Volume level problem - sound is too silent with MAX98357

[sir66]

Hi!

I have a small project:

https://oshwlab.com/sir066/max98357esp32_shield

It is the MAX98357 based 2*3Watt stereo amplifier shield for LILYGO® TTGO T7 V1.3 MINI 32 ESP32 and other pin-compatible ESP32 boards. The board contains two MAX98357 chips, SD slot and IR receiver. No DAC used, MAX98357 directly utilise i2s stream from esp32.

I have just trued to use ESP-AudioI2S library with my board. It work fine, but there is some problem. The sound level with setting audio.SetVolume(21) is definely much lower than when I use other audio libraries. I tried to double values in volumetable[]:

// const uint8_t volumetable[22]={ 0, 1, 2, 3, 4 , 6 , 8, 10, 12, 14, 17,
// 20, 23, 27, 30 ,34, 38, 43 ,48, 52, 58, 64}; //22 elements
const uint8_t volumetable[22]={ 0, 2, 4, 6, 8 , 12 , 16, 20, 24, 28, 34,
40, 46, 54, 60 , 68, 76, 86 , 96, 104, 116, 128}; //22 elements

and now I still have much more louder and absolutely clear sound. I tested it with number of track, including special full amplitude sine 1Khz 16bit sound test track. No any dшstortions noted. So I do not understand why such volumetable[] value uses in this library. I suppose that there is something wrong in this place. Who can explain me this feature?

[schreibfaul1]

The value of the output signal with 16bit/sample can be between INT16_MIN and INT16_MAX (-32768, 32767) at I2S for each channel. The lower or upper limit at the maximum level from the signal source must not be exceeded, as this would lead to distortion. Surely there is a margin that you can exploit, as in your example. Nevertheless, I am of the opinion that the level attenuation comes from the HW used.

[sir66]

I add the debug code to the library (in function Gain):

int32_t Audio::Gain(int16_t s[2]) {
int32_t v[2];
int32_t l = m_vol, r = m_vol;

/* balance is left -16...+16 right */
/* TODO: logarithmic scaling of balance, too? */
if(m_balance < 0){
    r -= (int32_t)m_vol * abs(m_balance) / 16;
} else if(m_balance > 0){
    l -= (int32_t)m_vol * abs(m_balance) / 16;
}

/* important: these multiplications must all be signed ints, or the result will be invalid */
v[LEFTCHANNEL] = (s[LEFTCHANNEL]  * l) / m_vol_step_div;
v[RIGHTCHANNEL]= (s[RIGHTCHANNEL] * r) / m_vol_step_div;

static unsigned long ccc;
static long min, max;

ccc++;

if (v[LEFTCHANNEL] > max) max = v[LEFTCHANNEL];
if (v[LEFTCHANNEL] < min) min = v[LEFTCHANNEL];

if (ccc > 44000) {
	Serial.print(min);
	Serial.print("\t");
	Serial.println(max);
	min = 0;
	max = 0;
	ccc = 0;
}

return (v[LEFTCHANNEL] << 16) | (v[RIGHTCHANNEL] & 0xffff);

}

Truing to play the following 1Khz sine file track099.mp3:

track0099.mp3

Track info:

Audacity.jpg

With standard values in volumetable[]:

const uint8_t volumetable[22]={   0,  1,  2,  3,  4 , 6 , 8, 10, 12, 14, 17,
                                 20, 23, 27, 30 ,34, 38, 43 ,48, 52, 58, 64}; //22 elements

we have the following output:

info PSRAM not found, inputBufferSize: 6399 bytes
info buffers freed, free Heap: 239292 bytes
info Reading file: "/track099.mp3"
info MP3Decoder has been initialized, free Heap: 211472 bytes
info Content-Length: 2401965
info ID3 framesSize: 45
info ID3 version: 2.3
info ID3 normal frames
id3data SettingsForEncoding: Lavf57.41.100
info Audio-Length: 2401920
info stream ready
info syncword found at pos 0
info Channels: 1
info SampleRate: 48000
info BitsPerSample: 16
info BitRate: 320000
-16384 16383
-16384 16383
-16384 16383

