New territory for sound design

On the stock PRO-800, modulation travels along paths printed on the panel: the LFO goes to pitch, pulse width or filter; poly-mod sends oscillator B and the filter envelope to oscillator A or the filter. Those are wonderful paths, and they're all still here. But they're fixed.

This firmware adds a mod matrix: any of 27 sources can be routed, with its own depth, to any of 46 destinations, 16 routes per layer, and two macros gather several sources into one hand. And because the voltages for pitch, filter and volume are worked out separately for each voice, most routes act per note: each note gets its own velocity, its own envelopes, its own random value, its own pressure, even its own LFO.

Every per-voice control now exists twice, as layer 1 and layer 2. Seq 1 and Seq 2 pick which one the knobs edit; a knob then waits until you turn it to the stored value before taking over, so nothing jumps. Each layer has its own oscillator tuning, filter, all four envelopes, glide, two LFOs, poly-mod envelope amount and 16 mod routes. Choose how notes reach them in Settings → 3 LAYr:

Some things stay shared, because the hardware has one wire for them: the waveform buttons, the mixer, resonance, pulse widths, poly-mod oscillator B and master volume. The art of layering here is to make the two layers different through what is per layer: tuning (an octave or a fifth apart), filter and envelopes, glide, LFOs and routes.

Each layer has two LFOs (Arp Up-Dn picks which one the LFO knobs and buttons edit), so four in all. Each has nine shapes, and three settings that change what kind of movement you get:

• Sync (Settings → 1): free-running from 0.05 to 51 Hz, or locked to the clock: MIDI clock from USB or the 5-pin socket, the firmware's own clock (tap the tempo on Sync Clock), or the SYNC IN jack. Locked, LFO FREQ picks one of 15 divisions from 8bAr down to 1-32, including dotted and triplet values. MIDI Start, or holding Sync Clock, lines it up with the bar.
• Retrigger (Settings → 2): when the LFO's cycle restarts. The most interesting one is PoLY: every voice runs its own LFO, started by its own note.
• Delay and fade-in (Shift + LFO FREQ, Shift + LFO AMOUNT): the LFO waits after a note starts and then fades in, per voice, from 20 ms to about 5 s. Vibrato that arrives late, with no route needed.

Pick a shape and a retrigger mode below. Three notes start at different moments (▼) and are held, then released; a new note follows.

Three of the shapes are new kinds of movement rather than new waveforms. drFt glides smoothly from one random value to the next: slow, it's the wandering of old hardware; fast, it's a liquid, unpredictable wobble. nOIS picks a new random value sixteen times per cycle and glides to it: a rough wander at low rates, a fizzing jitter at the top, up to about 800 values a second. StEP plays a pattern of 1 to 16 on/off steps per cycle, which is the next section.

The StEP shape turns an LFO into a little rhythm programmer: up to 16 steps per cycle, each on or off, with an 8 ms slope at every change so nothing clicks. Route it to the VCA at full depth and you have a trance gate; to the filter, a rhythmic wah; to pitch, octave jumps. Each LFO of each layer has its own pattern, saved with the patch.

Click the steps to change the pattern, then press Listen. (A browser approximation of a gated chord at 124 BPM, one cycle per bar, so each step is a 16th note.)

Every voice now has four ADSRs. Filter and amp work as always. Env 3 and Env 4 do nothing until a mod route uses them, and then they can shape anything per note: the depth of a vibrato, a pitch scoop, the poly-mod amount, an LFO's speed. Hold Shift (the Sync Source button) and the filter ADSR knobs edit Env 3, the amp ADSR knobs Env 4.

Better still, any envelope can cycle while the key is held: on Settings → 6, the envelope page, turn VALUE to CYCL or a clock division. The cycle is attack, hold at the top, fall, rest, and again: an LFO that belongs to one note and restarts with it. Free cycling uses the knob times. Synced cycling takes a clock division instead, and the four knobs become weights that share the cycle out, so it can never drift off the beat.

