dither goes Off at twenty and the quiet detail truncates. None of it is loud. All of it is there.

Plus four per octave

The pad runs dry for five seconds, then returns through the analyzer with Tilt at +4 dB/oct — the spectrum redrawn brighter than it arrived. FFT Size holds at 2048.

Gate rising

FFT Size at 512. Gate rises from −90 to −30 dB across fourteen seconds: bins below the threshold go dark and the pad thins to its loudest partials. The threshold falls away at twenty-two and the rest returns.

Held frame

FFT Size at 8192, Smooth raised to 80 %. Freeze engages at ten seconds and the analyzer holds its frame — the chord becomes a surface — until it lets go at twenty-four and the pad moves again.

One-sixty wide

The pad runs dry for five seconds, then through the M/S matrix at Width 160 % — pushed past unity, the sides carrying more than they were given. Nothing is corrected; the device does only what it is told.

Zero to two hundred

Width falls to mono by nine seconds, then opens to 200 % by eighteen. Rotate swings the field to +45° and snaps to −45° at the close — a line, a cloud, and a tilt, in that order.

The mono audition

Mono Below sums the low end under 300 Hz at six seconds. Fold auditions the whole mix in mono from twelve to eighteen. Display switches to Correlation at the end — a change of reading, not of signal.

Twenty-five per cent

The chords run dry for five seconds, then through the generator at Amount 25 % — Even and Odd both at 50, Order at 6. A measured warmth; the readout would call it THD, and it would be right.

The series, in order

Amount rises from 0 to 100 % over calibrated tones. The harmonic series climbs out of the sine, evens and odds together, up to Order 6 — exactly as commissioned.

Even against odd

Amount at 60 %: the even series alone, then the odd series alone at eight seconds. Order jumps to 12 and the edge brightens; back to 2 — with Even restored — only the octave remains. Bias tips the balance at the close.

The reference, blended

Pink noise at Level −18 dBFS, Blend at 100 — the reference alone. At eight seconds Blend falls to 50 and the material comes up through the noise; at twenty it drops to 25, the reference now a calibrated bed.

The source walk

White; Band noise 500 Hz wide around 1 kHz, the centre then moved to 4 kHz; a Sine at 440 and again at 1 kHz; a Sweep on a two-second period; and the Impulse, once a second, to close. Level holds at −18 dBFS throughout.

A dither bed

TPDF Dither blended at 30 under the material, Level at −18 dBFS. The bed falls away to −40 by fourteen seconds and returns to −18 at twenty — a calibrated hiss.

Room 04, as measured

The hits run dry for five seconds, then Mix rises to 50 % and Room 04 answers — four tenths of a second of measured space behind every strike. Mode stays on Convolve; nothing is embellished.

Plate 18, cut short

Plate 18 with Length at 45 % — the measured decay cut mid-fall, a gated read of the archive. Mix rises to 65 %; at twenty-two seconds the direct signal leaves and only the clipped plate remains.

Reversed, then measured

Chamber 11 played backwards — Reverse on, Tilt at +6 dB/oct, Pre-delay at 120 ms — every hit answered by a swell that builds instead of decays. At twenty-four seconds Mode switches to Measure and the Trigger fires the sweep: the instrument taking its own reading.

ATRAC, plainly

The mix runs dry, then through the recorder at SP from five seconds. What ATRAC discards at standard play is quiet — listen to the cymbal tails and the space around things, not the middle of the mix.

Three speeds down

Mode steps SP to LP2 to LP4, the allocator coarsening each time. Back at SP, Generations then climbs to eight — a dub of a dub of a dub, and the loss compounds.

Narrow, and dark

LP4 throughout. HF Trim shelves the top down eight decibels at eight seconds, and Width then narrows to 20 % — the intensity-stereo image folding toward the centre.

Warm, plainly

The mix runs dry, then through the deck at 48 kHz, 16-bit, Converter set to Warm. The change is small and lives mostly in the top — this is a healthy deck, recorded as found.

