reverb experiment

sources/Application/Player/CMakeLists.txt

@@ -4,6 +4,7 @@ add_library(application_player
   PlayerMixer.cpp
   SyncMaster.cpp
   TablePlayback.cpp
+  Reverb.h
 )
 
 target_link_libraries(application_player PUBLIC

sources/Application/Player/Reverb.h

@@ -0,0 +1,235 @@
+
+#pragma once
+
+#include "Externals/braids/dsp.h"
+// add here so as not to include whole of "Externals/braids/resources.h"
+const int16_t wav_sine[] = {
+    -32512, -32502, -32473, -32423, -32356, -32265, -32160, -32031, -31885,
+    -31719, -31533, -31331, -31106, -30864, -30605, -30324, -30028, -29712,
+    -29379, -29026, -28658, -28272, -27868, -27449, -27011, -26558, -26089,
+    -25604, -25103, -24588, -24056, -23512, -22953, -22378, -21793, -21191,
+    -20579, -19954, -19316, -18667, -18006, -17334, -16654, -15960, -15259,
+    -14548, -13828, -13100, -12363, -11620, -10868, -10112, -9347,  -8578,
+    -7805,  -7023,  -6241,  -5453,  -4662,  -3868,  -3073,  -2274,  -1474,
+    -674,   126,    929,    1729,   2527,   3326,   4123,   4916,   5707,
+    6495,   7278,   8057,   8833,   9601,   10366,  11122,  11874,  12618,
+    13353,  14082,  14802,  15512,  16216,  16906,  17589,  18260,  18922,
+    19569,  20207,  20834,  21446,  22045,  22634,  23206,  23765,  24311,
+    24842,  25357,  25858,  26343,  26812,  27266,  27701,  28123,  28526,
+    28912,  29281,  29632,  29966,  30281,  30579,  30859,  31118,  31361,
+    31583,  31788,  31973,  32139,  32286,  32412,  32521,  32608,  32679,
+    32725,  32757,  32766,  32757,  32725,  32679,  32608,  32521,  32412,
+    32286,  32139,  31973,  31788,  31583,  31361,  31118,  30859,  30579,
+    30281,  29966,  29632,  29281,  28912,  28526,  28123,  27701,  27266,
+    26812,  26343,  25858,  25357,  24842,  24311,  23765,  23206,  22634,
+    22045,  21446,  20834,  20207,  19569,  18922,  18260,  17589,  16906,
+    16216,  15512,  14802,  14082,  13353,  12618,  11874,  11122,  10366,
+    9601,   8833,   8057,   7278,   6495,   5707,   4916,   4123,   3326,
+    2527,   1729,   929,    126,    -674,   -1474,  -2274,  -3073,  -3868,
+    -4662,  -5453,  -6241,  -7023,  -7805,  -8578,  -9347,  -10112, -10868,
+    -11620, -12363, -13100, -13828, -14548, -15259, -15960, -16654, -17334,
+    -18006, -18667, -19316, -19954, -20579, -21191, -21793, -22378, -22953,
+    -23512, -24056, -24588, -25103, -25604, -26089, -26558, -27011, -27449,
+    -27868, -28272, -28658, -29026, -29379, -29712, -30028, -30324, -30605,
+    -30864, -31106, -31331, -31533, -31719, -31885, -32031, -32160, -32265,
+    -32356, -32423, -32473, -32502, -32512};
+
+#include "q15.h"
+#include <cstring>
+
+using namespace braids;
+#define VERB_LENGTH 8203
+#define VERB_AP1 113
+#define VERB_AP2 162
+#define VERB_AP3 241
+#define VERB_AP4 373
+
+#define VERB_AP5 615
+#define VERB_AP6 773
+#define VERB_D1 915
+
+#define VERB_AP7 513
+#define VERB_AP8 849
