SCIENCE DATA: GENERAL

On Paranal, the quick-look pipelines make an effort to automatically reduce all science data. The reduction is performed using standard calibration solutions from a local calibration database which is refreshed every few months. Generally any pipeline processing on the site is done on a best-effort basis. Its purpose is to offer a quick look to assess data quality etc.

Until the end of P87 (September 2011) science data have been processed by QC Garching with the best possible (certified) calibrations solutions. The products were ingested into the Science Archive and delivered to the PIs.

This service has been terminated with the begin of October 2011. The documentation of science pipelines and science processing is kept here for its heritage value.

GIRAFFE science data come in two main types: OzPoz (just the OBJECT and SKY fibres illuminated on the sky), and SimCal (with the additional SIMCAL fibres illuminated by the calibration unit). The data types are identified by the DPR TYPE keyword of the FITS headers. Find information about the general GIRAFFE data format and about its CCD here. The information contained in the simultaneous calibration fibres is applied by the pipeline since January 2008.

The GIRAFFE science frames come per slit which can be MEDUSA1, MEDUSA2, IFU1, IFU2, or Argus.

The SCIENCE varieties further split into two different gratings (HR and LR) with a set of central wavelengths (which are listed here).

Note: It is not visible from the DPR TYPE if a FLAMES OB uses UVES and GIRAFFE at the same time, or just GIRAFFE stand-alone. Data from these two spectrographs taken simultaneously are reduced, packed, and shipped independently.

DPR CATG DPR TYPE DPR TECH description FITS headers SCIENCE OBJECT,OzPoz MOS (for Medusa) or IFU (for IFU and Argus) only science and sky observations; no simultaneous calibrations taken science_ozpoz.hdr SCIENCE OBJECT,SimCal MOS (for Medusa) or IFU (for IFU and Argus) simultaneous calibrations taken science_simcal.hdr

SCIENCE PRODUCTS

Overview of science recipe and QC plots

The GIRAFFE calibration scheme, including SCIENCE data reduction, is shown here.

Recipe. A science raw file is pipeline-processed by the pipeline recipe giscience. Data are corrected for bias. Optionally a bad pixel map is used to correct for bad pixels (this is not done in QC operations; this step is not required for the new CCD). For the old CCD, the background (composed of dark current and CCD glow) was subtracted using a recent master_dark frame; the new CCD (since 2008-05-26) does not require this. The fibres are then extracted using the localization and width solutions from the flat field data. They are wavelength-calibrated using the dispersion-solution from the arc-lamp (wave) data. The spectra are then flat-fielded, to correct for overall efficiency, relative (fibre-to-fibre) efficiency, and fringing. For IFU and Argus data, a reconstructed image and a full 3D datacube are calculated.

Bias removal. For the new CCD (since 2008-05-26), the bias is averaged from the overscan region (columns 1-50) into a 1D profile, and subtracted as a 2D expanded function. This method is called PROFILE (--bsremove-method=PROFILE). It has the advantage that the overall BIAS level is always correct, despite the small fluctuations observed to sometimes occur between BIAS calibration files and SCIENCE files. On the other hand, the overscan region has small residual fluctuations which are probably of random nature. To avoid propagating them as dark and bright stripes and bands into the products of the giscience recipe (and actually also the gistandard recipe), there is the method PROFILE+CURVE which applies a 5th order polynomial fit to the 1d profile before applying the result to the science data. This improved method is used since version 2.6.7 (August 2010). It is applied both to science and to standard star data.

Result of bias removal method PROFILE. The displayed spectrum is an extracted but not yet binned ARGUS standard star product STD_EXTSPECTRA. Each column is an extracted fibre. The artefacts due to the low-amplitude bands in the overscan region are visible as horizontal structures (columns are still in pixel space). The spectrum displayed is extracted from the vertical cut marked by the white solid line.   Result of bias removal method PROFILE+CURVE. This is the same spectrum after applying the bias method PROFILE+CURVE with a 5th order polynomial. There are much less bands visible, most are removed by the smooth polynomial. Overscan BIAS profiles. Black is the raw BIAS profile (vertical overscan signal, collapsed over the first 50 columns), red is a smoothed version with a filter, green is the 5th order polynomial fit used by the PROFILE+CURVE method.

Extraction. By default, the GIRAFFE pipeline sums up all pixels within the mask defined in the flat-field localization file (extraction method 'SUM'). An optimum extraction algorithm is available (methods 'HORNE' or 'OPTIMAL'). This extraction mode has been developed to better handle the scattered light component in the background. The scattered light is more an issue for the IFU modes than for the Medusa data. Also, it performs very slow and does not always deliver stable solutions. For Medusa, the sky fibres are not marked in a standard scheme so the pipeline cannot find the and subtract the sky signal. Finally, this extraction mode requires a different flat field product than the one used for the SUM method. Therefore optimum extraction is not applied in the pipeline scheme used at QC.

Sky. There is no sky subtraction provided by the pipeline. The sky fibres (either the dedicated ones for IFU and Argus, or the user-defined ones for Medusa) are processed in exactly the same way as the target fibres. If sky subtraction is desired, the user has to identify the sky fibres, fit a model to their signal, and subtract this from the pipeline-provided solution. Find more information about the stability of the fibre transmission here.

Fibre-to-fibre efficiency. Differences of the fibre-to-fibre efficiency are corrected for by the flat-fielding step. The coefficients used for the correction are stored in the product binary table, column TRANSMISSION. (If you are unhappy with the applied correction, divide these factors into the fibre signals for undoing the correction.)

Fringing. Fringing is removed by the flat-fielding.

Cross-talk and background. With version 2.3, the pipeline can correct for the notorious glow of the old CCD using a close-in-time master_dark.

There is presently no pipeline correction for cross-talk.

Image reconstruction. In the IFU modes (IFU and Argus), a reconstructed image is calculated. It is obtained by collapsing the spectra in dispersion direction, and re-arrange the signal according to the X and Y sky coordinates as provided in the fibre table. With version 2.3, there is also the full 3D datacube delivered for Argus (X,Y,wavelength).

ACTUAL PROBLEMS AND ISSUES

Since December 2005, all setups (except for three very blue ones) are pipeline-supported. Before that date, most IFU1/2 settings have not been supported.

There is no scattered light correction.

The sky is not subtracted, sky emission lines and sky continuum are imprinted on the science data. The sky signal is available for subtraction by the user from the user-defined SKY fibres.

The following issues with the science reduction are known:

• wavelength calibration covers not always the maximum possible wavelength range (especially at the long end)
• the old CCD ("Bruce", used until 2008-03-13) had a bright glowing area on its upper right edge which does not cancel out upon flat-fielding; it caused an artificial background for fibres 105 - 125 (in the Medusa case) and 250+ (ARGUS, IFU), resp. This glow can be safely removed with a master dark frame. This is implemented in the QC processed data as of P80 (October 1, 2007).
  there is no such issue with the new CCD, no master-dark subtracted.
• three hot columns (at X coordinates 368 to 370, corresponds to fibre #24 for Medusa) are presently not suppressed. There is no meaningful signal in these columns. The extracted signal in this area may be corrupt.
• The pipeline extracts only complete fibres, i.e. the last one or two at the right edge of the chip are not extracted.
• The new CCD "Carreras" has a slightly more variable BIAS than the old one; therefore, it has been decided that the BIAS is removed from all raw frames, incl. SCIENCE data, by averaging the OVERSCAN region into a 1D column, expanding it, and removing it from the raw frames. No master bias is subtracted. However, created master bias frames are archived.