XIAO BLE Sense nRF52840 Support

[BobaTrek]

I have a small 2K buffer that I am trying to stream out to I2S. For some reason, the following line of code in I2SNanoSenseBLE.h appears to be locking up the processor:

NRF_I2S->TXD.PTR = (uint32_t) i2s_buffer.readEnd().address();

I broke up the line into a couple calls to narrow it down:

uint32_t address = 0;
address = (uint32_t) i2s_buffer.readEnd().address();
NRF_I2S->TXD.PTR = address;

The lockup occurs when i2s_buffer.readEnd().address() is called.

Here is the Arduino sketch I am using:

/**
 * @file streams-memory_raw-i2s.ino
 * @author Phil Schatzmann
 * @brief Compile with Partition Scheme Hughe APP!
 * @version 0.1
 * @date 2021-01-24
 * 
 * @copyright Copyright (c) 2021
 * 
 */
#include "AudioTools.h"
#include "alice.h"

//AudioInfo info(22050, 1, 16);
AudioInfo info(8000, 1, 16);
I2SStream i2s;  // Output to I2S
MemoryStream music(alice_wav, 2048);  // Only use first 2K of buffer for now
StreamCopy copier(i2s, music); // copies sound into i2s

void setup(){
    Serial.begin(115200);
    delay(5000);
    Serial.println("setup()");
    AudioLogger::instance().begin(Serial, AudioLogger::Info);

    auto config = i2s.defaultConfig(TX_MODE);
    config.copyFrom(info);


    config.pin_ws = 46;
    config.pin_data = 44;
    config.pin_bck = 45;
    i2s.begin(config);
}

void loop(){
    if (!copier.copy()){
      i2s.end();
      stop();
    }
}

I put a Serial.println() just before the call to i2s_buffer.readEnd().address() and just after the call. I only see the first println() message.

Using println() debugging in the readEnd() function of the BaseBuffer class (Buffers.h), I was able to narrow down the problem to when readEnd() returns. You get a null ptr crash because actual_read_buffer is null:

return *actual_read_buffer;  // Crashes for me

I am unsure why the actual_read_buffer is null as the logic for the buffers themselves is further than I wanted to dig for now. I hope I did not simply forget to set something up properly before calling i2s.begin()

Any help would be appreciated.

[pschatzmann]

I started to work on the I2S Support for this procesor, but never found the time to complete it.
Maybe you can try to dig into this ?

[BobaTrek]

I found at least "a" problem. The null pointer reference was basically because no data had been copied into the buffers yet. The trace looked like:
readEnd() ->
resetCurrent() ->
actual_read_buffer = getNextFilledBuffer() which returns nullptr ->
null pointer crash when resetCurrent() does "return *actual_read_buffer;"

The reason actual_read_buffer was still nullptr was because getNextFilledBuffer() got called with write_buffer_count still 0. In that case, it returns nullptr.

A potential fix is to make sure that the write_buffer_count > 0 so that actual_read_buffer gets a buffer value instead of nullptr.

This can be achieved by adding a copier.copy() prior to all of this i2s begin stuff.

I think a better fix would be to handle the case when i2s.being() is called and no buffers have been copied into yet.

Here is code that no longer locks up when i2s.begin() is called:

/**
 * @file streams-memory_raw-i2s.ino
 * @author Phil Schatzmann
 * @brief Compile with Partition Scheme Hughe APP!
 * @version 0.1
 * @date 2021-01-24
 * 
 * @copyright Copyright (c) 2021
 * 
 */
#include "AudioTools.h"
#include "alice.h"

//AudioInfo info(22050, 1, 16);
AudioInfo info(8000, 1, 16);
I2SStream i2s;  // Output to I2S
MemoryStream music(alice_wav, 4096);  // Only use first 2K of buffer for now
StreamCopy copier(i2s, music); // copies sound into i2s

void setup(){
    Serial.begin(115200);
    delay(5000);
    Serial.println("setup()");
    AudioLogger::instance().begin(Serial, AudioLogger::Info);

    auto config = i2s.defaultConfig(TX_MODE);
    config.copyFrom(info);


    config.pin_ws = 46;
    config.pin_data = 44;
    config.pin_bck = 45;
    copier.copy();  // Need to call at least once prior to calling i2s.begin()
    i2s.begin(config);
}

void loop(){
    Serial.println("loop()");
    
    if (!copier.copy()){
      i2s.end();
      //stop();
      delay(1000);
    }
}

And here is the output (including some of the prints I inserted in the i2s code):

setup()
addFilledBuffer() enter, write_buffer_count = 0, buffer_count = 5
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
I2SStream::begin()
I2SDriverNanoBLE::begin()
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
[I] I2SNanoSenseBLE.h : 202 - frequency requested 128000.000000 vs 125000.000000
setupData()
Calling address = (uint32_t) i2s_buffer.readEnd().address();
readEnd() enter
resetCurrent() enter
getNextFilledBuffer() enter, write_buffer_count = 1
getNextFilledBuffer() exit, write_buffer_count = 0
resetCurrent() exit
readEnd() exit, actual_read_buffer = 20010E38
Returned from uint32_t address = (uint32_t) i2s_buffer.readEnd().address();
addFilledBuffer() enter, write_buffer_count = 0, buffer_count = 5
loop()
addFilledBuffer() enter, write_buffer_count = 1, buffer_count = 5
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
loop()
addFilledBuffer() enter, write_buffer_count = 2, buffer_count = 5
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
loop()
addFilledBuffer() enter, write_buffer_count = 3, buffer_count = 5
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
loop()
loop()
loop()
loop()
loop()

I am going to try some larger audio files now and do a little more digging.

[pschatzmann]

Thats's good news. Would the following maybe help ?

Prevent potential NPE in IRQ

const uin8_t i2s_empty_array[i2s_buffer_size] = {0};

extern "C"  void I2S_IRQHandler(void) {
    if(NRF_I2S->EVENTS_TXPTRUPD != 0) {
      // writing to pins from buffer
      NRF_I2S->TXD.PTR = (uint32_t) !i2s_buffer.isEmpty() ? i2s_buffer.readEnd().address() : i2s_empty_array; // last buffer was processed
      NRF_I2S->EVENTS_TXPTRUPD = 0;

    } else if(NRF_I2S->EVENTS_RXPTRUPD != 0) {
      // reading from pins writing to buffer
      NRF_I2S->RXD.PTR = (uint32_t) i2s_buffer.writeEnd().address(); // last buffer was processed
      NRF_I2S->EVENTS_RXPTRUPD = 0;
    }
} 

I have the feeling that the setupData() should be commented out in begin().
Maybe it is also better to move the following start processing in the case of TX_MODE

        // start i2s
        NRF_I2S->ENABLE = 1;
        // start task
        NRF_I2S->TASKS_START = 1;

to the end of writeBytes(), so that the processing is only started when there is data available

[pschatzmann]

I committed some DRAFT corrections. However I noticed that the I2S_IRQHandler is never called.
Did this work for you ?

