diff --git a/assignment-client/src/audio/AudioMixer.cpp b/assignment-client/src/audio/AudioMixer.cpp
index b511c255b9..700642bf59 100644
--- a/assignment-client/src/audio/AudioMixer.cpp
+++ b/assignment-client/src/audio/AudioMixer.cpp
@@ -10,7 +10,6 @@
#include <fcntl.h>
#include <fstream>
#include <iostream>
-#include <limits>
#include <math.h>
#include <signal.h>
#include <stdio.h>
@@ -56,9 +55,6 @@ const short JITTER_BUFFER_SAMPLES = JITTER_BUFFER_MSECS * (SAMPLE_RATE / 1000.0)
const unsigned int BUFFER_SEND_INTERVAL_USECS = floorf((NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL / (float) SAMPLE_RATE) * 1000 * 1000);
-const int MAX_SAMPLE_VALUE = std::numeric_limits<int16_t>::max();
-const int MIN_SAMPLE_VALUE = std::numeric_limits<int16_t>::min();
-
const char AUDIO_MIXER_LOGGING_TARGET_NAME[] = "audio-mixer";
void attachNewBufferToNode(Node *newNode) {
diff --git a/interface/src/Audio.cpp b/interface/src/Audio.cpp
index e5f5df2311..bd7da88aef 100644
--- a/interface/src/Audio.cpp
+++ b/interface/src/Audio.cpp
@@ -47,6 +47,7 @@ Audio::Audio(Oscilloscope* scope, int16_t initialJitterBufferSamples, QObject* p
_desiredInputFormat(),
_inputFormat(),
_inputBuffer(),
+ _monoAudioSamples(NULL),
_numInputCallbackBytes(0),
_audioOutput(NULL),
_desiredOutputFormat(),
@@ -172,8 +173,9 @@ void linearResampling(int16_t* sourceSamples, int16_t* destinationSamples,
if (sourceAudioFormat == destinationAudioFormat) {
memcpy(destinationSamples, sourceSamples, numSourceSamples * sizeof(int16_t));
} else {
+ int destinationChannels = (destinationAudioFormat.channelCount() >= 2) ? 2 : destinationAudioFormat.channelCount();
float sourceToDestinationFactor = (sourceAudioFormat.sampleRate() / (float) destinationAudioFormat.sampleRate())
- * (sourceAudioFormat.channelCount() / (float) destinationAudioFormat.channelCount()) ;
+ * (sourceAudioFormat.channelCount() / (float) destinationChannels);
// take into account the number of channels in source and destination
// accomodate for the case where have an output with > 2 channels
@@ -184,40 +186,36 @@ void linearResampling(int16_t* sourceSamples, int16_t* destinationSamples,
// for now this only supports a mono output - this would be the case for audio input
for (int i = 2; i < numSourceSamples; i += 4) {
-
if (i + 2 >= numSourceSamples) {
destinationSamples[(i - 2) / 4] = (sourceSamples[i - 2] / 2)
- + (sourceSamples[i] / 2);
+ + (sourceSamples[i] / 2);
} else {
destinationSamples[(i - 2) / 4] = (sourceSamples[i - 2] / 4)
- + (sourceSamples[i] / 2)
- + (sourceSamples[i + 2] / 4);
+ + (sourceSamples[i] / 2)
+ + (sourceSamples[i + 2] / 4);
}
}
} else {
- int sourceIndex = 0;
-
// upsample from 24 to 48
- for (int i = 0; i < numDestinationSamples; i += destinationAudioFormat.channelCount()) {
- sourceIndex = i * sourceToDestinationFactor;
+ // for now this only supports a stereo to stereo conversion - this is our case for network audio to output
+ int sourceIndex = 0;
+ int destinationToSourceFactor = (1 / sourceToDestinationFactor);
+
+ for (int i = 0; i < numDestinationSamples; i += destinationAudioFormat.channelCount() * destinationToSourceFactor) {
+ sourceIndex = (i / destinationToSourceFactor);
- if (sourceIndex >= numSourceSamples) {
- sourceIndex -= destinationAudioFormat.channelCount();
- }
-
- destinationSamples[i] = sourceSamples[sourceIndex];
-
- if (sourceAudioFormat.channelCount() == 1) {
- destinationSamples[i + 1] = sourceSamples[sourceIndex];
- } else {
- destinationSamples[i + 1] = sourceSamples[(sourceIndex) + 1];
-
- if (destinationAudioFormat.channelCount() > 2) {
- // fill the rest of the channels with silence
- for (int j = 2; j < destinationAudioFormat.channelCount(); j++) {
- destinationSamples[i] = 0;
- }
+ // fill the L/R channels and make the rest silent
+ for (int j = i; j < i + (destinationToSourceFactor * destinationAudioFormat.channelCount()); j++) {
+ if (j % destinationAudioFormat.channelCount() == 0) {
+ // left channel
+ destinationSamples[j] = sourceSamples[sourceIndex];
+ } else if (j % destinationAudioFormat.channelCount() == 1) {
+ // right channel
+ destinationSamples[j] = sourceSamples[sourceIndex + 1];
+ } else {
+ // channels above 2, fill with silence
+ destinationSamples[j] = 0;
}
}
}
@@ -281,7 +279,8 @@ void Audio::handleAudioInput() {
static int numBytesPacketHeader = numBytesForPacketHeader((unsigned char*) &PACKET_TYPE_MICROPHONE_AUDIO_NO_ECHO);
static int leadingBytes = numBytesPacketHeader + sizeof(glm::vec3) + sizeof(glm::quat) + NUM_BYTES_RFC4122_UUID;
