circuitpython/shared-module/synthio/__init__.c

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

550 lines
21 KiB
C
Raw Normal View History

/*
* This file is part of the MicroPython project, http://micropython.org/
*
* The MIT License (MIT)
*
* Copyright (c) 2021 Artyom Skrobov
* Copyright (c) 2023 Jeff Epler for Adafruit Industries
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in
* all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
* THE SOFTWARE.
*/
#include "shared-module/synthio/__init__.h"
#include "shared-bindings/synthio/__init__.h"
#include "shared-module/synthio/Note.h"
#include "py/runtime.h"
#include <math.h>
#include <stdlib.h>
2023-05-15 11:49:42 -04:00
mp_float_t synthio_global_rate_scale;
uint8_t synthio_global_tick;
STATIC const int16_t square_wave[] = {-32768, 32767};
STATIC const uint16_t notes[] = {8372, 8870, 9397, 9956, 10548, 11175, 11840,
12544, 13290, 14080, 14917, 15804}; // 9th octave
STATIC int64_t round_float_to_int64(mp_float_t f) {
return (int64_t)(f + MICROPY_FLOAT_CONST(0.5));
}
mp_float_t common_hal_synthio_midi_to_hz_float(mp_float_t arg) {
return common_hal_synthio_onevo_to_hz_float(arg / 12.);
}
mp_float_t common_hal_synthio_onevo_to_hz_float(mp_float_t octave) {
return notes[0] * MICROPY_FLOAT_C_FUN(pow)(2., octave - 10);
}
STATIC int16_t convert_time_to_rate(uint32_t sample_rate, mp_obj_t time_in, int16_t difference) {
mp_float_t time = mp_obj_get_float(time_in);
int num_samples = (int)MICROPY_FLOAT_C_FUN(round)(time * sample_rate);
if (num_samples == 0) {
return 32767;
}
int16_t result = MIN(32767, MAX(1, abs(difference * SYNTHIO_MAX_DUR) / num_samples));
return (difference < 0) ? -result : result;
}
void synthio_envelope_definition_set(synthio_envelope_definition_t *envelope, mp_obj_t obj, uint32_t sample_rate) {
if (obj == mp_const_none) {
envelope->attack_level = 32767;
envelope->sustain_level = 32767;
envelope->attack_step = 32767;
envelope->decay_step = -32767;
envelope->release_step = -32767;
return;
}
mp_arg_validate_type(obj, (mp_obj_type_t *)&synthio_envelope_type_obj, MP_QSTR_envelope);
size_t len;
mp_obj_t *fields;
mp_obj_tuple_get(obj, &len, &fields);
envelope->attack_level = (int)(32767 * mp_obj_get_float(fields[3]));
envelope->sustain_level = (int)(32767 * mp_obj_get_float(fields[4]) * mp_obj_get_float(fields[3]));
envelope->attack_step = convert_time_to_rate(
sample_rate, fields[0], envelope->attack_level);
envelope->decay_step = -convert_time_to_rate(
sample_rate, fields[1], envelope->attack_level - envelope->sustain_level);
envelope->release_step = -convert_time_to_rate(
sample_rate, fields[2],
envelope->sustain_level
? envelope->sustain_level
: envelope->attack_level);
}
STATIC void synthio_envelope_state_step(synthio_envelope_state_t *state, synthio_envelope_definition_t *def, size_t n_steps) {
state->substep += n_steps;
while (state->substep >= SYNTHIO_MAX_DUR) {
// max n_steps should be SYNTHIO_MAX_DUR so this loop executes at most
// once
state->substep -= SYNTHIO_MAX_DUR;
switch (state->state) {