[BobaTrek]

I2S_IRQHandler is never called for me either.

So I added one line to setupIRQ to set NRF_I2S->INTENSET:

    void setupIRQ() {
        TRACED();
	NRF_I2S->INTENSET = I2S_INTENSET_TXPTRUPD_Enabled << I2S_INTENSET_TXPTRUPD_Pos;
        NVIC_EnableIRQ(I2S_IRQn);      
    }

The good news is that I believe the interrupt is being enabled now. The bad news is that it hangs the processor.

I think that the hang is not due to the code inside I2S_IRQHandler as it never reaches that code (I tried putting a set LED on in there). It appears the processor hands when the interrupt occurs, but before it reaches our I2S_IRQHandler.

I have another smaller project set up which simply pumps a static array out the I2S making all the required NRF_I2S calls, and it also hangs when the I2S_IRQn is enabled. So I think it is something I2S interrupt specific.

[pschatzmann]

Strange that NRF_I2S->INTENSET is not mentioned in chapter 6.11.8 Module operation of the Product Specification

[BobaTrek]

I found it in this forum discussion with Nordic Q&A:
https://devzone.nordicsemi.com/f/nordic-q-a/84624/i2s

The code the tech support recommended is copied below from that forum.

#include <nrf.h>

#define PIN_MCK    (13)
#define PIN_SCK    (14)
#define PIN_LRCK   (15)
#define PIN_SDOUT  (16)

int main(void)
{
  // Sine table (stereo, so every value is duplicated)
  int16_t sine_table[] = { 0, 0, 23170, 23170, 32767, 32767, 23170, 23170, 0, 0, -23170, -23170, -32768, -32768, -23170, -23170};
  
  // Enable transmission
  NRF_I2S->CONFIG.TXEN = (I2S_CONFIG_TXEN_TXEN_ENABLE << I2S_CONFIG_TXEN_TXEN_Pos);
  
  // Enable MCK generator
  NRF_I2S->CONFIG.MCKEN = (I2S_CONFIG_MCKEN_MCKEN_ENABLE << I2S_CONFIG_MCKEN_MCKEN_Pos);
  
  // MCKFREQ = 4 MHz
  NRF_I2S->CONFIG.MCKFREQ = I2S_CONFIG_MCKFREQ_MCKFREQ_32MDIV11  << I2S_CONFIG_MCKFREQ_MCKFREQ_Pos;
  
  // Ratio = 64 
  NRF_I2S->CONFIG.RATIO = I2S_CONFIG_RATIO_RATIO_64X << I2S_CONFIG_RATIO_RATIO_Pos;
    
  // Master mode, 16Bit, left aligned
  NRF_I2S->CONFIG.MODE = I2S_CONFIG_MODE_MODE_MASTER << I2S_CONFIG_MODE_MODE_Pos;
  NRF_I2S->CONFIG.SWIDTH = I2S_CONFIG_SWIDTH_SWIDTH_16BIT << I2S_CONFIG_SWIDTH_SWIDTH_Pos;
  NRF_I2S->CONFIG.ALIGN = I2S_CONFIG_ALIGN_ALIGN_LEFT << I2S_CONFIG_ALIGN_ALIGN_Pos;
  
  // Format = I2S
  NRF_I2S->CONFIG.FORMAT = I2S_CONFIG_FORMAT_FORMAT_I2S << I2S_CONFIG_FORMAT_FORMAT_Pos;
  
  // Use stereo 
  NRF_I2S->CONFIG.CHANNELS = I2S_CONFIG_CHANNELS_CHANNELS_STEREO << I2S_CONFIG_CHANNELS_CHANNELS_Pos;
  
  // Configure pins
  NRF_I2S->PSEL.MCK = (PIN_MCK << I2S_PSEL_MCK_PIN_Pos);
  NRF_I2S->PSEL.SCK = (PIN_SCK << I2S_PSEL_SCK_PIN_Pos); 
  NRF_I2S->PSEL.LRCK = (PIN_LRCK << I2S_PSEL_LRCK_PIN_Pos); 
  NRF_I2S->PSEL.SDOUT = (PIN_SDOUT << I2S_PSEL_SDOUT_PIN_Pos);
  
  
  // Configure data pointer
  NRF_I2S->TXD.PTR = (uint32_t)&sine_table[0];
  NRF_I2S->RXTXD.MAXCNT = sizeof(sine_table) / sizeof(uint32_t);
  
  //Enable interrupt and IRQ
  NRF_I2S->INTENSET = I2S_INTENSET_TXPTRUPD_Enabled << I2S_INTENSET_TXPTRUPD_Pos;

  
  NVIC_EnableIRQ(I2S_IRQn);
  
  NRF_I2S->ENABLE = 1;
  // Start transmitting I2S data
  NRF_I2S->TASKS_START = 1;
  
  
  // Since we are not updating the TXD pointer, the sine wave will play over and over again.
  // The TXD pointer can be updated after the EVENTS_TXPTRUPD arrives.
  while (1)
  {
    //__WFE();
  }
}

//IRQ handler

void I2S_IRQHandler(void)
{
  if (NRF_I2S->EVENTS_TXPTRUPD == 1)
  {
      NRF_I2S->EVENTS_TXPTRUPD = 0;
      NRF_I2S->TASKS_STOP = 1;

  }
}

[pschatzmann]

Calling NVIC_SetVector(I2S_IRQn, (uint32_t)I2S_IRQHandler); seems to make the difference.
Here is the corresponding Arduino Sketch:

#include <nrf.h>

#define PIN_MCK (13)
#define PIN_SCK (14)
#define PIN_LRCK (15)
#define PIN_SDOUT (16)
int count = 0;

void I2S_IRQHandler(void) {
  count++;
  if (NRF_I2S->EVENTS_TXPTRUPD == 1) {
    NRF_I2S->EVENTS_TXPTRUPD = 0;
    //NRF_I2S->TASKS_STOP = 1;
  }
}

void setup() {
  Serial.begin(115200);
  while (!Serial)
    ;
  Serial.print("start");
  delay(1000);

