Mirrors/overte
3
0
Fork 0
mirror of https://github.com/overte-org/overte.git synced 2025-04-21 09:44:21 +02:00
Code Issues Projects Releases Packages Wiki Activity Actions 2

Merge pull request #317 from birarda/audio-refactor

Browse source 
refactoring Audio in interface and audio-mixer
This commit is contained in:
ZappoMan 2013-05-15 12:40:05 -07:00
commit ab3ee22dc9
13 changed files with 262 additions and 441 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

28
audio-mixer/src/main.cpp
View file

@ -99,6 +99,11 @@ int main(int argc, const char* argv[]) {
int nextFrame = 0;
timeval startTime;
unsigned char clientPacket[BUFFER_LENGTH_BYTES + 1];
clientPacket[0] = PACKET_HEADER_MIXED_AUDIO;
int16_t clientSamples[BUFFER_LENGTH_SAMPLES_PER_CHANNEL * 2] = {};
gettimeofday(&startTime, NULL);
while (true) {
@ -129,8 +134,9 @@ int main(int argc, const char* argv[]) {
for (AgentList::iterator agent = agentList->begin(); agent != agentList->end(); agent++) {
AudioRingBuffer* agentRingBuffer = (AudioRingBuffer*) agent->getLinkedData();
int16_t clientMix[BUFFER_LENGTH_SAMPLES_PER_CHANNEL * 2] = {};
// zero out the client mix for this agent
memset(clientSamples, 0, sizeof(clientSamples));
for (AgentList::iterator otherAgent = agentList->begin(); otherAgent != agentList->end(); otherAgent++) {
if (otherAgent != agent || (otherAgent == agent && agentRingBuffer->shouldLoopbackForAgent())) {
@ -219,11 +225,11 @@ int main(int argc, const char* argv[]) {
}
int16_t* goodChannel = bearingRelativeAngleToSource > 0.0f
? clientMix + BUFFER_LENGTH_SAMPLES_PER_CHANNEL
: clientMix;
? clientSamples + BUFFER_LENGTH_SAMPLES_PER_CHANNEL
: clientSamples;
int16_t* delayedChannel = bearingRelativeAngleToSource > 0.0f
? clientMix
: clientMix + BUFFER_LENGTH_SAMPLES_PER_CHANNEL;
? clientSamples
: clientSamples + BUFFER_LENGTH_SAMPLES_PER_CHANNEL;
int16_t* delaySamplePointer = otherAgentBuffer->getNextOutput() == otherAgentBuffer->getBuffer()
? otherAgentBuffer->getBuffer() + RING_BUFFER_SAMPLES - numSamplesDelay
@ -249,7 +255,8 @@ int main(int argc, const char* argv[]) {
}
}
agentList->getAgentSocket().send(agent->getPublicSocket(), clientMix, BUFFER_LENGTH_BYTES);
memcpy(clientPacket + 1, clientSamples, sizeof(clientSamples));
agentList->getAgentSocket().send(agent->getPublicSocket(), clientPacket, BUFFER_LENGTH_BYTES + 1);
}
// push forward the next output pointers for any audio buffers we used
@ -268,9 +275,12 @@ int main(int argc, const char* argv[]) {
// pull any new audio data from agents off of the network stack
while (agentList->getAgentSocket().receive(agentAddress, packetData, &receivedBytes)) {
if (packetData[0] == PACKET_HEADER_INJECT_AUDIO) {
if (packetData[0] == PACKET_HEADER_INJECT_AUDIO || packetData[0] == PACKET_HEADER_MICROPHONE_AUDIO) {
char agentType = (packetData[0] == PACKET_HEADER_MICROPHONE_AUDIO)
? AGENT_TYPE_AVATAR
: AGENT_TYPE_AUDIO_INJECTOR;
if (agentList->addOrUpdateAgent(agentAddress, agentAddress, packetData[0], agentList->getLastAgentID())) {
if (agentList->addOrUpdateAgent(agentAddress, agentAddress, agentType, agentList->getLastAgentID())) {
agentList->increaseAgentID();
}

23
interface/src/Application.cpp
View file

@ -154,7 +154,7 @@ Application::Application(int& argc, char** argv) :
_oculusProgram(0),
_oculusDistortionScale(1.25),
#ifndef _WIN32
_audio(&_audioScope, &_myAvatar),
_audio(&_audioScope),
#endif
_stopNetworkReceiveThread(false),
_packetCount(0),
@ -202,10 +202,6 @@ Application::Application(int& argc, char** argv) :
// the callback for our instance of AgentList is attachNewHeadToAgent
AgentList::getInstance()->linkedDataCreateCallback = &attachNewHeadToAgent;
#ifndef _WIN32
AgentList::getInstance()->audioMixerSocketUpdate = &audioMixerUpdate;
#endif
#ifdef _WIN32
WSADATA WsaData;
int wsaresult = WSAStartup(MAKEWORD(2,2), &WsaData);
@ -913,11 +909,7 @@ void Application::terminate() {
// Close serial port
// close(serial_fd);
_myAvatar.writeAvatarDataToFile();
#ifndef _WIN32
_audio.terminate();
#endif
_myAvatar.writeAvatarDataToFile();
if (_enableNetworkThread) {
_stopNetworkReceiveThread = true;
@ -1278,7 +1270,7 @@ void Application::updateAvatar(float deltaTime) {
// Get audio loudness data from audio input device
#ifndef _WIN32
_myAvatar.setLoudness(_audio.getInputLoudness());
_myAvatar.setLoudness(_audio.getLastInputLoudness());
#endif
// Update Avatar with latest camera and view frustum data...
@ -1972,12 +1964,6 @@ void Application::attachNewHeadToAgent(Agent *newAgent) {
}
}
#ifndef _WIN32
void Application::audioMixerUpdate(in_addr_t newMixerAddress, in_port_t newMixerPort) {
static_cast<Application*>(QCoreApplication::instance())->_audio.updateMixerParams(newMixerAddress, newMixerPort);
}
#endif
// Receive packets from other agents/servers and decide what to do with them!
void* Application::networkReceive(void* args) {
sockaddr senderAddress;
@ -2009,6 +1995,9 @@ void* Application::networkReceive(void* args) {
printf("The rotation: %f, %f, %f\n", rotationRates[0], rotationRates[1], rotationRates[2]);
break;
case PACKET_HEADER_MIXED_AUDIO:
app->_audio.addReceivedAudioToBuffer(app->_incomingPacket, bytesReceived);
break;
case PACKET_HEADER_VOXEL_DATA:
case PACKET_HEADER_VOXEL_DATA_MONOCHROME:
case PACKET_HEADER_Z_COMMAND:

