chassis

a semi-modular synthesis system

chassis is a semi-modular synthesis system by Lux Cache. It has 34 modules for making and shaping sound.

The modules were designed together, as one set. Each module has several modes, and more than one way to use it.

The set of modules is fixed, so you can learn all of it. If you are new to modular synthesis, or used to hardware, expect it to take some time.

chassis and chassisFX

chassis comes as 2 plugins that share the same 34 modules.

chassis	an instrument. You play it from a keyboard or from its own sequencer, and it makes the sound
chassisFX	an effect. You insert it on a track, and the track’s audio goes through the modules

how chassisFX handles sound sources

When chassisFX loads, its sound sources stand down. A module that stands down has a white wash over its panel. Its outputs carry the track’s audio instead of its own sound.

The track’s audio then travels the same cables, filters, delays and mixer channels as the synthesised sound would. This is why a chassis preset also works as an effect, just as it is.

To wake a module that has stood down, press cmd E, use its right-click menu, or patch one of its outputs. It then plays its own sound again, under the track.

how the demo works

Until you sign in, chassis runs as a demo. The header shows this.

The demo includes every module and every page. It allows 20 minutes of audio on this computer. Presets by guest artists open with the full version, and their rows are grey in the demo.

The timer counts only the audio that chassis plays. Each time you launch chassis, it asks before it starts counting.

When the 20 minutes are used, the sound stops, and sortie and coupe record silence. Your patch stays, and you can still save it. Each computer gets its 20 minutes once.

sign in to unlock chassis

1. Choose log in. chassis opens luxcache.com in your browser, with a short code.
2. Approve the code on the website.
3. chassis installs your licence. From then on, it checks the licence on this computer, offline.

One licence covers 3 computers at a time. To free this computer, use sign out on the settings page.

You can open the sign-in card at any time from unlock in the footer, or from the account section of the settings page.

set up sound and midi

The standalone app plays through this computer’s sound hardware. You choose that hardware at the top of the settings page, in the audio and midi section.

1. Open the settings page. On a Mac, you can also choose Audio and MIDI Settings in the chassis menu, at the top left of the screen.
2. Click a row, such as output, to see its list.
3. Choose from the list. chassis changes to it at once.
4. Click test tone to check that you can hear the output.

output	the speakers, headphones or interface that chassis plays through
output channels	the pair of channels that chassis plays through. This row appears when the output has more than 2 channels
input	the microphone or interface that chassis takes sound from. Choose none when you only play chassis
input channels	the pair of channels that chassis listens to. This row appears when the input has more than 2 channels
sample rate	lists the rates that the output can run at
buffer size	a smaller buffer answers your playing sooner, but works the processor harder. If the sound crackles, choose a larger one
mute the input	stops a microphone feeding back through the speakers. Turn it off when you want sound to come in
midi inputs	one switch for each keyboard or controller. Turn a switch on, and its notes reach clavier
test tone	plays a short tone through the output

chassis remembers these choices for the next time you open it.

As a plugin, chassis uses your DAW’s sound hardware. You choose it in the DAW’s own settings.

how modular differs from a DAW

In a DAW, the parts of an instrument are already connected. The oscillator feeds the filter, the filter feeds the amplifier, and the knobs set how much of each.

In a modular system, you make every connection yourself. You connect the parts with cables, and the cables decide what the instrument is. Two people with the same system can end up with very different instruments.

The rest of this chapter explains the words you will see. Most are familiar ideas with older names.

signals

Every cable carries a voltage. What the voltage does depends on where the cable goes.

cv	control voltage. A signal that controls something, like an automation lane in a DAW. Any module can make it or read it, and it can move as fast as audio
gate	a held note. 10 volts while a key is down, 0 volts when it is released. Like a MIDI note on and note off, in one cable
trigger	a very short gate. It marks a moment, such as a drum hit, a step or a re-roll
v/oct	pitch as a voltage. 1 volt is 1 octave, so 1 semitone is a twelfth of a volt. Like a MIDI note number, but continuous
audio	the sound itself. It swings above and below 0 volts
bipolar	a signal that goes above and below zero, like an lfo
unipolar	a signal that only goes up from zero, like an envelope

types of module

oscillator, vco	makes a tone. Like the oscillator inside a DAW synth, with its connections exposed
filter, vcf	removes part of the sound. Resonance boosts the frequencies at the edge of the cut
vca	controls volume. Every DAW instrument has one inside. Here it is a module, and it turns a constant tone into separate notes
envelope, adsr	attack, decay, sustain and release. Here an envelope is a voltage, so it can control anything, volume included
lfo	a low frequency oscillator: a slow, repeating movement you can send anywhere
sequencer	plays notes one step at a time, like a piano roll or an arrangement
quantiser	moves incoming pitches onto the notes of a key
sample and hold	takes a reading when it is triggered, and keeps it until the next trigger. The usual source of stepped random values
attenuator	sets how much of a signal gets through, like a modulation amount control. An attenuverter can also turn the signal upside down

words on the panels

hp	the width of a module. 1 hp is a fifth of an inch
patch	the whole state of the system: every control and every cable. A DAW would call this a preset
normalled	a connection already made inside the module. Patching a cable into the jack replaces it
mult	one signal sent to several places. In chassis, drag another cable from the same output
sum	several signals added together at one input. chassis does this for you
poly	one cable carrying up to 16 voices, so a chord needs only one cable

your first settings

The first time chassis opens, a welcome card asks for your sample folders and a recordings folder, and offers a few first settings. To open it again, use the welcome card row on the settings page.

any output can connect to any input

Any output can connect to any input, and the instrument is whatever you patch.

Any order works. You can connect an oscillator to another oscillator’s pitch, just as you can connect it to the output.

a cable carries a voltage

A cable carries a changing voltage. The input it connects to decides what the voltage does.

For example, take one slow wobble and connect it to different places:

filter cutoff	the filter sweeps
oscillator pitch	you hear vibrato
the output	made fast enough to hear, the wobble is a tone

The signal stays the same. Only its destination changes.

audio and control are the same

In chassis, audio and control use the same jacks, cables and voltages.

If you use audio as control, you get audio-rate modulation, such as ring modulation and metallic tones. If you use control as audio, you hear clicks and thumps.

The conventions in the next section are habits. Breaking them on purpose is a common technique.

standard voltages

chassis follows the voltage conventions of modular synthesis, so the numbers here mean the same as they do elsewhere.

audio	swings around 0 volts. About 5 volts each way is a healthy level
pitch	1 volt per octave. 1 semitone is a twelfth of a volt
gate	0 volts is off and 10 volts is on. A gate counts as on above 1 volt, and as off below 0.1 volts
trigger	a gate that switches off straight away. It marks a moment
cv	any other control signal, either around zero or above it

jack kinds and their colours

Every jack has a kind, based on its name. When you drag a cable, the patching assists colour the cable and the jacks that suit it:

audio	teal
control	blue
pitch	violet
gate	amber
clock	pink

The colours are a guide. Any jack connects to any other, whatever its colour.

3 rows, 9 bands

The 34 modules sit in 3 rows. Each row is divided into bands. A band is a group of neighbouring modules that do one kind of job. There are 9 bands, and this chapter describes each one.

A patch usually starts on the bottom row and ends on the top row, so signals tend to travel upwards. You can patch in any direction.

sequencing and input

This band plays notes. It decides when notes play, and which notes. The voices on the middle row make the sound.

xtrack	an 8-track tracker. Each cell has a note, 2 cvs and ratchets. xtrack leads the system’s transport
xarrange	stores 16 scenes of xtrack, and switches the whole pattern on the bar
clavier	poly keys, a phrase player, an arpeggiator, and a microphone that feeds straight into the patch

modulation

This band moves controls for you, so the patch changes over time.

qmod, qmod+	16 slots each, 2 slots to a rate. Each slot is an lfo, sampled random, a smooth drift or a trigger
qmap	takes a cv and applies it to a knob it has learnt, so a cv can move any knob directly
volet	an 8-channel trance gate. Each channel is a wheel of steps with its own shape, depth and rotation

synth voices and samplers

This band makes sound. It is the widest band in the system.

corde	throws a packet across a field and reads it back as a wavetable, so the tone changes by itself
ton	4 oscillators playing the same note. You choose the wave by choosing which output jack to use
scie	a polyphonic supersaw
frappe	strikes an object instead of playing a sample. Choose the tiling, the size and the mallet, and it rings the way that shape would
coupe	an 8-head chain sampler. Each head has its own region, speed and direction, and hands playback to the next

effects

This band changes sound. The modules are roughly in the order you would chain them.

sillage	follows how loud a sound is
presse	a 3-band compressor that works in both directions: it raises quiet sounds as it lowers loud ones
crible	a filter bank. It works as a vocoder, or as a filter that keeps the notes of a scale
pli	an overdrive that clips the 2 ears differently, with a noise source inside
biais	shifts every frequency by the same number of hertz. The pitch stays, but the tuning changes
essaim	moves a group of resonances together
nef	a reverb, at the end of the row

the sample picker

bac is a file browser. You can listen to a file before you use it. Drag a file from bac into coupe, or onto your desktop.

bac works with the files on your disk.

percussion

fracas has 4 fm percussion voices, with 2 operators each. Its router sends any of its 4 cvs to any setting inside the voices, so each hit can sound different.

frappe, a similar module, is on the middle row with the other voices.

more effects

trame	16 separate spectral effects on one transform. You choose one at a time
foudre	a dual filter with the drive inside the feedback loop. Close the loop, and the filter becomes an oscillator
vanne	8 stereo rows, each with an input, a mode, a cv and an output, and a sum output that mixes them. The mode sets what each row does to its signal
ruban	a bucket-brigade echo. One control sets both the delay time and the tone

utilities

quatre	4 macros. Each is a playhead moving over 4 painted lanes
atelier	11 operators, each with 2 inputs and 1 output. Each cell does logic, routing or sample and hold
gamme	moves 8 lanes onto one scale, and shares the key with the whole system
grille	a drum computer. It reads 4 lanes from a map. As you move across the map, the pattern changes
liasse	selects modules anywhere in the system, so you can control them as a group

the output

tresse	an 8-channel stereo mixer. The channels stand in a room around a listener: distance sets the level, and moving the listener changes the whole mix
brin	a channel strip for one channel at a time, before the fader
envoi	one stereo send and return, shared by all 8 channels
sortie	the master output and the recorder. The mix leaves here for your speakers, your DAW and your disk

patch in any direction

The bands show where modules sit. Signals can go anywhere across them.

For example, a reverb can control an oscillator’s pitch. The mixer’s output can go back into a filter. A follower listening to the mix can control the sequencer that made it.

Learn the system from bottom to top. Then patch across it, down it and back into itself.

one cable carries left and right

Every audio cable in chassis is stereo. Each module’s left and right jacks are joined into one jack, and one cable carries both channels. A mono source in that jack feeds both sides.

Some jacks are mono on purpose, such as one of ton’s 4 waves, one of fracas’s 4 voices, or a volet channel. The panel labels these jacks.

practise on the bench

In the learn tab, this chapter shows a bench of 12 silent jacks. Practise patching here, and your own patch stays as it is. The bench clears each time you open the learn tab.

connect two jacks

1. Press on a jack and drag. A cable follows the pointer.
2. Release the cable on another jack to connect it.

If you release the cable anywhere else, it is removed. You can start from either end.

connect one output to several inputs

Every jack in chassis can hold several cables, so one output can feed many inputs.

An output can have any number of cables. Each cable carries the full signal.

An input can also take several cables. It adds their signals together.

Each pair of jacks takes one cable. This stops a signal arriving at double level by accident.

move, unplug or add a cable

To move or unplug a cable, drag its end away from the jack. Release it on another jack to move it, or anywhere else to unplug it.

To add a new cable from a jack that already has one, hold cmd and drag from the jack. This works on inputs and outputs.

patch by typing

Point at a jack and press q. A box opens on the jack. Type a destination, a module or a jack, and the list narrows. Enter patches the cable.

The box on the bench is live. It is open on the bench’s cv in. Type in it and move through it with the keys. It stays open while you read this step.

1. Point at a jack and press q.
2. Type part of a name. qmod lists all of qmod’s jacks. qmod 3 is one jack.
3. Press the up and down arrows to choose a row, and enter to patch it.

With nothing typed, the list shows the cables already on the jack, then the best fits, nearest first. Jacks of other kinds are listed too, in grey.

You can also type what a module does. keys finds clavier, delay finds ruban and mixer finds tresse. A jack answers to its other names too: pitch finds a v/oct jack, and trig finds a trigger. So keys pitch is clavier’s v/oct, and mixer 1 is tresse’s channel 1.

After enter, the same box shows the cable’s controls: attenuator, offset, slew, slew shape and colour. Tab moves between them. Type a value, or press the left and right arrows to nudge it. On the colour row, 1 to 5 picks a colour. Enter closes the box.

q, with a cable in hand	drops that cable and opens the box on the jack it came from, so the other end is typed
cmd enter	patches and closes at once
cmd left	back to the list, to add another cable to the same jack
cmd right	follows the cable, and routes from its other end
cmd backspace	disconnects the cable. Undo brings it back
esc	closes the box

signals into one input are added

When 2 or more cables go into one input, their signals are added together. Every input works like a small mixer.

Adding signals can make a signal too large. For example, 2 full-level sources into one input give twice the full level. Modules handle this in different ways: some clip, some fold, and some pass it on unchanged.

highlighting and magnetism

While you drag a cable, the jacks that suit it are ringed in their kind’s colour, and the other modules are dimmed. The pointer is pulled gently towards good targets.

The cable shows where it would connect. When you release, it connects exactly there.

Cables swing briefly when you plug, unplug or move them. You can adjust or turn off all of this in settings.

see a signal with the scope

Point at a jack to see its signal in a scope. The scope opens in a corner of the window, and shows up to 2 seconds of signal, so you can see the whole shape of an envelope or a slow lfo.

In settings, you can change the scope’s corner, size, time window and opacity, or invert it to a dark panel.

cables show their signal

Cables move with the voltage they carry. A cable with audio shivers. A cable with a slow envelope sags and swells. A cable with no signal hangs still.

If a patch is silent, look for a cable that hangs still.

shortcuts and undo

Every action is one undo step: a cable, a knob, a menu action, a dice roll or a preset load. cmd I, cmd R, cmd U and cmd E act on the module under the pointer.

cmd Z	undo. Add shift to redo
cmd I	initialise
cmd R	randomise
cmd U	disconnect every cable
cmd E	bypass or wake. Patching an audio output of a bypassed sound source also wakes it, and one undo removes the cable and the wake together
cmd shift D	the console: a live listing of the patch, for bug reports. c copies it, s saves it, t tests a save and reload, r resets the counters, esc closes it
cmd C, cmd V	copy a module’s settings, and paste them onto another module of the same kind
q	over a jack: patch by typing. The patch by typing step above has the keys
cmd 1 to cmd 8	the output under the pointer, or the cable in your hand, straight into that channel of tresse

cable colours

A new cable takes its colour from the jack you drag it from. These are the same colours the presets use.

audio	red
modulation	green
gate and clock	blue
pitch	yellow

While you are holding a cable, you can recolour it or remove it with these keys:

1	red
2	blue
3	green
4	yellow
5	grey
backspace	remove the cable

These keys work while you hold a cable, with cmd and option up. To make every new cable grey, turn off colour by what it carries in settings.

shape any cable from its jack

A cable can attenuate, offset and slew the signal it carries. Right-click either jack the cable is plugged into. The menu shows the cable’s attenuator, offset, slew and slew shape, and a reset cable mods item to put it back.

Use it where the source gives too much. Pull the attenuator down on the cable into a filter, and the same lfo can still drive something else at full depth. A cable can invert a signal, lift a unipolar signal to sit around 0 V, or round a gate into an envelope.

The right click menus chapter lists the 4 controls.

information every module shares

Some information is useful to every module, such as the tempo. Running cables for all of it would clutter the patch. Instead, chassis sends it on 4 buses. A bus is an invisible connection that any module can use.

A cable always takes priority. If you patch a cable into a jack that a bus was feeding, chassis uses the cable instead, straight away.

A module that follows a bus shows a chain mark next to what the bus controls. Click the mark to unlink the module. To link it again, use its right-click menu.

the 4 buses

transport	the tempo, whether the system is running, and the number of 16th notes since the last reset
scale	the key: a root note and the notes in the scale
notes	the notes being played, in 8 lanes of 16 voices
room	audio from outside chassis. In a plugin, this is the track’s audio

tempo, run and reset

The sequencer sends the tempo and whether it is running. Any module with a clock input follows it while nothing is patched into that input. This keeps drums, gates, samplers and tempo-synced delays in time by themselves.

Start and stop are shared too. Press play on one module, and all of them start.

In a DAW, the DAW’s tempo and play button drive the bus. Turn this off with follow DAW bpm and follow DAW transport, on the browse page.

the key

The quantiser sends a key, and every module that quantises follows it. Change the key, and melodies, arpeggios, tuned percussion and sampler notes all change with it.

The bus sends the notes themselves, so any set of notes works.

played notes

Played notes travel in 8 lanes, with 16 voices each, carrying pitch, velocity and gate. A synth voice can play the notes from a lane directly.

A cable into the voice takes priority over the bus.

audio from outside

The room bus carries audio from outside chassis. In a plugin, this is the audio on the track. clavier’s audio input reads it, and in chassisFX the sound sources that stand down play it.

play, tempo and key

The right of the footer shows play, the bpm and the key. Click play to start or stop, drag the bpm, and click the root or the scale to choose. A * after a word means a module, a cable or the DAW also sets it. Point at the word, and the footer line says which.

what each row shows

The browse tab lists every preset chassis can find, grouped by author, with Lux Cache first. Each row shows the preset’s name, its author, and a small map of the patch.

The map has one mark for each connected module, joined by its cables in their colours. It shows how large a patch is before you load it.

load a preset

Click a row to load the preset. The system changes straight away. To go back, use undo.

Loading a preset clears any sound left from the previous patch, such as reverb tails and echoes, so the new preset starts from silence. This also happens after a dice roll, and when a DAW session opens.

filter the list

all presets	every preset, in one list
sounds	presets you play from the keys. Their names end in (keys)
patches	whole systems with their own sequences. Their names end in (seq)
favourites	the presets you have starred
a folder	a folder you have made in the preset folder, and what is in it
an author	every preset credited to one person

The folders in the sidebar are folders on your disk. If you create a folder in the Finder, it appears in the sidebar. Click rescan to find files you added while chassis was open.

drag a preset into a folder

To organise your presets, drag them onto the folders in the sidebar.

1. Press a preset’s row and drag it. Its name follows the pointer.
2. Move over a folder in the sidebar. The folder is outlined, and the footer says where the preset will go.
3. Let go. The preset’s file moves into that folder.

a folder	the preset moves into it
sounds or patches	the preset moves to the top of that shelf, out of any folder
a shipped preset	it is copied, and the copy is yours. The shipped one stays where it is
a name already in the folder	the preset stays where it is, and the footer says why

A drag moves the preset. A click loads it: press the row and let go in the same place.

To make a new folder, save a preset into it from the save card, or create the folder in the Finder and click rescan.

step through presets

The strip at the top right of every page shows the current preset. Its arrows step through the presets the browser shows, one at a time.

Click the preset’s name to open the browser. A star next to the name means the patch has changed since you loaded it.

The shuffle button next to the arrows is the dice. It creates a new patch. See the randomisation chapter.

the row menu

Right-click a row to edit, favourite, duplicate or delete the preset, to use it as init, or to reveal in finder. edit… opens the preset’s name, tags, author and description.

use as init makes the init row load this preset. You then return to a patch you chose, instead of an empty system.

the patch name and the shelf

Right-click the patch name in the header to save, revert, rename or start a new patch, or to choose your init patch.

The bottom row of the system stays under the list, so you can play while you browse. To hide it, turn off shelf on browse in settings.

what a preset contains

A preset stores the whole system: every control, each module’s own data such as patterns, samples, phrases and painted lanes, and every cable with its colour.

A preset is a text file. It loads correctly every time, because every module keeps its place in the system.

save a preset

1. Click save in the header strip. The save card opens.
2. Type a name.
3. Choose a shelf: sounds or patches.
4. If you want, type a folder, tags, an author and a description.
5. Press enter to save.

Press escape to cancel. Press tab to move between fields.

names, folders and tags

The name is also the file name. If you rename a preset later, the file is renamed too.

Type a folder as a path, for example drums or drums/808. chassis creates the folder when you save. Existing folders appear under the field: click one to use it.

Tags are words separated by commas. Existing tags appear under the field: click a tag to add it or remove it.

The line above the buttons shows where the file will be saved, and whether chassis will create a new folder.

save over a preset

The card opens with the loaded preset’s name, shelf, folder and tags. To save your changes over it, press enter.

If a preset with the same name already exists in that place, chassis asks before replacing it.

The author field remembers the last author on this computer, so you only type your name once.

samples inside a preset

coupe and bac store where a sample is on your disk. Samples from the Lux Cache factory sets load on any computer that has the set. A sample of your own, or a recording made in coupe, exists only on your computer.

The save card has a tick box: embed used samples. When it is on, chassis puts a copy of your own samples inside the preset file. The line beside the box shows how many files this is and how large they are. You can then share the one file and it plays on another computer.

chassis embeds the samples coupe holds. It also embeds the sample selected in bac when bac has a cable. Samples from the factory sets stay in the sets, and load from there on each computer.

When you load the preset, chassis uses your original file if it is still there. If it is not, chassis copies the embedded sample to a preset samples folder in the application support folder, inside a folder with the preset’s name, and plays it from there. In bac, that folder appears as a new root.

A preset with samples inside is larger. Turn the box off if you only use the preset on this computer. chassis remembers your choice.

descriptions and credits

Type a description on the save card. To change it later, right-click the preset’s row in the browser and choose edit…. You can change the author and tags there too.

All of this is stored inside the file, so the credits stay with the preset when it is copied. This is what lets the browser filter by author and tag.

where presets are stored

Presets are .lxcp files in the application support folder, inside sounds or patches, and inside any folders you make. The settings page can open this folder for you.

chassis also comes with a library of presets, which stay as they are shipped. To change one, save it under the same name. chassis saves your own copy, and it appears in place of the shipped one.

To move a preset to another computer, copy its file. To organise presets, make folders.

choose the patch chassis opens with

The first row in the browser is init. By default, it resets every module and removes every cable, which gives you an empty system.

A reset returns each module to its init settings. If you have used save as init on a module, those are the settings it returns to. See the right click menus chapter.

1. Save a patch you would like to start from.
2. Right-click its row in the browser, and choose use as init.

chassis now opens with this patch, and the init row loads it. If you save over this preset, chassis opens with the new version.

the standalone app	opens where you left off: the patch that was open when you last quit. It also saves that once a minute, so a crash loses little. To start clean every time, set the app opens on to the init patch, in the system section of the settings page
a plugin	loads the patch saved in your DAW project. A new instance with nothing saved gets the init patch
no init chosen	the empty system

what the dice does

The dice creates a new patch. It rewires the cables and changes the settings of the whole system.

Every result makes sound. The dice always avoids a silent patch, whatever else it changes.

dice modes

full random	changes every setting on every module, within safe limits
spine	drums, a drone, effects for colour, and a quantised melody on the beat
synth voice	a playable instrument for the keys
fx web	keeps your sound sources, and changes only the effects and their routing
club anthem	four to the floor: a big kick, a bass that pumps under it, offbeat hats, a clap on 2 and 4, and stabs in key
sampler explorations	built around coupe: a sample from your library, or a bar of the system recorded into it, chopped and sequenced in key
mutate	small changes to the knobs already in use. Cables, pitches and switches stay as they are
chord garden	ton plays a chord from gamme’s lanes, over drones
percussion factory	every drum lane plays, with the toms tuned to the scale
feedback bloom	a riff feeds a tape loop that feeds back into itself
fm chaos	modules modulate each other at audio rate
vast minimal	a few hits in a large room, and a chord that swells

keep parts of your patch

Use the settings page to keep parts of your patch when you roll the dice.

keep the key	keeps the quantiser’s root and scale
keep the tempo	keeps the sequencer’s bpm. In a DAW, the DAW’s tempo is always kept
keep patterns	keeps sequences, samples and phrases, and changes only the sound
density	makes results sparse, dense, or the mode’s usual amount
hold these still	keeps the settings of the modules you name

A module you hold keeps its settings, and the dice can still give it new cables. For example, if you hold the sampler, it keeps your sample, and the dice can still use it.

randomise one module

Point at a module and press cmd R to randomise only that module. You can undo this.

To leave a module out of every dice roll, choose exclude from randomisation in its right-click menu. This is the same list as on the settings page.

three kinds of menu

Right-click to see more options. The menu depends on where you click: the panel, a jack or a control.

the module menu

Right-click an empty part of a module’s panel to open the module menu.

the name	the module’s name, at the top of the menu
learn this module	opens the manual at this module’s chapter
initialise	cmd I. Returns the module to its init settings. These are the factory settings, unless you have saved your own
save as init	saves the module’s current settings as its init settings. From then on, initialise returns to these. This applies to every module of the same kind, and the init preset uses them too
forget saved init	only shown when you have saved init settings. Removes them, so initialise returns to the factory settings
randomise	cmd R. Randomises this module only
disconnect cables	cmd U. Removes every cable from this module
bypass	cmd E. Takes the module out of the signal path, and passes its input straight through. A white sheet covers the module, with a line saying how to wake it. To hide that line and keep the sheet, turn off bypass message in the appearance section of the settings page
exclude from randomisation	the dice leaves this module as it is
copy settings, paste settings	cmd C and cmd V, between modules of the same kind

Below these items, each module adds its own settings, beyond the controls on its panel.

the jack menu

Right-click a jack to see its name and disconnect it. On an output, this removes every cable from the jack. On an input, it removes the cable arriving at it.

An output with 2 or more cables also lists each cable on its own, as disconnect and the jack the cable lands on. Choose one and only that cable leaves.

A jack with a cable in it also carries that cable’s own controls: 4 sliders and a reset. They are the same set qmod offers on its jacks, here on every cable in the system.

attenuator	-2 to 2. 1 is the wire as it was. A negative value inverts the signal
offset	-10 to 10 V, added after the attenuator
slew	0 to 1. 0 is instant. 1 is about a third of a second each way
slew shape	where the slew acts. Fully left, all of it on the fall. Fully right, all on the rise
reset cable mods	back to the plain cable

On an output with several cables, each cable’s entry opens onto its own 4 sliders. The shaping travels with the cable: it is saved in the preset, each drag is one undo step, and the console shows it beside the cable.

the control menu

Right-clicking a knob or a field names the control and reads its value back. The menu then offers:

initialise	back to the control’s default
type a value	click the row, type, and press enter. See below
the options	a control with a fixed set of choices lists them all, ticked at the current one. Choose one to set it
unmap from qmap	only on a control a mapper has claimed. Releases the mapping from the control itself

a control that steps through choices

Some controls hold 1 of a fixed set of choices: a wave shape, a mode, a direction, a routing. They all work the same way.

click	moves to the next choice. After the last choice, it goes back to the first
drag	where the control allows it, scrubs through the choices both ways
right-click	lists every choice, ticked at the current one. Choose one to set it

Use the list to see every choice, or to reach one that is several clicks away.

type a value in your own units

The type a value row reads what you type the way you meant it, and converts it to what the control measures.

a plain number	the units the readout shows. A control reading 52% takes 30 as 30%. A control reading 672 Hz takes 220 as 220 Hz
440 hz, 1.2 khz	a frequency. A control in kilohertz takes hertz, and the other way round
c3, a#4, bb2	a note name. A frequency control goes to that pitch. A v/oct control goes to that note, with c4 at 0 V. A semitone control counts from c4
250 ms, 1.5 s	a time, in either unit
50%	a share of the control’s range
-6 db	a gain control takes decibels
7 st, 1 oct	on a v/oct control, semitones and octaves become volts

What the hover tip prints is what the field takes back. Every control in the system follows this rule, whatever law sits behind its readout.

Press esc to leave the row as it was.

extra items on some controls

Some controls have their own menu items. xtrack uses these most: a pattern knob offers copy to all tracks, a note knob offers copy to all notes, and the scale knob has short root and scale lists instead of one long menu.

where to change the theme

The theme is light, dark or grey. It applies to the whole app.

Change the theme on the settings page, or press the section key (§) to cycle through the themes from anywhere.

when an update is ready

chassis looks for a newer version once a day. It asks luxcache.com for the version number only.

When one is ready, a dot appears beside the settings tab. The first time you open the settings page, a card tells you the version. Choose get the update to open your downloads page, or later to close the card. The card shows once for each version. The dot stays until you install the update, and the about section keeps an update row that opens the same page.

tell one module from the next

Modules sit side by side, edge to edge. To see where one ends and the next begins, turn on module borders in the appearance section of the settings page. It draws a thin grey line round every module.

a simulation read as a wavetable

corde reads its wave from a running simulation. It runs a small physical model of energy moving through a landscape. The model moves forward 240 times a second.

corde reads the state of the model as a wavetable, and ordinary oscillators play that table. The movement is already in the simulation. Each control steers it.