  // Sine table (stereo, so every value is duplicated)
  int16_t sine_table[] = { 0, 0, 23170, 23170, 32767, 32767, 23170, 23170, 0, 0, -23170, -23170, -32768, -32768, -23170, -23170 };

  //Define IRQ handler
  NVIC_SetVector(I2S_IRQn, (uint32_t)I2S_IRQHandler);

  // Enable transmission
  NRF_I2S->CONFIG.TXEN = (I2S_CONFIG_TXEN_TXEN_ENABLE << I2S_CONFIG_TXEN_TXEN_Pos);

  // Enable MCK generator
  NRF_I2S->CONFIG.MCKEN = (I2S_CONFIG_MCKEN_MCKEN_ENABLE << I2S_CONFIG_MCKEN_MCKEN_Pos);

  // MCKFREQ = 4 MHz
  NRF_I2S->CONFIG.MCKFREQ = I2S_CONFIG_MCKFREQ_MCKFREQ_32MDIV11 << I2S_CONFIG_MCKFREQ_MCKFREQ_Pos;

  // Ratio = 64
  NRF_I2S->CONFIG.RATIO = I2S_CONFIG_RATIO_RATIO_64X << I2S_CONFIG_RATIO_RATIO_Pos;

  // Master mode, 16Bit, left aligned
  NRF_I2S->CONFIG.MODE = I2S_CONFIG_MODE_MODE_MASTER << I2S_CONFIG_MODE_MODE_Pos;
  NRF_I2S->CONFIG.SWIDTH = I2S_CONFIG_SWIDTH_SWIDTH_16BIT << I2S_CONFIG_SWIDTH_SWIDTH_Pos;
  NRF_I2S->CONFIG.ALIGN = I2S_CONFIG_ALIGN_ALIGN_LEFT << I2S_CONFIG_ALIGN_ALIGN_Pos;

  // Format = I2S
  NRF_I2S->CONFIG.FORMAT = I2S_CONFIG_FORMAT_FORMAT_I2S << I2S_CONFIG_FORMAT_FORMAT_Pos;

  // Use stereo
  NRF_I2S->CONFIG.CHANNELS = I2S_CONFIG_CHANNELS_CHANNELS_STEREO << I2S_CONFIG_CHANNELS_CHANNELS_Pos;

  // Configure pins
  NRF_I2S->PSEL.MCK = (PIN_MCK << I2S_PSEL_MCK_PIN_Pos);
  NRF_I2S->PSEL.SCK = (PIN_SCK << I2S_PSEL_SCK_PIN_Pos);
  NRF_I2S->PSEL.LRCK = (PIN_LRCK << I2S_PSEL_LRCK_PIN_Pos);
  NRF_I2S->PSEL.SDOUT = (PIN_SDOUT << I2S_PSEL_SDOUT_PIN_Pos);


  // Configure data pointer
  NRF_I2S->TXD.PTR = (uint32_t)&sine_table[0];
  NRF_I2S->RXTXD.MAXCNT = sizeof(sine_table) / sizeof(uint32_t);

  //Enable interrupt and IRQ
  NRF_I2S->INTENSET = I2S_INTENSET_TXPTRUPD_Enabled << I2S_INTENSET_TXPTRUPD_Pos;
  NVIC_EnableIRQ(I2S_IRQn);

  NRF_I2S->ENABLE = 1;
  // Start transmitting I2S data
  NRF_I2S->TASKS_START = 1;
}

void loop() {
  // monitor irq handler
  delay(1000);
  Serial.println(count);
  count = 0;
}

I tried to add some corrections to the I2SNanoSenseBLE: It looks pretty promising...

[BobaTrek]

Great find!

It did not work initially on my version of the code until I also commented out the line that turned on the Blue LED in the I2S_IRQHandler routine. Lesson learned - Don't set output pins in interrupt handlers!

Great sleuthing! Thanks!

[BobaTrek]

FYI: I discovered that

NRF_I2S->TASKS_STOP = 1;

cannot be called within the I2S_IRQHandler() as it locks up the processor. I think because according to the documentation, calling NRF_I2S->TASKS_STOP causes the STOPPED event to be generated, so you get an interrupt within an interrupt?

So I tried:

NRF_I2S->INTENCLR = 1;
NRF_I2S->TASKS_STOP = 1;

but this also locked up the processor. I am unsure why.

NRF_I2S->ENABLE = 0;

however can be called to stop the audio.

The first interrupt and EVENTS_TXPTRUPD is received almost immediately. This is because the interrupt occurs right after the processor has copied the data to it's internal double buffer, but has yet to even transmit it. So if you call NRF_I2S->ENABLE = 0 without checking how many EVENTS_TXPTRUPD have been received, then it stops too early. In fact, if you call NRF_I2S->ENABLE = 0 at the first EVENTS_TXPTRUPD, it will be stopped before anything has been output at all yet.

I am not sure what mechanism to use to know that the processor has actually finished transmitting the entire audio buffer (and not just finished copying the data into the internal double buffers).

[pschatzmann]

Are you running into issues with my currently implemented solution ?

I am currently activating the events after the first write, but we can change this to be done when the p_i2s_buffer buffer is full, to give enough headroom to prevent that we are running out of audio data in the IRQ.

In the IRQ, it would also make sense to clear the array before calling readArray, to make sure that we don't output any audio signal, when there is no more data. So I don't think that any additional special logic would be needed.