5
interface/src/Application.h
View file

@ -57,6 +57,8 @@ public:
void mouseReleaseEvent(QMouseEvent* event);
void wheelEvent(QWheelEvent* event);
const Avatar& getAvatar() const { return _myAvatar; }
private slots:
@ -119,9 +121,6 @@ private:
void setMenuShortcutsEnabled(bool enabled);
static void attachNewHeadToAgent(Agent *newAgent);
#ifndef _WIN32
static void audioMixerUpdate(in_addr_t newMixerAddress, in_port_t newMixerPort);
#endif
static void* networkReceive(void* args);
QMainWindow* _window;

439
interface/src/Audio.cpp
View file

@ -10,19 +10,21 @@
#include <iostream>
#include <fstream>
#include <pthread.h>
#include <sys/stat.h>
#include <cstring>
#include <StdDev.h>
#include <UDPSocket.h>
#include <SharedUtil.h>
#include <PacketHeaders.h>
#include <AgentList.h>
#include <AgentTypes.h>
#include "Application.h"
#include "Audio.h"
#include "Util.h"
#include "Log.h"
Oscilloscope * scope;
const int NUM_AUDIO_CHANNELS = 2;
const int PACKET_LENGTH_BYTES = 1024;
@ -55,15 +57,8 @@ const float AUDIO_CALLBACK_MSECS = (float)BUFFER_LENGTH_SAMPLES / (float)SAMPLE_
const int AGENT_LOOPBACK_MODIFIER = 307;
const char LOCALHOST_MIXER[] = "0.0.0.0";
const char WORKCLUB_MIXER[] = "192.168.1.19";
const char EC2_WEST_MIXER[] = "54.241.92.53";
const int AUDIO_UDP_LISTEN_PORT = 55444;
int starve_counter = 0;
int numStarves = 0;
StDev stdev;
bool stopAudioReceiveThread = false;
int samplesLeftForFlange = 0;
int lastYawMeasuredMaximum = 0;
@ -71,43 +66,31 @@ float flangeIntensity = 0;
float flangeRate = 0;
float flangeWeight = 0;
timeval firstPlaybackTimer;
int packetsReceivedThisPlayback = 0;
float usecsAtStartup = 0;
/**
* Audio callback used by portaudio.
* Communicates with Audio via a shared pointer to Audio::data.
* Writes input audio channels (if they exist) into Audio::data->buffer,
multiplied by Audio::data->inputGain.
* Then writes Audio::data->buffer into output audio channels, and clears
the portion of Audio::data->buffer that has been read from for reuse.
*
* @param[in] inputBuffer A pointer to an internal portaudio data buffer containing data read by portaudio.
* @param[out] outputBuffer A pointer to an internal portaudio data buffer to be read by the configured output device.
* @param[in] frames Number of frames that portaudio requests to be read/written.
(Valid size of input/output buffers = frames * number of channels (2) * sizeof data type (float)).
* @param[in] timeInfo Portaudio time info. Currently unused.
* @param[in] statusFlags Portaudio status flags. Currently unused.
* @param[in] userData Pointer to supplied user data (in this case, a pointer to Audio::data).
Used to communicate with external code (since portaudio calls this function from another thread).
* @return Should be of type PaStreamCallbackResult. Return paComplete to end the stream, or paContinue to continue (default).
Can be used to end the stream from within the callback.
*/
int audioCallback (const void *inputBuffer,
void *outputBuffer,
// inputBuffer A pointer to an internal portaudio data buffer containing data read by portaudio.
// outputBuffer A pointer to an internal portaudio data buffer to be read by the configured output device.
// frames Number of frames that portaudio requests to be read/written.
// timeInfo Portaudio time info. Currently unused.
// statusFlags Portaudio status flags. Currently unused.
// userData Pointer to supplied user data (in this case, a pointer to the parent Audio object
int audioCallback (const void* inputBuffer,
void* outputBuffer,
unsigned long frames,
const PaStreamCallbackTimeInfo *timeInfo,
PaStreamCallbackFlags statusFlags,
void *userData)
{
AudioData *data = (AudioData *) userData;
void* userData) {
Audio* parentAudio = (Audio*) userData;
AgentList* agentList = AgentList::getInstance();
Application* interface = (Application*) QCoreApplication::instance();
Avatar interfaceAvatar = interface->getAvatar();
int16_t *inputLeft = ((int16_t **) inputBuffer)[0];
// Add Procedural effects to input samples
data->addProceduralSounds(inputLeft, BUFFER_LENGTH_SAMPLES);
parentAudio->addProceduralSounds(inputLeft, BUFFER_LENGTH_SAMPLES);
if (inputLeft != NULL) {
@ -118,35 +101,32 @@ int audioCallback (const void *inputBuffer,
}
loudness /= BUFFER_LENGTH_SAMPLES;
data->lastInputLoudness = loudness;
parentAudio->_lastInputLoudness = loudness;
// add data to the scope
scope->addSamples(0, inputLeft, BUFFER_LENGTH_SAMPLES);
parentAudio->_scope->addSamples(0, inputLeft, BUFFER_LENGTH_SAMPLES);
if (data->mixerAddress != 0) {
sockaddr_in audioMixerSocket;
audioMixerSocket.sin_family = AF_INET;
