Changing Buffer Size

[LukasJoswiak]

I have been searching for a way to change the buffer size to increase the number of data points received per second.

This issue (#50) is similar to what I am trying to do, but there does not appear to be an answer. How do I decrease the buffer size?

Thanks!

[syedhali]

For only the microphone component? If it's purely for a graphical display then you can adjust the rollingHistoryLength on the audio plot.

[self.audioPlot setRollingHistoryLength:128]

For the microphone it will default to the buffer size specified by the AVAudioSession, so before instantiating the microphone you could try adjusting the preferred IO buffer duration like:

NSError* err;
[[AVAudioSession sharedInstance] setPreferredIOBufferDuration:(DURATION IN SECONDS) error:&err];

Reference:
https://developer.apple.com/library/ios/documentation/avfoundation/reference/AVAudioSession_ClassReference/Reference/Reference.html#//apple_ref/occ/instm/AVAudioSession/setPreferredIOBufferDuration:error:

[LukasJoswiak]

Yes, this is for the microphone. When the microphone is on, in the microphone:hasAudioReceived:withBufferSize:withNumberOfChannels: method, I am calling createFFTwithBufferSize:WithAudioData: and updateFFTWithBufferSize:withAudioData:, taken from the FFT example project. In updateFFTWithBufferSize:withAudioData:, I am storing the max frequency each loop in an array, which eventually gets turned into a graph.

Setting the preferred IO buffer duration does not change bufferSize in updateFFTWithBufferSize:withAudioData: and does not effect the number of data points being received. What I want to do is increase the number of data points by calling updateFFTWithBufferSize:withAudioData: more often, which I believe can be achieved by decreasing the buffer size.

[JeanRintoul]

Now I can run it with any window size to get better resolution.

-(void)microphone:(EZMicrophone *)microphone

hasAudioReceived:(float **)buffer

withBufferSize:(UInt32)bufferSize

withNumberOfChannels:(UInt32)numberOfChannels {

// Getting audio data as an array of float buffer arrays. What does that mean? Because the audio is coming in as a stereo signal the data is split into a left and right channel. So buffer[0] corresponds to the float* data for the left channel while buffer[1] corresponds to the float* data for the right channel.



// See the Thread Safety warning above, but in a nutshell these callbacks happen on a separate audio thread. We wrap any UI updating in a GCD block on the main thread to avoid blocking that audio flow.

dispatch_async(dispatch_get_main_queue(),^{

    // All the audio plot needs is the buffer data (float*) and the size. Internally the audio plot will handle all the drawing related code, history management, and freeing its own resources. Hence, one badass line of code gets you a pretty plot :)



    [self.audioPlot updateBuffer:buffer[0] withBufferSize:bufferSize];



    // Decibel Calculation.

    float one       = 1.0;

    float meanVal   = 0.0;

    float tiny      = 0.1;

    vDSP_vsq(buffer[0], 1, buffer[0], 1, bufferSize);

    vDSP_meanv(buffer[0], 1, &meanVal, bufferSize);

    vDSP_vdbcon(&meanVal, 1, &one, &meanVal, 1, 1, 0);

    // Exponential moving average to dB level to only get continous sounds.

    float currentdb = 1.0 - (fabs(meanVal)/100);

    if (lastdbValue == INFINITY || lastdbValue == -INFINITY || isnan(lastdbValue)) {

        lastdbValue = 0.0;

    }

    dbValue =   ((1.0 - tiny)*lastdbValue) + tiny*currentdb;

    lastdbValue = dbValue;

    // NSLog(@"dbval:  %f",dbValue);

    //

    // Setup the FFT if it's not already setup

    int samplestoCopy = fmin(bufferSize, FFTLEN - _fftBufIndex);

    for ( size_t i = 0; i < samplestoCopy; i++ ) {

        _fftBuf[_fftBufIndex+i] = buffer[0][i];

    }

    _fftBufIndex        += samplestoCopy;

   _samplesRemaining    -= samplestoCopy;

    if (_fftBufIndex == FFTLEN) {

        if( !_isFFTSetup ){

            [self createFFTWithBufferSize:FFTLEN withAudioData:_fftBuf];

            _isFFTSetup = YES;

        }

        [self updateFFTWithBufferSize:FFTLEN withAudioData:_fftBuf];

        _fftBufIndex        = 0.0f;

        _samplesRemaining   = FFTLEN;

    }



});

}

[LukasJoswiak]

Jean, would you mind posting the Xcode project, so I can get a feel for how it works? If not, what are FFTLEN and fftBufIndex initialized as?

[JeanRintoul]

const UInt32 FFTLEN = 4410;
int _fftBufIndex = 0;
You may want to change FFTLEN to whatever you want, based on what resolution you want... This gave 5Hz bins.

[jamrader]

JeanRintoul or LikasJoswiak, did either of you manage to get a FFTViewController that could measure the Hz of a sound through the microphone with greater accuracy than that which is provided for with a buffer size of 512? I'm trying to manage increasing its accuracy (most likely by increasing buffer size) but am coming up short.

[syedhali]

As of 1.1.0 you can just use the EZAudioFFTRolling to get an FFT with a larger window. Please check out the answer for #185 for a code example of how to store more than 512 audio samples using an EZPlotHistoryInfo structure.

Also, the EZAudioFFTExample has been updated to be a high resolution pitch detector using a 4096 length FFT.