Cheetah at the European XFEL


Quick start

Cheetah is available for the AGIPD-1M detector at the European XFEL.

The European XFEL became available for experiments in September 2017 and everything is still developmental.  Some patience is required and be aware that things may be subject to change.

As mentioned: Auto experiment setup currently is not configured - you will have to copy and modify the template yourself.

Cheetah for EuXFEL is installed on Maxwell:

        > source /gpfs/cfel/cxi/common/public/development/cheetah/ (or setup.csh if using csh/tcsh)

        > cheetah-gui

    A template directory structure and sample calibration files are located in the same place:

        > tar -xvf /gpfs/cfel/cxi/common/public/development/cheetah/template.tar

    Auto-experiment-setup is working, so just use the ‘new experiment’ button in the GUI to set things up.

CrystFEL is also installed in a public version of the CFEL software stack:

        > source  /gpfs/cfel/cxi/common/public/cfelsoft-rh7-public/

        > module avail

        > module load crystfel

Some useful notes on the European XFEL computing environment

  1. Quick reference for data analysis at the European XFEL

  2. Maxwell cluster (offline analysis, DESY site)

  3. XFEL detector specific analysis notes (XFEL site)

  4. XFEL detector calibration (XFEL site)

Differences at the European XFEL


AGIPD detector calibration

Processing data from the AGIPD detector is currently one of the key challenges and differences in performing data analysis at the EuXFEL. It is important to understand the detector in order to make sense of your data and understand whether anything has gone wrong. This will, of course, get better over time.

Two options are implemented:

1) Perform AGIPD detector calibration in Cheetah

        Cheetah can process AGIPD detector directly from RAW data files, calibrating images on the fly.

        Please read the notes on AGIPD data calibration for instructions on how to create calibration files.

       This gives you direct control over the detector calibration at the expense of having to obtain and handle the
        calibration data directly.  Depending on your persistence and/or level of experience with detectors it may or may

        not give better results. It seems to run faster.

        The EuXFEL data file contents changed during initial experiments.  If working with old data, you may need to

        select the correct ‘data reading mode’ in order to switch between interleaved and parallel gain data, or different
        firmware frame vetoing strategies, for example.

-- or --

2) Read XFEL calibrated and corrected AGIPD data from ‘/proc’

        This is the easiest route.
        EuXFEL provides an official detector calibration pipeline (reads data in /raw and puts corrected data into /proc).

        Simply point Cheetah to the ‘/proc’ directory as the data source; any file name not beginning with ‘RAW_’ will be
        processed assuming it is pre-calibrated by EuXFEL.

        For data in /proc, calibration is outside of Cheetah and therefore outside of it’s control so please don’t ask us
        about any strangeness in the input data. Any problems with the data correction should be discussed directly with the
        EuXFEL detector team (Steffen Hauf or Stephan Stern).   This includes strange values, hot pixels, number and
        flagging of bad pixels, etc.

Broadly speaking, Cheetah looks and behaves as it does elsewhere. The GUI and image viewers are identical and data is organised by runs.  It should feel very familiar.  What changes is the front-end input data source - the interface to the saved data and computing environment.