Question

I was going over core audio conversion services in the Learning Core Audio and I was struck by this example in their sample code:

while(1)
{
    // wrap the destination buffer in an AudioBufferList
    AudioBufferList convertedData;
    convertedData.mNumberBuffers = 1;
    convertedData.mBuffers[0].mNumberChannels = mySettings->outputFormat.mChannelsPerFrame;
    convertedData.mBuffers[0].mDataByteSize = outputBufferSize;
    convertedData.mBuffers[0].mData = outputBuffer;

    UInt32 frameCount = packetsPerBuffer;

    // read from the extaudiofile
    CheckResult(ExtAudioFileRead(mySettings->inputFile,
                                 &frameCount,
                                 &convertedData),
                "Couldn't read from input file");

    if (frameCount == 0) {
        printf ("done reading from file");
        return;
    }

    // write the converted data to the output file
    CheckResult (AudioFileWritePackets(mySettings->outputFile,
                                       FALSE,
                                       frameCount,
                                       NULL,
                                       outputFilePacketPosition / mySettings->outputFormat.mBytesPerPacket, 
                                       &frameCount,
                                       convertedData.mBuffers[0].mData),
                 "Couldn't write packets to file");

    // advance the output file write location
    outputFilePacketPosition += (frameCount * mySettings->outputFormat.mBytesPerPacket);
}

notice how frameCount is defined as packetsPerBuffer.. packetsPerBuffer is defined here:

UInt32 outputBufferSize = 32 * 1024; // 32 KB is a good starting point
UInt32 sizePerPacket = mySettings->outputFormat.mBytesPerPacket;    
UInt32 packetsPerBuffer = outputBufferSize / sizePerPacket;

the part that stumped me is AudioFileWritePackets is called.. in the documentation AudioFileWritePackets third and fifth parameters are defined as:

inNumBytes
The number of bytes of audio data being written.

ioNumPackets
On input, a pointer to the number of packets to write. On output, a pointer to the number of packets actually written..

yet in the code both parameters are given frameCount.. how is this possible?? I know with PCM data 1 frame = 1 packet:

// define the ouput format. AudioConverter requires that one of the data formats be LPCM
audioConverterSettings.outputFormat.mSampleRate = 44100.0;
audioConverterSettings.outputFormat.mFormatID = kAudioFormatLinearPCM;
audioConverterSettings.outputFormat.mFormatFlags = kAudioFormatFlagIsBigEndian | kAudioFormatFlagIsSignedInteger | kAudioFormatFlagIsPacked;
audioConverterSettings.outputFormat.mBytesPerPacket = 4;
audioConverterSettings.outputFormat.mFramesPerPacket = 1;
audioConverterSettings.outputFormat.mBytesPerFrame = 4;
audioConverterSettings.outputFormat.mChannelsPerFrame = 2;
audioConverterSettings.outputFormat.mBitsPerChannel = 16;

but the same lPCM formatting also clearly states that there are 4 bytes per packet (= 4 bytes per frame)..

so how does this work? (the same applies to the other example in the same chapter that uses AudioConverterFillComplexBuffer instead of ExtAudioFileRead, and uses packets instead of frames.. but it’s the same thing)

Answer 1

I think you’re right, according to the definition in the AudioFile.h header file, AudioFileWritePackets should take the number of bytes of audio data being written as the third parameter, and in that Learning Core Audio example the framecount variable is defined as the number of packets, not the number of bytes.

I tried the examples out and got the exact same output with (framecount * 4), 0 and even -1 as the third parameter of the AudioFileWritePackets function call. So for me it would seem that the function doesn’t work exactly as defined in the .h file (does not require the third parameter), and that in that example the authors of the book have not noticed this error either – I might be wrong though.