Down to long play

Rate steps 48 to 44.1 to 32 kHz — long play, and the band closes in — then Word drops to 12-bit and the floor comes up under the quiet passages.

Three voicings, one bad transport

Converter walks Neutral, Warm, Bright; Jitter then climbs to 70 % and the transport begins to smear. Dither goes Off at the end — set for completeness; on material this dense it is felt more than heard.

Light wear, plainly

The mix runs dry, then on the player at Severity 25 % — the occasional held sample, a short mute, a disc that mostly still plays.

The escalation

Severity climbs from nought to full: the player passes, then interpolates, then mutes, then repeats what it has — the concealment order, in order.

Forced skips

Trigger forces the skip on demand. Stutter stretches from 60 to 300 ms, Seek hunts at 80 %, Mute sits at 70 % — and Sync then locks the stutter to the beat.

128k, plainly

The mix runs dry, then through the coder at 128 kbps. The loss sits at the top and in the space between things — present, but polite, exactly as the era heard it.

The bitrate ladder

Bitrate steps down the ladder — 320, 192, 128, 64, 32 kbps. Somewhere past 64 the swirl arrives and stays; the band narrows with every rung.

Eighth generation

Eight Generations at 64 kbps, Joint Stereo off, then on — the coder splitting, then pooling, its budget — and HF Cutoff finally pinned at 6 kHz. A file passed around until this is what remains.

The hall, plainly

Sticks and hits, dry, then into the Hall program at the factory Mix of 30 % — raised to 80 % at eighteen seconds so the tank itself can be heard. It is small, the tail is metallic, and neither is an accident.

Five programs, one chip

Program walks Hall, Room, Plate, Stage, Vocal — the same sparse chip wearing five names. On Vocal, Decay snaps to 0.2 seconds and then opens out to the full five.

Dark, sparse, twelve bits

Tone at 1 kHz, Density at 10 %, Pre-delay at 60 ms — a dark, clangy, deliberate space. At fourteen seconds Grain drops from 16-bit to 12-bit and the tail picks up its grit.

The ladder, plainly

Chords through the Analog Ladder: dry, then filtered at 400 Hz, then Cutoff swept up to 8 kHz and back down with Resonance at 45 %. The sampler's front panel, doing its plain job.

Three filters

Resonance at 80 % while Cutoff sweeps; Type steps Analog Ladder, Analog SVF, Digital IIR — the same move, three tempers.

The zipper

Smoothing at nought, Drive at 80 %. Cutoff moves in deliberate steps — an A-minor arpeggio of corner frequencies — and every step clicks. In Digital IIR, un-smoothed, that zipper is the point.

The voice, pulled

The song plays untouched for five seconds; the machine then takes the line and Vocal Cut, at 80 %, pulls the centred voice out of it. At eighteen seconds it goes to 100 and the null deepens.

Off-key, then empty

Key drops three semitones, rises three, and returns — the machine's granular shift wobbling as it goes. Vocal Cut then sweeps from nought to full, and the singer leaves the room.

The mic path

A voice arrives on the mic input over the cut backing — Mic Reverb at 80 %, Mic Echo at 60 %, splashy on purpose — and Tone tilts down eight decibels at twenty seconds.

OPL, plainly

A lead and its chords from the OPL3 engine at bench settings, through the card's own 8-bit converter. At fifteen seconds Attack and Decay open out, and the same voice swells instead of plucking.

Four operators, then two

The phrase repeats four ways: OPL3 at bench settings; OPL2's two-operator voice; OPL3 again with Ratio at 4.0; then FM Depth at 80 % with Feedback at 60 %. The card's whole vocabulary, in order.

Through the eight-bit DAC

Input Crush routes a full mix through the card's own converter. Rate falls to 11 kHz, Hiss climbs to 80 %, and Bits flips briefly to 16 before settling back where the card lives — at 8.

