Skip to content
Merged
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 1 addition & 1 deletion .github/ISSUE_TEMPLATE/bug.yml
Original file line number Diff line number Diff line change
Expand Up @@ -89,7 +89,7 @@ body:
attributes:
label: Version
description: The release you flashed (e.g. from its file name), or the commit if you built it yourself.
placeholder: v0.7.1
placeholder: v0.7.2
validations:
required: true

Expand Down
2 changes: 1 addition & 1 deletion .github/ISSUE_TEMPLATE/feature.yml
Original file line number Diff line number Diff line change
Expand Up @@ -76,6 +76,6 @@ body:
attributes:
label: Version
description: The release you are using, if the request is about something you've seen on hardware.
placeholder: v0.7.1
placeholder: v0.7.2
validations:
required: false
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,7 @@ REMEMBER: There is no way to go back to the base firmware once you've flashed it

## Status

**v0.7.1, pre-release for beta testing.** Report problems with the bug form under [Issues](https://github.com/Worldwave/Wakes/issues).
**v0.7.2, pre-release for beta testing.** Report problems with the bug form under [Issues](https://github.com/Worldwave/Wakes/issues).

## Flashing

Expand Down
1 change: 1 addition & 0 deletions config/midi.ini
Original file line number Diff line number Diff line change
Expand Up @@ -202,6 +202,7 @@ t_bias = cc 16
jitter = cc 85
deja_vu = cc 31
; t SHIFT page
gtlt = cc 89 ; t gate tilt
gate_length = cc 17
gate_length_random = cc 86
length = cc 103 ; stepped
Expand Down
2 changes: 1 addition & 1 deletion docs/DEFAULTS.md
Original file line number Diff line number Diff line change
Expand Up @@ -133,7 +133,7 @@ A ROTC on MARBLES now leaves the drive alone, deliberately.
| F2 | t BIAS `[D]` | 0…1, detent | centre | centre | |
| F3 | JITTER `[C]` | 0…1 | 0 | 0 | |
| F4 | DEJA VU `[H]` | 0…1, detent | centre = locked loop | centre | one value, shared with the X page |
| t SHIFT F1 | (free) | — | — | — | |
| t SHIFT F1 | GTLT (#20) | −1…+1, detent | centre = off | centre | t1 / t3 gate height; TRIG ignores it |
| t SHIFT F2 | gate length | 0…1 | 0.5 | 0.5 | |
| t SHIFT F3 | gate-length randomness | 0…1 | 0 | 0 | |
| t SHIFT F4 | LENGTH `[I]` | 1…16 steps | 8 | 8 | one value, shared with the X page |
Expand Down
3 changes: 2 additions & 1 deletion docs/MIDI.md
Original file line number Diff line number Diff line change
Expand Up @@ -56,7 +56,7 @@ Each CC reads the way its parameter works -- the same rule Marbles' INTELLIGENT
uses for its outputs:

- **centred** parameters, whose middle is "nothing happening" (FREQUENCY, the attenuverters,
RATE, BIAS, DEJA VU, STEPS): **64 = no change**, 0 and 127 = half a fader's travel down and
RATE, BIAS, DEJA VU, GTLT, STEPS): **64 = no change**, 0 and 127 = half a fader's travel down and
up -- so with the fader on its centre, the CC sweeps the whole range, end to end.
- **one-sided** parameters (LPG colour and decay, LEVEL, JITTER, gate length, SPREAD, and the
stepped ones): **0 = no change**, 127 = a whole travel up -- so with the fader at 0, the CC
Expand Down Expand Up @@ -173,6 +173,7 @@ half still works.
| 16 | 48 | t BIAS |
| 85 | | JITTER |
| 31 | 63 | DEJA VU |
| 89 | | GTLT, the t gate tilt (MARBLES t SHIFT F1) |
| 17 | 49 | gate length |
| 86 | | gate length randomness |
| 103 | | LENGTH |
Expand Down
11 changes: 8 additions & 3 deletions docs/UI-PAGES.md
Original file line number Diff line number Diff line change
Expand Up @@ -239,14 +239,20 @@ and one SETTINGS page (Y) reached the same way as on PLAITS and identical from e
| layer | how you get there | F1 | F2 | F3 | F4 |
|---|---|---|---|---|---|
| **t BASE** | T2 | **RATE** `[A]` (±5 oct) | **t BIAS** `[D]` * | **JITTER** `[C]` | **DEJA VU** `[H]` * |
| **t SHIFT** | T2, then `••` held | *(free)* | **gate length** | **gate length randomness** | **LENGTH** `[I]` |
| **t SHIFT** | T2, then `••` held | **GTLT** † * | **gate length** | **gate length randomness** | **LENGTH** `[I]` |
| **X BASE** | T3 | **SPREAD** `[K]` | **X BIAS** `[L]` * | **STEPS** `[M]` * | **DEJA VU** `[H]` * |
| **X SHIFT** | T3, then `••` held | *(free)* | *(free)* | *(free)* | **LENGTH** `[I]` |
| **SETTINGS (Y)** | `••` double-tap | **Y SPREAD** | **Y BIAS** * | **Y STEPS** * | **Y divider** (12 steps) |