Settings → 6 shows one envelope at a time (En.A, En.F, En.3, En.4; press 6 again for the next) and VALUE picks its mode. Besides the classic ADSR, each envelope can be a DAHDSR, with a delay before the attack and a hold at the top. Arp Up-Dn steps on to the envelope's trigger (tr.A…): dn, from the key going down, or UP, from key up: it rests at zero while you hold the key and plays its whole shape when you release it, in any mode, cycling included. Then its timing (tI.A…): SEC, or SYnC, where the time knobs step through note values (0, 1/32 … 8 bars) at the tempo, so a decay can be exactly an eighth note and stay one when the tempo changes. Then its curve (CU.A…): the analog one, straight lines, or a swell that starts slowly and accelerates. Last, the cycles (Cn.A…) of a cycling envelope: LOOP, or a count of 1 to 99, after which it rests. All of it is per layer and saved in the patch.

What they're for:

• A key-up amp envelope is a note that sounds only after the key: a release "thump", or on a second layer, a note that answers the one you played.
• A key-up Env 3 → CUTOFF opens the filter as a note ends, so the tail blooms instead of fading.
• A key-up cycling amp envelope with a count plays a burst of repeats after each note: synced at 1-16 with 4 cycles, four sixteenths that follow the key's release, like a tempo echo that belongs to each voice.
• A DAHDSR with a long delay on Env 4 → pitch is a pitch blip that arrives late in a held note.
• The swell curve on a pad's amp attack rises like a bowed string; the linear curve gives even, synthetic fades.

Things to try with cycling:

• A synced amp envelope at 1-16: every held chord pulses in sixteenths, with shaped edges and on the beat.
• A cycling Env 3 routed to CUTOFF: each note gets its own filter "wah" that restarts with the note.
• A cycling Env 4 routed to pitch A with a long rest: a recurring blip, like a bird call inside a pad.
• A count of 2 or 3 on a cycling filter envelope: each note gets a short ratchet of filter hits, then settles closed.
• Env 3 with a slow attack routed to LFO AMOUNT: vibrato that fades in on long notes only (or use the LFO's own delay and fade, Shift + the LFO knobs).

Each layer has 16 routes. Each route is a source, a destination and a depth from −99 to +99. You build them from the panel with MOD ASSIGN:

1. Press Arp Assign. Its LED blinks and the display shows SrC. The knobs stay in pickup while you assign, so touching one won't change the sound.
2. Pick the source. A number key (1 LFO, again for the other LFO; 2 filter env; 3 amp env; 4 velocity; 5 mod wheel; 6 aftertouch; 7 Env 3; 8 Env 4; 9 random; 0 note), or touch a knob that stands for one (an LFO knob, an ADSR knob), or turn VALUE to scroll through all 27.
3. Pick the destination. Turn its knob a little, or press its LFO routing button; the knob you touched last before pressing Arp Assign is already offered. AMP SUS means the VCA, the other ADSR knobs mean that envelope's stage, and VALUE scrolls through the rest.
4. Turn VALUE for the depth, then press Arp Assign to finish. Depth 0 removes the route. Hold Arp Assign for half a second to browse the layer's routes one by one.

Pick an idea below to see its routes in the patch bay:

Depths are tuned so the whole range is useful. Pitch routes follow a squared curve: depth 5 is 6 cents, 10 is a quarter of a semitone, 20 a semitone, 50 six semitones, 99 two octaves. Cutoff at 99 is eight octaves. An envelope stage or the glide time scales by up to 16× either way. The VCA destination is a gain on the amp envelope, so it can never open a voice that isn't playing: velocity → VCA at 99 is plain velocity sensitivity, and aftertouch → VCA at 99 keeps a note silent until you press.

Some destinations aren't sounds at all: they're the other modulators. LFO rates and amounts, the LFO routing depths, the filter and poly-mod envelope amounts, every envelope stage and the glide time can all be routed to. That's where sounds start to feel alive, because the movement itself changes over time or from note to note.

A macro is a source and a destination at once. Route whatever you like into Macro 1 (the wheel, pressure, a CC, an LFO) and the sum, −1 to 1, becomes the macro's value; then route Macro 1 out to as many destinations as you want, each with its own depth and sign. The M TUNE knob is Macro 1's own hand knob (master tune is gone; autotune keeps the voices in tune), and Shift + CUTOFF sets a resting offset, so a macro can be a performance knob, a pedal and a modulator all at once. Macro 2 works the same, with its offset on Shift + RESONANCE.

Velocity, aftertouch, a random value and the note's position are all per voice sources. That's what lets a single route make every note of a chord different.

Poly aftertouch is the standout. A controller that sends per-note pressure (the Ableton Move's pads do) moves only the voice playing that note. Route aftertouch to cutoff and one finger can lean into a single note of a chord and open just that note's filter. Channel pressure still works too and moves all voices; each voice uses whichever is larger.

