scipy.linalg.dft#

scipy.linalg.dft(n, scale=None)[source]#

Discrete Fourier transform matrix.

Create the matrix that computes the discrete Fourier transform of a sequence [1]. The nth primitive root of unity used to generate the matrix is exp(-2*pi*i/n), where i = sqrt(-1).

Parameters:

nint: Size the matrix to create.
scalestr, optional: Must be None, ‘sqrtn’, or ‘n’. If scale is ‘sqrtn’, the matrix is divided by sqrt(n). If scale is ‘n’, the matrix is divided by n. If scale is None (the default), the matrix is not normalized, and the return value is simply the Vandermonde matrix of the roots of unity.

Returns:

m(n, n) ndarray: The DFT matrix.

Notes

When scale is None, multiplying a vector by the matrix returned by dft is mathematically equivalent to (but much less efficient than) the calculation performed by scipy.fft.fft.

Added in version 0.14.0.

References

[1]

“DFT matrix”, https://en.wikipedia.org/wiki/DFT_matrix

Examples

>>> import numpy as np
>>> from scipy.linalg import dft
>>> np.set_printoptions(precision=2, suppress=True)  # for compact output
>>> m = dft(5)
>>> m
array([[ 1.  +0.j  ,  1.  +0.j  ,  1.  +0.j  ,  1.  +0.j  ,  1.  +0.j  ],
       [ 1.  +0.j  ,  0.31-0.95j, -0.81-0.59j, -0.81+0.59j,  0.31+0.95j],
       [ 1.  +0.j  , -0.81-0.59j,  0.31+0.95j,  0.31-0.95j, -0.81+0.59j],
       [ 1.  +0.j  , -0.81+0.59j,  0.31-0.95j,  0.31+0.95j, -0.81-0.59j],
       [ 1.  +0.j  ,  0.31+0.95j, -0.81+0.59j, -0.81-0.59j,  0.31-0.95j]])
>>> x = np.array([1, 2, 3, 0, 3])
>>> m @ x  # Compute the DFT of x
array([ 9.  +0.j  ,  0.12-0.81j, -2.12+3.44j, -2.12-3.44j,  0.12+0.81j])

Verify that m @ x is the same as fft(x).

>>> from scipy.fft import fft
>>> fft(x)     # Same result as m @ x
array([ 9.  +0.j  ,  0.12-0.81j, -2.12+3.44j, -2.12-3.44j,  0.12+0.81j])