audioMixerSocket.sin_addr.s_addr = data->mixerAddress;
audioMixerSocket.sin_port = data->mixerPort;
Agent* audioMixer = agentList->soloAgentOfType(AGENT_TYPE_AUDIO_MIXER);
if (audioMixer) {
int leadingBytes = 2 + (sizeof(float) * 4);
// we need the amount of bytes in the buffer + 1 for type
// + 12 for 3 floats for position + float for bearing + 1 attenuation byte
unsigned char dataPacket[BUFFER_LENGTH_BYTES + leadingBytes];
dataPacket[0] = PACKET_HEADER_INJECT_AUDIO;
dataPacket[0] = PACKET_HEADER_MICROPHONE_AUDIO;
unsigned char *currentPacketPtr = dataPacket + 1;
// memcpy the three float positions
memcpy(currentPacketPtr, &data->linkedAvatar->getHeadPosition(), sizeof(float) * 3);
memcpy(currentPacketPtr, &interfaceAvatar.getHeadPosition(), sizeof(float) * 3);
currentPacketPtr += (sizeof(float) * 3);
// tell the mixer not to add additional attenuation to our source
*(currentPacketPtr++) = 255;
// memcpy the corrected render yaw
float correctedYaw = fmodf(-1 * data->linkedAvatar->getAbsoluteHeadYaw(), 360);
float correctedYaw = fmodf(-1 * interfaceAvatar.getAbsoluteHeadYaw(), 360);
if (correctedYaw > 180) {
correctedYaw -= 360;
@ -154,29 +134,30 @@ int audioCallback (const void *inputBuffer,
correctedYaw += 360;
}
if (data->mixerLoopbackFlag) {
if (parentAudio->_mixerLoopbackFlag) {
correctedYaw = correctedYaw > 0
? correctedYaw + AGENT_LOOPBACK_MODIFIER
: correctedYaw - AGENT_LOOPBACK_MODIFIER;
? correctedYaw + AGENT_LOOPBACK_MODIFIER
: correctedYaw - AGENT_LOOPBACK_MODIFIER;
}
memcpy(currentPacketPtr, &correctedYaw, sizeof(float));
currentPacketPtr += sizeof(float);
currentPacketPtr += sizeof(float);
// copy the audio data to the last BUFFER_LENGTH_BYTES bytes of the data packet
memcpy(currentPacketPtr, inputLeft, BUFFER_LENGTH_BYTES);
data->audioSocket->send((sockaddr *)&audioMixerSocket, dataPacket, BUFFER_LENGTH_BYTES + leadingBytes);
agentList->getAgentSocket().send(audioMixer->getActiveSocket(), dataPacket, BUFFER_LENGTH_BYTES + leadingBytes);
}
}
int16_t *outputLeft = ((int16_t **) outputBuffer)[0];
int16_t *outputRight = ((int16_t **) outputBuffer)[1];
int16_t* outputLeft = ((int16_t**) outputBuffer)[0];
int16_t* outputRight = ((int16_t**) outputBuffer)[1];
memset(outputLeft, 0, PACKET_LENGTH_BYTES_PER_CHANNEL);
memset(outputRight, 0, PACKET_LENGTH_BYTES_PER_CHANNEL);
AudioRingBuffer *ringBuffer = data->ringBuffer;
AudioRingBuffer* ringBuffer = &parentAudio->_ringBuffer;
// if we've been reset, and there isn't any new packets yet
// just play some silence
@ -184,15 +165,16 @@ int audioCallback (const void *inputBuffer,
if (ringBuffer->getEndOfLastWrite() != NULL) {
if (!ringBuffer->isStarted() && ringBuffer->diffLastWriteNextOutput() < PACKET_LENGTH_SAMPLES + JITTER_BUFFER_SAMPLES) {
//printLog("Held back, buffer has %d of %d samples required.\n", ringBuffer->diffLastWriteNextOutput(), PACKET_LENGTH_SAMPLES + JITTER_BUFFER_SAMPLES);
//printLog("Held back, buffer has %d of %d samples required.\n",
// ringBuffer->diffLastWriteNextOutput(), PACKET_LENGTH_SAMPLES + JITTER_BUFFER_SAMPLES);
} else if (ringBuffer->diffLastWriteNextOutput() < PACKET_LENGTH_SAMPLES) {
ringBuffer->setStarted(false);
starve_counter++;
packetsReceivedThisPlayback = 0;
::numStarves++;
parentAudio->_packetsReceivedThisPlayback = 0;
// printLog("Starved #%d\n", starve_counter);
data->wasStarved = 10; // Frames to render the indication that the system was starved.
parentAudio->_wasStarved = 10; // Frames to render the indication that the system was starved.
} else {
if (!ringBuffer->isStarted()) {
ringBuffer->setStarted(true);
@ -206,21 +188,22 @@ int audioCallback (const void *inputBuffer,
// if we haven't fired off the flange effect, check if we should
// TODO: lastMeasuredHeadYaw is now relative to body - check if this still works.
int lastYawMeasured = fabsf(data->linkedAvatar->getLastMeasuredHeadYaw());
int lastYawMeasured = fabsf(interfaceAvatar.getLastMeasuredHeadYaw());
if (!samplesLeftForFlange && lastYawMeasured > MIN_FLANGE_EFFECT_THRESHOLD) {
if (!::samplesLeftForFlange && lastYawMeasured > MIN_FLANGE_EFFECT_THRESHOLD) {
// we should flange for one second
if ((lastYawMeasuredMaximum = std::max(lastYawMeasuredMaximum, lastYawMeasured)) != lastYawMeasured) {
lastYawMeasuredMaximum = std::min(lastYawMeasuredMaximum, MIN_FLANGE_EFFECT_THRESHOLD);
if ((::lastYawMeasuredMaximum = std::max(::lastYawMeasuredMaximum, lastYawMeasured)) != lastYawMeasured) {
::lastYawMeasuredMaximum = std::min(::lastYawMeasuredMaximum, MIN_FLANGE_EFFECT_THRESHOLD);
samplesLeftForFlange = SAMPLE_RATE;
::samplesLeftForFlange = SAMPLE_RATE;
