kopia lustrzana https://gitlab.com/eliggett/wfview
361 wiersze
11 KiB
C++
361 wiersze
11 KiB
C++
/*
|
|
This class handles both RX and TX audio, each is created as a separate instance of the class
|
|
but as the setup/handling if output (RX) and input (TX) devices is so similar I have combined them.
|
|
*/
|
|
|
|
#include "audiohandler.h"
|
|
|
|
#include "logcategories.h"
|
|
#include "ulaw.h"
|
|
|
|
|
|
|
|
audioHandler::audioHandler(QObject* parent) : QObject(parent)
|
|
{
|
|
Q_UNUSED(parent)
|
|
}
|
|
|
|
audioHandler::~audioHandler()
|
|
{
|
|
|
|
if (isInitialized) {
|
|
stop();
|
|
}
|
|
|
|
if (audioInput != Q_NULLPTR) {
|
|
audioInput = Q_NULLPTR;
|
|
delete audioInput;
|
|
}
|
|
if (audioOutput != Q_NULLPTR) {
|
|
delete audioOutput;
|
|
audioOutput = Q_NULLPTR;
|
|
}
|
|
|
|
if (converterThread != Q_NULLPTR) {
|
|
converterThread->quit();
|
|
converterThread->wait();
|
|
}
|
|
}
|
|
|
|
bool audioHandler::init(audioSetup setup)
|
|
{
|
|
if (isInitialized) {
|
|
return false;
|
|
}
|
|
|
|
this->setup = setup;
|
|
qInfo(logAudio()) << (setup.isinput ? "Input" : "Output") << "audio handler starting:" << setup.name;
|
|
if (setup.port.isNull())
|
|
{
|
|
qInfo(logAudio()) << (setup.isinput ? "Input" : "Output") << "No audio device was found. You probably need to install libqt5multimedia-plugins.";
|
|
return false;
|
|
}
|
|
|
|
qDebug(logAudio()) << "Creating" << (setup.isinput ? "Input" : "Output") << "audio device:" << setup.name <<
|
|
", bits" << inFormat.sampleSize() <<
|
|
", codec" << setup.codec <<
|
|
", latency" << setup.latency <<
|
|
", localAFGain" << setup.localAFgain <<
|
|
", radioChan" << inFormat.channelCount() <<
|
|
", resampleQuality" << setup.resampleQuality <<
|
|
", samplerate" << inFormat.sampleRate() <<
|
|
", uLaw" << setup.ulaw;
|
|
|
|
inFormat = toQAudioFormat(setup.codec, setup.sampleRate);
|
|
|
|
if(!setup.isinput)
|
|
{
|
|
this->setVolume(setup.localAFgain);
|
|
}
|
|
|
|
outFormat = setup.port.preferredFormat();
|
|
qDebug(logAudio()) << (setup.isinput ? "Input" : "Output") << "Preferred Format: SampleSize" << outFormat.sampleSize() << "Channel Count" << outFormat.channelCount() <<
|
|
"Sample Rate" << outFormat.sampleRate() << "Codec" << outFormat.codec() << "Sample Type" << outFormat.sampleType();
|
|
if (outFormat.channelCount() > 2) {
|
|
outFormat.setChannelCount(2);
|
|
}
|
|
else if (outFormat.channelCount() < 1)
|
|
{
|
|
qCritical(logAudio()) << (setup.isinput ? "Input" : "Output") << "No channels found, aborting setup.";
|
|
return false;
|
|
}
|
|
|
|
/* if (outFormat.channelCount() == 1 && inFormat.channelCount() == 2) {
|
|
outFormat.setChannelCount(2);
|
|
if (!setup.port.isFormatSupported(outFormat)) {
|
|
qCritical(logAudio()) << (setup.isinput ? "Input" : "Output") << "Cannot request stereo input!";
|
|
outFormat.setChannelCount(1);
|
|
}
|
|
}
|
|
|
|
if (outFormat.sampleType()==QAudioFormat::SignedInt) {
|
|
outFormat.setSampleType(QAudioFormat::Float);
|
|
outFormat.setSampleSize(32);
|
|
if (!setup.port.isFormatSupported(outFormat)) {
|
|
qCritical(logAudio()) << (setup.isinput ? "Input" : "Output") << "Attempt to select 32bit Float failed, reverting to SignedInt";
|
|
outFormat.setSampleType(QAudioFormat::SignedInt);
|
|
outFormat.setSampleSize(16);
|
|
}
|
|
|
|
}
|
|
*/
|
|
|
|
if (outFormat.sampleSize() == 24) {
|
|
// We can't convert this easily so use 32 bit instead.