case SYNTHIO_ENVELOPE_STATE_SUSTAIN:
break;
case SYNTHIO_ENVELOPE_STATE_ATTACK:
state->level = MIN(state->level + def->attack_step, def->attack_level);
if (state->level == def->attack_level) {
state->state = SYNTHIO_ENVELOPE_STATE_DECAY;
}
break;
case SYNTHIO_ENVELOPE_STATE_DECAY:
state->level = MAX(state->level + def->decay_step, def->sustain_level);
if (state->level == def->sustain_level) {
state->state = SYNTHIO_ENVELOPE_STATE_SUSTAIN;
}
break;
case SYNTHIO_ENVELOPE_STATE_RELEASE:
state->level = MAX(state->level + def->release_step, 0);
}
}
}
STATIC void synthio_envelope_state_init(synthio_envelope_state_t *state, synthio_envelope_definition_t *def) {
state->level = 0;
state->substep = 0;
state->state = SYNTHIO_ENVELOPE_STATE_ATTACK;
synthio_envelope_state_step(state, def, SYNTHIO_MAX_DUR);
}
STATIC void synthio_envelope_state_release(synthio_envelope_state_t *state, synthio_envelope_definition_t *def) {
state->state = SYNTHIO_ENVELOPE_STATE_RELEASE;
}
STATIC synthio_envelope_definition_t *synthio_synth_get_note_envelope(synthio_synth_t *synth, mp_obj_t note_obj) {
synthio_envelope_definition_t *def = &synth->global_envelope_definition;
if (!mp_obj_is_small_int(note_obj)) {
synthio_note_obj_t *note = MP_OBJ_TO_PTR(note_obj);
if (note->envelope_obj != mp_const_none) {
def = &note->envelope_def;
}
}
return def;
}
#define RANGE_LOW (-28000)
#define RANGE_HIGH (28000)
#define RANGE_SHIFT (16)
#define RANGE_SCALE (0xfffffff / (32768 * CIRCUITPY_SYNTHIO_MAX_CHANNELS - RANGE_HIGH))
// dynamic range compression via a downward compressor with hard knee
//
// When the output value is within the range +-28000 (about 85% of full scale),
// it is unchanged. Otherwise, it undergoes a gain reduction so that the
// largest possible values, (+32768,-32767) * CIRCUITPY_SYNTHIO_MAX_CHANNELS,
// still fit within the output range
//
// This produces a much louder overall volume with multiple voices, without
// much additional processing.
//
// https://en.wikipedia.org/wiki/Dynamic_range_compression
STATIC
int16_t mix_down_sample(int32_t sample) {
if (sample < RANGE_LOW) {
sample = (((sample - RANGE_LOW) * RANGE_SCALE) >> RANGE_SHIFT) + RANGE_LOW;
} else if (sample > RANGE_HIGH) {
sample = (((sample - RANGE_HIGH) * RANGE_SCALE) >> RANGE_SHIFT) + RANGE_HIGH;
}
return sample;
}
static void synth_note_into_buffer(synthio_synth_t *synth, int chan, int32_t *out_buffer32, int16_t dur) {
mp_obj_t note_obj = synth->span.note_obj[chan];
if (note_obj == SYNTHIO_SILENCE) {
synth->accum[chan] = 0;
return;
}
if (synth->envelope_state[chan].level == 0) {
// note is truly finished, but we only just noticed
synth->span.note_obj[chan] = SYNTHIO_SILENCE;
return;
}
int32_t sample_rate = synth->sample_rate;
// adjust loudness by envelope
uint16_t loudness[2] = {synth->envelope_state[chan].level,synth->envelope_state[chan].level};
uint32_t dds_rate;
const int16_t *waveform = synth->waveform_bufinfo.buf;
uint32_t waveform_length = synth->waveform_bufinfo.len / sizeof(int16_t);
uint32_t ring_dds_rate = 0;