flangeIntensity = MIN_FLANGE_INTENSITY +
((lastYawMeasuredMaximum - MIN_FLANGE_EFFECT_THRESHOLD) / (float)(MAX_FLANGE_EFFECT_THRESHOLD - MIN_FLANGE_EFFECT_THRESHOLD)) *
::flangeIntensity = MIN_FLANGE_INTENSITY +
((::lastYawMeasuredMaximum - MIN_FLANGE_EFFECT_THRESHOLD) /
(float)(MAX_FLANGE_EFFECT_THRESHOLD - MIN_FLANGE_EFFECT_THRESHOLD)) *
(1 - MIN_FLANGE_INTENSITY);
flangeRate = FLANGE_BASE_RATE * flangeIntensity;
flangeWeight = MAX_FLANGE_SAMPLE_WEIGHT * flangeIntensity;
::flangeRate = FLANGE_BASE_RATE * ::flangeIntensity;
::flangeWeight = MAX_FLANGE_SAMPLE_WEIGHT * ::flangeIntensity;
}
}
@ -229,13 +212,14 @@ int audioCallback (const void *inputBuffer,
int leftSample = ringBuffer->getNextOutput()[s];
int rightSample = ringBuffer->getNextOutput()[s + PACKET_LENGTH_SAMPLES_PER_CHANNEL];
if (samplesLeftForFlange > 0) {
float exponent = (SAMPLE_RATE - samplesLeftForFlange - (SAMPLE_RATE / flangeRate)) / (SAMPLE_RATE / flangeRate);
int sampleFlangeDelay = (SAMPLE_RATE / (1000 * flangeIntensity)) * powf(2, exponent);
if (::samplesLeftForFlange > 0) {
float exponent = (SAMPLE_RATE - ::samplesLeftForFlange - (SAMPLE_RATE / ::flangeRate)) /
(SAMPLE_RATE / ::flangeRate);
int sampleFlangeDelay = (SAMPLE_RATE / (1000 * ::flangeIntensity)) * powf(2, exponent);
if (samplesLeftForFlange != SAMPLE_RATE || s >= (SAMPLE_RATE / 2000)) {
if (::samplesLeftForFlange != SAMPLE_RATE || s >= (SAMPLE_RATE / 2000)) {
// we have a delayed sample to add to this sample
int16_t *flangeFrame = ringBuffer->getNextOutput();
int flangeIndex = s - sampleFlangeDelay;
@ -251,13 +235,13 @@ int audioCallback (const void *inputBuffer,
int16_t leftFlangeSample = flangeFrame[flangeIndex];
int16_t rightFlangeSample = flangeFrame[flangeIndex + PACKET_LENGTH_SAMPLES_PER_CHANNEL];
leftSample = (1 - flangeWeight) * leftSample + (flangeWeight * leftFlangeSample);
rightSample = (1 - flangeWeight) * rightSample + (flangeWeight * rightFlangeSample);
leftSample = (1 - ::flangeWeight) * leftSample + (::flangeWeight * leftFlangeSample);
rightSample = (1 - ::flangeWeight) * rightSample + (::flangeWeight * rightFlangeSample);
samplesLeftForFlange--;
::samplesLeftForFlange--;
if (samplesLeftForFlange == 0) {
lastYawMeasuredMaximum = 0;
if (::samplesLeftForFlange == 0) {
::lastYawMeasuredMaximum = 0;
}
}
}
@ -267,8 +251,8 @@ int audioCallback (const void *inputBuffer,
}
// add data to the scope
scope->addSamples(1, outputLeft, PACKET_LENGTH_SAMPLES_PER_CHANNEL);
scope->addSamples(2, outputRight, PACKET_LENGTH_SAMPLES_PER_CHANNEL);
parentAudio->_scope->addSamples(1, outputLeft, PACKET_LENGTH_SAMPLES_PER_CHANNEL);
parentAudio->_scope->addSamples(2, outputRight, PACKET_LENGTH_SAMPLES_PER_CHANNEL);
ringBuffer->setNextOutput(ringBuffer->getNextOutput() + PACKET_LENGTH_SAMPLES);
@ -278,140 +262,108 @@ int audioCallback (const void *inputBuffer,
}
}
gettimeofday(&data->lastCallback, NULL);
gettimeofday(&parentAudio->_lastCallbackTime, NULL);
return paContinue;
}
void Audio::updateMixerParams(in_addr_t newMixerAddress, in_port_t newMixerPort) {
audioData->mixerAddress = newMixerAddress;
audioData->mixerPort = newMixerPort;
}
struct AudioRecThreadStruct {
AudioData *sharedAudioData;
};
void *receiveAudioViaUDP(void *args) {
AudioRecThreadStruct *threadArgs = (AudioRecThreadStruct *) args;
AudioData *sharedAudioData = threadArgs->sharedAudioData;
int16_t *receivedData = new int16_t[PACKET_LENGTH_SAMPLES];
ssize_t receivedBytes;
// Init Jitter timer values
timeval previousReceiveTime, currentReceiveTime = {};
gettimeofday(&previousReceiveTime, NULL);
gettimeofday(&currentReceiveTime, NULL);
int totalPacketsReceived = 0;
stdev.reset();
while (!stopAudioReceiveThread) {
if (sharedAudioData->audioSocket->receive((void *)receivedData, &receivedBytes)) {
gettimeofday(&currentReceiveTime, NULL);
totalPacketsReceived++;
double tDiff = diffclock(&previousReceiveTime, &currentReceiveTime);
//printLog("tDiff %4.1f\n", tDiff);
// Discard first few received packets for computing jitter (often they pile up on start)
if (totalPacketsReceived > 3) stdev.addValue(tDiff);
if (stdev.getSamples() > 500) {
sharedAudioData->measuredJitter = stdev.getStDev();
//printLog("Avg: %4.2f, Stdev: %4.2f\n", stdev.getAverage(), sharedAudioData->measuredJitter);
stdev.reset();
}
AudioRingBuffer *ringBuffer = sharedAudioData->ringBuffer;
if (!ringBuffer->isStarted()) {
packetsReceivedThisPlayback++;
}
else {
//printLog("Audio packet received at %6.0f\n", usecTimestampNow()/1000);
}
if (packetsReceivedThisPlayback == 1) gettimeofday(&firstPlaybackTimer, NULL);
ringBuffer->parseData((unsigned char *)receivedData, PACKET_LENGTH_BYTES);
