taper window type in H/V computation

[luigiV]

Hi Marc,
while for the smoothing function in H/V computation SESAME guidelines suggest the adoption of the Konno&Ohmachi filter, there is no indication regarding the type of window for the taper (this choice is present only since version 3.0). So far I have used Konno&Ohmachi because the adoption of any other type window results in a dramatic increase in sharpness of the H/V peak in many situations.
I compared the same data file processed with Geopsy 2.10 and with Geopsy 3.4.2 (same parameters for waveform filter and taper, antitriggering, smoothing constant, time windows and windows length). In the first case it is not possible to choose the window type for the taper, while in the second case the window type that gives the H/V curve closest to the first is the rectangular one.
Is there a "rule" to adopt in H/V processing with Geopsy versions greater than 3.0?
Thank you

Luigi

[admin]

Hi Luigi,

I did the same comparison between 2.5.0 and 3.5.0-preview that should be similar to 3.4.2.
The plot (a) and (b) should be quasi equal. There might be minor differences. To get a strict equivalence, a more in-depth analysis should be conducted. If you look at the individual windows, you can find a one-to-one correspondence between the two versions.

The smoothing Konno log-log is now the default, it is more natural than the old log-linear. log-linear means, log scale for frequency (width of smoothing window impacted), linear scale for averaging values.The smoothing constant b=40 in the old versions is almost equivalent to a relative width of 20% (see "What's this" for details). You can see that there are some visible differences between log-linear and log-log.

Plot (c) is computed with rectangular window for smoothing, there are significant differences, namely, a decrease if the average peak amplitude.

[luigiV]

Hi Marc,
now everything is clear. In fact, after sending my post, I found your answer to Luke on the same topic where you explained that with version 3, to get the same response as version 2, the window type Tukey must be used for the taper. Now, in addition, I learned that with version 2 the button cosine must also be selected at 10% and with 3 Tukey width is 10% . Doing some tests, I noticed that the most significant differences occur when passing from an alpha/2 of 0.5% to that of 10%. Beyond this percentage the H/V curves remain substantially identical (tukey_comparison.jpg). Therefore it is within the interval 0.5%-10% that the choice must be made to obtain the best result which, in practice, translates into an increase in the amplitude of the peak which becomes maximum with 10% of alpha/2.
Concerning smoothing it is clear that Konno-Ohmachi is the function of choice for both version 2 and version 3. In my comparison.jpg figure I had used the rectangular filter for the version 3 taper and not for smoothing.
Thanks
Regards
Luigi