The panel follows the same order. One large view runs along the top, with the voice row under it. Below are 5 tiles: landscape, energy, motion, binaural and probability. motion is the largest tile, with an XY pad across it.

2 fields make the stereo width

2 simulations, called fields, run side by side. Each moves forward on its own, and they start from very slightly different seeds.

The left output reads one field, and the right output reads the other. So the stereo width is 2 real simulations drifting apart.

This also makes polyphony cheap. 16 voices read the tables 16 times, but the physics is still the same 2 fields. Each voice has its own phase, envelope, filter and drive.

All voices read the same 2 tables. Each table is band-limited for the highest note playing.

what the view shows

The view along the top shows the field the oscillator is playing. It draws 4 things at once:

the terrain	the faintest layer: the landscape the energy moves through
the two fields	one teal and one magenta. The filled body is the magnitude, and the line through it is the real part
the tables	dark lines: the 2 tables the oscillators read, laid over their span
the span marks	2 blue vertical lines. Everything outside them is veiled

The terrain is drawn against a fixed ceiling. So a well grows as relief rises, and flattens at 0.

Where the 2 fields overlap, the output is mono. Where they separate, the gap is the stereo image.

the span marks are measured

The span is the stretch of the field the oscillator reads. corde measures it.

A flat stretch of field would be silence inside every cycle. So corde finds where the field rises above a small fraction of its peak. It pads that area, keeps it above a minimum width, and moves it a few per cent at a time.

The span moves in whole field points. A jump of even one point would cause a step in the waveform, 240 times a second, which is why it moves gradually.

When the marks close in, the sound is more compact. When they move apart, the table is taking in parts of the field it was ignoring.

relief sets the depth of the terrain

The first tile’s header cycles through 6 terrain shapes. The terrain scatters the energy into harmonics.

relief sets how deeply the chosen shape is cut. It runs from nothing to the deepest shape the field can hold, just short of the wall that contains it.

At relief 0, every landscape is flat, whatever the header says. The swell on the display is then the sound.

With a sidechain patched, relief sets how hard the incoming signal deforms the ground.

warp leans and ripples the terrain

warp is bipolar, with a dead centre. It brightens the sound in either direction, and does 2 things at once.

First, it leans the terrain. The middle stays fixed, and the 2 sides scale against each other: one gets steeper as the other stretches.

Second, it adds a ripple of a few cycles across the field, which grows with the lean.

size sets the packet width

The second tile’s header chooses one of 3 packet modes: how many packets there are, and where they land.

The mode only takes effect at a re-seed, because that is when packets are added. So the tile shows the next throw. The field changes to match at the next re-seed.

throw sets the starting push

throw sets how hard the packet is pushed when it is added. corde reads it only at that moment. At dead centre, the packet starts still.

the XY pad sets speed and depth

The large square is a map with 2 axes.

horizontal	speed: the rate of the field’s own clock. It is squared, so slow movement takes the bottom half of the travel
vertical	depth: how much of the motion mode’s shape is applied to the clock. At 0 the clock is flat and steady. At the top, the mode applies in full

The tile’s header chooses 1 of 6 motion shapes, and each shape moves with the clock.

re-seed and dir

re-seed sits below the pad, as text you drag. It is off by default, at the bottom of its travel.

Above off, it sets a period: from 30 seconds at the slow end, to about a sixtieth of a second at the fast end. At the slow end, you get a drone that rewrites itself over minutes. At the fast end, re-seed changes the tone.

dir sets the direction of the clock.

rev reverses the field’s history

rev un-evolves the field.

The simulation is built from phase rotations, and a rotation one way cancels a rotation the other way. So a negative clock exactly reverses a positive one. The field walks back down the path it came. This is why png returns the field to where it was.

couple sets how closely the fields agree

The binaural tile has 5 modes. The mode decides what the 2 fields agree about, and couple decides how firmly.

In 2 of the modes, raising couple narrows the image. In the other 3, it widens it.

spread does 3 things

spread changes the right field in 3 ways at once:

its clock	runs a little faster than the left field’s
its seed	moves further from the left field’s seed
its terrain	becomes its own, leaned further and cut deeper than the left’s

smooth, pinch and the readout

The field is made of complex numbers, but a wavetable needs real numbers. The tile’s header chooses 1 of 3 readouts, which convert the field into a table.

smooth and pinch both remove detail. Their effect is measured on the default landscape.

pitch has 3 editors

frequency, note and octave are 3 editors for one value, so they always agree. They differ only in what a drag snaps to.

the filter is off by default

The filter has a type, and it starts at off. While it is off, cutoff and res are dimmed.

Cycle the type to low-pass, band-pass or high-pass, and the filter turns on. Each voice on each channel gets its own filter.

clip adds brightness

clip is drive, and it is the brightest single control on corde. It runs from 0% to 200%, and starts at 100%.

Measured, it takes the table from about 45 harmonics to 109.

The band-limiter before it reads the clip setting and leaves room for the harmonics the drive will add. clip is also eased in over the first half of its travel, so leaving 0 is smooth.

The voice row also holds the envelope: attack and release.

the default patch and the brightness chain

The brightness chain is the set of controls that make the default sound brighter or darker.

With nothing patched, the default settings are:

landscape	double well, relief just over half, warp centred
energy	gaussian, size about a fifth, throw a little right of centre
motion	breathe, speed and depth both in the middle of the pad, re-seed off, dir forward
binaural	drift, couple about a quarter, spread about a fifth
probability	rectified, smooth 0, pinch 0
voice row	tune C2, filter off, clip 100%

It sounds with no cable patched. The packet keeps crossing between the 2 wells, and the 2 fields drift apart on their own. Drag the pad straight down to depth 0, and the breathing stops.

For the bright end, set clip to 200%, size moderate, throw firm, relief high and the filter off. Then try these one at a time:

1. Raise smooth to about 12%. The number of harmonics halves.
2. Set smooth back to 0, and raise pinch to half. The detail crowds towards the middle of the table. The loss sounds different.
3. Switch probability to magnitude. The sound dulls again by about a fifth.

The chain runs in this order: readout, smooth, pinch, clip. clip is the only one that adds brightness. The other 3 remove it.

a pluck and a self-rewriting drone

Use these for a pluck that changes with every note, or a drone that keeps changing.

For the pluck, set:

landscape	well, with deep relief
energy	gaussian, small size, hard throw
motion	pulse, speed and depth both high
re-seed knob	off
gate re-seeds the field	on, in the menu
envelope	shortest attack, release about a tenth of a second
filter	low-pass, with some resonance
clip	pulled back

Patch a keyboard’s v/oct and gate. Every note clears the field and throws a new packet at a hard wall. The notes are related, but each is different. pulse makes the field move, then hold completely still between throws.

You need the hard throw. With a small size in a single well, the sound would settle towards a sine.

For the drone, remove both cables, and set:

re-seed knob	a period of a few seconds
motion	steady, slow speed, any depth
landscape	barrier, medium relief, warp well below centre
energy	comb, wide size, throw at dead centre
binaural	lag, couple high, spread high
probability	a little smooth

It sounds all the time and keeps changing. 4 still packets meet a barrier that splits them. The right channel hears about 45 ms in the past.

This also shows that throw at centre really is still, and that the voice sounds with the gate unpatched.

4 fm drum voices with feedback noise

fracas is a percussion module, 15 hp wide. It has 4 voices that use frequency modulation (fm). You play them from outside the module.

Each voice is a pair of operators. The carrier is the one you hear. The modulator changes the carrier’s tone.

Each operator has its own envelope. One envelope shapes the carrier’s volume. The other shapes the modulator’s level, which is the modulation index. A drum sound comes from driving this pair hard, then letting it go quickly.

Every rattle, hiss, snap and crack comes from feedback on the modulator. So the feedback field sets the noise.

how the panel is laid out

Each voice is a column, with the same controls in the same order. The left lane is the carrier: bend, time, the amplitude slider, and the voice’s trigger input. The right lane is the modulator: feedback, mult, the fm slider, and the voice’s output.

Pitch and wave sit above the point where the 2 lanes divide. The pitch jack is inside the pitch ring. The outer 2 columns sit high and the inner 2 sit low, so the panel dips towards the display.

Each voice has a colour. You see it in 5 places: its 2 rings, the lines under its 4 fields, its jack collars, its slider band, and the cable from its output.

the starting kit

When fracas loads, the 4 voices are tuned as a kit:

voice one	kick, near 50 Hz
voice two	snare, near 180 Hz
voice three	hat, near 600 Hz
voice four	high tom, near 260 Hz

Each voice also starts with its own bend, ratio and pair of envelopes.

Retune voice one, and it is a bass. Set the hat’s feedback to 0, and it is a wood block.

pitch, and the jack inside its ring

The pitch ring has a jack in its centre, at 1 V/oct. The voltage at the jack is added to the ring’s setting. So the control and its modulation input are one object.

All 4 columns repeat the same 10 controls.

wave morphs between 4 shapes

wave moves smoothly from sine, to triangle, to square, to saw.

The centre of the wave ring draws one cycle of the wave you will hear, and it redraws as you turn the ring.

bend sets how far the pitch travels

bend sets how far the pitch moves during the hit. It is bipolar, centred on zero. It multiplies the pitch.

Above the middle, the hit starts sharp and falls into the note. The higher you set it, the deeper the fall. This swoop is what makes a sound feel struck.

Below the middle, the hit starts up to 3 octaves flat and rises into the note. It gives a rising zap.

time sets how fast the pitch travels

time sets how fast the pitch moves. The pitch covers its distance within that time, then stops. Together, the 2 fields set the whole tuning of the drum.

At a few milliseconds, the swoop ends before you hear it as pitch. You hear a click: the beater. At 200 ms, you hear the dive as a note.

mult sets the material

mult sets the modulator’s frequency as a ratio of the carrier’s frequency. The ratio decides what material the drum seems to be made of.

The scale is geometric. The simple ratios sit on the quarter marks.

feedback sets the noise

feedback is the noise control. It feeds the modulator’s phase back into itself. As you raise it, a clean tone becomes rough, then turns into broadband hiss.

Raise it, and a tom becomes a snare. Set it to 0, and a snare dries out to a tick. Both directions from the middle add richness.

each slider holds 2 values

Each slider is one track that holds 2 values. The coloured band between them is the envelope, drawn to scale.

The top edge of the band is the attack. The bottom edge is the decay. They are in the order the envelope happens.

the fm slider sets brightness over the hit

The second slider sets how hard the modulator drives the carrier during the hit. So it controls how bright the hit is over time.

Real struck objects lose their high frequencies faster than their body. To copy this, make the fm decay shorter than the amplitude decay. The starting kit is set up this way.

Make the fm decay longer, and the hit gets brighter as it fades.

Double-click a slider to reset both ends.

map a control to a router slot

The 4 dots around the display are the router’s arm slots. Each slot lets a cv drive one control.

1. Click a dot to arm that slot. The display switches to router view.
2. Touch any control on any voice: a ring, a field, or either end of either slider. The slot now drives that control.

This maps any of the panel’s 40 controls in 2 clicks. To disarm a slot, click its dot again. … The router’s live value is drawn as a second reading. On an envelope slider, it is a thin line at the value the router holds right now. The line appears only while the router pushes the value away from the knob. … CUT (L133)

set the amount for each slot

Each slot has an amt field. It runs from full negative, through 0, to full positive, as a share of the destination’s travel.

The slot’s source is the cv jack beside the field. It reads 5 V as full scale, on the first channel only. A full swing at full amount sweeps the destination from one end to the other.

The field, the jack and the arm dot sit together on the jack rail. Read left to right, they show the source, the amount and the destination. … Click the display to switch views. The display reads values in the destination’s units.

cables and polyphony

CUT (L193)

Its link to the system is polyphony. Every trigger jack and every pitch jack reads polyphonic cables correctly.

material, size and swoop

Try mult, pitch, bend and time in turn, and hear how each shapes a drum.

1. Start on voice four, the high tom. Set mult to a ratio of 1.5, and feedback low.
2. Move the ratio to about 1.7, then to just under 3. Each ratio sounds like a different material.
3. Move the pitch ring. The material stays the same, and only the size changes.
4. Go to voice three, the hat. Drag bend well below the middle, to about a fifth of the way up (about 20 semitones flat).
5. Set time to about 60 ms. The pitch sweeps up into the note from well below: the rising zap.
6. Push bend to its extreme, and set time under 10 ms. The sweep becomes a pure click: the beater alone.
7. Go to voice two, the snare. Leave the amplitude slider alone.
8. Drag the fm slider’s bottom end down until the fm decay is longer than the amplitude decay.

On a snare, this sounds like a decay played backwards. While you drag, the bottom line of the display follows the fm pair.

2 slots on one control

Map 2 sources to one control, pushing it in opposite directions.

1. Arm slot one. Touch the bottom end of voice two’s fm slider (the fm decay).
2. Set slot one’s amt to about half negative.
3. Arm slot three. Touch the same slider end.
4. Set slot three’s amt to about 70% positive.
5. Patch a velocity signal into slot one’s jack, and a slow oscillator into slot three’s jack.

Harder hits now sound darker, while the slow source moves the same setting underneath. The 2 slots add together.

To unmap both slots at once, click the amber mark beside that slider end.

a struck object whose shape sets its pitches

frappe is a physical modelling percussion voice. A mallet strikes a resonant object, and the object rings.

The object is a tiling: a pattern of shapes that fills a flat surface. The pattern in the window is the object itself. The window and the sound read from the same table.

how the shapes make the sound

Every polygon in the tiling is a resonator, and its size sets its pitch. Smaller polygons ring higher: frequency goes as 1 over the polygon’s inradius (the radius of the circle that fits inside it).

Each resonator rings at several frequencies, called modes. frappe collects the modes of every polygon family in the tiling. It weights each family by how much of the surface it covers.

triangles	use the exact spectrum of an equilateral triangle
squares	use the exact spectrum of a square
6 or more sides	use the spectrum of the disc they fit inside

The round shapes sound like drums, because they very nearly are.

why each tiling is its own instrument

Because size sets pitch, the 7 tilings sound like 7 instruments.

In the tiling of 12-sided shapes and triangles, the 2 shapes differ in size by more than 6 times. So the triangles sparkle about 2 octaves above the 12-sided drum, with a gap in between. A tiling of one square size has a dense, even set of modes, full of twinned pairs.

frappe merges and sorts all the modes, and keeps the lowest 48. Each mode remembers which polygon made it. This is how the window can light the triangles and hexagons on separate timings.

what changes while the object rings

When you move a control that shapes the object, frappe rebuilds the sound within 64 samples, and the filters follow within 16. So you can retune a ringing object by hand.

hit and exciter describe the contact, so they wait for the next strike.

live under a ring	pitch, the pitch input, pos, decay, matr, disp, aspt, body and bend
latched at the strike	hit and exciter
noiz	acts on the strike force, so it also reaches a ringing object. It sets the density of the geiger strike
the audio input	acts on the strike force for the same reason, and gates the geiger strike

If you sweep body while the object rings, the ring bends to the new object.

body chooses the object

body chooses 1 of 9 objects: 7 tilings, then a rod and a chime.

Click the field to step to the next object. After the last, it wraps back to the first. The window redraws to show the new object. body is an ordinary field, and the window shows its value.

disp stretches the modes

disp spreads the modes further apart, and makes the ripple in the window wider. It is an exponent applied to every mode’s ratio.

0	the plain geometric spacing. The object behaves like a wave, and each mode climbs in proportion to its number
full	every ratio is squared. The object behaves like a stiff plate: very inharmonic and much brighter, with the top modes thrown far up

matr sets how fast the top dies

matr sets how much faster the high modes decay than the fundamental. It uses a square law on each mode’s ratio. Every mode rings for at least 4 ms.

matr changes decay time only. The spacing of the modes stays where body, disp and aspt put it.

aspt splits twinned modes

Regular shapes have modes in identical pairs: the same shape, rotated. aspt makes the object slightly uneven, and pulls each pair apart by up to 18% either side.

The split pairs beat against each other. The tuning stays the same.

pitch sets the fundamental

pitch sets the fundamental, from 24 Hz to 3 kHz, on an exponential scale. The field prints hertz. The default is 110 Hz.

Every mode is a ratio of the fundamental. pitch moves them all together, so the spacing stays where body, disp and aspt put it.

decay sets the ring time

decay sets how long the fundamental rings, from 20 ms to 6 s, on an exponential scale. The field prints milliseconds or seconds. decay also zooms the display.

It sets the fundamental’s decay only. The modes above it decay faster, by the amount matr sets.

bend adds a falling pitch

bend adds tension to the hit, from 0 to 24 semitones. The default is 5.

Every mode’s frequency follows the object’s own envelope. So the pitch starts sharp and falls back as the hit fades. bend always falls, and it also reaches an object that is already ringing.

exciter chooses the strike

exciter chooses 1 of 9 kinds of strike. The field is continuous, but its travel is split into 9 named zones, and each zone gives its strike in full. The default is tap.

The first 6 are single contacts. Each presses a force shape against the object for the length of the contact.

The last 3 are patterns of contact: a bounce, a geiger counter and a bow.

hit sets the contact time

hit sets how long the mallet touches the object: 5 ms at the bottom, down to 0.25 ms at the top, on an exponential scale. A shorter contact reaches further up the modes, so the hit sounds harder and brighter.

The contact model changes with it. frappe calculates the mallet’s stiffness to match the contact time you set.

noiz adds noise to the strike

noiz mixes noise into the strike force, live, sample by sample. It only sounds while a contact or a brush is happening. The default is 5%.

The noise follows the force. It is loud where the force is strong, and silent between hits.

pos sets where the mallet lands

pos sets where the mallet lands on the shape. It is a real position on the polygon.

A mode with a still point (a node) under the mallet stays silent. A mode whose shape is negative at that point starts moving the other way.

At dead centre, every mode that is not symmetrical around the centre cancels. This is where a kick sound lives.

4 layers

The window shows 4 layers, each carrying a different quantity.

In the light, the hue shows which polygon, and the brightness shows how much energy. Because hue shows the polygon, a shape keeps its colour as it fades. The families flash and fade separately.

With nothing patched, the bowl is completely flat.

drag the mallet

You can drag the mallet in the window to set 2 controls at once.

horizontal distance from the centre	pos
vertical position	hit

Press anywhere in the window with the left button, and drag. The press sets both values at the point you press, before you move.

frappe works through its jacks

frappe connects to the system through its jacks.

Its window runs on its own energy. So it draws a full tiling as soon as you add the module, with nothing patched. The browser card also draws a tiling, using the default settings.

a kick and a beating bell

Use these to hear the 2 extremes: a clean kick from the geometry, and a metal bell.

For the kick, set:

body	skin
pos	exactly 0. Drag the mallet to dead centre, or right-click the knob and type it
pitch	about 55 Hz
decay	about 450 ms
matr	about 75%
bend	about 14 semitones
disp	0
hit	about 55%
exciter	press
noiz	0

The centre position cancels every mode that is not symmetrical around the centre. This leaves the 3 modes of a disc. matr removes the top, and bend supplies the dive.

For the bell, set:

body	bell, which has 3 polygon families
matr	about 10%, so everything rings
decay	about 4 s
disp	about 35%
aspt	about 45%
pos	about 55%
hit	about 85%
exciter	tick
bend	0

Low matr makes it sound like metal. aspt splits the twinned modes, so the partials beat. disp stretches the modes past any harmonic spacing.

Watch the window. The 3 polygon types light and fade at different rates, which shows the inharmonic sound on the surface.

a scrape, and another sound through the object

Use these for a long noisy scrape, or to play the object with another sound instead of a mallet.

For the scrape, set:

exciter	brush, the sixth zone
hit	about 20%
body	box
noiz	about 35%
decay	about 250 ms
matr	about 60%
pos	about 70%

With brush, hit sets the length of the scrape. So 20% gives a long scrape. box has twinned pairs all the way up, and they beat quietly under the noise.

To play the object with another sound:

1. Patch a sustained sound into the audio input.
2. Set body to tom, decay to about 2 s, matr to about 15%, disp to about 5% and pos to about 30%.
3. Leave the trigger unpatched.

The patched sound alone drives the object. The window bows with its level.

detuned saws played as one voice

scie is a polyphonic supersaw, 3 hp wide. One note builds a stack of up to 16 band-limited sawtooth waves.

The saws are spread in pitch around the note you play. 2 gain curves that cross each other set their levels. They are laid across the stereo field in pitch order. A rounded saw plays an octave below.

The stack starts at 7 saws, but you can set anything from 1 to 16. A single saw and a full wall of saws are the same module at 2 settings.

One saw sounds thin and static. Several saws a few cents apart beat against each other all the time. The beating rate differs at every harmonic, so the sound never settles into a steady tone. This constant movement is what makes the sound big.

how the saws are spread

scie spreads the stack in 2 different ways at once:

in pitch	a curve packs the inner saws much closer to the centre than an even spread would. The result is a crowded core with 2 outliers. The flat side reaches about 2.5% further than the sharp side
across the image	each saw’s pan position is its place in the stack multiplied by the width knob. This is a straight line

The pitch curve puts most of the beating near the note, where the clusters stay distinct. Only the 2 outer saws go far out. This is the difference between a chorus and a wall.

Together, the 2 spreads give a note that is dense in the middle of the spectrum and even across the speakers. The pitch order stays the same: the flattest saw is hard left, and the sharpest is hard right.

tune shares the pitch jack

tune sets the pitch, from C0 to C6, in whole semitones. That is 4 octaves down and 2 octaves up from C4. It prints a note name.

At the bottom of the range, scie is below hearing. C0 is 16 Hz, and you hear each tooth of the saw as a click.

tune is added to the incoming pitch. So tune and the 1 V/oct input are the same thing, arriving 2 ways. This is why tune’s cv jack is the pitch input.

Every other control has its jack in the row directly above it. tune’s jack is the v/oct jack.

voices sets how many saws

voices sets the number of saws, from 1 to 16, in whole numbers. As you sweep it slowly, the stack changes from a detuned pair into a texture.

The count changes how detune sounds. Fewer voices with more detune sound rougher, with audible beating. More voices with less detune sound smoother and denser.

detune sets the spread in cents

detune sets how far the saws spread in pitch. The field prints cents.

The knob is far from linear in cents: the curve is steep. Turning the knob up always adds more detune. Listen for the point where the clusters merge.

width sets the stereo spread

width places each saw at its own angle across the stereo field.

At width 0, the left and right outputs are exactly the same, sample for sample. The engine writes the same equal-power level into both sides.

blend sets centre and side levels

blend moves 2 gain curves at once.

The centre saw of an odd-sized stack follows one curve. Every other saw follows the other curve. The 2 curves cross about three quarters of the way up.

blend needs an odd number of saws

blend works on the difference between the centre saw and the side saws. Only an odd number of voices has a centre saw. So blend works on counts of 3 and up.

the level stays steady

Before you hear anything, scie divides every gain by the square root of the sum of their squares. This keeps the output level within about 3 dB, from 1 voice to 16.

sub adds a saw an octave down

sub adds a rounded saw an octave below the stack, after the filtering. The field prints decibels. At the bottom it reads off, and scie skips the sub completely.

The rounded shape has weaker upper harmonics than a plain saw. They fall away as the square of the harmonic number. So the sub adds weight underneath.

The sub is mono. It goes into both channels equally, so the low end stays centred however wide the stack is.

read the stack from the picture

The window at the top draws one sawtooth for each voice. Each is shifted sideways by its place in the detune spread.

the number of teeth	the voice count
how far the fan opens	the detune
the heights	the blend. A peak in the middle below the crossing, a dip in the middle above it
the filled shape underneath	the sub, drawn with the engine’s own rounded saw

the teeth move at the played pitch

Each tooth wobbles 1.5 pixels around its position. The wobble rate comes from the note you play: about 1 cycle a second at 2 octaves below middle A, doubling with every octave up.

On a fast line, the picture shimmers. On a bass note, it moves slowly enough to count. On a chord, the shimmer follows the first polyphonic channel only.

drift

drift is on the right of the top row. It sets how far each saw wanders out of tune. The steps are named off and a to d, each wandering further than the last.

off	the stack is exact. This is the default
a	a shimmer, 2 cents
b	an analog stack, 5 cents
c	coming apart, 9 cents
d	seasick, 16 cents

Click the field to step through the 5 settings, drag it to scrub, or right-click it for the list. The menu names the cents, the panel names the step.

Each saw wanders on its own slow path, taking between 1.5 and 3.5 seconds to reach a new target. They never move together, so the stack comes apart and gathers again instead of bending as one. The sub stays where it is, so the bottom of the sound holds still.

a wide wall and a focused lead

The 2 settings sit on opposite sides of the blend crossing, so you can hear what blend does on each.

For a wide wall, set:

voices	16
detune	about a third of the way up (16 cents)
blend	well above the crossing
width	fully open
sub	a few decibels down
tune	unison

The outer saws are loudest and the centre is scooped. You hear a hollow sound spread across the field in pitch order. It is wider than a single saw, at the same level. The width is in the source, so it survives being summed to mono.

For a focused lead, set:

voices	7
detune	a fifth of the way up (5 cents)
blend	well below the crossing
width	about a quarter open
sub	a little way up

The centre saw is loudest, so the pitch is clear. The stack sounds like one note with a halo around it. Below the crossing, blend gives one note. Above it, blend gives a texture.

a noise cluster and a note that thickens

For the cluster, set:

voices	15
detune	maximum (181 cents)
blend	at the crossing
width	most of the way open
sub	off

The outer saws are nearly 2 semitones out, and the clusters merge by the fourth harmonic. The result is a noise source with a pitch centre. It is even, because blend at the crossing lets no saw lead. The count must be odd for blend to work.

For a note that thickens, set:

voices	1
the voices jack	a slow attack envelope
detune	about half
blend	near the default
width	well open

10 V adds 15 voices. The note starts as a single saw and thickens into the full stack over the envelope. The count is a whole number, so the voices arrive in 15 separate steps.

A moving cv on voices makes scie rebuild the stack every sample. This patch uses the most processing power.

4 oscillators, chosen by output jack

ton is a bank of 4 oscillators, 3 hp wide. You choose a wave by patching its output jack.

sine	needs no anti-aliasing, because its shape and slope have no sudden jumps
triangle	made by integrating the corrected square wave, with a servo that removes drift
saw	uses one band-limited step correction
pulse	uses 2 corrections, one for each edge. It is the only section that the menu’s width slider affects

Patch 2 cables, and you get both waves, running together at the same pitch.

The rest of the panel follows from this. There are 4 sections, each with a small window and its own tuning offset. The pitch inputs are normalled downwards, so one cable can feed all 4 sections.

each section is an oscillator or an lfo

Each section’s tune field runs from 120 semitones down to 24 semitones up: 10 octaves down and 2 up. With nothing patched, that is about 0.25 Hz at the bottom and 1046 Hz at the top.

So a section is whatever you tune it to be. The same waves, jacks and detents do both jobs.

For example, the saw at 60 semitones down runs at a little over 8 Hz. Patched to a filter cutoff, it is a sweep. Meanwhile, the pulse left at unison is the audio.

5 sixths of the field’s travel is below 20 Hz. So most of this control’s range sets the speed of a slow shape.

tune sets each section’s offset

Each of the 4 sections has its own tune field. The number is an offset in semitones from the pitch that section receives. 0 is unison with the input.

The field snaps to detents.

the window shows the interval

The small window above each jack draws that section’s wave, at its own ratio. It always shows 2 periods of the root pitch, and scrolls at the pitch.

So you can read the interval from the number of cycles:

24 semitones down	half a cycle
unison	2 cycles
an octave up	4 cycles
the top of the travel	8 cycles

Below 24 semitones down, the window stays at half a cycle.

the line under each number

Under each tune field is a short coloured line. It shows where that section sits across the full 144 semitones of travel.

A faint fixed notch marks unison. The notch is 5 sixths of the way along, because the travel goes 10 octaves down but only 2 up.

whole numbers above 10 semitones

Below 10 semitones, the tune field shows a decimal. At 10 or more, it shows a whole number only. A longer reading, such as minus 12.4, would run into the jack beside it.

Drag the field left or right, like a knob.

one cable plays all 4 sections

The 4 pitch inputs are normalled downwards. Each sample, ton checks them from the top. The last input with a cable passes its voltage to every section below it.

So one cable in the top jack plays all 4 sections. A second cable lower down splits the stack into 2.

the sum output is a fixed mix

The fifth output mixes all 4 sections at fixed levels. The sine is loudest, then the triangle, then the saw, then the pulse. The mix is divided by 4 and goes through a gentle soft clip.

