Telescope Data Center

Hectochelle Data Pipeline

Status as of 2003-03-06
Hectospec Image
[TDC Home]  [TDC Search]  [TDC Help] [OIR Home]  [OIR Search]  [OIR Help]

I. Preparation

  1. Get positions to better than 0.25 arcseconds
  2. Find fiber assignments and select guide stars
    Use John Roll's fitfibs or xfitfibs to create fiber assignment tables to be taken to the telescope.
    This step needs further testing and documentation before being unleashed on the world.
    It is not clear how guide stars are selected, and we need to know the magnitude limits for guide stars. We should be able to use existing software, such as scat to prepare the input for this stage.

II. Observation

III. Reduction

  1. Remove bias and combine amplifiers
    Use mscred ccdproc in IRAF to remove the bias and dark levels from the frames and combine amplifiers into a two images, one per detector, each containing up to 150 fiber targets.
    Existing IRAF code should work perfectly, but needs to be tested.
    We need raw images from the separate chips for testing and long exposure test data to see if the dark current is significant.
  2. Remove Cosmic Rays (Optional)
    Use imcombine to combine the multiple exposures at a single pointing, removing cosmic rays in the process.
    Existing IRAF code should work perfectly, but needs to be tested. This may not need to be done for Hectochelle; the plethora of features overrides a few bright cosmic rays.
  3. Extract Spectra
    dohspec is the main processing program. Before running, check the parameter settings by typing epar dofibers and typing params. Run dofibers on each of the images created in step 1. The output will be a multispec file in which each of the fibers has been normalized, perhaps scattered light corrected, traced, extracted and wavelength calibrated.
    This task will have to be rewritten or the wavelength dispersion correction done as a separate step afterward if we have to correct for shifts in spectrum position. The existing form of dofibers can be used for quick-look reduction. How will sky flats be taken? What sort of scattered light correction is needed?
  4. Recalibrate Spectra
    Find and repair any shift which occured to the spectra.
    Use an IRAF task (to be written) similar to reref which will be used, once it is debugged, to analyze and fix calibration shifts. For Hectochelle, reference fibers will be used rather than sky lines.
  5. Rebin Spectra for sky subtraction (Optional)
    Use rvsao.sumspec to rebin each of the 300 extracted spectra from the multispec file into a stacked-spectrum 2-D file so they all have the same number of pixels and cover the same wavelength range.
    This task works. This is where quick look for all of the spectra could stop.
  6. Create individual spectrum files
    Use the hectospec.hsplit task to split the two files of stacked spectra into individual multispec files, putting these newly created files into a new directory.
    This task works.
  7. Remove Background/Sky (Optional)
  8. Compute redshifts (Optional)
    Use IRAF rvsao.xcsao to find the redshifts of all of the object spectra.
    We will have to make or find good templates for Hectochelle.
  9. Quality Control
    Methods are yet to be determined.

IV. Distribution

Use the gather scripts written by Bill Wyatt to distribute data to principal investigators.
This will be handled almost exactly the way FAST distribution is done, given that the final data are FITS multispecs file just like FAST.

V. Archive

Individual spectra will be stored in a secure archive in a structure to be determined. Pointing directories within night directories could be easily accessed through reduced file numbers yyyymmdd.pppnnn, where yyyymmdd is the UT date, ppp is the pointing sequence number (or maybe the sequence number of the first exposure within a pointing), and nnn is the fiber number.
Calibration files and velocity standards should go into a parallel public archive immediately; PI files should be copied from the secure archive to the public archive when their proprietary period expires or when the PI releases the data, whichever comes first.
This is pretty close to how FAST works now. There would have to be a new web interface and RFN to path code would have to be written for the new RFN format, but not much else would be different. A database needs to be set up to enable access by a variety of characteristics allowing searches for observations with similar orders or order separating filters or grating settings.
[Instruments] [Hectospec Pipeline]