Just committed the corresponding correction

[BobaTrek]

I did a get latest from github and the processor is locking up as soon as startI2SActive() is called. Here is the debug output:

setup()
[D] I2SStream.h : 58 - bool audio_tools::I2SStream::begin(audio_tools::I2SConfig)
[I] I2SNanoSenseBLE.h : 118 - bool audio_tools::I2SDriverNanoBLE::begin(audio_tools::I2SConfig)
[D] I2SNanoSenseBLE.h : 352 - bool audio_tools::I2SDriverNanoBLE::setupBuffers()
[D] I2SNanoSenseBLE.h : 202 - void audio_tools::I2SDriverNanoBLE::setupRxTx(audio_tools::I2SConfig)
[D] I2SNanoSenseBLE.h : 214 - void audio_tools::I2SDriverNanoBLE::setupClock(audio_tools::I2SConfig)
[D] I2SNanoSenseBLE.h : 233 - frequency requested 16000.000000 vs 31250.000000
[D] I2SNanoSenseBLE.h : 233 - frequency requested 16000.000000 vs 20833.333984
[D] I2SNanoSenseBLE.h : 233 - frequency requested 16000.000000 vs 15625.000000
[D] I2SNanoSenseBLE.h : 233 - frequency requested 16000.000000 vs 15873.015625
[D] I2SNanoSenseBLE.h : 233 - frequency requested 16000.000000 vs 15873.016602
[I] I2SNanoSenseBLE.h : 237 - Frequency req. 16000.000000 vs eff. 15873.016602
[D] I2SNanoSenseBLE.h : 242 - void audio_tools::I2SDriverNanoBLE::setupBitWidth(audio_tools::I2SConfig)
[D] I2SNanoSenseBLE.h : 261 - void audio_tools::I2SDriverNanoBLE::setupMode(audio_tools::I2SConfig)
[D] I2SNanoSenseBLE.h : 285 - void audio_tools::I2SDriverNanoBLE::setupPins(audio_tools::I2SConfig)
[D] D:\Robert\DEV\Arduino\Bob\streams_memory_raw_i2s\streams_memory_raw_i2s.ino : 38 - void loop()
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] D:\Robert\DEV\Arduino\Bob\streams_memory_raw_i2s\streams_memory_raw_i2s.ino : 38 - void loop()
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] D:\Robert\DEV\Arduino\Bob\streams_memory_raw_i2s\streams_memory_raw_i2s.ino : 38 - void loop()
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] D:\Robert\DEV\Arduino\Bob\streams_memory_raw_i2s\streams_memory_raw_i2s.ino : 38 - void loop()
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] D:\Robert\DEV\Arduino\Bob\streams_memory_raw_i2s\streams_memory_raw_i2s.ino : 38 - void loop()
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] D:\Robert\DEV\Arduino\Bob\streams_memory_raw_i2s\streams_memory_raw_i2s.ino : 38 - void loop()
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] D:\Robert\DEV\Arduino\Bob\streams_memory_raw_i2s\streams_memory_raw_i2s.ino : 38 - void loop()
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] I2SStream.h : 92 - I2SStream::write: 1024
[D] Buffers.h : 63 - writeArray 1024 -> 0
[D] I2SNanoSenseBLE.h : 328 - void audio_tools::I2SDriverNanoBLE::startI2SActive()

The .ino code I am running is streams_memory_raw_i2s.ino from the examples directory. I have only added a few lines to change I2S pins and some prints:

/**
 * @file streams-memory_raw-i2s.ino
 * @author Phil Schatzmann
 * @brief Compile with Partition Scheme Hughe APP!
 * @version 0.1
 * @date 2021-01-24
 * 
 * @copyright Copyright (c) 2021
 * 
 */
#include "AudioTools.h"
#include "StarWars30.h"

AudioInfo info(22050, 1, 16);
I2SStream i2s;  // Output to I2S
MemoryStream music(StarWars30_raw, sizeof(StarWars30_raw));
StreamCopy copier(i2s, music); // copies sound into i2s

void setup(){
    Serial.begin(115200);
    delay(5000);
    Serial.println("setup()");
    AudioLogger::instance().begin(Serial, AudioLogger::Debug);

    auto config = i2s.defaultConfig(TX_MODE);
    config.copyFrom(info);

    config.pin_ws = 44;  // Pin D7 on XIAO nRF52840
    config.pin_data = 46;  // Pin D9 on XIAO nRF52840
    config.pin_bck = 45;  // Pin D8 on XIAO nRF52840

    i2s.begin(config);
}

void loop(){
    TRACED();
    //i2s.logIRQCount();
    
    if (!copier.copy()){
      i2s.end();
      //stop();
      delay(1000);
    }
}

I also do not get any audio output, so the call to startI2SActive() is locking the processor before the I2S gets started.

I have played around a little with the latest from github. I commented everything out of the I2S_IRQHandler() handler except for the irq_count++. I then added a function to print the value from the .ino loop, and it is always 0. This makes sense since never hear anything.

I am now closely comparing the NRF_I2S calls between the simple example we got working and the audiotools version.

[BobaTrek]

Do you have a physical XIAO nRF52840 to run the code on?

[pschatzmann]

I suggest to use the simplest possible sketch for testing as a starting point: https://github.com/pschatzmann/arduino-audio-tools/blob/main/examples/examples-stream/streams-generator-i2s/streams-generator-i2s.ino

Your pins are wrong! Use Arduino Conventions!

For me the above sketch was working. I did not test the signal though...

[BobaTrek]

I will switch to streams-generator.

What is the Arduino convention for specifying the pins? I tried all sorts of variation, but I could only get the calculated numbers to work, i.e. on the XIAO nRF52840, the pad labelled D7 is connected to the processor's port 1.12. So 1*32 + 12 = 44. I also tried "D7" but it did not work. The pin assignments 44, 45, and 46 are definitely working at least (on the simple example)

[pschatzmann]

https://github.com/arduino/ArduinoCore-mbed/blob/main/variants/ARDUINO_NANO33BLE/pins_arduino.h

If you check in my code, I am using the digitalPinToPinName() method to convert the Arduino Pin.
When I tested last, I just used the default pins and it was not locking up....

[BobaTrek]

streams-generator-i2s.ino behaves the same with regard to the I2S. The processor appears to lockup right at the same place - the call to startI2SActive().

Here is the debug output for streams-generator-i2s.ino:

