7.1.2 Processing parameters

Raw spectra contain a lot of frequency samples (e.g. for 50 s time windows and 100 Hz as sampling rate, the natural number of frequency samples is 2500 and the frequency step is 0.02 Hz). They generally show lot of very narrow oscillations and spikes which alter their readability. Moreover, while computing the ratio H/V, the very small values reached by raw spectra lead to unusual values of the ratio.

For this reason, before computing the H/V ratio, the Fourier spectra amplitude of the three components are smoothed. You can choose between three types of algorithm (see table here below). Note that smoothing the Fourier spectra is not mandatory, but strongly recommended.

You can also use a cosine taper to minimise the border effects due to the windowing of the Fourier amplitude spectra. in this case check the "cosine taper" box and give the appropriate value in percentage inside the empty box.

Different types of smoothing function

Type Description Konno & Ohmachi Smoothing with the "Konno-Ohmachi" function: where,    f is the frequency,    fc is the central frequency where the smoothing is performed,    b is the bandwidth coefficient. The bandwidth of the smoothing function is constant on a logarithmic scale. A small value of b will lead to a strong smoothing, while a large value of b will lead to a low smoothing of the Fourier spectra This is the recommended option. The default (and generally used) value for b is 40. Constant The smoothing function has a triangular shape centered on the current frequency and its width is equal to "Band width" Proportional The smoothing function has a triangular shape and its width depends upon the current frequency. The half width is defined by percentage*Frequency. The value of "percentage" cannot be greater or equal to 100%. No smoothing No smoothing function is applied on the Fourier spectra. This option is not recommended.

Figure 2: Define the output parameters for the computation of the H/V ratio.

The Output sub-section (Figure 2) helps you defining the frequency range for the computation of the H/V curve. As mentionned above the natural sampling for spectra contain a lot of samples. For H/V output, such a precision is useless and sometimes heavy to handle. Furthermore, comparison of spectra obtained with different time window lenghts (hence with different natural sampling) is not easy.

After computing the raw spectra with the full precision, they are output with a restricted "number of samples" ranging from the "Minimum" to "Maximum". These two limits have to be in agreement with the minimum time window length choosen in "Windowing" tab: at least ten cycles in one time window (e.g. for a H/V between 0.2 and 20 Hz, time window lenght should at least 50 s).

Linear or logarithmic scale controls the distribution of samples along the frequency axis. The recommended scale is logarithmic, and the number of frequency ("number of samples").

Figure 3: Upload make up files for the H/V ratio output

Here you can upload the setting files for automatic update of the window's appearance of the H/V ratio and Fourier spectra amplitude results.