104 lines
2.7 KiB
Text
104 lines
2.7 KiB
Text
|
#lang scribble/manual
|
||
|
|
@(require (for-label racket))
|
||
|
|
|
||
|
|
@title{UnpackFFT}
|
||
|
|
Unpack an FFT chain into separate demand-rate FFT bin streams@section{categories}
|
||
|
|
UGens>FFT
|
||
|
|
@section{related}
|
||
|
|
Classes/PackFFT, Classes/Unpack1FFT
|
||
|
|
|
||
|
|
@section{description}
|
||
|
|
|
||
|
|
Takes an FFT chain and separates the magnitude and phase data into separate demand-rate streams, for arithmetic manipulation etc.
|
||
|
|
|
||
|
|
This is technically a demand-rate UGen. The actual "demand" is usually created by PackFFT later on in the graph, which requests the values in order to re-pack the data. This allows for processing to occur in between...
|
||
|
|
|
||
|
|
See also pvcollect, pvcalc and pvcalc2 methods ( in link::Classes/PV_ChainUGen:: ) which provide convenient ways to process audio in the frequency domain. The help for pvcollect includes notes on efficiency considerations.
|
||
|
|
|
||
|
|
@section{classmethods}
|
||
|
|
|
||
|
|
@section{private}
|
||
|
|
categories
|
||
|
|
|
||
|
|
@section{method}
|
||
|
|
new
|
||
|
|
@section{argument}
|
||
|
|
chain
|
||
|
|
FFT chain
|
||
|
|
@section{argument}
|
||
|
|
bufsize
|
||
|
|
FFT buffer size
|
||
|
|
@section{argument}
|
||
|
|
frombin
|
||
|
|
limiting analysis to the bins of interest
|
||
|
|
@section{argument}
|
||
|
|
tobin
|
||
|
|
limiting analysis to the bins of interest
|
||
|
|
@section{returns}
|
||
|
|
|
||
|
|
A list from DC up to Nyquist of
|
||
|
|
@racketblock[ [mag[0], phase[0], mag[1], phase[1], ... mag[nyquist], phase[nyquist]]. ::
|
||
|
|
]
|
||
|
|
@section{discussion}
|
||
|
|
|
||
|
|
Note that you do have to decide your FFT buffer size in advance, since this determines how many values the UGen will output.
|
||
|
|
|
||
|
|
@racketblock[
|
||
|
|
#magsphases = UnpackFFT(chain, bufsize)
|
||
|
|
::
|
||
|
|
|
||
|
|
]
|
||
|
|
@section{examples}
|
||
|
|
|
||
|
|
|
||
|
|
@racketblock[
|
||
|
|
(
|
||
|
|
s.waitForBoot({
|
||
|
|
var fftsize = 1024;
|
||
|
|
b = Buffer.alloc(s, fftsize, 1);
|
||
|
|
c = Buffer.read(s, Platform.resourceDir +/+ "sounds/a11wlk01.wav");
|
||
|
|
})
|
||
|
|
)
|
||
|
|
|
||
|
|
// This one just drags out various the values and posts them - a little bit pointless!
|
||
|
|
(
|
||
|
|
x = {
|
||
|
|
var sig, chain, unp;
|
||
|
|
sig = SinOsc.ar;
|
||
|
|
sig = PlayBuf.ar(1, c, BufRateScale.kr(c), loop: 1);
|
||
|
|
chain = FFT(b, sig);
|
||
|
|
|
||
|
|
// Using the frombin & tobin args makes it much more efficient, limiting analysis to the bins of interest
|
||
|
|
unp = UnpackFFT(chain, b.numFrames, frombin: 0, tobin: 4);
|
||
|
|
|
||
|
|
// Demand some data from the unpacker.
|
||
|
|
// NOTE: At present, Demand.kr is unable to handle more than 32 inputs,
|
||
|
|
// so using frombin & tobin to limit the number of bins is compulsory.
|
||
|
|
Demand.kr(chain>=0, 0, unp).collect{|anunp, index|
|
||
|
|
anunp.poll(chain>=0, if(index % 2 == 0, "Magnitude", "Phase")+(index/2).floor);
|
||
|
|
};
|
||
|
|
|
||
|
|
(sig*0.1).dup;
|
||
|
|
}.play(s);
|
||
|
|
)
|
||
|
|
x.free;
|
||
|
|
|
||
|
|
// Now a simple frequency-domain manipulation, square-rooting the magnitudes AND phases.
|
||
|
|
(
|
||
|
|
x = {
|
||
|
|
var in, chain, magsphases;
|
||
|
|
in = PlayBuf.ar(1, c, BufRateScale.kr(c), loop: 1);
|
||
|
|
chain = FFT(b, in);
|
||
|
|
magsphases = UnpackFFT(chain, b.numFrames);
|
||
|
|
magsphases = magsphases.collect(_.sqrt);
|
||
|
|
PackFFT(chain, b.numFrames, magsphases);
|
||
|
|
Out.ar(0, 0.25 * IFFT(chain).dup);
|
||
|
|
}.play(s);
|
||
|
|
)
|
||
|
|
x.free;
|
||
|
|
::
|
||
|
|
|
||
|
|
]
|
||
|
|
|
||
|
|
|