const int16_t *ring_waveform = NULL;
uint32_t ring_waveform_length = 0;
if (mp_obj_is_small_int(note_obj)) {
uint8_t note = mp_obj_get_int(note_obj);
uint8_t octave = note / 12;
uint16_t base_freq = notes[note % 12];
// rate = base_freq * waveform_length
// den = sample_rate * 2 ^ (10 - octave)
// den = sample_rate * 2 ^ 10 / 2^octave
// dds_rate = 2^SHIFT * rate / den
// dds_rate = 2^(SHIFT-10+octave) * base_freq * waveform_length / sample_rate
dds_rate = (sample_rate / 2 + ((uint64_t)(base_freq * waveform_length) << (SYNTHIO_FREQUENCY_SHIFT - 10 + octave))) / sample_rate;
} else {
synthio_note_obj_t *note = MP_OBJ_TO_PTR(note_obj);
int32_t frequency_scaled = synthio_note_step(note, sample_rate, dur, loudness);
if (note->waveform_buf.buf) {
waveform = note->waveform_buf.buf;
waveform_length = note->waveform_buf.len / sizeof(int16_t);
}
dds_rate = synthio_frequency_convert_scaled_to_dds((uint64_t)frequency_scaled * waveform_length, sample_rate);
if (note->ring_frequency_scaled != 0 && note->ring_waveform_buf.buf) {
ring_waveform = note->ring_waveform_buf.buf;
ring_waveform_length = note->ring_waveform_buf.len / sizeof(int16_t);
2023-05-15 11:49:42 -04:00
ring_dds_rate = synthio_frequency_convert_scaled_to_dds((uint64_t)note->ring_frequency_bent * ring_waveform_length, sample_rate);
uint32_t lim = ring_waveform_length << SYNTHIO_FREQUENCY_SHIFT;
if (ring_dds_rate > lim / sizeof(int16_t)) {
ring_dds_rate = 0; // can't ring at that frequency
2023-05-10 13:07:01 -04:00
}
}
}
int synth_chan = synth->channel_count;
if (ring_dds_rate) {
uint32_t lim = waveform_length << SYNTHIO_FREQUENCY_SHIFT;
uint32_t accum = synth->accum[chan];
if (dds_rate > lim / 2) {
// beyond nyquist, can't play note
return;
}
// can happen if note waveform gets set mid-note, but the expensive modulo is usually avoided
if (accum > lim) {
accum %= lim;
}
int32_t ring_buffer[dur];
// first, fill with waveform
for (uint16_t i = 0; i < dur; i++) {
accum += dds_rate;
// because dds_rate is low enough, the subtraction is guaranteed to go back into range, no expensive modulo needed
2023-05-10 13:07:01 -04:00
if (accum > lim) {
accum -= lim;
2023-05-10 13:07:01 -04:00
}
int16_t idx = accum >> SYNTHIO_FREQUENCY_SHIFT;
ring_buffer[i] = waveform[idx];
}
synth->accum[chan] = accum;
2023-05-10 13:07:01 -04:00
// now modulate by ring and accumulate
accum = synth->ring_accum[chan];
lim = ring_waveform_length << SYNTHIO_FREQUENCY_SHIFT;
2023-05-10 13:07:01 -04:00
// can happen if note waveform gets set mid-note, but the expensive modulo is usually avoided
if (accum > lim) {
accum %= lim;
}
2023-05-10 13:07:01 -04:00
for (uint16_t i = 0, j = 0; i < dur; i++) {
accum += ring_dds_rate;
// because dds_rate is low enough, the subtraction is guaranteed to go back into range, no expensive modulo needed
if (accum > lim) {
accum -= lim;
2023-05-10 13:07:01 -04:00
}
int16_t idx = accum >> SYNTHIO_FREQUENCY_SHIFT;
int16_t wi = (ring_waveform[idx] * ring_buffer[i]) / 32768;
for (int c = 0; c < synth_chan; c++) {