starting I2S...
[D] I2SStream.h : 58 - bool audio_tools::I2SStream::begin(audio_tools::I2SConfig)
[I] I2SNanoSenseBLE.h : 121 - bool audio_tools::I2SDriverNanoBLE::begin(audio_tools::I2SConfig)
[D] I2SNanoSenseBLE.h : 356 - bool audio_tools::I2SDriverNanoBLE::setupBuffers()
[D] I2SNanoSenseBLE.h : 204 - void audio_tools::I2SDriverNanoBLE::setupRxTx(audio_tools::I2SConfig)
[D] I2SNanoSenseBLE.h : 216 - void audio_tools::I2SDriverNanoBLE::setupClock(audio_tools::I2SConfig)
[D] I2SNanoSenseBLE.h : 235 - frequency requested 44100.000000 vs 83333.335938
[D] I2SNanoSenseBLE.h : 235 - frequency requested 44100.000000 vs 62500.000000
[D] I2SNanoSenseBLE.h : 235 - frequency requested 44100.000000 vs 41666.667969
[D] I2SNanoSenseBLE.h : 235 - frequency requested 44100.000000 vs 45454.546875
[D] I2SNanoSenseBLE.h : 235 - frequency requested 44100.000000 vs 44444.449219
[I] I2SNanoSenseBLE.h : 239 - Frequency req. 44100.000000 vs eff. 44444.449219
[D] I2SNanoSenseBLE.h : 244 - void audio_tools::I2SDriverNanoBLE::setupBitWidth(audio_tools::I2SConfig)
[D] I2SNanoSenseBLE.h : 263 - void audio_tools::I2SDriverNanoBLE::setupMode(audio_tools::I2SConfig)
[D] I2SNanoSenseBLE.h : 287 - void audio_tools::I2SDriverNanoBLE::setupPins(audio_tools::I2SConfig)
[I] SoundGenerator.h : 173 - SineWaveGenerator::begin(channels=2, sample_rate=44100, frequency=493.88)
[I] SoundGenerator.h : 159 - bool audio_tools::SineWaveGenerator<T>::begin() [with T = short int]
[D] SoundGenerator.h : 43 - bool audio_tools::SoundGenerator<T>::begin() [with T = short int]
[I] SoundGenerator.h : 46 - bits_per_sample: 16
[I] SoundGenerator.h : 47 - channels: 2
[I] SoundGenerator.h : 48 - sample_rate: 44100
[I] Buffers.h : 321 - resize: 4
[I] SoundGenerator.h : 200 - setFrequency: 493.88
[I] SoundGenerator.h : 201 - active: true
started...
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] AudioStreams.h : 661 - GeneratedSoundStream::readBytes: 512
[D] SoundGenerator.h : 71 - readBytes: 512
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] AudioStreams.h : 661 - GeneratedSoundStream::readBytes: 512
[D] SoundGenerator.h : 71 - readBytes: 512
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] AudioStreams.h : 661 - GeneratedSoundStream::readBytes: 512
[D] SoundGenerator.h : 71 - readBytes: 512
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] AudioStreams.h : 661 - GeneratedSoundStream::readBytes: 512
[D] SoundGenerator.h : 71 - readBytes: 512
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] AudioStreams.h : 661 - GeneratedSoundStream::readBytes: 512
[D] SoundGenerator.h : 71 - readBytes: 512
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] AudioStreams.h : 661 - GeneratedSoundStream::readBytes: 512
[D] SoundGenerator.h : 71 - readBytes: 512
[D] I2SStream.h : 92 - I2SStream::write: 512
[D] Buffers.h : 63 - writeArray 512 -> 512
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[D] StreamCopy.h : 141 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 149 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 155 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] StreamCopy.h : 97 - size_t audio_tools::StreamCopyT<T>::copy() [with T = unsigned char; size_t = unsigned int]
[D] AudioStreams.h : 661 - GeneratedSoundStream::readBytes: 1024
[D] SoundGenerator.h : 71 - readBytes: 1024
[D] I2SStream.h : 92 - I2SStream::write: 1024
[D] Buffers.h : 63 - writeArray 1024 -> 0
[D] I2SNanoSenseBLE.h : 330 - void audio_tools::I2SDriverNanoBLE::startI2SActive()

[BobaTrek]

I am using the "XIAO nRF52840 BLE Sense". It is identical to the "Nano 33 BLE Sense" except it is smaller and exposes fewer pins.

I am not aware of a variant package for the XIAO nRF52840 BLE Sense. Nordic Q&A recommended either making one myself, or just use the Pin to Port mapping formula.

[BobaTrek]

For now, I removed the digitalPinToPinName() call and just use the values that I inserted into the config object. I do get sound now! But the processor is still locking up right at the startI2SActive() call.

Still digging, but it is good to know that the I2S has actually been running, just not the pins I intended.

[BobaTrek]

I found one minor bug, but not the cause of the lockup.

The lines:

case I2S_PHILIPS_FORMAT:
  NRF_I2S->CONFIG.ALIGN = I2S_CONFIG_FORMAT_FORMAT_I2S << I2S_CONFIG_FORMAT_FORMAT_Pos;

should be:

case I2S_PHILIPS_FORMAT:
  NRF_I2S->CONFIG.FORMAT = I2S_CONFIG_FORMAT_FORMAT_I2S << I2S_CONFIG_FORMAT_FORMAT_Pos;
  NRF_I2S->CONFIG.ALIGN = I2S_CONFIG_ALIGN_ALIGN_LEFT << I2S_CONFIG_ALIGN_ALIGN_Pos;

[BobaTrek]

I wrote something to just trace the NVIC and the NRF_I2S calls, and here is the output. It looks pretty much the same as the simple example:

starting I2S...
[D] I2SNanoSenseBLE.h : 139 - NVIC_SetVector(I2S_IRQn, (uint32_t)I2S_IRQHandler)
[D] I2SNanoSenseBLE.h : 210 - NRF_I2S->CONFIG.TXEN = (I2S_CONFIG_TXEN_TXEN_Enabled << I2S_CONFIG_TXEN_TXEN_Pos)
[D] I2SNanoSenseBLE.h : 225 - NRF_I2S->CONFIG.MCKEN = (I2S_CONFIG_MCKEN_MCKEN_Enabled << I2S_CONFIG_MCKEN_MCKEN_Pos)
[D] I2SNanoSenseBLE.h : 241 - frequency requested 44100.000000 vs 83333.335938
[D] I2SNanoSenseBLE.h : 241 - frequency requested 44100.000000 vs 62500.000000
[D] I2SNanoSenseBLE.h : 241 - frequency requested 44100.000000 vs 41666.667969
[D] I2SNanoSenseBLE.h : 241 - frequency requested 44100.000000 vs 45454.546875
[D] I2SNanoSenseBLE.h : 241 - frequency requested 44100.000000 vs 44444.449219
[I] I2SNanoSenseBLE.h : 245 - Frequency req. 44100.000000 vs eff. 44444.449219
[D] I2SNanoSenseBLE.h : 258 - NRF_I2S->CONFIG.SWIDTH = (I2S_CONFIG_SWIDTH_SWIDTH_16Bit << I2S_CONFIG_SWIDTH_SWIDTH_Pos)
[D] I2SNanoSenseBLE.h : 277 - NRF_I2S->CONFIG.FORMAT = I2S_CONFIG_FORMAT_FORMAT_I2S << I2S_CONFIG_FORMAT_FORMAT_Pos
[D] I2SNanoSenseBLE.h : 279 - NRF_I2S->CONFIG.ALIGN = I2S_CONFIG_ALIGN_ALIGN_Left << I2S_CONFIG_ALIGN_ALIGN_Pos
[D] I2SNanoSenseBLE.h : 324 - NRF_I2S->PSEL.MCK = (cfg.pin_mck) << I2S_PSEL_MCK_PIN_Pos
[D] I2SNanoSenseBLE.h : 328 - NRF_I2S->PSEL.SCK = (cfg.pin_bck) << I2S_PSEL_SCK_PIN_Pos
[D] I2SNanoSenseBLE.h : 331 - NRF_I2S->PSEL.LRCK = (cfg.pin_ws) << I2S_PSEL_LRCK_PIN_Pos
[D] I2SNanoSenseBLE.h : 335 - NRF_I2S->PSEL.SDOUT = (cfg.pin_data) << I2S_PSEL_SDOUT_PIN_Pos

