path: root/platforms/chibios/drivers/audio_dac_additive.c

/* Copyright 2016-2019 Jack Humbert
 * Copyright 2020 JohSchneider
 *
 * This program is free software: you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation, either version 2 of the License, or
 * (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
 * GNU General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program.  If not, see <http://www.gnu.org/licenses/>.
 */

#include "audio.h"
#include "gpio.h"
#include <math.h>
#include "util.h"

// Need to disable GCC's "tautological-compare" warning for this file, as it causes issues when running `KEEP_INTERMEDIATES=yes`. Corresponding pop at the end of the file.
#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wtautological-compare"

/*
  Audio Driver: DAC

  which utilizes the dac unit many STM32 are equipped with, to output a modulated waveform from samples stored in the dac_buffer_* array who are passed to the hardware through DMA

  it is also possible to have a custom sample-LUT by implementing/overriding 'dac_value_generate'

  this driver allows for multiple simultaneous tones to be played through one single channel by doing additive wave-synthesis
*/

#if !defined(AUDIO_PIN)
#    error "Audio feature enabled, but no suitable pin selected as AUDIO_PIN - see docs/feature_audio under 'ARM (DAC additive)' for available options."
#endif
#if defined(AUDIO_PIN_ALT) && !defined(AUDIO_PIN_ALT_AS_NEGATIVE)
#    pragma message "Audio feature: AUDIO_PIN_ALT set, but not AUDIO_PIN_ALT_AS_NEGATIVE - pin will be left unused; audio might still work though."
#endif

#if !defined(AUDIO_PIN_ALT)
// no ALT pin defined is valid, but the c-ifs below need some value set
#    define AUDIO_PIN_ALT PAL_NOLINE
#endif

#if !defined(AUDIO_DAC_SAMPLE_WAVEFORM_SINE) && !defined(AUDIO_DAC_SAMPLE_WAVEFORM_TRIANGLE) && !defined(AUDIO_DAC_SAMPLE_WAVEFORM_SQUARE) && !defined(AUDIO_DAC_SAMPLE_WAVEFORM_TRAPEZOID)
#    define AUDIO_DAC_SAMPLE_WAVEFORM_SINE
#endif