with doubled values in volumetable[]:

   const uint8_t volumetable[22]={   0,  2,  4,  6,  8 , 12 , 16, 20, 24, 28, 34,
                                 40, 46, 54, 60 , 68, 76, 86 , 96, 104, 116, 128}; //22 elements

we have output:

info PSRAM not found, inputBufferSize: 6399 bytes
info buffers freed, free Heap: 239292 bytes
info Reading file: "/track099.mp3"
info MP3Decoder has been initialized, free Heap: 211472 bytes
info Content-Length: 2401965
info ID3 framesSize: 45
info ID3 version: 2.3
info ID3 normal frames
id3data SettingsForEncoding: Lavf57.41.100
info Audio-Length: 2401920
info stream ready
info syncword found at pos 0
info Channels: 1
info SampleRate: 48000
info BitsPerSample: 16
info BitRate: 320000
-32768 32766
-32768 32766
-32768 32766
-32768 32766
-32768 32766
-32768 32766
-32768 32766

Thus, doubled values seems right.

[sir66]

I am using digital input chips. I do not change anything and I really can't change anything in I2S stream, it moves directly from I2S ESP32 output pins to MAX98357 (you could see a schematic). And I have no any distortion when playing full amplitude test 1 KHZ sine file with twice values in volumetable[]. It is notes that something wrong, I suppouse. Or, we deal with some strange feature of MAX98357 chip.

[schreibfaul1]

Your measurements are correct. What if you comment out this code?

This is to be able to use the digital filters.

[sir66]

There are no any miracles, if we comment out first 2 lines we will have normal output with stanard values in volumetable[].

id3data SettingsForEncoding: Lavf57.41.100
info Audio-Length: 2401920
info stream ready
info syncword found at pos 0
info Channels: 1
info SampleRate: 48000
info BitsPerSample: 16
info BitRate: 320000
-32768 32767
-32768 32767
-32768 32767
-32768 32767
-32768 32767
-32768 32767
-32768 32767
-32768 32767

I was originally sure that this is not a bug, but a feature :) I understand, that this conversion is helpful for digital filtering, but I am not sure, that it is a reason for decreasing digital levels anytime. It seems to me, that it may be switchable compile time or runtime option. In any case this conversion may be implemented by additional digital filter which applies before others. It is very uncomfortable to use such reduced stream in audio application based on digital input amplifiers.

According my experience, most of the digital track we are dealing with have a levels lower them it must have. Full amplitude audio track, such I use in my test is rather rare case. So it is not a good practice to decrease level of digital stream. Really in most application we had to increase digital levels. We had to engage to reserve of amplification for "silent" tracks. For example, I have just appeared that my ASUS VivoBook can't play test track mentioned above at maximum volume, I hear a huge distortions. Wherein I have never heard any distortions in real track before.

In any case, I want to thank you for your job, your library is excellent, It work fine, better then many others.

[schreibfaul1]

Thanks for your answer. The digital filters can not only reduce the level, but also increase it. Increasing the level can lead to distortion due to overflow. Therefore, when installing the filters, I lowered the volume by 'shift it to the right'. That was a long time ago and nobody has it on their screen yet. I have to admit this is not a good solution. I think I have an idea how I can improve this significantly.
Until then and best regards
Wolle

[gauravacad]

Dear schreibfaul1 sir

I AM USING A2DP SINK AND SOURCE LIBRARY.
USING INMP441. ITS WORKING CAUSE ANOTHER A2DP AUDIO TOOLS STRAMING IS NOT WORKING IN MY CASE.
I need your advice or help in two things. First how to work on audio control as gain vs sensitivity has issue. Present gain is 12. But mic pics noise white at gain 12 and lower. Can you please help me.

[gauravacad]

https://github.com/schreibfaul1/ESP32_A2DP_Source
Above question is related to this repo. Also is there a method to fix the a2dp delay which is meseaurable in 60 to 120 ms. Can we use other codec. Also please suggest a DC blocking filter, HW to plan it to remove constant white noise audible on gain 11 or 12 or 13 mostly.

