[ameat]

Hey folks

Wondering if anyone could lead me in the right direction for working with a sound input device to determine the frequency; basically for instrument tuning type purposes. Additionally, how much of a factor is something like background noise? And what type of filter options (using CoreAudio?) are available to gather the best source for frequency polling?

thanks for any help.

[parallax]

Background noise will generally not impact your measurements. White noise will add a low-level spread across the entire spectrum. In the time domain this alters the wave, but in the frequency domain the frequency you're looking for will not be shifted or anything.

If you just execute a Discrete Fourier Transform on your sound sample this should be sufficient.

Sorry, what are you looking for additionally?

[ameat]

Specifically, which API toolset to even begin learning and where to search in the documentation would be helpful. I've done lots of little cocoa demo projects and such, but nothing dealing at all with audio IO.

thanks.

[parallax]

The last time I was hacking with CoreAudio stuff was a long time ago and documentation was scarce/none at that point. :-) However, collected from some old source, here are some things you may want to check out:

Callback you need to write:

Code:
OSStatus inputIOProc (AudioDeviceID  inDevice, const AudioTimeStamp*  inNow, const AudioBufferList*   inInputData, const AudioTimeStamp*  inInputTime, AudioBufferList*  outOutputData, const AudioTimeStamp* inOutputTime, void* defptr) {
   float *in = inInputData->mBuffers[0].mData;
   int numSamples = def->deviceBufferSize / def->deviceFormat.mBytesPerFrame;
   for (int i = 0; i < numSamples; ++i) {
     float sample = *in++;
     // now go to town with your sample
   }
}

Setting up the device:

Code:
AudioDeviceID inputDevice;
UInt32 count = sizeof(inputDevice);
AudioHardwareGetProperty(kAudioHardwarePropertyDefaultInputDevice, &count, &inputDevice);

You'll need

Code:
UInt32 deviceBufferSize;
count = sizeof(deviceBufferSize);
 err = AudioDeviceGetProperty(inputDevice, 0, false, kAudioDevicePropertyBufferSize, &count, &deviceBufferSize);
AudioStreamBasicDescription deviceFormat;
count = sizeof(deviceFormat);
 err = AudioDeviceGetProperty(inputDevice, 0, false, kAudioDevicePropertyBufferSize, &count, &deviceBufferSize);

to be able to read properties of the device.

Starting the callback:

Code:
err = AudioDeviceAddIOProc(inputDevice, inputIOProc, (void *) def);
if (err != kAudioHardwareNoError) return false;
err = AudioDeviceStart(inputDevice, inputIOProc);

Stopping the callback:

Code:
AudioDeviceStop(inputDevice, inputIOProc);

(void *)def are the defs of your class. This is if you want to directly access Objective-C class fields from your C callback. (in inputIOProc)

Code:
typedef struct {
   @defs(WhateverYourClassIs);
} classdef;

classdef *def = (classdef *)self;

[ameat]

Fun... thanks. Hopefully some of the docs have been beefed up since then, or the API is more object oriented, but guess I just need to dig in.

[ameat]

Some API updates for Tiger:
http://developer.apple.com/technotes/tn2005/tn2113.html

[ameat]

And some more reference links (for anyone else curious, and since i'm at work):

• http://aldebaran.armory.com/~zenomt/macosx/MTCoreAudio/
• http://www.mat.ucsb.edu:8000/CoreAudio/13
• http://lists.apple.com/archives/core.../msg00016.html
• http://borkware.com/rants/sound/
• http://lists.apple.com/archives/Coreaudio-api/