Manual

Everything the FM-1 does under Tonitor. New here? Start with Playing it; to install, see Install.

Not a groovebox

No tempo, no steps, no patterns, no drum kit.

Tonitor is for sound heard for itself: texture, grain, movement and slow transformation. You play drones, gestures and textures, layer them, loop them freely out of time with one another, and shape them while they play. It suits acousmatic and electroacoustic work as much as live ambient music, dark ambient above all.

Playing it

Generative paths

Each track has a path that leans what HOME draws, on every engine. A path changes nothing already playing: it shapes the next sounds drawn on that track. The path's key is lit while that track is chosen.

Sessions keep each track's path, and the path each of its sounds was drawn on.

Macros

Press EDIT (the black key just left of SHIFT; lit while on) and the four knobs across the top stop being the filter, the granulator, the crush and the send: they become four macros that reshape the chosen track's sound, while it plays, under a held note. Every macro answers early in the turn (a quarter of the way, half the change). Each is turned from the middle, which is the sound as HOME drew it; either way transforms it. While a knob turns, the whole screen is its dial, with the macro's name for this sound and how far it is turned (+40 %, -25 %, AS DRAWN). Press EDIT again to give the knobs back.

The macros are chosen for each engine, and for some by the kind of sound:

Each track keeps the macros of each of its four sounds. HOME draws a new sound as it comes (its macros in the middle). Sessions and presets keep the macros with the sound.

Latched keys

Hold one or more white keys and press SHIFT (the rightmost black key): they latch, sounding on after you let go, and they stay lit. A key pressed while SHIFT is already held plays as usual (unless it is lit). Latches belong to the track they were made on: the keys are lit while that track is chosen, and on another track the same keys play as usual, so a drone can be left on one track while you play another. Pressing a latched key without SHIFT does nothing (it sounds already); SHIFT and the lit key lets go of it, and its note ends as a lifted key's would. Loading a session lets go of every latch. Latched notes hold voices of the shared pool: with all twelve busy, a new note can take one, even a latched one's.

Direction and speed

The 4th to 8th black keys set how the chosen track's loops play, all its overdubs together, each within its own length:

REVERSE, PENDULUM and RANDOM exclude each other, as do HALF and DOUBLE; pressing the lit key again returns to forwards or to normal speed. Direction and speed combine (backwards at half speed, pendulum at double, and so on). Changing the direction while a loop plays continues from where it is: a note sounding at that moment carries on and ends where its start was. An overdub recorded at half or double speed, or while the loops play backwards, is stored so that it plays back as it was played. Sessions keep the settings; copying a track copies them; emptying a track's loops (REC held) returns them to forwards and normal speed.

Copying a track

Hold a track's button until the screen flashes COPIED / TRACK n: the track is on the clipboard, and its button blinks. Choose another track and hold its button until PASTED / TRACK n: it now has the copied track's tone and loops. The clipboard is then empty: the next hold copies again. Holding the copied track's own button again copies it again (as it is now), pasting nothing.

What is copied: the tone — the engine, the sounds of all four engines with their macros, the filter, the granulator, the crush, the reverb send, the octave and the generative path — and all the loops, every overdub, those taken out still out, playing or stopped as they were (pasted, they start from their beginnings). The copy is taken when you hold the button: what the track does afterwards is not pasted. The track pasted onto keeps its level and pan (its place in the mix); whatever it had before is replaced, and its latched notes are let go.

The loopers

Every track has its own looper of up to ten loops, and they are all free: no tempo, no grid, no quantising — not even between a track's own loops. Each overdub is a loop of its own, as long as you made it (from its first note to REC), going round in its own length over the others, so they drift in and out of phase. Each plays back exactly what was played, to 0.73 ms. A note still held when REC closes the loop is tied over its end: it sounds on unbroken as the loop goes round, so a drone stays one long note rather than being struck again each time. The keys go on playing the track over its loops, and the loops' notes and the keys' never cut each other's notes off — but all of them share the pool's twelve voices with the other tracks, so many loops and tracks at once take voices from each other. A track's loops hold about 500 notes between them; nothing is lost when the FM-1 is switched off if the session is saved.