out_buffer32[j] += (wi * loudness[c]) / 32768;
j++;
2023-05-10 13:07:01 -04:00
}
}
synth->ring_accum[chan] = accum;
} else {
uint32_t lim = waveform_length << SYNTHIO_FREQUENCY_SHIFT;
uint32_t accum = synth->accum[chan];
if (dds_rate > lim / 2) {
// beyond nyquist, can't play note
return;
}
// can happen if note waveform gets set mid-note, but the expensive modulo is usually avoided
if (accum > lim) {
accum %= lim;
}
2023-05-10 13:07:01 -04:00
for (uint16_t i = 0, j = 0; i < dur; i++) {
accum += dds_rate;
// because dds_rate is low enough, the subtraction is guaranteed to go back into range, no expensive modulo needed
2023-05-10 13:07:01 -04:00
if (accum > lim) {
accum -= lim;
2023-05-10 13:07:01 -04:00
}
int16_t idx = accum >> SYNTHIO_FREQUENCY_SHIFT;
int16_t wi = waveform[idx];
for (int c = 0; c < synth_chan; c++) {
out_buffer32[j] += (wi * loudness[c]) / 65536;
j++;
}
}
synth->accum[chan] = accum;
}
}
STATIC void run_fir(synthio_synth_t *synth, int32_t *out_buffer32, uint16_t dur) {
int16_t *coeff = (int16_t *)synth->filter_bufinfo.buf;
size_t fir_len = synth->filter_bufinfo.len / sizeof(int16_t);
int32_t *in_buf = synth->filter_buffer;
int synth_chan = synth->channel_count;
// FIR and copy values to output buffer
for (int16_t i = 0; i < dur * synth_chan; i++) {
int32_t acc = 0;
for (size_t j = 0; j < fir_len; j++) {
// shift 5 here is good for up to 32 filtered voices, else might wrap
acc = acc + (in_buf[j * synth_chan] * (coeff[j] >> 5));
}
*out_buffer32++ = acc >> 10;
in_buf++;
}
// Move values down so that they get filtered next time
memmove(synth->filter_buffer, &synth->filter_buffer[dur * synth_chan], fir_len * sizeof(int32_t) * synth_chan);
}
STATIC bool synthio_synth_get_note_filtered(mp_obj_t note_obj) {
if (note_obj == mp_const_none) {
return false;
}
if (!mp_obj_is_small_int(note_obj)) {
synthio_note_obj_t *note = MP_OBJ_TO_PTR(note_obj);
return note->filter;
}
return true;
}
void synthio_synth_synthesize(synthio_synth_t *synth, uint8_t **bufptr, uint32_t *buffer_length, uint8_t channel) {
if (channel == synth->other_channel) {
*buffer_length = synth->last_buffer_length;
*bufptr = (uint8_t *)(synth->buffers[synth->other_buffer_index] + channel);
return;
}
2023-05-15 11:49:42 -04:00
shared_bindings_synthio_lfo_tick(synth->sample_rate);
synth->buffer_index = !synth->buffer_index;
synth->other_channel = 1 - channel;
synth->other_buffer_index = synth->buffer_index;
uint16_t dur = MIN(SYNTHIO_MAX_DUR, synth->span.dur);
synth->span.dur -= dur;
int32_t out_buffer32[dur * synth->channel_count];
if (synth->filter_buffer) {
int32_t *filter_start = &synth->filter_buffer[synth->filter_bufinfo.len * synth->channel_count / sizeof(int16_t)];
memset(filter_start, 0, dur * synth->channel_count * sizeof(int32_t));
for (int chan = 0; chan < CIRCUITPY_SYNTHIO_MAX_CHANNELS; chan++) {
mp_obj_t note_obj = synth->span.note_obj[chan];
if (!synthio_synth_get_note_filtered(note_obj)) {
continue;
}
synth_note_into_buffer(synth, chan, filter_start, dur);
}
run_fir(synth, out_buffer32, dur);
} else {
