Next: Automated Object Classification with ClassX
Up: Algorithms
Previous: Desmearing Solar System Objects in Chandra Data
Table of Contents -
Subject Index -
Author Index -
Search -
PS reprint -
PDF reprint
koda, P. & Hensberge, H. 2003, in ASP Conf. Ser., Vol. 295 Astronomical Data Analysis Software and Systems XII, eds. H. E. Payne, R. I. Jedrzejewski, & R. N.
Hook (San Francisco: ASP), 415
Merging of Spectral Orders from Fiber Echelle Spectrographs
Petr koda
Astronomical Institute of the Academy of Sciences of the Czech Republic
Herman Hensberge
Royal Observatory of Belgium
Abstract:
We review the data reduction of two fiber-based echelle
spectrographs (HEROS and FEROS) with emphasis on the similarity
of the inconsistencies between the overlap of spectral orders
before merging.
The literature on echelle data reduction shows that such
inconsistencies are commonly observed and usually handled by
rather heuristic interactive procedures.
For both instruments it seems to be the calibration unit that
introduces the bulk of the problem through errors in the flat field.
We discuss strategies to treat the problem and to remove the
inconsistencies before merging the spectral orders with minimal use
of interactive, subjective algorithms.
Merging spectral orders from echelle spectra is known to be
non-trivial. Some manuals of echelle spectroscopy even advise
against trying it. Authors who do the merging usually report problems
and try to solve the order inconsistencies by some heuristic interactive
method (Churchill 1994; Hall et al. 1994; Erspamer & North 2002).
Recently, De Cuyper & Hensberge (2003)
commented on a study of calibration flat-fields taken with
the FEROS fiber-fed echelle spectrograph at ESO and pointed
out several effects influencing the merging. In this contribution,
we present the results of an extended analysis including special
calibration flat-fields, dome flat-fields and science exposures
of bright objects. The results obtained from FEROS were compared,
where possible, to red-channel exposures made with HEROS at Ondejov
observatory.
A detailed description of both instruments can be
found on web pages of FEROS
and HEROS
respectively. FEROS is a 2-fiber spectrograph covering a wide spectral
region (3600-9000 Å) in 39 spectral orders. It is particularly
suited to distinguish effects that are related to blazed spectral orders
from optical projection effects since the bluest spectral orders are much
narrower than the size of the detector while the reddest orders are
covered over a bit less than their free spectral range. FEROS operates in
a temperature and humidity controlled room. HEROS is a one-fiber
instrument developed earlier by the same team in Heidelberg, as a compact
echelle spectrograph that has been used at several telescopes.
Since August 2000 it has been connected for more than two years to the 2m telescope of
Ondejov observatory. The light from the telescope Cassegrain focus is
fed by the 10m long fiber to the echelle grating and then, after the
beam-splitter, it goes to two independent channels: blue (3600-5600 Å
in 70 orders) and red (5800-8400 Å in 32 orders).
The projection of the spectral orders on the detector over an observing
run of 4 nights was stable in the wavelength direction at the level of 0.1
pixel (except for an explained, and meanwhile removed oscillation, due to
short-term temperature fluctuations of 1 K in the FEROS room) and in the
spatial (cross-order) direction at the level of 0.5 pixel (Figure 1).
Figure 1:
FEROS -- time evolution of the position of the spectral orders
in cross-order direction (lower part)
and the shift of the blaze function along the spectral order (upper part)
during an observing run. Flat-field images (+) and
science frames (o) are indicated with different symbols.
|
However, the position of the blaze profile in wavelength changed by of the
order of 10 pixels (i.e., of the wavelength) in a highly correlated
way with the changes in spatial direction (Figure 1). These slow temporal
changes apply as well to calibration unit flat-fields as to dome
flat-fields or scientific exposures. Such changes, if not taken into
account when flat-fielding a science frame, lead to inconsistencies at the
level of several percent of the flux in the overlap of spectral orders.
The data suggest a very similar shift of the blaze function over all orders in the case of FEROS, leading to larger overlap mismatch in the narrower orders in the blue spectral region i.e., with larger gradients in the blaze function.
Temporal changes in the blaze profile are detected on shorter time-scales,
which is presumably related to the fact that the spectrograph is operating in the dome
and not in a controlled room. The changes vary smoothly over
subsequent orders, but cannot be represented by a simple small shift in
wavelength of the blaze function. This may be related to the fact that the
blaze profiles produced by HEROS are not so near to the theoretical
predictions than in the case of FEROS: the blaze profiles of different
spectral orders are almost identical when expressed in the coordinate
where refers to the spectral order, to the wavelength and
to in order at the peak of the blaze
intensity. De Cuyper & Hensberge (2003) discuss the similar case for FEROS,
but the shape and the width do not scale accurately in the same way for HEROS.
Since flat-fields are commonly taken with each science exposure at
Ondejov (because of the fast low-frequency temporal changes of the
calibration images), efforts to address the order merging problems were
directed to the study of the unblazed science frames rather than
considering the calibration images and the science frames separately, as
in the FEROS case.
In order to visualize the lack of consistency in the order overlap regions
more clearly, we present figures where the global wavelength dependency
of instrument and object is removed from the separate orders. This step,
the normalization of the merged spectrum, comes last in a real data reduction
chain.
Figure 2 shows the separate spectral orders and the level of inconsistency
in the regions of spectral order overlap. Overplotted is a correction
function with identical shape (in pixel space) in all 32 spectral orders,
but smoothly varying amplitude. Dividing by this function reduces
the inconsistencies in the overlap of spectral
orders to well below the 1% level.
Figure 2:
HEROS -- seven spectral orders starting from the longward wing of H before (left panel) and after (right panel) correction
|
Analysis of data obtained by the fiber echelle
spectrographs FEROS at ESO, La Silla, Chile and HEROS at Ondejov, Czech Republic,
identifies the high sensitivity of the shape and position of the blaze
function as the primary source of order overlap inconsistencies. Since
changes in the blaze function are very consistent over many spectral
orders, and highly correlated with positional changes in the projection of
orders on the detector, a robust empirical model can be developed.
However, on a longer term, the origin of these effects should be
understood such that action can be taken to stabilize the blaze function
sufficiently.
If sufficient attention is paid to understand the
calibration unit, in order not to introduce spurious low-frequency
patterns in the flat-fielded science data, merging spectral orders becomes
a trivial exercise.
Acknowledgments
Part of this research is done in the framework of
the IUAP P5/36 project
financed by the Belgian Federal DWTC/SSTC and project K2043105 of the
Academy of Sciences of the Czech Republic.
References
Churchill, C. W. 1994, Lick Obs. Techn. Rep., 74, 1
De Cuyper J. P. & Hensberge, H. 2003, in ASP Conf. Ser., Vol. 281, Astronomical Data Analysis Software and Systems
XI, ed. David A.
Bohlender, Daniel Durand and T. H. Handley (San Francisco: ASP), 324
Erspamer, D. & North, P. 2002, A&A, 383, 227
Hall, J. C., Fultoni, E. E., Huenemoerder, D. P., Welty, A. D. & Neff, J. E. 1994, PASP, 106, 315
© Copyright 2003 Astronomical Society of the Pacific, 390 Ashton Avenue, San Francisco, California 94112, USA
Next: Automated Object Classification with ClassX
Up: Algorithms
Previous: Desmearing Solar System Objects in Chandra Data
Table of Contents -
Subject Index -
Author Index -
Search -
PS reprint -
PDF reprint