Bringing loops in and out. Hold PLAY (it does not start or stop anything held): the chosen track's loops appear on the white keys, the first loop — the foundation — on the leftmost, each overdub on the next, up to ten, lit while they play; the screen shows them too (in the track's colour: in; a grey frame: out). Press a key to take its loop out — its notes stop, the others go on — or to bring it back in: it comes in at once, from its start. With the loops stopped it chooses which will play: take out all but one, let go of PLAY, press PLAY, and only that one starts; then bring the others in, one by one, as the piece unfolds. Every loop is in when recorded. Sessions keep which loops are out.

Moving the loops along the scale. Hold SHIFT and press OCT+ or OCT−: every loop of every track moves one step of the chosen scale up or down — from C to D in C major, from C to D♯ in C minor pentatonic — so each loop keeps its shape within the scale. The notes sounding move at once; the screen shows how far the loops are moved (TRANSP +2, TRANSP -1, TRANSP 0), up to two octaves of the scale either way. Choosing another scale or root while they are moved walks the new one. A note outside the scale keeps its distance from the scale note below it. The white keys move with them: on a track, they play the scale moved as far as that track's loops are (all the loops' move and its own), so what you add fits what is there — and an overdub recorded while the loops are moved plays back as you played it, the others staying where they are. (USB MIDI notes are not moved.) Sessions keep how far the loops are moved.

One track's loops along the scale. Hold SHIFT, hold a track's button, and press OCT+ or OCT−: only that track's loops move (every overdub with them), a step of the scale at a time; the screen says TRANSP T3 +2. It adds to the move of all the loops, and that track's white keys move with it. Holding the track's button with SHIFT down never copies it. Emptying a track's loops (REC held) takes its own move away; copying a track takes it along; sessions keep it.

The screen

From top to bottom: the track number (very large, in the track's colour), its engine (FM or ANIMA, in white), the family of its sound (in the track's colour). When something changes (NEW, OCT +1, LOOP), a black-on-white message replaces the family for a moment. While the chosen track's looper is armed the band below the engine flashes ARMED in red; while it records it shows REC, white on red. A message too long for the band at its usual size (EXPERIMENTAL) is drawn a size smaller. Under it, across the screen, the track's level meter: a bar filling from the left with the track's colour as it sounds (-48 to +6 dB), and a white mark at the level the SELECT knob sets (red at the left when the track is off). At the very bottom, the track's pan: a block in its colour on a track from left to right, with a tick at the middle.

KNOB 1 to 4 (filter, granulator, crush, reverb send) each show, while they turn and for a second after, the same full-screen dial: the knob's name and the track, an arc from 7 to 5 o'clock lit in the track's colour, a white pointer, and in large type what it does.

The REVERB page shows the title and its four values in rows, each with a bar; the row you just turned is drawn black on white for a moment.

Scales

One scale and root for the whole keyboard (and saved with the session); C major at the start. The root sits on the first white key, in the octave from C3 to B3 (each track's OCT moves it); the scale climbs from there across the 16 white keys, so a pentatonic scale spans more than three octaves.

USB MIDI notes are not quantised.

Presets

Each engine has 64 presets, P1 to P64, for the sounds you want to keep and bring back on any track. Hold SHIFT and press SAVE: the PRESETS page of the chosen track's engine opens, four slots a page. PRESETS turns through them; a slot holding a sound shows its number in red and the sound's name, an empty one EMPTY in grey.

• SAVE held saves the track's sound into the slot chosen. Over a sound already there it asks OVERWRITE? first, on NO until PRESETS turns it to YES; SAVE answers.
• SAVE loads the slot chosen into the track (which crossfades to it, as to a new sound) and goes back to playing.

A preset keeps what draws the sound again, exactly — its seed, its path and the generation of the generators that drew it — and its level: all 256 take one sector of flash, beside the sessions. Saving erases flash, which silences the audio for a moment, as a session save does.

Sessions

The FM-1 always starts with no session loaded. There are 64 slots, numbered 1–64.

A session stores in full the sound of the engine each track is on, and the other three as the seed that made them and their level (the same sounds, drawn again at once), so a session usually takes one 4 KB sector of flash. A session is the whole machine: every track's engine and its four sounds, its path, octave, level, pan, reverb send, filter, granulator (and the character it was drawn with), crush, latched notes and loops (playing or stopped, each overdub in or out); the scale and root, the reverb, and the track chosen. They are kept in the FM-1's flash, in its user data area (saving overwrites any projects, presets or samples another firmware left there): dozens of typical sessions, at least ten even with six full loops. A session also keeps the seeds its sounds were made from, with the path and the generation of the generators that drew them: every generation is kept exactly as it was (a test checks thousands of sounds of each), so sessions and presets saved under an older Tonitor load with the same sounds after an update that brings new generators. Erasing flash silences the audio for a moment: a save uses blank flash when there is some, and the FM-1 blanks free flash by itself while nothing sounds.