memset(out_buffer32, 0, sizeof(out_buffer32));
}
for (int chan = 0; chan < CIRCUITPY_SYNTHIO_MAX_CHANNELS; chan++) {
mp_obj_t note_obj = synth->span.note_obj[chan];
if (synth->filter_buffer && synthio_synth_get_note_filtered(note_obj)) {
continue;
}
synth_note_into_buffer(synth, chan, out_buffer32, dur);
}
int16_t *out_buffer16 = (int16_t *)(void *)synth->buffers[synth->buffer_index];
// mix down audio
for (size_t i = 0; i < MP_ARRAY_SIZE(out_buffer32); i++) {
int32_t sample = out_buffer32[i];
out_buffer16[i] = mix_down_sample(sample);
}
// advance envelope states
for (int chan = 0; chan < CIRCUITPY_SYNTHIO_MAX_CHANNELS; chan++) {
mp_obj_t note_obj = synth->span.note_obj[chan];
if (note_obj == SYNTHIO_SILENCE) {
continue;
}
synthio_envelope_state_step(&synth->envelope_state[chan], synthio_synth_get_note_envelope(synth, note_obj), dur);
}
*buffer_length = synth->last_buffer_length = dur * SYNTHIO_BYTES_PER_SAMPLE * synth->channel_count;
*bufptr = (uint8_t *)out_buffer16;
}
void synthio_synth_reset_buffer(synthio_synth_t *synth, bool single_channel_output, uint8_t channel) {
if (single_channel_output && channel == 1) {
return;
}
synth->other_channel = -1;
}
bool synthio_synth_deinited(synthio_synth_t *synth) {
return synth->buffers[0] == NULL;
}
void synthio_synth_deinit(synthio_synth_t *synth) {
synth->filter_buffer = NULL;
synth->buffers[0] = NULL;
synth->buffers[1] = NULL;
}
void synthio_synth_envelope_set(synthio_synth_t *synth, mp_obj_t envelope_obj) {
synthio_envelope_definition_set(&synth->global_envelope_definition, envelope_obj, synth->sample_rate);
synth->envelope_obj = envelope_obj;
}
mp_obj_t synthio_synth_envelope_get(synthio_synth_t *synth) {
return synth->envelope_obj;
}
void synthio_synth_init(synthio_synth_t *synth, uint32_t sample_rate, int channel_count, mp_obj_t waveform_obj, mp_obj_t filter_obj, mp_obj_t envelope_obj) {
synthio_synth_parse_waveform(&synth->waveform_bufinfo, waveform_obj);
synthio_synth_parse_filter(&synth->filter_bufinfo, filter_obj);
mp_arg_validate_int_range(channel_count, 1, 2, MP_QSTR_channel_count);
synth->buffer_length = SYNTHIO_MAX_DUR * SYNTHIO_BYTES_PER_SAMPLE * channel_count;
synth->buffers[0] = m_malloc(synth->buffer_length, false);
synth->buffers[1] = m_malloc(synth->buffer_length, false);
if (synth->filter_bufinfo.len) {
synth->filter_buffer_length = (synth->filter_bufinfo.len / 2 + SYNTHIO_MAX_DUR) * channel_count * sizeof(int32_t);
synth->filter_buffer = m_malloc(synth->filter_buffer_length, false);
}
synth->channel_count = channel_count;
synth->other_channel = -1;
synth->waveform_obj = waveform_obj;
synth->sample_rate = sample_rate;
synthio_synth_envelope_set(synth, envelope_obj);
for (size_t i = 0; i < CIRCUITPY_SYNTHIO_MAX_CHANNELS; i++) {
synth->span.note_obj[i] = SYNTHIO_SILENCE;
}
}
void synthio_synth_get_buffer_structure(synthio_synth_t *synth, bool single_channel_output,
bool *single_buffer, bool *samples_signed, uint32_t *max_buffer_length, uint8_t *spacing) {
*single_buffer = false;
*samples_signed = true;