previousReceiveTime = currentReceiveTime;
}
void outputPortAudioError(PaError error) {
if (error != paNoError) {
printLog("-- portaudio termination error --\n");
printLog("PortAudio error (%d): %s\n", error, Pa_GetErrorText(error));
}
pthread_exit(0);
}
void Audio::setMixerLoopbackFlag(bool newMixerLoopbackFlag) {
audioData->mixerLoopbackFlag = newMixerLoopbackFlag;
}
bool Audio::getMixerLoopbackFlag() {
return audioData->mixerLoopbackFlag;
}
/**
* Initialize portaudio and start an audio stream.
* Should be called at the beginning of program exection.
* @seealso Audio::terminate
* @return Returns true if successful or false if an error occurred.
Use Audio::getError() to retrieve the error code.
*/
Audio::Audio(Oscilloscope* s, Avatar* linkedAvatar) {
paError = Pa_Initialize();
if (paError != paNoError) goto error;
scope = s;
audioData = new AudioData();
audioData->linkedAvatar = linkedAvatar;
// setup a UDPSocket
audioData->audioSocket = new UDPSocket(AUDIO_UDP_LISTEN_PORT);
audioData->ringBuffer = new AudioRingBuffer(RING_BUFFER_SAMPLES, PACKET_LENGTH_SAMPLES);
AudioRecThreadStruct threadArgs;
threadArgs.sharedAudioData = audioData;
pthread_create(&audioReceiveThread, NULL, receiveAudioViaUDP, (void *) &threadArgs);
paError = Pa_OpenDefaultStream(&stream,
2, // input channels
2, // output channels
(paInt16 | paNonInterleaved), // sample format
SAMPLE_RATE, // sample rate (hz)
BUFFER_LENGTH_SAMPLES, // frames per buffer
audioCallback, // callback function
(void *) audioData); // user data to be passed to callback
if (paError != paNoError) goto error;
initialized = true;
Audio::Audio(Oscilloscope* scope) :
_stream(NULL),
_ringBuffer(RING_BUFFER_SAMPLES, PACKET_LENGTH_SAMPLES),
_scope(scope),
_averagedLatency(0.0),
_measuredJitter(0),
_wasStarved(0),
_lastInputLoudness(0),
_mixerLoopbackFlag(false),
_lastVelocity(0),
_lastAcceleration(0),
_totalPacketsReceived(0),
_firstPlaybackTime(),
_packetsReceivedThisPlayback(0)
{
outputPortAudioError(Pa_Initialize());
outputPortAudioError(Pa_OpenDefaultStream(&_stream,
2,
2,
(paInt16 | paNonInterleaved),
SAMPLE_RATE,
BUFFER_LENGTH_SAMPLES,
audioCallback,
(void*) this));
// start the stream now that sources are good to go
Pa_StartStream(stream);
if (paError != paNoError) goto error;
return;
error:
printLog("-- Failed to initialize portaudio --\n");
printLog("PortAudio error (%d): %s\n", paError, Pa_GetErrorText(paError));
initialized = false;
delete[] audioData;
outputPortAudioError(Pa_StartStream(_stream));
gettimeofday(&_lastReceiveTime, NULL);
}
float Audio::getInputLoudness() const {
return audioData->lastInputLoudness;
Audio::~Audio() {
if (_stream) {
outputPortAudioError(Pa_CloseStream(_stream));
outputPortAudioError(Pa_Terminate());
}
}
void Audio::render(int screenWidth, int screenHeight)
{
if (initialized) {
// Take a pointer to the acquired microphone input samples and add procedural sounds
void Audio::addProceduralSounds(int16_t* inputBuffer, int numSamples) {
const float MAX_AUDIBLE_VELOCITY = 6.0;
const float MIN_AUDIBLE_VELOCITY = 0.1;
const int VOLUME_BASELINE = 400;
const float SOUND_PITCH = 8.f;
float speed = glm::length(_lastVelocity);
float volume = VOLUME_BASELINE * (1.f - speed / MAX_AUDIBLE_VELOCITY);
// Add a noise-modulated sinewave with volume that tapers off with speed increasing
if ((speed > MIN_AUDIBLE_VELOCITY) && (speed < MAX_AUDIBLE_VELOCITY)) {
for (int i = 0; i < numSamples; i++) {
inputBuffer[i] += (int16_t)((cosf((float) i / SOUND_PITCH * speed) * randFloat()) * volume * speed);
}
}
}
void Audio::addReceivedAudioToBuffer(unsigned char* receivedData, int receivedBytes) {
const int NUM_INITIAL_PACKETS_DISCARD = 3;
timeval currentReceiveTime;
gettimeofday(&currentReceiveTime, NULL);
_totalPacketsReceived++;
double timeDiff = diffclock(&_lastReceiveTime, &currentReceiveTime);
// Discard first few received packets for computing jitter (often they pile up on start)
if (_totalPacketsReceived > NUM_INITIAL_PACKETS_DISCARD) {
::stdev.addValue(timeDiff);
}
if (::stdev.getSamples() > 500) {
_measuredJitter = ::stdev.getStDev();
//printLog("Avg: %4.2f, Stdev: %4.2f\n", stdev.getAverage(), sharedAudioData->measuredJitter);
::stdev.reset();
}
if (!_ringBuffer.isStarted()) {
_packetsReceivedThisPlayback++;
}
if (_packetsReceivedThisPlayback == 1) {
gettimeofday(&_firstPlaybackTime, NULL);
}
_ringBuffer.parseData((unsigned char *)receivedData, PACKET_LENGTH_BYTES);
_lastReceiveTime = currentReceiveTime;
}
void Audio::render(int screenWidth, int screenHeight) {
if (_stream) {
glLineWidth(2.0);
glBegin(GL_LINES);
glColor3f(1,1,1);
