Figure 18 shows a comparison between different windows. Specifically, this shows the Discrete Time Fourier Transform of a cosine with various windows applied. Note that this is not a DFT or FFT. In other words if you do the following signal = cos(t); x = fft(hann(n) .* signal); plot(x); in matlab, you will not get these plots.
There are several things to consider about this plot. One way to look at it is to thing of each bin in an FFT as a bandpass filter. Then this plot shows the frequency response of the filter. So, pretend that you are looking at exactly one and only one FFT bin. If your input is a unit amplitude sine wave centered exactly at the bin frequency, your FFT bin has amplitude 1 for any window. That's what you'd expect. And if you have a sine wave at, say, exactly 9 bins away from your bin frequency, your FFT bin shows basically 0 amplitude for any window. Good.
Now, what if you have a sine wave centered exactly 1/2 a bin away? Or 3.5 bins away. You will get response in your FFT. How much depends on which window you choose. Which window is ``best'' depends on what you are trying to do. In the next sections we talk about a few different applications, and what windows work best.
http://en.wikipedia.org/wiki/Window_function has a similar comparison for many more windows.
|
