ST_CM_PowerCepstrum_InitPars

Function block-specific structure with initialization parameters, which are analyzed when the function block is initialized.

TYPE ST_CM_PowerCepstrum_InitPars EXTENDS ST_CM_Object_InitPars :
STRUCT
    nFFT_Length        : UDINT := 512;                              (* Length of FFT. *) 
    nWindowLength      : UDINT := 400;                              (* Length of analysis window. *)
    eWindowType        : E_CM_WindowType := eCM_HannWindow;         (* Window type. *)
    aWindowParameters  : T_CM_WindowParameters := [2.5,1,1,1,1];    (* Window parameters for specific windows, e.g. FlatTop. *)
    nOverlap           : UDINT := -1;                               (* Size of overlap in samples. *)
    fLogThreshold      : LREAL := cCM_MinArgLogN;                   (* Minimum argument of decadic logarithm for 64- bit IEEE 754 arithmetic     eScalingType       : E_CM_ScalingType := eCM_DiracScaling;      (* Scaling type. *)
    nChannels          : UDINT := 1;                                (* Number of channels. *)
 *)
END_STRUCT
END_TYPE

• nFFT_Length is the length of the FFT. It must be greater than one and an integral power of two.
• nWindowLength is the length of the analysis window in samples. The length must be greater than one and an even number.
• eWindowType defines the used window function (of the type E_CM_WindowType). A good default value is the window type eCM_HannWindow.
• aWindowParameters contains the free parameters of selected window functions. When using eCM_KaiserWindow, the first entry defines the parameter beta; if eCM_FlatTopWindow is used, all parameters are used. See section Window functions.
• nOverlap defines the number of overlapping samples. This must be greater than or equal to zero. If the value cCM_OverlapRecommended is selected, a recommended overlap is calculated internally (see F_CM_CalculateRecommendedOverlap); the value cCM_OverlapInactive deactivates the internally used buffer and sets the value to zero.
• fLogThreshold is a very small floating point value greater than zero. The smallest possible value is 3.75e-324, which is equivalent to the constant cCM_MinArgLogN.
  Spectral values with absolute values that are smaller than this number are replaced with this value before the spectrum is logarithmized. The purpose is the avoidance of value range errors. The logarithm of zero is not defined and strives infinitely towards minus for the limit value of small arguments. The same applies to the argument of the number zero, arg(0).
• eScalingType enables the selection of the scaling to be used (of the type E_CM_ScalingType), in case absolute scaling is required. The default value is eCM_DiracScaling. When selecting the scaling the type of signal should be considered: either deterministic signals or wide-band signals with stochastic portion. Both types require different scalings.
• nChannels defines the number of independent channels. This must be greater than zero.

Window length The value of nFFT_Length must be equal or greater the value of nWindowLength. The length of the FFT can orient itself to the required frequency resolution. Typically a value of about 3/4 of the FFT length is often used as the window length.

If nFFT_Length is greater than nWindowLength, the frequency resolution of the FFT (and therefore also the length of the return values vector) is increased. The length difference is filled with zeros before the Fourier transform. This can be useful for achieving a higher frequency resolution or, e.g. when calculating with inverse transformation in the time domain, in order to avoid circular aliasing . Despite the higher frequency resolution, however, the result contains no more information.