@ -438,25 +390,23 @@ void Audio::render(int screenWidth, int screenHeight)
}
glEnd();
// Show a bar with the amount of audio remaining in ring buffer beyond current playback
float remainingBuffer = 0;
timeval currentTime;
gettimeofday(&currentTime, NULL);
float timeLeftInCurrentBuffer = 0;
if (audioData->lastCallback.tv_usec > 0) {
timeLeftInCurrentBuffer = AUDIO_CALLBACK_MSECS - diffclock(&audioData->lastCallback, &currentTime);
if (_lastCallbackTime.tv_usec > 0) {
timeLeftInCurrentBuffer = AUDIO_CALLBACK_MSECS - diffclock(&_lastCallbackTime, &currentTime);
}
// /(1000.0*(float)BUFFER_LENGTH_SAMPLES/(float)SAMPLE_RATE) * frameWidth
if (audioData->ringBuffer->getEndOfLastWrite() != NULL)
remainingBuffer = audioData->ringBuffer->diffLastWriteNextOutput() / PACKET_LENGTH_SAMPLES * AUDIO_CALLBACK_MSECS;
if (_ringBuffer.getEndOfLastWrite() != NULL)
remainingBuffer = _ringBuffer.diffLastWriteNextOutput() / PACKET_LENGTH_SAMPLES * AUDIO_CALLBACK_MSECS;
if (audioData->wasStarved == 0) glColor3f(0, 1, 0);
else {
glColor3f(0.5 + (float)audioData->wasStarved/20.0, 0, 0);
audioData->wasStarved--;
if (_wasStarved == 0) {
glColor3f(0, 1, 0);
} else {
glColor3f(0.5 + (_wasStarved / 20.0f), 0, 0);
_wasStarved--;
}
glBegin(GL_QUADS);
@ -466,26 +416,29 @@ void Audio::render(int screenWidth, int screenHeight)
glVertex2f(startX, bottomY - 2);
glEnd();
if (audioData->averagedLatency == 0.0) audioData->averagedLatency = remainingBuffer + timeLeftInCurrentBuffer;
else audioData->averagedLatency = 0.99*audioData->averagedLatency + 0.01*((float)remainingBuffer + (float)timeLeftInCurrentBuffer);
if (_averagedLatency == 0.0) {
_averagedLatency = remainingBuffer + timeLeftInCurrentBuffer;
} else {
_averagedLatency = 0.99f * _averagedLatency + 0.01f * (remainingBuffer + timeLeftInCurrentBuffer);
}
// Show a yellow bar with the averaged msecs latency you are hearing (from time of packet receipt)
glColor3f(1,1,0);
glBegin(GL_QUADS);
glVertex2f(startX + audioData->averagedLatency/AUDIO_CALLBACK_MSECS*frameWidth - 2, topY - 2);
glVertex2f(startX + audioData->averagedLatency/AUDIO_CALLBACK_MSECS*frameWidth + 2, topY - 2);
glVertex2f(startX + audioData->averagedLatency/AUDIO_CALLBACK_MSECS*frameWidth + 2, bottomY + 2);
glVertex2f(startX + audioData->averagedLatency/AUDIO_CALLBACK_MSECS*frameWidth - 2, bottomY + 2);
glVertex2f(startX + _averagedLatency / AUDIO_CALLBACK_MSECS * frameWidth - 2, topY - 2);
glVertex2f(startX + _averagedLatency / AUDIO_CALLBACK_MSECS * frameWidth + 2, topY - 2);
glVertex2f(startX + _averagedLatency / AUDIO_CALLBACK_MSECS * frameWidth + 2, bottomY + 2);
glVertex2f(startX + _averagedLatency / AUDIO_CALLBACK_MSECS * frameWidth - 2, bottomY + 2);
glEnd();
char out[40];
sprintf(out, "%3.0f\n", audioData->averagedLatency);
drawtext(startX + audioData->averagedLatency/AUDIO_CALLBACK_MSECS*frameWidth - 10, topY-10, 0.10, 0, 1, 0, out, 1,1,0);
sprintf(out, "%3.0f\n", _averagedLatency);
drawtext(startX + _averagedLatency / AUDIO_CALLBACK_MSECS * frameWidth - 10, topY - 10, 0.10, 0, 1, 0, out, 1,1,0);
//drawtext(startX + 0, topY-10, 0.08, 0, 1, 0, out, 1,1,0);
// Show a Cyan bar with the most recently measured jitter stdev
int jitterPels = (float) audioData->measuredJitter/ ((1000.0*(float)PACKET_LENGTH_SAMPLES/(float)SAMPLE_RATE)) * (float)frameWidth;
int jitterPels = _measuredJitter / ((1000.0f * PACKET_LENGTH_SAMPLES / SAMPLE_RATE)) * frameWidth;
glColor3f(0,1,1);
glBegin(GL_QUADS);
@ -495,7 +448,7 @@ void Audio::render(int screenWidth, int screenHeight)
glVertex2f(startX + jitterPels - 2, bottomY + 2);
glEnd();
sprintf(out,"%3.1f\n", audioData->measuredJitter);
sprintf(out,"%3.1f\n", _measuredJitter);
drawtext(startX + jitterPels - 5, topY-10, 0.10, 0, 1, 0, out, 0,1,1);
sprintf(out, "%3.1fms\n", JITTER_BUFFER_LENGTH_MSECS);
@ -503,34 +456,4 @@ void Audio::render(int screenWidth, int screenHeight)
}
}
/**
* Close the running audio stream, and deinitialize portaudio.
* Should be called at the end of program execution.
* @return Returns true if the initialization was successful, or false if an error occured.
The error code may be retrieved by Audio::getError().
*/
bool Audio::terminate() {
stopAudioReceiveThread = true;
pthread_join(audioReceiveThread, NULL);
if (initialized) {
initialized = false;
paError = Pa_CloseStream(stream);
if (paError != paNoError) goto error;
paError = Pa_Terminate();
if (paError != paNoError) goto error;
}
delete audioData;
return true;
error:
printLog("-- portaudio termination error --\n");
printLog("PortAudio error (%d): %s\n", paError, Pa_GetErrorText(paError));
return false;
}
#endif