[D] I2SNanoSenseBLE.h : 368 - void audio_tools::I2SDriverNanoBLE::startI2SActive()
[D] I2SNanoSenseBLE.h : 370 - NRF_I2S->CONFIG.CHANNELS = I2S_CONFIG_CHANNELS_CHANNELS_Stereo << I2S_CONFIG_CHANNELS_CHANNELS_Pos
[D] I2SNanoSenseBLE.h : 373 - NRF_I2S->CONFIG.MODE = cfg.is_master (=1) ? I2S_CONFIG_MODE_MODE_MASTER << I2S_CONFIG_MODE_MODE_Pos : I2S_CONFIG_MODE_MODE_Slave << I2S_CONFIG_MODE_MODE_Pos
[D] I2SNanoSenseBLE.h : 377 - NRF_I2S->TXD.PTR = (uint32_t) p_i2s_array (= 536942392)
[D] I2SNanoSenseBLE.h : 379 - NRF_I2S->RXD.PTR = (uint32_t) p_i2s_array (= 536942392)
[D] I2SNanoSenseBLE.h : 382 - NRF_I2S->RXTXD.MAXCNT = i2s_buffer_size / frame_size() (= 256)
[D] I2SNanoSenseBLE.h : 385 - NRF_I2S->INTENSET = getINTENSET() (=32)
[D] I2SNanoSenseBLE.h : 388 - NVIC_EnableIRQ(I2S_IRQn)
[D] I2SNanoSenseBLE.h : 392 - NRF_I2S->ENABLE = 1
[D] I2SNanoSenseBLE.h : 395 - NRF_I2S->TASKS_START = 1

[pschatzmann]

Strange: For me it is not hanging at all!

My gut feeling is that this might be related with your pins or an issue with your specific processor.

This was also the reason, why it did not work for me when I tried first, because I did not convert the Arduino Pin numbers.
I had to update the default pins, because pin 0 and pin 1 were not available on my board: The generated i2s signals in the Oscilloscope are looking good to me.

To support your pin convention, I extended the logic a bit so that you can do the following:

    config.is_arduino_pin_numbers = false;
    config.pin_ws = 44;  // Pin D7 on XIAO nRF52840
    config.pin_data = 46;  // Pin D9 on XIAO nRF52840
    config.pin_bck = 45;  // Pin D8 on XIAO nRF52840

With the standard sine test, I am getting:

08:13:07.340 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 3 hops
08:13:07.340 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.342 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.342 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.342 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.342 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.379 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.389 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.389 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.389 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.389 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.389 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.389 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.389 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.389 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.435 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.439 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.439 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.439 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.439 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.439 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.439 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.439 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.439 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.481 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.483 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.483 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.483 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.483 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.483 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.483 -> [I] StreamCopy.h : 305 - try write - 2 (open 512 bytes) 
08:13:07.483 -> [I] StreamCopy.h : 305 - try write - 3 (open 0 bytes) 
08:13:07.531 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 3 hops
08:13:07.531 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
08:13:07.531 -> [I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops

etc

[BobaTrek]

I have gotten it to work. I had to make the following changes to the I2S_IRQHandler():

void I2S_IRQHandler(void) {
  int eventTxPtrUpd = NRF_I2S->EVENTS_TXPTRUPD;
  int eventRxPtrUpd = NRF_I2S->EVENTS_RXPTRUPD;
  
  // No logic can allow these to not be cleared, otherwise you get locked up
  if( eventTxPtrUpd != 0 )
      NRF_I2S->EVENTS_TXPTRUPD = 0;
  
  if( eventRxPtrUpd != 0 )
      NRF_I2S->EVENTS_RXPTRUPD = 0;
  
  
    // to validate if IRQ is called
  irq_count++;

  // prevent NPE
  if (p_i2s_buffer==nullptr || p_i2s_array==0)  return;

  //Handle Wrtie
  if(eventTxPtrUpd != 0) {
    memset(p_i2s_array,0, i2s_buffer_size);
    // fill will audio data, if available
    p_i2s_buffer->readArray(p_i2s_array, i2s_buffer_size);
  }  

  //Handle Read
  if(eventRxPtrUpd != 0) {
    // reading from pins writing to buffer - overwrite oldest data on overflow
    p_i2s_buffer->writeArrayOverwrite(p_i2s_array, i2s_buffer_size);
  }

} 

My hangups were apparently any time the I2S_IRQHandler() did get called, but did not get cleared, i.e. by calling NRF_I2S->EVENTS_TXPTRUPD = 0.

A lot of effort was wasted when I commented out all of the I2S_IRQHandler() (including the NRF_I2S->EVENTS_TXPTRUPD = 0) and nothing worked. It led me very astray.

So I am not hanging up any more for that reason. And my I2S pins seem to work fine.

I am still having problems though. I think it may be related to trying to do the readArray() call within the I2S_IRQHandler(). I think it may be taking too long. The intent of the chipset was to only update the TXD.PTR in the interrupt routine, and then return. So I hear bursts of the audio, and a very consistent chopping noise.

The simple version of code I am using has a large array of memory with Wav data in it. I set RXTXD.MAXCNT to 1024. Then every EVENTS_TXPTRUPD, I set TXD.PTR to the start of the array + n*1024 where n is the number of EVENTS_TXPTRUPD I have received. This is performing flawlessly, but lacks all of the features of audiotools!