*max_buffer_length = synth->buffer_length;
if (single_channel_output) {
*spacing = synth->channel_count;
} else {
*spacing = 1;
}
}
STATIC void parse_common(mp_buffer_info_t *bufinfo, mp_obj_t o, int16_t what, mp_int_t max_len) {
if (o != mp_const_none) {
mp_get_buffer_raise(o, bufinfo, MP_BUFFER_READ);
if (bufinfo->typecode != 'h') {
mp_raise_ValueError_varg(translate("%q must be array of type 'h'"), what);
}
mp_arg_validate_length_range(bufinfo->len / sizeof(int16_t), 2, max_len, what);
}
}
void synthio_synth_parse_waveform(mp_buffer_info_t *bufinfo_waveform, mp_obj_t waveform_obj) {
*bufinfo_waveform = ((mp_buffer_info_t) { .buf = (void *)square_wave, .len = 4 });
parse_common(bufinfo_waveform, waveform_obj, MP_QSTR_waveform, 16384);
}
void synthio_synth_parse_filter(mp_buffer_info_t *bufinfo_filter, mp_obj_t filter_obj) {
*bufinfo_filter = ((mp_buffer_info_t) { .buf = NULL, .len = 0 });
parse_common(bufinfo_filter, filter_obj, MP_QSTR_filter, 128);
}
STATIC int find_channel_with_note(synthio_synth_t *synth, mp_obj_t note) {
for (int i = 0; i < CIRCUITPY_SYNTHIO_MAX_CHANNELS; i++) {
if (synth->span.note_obj[i] == note) {
return i;
}
}
int result = -1;
if (note == SYNTHIO_SILENCE) {
// replace the releasing note with lowest volume level
int level = 32768;
for (int chan = 0; chan < CIRCUITPY_SYNTHIO_MAX_CHANNELS; chan++) {
if (!SYNTHIO_NOTE_IS_PLAYING(synth, chan)) {
synthio_envelope_state_t *state = &synth->envelope_state[chan];
if (state->level < level) {
result = chan;
level = state->level;
}
}
}
}
return result;
}
bool synthio_span_change_note(synthio_synth_t *synth, mp_obj_t old_note, mp_obj_t new_note) {
int channel;
if (new_note != SYNTHIO_SILENCE && (channel = find_channel_with_note(synth, new_note)) != -1) {
// note already playing, re-enter attack phase
synth->envelope_state[channel].state = SYNTHIO_ENVELOPE_STATE_ATTACK;
return true;
}
channel = find_channel_with_note(synth, old_note);
if (channel != -1) {
if (new_note == SYNTHIO_SILENCE) {
synthio_envelope_state_release(&synth->envelope_state[channel], synthio_synth_get_note_envelope(synth, old_note));
} else {
synth->span.note_obj[channel] = new_note;
synthio_envelope_state_init(&synth->envelope_state[channel], synthio_synth_get_note_envelope(synth, new_note));
synth->accum[channel] = 0;
}
return true;
}
return false;
}
uint64_t synthio_frequency_convert_float_to_scaled(mp_float_t val) {
return round_float_to_int64(val * (1 << SYNTHIO_FREQUENCY_SHIFT));
}
uint32_t synthio_frequency_convert_float_to_dds(mp_float_t frequency_hz, int32_t sample_rate) {
return synthio_frequency_convert_scaled_to_dds(synthio_frequency_convert_float_to_scaled(frequency_hz), sample_rate);
}
uint32_t synthio_frequency_convert_scaled_to_dds(uint64_t frequency_scaled, int32_t sample_rate) {
return (sample_rate / 2 + frequency_scaled) / sample_rate;
}
2023-05-15 11:49:42 -04:00
void shared_bindings_synthio_lfo_tick(uint32_t sample_rate) {
synthio_global_rate_scale = (mp_float_t)SYNTHIO_MAX_DUR / sample_rate;
synthio_global_tick++;
}