add bandpass filter for PDM mics

This introduces an optional band pass filtering step to remove low frequency noise. Very effective for PDM mics - reduces noise floor by a factor of 100 !
This commit is contained in:
Frank
2022-11-17 15:09:54 +01:00
parent eb2bf421b0
commit d3af3446f7
2 changed files with 68 additions and 8 deletions

View File

@@ -91,6 +91,7 @@ const float agcSampleSmooth[AGC_NUM_PRESETS] = { 1/12.f, 1/6.f, 1/16.f}; //
static AudioSource *audioSource = nullptr;
static volatile bool disableSoundProcessing = false; // if true, sound processing (FFT, filters, AGC) will be suspended. "volatile" as its shared between tasks.
static bool useBandPassFilter = false; // if true, enables a bandpass filter 80Hz-16Khz to remove noise. Applies before FFT.
// audioreactive variables shared with FFT task
static float micDataReal = 0.0f; // MicIn data with full 24bit resolution - lowest 8bit after decimal point
@@ -225,6 +226,7 @@ static float mapf(float x, float in_min, float in_max, float out_min, float out_
return (x - in_min) * (out_max - out_min) / (in_max - in_min) + out_min;
}
#if 1 // linear average
static float fftAddAvg(int from, int to) {
float result = 0.0f;
for (int i = from; i <= to; i++) {
@@ -233,6 +235,16 @@ static float fftAddAvg(int from, int to) {
return result / float(to - from + 1);
}
#else // RMS average
static float fftAddAvg(int from, int to) {
double result = 0.0;
for (int i = from; i <= to; i++) {
result += vReal[i] * vReal[i];
}
return sqrtf(result / float(to - from + 1));
}
#endif
// FFT main task
void FFTcode(void * parameter)
{
@@ -270,6 +282,37 @@ void FFTcode(void * parameter)
xLastWakeTime = xTaskGetTickCount(); // update "last unblocked time" for vTaskDelay
// band pass filter - can reduce noise floor by a factor of 50
// downside: frequencies below 100Hz will be ignored
if (useBandPassFilter) {
// low frequency cutoff parameter
//constexpr float alpha = 0.04f; // 100hz
constexpr float alpha = 0.03f; // 80hz
//constexpr float alpha = 0.0225f; // 60hz
// high frequency cutoff parameter
//constexpr float beta1 = 0.75; // 11Khz
//constexpr float beta1 = 0.82; // 15Khz
//constexpr float beta1 = 0.8285; // 18Khz
constexpr float beta1 = 0.85; // 20Khz
constexpr float beta2 = (1.0f - beta1) / 2.0;
static float last_vals[2] = { 0.0f }; // FIR high freq cutoff filter
static float lowfilt = 0.0f; // IIR low frequency cutoff filter
for (int i=0; i < samplesFFT; i++) {
// FIR lowpass, to remove high frequency noise
float highFilteredSample;
if (i < (samplesFFT-1)) highFilteredSample = beta1*vReal[i] + beta2*last_vals[0] + beta2*vReal[i+1]; // smooth out spikes
else highFilteredSample = beta1*vReal[i] + beta2*last_vals[0] + beta2*last_vals[1]; // spcial handling for last sample in array
last_vals[1] = last_vals[0];
last_vals[0] = vReal[i];
vReal[i] = highFilteredSample;
// IIR highpass, to remove low frequency noise
lowfilt += alpha * (vReal[i] - lowfilt);
vReal[i] = vReal[i] - lowfilt;
}
}
// find highest sample in the batch
float maxSample = 0.0f; // max sample from FFT batch
for (int i=0; i < samplesFFT; i++) {
@@ -365,10 +408,22 @@ void FFTcode(void * parameter)
#else
/* new mapping, optimized for 22050 Hz by softhack007 */
// bins frequency range
fftCalc[ 0] = fftAddAvg(1,2); // 1 43 - 86 sub-bass
fftCalc[ 1] = fftAddAvg(2,3); // 1 86 - 129 bass
fftCalc[ 2] = fftAddAvg(3,5); // 2 129 - 216 bass
fftCalc[ 3] = fftAddAvg(5,7); // 2 216 - 301 bass + midrange
if (useBandPassFilter) {
// skip frequencies below 100hz
fftCalc[ 0] = 0.8f * fftAddAvg(3,4);
fftCalc[ 1] = 0.9f * fftAddAvg(4,5);
fftCalc[ 2] = fftAddAvg(5,6);
fftCalc[ 3] = fftAddAvg(6,7);
// don't use the last bins from 206 to 255.
fftCalc[15] = fftAddAvg(165,205) * 0.75f; // 40 7106 - 8828 high -- with some damping
} else {
fftCalc[ 0] = fftAddAvg(1,2); // 1 43 - 86 sub-bass
fftCalc[ 1] = fftAddAvg(2,3); // 1 86 - 129 bass
fftCalc[ 2] = fftAddAvg(3,5); // 2 129 - 216 bass
fftCalc[ 3] = fftAddAvg(5,7); // 2 216 - 301 bass + midrange
// don't use the last bins from 216 to 255. They are usually contaminated by aliasing (aka noise)
fftCalc[15] = fftAddAvg(165,215) * 0.70f; // 50 7106 - 9259 high -- with some damping
}
fftCalc[ 4] = fftAddAvg(7,10); // 3 301 - 430 midrange
fftCalc[ 5] = fftAddAvg(10,13); // 3 430 - 560 midrange
fftCalc[ 6] = fftAddAvg(13,19); // 5 560 - 818 midrange
@@ -380,8 +435,6 @@ void FFTcode(void * parameter)
fftCalc[12] = fftAddAvg(70,86); // 16 3015 - 3704 high mid
fftCalc[13] = fftAddAvg(86,104); // 18 3704 - 4479 high mid
fftCalc[14] = fftAddAvg(104,165) * 0.88f; // 61 4479 - 7106 high mid + high -- with slight damping
fftCalc[15] = fftAddAvg(165,215) * 0.70f; // 50 7106 - 9259 high -- with some damping
// don't use the last bins from 216 to 255. They are usually contaminated by aliasing (aka noise)
#endif
} else { // noise gate closed - just decay old values
for (int i=0; i < NUM_GEQ_CHANNELS; i++) {
@@ -854,7 +907,11 @@ class AudioReactive : public Usermod {
#endif
//micLev = ((micLev * 8191.0f) + micDataReal) / 8192.0f; // takes a few seconds to "catch up" with the Mic Input
//if (useBandPassFilter)
// micLev += (micDataReal-micLev) / 8192.0f; // we expect some more fluctuations with mics that need pre-filtering
//else
micLev += (micDataReal-micLev) / 12288.0f;
if(micIn < micLev) micLev = ((micLev * 31.0f) + micDataReal) / 32.0f; // align MicLev to lowest input signal
micIn -= micLev; // Let's center it to 0 now
@@ -1119,6 +1176,8 @@ class AudioReactive : public Usermod {
periph_module_reset(PERIPH_I2S0_MODULE); // not possible on -C3
#endif
delay(100); // Give that poor microphone some time to setup.
useBandPassFilter = false;
switch (dmType) {
#if defined(CONFIG_IDF_TARGET_ESP32S2) || defined(CONFIG_IDF_TARGET_ESP32C3) || defined(CONFIG_IDF_TARGET_ESP32S3)
// stub cases for not-yet-supported I2S modes on other ESP32 chips
@@ -1156,6 +1215,7 @@ class AudioReactive : public Usermod {
case 5:
DEBUGSR_PRINT(F("AR: I2S PDM Microphone - ")); DEBUGSR_PRINTLN(F(I2S_PDM_MIC_CHANNEL_TEXT));
audioSource = new I2SSource(SAMPLE_RATE, BLOCK_SIZE, true, 1.0f/4.0f);
useBandPassFilter = true; // this reduces the noise floor on SPM1423 from 5% Vpp (~380) down to 0.05% Vpp (~5)
delay(100);
if (audioSource) audioSource->initialize(i2swsPin, i2ssdPin);
break;