It is divided by 4 because the 4 sections are tuned in simple ratios to each other. Related waves like these add up to a higher level than unrelated sources would.

Polyphony comes from the cables.

ton connects by cable

ton connects only by cable. It gets its multiple voices from the normalled chain and from polyphony.

The only thing it reads is 1 V/oct pitch, on up to 4 jacks. So it works best with modules that make pitch: a keyboard’s polyphonic pitch and gate, a sequencer, or the quantiser.

Put a quantiser before ton. The whole stack follows one cable, so one quantiser tunes all 4 sections at once.

audio and control from one module

Any input that expects a 5 V audio signal can take the 4 wave outputs and the sum. Any input that expects a cv can take a section tuned below about 60 semitones down, where it becomes a slow shape.

One ton can be an oscillator and a modulation source at the same time.

a 4-note chord from one key

1. Patch a pitch cable into the sine jack only.
2. Set the 4 offsets to unison, an octave up, a fifth, and an octave and a fifth. These are all wide detents, so a normal drag lands on them easily.
3. Take the sum output.

One key now plays the root, octave, fifth and twelfth. The offsets are per section, so each note has its own wave: the root is a sine, the fifth is a saw, and the top note is a pulse.

Move any tune field, and the chord changes with it. The windows show the new ratio as a cycle count before you play it.

Make the cable 6 channels, and the same patch becomes a 6-voice pad. The offsets apply to every channel, so every held note gets the same chord. Add a 6-channel gate through a bank of amplifiers, and the whole polyphonic voice is one ton plus one amplifier.

2 voices, or an oscillator and a sweep

For 2 voices:

1. Patch pitch source A into the sine jack, and pitch source B into the saw jack.
2. Set the sine and triangle offsets to unison and an octave down.
3. Set the saw and pulse offsets to unison and a fifth.
4. Send the sine and triangle outputs to one destination, and the saw and pulse outputs to another.

The split happens at the second cable. Remove it, and ton returns to one voice in unison, with no change on the panel.

For an oscillator and a sweep:

1. Patch one pitch cable into the top jack only.
2. Set the saw tune field to 60 semitones down. Its window slows to a crawl.
3. Patch the saw output to a filter cutoff.
4. Leave the pulse at unison, and patch its output to the audio path.

Take the saw lower still: 96 semitones down is about 1 Hz. The bottom of the range, 120 semitones down, gives about 0.25 Hz, one sweep every 4 seconds. The window still moves, at its minimum rate.

one sample played by 8 heads in a chain

coupe is a sampler with one stereo buffer and 8 playback heads. Each head is a region of that buffer, with its own start and end. Each head also has its own speed, direction, tone, level and pan. The result is 8 windows onto one sound.

The chain is a ninth voice. It plays the active head, finishes that head’s pass, and then hands on to the next head. This walk is what you hear when nothing is patched.

Most of the panel shapes the chain. It sets which regions play, in what order, how many times each, and how one region hands on to the next.

how to read the panel

The panel shows the 8 regions in 3 views:

the waveform	at the top. Shows where each region sits in the buffer
the 8 lane rows	the transport of each head
the deck	9 columns showing the selected head in detail

controls and undo

coupe puts its whole interface on one display. Every control is a click or a drag, and every value is stored in the module’s own saved data. This has 3 results:

the right-click menu	is short. It only holds the items coupe adds
initialise	restores the module’s saved data
undo	each drag is one undo step

Undo steps are named after the action: move slice region, slice parameter, assign modulation, cv1 mode, load kit, send from bac.

the waveform edits the selected head

The strip across the top is the whole buffer. The 8 regions are drawn on it, each in its own colour.

The waveform edits the selected head. Select a head first, on its lane or on the deck.

the slicers

6 icons sit either side of the waveform. 3 of them are slicers. The icons light blue as the pointer crosses them.

The transient slicer starts from 8 even divisions. Each slice point moves to the strongest onset within about a third of a division either side. Nearer onsets score higher. So a slice lands on a real hit, but stays near the grid.

If there are no transients, or the buffer is very short, the transient slicer uses even eighths instead.

the deck shows 9 values

The deck is a row of 9 values below the waveform. Above it, a page line reads global, lane and mod. The page you are on is blue. Drag a value to change it.

lane	the selected head’s parameters. Selecting a head on its lane always returns to this page
global	the master: pitch, tone, res, lvl and pan for the whole mix, then the settings for the whole module
mod	faint until you pick a modulator tile. The 9 columns then become that modulator’s editor

the global page

The global page starts with the master controls, then 4 settings for the whole module.

pitch	moves every slice together
tone, res	one filter on the whole mix
lvl	the master level
pan	the balance

After these come crossfade, stretch and quantise. Crossfade starts at 4 ms, so one slice hands over smoothly to the next. Drag it to the bottom to turn it off. The last setting is choke, which is off by default. With choke on, firing a lane stops every other lane that is still sounding, in any sequence mode.

The input monitor joins the outs after the master controls. So you hear what you are about to record as it arrives.

tone holds 2 filters

tone is one knob with 2 filters. res sharpens whichever filter is open.

below centre	a low-pass sweeps down, to about 40 Hz at the bottom
above centre	a high-pass sweeps up, to about 8 kHz at the top

the lane rows

The 8 lane rows sit under the deck. Each row is one head’s transport, and it reads from the left. Click anywhere on a lane to select that head.

when the chain stops

speed can freeze a head. At 0%, the head’s phase stops moving, so its pass holds. The chain waits for each pass to end, so it stops too.

repeats can also hold the chain. Infinity in the repeats column overrides everything, including a sequence mode that would stop. A head with infinite repeats holds the chain until you change a setting.

6 modulators

coupe has 6 modulators: 3 lfos, 2 envelopes and a sample and hold. Their kinds are fixed, but you can edit everything about how they behave. Each tile has an arm dot in its corner.

what a tile draws

Each tile is a scope of what its modulator is really sending. The newest value is on the right.

an lfo	the real signal, scrolling. It shows the lfo’s true phase, the random values it actually drew, and the effect of any modulator bound to its rate. A synced lfo stands still while the transport is stopped, and its trace goes flat
an envelope	the envelope’s shape from its settings, with a line that shows how far the last trigger has got
the sample and hold	the real signal, over about 1 second

An lfo’s tile holds more time when the lfo is slow, and less when it is fast. A slow lfo shows about half a cycle, and a fast one about 8. So a faster lfo still looks tighter.

assign a modulator

1. Arm a modulator with the dot on its tile.
2. Click a deck column to bind that parameter to it.
3. To unbind it, click the same column again with the same modulator armed.

A new binding modulates the current value by about 15% either way.

A modulator binds to a lane’s parameters: start, end, pitch, tone, res, level and pan, on the deck’s lane page. It also binds to the master’s pitch, tone, res, level and pan on the global page. Arm a modulator on either page and click the cell you want. A bound cell shows a range bar in place of its line, with a dot in the modulator’s colour beside its name.

A free lfo always runs.

the modulator editor

In a modulator’s editor, the leftmost column names the modulator you are editing. Drag it to move to another modulator, about 10 pixels for each one.

An lfo’s sync column is a switch. Click it to turn sync on or off. With sync off, rate is a speed in hz. With sync on, rate is a length, from 8 bars down to a 16th of a bar, and the lfo follows the system’s transport. It takes the transport’s tempo, it starts each cycle on the bar line, and it stops when the transport stops. A cable in the clock input takes over the tempo, as it does for the rest of coupe. sync remembers its length when you turn it off and on again.

shape, a synced rate and trig each hold 1 of a fixed set of choices. Click one to move to the next choice, or drag to scrub. Right-click it to see every choice in a list. The cv1 and cv2 columns on the lane page work the same way, and so do each lane’s direction and mode, and chain and write in the chain row.

Some columns only appear when they apply. For example, an envelope’s rate field only appears when its trigger is set to rate. The sample and hold has its own poly field.

run, slice and chain

The top row holds 3 controls for the chain.

chain	sets how the chain walks. There are 7 sequence modes
run	the teal light. It stops the chain only. Slices you fire by hand, lane triggers and voices played by gate still work
slice	shows which head the chain is on. Drag it to jump the chain to another head

run also controls the whole system, so pressing it can start and stop everything.

the gate jack

What the gate jack does depends on the sequence mode.

A long mono gate plays many slices. Stacked poly notes play chord voicings of the same running sequence. With a poly gate, the gates start the chain.

the key modes

In key poly and key mono, the pitch input chooses which slice plays.

The slice always plays once, in full.

crossfade

crossfade is the sixth column of the global page.

It goes up to 1 second. The first quarter of its travel covers most of the range you can hear.

arm and record

coupe records into its own buffer. When you arm it, it waits, and the first sound starts the take.

To start the take the moment you arm it, right-click coupe and turn off a take waits for sound. The take then keeps the silence, and a take held open by a gate is exactly as long as the gate.

The red light arms recording, and so does the rec jack. They are 2 halves of one control, and either one keeps a take going.

the write modes

The write field chooses what a take does to the buffer. It has 3 modes. 2 of them, dub and punch, play onto the running sequence.

dub and punch write under the chain’s playhead, so the overdub lands on the slice that is sounding. They fill the whole span the head crossed since the last sample. So slices at a changed speed, or stretched to the clock, still record in full.

Both keep the regions and playheads on screen, and only turn the playhead red. The display shows which of the 3 modes is running. With an empty buffer, dub and punch work like new.

a finished take becomes a file

When a take ends, coupe writes the buffer to a stereo wav file in its takes folder. It then points itself at that file, as if you had loaded it. This applies to new, dub and punch takes.

The deck names the file. The preset saves the path, like any sample’s path. When you reopen the preset, the take comes back.

A dub over a kit is saved as one file, holding what the kit had become.

The takes folder is called coupe takes. It sits beside the presets folder. On a Mac, chassis moves unused takes older than 30 days to the Trash at launch.

the shared transport

coupe’s run button starts and stops the whole system. coupe joins the shared transport, and is linked by default.

The bus gives coupe tempo, run and reset. One slice per pulse needs a cable in clk.

pairing with bac

bac, the sample browser, sends selections straight to coupe. The link carries 3 things:

the head count	so bac’s selection limit always matches coupe
every coupe in the patch	in the system’s reading order, each labelled with what it holds
the send	the selection itself

a break sliced to the grid

1. Drop a file on the panel.
2. Click the transient slicer.
3. Patch a clock into the clock jack, and leave clock-stretch on.
4. Set chain to forward, every repeat to 1, and crossfade to 0.

Each pass now lasts one pulse. Set one head’s speed to 50%: that slice plays at half speed, and still fits the pulse. This is because the stretch comes from the region’s natural length, and speed multiplies on top of it.

a free-running stutter

Start from the break above. Remove the clock cable, then set:

chain	ping-pong
one head’s repeats	4
that head’s direction	ping-pong
that head’s speed	75%
crossfade	about 40 ms

With no quantising clock, the crossfade lead-in is active. The 4 repeats blend into each other. The ping-pong walk turns round at the ends.

Push crossfade past about 120 ms and the tails start to stack. Above that, the chain sounds smeared rather than sliced.

8 one-shots played from a keyboard

1. Load a folder of 8 sounds as a kit.
2. Set chain to key poly.
3. Patch pitch and gate from a keyboard.

Each head already sits on its own sound, with the seams drawn on the waveform. Each key press gets its own voice and plays in full. One sound can roll, or 3 can stack as a chord.

Switch to key mono, and the kit becomes a choke group. Add 25 ms of crossfade, and the cut becomes a duck.

a file browser that shows every waveform

bac is a sample browser. Every visible file row draws its own waveform at all times: on hover, when selected and at rest. So a folder of kicks and a folder of field recordings look different before you read a file name.

how the waveforms are made

3 processes make the waveforms:

a background scan	finds the files
a cache for each folder	keeps the results
an analysis pass	works on the quickest files first, so the list fills in from those

A folder you have just opened fills in over a few seconds. A row waiting for analysis is a faint ghost, with 3 dots in place of its duration.

what reaches the patch

You browse on the panel, by hand and with the keyboard. Only the audition reaches the patch, through 2 preview outputs.

The only other link to a patch is the sampler that bac sends files to. You set up that connection in the menu.

roots

A root is a folder that bac reads. Every bac in the system shares the same roots.

The bar across the top names the folder being read.

the arrows	at the left of the bar. Step between the folders you have added. They wrap round, and each step rescans
the folder name	click it to open a picker and add another folder

You can also drop a folder on the panel to add it.

If you have not added a folder yet, bac opens on the factory sample content. This is the free Lux Cache samples that came with the install. Where the full set is installed, bac opens on that instead. The full set is shared with every Lux Cache product on the computer.

crates

Inside a root, each first-level folder becomes a crate. The scan shows which stage it is at.

Files in deeper folders are gathered into their crate, with their path shown in the name. Loose files in the root gather into a crate of their own.

Crates start closed. When you reopen a saved patch, the crates that were open open again.

the parts of a row

A file row has 5 zones. From left to right:

the name	clipped to fit
the waveform	dimmer while the row waits for analysis, brighter when it is selected
the right-hand column	its meaning changes
the star	marks a favourite. Drawn when the file is starred, selected or under the pointer
the ear	at the outer edge. Ranks the library against this row

search

Search matches file names, and opens the crates that match.

Click the search field to type in it. It takes plain characters and changes everything to lower case.

backspace	deletes a character
escape	clears the search and leaves the field
enter	leaves the field and keeps the search
clicking elsewhere	leaves the field

sort

The word beside the count changes the sort. Click it to cycle through the 6 sorts.

Crates follow 4 of them: the name and date sorts. Under shortest and longest, crates stay in alphabetical order, and the files inside them sort by length.

rank by ear

The ear ranks the whole library by how a sound behaves. Click the ear at the outer edge of a row to rank the library against that row.

The sort word changes to "by ear". A new row appears on the rail. It names the reference file, with a small cross to clear it.

scrolling and the rail

Use the scroll wheel to scroll the list.

When you scroll into a crate’s files, the crate’s header pins to the top edge of the list, with a soft shadow under it. Click the pinned header to close the crate and scroll back to it. A click in that top row always goes to the pinned crate, whatever file has scrolled under it.

The divider of the rail is a drag handle.

3 kinds of click

There are 3 kinds of click for selecting files. The most files you can select is the head count of the sampler.

click	selects one file
cmd-click (ctrl-click on Windows)	adds or removes one file
shift-click	selects a range

The caption, the waveform at the bottom and the audition voice each show one file at a time.

If a shift-click range goes over the limit, bac drops the far end. It keeps the rows next to the first row you clicked. The last row it keeps is the one the audition voice plays.

order on screen is order sent

The order on screen is the order the files arrive in, and sending keeps it.

Each time the list rebuilds, the selection is put back into the order the rows are drawn in. This is the head order when the files reach the sampler.

So if you change the sort while files are selected, the heads are renumbered. The head-n column shows this as it happens. This is how you arrange the order of a kit.

the caption and the large waveform

The caption under the list names the crate and file, in the crate’s colour. It also shows the duration, sample rate and number of channels.

Below it, a full-width waveform shows the row’s picture at full size, with a playhead. Click it to jump to that point and start playing at once.

When you drag a file out, it always carries the whole file at full length.

step through with the arrow keys

With the pointer over the panel, the arrow keys move through the list. With auto-audition on, each row plays as you land on it. The list scrolls to keep the cursor in view.

The keyboard cursor is a thin outline on its row, and looks different from the selection.

the preview with nothing patched

In the plugin, bac’s 2 preview outputs go to the inputs of sortie. So the audition works as soon as you add the module.

This route stops as soon as a real cable feeds either end, or either module is bypassed.

If you patch a preview output by hand, the loaded file plays from the start, so you hear the patched preview at once.

One library feeds every bac.

audition files in context

1. Patch the preview outputs into the same chain the sampler feeds: its filter, its delay, its levels.
2. Leave auto-audition on, and turn loop off.
3. Point at the panel and hold the down arrow.

Each row loads and plays as the cursor lands on it. You can compare 40 similar files with one key press each.

Turn auto-audition off, and the same steps run in silence. Use this when you are choosing files to send.

rank by what you hear

Start from a file you already like, and rank the others by how alike they behave.

1. Wait for the reference row’s waveform to reach full strength. The ear only ranks analysed files, and a ghost row is still waiting.
2. Click the row’s ear.

The ranking uses 8 descriptors of the sound. So a file of any name can rank close to the reference.

To leave the ranking, click the cross on the rail’s reference row, the sort word, or the same ear again.

build and send an 8-slice kit

The kit goes to the sampler.

1. Open a crate and click the first row.
2. Shift-click the eighth row.
3. Read the head-n column down the rows to check the order.
4. Right-click inside the selection and choose send.

The count line shows how many of the 8 slots are filled, so you can always see how much room is left.

Some selected files may be in a compressed format. If so, the menu says how many it will skip.

a keyboard with 7 pages of controls

clavier is a keyboard you can play, with a screen that holds 7 pages of controls.

The jacks run along the top edge. The keys have the bottom of the panel to themselves.

keys are always keys

Every other function is a page on the screen. So pressing a key always plays a note.

Every value on the screen is a plain setting on the module. You edit it on the panel, and it is saved with the patch.

velocity and slides

Where you click on a key sets the velocity. Near the top of a key gives 127. The very bottom gives the lowest velocity above 0. Everything between is in proportion.

A held key shows a blue wash up to the height you clicked, with the velocity printed on it. So the wash also shows you the velocity.

Black keys are shorter. The same click position gives a different velocity on a black key than on the white key beside it.

Drag across the keys to play a glissando, and to slide between notes.

the scale tint

When a scale is live on the shared bus, the keys in that scale have a teal tint. The root note has a coloured dot. These marks only appear while something is reading the bus.

Keys outside the scale still play.

Only the C keys are named. When the root dot sits on a C, the name moves out of the way.

buttons, strips and octave

The keys have 4 buttons, 2 strips and the octave control.

Turning hold off releases every held note at once, and closes the gate of every voice the keys own. You always hear this as a release.

pages and rows

The rail down the left of the screen chooses a page. The area beside it holds that page’s rows. There are 7 pages, with up to 5 rows each.

There are 3 kinds of row. Their marks tell them apart:

a bar under the row	drag to change it
a word	click to step to the next one
a link mark	shows a value from somewhere else, such as the shared tempo, the shared scale or the state of the audio input. The mark is a button

Each page has 5 rows.

A dot beside a page’s name on the rail means that part is running: the arp is on, the phrase is playing or recording, a chord type is set, a voice is sounding, clavier leads its note lane, or the input is open.

the display

The darker right half of the screen is the display. It is laid out the same way on every page. The top line says what is set, and what is happening. The middle is a drawing. The bottom line holds the page’s verbs on the left and its one action on the right. A verb out of use goes grey, and stays where it is.

On the seq page the drawing is the phrase. Click a step to go there. While rec is armed that moves the writing head. Drag a note up or down to change its pitch, and double-click a note to remove it. The 8 marks on the top line are the 8 phrases. A filled mark holds notes. Click one to go to it.

On the chord page each voice of the chord has its own line, named by note and by interval. With the arp off, a bar starts where its voice comes in, so level bars are a block chord and a staircase is the strum. With the arp on, each note enters at the right as it sounds and drifts left.

On the in page the drawing is the last 10 seconds of the input, newest at the right. A bar that reaches the edge turns red, and so does the peak reading.

The arp and seq buttons on the left also show their page when you click them.

the event log

The right part of the screen, from about three quarters of the way across, is a live list of what just happened. The newest event is at the top, and each row fades with age.

The log holds 24 events and shows 8.

arp modes and timing

The arp has 9 modes, which walk over the range set by the octaves row.

order	plays the notes in the order you pressed them
up-dn	plays the notes at each turn once
incl	plays the notes at each turn twice
the 2 ×2 modes	play every note twice
walk	wanders one note at a time

The rate row has 15 divisions. The 3 slowest are 2 bars, 1 bar and a half note per step. The other 12 are each straight division, with its triplet and its dotted version. The phrase’s rate row has the same 15, so a phrase can move as slowly as one step every 2 bars.

gate and swing are shared with the phrase. swing above halfway pushes every second step late.

the arp plays alone

While the arp plays, clavier closes the gate of every voice the keys own, at every sample, so you hear the arp alone.

When the arp stops, the keys play normally again.

8 phrases in time with the bar

clavier holds 8 phrases of up to 64 steps each, saved with the patch.

The loop position comes from the absolute bar. So 2 claviers playing phrases of the same length stay in the same bar.

record a phrase step by step

Press rec while the phrase is stopped. A take opens, and you write steps by hand.

Each key press writes the step at the cursor, and you hear it as it is written. When you release every key, the cursor moves on.

The cursor counts physical keys. However many chord tones one finger adds, the cursor moves on when the last key comes up.

A take is one undo step, from rec on to rec off. Undo removes the whole take and redo brings it back, whether you played it on the panel, on the computer keyboard or over midi. Edits on the display, such as moving or deleting a note, clear and mutate, are each one step.

The plot steers the take. The strip also has 2 more actions:

to seq	writes the arp over the held keys into the phrase, one note per step
mutate	changes the notes already there

With keys set to transpose on the midi page, a key you play while the phrase runs moves the phrase to that key.

one key plays up to 5 notes

One key can play up to 5 notes. The added tones belong to the key you pressed, so releasing that key releases all of them.

spread	lifts every other added voice by whole octaves. This opens the voicing, but the notes stay the same
octave	moves the added tones up to 2 octaves either way
velocity	the velocity of the chord tones. It is an offset from the key’s velocity, so it still follows how hard you play

poly and out

poly runs from 1 to 16. It is the most voices clavier will use, and so the most channels the v/oct, gate and velocity jacks can carry.

Lowering poly takes effect straight away. At every sample, any voice above the poly count has its gate closed.

out, on the bus page, has 4 shapes.

how voices are given out

In poly, each new note takes the first free channel. So a single melody line always uses channel 1. A module that reads only channel 1 hears every note, such as a drum pitched from here and triggered elsewhere, or a filter that follows the keys. Chords spread upwards.

In poly, the jacks carry only the voices in use. A single line is one channel, and a 3-note chord is 3. A released voice keeps its channel for 8 seconds, so its release can finish. An oscillator sounds every channel it receives, so scie or ton on the v/oct jack plays one voice for each key.

In rotate, each new note takes the voice that has been silent longest. This keeps the last note’s release off the channel the new note uses. Use it with synths that keep their release on that channel. In rotate, the jacks always carry every channel, so give each voice its own envelope and level control.

quant holds a note until the grid

quant sits between the 2 groups of jacks. It steps through off and 6 divisions.

It quantises timing by holding notes. A note you play waits until the next line of the chosen division, so it can be up to one division late.

bpm and scale are editors

bpm and scale look like displays, but you can drag both.

bpm moves 1 beat per 2 pixels of drag, and writes the shared tempo, so the sequencer’s knob moves with it. It holds while a cable is in the clock jack, while this module is unlinked, or while a master somewhere else takes priority.

the input

The input block is to the right of the outputs, in the same top row. It is a grille with a lamp, and it reads the room from the audio-in bus. The in page draws the input’s last 10 seconds.

The audio jack, to the right of mod, carries the input after gain. It is a stereo pair. A stereo track or interface keeps both sides; a mono source fills both. The lamp and the in page read the pair as one level.

Click the grille to mute the input. The same switch is on the in page, beside gain.

clavier sends the note bus

The note bus is the third shared bus, after the transport and the scale. It works the same way: 8 lanes, and the first leader to send owns a lane until it goes quiet.

Each lane carries 16 voices.

the clock runs while stopped

While the transport is stopped, clavier’s clock keeps running at the leader’s tempo.

run still reaches clavier, but it only starts and stops the phrase.

a held arp, then strummed chords

hold	on
arp	on
mode	up-dn
rate	a 16th
octaves	2
gate	a little over half
swing	about 60%

Put a scale on the shared bus, from the quantiser or by dragging clavier’s own scale reading.

Play a triad and release the keys. hold holds back the note-offs, so the arp keeps walking. A key outside the scale plays the nearest note in the scale.

swing is also the phrase’s swing. A phrase you build later already swings to match.

Now turn the arp off and set:

chord type	7th
spread	1
octave	0
strum	about 30 ms
velocity	high
out	poly
poly	4

Play single keys, and each one plays a rolled chord.

record a phrase, then fix one step

On the seq page, set phrase 1, length 16, rate a 16th and timing nearest.

1. Press rec with seq stopped.
2. Play notes. Each release moves the cursor on.
3. For a gap, release every key and press rest.
4. To carry a chord into the next step, hold it and press tie.
5. Press rec again to close the take.

To fix a step:

1. Press rec again.
2. Before you play anything, click step 5 on the plot. The phrase and its length stay as they are, so this take is an edit.
3. Play the fix.
4. Close the take.
5. Press seq with the transport running to hear it.

4 drum lanes read from a map

grille is a drum computer with 4 lanes: kick, snare, hat and aux. Its steps are already written into a map.

Each bank holds a 5 by 5 grid of 32-step patterns. x and y place a reading point anywhere on that grid. grille works out the velocity of every step on every lane from the 4 grid points around the reading point.

Neighbouring positions share most of their hits. So when you sweep slowly across the map, one pattern turns into another by degrees.

ways to leave the map

The rest of the panel gives you ways to move away from the written patterns:

density	a threshold for each lane
chaos	flips hits on and off
brk	sends a lane’s reading point across the map for a while
rat	adds repeats and rolls
humanise	moves hits late
the fill	a window where extra hits are added
turing	decides whether all this randomness repeats or changes

what grille does at factory settings

At factory settings, grille plays the map and the 4 densities only. chaos, humanise, ratchet and break all start at 0.

The fill section starts armed but silent. Its window opens every 4 bars, but dur and amt both start at 0. Raise both of them to put a fill in the bar.

turing holds the dice. It acts once chaos, humanise, ratchet, break or the fill amount is above 0.

click and drag on the map

Press anywhere on the map to set x and y to that position. If you keep dragging, the position moves from the point you pressed. The drag follows the pointer at any zoom level.

x and y are also the first 2 knobs in the band below the map. They work like any other knob.

how neighbouring grid points combine

The 25 grid points are separate patterns. Between them, grille chooses.

Where a column sits between 2 written variants, each step picks one variant or the other. The chance of each depends on how far across the column the point is. The rows then thin or thicken the pattern. Finally, every velocity that remains moves by up to 12 either way, at random.

chaos flips hits both ways

chaos sets the chance that a hit flips, rolled for each step on each lane. At full, it flips about 1 step in 5. At a third, it flips a little under 1 step in 10.

A flip can add a hit or remove one. So the density stays about the same, while the placement moves.

humanise pushes hits late

humanise delays hits by a random amount. Every hit it moves lands behind the grid. It adds to swing.

At full, it can push a hit back by more than half a step. The bottom quarter of the knob covers the small delays.

turing repeats or renews the dice

Every random decision grille makes inside one window comes from a single seed. turing is the chance that the next window uses the same seed again.

full	the dice stay locked, and the pattern repeats for ever
0	new dice every window
about four fifths	about 6 bars in 7 reuse the previous bar’s dice, so the pattern develops one bar at a time

The window length is set in the menu. By default it is 1 bar.

run waits for the bar

The lamp at the top left is run. By default, when you start grille again, it waits for the next downbeat.

While it waits, the step counter moves on silently, one step per tick, until the next step is a downbeat. This is the menu setting run restart waits for the bar. Turn it off to restart straight away, out of phase with the bar.

choose a bank

The text in the middle of the top row is the bank. Click its left half to go back, or its right half to go forward. The list wraps round.

There are 7 banks: émilie, jungle, alt club, riddim, 4/4, glitch and og. og holds 3 voices.

The bank is a real parameter with fixed steps. You can map and modulate it like any other control.

record a movement on the map

The button on the right of the top row records what you do on the map, and plays it back. Each click moves it to the next state:

idle	nothing recorded
recording	red and glowing
looping	red and lit, steady. Click again to return to idle and clear the recording

density sets how much of each lane plays

The panel has 4 columns of 4 knobs, with one knob per lane in each column.

density is a threshold. At full, every cell in the map that has any velocity plays. At 0, only a cell at maximum velocity plays.

As you turn density down, the ghost notes go first, then the ordinary hits. The accents stay longest.

The 4 lanes start at different values. The kick starts high, the snare lower, the hat high again, and the aux lowest.

rat and brk leave the written pattern

The ratchet column does 2 separate things, at different settings. Both run all the time. The setting changes how often each happens, and the trailing hit is always the more likely.

low settings	after a hit, a silent step is filled at a lower velocity. The groove gets thicker but keeps its shape
higher settings	a real double hit is added at the half step, at three quarters of the velocity. You hear it as a roll

brk is the chance, each bar, that a lane’s reading point leaves for a trip across the map. While away, the lane rewrites about 3 hits in 10.

fills every few bars, or by chance

The fill section has 4 fields, a button and a jack.

