Core Audio sine wave example

core-audio-sine-wave.c

// To run:
//   clang core-audio-sine-wave.c -framework AudioUnit && ./a.out
#include <AudioUnit/AudioUnit.h>

#define SAMPLE_RATE 48000
#define TONE_FREQUENCY 440
#define M_TAU 2.0 * M_PI

OSStatus RenderSineWave(
        void *inRefCon,
        AudioUnitRenderActionFlags *ioActionFlags,
        const AudioTimeStamp *inTimeStamp,
        UInt32 inBusNumber,
        UInt32 inNumberFrames,
        AudioBufferList *ioData)
{
    static float theta;

    SInt16 *left = (SInt16 *)ioData->mBuffers[0].mData;
    for (UInt32 frame = 0; frame < inNumberFrames; ++frame) {
        left[frame] = (SInt16)(sin(theta) * 32767.0f);
        theta += M_TAU * TONE_FREQUENCY / SAMPLE_RATE;
        if (theta > M_TAU) {
            theta -= M_TAU;
        }
    }

    // Copy left channel to right channel
    memcpy(ioData->mBuffers[1].mData, left, ioData->mBuffers[1].mDataByteSize);

    return noErr;
}

int main() {
    OSErr err;

    AudioComponentDescription acd = {
        .componentType = kAudioUnitType_Output,
        .componentSubType = kAudioUnitSubType_DefaultOutput,
        .componentManufacturer = kAudioUnitManufacturer_Apple,
    };

    AudioComponent output = AudioComponentFindNext(NULL, &acd);
    if (!output) printf("Can't find default output\n");

    AudioUnit toneUnit;
    err = AudioComponentInstanceNew(output, &toneUnit);
    if (err) fprintf(stderr, "Error creating unit: %d\n", err);

    AURenderCallbackStruct input = { .inputProc = RenderSineWave };
    err = AudioUnitSetProperty(toneUnit, kAudioUnitProperty_SetRenderCallback,
            kAudioUnitScope_Input, 0, &input, sizeof(input));
    if (err) printf("Error setting callback: %d\n", err);

    AudioStreamBasicDescription asbd = {
        .mFormatID = kAudioFormatLinearPCM,
        .mFormatFlags = 0
            | kAudioFormatFlagIsSignedInteger
            | kAudioFormatFlagIsPacked
            | kAudioFormatFlagIsNonInterleaved,
        .mSampleRate = 48000,
        .mBitsPerChannel = 16,
        .mChannelsPerFrame = 2,
        .mFramesPerPacket = 1,
        .mBytesPerFrame = 2,
        .mBytesPerPacket = 2,
    };

    err = AudioUnitSetProperty(toneUnit, kAudioUnitProperty_StreamFormat,
            kAudioUnitScope_Input, 0, &asbd, sizeof(asbd));
    if (err) printf("Error setting stream format: %d\n", err);

    err = AudioUnitInitialize(toneUnit);
    if (err) printf("Error initializing unit: %d\n", err);

    err = AudioOutputUnitStart(toneUnit);
    if (err) printf("Error starting unit: %d\n", err);

    usleep(500000);

    AudioOutputUnitStop(toneUnit);
    AudioUnitUninitialize(toneUnit);
    AudioComponentInstanceDispose(toneUnit);
}

Thanks! Great example!
I'd suggest some optimization of the rendering callback:

static float theta = 0.01;
static float amplitude = 0.25; // SignalLevel - Volume [0; 1]
static float thetaIncrement = M_TAU * TONE_FREQUENCY / SAMPLE_RATE;

float *left = (float *)ioData->mBuffers[0].mData;

for (UInt32 frame = 0; frame < inNumberFrames; ++frame) {
    left[frame] = (float)(sin(theta) * amplitude);
    theta += thetaIncrement;
}

You can replace float back to SInt16. I use float because all my audio unit logic is based on float values. I hear pure and beautiful sine wave!

@gcatlin great example thanks

@emartinson I'm confused by your optimization which doesn't provide clean sine tone at all if implementing the callback as you have suggested and I get a fuzzy tone instead. Am I missing something?

OSStatus RenderSineWave(
        void *inRefCon,
        AudioUnitRenderActionFlags *ioActionFlags,
        const AudioTimeStamp *inTimeStamp,
        UInt32 inBusNumber,
        UInt32 inNumberFrames,
        AudioBufferList *ioData)
{
    static float theta = 0.01;
    static float amplitude = 0.25; // SignalLevel - Volume [0; 1]
    static float thetaIncrement = M_TAU * TONE_FREQUENCY / SAMPLE_RATE;

    float *left = (float *)ioData->mBuffers[0].mData;

    for (UInt32 frame = 0; frame < inNumberFrames; ++frame) {
        left[frame] = (float)(sin(theta) * amplitude);
        theta += thetaIncrement;
    }

    // Copy left channel to right channel
    memcpy(ioData->mBuffers[1].mData, left, ioData->mBuffers[1].mDataByteSize);

    return noErr;
}

@shakfu
The reason of your fuzzy tone is probably related to either wrong Sample Rate or (most probable) audio buffer format. Sorry for that. I'm using Float sample data in my project and copy-pasted, but in this example uses Integer buffer values (kAudioFormatFlagIsSignedInteger) and casting to SInt16 in: (SInt16)(sin(theta) * 32767.0f) - that's why it's necessary to multiply to 32767.0f.

The main idea of the optimization is moving all unnecessary math out of for-loop:
theta += M_TAU * TONE_FREQUENCY / SAMPLE_RATE; -> this is a constant - no need to compute it in loop, no need to compute it even within renderer callback - is renderer callback must work as fast as possible (as it affects sound delay and audio accuracy - sometimes it's critical).

@emartinson Thanks for taking the time to explain this. 👍