#ifdef AUDIO_DAC_SAMPLE_WAVEFORM_SINE
/* one full sine wave over [0,2*pi], but shifted up one amplitude and left pi/4; for the samples to start at 0
 */
static const dacsample_t dac_buffer_sine[AUDIO_DAC_BUFFER_SIZE] = {
    // 256 values, max 4095
    0x0,   0x1,   0x2,   0x6,   0xa,   0xf,   0x16,  0x1e,  0x27,  0x32,  0x3d,  0x4a,  0x58,  0x67,  0x78,  0x89,  0x9c,  0xb0,  0xc5,  0xdb,  0xf2,  0x10a, 0x123, 0x13e, 0x159, 0x175, 0x193, 0x1b1, 0x1d1, 0x1f1, 0x212, 0x235, 0x258, 0x27c, 0x2a0, 0x2c6, 0x2ed, 0x314, 0x33c, 0x365, 0x38e, 0x3b8, 0x3e3, 0x40e, 0x43a, 0x467, 0x494, 0x4c2, 0x4f0, 0x51f, 0x54e, 0x57d, 0x5ad, 0x5dd, 0x60e, 0x63f, 0x670, 0x6a1, 0x6d3, 0x705, 0x737, 0x769, 0x79b, 0x7cd, 0x800, 0x832, 0x864, 0x896, 0x8c8, 0x8fa, 0x92c, 0x95e, 0x98f, 0x9c0, 0x9f1, 0xa22, 0xa52, 0xa82, 0xab1, 0xae0, 0xb0f, 0xb3d, 0xb6b, 0xb98, 0xbc5, 0xbf1, 0xc1c, 0xc47, 0xc71, 0xc9a, 0xcc3, 0xceb, 0xd12, 0xd39, 0xd5f, 0xd83, 0xda7, 0xdca, 0xded, 0xe0e, 0xe2e, 0xe4e, 0xe6c, 0xe8a, 0xea6, 0xec1, 0xedc, 0xef5, 0xf0d, 0xf24, 0xf3a, 0xf4f, 0xf63, 0xf76, 0xf87, 0xf98, 0xfa7, 0xfb5, 0xfc2, 0xfcd, 0xfd8, 0xfe1, 0xfe9, 0xff0, 0xff5, 0xff9, 0xffd, 0xffe,
    0xfff, 0xffe, 0xffd, 0xff9, 0xff5, 0xff0, 0xfe9, 0xfe1, 0xfd8, 0xfcd, 0xfc2, 0xfb5, 0xfa7, 0xf98, 0xf87, 0xf76, 0xf63, 0xf4f, 0xf3a, 0xf24, 0xf0d, 0xef5, 0xedc, 0xec1, 0xea6, 0xe8a, 0xe6c, 0xe4e, 0xe2e, 0xe0e, 0xded, 0xdca, 0xda7, 0xd83, 0xd5f, 0xd39, 0xd12, 0xceb, 0xcc3, 0xc9a, 0xc71, 0xc47, 0xc1c, 0xbf1, 0xbc5, 0xb98, 0xb6b, 0xb3d, 0xb0f, 0xae0, 0xab1, 0xa82, 0xa52, 0xa22, 0x9f1, 0x9c0, 0x98f, 0x95e, 0x92c, 0x8fa, 0x8c8, 0x896, 0x864, 0x832, 0x800, 0x7cd, 0x79b, 0x769, 0x737, 0x705, 0x6d3, 0x6a1, 0x670, 0x63f, 0x60e, 0x5dd, 0x5ad, 0x57d, 0x54e, 0x51f, 0x4f0, 0x4c2, 0x494, 0x467, 0x43a, 0x40e, 0x3e3, 0x3b8, 0x38e, 0x365, 0x33c, 0x314, 0x2ed, 0x2c6, 0x2a0, 0x27c, 0x258, 0x235, 0x212, 0x1f1, 0x1d1, 0x1b1, 0x193, 0x175, 0x159, 0x13e, 0x123, 0x10a, 0xf2,  0xdb,  0xc5,  0xb0,  0x9c,  0x89,  0x78,  0x67,  0x58,  0x4a,  0x3d,  0x32,  0x27,  0x1e,  0x16,  0xf,   0xa,   0x6,   0x2,   0x1};
#endif // AUDIO_DAC_SAMPLE_WAVEFORM_SINE
#ifdef AUDIO_DAC_SAMPLE_WAVEFORM_TRIANGLE
static const dacsample_t dac_buffer_triangle[AUDIO_DAC_BUFFER_SIZE] = {
    // 256 values, max 4095
    0x0,   0x20,  0x40,  0x60,  0x80,  0xa0,  0xc0,  0xe0,  0x100, 0x120, 0x140, 0x160, 0x180, 0x1a0, 0x1c0, 0x1e0, 0x200, 0x220, 0x240, 0x260, 0x280, 0x2a0, 0x2c0, 0x2e0, 0x300, 0x320, 0x340, 0x360, 0x380, 0x3a0, 0x3c0, 0x3e0, 0x400, 0x420, 0x440, 0x460, 0x480, 0x4a0, 0x4c0, 0x4e0, 0x500, 0x520, 0x540, 0x560, 0x580, 0x5a0, 0x5c0, 0x5e0, 0x600, 0x620, 0x640, 0x660, 0x680, 0x6a0, 0x6c0, 0x6e0, 0x700, 0x720, 0x740, 0x760, 0x780, 0x7a0, 0x7c0, 0x7e0, 0x800, 0x81f, 0x83f, 0x85f, 0x87f, 0x89f, 0x8bf, 0x8df, 0x8ff, 0x91f, 0x93f, 0x95f, 0x97f, 0x99f, 0x9bf, 0x9df, 0x9ff, 0xa1f, 0xa3f, 0xa5f, 0xa7f, 0xa9f, 0xabf, 0xadf, 0xaff, 0xb1f, 0xb3f, 0xb5f, 0xb7f, 0xb9f, 0xbbf, 0xbdf, 0xbff, 0xc1f, 0xc3f, 0xc5f, 0xc7f, 0xc9f, 0xcbf, 0xcdf, 0xcff, 0xd1f, 0xd3f, 0xd5f, 0xd7f, 0xd9f, 0xdbf, 0xddf, 0xdff, 0xe1f, 0xe3f, 0xe5f, 0xe7f, 0xe9f, 0xebf, 0xedf, 0xeff, 0xf1f, 0xf3f, 0xf5f, 0xf7f, 0xf9f, 0xfbf, 0xfdf,