Two layer modes go further. MPE gives every note its own MIDI channel, so each has its own pitch bend (48 semitones), its own pressure and its own slide (CC 74, the tInb source) from an MPE controller. 8CH gives eight mono lines, one per channel, each with its own bend and pressure, for a sequencer that drives them.

Autotune puts every oscillator within a cent, which is great for chords that stay in tune all evening, but perfect can sound static. Drift (Settings → 0, 0–99) puts the life back on purpose: each voice's pitch (up to 15 cents) and cutoff (up to a semitone) wander slowly and independently, picking a new target every second or so and gliding towards it.

Related controls: unison detune (Settings → 8, or Shift + M TUNE) sets how far apart the outermost unison voices are, 0–50 cents; Shift + GLIDE nudges the drift. An LFO on drFt or nOIS with PoLY retrigger, routed to pitch at a small amount, gives a faster, more obvious per-voice wander than the Drift setting. Voice modes per layer (poly, mono, legato, with last, low or high note priority) shape how lines connect.

Set a scale and root on the web page, and each voice's pitch snaps to the nearest note of that scale: the note itself plus the pitch modulation both oscillators get (LFO → Freq, LFO 2 → pitch, routes to pitch A + B). Suddenly random sources play tunes: a sample-and-hold LFO on pitch becomes a melody in key, and a sequence's transpositions can't go wrong. Twelve scales are built in, from major and minor through the modes to pentatonics, blues and whole tone.

The stock sequencers are replaced by something different: each of the 128 patches carries its own step sequence, played from the keyboard. Press a key and the sequence starts, transposed so that a step of 0 plays that key. Press another key while holding and it transposes from the next step without restarting. Let go and it stops. The notes go through the normal voices, envelopes and modulation, so the sequence is part of the sound, not a separate machine.

• Probability from 100 % down to 12.5 %: some steps sometimes don't play, so a loop never quite repeats.
• Ratchets: a step plays as 2, 3 or 4 quick notes.
• Arpeggio mode: instead of transposing, the steps walk the held notes (up, down, up-down, random or as played), and each step's offset shifts the note it plays (+12 for an octave jump).
• Automation lanes: up to four rows of values, optionally gliding between steps, routed like any mod source. Lane → cutoff gives a different brightness per step; lane → Morph gives a different sound per step.
• Recording on the panel: hold Rec and press Seq 1 (rSEq), then play the steps one by one, Seq 2 for a rest, and Rec to finish. The sequence is on, and Rec saves it with the patch.
• A pad with an arpeggio on top: in stack mode, a sequence on layer 2's keys plays layer 2 while layer 1 plays the keys as a pad. Put layer 2 in mono and the pad gets seven voices, the arp one.
• Timing: a clock division at the MIDI clock's tempo or the sequence's own BPM, with swing. Euclid fills, step recording and a layer filter (a sequenced bass in one half of a split, played normally in the other) are on the web page.

Pick another patch as the morph target (web page, Voice section) and route any source to the Morph destination. That source now fades every knob of this patch towards the target's values: cutoff, envelopes, tuning, LFO rates, all at once. Buttons, routes and settings stay this patch's own.

Sixty-four ready-made sounds cover all of this: the bank in presets/ (installed with presets/install.sh) fills P1–P64 with eight each of basses, leads, polys, pads, plucks, sequences, effects and curiosities, and its DESIGN.md explains the techniques. The recipes below are the same ideas by hand. Start from Perf (a plain sawtooth, filter open) unless it says otherwise. Display codes are what you'll see on the synth.

The firmware sets voltages; it can't rewire the voice board. Knowing the edges helps you design around them:

• Shared wires. Waveforms, sync, mixer levels, noise, resonance, pulse widths and the poly-mod oscillator B amount are one setting for all eight voices. Routes to them follow the most recently played note, which works well for mono lines and as a "last note sets the colour" effect in chords.
• One audio output. The voices are summed on the voice board, so the two layers can't go to separate outputs.
• Modulation speed. Software modulation updates every 0.3–2.4 ms (the VCAs fastest), plenty for LFOs, envelopes and gates, but not audio-rate FM. For that, the analog poly-mod path is still there: oscillator B and the filter envelope into oscillator A's pitch and the filter, at audio rate.

For how all of this works inside, see How the firmware works.