Possible Bug: Using ADCInput and PWMAudio simultanously...

[Locutus74]

Hello!

I have tried to use ADCInput and PWMAudio simultaneously. The result was a little bit unexpected. It works, but the ADCInput Callback delivers by far too few samples for the requested sample rate. Nice unintended chipmunk-echo-effect, though...

This might be a bug or me using some component in a way it was not intended to be used (--> /me doing something stupid).

Albeit I did not expect both sample clocks to be absolutely synchronous, I however did expected that both callbacks would at least approximately run at the same speed.

// Minimalistic full-duplex audio with the Pi pico
// just copies samples from ADC-0 to PWM on GPIO-0
//
// What it should do: Copy samples from ADC-0 to GPIO-0
// "unaltered" if you disregard the x-runs happening from
// time to time.
//
// What it does: Copy samples from ADC-0 to GPIO-0. Due
// to a massive speed difference in input and output sample-
// rates (more than 2x) lots of x-runs happen and a "nice"
// chipmunk-echo-effect is produced.

#include <PWMAudio.h>
#include <ADCInput.h>

// PWM-Rate is audible but this is not important for this test
#define SAMPLERATE        12000
#define NR_OF_PWM_BUFFERS 8
#define PWM_BUFFER_SIZE   32

ADCInput g_ADCInput1( A0 );
PWMAudio g_PWMAudio1( 0 );

void ADCCallback(void);
void DACCallback(void);

volatile int16_t g_SampleBuffer[16384] = {0};
volatile uint16_t g_wPtr = 0;
volatile uint16_t g_rPtr = 0;

void setup()
{
  // PSU-Mode 
  //
  // High == PWM (low noise but maybe a single "birdy"/"whistle" around 10-15kHz),
  // Low  == PFS (default, no "birdy"/"whistle" but loads of noise)
  pinMode( 23, OUTPUT );
  digitalWrite( 23, HIGH );

  // setup ADC
  g_ADCInput1.setFrequency( SAMPLERATE );

  // all these have been "in". Depending on the chosen samplerate and the load(?)
  // of the callbacks...
  // g_ADCInput1.setFrequency( SAMPLERATE*2 );
  // g_ADCInput1.setFrequency( SAMPLERATE*3 );
  // g_ADCInput1.setFrequency( SAMPLERATE*4 );
  // g_ADCInput1.setFrequency( SAMPLERATE*6 );
  // g_ADCInput1.setFrequency( SAMPLERATE*8 );

  g_ADCInput1.setBuffers( NR_OF_PWM_BUFFERS, PWM_BUFFER_SIZE );
  g_ADCInput1.onReceive( ADCCallback );

  // setup DAC
  g_PWMAudio1.setFrequency( SAMPLERATE );
  g_PWMAudio1.setBuffers( NR_OF_PWM_BUFFERS, PWM_BUFFER_SIZE );
  g_PWMAudio1.setStereo( false );
  g_PWMAudio1.onTransmit( DACCallback );

  // start ADC + DAC
  g_ADCInput1.begin();
  g_PWMAudio1.begin();
}

void loop()
{
  // nothing
}

void ADCCallback(void)
{
  // This one runs too slow, that is: It delivers approximately in between
  // half and one eighth of the samples per second than the DAC callback
  // needs.
  while(g_ADCInput1.available())
  {
    g_SampleBuffer[g_wPtr++] = (g_ADCInput1.read()-2048)*16;
    g_wPtr %= 16384;
  }
}

void DACCallback(void)
{
  // As far as I can tell: Runs normally...
  while( g_PWMAudio1.availableForWrite() )
  {
    g_PWMAudio1.write( g_SampleBuffer[g_rPtr++] );
    g_rPtr %= 16384;
  }
}

// PS: It is intentional, that the read/write pointers of the ring-buffer
//     can overtake each other. This is not what I usually would do, but
//     here it helps to find out the "speed-difference" in between ADC
//     and DAC by ear.

[earlephilhower]

I need to dig deeper to verify, but I don't think it's physically possible to have both running at the same time. The issue is that, unlike I2S audio, these two methods need a special, self-timed DMA. That timer is a shared resource and when you init it a second time the clocking for the first DMA is going to get weird.

[Locutus74]

If so, could this maybe be solved by some sort of a combined "ADC_PWM_audio_IO"-class? Which because of doing this in a synchronous way only needs one self-timed DMA for both? The need of audio-in/-out in a duplex fashion isn't that unusual I guess? (guitar pedals, general audio-FX/-filters; sample-input --> process_somehow --> sample-output)

[Geekachuqt]

Hi,

I'm encountering this issue when attempting to use ADCInput with I2S aswell. It works fine if I use a single pin for the ADCInput, but when I add more pins the sampling gets messed up.

Am I right in thinking that this simply won't work, and that I have to roll my own solution through, per example, interrupts?