54
interface/src/Audio.h
View file

@ -10,44 +10,46 @@
#define __interface__Audio__
#include <portaudio.h>
#include "AudioData.h"
#include <AudioRingBuffer.h>
#include "Oscilloscope.h"
#include "Avatar.h"
class Audio {
public:
// initializes audio I/O
Audio(Oscilloscope *s, Avatar *linkedAvatar);
void render();
Audio(Oscilloscope* scope);
~Audio();
void render(int screenWidth, int screenHeight);
bool getMixerLoopbackFlag();
void setMixerLoopbackFlag(bool newMixerLoopbackFlag);
void setMixerLoopbackFlag(bool mixerLoopbackFlag) { _mixerLoopbackFlag = mixerLoopbackFlag; }
float getInputLoudness() const;
void updateMixerParams(in_addr_t mixerAddress, in_port_t mixerPort);
float getLastInputLoudness() const { return _lastInputLoudness; };
void setLastAcceleration(glm::vec3 a) { audioData->setLastAcceleration(a); };
void setLastVelocity(glm::vec3 v) { audioData->setLastVelocity(v); };
void setLastAcceleration(glm::vec3 lastAcceleration) { _lastAcceleration = lastAcceleration; };
void setLastVelocity(glm::vec3 lastVelocity) { _lastVelocity = lastVelocity; };
// terminates audio I/O
bool terminate();
void addProceduralSounds(int16_t* inputBuffer, int numSamples);
void addReceivedAudioToBuffer(unsigned char* receivedData, int receivedBytes);
private:
bool initialized;
AudioData *audioData;
// protects constructor so that public init method is used
Audio();
// hold potential error returned from PortAudio functions
PaError paError;
// audio stream handle
PaStream *stream;
// audio receive thread
pthread_t audioReceiveThread;
PaStream* _stream;
AudioRingBuffer _ringBuffer;
Oscilloscope* _scope;
timeval _lastCallbackTime;
timeval _lastReceiveTime;
float _averagedLatency;
float _measuredJitter;
int _wasStarved;
float _lastInputLoudness;
bool _mixerLoopbackFlag;
glm::vec3 _lastVelocity;
glm::vec3 _lastAcceleration;
int _totalPacketsReceived;
timeval _firstPlaybackTime;
int _packetsReceivedThisPlayback;
// give access to AudioData class from audioCallback
friend int audioCallback (const void*, void*, unsigned long, const PaStreamCallbackTimeInfo*, PaStreamCallbackFlags, void*);

48
interface/src/AudioData.cpp
View file

@ -1,48 +0,0 @@
//
// AudioData.cpp
// interface
//
// Created by Stephen Birarda on 1/29/13.
// Copyright (c) 2013 HighFidelity, Inc. All rights reserved.
//
#ifndef _WIN32
#include "AudioData.h"
AudioData::AudioData() {
mixerAddress = 0;
mixerPort = 0;
averagedLatency = 0.0;
lastCallback.tv_usec = 0;
wasStarved = 0;
measuredJitter = 0;
jitterBuffer = 0;
mixerLoopbackFlag = false;
audioSocket = NULL;
}
AudioData::~AudioData() {
delete audioSocket;
}
// Take a pointer to the acquired microphone input samples and add procedural sounds
void AudioData::addProceduralSounds(int16_t* inputBuffer, int numSamples) {
const float MAX_AUDIBLE_VELOCITY = 6.0;
const float MIN_AUDIBLE_VELOCITY = 0.1;
float speed = glm::length(_lastVelocity);
float volume = 400 * (1.f - speed/MAX_AUDIBLE_VELOCITY);
// Add a noise-modulated sinewave with volume that tapers off with speed increasing
if ((speed > MIN_AUDIBLE_VELOCITY) && (speed < MAX_AUDIBLE_VELOCITY)) {
for (int i = 0; i < numSamples; i++) {
inputBuffer[i] += (int16_t) ((cosf((float)i / 8.f * speed) * randFloat()) * volume * speed) ;
}
}
return;
}
#endif

54
interface/src/AudioData.h
View file

@ -1,54 +0,0 @@
//
// AudioData.h
// interface
//
// Created by Stephen Birarda on 1/29/13.
// Copyright (c) 2013 HighFidelity, Inc. All rights reserved.
//
#ifndef __interface__AudioData__
#define __interface__AudioData__
#include <stdint.h>
#include <glm/glm.hpp>
#include "AudioRingBuffer.h"
#include "UDPSocket.h"
#include "Avatar.h"
class AudioData {
public:
AudioData();
~AudioData();
AudioRingBuffer *ringBuffer;
UDPSocket *audioSocket;
Avatar *linkedAvatar;
// store current mixer address and port
in_addr_t mixerAddress;
in_port_t mixerPort;
timeval lastCallback;
float averagedLatency;
float measuredJitter;
float jitterBuffer;
int wasStarved;
float lastInputLoudness;
bool mixerLoopbackFlag;
// Added avatar acceleration and velocity for procedural effects sounds from client
void setLastVelocity(glm::vec3 v) { _lastVelocity = v; };
void setLastAcceleration(glm::vec3 a) { _lastAcceleration = a; };
void addProceduralSounds(int16_t* inputBuffer, int numSamples);
private:
glm::vec3 _lastVelocity;
glm::vec3 _lastAcceleration;
};
#endif /* defined(__interface__AudioData__) */