[gauravacad]

Sorry for posting here. I can not find any other method to write to you sir.

[schreibfaul1]

This is a difficult thing. The microphone used has a processing width of 24 bits, I2S must be set to 32 bits (because 24 does not exist), and the actual BT transmission uses 16 bits per channel. The simplest method is to cut out the lower 8 bits.

            //int32_t sample = (buf_ptr_read1[3] << 24) + (buf_ptr_read1[2] << 16) + (buf_ptr_read1[1] << 8);
            //sample = sample >> gain;
            int16_t sample = (buf_ptr_read1[3] << 8) + (buf_ptr_read1[2] );

[gauravacad]

Thank you for your reply But Dear its sound goes very low Any middle solution using a filter Where and how shall I add a filer after samples. Also any where I can reduce LAtency in A2dp Tx?

…

On Sun, Aug 27, 2023 at 10:19 PM Wolle ***@***.***> wrote: This is a difficult thing. The microphone used has a processing width of 24 bits, I2S must be set to 32 bits (because 24 does not exist), and the actual BT transmission uses 16 bits per channel. The simplest method is to cut out the lower 8 bits. //int32_t sample = (buf_ptr_read1[3] << 24) + (buf_ptr_read1[2] << 16) + (buf_ptr_read1[1] << 8); //sample = sample >> gain; int16_t sample = (buf_ptr_read1[3] << 8) + (buf_ptr_read1[2] ); — Reply to this email directly, view it on GitHub <#475 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACV6JJDPZ7E72UUSD76YMS3XXN3BPANCNFSM6AAAAAAUOHZ7Y4> . You are receiving this because you commented.Message ID: ***@***.*** com>

-- SIMPLY GAURAV

[gauravacad]