A fill due every 4 bars happens every 4 bars, always. chc adds extra fills, at random, on top of the scheduled ones.

For fills that only happen sometimes, set evry to 0 and use chc alone.

run starts and stops the whole system

With clk empty and the system’s transport live, grille follows the system with no cables:

steps	on the system’s 16th notes
reset	the system’s reset
step length	worked out from the system’s tempo

The link works both ways for run. Press run on grille, and it starts or stops the shared transport.

grille reads the tempo from the system.

play the map, then add the dice

bank	émilie
x and y	factory position
densities	factory values
everything else	0
all 4 lengths	32

Drag on the map. Left is the first written variant and right is the fourth. Up thins the pattern and down thickens it.

Then add the dice:

chaos	about a third
turing	four fifths
turing length	1 bar
humanise	low
ratchet	about a fifth, on all 4 lanes

Now a little under 1 step in 10 flips, and those hits land slightly late. About 1 step in 8 gets a trailing hit. Most bars repeat their dice into the next bar.

turing needs chaos, humanise or ratchet above 0 to have anything to hold.

fills, and lanes that drift apart

evry	4
chc	0
dur	6
amt	just over half
break	0 on every lane

The last 6 steps of every fourth bar now get an extra hit on about 1 silent step in 4. The lamp beside the fill row lights while the window is open. Set chc to a quarter to add extra fills at random, on top.

A fill needs both dur and amt above 0.

Next, set the lengths and break values:

kick	length 32, break 0
snare, hat, aux	lengths 12, 7 and 10, with rising break values

The kick holds the beat. The other 3 lanes loop at lengths that do not divide 32, so the groove comes apart and comes back together every 2 bars.

The break values send hat and aux away from the map for a bar or two. While away, they rewrite about a third of their own hits. To see these excursions, set lane crosshairs to always.

8 gate patterns, each drawn as a wheel

volet is an 8-channel gate. Its name means a shutter: slats that open and close to let light through.

Its face is a wheel. The angle is the position in the bar, and the distance from the centre is how far the gate is open. Every control on volet is a drag on this one display, including the wheel and the value fields.

The curve on screen is the curve the gate sends. It includes the smoothing, and it is flipped if the channel is inverted.

With poly audio on, the audio is polyphonic. There is one stereo input and one stereo output for all 8 channels, plus a separate input and output for each channel.

the envelope stages are curves

Attack, decay and release are exponential curves, like an analogue envelope. The rise eases into the peak. The fall drops fast, then slows.

Each stage still reaches its corner exactly, so the stages join smoothly.

Straight ramps are also available for each channel. Choose them in the right-click menu, next to the output mode. lane shapes sets all 8 channels at once, and shows exponential, linear or mixed.

one curve feeds every output

Each channel has a mix. The module also has a master amount, on a jack. Both set how far the gate is allowed to close.

They work the same way on the audio outputs, the lane outputs and the gate output. So a depth you set by ear on the audio gives the same result on the voltage.

run, tempo and the space bar

Click the green light to start or stop volet.

The space bar does the same while the pointer is anywhere over the module, including over the jacks. It only works with no modifier keys held.

Clicking the light adds an undo step, because run is saved with the module. The space bar leaves the undo history alone.

Drag the rest of the title strip to move the module.

8 channel tabs

Click one of the 8 tabs to choose a channel. The wheel, the 8 value fields and the wheel’s right-click menu all switch to that channel.

Each channel has its own settings:

pattern and length	the steps, and how many of them play
res	the length of each step
envelope	attack, decay, sustain and release
slew, shift, mix	smoothing, rotation and depth
invert, output mode	flips the curve, and sets what the lane output sends

All 8 channels take their timing from one shared counter. So channels at different resolutions stay locked to one grid, and none drifts during a long set.

Each channel starts with a pattern already drawn. A new volet plays straight away.

switch a step on or off, and set its level

Click the wheel without moving the pointer to switch a step on or off. Drag a segment to set its level: away from the centre raises it, towards the centre lowers it, and the segment’s edge follows the pointer.

Each gesture leaves the other alone. Switching a step off keeps its level. Switching it on restores that level, even if the level is 0.

tie steps together

Each open segment has a short grip just inside the rim, at its trailing edge. The grips appear only while the pointer is over the wheel, or while you are dragging a tie.

1. Drag a grip clockwise. The segment extends over the steps after it.
2. Drag it back to release those steps.

The steps under a tie keep their settings, so they return unchanged when you release them.

the step count and inv

The large number in the centre is how many steps this channel plays, from 4 to 32. Drag anywhere inside the centre to change it, 1 step for every 7 pixels. inv is also in the centre.

The channel always stores 32 steps. If you reduce the count, the extra steps are kept. Increase the count again, and they return as they were.

When you change the count, the channel’s shift is adjusted to fit the new length. res sets how long each step lasts.

the shading behind a segment

The shading behind an open segment fills only the space between the curve and the rim. It shows the part of the step where the gate is closed.

A step at full level has no shading. A tie shows one long shaded area across all of its steps.

the wheel’s right-click menu

A right-click on the wheel does one of 4 things, depending on where you click.

It is the only way to copy a single value to the other channels. For example, to give all 8 channels the same envelope but keep their patterns:

1. Set one channel’s envelope and slew.
2. Right-click each of those values in turn, and copy it to the other 7 channels.

The patterns stay as they were. To copy the pattern as well, use copy this channel to all in the module menu.

the 8 value fields

The 8 fields under the wheel belong to the channel you are viewing. Drag a field up or down to change it. The continuous fields cover their full range in about 160 pixels.

Every channel starts as a hard gate:

att	0
dec	67%
sus	100%
rel	0

envelope stages are part of a step

The 4 envelope stages are set as a share of a step. So a shape stays the same when you change the tempo or the resolution. A 6% attack is 6% of the step, whether the step is a quarter note or a 32nd note.

attack plus decay over 100%	both are scaled down to fit the step
a stage set to 0	the stage is skipped, so a fully open pattern holds steady instead of dipping at each step
release	runs on after the end of the step. volet uses the higher of the previous step’s level and the current one, so a long tail carries into the next step

res and slew

res sets the length of each step. It has 6 values, and stops at both ends.

In order, the values are a quarter note, an eighth note and an eighth note triplet. Then come a 16th note, a 16th note triplet and a 32nd note.

The length of the bar is the step count times res. This is the only thing that sets how fast the wheel turns. For example, 16 steps of a 16th note is one 4-beat bar. 12 steps of a 16th note is three quarters of a bar.

By default, every channel uses a 32nd note.

slew is the only value set in seconds.

shift and mix

shift rotates the pattern. It wraps round at the channel’s own length. The step at 12 o’clock changes, and the pattern stays the same.

mix sets the depth of the gate.

shared run and tempo

When volet is linked, it reads and writes the tempo and the run state on the transport bus. Every linked module follows them.

Press run here, or tap space with the pointer over volet, and every linked module starts. The sequencer treats this as play from the top.

volet also takes its position in the bar from the bus. If either transport is stopped, the wheel holds where it is.

volet takes reset from the bus.

the wheel runs before you patch

The wheel is drawn from the pattern, so it moves as soon as you add the module.

a 16th note gate, and a polyrhythm

tab	1
steps	16
res	a 16th note. The default is a 32nd note, so drag down 2 values
att, dec	0
sus	100%
rel	0
slew	0
mix	100%
inv	off
poly audio	off

Patch audio into the main input, and take it from the main output. At 120 bpm, the bar is 2 seconds long. With poly audio off, the main output sends only the current tab.

Then set tab 2 to 12 steps of a 16th note. Take both channels from their own lane outputs in row 2.

The 2 channels share one counter, so they stay on one grid. They drift apart, and line up again after 3 of tab 1’s bars.

a swell and a duck

Use these to shape a square gate into a slow swell, or into a sidechain-style duck.

For both, set attack to 0, decay to 0 and sustain to 100%. The envelope is then a plain square, and slew and invert do the work.

For the swell:

res	an eighth note
steps	8
rel	0
mix	100%
slew	60%

At this setting, slew is a 90 ms time constant against a 250 ms step. The gate almost reaches full level, and the wheel shows the same limit. Turn slew up to 100%, and the pattern smooths into one slow rise and fall.

For the duck:

inv	on
res	a quarter note
steps	4
pattern	one step on at full level, tied across all 4 steps with its grip
rel	40%
slew	25%

Invert comes after the slew, so a fast edge stays a fast duck. The release runs past the end of the tie, so the gate recovers smoothly instead of jumping back.

For a sidechain voltage that matches the audio, including its depth, right-click that channel’s lane output and set it to CV.

16 stored patterns for xtrack

xarrange is a 4 hp arranger. It fits flush against an xtrack.

It holds 16 scene slots. Each slot is a complete, separate copy of a pattern: all 8 tracks, every trigger, every off marker, and every length, speed, direction, key, mute and name.

how a scene change works

A scene change is a swap. xarrange puts the pattern in the tracker back into the slot it came from. It then copies the new slot into the tracker. All 8 tracks are ruled again at the tracker’s current resolution, so both views redraw at once.

Editing scene 4 leaves scene 5 as it is.

The tracker plays the patterns. xarrange moves them, and writes straight into the xtrack beside it.

when a scene change happens

A scene change is a request. You can make it at any time, and the quantise grid decides when it happens. A request made most of a bar early still lands on the line.

Requests wait in a queue while the tracker is running. When it is stopped, click a row and the scene changes straight away. Setting quantise to off does the same.

how to read the panel

the list	a tall dark rectangle of 16 rows, under the title
CV	a jack below the list, with the quant field under it
scene	a knob below the list, with the chain field under it

every row is a control

Left-click a row to set the scene knob. A click is the same as turning the knob, so it is a request that goes through quantise.

Each row takes its colour from the 8 track colours, in order. Scenes 1 to 8 and 9 to 16 use the same 8 colours. The colour helps you tell rows apart.

the row menu

Right-click a row to open its clipboard menu. The menu shows the scene number at the top.

It acts on the row you clicked, whichever scene is playing.

the knob sets the target scene

Every frame, xarrange adds the knob and the CV jack together to get a target scene. It compares the target with the scene playing in the tracker.

If they differ, xarrange switches now, or queues the change and waits.

chain modes

chain has 6 modes: off, and 5 that play through the scenes on their own. The 5 match the directions a single track can read in, one level up. off is the factory setting.

The chain moves on each time track 1 wraps. A wrap is when track 1 goes from its last cell back to its first (or the other way, on a reverse track).

If track 1 reads ping-pong, random or brownian, the chain stays where it is.

loop scene holds one section

loop scene is in the right-click menu. It stops the chain moving on at each wrap, so the current scene repeats.

Clicks, CV changes, queued requests and triggers still change the scene as normal.

Setting chain to off loses the direction, and you then have to set it again. loop scene keeps it. When you turn loop scene off, the chain carries on forwards from the current scene.

1 bar is the default

The coarsest quantise setting is 1 bar, and this is the default.

At 1 bar, you can make a request a long way early and it still lands on the line. Finer settings give you less time. At a 16th of a bar, you are playing the change by hand.

how the grid is measured

A bar here is 4 times the tracker’s lines per beat. It is measured on track 1’s cursor, counting from track 1’s first cell.

The grid counts inside track 1’s own loop. This has 3 effects: …

coarse resolutions

fine settings become the same, because the shortest division is 1 line. At 1 line per beat, a quarter, an eighth and a 16th of a bar all mean every row

Keep the quantise within track 1’s loop, or the queued row waits for ever.

Turn the scene knob back to the playing scene to cancel a queued change.

xarrange docks on either side of xtrack

Every frame, xarrange looks for an xtrack on both sides, left first. It works against either edge of an xtrack. … With no tracker beside it, the list draws, the knob turns and the fields step, and the arrangement waits.

play sections by hand

chain	off
quant	1
the tracker	running, with track 1 reading forwards, 16 cells, at a quarter-note resolution

Click any row at any time. It is outlined in its own colour straight away. The playing row’s fill runs on to the end of the bar, and the swap happens there. This is the factory setup.

cut between scenes while the drums run on

… A swap leaves the cursor where it is.

Build the 16 scenes as variations of the same length. They cut between each other a beat at a time, and the drums run on. At 1/8 or 1/16, the cuts become a stutter with an arrangement under it.

With reset playhead on scene change left on, the same click restarts every track from its first cell. The swap becomes a section change.

an arrangement that plays itself

The chain plays through the scenes, and loop scene can then hold one.

chain	fwd
quant	1
loop scene	off

Fill the scenes with variations and press play. The arrangement plays 1 to 16 and back to 1, one scene per wrap, empty slots included. The scene knob follows. Track 1 must read forwards or in reverse.

Set chain to png, and the arrangement bounces 1 to 16 and back. Set it to brn, and it drifts forwards. Now and then it stays on a scene for 2 passes, or goes back one.

Now turn loop scene on. The chain keeps its setting, but the current scene repeats for as long as loop scene is on. Turn it off, and the chain carries on forwards from where it is.

To stop the arrangement in one click, set chain to off. This also clears the direction, so you have to set it again afterwards.

8 tracks in one pattern

@view 0 xtrack holds one pattern of 8 tracks. You can look at it in 3 views: the tracker grid, the piano roll and the overview. The view button on the left of the panel steps through them.

The 3 views show the same notes. A note you draw in the piano roll is in the grid, and a note you type in the grid is in the piano roll.

Every note is a trigger at an exact point in time. There are 960 ticks to a quarter note. The resolution knob divides those ticks into rows. A pattern typed at one resolution and read at another is the same music, with a different number of lines to look at.

Each track holds up to 128 rows and has 7 output jacks of its own, 56 in all.

what belongs to a track, a step or the pattern

@view 0 Some settings belong to a track, some to a single step, and a few to the whole pattern.

per track	the row count, the speed, the read direction, the root and scale, the cv1 mode, mute, the name, and the smooth and random amounts of the automation lanes
per step	the pitch, the gate length, glide, roll, chance and both control voltages
global	the tempo, the resolution, which view is on screen, and which track is selected

Each of the 8 tracks has its own copy of the per-track settings. Each per-step setting belongs to one note only.

rows, tracks and cells

@view 0 The tracker grid reads from top to bottom, like a list. It is the first view.

row	one line of the grid, and one step in time. Rows are numbered down both sides, starting at 00
track	one column. There are 8, named TRACK 01 to TRACK 08 along the top
cell	one track at one row. A cell holds at most one note, with its settings
beat	every few rows a band is shaded darker. Each band starts a beat. At res 1/4 there are 4 rows to a beat
playing row	the row with the blue wash. The pattern scrolls past it while it plays

The selected track is wider than the others and has blue corner marks. It shows 2 more fields than the other tracks.

the parts of a cell

@view 0 A cell reads from left to right in 4 parts. The selected track adds 2 more.

note	3 characters: the note name, then a sharp or a space, then the octave. C#4 is C sharp in octave 4. C 5 is C in octave 5, with a space where the sharp would go
len	2 digits: how many rows the note lasts. 04 lasts 4 rows. 00 is a tie: the note holds until the next note in the track
FX1 and FX2	2 effect slots. Each holds a letter and a 2-digit value. C 60 is a 60% chance
cv1 and cv2	2 digits each, from 00 to 99. These show on the selected track only

Dashes mean empty: --- is an empty note part, and ---- an empty effect slot.

examples of whole cells

@view 0 Here are some cells as the grid shows them, and what each one plays.

a plain note	C 3 04 ---- ---- is C in octave 3, for 4 rows
a tie and a glide	A#2 00 G 50 ---- is A sharp in octave 2, tied, and gliding from the note before it
repeats and chance	E 4 01 R 04 C 50 is E in octave 4 for 1 row, played as 4 fast repeats, half of the time
a chord	C 3 08 qM7 ---- is a C major 7 chord: the note and 3 more voices, for 8 rows
a late note	G 3 04 m 70 ---- is G in octave 3, played a little late
an empty row	--- -- ---- ---- is empty. A note from a row above can still be sounding

The track’s own settings, such as its speed or scale, are on the deck and in the track menu.

the effect letters

@view 0 An effect is a letter and a 2-digit value. Type the letter into an effect slot, then the value.

C	chance. The value is the percentage of times the note plays. C 50 plays about half of the time
R	roll. The note repeats this many times inside its row. R 04 is 4 fast repeats
G	glide. The pitch slides from the note before, and the gate stays open. Any value above 00 gives the same glide
m	micro timing. 50 is on the row. Lower values play the note early, and higher values play it late
F and J	slide up and slide down. The pitch slides this many semitones across the row. F 12 slides up an octave
q	chord. The value picks the chord: 01 major, 02 minor, 03 seventh, 04 major 7, 05 minor 7, and more up to 39. The slot then shows the chord’s name, such as qM7
i	inversion, used with a chord. i 01 moves the lowest voice up an octave, and i 02 moves the lowest 2
!	off. The note stops at this row, even if its length runs on

chance, roll and glide can also be set with the deck’s knobs. The knobs write into the same 2 slots.

type a note, a length and an effect

@view 0 Click a part of a cell to put the cursor on it, then type.

1. Click the note part of a cell.
2. Type C, then #, then 4. The cell now reads C#4. The octave digit writes the note.
3. Click the len part and type 0 then 8. The note now lasts 8 rows.
4. Click the FX1 letter and type R. The slot reads R 02, a default value.
5. Click the value next to it and type 0 then 4. The slot now reads R 04.

Digits come in from the right, so typing 4 then 7 gives 47. Backspace removes the last digit, and Delete sets the value to 00.

In the note part, type 1, or the key to the left of 1, for an off, and press Delete to clear the cell. The grid draws an off as an empty cell, because a note’s length already marks where it ends.

set the tempo, or follow a clock

@view 0 bpm is a number you drag. It has 2 zones:

20 to 240	xtrack runs its own clock at that tempo
below 20	the field reads ext, and xtrack moves on at each rising edge at the clk jack

set the resolution

@view 0 res steps through 1, 1/2, 1/3, 1/4, 1/6, 1/8, 1/12 and 1/16.

In the grid, res sets the rows per beat. The internal pulse rate becomes the tempo multiplied by that figure. The pattern is divided into more or fewer rows, and note lengths change to keep their musical length. In the other views, res only draws the grid lines.

run, rec, clear and view

@view 0 run, rec, clear and view are buttons. The space bar presses run.

Press run while xtrack is running, and it stops where it is. Press run again, and it starts from the top.

To stop and return the cursor to the top in one action, use stop & rewind in the right-click menu.

view steps through the tracker grid, the piano roll and the overview.

count yourself in before recording

@view 0 Press rec while xtrack is stopped, and you get 4 beats before anything is recorded. The rec button flashes each beat, and you hear a click.

On the fifth beat, rec turns on and xtrack starts playing from the top. So you can count yourself in and play the first note in time.

1. Stop xtrack.
2. Press rec.
3. Play along with the 4 clicks, and start on the fifth.

The count follows the tempo the system is on. The first beat is a higher click, held a little longer, so you can hear and see where the bar starts.

press rec while running	records straight away
press rec during the count	cancels the count
count in before recording	in the right-click menu. Turn it off and rec turns on the moment you press it
metronome	in the right-click menu. How loud the click is, from 0 to 100%. At 0 the count runs silently, and the button still flashes

12 knobs in 2 banks

@view 0 The deck on the right has 6 rows of 2 knobs, with a value strip down the middle.

pattern bank	the top 2 rows, 4 knobs. They edit the selected track. Select another track, and all 4 reload from it
note bank	the 4 rows below, 8 knobs. They edit the note you picked: the note under the cursor when you clicked or typed your way to it

While the transport runs, the cursor follows the playing row. The note bank stays on the note you picked.

steps, speed and read

@view 0 steps, speed and read set the shape of one track. Every track keeps its own.

At a speed of a half, the track moves on every second pulse. Above 1, it fires extra pulses inside each master pulse.

Brownian reading is a walk that leans forwards. It usually moves on, but sometimes goes back and sometimes stays.

scale

@view 0 scale is off by default. It offers none, or 12 roots in 5 modes.

Right-click the knob for none, root and scale. A track’s right-click menu offers more modes than the knob reaches.

While xtrack is linked, every track uses the system’s key.

length

@view 0 length is counted in rows. The stored value, the width in the piano roll and the gate’s countdown are all the same number. So when the tempo changes, the note changes length with the grid.

A length of 00 is a tie: the gate holds until the next note or the next off. Lengths go up to 99 rows, although patterns can run to 128.

glide, roll and chance

@view 0 Each cell has 2 effect slots. roll and chance share one, and chance wins when both are set. glide takes the other.

The pitch slides from the previous note across the whole step, on a soft curve. The gate stays open.

how the knobs and the pattern stay in step

@view 0 The knobs and the pattern follow each other. Whichever moved last wins.

A knob turn writes to the track or the note. A change from anywhere else pulls the knob to match on the next frame. This includes the track’s right-click menu, the overview’s lanes, a scene change from xarrange, and undo.

If a track uses a scale from the track menu, such as Lyd or Blu, the track keeps it and the scale knob stays where it is.

When you select a box of notes in the piano roll, the note bank writes every selected note, and only those. octave and note then move the selection by the distance you turn.

draw a note with the mouse

@view 0 To draw a note, command-click or control-click an empty cell. You can also double-click it within about a third of a second.

The note appears at C3 with a length of 4 rows, which is 1 beat at res 1/4. You hear it as it lands.

the playing row and dynamic mode

@view 0 The playing row stays 8 rows down from the top of the grid. The pattern scrolls past it, so you see 8 rows that have played above it and 8 rows still to come below.

While the transport runs, the highlight and the typing position move with the playing row.

Dynamic mode frees the cursor. xtrack switches to it when any track’s length, speed or read differs from track 1, until the next start from the top.

amber dots and chords

@view 0 An amber dot in the bottom-right corner of a cell means the row holds more than this resolution can show. Either more than one note starts inside the cell, or a note in it starts partway through the row. The same dots run along the top of the automation plot.

A chord in FX1 is shown by its name, such as qM7, in teal.

where your typing goes

@view 0 When the pointer rests over the grid, you can type into it. xtrack takes keys from the part of the panel under the pointer, as well as from the part that has focus.

Any left click on the module also gives it focus.

2 ways to enter notes

@view 0 There are 2 ways to type notes.

rec off	type the note’s name. A letter from A to G, then # or b if you want one, then the octave digit. The digit writes the note
rec on	the keyboard is a piano. Each key is one note in the current octave. It goes in at the cursor, and the cursor moves on by the step jump setting

With rec on and the piano layout, Z is octave down and X is octave up. On the tracker layout, Z and X play notes.

A midi keyboard works like the piano keys: with rec on, each note you play goes in at the cursor. To play another module from the same keyboard with rec on, turn off midi notes enter the grid while rec is on in the right-click menu.

copy and paste

@view 0 The clipboard keys copy one cell in the grid and one note in the piano roll.

To copy a whole track, right-click its column in the grid, or its tab in the piano roll, and choose copy track contents. Then right-click another track and choose paste track contents. The notes, their settings and the track’s own settings all move across. The track you paste into keeps its name.

the second view

@view 1 The piano roll shows one track at a time. Pitch runs up the page and time runs across it. Each note is a bar as long as its length.

The tabs along the top are the 8 tracks. Click a tab to show that track.

The strip on the left is drawn as a keyboard, light for naturals and dark for sharps, with every C labelled. Other tracks show faintly behind the current one.

find your notes

@view 1 When you open the piano roll, or switch to another track’s tab, the view centres on that track’s notes. Leave and come back, and it centres again.

The mouse wheel scrolls through pitch, across 128 semitones around middle C.

resize and move notes

@view 1 Point at a note or select it, and it gets a blue outline and thick white grips at both ends. Drag a grip to change the note’s length, or drag the middle of the note to move it.

When 2 notes start on the same row at different pitches, each has its own grips. Point at the one you want.

A note 1 row long keeps a thin strip at its right edge for resizing. The rest of it moves the note. When you move a note, everything it owns moves with it.

draw or select

@view 1 To draw a note, use the same command-click or double-click as in the grid.

Click and drag on empty space to draw a selection box, or shift-click notes to add them. The note bank then edits every selected note. Hold cmd while you drag a note to drag a copy.

the third view

@view 2 The overview shows all 8 tracks at once, as narrow lanes. Each note is drawn at its length, the same span as its len in the grid and its width in the piano roll.

playheads	each lane has its own playhead, moving at that track’s speed
number and name	click to mute the lane
a step	click to turn it on or off
grid ticks	fall on the beat, at the grid’s resolution

Here the mouse wheel moves the edit cursor.

the lane rail and the grid cursor

@view 0 The strip under the pattern has 6 lanes, listed in a rail on the left. The piano roll shows the last 4. The ✱ and ∿ values above the plot add random offsets and smoothing to PITCH, CV1 and CV2.

Clicking a row in the rail does the same as moving the grid’s cursor sideways to that field. Each one selects the other.

paint and erase

@view 0 Click and drag to paint. Values between the points you paint are filled in.

A fast stroke fills every column it skips. If you drag past either edge, the value holds at the first or last column. One stroke is one undo step.

To erase, shift-drag.

xtrack leads the system’s transport

@view 0 At every sample, xtrack sends 4 things to the shared transport: a tempo, whether it is running, a reset count and a position in 16th notes.

It sends 16th notes, because res changes the pulse rate but keeps the tempo. The position is worked out from the tempo, so followers move smoothly even when an external clock is unsteady.

The first module to send owns the transport. A second xtrack only takes over after the first has been quiet for a while. If you remove one, it hands the transport back at once.

Any cabled input in the system can also drive xtrack. Right-click the jack and choose drive the system, then tempo, run or reset. What arrives at that jack also goes to xtrack’s own jack, and the jack gets a blue ring.

the key, xarrange and track voices

@view 0 xtrack reads the system’s key, xarrange docks beside it, and each track column is a whole voice.

While xtrack is linked, the system’s key is every track’s key, even for a track set to none. The scale field shows that key next to a chain mark. When you unlink the module, each track’s own scale applies again.

3 lengths at once, and the system’s clock

@view 2

1. Select track 1 and set steps to 16.
2. Set track 2 to 12 steps and track 3 to 15.
3. Leave all 3 at speed 1, reading forwards, at res 1/4.
4. Press run.

The overview shows the 3 lanes drifting apart. As soon as the lengths differ, xtrack switches to dynamic mode, so you can edit anywhere while it plays. Patch each track’s eoc to a drum trigger to hear the end of each phrase. The 3 tracks line up again after 240 rows.

Set bpm to 124 with res at 1/4. With the transport link on, every module in the system with an empty clock jack runs at 124.

Turn res to 1/8, and the system’s tempo stays the same. The internal pulse rate doubles, the pattern has twice as many rows, and notes keep their length.

To follow another clock, drag bpm below 20 until it reads ext. Patch a clock into clk with one pulse per row. xtrack measures it and sends the tempo on.

a chord, and repeats with chance

@view 1 Both come from the 2 effect slots in a cell.

1. Put a note at row 00 of track 1: C 3 with length 04.
2. Click the FX1 letter and type q.
3. Click the value and type 0 then 4. The slot shows qM7, and the piano roll shows 4 notes stacked on that row.
4. Click the FX2 letter and type i, then the value 0 then 1. The lowest voice jumps up an octave.

Patch that track’s gate and v/o to a polyphonic voice. You get 4 channels and 4 envelopes, and you can drag each note separately in the piano roll. Set chance to 75 on the lowest note, and the whole chord takes that chance.

The same 2 slots hold repeats. Set roll to 4 with the deck knob for 4 repeats in one row. Now set chance to 60, and the roll disappears, because both use the same slot.

To keep both, type them into separate slots: C 60 in FX1 and R 04 in FX2. The trig output then gives 4 clean edges per row, whatever the gate does.

shifting by hertz

… Transposing multiplies every partial by the same ratio, so the intervals between them stay the same. After biais, a harmonic series is no longer made of whole multiples of one fundamental.

A shifted note stays close to its original pitch, because the fundamental moves only a little. But it loses its tuning, because the overtones move by the same number of hertz.

Small shifts thicken the sound and make it beat. Large shifts turn pitched sounds metallic. The shift can go up or down, and it can be exactly zero.

a delay with the shifter in its feedback loop

The second half of biais is a delay line. The shifter sits inside its feedback path, so each pass through the loop is shifted again by the same amount.

The result is a staircase, with one step of shift per lap. Each repeat arrives one delay time later, and one shift further up or down. It keeps climbing until the feedback runs out.

shift	sets the size of one step
dir, spread and the delay time	set the direction and speed of the climb

A damper and a soft clip sit after the feedback. Without them, a loop that shifts upwards would push all its energy to the top of the range in about 2 seconds.

what is on the panel

The top of the panel is a spectrum window. The window is also a control. Under it are 9 controls and 12 jacks. … CUT (L1272–1277)

set the shift

shift runs from about 2,000 Hz down to about 2,000 Hz up. At dead centre, the shift is exactly 0.