The console, plainly

A lead over a bass line with Chip set to Both — the FM voice and its sampled double sharing one tune — at the console's default converter, the last chord ringing out through the echo.

FM, Sample, Both

The same phrase three ways: the raw FM voice, FM Depth and Ratio pushed as it plays; then its sampled double, Interpolation stepping Gaussian to None to Linear; then Both, the two chips layered on one tune.

The cave

Echo at 80 % with Echo Time at 240 ms — every note answered twice. Bits steps 16, 12, 9 — the 9-bit ladder grinding — while Rate falls to 16 kHz.

Eight bits, plainly

The break runs dry, then through the machine at 16-bit, then — at fifteen seconds — the true 8-bit depth. Quiet tails fall below the eighth bit and simply stop.

Off the root

Pitch moves the resample grid off the root — down twelve, up twelve, home again — while the Ceiling falls from 28 to 9 kHz. The aliasing hardens and eases; the notes never move.

The tracker, working

The full mix on the tracker's terms: Linear interpolation gives way to None at ten seconds, the Channels go to hard pans, and Loop Buzz rises to 80 %.

The hall, plainly

Sparse percussion, dry, then the Hall program at 35 % mix. Each hit blooms into the tank and is left to finish.

Decay, opened up

Decay opens from 0.8 to 12 seconds across the take. Diffusion begins sparse enough to count the reflections, then goes dense; Size widens last.

The plate, pushed

The Plate, pushed to the murky end: 12-bit Grain in the delay memory, Mod at 80 %, HF Damp down to 2 kHz, 120 ms of Pre-Delay.

The gate, plainly

Drums, dry, then the Gated program at 40 % mix. Each hit blooms into a dense burst and is cut off at 180 ms of Gate Time — the tail removed, by design.

Gate Time, walked

Gate Time opens from 20 to 500 milliseconds across the take. Density goes full, then sparse enough to count the reflections; Slope moves from a hard chop to a soft close.

The burst, reversed

The Nonlinear program with 12-bit Grain in the delay memory, 60 ms of Pre-delay and a 200 ms Hold. At fifteen seconds Reverse flips the envelope: each burst swells instead of falling.

The delay, plainly

Plucked phrases, dry, then the delay at 30 % mix — Time at 375 milliseconds, Feedback at 35 %, the repeats left to fall away.

Feedback, past the rail

Feedback climbs from 35 to 105 % — past self-oscillation — and is brought back down. The Time control is then jumped mid-ring, 375 to 125 to 500 milliseconds, and the repeats jump with it.

Eight bits in the loop

Every repeat passes through 8 Bits and a 2 kHz Bandwidth, losing resolution and highs as it circles. Mod starts Stepped — sample-jump glitches — and turns Smooth at twelve seconds; Sync locks the Time to the pulse at the end.

A fifth above

Plucked phrases, dry, then a parallel line at Pitch +7 semitones from five seconds, at 50 % Mix — the shifted voice carrying the grain of the machine.

Two octaves, spoken

The restorer's note, shifted in steps: Pitch walks down to −24 semitones, up to +24, then home. Grain then closes to 5 milliseconds — a flutter — and opens to 200, loosening the voice.

Doubler, then the spiral

Pitch 0 with Fine at +12 cents and 30 milliseconds of Delay — the doubler. Then Delay opens to 220, Pitch goes to +12, and Feedback sends each repeat another octave up: the spiral.

The smear, plainly

The pad, dry, then through the machine at Stretch 200 % from five seconds — the same material smeared to twice its internal length, FFT Size at 2048.

Coherence, let go

FFT Size at 256 — the warble end of the machine. Coherence unlocks from 100 to 0 across the take and the phases run free; Stretch then climbs from 250 to 400 %.

A frame, held

Stretch at 400 % with FFT Size 4096 and Overlap at 2×. At ten seconds Freeze holds a single spectral frame and the pad becomes a drone; at twenty-four it is released.