    0xfff, 0xfdf, 0xfbf, 0xf9f, 0xf7f, 0xf5f, 0xf3f, 0xf1f, 0xeff, 0xedf, 0xebf, 0xe9f, 0xe7f, 0xe5f, 0xe3f, 0xe1f, 0xdff, 0xddf, 0xdbf, 0xd9f, 0xd7f, 0xd5f, 0xd3f, 0xd1f, 0xcff, 0xcdf, 0xcbf, 0xc9f, 0xc7f, 0xc5f, 0xc3f, 0xc1f, 0xbff, 0xbdf, 0xbbf, 0xb9f, 0xb7f, 0xb5f, 0xb3f, 0xb1f, 0xaff, 0xadf, 0xabf, 0xa9f, 0xa7f, 0xa5f, 0xa3f, 0xa1f, 0x9ff, 0x9df, 0x9bf, 0x99f, 0x97f, 0x95f, 0x93f, 0x91f, 0x8ff, 0x8df, 0x8bf, 0x89f, 0x87f, 0x85f, 0x83f, 0x81f, 0x800, 0x7e0, 0x7c0, 0x7a0, 0x780, 0x760, 0x740, 0x720, 0x700, 0x6e0, 0x6c0, 0x6a0, 0x680, 0x660, 0x640, 0x620, 0x600, 0x5e0, 0x5c0, 0x5a0, 0x580, 0x560, 0x540, 0x520, 0x500, 0x4e0, 0x4c0, 0x4a0, 0x480, 0x460, 0x440, 0x420, 0x400, 0x3e0, 0x3c0, 0x3a0, 0x380, 0x360, 0x340, 0x320, 0x300, 0x2e0, 0x2c0, 0x2a0, 0x280, 0x260, 0x240, 0x220, 0x200, 0x1e0, 0x1c0, 0x1a0, 0x180, 0x160, 0x140, 0x120, 0x100, 0xe0,  0xc0,  0xa0,  0x80,  0x60,  0x40,  0x20};
#endif // AUDIO_DAC_SAMPLE_WAVEFORM_TRIANGLE
#ifdef AUDIO_DAC_SAMPLE_WAVEFORM_SQUARE
static const dacsample_t dac_buffer_square[AUDIO_DAC_BUFFER_SIZE] = {
    [0 ... AUDIO_DAC_BUFFER_SIZE / 2 - 1]                     = AUDIO_DAC_OFF_VALUE,  // first and
    [AUDIO_DAC_BUFFER_SIZE / 2 ... AUDIO_DAC_BUFFER_SIZE - 1] = AUDIO_DAC_SAMPLE_MAX, // second half
};
#endif // AUDIO_DAC_SAMPLE_WAVEFORM_SQUARE
/*
// four steps: 0, 1/3, 2/3 and 1
static const dacsample_t dac_buffer_staircase[AUDIO_DAC_BUFFER_SIZE] = {
    [0 ... AUDIO_DAC_BUFFER_SIZE/3 -1 ]                               = 0,
    [AUDIO_DAC_BUFFER_SIZE / 4 ... AUDIO_DAC_BUFFER_SIZE / 2 -1 ]     = AUDIO_DAC_SAMPLE_MAX / 3,
    [AUDIO_DAC_BUFFER_SIZE / 2 ... 3 * AUDIO_DAC_BUFFER_SIZE / 4 -1 ] = 2 * AUDIO_DAC_SAMPLE_MAX / 3,
    [3 * AUDIO_DAC_BUFFER_SIZE / 4 ... AUDIO_DAC_BUFFER_SIZE -1 ]     = AUDIO_DAC_SAMPLE_MAX,
}
*/
#ifdef AUDIO_DAC_SAMPLE_WAVEFORM_TRAPEZOID
static const dacsample_t dac_buffer_trapezoid[AUDIO_DAC_BUFFER_SIZE] = {0x0,   0x1f,  0x7f,  0xdf,  0x13f, 0x19f, 0x1ff, 0x25f, 0x2bf, 0x31f, 0x37f, 0x3df, 0x43f, 0x49f, 0x4ff, 0x55f, 0x5bf, 0x61f, 0x67f, 0x6df, 0x73f, 0x79f, 0x7ff, 0x85f, 0x8bf, 0x91f, 0x97f, 0x9df, 0xa3f, 0xa9f, 0xaff, 0xb5f, 0xbbf, 0xc1f, 0xc7f, 0xcdf, 0xd3f, 0xd9f, 0xdff, 0xe5f, 0xebf, 0xf1f, 0xf7f, 0xfdf, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff, 0xfff,
                                                                        0xfff, 0xfdf, 0xf7f, 0xf1f, 0xebf, 0xe5f, 0xdff, 0xd9f, 0xd3f, 0xcdf, 0xc7f, 0xc1f, 0xbbf, 0xb5f, 0xaff, 0xa9f, 0xa3f, 0x9df, 0x97f, 0x91f, 0x8bf, 0x85f, 0x7ff, 0x79f, 0x73f, 0x6df, 0x67f, 0x61f, 0x5bf, 0x55f, 0x4ff, 0x49f, 0x43f, 0x3df, 0x37f, 0x31f, 0x2bf, 0x25f, 0x1ff, 0x19f, 0x13f, 0xdf,  0x7f,  0x1f,  0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0,   0x0};
#endif // AUDIO_DAC_SAMPLE_WAVEFORM_TRAPEZOID