+#define VERB_D2 1515
+
+typedef struct AllPassFilter {
+  int16_t *buf;
+  int count = 0;
+  int length = 0;
+  q15_t c;
+  uint32_t phase;
+  uint16_t phaseOffset;
+} AllPassFilter;
+
+class Reverb2 {
+public:
+  Reverb2() {
+    // --- EXPERIMENT WITH THESE VALUES ---
+
+    // 1. All-pass coefficient. Controls diffusion. Higher values =
+    // smoother/more diffuse. Range: ~0.6f to ~0.85f. Default was 0.7f.
+    const float all_pass_coeff = 0.7f;
+
+    // 2. Damping. Controls high-frequency decay. Lower values = darker reverb.
+    // Range: 0.0f (no highs) to 0.99f (bright). Default was 0.9f.
+    dampAmount = f32_to_q15(0.2f);
+
+    // 3. Feedback. Controls the decay time / size of the room.
+    // Range: 0.0f (no decay) to <1.0f. Default was 0.87f.
+    feedbackAmount = f32_to_q15(0.85f);
+
+    // --- END OF EXPERIMENTAL VALUES ---
+
+    for (int i = 0; i < 10; i++) {
+      allpass[i].count = 0;
+      allpass[i].c = f32_to_q15(all_pass_coeff);
+      allpass[i].phaseOffset = phaseOffset[i];
+      allpass[i].phase = 0;
+    }
+    int bufCount = 0;
+    allpass[0].buf = buf + bufCount;
+    bufCount += VERB_AP1;
+    allpass[0].length = VERB_AP1;
+    allpass[1].buf = buf + bufCount;
+    bufCount += VERB_AP2;
+    allpass[1].length = VERB_AP2;
+    allpass[2].buf = buf + bufCount;
+    bufCount += VERB_AP3;
+    allpass[2].length = VERB_AP3;
+    allpass[3].buf = buf + bufCount;
+    bufCount += VERB_AP4;
+    allpass[3].length = VERB_AP4;
+
+    allpass[4].buf = buf + bufCount;
+    bufCount += VERB_AP5;
+    allpass[4].length = VERB_AP5;
+    allpass[5].buf = buf + bufCount;
+    bufCount += VERB_AP6;
+    allpass[5].length = VERB_AP6;
+    allpass[6].buf = buf + bufCount;
+    bufCount += VERB_D1;
+    allpass[6].length = VERB_D1;
+
+    allpass[8].buf = buf + bufCount;
+    bufCount += VERB_AP7;
+    allpass[8].length = VERB_AP7;
+    allpass[9].buf = buf + bufCount;
+    bufCount += VERB_AP8;
+    allpass[9].length = VERB_AP8;
+    allpass[7].buf = buf + bufCount;
+    bufCount += VERB_D2;
+    allpass[7].length = VERB_D2;
+
+    position[0] = f32_to_q15(0.9f);
+    position[1] = f32_to_q15(-0.9f);
+    position[2] = f32_to_q15(1.f);
+    position[3] = f32_to_q15(-1.f);
+  }
+  int16_t last_l, last_r;
+  inline void process(int16_t in, int16_t &l, int16_t &r) {
+    if ((++count) % 2 == 0) {
+      l = last_l;
+      r = last_r;
+      return;
+    }
+    for (int i = 0; i < 4; i++) {
+      in = ProcessAllPass(in, &allpass[i]);
+    }
+    // low pass filter the feedback
+    feedback = add_q15(mult_q15(feedback, dampAmount),
+                       mult_q15(damp[0], sub_q15(Q15_MAX, dampAmount)));
+    damp[0] = feedback;
+
+    in = ProcessAllPass(add_q15(in, mult_q15(feedbackAmount, feedback)),
+                        &allpass[4]);
+    in = ProcessAllPass(in, &allpass[5]);
+    in = DelayProcess(in, &allpass[6]);
+    in = ProcessAllPass(in, &allpass[8]);
