The data sets provided by INES consist of low-resolution spectra extracted with an improved method from the line-by-line images of the IUE Final Archive, and of high-resolution spectra resampled to the low-resolution wavelength step. The new extraction of the low-resolution spectra includes:
Another important aspect of INES vs. ULDA is the presentation of spectra in standard FITS format; this implies that the spectra can easily be manipulated by the standard image-processing software packages in general use by the astronomical community.
The noise model used for the production of INES is derived empirically, from hundreds of flat-field images with different exposure times. This was required, given that the IUE detectors are television-type and are different in their behavior from CCD detectors. An improved noise model for the IUE spectra was a necessary first-step of an improved extraction procedure, because the determination of the spatial extent of the object spectrum in the cross-dispersion direction is based on the signal-to-noise (S/N) ratio along the entrance aperture. Also, the determination of the errors in the final extracted spectra is based on propagation of the errors from all the previous stages of the extraction.
The background determination for INES is performed by fits of Chebyshev polynomials across the IUE spectra external to the entrance aperture. The polynomial fit is assumed to be constant across the aperture. The extraction profile is determined by finding the extent of the spectrum through a spline fit along the spatial direction, for extents of S/N30 along the direction of the dispersion, and with a minimum of seven wavelength steps. In cases of weak spectra, where the S/N is below a certain threshold, the extraction is done by adding up all the flux within the entire aperture. In many instances relevant to galaxies, when the aperture is much larger than the dimensions of the target and the surface brightness is relatively low, this is the proper procedure.
The accuracy of INES flux extraction has been tested by Schartel & Skillen (1998) and was found that, in general, NEWSIPS and INES gave consistent results. The definite advantage of INES was demonstrated for spectra with strong and narrow emission lines, where INES gave consistently more reliable results.
In addition, the production of INES was done through a strict configuration control of all the re-processing stages. This resulted in improved quality control of the spectra and also of the information entered in the image headers. This included interactive verification of the input parameters against the hand-written observing logs, as well as automatic checks of the validity of values, correct sizes, and input formats of the raw images.