ROACH save_trace method

[wcpettus]

The ROACH should have save_trace method functionality parallel to the RSA.

As a first guess, this will involve several different "modes":

• use psyllid to build and save a mask for a 100 MHz span, uses the full digital processing chain
• in the roach2_interface, take raw time series data and do some intelligent FFT to reduce data size
  • template off Raphael's current analysis
• in the roach2_interface, save raw time series data as we currently do (but it's ~1 GB per shot)

In all cases, an individual "trace" should be saved in a timestamped folder following the naming convention for DIRAC to pickup, and should save appropriate metadata/setup information.

[wcpettus]

This data will be saved to zeppelin in some /data/secondary structure. A DIRAC agent will then move it to the catalog and PNNL storage. What should be the storage path in the catalog?

• Will we stop taking RSA rf_bkgd data and the ROACH will just take over that space on DIRAC?
• Will we resolve this with metadata? (I don't think we currently have metadata keyed into the RF background data)
• Will we make users wade through and discern type based on filenames/extensions? (this just sounds cruel)
• Do we just make a parallel directory for roach_rf?

@guiguem: has anyone proposed this new data stream will get sorted in the catalog?

[guiguem]

Hey Walter,
here are my guesses/thoughts about these questions:

• eventually yes. I am not sure I understand the "take over that space" on DIRAC.
  I believe the RSA and the ROACH can live in the same subfolder on dirac and the distinction from a processing perspective can be the format (.dpt or .egg) and...
• the metadata. There should be several metadata saved with each file:
  • (NEW) Datalevel: RAW (in contrast with Processed, Merged, Monitoring)
  • (NEW) Datatype: Data (in contrast with MC or Context)
  • DAQ: ROACH or RSA
    Note that these metadata are identical to the normal electron data, the distinction should be done via the location of the file (/calib vs /data)
• I think metadata and location will do the trick.
• I am not sure I want to separate the rsa from the roach if both are recording the rf_background.

Does this help?

[wcpettus]

• I believe the DIRAC transfer agent is handling the file-level metadata, so that doesn't impact this implementation
• If we are implementing real metadata on RF data, we should structure this properly for the RSA and ROACH going forward, this should be scoped into a new issue though.