Is this idea correct?

[enzo ternavasio]

I have developped a new way to try to measure the resonance site frequency (if this is the correct way to call it).
I elaborate a saf file, whose improves the measure.

The responce from the app based on this idea is sometimes in contrast with what you get from the HV curve, evaluated in the usual way (Nakamura).

I would be glad to share my ideas with someone that has a theoritical interest, to discuss the validity of all the porject.

Enzo

[admin]

Hi Enzo,

Do you have a paper or a report that describes the details of your approach? Without any indication it is difficult to provide an opinion.

[enzo ternavasio]

https://drive.google.com/file/d/1W4H7j1 ... =drive_web
https://drive.google.com/file/d/10HhwEF ... =drive_web

these are 2 screenshots after processing 2 different .saf files.
In the bottom left you can see what should be the resonant frequencies fro axis x,y,z.
Sometimes there is not a sharp curve (so there is not res. freq.), but often it is very evident.
You can expect that x,y,z res.freq. should correspond to HV peak, but more often values are similar or different. If this method works, it gives a by far better information on resonant frequencies of site.

The basis of this processing lays in the fact that we have a resonant sistem (ground+some layer).
If in this area the conditions are that a persistent oscillations can arise (taking energy from environemental noise), this permanent signal must be found in an x,y,z acquisition.
In saf file, the other signals are impulsive and incoherent. The exeption should be a persistent periodoic noise (antropic) that exist for all the acquisition. This can give misunderstanding.

The software looks for a continuous sinusoid in the .saf file, trying different segments of different length. This operation should find,if it exist, a segment with an exact wave length, that if summed for all windows, it decreases random noise, other incoherent signals and enhance that exact wave length.
FFT cannot be used beacouse segments are not multiple of 2 and filling with zeros gives not good results. So the process is can last some time.

I'm trying to arrange some testing, but at this moment a have no other means than comparing rasults with HV; i'm not a geophisic, but theoritiacal basis of HV are absent and any disussion i've read not conveincing.

Ask for every thing you are interested. Thanks for your attention.

Enzo Ternavasio

[admin]

In the bottom left you can see what should be the resonant frequencies fro axis x,y,z.

What are X, Y, Z? Are they the three components of the particle motion (Z for vertical)? I guess so but better to be explicit. Is it the simply the spectrum of the three components that are shown?

The software looks for a continuous sinusoid in the .saf file, trying different segments of different length. This operation should find,if it exist, a segment with an exact wave length, that if summed for all windows, it decreases random noise, other incoherent signals and enhance that exact wave length.
FFT cannot be used beacouse segments are not multiple of 2 and filling with zeros gives not good results. So the process is can last some time.

I can't see the real difference between a time consuming fit of a continuous sinusoid and a fast and efficient FFT like FFTW. Note that FFT limited to power of 2 is something that people used in the Eighties. Nowadays, FFTs on an arbitrary number of samples are usual (geopsy does it without 0 padding, it is based on FFTW library).

[enzo ternavasio]

Hy.
Yes, what is shown is the spectrum of x,y and z components.
But it is not the same as the spectrum of the original signal. The signal present in the original saf file does not give any information.
What is triggering my interest is the presence of a continuous frequency that should be in the original signal.
Doing the process of searching a segment that shows some peak in frequency and adding for all the time, gives a spectrum different, generally shwing some peak.
Do you know if someone has done something like this? Or has abbandoned this beacouse of some reason?

The theoretical basis on witch all lays do you think is correct?

[admin]

Note that the spectra of Z, N and E do not necessarily have a peak. A typical shape is the the so-called "eye shape" around the H/V peak where the vertical component vanishes while the horizontal component not. The result is a peak on the H/V spectral ratio. Hence, peaks appear on the spectral ratio but not necessarily on the individual components. A peak on the single component spectrum is usually induced by machinery or other oscillators close-by (tower, trees, buildings,...) but they are not directly linked to the ground structure properties.

Have you search in the literature (scholar.google.com or researchgate for instance) for a similar process?

[enzo ternavasio]

Hy Admin,

i looked not intensively for something related to "site resonance".
Every reading conducted to HV ratio. I will look better.

If a see some peek in the spectrum of an acquired signal, it is easy to conduct it to an external peridoic noise that has lasted for a good portion of time.
If only after my processing I find some peek in the spectrum, I am more interested but in fact it could still be a noise.

Phisically, if HV gives a value of 3Hz, there must be 3Hz signal in the X,Y or Z or all axis.
This is if we trust in that HV procedure.

Geologists told me about the "eye" in x,y spectrum, but i do not understand its relation with some resonance.
Spectrum like that used for HV is smoothed and filtered and it is impossible to see any resonance.

I will keep on my way until i will find some testing reliable.

[admin]

Did you read Scherbaum et al. 2003 in GJI, in particular fig 5?