what is the maximum value of audio data returned by AudioUnitRender() function in ios

Question

I am using the AudioUnitRender() function in my render call back function to get audio data from the microphone in real time in iphone

err = AudioUnitRender(player->outputUnit, ioActioanFlags, inTimeStamp, 1, inNumberFrames, ioData);

The audio data comes into ioData when the callback function is called automatically. I am using the audio data returned in ioData as shown below:

for(frame = 0; frame<inNumberFrames; ++frame){  
    Float32 *data = (Float32*)ioData->mBuffers[0].mData;  
    myvar[k++] = (data)[frame];  
.  
.  
.  
}

Here myvar is a array of Float32 type. I had guessed that the input audio is within the +1.0/-1.0 range since the values in myvar[] were always something within that range. I recently found out that if I make loud sounds close to the microphone, sometimes I get values in myvar[]that are outside the +1.0/-1.0 range.

What exactly is the range of the Float32 type data returned by AudioUnitRender() as the microphone audio data?

Is it possible to get whatever raw audio is being returned by AudioUnitRender() as an integer? The AudioRecord class in android gives me the raw microphone audio as signed short numbers (16-bits). I am looking for it's equivalent in ios, in objective C.

--- EDIT 1 ---

The current configuration used for the audio is given below:

// Configure the audio session
AVAudioSession *sessionInstance = [AVAudioSession sharedInstance];

// we are going to play and record so we pick that category
NSError *error = nil;
[sessionInstance setCategory:AVAudioSessionCategoryPlayAndRecord error:&error];

// set the buffer duration to 5 ms
NSTimeInterval bufferDuration = .004;        // with setPreferredSampleRate:16000 gives inNumberFrames =  64 in SineWaveRenderProc()
//    NSTimeInterval bufferDuration = .016;        // with setPreferredSampleRate:16000 gives inNumberFrames =  256 in SineWaveRenderProc() ;; NOTE: 0.004*4 = 0.016
[sessionInstance setPreferredIOBufferDuration:bufferDuration error:&error];

// set the session's sample rate
// [sessionInstance setPreferredSampleRate:44100 error:&error];     // ORIGINAL     // inNumberFrames =  256 in SineWaveRenderProc() with bufferDuration = .005; above
[sessionInstance setPreferredSampleRate:16000 error:&error];         // inNumberFrames =  64 in SineWaveRenderProc() with bufferDuration = .005; above


// activate the audio session
[[AVAudioSession sharedInstance] setActive:YES error:&error];
// XThrowIfError((OSStatus)error.code, "couldn't set session active");

// NOTE: looks like this is necessary
UInt32 one = 1;
AudioUnitSetProperty(player->outputUnit, kAudioOutputUnitProperty_EnableIO, kAudioUnitScope_Input, 1, &one, sizeof(one) );
AudioUnitSetProperty(player->outputUnit, kAudioOutputUnitProperty_EnableIO, kAudioUnitScope_Output, 0, &one, sizeof(one) );

The range should be -1.0/1.0, but if you are using a different sample rate than the data source or built-in iPhone hardware, the resampling filter might generate ringing/ripple outside that range. — hotpaw2
@hotpaw I have edited in the configuration I used to set the sampling frequency — user13267
The audio hardware on newer iOS devices seems to run at 48ks/s. All other sample rates may be filtered resample conversions from 48k (or higher?). — hotpaw2

hotpaw2 hotpaw2 · Accepted Answer · 2018-11-30T21:12:19

To the 2nd part of your question, yes. You can configure many Audio Units (such as RemoteIO) to output audio samples in 16-bit signed integer format.

set the Audio Unit kAudioUnitProperty_StreamFormat property to something like:

AudioStreamBasicDescription   audioFormat;
audioFormat.mFormatID         =  kAudioFormatLinearPCM;
audioFormat.mFormatFlags      =  kAudioFormatFlagIsSignedInteger;
audioFormat.mBitsPerChannel   =  16;
// etc.

This used to be the default format, back in the iPhone 3Gs days. But newer iPhones can do audio DSP faster using 32-bit floats, so the default changed.

what is the maximum value of audio data returned by AudioUnitRender() function in ios

1 Answers