2.2 Initial Background Flux Estimates (Pass 1)

  Most of the work and decision-making of the algorithm is done in the Pass 1 stage of BCKGRD. NEWSIPS does not provide the results of this critical operation in the final-archive data products. A certain complexity is required for this step because of the logic required to estimate the amount of contaminating flux from neighboring orders from a model PSF (estimated percentage of flux overlap for each echelle order) and by checks for unwanted ``ringing" in the solution which may occur for high-degree polynomial fits to ill-conditioned data (e.g., background fluxes containing bright spots or negative extrapolations). As a practical matter, we have found that a 7th-degree fit (8 constants) is the limiting degree to which IUE data can be fit by an automated algorithm without appreciable ringing.

The initial Pass 1 proceeds in two steps. In the first step a background estimate is obtained by fitting a 7th-degree polynomial through fluxes of selected pixels in the spatial direction. In this step interorder fluxes are ignored between the short-wavelength orders for which the overlap is most severe, a triangular-shaped region easily seen in Figure 3 and which we will call the interorder overlap region (``IOR"). For the SWP camera the IOR sets in at $\lambda$1520 (line 373). The percentage contamination gets progressively worse below this wavelength, but its importance in arbitrary flux units is greatest for the SWP camera at $\sim$1280Å  (line 233). Below $\sim$$\lambda$1300, the instrumental sensitivity rapidly declines, so the contaminated flux becomes less of an interfering factor in determining local backgrounds. Analogous conditions and defined regions apply for the long-wavelength camera images but are less pronounced.

In the presence of certain image pathologies and for first few and last Pass 1 swaths across the short and long-wavelength ends of echelle orders, the first step is the only one for the Pass 1 procedure. However, for most swaths it serves as reference background function for a more detailed determination. In the second step this estimate is used as a floor from which heights of local order fluxes are combined with a Point Spread Function model¹ to estimate the interorder flux contribution. The steepness of the rising leg of the IOR flux elevations near orders in the wavelength range $\lambda\lambda$1280-1520 is compared with a PSF model derived from well-exposed images of exposures of hot stars. If the measured slope of overlap among a group of echelle orders agrees within a tolerance factor of the slope of the model, the measured value is used. If the tolerance test fails, the program uses the model slope between points nd and nf (Figure 3) from the model as a default. The PSF also contains an extended flat ``ramp" due to halation, which is the reflection of photoelectrons up to 7 pixels from an illuminated source (Coleman et al. 1977). Fluxes for a large number of orders, comprising wavelengths over $\lambda\lambda$1520-1860 for SWP echellograms, are elevated by halation, so interorder contamination of background fluxes actually extends over most of the useable region of SWP images. The product of the PSF values and heights of each adjacent order is computed and subtracted from the raw interorder fluxes (squares) in Figure 3 to create ``working" background values. A new 7th-degree polynomial is then fit through both the working and raw fluxes outside the regions affected by interorder contamination, as depicted by the solid line in Figure 3. It is important to note that BCKGRD attempts to remove the contaminant flux from electro-optic halation and from adjacent orders but not light scattered from the two gratings. The omission of the latter component means that the background fluxes can be too low by a few percent for the short-wavelength orders relative to the true zero-flux level, e.g., as determined from interstellar or lines.

After a solution is computed for a particular Pass 1 data swath, a series of data pathology tests is performed on the solution. These criteria and the action taken by BCKGRD if they fail are described in the NEWSIPS manual appendix. Typical types of failures are maxima or minima of a Pass 1 solution occurring beyond empirically determined tolerance factors for spatial regions of the swath where solutions do not normally exhibit large extrema. The typical response of BCKGRD to a failure of one of these tests is to decrement the trial Chebyshev degree by one and to repeat the Pass 1 step until no failures occur or, in rare instances, a degree of 3 is reached; see the appendix in Garhart et al. 1997 for a summary of these tests. One of these pathology tests is particularly relevant to unfavorable conditions often occurring near the short-wavelength corner of the SWP camera image, where the Lyman $\alpha$ feature lies (see $\S$2.4).