Hard lower frequency limit in weighting function

[krisvanneste]

In the _build_weight_from_ratio function in ssp_build_spectra.py, there is a hard lower frequency limit of 0.2 Hz. Weights below this limit are set to a small value, so that lower frequencies do not contribute to the fitting. This will likely affect magnitude estimation for large events with corner frequencies below 0.2 Hz (but perhaps only if weighting based on SNR is used).

We should investigate if this lower bound can be more flexible or user-defined.

[claudiodsf]

Hi Kris,

This turned out to be a non-issue. The original comment incorrectly stated that everything below 0.2 Hz was set to zero. In reality, the weight amplitudes below 0.2 weight units were set to zero.

While reviewing this part of the code, I also fixed a logical error where small weights were being set to a small exponential value before applying log10. This caused negative weights in log scale. The issue has been corrected in commit 793a35a.

Note that I’ve made additional modifications after this commit, which require updating both the config file and the database. I’ll open a separate issue to cover these new features—basically, an initial attempt at defining inversion quality. Stay tuned for updates early next week! 😉

Bon weekend!
Claudio

[krisvanneste]

OK, that is good to know. Thank you for looking into this!

Kris