- static int16_t* monoAudioSamples = (int16_t*) (monoAudioDataPacket + leadingBytes);
+
+ _monoAudioSamples = (int16_t*) (monoAudioDataPacket + leadingBytes);
static float inputToNetworkInputRatio = _numInputCallbackBytes * CALLBACK_ACCELERATOR_RATIO
/ NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL;
@@ -298,37 +297,23 @@ void Audio::handleAudioInput() {
_inputRingBuffer.readSamples(inputAudioSamples, inputSamplesRequired);
// zero out the monoAudioSamples array
- memset(monoAudioSamples, 0, NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL);
+ memset(_monoAudioSamples, 0, NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL);
if (!_muted) {
// we aren't muted, downsample the input audio
linearResampling((int16_t*) inputAudioSamples,
- monoAudioSamples,
+ _monoAudioSamples,
inputSamplesRequired,
NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL,
_inputFormat, _desiredInputFormat);
// add input data just written to the scope
QMetaObject::invokeMethod(_scope, "addSamples", Qt::QueuedConnection,
- Q_ARG(QByteArray, QByteArray((char*) monoAudioSamples,
+ Q_ARG(QByteArray, QByteArray((char*) _monoAudioSamples,
NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL)),
Q_ARG(bool, false), Q_ARG(bool, true));
}
-// if (Menu::getInstance()->isOptionChecked(MenuOption::EchoLocalAudio)) {
-// // if local loopback enabled, copy input to output
-// QByteArray samplesForOutput;
-// samplesForOutput.resize(inputSamplesRequired * outputToInputRatio * sizeof(int16_t));
-//
-// linearResampling(monoAudioSamples, (int16_t*) samplesForOutput.data(),
-// NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL,
-// inputSamplesRequired,
-// _desiredInputFormat, _outputFormat);
-//
-// _outputDevice->write(samplesForOutput);
-// }
-
-
// add procedural effects to the appropriate input samples
// addProceduralSounds(monoAudioSamples + (_isBufferSendCallback
// ? BUFFER_LENGTH_SAMPLES_PER_CHANNEL / CALLBACK_ACCELERATOR_RATIO : 0),
@@ -440,6 +425,17 @@ void Audio::addReceivedAudioToBuffer(const QByteArray& audioByteArray) {
int16_t ringBufferSamples[NETWORK_BUFFER_LENGTH_SAMPLES_STEREO];
_ringBuffer.readSamples(ringBufferSamples, NETWORK_BUFFER_LENGTH_SAMPLES_STEREO);
+ if (!_muted && Menu::getInstance()->isOptionChecked(MenuOption::EchoLocalAudio)) {
+ // copy whatever is pointed to at _monoAudioSamples into our ringBufferSamples
+ // so that local audio is echoed back
+
+ for (int i = 0; i < NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL; i++) {
+ ringBufferSamples[i * 2] = glm::clamp(ringBufferSamples[i * 2] + _monoAudioSamples[i],
+ MIN_SAMPLE_VALUE, MAX_SAMPLE_VALUE);
+ ringBufferSamples[(i * 2) + 1] = ringBufferSamples[i * 2];
+ }
+ }
+
// copy the packet from the RB to the output
linearResampling(ringBufferSamples,
(int16_t*) outputBuffer.data(),
@@ -460,6 +456,7 @@ void Audio::addReceivedAudioToBuffer(const QByteArray& audioByteArray) {
} else if (_audioOutput->bytesFree() == _audioOutput->bufferSize()) {
// we don't have any audio data left in the output buffer, and the ring buffer from
// the network has nothing in it either - we just starved
+ qDebug() << "Audio output just starved.\n";
_ringBuffer.setIsStarved(true);
_numFramesDisplayStarve = 10;
}
diff --git a/interface/src/Audio.h b/interface/src/Audio.h
index 8557120ab7..9a051b246b 100644
--- a/interface/src/Audio.h
+++ b/interface/src/Audio.h
@@ -76,6 +76,7 @@ private:
QAudioFormat _inputFormat;
QIODevice* _inputDevice;
QByteArray _inputBuffer;
+ int16_t* _monoAudioSamples;
int _numInputCallbackBytes;
QAudioOutput* _audioOutput;
QAudioFormat _desiredOutputFormat;
diff --git a/libraries/audio/src/AudioRingBuffer.h b/libraries/audio/src/AudioRingBuffer.h
index d715e63a97..3e6917456f 100644
--- a/libraries/audio/src/AudioRingBuffer.h
+++ b/libraries/audio/src/AudioRingBuffer.h
@@ -9,6 +9,7 @@
#ifndef __interface__AudioRingBuffer__
#define __interface__AudioRingBuffer__
+#include <limits>
#include <stdint.h>
#include <glm/glm.hpp>
@@ -24,6 +25,9 @@ const int NETWORK_BUFFER_LENGTH_SAMPLES_STEREO = NETWORK_BUFFER_LENGTH_BYTES_STE
const int NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL = 512;
const int NETWORK_BUFFER_LENGTH_SAMPLES_PER_CHANNEL = NETWORK_BUFFER_LENGTH_BYTES_PER_CHANNEL / sizeof(int16_t);
+const int MAX_SAMPLE_VALUE = std::numeric_limits<int16_t>::max();
+const int MIN_SAMPLE_VALUE = std::numeric_limits<int16_t>::min();
+
class AudioRingBuffer : public NodeData {
Q_OBJECT
public: