Spectral analysis defaults
   SIGVIEW Help Home

Most of spectral analysis tools in SIGVIEW are based on the FFT algorithm. There are many different parameters which can be applied to the signal before the FFT analysis is performed, to the FFT calculation itself or its result.

If you want to change the default settings for spectral analysis calculation in SIGVIEW, use Signal tools/Spectral analysis defaults option from the main menu. These settings will be applied to all new FFT-based calculations, including FFT, Time-FFT, cross spectral analysis, etc.

Once the FFT or Time FFT is already calculated according to your current spectral analysis defaults, you can edit those anytime by choosing Edit/Properties menu option for calculated window (FFT, Time FFT,…). The same option is also available in the windows context menu.
These settings will apply only to that single window.

graphic

The following parameters can be changed for each FFT based spectral analysis operation in SIGVIEW:

Subtract mean check box: Select it if you want to normalize the signal before processing. It simply subtracts the mean value of the signal from each sample

Remove linear trend check box: If linear trend appears in signal ( i.e. the whole signal raises or falls monotonously), it can affect evaluation of low frequency components in the FFT. This option removes linear trend by subtracting linear least squares approximation from the signal

Remove values > : This option automatically removes values from the signal that are not in range ( mean -N*StDev , mean + N*StDev ) where StDev is a standard deviation of the signal and N is a user-defined coefficient entered in the corresponding edit box. Values are removed by replacing them with the mean value.

Apply window: To avoid some undesirable effects of discrete Fourier transform, it is recommended to lower the signal values near the end of the signal by multiplying it with some appropriate weighting function. This technique is called “windowing”. SIGVIEW supports several standard weight functions: Hann, Haming, Blackman, Triangle, Tukey. Choosing “Rectangle” window has the same effect as turning windowing option off.

Zero padding: Due to the nature of the FFT algorithm, the fastest calculation will be performed for signals having power-of-2 length (for example 128, 256, 512,…). If your signal or the part you would like to analyze have some other length, there are several options you can choose:

·     Never use zero padding: FFT or slower DFT algorithm will be applied on your signal without altering or extending it with zeros. This is the slowest calculation method, but the advantage is that your signal is not changed in any way. For prime number signal lengths the calculation will be performed by using very slow DFT algorithm.
·     Optimal method: Your signal will be expanded with zeros to the next possible length allowing the usage of the FFT algorithm instead of very slow DFT.
·     Next power of 2: Your signal will be expanded with zeros until the next power-of-2 length. This will enable the usage of fastest FFT algorithm.

To increase the precision of the spectral analysis without taking longer signal segments, you can also expand existing signal segment with zeros, even beyond the next power-of-2 length. That will not add essentially new information in the FFT result, but will increase its precision, i.e. will reduce the size of the one frequency bin. You will simply get FFT result with more points -  from the same signal segment. (If signal has 1024 samples and you choose expanding by factor 8, resulting spectrum will have 4096 pt. instead of 512 without using option). The result will be comparable to interpolation of the normal FFT result. To use this feature, you can choose 2x, 4x or 8x zero padding option.

Spectrum type:

·     Instantaneous spectrum (no averaging): Each change in the signal will cause the FFT to recalculate. This is the default behavior.

·     Average last X spectrum results: This option is useful only for the FFT of the live input signal or other fast changing signals. Last X complex FFTs will be averaged to calculate the final result. That will decrease the influence of the noise in the signal and is usually recommended for all spectral analysis functions on the live signal.

Show result as:  After applying the FFT algorithm to the signal, the result will be an array of complex numbers. These radio-buttons determine the way these complex numbers will be converted to real values and displayed as FFT graphics.

·     Magnitude: value generally regarded as “spectrum”; calculated as sqrt(R^2 + I^2)
·     Power spectrum: Magnitude squared.
·     Power spectral density (PSD): Power spectrum is calculated first and its values are divided with the width of the frequency bin. This “normalization” makes comparisons between FFT sequences from different signals with different sampling rates possible.
·     Phase: show signal phase angle (in degrees)
·     Real & Imaginary part:  show only real or only imaginary part of the spectrum

Apply custom filter curve: You can apply custom filter curve to change the values of the spectrum. See Custom filter curves section for more information. This option is applicable only to magnitude and power spectrum result types.

Logarithmic Y-Axis (dB units) option can be applied only to Magnitude, Power spectrum and PSD spectrum results. It shows a logarithmic values for all FFT result points, causing Y-axis to be logarithmic.

Y-axis in dBFS: This option can be applied only if Logarithmic Y-Axis is used. It will calculate dB values relative to some maximal, i.e. full scale value. You can freely define this Full-scale value in signal units (for example 32767 for 16-bit sound card or max. voltage for NiDAQ device). There is also a combo-box with the choice of some predefined values for the maximal amplitude.  If you set the full scale value to 0, dB values will always be calculated relative to the currently highest FFT value so that the highest value is always equal to 0 dB and all other FFT values are negative.

X-Axis units combo box: allows you to switch between following X-axis units: “Hz (Cycles/sec)”, “CPM (Cycles/minute)”, “KHz”, “MHz”.

Smoothing: If the result of FFT is noisy and has many values, it can be useful to smooth it by using weighted moving average function. That will remove small noisy details from the FFT and give you a better overview of the important spectrum events. You can choose some of standard weighting functions here and determine the length of a smoothing window (longer window means more smoothing).

Test for confidence: Siegels test for confidence of spectral peaks tells you if some peak in spectrum is statistically significant or it could be product of some random fluctuation of the signal. If you turn this option on, only significant peaks will be shown, while all other values in the spectrum will be set to zero. You can choose two levels of confidence for this test: 95% or 99%. This test is purely statistics; it does not use any artificial intelligence methods.
  
Spectral analysis defaults