Dear Sir , how can I use the Pink colour text having high gain and reducing noise. Please /* This example code is in the Public Domain (or CC0 licensed, at your option.) Unless required by applicable law or agreed to in writing, this software is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. Arduino port of https://github.com/espressif/esp-idf/tree/master/examples/bluetooth/a2dp_source origin URL: https://github.com/dgm3333/Arduino_A2DP_Source After connection with A2DP sink is established, the example performs the following running loop 1-2-3-4-1: 1. audio transmission starts and lasts for a while 2. audio transmission stops 3. disconnect with target device 4. reconnect to target device The example implements an event loop triggered by a periodic "heart beat" timer and events from Bluetooth protocol stack callback functions. For current stage, the supported audio codec in ESP32 A2DP is SBC (SubBand Coding). SBC specification is in Appendix B (page 50) of the document A2DP_Spec_V1_0 (can be found with search engine, although the original is behind the Bluetooth firewall) SBC audio stream is encoded from PCM data normally formatted as 44.1kHz sampling rate, two-channel 16-bit sample data. Other SBC configurations can be supported but there is a need for additional modifications to the protocol stack. */ #include "sos-iir-filter.h" #include "Arduino.h" #include "a2dp_source.h" #include <driver/i2s.h> //* SCK 5 = bit clock line (BCLK) //* FS 25 = WS PIN //* DIN 35 MIC #define RX_I2S_DIN 35 // connect with I2S microphone pin SO (signal out) #define RX_I2S_BCLK 5 // connect with I2S microphone pin SCK (bit clock) #define RX_I2S_LRC 33 // connect with I2S microphone pin WS (word select) //#define RX_I2S_DIN 33 // connect with I2S microphone pin SO (signal out) //#define RX_I2S_BCLK 15 // connect with I2S microphone pin SCK (bit clock) //#define RX_I2S_LRC 14 // connect with I2S microphone pin WS (word select) char BT_SINK_NAME[] = "Airdopes 141"; //"Infinity Glide 120, Nirvana Ion, Airdopes 141/boAt Rockerz 510";Smartbuy EB4MVBT // set your sink devicename here char BT_SINK_PIN[] = "0000"; // sink pincode char BT_DEVICE_NAME[] = "ESP_A2DP_SRC"; // source devicename const i2s_port_t I2S_PORT_RX = I2S_NUM_0; const uint32_t sample_rate = 44100; //44100; //43945; const uint16_t buf_len = 1024; size_t bytes_written = 0; char readBuff[buf_len]; uint16_t buffSize; uint8_t buffStat; uint8_t gain = 12; // reduce volume -> increase gain enum : uint8_t {BUFF_FULL, BUFF_EMPTY}; // DC-Blocker filter - removes DC component from I2S data // See: https://www.dsprelated.com/freebooks/filters/DC_Blocker.html // a1 = -0.9992 should heavily attenuate frequencies below 10Hz //SOS_IIR_Filter DC_BLOCKER = { //gain: 1.0, //sos: {{-1.0, 0.0, +0.9992, 0}} //}; // TDK/InvenSense INMP441 // Datasheet: https://www.invensense.com/wp-content/uploads/2015/02/INMP441.pdf // B ~= [1.00198, -1.99085, 0.98892] // A ~= [1.0, -1.99518, 0.99518] //SOS_IIR_Filter INMP441 = { //gain: 1.00197834654696, //sos: { // Second-Order Sections {b1, b2, -a1, -a2} //{-1.986920458344451, +0.986963226946616, +1.995178510504166, -0.995184322194091} //} //}; //--------------------------------------------------------------------------------------------------------------------- void i2s_install(){ /* RX: I2S_NUM_1 */ i2s_config_t i2s_config_rx = { .mode = (i2s_mode_t) (I2S_MODE_MASTER | I2S_MODE_RX), // Only TX I2S_COMM_FORMAT_STAND_PCM_SHORT .sample_rate = sample_rate, .bits_per_sample = I2S_BITS_PER_SAMPLE_32BIT, // Only 8-bit DAC support .channel_format = I2S_CHANNEL_FMT_RIGHT_LEFT,//I2S_CHANNEL_FMT_ALL_RIGHT, //I2S_CHANNEL_FMT_RIGHT_LEFT, // 2-channels .communication_format = (i2s_comm_format_t) I2S_COMM_FORMAT_STAND_I2S, // I2S_COMM_FORMAT_STAND_I2S, //I2S_COMM_FORMAT_STAND_PCM_SHORT, .intr_alloc_flags = ESP_INTR_FLAG_LEVEL2, // Interrupt level 1 .dma_buf_count = 4, // number of buffers, 128 max. .dma_buf_len = 512, // size of each buffer60 60 pe good hain // 4-512; .use_apll = false, .tx_desc_auto_clear = true, // new in V1.0.1 .fixed_mclk = I2S_PIN_NO_CHANGE, .mclk_multiple = I2S_MCLK_MULTIPLE_DEFAULT }; i2s_pin_config_t pin_config_rx = { .bck_io_num = RX_I2S_BCLK, .ws_io_num = RX_I2S_LRC, .data_out_num = I2S_PIN_NO_CHANGE, .data_in_num = RX_I2S_DIN }; i2s_driver_install(I2S_PORT_RX, &i2s_config_rx, 0, NULL); i2s_set_pin(I2S_PORT_RX, &pin_config_rx); } //---------------------------------------------SETUP-------------------------------------------------------------------- void setup(){ Serial.begin(115200); i2s_install(); buffStat = BUFF_EMPTY; log_i("free heap %i", esp_get_free_heap_size()); a2dp_source_init(BT_SINK_NAME, BT_SINK_PIN); log_i("free heap %i", esp_get_free_heap_size()); } //----------------------------------------------LOOP-------------------------------------------------------------------- void loop() { //char *buf_ptr_read1 = readBuff + 4; // connect L/R with VDD char *buf_ptr_read1 = readBuff; // connect L/R with GND char *buf_ptr_write1 = readBuff; if(buffStat == BUFF_EMPTY) { size_t bytes_read = 0; while(bytes_read == 0) { i2s_read(I2S_PORT_RX, readBuff, buf_len, &bytes_read, portMAX_DELAY); } uint32_t samples_read = bytes_read / 2 / (I2S_BITS_PER_SAMPLE_32BIT / 8); // convert 2x 32 bit stereo -> 1 x 16 bit mono for(int i = 0; i < samples_read; i++) { // left channel //int32_t sample = (buf_ptr_read1[3] << 24) + (buf_ptr_read1[2] << 16) + (buf_ptr_read1[1] << 8); //(buf_ptr_read1[4] << 32)+ //sample = sample >> gain; int32_t sample = (buf_ptr_read1[3] << 24) + (buf_ptr_read1[2]<<16); sample = sample >> gain; buf_ptr_write1[0] = sample & 0x00FF; buf_ptr_write1[1] = (sample >>8) & 0x00FF; // right channel buf_ptr_write1[2] = buf_ptr_write1[0]; // mid buf_ptr_write1[3] = buf_ptr_write1[1]; // high buf_ptr_write1 += 2 * (I2S_BITS_PER_SAMPLE_16BIT / 8); buf_ptr_read1 += 2 * (I2S_BITS_PER_SAMPLE_32BIT / 8); buffSize = samples_read * 2 * (I2S_BITS_PER_SAMPLE_16BIT / 8); } buffStat = BUFF_FULL; } bt_loop(); } //---------------------------------------------EVENTS------------------------------------------------------------------- int32_t bt_data(uint8_t *data, int32_t len, uint32_t* sr){ // BT data event *sr = 44100; if (len < 0 || data == NULL) { buffStat = BUFF_EMPTY; return 0; } if(!buffSize) return 0; memcpy(data, readBuff, buffSize); buffStat = BUFF_EMPTY; return buffSize; } void bt_info(const char* info){ Serial.printf("bt_info: %s\n", info); } On Mon, Aug 28, 2023 at 2:07 PM Dr. Gaurav Purohit ***@***.***> wrote:

