Ethernet Audio Streamer

#include <Arduino.h>
#include <SPI.h>
#include <Ethernet.h>
#include <EthernetUdp.h>
#include <driver/i2s.h>


// Forward declarations
void setupI2S();
void setupEthernet();
void playRtpPacket(int packetSize);

constexpr int ETH_CS_PIN = 4;
constexpr int ETH_RST_PIN = 13;
constexpr int ETH_SCK_PIN = 18;
constexpr int ETH_MISO_PIN = 19;
constexpr int ETH_MOSI_PIN = 23;
constexpr int I2S_BCK_PIN = 27;
constexpr int I2S_LRCK_PIN = 26;
constexpr int I2S_DOUT_PIN = 25;
constexpr uint16_t RTP_PORT = 5004;
constexpr size_t MAX_UDP_PACKET = 1472;

byte macAddress[] = {0x02, 0xA0, 0x10, 0x00, 0x00, 0x50};
IPAddress localIp(192, 168, 1, 50);
EthernetUDP rtpUdp;
uint8_t packetBuffer[MAX_UDP_PACKET];
int16_t audioBuffer[MAX_UDP_PACKET / 2];

void setupI2S() {
  i2s_config_t config = {};
  config.mode = (i2s_mode_t)(I2S_MODE_MASTER | I2S_MODE_TX);
  config.sample_rate = 48000;
  config.bits_per_sample = I2S_BITS_PER_SAMPLE_16BIT;
  config.channel_format = I2S_CHANNEL_FMT_RIGHT_LEFT;
  config.communication_format = I2S_COMM_FORMAT_STAND_I2S;
  config.intr_alloc_flags = ESP_INTR_FLAG_LEVEL1;
  config.dma_buf_count = 8;
  config.dma_buf_len = 256;
  config.use_apll = false;
  config.tx_desc_auto_clear = true;
  config.fixed_mclk = 0;
  i2s_driver_install(I2S_NUM_0, &config, 0, nullptr);

  i2s_pin_config_t pins = {};
  pins.bck_io_num = I2S_BCK_PIN;
  pins.ws_io_num = I2S_LRCK_PIN;
  pins.data_out_num = I2S_DOUT_PIN;
  pins.data_in_num = I2S_PIN_NO_CHANGE;
  i2s_set_pin(I2S_NUM_0, &pins);
  i2s_zero_dma_buffer(I2S_NUM_0);
}

void setupEthernet() {
  pinMode(ETH_RST_PIN, OUTPUT);
  digitalWrite(ETH_RST_PIN, LOW);
  delay(20);
  digitalWrite(ETH_RST_PIN, HIGH);
  delay(100);

  SPI.begin(ETH_SCK_PIN, ETH_MISO_PIN, ETH_MOSI_PIN, ETH_CS_PIN);
  Ethernet.init(ETH_CS_PIN);
  Ethernet.begin(macAddress, localIp);
  rtpUdp.begin(RTP_PORT);
}

void playRtpPacket(int packetSize) {
  if (packetSize < 12 || packetSize > (int)sizeof(packetBuffer)) return;
  int received = rtpUdp.read(packetBuffer, packetSize);
  if (received != packetSize) return;

  const uint8_t version = packetBuffer[0] >> 6;
  const bool hasExtension = (packetBuffer[0] & 0x10) != 0;
  const uint8_t csrcCount = packetBuffer[0] & 0x0F;
  if (version != 2) return;

  size_t offset = 12 + (size_t)csrcCount * 4;
  if (offset > (size_t)received) return;
  if (hasExtension) {
    if (offset + 4 > (size_t)received) return;
    const uint16_t extensionWords = ((uint16_t)packetBuffer[offset + 2] << 8) | packetBuffer[offset + 3];
    offset += 4 + (size_t)extensionWords * 4;
  }
  if (offset >= (size_t)received) return;

  size_t payloadBytes = (size_t)received - offset;
  payloadBytes &= ~((size_t)3); // Whole 16-bit left/right sample pairs only.
  if (payloadBytes == 0 || payloadBytes / 2 > sizeof(audioBuffer) / sizeof(audioBuffer[0])) return;

  for (size_t i = 0; i < payloadBytes / 2; ++i) {
    audioBuffer[i] = (int16_t)(((uint16_t)packetBuffer[offset + i * 2] << 8) |
                               packetBuffer[offset + i * 2 + 1]);
  }

  size_t written = 0;
  i2s_write(I2S_NUM_0, audioBuffer, payloadBytes, &written, portMAX_DELAY);
}

void setup() {
  Serial.begin(115200);
  setupI2S();
  setupEthernet();
  Serial.println("AoIP receiver ready: RTP/L16 stereo, 48 kHz, UDP port 5004, IP 192.168.1.50");
}

void loop() {
  int packetSize = rtpUdp.parsePacket();
  if (packetSize > 0) playRtpPacket(packetSize);
}