The middle 40% of the knob stays within 4 Hz either way. Outside that region, every 2% of travel doubles the shift.

drag the window to set the shift

The window covers the whole top of the module. Drag anywhere in it to change the shift, at about 17 Hz per pixel. The system’s zoom scales this.

the shift CV changes with the knob position

The shift CV jack works in 2 ways. The position of the shift knob decides which.

in the beating region	the jack adds an absolute number of hertz
in the metallic region	the jack works as a ratio. At the ends of the knob, 1 volt is 1 octave

This is because a fixed number of hertz per volt is too small an interval at a shift of 2 kHz, and a fixed ratio never reaches zero.

set the delay in ms or in beats

ms and sync set the same delay time, in 2 different units.

ms	4 ms to 1 second
sync	a division of the shared tempo, in 8 named steps from a 64th note up to a quarter note

The field you touched last is the one in control.

switch between ms and sync

Both fields always show the same delay time. The field in control shows it in its own unit. The other field is greyed, and shows the same time in its unit. Both update all the time.

Touch the greyed field to put it in control. It starts from the value it was already showing.

By default, ms is live at 149 ms, and sync shows a triplet eighth, greyed. At 120 bpm, these are nearly the same length.

bal sets direct shift against the loop

bal sets how much direct shifted signal you hear against the delayed return. It uses an equal-power blend.

fully down

biais is a plain frequency shifter. The loop still runs behind it, out of earshot

… dir balances the 2 sidebands. … Each lap is shifted again, so feed sets how many steps the staircase has. … feed changes the structure of the sound. … CUT

At high settings, each returning pass arrives smeared, and a fast staircase blends into a wash.

dir chooses up, down or both

A frequency shifter makes 2 results at once: a copy shifted up and a copy shifted down. dir sets how much of each you hear.

dir’s ring grows out from 12 o’clock. The centre is a setting in its own right. There, both copies are present in equal amounts, so you can barely hear whether shift is up or down.

spread widens the blend between channels

Both channels are shifted by the same number of hertz, from one shared oscillator.

spread changes the blend each channel gets. The left channel leans one way from the centre that dir sets. The right channel leans the same distance the other way.

how spread changes the climb

Each lap adds one shift and takes one delay time, so the climb is a rate in hertz per second. How much of the shift counts as up depends on where each channel’s blend sits.

So opening spread changes how each channel climbs. In the last configuration, the 2 channels climb in opposite directions.

mix sets dry against wet

mix blends the dry input with everything biais has done to it. It uses an equal-power law.

At any shift other than zero, the dry and wet signals share no frequencies. They add in power, so a linear fade would lose 3 dB in the middle of its travel.

how the spectrum is drawn

The window draws the spectrum of each channel as bars. The bars grow out from a vertical centre line.

rose	the left channel
amber	the right channel
brightness	the comets

The range rises at once and falls back over several seconds. This stops a transient from flattening the picture that follows it.

the comets show the climb

Each channel has a comet: a bright head a few pixels long, with a trail of 20 pixels. It moves up or down that channel’s half of the window, at the speed the loop is climbing. The speed is compressed.

The comets show the direction of the climb before 3 of its steps have sounded.

the large number and the shift field

The panel shows the shift in 2 places.

the large number	at the top of the window
the shift field	below the window

Both show the shift as it is happening, including the CV.

slow beating, then a ring modulator

shift	about 2 Hz, just above the middle of the knob
feed	fully closed
bal	fully down
mix	50%

Nothing sounds shifted. The shifted copy beats against the dry signal twice a second, across the whole spectrum at once. The shifted partials are no longer in tune with anything.

Move shift down through zero to 2 Hz the other way. The beating stops at the centre, and starts again as the sidebands swap. The middle 40% of the knob covers this move.

To make a ring modulator, use these settings:

dir	centre, so it reads ring
spread	fully closed
feed	fully closed
bal	fully down
mix	100%
shift	about 300 Hz up

With the loop closed, biais is an amplitude modulator.

You can tune the carrier across the whole range with the knob. Outside the centre region, the shift CV gives 1 volt per octave, so you can play the ring modulation from a keyboard.

the staircase, then climbing both ways

shift	about 40 Hz up
ms	400
feed	85%
bal	60%
blur	10%
dir	the up end
spread	fully closed
mix	80%

Each repeat arrives 400 ms later and 40 Hz higher, for as many laps as the feedback allows. Both channels climb together at 100 Hz a second, and the comets show the rate. The staircase works in mono.

Lower ms to 30, and the same settings give a metallic resonator. … CUT (L1450’s last sentence, and L1452–1459)

On a mono input, spread creates width from a perfectly matched pair.

a vocoder made from a filter bank

crible is a vocoder built the traditional way. Bandpass filters listen to one signal, and envelope followers read their outputs. Amplifiers then apply those envelopes to another signal.

The bank works one sample at a time, so transients arrive intact.

A filter bank has a different cost: ring time. A filter narrow enough to resolve 1 semitone rings for about 67 ms at 80 Hz, and about 4 ms at 1.3 kHz. So the low end sounds slower than the top.

where the bands go is the instrument

A filter bank has to be told where to put its bands. That choice decides what crible is.

Spread the bands evenly across the octaves, and crible is an ordinary vocoder. Put them on the notes of a key, and it is a harmonic sieve.

In a sieve, sound between the bands is reduced twice. … CUT (L422)

1 semitone, and the default carrier

The module is built around 1 semitone. A semitone is just under 6% in frequency, so a filter that resolves one needs a resonance of about 17.

By default, the carrier is the input itself. A 5 V note on a band comes out at 5 V. A 1 V note comes out at 0.04 V, because its band opens 5 times less and the note itself is 5 times quieter. That squaring is what self-vocoding is. Two detuned oscillators that beat by 6 dB going in beat by 12 dB coming out. depth and sens are the 2 controls over it.

placement lays out the bands

placement is a menu setting with 2 options: log and scale. It decides how the bands are laid out.

log	bands spread evenly across the octaves
scale	bands sit on the notes of the key

Changing placement rebuilds the bank.

bands sets how many bands there are

bands is the number of bands, from 1 to 64. Drag the field to change it 1 band at a time. It means the same thing in both placements.

log	that many bands, spread evenly across the span
scale	that many notes of the key, spread evenly across the span

Under scale, the key only has so many notes inside the span. A minor key from 60 Hz to 12 kHz has 53. Ask for that many or more, and the field reads all. This is the starting setting.

Ask for fewer, and crible keeps notes from the whole span, bottom to top. A small number gives a sparse, hollow sound. You can change bands while a sound plays. Bands that stay keep sounding, and new bands fade in.

span sets where the bank starts and stops

The span sets the low limit where the bank starts, and the high limit where it stops. crible builds up to 64 bands.

Chromatic under scale is a useful setting. … If the mask fits no notes, you get 1 band at the low limit.

width sets how wide the bands are

width is each band’s width in semitones. It runs from 0.25 to 4, and starts at 1. The higher the number, the wider each band. It is a ring with its own cv jack, because it changes the tone and is worth sweeping. 10 V covers the whole travel.

A narrow band rings. At 0.25, each band is a needle that rings on after the sound. At 4, the bands smear together and just pass.

Each time the bank rebuilds, crible sets a trim so the output level stays the same. For example, 48 overlapping bands are much louder than 8, so they get more trim.

key sets the root and scale

The key field holds the root and the scale in one control. Turning it retunes the system.

It runs from chromatic through every root with 5 scale types: minor, major, pentatonic, harmonic and fifths. It starts at c minor. Click it to step to the root’s next scale.

Under scale, chromatic gives the dark comb.

the key input moves the sieve

The key input transposes the sieve. It moves the root and the notes together, to the nearest whole semitone.

It follows 1 V/oct. A twelfth of a volt moves the sieve 1 semitone, and 1 V returns it to where it started.

The key input acts after crible reads the scale bus. So a sequence here moves the sieve around whatever key the system is in.

the carrier strip chooses the carrier

The carrier strip shows its 3 modes side by side, with the current one lit in its colour. Click a mode to choose it. It starts on self.

self	the input is its own carrier
ext	the carrier comes from the carrier jacks

In the third mode, tone, the carrier is a saw on the root of the key, 2 octaves below middle C. It follows the key field, the scale bus and the key input. This makes it a drone in key to put under a voice.

For a carrier that moves, patch it into the carrier jacks and select ext.

add noise for consonants

unvoiced mixes noise into the carrier. It is the field at the right-hand end of the carrier strip. Drag it up for more. It starts at 0, and it works in all 3 carrier modes. The unvoiced slider in the menu is the same control. … Its level is fixed.

attack and release for every band

Each band has its own envelope follower. All the followers share one attack and one release.

Fast attack and release make speech easy to understand.

A slow release holds each band’s last level as the source moves on. A phrase then leaves a chord behind it, in key, because the bank is tuned. The longest release is 600 ms. Even at that setting, the chord fades away.

sens sets what every band reads

sens is the gain into the detectors, in dB. It runs from -12 dB to +12 dB, and starts at 0 dB. You set it once for a source. … CUT (L512)

gate mutes quiet bands

gate closes any band that reads below a threshold. It starts at off. Above that, the threshold runs from -48 dB to 0 dB. 0 dB is the reading of a 5 V note.

A band is fully open at the threshold, fully closed 12 dB below it, and eased in between. So a fading sound closes its bands smoothly.

about -30 dB	removes the faint wash that every band picks up from its neighbours. Notes outside the key drop further
about -12 dB	keeps only the loudest few harmonics of each note
near 0 dB	almost nothing opens

sens raises what every band reads, so it also moves what gate closes.

mix sets dry against wet

mix runs from 0 to 100%. At 0% you hear the input. At 100% you hear only the bank. Its cv jack covers the whole travel with 10 V.

ring makes the sieve stricter

ring is a slider in the right-click menu. It runs from 0 to 100%, and starts at 0%.

ring narrows the bands on the way out, by up to 4 times. The bands that listen keep their width, so crible still reads each semitone.

a note in the key	hardly changes level. It rings on a little after it stops
a note outside the key	drops by up to 17 dB more

So ring sets how strict the sieve is. Turn it up when notes outside the key still come through.

depth sets how hard the bank bites

Each band reads its input against a 5 V note: a full reading is a gain of 1. depth is the power that reading is raised to.

0	every band’s gain is 1. The bank passes the carrier flat, a plain filter bank on the key
100%	the vocoder. Each band’s gain is its reading
200%	each band’s gain is its reading squared. The sieve bites twice as hard, in dB

With the input as its own carrier, a quiet partial is turned down twice: once by the carrier band, once by its own reading. So the output is thinner and quieter than the input, and more so as depth rises. Raise sens, or lower depth, to bring it back.

With mix at 100% and depth at 0, you hear the input through the bank. For the dry signal, turn mix down.

How strongly the sieve refuses notes outside the key depends on placement, width and depth.

formant shifts the carrier bands

formant shifts the carrier bands up to 2 octaves either way. CUT … formant is continuous. Any shift other than whole octaves leaves the key. For example, 7 semitones puts every carrier band between the notes.

formant costs some cpu.

what the window shows

The window sits under the first row of rings: depth, mix and formant. It draws every band in the bank as a slot, sounding or not. The grey empty slots show the shape of the bank.

A band’s colour is its pitch, on a smooth ramp down the frequency axis. When you move formant, the coloured bands slide across a spectrum that stays still.

bar height follows mix

The window draws one channel. A bar’s full height is the reading of a 5 V note, and its full height follows mix.

The bars shrink as mix goes dry. They grow by up to a third as ring comes up.

If the display looks too small, check mix and sens before you check the source.

crible leads and follows the key

crible leads the system’s key as well as following it. Turn the key field, and the quantisers, the arpeggiators and anything else that follows the scale change with it, with no cable.

crible and a scale module retune each other. When they disagree, the one set as master wins.

Everything else reaches crible on a cable.

sing a voice in key

placement	scale
key	a minor key
carrier	ext
bands	all notes
span	60 Hz to 12 kHz
width	1 semitone
mix	100%
ring	0%
depth	100%
attack	about 10 ms
release	about 150 ms

Patch a voice into the audio inputs and a saw pad into the carrier. Speech comes out harmonised, because only frequencies in the key survive. crible builds about 50 bands.

leave a chord behind the phrase

Start from the setting above, and change these:

release	its top, 600 ms
attack	about 6 ms
gate	about -12 dB
sens	about +6 dB
mix	100%
ring	about 60%

To hear what gate and sens each do, move one at a time.

pull a loop into key

With only the input and output patched, a drum loop or synth line comes out gently in key.

carrier	self
placement	scale
mix	100%
ring	about 35%
depth	about 80%
width	a little under 1 semitone
attack	fast, about 4 ms
release	long, about 300 ms

The notes in the key ring on after each transient. If the loop comes out quiet, raise sens.

vary the width

From this setting, or from any setting above, move only width.

At the bottom, the bands are quarter-semitone needles that barely overlap, and ring. At the top, they are 4 semitones wide and smear together.

The level stays the same, because the trim moves the opposite way.

a filter made of many resonances

essaim is a stereo filter. Instead of a cutoff and a slope, it uses a crowd of resonant peaks, called nodes.

placed nodes	7 nodes that you place by dragging them on the curve
extra nodes	up to 25 more, which the rake macro adds on a harmonic series

The filters run in series, so the cascade can have as many as 32 peaking sections.

essaim is 7 hp wide. The 8 fields along the bottom are macros. Each macro moves every node at once: all their frequencies, all their widths, or the tilt across all of them.

The curve on the display is worked out from the filter’s own settings, so what you see is what you hear.

morph blends 2 whole filter states

morph is a straight blend between 2 complete states, called a and b. You fill both ends yourself.

A state is everything the filter is set to:

node frequencies	the 7 placed nodes’ frequencies
node gains	the 7 placed nodes’ gains
macros	all 8 macro values

As you move from a to b, the frequencies move in octaves, so a sweep stays musical. The gains and macros move in a straight line.

step through the presets

The box under the title shows the current preset.

click the box	steps forward
click the left quarter of the box	steps back

The list wraps both ways. To flick through all 10 presets, click the box repeatedly.

the 10 presets

flat	the placed curve near neutral, with a gentle lean
lowpass	shift at -80, tilt at -85 and bite at +95: a real lowpass
highpass	the mirror of lowpass
bandpass	a hump
notch	a scoop, like a phaser you move by hand
comb	rake full and tooth full, so every harmonic is a tooth, with 9 dB of teeth. shift moves the root, and every tooth follows
ring	the same series stretched by 4%, like a bell, with pinch higher, so the sound starts ringing
scoop	the placed nodes cut in the middle and lifted at the edges
vowel	the 7 nodes on formant positions
swarm	nodes scattered wide. The series is crowded by 6%

A preset fills b, so a sound you captured in a stays there.

Loading a preset sets morph to 100%. It also hands every macro back to the states, and releases any macro you set by hand.

drag the dots on the curve

The curve has 7 dots, one for each placed node. When you drag a dot, you edit the end of the morph that the bar is nearer to.

Both axes allow for the system zoom, so zooming in gives you finer control.

above 50% morph, the dots you grab follow a

The dots on screen switch from a’s positions to b’s at 50% morph. A drag writes to the matching end.

To find the dot you grabbed, essaim uses a’s positions. So above 50%, press where a’s dot sits.

the faint shape under the curve is the output

Under the curve is a dim filled shape, on the same 20 Hz to 20 kHz scale. It is a live spectrum of the output, left and right summed. It still moves with nothing patched to the output.

The curve shows what the filter is set to. The shape underneath shows what it has just done to the audio. When a resonance bites, you see a peak directly below the node that made it.

the morph bar

The thin strip across the top of the display is morph.

1. Press anywhere on the strip. morph jumps to that point straight away.
2. Keep dragging. The bar now moves relative to where you pressed, across its width.

While the pointer is on the strip, the bar shows its name and value between the letters a and b. The value is the morph the filter is actually using, which can differ from where the bar is set.

qmap can map morph: arm a slot, then press the bar. To automate morph, use its CV jack.

capture the current sound into a or b

The letters a and b at the top corners of the display are buttons.

Click one to write the current sound into that end. morph then moves to that end. The current sound is the blend the filter is on, including any macro you set by hand, so the sound stays the same.

the 8 macro fields

Under the display are 8 number fields that you drag. Each has a CV jack: the top row’s jacks sit above it, and the bottom row’s jacks sit below.

When you move a field by hand, it takes over from the morph at once.

at rest	a field reads 0. scale rests at 1 instead
at the ends	a field reads +100 or -100

pinch goes further up than down, and spread goes further down than up. The panel shows each reading as a percentage, so it tells you how much of the range is left.

The fields are real parameters. You can drag them, double-click to return to rest, right-click to type a value, and use click-to-learn mapping.

shift moves every node

shift moves every node, up to 2 octaves either way. On the pass-filter presets, shift is the corner frequency, and it is the macro that the morph moves most.

shift applies after rake has bent the spacing, and before spread moves it. On a comb, shift moves the root, and every tooth follows. On a scattered set, it slides the whole set without changing the arrangement. A node you dragged somewhere unusual keeps its place relative to the others.

pinch sets the resonance of every node

pinch sets the resonance of every node at once. Its range goes far beyond an equaliser’s. spread works around the centre.

On a harmonic series, the teeth are spaced by the root. So each tooth must be narrow compared with its own frequency: it needs a width setting of about 35 to be a tooth at all.

Anything wider is a bump. This is the difference between resonances that colour the sound and resonances that ring.

scale changes only the placed nodes

scale multiplies the gains of the 7 placed nodes. It leaves the extra nodes from rake at their own level. So a comb keeps its teeth while you change the shape underneath.

rake adds a harmonic series

rake does 2 things. It bends the node spacing away from where you placed the nodes, towards whole multiples of the lowest node: a harmonic series. It also adds up to 25 extra nodes at full travel.

The extra nodes are extra filters, so they cost more processing.

tilt leans one end only

tilt adds a ramp of up to 20 dB to the end being kept, and leaves the other end alone. It is one-sided because a pass filter is flat in the band it keeps.

bite turns the bells into a real filter

bite multiplies every gain, by up to about 2.5 times. It also fades the 2 outermost nodes from bell shapes into shelves.

A bank of peaking filters passes everything at full level outside its bumps. So beyond the last node, the response comes back up, where a shelf would keep going. For example, tilt at -100 alone gives a lean. Add bite at +100, and you get a real lowpass.

tooth, stretch and teeth

Under the 8 macros is a row of 3 fields. Each sets a rule that every node follows at once. This is what lets you dial in an exact comb or bell.

tooth	how fast the resonance narrows with a node’s place in the set. At 0, every node has the same width. At 100, the 20th node is 20 times narrower than the first, which a tooth on a harmonic series needs
stretch	bends the series, by up to 15% either way from whole-number spacing. Up, the partials spread out like a bell or a piano string. Down, they crowd together
teeth	how deep rake’s alternating lift and cut goes, up to 18 dB

what essaim shares

essaim stands on its own in the system.

What it shares is code. The spectrum shape uses the system’s own analyser, and auto wander uses the mixer’s orbit movement.

a 3-octave lowpass sweep from one cable

1. Click through to the lowpass preset. It sets morph to 100%, with shift reading -80.
2. Drag shift to +70, then click a. morph moves to 0.
3. Patch a slow oscillator into the morph jack. 10 volts covers the whole bar, so one cable gives the whole sweep.

tilt at -85 and bite at +95 are what make it a real filter.

keep one end, audition the other

Your current sound stays safe in a while you try presets in b.

1. Load a preset, push a few macro fields, and drag a node or two.
2. Click a. The sound stays the same.
3. Flick through the presets. Each one lands in b and leaves a as it was.
4. Drag the bar back from 100%, to hear the blend as well as the destination.

The bar now sweeps from the sound you kept to the preset you are trying.

vowels made of harmonics, then pinch on a comb

Load vowel. Sweep the morph once to hear the glide into the 7 formant positions.

Then drag rake to about +40. You hear a vowel made of harmonics.

rake is now an override. Click b to store it in that end. Or click a preset arrow to release every override at once.

Now load comb, and read the macros. rake is at +100, which adds 25 extra nodes for 32 in all. pinch is high enough that each tooth is narrower than its own spacing.

Lower pinch to about +30, and the teeth smear into one flat bump. Go on to ring, where pinch is higher still, and the whole series rings.

a dual filter with drive in the loop

foudre is a dual resonant filter with distortion, 8 hp wide. It has 2 state-variable filters, with the drive inside the resonance loop.

The routing decides where the distortion sits. It can sit between the 2 filters, inside each filter’s resonance, across the pair, or around the whole module.

So a filter can shape a signal it has already shaped, saturated on the way round. As the loop gain rises, the loop rings, then howls. At the top, it oscillates by itself.

Both filters run at twice the sample rate. This keeps aliasing low under heavy drive.

how to read the display

The display shows an arc between 2 posts, called electrodes. Each electrode is one filter.

An electrode’s horizontal position is that filter’s cutoff. The axis is logarithmic, from 20 Hz at the left edge to 16 kHz at the right edge.

The 2-letter label under each electrode is that filter’s mode.

The bolt between the electrodes is the output waveform. It is normalised, so a quiet signal still draws a shape.

move one filter or change its mode

1. Point within a few pixels of an electrode. Use the lower half of the arc for the first filter, and the upper half for the second.
2. Drag left or right to move that filter’s cutoff.
3. Click the electrode to cycle through its 4 modes.

A drag moves cutoff and spacing together, so the other filter stays still. To move both filters together, use the cutoff ring.

what the bolt shows

CUT (L253) … The picture is taken before level. The arc shows what the module is doing. The jacks carry what leaves it.

drive and the fixed knee

The knee is the saturator’s threshold inside the resonance loop. It is the point where the loop stops growing and starts to squash.

The knee is fixed. This fixes the loudness of the oscillation, so drive does more than change the volume.

resonance and where the filter sings

resonance lowers a fixed loss and raises a saturating regeneration, both at once. Where the 2 meet, the filter sings.

With drive at 0, this happens a little over 70% of the way up. With drive at the top, it moves to about 75%, because drive raises the loss.

level sets the output gain

level is the last stage in the signal path, outside the loop. It runs from a quarter of the gain at the bottom, through unity at the centre, to 4 times the gain at the top.

That is 12 dB either way.

move the pair with cutoff and spacing

cutoff moves both filters together. It runs from 20 Hz to 16 kHz on a logarithmic scale, and starts a little over 300 Hz. The field shows the effective cutoff.

The filters keep the ratio that spacing sets. So a cutoff sweep moves the whole shape across the spectrum. … Its 2 ends mirror each other. … fm is exponential, so a given depth is a fixed interval. With no cable in its jack, fm modulates the cutoff with the input signal itself.

the 4 routings

The routing field has 4 words: série, cœur, meute and boucle. Each one changes how the filters and the saturators connect.

série turns the regeneration off and keeps the damping above zero. resonance still narrows both filters, and Q rises to about 16.

In série the loop is open, so it stops short of self-oscillation. The switched-capacitor grit is clearest here.

step through the routings

Click the right half of the routing field to step forwards through the 4 routings. Click the left half to step backwards.

the follower amounts

foudre has its own envelope follower, which reads the raw input. The 3 bipolar bars under the 3 rings set how far the follower moves each of those controls.

They let foudre’s own input modulate it.

rise, fall and invert

The follower’s 2 times are sliders in the menu. rise runs from a fraction of a millisecond to 100 ms, and starts at 4 ms. fall runs from 10 ms to 2 seconds, and starts at 180 ms.

The sliders move smoothly while audio is playing. Each of the 3 amounts has its own invert.

spread the channels with stereo

A second row of fields sits under the follower bars.

stereo moves the 2 channels’ cutoffs apart, in semitones. The left cutoff goes down by half the value, and the right goes up by half. The pair stays centred where the knob and the arc show.

A negative value swaps which side is brighter.

make the cutoff follow notes with key

key sets how far the cutoff follows pitch. At 100%, it follows at 1 V/oct.

With a cable in the v/oct jack, the pitch comes from that cable. With no cable, it comes from the system’s keyboard: the highest note that clavier holds on the first live lane.

hear the knee stand still

drive changes the colour of the sound.

routing	cœur
cutoff	near its default
resonance	a little over half
spacing	its default
level	the centre

Play something and sweep drive from low to high. The resonant peak squashes instead of climbing, and the shape stays put. Across the whole drive knob, it slides down by only about 1.5 semitones.

make a vowel from 2 peaks

Start from the setting above, and change these:

filters in parallel	on, in the menu
both electrodes	clicked round to bandpass
spacing	about 75%
resonance	a little over half
drive	about a third

Drag the lower electrode left and right. The 2 peaks stand out from the valley between them. One peak moving while the other holds is what a vowel does.

Turn on harmonic spacing, and the pair snaps onto whole ratios. The vowel becomes an interval as well as a shape.

a howl that ducks

Use this setting for a howl that the drums push out of the way.

routing	boucle
resonance	near the top
drive	half
cutoff	the centre
level	pulled back

It sings with nothing patched, because boucle is one of the 2 routings that pass the wake-up gate.

Patch a kick drum into the left input. Every hit ducks the howl through the shared saturator, and the howl swells back.

drive changes the colour. resonance changes how fierce it is: in this routing, resonance is the loop gain.

play the filter as a voice

meute sings at a lower resonance than cœur.

routing	meute
resonance	about 80%
drive	low
spacing	the centre, so the 2 filters meet
first electrode	clicked round to bandpass
level	a little under half

Set key to 100%, and patch a sequencer’s pitch into v/oct. An octave on the cable is an octave on the filter, so it plays in tune.

Add drive, and the pitch drops about 1.5 semitones. Use the cutoff ring to correct it.

a reverb made of two engines

nef is a reverb made of 2 engines: the head, which writes the early reflections, and the network, which turns what is left into a wash.

the head	a velvet-noise engine. It makes up to 900 separate pulses per channel. Each pulse is at plus or minus 1 and lands on a whole sample. Its level comes from a drawn envelope, and its tone tilt comes from colour at that moment
the network	8 delay lines in a feedback loop, fed from the late part of the head. The lines trade energy through a lossless matrix, and smear the sound into a wash

… The network’s gain per pass is at most 1 at every frequency, so the network only darkens.

the horizon, where one engine hands to the other

The horizon is the point where the head hands over to the network. nef works it out from time, colour and texture together, and draws it in the window as a thin line.

Above the line, you hear the head’s pulses. Below it, you hear the loop.

colour sets the pulse density: 600 pulses a second at the dark end, and 2,000 at the bright end. texture moves between that figure and a floor of 8 pulses across the whole span.

The head stops at whichever comes first:

the whole span	the head covers the full length of the reverb
the budget	900 pulses have been used
6 seconds	the head’s longest possible run

While the head covers the whole span, only the head plays. … These figures were measured with a span of 3 seconds and texture at full:

colour at zero	the head runs 1.5 seconds
colour halfway	the head runs 820 ms
colour at the top	the head runs 450 ms

Turning colour up hands more of the tail to the network. … So time sets a span. It is also one of the 3 controls that set the horizon.

time sets the length of the span

time sets the length of the reverb, from 100 ms to 60 seconds, on an exponential scale. At long settings, the head stops at 6 seconds, and the network plays everything after that.

shape sets the swell

shape sets where the envelope’s peak sits in the span, and how low the envelope starts before that peak.

shape at the top	the reverb starts at about a tenth of its peak and swells into it
shape at the bottom	the reverb is at full level from the first pulse

colour does 3 jobs

the horizon	colour is one of the 3 controls that set it, as described above
tilt in the head	each pulse gets a tone tilt. The tilt ramps up across the head to full strength at the horizon. The brightening lands where the reflections are thinning, and the first reflections stay as they are
damping in the network	colour also sets the tone tilt of the network’s damping

texture sets the pulse density

texture sets how many pulses the head makes. It uses a square-root curve: the first quarter of its travel reaches the middle of the density range, and the other three quarters cover the dense half.

At zero, the head makes 8 pulses across the whole span. The head is then the whole reverb. nef becomes a scattered multi-tap delay.

Turn texture up, and the taps multiply and fuse. The far end of the span passes to the network.

Loudness stays the same across the range: 8 taps and 900 taps come out at the same level. texture is also nef’s CPU control, because the cost depends mainly on how many pulses the head gathers per sample.

mod moves the delay lines

mod moves the network’s delay lines. One knob sets both rate and depth: from 0.3 Hz to 3 Hz … CUT

warp sets what moving time sounds like

warp has 2 settings, and its cell shows the name of the current one. The default is repitch.

repitch	nef rebuilds the pulses as time moves, so the tail bends in pitch between sizes
crossfade	nef dissolves from one size to the next, with a small dip in level

… Changes are smoothed in the same way, so a move scrubs.

what the window shows

The window shows the sound falling through the space.