…

Thank you for your reply But Dear its sound goes very low Any middle solution using a filter Where and how shall I add a filer after samples. Also any where I can reduce LAtency in A2dp Tx? On Sun, Aug 27, 2023 at 10:19 PM Wolle ***@***.***> wrote: > This is a difficult thing. The microphone used has a processing width of > 24 bits, I2S must be set to 32 bits (because 24 does not exist), and the > actual BT transmission uses 16 bits per channel. The simplest method is to > cut out the lower 8 bits. > > //int32_t sample = (buf_ptr_read1[3] << 24) + (buf_ptr_read1[2] << 16) + (buf_ptr_read1[1] << 8); > //sample = sample >> gain; > int16_t sample = (buf_ptr_read1[3] << 8) + (buf_ptr_read1[2] ); > > — > Reply to this email directly, view it on GitHub > <#475 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ACV6JJDPZ7E72UUSD76YMS3XXN3BPANCNFSM6AAAAAAUOHZ7Y4> > . > You are receiving this because you commented.Message ID: > ***@***.*** > .com> > -- SIMPLY GAURAV

-- SIMPLY GAURAV

[schreibfaul1]

The microphone runs in 24 bit mode and BT uses 16 bit. Thus, the value range of the microphone is 256 times larger. This causes a right shift of 8 bits. In my example it is less, which increases the sensitivity, but an overflow (larger INT16_MAX or smaller INT16_MIN) You can build a filter with a curve that amplifies low signals and attenuates strong pages.

[gauravacad]

Thanks sir, But i am unable to do the same as you suggested. Meanwhile I get this suggestion ********* Now what i recommend is converting it to 16-bit signed samples: CODE: SELECT ALL <https://www.esp32.com/viewtopic.php?t=15185#> int16_t buffer16[ESP_NOW_MAX_DATA_LEN] = {0};for (int i=0; i<samplesRead; i++) { uint8_t mid = buffer32[i * 4 + 2]; uint8_t msb = buffer32[i * 4 + 3]; uint16_t raw = (((uint32_t)msb) << 8) + ((uint32_t)mid); memcpy(&buffer16[i], &raw, sizeof(raw)); // Copy so sign bits aren't interfered with somehow. }************By Chris_oz link: https://www.esp32.com/viewtopic.php?t=15185Can you lead to use the same in your existing code.

