limitations in FFT in matlab?

[Piierre]

I have a question regarding using FFT i MATLAB

When I make an FFT of a small number of periods of a signal the frequencys are moved up to a larger value, for example a signal with the frequency 3Hz are shown in the FFT plot as 3.5Hz and so on. If I increase the number of periods used to make the FFT the value closes in to the real one. It seems like the offset has an exponential relation to the number of periods
Anybody knows why this is? Is it rounding in Matlab or limitations in the FFT algorithm or what?

 

[wpchan05]

I have a question regarding using FFT i MATLAB

When I make an FFT of a small number of periods of a signal the frequencys are moved up to a larger value, for example a signal with the frequency 3Hz are shown in the FFT plot as 3.5Hz and so on. If I increase the number of periods used to make the FFT the value closes in to the real one. It seems like the offset has an exponential relation to the number of periods
Anybody knows why this is? Is it rounding in Matlab or limitations in the FFT algorithm or what?

I guess the total time of your time simulation gives you the lowest bound of your frequency signal that you can observe in fft. Eg if you run 1 sec, you get 1Hz. If you run 10 sec, you get 0.1 Hz, etc

Correct me if I am wrong

 

[Piierre]

well that is correct but what has it got to do with my question?

 

[wpchan05]

well that is correct but what has it got to do with my question?

the time to compute the freq spectrum gives you the lowest bound of freq, and this is nothing to correlate matlab's computation power. It is the loss of accuracy when you map a continuous signal by a discrete signal.