Problem with ifftn and ifftshift??

Hi there I was wondering if someone could help me. I am using the fftn fftshift functions as part of my alogrithm;

tmp = fftn(MAT); Adash = fftshift(tmp);

 Adashup=ifftshift(Adash);
 Adashup2new = ifftn(Adashup);

However, When I took everything else out and left these in as a tester I noticed that instead of the above taking me back to the original MAT, it alters the numbers quite a lot. What am I doing wrong?? Example original file MAT has a min of 0 and a max of 1.0856, however the file Adashup2new has a min of -6.792 and a max of 1.0856

[Mehri Mehrnia]

why we should use fftshit?when I dont use fftshit, scale is good and untouched

[Walter Roberson]

@Mehri Mehrnia

fft() returns back the 2-sided fft. By convention, the order of frequency bins returns back from fft() of real input signals is like

0, 1, 2, 3, 4, ... n-1, n, conj(-n), conj(-(n-1)), ... conj(-(4)), conj(-(3)), conj(-(2)), conj(-1)

but people often prefer to visualize it in the order

conj(-(n-1)), ... conj(-(4)), conj(-(3)), conj(-(2)), conj(-1), 0, 1, 2, 3, 4, ... n-1, n

The change is not mathematically significant: it is just easier for people to look at the plots.

Answer 1

[Nitin]

Sounds like Imaginary components are there due to round off errors, try extracting the real part

f = real(ifft2(F));

Answer 2

[Bran]

Thanks so much for the help! I will try your suggestion! :D