Adding a new camera

The policy directory in the obs_lsst package contains the files needed to describe cameras made up of LSST chips. The eventual goal is to describe the real camera, but for now we also have variants to handle LATISS data, phosim and imsim simulations (they differ in e.g., the gain and crosstalk values) and data from various test stands.

Once Butler Gen3 is ready this configuration data will be moved out of the obs package and into the calibration registry, which will allow us to track evolution of the system (including replacing failed rafts – not that that’s going to happen).

The basic strategy is that the SConscript file in the directory assembles a suitable camera.yaml file (e.g. phosim.yaml) and we put the appropriate entry in the _mapper file in the data repository.

To add a new camera (e.g., fooCam, made up of 9 CCDs in a single “raft” – call it RXX for now):

  • Add a directory policy/fooCam

  • Put a file rafts.yaml in that directory describing RXX (you can start with policy/rafts.yaml).

  • Put a file RXX.yaml in policy/fooCam (you can start with policy/lsstCam/R11.yaml). Note that you can choose an ITL or E2V device. Note that you must provide a serial number for each CCD in the raft as that’s how I know how many CCDs there are in the “raft” (e.g. LATISS has only one)

    The geometryWithinRaft field may be omitted, in which case offsets default to 0.0 and the yaw entry is not generated. These offsets are interpreted relative to the nominal positions given in cameraHeader.yaml for each type of raft, as adjusted for the centre of the raft; these values are therefore reasonable.

    The yaw (rotation in the plane of the detector) is measured in degrees, anticlockwise as shown in cameraGeomUtils.plotFocalPlane (i.e. with R00 in the bottom left, and R04 in the bottom right)

  • If you want to add a camera-specific set of transformations, put a file cameraTransforms.yaml in policy/fooCam. Look at the one in policy/cameraTransforms.yaml for inspiration.

  • If you’re plagued by crosstalk, put a file ccdData.yaml in policy/fooCam. Look at the one in policy/phosim for the format – basically a dict crosstalk indexed by the CCD type. There are 256 coefficients (16 amplifiers and 256 == 16^2), and we assume for now that all CCDs from a given vendor are the same (but the fix to bin.src/generateCamera.py to handle per-CCD coefficients would be easy)

  • Edit policy/SConscript to add your new camera “fooCam” to the for camera in ... loop. note the magic --path options – it tells the code to use your data in fooCam to override lsstCam values. This is why there’s no imsim directory and phosim only contains the crosstalk coefficients; they take almost everything from policy/` and`policy/lsstCam`

  • run scons in the obs_lsst directory (or scons -u in policy)

  • add fooCam.yaml to policy/.gitignore

  • Create a new file python/lsst/obs/lsst/fooCam.py (see latiss.py for an example)

    Add a class FooCamMapper to the same file, setting a class-level string _cameraName to “fooCam”. You also need to specify the metadata translation class to use such as LsstFooCamTranslator. Look at the example in python/lsst/obs/lsst/latiss.py – you’ll see that this overrides some entries in lsstCamMapper.yaml (in the class data member yamlFileList) with latissMapper.yaml. If you want to provide your own templates you’ll need to do the same thing, adding a file policy/fooCam/fooCamMapper.yaml

    The name you provided as _cameraName is also used to e.g., provide per-camera configuration files (for example config/latiss/ingest.py)

    Don’t forget to add FooCamMapper to __all__

  • Write a header translator for your instrument. This should be placed in python/lsst/obs/lsst/translators/fooCam.py. You can follow the examples from other translators there. Remember to add the new file to python/lsst/obs/lsst/translators/__init__.py. You can test the translator by running

    translate_header.py -p lsst.obs.lsst.translators testfile.fits
    

    The translators must define the properties specified and defined by ObservationInfo. Pay careful attention to how you decide to define detector_group and detector_name. You can read your detector IDs out of the camera YAML file once it’s created or hard code them into your translator. exposure_id and detector_exposure_id should be written in such a way that they can be called from the mapper class to allow the values to be determined from dataIds.

  • Add a FooCamParseTask to python/lsst/obs/lsst/fooCam.py. It’ll need to set a class variable _mapperClass = FooCamMapper and also a class variable _translatorClass = LsstFooCamTranslator (the same class as used in the mapper). Most of the metadata translations are inherited from the base class so you only need to add translations here if you are adding something non-standard. Don’t forget to add FooCamParseTask to __all__

  • Put a _mapper file in the root of your repository, containing lsst.obs.lsstCam.fooCam.FooCamMapper

  • Retarget config.parse in config/fooCam/ingest.py to FooCamParseTask

  • You will probably also want to copy e.g., config/latiss/latiss.py to config/fooCam/fooCam.py and also files such as config/latiss/bias.py – don’t forget to modify them to import fooCam.py!

  • Add FooCam.yaml to policy/.gitignore

  • Add test data and associated unit tests following the instructions in Testing the Package.

If these instructions are complete and correct you are good to go. Once you’re happy commit all your changes to a branch git, and push, and make a pull request.

We would ideally like to be able to run the integration tests to include the new camera. To enable this please also add some test data to the testdata_lsst repository, and update the ci_lsst package so that the new instrument is included in the integration test.