Next: Echelle Reduction
Up: No Title
Previous: CCD-Commands Summary
The reduction of CES spectra is, especially for point sources, very
straightforward, and the general instructions given in Chapter 6 for
the standard reduction of spectra are fully adequate. The following
merely adds a few instrument specific details and hints.
- Cameras:
- Except where noted otherwise, the reduction of data
obtained with the Short or the Long Camera is exactly the same.
- Detectors:
- As far as the data reduction goes, the main difference
between Reticon and CCD is the data format. Since the Reticon is a
one-dimensional array, no attention has to (can) be paid to the extraction
of the object spectra. The data acquisition system used with the Reticon
permits multiple exposures to be combined into a pseudo two-dimensional
spectrum. For operations on such data, all commands which work on image
rows are useful, e.g., COMPUTE/ROW, AVERAGE/ROW or also
EXTRACT/IMAGE.
- Blemishes, etc.:
-
-
In many echelle orders a prominent ghost appears which, with some bad luck,
may sit right on top of the object spectrum. It cannot be `flatfielded away'
or otherwise corrected.
-
The first 10--15% (in wavelength) of most spectra suffer from vignetting
which divison by a flatfield may even enhance rather than remove. The most
successful way of correcting it is with a flatfield standard star (see item
flatfielding below).
-
Reticon data suffer from a periodic ripple which flatfielding does not
always remove adequately. For the `normal' 4- or 8-pixel ripple, the
command FILTER/RIPPLE can be tried. However, occasionally, also
very weird periods occur which can be identified in a Fourier transform
(see the various FFT/ commands). Spikes in the real part of the
Fourier transform can be removed with the interactive command
MODIFY/GCURSOR before the data is transformed back to the
original pixel space. Alternatively, FILTER/RIPPLE can be used
once the period is known. Note that these corrections have to be
done prior to any rebinning (wavelength calibration in particular)!
-
The dome flatfield lamp has an emission line at 670.7 nm, apparently due to
lithium. Other deviations from a pure continuum are not known, but you may
wish to watch for them (this advice applies also to the `internal' FF lamp).
- Some of the strongest absorption lines appear not only in
stellar but occasionally also in flatfield exposures!
- Background determination:
- For Reticon data, separate observations
are required. On CCD spectra, the various background components (bias,
scattered light, ghosts, sky, etc.) can usually be estimated from the
signal on either side of the object spectrum.
Unless you are sure that features in the background spectra are significant,
only subtract the mean value as number or a strongly smoothed background
spectrum in order not to add noise to your object spectra.
- Flatfielding:
-
- `Internal' FF lamp
- is perfect (apart from vignetting, see above)
for the Reticon and usually fully adequate for CCD spectra over most
of the wavelength range accessible in the blue pass of the CES. In the red,
the phases of the fringes in flatfield and object spectra may be so
different that division by such a `flatfield' only makes things much worse.
- Dome flats:
- Observers have reported that the position of fringes may
depend slightly on telescope position.
- Bright stars
- without disturbing spectral features and if observed
sufficiently close to the target in both position and time, may be used for
three purposes:
- High spatial frequencies:
- For this application, the spectrum of the
flatfield star must have been trailed so that its well exposed part
(along the spatial axis) fully covers the relevant positions of object
spectra. Division of the object spectrum by the standard spectrum (both
assumed bias corrected) will also take care of low spatial frequencies and,
perhaps, telluric features as described below.
- Low spatial frequencies:
- Command NORMALIZE/SPECTRUM can be
used to obtain an approximation to the continuum of the comparison star.
Division of this curve into the extracted target spectra will be useful
in correcting for vignetting problems and other residual curvatures
(echelle ripple) of the flatfielded spectra.
- Telluric lines
- can, with some luck, be removed by dividing
the extracted object spectrum by a suitably normalised extracted flatfield
star spectrum. Wherever possible, this should be done prior to wavelength
calibration.
- Wavelength calibration:
-
If you have to be worried about artifacts introduced by the non-linear
rebinning, try to be innovative and do not rebin your data at all! If this
is not practical, the following details should be considered:
-
For high precision, always flatfield your arc spectra.
-
Almost the only relevant comparison source is a thorium lamp. Do not use the
argon lines; the lamp contains argon only to start the gas discharge process.
-
By far the best laboratory wavelengths are those by Palmer and Engleman (1983).
Their list is availabe as MIDAS table TH in directory MID_ARC.
Copy this table to your MID_WORK, use SELECT/TAB on column :WAVE
and COPY/TT to reduce the size of the table to the range in wavelength
of your spectra, then delete the first copy to recover disk space.
-
The information given in the descriptor O_COM about the wavelength
(i.e., `CRVALX') of the central pixel (i.e., `CRPIXX') and the mean channel
width (i.e., `CDELTX') usually is extremely reliable and therefore very
useful in the interactive part of identifying the comparison lines.
-
Select the threshold in the line searching step so that about 10-25 lines
are detected. (As a rule of thumb, it is for normally exposed arc spectra
both necessary and sufficient to use all lines which in column :INTENSITY
of table TH are listed with a laboratory strength of about 3 units
or more.) Only the GAUSSIAN option in command SEARCH/LINE
will give useful line positions.
-
Identify ( IDENTIFY/LINE) about five lines interactively; they should be
well distributed over the wavelength range you are interested in.
-
Always start with a parabola for the approximation of the dispersion
curve ( CALIBRATE/WAVE), never user polynomials with degree > 3.
-
For normally exposed arc spectra, the automatic identification should
identify most of the lines which in table TH are listed (column
:INTENSITY) with a laboratory strength of 3 units or more (but, of course,
reject blends).
-
With the Long Camera, the rms scatter of the computed wavelengths about a
fitted second-order polynomial would usually be 1--2 10 nm (if not
better). In spectra taken with the Short Camera, the corresponding value
may be about two times higher.
-
Rebin your spectra to a step in wavelength which is at least two times
smaller than the detector pixel width.
-
For the rebinning of such very high resolution spectra it is important
that the descriptors START and STEP and the relevant variables
of programs are of double precision (often applies also to the subsequent
analysis of the calibrated spectra). If in doubt or in order to check
possible problems, suppress the leading two digits from the laboratory
wavelengths (e.g., COMPUTE/TABLE) and later re-introduce them in
descriptor START of the calibrated spectra.
.
- Flux calibration
- is not possible for CES spectra unless you
managed to observe a standard star with flux data that is extremely well
sampled in wavelength.
Next: Echelle Reduction
Up: No Title
Previous: CCD-Commands Summary
Pascal Ballester
Tue Mar 28 16:52:29 MET DST 1995