Purpose
To compute the discrete Fourier transform, or inverse Fourier transform, of a real signal.Specification
      SUBROUTINE DG01ND( INDI, N, XR, XI, INFO )
C     .. Scalar Arguments ..
      CHARACTER         INDI
      INTEGER           INFO, N
C     .. Array Arguments ..
      DOUBLE PRECISION  XI(*), XR(*)
Arguments
Mode Parameters
  INDI    CHARACTER*1
          Indicates whether a Fourier transform or inverse Fourier
          transform is to be performed as follows:
          = 'D':  (Direct) Fourier transform;
          = 'I':  Inverse Fourier transform.
Input/Output Parameters
  N       (input) INTEGER
          Half the number of real samples.  N must be a power of 2.
          N >= 2.
  XR      (input/output) DOUBLE PRECISION array, dimension (N+1)
          On entry with INDI = 'D', the first N elements of this
          array must contain the odd part of the input signal; for
          example, XR(I) = A(2*I-1) for I = 1,2,...,N.
          On entry with INDI = 'I', the first N+1 elements of this
          array must contain the the real part of the input discrete
          Fourier transform (computed, for instance, by a previous
          call of the routine).
          On exit with INDI = 'D', the first N+1 elements of this
          array contain the real part of the output signal, that is
          of the computed discrete Fourier transform.
          On exit with INDI = 'I', the first N elements of this
          array contain the odd part of the output signal, that is
          of the computed inverse discrete Fourier transform.
  XI      (input/output) DOUBLE PRECISION array, dimension (N+1)
          On entry with INDI = 'D', the first N elements of this
          array must contain the even part of the input signal; for
          example, XI(I) = A(2*I) for I = 1,2,...,N.
          On entry with INDI = 'I', the first N+1 elements of this
          array must contain the the imaginary part of the input
          discrete Fourier transform (computed, for instance, by a
          previous call of the routine).
          On exit with INDI = 'D', the first N+1 elements of this
          array contain the imaginary part of the output signal,
          that is of the computed discrete Fourier transform.
          On exit with INDI = 'I', the first N elements of this
          array contain the even part of the output signal, that is
          of the computed inverse discrete Fourier transform.
Error Indicator
  INFO    INTEGER
          = 0:  successful exit;
          < 0:  if INFO = -i, the i-th argument had an illegal
                value.
Method
  Let A(1),....,A(2*N) be a real signal of 2*N samples. Then the
  first N+1 samples of the discrete Fourier transform of this signal
  are given by the formula:
               2*N           ((m-1)*(i-1))
       FA(m) = SUM ( A(i) * W              ),
               i=1
                                               2
  where m = 1,2,...,N+1, W = exp(-pi*j/N) and j = -1.
  This transform can be computed as follows. First, transform A(i),
  i = 1,2,...,2*N, into the complex signal Z(i) = (X(i),Y(i)),
  i = 1,2,...,N. That is, X(i) = A(2*i-1) and Y(i) = A(2*i). Next,
  perform a discrete Fourier transform on Z(i) by calling SLICOT
  Library routine DG01MD. This gives a new complex signal FZ(k),
  such that
                N            ((k-1)*(i-1))
       FZ(k) = SUM ( Z(i) * V              ),
               i=1
  where k = 1,2,...,N, V = exp(-2*pi*j/N).  Using the values of
  FZ(k), the components of the discrete Fourier transform FA can be
  computed by simple linear relations, implemented in the DG01NY
  subroutine.
  Finally, let
       XR(k) = Re(FZ(k)), XI(k) = Im(FZ(k)),   k = 1,2,...,N,
  be the contents of the arrays XR and XI on entry to DG01NY with
  INDI = 'D', then on exit XR and XI contain the real and imaginary
  parts of the Fourier transform of the original real signal A.
  That is,
       XR(m) = Re(FA(m)),  XI(m) = Im(FA(m)),
  where m = 1,2,...,N+1.
  If INDI = 'I', then the routine evaluates the inverse Fourier
  transform of a complex signal which may itself be the discrete
  Fourier transform of a real signal.
  Let FA(m), m = 1,2,...,2*N, denote the full discrete Fourier
  transform of a real signal A(i), i=1,2,...,2*N. The relationship
  between FA and A is given by the formula:
              2*N            ((m-1)*(i-1))
       A(i) = SUM ( FA(m) * W              ),
              m=1
  where W = exp(pi*j/N).