The granulator

Each track keeps the last half second of its own sound (at 22 kHz, 8-bit mu-law: a little grit is part of it) and plays grains out of it, after its filter. Each track's granulator has a character of its own, drawn at random when the FM-1 starts and every time a session is loaded: the intervals its grains are pitched to, how often they reverse, stutter, drop out or freeze, the shape of its scattered grains, how far its feedback and darkening go, how it drifts. The knob's position is saved with the session; the character is drawn anew. When the track's notes end, its grains ring on for about two seconds.

The delay

A master delay, mono, placed before the reverb: each track's delay send (SHIFT + KNOB 4) feeds it, its echoes go to both sides of the master, and half of their level also goes into the reverb, so the echoes are reverberated. Its line holds 2 seconds at 22.05 kHz in 8 bits (μ-law, as the granulator's), so the repeats are slightly dark and grainy.

The delay's parameters and every track's send are kept in sessions; copying a track copies its send.

The reverb

A plate after Jon Dattorro (1997): predelay, input diffusion, and a figure-of-eight tank whose two allpasses sweep slowly out of step, so the tail stays in motion and never rings. Each track sends to it after its level; the reverb's return is added to the mix before the master.

Track colours: 1 yellow, 2 cyan, 3 pink, 4 green, 5 orange, 6 lilac.

The engines

FM — six operators, DX7 style

The classic six-operator FM (msfa, the core of Dexed: 32 algorithms, rate/level envelopes, ratio or fixed frequencies, feedback, LFO, pitch envelope). HOME draws a patch from one of twenty-two families of sound objects (since 0.22 with their operators on the ratios they were designed with — before, a slip in the generator put nearly every ratio near an even multiple and the patches two octaves above the key; sounds saved then come back as they were):

Each patch is then shaped with care: modulators on high ratios turned down (their sidebands alias and scream), the brightness scaled down the keyboard so high notes are not shrill nor low ones muddy, rate scaling on the carriers. HOME draws several patches, measures each, and keeps one whose brightness suits the track's path (darker on AMBIENT, brighter on EXPERIMENTAL) and that is not of the family the track had.

Since 0.29 (generation 3) a patch may also be made plucked whatever its family, have a modulator moved to another ratio, or an operator detuned against the rest; the last five families are drawn only from then. About half the patches are then bent once more, gently (an operator detuned to beat, an envelope turned over to grow as it holds, the LFO waiting and then moving the timbre, an octave down, more feedback, a modulator's ratio moved to a neighbour, a partial nailed to a frequency). Every patch is rendered offline as it is made and given the gain that brings it to one loudness.

ANIMA — a lo-fi synthesizer played by a simulated hand

ANIMA is the lo-fi engine: each voice runs at half the sample rate (22.05 kHz, so nothing above 11 kHz), its samples drawn out to the full rate by straight lines between them. Its pitches, rates and filters are scaled to sound where they would at the full rate; what goes is the top octave of air — and half of the chip's time, so more voices sound at once.

A fixed-point port of the anima opcode from Vascular, rebuilt for the FM-1. While a key is held, a simulated finger rests on a virtual surface and drags one figure after another — orbits, figures of eight, scribbles, slow glides, now and then a flick — pressing in slow swells, and now and then lifting to land elsewhere. A ghost finger comes and goes on an irregular clock. Where the finger is, how hard it presses, how fast it moves and how long it has been down move the timbre, never the pitch.

