some audio processing improvements and bugfixes from SR WLED

- smoothing FFTResult (don't have a matrix to test)
- UDP sound sync improvements
- some bugfixes from SR WLED
- button.cpp: avoid starvation: strip.isUpdating() can be true for a long time.

work in progress - still needs testing!!
This commit is contained in:
Frank 2022-08-14 13:58:07 +02:00
parent d05b49496c
commit 968721a515
4 changed files with 62 additions and 30 deletions

View File

@ -23,10 +23,13 @@
// Comment/Uncomment to toggle usb serial debugging // Comment/Uncomment to toggle usb serial debugging
// #define MIC_LOGGER // MIC sampling & sound input debugging (serial plotter) // #define MIC_LOGGER // MIC sampling & sound input debugging (serial plotter)
// #define FFT_SAMPLING_LOG // FFT result debugging // #define FFT_SAMPLING_LOG // FFT result debugging
// #define SR_DEBUG // generic SR DEBUG messages // #define SR_DEBUG // generic SR DEBUG messages (including MIC_LOGGER)
// #define NO_MIC_LOGGER // exclude MIC_LOGGER from SR_DEBUG
// hackers corner // hackers corner
//#define SOUND_DYNAMICS_LIMITER // experimental: define to enable a dynamics limiter that avoids "sudden flashes" at onsets. Makes some effects look more "smooth and fluent" #if !defined(SOUND_DYNAMICS_LIMITER) && !defined(NO_SOUND_DYNAMICS_LIMITER)
#define SOUND_DYNAMICS_LIMITER // experimental: define to enable a dynamics limiter that avoids "sudden flashes" at onsets. Makes some effects look more "smooth and fluent"
#endif
#ifdef SR_DEBUG #ifdef SR_DEBUG
#define DEBUGSR_PRINT(x) Serial.print(x) #define DEBUGSR_PRINT(x) Serial.print(x)
@ -60,6 +63,10 @@ static uint8_t sampleGain = 60; // sample gain (config value)
static uint8_t soundAgc = 0; // Automagic gain control: 0 - none, 1 - normal, 2 - vivid, 3 - lazy (config value) static uint8_t soundAgc = 0; // Automagic gain control: 0 - none, 1 - normal, 2 - vivid, 3 - lazy (config value)
static uint8_t audioSyncEnabled = 0; // bit field: bit 0 - send, bit 1 - receive (config value) static uint8_t audioSyncEnabled = 0; // bit field: bit 0 - send, bit 1 - receive (config value)
// user settable parameters for limitSoundDynamics()
static int attackTime = 80; // int: attack time in milliseconds. Default 0.1sec
static int decayTime = 1400; // int: decay time in milliseconds. Default 1.4sec
// //
// AGC presets // AGC presets
// Note: in C++, "const" implies "static" - no need to explicitly declare everything as "static const" // Note: in C++, "const" implies "static" - no need to explicitly declare everything as "static const"
@ -98,7 +105,7 @@ static float multAgc = 1.0f; // sample * multAgc = sampleAgc.
// FFT Variables // FFT Variables
constexpr uint16_t samplesFFT = 512; // Samples in an FFT batch - This value MUST ALWAYS be a power of 2 constexpr uint16_t samplesFFT = 512; // Samples in an FFT batch - This value MUST ALWAYS be a power of 2
constexpr uint16_t samplesFFT_2 = 256; // meaningfull part of FFT results - nly the "lower half" contains usefull information. constexpr uint16_t samplesFFT_2 = 256; // meaningfull part of FFT results - only the "lower half" contains useful information.
static float FFT_MajorPeak = 0.0f; static float FFT_MajorPeak = 0.0f;
static float FFT_Magnitude = 0.0f; static float FFT_Magnitude = 0.0f;
@ -274,9 +281,12 @@ void FFTcode(void * parameter)
// Manual linear adjustment of gain using sampleGain adjustment for different input types. // Manual linear adjustment of gain using sampleGain adjustment for different input types.