|
|
outFormat.setSampleSize(32);
|
|
if (!setup.port.isFormatSupported(outFormat)) {
|
|
qCritical(logAudio()) << (setup.isinput ? "Input" : "Output") << "24 bit requested and 32 bit audio not supported, try 16 bit instead";
|
|
outFormat.setSampleSize(16);
|
|
}
|
|
}
|
|
|
|
qDebug(logAudio()) << (setup.isinput ? "Input" : "Output") << "Selected format: SampleSize" << outFormat.sampleSize() << "Channel Count" << outFormat.channelCount() <<
|
|
"Sample Rate" << outFormat.sampleRate() << "Codec" << outFormat.codec() << "Sample Type" << outFormat.sampleType();
|
|
|
|
// We "hopefully" now have a valid format that is supported so try connecting
|
|
|
|
converter = new audioConverter();
|
|
converterThread = new QThread(this);
|
|
if (setup.isinput) {
|
|
converterThread->setObjectName("audioConvIn()");
|
|
}
|
|
else {
|
|
converterThread->setObjectName("audioConvOut()");
|
|
}
|
|
converter->moveToThread(converterThread);
|
|
|
|
connect(this, SIGNAL(setupConverter(QAudioFormat,QAudioFormat,quint8,quint8)), converter, SLOT(init(QAudioFormat,QAudioFormat,quint8,quint8)));
|
|
connect(converterThread, SIGNAL(finished()), converter, SLOT(deleteLater()));
|
|
connect(this, SIGNAL(sendToConverter(audioPacket)), converter, SLOT(convert(audioPacket)));
|
|
converterThread->start(QThread::TimeCriticalPriority);
|
|
|
|
if (setup.isinput) {
|
|
audioInput = new QAudioInput(setup.port, outFormat, this);
|
|
connect(audioInput, SIGNAL(stateChanged(QAudio::State)), SLOT(stateChanged(QAudio::State)));
|
|
emit setupConverter(outFormat, inFormat, 7, setup.resampleQuality);
|
|
connect(converter, SIGNAL(converted(audioPacket)), this, SLOT(convertedInput(audioPacket)));
|
|
}
|
|
else {
|
|
audioOutput = new QAudioOutput(setup.port, outFormat, this);
|
|
connect(audioOutput, SIGNAL(stateChanged(QAudio::State)), SLOT(stateChanged(QAudio::State)));
|
|
emit setupConverter(inFormat, outFormat, 7, setup.resampleQuality);
|
|
connect(converter, SIGNAL(converted(audioPacket)), this, SLOT(convertedOutput(audioPacket)));
|
|
}
|
|
|
|
|
|
|
|
qInfo(logAudio()) << (setup.isinput ? "Input" : "Output") << "thread id" << QThread::currentThreadId();
|
|
|
|
underTimer = new QTimer();
|
|
underTimer->setSingleShot(true);
|
|
connect(underTimer, SIGNAL(timeout()), this, SLOT(clearUnderrun()));
|
|
|
|
this->start();
|
|
|
|
return true;
|
|
}
|
|
|
|
void audioHandler::start()
|
|
{
|
|
qInfo(logAudio()) << (setup.isinput ? "Input" : "Output") << "start() running";
|
|
|
|
if (setup.isinput) {
|
|
//this->open(QIODevice::WriteOnly);
|
|
//audioInput->start(this);
|
|
#ifdef Q_OS_WIN
|
|
audioInput->setBufferSize(inFormat.bytesForDuration(setup.latency * 100));
|
|
#else
|
|
audioInput->setBufferSize(inFormat.bytesForDuration(setup.latency * 1000));
|
|
#endif
|
|
audioDevice = audioInput->start();
|
|
connect(audioInput, SIGNAL(destroyed()), audioDevice, SLOT(deleteLater()), Qt::UniqueConnection);
|
|
connect(audioDevice, SIGNAL(readyRead()), this, SLOT(getNextAudioChunk()), Qt::UniqueConnection);
|
|
//audioInput->setNotifyInterval(setup.blockSize/2);
|
|
//connect(audioInput, SIGNAL(notify()), this, SLOT(getNextAudioChunk()), Qt::UniqueConnection);
|
|
}
|
|
else {
|
|
// Buffer size must be set before audio is started.
