BassBASS_ChannelGetData Method (Int32, Single, Int32)

[DllImportAttribute("bass")]
public static int BASS_ChannelGetData(
	int handle,
	float[] buffer,
	int length
)

<DllImportAttribute("bass">]
Public Shared Function BASS_ChannelGetData ( 
	handle As Integer,
	<OutAttribute> buffer As Single(),
	length As Integer
) As Integer

Parameters

handle

Type: SystemInt32
The channel handle... a HCHANNEL, HMUSIC, HSTREAM, or HRECORD.

buffer

Type: SystemSingle
The array (float[]) to receive the data, use BASS_SAMPLE_FLOAT when creating the channel stream!

length

Type: SystemInt32
Number of bytes wanted, and/or the following flags (BASSData):

BASS_DATA_FLOAT	Return floating-point sample data.
BASS_DATA_FIXED	Return 8.24 fixed-point data.
BASS_DATA_FFT256	256 sample FFT (returns 128 floating-point values)
BASS_DATA_FFT512	512 sample FFT (returns 256 floating-point values)
BASS_DATA_FFT1024	1024 sample FFT (returns 512 floating-point values)
BASS_DATA_FFT2048	2048 sample FFT (returns 1024 floating-point values)
BASS_DATA_FFT4096	4096 sample FFT (returns 2048 floating-point values)
BASS_DATA_FFT8192	8192 sample FFT (returns 4096 floating-point values)
BASS_DATA_FFT16384	16384 sample FFT (returns 8192 floating-point values)
BASS_DATA_FFT32768	16384 sample FFT (returns 16384 floating-point values)
BASS_DATA_FFT_COMPLEX	Return the complex FFT result rather than the magnitudes. This increases the amount of data returned (as listed above) fourfold, as it returns real and imaginary parts and the full FFT result (not only the first half). The real and imaginary parts are interleaved in the returned data.
BASS_DATA_FFT_INDIVIDUAL	Perform a separate FFT for each channel, rather than a single combined FFT. The size of the data returned (as listed above) is multiplied by the number of channels.
BASS_DATA_FFT_NOWINDOW	Prevent a Hann window being applied to the sample data when performing an FFT.
BASS_DATA_FFT_NYQUIST	Return an extra value for the Nyquist frequency magnitude. The Nyquist frequency is always included in a complex FFT result.
BASS_DATA_FFT_REMOVEDC	Remove any DC bias from the sample data when performing an FFT.
BASS_DATA_NOREMOVE	Do not remove the data from a recording channel's buffer. This is automatic if the recording channel is using a RECORDPROC callback function. This flag is ignored on other channel types.
BASS_DATA_AVAILABLE	Query the amount of data the channel has buffered. This flag is primarily of use when recording, and can't be used with decoding channels as they do not have playback buffers. buffer can be when using this flag.

Return Value

When requesting FFT data, the number of bytes read from the channel (to perform the FFT) is returned.

When requesting sample data, the number of bytes written to buffer will be returned (not necessarily the same as the number of bytes read when using the BASS_DATA_FLOAT flag).

When using the BASS_DATA_AVAILABLE flag, the number of bytes in the channel's buffer is returned.

This function can only return as much data as has been written to the channel's buffer, so it may not always be possible to get the amount of data requested, especially if you request large amounts. If you really do need large amounts, then increase the buffer lengths (BASS_CONFIG_BUFFER). The BASS_DATA_AVAILABLE flag can be used to check how much data a channel's buffer contains at any time, including when stopped or stalled.

When requesting data from a decoding channel, data is decoded directly from the channel's source (no playback buffer) and as much data as the channel has available can be decoded at a time.

When retrieving sample data, 8-bit samples are unsigned (0 to 255) , 16-bit samples are signed (-32768 to 32767), 32-bit floating-point samples range from -1 to +1 (not clipped, so can actually be outside this range). That is unless the BASS_DATA_FLOAT flag is used, in which case, the sample data will be converted to 32-bit floating-point if it is not already.

Unless complex data is requested via the BASS_DATA_FFT_COMPLEX flag, the magnitudes of the first half of an FFT result are returned. For example, with a 2048 sample FFT, there will be 1024 floating-point values returned. Each value, or "bin", ranges from 0 to 1 (can actually go higher if the sample data is floating-point and not clipped). The 1st bin contains the DC component, the 2nd contains the amplitude at 1/2048 of the channel's sample rate, followed by the amplitude at 2/2048, 3/2048, etc. A Hann window is applied to the sample data to reduce leakage, unless the BASS_DATA_FFT_NOWINDOW flag is used. When a window is applied, it causes the DC component to leak into the next bin, but that can be removed (reduced to 0) by using the BASS_DATA_FFT_REMOVEDC flag. Doing so slightly increases the processing required though, so it should only be done when needed, which is when a window is applied and the 2nd bin value is important.

Channels that have 2 or more sample channels (ie. stereo or above) may have FFT performed on each individual channel, using the BASS_DATA_FFT_INDIVIDUAL flag. Without this flag, all of the channels are combined, and a single mono FFT is performed. Performing the extra individual FFTs of course increases the amount of processing required. The return values are interleaved in the same order as the channel's sample data, eg. stereo = left,right,left,etc.

This function is most useful if you wish to visualize (eg. spectrum analyze) the sound.

FFT processing hint 1: DC stands for direct current (same as what a flashlight cell gives out) and is represents a 0 Hz sound, which cannot exist in real life. A sound sample should not have any DC component but probably will due to inaccuracies in the recording equipment. You won't hear it (except a click when playing and stopping the sample). The DC component is basically the average of all the samples that the FFT was applied to, and is pretty useless. You'll have to halve the DC component that BASS returns to get the actual DC component, as BASS doubles all the bin values.

FFT processing hint 2: How to read out the amplitude of a single frequency/a frequency band? FFT[0] maps to amplitude at 0 Hz, FFT[length-1] maps to amplitude at Nyquist's frequency. So the index to FFT data at an arbitrary band is:

idx = length*freq/Nyquist
where:
length : length of the returned FFT buffer (in samples)
freq : required frequency (Hz)
Nyquist : Nyquist's frequency of the signal (half the sampling rate) (in Hz)

// a 30ms window in bytes to be filled with sample data
int length = (int)Bass.BASS_ChannelSeconds2Bytes(channel, 0.03);

// first we need a mananged object where the sample data should be placed
// length is in bytes, so the number of floats to process is length/4 
float[] data = new float[length/4];

// get the sample data
length = Bass.BASS_ChannelGetData(channel, data, length);

' a 30ms window in bytes to be filled with sample data
Dim length As Integer = CInt(Bass.BASS_ChannelSeconds2Bytes(channel, 0.03))

' first we need a mananged object where the sample data should be placed
' length is in bytes, so the number of floats to process is length/4 
Dim data(length/4 - 1) As Single

' get the sample data
length = Bass.BASS_ChannelGetData(channel, data, length)

float[] fft = new float[2048];
Bass.BASS_ChannelGetData(channel, fft, (int)BASSData.BASS_DATA_FFT4096)
// assuming the channel's samplerate is 44.1kHz,
// this will return the frequency represented by bucket 51
int hz = Utils.FFTIndex2Frequency(51, 4096, 44100);

Dim fft(2048 - 1) As Single
Bass.BASS_ChannelGetData(channel, fft, CInt(BASSData.BASS_DATA_FFT4096))
' assuming the channel's samplerate is 44.1kHz,
' this will return the frequency represented by bucket 51
Dim hz As Integer = Utils.FFTIndex2Frequency(51, 4096, 44100)

Reference