/// <summary>
/// This function is the one where all the cool stuff happens :)
/// It returns the requested number of audio frames, with time compensation for the latency values (server buffer, local buffer, DAC latency)
/// </summary>
/// <param name="format">The format being used</param>
/// <param name="bufferFrameCount">The number of frames the audio device wants</param>
/// <param name="bufferMs">Server buffer minus our local player buffer</param>
/// <param name="dacLatency">The dac latency</param>
/// <returns>The requested number of audio frames for the supplied playTime</returns>
public byte[] GetNextBuffer(SampleFormat format, int bufferFrameCount, int bufferMs, int dacLatency)
{
lock (m_Chunks)
{
// if we don't currently have a chunk ready to read from, grab the first one from the queue
if (m_Chunks.Count > 0 && m_Chunk == null)
{
m_Chunk = m_Chunks.Dequeue();
}
}
int frames = bufferFrameCount; // number of frames we'll send to the audio device
if (m_Chunk != null)
{
double serverPlayTimeMs = m_TimeProvider.ServerNow() - bufferMs + dacLatency;
double age = serverPlayTimeMs - m_Chunk.StartMs(); // age = server time - the time the server says this chunk needs to start playing
double requestedFramesDurationMs = frames / m_SampleFormat.MsRate; // duration in milliseconds of the frames being requested
int read = 0; // this will hold the number of frames we've read from the queued chunks (we often get asked to read more frames than a single chunk holds, which means we need to read across chunks)
int pos = 0; // this will keep track of our position within the output buffer as we write to it (in bytes)
if (age < -requestedFramesDurationMs)
{
// eg. we're being asked to play 80ms worth of audio, and the current chunk is 80ms too young,
// in this case just return 80ms of silence (which puts us in sync next time this function gets called)
return(new byte[bufferFrameCount * format.FrameSize]);
}
//Console.WriteLine(age); // this value is the best indicator we have to measure how tight our audio loop is. the less this value strays from zero, the better
// if the current chunk is (x)ms too young or old, seek to the correct position (seeking causes audible pops, but can't be helped if we somehow get this far out of sync)
if (Math.Abs(age) > m_OffsetToleranceMs)
{
// first, just skip through chunks as long as they're too old (eg if this chunk lasts 16ms but it's 16ms too old, just skipping this one puts us back in sync)
while (m_Chunk != null && age > m_Chunk.Duration())
{
// this chunk is too old, just skip it
//Console.WriteLine("chunk too old, dropping (age: " + age + " > " + m_Chunk.Duration() + ") " + m_Chunk.StartMs());
lock (m_Chunks)
{
if (m_Chunks.Count == 0) // no chunks left, return silence
{
m_Chunk = null;
return(new byte[bufferFrameCount * format.FrameSize]);
}
m_Chunk = m_Chunks.Dequeue();
}
age = serverPlayTimeMs - m_Chunk.StartMs();
}
// at this point we've got the right chunk, now we just need to seek to the correct position inside of it
if (m_Chunk != null)
{
if (age > 0)
{
// we're reading frames from the chunk but not actually doing anything with them (seeking forward)
// (Chunk.ReadFrames keeps an internal position pointer, so next time we call ReadFrames we should be in sync)
m_Chunk.ReadFrames((int)Math.Floor(age * m_Chunk.SampleFormat.MsRate));
}
else if (age < 0)
{
// the current chunk is slightly ahead, so we need to play a bit of silence until we're synced up (seeking backward)
int silentFrames = (int)Math.Floor(-age * m_Chunk.SampleFormat.MsRate);
read = silentFrames; // this makes sure we offset our output bytes so the first x frames are zeroes
pos = silentFrames * m_SampleFormat.FrameSize;
}
age = 0; // we don't want any frame corrections to be applied after this
}
}
int addFrames = 0; // number of frames we'll add / remove
int everyN = 0; // interval at which we'll add / remove frames
if (age > 0.1f)
{
addFrames = (int)Math.Ceiling(age);
}
else if (age < -0.1f)
{
addFrames = (int)Math.Floor(age);
}
// number of frames we'll need to actually read from the buffer
// when we want to play faster (aka age > 0), we'll read more frames from the buffer, but skip a few in playback
// when we want to play slower, we read less frames from the buffer and duplicate a few in playback
int numFramesToRead = frames + addFrames - read;
if (addFrames != 0)
{
everyN = (int)Math.Ceiling((frames + addFrames - read) / (Math.Abs(addFrames) + 1.0f));
}
byte[] output = new byte[bufferFrameCount * format.FrameSize]; // prepare the actual output bytes!
while (read < numFramesToRead) // as long as we haven't read in the number of frames requested
{
lock (m_Chunks)
{
// if the current chunk runs out, grab the next one from the queue
if (m_Chunk != null && m_Chunk.IsEndOfChunk())
{
if (m_Chunks.Count > 0)
{
m_Chunk = m_Chunks.Dequeue();
}
else // the chunk has run out, but there's no chunks waiting - return output to the audio device (this will effectively mean we've padded with zeroes, but not doing this would keep us in an infinite while loop here)
{
return(output);
}
}
}
// try to read the number of frames we need, minus the number of frames we've already read
// (often the chunk has less frames than we're requesting, which is the reason this is inside of a while loop)
byte[] chunk = (m_Chunk.ReadFrames(numFramesToRead - read));
for (int i = 0; i < chunk.Length; i += m_SampleFormat.FrameSize) // iterate through the chunk frame by frame
{
read++;
Buffer.BlockCopy(chunk, i, output, pos, m_SampleFormat.FrameSize); // copy the frame from the chunk to the output buffer
if (everyN != 0 && read % everyN == 0) // if we reached a point where a frame needs to be added/removed
{
if (addFrames > 0)
{
// adding a frame: move back our position pointer.
// this means we'll effectively skip a frame from being played, while reading an extra one from the buffer instead
// (playing faster)
pos -= m_SampleFormat.FrameSize;
}
else
{
// removing a frame: copy the current frame to the next position, and move forward our position pointer
// this means we'll duplicate one frame, while reading a frame less from the chunk
// (playing slower)
Buffer.BlockCopy(output, pos, output, pos + m_SampleFormat.FrameSize, m_SampleFormat.FrameSize);
pos += m_SampleFormat.FrameSize;
}
}
pos += m_SampleFormat.FrameSize;
}
}
return(output);
}
// return silence for the requested number of frames
return(new byte[bufferFrameCount * format.FrameSize]);
}