static dacsample_t dac_buffer_empty[AUDIO_DAC_BUFFER_SIZE] = {AUDIO_DAC_OFF_VALUE};

/* keep track of the sample position for for each frequency */
static float dac_if[AUDIO_MAX_SIMULTANEOUS_TONES] = {0.0};

static float   active_tones_snapshot[AUDIO_MAX_SIMULTANEOUS_TONES] = {0};
static uint8_t active_tones_snapshot_length                        = 0;

typedef enum {
    OUTPUT_SHOULD_START,
    OUTPUT_RUN_NORMALLY,
    // path 1: wait for zero, then change/update active tones
    OUTPUT_TONES_CHANGED,
    OUTPUT_REACHED_ZERO_BEFORE_TONE_CHANGE,
    // path 2: hardware should stop, wait for zero then turn output off = stop the timer
    OUTPUT_SHOULD_STOP,
    OUTPUT_REACHED_ZERO_BEFORE_OFF,
    OUTPUT_OFF,
    OUTPUT_OFF_1,
    OUTPUT_OFF_2, // trailing off: giving the DAC two more conversion cycles until the AUDIO_DAC_OFF_VALUE reaches the output, then turn the timer off, which leaves the output at that level
    number_of_output_states
} output_states_t;
output_states_t state = OUTPUT_OFF_2;

/**
 * Generation of the waveform being passed to the callback. Declared weak so users
 * can override it with their own wave-forms/noises.
 */
__attribute__((weak)) uint16_t dac_value_generate(void) {
    // DAC is running/asking for values but snapshot length is zero -> must be playing a pause
    if (active_tones_snapshot_length == 0) {
        return AUDIO_DAC_OFF_VALUE;
    }

    /* doing additive wave synthesis over all currently playing tones = adding up
     * sine-wave-samples for each frequency, scaled by the number of active tones
     */
    uint16_t value     = 0;
    float    frequency = 0.0f;

    for (uint8_t i = 0; i < active_tones_snapshot_length; i++) {
        /* Note: a user implementation does not have to rely on the active_tones_snapshot, but
         * could directly query the active frequencies through audio_get_processed_frequency */
        frequency = active_tones_snapshot[i];

        dac_if[i] = dac_if[i] + ((frequency * AUDIO_DAC_BUFFER_SIZE) / AUDIO_DAC_SAMPLE_RATE) * 2 / 3;
        /*Note: the 2/3 are necessary to get the correct frequencies on the
         *      DAC output (as measured with an oscilloscope), since the gpt
         *      timer runs with 3*AUDIO_DAC_SAMPLE_RATE; and the DAC callback
         *      is called twice per conversion.*/

        dac_if[i] = fmod(dac_if[i], AUDIO_DAC_BUFFER_SIZE);

        // Wavetable generation/lookup
        uint16_t dac_i = (uint16_t)dac_if[i];

#if defined(AUDIO_DAC_SAMPLE_WAVEFORM_SINE)
        value += dac_buffer_sine[dac_i] / active_tones_snapshot_length;
#elif defined(AUDIO_DAC_SAMPLE_WAVEFORM_TRIANGLE)
        value += dac_buffer_triangle[dac_i] / active_tones_snapshot_length;
#elif defined(AUDIO_DAC_SAMPLE_WAVEFORM_TRAPEZOID)
        value += dac_buffer_trapezoid[dac_i] / active_tones_snapshot_length;
#elif defined(AUDIO_DAC_SAMPLE_WAVEFORM_SQUARE)
        value += dac_buffer_square[dac_i] / active_tones_snapshot_length;
#endif
        /*
        // SINE
        value += dac_buffer_sine[dac_i] / active_tones_snapshot_length / 3;
        // TRIANGLE
        value += dac_buffer_triangle[dac_i] / active_tones_snapshot_length / 3;
        // SQUARE
        value += dac_buffer_square[dac_i] / active_tones_snapshot_length / 3;
        //NOTE: combination of these three wave-forms is more exemplary - and doesn't sound particularly good :-P
        */