fftCalc[i] *= soundAgc ? multAgc : ((float)sampleGain/40.0f * (float)inputLevel/128.0f + 1.0f/16.0f); //with inputLevel adjustment fftCalc[i] *= soundAgc ? multAgc : ((float)sampleGain/40.0f * (float)inputLevel/128.0f + 1.0f/16.0f); //with inputLevel adjustment
// smooth results
//fftAvg[i] = fftCalc[i]*0.05f + 0.95f*fftAvg[i]; // will need approx 10 cycles (250ms) for converging against fftCalc[i]
fftAvg[i] = fftCalc[i] *0.1f + 0.9f*fftAvg[i]; // will need approx 5 cycles (125ms) for converging against fftCalc[i]
// Now, let's dump it all into fftResult. Need to do this, otherwise other routines might grab fftResult values prematurely. // Now, let's dump it all into fftResult. Need to do this, otherwise other routines might grab fftResult values prematurely.
fftResult[i] = constrain((int)fftCalc[i], 0, 254); //fftResult[i] = constrain((int)fftCalc[i], 0, 254);
fftAvg[i] = (float)fftResult[i]*0.05f + 0.95f*fftAvg[i]; fftResult[i] = constrain((int)fftAvg[i], 0, 254);
} }
#ifdef WLED_DEBUG #ifdef WLED_DEBUG
@ -602,10 +612,13 @@ class AudioReactive : public Usermod {
// this is the minimal code for reading analog mic input on 8266. // this is the minimal code for reading analog mic input on 8266.
// warning!! Absolutely experimental code. Audio on 8266 is still not working. Expects a million follow-on problems. // warning!! Absolutely experimental code. Audio on 8266 is still not working. Expects a million follow-on problems.
static unsigned long lastAnalogTime = 0; static unsigned long lastAnalogTime = 0;
static float lastAnalogValue = 0.0f;
if (millis() - lastAnalogTime > 20) { if (millis() - lastAnalogTime > 20) {
micDataReal = analogRead(A0); // read one sample with 10bit resolution. This is a dirty hack, supporting volumereactive effects only. micDataReal = analogRead(A0); // read one sample with 10bit resolution. This is a dirty hack, supporting volumereactive effects only.
lastAnalogTime = millis(); lastAnalogTime = millis();
} lastAnalogValue = micDataReal;
yield();
} else micDataReal = lastAnalogValue;
micIn = int(micDataReal); micIn = int(micDataReal);
#endif #endif
#endif #endif
@ -618,6 +631,7 @@ class AudioReactive : public Usermod {
float micInNoDC = fabs(micDataReal - micLev); float micInNoDC = fabs(micDataReal - micLev);
expAdjF = (weighting * micInNoDC + (1.0-weighting) * expAdjF); expAdjF = (weighting * micInNoDC + (1.0-weighting) * expAdjF);
expAdjF = (expAdjF <= soundSquelch) ? 0: expAdjF; // simple noise gate expAdjF = (expAdjF <= soundSquelch) ? 0: expAdjF; // simple noise gate
if ((soundSquelch == 0) && (expAdjF < 0.25f)) expAdjF = 0; // do something meaningfull when "squelch = 0"
expAdjF = fabsf(expAdjF); // Now (!) take the absolute value expAdjF = fabsf(expAdjF); // Now (!) take the absolute value
tmpSample = expAdjF; tmpSample = expAdjF;
@ -664,14 +678,9 @@ class AudioReactive : public Usermod {
/* Limits the dynamics of volumeSmth (= sampleAvg or sampleAgc). /* Limits the dynamics of volumeSmth (= sampleAvg or sampleAgc).
* It does not affect FFTResult[] or volumeRaw ( = sample or rawSampleAgc) * does not affect FFTResult[] or volumeRaw ( = sample or rawSampleAgc)
*/ */
// effects: Gravimeter, Gravcenter, Gravcentric, Noisefire, Plasmoid, Freqpixels, Freqwave, Gravfreq, (2D Swirl, 2D Waverly) // effects: Gravimeter, Gravcenter, Gravcentric, Noisefire, Plasmoid, Freqpixels, Freqwave, Gravfreq, (2D Swirl, 2D Waverly)
