i am trying to get the same buffersize after i converted a stereosignal to mono that i get using a stereo signal.
When i set buffersize to 256 then i will get 128 values from the FFT with mono.
With node “abs” i am cut down to 64 values.
When using a stereo signal with buffersize 256 i get a spread of two with each 128 values.
I think it is connected to the bin size the FFt outputs on default. Or am i wrong?
Is it possible to use the whole buffersize for a mono signal?
And is it also possible to use logarithmic scaling for the FFT like it is possible the FFT(Dx9)vAudio?
the history of FFT is a history full of misundertsnadings. here are two fun facts:
the real output of a FFT is always half of the the input size
FFT has always linear frequency distribution
so to get another frequency distribution you just have to recalculate the output values somehow. if you have a look into the vvvv girlpower folder you will find some examples. just copy (or clone) this patch and replace the DShow FFT with the VAudio one.
if you mean the “Frequency Scaling” pin of the DShow node, that is just a multiplication of the values with an I (Spread) from 1…SpreadCount divided by the SpreadCount (see patch).
also:
the output of the FFT (VAudio) node is already the magnitude of the complex output, no need for the Abs node.
converting to mono can be done with a + (VAudio Spectral)
I thought the output would be complex. And the linear frequenzy distribution is understandable for me too.
But why do i get 1024 values with stereo compared to 512 with mono while using the same buffersize?
I see in the renderer that there are left and right channel but because of stereo they arent the same i would guess
So my thinking would be that when using mono i have those other 512 values from - lets say - the right channel left for analysis of the whole mono signal.
