National Instruments NI Spectral Measurements Toolkit - Manual (Page 24)

NI Spectral Measurements Toolkit User Guide 24 ni.com
Spectral Domain Averaging
Averaging is an important part of spectrum-domain measurements because
of the effects of noise on a signal and its spectrum. The Spectral
Measurements Toolkit includes averaging VIs that average several
records of data to reduce the noise effects. You can use the three different
averaging types: vector, RMS, and peak-hold.
Vector averaging lowers the noise floor while retaining the signal
spectrum. In the time domain, a running average reduces the effect of
zero-mean white noise on a signal. The noise is averaged out while the
signal is retained. The signal must be triggered, meaning that each data
record starts at a consistent point in the periodic signal, preserving the
signal integrity during an averaging process. Because the FFT is a linear
transform, averaging spectral records in the frequency domain is equivalent
to averaging data records in the time domain. The signal must be triggered
for vector averaging to work properly. Vector averaging requires a complex
spectrum and produces a complex result that you can convert to a real
power spectrum.
If the signal is not triggered in the time domain, phase noise appears in the
resulting spectrum. You can use RMS averaging to eliminate the effect of
phase noise. The magnitude of the spectrum is independent of time shifts
of the input signal, but the phase can change dramatically with each data
record. If you average the power spectra and take the square root of the
result, you eliminate the effect of phase variations. You can no longer
reduce the noise floor, but you can reduce the magnitude variance of the
noise. Reducing the noise variance helps to distinguish small frequency
peaks from the largest noise peaks. RMS averaging eliminates all phase
information and returns a real spectrum. If the averaging process
returns results in a complex data type, the imaginary portion is zero.
Peak-hold averaging refers to a method of retaining the maximum
magnitude value of every frequency bin over several data records.
Peak-hold averaging is most useful for capturing transient phenomena
that do not appear in individual spectra. In a monitoring application, the
peak-hold display allows an operator to tell at a glance if a transient at
a certain frequency occurred since the last reset. However, peak-hold
averaging cannot specify when the transient happened. Like RMS
averaging, peak-hold averaging results in a real spectrum.
When you apply a zoom FFT VI to a signal, you receive the complex
FFT spectrum. The spectrum domain averaging functions can operate on
the FFT spectrum to return different types of spectra, such as averaged
FFT spectrum, power spectrum, cross spectrum, and frequency response.