3. Data reduction
The parallel mode of operation of the Hitchhiker camera, without any control of the pointing of the telescope, requires that the data be reduced in a slightly non-standard manner. These methods are described in full by Driver (1994) and summarised by Driver et al. (1994a). Firstly, bias signal subtraction was performed by removing a mean bias level calculated from the CCD bias strips of each raw data frame. The validity of this approach has been verified by testing the intensity of the bias signal across the whole frame; no significant structure has been found. The data frames suffer from an essentially circularly-symmetric vignetting pattern (caused by instrumental optical components). Because this varies with the focussing of the camera lenses, and because it is not always possible to obtain nightly flatfielding data, it is necessary to correct for this vignetting at the outset of data reduction. For the 1991 observations the vignetting pattern was modelled by measuring the sky background at a set of points on a rectangular grid, performing two-dimensional bicubic spline interpolation between the sampling points. A relatively coarse grid was chosen, 5 by 7 sampling points in extent, in order to lessen the possibility of removing any extended astronomical structures. The frames were then corrected for vignetting by division by the model. For the 1993 data a median filtering technique using a 45 arcsec wide filter box was employed. Flatfielding was accomplished for the 1991 data using a superflat constructed from over 100 long exposures of the night sky. For the less extensive 1993 observations the superflats were generated by coadding a number of twilight sky frames. The individual flatfield frames were vignetting-corrected before coaddition to produce the superflats; in this application the flatfielding process corrects only for the pixel-to-pixel efficiency variations and not for large-scale effects.
Cosmic ray detections were removed from the individual dark sky frames before alignment and coaddition. Candidate cosmic ray detections were identified as localised peaks rising higher than seven standard deviations above the sky background. The intensity profiles of the peaks were determined, with those steeper than a typical seeing disc being labelled as a genuine cosmic ray event. The pixel intensities in each of these detections were set to the median in surrounding pixels.
Finally the coadded images were cleaned by performing an additional median filter sky subtraction (with a 40 arcsec wide square box). Remaining spurious, low-intensity artifacts (due to dust particles on the instrumental optical surfaces) were removed via inverse unsharp masking (cf. Driver et al. 1994a). The photometric calibration of the data was accomplished using the results of pointed Hitchhiker observations of standard stars in February 1991 and May 1993, corrected to the date of observation using the nightly extinction coefficients from the Carlsberg Automated Meridian Transit Circle Telescope on La Palma. Magnitudes are expressed on the Cousins system (Cousins 1976; Bessell 1979), calibrated with Landolt (1983) photometric standards. The accuracy of the magnitude scale has been shown to be in each filter (Driver et al. 1994b). Colour indices are, however, more accurate because of the simultaneous recording of data through the two filters.
© European Southern Observatory (ESO) 1997
Online publication: July 3, 1998