3 INES processing of high dispersion spectra

 

The starting point for the INES processing of high resolution data are the NEWSIPS MXHI files. The spectra have not been re-extracted from the bi-dimensional files, as in the case of the low resolution data (see Paper I). The INES system provides two output spectra for each high dispersion image: the "concatenated'' and the "rebinned'' spectra. Both include a modified wavelength scale and the error vector calibrated in absolute flux units. All these aspects are discussed in the following sections.

3.1 The high resolution concatenated spectra

 

The main features of the INES concatenated spectra are:

a) The overlap regions between adjacent orders are suppressed in such a way that the less noisy portion of the orders is retained.

b) The error vector is calibrated in absolute flux units.

c) The wavelength scale is modified to make consistent the radial velocities obtained from the short and the long wavelength cameras. The wavelength sampling of the MXHI files has been retained.

d) The spectra are provided as FITS tables having the same format as the low resolution spectra, i.e. they contain only four columns: wavelength, absolute flux, error and quality factor. This setting reduces significantly the download time for remote data retrieval, and simplifies considerably the structure of the NEWSIPS MXHI files, which contain additional information, not relevant for most investigations, such as the position of the orders on the bi-dimensional image, the height of the extraction slit and the order number for each extracted point.

The procedure followed to concatenate the spectral orders is described in detail in the next paragraphs.

Figure 4: Example of the concatenation procedure for the SWP camera. The top panel shows the individual orders from the MXHI file. The bottom panel shows the final concatenated spectrum and the calibrated errors (dotted line)

3.1.1 The concatenation procedure

The most critical point in the concatenation of adjacent echelle orders is a suitable definition of the "cut wavelengths" so that only the highest quality data points of the overlap region are retained.

Figure 5: As Fig. 4 for the LWP camera

Figure 6: As Fig. 4 for the LWR camera. In the spectral region shown orders do not overlap. The INES concatenated spectrum has no data in that region

In IUE high resolution spectra, the signal-to-noise level at the edge of the orders is different in the short and long wavelength cameras. In the LWP and LWR cameras the signal-to-noise is always lower at the short wavelength end of the orders (except in the highest orders, see below). The contrary happens in the SWP camera, where the long wavelength edge of the orders is much noisier than the short wavelength one. Taking this into account, the cut wavelengths have been defined as follows:

SWP:

$$\lambda_{\rm cut} = \lambda_{\rm start} + (\lambda_{\rm end}-\lambda_{\rm start})/3.$$ (1)

LWP/LWR:

$$\lambda_{\rm cut} = \lambda_{\rm start} + 2\times(\lambda_{\rm end}-\lambda_{\rm start})/3.$$ (2)

where
$\lambda_{\rm cut}$: cut wavelength between orders m and m-1
$\lambda_{\rm start}$: start wavelength of overlap region (order m-1)
$\lambda_{\rm end}$: end wavelength of overlap region (order m).

These expressions are valid for all spectral orders except order 125 in the LWP camera and orders 120 to 125 in the LWR camera, where the S/N ratio in order m is systematically higher than in order m-1 in the overlap region. In these cases only the points of order m are taken.

The above defined cut wavelengths (i.e. end wavelengths of order m) can be computed as a function of order number as:

$$\lambda_{\rm cut}(m) = A+\frac{B}{m}+\frac{C}{m^2}$$ (3)

with the values of A, B and C given in Table 7.

For non overlapping orders (lower than 73, 77 and 76 for SWP, LWP and LWR, respectively) only photometrically corrected pixels have been included in the concatenated spectra (see Fig. 6). The concatenated spectra cover the same spectral range as the INES low resolution spectra, i.e. 1150-1980 Å for the SWP camera, and 1850-3350 Å for LWP and LWR.

Figures 4, 5 and 6 show examples of the concatenation procedure for the three cameras.

