Measurement Computing eZ-Record rev.2.1 - Manual (Page 20)

20
eZ-Record Manual January 2001
Miscellaneous Tab - FFT Window (Response) Panel
The FFT (Response) window is a time-domain weighting window. A response
window is usually applied to data to reduce FFT leakage errors. FFT theory
assumes that the signal being analyzed is periodic in the data acquisition block.
When this is not the case, energy from a signal at a specified frequency can leak
into nearby spectral bins causing spectral amplitude inaccuracies. Applying a
windowing function controls, but doesn’t completely eliminate, the error by
multiplying each data frame by a suitable time-domain weighting window. This
calculation reduces the amplitude/magnitude of the data near the ends of each
data frame prior to performing the FFT and forces the data to be nearly periodic in
the window, thus reducing leakage errors. Response window options are:
None: No weighting window is applied.
Hanning: The Hanning window is typically used to analyze continuous
signals. It offers a reasonable trade-off of frequency accuracy versus
amplitude accuracy.
FlatTop: Compared to the very similar 4-term “Max Flat Top,” this window also
has a very low peak amplitude error, and its frequency resolution is
somewhat better. Its side lobes are considerably higher. Its effective noise
bandwidth is still almost twice that of the Hanning window, therefore this
window is used mainly to measure accurate peak amplitudes of discrete
spectral components that are known to be separated by at least several
spectral lines.
Blackman-Harris: This window function was designed to provide the
minimum side lobe level of any three-term window. Compared with the very
similar Hanning window, it has a slightly wider main lobe but much better
dynamic range. This window has the smallest 60 dB bandwidth of any
window listed. The Blackman-Harris window may be preferred over the
Hanning for measurements requiring better dynamic range.
Exponential: An exponential weighting window is equal to 1.0 at the beginning
of the block and decays exponentially to a smaller value at the end of the
block. Exponential is used only with transient data that is captured with
pre-trigger to assure that the initial values in all data channels are very