\* 10 % centre detent. **DEJA VU's centre is LOCK**, not its bottom: below the centre the loop is
recorded, above it the loop is replayed, at it the loop is frozen.

† **GTLT** (#20), Wakes' own: tilts the HEIGHT of t1's and t3's gates around t2, which it never
touches. At its detent it is off. Just past the detent both drop steeply to 50 %; towards + t1
falls to 0 % and t3 returns to 100 %, towards − the reverse. **TRIG ignores it** — a t output on
TRIG fires on every gate whatever GTLT says — so it only shapes t outputs routed to LEVEL, FM,
TIMBRE, MORPH or HARMONICS.

- **F4 is ONE control on both pages, not two that are kept in step.** Marbles has one DEJA VU
knob and one LENGTH knob, so the X page's F4 addresses the t page's value rather than
mirroring it. Mirroring made the two pages keep separate pickup state, which could put F4
Expand Down Expand Up @@ -401,7 +407,7 @@ is the most predictable starting point, and t1 / t3 are then yours to route.

| | |
|---|---|
| **model row, BASE** | **Marbles' own output LEDs**: on the t page T1–T3 are the t1–t3 gates; on the X page they are \|X1\|…\|X3\| against the selected range. **T4 is always \|Y\|.** Moving a fader shows that page's four values instead, for 1.2 s. |
| **model row, BASE** | **Marbles' own output LEDs**: on the t page T1–T3 are the t1–t3 gates, each lit at its GTLT height (0 % shows as 8 %); on the X page they are \|X1\|…\|X3\| against the selected range. **T4 is always \|Y\|.** Moving a fader shows that page's four values instead, for 1.2 s. |
| **model row, SHIFT / SETTINGS** | that layer's four values; free faders read dark. SETTINGS breathes. |
| **play row** | **Marbles' clock** in front, stepping once per t2 tick from the `••` end, with the dB meter behind it at ~10 %. (On PLAITS it is the other way round.) |

Expand All @@ -420,7 +426,6 @@ that has nothing on it. This is the supply side for M5 and M6.
| where | what |
|---|---|
| PLAITS | a short `••` + T1, T2, T3 or T4 (held, each is UNPATCH) |
| MARBLES | t SHIFT F1 |
| MARBLES | X SHIFT F1, F2, F3 |
| MARBLES | t SHIFT T4 |
| MARBLES | `••` + FFWD / RWD on the t page and on SETTINGS |
Expand Down
20 changes: 17 additions & 3 deletions docs/UI-SPEC.md
Original file line number Diff line number Diff line change
@@ -1,10 +1,23 @@
# wakes-sp1 — UI specification v0.16
# wakes-sp1 — UI specification v0.17

> **For the flat control map — every control on every page, with nothing else in the way —
> read `docs/UI-PAGES.md`.** This file is the design record: why each decision went the way it
> did, what was rejected, and what is still open. `docs/MARBLES-SETTINGS.md` is the same split
> for Marbles' models, ranges and scales.

