The first element of the output of the fft function is the constant or DC offsets (mean values) of the original time domain vector. As a general rule, the second half of the original vector is the flipped complex conjugate of the first half of the vector. For a one-sided Fourier transform, the frequencies go from 0 Hz (DC) to the Nyquist frequency (half the sampling frequency). The sampling frequency (assuming a time vector with uniform increments, so a constant sampling frequency) is the inverse of the common sampling interval.
For a two-sided fft result (after using the fftshift function so that the centre frequency is 0 Hz) the frequency vector is:
Fv = linspace(-Fs/2, Fs/2-Fs/length(s), length(s)); % EVEN 'length(s)' (Asymmetric)
Fv = linspace(-Fs/2, Fs/2, length(s)); % ODD 'length(s)' (Symmetric)
This results in a two-sided Fourier transform with the ‘negative’ frequencies to the left of centre and the positive frequencies to the right of centre.
I generally zero-pad the fft argument using:
NFFT = 2^nextpow2(L);
FTs = fft(s, NFFT)/L;
where ‘L’ is the length of the associated time vector (since the signal can be either a vector or a column-oriented matrix where each column is a separate signal) and ‘s’ is the signal (vector or matrix). This improves the efficiency of the fft calculation, and also makes the subsequent analysis a bit easier.
.
