about filtering and removing instrument response

[jlsanhueza]

Hi all,

I've been using this module to process some data from ZEN equipments (ex, ey, hx, hy, hz), which was pretty easy to follow.
First, I built the .h5 file using the mth5 module, which provides readers for the ZEN outputs.
Then, aurora made the trick to obtain the TFs into edi files, which completely solved my problem.

Now, I'm trying to dig some processing on my own and I saw the filtering guide in the tutorials, which allows to get slices of data, choose frequencies for the filtering and remove the instrument response. I could do all of these but it isn't clear to me if the already filtered data is automatically updated into the .h5 file or if I need to build another .h5 file with the filtered channels (cal_ex, cal_ey, etc) and then run the aurora processing scheme to obtain the TFs from the filtered/calibrated channels.

Also, for the future I'm planning to have a coherence check to try to crop some bad coherence points. I know it can be super tricky to do but is there a guide where I can begin using aurora/mth5?

Thanks in advance and for the nice guides provided!

[kkappler]

Hi @jlsanhueza ,

The way things are currently setup, the data are not automatically stored back into the .h5 file. (The calibrate function here just operates on the RunTS object, returning a new one with calibrated channels).

Both options (building a separate .h5 for the calibrated time series, or storing the calibrated data in the same file) are possible. It depends on your use case.

To store the calibrated data in the same file, you would need to create a place in the hierarchy for the data. A natural (but not the only) choice would be for example to create a run with a similar name as the un-calibrated data, for example, if you had station MT01, with a run called 001, you may create an additional 001_calibrated run under station MT01, and store the calibrated data there. This could make the .h5 roughly double the size. If that were an issue for any reason, then you could instead create another .h5 archive, maybe called survey_xx_calibrated.h5.

While possible, I don't think that there are example notebooks or workflows showing explicitly how the re-archiving would be done. If you give it a try and get stuck feel free to share a notebook or script via issues or a fork or PR with an encountered error and someone can likely help get it unstuck.

Also, do you want to discuss your use case? Visualizing calibrated time series data can be great for building intuition, but calibrating data in time-domain before processing comes with come caveats about spectral leakage that can make things tricky to debug. If you want to look at calibrated data in frequency domain, it maybe better to create an FC layer in your archive with the Fourier coefficients, and then vizualize the data via power or amplitude spectra, or spectrograms. These are a work in progress, but an example can be found around cell 43 in this example notebook.