[BobaTrek]

Here is what my standard sine testing run looks like:
(Note: I commented out the LOGI at StreamCopy.h line 305 because they were flooding the console)

starting I2S...
[I] I2SNanoSenseBLE.h : 128 - bool audio_tools::I2SDriverNanoBLE::begin(audio_tools::I2SConfig)
[I] I2SNanoSenseBLE.h : 248 - Frequency req. 44100.000000 vs eff. 44444.449219
[I] SoundGenerator.h : 173 - SineWaveGenerator::begin(channels=2, sample_rate=44100, frequency=493.88)
[I] SoundGenerator.h : 159 - bool audio_tools::SineWaveGenerator<T>::begin() [with T = short int]
[I] SoundGenerator.h : 46 - bits_per_sample: 16
[I] SoundGenerator.h : 47 - channels: 2
[I] SoundGenerator.h : 48 - sample_rate: 44100
[I] Buffers.h : 321 - resize: 4
[I] SoundGenerator.h : 200 - setFrequency: 493.88
[I] SoundGenerator.h : 201 - active: true
started...
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 5 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 3 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 3 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 5 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 3 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 3 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 512 -> 512 -> 512 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 4 hops

The tone coming out the speaker is not right though. It is chopped up as if it is playing a short segment of the sinewave over and over.

[BobaTrek]

I tried increasing the buffer_size to 16k:

  config.buffer_size = 16*1024;

Two things happen:

1. I can hear the tone nice and clear, but there is a tick about 4 times per second.

2. The number of hops is always 1.

starting I2S...
[I] I2SNanoSenseBLE.h : 128 - bool audio_tools::I2SDriverNanoBLE::begin(audio_tools::I2SConfig)
[I] I2SNanoSenseBLE.h : 248 - Frequency req. 44100.000000 vs eff. 44444.449219
[I] SoundGenerator.h : 173 - SineWaveGenerator::begin(channels=2, sample_rate=44100, frequency=493.88)
[I] SoundGenerator.h : 159 - bool audio_tools::SineWaveGenerator<T>::begin() [with T = short int]
[I] SoundGenerator.h : 46 - bits_per_sample: 16
[I] SoundGenerator.h : 47 - channels: 2
[I] SoundGenerator.h : 48 - sample_rate: 44100
[I] Buffers.h : 321 - resize: 4
[I] SoundGenerator.h : 200 - setFrequency: 493.88
[I] SoundGenerator.h : 201 - active: true
started...
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops
[I] StreamCopy.h : 139 - StreamCopy::copy 1024 -> 1024 -> 1024 bytes - in 1 hops

I think the tick noise occurs after every buffer_size has transferred. So with a buffer_size of 16K, I am hearing it 4'ish times per second. When the buffer_size is 1024, it is like 44 times per second and makes the "chopping" noise I was describing.

[pschatzmann]

Did you deactivate the Info log ? This might avoid any sound quality issues.

You could also try to define a fixed i2s_buffer_size instead of updating it from the config.buffer_size. I suggest to try with smaller values then 512 to keep the time low

[BobaTrek]

I did deactivate ALL logging, but the tick is exactly the same with or without logs.

[BobaTrek]

There is a 1:1 correlation between the buffer_size and the ticking rate.

[pschatzmann]

It would also be interresting to see if lower sample rates work better.

[pschatzmann]

Not sure if your logic is sound: if you clear EVENTS_TXPTRUPD and update the memory later, you risk that this is done too late. This will lead to an audible issue...

[BobaTrek]

Three experiments:

Leaving the buffer_size alone:

1. Going from MONO to STEREO doubles the tick rate x2.
2. Going from 44100 to 22000 sample rate cuts the tick rate in half.

Leaving all else alone:
3) Doubling the buffer_size cuts the tick rate in half.

[BobaTrek]

Good point, so I changed it to:
(pardon the indentation, the code wrapper on Github does not like something about my indents)

void I2S_IRQHandler(void) {
	int eventTxPtrUpd = NRF_I2S->EVENTS_TXPTRUPD;
	int eventRxPtrUpd = NRF_I2S->EVENTS_RXPTRUPD;
	
    // to validate if IRQ is called
	irq_count++;
	
    // prevent NPE
    if (p_i2s_buffer==nullptr || p_i2s_array==0)
	{
	  if( eventTxPtrUpd != 0 )
	    NRF_I2S->EVENTS_TXPTRUPD = 0;

	  if( eventRxPtrUpd != 0 )
        NRF_I2S->EVENTS_RXPTRUPD = 0;
	  
	  return;
	}

    //Handle Wrtie
    if(eventTxPtrUpd != 0) {
      memset(p_i2s_array,0, i2s_buffer_size);
      // fill will audio data, if available
      p_i2s_buffer->readArray(p_i2s_array, i2s_buffer_size);
      NRF_I2S->EVENTS_TXPTRUPD = 0;
	}  

    //Handle Read
    if(eventRxPtrUpd != 0) {
      // reading from pins writing to buffer - overwrite oldest data on overflow
      p_i2s_buffer->writeArrayOverwrite(p_i2s_array, i2s_buffer_size);
      NRF_I2S->EVENTS_RXPTRUPD = 0;
    }
} 

and it is behaving exactly the same with the audible ticks.

[pschatzmann]

Did you try this: You could also try to define a fixed i2s_buffer_size instead of updating it from the config.buffer_size. I suggest to try with smaller values then 512 to keep the time low (e.g. with 64).
To double check if the issue is coming from the timing you could also comment out the memset to 0

[BobaTrek]

Commenting out memset made the ticks go from loud chops to softer taps.

[BobaTrek]

As a hack, I thought I would try this:

    if(eventTxPtrUpd != 0) {
      //memset(p_i2s_array,0, i2s_buffer_size);
      // fill will audio data, if available
      if(irq_count == 1)
	    p_i2s_buffer->readArray(p_i2s_array, i2s_buffer_size);
      NRF_I2S->EVENTS_TXPTRUPD = 0;
    }  

That way the sinewave only gets copied into the buffer on the first interrupt, and then will just be there for all the remaining interrupts. The tick is still there! I realize there will be a disjoint when the signal wraps around, but I would have thought that would be rather low in volume.

[BobaTrek]

And now I set the sinewave to 1Hz. I don't hear the sinewave at all anymore, but I still hear the tick.

[BobaTrek]

I used the following so I could "hear" every time the buffer_size passed and correlate it with the silenced section from the memset(0...)