The voice: two wavetable oscillators (24 band-limited single cycles), an FM pair, a stack of copies, a grain cloud, a filtered noise, a resonant filter, a wavetable tremolo; then the hand's chain: a downsampler, a waveshaper, a phaser, a second filter and a comb resonator tuned to the note. Each new sound is a design: which of these are in, on which waveforms, and for every parameter which gesture moves it, over what range, along what curve, plus a mod matrix of three to seven routes and Anima's eight sliders, drawn — often far from their middle. Twenty-two kinds: HAZE, VOYAGE, GLASS, BELLS, GOLDEN (Anima's own) and BREATH, GRIT, SWARM, TIDE (waves of noise over a low drone), CHOIR (vowels, a formant the hand moves), SHARDS (glass grains scattered and crushed), BOWL (a beating inharmonic hum in a ringing comb), DUST (sparse crackle over a dark drone), MACHINE (a pulsing, folded mechanism), and since 0.29 PLUCK (a plucked string: a burst through the comb tuned to the note, dying away), REED (nasal, through a formant the hand moves, a breath in it), FOG (a stack under fog, a slow phaser drifting through it), ORGAN (drawbars, a rotary tremolo, a little drive), BITS (pulses in few bits: the rate crushed and dragged about, folded), DRIFT (two saws drifting against each other, a chorus the hand moves), MURMUR (a dark cloud of noisy grains) and HOWL (noise through a resonance so sharp it whistles, swept by the hand). Every kind draws wider since 0.29 (generation 3): any of the 24 waveforms, wider intervals and ratios, a detuning, now and then a plucked envelope or a slow swell, another way for the hand to move, a pulsing tremolo, a crushed rate or a sharp resonance. The filter follows the keyboard, so low notes are not muffled nor high ones shrill; the second oscillator's interval suits the path (consonant on AMBIENT, anything on EXPERIMENTAL); and an AMBIENT sound nearly always has a resonant space to ring in.

In Vascular a design plays up to six fingers, each a voice; here a track has one voice, so the hand has one finger on the sound.

CALIGULA — an organism

The third engine, after the architecture of SOMA's LYRA-8 as Mike Moreno's LIRA-8 emulates it (BSD licence), rebuilt for the FM-1 and bent towards dark, restless textures. Eight oscillators in four pairs on two sides; each a triangle that its pair's SHARP pushes towards a square; each with a sensor and an envelope, slow (a long swell, a longer fall) or fast; each pair frequency-modulated at audio rate by its partner, the other pair of its side, a pair of the other side or the total feedback (the whole organism, saturated, fed back in). A hyper LFO of two LFOs (alone, AND, OR, XOR) opens and closes the modulation and gates chosen pairs into irregular rhythms; a short modulated delay (two lines, ~35 ms) choruses, flanges or warps it; a drive.

A CALIGULA track is played like the others: the white keys play the scale's notes, and it is silent until a key is down. A key wakes the organism at that note; a second key on top moves the whole organism to the new note at once (a snap, not a glide), and letting go of it goes back to the key still held; letting go of the last key lifts the hand and the organism rings out. It is rooted: tuned around the note played, in intervals of it (octaves, fifths, unisons and semitone clusters that beat, tritones, the depths). The hyper LFO moves timbre, never pitch: no sirens.

Nobody touches the eight plates: a simulated hand does, while a key is held. Each species has its own hand: how often it touches, how long a finger stays, how many fingers at most, its favourite plates, quick taps against long holds. The first finger stays down until the key is let go, so the sound never drops out under a held key. Around the hand the organism has a life of its own: the modulation depths and the sharpness wander, oscillators fire in bursts at irregular times, and now and then the feedback and the drive surge for a moment. A pair's envelope may be slow: a swell of seconds, a release of up to nine seconds after the key is let go.

The hand often has a phrase: two to four plates it keeps coming back to, varied now and then, often on a pulse of its own (long, short, short, longer) — so it plays like someone rather than like chance. A pair the hyper LFO's gates never close stays under it all. HOME draws a new organism from ten species, tuned on one of ten sets of intervals (pillars of octaves and fifths, semitone clusters, the depths, the harmonics, unisons, tritones, fourths, a minor seventh, open fifths, a quarter-tone cluster):

RUSSOLO — noise, logic and broken buffers

The fourth engine re-makes the techniques of Robert Heel's Noisferatu (GPL-3.0) in fixed point, for the keys. It runs at a third of the sample rate (about 14.7 kHz) in 10-bit values, each held until the next — the grit of a small DAC — then reduces its bits and holds its samples further as the sound says.

A RUSSOLO track is played like the others and is silent until a key is down: the key opens an envelope and sets the pitch (the oscillators' frequency, the speed a buffer is read at, the rate noise is sampled at); a second key on top moves it there at once; letting go of the last key lets it fade. Like CALIGULA it is one organism a track, outside the pool of voices. HOME draws one of seventeen kinds:

Each kind has two parameters of its own, the first two macros; CRUSH and CHAOS are the other two. Its level is measured when it is drawn, and never raised so far that its loudest click could clip.