Geopsy-hv parameter file related problem

[yuqiao]

Hi Marc,

I'm using Geopsy-hv command line tool to process ambient noise data and to get HVSR results. I want to set a parameter file to set up the anti-triggering on raw data parameters(STA LTA MAX MIN). In the Geopsy-gui output log file which can be used as parameter file, there are no such parameter values to set. There is only one line says:

[PARAM] ANTI-TRIGGERING_ON_RAW_SIGNAL (y/n)=y

I noticed a page in Geopsy wiki(https://www.geopsy.org/wiki/index.php/G ... meter_file). Parameter values to set STL LTA MIN MAX are present in the example file. Like this:

USED RAW COMPONENTS = y, y, y, n, y
RAW STA (s) = 1
RAW LTA (s) = 30
RAW MIN SLTA = 0.2
RAW MAX SLTA = 2.5

But this file does not work with Geopsy 3.4.2. I copied those lines to output result log file, it did not work either.
Geopsy-hv command line tool is a greate tool to process a list of noise data, so i want to figure out how to set anti-triggering on raw data's parameters in parameter file.
I will appreciate your reply.

[admin]

The correct syntax is:

Code: Select all

ANTI-TRIGGERING_ON_RAW_SIGNAL (y/n)=y
USED_RAW_COMPONENTS=y, y, y
RAW_STA (s)=1
RAW_LTA (s)=30
RAW_MIN_SLTA=0.2
RAW_MAX_SLTA=2.5

But even with 3.5.0 (the next development release) the anti-triggering option is not correctly handled by geopsy-hv. I will try to fix it in 3.4.3-preview (available directly after correction through git if you are compiling under Linux).
Thanks for your report.

Best regards,

Marc

[admin]

Hi,

The import/export of parameter files for HV are now fixed for 3.4.3-preview. It can be accessed from now through git (see download page).

Best regards,

Marc

[luigiV]

Hi Marc,
what is your opinion about the "frequency rejection" option compared to the anti-triggering one? I have read the work of Cox et al. which I found convincing enough.
Personally I have used it several times, but the main problem is that it is sometimes impossible to correctly calibrate the standard deviation value. With the decrease of even a single decimal the number of windows is excessively reduced and the peak becomes unrealistically sharp.

Luigi

[admin]

Hi Luigi,

After implementing it, I played a little bit with it as well. Effectively, it can a bit parameter dependent like other selections (anti-trigger). As far as I can remember I did some tests on synthetic data but it was not really helping a lot because the synthetic data were too "good" even without selection, there were not enough transient signal to reject something. Did you check the citations of this paper from Cox et al.? Is there any interesting application from other authors?

Best regards,

Marc

[luigiV]

Hi Marc,
No, I just experimented with my data and compared it with anti-triggering.
It seems to me that this tool works quite well when the raw data is of low quality, in the sense that it does "emerge" a peak anyway. It is its form that allows me to judge whether it is an artifact or not. If the peak is evident, the standard deviation of the peak frequency must be evaluated. If it appears too small, I prefer not to accept it.
I am attaching an example of a recording very close to an HT power line where I reported the H / V curve of all the time windows, the selected time windows with anti-triggering and frequency rejection and related H / curves. The standard deviation was set to 1.5 because at larger deviations (starting from 1.55) there was no selection (104/104 time windows). In this case I used the anti-triggering solution.

What do you think about?