    //Handle Wrtie
    if(eventTxPtrUpd != 0) {
      if(irq_count%2 == 0)
        memset(p_i2s_array,0, i2s_buffer_size);
      // fill will audio data, if available
      if(irq_count%2 == 1)
	 p_i2s_buffer->readArray(p_i2s_array, i2s_buffer_size);
      NRF_I2S->EVENTS_TXPTRUPD = 0;
    }  

It confirms that the tick comes each and every time I2S_IRQHandler() is called.

[BobaTrek]

I have been up all night working on this and I am bushed. I need to take a break and get some sleep. It is 5:30am!

[BobaTrek]

I tried the following experiment.

I made my own non-streaming source of a sinewave int16 array. I then changed the I2S_IRQHandler() to copy from this array instead, and the ticks are completely gone. Just pure tone.

//Handle Write
if(eventTxPtrUpd != 0) {
   // fill with audio data, if available
   for(int i=0; i<i2s_buffer_size/2; i++)
   {
     ((int16_t *)p_i2s_array)[i] = sinewave[i];
   }
   NRF_I2S->EVENTS_TXPTRUPD = 0;
}  

I am wondering if the problem might be upstream of the I2S buffer?

[pschatzmann]

I was on a similar quest: I extended the I2SDriverNanoBLE to support the data directly from a Stream. I also added 2 alternative IRQ implementations: one with readArray and one using directly the addresses and provided different buffer implementations.

I am confused as each alternative implementation behaves totally different.

Your version should correspond to the following:

#include "AudioTools.h"

AudioInfo info(8000, 2, 16);
SineWaveGenerator<int16_t> sineWave(32000);                // subclass of SoundGenerator with max amplitude of 32000
GeneratedSoundStream<int16_t> sound(sineWave);             // Stream generated from sine wave
I2SDriverNanoBLE out;

void setup() {
  // Open Serial 
  Serial.begin(115200);
  while(!Serial);
  AudioLogger::instance().begin(Serial, AudioLogger::Warning);

  // Setup sine wave
  sineWave.begin(info, N_B4);

  // define direct input from stream
  out.setStream(sound);

  // start I2S
  Serial.println("starting I2S...");
  auto config = out.defaultConfig(TX_MODE);
  config.copyFrom(info); 
  config.buffer_size = 64; // defines the MAXCNT in bytes
  config.buffer_count = 1; // not relevant
  out.begin(config);

  Serial.println("started...");
}

void loop() {
  //delay(1000);
}

It is just replacing your fixed array with the input from sine stream.
I have the feeling that the issue is related to the DMA timings in the IRQ since changing the buffer_size and sample rate is giving different effects

[BobaTrek]

OK, I downloaded all the latest and am using the test code above. The only thing I added is:

  config.is_arduino_pin_numbers = false;
  config.pin_mck = 0;
  config.pin_ws = 44;  // Pin D7 on XIAO nRF52840
  config.pin_data = 46;  // Pin D9 on XIAO nRF52840
  config.pin_bck = 45;  // Pin D8 on XIAO nRF52840

I am still getting ticks, but at a rate commensurate with the new DMA settings.

The tick sound and rate seems the same regardless of setting USE_FAST_READ or USE_RINGBUFFER true.

If I set the buffer_size to 8K, the ticks are a couple times per second.
If I set the buffer_size smaller, the tick rate increases.
If I set the buffer_size down to 64, the tick rate is so fast, it is like having a second tone.

[pschatzmann]

Right, USE_FAST_READ or USE_RINGBUFFER is only relevant for the standard scenario (with the copy in the sketch) and has no impact on the above code. The ticking seems to come from the (small) delay that is added to provide the data.

In the standard copy sceanrio, USE_FAST_READ should (in theory) add only a minimal delay because we just need to provide the available address with the already available data in the IRQ.

I think the next step is to start to measure the time each step takes and how much time we have available...

[pschatzmann]

I think, I found it: It seems that if the buffer is just updated but the address not changed in the IRQ, bad things happen!
With my latest correction the noise has gone...

[BobaTrek]

I downloaded latest and tried with USE_FAST_READ set to both false and to true, and I am still getting the ticking sound at the buffer_size rate. I noticed NRF_I2S->TXD.PTR only gets re-written if USE_FAST_READ is true?

I tried various buffer_sizes between 64 and 8192.

[pschatzmann]

You must have done something wrong: USE_FAST_READ does not exist any more...

The magical correction is

      p_i2s_array = p_i2s_array==p_i2s_array_1?p_i2s_array_2:p_i2s_array_1;

I left the alternative setStream() API because I think it is a kind of cool that you can e.g. give a pcm file as argument that will play itself automatically.

[BobaTrek]

Wow - that is magic! I do like the setStream() API option.

I downloaded latest again, compiled, and the sinewave test is working without any ticks!

Two questions though:

1. If I set config.buffer_size to 64, 128, or 256 the sinewave plays nice and steady. If I set config.buffer_size to 512 or greater, I do not hear anything and the "started..." message never prints, so I think the processor is locking up.

2. The sound is great unless I do any serial prints. Then I hear severe distortion. (This is why I tried increasing buffer size). Any thoughts?

[pschatzmann]

Can it be that your processor has less RAM then mine ?
My first guess would be that you ask form more RAM that is available.
Did you try to reduce the buffer_count ?

You could actally optimize the memory usage tremendously:

  if (p_i2s_buffer==nullptr && p_nano_ble_stream==nullptr){
    p_i2s_buffer = new NBuffer<uint8_t>(cfg.buffer_size, i2s_buffer_size);
  }

The p_i2s_buffer is not needed when you use the setStream option....

I guess you would reduce the risk of distortions by using a lower sampling rate. At least that's what's happening on other processors. My regular API should also protect you from that because a delay from a Serial print would usually not have any impact if the buffer is large enough and can provide audio data during this delay

[BobaTrek]

I tried setStream() and it also works, albeit with some random noise at the end of the sound clip. I'll dig into that a little later.

I did add the following code to the I2SStream class in I2SStream.h to allow me a mechanism to call setStream(). Was there another more direct method?

void setStream(Stream &stream){
	i2s.setStream(stream);
}

[pschatzmann]

This is a bad idea, since this will lead to compile errors im my framework on other platforms. In my latest commit, I have added a driver() method to give access to the underlying driver: so i2s.driver()->setStream(stream), should work, but I am not convinced if this is a good idea.

Please note that this also gives a compile error, but it is in your sketch if you try to compile it on any other platfroms. So to highlight that this is only working on the "nano ble", it might be recommended to use the driver class directly, as I did in my example and not mix the approaches.