The window has 52 rows and 10 bands:

down	time, from the moment of the sound at the top to the end of the span at the bottom
across	frequency, dark at the left and bright at the right, from about 60 Hz to 14 kHz
the thin line	the horizon

A bright transient starts at the top right. As it falls, you can see how long it lasts, how far left it moves as the room takes its top, and where it crosses the horizon.

see the reflections come apart

The window lights one row per pulse, until there are more than about 1.2 pulses per row. Above that, the rows join into a solid field.

As you lower texture, the top of the picture breaks into separate reflections before you can hear them. The horizon line moves up as the taps fuse.

To see this clearly, use a short, bright sound. A continuous sound hides the structure.

use nef as a send or an insert

mix starts at wet only, so nef works as a send. … An envelope into texture changes the balance between reflections and wash.

a scattered delay, then a colour sweep

These settings take the horizon away, and then let colour move it.

texture	0
time	2 seconds
shape	about 45%
colour	halfway
mod	low
mix	full
warp	crossfade

The horizon line drops off the bottom of the window. The rows separate into 8 countable events, and nef plays as a scattered delay.

Now set texture to full and time to about 3 seconds. Sweep colour from bottom to top, and touch nothing else.

At the dark end, the head runs 1.5 seconds, and the line sits about halfway down the window. At the top, the head runs under half a second, and the line sits near the top. Somewhere on the sweep, more colour gives a duller tail.

a room that follows the note, then a freeze

A keyboard sets the size of the room, note by note.

warp	repitch
time input	a keyboard’s note CV
time	about 1 second
texture	high
mod	low
mix	full

Each octave halves or doubles the space. The tail glides between sizes, and held chords give a stack of rooms of different sizes.

Replace the keyboard with a slow triangle wave of about 1 volt each way, and you get a tape bend. Switch warp to crossfade while it runs, and the bending stops. You hear a dissolve with a small dip in it instead.

For a freeze, use these settings:

frz	a gate of 1 volt or more
texture	high
mod	around 60%
colour	halfway
mix	full

The window reads held. The loop holds without loss, and mod’s movement drops to zero. Play into it, and you add early reflections over the frozen wash.

When you release the gate, the held sound decays normally from where it was.

a limiter in front of a clipper

pli is an overdrive. Each of its stages does 4 things in order:

high-pass	removes everything below about 30 Hz
gain	drive sets how much
limiter	turns the signal down when it is loud, slowly
clipper	flattens whatever is still above the ceiling, which is 5 V

The limiter is slow on purpose. It takes 10 ms to react, so the front of a note reaches the clipper at full height. After that, the limiter holds the rest of the note just under the clipper.

So hits are clipped hard, held notes are clipped a little, and the gap after a hit swells back up.

the 2 ears get different waves

The clipper treats the top and bottom of a wave the same. bias moves the wave off centre before the clipper, so one half is flattened more than the other. That makes even harmonics.

spread pushes the bias opposite ways in the 2 channels. The left channel flattens at the top, and the right channel flattens at the bottom.

Fold the 2 channels to mono, and the even harmonics cancel exactly. What is left has only odd harmonics, like the plain overdrive with a slightly softer clip. The difference between the channels holds only the even harmonics.

the panel

The panel has 6 controls, in 3 rows of 2:

drive and bias	the top row. Each has a cv jack above it
stages and spread	the middle row, in grey. You set them for a patch
noise and mix	the bottom row. Each has a cv jack below it

The jacks are labelled by the row of names beside them. Each ring’s colour matches the line under the value it controls.

drive sets the gain

drive sets the gain in front of the limiter. At 0% the gain is 1 dB. Each 1% adds 0.39 dB, so 100% is 40 dB.

drive also raises the limiter’s threshold a little, by 5 dB across its travel. So at high drive, more of the signal reaches the clipper.

The wet signal stays at or under 5 V. So drive changes the sound and not the level.

stages repeats the chain

stages runs the whole chain again, from 1 to 16 times. Every stage has its own high-pass, gain, limiter and clipper. Every stage after the first adds the drive gain again, plus 0.9 dB.

At 1 stage, pli is an overdrive. At 4 stages, it is already a wall. Each stage also removes a little more bass, because each has its own high-pass.

stages is continuous. Between 2 whole numbers, pli crossfades from one stage to the next. Both sit under the same ceiling, so stages moves smoothly.

Each stage costs processor time. 1 stage uses about 0.4% of a core, and 16 stages use about 6%.

bias moves the wave off centre

bias runs from minus 1 through 0 to plus 1. It adds an offset to the wave just before each clipper, and takes it off again afterwards.

A positive bias flattens the top of the wave more than the bottom. A negative bias does the opposite. At full travel, the offset is 60% of the ceiling.

bias is read on every sample. So an oscillator into the bias jack works at audio rate, and the clipping point moves across the wave as it plays.

A biased channel can reach 8 V for a moment, at the start of a note. pli then removes the offset this creates, and the peaks return to 5 V.

spread pushes the channels apart

spread adds to the left channel’s bias, and subtracts the same amount from the right channel’s bias. At 100%, it moves each side by three quarters of the bias range.

The limiter is shared by the 2 channels, so they are always turned down together. The stereo image holds still when a note hits.

spread also separates the noise. At 0%, both channels carry the same noise. At 100%, each channel carries its own.

spread needs the clipper to be working.

a noise layer before the drive

noise adds pink noise to the signal, in front of the first stage. The noise has a high-pass filter at 250 Hz, so it stays out of the bass.

The right-click menu changes the noise to white or brown, and can make it follow the input.

Because the noise goes in before the limiter, a loud note turns it down. When the note ends, the noise swells back up over the recover time.

The knob has a squared curve, so the bottom half is fine control. At 100%, the noise goes in 12 dB below full scale.

pli as a noise source

With nothing patched into the audio inputs, pli still runs when noise is above 0%, or when a cable is in the noise jack. The outputs then carry the driven noise alone.

This needs noise follows input to be off, which is how it starts.

1. Remove every cable from the audio inputs.
2. Patch an envelope into the noise jack, and leave the noise control at 0%.
3. Patch the left output to a mixer channel.

Each envelope now plays a burst of driven noise. Short envelopes give hi-hats. drive and stages set how hard the burst is, and bias changes its tone.

mix blends in the input

mix blends the input with the driven signal. At 100%, you hear only the driven signal. At 0%, you hear the input exactly.

The noise is part of the driven signal only.

The 2 signals stay in time with each other at every setting.

3 traces on a scope

The display is a scope showing about 20 ms.

The input is drawn faintly underneath. The 2 driven channels are drawn over it in their own colours, amber for the left and blue for the right. They are transparent, so they look darker where the 2 sides agree.

Short lines join the input to each driven trace. They are densest where the clipper is doing the most work. With spread up, they lean one way on the amber trace and the other way on the blue trace.

All 3 traces are scaled together, to the tallest one. Otherwise the display would be a loudness meter.

the sweep follows the input

The scope restarts each time the input rises through zero. So a held note stands still on screen while you move a control.

If no restart happens for about 50 ms, the scope restarts anyway.

an insert that stays under 5 V

pli drives whatever it is given, and the driven signal stays at or under 5 V. So you can put it inside a feedback loop, or modulate it hard.

When the input is silent and noise is at 0%, pli stops working and costs nothing.

the plain overdrive

Use this setting as a reference for every other setting.

drive	50%
bias	0
stages	1
spread	0%
noise	0%
mix	100%

With bias and spread at 0, both outputs carry the same signal.

hear the limiter work

1. Start from the plain overdrive, and play short notes with gaps between them.
2. Set noise to 50%. The noise drops under each note and swells back in the gaps.
3. In the menu, set recover to its shortest. The noise now returns almost at once.
4. Set recover to its longest. The noise stays down for seconds after each note.

width that folds to mono

drive	60%
bias	0
spread	100%
stages	1

The 2 channels now carry different waves. Sum them to mono, and the even harmonics cancel. You hear an overdrive with a slightly softer clip.

Pull drive down to 10% with a quiet source. Most of the width goes, because the signal stays under the clipper’s ceiling.

a wall

drive	70%
stages	4
bias	a little above 0
mix	100%

Every note is now a square wave held at 5 V. bias changes the width of the square, so a slow cv into the bias jack sweeps the tone.

Bring mix down to 70% to let the attack of the input back through.

a walking clip point

1. Set drive to 60% and stages to 1.
2. Patch a sine oscillator into the bias jack, tuned a fifth above the note you play.
3. Set spread to 50%.

The clipping point now moves across the wave many times a cycle. This adds tones that follow both pitches, and the 2 ears hear them differently.

compression upwards and downwards at once

presse splits the signal into 3 bands. It compresses each band from both ends at once.

below the lower threshold	the signal is lifted up towards the threshold
above the upper threshold	the signal is pulled down towards the threshold
between the thresholds	the signal is left alone

A normal downward compressor makes loud sounds quieter, and then adds gain afterwards. presse closes the gap from both sides at the same time, in each band. So a quiet part of the spectrum comes up while a loud part comes down, in the same instant.

fixed settings, scaled by 2 knobs

presse’s detailed settings are fixed inside the module. These are the thresholds, ratios, attack and release times, the input trim and each band’s make-up gain.

amount	scales how much of the compression happens
time	scales how fast all of it happens

The rest is up to you: where the 3 bands sit, which bands are working, and how much output each band gets. You set these by dragging on the picture above the 2 knobs.

what the display analyses

The picture analyses the input only. It uses a 512-point analysis, about 50 times a second, shown as 56 columns spaced logarithmically across the frequencies. The output shown is worked out from the input.

The columns rise fast and fall slowly, so you can read the peaks.

read the gap, then the numbers

Look at the gap between the input and output first. Then read the numbers under it.

Each band’s bar shows its live gain in dB, or the word off.

a positive number	the band is being lifted
a negative number	the band is being held down
off	the band is switched off

The number is the actual gain the audio got, measured to within half a dB.

how an off band looks

When a band is on, its output curve is filled solid. When it is off, the curve is nearly transparent. An off band passes through untouched, so its output curve sits exactly on its input curve.

A band takes up as much of the picture as the slice of spectrum it covers. The bands and bars resize as you move the crossovers. For example, drag the low line up, and the left third of the picture becomes the left half.

the gestures on the picture

There are 6 gestures on the picture. You grab a band anywhere inside it.

The line inside a band shows its level.

Drags allow for the system’s zoom, so a gesture does the same thing at any zoom. They all run at one speed: the fine-drag modifier only works on the 2 knobs.

drag a band to set its gain

Each band’s gain is added on top of the compressor’s move, and on top of the fixed make-up gain. It is accurate to a quarter of a dB.

move the crossovers

Both crossovers run from about 25 Hz to 16 kHz, on an exponential scale, so the musical range is spread evenly across the travel.

The splits are eighth order, and the 2 halves of each split add back together without changing the frequency response. The 3 bands rejoin flat, measured to a quarter of a dB across the whole range.

The upper bands pass through a second split. So the low band gets an extra stage that corrects its phase.

the crossovers keep a fixed distance apart

Each crossover line stops well short of the other, at a fixed ratio apart. A drag stops at this limit.

The limit applies while you drag, and again inside the engine. The 3 bands always stay in order, and the middle band always has some width.

switch a band off

Click a band’s number under the bars to switch it off. An off band goes straight to the output, untouched. It measures the same as a plain cable.

The band is still split, and it still rejoins flat with its neighbours. It just arrives unchanged.

amount scales the move and the make-up

amount multiplies each band’s whole gain move before it is applied. This includes the fixed make-up gain that goes with the move.

At zero, presse measures flat to within a third of a dB, from 30 Hz to 15 kHz.

time scales the attack and release

presse has 6 fixed attack and release times, one pair per band. time multiplies all 6 by the same factor, from a twentieth up to 4 times.

Attack and release scale together, so the balance between them stays the same.

ratios, knee and detectors per band

The ratios, knee and detector windows are set for each band.

the top band	a limiter
the lower 2 bands	very hard compressors

On a full mix, this holds the top transients firmly, while the body is squeezed hard without sounding flat.

Every threshold has a 6 dB soft knee. A 5 dB trim on the input sets where the fixed thresholds sit against the system’s line level.

the upward lift has a limit

Upward compression adds gain in proportion to how far below the threshold the signal is. An empty band would ask for 148 dB of lift.

So the lift has a ceiling. The number on the bar shows when the lift reaches it.

keep the bass untouched

low crossover	about 120 Hz
high crossover	about 2.5 kHz
amount	full
time	normal
low band	switched off at its bar

The left third of the display goes hollow, and its bar reads off. The low band passes through as a plain cable, while the middle and top bands get the full effect.

Where you put the low line decides what off protects. Drag it up to 200 Hz, and the whole bottom of a mix is left alone. Drag it down to 40 Hz, and almost nothing is.

the top band as a limiter, then staging the mix

high crossover	about 5 kHz
amount	full
time	low
high band gain	a few dB down, by dragging inside the top of the display

With time low, the top band’s attack is only a few ms, and its release is under a tenth of a second. This catches cymbal transients without audible pumping. Its bar number sits at 0 most of the time, and dips sharply on transients without lifting.

The same drag sets the balance of the whole mix. Set the low band a little down, the middle at 0 and the high band a little up. The shape of the 3 lines shows the balance.

These 3 gains sit on top of the fixed make-up gain. Randomise leaves them as you set them.

one clock sets both time and tone

ruban is a bucket-brigade delay. Its line is a chain of capacitors, called buckets, that pass a sample along under a clock.

The delay time is the number of buckets divided by twice the clock rate. The bandwidth is half that same clock. So one number sets both, and a longer delay gives a darker repeat.

The line has 2048 cells. Its anti-alias and reconstruction filters track just under half the clock. These filters are the wall, the point where the top end is cut. The wall moves with the delay.

varispeed comes from the same clock

The 4 playback heads are placed as fractions of the way along the line. … The pitch of everything in the line changes at once. …

1 ms	above the audio band, and clamped there
20 ms	still clamped
about 41 ms	the clamp lets go
100 ms	about 9 kHz
300 ms	about 3 kHz
the top of the knob	about 1 kHz

Measured at 300 ms, a 5 kHz tone comes back more than 12 dB below a 1 kHz tone. At 40 ms, the same pair is within 3 dB.

So the bottom third of the delay knob is a comb filter on a clean, bright line. The top two thirds is an echo that gets darker as the delay gets longer.

delay sets the time

delay runs from 1 ms to 900 ms, on a logarithmic scale. The range from 1 to 10 ms, where combs and flanging happen, takes about a third of the travel.

The ring shows its value. Below 10 ms it shows decimals, and above 10 ms it shows whole milliseconds.

hz or sync: a time, or a division of the beat

The round button between the 2 readouts sets what the delay knob means. It shows hz when the knob is a time, and the link mark when it is synced.

hz	the knob is a time, 1 to 900 ms. The default
sync	the knob walks divisions of the beat, from 1/32 up to a whole bar: straight, dotted and triplet. The readout names the division

sync counts in the system’s tempo, from the transport bus. … The clock jack writes the knob only while sync is off.

feedback sets how long the tail lasts

feedback is taken from the very end of the line. Past unity, the loop sings.

Because it is taken from the end, spread and feedback work independently.

spread places the 4 heads

ruban has 4 playback heads. The first is at full level, and each one after it is quieter. spread decides where the heads land, from all together to spread across the whole delay.

So spread sets a rhythm. For example, with a delay of 500 ms and spread at full, the heads land at 120, 240, 360 and 480 ms. This gives a run of 4 with a fading accent.

The last head is always the quietest, so the pattern leans towards its first hit.

width pans the heads

width pans the 4 heads to alternate sides. Each head is panned further than the one before. The first pair sit a little either side, and the last pair go almost hard left and right.

width turns sides down. … A random roll keeps your dry and wet balance.

instability wobbles the clock

instability is in the right-click menu, and it is on by default. It wobbles the clock. Because the clock is the delay, it wobbles the time and pitch of everything in the line together.

It combines 3 motions:

wow	slow, at about 0.5 Hz
flutter	faster, at about 5.5 Hz
drift	a slow random wander, within limits

spray scatters the heads

spray adds random error to the head positions. Each time the line completes one full trip, all 8 offsets get new values: 4 heads on each side. The heads crossfade to their new places over a few milliseconds.

The repeats scatter around their places, and their pitch stays steady.

diffuse smears each pass

diffuse passes each repeat through 3 all-pass filters, at about 7, 12 and 19 ms. The control blends them against the dry path.

At 0, the stage is skipped completely.

drive saturates the feedback

drive is a soft saturation stage. As it comes up, the stage bends harder, and the level stays matched.

By default, drive sits on the feedback. At 0, the stage is skipped completely.

cross and ping-pong

cross is a slider in the menu. It sends each channel’s feedback into the opposite channel’s loop.

At 0, the 2 lines are independent. At full, they swap completely on each pass.

With ping-pong off, cross works as a width control for the tail. A little makes the 2 sides answer each other. Full makes every repeat alternate. ping-pong overrules cross.

ping-pong is on to start with. The input is summed to mono and goes into the left line. Each repeat then bounces to the other side: left, right, left, right.

The 2 sides are equally loud at every feedback setting. The first bounce on the left crosses to the right at full level, and feedback is applied once for each pair of bounces. So even with feedback at 0, you hear 1 repeat in each ear.

the prediction band

The band of vertical bars above the tabs shows every repeat before it happens.

The bars are calculated from the current settings, with cv included. feedback sets the band’s zoom.

the filter tab

The filter tab has one handle. Press anywhere in the tab to move the handle there and start dragging.

left and right	the centre frequency, on a logarithmic scale across the whole audio range
up and down	the width, from about a sixth of an octave to 9 octaves. Up is wider
the order	also set by up and down. The narrowest setting is 4 poles each way, 24 dB per octave. The widest is 1 pole, 6 dB per octave

the route tab

The route tab shows 4 processing stages: the filter, the diffuser, the drive and the external loop insert. You can change their order.

A bracket sits under the stages. Drag its ends to choose which stages are inside the loop. A stage inside the bracket happens again on every repeat. A stage outside it happens once.

a dark echo

ruban starts with these settings.

delay	300 ms
feedback	about 60%
spread	55%
width	70%
mix	40%

…

instability	15%
the bracket	covering every stage

The wall sits at about 3 kHz. Each repeat is nearly a third darker than the last. By the 5th repeat, almost nothing above 1 kHz is left.

Sweep the delay ring while it plays, and the whole tail smears in pitch.

a drone that sings

Start from the setting above, and change these:

feedback	the top of its travel, 110%
mix	the top
drive	about a quarter
spread	low

Leave delay at 300 ms and instability at its default. Play one sound, then stop the input. The loop builds, and the ceiling holds it. The wall stops it becoming white noise, and the drone settles dark.

Move the delay ring, and the drone transposes. Add a little spray, and it starts to breathe.

4 heads across the bar

These settings spread a rhythm of 4 repeats across the stereo image.

ping-pong	off, in the menu. Do this first
cross	0
spread	full
width	about 70%
delay	about 500 ms
feedback	about 45%
mix	about 45%

Turn ping-pong off before anything else. Otherwise every repeat lands on one side.

spread sets the rhythm, and width sets the placement. ping-pong decides whether either one reaches the stereo image.

a flanger

delay	the bottom third of the ring, between 1 and 10 ms
feedback	near unity
mix	half
spread, width, diffuse	0
instability	about 40%

The wall is clamped above the audio band, so the line is clean and bright. All the character comes from comb filtering. feedback near unity makes the comb resonant.

instability is the only movement. At 40%, the clock wobbles by about 10 cents, at 0.5 Hz and at 5.5 Hz. That wobble is the sweep.

16 spectral effects on one transform

trame is a spectral multi-effect. It turns the sound into a spectrum with one short-time transform, processes it, and turns it back into sound. There are 16 different processes in between.

Each process is its own effect. The 16 effects share the same engine.

the panel relabels itself for each effect

The 3 rings take their names from the effect. Each effect sets what its 3 controls are called, what each one shows, and where each one sits. So the labels on the panel tell you what the controls do.

how the engine works

The engine uses square-root windows, a hop of a quarter of a window, and an overlap that adds up to exactly 2.

To keep the phase clean, it finds the peaks in each frame. It moves each peak on at its own measured frequency. Every bin around a peak stays locked to that peak’s offset.

what happens with nothing patched

With both inputs empty, trame plays itself a silent drone of 5 partials over a low root. The drone goes through the real engine, so both displays draw what they would draw for real audio.

The outputs are held at exactly 0 V. So you can watch all 16 effects work before you patch a cable.

When a cable goes into either input, or comes out, the transform’s memory is cleared and the drone resets. So pulling the last cable out of a long spectral reverb tail cuts the tail dead. Putting a cable back starts from silence, with a short fade.

fft size

The dim cycler at the top sets the transform size. The size trades timing detail against frequency detail, and it also sets the latency.

It has 7 settings: snap, edge, near, even, wide, bloom and wash. The default is even. wash is the largest, at 8192 points: the finest frequency detail, and about 170 ms of delay through the module.

the 2 displays

The upper display compares input and output. The lower display, the spectrogram, shows the output over time.

In the upper display, the input is a grey outline. The output is filled in the effect’s own colour. Both are scaled against one shared maximum, on a curve that lifts quiet detail off the bottom.

a gate	looks like holes punched through the outline
a spread	looks like the outline smeared sideways, under a wider fill
a freeze	looks like a fill that has stopped following the outline

Each effect has its own colour. Colours step 3 places along the palette for each effect, so neighbouring effects look different. thru is drawn in grey, because it is only a wire.

the effect cycler

The effect cycler is the only control that changes what the other 3 controls mean.

The effect name sits under the displays. Click its left half to step down the list, or its right half to step up. It wraps at both ends, from freeze round to thru.

Switching clears that effect’s memory: the held magnitudes, the frozen frame, the scatter history and the stretch capture. Old history would otherwise arrive as noise. The audio keeps running and the window overlap smooths the change, so switching is smooth.

the 3 rings

Each ring runs from 0 to 1 inside, and shows whatever unit the effect asks for:

frequency	hertz or octaves
time	seconds or milliseconds
other	a multiplier, a signed number either side of zero, or a percentage

Each ring has its cv jack inside it. The effect decides whether the arc is drawn from the left or from 12 o’clock.

the follower bars

Under each ring is a thin horizontal bar. It sets how far an envelope follower reaches into that ring’s control, from minus 100% to plus 100%. It is bipolar, with a centre detent.

At zero the bar is invisible. So on an untouched panel, the bars are the strip of empty space under each ring.

thru, the 2 movers and the 3 band effects

In all 3 band effects, freq and width work the same way.

freq	exponential, from 40 Hz to 8 kHz, shown in hertz
width	from a quarter of an octave to 6 octaves, spread evenly either side of freq
twist	wraps the band around itself
mirror	flips the band
bandpass	keeps the band and removes the rest

The 2 movers, shift and rotate, shift every partial. Scaling transposes, and adding shears. So a multiplied harmonic series stays harmonic, and an offset one turns inharmonic. rotate wraps what passes the top round to the bottom.

On both movers, mix takes the louder of the 2 signals in each bin. This keeps the level steady, where adding them would double it.

thru is a plain wire only while sieve is off. With sieve below 100%, even thru goes through the phase-locking path.

the 4 frame rewriters

spread, compress, corrupt and lofi rewrite each frame. lofi removes harmonics in groups.

spread joins neighbouring bins by taking the maximum. This is what makes it ring. Its kernel is nearly flat, so even far neighbours come up strongly.

Diffused bins return to their own bin centres, and the phase engine treats each one on its own.

the 6 effects with memory

stretch, scatter, cascade, smear, reverb and freeze hold part of the past.

At stretch’s dead centre, the read head jumps back to live and stops. The output is then a still frame, refreshed once every tempo period. A little either side of centre sets the stills crawling.

stretch runs on its own clock, whatever the frame rate.

In cascade, a gain of 0.9 per frame would already fall by 160 dB a second. That kills the slide before it crosses the spectrum.

cascade and freeze join their memory with the live spectrum by taking the maximum. So new material wins wherever it is louder.

sieve and overlap

2 fields sit under the envelope bars.

sieve	keeps only the loudest bins in each frame, and sets the rest to zero. From 2% to 100%. 100% is off, and is the default. Turn it down and the sound thins to its bones, under whatever effect is running
overlap	how many analysis frames cover each window: 2, 4 or 8

choosing an overlap

2	half the processing cost. You can hear the frames as frames, so the sound is grainier
4	the original setting
8	twice the processing cost. Smooths every effect that rewrites frames

The window sum follows the setting, so thru stays a clean wire at every overlap. Changing it resets the transform, as changing the fft size does.

freeze played by the follower

The frozen chord loosens whenever the input gets loud.

effect	freeze
fft size	wide
grip	100%
chaos	15
mix	100
ring 1’s amount bar	minus 60%
follower	4 ms rise, 400 ms fall
sieve	off

The memory fills from the first sound it hears, and locks once it outweighs the input.

The negative bar on ring 1 loosens the grip on a loud passage, so new material pushes into the frozen chord. The long fall lets it lock again as the passage dies away. wide suits a freeze, because a freeze is pitch work and needs fine bins.

corrupt driven by a sidechain

effect	corrupt
amount	about minus 60, so it removes the quietest
decay	40%
mix	100
sidechain on in r	on
ring 1’s amount bar	plus 50%
follower	4 ms rise, 120 ms fall
fft size	near

Each hit on the right input strips more of the quiet partials. The sound thins with the sidechain, and fills back in as it decays.

decay at 40% holds each removal for about 200 ms. That is long enough to sound like a glitch.

a rising slide

effect	cascade
length	70%
feedback	40
the third ring	about minus 40, so it slides upwards
fft size	even
sieve	60%

The sieve runs after the effect. It thins the slide’s own trail as well as the input, so the climb sounds like a few clean partials rising.

a coarse frame rate

effect	smear
length	55%
chaos	45
mix	85
frame rate	x1/4
fft size	edge

A new analysis frame arrives every fourth hop, and the 3 between are repeats. So the smear moves in steps, and the spectrogram’s columns clump into fours.

stretch on its own clock

effect	stretch
speed	dead centre
tempo	500 ms
chaos	25%
fft size	wide

The output is half-second stills of the input. stretch reads them on its own clock, whatever the frame rate is set to.

a channel strip for tresse’s 8 channels

brin is a channel strip for the 8 channels of tresse, the mixer. You edit one channel at a time.

All 8 strips run all the time. Selecting a channel only points the 12 fields and the pan rail at that channel’s settings.

The ladder at the top selects the channel. It has 8 rows, one for each channel. Each row shows the channel’s number, its live gain reduction and its frequency response.

Click anywhere on a row to select it. The controls below then edit that channel.

brin sits on tresse’s left

brin works when its right edge touches tresse’s left edge. … brin processes it, and hands the processed pair back. … Because brin works before the fader, the compressor hears the level before tresse’s field.

Inside each strip, the signal goes through these stages in order:

high-pass	removes low frequencies
low shelf, bell, high shelf	the 3 tone bands
low-pass	removes high frequencies
compressor	evens out the level
pan	rotates the stereo pair
trim	sets the final level

the curve shows the tone

Each row draws a small frequency response. brin calculates it from the filters themselves, at 45 points across the audible range, so the curve matches what you hear.

The curve shows the overall tilt, at a little under 4 pixels an octave and about half a pixel per dB. For exact values, read the fields below.

selected and unselected rows

A row has 3 states, told apart by line weight alone. The curve keeps the channel’s own colour in every state.

The selected channel’s colour also appears in the line under every field, and on the cap of the pan rail.

Each selection is an undo step. Undo steps back through the channels you looked at.

the gain reduction bar

The bar hangs from the top of the row, at 1 pixel per dB. It stops growing at 14 dB.

The channel number lights while signal arrives.

A row whose equaliser is at its defaults draws faintly, even when its compressor is working hard.

brin goes dormant away from tresse

If brin is not against tresse, or tresse is bypassed, brin goes dormant:

the ladder	dims to about 40%
the numbers	turn grey
the gain reduction bars	read zero

To fix this, move brin back against tresse’s left edge.

If you bypass brin, tresse stops sending it audio, and uses the dry input straight away.

5 sections, 2 of them off at first

The equaliser has 5 sections: a high-pass, 3 tone bands and a low-pass.

The high-pass and the low-pass both start off. Off is a true bypass: the section leaves the signal unchanged. … CUT … If you sweep a band’s frequency while its gain is at zero, the sound and the curve stay flat.

the frequency ranges

The frequency ranges stop where a neighbouring section takes over.

below 40 Hz	the low shelf would overlap the high-pass
above 500 Hz	a shelf becomes a tilt across the whole voice
above 16 kHz	the low-pass covers the same range

one field sets threshold and ratio

The threshold field sets the threshold and the ratio together.