v0.17 — 2026-10-06: **GTLT, the t gate tilt, on MARBLES t SHIFT F1** (#20, Adara). Wakes' own
parameter: bipolar with a 10 % centre detent, where it does nothing. Just past the detent t1's and
t3's gates drop steeply to half height; towards + t1 falls to 0 and t3 returns to full, towards −
the reverse. t2 is never touched. A gate stays 0 V / +5 V — +5 V is the positive maximum of every
destination a t output can reach — and GTLT scales the +5 V; the attenuverters and fader / CC
positions do the rest. **TRIG ignores GTLT**, so rhythm stays predictable. Unsmoothed. MIDI: CC 89,
7-bit, centred. Costs nothing per sample: the gain is folded into the routing once per audio block.
**Pickup gets a 3 % lock on every layer change, on both modules** (Adara: a GTLT edit was carrying
into CLOK). After `••` goes down or comes up, a fader's new parameter takes nothing until the fader
has moved 3 % from where the finger is — Plaits' own `POT_STATE_LOCKING` threshold — so the fader's
smoothing tail and a finger still resting on it no longer reach the other layer. Whether the fader
then tracks or catches up is unchanged.

v0.16 — 2026-10-06: **OCTV loses Plaits' LFO range** (#18, Adara). The bottom 1/11 of SETTINGS F1
is now the **full range with no centre detent**: F1 sweeps through C4 without the flat spot and can
be set finely around it. The scale quantizer applies there exactly as in the full range. Routing
Expand Down Expand Up @@ -314,7 +327,8 @@ free-form full-brightness glyph per model, tabulated in `docs/MARBLES-SETTINGS.m
half-brightness "other bank" glyphs of v0.9 are gone: Adara found dimmed glyphs hard to tell
apart in room light, which is the same finding that took the engine flash's `◐` to 33 %.

**Track row, default: Marbles' own output LEDs.** t page: T1–T3 = t1–t3 gates; X page:
**Track row, default: Marbles' own output LEDs.** t page: T1–T3 = t1–t3 gates, a high gate lit
at its GTLT height (v0.17, #20) with 0 % shown at 8 % so a gate firing TRIG never goes dark; X page:
T1–T3 = |X1|–|X3| over the selected range; **T4 = |Y| on both pages.** Stopped, X and Y hold
their last voltage (as the module does). Moving a fader shows the page's four values for
1.2 s (bipolar ones as magnitude, as on PLAITS). The page pattern flashes on a page change and
Expand All @@ -326,7 +340,7 @@ on arriving from PLAITS.
|---|---|---|
| T1 / T2 / T3 | **t1 / t2 / t3 destination**: none `○○○○` → TRIG `○●●●` → LEVEL `○○●●` → FM `●○○○` → TIMBRE `○●○○` → MORPH `○○●○` → HARMONICS `○○○●`. v0.11: the same glyphs as the X side (Adara) — the counting bar v0.10 used on t is gone, and TRIG deliberately shares `○●●●` with the X page's V/Oct | X1 / X2 / X3 destination: FM `●○○○` → TIMBRE `○●○○` → MORPH `○○●○` → HARMONICS `○○○●` → V/Oct `○●●●` → LEVEL `○○●●` (v0.10) → none `○○○○`. One LED under the Plaits fader of that parameter |
| T4 | **Y destination** (v0.14) · held 2 s: **UNPATCH Y** | **Y destination** (v0.14) · held 2 s: **UNPATCH Y**. [J] moved to SETTINGS T4 and now covers X **and** Y |
| F1 | reserved | **free** (v0.10) — was SCALE, now `••` + FFWD / RWD |
| F1 | **GTLT** (v0.17, #20): tilts t1's and t3's gate height around t2; 10 % centre detent = off; not on TRIG | **free** (v0.10) — was SCALE, now `••` + FFWD / RWD |
| F2 | gate length (Marbles' [E]-hold) | **free** (v0.10) — was X CLOCK; X keeps Marbles' own clocking and there is no control for it |
| F3 | gate-length randomness | reserved |
| F4 | LENGTH [I] (5 … 16 steps) | LENGTH [I] — the same value |
Expand Down
Binary file modified docs/Wakes - Manual.pdf
Binary file not shown.
2 changes: 1 addition & 1 deletion firmware/VERSION
Original file line number Diff line number Diff line change
@@ -1,5 +1,5 @@
VERSION_MAJOR = 0
VERSION_MINOR = 7
PATCHLEVEL = 1
PATCHLEVEL = 2
VERSION_TWEAK = 0
EXTRAVERSION =
1 change: 1 addition & 0 deletions firmware/src/main.c
Original file line number Diff line number Diff line change
Expand Up @@ -1767,6 +1767,7 @@ int main(void)
for (int k = 0; k < 4; k++) {
sp.mrb_dest[k] = rt.dest[k];
}
sp.mrb_gtlt = rt.gtlt;
sp1_synth_set_params(&sp);

