3 spectrum analyzer application, 1 hamming window – Measurement Computing WavePort 312P rev.1.0 User Manual
Page 62

7-2 Spectrum Analyzer
PowerVista/312 User’s Manual
• Input Channel: is the actual input channel (V1, V2, … I3, E1) to the hardware. This routine only uses
results from a single channel, and performs no differential calculations for a line-to-line voltage. If line-
to-line voltage is desired, it is recommended that all other inputs be disconnected and the single line-to-
line connection be made to a differential input.
• Apply Hamming Window: when checked, multiplies the input wave data by a Hamming Window.
This reduces excessive spectrum energy leakage when the Sampling Frequency and Number of Samples
have not been defined to generate an integer multiple of cycles of the input waveshape within the
number of points. See the Hamming Window application discussion for additional information.
• Spectrum Output Units: is either dB or Actual Values. If Actual Values, results are presented on a
linear scale in Volts or Amps according to the Input Channel. If dB, the results are presented on a log
scale where the results are in dB relative to the total input range for the given channel.
For example, if the input voltage fundamental (1
st
harmonic) is 480V for a 1250V peak input range. Its
dB value is 20log( (sqrt(2)*480) / 1250 ) = -5.303. And of course, if the input signal occupies the total
input range, its dB value is 0, as log of 1 is zero.
• Continuous Update: activates the Spectrum Analyzer continuous update feature. When checked, the
routine will capture, process, display and then repeat as fast as the PC will go.
7.3 Spectrum Analyzer Application
The Spectrum Analyzer should be used when higher frequency waveshape and spectral information is desired.
With the added frequency range capability supplied in the Spectrum Analyzer, the routine can be used for both
spectral decomposition as well as a high speed single channel digital scope. Two clarifying application notes
are discussed below.
7.3.1
Hamming Window
The application of the Spectrum Analyzer is fairly straight forward. Wave data information is captured,
processed through a Fast Fourier Transform (FFT), and displayed. This is a simple and elegant process. One
item in the Spectrum Analyzer Dialog that adds a feature which may not be well understood is the Hamming
Window.
The Hamming Window is one of several windowing pre-processes that aids in reducing “leakage” of spectral
energy into adjacent frequencies. This “leakage” or spreading of spectral energy over a broad range of
frequencies is noticed most when the captured data does not have an integer number of cycles within the data
window. Since the FFT generates results that are discrete (results at frequency points determined by the sample
rate and number of sample points), if the signal has frequencies not occurring at these defined frequency points,
then their energy has to be distributed among available frequencies.
For most harmonics capturing devices, only integer harmonics are supplied. This can be done accurately and
simply because, for a 128 point wave of one cycle, exactly one cycle is captured within the power two data
window. This meets the criteria of the FFT where only integer multiples of the base frequency are produced. If
however, 1.23 cycles were processed within the 128 point window, leakage would reign, and resultant peak
magnitudes would not represent the input signals well. The result would generally indicate where frequencies
are located, but energy for each of those frequencies would spread and make accurate magnitudes very difficult
to obtain.
The effect of the Hamming Window is to take energy leakage spread over many frequencies and compress that
energy to be mostly within 3 to 4 frequencies. This is illustrated in
Figures 7-2 and 7-3.
Figure 7-2
shows the
results of a 250 V RMS wave input at 700 Hz. Sampling is set at 1024 points and 50 kHz. Notice that the peak
in the resultant spectrum is 204.34 V and occurs at 695.8 Hz (using the Graphical Result Window digital
cursor). A list of 10 values around this point are also displayed in Table 8, where the root sum square of the
points only yields a total value of 246.10 V, a very poor estimate of the actual input waveform magnitude of 250