Drag it up for more compression. The printed number goes down as you drag up. The tooltip shows the threshold and the ratio side by side.

Attack works the same way: higher values are faster.

make-up gain starts at -12 dB

Make-up gain stays at zero until the printed threshold goes below -12 dB.

brin predicts the make-up gain from the threshold and the ratio. Its reference level is 12 dB below full scale.

release and attack

Release has its own control.

Attack has a large effect on drums. A very slow attack lets the start of each hit through, then clamps down behind it. A very fast attack flattens the start of every hit. On a drum, attack changes the result more than the threshold does.

pan rotates the stereo pair

brin’s pan rotates the stereo pair, up to 45 degrees either way. The left side picks up some of the right, and the right picks up some of the left.

The total energy stays exactly the same. If you turn the pan fully and back again, you get the original signal.

This is the pan for a tresse channel. tresse’s field sets level only.

trim sets the final level

The trim is the last stage in the strip. It ranges from -12 dB to +12 dB.

The trim uses the same number format as the tone bands.

brin works with tresse

… A second brin placed to the left of the first stays dormant. The same applies to envoi on tresse’s right.

The processed signal from brin is what envoi’s sends pick up.

The round trip through brin takes 2 samples, about 40 microseconds at the usual sample rate. The delay is the same on every channel, so the channels stay aligned.

clean up channels with the high-pass

The high-pass removes low rumble.

1. Select a row in the ladder.
2. Set the high-pass to about 80 Hz, and leave everything else alone.
3. Watch the row’s curve as you sweep.
4. Repeat on 6 of the 8 channels.

Apart from the high-pass, each channel is bit-identical to what arrived.

shape a voice with 3 bands and trim

The same 3 moves work on a submix.

low shelf	up about 3 dB at 75 Hz
bell	down about 4 dB at around 450 Hz
high shelf	up about 3 dB at 9 kHz

gentle glue compression

threshold	about -18 dB
attack	around 10 ms
release	about 200 ms

The gain reduction bar sits a few pixels deep on peaks.

Now move the threshold back to -12 dB. The ratio hardly changes, but the make-up gain drops to zero and the channel gets noticeably quieter.

compression for a drum

threshold	about -24 dB
attack	fully slow, at 30 ms
release	around 60 ms, which recovers before the next hit at most tempos

Now change only the attack, down to 0.3 ms. The start of every hit flattens.

Watch the row’s bar. With the slow attack, it fills after the hit. With the fast attack, it fills on the hit.

one stereo send for tresse’s 8 channels

envoi is a send bus for the 8 channels of tresse, the mixer. It adds a send and return to the whole mixer.

1. envoi scales each channel by its own send amount.
2. It adds the 8 results together into one stereo bus.
3. The bus leaves on the 2 send jacks, into an effect you patch in the system.
4. The effect comes back into the 2 return jacks, through one return level, and goes straight into tresse’s output.

envoi has 10 controls, 6 jacks and one display. The 8 send amounts and the tap switch act on the way out. The return level acts on everything coming back. The chnl and cv jacks move one channel’s send from a cable.

The bus is an ordinary stereo pair, so any stereo effect can take it.

envoi sits on tresse’s right

envoi only works while its left edge touches tresse. This position is what makes envoi a send. brin, on tresse’s other side, is an insert.

The caption at the bottom of the display shows the status:

eight to stereo	envoi is connected to tresse
no tresse	envoi is waiting for tresse

While envoi sits away from tresse, or tresse is bypassed, the panel dims to about 30% and the send bus goes silent. envoi is waiting for tresse.

every send starts closed

All 8 send amounts start at none. A new envoi shows 8 empty bars, a live caption and a silent bus. Turn up a send to hear anything.

Each send amount changes through a 12 ms declick, the same one tresse uses on its faders. A channel’s mix level and its send level move together.

the send knobs

The 8 send knobs are rings with the panel showing through the middle. A small digit in the middle gives the channel number. The ring is the channel’s colour on tresse.

randomise changes the sends and the return

Randomise changes the 8 send amounts and the return level. It leaves the tap switch as it is.

A random tap would change how the module works without warning. 8 random send amounts give you a starting balance to adjust.

the tap switch: post or pre

The tap switch is a word halfway down the panel. It reads post or pre. Click the word to switch. The word tap behind it names the control.

post is drawn in plain ink. pre is drawn in amber, the solo colour, as a warning.

what post and pre do

post	the send follows the mix. Pull a channel’s dot away in tresse’s field, and its send goes down with it
pre	the send ignores the mix. Pull the dot away, and the send stays at full strength

With pre, a source can leave the mix while its reverb tail stays.

the return level goes up to about +6 dB

Everything coming back passes through one return level. Unity gain is at the middle of its travel. The top is double that, a little over 6 dB of gain, so the return can be louder than the dry mix.

The return level changes through the same 12 ms declick as the sends.

bars and ticks

The display in the middle of the panel has one column for each channel. Each column shows 2 readings:

the bar	the signal the channel is actually sending
the tick above it	the send knob’s setting

The gap between them is headroom. A tall tick over a short bar is a high send on a channel that is sending little. A short tick with the bar almost touching it is a low send, used in full.

The ticks show even with nothing playing. This lets you tell an empty display from a disconnected one.

One column has a blue wash behind it. That is the selected channel, the one the cv jack moves. Click a column to select it.

the gains keep moving when envoi is dormant

The send gains and the return level still move to match their knobs, out of sight.

envoi works only with tresse

envoi’s only connection beyond its 6 jacks is to tresse. envoi looks for tresse by name, in the module on its left. It checks only when the system layout changes. If the module on its left is not tresse, envoi stays dormant and its caption says so.

one envoi for each tresse

envoi checks its immediate left neighbour, and tresse its immediate right one. A second envoi to the right of the first stays dormant, and its caption reads no tresse. brin follows the same rule on tresse’s left.

one reverb for the whole mixer

tap	post
return level	unity
sends 1 and 2	about a third
send 5	about 60%
the other 5 sends	none

1. Patch the 2 send jacks into a reverb.
2. Patch the reverb’s outputs back into the 2 return jacks.

The display shows 3 bars, 3 ticks and 5 empty columns. Once the 3 sends are balanced, the return level makes the whole mix wetter or drier without changing the balance.

a mono effect in the return

Use this with a mono delay or a single-output reverb.

Patch the effect’s one output into either return jack. It sits in the centre. If the effect is polyphonic, sum it before the return.

read the gap between bar and tick

Use the display to set the 8 sends by eye.

Turn a quiet channel’s send to 80%. The tick sits near the top, with the bar barely off the floor. Turn a loud channel’s send down to 20%. Its bar climbs almost to its tick.

A bar on the floor under a high tick just means the channel is quiet.

pre-fader tails

1. Set the tap to pre.
2. Set channel 4’s send to about half.
3. In tresse’s field, pull channel 4’s dot out to the edge, or mute the channel.

The channel leaves the mix, but its send keeps running at full strength. The tail stays after the source has gone.

Set the tap back to post, and the same move takes the reverb away with the channel.

brin’s processing is in both taps. pre means before the fader.

monitor, send to your DAW and record

sortie is the master output of the system, at the end of it, on a 4 hp panel. It has 2 inputs. …

the output

sends the audio to your DAW or the standalone app

what sortie keeps out of the patch

The dice leave sortie’s settings as they are, including the folder and the mute.

the mode lamp changes after 2 seconds

The lamp at the top has 3 states. It changes state 2 seconds after you click it.

off	sortie records when you click the big button
grab	the one-shot recorder is armed
snip	the rolling buffer pauses during silence

1. Click the lamp to choose the next state.
2. Click again to move on. Each click restarts the 2-second timer.
3. Wait. The state changes 2 seconds after your last click.

While the change is waiting, the lamp flashes on and off every 200 ms, and its glow dims. The delay stops a fumbled click from starting a recording straight away.

The same 3 states are in the right-click menu. Choosing one there takes effect at once.

the big button does 2 things

The big button is 2 controls in one:

click	turns force-record on or off
press and hold	saves the rolling buffer, while you are still holding

Once a hold has saved, letting go leaves force-record as it was. Sliding the pointer off the button counts as letting go.

force-record overrides the lamp

Force-record records even with the lamp off. The lamp sets what sortie does by itself, and the button overrides it.

Whatever started a take also stops it:

a take you started by hand	runs through silence, until you click again
a take the threshold started	stops by the threshold’s own rules

meters, waveform and mute

The columns down the 2 edges are the meters. The strip in the centre shows the rolling buffer. The slat is the mute.

Each meter has 10 segments, at these levels:

green	-54, -42, -30, -21, -15, -12 and -9 dB
yellow	-6 and -3 dB
red	0 dB

A peak hold falls away with a quarter-second time constant.

A mono source lights both columns. So the right column can light with nothing in the right input: it shows that mono recording works.

grab records when signal arrives

grab is the one-shot recorder. Set the lamp to grab to arm it.

1. grab waits for the input to pass the threshold. The default threshold is -65 dB.
2. grab needs 3 ms of signal above the threshold before it opens.
3. It records until the input has stayed quiet for the whole hangover time.
4. It closes and writes the file.

take keeps the last minute

take is the rolling buffer. It records all the time. … To save it, hold the big button, or choose save the rolling buffer now from the menu. sortie writes the buffer as it stands.

snip pauses take during silence

Set the lamp to snip, and take’s buffer stops whenever the input stays below the silence threshold for longer than the hangover.

The buffer and the waveform strip both stop. The still strip shows that snip is pausing the recording.

3 kinds of recording, 3 folders

The 3 kinds of recording go into 3 folders, with 3 kinds of name.

sortie looks for the highest number in the folder, and adds 1. Delete the newest file, and the next recording reuses its number.

where recordings are saved

grab and take each have their own recording settings.

The destination folder is shared. Choose it on the settings page, under recordings. It starts as ~/Music/chassis, and sortie remembers the last one you chose. … CUT

where the sound goes

sortie is the end of the system, and chassis plays what reaches it.

in a DAW	the sound goes to the track chassis is on. The DAW chooses the device
in the standalone app	choose the output, sample rate and buffer size at the top of the settings tab

sortie’s jacks, meters, soft clip and recorder work the same either way.

the end of the signal path

Signal arrives at sortie and stops there. 3 routes reach sortie by themselves: … A typical patch sends everything into tresse, and tresse into sortie.

save the last 3 minutes

lamp	off
buffer length	180 seconds
bit depth	32-bit float
normalise	off
save to sub folder	on

Play, and hold the big button for half a second whenever something good happens. sortie writes the last 3 minutes into a dated folder.

Float keeps peaks above 0 dB, so a hot take survives. With normalise off, takes stay comparable.

record only the sound

snip suits sparse material with long silences.

lamp	snip
silence threshold	-52 dB
hangover	220 ms
buffer length	120 seconds

The buffer and the waveform stop during silence, so the 2 minutes hold only sound. Hold the button at the end to save it.

slow the whole system down

tape speed moves a whole arrangement into a new key in one move.

1. Start a take rolling.
2. Drag tape speed to 0.75. The system drops a fourth and runs a third longer. The footer shows the speed.
3. Save the take.
4. Drag back. The slider catches at 1.

sortie stores the speed in its own patch data, so the patch reopens at the speed you left it.

8 channels placed in a field

tresse is an 8-channel stereo mixer, drawn as a view from above. 8 coloured dots stand in a field, and a crosshair marks the listener.

A dot’s distance from the crosshair sets that channel’s level.

the fader and the dot are one value

Each channel’s fader and dot show the same value in 2 ways, and they stay linked:

pull a fader down	the dot slides straight out along its line
drag a dot inwards	the fader rises to match
move the crosshair	all 8 distances change at once

Whichever you moved last sets the value.

a dot’s direction is its fader’s line

A dot’s direction from the crosshair is the line the dot slides along when you move its fader.

What leaves tresse keeps left as left and right as right, changed only in level. Moving a dot sideways changes the level, because it moves the dot further away.

To pan a tresse channel, use brin, the channel strip on tresse’s left. brin’s pan rotates the stereo pair.

one display covers the panel

The field, the 8 faders and the 16 solo and mute switches are one display that covers the panel. Apart from the display, tresse has only jacks.

The display has its own dragging, its own fine mode and its own undo, in place of the usual control gestures.

The values underneath are real parameters. Anything that reaches parameters by number can control them.

silence is the same distance in every direction

Level falls with distance, on a fixed law very close to a square law.

One pixel across counts the same as one pixel down, so the edge of silence is a true circle. The reach and steepness are fixed: 35 pixels from the crosshair is always half level.

press anywhere else to move the crosshair

Press inside the field, away from every dot, and the crosshair jumps there. Keep dragging to move it.

The field is the upper part of the panel, from the title strip down to the tops of the faders.

spokes and hollow dots

The spokes show 2 things. A channel that is out of the mix is drawn hollow.

The quietest dots are drawn first. A click picks the nearest dot, so a loud dot on top of a quiet one is the one you grab.

the field shows the levels you set

Each dot’s position comes from the level you set. The field stays still while the music plays.

While you drag, the channel number and its level in per cent appear at the bottom left. They disappear when you let go.

a fader, a solo and a mute for each channel

Below the field, the same 8 channels appear as vertical faders, in the same colours. Each channel has a fader, a solo switch and a mute switch.

changes always fade

Every level change, including a mute or a solo, fades over 12 ms. The fade is too short to hear as a slide, but long enough to stop clicks.

For slow movement, use walk in the menu.

the meters are post-fader

The 2 thin lines beside each fader are the channel’s meters, left and right. They show peaks, with 5 V as full scale. They fall at a steady rate, so a quiet tail stays readable.

The meters are post-fader. So on a channel with signal, a low meter means a low fader.

the limiter holds the output at -0.3 dB

tresse’s output goes through a limiter, which is on by default.

ceiling	-0.3 dBFS, which is 4.83 V at the jacks
attack	instant
release	100 ms

One detector serves both sides. The mix dips in both ears together, so the stereo image holds still.

3 sliders make the listener wander

3 sliders in the menu make the listener wander around the field. All 3 show whole percentages.

walk is added to the listener position. If you drag the crosshair while walk runs, you move the centre it circles. Turn walk back down, and the crosshair returns to that centre over about half a second.

At full depth, the listener moves about 40% of the field either way. The 2 directions move at rates that are not whole multiples of each other, so the path keeps changing, even with jitter at zero.

wandering moves all 8 levels together

The wander moves the listener, so every channel changes level at once, in relation to the others.

A channel straight ahead of the listener’s path changes much more than one off to the side. A 4-second lap works as a tremolo across the whole mix.

an expander on each side

tresse hosts an expander on each side, and each side takes exactly one:

on the left	brin, the channel strip
on the right	envoi, the stereo send and return

tresse recognises them by name. It checks only when the system layout changes.

brin is an insert, envoi is a send

brin, on the left, is a pre-fader insert. tresse hands each channel’s dry stereo pair to brin, and reads the processed pair back 2 samples later.

From then on, the processed pair is the channel. It feeds the fader, the meters, the mix and envoi’s sends.

Touch a channel on tresse, and brin selects it. Press the channel’s fader, its dot, its solo or its mute, and brin shows that channel’s strip. A cable in brin’s chnl jack stops this, because the voltage then decides the selection.

brin’s side of tresse is what makes it pre-fader. The round trip takes 2 samples, about 40 microseconds at the usual sample rate.

bypass disconnects both ways

Bypass disconnects an expander from tresse, in both directions.

bypass brin	tresse uses the dry inputs
bypass tresse	brin and envoi go dormant and silent

the desk

Each fader and its dot move together.

1. Pull channel 3’s fader down about 20 pixels. Its dot slides straight out along its own line.
2. Drag that dot back in. The fader rises to match.

a solo without the solo switch

1. Push one fader to full. Its dot lands on the crosshair and keeps its stored direction.
2. Drag every other dot into a corner, down to 1 or 2%.
3. Pull the featured fader back to 60%.

The other channels are still faintly there, and their spokes still show where they are. Pulling the featured fader back gives a real balance, without switching 7 channels back on.

The solo switches are 11 pixels below the faders.

glue the mix with the limiter

Bring 3 or 4 dots above 90%, within a few pixels of the crosshair. The mix now passes the ceiling often.

The whole stereo image dips and returns together, without moving from side to side. This glued mix is why the limiter is on by default.

walk towards the drums

The listener moves closer to one group of channels.

2 or 3 close channels	around 85%, about 10 pixels from the crosshair
the other channels	between a quarter and a third, 50 to 60 pixels out

Drag the crosshair towards the close group, and leave every dot where it is. One drag changes all 8 distances, and one undo puts the whole move back.

walk with a control voltage

1. Patch a very slow oscillator into the front-to-back position input. Leave the left-to-right input empty.
2. Place the steady parts out to the left and right.
3. Place the moving parts straight ahead and behind.

The whole mix travels forward and back. One control voltage decides which parts change level and which stay steady. Add a little walk, and the listener keeps moving instead of stopping.

8 small utilities on one panel

… The verb is what the cell does. You choose it from a list of 14.

A 9th row sits at the top: the transport clock, divided.

how to read a row

Read each row from left to right. Every row has the same 5 parts, in the same order.

a, b	the 2 inputs. The verb works on these
cv	controls the verb’s number, if the verb has one
the verb	3 letters over a coloured line. Click to step through the verbs, or drag to choose one
out	the result. The jack limits it to between -12 V and +12 V

connect cells with cables

To feed one cell from another, patch the first cell’s out into the next cell’s a with an ordinary cable.

the foot shows the cell you last touched

The display at the foot of the panel shows one cell as the sentence it computes. On the top line, a small scope of a, the verb’s word with its number under it as a bar, and a small scope of b. Below, across the whole width, the out, filled in the verb’s family colour.

It shows the cell you last touched: the one whose verb or number you changed, or whose jack a cable last arrived at. Click the display to step to the next cell. The header names the cell and its verb in full.

choose a verb

The verb field is the same control qmod uses for its modes, at the same size.

click	steps to the next verb
right-click	opens the list of verbs
drag up or down	selects a verb

what the colours mean

The colour of the line shows the verb’s family. There are 5 families. The 3 letters tell you which verb in the family it is.

the 6 logic verbs

The 6 logic verbs do arithmetic on signals from 0 to 10 V.

and, or	the lower or the higher of a and b
xor	the difference between a and b
nnd, nor, xnr	10 V minus and, or and xor

With gates as inputs, they work as logic.

sum and rng

sum and rng pass bipolar signals through unchanged. …

cmp, xfd and cn% choose between inputs

cmp, xfd and cn% decide between the 2 inputs.

cn% tosses a coin …

s+h and sft remember a value

s+h and sft remember a value.

If nothing is patched into a, s+h samples its own noise instead, between -5 V and +5 V. With a clock into b and a left empty, the cell is a random voltage source. … cmp, xfd, cn%, s+h, sft and r+f each have one number. It is shown above the verb field. Drag it to change it.

control the number with cv

The voltage at the cv jack is added to the number you set. The result stays within the range the verb can use.

10 V covers the whole range, and 5 V covers half of it.

clock divisions from the transport

The top row gives you the transport clock 4 times over. Each of its 4 jacks sends the clock divided or multiplied by the number above it.

The numbers go from times 4, through every division, to 16. Each output is a gate that lasts half of its own period.

By default, the 4 jacks read as follows, from left to right:

8	a half bar
4	a beat
2	an eighth note
times 2	a 32nd note

a clock jack, normalled to the transport

The jack at the left of the bus row is its clock. Empty, the row counts the system’s 16ths, and the word bus with the link mark above it says so. Patch a cable in, and the row counts the rising edges of that signal instead.

Any signal works. A slower clock or a gate line makes the dividers follow it. An oscillator makes the row a sub-oscillator: ÷2 is a square wave an octave below it, ÷4 two octaves, and ×2 an octave above.

the transport bus, and saved data

atelier takes the tempo, run state and reset from the transport bus for its top row.

atelier’s verbs, numbers and divisions are saved in the module’s own data. So randomising the system leaves atelier as it was.

a random voltage, and a clock with gaps

1. Set one cell to s+h, and leave its a empty.

…

1. Set a second cell to cn%. Patch the clock into its a, and leave its b empty.

The second cell’s out is the clock with gaps in it. Its number sets how many pulses get through.

To make the odds change over time, patch the first cell’s out into the second cell’s cv.

a window comparator

A window comparator gives a gate that is high only while a signal is between 2 levels.

1. Set 2 cells to cmp: the first for the lower level, the second for the upper level.
2. Feed the second cell’s out through a cell set to nnd.
3. Set another cell to and. Patch the nnd cell’s out and the first cmp cell’s out into it.

The first cmp gates while the signal is above the lower level. The second gates while it is above the upper level. The final and is high only between the two.

one scale for 8 quantising lanes

gamme is 8 hp wide. It has 8 quantising lanes and a wheel.

Each lane has its own attenuverter, semitone offset, octave and glide. All 8 lanes snap to the same set of notes. You draw that set by clicking dots on the wheel.

gamme stores a root, plus which of the 12 notes are in the scale and which are out. This set of notes is called the mask. gamme shares the root and the mask with the rest of the system.

A mask can hold any named scale, or any set of notes at all. So every module that reads it agrees with gamme. The wheel is the main control, and the presets are starting points.

8 lanes, up to 16 voices each

Every lane is polyphonic. A lane’s output has the same number of channels as its input, up to 16, quantised. Each lane works independently, so gamme can quantise up to 128 voices.

A lane with no cable still makes a pitch. It quantises 0 V plus its offset, and adds its octave. So an unused lane gives a tuned drone or a reference pitch.

The dice takes modulation only from lanes with a cable in their input.

click a dot to add or remove a note

A click only counts within a few pixels of a dot’s centre. This stops a near miss from becoming a wrong note.

You can always add a note. The last note always stays.

a custom set moves with the root

When you click a dot, gamme stores it as an interval above the current root. So the mask is a set of intervals.

If you change the root, a set you drew moves with it.

the root dot and the live display

The root is always the big blue dot, even if you remove the root from the scale. Removing it only dims its label.

A scale without its root is allowed, and it sounds like one: a set that keeps stepping past its home note.

The wheel also shows the notes that are passing through the lanes.

the root and scale cyclers

Under the wheel are 2 text buttons, each made of 2 halves.

The left button sets the root, from c to b. The right button sets the scale. It has 15 presets: chromatic, the 7 modes, harmonic and melodic minor, the 2 pentatonics, blues, whole tone and an 8-note diminished scale.

off: semitones added before quantising

off adds semitones to the input before the quantiser. The quantiser then puts the result back in key.

off counts semitones. In a 7-note scale, a degree is about 2 semitones, so about half the values give the same note as their neighbour.

oct: octaves added after quantising

oct runs from minus 4 to plus 4 octaves. gamme adds it after quantising.

So the octave is exact, and the lane stays in key whatever the mask.

slw: glide time

slw sets the glide, from 0 to just under 1 second. The value is roughly the time to arrive at the new note, so the top of the range settles in about a second. At 0.5 ms or less, the glide is switched off.

A switch in the menu makes the glide climb through the scale.

atn: before or after the quantiser

atn scales the incoming pitch, from minus 100% to plus 100%. The default is plus 100%.

A menu switch sets which side of the quantiser atn works on. This changes what the control does.

gamme shares the key

gamme sends its root and mask on every processing cycle. When it is linked and its key changes, it writes the system’s shared key.

Any linked module can write the key, and every linked module follows it. crible also writes the key when its own key control moves.

These modules follow the key: crible’s filter bank, xtrack’s quantiser on each track and clavier’s note engine.

the key stays when gamme leaves

gamme also sends an older key signal, whether it is linked or not. Followers use the shared key instead, as soon as anything has written to it.

So if you unlink gamme, the system keeps the last key it was given, and stops following gamme. If you delete gamme, the system also stays in its current key.

A cable in root or scal makes gamme the source of the key.

a 4-voice harmoniser from one melody

One sequencer is enough for all 4 voices.

Set the root to c and the scale to minor. Patch one melody into lane 1, and also into lanes 2, 3 and 4. Set atn to full on all 4 lanes.

lane 1	oct minus 1, off 0, slw 0
lane 2	oct 0, off 3, slw 0
lane 3	oct plus 1, off 7, slw 40
lane 4	oct plus 2, off 10, slw 90

3, 7 and 10 are all in the minor scale, so they pass through the quantiser unchanged. An offset of 4 would tie between 3 and 5, and snap down to 3.

The 2 lanes with slew lag a little, so the chord arrives gradually instead of all at once.

draw a scale for the whole system

1. Set the scale to the minor pentatonic.
2. Click 3 or 4 dots off the wheel, until 4 notes are left. The scale cycler now reads custom.
3. Turn on wheel in fifths to see the shape as an arc.
4. Leave link on in the scale row. Set crible’s key control and xtrack’s tracks to any scale other than off.

Every module that quantises takes on those 4 notes at once, because gamme shares the notes. A module follows while it is quantising.

On the fifths layout, step the root one dot clockwise. This moves the system to the dominant key in one click.

To go back to the presets, use reset custom scale. Turn on this module is the master to make gamme the only module that sets the key. When it is off, any other linked module can change the key too.

a module that acts on other modules

liasse acts on the rest of the system. It holds a list of up to 8 modules that you pick, and it is 3 hp wide.

It does 3 things to that group:

randomise	changes the settings of the whole group at once
walk	fires the members one at a time
bypass	takes the whole group out of the patch, or brings it back

Randomising is the main job. You fire it with the button or a trigger, and it can glide to the new settings. The walk, skew, quantise and deep randomise all support it. The fire button is the largest control, below the display.

3 inputs take triggers and gates, and 1 output sends triggers.

how the display draws the group

The display plots each member as a point. Across is the member’s position in the system. Down is its pick order.

A group from one voice, picked in the order it sits, draws a tight diagonal. A group spread across the system, or picked out of order, zig-zags.

The horizontal axis spans the members.

pick the modules in the group

1. Turn on pick, the blue toggle on the left. A transparent layer covers the whole system.
2. Click the modules you want.

the pick order shown on the system

Each member gets a thin outline in its own colour. A small tag at its top left shows its pick number in white.

This number is the member’s place in the display’s gutter, in the walk and in the skew order. So the order of your clicks stays visible on the system.

up to 8 members

While the picker is on, point at a module that is not a member, and it takes a white outline. This means a click will do something.

When 8 modules are picked, only members highlight, so you can remove one and pick another. The outlines only show while the picker is on.

To control more than 8 modules, patch this liasse’s output into a second liasse.

hold a member

Click a member’s node on the display to hold it. A held member draws hollow, and its number in the gutter dims.

The group randomise and the walk both skip a held member. It still bypasses with the rest of the group.

Use a hold to keep a module out of the randomise without removing it. Click the node again to release it.

drag a node to set how far it moves

Press a member’s node and drag. Up is more, down is less, and the whole range takes about 90 pixels. The node’s size shows the amount. A faint ring stays at full size, so you can see how much is held back. While you drag, the gutter shows the amount as a percentage.

The amount sets how far a fire moves that member. At 100% each control goes to the value it rolled. At 20% it goes a fifth of the way from where it is towards that value. For a switch, the amount is the chance that it changes at all.

Double-click a node to set it back to 100%. A press that does not move still holds the member.

the halo and the ring

The halo shows where the next step will land. The ring shows the dwell.

liasse draws both by running a copy of the real walk. So the picture always matches what happens.

members that are still loading

When a patch loads, modules arrive over several frames. liasse waits about 2 seconds for them.

During that time, a member still loading draws as a hollow dash in the centre.

per: triggers on each member

per sets how many step triggers land on the same member before the walk moves on. It is the number field on the left, from 1 to 8 in whole steps. The default is 1.

Each step trigger always fires exactly one member. At 4, for example, 4 triggers in a row fire the same module, and the fifth moves to the next.

dir: the order of the walk

dir is the field on the right. It picks one of 4 orders for the walk. The shuffle makes a new random order every cycle.

liasse makes a new shuffle whenever the set of active members changes.

To stay longer on any member, including an end, use per. You can randomise both fields.

randomise the whole group

Press the large violet button at the bottom. Every un-held member of the group is randomised at the same instant.

The button is a real parameter, so you can map it to a hardware controller like any knob.

what a randomise changes

For each un-held member, liasse goes through the module’s parameters. It changes each one that allows randomisation and has a fixed range.

slew glides to the new values

At a slew of 0, every new value is set at the instant of the fire.

Above 0, each parameter moves in a straight line to its new value, at an even speed. A slew of 500 ms reaches the target at exactly half a second. So a group of 5 modules gliding together arrives together.

Parameters that snap to steps still jump.

bypass the whole group

The teal on toggle switches the whole group in or out of the patch at one instant.