struct sp1_marbles_params mp;
Expand Down
58 changes: 58 additions & 0 deletions firmware/src/sp1_gtlt.h
Original file line number Diff line number Diff line change
@@ -0,0 +1,58 @@
/*
* wakes-sp1 — GTLT, the t gate tilt (issue #20). MARBLES t SHIFT F1.
*
* Wakes' own parameter, not Marbles': it scales the HEIGHT of t1's and t3's gates around
* t2, which is never touched. Pure C, no Zephyr; sp1_synth.cc applies it and the host
* suites check it.
*
* tilt 0 (the 10 % centre detent) t1 100 % t3 100 % GTLT disengaged
* just past the detent t1 50 % t3 50 % a steep ramp, SP1_GTLT_RAMP
* towards +1 t1 -> 0 % t3 -> 100 %
* towards -1 t1 -> 100 % t3 -> 0 %
*
* Decided with Adara on the issue (2026-10-06):
* - A gate stays unipolar: 0 V low, +5 V high scaled by this gain. +5 V is already the
* positive maximum of every destination a t output can reach, so no INTELLIGENT range
* is involved; Plaits' attenuverters and the fader / CC positions do the rest.
* - ⚠️ TRIG IGNORES GTLT. A t output routed to TRIG fires on every gate whatever GTLT is
* set to, so rhythm stays predictable. Only the CV destinations (LEVEL, FM, TIMBRE,
* MORPH, HARMONICS) see the gain -- sp1_synth.cc reads TRIG from the raw gate bits.
* - Unsmoothed: it changes once per DMA block, like every other routed parameter.
*
* Cost: none per sample. The gain folds into each route's scale, which sp1_synth.cc
* already resolves once per DMA block; at the detent the gain is exactly 1.0f, so the
* output is bit-identical to a firmware without GTLT.
*/
#ifndef SP1_GTLT_H
#define SP1_GTLT_H

#include <stdbool.h>

/* How much of each half of the (post-detent) travel the drop to 50 % takes. 0.1 of a
* half is ~4.5 % of the whole fader: steep, but a ramp rather than a step, so leaving
* the detent never jumps a routed CV by half its height (Adara: "steep ramp"). */
#define SP1_GTLT_RAMP 0.1f

/* The t page's track row draws a high gate at its GTLT height, 0 % shown as this
* brightness (perceptual, before the LED gamma table) rather than dark. 8 % (Adara: 5 % was
* too dim to see at GTLT's extremes). */
#define SP1_GTLT_LED_FLOOR 0.08f

/* tilt: -1..+1, 0 = centre (the UI applies the detent). t: 0..2 = t1..t3. */
static inline float sp1_gtlt_gain(float tilt, int t)
{
if (t == 1 || tilt == 0.0f) {
return 1.0f; /* t2, or GTLT disengaged: exact */
}
const float a = tilt < 0.0f ? -tilt : tilt;
if (a < SP1_GTLT_RAMP) {
return 1.0f - 0.5f * (a / SP1_GTLT_RAMP); /* both: 100 % -> 50 % */
}
float s = (a - SP1_GTLT_RAMP) / (1.0f - SP1_GTLT_RAMP); /* 0..1 past the ramp */
s = s > 1.0f ? 1.0f : s;
/* + attenuates t1 and pushes t3; - the reverse. */
const bool falls = (tilt > 0.0f) == (t == 0);
return falls ? 0.5f * (1.0f - s) : 0.5f * (1.0f + s);
}

#endif /* SP1_GTLT_H */
55 changes: 49 additions & 6 deletions firmware/src/sp1_marbles_ui.c
Original file line number Diff line number Diff line change
Expand Up @@ -8,6 +8,7 @@

