72 lines
1.9 KiB
Text
72 lines
1.9 KiB
Text
|
#lang scribble/manual
|
||
|
|
@(require (for-label racket))
|
||
|
|
|
||
|
|
@title{IFFT}
|
||
|
|
Inverse Fast Fourier Transform@section{related}
|
||
|
|
Classes/FFT, Guides/FFT-Overview
|
||
|
|
@section{categories}
|
||
|
|
UGens>FFT
|
||
|
|
|
||
|
|
|
||
|
|
@section{description}
|
||
|
|
|
||
|
|
The inverse fast fourier transform converts from frequency content to a
|
||
|
|
signal.
|
||
|
|
|
||
|
|
The fast fourier transform analyzes the frequency content of a signal. The IFFT UGen converts this emphasis::frequency-domain:: information back into emphasis::time-domain:: audio data. Most often this is used as the end of a process which begins with link::Classes/FFT::, followed by frequency-domain processing using PV (phase-vocoder) UGens, followed by IFFT.
|
||
|
|
|
||
|
|
@section{classmethods}
|
||
|
|
|
||
|
|
|
||
|
|
@section{method}
|
||
|
|
new, ar, kr
|
||
|
|
returns a time domain signal from converting the FFT frequency domain signal chain. The *new method is equivalent to the *ar message returns an audio rate signal.
|
||
|
|
|
||
|
|
@section{argument}
|
||
|
|
buffer
|
||
|
|
|
||
|
|
The FFT "chain" signal coming originally from an FFT UGen, perhaps via other PV UGens.
|
||
|
|
|
||
|
|
@section{argument}
|
||
|
|
wintype
|
||
|
|
Defines how the data is windowed:
|
||
|
|
@section{table}
|
||
|
|
|
||
|
|
## -1 || strong::rectangular:: windowing, simple but typically not recommended;
|
||
|
|
## 0 || (the default) strong::Sine:: windowing, typically recommended for phase-vocoder work;
|
||
|
|
## 1 || strong::Hann:: windowing, typically recommended for analysis work.
|
||
|
|
::
|
||
|
|
|
||
|
|
@section{argument}
|
||
|
|
winsize
|
||
|
|
Can be used to account for zero-padding, in the same way as the link::Classes/FFT:: UGen.
|
||
|
|
|
||
|
|
@section{returns}
|
||
|
|
|
||
|
|
The emphasis::time-domain:: audio signal.
|
||
|
|
|
||
|
|
@section{discussion}
|
||
|
|
|
||
|
|
The IFFT UGen converts the FFT data in-place (in the original FFT buffer) and overlap-adds the result to produce a continuous signal at its output.
|
||
|
|
|
||
|
|
@section{Examples}
|
||
|
|
|
||
|
|
|
||
|
|
|
||
|
|
@racketblock[
|
||
|
|
|
||
|
|
// without any modification, convert FFT chain (frequency domain signal) back to audio (time domain signal)
|
||
|
|
(
|
||
|
|
{ var in, chain;
|
||
|
|
in = WhiteNoise.ar;
|
||
|
|
chain = FFT(LocalBuf(2048), in);
|
||
|
|
IFFT.ar(chain) * -20.dbamp // inverse FFT
|
||
|
|
}.play;
|
||
|
|
)
|
||
|
|
|
||
|
|
::
|
||
|
|
|
||
|
|
]
|
||
|
|
|
||
|
|
|