// experimental, as it still has side-effects on AGC - AGC detects "silence" to late (due to long decay time) and ditches up the gain multiplier.
// values below will be made user-configurable later
const float attackTime = 200; // attack time -> 0.2sec
const float decayTime = 2800; // decay time -> 2.8sec
void limitSampleDynamics(void) { void limitSampleDynamics(void) {
#ifdef SOUND_DYNAMICS_LIMITER #ifdef SOUND_DYNAMICS_LIMITER
const float bigChange = 196; // just a representative number - a large, expected sample value const float bigChange = 196; // just a representative number - a large, expected sample value
@ -681,8 +690,8 @@ class AudioReactive : public Usermod {
if ((attackTime > 0) && (decayTime > 0)) { // only change volume if user has defined attack>0 and decay>0 if ((attackTime > 0) && (decayTime > 0)) { // only change volume if user has defined attack>0 and decay>0
long delta_time = millis() - last_time; long delta_time = millis() - last_time;
delta_time = constrain(delta_time , 1, 1000); // below 1ms -> 1ms; above 1sec -> sily lil hick-up delta_time = constrain(delta_time , 1, 1000); // below 1ms -> 1ms; above 1sec -> sily lil hick-up
float maxAttack = bigChange * float(delta_time) / attackTime; float maxAttack = bigChange * float(delta_time) / float(attackTime);
float maxDecay = - bigChange * float(delta_time) / decayTime; float maxDecay = - bigChange * float(delta_time) / float(decayTime);
float deltaSample = volumeSmth - last_volumeSmth; float deltaSample = volumeSmth - last_volumeSmth;
if (deltaSample > maxAttack) deltaSample = maxAttack; if (deltaSample > maxAttack) deltaSample = maxAttack;
@ -704,8 +713,11 @@ class AudioReactive : public Usermod {
audioSyncPacket transmitData; audioSyncPacket transmitData;
strncpy_P(transmitData.header, PSTR(UDP_SYNC_HEADER), 6); strncpy_P(transmitData.header, PSTR(UDP_SYNC_HEADER), 6);
transmitData.sampleRaw = volumeRaw; //transmitData.sampleRaw = volumeRaw;
transmitData.sampleSmth = volumeSmth; //transmitData.sampleSmth = volumeSmth;
// transmit samples that were not modified by limitSampleDynamics()
transmitData.sampleRaw = (soundAgc) ? rawSampleAgc: sampleRaw;
transmitData.sampleSmth = (soundAgc) ? sampleAgc : sampleAvg;
transmitData.samplePeak = udpSamplePeak ? 1:0; transmitData.samplePeak = udpSamplePeak ? 1:0;
udpSamplePeak = false; // Reset udpSamplePeak after we've transmitted it udpSamplePeak = false; // Reset udpSamplePeak after we've transmitted it
transmitData.reserved1 = 0; transmitData.reserved1 = 0;
@ -744,9 +756,11 @@ class AudioReactive : public Usermod {
if (packetSize == sizeof(audioSyncPacket) && !(isValidUdpSyncVersion((const char *)fftBuff))) { if (packetSize == sizeof(audioSyncPacket) && !(isValidUdpSyncVersion((const char *)fftBuff))) {
audioSyncPacket *receivedPacket = reinterpret_cast<audioSyncPacket*>(fftBuff); audioSyncPacket *receivedPacket = reinterpret_cast<audioSyncPacket*>(fftBuff);
// update samples for effects
volumeSmth = receivedPacket->sampleSmth; volumeSmth = receivedPacket->sampleSmth;
volumeRaw = receivedPacket->sampleRaw; volumeRaw = receivedPacket->sampleRaw;
// update internal samples
sampleRaw = volumeRaw; sampleRaw = volumeRaw;
sampleAvg = volumeSmth; sampleAvg = volumeSmth;
rawSampleAgc = volumeRaw; rawSampleAgc = volumeRaw;
@ -945,6 +959,7 @@ class AudioReactive : public Usermod {