        // STAIRS (mostly usefully as test-pattern)
        // value_avg = dac_buffer_staircase[dac_i] / active_tones_snapshot_length;
    }

    return value;
}

/**
 * DAC streaming callback. Does all of the main computing for playing songs.
 *
 * Note: chibios calls this CB twice: during the 'half buffer event', and the 'full buffer event'.
 */
static void dac_end(DACDriver *dacp) {
    dacsample_t *sample_p = (dacp)->samples;

    // work on the other half of the buffer
    if (dacIsBufferComplete(dacp)) {
        sample_p += AUDIO_DAC_BUFFER_SIZE / 2; // 'half_index'
    }

    for (uint8_t s = 0; s < AUDIO_DAC_BUFFER_SIZE / 2; s++) {
        if (OUTPUT_OFF <= state) {
            sample_p[s] = AUDIO_DAC_OFF_VALUE;
            continue;
        } else {
            sample_p[s] = dac_value_generate();
        }

        /* zero crossing (or approach, whereas zero == DAC_OFF_VALUE, which can be configured to anything from 0 to DAC_SAMPLE_MAX)
         * ============================*=*========================== AUDIO_DAC_SAMPLE_MAX
         *                          *       *
         *                        *           *
         * ---------------------------------------------------------
         *                     *                 *                  } AUDIO_DAC_SAMPLE_MAX/100
         * --------------------------------------------------------- AUDIO_DAC_OFF_VALUE
         *                  *                       *               } AUDIO_DAC_SAMPLE_MAX/100
         * ---------------------------------------------------------
         *               *
         * *           *
         *   *       *
         * =====*=*================================================= 0x0
         */
        if (((sample_p[s] + (AUDIO_DAC_SAMPLE_MAX / 100)) > AUDIO_DAC_OFF_VALUE) && // value approaches from below
            (sample_p[s] < (AUDIO_DAC_OFF_VALUE + (AUDIO_DAC_SAMPLE_MAX / 100)))    // or above
        ) {
            if ((OUTPUT_SHOULD_START == state) && (active_tones_snapshot_length > 0)) {
                state = OUTPUT_RUN_NORMALLY;
            } else if (OUTPUT_TONES_CHANGED == state) {
                state = OUTPUT_REACHED_ZERO_BEFORE_TONE_CHANGE;
            } else if (OUTPUT_SHOULD_STOP == state) {
                state = OUTPUT_REACHED_ZERO_BEFORE_OFF;
            }
        }

        // still 'ramping up', reset the output to OFF_VALUE until the generated values reach that value, to do a smooth handover
        if (OUTPUT_SHOULD_START == state) {
            sample_p[s] = AUDIO_DAC_OFF_VALUE;
        }

        if ((OUTPUT_SHOULD_START == state) || (OUTPUT_REACHED_ZERO_BEFORE_OFF == state) || (OUTPUT_REACHED_ZERO_BEFORE_TONE_CHANGE == state)) {
            uint8_t active_tones         = MIN(AUDIO_MAX_SIMULTANEOUS_TONES, audio_get_number_of_active_tones());
            active_tones_snapshot_length = 0;
            // update the snapshot - once, and only on occasion that something changed;
            // -> saves cpu cycles (?)
            for (uint8_t i = 0; i < active_tones; i++) {
                float freq = audio_get_processed_frequency(i);
                if (freq > 0) { // disregard 'rest' notes, with valid frequency 0.0f; which would only lower the resulting waveform volume during the additive synthesis step
                    active_tones_snapshot[active_tones_snapshot_length++] = freq;
                }
            }

            if ((0 == active_tones_snapshot_length) && (OUTPUT_REACHED_ZERO_BEFORE_OFF == state)) {
                state = OUTPUT_OFF;
            }
            if (OUTPUT_REACHED_ZERO_BEFORE_TONE_CHANGE == state) {
                state = OUTPUT_RUN_NORMALLY;
            }
        }
    }

