Table of Contents
Advertisement
Table 31 - gSE (continued)
Parameter
Values
Number of Spectrum Lines
Select from:
• 100
• 200
• 400
• 800
• 1600
FFT Window Type
Select from:
• Rectangular
• Flat top
• Hanning
• Hamming
Number of Averages
Select from:
• 1
• 2
• 3
• 6
• 12
• 23
• 45
• 89
• 178
Comment
Select the number of lines of resolution to be provided in the FFT.
Select the window function to apply in the FFT signal processing.
• FFT Windows Purpose:
FFT Windows are applied to address the problem of signals that occur at frequencies that are not centered within a
frequency bin. In these cases, energy from the signal can be dispersed among adjacent bins such that the amplitude
of neither bin represents the actual magnitude of the signal. For example:
If no window is applied (the Rectangular Window): If the frequency of a signal is precisely centered between bins,
with no other signals present, then the magnitude of each bin is precisely ½ that of the actual signal. When viewing
the FFT, two adjacent bins are presented with equal and comparatively small peak amplitudes, rather than one bin
with 2x that amplitude, which is the actual amplitude of the signal.
As the frequency of the signal moves across a bin, the proportion of its energy that "bleeds" into adjacent bins
changes. For example, if you use a Rectangular Window, and a signal with a constant amplitude moves 50...60 Hz (10
bins), then a Waterfall display shows the bins enlarging as the signal enters the bin. The display grows to a maximum
that is equal to the actual signal amplitude when the signal is centered in the bin, and then shrinks to zero as the
signal moves above the bin.
FFT Windows are used to "smooth" this effect such that the amplitude of the signal, as represented by the amplitude
of the bin that it is in, is better represented. But there are trade-offs as these techniques all tend to make it more
difficult to ascertain the specific frequency of a signal (which bin is it). So when selecting an FFT Window the key is to
understand the intent: Is it more important to know the exact amplitude of the signals that the FFT measures, or is it
more important to know the exact frequencies of the signals within the FFT?
• Available FFT Windows:
Rectangular
– Description: No window is applied
– Other Terms: Normal, Uniform
– Performance: Gives poor peak amplitude accuracy, good peak frequency accuracy.
– Usage: Only for transient signals that die out before the end of the time sample, or for exactly periodic signals
within the time sample (such as integer order frequencies in synchronously sampled data).
Flat Top
– Other Terms: Sinusoidal
– Performance: Gives good peak amplitude accuracy, poor peak frequency accuracy for data with discrete frequency
components.
– Use when amplitude accuracy is more important than frequency resolution. In data with closely spaced peaks, a
Flat Top window can smear the peaks together into one wide peak.
Tip: The Bands FFT is exclusive to the bands function, so is not stored or communicated externally in any way; the Flat
Top FFT Window is recommended to verify the best possible measurement accuracy.
Hanning
– Description: A general-purpose window that is similar to a Hamming window.
– Performance: Gives fair peak amplitude accuracy, fair peak frequency accuracy.
– Usage: It is used on random type data when frequency resolution is more important than amplitude accuracy.
Most often used in predictive maintenance.
Hamming
– Performance: A general-purpose window that is similar to a Hanning window.
– Gives fair peak amplitude accuracy, fair peak frequency accuracy. It provides better frequency resolution but
decreased amplitude accuracy when compared to the Hanning window.
– Usage: Use it to separate close frequency components.
Select the number of averages for the gSE FFT or Time Waveform
• When averaging, the individual gSE FFTs are updated as quickly as possible. The timing of the update is dependent on
the overall processing demands on the module, which is a function of the module configuration and, to some degree,
the circumstance of the moment. This factor, along with the fact that the waveforms are always captured without
respect to an overlap requirement (so always "max overlap"), makes it impossible to define precisely how long (in
time) it takes to acquire any specific number of samples that are used in the averaging.
• Averaging is exponential, meaning that once the specified number of samples has been acquired that the averaged
sample (result) is available upon each subsequent update.
Rockwell Automation Publication 1444-UM001D-EN-P - June 2018
Measurement Definition
(Average TWF on page
320)
Chapter 4
149
Hide quick links:
Advertisement
Table of Contents