+    in = ProcessAllPass(in, &allpass[9]);
+    feedback = DelayProcess(in, &allpass[7]);
+
+    // 4. Delay Tap Positions. These are other good candidates for
+    // experimentation. I've changed them to different prime numbers to break up
+    // regularities.
+    int16_t dt1 = DelayTap(310, &allpass[6]);
+    int16_t dt2 = DelayTap(611, &allpass[6]);
+    int16_t dt3 = DelayTap(937, &allpass[7]);
+    int16_t dt4 = DelayTap(1201, &allpass[7]);
+
+    dt3 = mult_q15(dt2, f32_to_q15(0.8));
+    dt4 = mult_q15(dt2, f32_to_q15(0.8));
+
+    l = 0;
+    r = 0;
+
+    l = add_q15(l, Mix(0, dt1, position[0]));
+    r = add_q15(r, Mix(0, dt1, 0xffff - position[0]));
+
+    l = add_q15(l, Mix(0, dt2, position[1]));
+    r = add_q15(r, Mix(0, dt2, 0xffff - position[1]));
+
+    l = add_q15(l, Mix(0, dt3, position[2]));
+    r = add_q15(r, Mix(0, dt3, 0xffff - position[2]));
+
+    last_l = l = add_q15(l, Mix(0, dt4, position[3]));
+    last_r = r = add_q15(r, Mix(0, dt4, 0xffff - position[3]));
+  }
+
+private:
+  inline int16_t ProcessAllPass(int16_t in, AllPassFilter *ap) {
+    int16_t apdelayed = ap->buf[ap->count];
+    int16_t inSum = ap->buf[ap->count] =
+        add_q15(mult_q15(-ap->c, apdelayed), in);
+    ap->count = (ap->count + 1) % ap->length;
+    return add_q15(mult_q15(inSum, ap->c), apdelayed);
+  }
+  inline int16_t DelayWobbleLookup(AllPassFilter *d) {
+    int16_t lfo = Interpolate824(wav_sine, d->phase);
+    int32_t offset = ((int32_t)d->length * (int32_t)(lfo >> 1));
+    int16_t a = DelayTap(offset >> 16, d);
+    int16_t b = DelayTap((offset >> 16) + 1, d);
+    return Mix(a, b, lfo);
+  }
+  inline int16_t DelayProcessWobble(int16_t in, AllPassFilter *d) {
+    DelayProcess(in, d);
+    d->phase += d->phaseOffset;
+    return DelayWobbleLookup(d);
+  }
+  inline int16_t DelayProcess(int16_t in, AllPassFilter *d) {
+    int16_t delayed = d->buf[d->count];
+    d->buf[d->count] = in;
+    d->count = (d->count + 1) % d->length;
+    return delayed;
+  }
+  inline int16_t DelayTap(int16_t tap, AllPassFilter *d) {
+    int offset = tap + d->count;
+    if (offset > d->length - 1) {
+      offset -= d->length;
+    }
+    return d->buf[offset % d->length];
+  }
+
+  // int16_t TapAp(int tap, AllPassFilter *ap)
+  // {
+  //     int offset = tap+ap->count;
+  //     while (offset < 0)
+  //         offset += ap->length;
+  //     while(offset > ap->length-1)
+  //     {
+  //         offset -= ap->length;
+  //     }
+  //     return ap->buf[offset];
+  // }
+  q15_t damp[2] = {0};
+  q15_t dampAmount, feedbackAmount;
+  uint16_t phaseOffset[10] = {803, 447,  1243, 1089, 304,
+                              183, 4128, 481,  513,  713};
+
+  AllPassFilter allpass[10];
+  int16_t buf[VERB_LENGTH] = {0};
+  int16_t feedback = 0;
+  uint8_t count = 0;
+  int length = VERB_LENGTH;
+
+  uint16_t position[4] = {0x7fff, 0x7fff, 0x4000, 0xffff};
+};

