FFT842(3)

NAME

     fft842, newfft842, trig_init:  Functions to perform  a  fast
     Fourier transform on a one-dimensional array of data.

Syntax

     #include <soi_util.h>

     int fft842 (in, n, x, y)
     int in, n;
     float *x, *y;

     void newfft842 (in, n, Wr, Wi, x, y)
     int in, n;
     double *Wr, *Wi;
     float *x, *y;

     int trig_init (int n, double *Wr, double *Wi)
     int n;
     double *Wr, double *Wi;

DESCRIPTION

     fft842() replaces the vector z  =  x  +  iy  by  its  finite
     discrete,  complex Fourier transform if in = 0.  The inverse
     transform is calculated for in = 1.   It  performs  as  many
     base-8  iterations  as  possible  and  then  finishes with a
     base-4 iteration or a base-2 iteration, if needed.

     The integer n should be a power of 2, and x and  y  pointers
     to  arrays  of  n  floats  each.   If  n is non-positive, no
     transform is performed and the function returns -1.  If n is
     not a power of 2, only the values of x and y up to the larg-
     est power of 2 smaller than n are transformed and the  func-
     tion  returns  1.   Otherwise,  the function returns 0.  The
     function arguments x and y must point  to  float  arrays  of
     length  n  representing the real and imaginary parts respec-
     tively of the original data on input and transformed data on
     return.

     The function newfft842()  behaves  similarly,  but  must  be
     called  with  the precomputed array of complex trigonometric
     "twiddle factors" Wr + i * Wi.  This array, of length 7n/8 -
     6, is computed with a call to trig_init.

BUGS

     The duplicate  functions  fft842  and  newfft842  should  be
     merged;  the  time  saving in precomputing the trigonometric
     factors is seldom worth the effort.

     It is dangerous to call the functions with arrays of  length
     other than a power of 2.

SEE ALSO

     fft2d(3), rowfft(3)

AUTHORS

     Lyle Bacon, Rick Bogart

HISTORY

     1994-11-10     SOI Version 0.9