  Let
       XR(m) = Re(FA(m)) and XI(m) = Im(FA(m)) for m = 1,2,...,N+1,
  be the contents of the arrays XR and XI on entry to the routine
  DG01NY with INDI = 'I', then on exit the first N samples of the
  complex signal FZ are returned in XR and XI such that
       XR(k) = Re(FZ(k)), XI(k) = Im(FZ(k)) and k = 1,2,...,N.
  Next, an inverse Fourier transform is performed on FZ (e.g. by
  calling SLICOT Library routine DG01MD), to give the complex signal
  Z, whose i-th component is given by the formula:
               N             ((k-1)*(i-1))
       Z(i) = SUM ( FZ(k) * V              ),
              k=1
  where i = 1,2,...,N and V = exp(2*pi*j/N).
  Finally, the 2*N samples of the real signal A can then be obtained
  directly from Z. That is,
       A(2*i-1) = Re(Z(i)) and A(2*i) = Im(Z(i)), for i = 1,2,...N.
  Note that a discrete Fourier transform, followed by an inverse
  transform will result in a signal which is a factor 2*N larger
  than the original input signal.
References
  [1] Rabiner, L.R. and Rader, C.M.
      Digital Signal Processing.
      IEEE Press, 1972.
Numerical Aspects
The algorithm requires 0( N*log(N) ) operations.Further Comments
NoneExample
Program Text
*     DG01ND EXAMPLE PROGRAM TEXT
*     Copyright (c) 2002-2010 NICONET e.V.
*
*     .. Parameters ..
      INTEGER          NIN, NOUT
      PARAMETER        ( NIN = 5, NOUT = 6 )
      INTEGER          NMAX
      PARAMETER        ( NMAX = 128 )
*     .. Local Scalars ..
      INTEGER          I, IEND, INFO, N
      CHARACTER*1      INDI
*     .. Local Arrays ..
      DOUBLE PRECISION A(2*NMAX), XI(NMAX+1), XR(NMAX+1)
*     .. External Functions ..
      LOGICAL          LSAME
      EXTERNAL         LSAME
*     .. External Subroutines ..
      EXTERNAL         DG01ND
*     .. Executable Statements ..
*
      WRITE ( NOUT, FMT = 99999 )
*     Skip the heading in the data file and read the data.
      READ ( NIN, FMT = '()' )
      READ ( NIN, FMT = * ) N, INDI
      IF ( N.LE.0 .OR. N.GT.NMAX ) THEN
         WRITE ( NOUT, FMT = 99995 ) N
      ELSE
         READ ( NIN, FMT = * ) ( A(I), I = 1,2*N )
*        Copy the odd and even parts of A into XR and XI respectively.
         DO 20 I = 1, N
            XR(I) = A(2*I-1)
            XI(I) = A(2*I)
   20    CONTINUE
*        Find the Fourier transform of the given real signal.
         CALL DG01ND( INDI, N, XR, XI, INFO )
         IF ( INFO.NE.0 ) THEN
            WRITE ( NOUT, FMT = 99998 ) INFO
         ELSE
            WRITE ( NOUT, FMT = 99997 )
            IEND = N
            IF ( LSAME( INDI, 'D' ) ) IEND = N + 1
            DO 40 I = 1, IEND
               WRITE ( NOUT, FMT = 99996 ) I, XR(I), XI(I)
   40       CONTINUE
         END IF
      END IF
      STOP
*
99999 FORMAT (' DG01ND EXAMPLE PROGRAM RESULTS',/1X)
99998 FORMAT (' INFO on exit from DG01ND = ',I2)
99997 FORMAT (' Components of Fourier transform are',//'   i',6X,
     $       'XR(i)',6X,'XI(i)',/)
99996 FORMAT (I4,3X,F8.4,3X,F8.4)
99995 FORMAT (/' N is out of range.',/' N = ',I5)
      END
Program Data
DG01ND EXAMPLE PROGRAM DATA 8 D -0.1862 0.1288 0.3948 0.0671 0.6788 -0.2417 0.1861 0.8875 0.7254 0.9380 0.5815 -0.2682 0.4904 0.9312 -0.9599 -0.3116Program Results
DG01ND EXAMPLE PROGRAM RESULTS Components of Fourier transform are i XR(i) XI(i) 1 4.0420 0.0000 2 -3.1322 -0.2421 3 0.1862 -1.4675 4 -2.1312 -1.1707 5 1.5059 -1.3815 6 2.1927 -0.1908 7 -1.4462 2.0327 8 -0.5757 1.4914 9 -0.2202 0.0000
Click here to get a compressed (gzip) tar file containing the source code of the routine, the example program, data, documentation, and related files.
Return to index