Документ взят из кэша поисковой машины. Адрес
оригинального документа
: http://www.stsci.edu/stsci/meetings/irw/proceedings/walshj.dir/section3_1.html
Дата изменения: Wed Jun 15 18:11:48 1994 Дата индексирования: Sun Dec 23 19:34:38 2007 Кодировка: Поисковые слова: planetary nebula |
The diodes of both GHRS detectors are spaced on 50 m centers. The
limiting resolution of the instrument with the small science aperture (0."25
0."25, 0."22
0."22 post-COSTAR) is
55.5
6.5
m (FWHM), and slightly larger for the echelle gratings (GHRS
SV Report, § 7.1). The spectrum projected on the 500 diodes is therefore not
well sampled, with only one diode per resolution element. In order to ensure
adequate sampling, the array can be electronically shifted by one half or one
quarter of a diode width. The resulting spectra are called half-stepped or
quarter-stepped respectively, and can be reconstituted to give two or four
samples per resolution element, thus ensuring proper sampling. In imaging such
a strategy is called dithering. This procedure, which can also be accomplished
together with sampling of the adjacent background, is fast and efficient in
terms of observing time.
Depending on the sub-stepping pattern chosen, the PODPS pipeline
calibration procedure calhrs delivers calibrated spectra where
the substep bins have been ``merged''. This is a simple reconstitution
of the data; thus for quarter sub-stepped spectra the merged
spectrum consists of
| D1s1 | D1s2 | D1s3 | D1s4 | D2s1 | D2s2 | D2s3 | D2s4 | D3s1 | D3s2 |
,
where Dn refers to diode n and sn to substep n. Thus D2s3 is the second diode from the third sub-stepped spectrum. Experiments show that this method of combination is preferable to those which involve some assumptions about the behavior of the data within the ``sub-channels'' of a diode, such as linear or spline interpolation onto a finer grid. In addition this merging procedure is fully reversible, allowing the individual spectra at each substep to be reconstituted. However the resulting spectrum still suffers from undersampling.