    // update audio internal state (note position, current_note, ...)
    if (audio_update_state()) {
        if (OUTPUT_SHOULD_STOP != state) {
            state = OUTPUT_TONES_CHANGED;
        }
    }

    if (OUTPUT_OFF <= state) {
        if (OUTPUT_OFF_2 == state) {
            // stopping timer6 = stopping the DAC at whatever value it is currently pushing to the output = AUDIO_DAC_OFF_VALUE
            gptStopTimer(&GPTD6);
        } else {
            state++;
        }
    }
}

static void dac_error(DACDriver *dacp, dacerror_t err) {
    (void)dacp;
    (void)err;

    chSysHalt("DAC failure. halp");
}

static const GPTConfig gpt6cfg1 = {.frequency = AUDIO_DAC_SAMPLE_RATE * 3,
                                   .callback  = NULL,
                                   .cr2       = TIM_CR2_MMS_1, /* MMS = 010 = TRGO on Update Event.  */
                                   .dier      = 0U};

static const DACConfig dac_conf = {.init = AUDIO_DAC_OFF_VALUE, .datamode = DAC_DHRM_12BIT_RIGHT};

/**
 * @note The DAC_TRG(0) here selects the Timer 6 TRGO event, which is triggered
 * on the rising edge after 3 APB1 clock cycles, causing our gpt6cfg1.frequency
 * to be a third of what we expect.
 *
 * Here are all the values for DAC_TRG (TSEL in the ref manual)
 * TIM15_TRGO 0b011
 * TIM2_TRGO  0b100
 * TIM3_TRGO  0b001
 * TIM6_TRGO  0b000
 * TIM7_TRGO  0b010
 * EXTI9      0b110
 * SWTRIG     0b111
 */
static const DACConversionGroup dac_conv_cfg = {.num_channels = 1U, .end_cb = dac_end, .error_cb = dac_error, .trigger = DAC_TRG(0b000)};

void audio_driver_initialize(void) {
    if ((AUDIO_PIN == A4) || (AUDIO_PIN_ALT == A4)) {
        palSetLineMode(A4, PAL_MODE_INPUT_ANALOG);
        dacStart(&DACD1, &dac_conf);
    }
    if ((AUDIO_PIN == A5) || (AUDIO_PIN_ALT == A5)) {
        palSetLineMode(A5, PAL_MODE_INPUT_ANALOG);
        dacStart(&DACD2, &dac_conf);
    }

    /* enable the output buffer, to directly drive external loads with no additional circuitry
     *
     * see: AN4566 Application note: Extending the DAC performance of STM32 microcontrollers
     * Note: Buffer-Off bit -> has to be set 0 to enable the output buffer
     * Note: enabling the output buffer imparts an additional dc-offset of a couple mV
     *
     * this is done here, reaching directly into the stm32 registers since chibios has not implemented BOFF handling yet
     * (see: chibios/os/hal/ports/STM32/todo.txt '- BOFF handling in DACv1.'
     */
    DACD1.params->dac->CR &= ~DAC_CR_BOFF1;
    DACD2.params->dac->CR &= ~DAC_CR_BOFF2;

    if (AUDIO_PIN == A4) {
        dacStartConversion(&DACD1, &dac_conv_cfg, dac_buffer_empty, AUDIO_DAC_BUFFER_SIZE);
    } else if (AUDIO_PIN == A5) {
        dacStartConversion(&DACD2, &dac_conv_cfg, dac_buffer_empty, AUDIO_DAC_BUFFER_SIZE);
    }

    // no inverted/out-of-phase waveform (yet?), only pulling AUDIO_PIN_ALT to AUDIO_DAC_OFF_VALUE
#if defined(AUDIO_PIN_ALT_AS_NEGATIVE)
    if (AUDIO_PIN_ALT == A4) {
        dacPutChannelX(&DACD1, 0, AUDIO_DAC_OFF_VALUE);
    } else if (AUDIO_PIN_ALT == A5) {
        dacPutChannelX(&DACD2, 0, AUDIO_DAC_OFF_VALUE);
    }
#endif

    gptStart(&GPTD6, &gpt6cfg1);
}

void audio_driver_stop(void) {
    state = OUTPUT_SHOULD_STOP;
}

void audio_driver_start(void) {
    gptStartContinuous(&GPTD6, 2U);

    for (uint8_t i = 0; i < AUDIO_MAX_SIMULTANEOUS_TONES; i++) {
        dac_if[i]                = 0.0f;
        active_tones_snapshot[i] = 0.0f;
    }
    active_tones_snapshot_length = 0;
    state                        = OUTPUT_SHOULD_START;
}

#pragma GCC diagnostic pop