sources/Application/Player/q15.h

@@ -0,0 +1,84 @@
+// adapted from https://github.com/ARM-software/CMSIS_5
+/*
+Copyright (C) 2010-2022 ARM Limited or its affiliates. All rights reserved.
+Christophe Favergeon
+*/
+
+#ifndef _FIXED_POINT_Q15_H_
+#define _FIXED_POINT_Q15_H_
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+#ifndef __STATIC_FORCEINLINE
+#define __STATIC_INLINE static inline
+#endif
+
+#include <stdint.h>
+
+/**
+ * @brief 8-bit fractional data type in 1.7 format.
+ */
+typedef int8_t q7_t;
+
+/**
+ * @brief 16-bit fractional data type in 1.15 format.
+ */
+typedef int16_t q15_t;
+
+/**
+ * @brief 32-bit fractional data type in 1.31 format.
+ */
+typedef int32_t q31_t;
+
+#if !defined(__SSAT)
+__STATIC_INLINE int32_t __SSAT(int32_t val, uint32_t sat) {
+  if ((sat >= 1U) && (sat <= 32U)) {
+    const int32_t max = (int32_t)((1U << (sat - 1U)) - 1U);
+    const int32_t min = -1 - max;
+    if (val > max) {
+      return max;
+    } else if (val < min) {
+      return min;
+    }
+  }
+  return val;
+}
+#endif
+
+#define Q31_MAX ((q31_t)(0x7FFFFFFFL))
+#define Q15_MAX ((q15_t)(0x7FFF))
+#define Q7_MAX ((q7_t)(0x7F))
+#define Q31_MIN ((q31_t)(0x80000000L))
+#define Q15_MIN ((q15_t)(0x8000))
+#define Q7_MIN ((q7_t)(0x80))
+
+#define Q31_ABSMAX ((q31_t)(0x7FFFFFFFL))
+#define Q15_ABSMAX ((q15_t)(0x7FFF))
+#define Q7_ABSMAX ((q7_t)(0x7F))
+#define Q31_ABSMIN ((q31_t)0)
+#define Q15_ABSMIN ((q15_t)0)
+#define Q7_ABSMIN ((q7_t)0)
+
+inline q15_t add_q15(q15_t srcA, q15_t srcB) {
+  return (q15_t)__SSAT(((q31_t)srcA + srcB), 16);
+}
+inline q15_t mult_q15(q15_t srcA, q15_t srcB) {
+  q31_t mul = (q31_t)((q15_t)(srcA) * (q15_t)(srcB));
+  return (q15_t)__SSAT(mul >> 15, 16);
+}
+inline q15_t sub_q15(q15_t srcA, q15_t srcB) {
+  return (q15_t)__SSAT(((q31_t)srcA - srcB), 16);
+}
+inline q15_t f32_to_q15(float in) { return (q15_t)(in * 32767.0f); }
+inline float q15_to_f32(q15_t in) { return ((float)in / 32767.0f); }
+inline q15_t lerp_q15(q15_t a, q15_t b, q15_t amt) {
+  return add_q15(a, mult_q15(sub_q15(a, b), amt));
+}
+
+#ifdef __cplusplus
+}
+#endif
+
+#endif // _FIXED_POINT_Q15_H_

sources/Externals/CMakeLists.txt

@@ -1,6 +1,7 @@
 add_subdirectory(TinyXML2)
 add_subdirectory(yxml)
 add_subdirectory(etl)
+# add_subdirectory(braids)
 target_include_directories(etl INTERFACE "${PROJECT_SOURCE_DIR}/config")
 add_subdirectory(cRSID)
-add_subdirectory(opal)
+add_subdirectory(opal)

sources/Externals/braids/dsp.h

@@ -31,7 +31,7 @@
 #define STMLIB_UTILS_DSP_H_
 
 // #include "stmlib/stmlib.h"
-#include "pico/stdlib.h"
+// #include "pico/stdlib.h"
 
 namespace braids {
 

sources/Externals/braids/resources.h

@@ -33,7 +33,8 @@
 #define BRAIDS_RESOURCES_H_
 
 // #include "stmlib/stmlib.h"
-#include "pico/stdlib.h"
+// #include "pico/stdlib.h"
+#include <cstdint>
 
 namespace braids {