|
|
#ifdef Q_OS_WIN
|
|
audioOutput->setBufferSize(outFormat.bytesForDuration(setup.latency * 100));
|
|
#else
|
|
audioOutput->setBufferSize(outFormat.bytesForDuration(setup.latency * 1000));
|
|
#endif
|
|
audioDevice = audioOutput->start();
|
|
connect(audioOutput, SIGNAL(destroyed()), audioDevice, SLOT(deleteLater()), Qt::UniqueConnection);
|
|
}
|
|
if (!audioDevice) {
|
|
qInfo(logAudio()) << (setup.isinput ? "Input" : "Output") << "Audio device failed to start()";
|
|
return;
|
|
}
|
|
|
|
}
|
|
|
|
|
|
void audioHandler::stop()
|
|
{
|
|
qDebug(logAudio()) << (setup.isinput ? "Input" : "Output") << "stop() running";
|
|
|
|
if (audioOutput != Q_NULLPTR && audioOutput->state() != QAudio::StoppedState) {
|
|
// Stop audio output
|
|
audioOutput->stop();
|
|
}
|
|
|
|
if (audioInput != Q_NULLPTR && audioInput->state() != QAudio::StoppedState) {
|
|
// Stop audio output
|
|
audioInput->stop();
|
|
}
|
|
audioDevice = Q_NULLPTR;
|
|
}
|
|
|
|
void audioHandler::setVolume(unsigned char volume)
|
|
{
|
|
this->volume = audiopot[volume];
|
|
qDebug(logAudio()) << (setup.isinput ? "Input" : "Output") << "setVolume: " << volume << "(" << this->volume << ")";
|
|
}
|
|
|
|
|
|
void audioHandler::incomingAudio(audioPacket packet)
|
|
{
|
|
|
|
if (audioDevice != Q_NULLPTR && packet.data.size() > 0) {
|
|
packet.volume = volume;
|
|
|
|
emit sendToConverter(packet);
|
|
}
|
|
|
|
return;
|
|
}
|
|
|
|
void audioHandler::convertedOutput(audioPacket packet) {
|
|
|
|
if (packet.data.size() > 0 ) {
|
|
|
|
currentLatency = packet.time.msecsTo(QTime::currentTime()) + (outFormat.durationForBytes(audioOutput->bufferSize() - audioOutput->bytesFree()) / 1000);
|
|
if (audioDevice != Q_NULLPTR) {
|
|
audioDevice->write(packet.data);
|
|
if (lastReceived.msecsTo(QTime::currentTime()) > 100) {
|
|
qDebug(logAudio()) << (setup.isinput ? "Input" : "Output") << "Time since last audio packet" << lastReceived.msecsTo(QTime::currentTime()) << "Expected around" << setup.blockSize;
|
|
}
|
|
lastReceived = QTime::currentTime();
|
|
}
|
|
/*if ((packet.seq > lastSentSeq + 1) && (setup.codec == 0x40 || setup.codec == 0x80)) {
|
|
qDebug(logAudio()) << (setup.isinput ? "Input" : "Output") << "Attempting FEC on packet" << packet.seq << "as last is" << lastSentSeq;
|
|
lastSentSeq = packet.seq;
|
|
incomingAudio(packet); // Call myself again to run the packet a second time (FEC)
|
|
}
|
|
*/
|
|
lastSentSeq = packet.seq;
|
|
emit haveLevels(getAmplitude(), setup.latency, currentLatency, isUnderrun);
|
|
|
|
amplitude = packet.amplitude;
|
|
}
|
|
}
|
|
|
|
void audioHandler::getNextAudioChunk()
|
|
{
|
|
if (audioDevice) {
|
|
tempBuf.data.append(audioDevice->readAll());