if (audioSyncEnabled & 0x02) disableSoundProcessing = true; // make sure everything is disabled IF in audio Receive mode if (audioSyncEnabled & 0x02) disableSoundProcessing = true; // make sure everything is disabled IF in audio Receive mode
if (audioSyncEnabled & 0x01) disableSoundProcessing = false; // keep running audio IF we're in audio Transmit mode if (audioSyncEnabled & 0x01) disableSoundProcessing = false; // keep running audio IF we're in audio Transmit mode
// Only run the sampling code IF we're not in Receive mode or realtime mode // Only run the sampling code IF we're not in Receive mode or realtime mode
if (!(audioSyncEnabled & 0x02) && !disableSoundProcessing) { if (!(audioSyncEnabled & 0x02) && !disableSoundProcessing) {
bool agcEffect = false; bool agcEffect = false;
@ -981,9 +996,8 @@ class AudioReactive : public Usermod {
limitSampleDynamics(); // optional - makes volumeSmth very smooth and fluent limitSampleDynamics(); // optional - makes volumeSmth very smooth and fluent
// update UI // update WebServer UI
uint8_t knownMode = strip.getFirstSelectedSeg().mode; // 1st selected segment is more appropriate than main segment uint8_t knownMode = strip.getFirstSelectedSeg().mode; // 1st selected segment is more appropriate than main segment
if (lastMode != knownMode) { // only execute if mode changes if (lastMode != knownMode) { // only execute if mode changes
char lineBuffer[4]; char lineBuffer[4];
extractModeName(knownMode, JSON_mode_names, lineBuffer, 3); // use of JSON_mode_names is deprecated, use nullptr extractModeName(knownMode, JSON_mode_names, lineBuffer, 3); // use of JSON_mode_names is deprecated, use nullptr
@ -1024,11 +1038,20 @@ class AudioReactive : public Usermod {
} }
} }
// Begin UDP Microphone Sync
if ((audioSyncEnabled & 0x02) && millis() - lastTime > delayMs) { // Only run the audio listener code if we're in Receive mode // UDP Microphone Sync - receive mode
(void) receiveAudioData(); // ToDo: use return value for something meaningfull if ((audioSyncEnabled & 0x02) && udpSyncConnected) {
// Only run the audio listener code if we're in Receive mode
static float syncVolumeSmth = 0;
bool have_new_sample = false;
if (millis() - lastTime > delayMs) {
have_new_sample = receiveAudioData();
lastTime = millis(); lastTime = millis();
} }
if (have_new_sample) syncVolumeSmth = volumeSmth; // remember received sample
else volumeSmth = syncVolumeSmth; // restore originally received sample for next run of dynamics limiter
limitSampleDynamics(); // run dynamics limiter on received volumeSmth, to hide jumps and hickups
}
#if defined(MIC_LOGGER) || defined(MIC_SAMPLING_LOG) || defined(FFT_SAMPLING_LOG) #if defined(MIC_LOGGER) || defined(MIC_SAMPLING_LOG) || defined(FFT_SAMPLING_LOG)
EVERY_N_MILLIS(20) { EVERY_N_MILLIS(20) {
@ -1036,12 +1059,13 @@ class AudioReactive : public Usermod {
} }
#endif #endif
if ((audioSyncEnabled & 0x01) && millis() - lastTime > 20) { // Only run the transmit code IF we're in Transmit mode //UDP Microphone Sync - transmit mode
if ((audioSyncEnabled & 0x01) && millis() - lastTime > 20) {
// Only run the transmit code IF we're in Transmit mode
transmitAudioData(); transmitAudioData();
lastTime = millis(); lastTime = millis();
} }
//limitSampleDynamics(); // If done as the last step, it will also affect audio received by UDP sound sync. Problem: effects might see inconsistent intermediate values and start flickering :-(
} }

