9 Low-Dispersion Flux Extraction (SWET)

 

Spectral data are extracted from the two-dimensional (2-D) low-dispersion resampled images (SI) using a signal weighted extraction technique (SWET). This technique provides an increase in the signal-to-noise ratio (S/N) of the extracted spectrum, as compared to simple ``boxcar'' extractions, by utilizing information on the cross-dispersion spectral profile, while preserving the total flux. The output of this technique also includes an error estimate associated with each point in the extracted spectrum and cosmic ray hits are, in many cases, automatically rejected from the calculation of extracted fluxes. The technique, which has been developed for low-dispersion IUE data, is based on the weighted slit extraction code originally developed by Horne (1986) for use with optical CCD data. A detailed explanation of the adaptation of this software for use with IUE data processed using IUESIPS is given by Kinney et al., (1991) who supplied the software to the IUE Project. Various threshold values in the code were modified by the IUE Project to account for the very different flux scales of NEWSIPS and IUESIPS.

Briefly, the processing steps involved in the weighted extraction of a one-dimensional (1-D) spectrum from the low-dispersion SI are as follows:

• The background flux as a function of wavelength in the low-dispersion SI is characterized by a sixth-order Chebyshev polynomial.
• The spectral signal is located within the image and the average signal strength is evaluated in order to determine whether or not a spectral profile can be computed empirically.
• If there is sufficient signal strength, the cross-dispersion instrumental flux profile is derived for each wavelength sample. A default profile shape is used for spectra that are too faint to allow an empirical determination.
• The cross-dispersion profile is used in combination with a model of the detector noise characteristics to identify anomalous data points which are to be excluded from the computation of extracted fluxes.
• The good data points are weighted according to their positions on the cross-dispersion profile and according to their relative noise in order to calculate the total net flux and its associated one-sigma uncertainty at each wavelength sample.
• A 1-D $\nu$ flag spectrum is calculated to show where flagged pixels may have precluded the calculation of a reasonable flux value.