View File

@@ -220,7 +220,7 @@ class I2SSource : public AudioSource {
// pin as DATA
_config.mode = i2s_mode_t(I2S_MODE_MASTER | I2S_MODE_RX | I2S_MODE_PDM); // Change mode to pdm if clock pin not provided. PDM is not supported on ESP32-S2. PDM RX not supported on ESP32-C3
_config.channel_format =I2S_PDM_MIC_CHANNEL; // seems that PDM mono mode always uses left channel.
//_config.use_apll = true; // experimental - use aPLL clock source to improve sampling quality
_config.use_apll = true; // experimental - use aPLL clock source to improve sampling quality
//_config.bits_per_sample = I2S_BITS_PER_SAMPLE_16BIT; // not needed
#endif
}
@@ -269,7 +269,7 @@ class I2SSource : public AudioSource {
}
DEBUGSR_PRINTF("AR: I2S#0 driver %s aPLL; fixed_mclk=%d.\n", _config.use_apll? "uses":"without", _config.fixed_mclk);
DEBUGSR_PRINTF("AR: Sample scaling factor = %6.4f\n", _sampleScale);
DEBUGSR_PRINTF("AR: %d bits, Sample scaling factor = %6.4f\n", _config.bits_per_sample, _sampleScale);
if(_config.mode & I2S_MODE_MASTER) {
if (_config.mode & I2S_MODE_PDM)
DEBUGSR_PRINTLN(F("AR: I2S#0 driver installed in PDM MASTER mode."));