#include "sp1_ui_timing.h"
#include "sp1_midi.h" /* MIDI CC offsets (M5a) */
#include "sp1_gtlt.h" /* GTLT's gain curve, for the t LEDs (#20) */

#include <math.h>
#include <stddef.h>
Expand All @@ -21,6 +22,11 @@
* and the bug it is the second half of the fix for, is in sp1_plaits_ui.c -- kept there
* because that is where the symptom was, and it is the same number for the same reason. */
#define CATCH_JUMP 0.25f
/* After a layer change a fader must move this far from where the finger is before its new
* parameter takes anything (pot_controller.h, POT_STATE_LOCKING's 0.03). Without it the
* low-pass tail of the last move, and a finger still resting on the fader, carried a GTLT
* edit into CLOK as the "••" came up (#20). Same rule in sp1_plaits_ui.c. */
#define CATCH_LOCK 0.03f
#define DETENT_BIPOLAR 0.10f
#define TAP_MAX_MS 300u
#define DOUBLE_TAP_MS 400u
Expand Down Expand Up @@ -104,6 +110,9 @@ static float stored[SP1_MUI_LAYERS][4];
static bool catching[SP1_MUI_LAYERS][4];
static float pos[4];
static float prev[4];
static float now01[4]; /* this tick's UNFILTERED positions */
static float lock_ref[4]; /* where each finger was at the change */
static bool locked[4]; /* inside CATCH_LOCK since that change */
static float shown[4]; /* fader positions last reflected */
static uint32_t show_ms; /* value overlay on BASE, counts down */

Expand Down Expand Up @@ -139,7 +148,7 @@ static enum sp1_mui_layer canon(enum sp1_mui_layer l, int i)
static const int8_t midi_dest[SP1_MUI_LAYERS][4] = {
[SP1_MUI_T_BASE] = { SP1_MIDI_D_RATE, SP1_MIDI_D_T_BIAS, SP1_MIDI_D_JITTER,
SP1_MIDI_D_DEJA_VU },
[SP1_MUI_T_SHIFT] = { -1, SP1_MIDI_D_GATE_LENGTH, SP1_MIDI_D_GATE_LENGTH_RANDOM,
[SP1_MUI_T_SHIFT] = { SP1_MIDI_D_GTLT, SP1_MIDI_D_GATE_LENGTH, SP1_MIDI_D_GATE_LENGTH_RANDOM,
SP1_MIDI_D_LENGTH },
[SP1_MUI_X_BASE] = { SP1_MIDI_D_SPREAD, SP1_MIDI_D_X_BIAS, SP1_MIDI_D_STEPS, -1 },
[SP1_MUI_X_SHIFT] = { -1, -1, -1, -1 },
Expand Down Expand Up @@ -208,17 +217,19 @@ static bool is_bipolar(enum sp1_mui_layer l, int i)
{
switch (l) {
case SP1_MUI_T_BASE: return i == 1 || i == 3;
case SP1_MUI_T_SHIFT: return i == 0; /* GTLT (issue #20) */
case SP1_MUI_X_BASE: return i >= 1;
case SP1_MUI_Y: return i == 1 || i == 2;
default: return false;
}
}

/* Faders with nothing bound to them. They read dark and are ignored. X SHIFT F1 (was
* SCALE) and F2 (was X CLOCK) joined the list in M4a -- see the header. */
* SCALE) and F2 (was X CLOCK) joined the list in M4a -- see the header. t SHIFT F1 left
* it when it became GTLT (issue #20). */
static bool is_reserved(enum sp1_mui_layer l, int i)
{
return (l == SP1_MUI_T_SHIFT && i == 0) || (l == SP1_MUI_X_SHIFT && i <= 2);
return l == SP1_MUI_X_SHIFT && i <= 2;
}

static enum sp1_mui_layer base_of(enum sp1_mui_page p)
Expand All @@ -233,12 +244,14 @@ static void activate(enum sp1_mui_layer l)
prev[i] = pos[i];
CAT(l, i) = fabsf(VAL(l, i) - pos[i]) >= CATCH_MATCH;
shown[i] = pos[i];
lock_ref[i] = now01[i];
locked[i] = true;
}
}