View File

@ -339,9 +339,12 @@ class I2SAdcSource : public I2SSource {
#else #else
.communication_format = i2s_comm_format_t(I2S_COMM_FORMAT_I2S | I2S_COMM_FORMAT_I2S_MSB), .communication_format = i2s_comm_format_t(I2S_COMM_FORMAT_I2S | I2S_COMM_FORMAT_I2S_MSB),
#endif #endif
.intr_alloc_flags = ESP_INTR_FLAG_LEVEL2, .intr_alloc_flags = ESP_INTR_FLAG_LEVEL1,
.dma_buf_count = 8, .dma_buf_count = 8,
.dma_buf_len = _blockSize .dma_buf_len = _blockSize,
.use_apll = false,
.tx_desc_auto_clear = false,
.fixed_mclk = 0
}; };
} }
@ -350,6 +353,7 @@ class I2SAdcSource : public I2SSource {
void initialize(int8_t audioPin, int8_t = I2S_PIN_NO_CHANGE, int8_t = I2S_PIN_NO_CHANGE, int8_t = I2S_PIN_NO_CHANGE, int8_t = I2S_PIN_NO_CHANGE, int8_t = I2S_PIN_NO_CHANGE) { void initialize(int8_t audioPin, int8_t = I2S_PIN_NO_CHANGE, int8_t = I2S_PIN_NO_CHANGE, int8_t = I2S_PIN_NO_CHANGE, int8_t = I2S_PIN_NO_CHANGE, int8_t = I2S_PIN_NO_CHANGE) {
if(!pinManager.allocatePin(audioPin, false, PinOwner::UM_Audioreactive)) { if(!pinManager.allocatePin(audioPin, false, PinOwner::UM_Audioreactive)) {
DEBUGSR_PRINTF("failed to allocate GPIO for audio analog input: %d\n", audioPin);
return; return;
} }
_audioPin = audioPin; _audioPin = audioPin;
@ -376,7 +380,7 @@ class I2SAdcSource : public I2SSource {
DEBUGSR_PRINTF("Failed to set i2s adc mode: %d\n", err); DEBUGSR_PRINTF("Failed to set i2s adc mode: %d\n", err);
return; return;
} }
// adc1_config_channel_atten(adc1_channel_t(channel), ADC_ATTEN_DB_11)); //see https://github.com/espressif/arduino-esp32/blob/master/libraries/ESP32/examples/I2S/HiFreq_ADC/HiFreq_ADC.ino
#if defined(ARDUINO_ARCH_ESP32) #if defined(ARDUINO_ARCH_ESP32)
// according to docs from espressif, the ADC needs to be started explicitly // according to docs from espressif, the ADC needs to be started explicitly
// fingers crossed // fingers crossed
@ -408,7 +412,7 @@ class I2SAdcSource : public I2SSource {
#if !defined(ARDUINO_ARCH_ESP32) #if !defined(ARDUINO_ARCH_ESP32)
// old code - works for me without enable/disable, at least on ESP32. // old code - works for me without enable/disable, at least on ESP32.
err = i2s_adc_disable(I2S_NUM_0); err = i2s_adc_disable(I2S_NUM_0); //i2s_adc_disable() may cause crash with IDF 4.4 (https://github.com/espressif/arduino-esp32/issues/6832)
//err = i2s_stop(I2S_NUM_0); //err = i2s_stop(I2S_NUM_0);
if (err != ESP_OK) { if (err != ESP_OK) {
DEBUGSR_PRINTF("Failed to disable i2s adc: %d\n", err); DEBUGSR_PRINTF("Failed to disable i2s adc: %d\n", err);

View File

@ -6711,7 +6711,8 @@ uint16_t mode_freqmap(void) { // Map FFT_MajorPeak to SEGLEN.
SEGMENT.fade_out(SEGMENT.speed); SEGMENT.fade_out(SEGMENT.speed);
uint16_t locn = (log10f((float)FFT_MajorPeak) - 1.78f) * (float)SEGLEN/(3.71f-1.78f); // log10 frequency range is from 1.78 to 3.71. Let's scale to SEGLEN. int locn = (log10f((float)FFT_MajorPeak) - 1.78f) * (float)SEGLEN/(3.71f-1.78f); // log10 frequency range is from 1.78 to 3.71. Let's scale to SEGLEN.
if (locn < 1) locn = 0; // avoid underflow
if (locn >=SEGLEN) locn = SEGLEN-1; if (locn >=SEGLEN) locn = SEGLEN-1;
uint16_t pixCol = (log10f(FFT_MajorPeak) - 1.78f) * 255.0f/(3.71f-1.78f); // Scale log10 of frequency values to the 255 colour index. uint16_t pixCol = (log10f(FFT_MajorPeak) - 1.78f) * 255.0f/(3.71f-1.78f); // Scale log10 of frequency values to the 255 colour index.

View File

@ -216,10 +216,13 @@ void handleAnalog(uint8_t b)
void handleButton() void handleButton()
{ {
static unsigned long lastRead = 0UL; static unsigned long lastRead = 0UL;
static unsigned long lastRun = 0UL;
bool analog = false; bool analog = false;
unsigned long now = millis(); unsigned long now = millis();
if (strip.isUpdating()) return; // don't interfere with strip updates. Our button will still be there in 1ms (next cycle) //if (strip.isUpdating()) return; // don't interfere with strip updates. Our button will still be there in 1ms (next cycle)
if (strip.isUpdating() && (millis() - lastRun < 400)) return; // be niced, but avoid button starvation
lastRun = millis();
for (uint8_t b=0; b<WLED_MAX_BUTTONS; b++) { for (uint8_t b=0; b<WLED_MAX_BUTTONS; b++) {
#ifdef ESP8266 #ifdef ESP8266