…

On Fri, Sep 1, 2023 at 12:52 AM Wolle ***@***.***> wrote: The microphone runs in 24 bit mode and BT uses 16 bit. Thus, the value range of the microphone is 256 times larger. This causes a right shift of 8 bits. In my example it is less, which increases the sensitivity, but an overflow (larger INT16_MAX or smaller INT16_MIN) You can build a filter with a curve that amplifies low signals and attenuates strong pages. — Reply to this email directly, view it on GitHub <#475 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACV6JJEDXBIWWN7N5DP2EJ3XYDP7JANCNFSM6AAAAAAUOHZ7Y4> . You are receiving this because you commented.Message ID: ***@***.*** com>

-- SIMPLY GAURAV

[gauravacad]

The above mail is regarding The microphone is a 24-bit one, but if you use bits_per_sample = I2S_BITS_PER_SAMPLE_24BIT it doesn't work, maybe an ESP bug? Anyway 32 bits works with some workarounds. The 4 bytes that make up each sample in buffer32 are as follows: First byte: E0 or 00, regardless of the sign of the other bytes, discard these! Byte 2: Lowest significant byte of the sample. You can also discard this because it's just fuzz even in a quiet room. Byte 3: Middle byte of the sample (signed) Last byte: Most significant byte of the sample (signed). Now what i recommend is converting it to 16-bit signed samples:

…

On Tue, Sep 5, 2023 at 2:25 PM Dr. Gaurav Purohit ***@***.***> wrote: Thanks sir, But i am unable to do the same as you suggested. Meanwhile I get this suggestion ********* Now what i recommend is converting it to 16-bit signed samples: CODE: SELECT ALL <https://www.esp32.com/viewtopic.php?t=15185#> int16_t buffer16[ESP_NOW_MAX_DATA_LEN] = {0};for (int i=0; i<samplesRead; i++) { uint8_t mid = buffer32[i * 4 + 2]; uint8_t msb = buffer32[i * 4 + 3]; uint16_t raw = (((uint32_t)msb) << 8) + ((uint32_t)mid); memcpy(&buffer16[i], &raw, sizeof(raw)); // Copy so sign bits aren't interfered with somehow. }************By Chris_oz link: https://www.esp32.com/viewtopic.php?t=15185Can you lead to use the same in your existing code. On Fri, Sep 1, 2023 at 12:52 AM Wolle ***@***.***> wrote: > The microphone runs in 24 bit mode and BT uses 16 bit. Thus, the value > range of the microphone is 256 times larger. This causes a right shift of 8 > bits. In my example it is less, which increases the sensitivity, but an > overflow (larger INT16_MAX or smaller INT16_MIN) You can build a filter > with a curve that amplifies low signals and attenuates strong pages. > > — > Reply to this email directly, view it on GitHub > <#475 (comment)>, > or unsubscribe > <https://github.com/notifications/unsubscribe-auth/ACV6JJEDXBIWWN7N5DP2EJ3XYDP7JANCNFSM6AAAAAAUOHZ7Y4> > . > You are receiving this because you commented.Message ID: > ***@***.*** > .com> > -- SIMPLY GAURAV

-- SIMPLY GAURAV

[gauravacad]

