Audio output with systematic sample loss

[lutzray]

After calibrating the RP2040 onboard 1 MHz clock used by systick, I can precisely count (±5 ppm) how many samples have been produced by readCB()

#include "Adafruit_TinyUSB.h"
#include "hardware/structs/systick.h"

Adafruit_USBD_Audio usb;
volatile size_t sample_count = 0;

size_t readCB(uint8_t* data, size_t buff_len, Adafruit_USBD_Audio& ref) {
  int16_t* data16 = (int16_t*)data;
  size_t samples = buff_len / sizeof(int16_t);
  size_t result = 0;

  for (int j = 0; j < samples; j++) {
    sample_count += 1;
    // period should be 200 samples, it is less (196)
    data16[j] = 32000 * sin(6.2831853 * sample_count / 200);
    result += sizeof(int16_t);
  }
  return result;
}

void setup() {
  // Manual begin() is required on core without built-in support e.g. mbed rp2040
  if (!TinyUSBDevice.isInitialized()) {
    TinyUSBDevice.begin(0);
  }
  systick_hw->csr = 0;        //Disable
  systick_hw->rvr = 1000002;  // calibrated with GNSS = 1.000000 second period
  systick_hw->cvr = 0;        //clear the count to force initial reload
  systick_hw->csr = 0x1;      //Enable, with External clock at 1 MHz

  Serial.begin(115200);
  while (!Serial)
    ;  // wait for serial

  // Start USB device as Audio Source
  usb.setReadCallback(readCB);
  usb.begin(44100, 1, 16);

  if (TinyUSBDevice.mounted()) {
    TinyUSBDevice.detach();
    delay(10);
    TinyUSBDevice.attach();
  }
}

void loop() {
  if ((systick_hw->csr >> 16) & 1) {
    // print sample_count each 1.000000 second
    // should be 44100 but is 46000, systematically
    Serial.println(sample_count);
    sample_count = 0;
  }
}

typical Arduino Serial print out:

14:31:42.116 -> 46000
14:31:43.129 -> 46000
14:31:44.135 -> 45954
14:31:45.115 -> 45954
14:31:46.121 -> 46000

And a manual count of samples in the recorded wav shows coherent discontinuities in the produced sin wave and a shorter period.

Host computer:

Model Identifier:	Macmini7,1
Processor Name:	Dual-Core Intel Core i7
System Version:	macOS 12.7.6 (21H1320)

[pratikrath126]

Sample loss usually means the readCB isn't draining the buffer fast enough or the host's audio pull rate differs. Try increasing the TinyUSB audio buffer size (AUDIO_BUFFER_SAMPLES) or optimizing the callback to copy samples faster. Also verify sample format (mono/stereo, bits) matches both sides. If the host clock drifts slightly, enabling feedback endpoint (freq adjustment) can help. Check that your SYSTICK calibration isn't interfering with interrupt timing. Good luck!

[lutzray]

Also verify sample format (mono/stereo, bits) matches both sides

They do:

        TinyUSB Audio:

          Default Input Device: Yes
          Input Channels: 1
          Manufacturer: Seeed
          Current SampleRate: 44100
          Transport: USB
          Input Source: Default

Check that your SYSTICK calibration isn't interfering with interrupt timing

This is done with another program, before running the test.

optimizing the callback to copy samples faster

Hmm, readCB() can't be any simpler...

Try increasing the TinyUSB audio buffer size (AUDIO_BUFFER_SAMPLES)

Where do I do this?

thanks for your suggestions!

[lutzray]

audio USB is downpitched 2%

I'm following up on my own post. Here's another way to state the problem. To reproduce, you need to apply a 1 PPS from any GPS board on pin D10

// code illustrating time error with audio USB from pschatzmann's
// Adafruit_TinyUSB_Arduino fork, on a XIAO RP2040 with 1 PPS
// from GPS on pin D10
#include "Adafruit_TinyUSB.h"
Adafruit_USBD_Audio usb;
volatile bool PPSarrived = false;

size_t readCB(uint8_t* data, size_t buff_len, Adafruit_USBD_Audio& ref) {
  int16_t* data16 = (int16_t*)data;
  size_t samples = buff_len / sizeof(int16_t);
  size_t result = 0;
  for (int j = 0; j < samples; j++) {
    if (PPSarrived) {
      PPSarrived = false;
      data16[j] = 32700;
    } else {
      data16[j] = 0;
    }
    result += sizeof(int16_t);
  }
  return result;
}
void PPS_isr(void) {
  PPSarrived = true;
}
void setup() {
  if (!TinyUSBDevice.isInitialized()) {
    TinyUSBDevice.begin(0);
  }
  // GPS 1 PPS pulse on pin D10
  attachInterrupt(D10, PPS_isr, RISING);
  Serial.begin(115200);
  while (!Serial) ;  // wait for serial
  usb.setReadCallback(readCB);
  usb.begin(44100, 1, 16);
  if (TinyUSBDevice.mounted()) {
    TinyUSBDevice.detach();
    delay(10);
    TinyUSBDevice.attach();
  }
}
void loop() {
}

The interval between those spike should be 10.00 s, not 10.210 (yes, the framerate of the captured wav is coherent: 44100)

I want to use a RP2040 for precise timing but this won't cut it. I did code a working solution on a Teensy4.0 but it is costly...