When you click it, the change is one undo step: liasse’s own setting plus each member’s bypass.

patch a clock to drive liasse

A clock reaches liasse only through the rnd and step jacks.

lead one liasse from another

pas sends one trigger per member fired, so you can hear the walk as a rhythm.

pas into an envelope’s trigger	every hand-off gets a shape
pas into a sampler’s trigger	every change of member plays a sound
pas into a second liasse’s step or rnd	the pair control more than 8 modules, and the first sets the pace of the second

a slow drift from 5 modules

Pick 5 modules from one voice.

slew	about 1.2 seconds
skew	about 800 ms
quantise	off
rnd	a clock divided down to 1 pulse every couple of bars

The slew is linear, so every parameter arrives exactly on time. The skew spreads the 5 fires across most of a second, so the voice walks rather than lurches. The 5 pulses on pas are about 200 ms apart.

a hard cut from the same modules

slew	0
skew	0
quantise	4 steps
rnd	a trigger on the downbeat

With 4 grid steps across every range, each fire sounds like a change of preset. With quantise off, the same fire lands somewhere new each time.

a rhythm made of modules

Pick 6 modules, then set:

dir	png
per	2
step	an eighth-note clock
pas	into an envelope’s trigger
held members	3 and 5. Click their nodes

The walk plays 1, 1, 2, 2, 4, 4, 6, 6, 4, 4, and back. It closes over the held gaps, and each end fires once per visit. With per at 2, the ring takes 2 triggers to fill.

chain 2 liasses past 8 modules

A chained pair controls 16 modules from one clock.

1. On liasse A, pick 8 modules. Set dir to fwd and per to 1, and patch a clock into step.
2. Patch A’s pas into liasse B’s step.
3. On liasse B, pick 8 more modules, and set dir to shf.

A sends one pulse for every member it fires, so B moves on once per step of A. A walks in order, and B shuffles.

To make each step of A fire a burst on B, patch pas into B’s rnd instead, and give B some skew.

16 slots that link a cv to a knob

qmap is a 3 hp panel with 16 slots. A slot is one jack, with a small round button above it. The slots are arranged in 8 rows of 2.

The jack takes a voltage. The button links the slot to a control on another module. While the slot receives a voltage, that voltage moves the control.

Signals go from the cv into the control.

The number next to each button counts across the whole run of q modules. A second qmap next to this one numbers its slots from 17.

what the lights show

The button above each jack is the slot’s control, and also its light. It shows 4 states.

amber, breathing	armed, and waiting for you to touch a control
solid amber	mapped. Its brightness follows the incoming voltage
grey	an empty slot, with no voltage coming in
white	a voltage is arriving at an unmapped slot

Next to a modulator, every unmapped slot shows white.

A small white dot in the centre of a jack means something different. It shows that a partner in the array feeds this slot. A plain jack takes a cable.

how slots pair with modulators

Slots pair by position in the run.

For example, take a run of 2 qmaps and 1 modulator. The first qmap’s jacks all have dots. The second qmap’s slots are 17 to 32, and a second modulator feeds them.

initialise and randomise

qmap has 16 inputs. It saves its 16 mappings as its own data. This changes 2 menu items.

randomise	keeps every mapping. The dice also skips qmap
initialise	clears all 16 mappings. It sets every slot to unipolar, with an attenuator of 1 and no offset, and joins the array again

The menu item clear all mappings does less: it removes the mappings, and keeps each slot’s settings.

arm a slot

Click a slot’s round button to arm it. The light breathes amber. Arming a slot disarms any other, so one slot is armed at a time.

To disarm the slot, click its button again.

An armed slot waits for as long as it takes. It maps the next control you press anywhere in the system, whatever you pressed it for.

A reopened patch starts with every slot disarmed.

map a control

1. Arm a slot.
2. Press a knob, slider, switch or number field on any other module, with the left mouse button.
3. Release the mouse button.

The slot is now mapped. The light turns solid amber.

what a slot maps

A slot maps any knob, field or switch in the system, except those on qmap, qmod and qmod+.

the amber square on a mapped control

A mapped control shows a small amber square at its top right corner.

Right-click a slot’s button to clear its mapping.

settings for each slot

Right-click any aux jack to open that slot’s settings. Each of the 16 slots has its own settings.

how a slot changes the voltage

A slot works in this order:

1. It multiplies the voltage by the attenuator.
2. It adds the offset.
3. It uses the polarity to turn the result into a position on the control.

On a unipolar slot, 0 V sets the control to its minimum, and 10 V sets it to its maximum. 7.5 V sets it to three quarters of its travel.

attenuator 0	removes the incoming signal. The offset alone sets the value, so the offset slider works as a remote knob for the mapped control
attenuator 0.5, offset 2.5 V	a 0 to 10 V source moves the control over the middle half of its travel
attenuator -1, offset 10 V	a 0 to 10 V source moves the control from its top to its bottom

when a mapping uses only half its travel

Check the polarity if a mapping only moves the control over half its range.

A signal that swings above and below zero, into a unipolar slot, spends half its time below zero. The slot holds the control at its minimum for that half, so the control uses only the bottom half of its travel.

A 0 to 10 V envelope into a bipolar slot uses only the top half.

the array

An array is a row of qmap and modulator panels, side by side and touching. Placing them together is all the setup it needs.

To take a panel out of the array, turn off its own join array item.

a modulator next to qmap feeds every slot

Place a modulator from the qmod family next to a qmap, and it feeds all 16 slots:

each jack	shows a white dot in its centre
each slot that is not mapped	glows white
the status line	counts the pairing

A cable into one jack replaces the modulator for that slot only. So you can mix sources: 15 slots from the array, and 1 from anywhere in the system.

In the plugin, the system is already set up this way. One wide modulator, 2 compact ones and 3 qmaps stand together. All 48 mapping slots are paired, and the wide module’s 8 rate knobs set all their speeds.

map 16 controls with no cables

1. Open the module menu, and choose arm all. A light flashes amber.
2. Click 16 controls anywhere in the system, in any order. After each click, the next slot is armed.
3. Place a modulator from the qmod family right next to the qmap, with join array on for both. It feeds all 16 slots.
4. Set each slot’s polarity to suit the modulator’s outputs. Use bipolar for the default range of -5 V to +5 V, so each mapping uses its full travel.

You now have 16 modulation routes, and no cables. For slots 17 to 32, add a second modulator and a second qmap. For slots 33 to 48, add a third pair. The numbers of modulators and qmaps need to match.

Instead of the modulator, you can also patch one cable into each slot.

one movement, 3 different responses

1. Feed the same slow source into aux 1, 2 and 3. Use a splitter, or 3 slots from one modulator row.
2. Map the 3 slots to 3 unrelated controls.
3. Set the slots as follows.

aux 1	attenuator 1, no offset. The first control sweeps its full travel
aux 2	attenuator about 0.33, offset about 3 V. The second control moves through the middle third of its travel
aux 3	a negative attenuator, offset near 10 V. The third control closes as the other two open

To add a remote knob, set a 4th slot to an attenuator of 0, and leave it unipolar. Its offset slider now sets the mapped control.

That slot still needs an input. A slot writes only while it has an input. A slot that writes a fixed value also holds its control in place when you roll the dice.

16 modulation outputs in rows and columns

qmod is a 5 hp module with 16 separate modulation sources. The column sets the mode, and the row sets the rate.

the control strip	2 hp on the left. It holds the trigger jack at the top, and 8 rate knobs, one level with each row
the output block	3 hp on the right. It holds 2 columns of 8 jacks, each with a small round light above it
the mode fields	one at the head of each column

Each of the 8 rows holds 2 slots, left and right.

the rate knobs

Each knob multiplies the module’s global rate for one row. It shows the multiplier.

centre	1 times
fully left	0.01 times
fully right	100 times

The knob is logarithmic, so each side covers 2 decades. Double-click a knob to return it to 1 times.

Together, the global rate slider and the knobs reach from 0.0001 Hz to 1000 Hz. A row turned fully right can run at audio rate.

Both slots in a row always run at the same speed.

rate knobs in an array

In an array, the leftmost qmod’s row knobs set the rates for every member, including a second qmod. So 2 qmods side by side have 8 working knobs between them.

The second qmod’s slots use the first qmod’s row rates, multiplied by its own global rate slider. Turn off join array to give it back its own knobs.

what initialise and randomise change

The 8 knobs are qmod’s parameters. The modes, ranges, attenuators, offsets and slews are saved as the module’s own data.

randomise	moves the 8 knobs only. The speeds change, and the character of the bank stays the same
the dice	also changes the modes, ranges, trims and slews. It favours the smooth shapes, and keeps each rate above a minimum, so a slow row never looks like a dead cable
initialise	both columns to smooth random. Every slot to -5 V to +5 V, with no attenuation, offset or slew. The global rate to 4.5 Hz, smoothness off, and the waveshape to sine

the mode fields

The field at the head of each column sets the mode for all 8 slots in that column.

click	steps to the next mode
drag up or down	selects a mode, about 12 pixels for each mode. It wraps round at both ends

A click or a whole drag is one undo step. The field shows 3 letters over a short line in the mode’s colour.

the slot lights

The light above each jack is also a button. Click it to step that slot to the next mode, apart from the rest of its column. Each click is one undo step.

Point at a light to see its slot number and its mode in full. The light’s brightness shows the raw shape of the slot.

the 6 modes

Each mode shows 3 letters in its own colour.

smo	smooth random
lfo	a repeating wave, in the module’s waveshape
s+h	sample and hold: a stepped random value
trg	triggers
gat	gates
tsh	triggered sample and hold

settings that apply beyond one slot

The waveshape is one setting for the whole module. For 2 waveshapes at once, use 2 modules.

There are 7 waveshapes, on the menu under lfo waveshape.

sine	the smooth default, and the one to use at audio rates
triangle	straight lines up and down
square	full up for half the cycle, full down for the other half
saw up	climbs the whole cycle, then jumps back
saw down	falls the whole cycle, then jumps back
peaks	rests near the bottom and rises to a point, an exponential curve each way
valleys	the same curve upside down: it rests near the top and dips

trg and gat fire at a varying interval. The rate sets the average interval, so a row of triggers or gates lands unevenly.

trg and gat send 0 to 5 V, whatever the range. Their attenuator and offset still apply. Every other mode uses the range.

what the array shares

qmod’s timing comes from its global rate slider and its 8 knobs.

A pulse into the trigger jack restarts all the slots on each edge.

Two things travel along the array from qmod: the 8 row knobs, and the trigger edge.

feeding a qmap

qmod feeds a qmap by standing next to it in an array.

Each output feeds the qmap slot at the same position in the array. The white dot in a qmap jack shows this pairing.

Modulators and qmaps are numbered separately. So the first modulator’s slot 4 always pairs with the first qmap’s slot 4, in whatever order the panels sit.

In the plugin, qmod heads a run of 6: itself, 2 qmod+ modules, then 3 qmaps. Its 8 knobs set the rates of all 48 slots, and its trigger jack resyncs every one.

8 speeds that drift apart

1. Set both column fields to smo.
2. In the menu, set global rate to 0.60 Hz, and stagger spread to 0.100.
3. Choose apply stagger to row knobs.

Row 1 turns over in under 2 seconds, and row 8 takes about 17 seconds. The 16 bipolar outputs drift apart.

one row at audio rate

An audio-rate row works for ring modulation, or to modulate another voice.

one column	lfo
waveshape	sine
global rate	10 Hz
row 1 knob	20 times

Both of row 1’s jacks now run at 200 Hz, while the other 7 rows stay slow. Turn the knob to 100 times for 1 kHz.

Use sine at audio rates.

struck envelopes and sloped gates

1. Set the left column to trg.
2. On each left jack, set slew to about 0.50, and slew shape fully left, to -1.
3. Set the right column to gat.
4. On each right jack, set slew to about 0.60, and slew shape to 0.

The left jacks now rise instantly and decay over about 150 ms. The right jacks send gates with sloped edges.

Leave the module’s smoothness at 0. If you skip the slew shape, or raise smoothness, the trigger is too short to drive the slew. The row then goes almost silent.

one speed heard 2 ways

left column	lfo
right column	s+h
waveshape	triangle
row knobs 1 to 4	1, 0.5, 0.25 and 0.125 times
row knobs 5 to 8	the same again
global rate	2 Hz

Each row sends a smooth triangle on its left jack and a stepped random value on its right jack. Both slots share the row’s knob, so they stay locked to the same period.

16 modulation outputs in 3 hp

qmod+ has 2 columns of 8 output jacks, each with a small round light above it. Each column has one mode field at its head. … The number next to each light counts across the whole run of q modules. A second modulator in the run numbers its slots from 17.

the plus means more slots

The plus means the extra slots qmod+ adds to a system. …

on its own

it is still complete: 16 sources, all 6 modes, the same ranges and the same settings, on its own internal rate

what initialise and randomise change

All qmod+’s settings are saved as the module’s own data.

the dice	changes the module’s data directly: modes, rate, stagger, spread, waveshape, smoothness, ranges, attenuators, offsets, slews and slew shapes
initialise	both columns to smooth random. Every slot to -5 V to +5 V, with no attenuation, offset or slew. The global rate to 4.5 Hz, stagger on at 0.100, smoothness off, and the waveshape to sine

turn off the stagger to use qmod+ as an expander

With stagger rates off, all 16 slots run at exactly the global rate.

If a qmod is also in the array, each slot runs at the global rate multiplied by that qmod’s knob for its row. qmod+ then gives you 16 more slots on the same 8 row rates.

The global rate of qmod+ scales all 16 slots together. For example, set it to half of the qmod’s global rate, and you get a second bank at half speed on every row.

resync from a qmod in the array

A qmod’s trigger jack in the same array resyncs qmod+. … The qmod’s other readings, gate, rate, smoothness and mode, act on the qmod alone.

16 speeds from the stagger

Standing alone, qmod+ can give 16 slow, smooth outputs at different speeds.

1. Set both column fields to smo.
2. In the menu, turn stagger rates on.
3. Set global rate to 1.20 Hz, and stagger spread to 0.020.

Each slot runs at about 77% of the speed of the slot before it. Slot 1 turns over in under a second, and slot 16 takes about 40 seconds. That is a spread of 50 to 1. In every row, the right slot is a step slower than the left.

qmod+ as an expander

1. Place qmod+ next to a qmod.
2. Turn join array on for both modules.
3. On qmod+, turn stagger rates off.

The spread closes, and each row of qmod+ follows the qmod’s knob for that row. The global rate of qmod+ scales all of its slots together.

3 accents among slow drift

1. Set both column fields to smo.
2. Click 3 slot lights until they show trg.
3. On each of those 3 jacks, set slew to about 0.50, and slew shape fully left, to -1.

Each of the 3 slots now rises instantly and decays over about 150 ms. The other 13 slots drift.

Leave smoothness at 0. Smoothness sets a minimum for every slot, and the 3 trigger slots would go silent under it.

a stepped random value on each bar

Use this, with a qmod in the array, to get a random voltage that changes once a bar.

1. Click a 4th slot light until it shows tsh.
2. Patch a pulse on each bar into the qmod’s trigger jack. Leave its reading at the default, trigger / resync.

The slot holds a random voltage, and takes a new one on each bar.

Set the slot’s range to -1 V to +1 V, and its attenuator to 0.17, to transpose a pitch input by up to a tone.

each ring moves a playhead over 4 lanes

Each macro ring on quatre sets a position. Turning the ring slides a vertical playhead across a picture of 4 painted lanes.

Where the playhead stands, each lane’s painted height becomes that lane’s output voltage. So one movement of your hand does 4 things at once.

A lane painted as a rising ramp goes up as you turn the ring. A falling one goes down. A hump rises and then falls again.

quatre has 4 macros. Each is a large ring with its own cv jack in the centre, and each has 4 lanes and 4 output jacks. The 16 outputs sit in 4 rows of 4: one row per macro, in 4 shades of that macro’s colour.

The window shows the 4 lanes of one macro at a time: the macro whose ring you touched last.

the shape shows the output

Each lane is drawn as one filled shape. The shape is the lane read exactly as the output reads it, with the same smoothing, across the whole travel of the ring.

What sits under the playhead is what leaves the jack. If you paint a hard edge and it is drawn as a slope, the output is a slope too. The smooth setting causes this.

So you can read the output at any ring position before you turn the ring.

the 4 maps quatre starts with

A new quatre has all 16 lanes already painted, with a different map on each macro:

macro one	a four-way scanner. Its 4 lanes always add up to the same total
macro two	overlapping ramps and shelves
macro three	4 hard gates with gaps between them and no smoothing, so a painted edge acts as a switch
macro four	a fall, two islands, a rise and a flat shelf

Initialise restores these 4 maps. To empty one lane or one macro, use the clear items in the menu.

the lane you touched opens

The 4 lanes share the window’s height unequally. The lane you touched last opens to about half the window. The other 3 close to narrow strips.

The open lane has nearly 30 pixels of height for the full range: enough to paint a level by eye. A closed lane has 10 pixels: enough to read a shape.

To show a different macro, touch its ring.

3 values from different stages

The window shows 3 different values at once, and each comes from a different stage.

During a fast turn, the dots lead the arcs by a few milliseconds. This is the output smoothing, which stops zipper noise.

The arcs are the only place that shows all 4 macros at once. You can read the whole module from a distance.

The pill sits at the foot of the window, and clears when you release. Its position matches the ring: 60% in the readout is where the ring is 60% round. When you erase, the height reads 0, because erasing paints 0.

every press paints

Every press paints, and the first press is already a stroke.

A click meant only to change lanes still leaves a mark, 1 cell wide (2% of the travel).

However fast you move, the line stays continuous. A stroke that jumps 40 cells in one frame fills all 40. The brush follows the pointer at any zoom level.

shift is read once per stroke

quatre checks shift when you press. The whole stroke either paints or erases. To switch, release and press again.

Hold shift as you press to erase instead of paint. The header shows this in small type.

undo works per stroke

Each stroke is one undo step, however many cells it wrote. The step is added when you release, and only if a cell changed.

The menu’s clear items are permanent.

a lane is 50 cells

A lane is 50 cells, each with one level. Each cell is 2% of the travel. One cell is the smallest shape you can paint.

Reading a lane gives a line.

smooth: a gate or a crossfade

smooth sets how many neighbouring cells are averaged into each reading. It has 6 stops: 1, 3, 5, 7, 9 and 11 cells.

This setting decides whether a painted edge becomes a gate or a crossfade.

a front panel for your patch

quatre has 16 output jacks. Everything that reaches it arrives by cable or by hand, and everything it moves, it moves by cable.

This makes quatre the front panel of a patch. Its 4 rings, on one small module, stand in for 16 knobs spread across the system.

a scanner that adds up to 1

One ring crossfades between 4 destinations.

1. Touch macro one’s ring. Open the menu and choose hand-off preset.
2. Leave the output range at 0 to 10 V, and leave the smoothing as the preset set it.
3. Patch the macro’s 4 outputs to 4 destinations.

Turn the ring to crossfade between the 4. The level stays even through the seams, and clean at both ends, where the smoothing repeats the edge value. This shape is hard to paint by eye.

4 gates with gaps between them

The ring becomes a 4-way selector, or a step sequencer you play by hand.

Touch macro three’s ring, and leave the map as it starts. It has 4 windows: full height, about 70%, full again, and just over half. There are gaps between them, and no smoothing on any lane.

Turn the ring to step through 4 states. Between them, every output sends 0.

Patch the 4 outputs into 4 gate inputs. The ring becomes a performance control with 4 positions.

one painting as a gate or a crossfade

1. Touch a ring. On the top lane, paint a full-height block from about a third of the way along to about two thirds. Drag above the lane so the top comes out clean.
2. Set that lane’s smoothing to the lowest stop, and turn the ring through the block.
3. Set the smoothing to the highest stop, and turn the ring through the block again.

At the lowest stop, the block is a gate with a hard on and off. At the highest, it is a crossfade. Each edge becomes a ramp about a tenth of a turn wide, with a flat top between them.

a macro that pulls down as well as up

1. Set the output range to ±5 V.
2. Paint one lane flat at 50% across the whole travel.
3. Drag the left third down to 0, and the right third up to full.
4. Set that lane’s smoothing to the second or third stop.

With the ring in the middle, the destination is untouched, because 50% is 0 V in this range. Turn left to pull it to minus 5 V. Turn right to push it to plus 5 V. The swing is still a full 10 V.

16 lfo shapes that you draw

A rising saw into a macro’s cv jack sweeps the playhead forward on its own. Each lane then repeats its painted shape, so a lane becomes an lfo with a shape you drew.

1. Turn one macro’s ring to 0%.
2. Patch a rising saw into its cv jack.
3. Paint that macro’s 4 lanes with the shapes you want to hear.
4. Patch its 4 outputs to 4 destinations.

The saw’s rate sets the speed of the cycle. The playhead runs left to right and jumps back, so the lanes stay in step with each other and with whatever clocks the saw.

A painted lane can hold flat for half the cycle, rise in 3 steps, or swell once and stop. Set a lane’s smoothing to the lowest stop and its edges become a step sequencer. Set it high and the same painting becomes a slow sweep.

Each macro has its own cv jack, so 4 saws at 4 rates give you 16 modulation lanes, each with its own shape and its own speed.

2 envelope followers with separate timing

sillage is a pair of envelope followers. An envelope follower tracks how loud a signal is, and turns that level into a cv. The 2 followers share only the panel.

Column a is one follower and column b is the other. Each column reads straight down: in, atk, rel, thr, env, gate. Every control on the module is a field, a menu item or a drag on the display.

why there are 2

Different sounds need different timing, called ballistics.

a kick	attack near 0.1 ms and release around 65 ms, so the gate closes between hits
a pad	attack of 20 ms and release of 300 ms, so a filter breathes

These are 2 different followers. The defaults already set them up this way: lane a for percussion, and lane b for sustained sounds.

how to read the panel

The window at the top draws 2 lanes. Each lane belongs to the column beneath it.

Each label is drawn once, centred between the columns. The line under each value is in its lane’s colour. The jack rows work the same way: one in, one env and one gate label, with coloured rings. So the colour tells you which follower, and the label tells you which value.

The threshold line on a lane is the same blue as that lane’s gate jack ring, because the line is what makes the jack fire.

the band shows the ripple

Each lane shows 2 seconds of history in 48 columns, so each column is about 42 ms wide.

Each column stores the lowest and highest level the envelope reached in that time, and draws a band between them. So the thickness of the band shows the ripple in the envelope.

set release by watching the band

Set rel by watching the band, and set atk by ear.

A ripple of 4 dB means the follower buzzes at twice the source’s pitch, and everything it controls buzzes too. A rectified tone ripples at twice its own frequency, and the smoothing leaves a predictable part of that behind.

For example, on a bass line around 110 Hz:

a release of 3 ms	leaves a band about 4 dB thick
a release of 65 ms	shrinks the band to 0.2 dB, a hairline

The hairline is always drawn just under a pixel thick.

drag a lane to set the threshold

Press anywhere in a lane and drag up or down to move that lane’s threshold. It moves 1 dB per pixel, adjusted for the system’s zoom.

The whole band, 66 pixels tall, is the grab area. The lane you press decides which follower you edit.

Dragging on the display is the coarse adjustment. To type a value, reset to the default, or adjust slowly, use the thr field below the lane.

atk

atk sets how fast the envelope rises. It also sets how fast the gate opens.

It runs from 0.1 ms to 300 ms, on a logarithmic scale, so the fast end has as much travel as the slow end. The default is 3 ms on lane a and 20 ms on lane b.

The envelope rises at this rate whenever the detector is above it. The rise is an exact exponential.

rel

rel runs from 1 ms to 2 seconds. It does 3 jobs at once:

the fall	how fast the envelope falls
rms detector	the averaging window with rms detector on
the door	how fast the door closes when the gate jack carries audio

The minimum is 1 ms, because below that the envelope becomes the signal again.

thr

thr is the level the gate opens at. It runs from minus 60 dBFS to 0 dBFS, where 0 dBFS is 10 V.

The default is minus 24 on lane a and minus 30 on lane b.

The blue line on the display and this field set the same value. Drag the line and the number changes. Type a number and the line moves.

what sillage feeds

sillage only ever reads its input.

It makes other modules respond to dynamics. Any module control that takes an amount, drive, mix or cutoff from 0 V to 10 V takes exactly what env produces.

A short silent phrase keeps an empty lane moving. It stops as soon as a cable goes in.

an envelope generator for the synth voices

With dc-coupled on, sillage is the system’s envelope generator. It turns a gate into an attack and release shape.

The voice decides the times. A pluck wants an attack near the fast end and a short release. A pad wants a slower attack and a release of hundreds of milliseconds.

a kick opens a filter

atk	0.1 ms
rel	65 ms
thr	minus 24
hysteresis	3
rms detector	off
dB-linear	on

Set these on lane a, and send env a to a filter cutoff. The fast attack catches the click. The release lets the envelope fall between hits. dB-linear keeps the quiet part of the kick inside the sweep.

Watch the band. A hairline between hits, with a spike on each hit, means the release is right. A thick ribbon all the time means the release is too short.

the same lane on a bass line

Move the same lane onto a bass line around 110 Hz, and turn rms detector on. The band gets thinner, at the same release.

a voice envelope

A gate becomes an attack and release envelope.

dc-coupled	on
atk	8 ms
rel	300 ms
thr	minus 30

Set these on lane b, and feed in b a 10 V gate from a keyboard or a sequencer. env b rises over 8 ms, holds while the key is down, and falls over 300 ms.

Send env b to a level for amplitude. Then turn dB-linear on, and send a second copy to a cutoff for the bloom.

With dc-coupled off, the same patch decays to zero while the gate is still held.

a stereo width detector

inputs	left into in a, right into in b
rms detector	on, on both lanes
atk	20 ms
rel	300 ms
thr	the same on both lanes

When the source is mono, the 2 lanes draw the same shape. When it is wide, they separate. So the difference between env a and env b is the width at that moment.

The lanes start with different default thresholds. So first drag both blue lines to the same height. You can line them up by eye, because the display is 1 pixel per decibel.

8 stereo rows, each a signal processor

vanne is a bank of 8 stereo rows. Each row has an in, a mode, a cv and an out, and a sum out under the rows carries all 8.

A row’s mode sets what it does to the signal. The default is vca. The other modes are an attenuverter, an offset, a slew, a rectifier, a sample and hold, and off.

Every row is stereo. One cable into a row’s in feeds both sides.

how to read the panel

The tile at the top holds 8 lanes, one per row, in the row’s colour. Below it the 8 rows read left to right: in, mode, cv, out. The sum jack sits centred under the rows, ringed in all 8 colours.

the lane is the amount, the cv scales it

Each row has one amount, set by dragging its lane. The cv jack scales that amount from nothing to the full amount. With no cable in the cv, the amount stands as you set it, so a lane is a plain manual control until a cable makes it a depth.

drag a lane to set a row’s amount

Drag a lane up or down. The lane reads out in the terms of the row’s mode: a gain in vca, a percentage in atn, volts in ofs, milliseconds in slw.

The outline is the amount you set. The fill is what the row is doing now, after the cv.

unity is halfway up

In vca, the lane runs to 2, so a row can boost by 6 dB. Unity sits halfway, marked with a hairline, and the drag snaps to it.

a lane is a knob

A lane takes every knob gesture: cmd for a fine drag, right-click for its menu and the typed value, and a qmap mapping.

the 7 modes

Click a row’s mode field to step through the modes, drag it up or down to choose, or right-click it for the list.

vca	the signal times the amount, 0 to 2. The cv opens it
atn	an attenuverter. The lane runs from -1 at the foot to +1 at the top, with 0 in the middle. The cv sets its depth
ofs	adds an offset of -5 V to +5 V to the signal. The cv sets its depth
slw	slews the signal. The lane sets the time, 0 to 500 ms, and the cv scales it
rct	full-wave rectifies the signal, then scales it like vca
s+h	samples the signal each time the cv rises through 1 V and holds it. The lane scales the held value. With no cv, the row passes the signal
off	silence

the sum carries every row

The sum jack adds every row’s out together. A row whose own out is patched leaves the sum. Set a row to off to take it out of the sum.

mix 4 sources to one pair

1. Patch 4 sources into rows 1 to 4.
2. Leave the modes on vca.
3. Set each lane to taste.
4. Take the sum out to the mixer.

Each lane is that source’s level, and the sum is the mix. A cv into a row’s cv jack ducks or fades that source.

one envelope, 4 depths

1. Patch one envelope into row 1’s cv.
2. Turn cv normalling on in the menu.
3. Patch 4 signals into rows 1 to 4 and take their outs.

The one envelope reaches all 4 rows. Each lane sets how deep it goes on that row.

a slewed cv, a rectified lfo

1. Patch an lfo into row 1’s in and set its mode to rct. The out is the lfo folded up, at twice the rate.
2. Patch a stepped cv into row 2’s in and set its mode to slw. Set the lane to about 100 ms. The out glides between the steps.