The staircase, plainly

The restorer's note, read plainly, then held to A Minor at Speed 0 from five seconds — every slide in the voice snapped to the nearest scale note, the staircase in full.

Speed, opened

Speed opens from 0 to 500 milliseconds — the staircase softens into a slow glide toward each note — then Amount falls to 0 and the correction lets go of the voice.

Pulled to the wrong key

The reading, held at Speed 0 and remapped as it goes: Scale to Major at eight seconds, the whole grid to A# Minor at fourteen, home to A Minor at nineteen. Formant then tilts the voice −8 and +8 semitones.

The cathedral, plainly

Sparse percussion, dry, then the Cathedral program at 35 % mix. Each hit blooms into the flagship tank and is left to finish.

Early, then late

Early/Late crosses from all reflections to all tail. Diffusion thickens from 35 to 90 %, then Decay (RT60) opens from 1.5 to 30 seconds and Size widens to full.

The tank, clouded

300 milliseconds of Pre-Delay before a 12-second tail. HF Damp at 3 kHz and LF Damp at 80 % close the room in; Mod Depth at 70 % and Mod Rate at 3 Hz set the tail chorusing; Low Cut holds the send at 150 Hz and the delay memory runs 12-bit Grain.

Seven programs, in order

The seven programs, four seconds apiece — Large Hall, Concert, Cathedral, Plate, Chamber, Room, Ambience. Same material, same Decay; only the topology changes.

The ensemble, plainly

Held chords, dry, then the Ensemble at 50 % mix from five seconds — four modulated voices at 0.6 Hz, the single line becoming a section.

Two voices, then eight

Chorus Mode with two Voices, thin and even. Eight Voices arrive at eight seconds, the Ensemble at fourteen, Dimension at twenty-one — the same phrase, thickening.

Warble, then width

Rate at 6 Hz with Depth at 90 % — a fast warble on 12-bit Grain, held in mono — settles to 0.4 Hz and 60 %; Spread then opens the image from mono to full width.

A curve, in and out

A working curve: Low Shelf +3 dB at 180 Hz, Mid 1 dipped 2.5 dB at 300 Hz, Mid 2 lifted 2.5 dB at 2.2 kHz, High Shelf +2.5 dB. It enters at five seconds, steps out at nineteen for reference, and returns.

Character, raised

A +4 dB High Shelf at 16 kHz and a Mid 2 peak at 9 kHz — close to the edge of the band, where this EQ has opinions. Character then steps up, 0 to 35 to 70 to 100 %, and the coefficient quantization pulls the curve into its period shape.

Moved live, unsmoothed

Smoothing at 0: a high-Q Mid 1 boost swept from 200 Hz to 6 kHz steps audibly as it moves — the zipper left in, as the map requires. HP then rises to 250 Hz and releases; LP closes to 3.5 kHz and opens.

The air, plainly

The mix, then the mix with 25 % Harmonics generated above the 3 kHz split — the air switched in at five seconds, out at eighteen for reference, and in again.

Harmonics, from nothing

Harmonics climbs from 0 to 100 % above the 3 kHz split. Frequency then drops the split to 1 kHz — brash, into the mids — and lifts it to 8 kHz, where only the top of the mix is excited.

The bottom, enlarged

Big Bottom at 80 % under 40 % Harmonics. Bass Freq moves the low corner from 100 to 40 to 220 Hz, then Tune sweeps the harmonic character from one end to the other.

The choir, plainly

A line in A Minor, alone, then joined from five seconds by Voice 1 at +3rd and Voice 2 at +5th — the harmony cut to the key, at 50 % Mix.

Three voices, built

The harmony assembled one Voice at a time: +3rd at five seconds, +5th at eleven, -Oct at seventeen. Voice 1 then lifts to +Oct and the chord opens.

The spread, pushed

Detune at 50 cents and Delay at 50 milliseconds smear the two voices around the line. Formant sits at −12 semitones, jumps to +12 at sixteen seconds, and returns to 0.