Hi Luke,
Taper
Cosine taper is a cosine variation over the whole time window. Only the sample at the middle of the time window is unaffected. There is no parameter for such window function. To get the same results as in the previous releases, you have to choose
Tukey taper: a cosine function over a parametrisable range at the beginning and at the end of the time window. There you can adjust the percentage.
Smoothing
The classical
Konno-Ohmachi is obtained by, obviously, selecting the
Konno-Ohmachi window but also by choosing a
log width. The original
b constant value is inversely proportional to the width of the smoothing window, which was somehow counter-intuitive. A value of
b=40 corresponds to a 20% log width. A tooltip is displayed if you stop with the mouse anywhere on the smoothing box. It provides the following equivalences:
Code: Select all
b=20 ==> width=40%
b=40 ==> width=20%
b=80 ==> width=10%
Note that it is not an exact equality but very close. To be precise, the exact formula is
Hence more precise values are:
Code: Select all
b=20 ==> width=43.575%
b=40 ==> width=19.823%
b=80 ==> width=9.4636%
The
scale type is a new option that was not available previously. H/V amplitudes have always been averaged on a log scale (geometrical average) since SESAME. However the original smoothing algorithm was summing amplitudes on a linear scale. This sounded like inconsistent. The
Log scale type is probably better. This is now the default value.
Frequency samples
Effectively,
sample number is replaced by two options:
- Step value: defines a fixed step between samples. For a log scale, the default value is 1.025. This is the common ratio of a geometric progression. A sample value multiplied by the step value provides the next value.
- Step count: you can specify a fixed number of samples as in the previous releases
The first option has the advantage that it fully controls the density of samples. The minimum and maximum values can be changed without affecting the density of samples which is not the case with the second option. Additionally samples are always aligned to 0 and 1 for linear and log scales, respectively. Hence you can change the minimum and the maximum frequency, you will always get the same values, which is easier for comparing curves obtained with various processing.
I hope I answered all your questions.
Best regards,
Marc