 

 

Table 7: Parameters defining the cut wavelengths for the order concatenation

Camera	A	B	C	Orders
SWP LAP	24.3952	132875.4838	325840.9715	120-73
SAP	22.2095	133293.4862	300351.2209	120-73
LWP LAP	-7.9697	233257.6280	0	124-77
SAP	-7.7959	233382.6450	0	124-77
LWR LAP	-11.3459	233737.5903	0	119-76
SAP	-11.2214	233876.9950	0	119-76

3.1.2 The error vector

The NEWSIPS processing provides an error vector for the high resolution spectra which is computed simply as the sum along the extraction slit of the noise values for the individual pixels, as derived from the camera noise model. Unlike the "sigma'' of the low resolution data, the "sigma'' vector in the MXHI files is not flux calibrated but given in FN (Flux Number) units.

In the INES high resolution data, the "sigma'' spectrum is provided in absolute flux units. The calibration is performed by applying to the MXHI error vector the high resolution calibration and the time sensitivity degradation correction.

3.2 The resampled spectra

 

In the INES Archive, each high resolution image has an associated "rebinned" spectrum, which is obtained by rebinning the "concatenated" spectrum at the same wavelength step size as low resolution data. This data set represents an important complement to the low resolution archive, and it is especially useful for time variability studies. The rebinned spectra have not been convolved with the low resolution Point Spread Function, and therefore have a better spectral resolution than low dispersion spectra. Examples of rebinned spectra are shown in Figs. 7 and 8.

The high resolution concatenated spectra (derived as described in the previous section) are resampled into the low resolution wavelength space following the procedure detailed below.

Figure 7: Comparison of low resolution (thin line) and high resolution rebinned (thick line) spectra of the standard star BD+28 4211. The top panel shows the flux spectra. Flagged pixels are marked with diamonds (low resolution) and squares (rebinned). The two gaps in the high resolution spectra longward 1900 Å correspond to the regions where spectral orders do not overlap, which have been assigned zero flux. The inset in the top panel shows in more detail the region 1300-1400 Å. The error spectrum is shown in the bottom panel (thin line: low resolution, thick line: rebinned). The characteristic pattern of the errors of the rebinned spectrum is due to the lower signal-to-noise ratio at the edges of the individual echelle orders

Figure 8: Similar to Fig. 7 but comparing two LWP spectra of Nova Cygni 1992. The errors are shown in the bottom panel

3.2.1 The rebinning procedure

The concatenated spectra have been resampled into the INES low resolution wavelength domain as defined in Paper I. The sampling interval is 1.6764 Å/pixel and 2.6693 Å/pixel, for the short wavelength and the long wavelength ranges, respectively, and the wavelength coverage is 1150-1980 Å for SWP and 1850-3350 Å for LWP and LWR.

The resampling has been performed so that the total flux is conserved, that is, if n pixels with fluxes f₁, f₂, $\cdots$, f_n are rebinned into one, the total flux in the bin is:

$$F = \sum^{i=n}_{i=1} (\lambda_{i}-\lambda_{i-1})(f_{i}+f_{i-1})/2$$ (4)

where the flux at the bin edges (i=1, i=n) is calculated by linearly interpolating between the two adjacent pixels. The flux of the final pixel is:

$${\rm Flux} = F/{\rm step}$$ (5)

being "step'' the low resolution pixel size defined above.

3.2.2 Errors

The rebinned error spectrum is computed from the concatenated error spectrum according to the following expression:

$$E = \frac{\sqrt{\sum e_{i}^{2}}}{n}$$ (6)

where e_i are the errors of the original pixels in the concatenated spectrum.

3.2.3 Flagging

The quality flag assigned to each pixel in the resampled spectrum is the sum of the flags of the original high resolution pixels. Only the most relevant quality flags present in the high resolution spectrum have been transmitted to the rebinned spectrum:

-8192: Missing minor frames in extracted spectrum,

-1024: Saturated pixel,

-16: Microphonic noise (for the LWR camera only),

-8: Potential DMU corrupted pixel,

-2: Uncalibrated data point.

Other flags (e.g. reseau marks) have not been taken into account to avoid that a too large fraction of the pixels in the output spectrum come out flagged with error conditions, despite their quality is not significantly affected. Pixels corresponding to the gaps between non-overlapping orders are flagged with "-2''.