static void defaults_pages(void)
{
stored[SP1_MUI_T_SHIFT][0] = 0.0f;
stored[SP1_MUI_T_SHIFT][0] = 0.5f; /* GTLT on its detent: disengaged */
stored[SP1_MUI_T_SHIFT][1] = 0.5f; /* gate length 128/256 (Marbles) */
stored[SP1_MUI_T_SHIFT][2] = 0.0f; /* no randomness */
stored[SP1_MUI_T_SHIFT][3] = DEF_LENGTH;
Expand Down Expand Up @@ -333,6 +346,7 @@ void sp1_mui_enter(const uint16_t raw[4], bool fnc)
{
for (int i = 0; i < 4; i++) {
pos[i] = to01(raw[i]);
now01[i] = pos[i];
}
fnc_was = fnc;
press_dirty = true; /* a "••" already down is not a tap */
Expand All @@ -350,7 +364,8 @@ uint32_t sp1_mui_tick(uint32_t elapsed_ms, const uint16_t raw[4], bool valid,

if (valid) {
for (int i = 0; i < 4; i++) {
pos[i] += (to01(raw[i]) - pos[i]) * FADER_LP;
now01[i] = to01(raw[i]);
pos[i] += (now01[i] - pos[i]) * FADER_LP;
}
}

Expand Down Expand Up @@ -402,6 +417,16 @@ uint32_t sp1_mui_tick(uint32_t elapsed_ms, const uint16_t raw[4], bool valid,
* rests, and catches up after the port goes (sp1_plaits_ui.c). */
prev[i] = pos[i];
CAT(active, i) = fabsf(*s - pos[i]) >= CATCH_MATCH;
} else if (locked[i]) {
/* Since the layer change: the filter's tail and a resting finger
* are not movement (CATCH_LOCK). Tracking or catching up was decided
* at the change and is NOT re-decided here: re-checking against a
* fader that has already moved would put a tracking fader into
* catch-up and leave its value short of the fader. */
prev[i] = pos[i];
if (fabsf(now01[i] - lock_ref[i]) > CATCH_LOCK) {
locked[i] = false;
}
} else if (!CAT(active, i)) {
*s = pos[i];
prev[i] = pos[i];
Expand Down Expand Up @@ -678,9 +703,19 @@ void sp1_mui_params(struct sp1_marbles_params *p)
p->y_range = range; /* one [J] (M4e) */
}

/* GTLT (issue #20): -1..+1 past the 10 % detent, exactly 0 inside it. The curve and
* the TRIG exemption are in sp1_gtlt.h / sp1_synth.cc. Its CC is centred, like t BIAS:
* in `sum` it offsets the post-detent value, so a CC alone can engage it. */
static float gtlt_tilt(void)
{
return 2.0f * (clamp01(detent(stored[SP1_MUI_T_SHIFT][0], DETENT_BIPOLAR) +
mo(SP1_MIDI_D_GTLT)) - 0.5f);
}

void sp1_mui_routing(struct sp1_mui_routing *out)
{
*out = route;
out->gtlt = gtlt_tilt();
}

float sp1_mui_bpm(void)
Expand All @@ -703,9 +738,17 @@ static void leds_of(enum sp1_mui_layer l, uint8_t out[4])
if (base && show_ms == 0u) {
/* Marbles' own output LEDs; T4 = Y on both pages. */
if (page == SP1_MUI_PAGE_T) {
/* A high gate shows its HEIGHT under GTLT (#20), on every output whatever
* it is routed to, from 8 % for a 0 % gate up to full (Adara: the floor is
* "signal strength 0", so a gate still firing TRIG never goes dark). Low
* gates are off. On the detent every gain is 1, so this is the old
* on/off display exactly. */
const uint8_t g = sp1_marbles_last_gates();
const float tilt = gtlt_tilt();
for (int i = 0; i < 3; i++) {
out[i] = (g & (1u << i)) ? 255u : 0u;
const float x = SP1_GTLT_LED_FLOOR +
(1.0f - SP1_GTLT_LED_FLOOR) * sp1_gtlt_gain(tilt, i);
out[i] = (g & (1u << i)) ? (uint8_t)(x * 255.0f + 0.5f) : 0u;
}
} else {
for (int i = 0; i < 3; i++) {
Expand Down
Loading
Loading