Problems with ifft in general

I am having great difficulty using ifft. Basically when I use it on my original spectrum I do indeed get back to my original signal in the time domain, however, when I 'clean up' my spectra (Taking out all peaks from noise) and then apply ifft I get back to a singal that makes no sense in the time domain. Any idea what could be going wrong.

[Honglei Chen]

It's hard to judge without context, could you post some figures or code snippet?

Answer 1

[Image Analyst]

Do you know what spikes in the Fourier domain mean? It means that there is periodic information back in the time domain. Whether that periodic structure in the signal or image is noise or not depends on what you want your image to look like.

Usually people post their image or a screenshot of their signal or something to help us visualize what you're talking about. You have not posted anything so you cannot get much help yet until you do.

[Bran]

Hi there both! Man apologies for not supplying enough information! So OK my signal is of someone running and so the spectra originally looked like it does in the first image. So I decided to look at the intensities of the fundamental and harmonics. I then accepted those over 186 and so i was left with just four peaks and I 'zero' everything else. Then on this data I used ifft but I do not get back to my original signal or anything similar really. I get the second image attached.

Really apologies for such little info and thanks guy for pointing it out :D

Answer 2

[Bran]

Hi there both! Man apologies for not supplying enough information! So OK my signal is of someone running and so the spectra originally looked like it does in the first image. So I decided to look at the intensities of the fundamental and harmonics. I then accepted those over 186 and so i was left with just four peaks and I 'zero' everything else. Then on this data I used ifft but I do not get back to my original signal or anything similar really. I get the second image attached.

Really apologies for such little info and thanks guy for pointing it out :D

[Star Strider]

What does your original signal look like?

[Bran]

It is meant to look like this. One thing that came to mind was Gibbs??