12
libraries/shared/src/AgentList.cpp
View file

@ -243,14 +243,10 @@ bool AgentList::addOrUpdateAgent(sockaddr *publicSocket, sockaddr *localSocket,
// set the agent active right away
newAgent->activatePublicSocket();
}
if (newAgent->getType() == AGENT_TYPE_AUDIO_MIXER && audioMixerSocketUpdate != NULL) {
// this is an audio mixer
// for now that means we need to tell the audio class
// to use the local socket information the domain server gave us
sockaddr_in *publicSocketIn = (sockaddr_in *)publicSocket;
audioMixerSocketUpdate(publicSocketIn->sin_addr.s_addr, publicSocketIn->sin_port);
} else if (newAgent->getType() == AGENT_TYPE_VOXEL || newAgent->getType() == AGENT_TYPE_AVATAR_MIXER) {
if (newAgent->getType() == AGENT_TYPE_VOXEL ||
newAgent->getType() == AGENT_TYPE_AVATAR_MIXER ||
newAgent->getType() == AGENT_TYPE_AUDIO_MIXER) {
// this is currently the cheat we use to talk directly to our test servers on EC2
// to be removed when we have a proper identification strategy
newAgent->activatePublicSocket();

1
libraries/shared/src/AgentList.h
View file

@ -46,7 +46,6 @@ public:
AgentListIterator end() const;
void(*linkedDataCreateCallback)(Agent *);
void(*audioMixerSocketUpdate)(in_addr_t, in_port_t);
int size() { return _numAgents; }

3
libraries/shared/src/AgentTypes.h
View file

@ -20,8 +20,9 @@
// Agent Type Codes
const char AGENT_TYPE_DOMAIN = 'D';
const char AGENT_TYPE_VOXEL = 'V';
const char AGENT_TYPE_AVATAR = 'I'; // could also be injector???
const char AGENT_TYPE_AVATAR = 'I';
const char AGENT_TYPE_AUDIO_MIXER = 'M';
const char AGENT_TYPE_AVATAR_MIXER = 'W';
const char AGENT_TYPE_AUDIO_INJECTOR = 'A';
#endif

7
libraries/shared/src/AudioInjector.cpp
View file

@ -22,12 +22,9 @@ const float BUFFER_SEND_INTERVAL_USECS = (BUFFER_LENGTH_SAMPLES / SAMPLE_RATE) *
AudioInjector::AudioInjector(const char* filename) :
_numTotalBytesAudio(0),
_position(),
_bearing(0),
_attenuationModifier(255)
{
_position[0] = 0.0f;
_position[1] = 0.0f;
_position[2] = 0.0f;
_attenuationModifier(255) {
std::fstream sourceFile;

27
libraries/shared/src/AudioRingBuffer.cpp
View file

@ -7,6 +7,9 @@
//
#include <cstring>
#include "PacketHeaders.h"
#include "AudioRingBuffer.h"
AudioRingBuffer::AudioRingBuffer(int ringSamples, int bufferSamples) :
@ -46,18 +49,22 @@ AudioRingBuffer* AudioRingBuffer::clone() const {
const int AGENT_LOOPBACK_MODIFIER = 307;
int AudioRingBuffer::parseData(unsigned char* sourceBuffer, int numBytes) {
if (numBytes > (_bufferLengthSamples * sizeof(int16_t))) {
unsigned char* dataBuffer = sourceBuffer + 1;
if (sourceBuffer[0] == PACKET_HEADER_INJECT_AUDIO ||
sourceBuffer[0] == PACKET_HEADER_MICROPHONE_AUDIO) {
// if this came from an injector or interface client
// there's data required for spatialization to pull out
unsigned char *dataPtr = sourceBuffer + 1;
memcpy(&_position, dataBuffer, sizeof(_position));
dataBuffer += (sizeof(_position));
memcpy(&_position, dataPtr, sizeof(_position));
dataPtr += (sizeof(_position));
unsigned int attenuationByte = *(dataPtr++);
unsigned int attenuationByte = *(dataBuffer++);
_attenuationRatio = attenuationByte / 255.0f;
memcpy(&_bearing, dataPtr, sizeof(float));
dataPtr += sizeof(_bearing);
memcpy(&_bearing, dataBuffer, sizeof(float));
dataBuffer += sizeof(_bearing);
if (_bearing > 180 || _bearing < -180) {
// we were passed an invalid bearing because this agent wants loopback (pressed the H key)
@ -70,8 +77,6 @@ int AudioRingBuffer::parseData(unsigned char* sourceBuffer, int numBytes) {
} else {
_shouldLoopbackForAgent = false;
}
sourceBuffer = dataPtr;
}
if (!_endOfLastWrite) {
@ -82,7 +87,7 @@ int AudioRingBuffer::parseData(unsigned char* sourceBuffer, int numBytes) {
_started = false;
}
memcpy(_endOfLastWrite, sourceBuffer, _bufferLengthSamples * sizeof(int16_t));
memcpy(_endOfLastWrite, dataBuffer, _bufferLengthSamples * sizeof(int16_t));
_endOfLastWrite += _bufferLengthSamples;

2
libraries/shared/src/PacketHeaders.h
View file

@ -20,6 +20,8 @@ const PACKET_HEADER PACKET_HEADER_PING_REPLY = 'R';
const PACKET_HEADER PACKET_HEADER_HEAD_DATA = 'H';
const PACKET_HEADER PACKET_HEADER_Z_COMMAND = 'Z';
const PACKET_HEADER PACKET_HEADER_INJECT_AUDIO = 'I';
const PACKET_HEADER PACKET_HEADER_MIXED_AUDIO = 'A';
const PACKET_HEADER PACKET_HEADER_MICROPHONE_AUDIO = 'M';
const PACKET_HEADER PACKET_HEADER_SET_VOXEL = 'S';
const PACKET_HEADER PACKET_HEADER_SET_VOXEL_DESTRUCTIVE = 'O';
const PACKET_HEADER PACKET_HEADER_ERASE_VOXEL = 'E';