DASCore fails to read training data for phasenet DAS

[d-chambers]

Description

@Shihao-Yuan reported that DASCore has issues reading some of the files associated with the phasenet DAS training data found here. We need to look into it more.

Versions

• OS [e.g. Ubuntu 20.04]:
• DasCore Version [e.g. 0.0.5]:
• Python Version [e.g. 3.10]:

[d-chambers]

After some quick digging, I noticed that the Eureka data are in prodml and the Ridgecrest data are a custom format described here.

We could technically support the custom format. The only hangup is that the distance values for the cable are not provided, and there are other missing metadata (such as the gauge length) we might want. We could just assume the fiber distance starts at 0 I guess 🤷🏻.

I also found an issue in getting the metadata from prodml files likely introduced in #394. Currently, metadata like gauge length are not found due to this issue.

[d-chambers]

For the Ridgecrest data, @Shihao-Yuan mentioned the DAS data didn't look "right", and so there was some concern the data were being scrambled in some way. Plotting one of the files we get:

but this is similar to to what this notebook gets, so I think we are just seeing common-mode noise.

doing a high pass filter improves the image:

so after fixing the attribute issue above, I think we are ok for the prodml files.

[d-chambers]

Closed for now by #410, but we may implement read support for the custom DAS format later on.