Fast Fourier Transforms by using FFTW Ver.3.

Takeshi Horinouchi

(C) Takeshi Horinouchi / GFD Dennou Club, 2003

NO WARRANTY

- Uses NArray.
- Multi-dimensional complex FFT. (Real data are coerced to complex).
- Supports both double and single float transforms.
- Not normalized as in FFTW

- Sine / cosine transforms
- User choice of optimization levels (i.e., FFTW_MEASURE etc in addition to FFTW_ESTIMATE).
- Multi-threaded FFT3 support -- don't know whether it's really feasible.

- Install FFTW Ver.3.
- NOTE: To activate the single-float transform, you have to install FFTW3 with the single-float compilation, in addition to the default double-float version. This can be done by configuring FFTW3 with the --enable-float option, and install it again. The single-float version will coexist with the double-float version. If you do not install the single-float version, FFT is always done with the double precision, which is not bad if you are not time- and memory-conscious.

- Install NArray.
Then, install this library as follows (replace "version" with the actual version number):

% tar xvzf fftw3-version.tar.gz % cd fftw3-version % ruby extconf.rb % make % make site-install

Or

% make install

(If you are using Ruby 1.8, make install is the same make site-install.)

See the following peice of code. (Install this library and copy and paste the following to the interactive shell irb).

require "numru/fftw3" include NumRu na = NArray.float(8,6) # float -> will be corced to complex na[1,1]=1 # <example 1> fc = FFTW3.fft(na, -1)/na.length # forward 2D FFT and normalization nc = FFTW3.fft(fc, 1) # backward 2D FFT (complex) --> nb = nc.real # should be equal to na except round errors # <example 2> fc = FFTW3.fft(na, -1, 0) / na.shape[0] # forward FFT with the first dim # <example 3> fc = FFTW3.fft(na, -1, 1) / na.shape[1] # forward FFT with the second dim

`fft(`

`narray`,`dir`[,`dim`,`dim`,...])-
Complex FFT.

The 3rd, 4th,... arguments are optional.

ARGUMENTS

- narray (NArray or NArray-compatible Array) : array to be transformed. If real, coerced to complex before transformation. If narray is single-precision and the single-precision version of FFTW3 is installed (before installing this module), this method does a single-precision transform. Otherwise, a double-precision transform is used.
- dir (-1 or 1) : forward transform if -1; backward transform if 1.
- optional 3rd, 4th,... arguments (Integer) : Specifies dimensions to apply FFT. For example, if 0, the first dimension is transformed (1D FFT); If -1, the last dimension is used (1D FFT); If 0,2,4, the first, third, and fifth dimensions are transformed (3D FFT); If entirely omitted, ALL DIMENSIONS ARE SUBJECT TO FFT, so 3D FFT is done with a 3D array.

RETURN VALUE

- a complex NArray

NOTE

- As in FFTW, return value is NOT normalized. Thus, a consecutive forward and backward transform would multiply the size of data used for transform. You can normalize, for example, the forward transform FFTW.fft(narray, -1, 0, 1) (FFT regarding the first (dim 0) & second (dim 1) dimensions) by dividing with (narray.shape[0]*narray.shape[1]). Likewise, the result of FFTW.fft(narray, -1) (FFT for all dimensions) can be normalized by narray.length.