|
|
}
|
|
if (tempBuf.data.length() >= outFormat.bytesForDuration(setup.blockSize * 1000)) {
|
|
audioPacket packet;
|
|
packet.time = QTime::currentTime();
|
|
packet.sent = 0;
|
|
packet.volume = volume;
|
|
memcpy(&packet.guid, setup.guid, GUIDLEN);
|
|
//QTime startProcessing = QTime::currentTime();
|
|
packet.data.clear();
|
|
packet.data = tempBuf.data.mid(0, outFormat.bytesForDuration(setup.blockSize * 1000));
|
|
tempBuf.data.remove(0, outFormat.bytesForDuration(setup.blockSize * 1000));
|
|
|
|
emit sendToConverter(packet);
|
|
}
|
|
|
|
/* If there is still enough data in the buffer, call myself again in 20ms */
|
|
if (tempBuf.data.length() >= outFormat.bytesForDuration(setup.blockSize * 1000)) {
|
|
QTimer::singleShot(setup.blockSize, this, &audioHandler::getNextAudioChunk);
|
|
}
|
|
|
|
return;
|
|
}
|
|
|
|
|
|
void audioHandler::convertedInput(audioPacket audio)
|
|
{
|
|
if (audio.data.size() > 0) {
|
|
emit haveAudioData(audio);
|
|
if (lastReceived.msecsTo(QTime::currentTime()) > 100) {
|
|
qDebug(logAudio()) << (setup.isinput ? "Input" : "Output") << "Time since last audio packet" << lastReceived.msecsTo(QTime::currentTime()) << "Expected around" << setup.blockSize ;
|
|
}
|
|
lastReceived = QTime::currentTime();
|
|
amplitude = audio.amplitude;
|
|
emit haveLevels(getAmplitude(), setup.latency, currentLatency, isUnderrun);
|
|
}
|
|
}
|
|
|
|
void audioHandler::changeLatency(const quint16 newSize)
|
|
{
|
|
|
|
setup.latency = newSize;
|
|
|
|
if (!setup.isinput) {
|
|
stop();
|
|
start();
|
|
}
|
|
qDebug(logAudio()) << (setup.isinput ? "Input" : "Output") << "Configured latency: " << setup.latency << "Buffer Duration:" << outFormat.durationForBytes(audioOutput->bufferSize())/1000 << "ms";
|
|
|
|
}
|
|
|
|
int audioHandler::getLatency()
|
|
{
|
|
return currentLatency;
|
|
}
|
|
|
|
|
|
|
|
quint16 audioHandler::getAmplitude()
|
|
{
|
|
return static_cast<quint16>(amplitude * 255.0);
|
|
}
|
|
|
|
|
|
|
|
void audioHandler::stateChanged(QAudio::State state)
|
|
{
|
|
// Process the state
|
|
switch (state)
|
|
{
|
|
case QAudio::IdleState:
|
|
{
|
|
isUnderrun = true;
|
|
if (underTimer->isActive()) {
|
|
underTimer->stop();
|
|
}
|
|
break;
|
|
}
|
|
case QAudio::ActiveState:
|
|
{
|
|
//qDebug(logAudio()) << (setup.isinput ? "Input" : "Output") << "Audio started!";
|
|
if (!underTimer->isActive()) {
|
|
underTimer->start(500);
|
|
}
|
|
break;
|
|
}
|
|
case QAudio::SuspendedState:
|
|
{
|
|
break;
|
|
}
|
|
case QAudio::StoppedState:
|
|
{
|
|
break;
|
|
}
|
|
default: {
|
|
}
|
|
break;
|
|
}
|
|
}
|
|
|
|
void audioHandler::clearUnderrun()
|
|
{
|
|
isUnderrun = false;
|
|
}
|