// Define a simple compressor function int32_t compress(int32_t input, float threshold, float ratio) { // Apply compression only if the input exceeds the threshold if (input > threshold) { // Apply compression ratio input = threshold + (input - threshold) / ratio; } return input; } int main() { // Sample byte array buf_ptr_read1 uint8_t buf_ptr_read1[4] = {0x11, 0x22, 0x33, 0x44}; // Combine bytes at indices 3 and 2 to create a 32-bit integer int32_t sample = (buf_ptr_read1[3] << 24) + (buf_ptr_read1[2] << 16); // Right-shift the combined sample (as in your original code) sample = sample >> 14; // Apply compression (you can adjust threshold and ratio as needed) float threshold = 0x1000; // Adjust this threshold value as needed float ratio = 2.0; // Adjust this compression ratio as needed sample = compress(sample, threshold, ratio); // Extract and store the compressed sample in buf_ptr_write1 buf_ptr_write1[0] = sample & 0x00FF; buf_ptr_write1[1] = (sample >> 8) & 0x00FF; // Print the result printf("Sample: 0x%X%X\n", buf_ptr_write1[1], buf_ptr_write1[0]); return 0; } is this want to say for dynamic range compression am I doing right

…

On Tue, Sep 5, 2023 at 2:26 PM Dr. Gaurav Purohit ***@***.***> wrote: The above mail is regarding The microphone is a 24-bit one, but if you use bits_per_sample = I2S_BITS_PER_SAMPLE_24BIT it doesn't work, maybe an ESP bug? Anyway 32 bits works with some workarounds. The 4 bytes that make up each sample in buffer32 are as follows: First byte: E0 or 00, regardless of the sign of the other bytes, discard these! Byte 2: Lowest significant byte of the sample. You can also discard this because it's just fuzz even in a quiet room. Byte 3: Middle byte of the sample (signed) Last byte: Most significant byte of the sample (signed). Now what i recommend is converting it to 16-bit signed samples: On Tue, Sep 5, 2023 at 2:25 PM Dr. Gaurav Purohit ***@***.***> wrote: > Thanks sir, But i am unable to do the same as you suggested. > Meanwhile I get this suggestion > > ********* > Now what i recommend is converting it to 16-bit signed samples: > > CODE: SELECT ALL <https://www.esp32.com/viewtopic.php?t=15185#> > > int16_t buffer16[ESP_NOW_MAX_DATA_LEN] = {0};for (int i=0; i<samplesRead; i++) { > uint8_t mid = buffer32[i * 4 + 2]; > uint8_t msb = buffer32[i * 4 + 3]; > uint16_t raw = (((uint32_t)msb) << 8) + ((uint32_t)mid); > memcpy(&buffer16[i], &raw, sizeof(raw)); // Copy so sign bits aren't interfered with somehow. > }************By Chris_oz link: https://www.esp32.com/viewtopic.php?t=15185Can you lead to use the same in your existing code. > > > On Fri, Sep 1, 2023 at 12:52 AM Wolle ***@***.***> wrote: > >> The microphone runs in 24 bit mode and BT uses 16 bit. Thus, the value >> range of the microphone is 256 times larger. This causes a right shift of 8 >> bits. In my example it is less, which increases the sensitivity, but an >> overflow (larger INT16_MAX or smaller INT16_MIN) You can build a filter >> with a curve that amplifies low signals and attenuates strong pages. >> >> — >> Reply to this email directly, view it on GitHub >> <#475 (comment)>, >> or unsubscribe >> <https://github.com/notifications/unsubscribe-auth/ACV6JJEDXBIWWN7N5DP2EJ3XYDP7JANCNFSM6AAAAAAUOHZ7Y4> >> . >> You are receiving this because you commented.Message ID: >> <schreibfaul1/ESP32-audioI2S/repo-discussions/475/comments/6879204@ >> github.com> >> > > > -- > SIMPLY GAURAV > -- SIMPLY GAURAV

-- SIMPLY GAURAV

[gauravacad]

Dear Sir any comment or help on improving the Lag on A2DP

[gauravacad]

Respected Why Cant I change const uint16_t buf_len In the A2dp Code , I need to experiment whether I can improve the latency issue in A2dp Can I change SBS codec functionalities anywhere? Regards gaurav

[gauravacad]

Dear Respected

Why Cant I change const uint16_t buf_len 1024 to 512

can I improve the latency by reducing this value?

…

In the A2dp Code , I need to experiment whether I can improve the latency issue in A2dp Can I change SBS codec functionalities anywhere? Regards gaurav

-- SIMPLY GAURAV