rfftip¶

Purpose¶

Computes an inverse real 1- or 2-D FFT. Takes a packed format FFT as input.

Format¶

y = rfftip(x)
Parameters: x (NxK matrix or K-length vector) – data y (LxM real matrix or M-length vector) –

Remarks¶

rfftip() assumes that its input is of the same form as that output by rfftp() and rfftnp().

rfftip() uses the Temperton prime factor FFT algorithm. This algorithm can compute the inverse FFT of any vector or matrix whose dimensions can be expressed as the product of selected prime number factors. GAUSS implements the Temperton algorithm for any integer power of 2, 3, and 5, and one factor of 7. Thus, rfftip() can handle any matrix whose dimensions can be expressed as:

\begin{align}\begin{aligned}2^p \times 3^q \times 5^r \times 7^s\\\begin{split}p, q, r \geq 0\\ s = 0 \text{ or } 1\end{split}\end{aligned}\end{align}

If a dimension of x does not meet this requirement, it will be padded with zeros to the next allowable size before the inverse FFT is computed. Note that rfftip() assumes the length (for vectors) or column dimension (for matrices) of x is $$K-1$$ rather than $$K$$, since the last element or column does not hold FFT information, but the Nyquist frequencies.

The sizes of x and y are related as follows: $$L$$ will be the smallest prime factor product greater than or equal to $$N$$, and $$M$$ will be twice the smallest prime factor product greater than or equal to $$K-1$$. This takes into account the fact that x contains both positive and negative frequencies in the row dimension (matrices only), but only positive frequencies, and those only in the first $$K-1$$ elements or columns, in the length or column dimension.

It is up to the user to guarantee that the input will return a real result. If in doubt, use ffti(). Note, however, that ffti() expects a full FFT, including negative frequency information, for input.

Do not pass rfftip() the output from rfft() or rfftn() - it will return incorrect results. Use rffti() with those routines.