Do you have to have an FFT to do an IFFT?

I know that you can do a Fast Fourier Transformation from time to frequency domain and then back to the time domain with an Inverse Fast Fourier Transformation, but is it possible to just have an Inverse Fast Fourier Transformation converting from the frequency to the time domain without initially having the Fast Fourier Transformation?

Answer 1

[David Sanchez]

You do not have to have used fft before using ifft. The following will work:

x=rand(100,1);
b=ifft(x);

The subject of what you get at the end is another issue. To obtain anything coherent, your initial data should be data in frequency domain.

[Joe]

Thank you.

Answer 2

[Walter Roberson]

Yes. ifft() works on any frequency data, no matter where it came from. Just be careful that the frequency data is not stored in shifted format (with 0 in the center)

For example you can find out how a 50 Hz and 50.1 Hz tone "beat" together by generating an array of zeros that is long enough, setting the bins for 50 Hz and 50.1 Hz to the same value (remember to set for both half-arrays), and then ifft() to see what the result is over time.

[Joe]

Ok thanks. I think I have a more specific question, but I'll probably ask it in a new thread with my code.