Fast Fourier Transforms
TDS 620A, 640A, & 644A User Manual
Window Characteristics — When evaluating a window for use, you may
want to examine how it modifies the FFT time domain data. Figure 3-28
shows each window, its bandpass characteristic, bandwidth, and highest side
lobe. Consider the following characteristics:
The narrower the central lobe for a given window, the better it can resolve
a frequency.
The lower the lobes on the side of each central lobe are, the better the
amplitude accuracy of the frequency measured in the FFT using that
window.
Narrow lobes increase frequency resolution because they are more
selective. Lower side lobe amplitudes increases accuracy because they
reduce leakage.
Leakage results when the FFT time domain waveform delivered to the
FFT function contains a non-integer number of waveform cycles. Since
there are fractions of cycles in such records, there are discontinuities at
the ends of the record. These discontinuities cause energy from each
discrete frequency to “leak” over on to adjacent frequencies. The result is
amplitude error when measuring those frequencies.
The rectangular window does not modify the waveform record points; it
generally gives the best frequency resolution because it results in the most
narrow lobe width in the FFT output record. If the time domain records you
measured always had an integer number of cycles, you would only need this
window.
Hamming, Hanning, and Blackman-Harris are all somewhat bell-shaped
widows that taper the waveform record at the record ends. The Hanning and
Blackman/Harris windows taper the data at the end of the record to zero;
therefore, they are generally better choices to eliminate leakage.