MOS Data Reduction Michael Balogh University of Durham Outline 1. 2. 3. 4. 5. (Automatic) identification of slits and galaxies Distortion correction Background subtraction Wavelength calibration Flat fields and flux calibration.

Report
MOS Data Reduction
Michael Balogh
University of Durham
Outline
1.
2.
3.
4.
5.
(Automatic) identification of slits and galaxies
Distortion correction
Background subtraction
Wavelength calibration
Flat fields and flux calibration
Data Reduction software
1. IRAF: Can deal with multiobject spectroscopy, but
handles the following inelegantly:
•
•
wavelength calibration
distortion corrections
2. Dan Kelson’s recently public software:
http://www.ociw.edu/~kelson/
•
•
•
•
designed for use with MOS data
handles wavelength calibration and distortion corrections
easily
Employs new technique for optimal background subtraction
But is somewhat obscure
Note: neither software package deals easily with ultraplex
data
MOS data: the spectra
MOS data: flats
MOS data: arc lamps
Ultraplex data
Identification of Objects
Identification of Objects: IRAF
Identification of Objects: IRAF
Interactively identify
object(s) in each slit
Specify extent to
extract in 1D spectrum
Can be tricky for faint
spectra because
optimal columns to
extract will vary from
slit to slit (in some
cases will hit bright sky
lines, in other cases
miss bright part of
spectrum)
Identification of Objects: Kelson
Identify slits in flat
field image
Laplacian filter helps
define slit edges
Pick object location on
2D image (using ds9,
for example)
Kelson 2003
Distortion correction
NIRSPEC: Kelson 2003
Spectra are usually curved, due to instrument distortions
Distortion correction
Two options:
1. Rectify image before extracting spectra.
Makes reduction easier, but introduces residuals
in sky subtraction.
2. Measure distortion, but extract spectra from
original frame and map to rectified coordinate
frame.
Distortion correction: IRAF
Curvature in spatial
direction is tricky to
correct; not easily
implemented.
Curvature in spectral
direction can be traced
when extracting
spectrum. Must be
done interactively and
probably not used
when extracting arc
spectrum
Need to be able to see
the spectrum…
d
d
Distortion Corrections: Kelson
1. Trace slits in flat field to map distortion in
spectral direction
2. For each slit, trace sky lines (or arc lines) to
map distortion in spatial direction
Kelson 2003
Background Correction
Background Correction
Usual procedure:
Define background
region on either side of
object
Fit polynomial across
dispersion
Assumes no distortion
in spatial direction, so
must correct first
Background Correction
Rebinning: introduces correlated noise, smears bad pixels, produces
artifacts/residuals, and forces sky spectrum to have common pixelization
Instead: perform least-squares fit to sky spectrum in original coordinates. This
provides better sampling in rectified coordinates.
Kelson, PASP, in press
Background Correction
Rebinning: introduces correlated noise, smears bad pixels, produces
artifacts/residuals, and forces sky spectrum to have common pixelization
Instead: perform least-squares fit to sky spectrum in original coordinates. This
provides better sampling in rectified coordinates.
Kelson, PASP, in press
Background Subtraction
2D LRIS spectrum
Spectrum profile in
rectified coordinates
Compare smoothed
version of above with
profile from single
pixel width
Kelson, PASP in press
Background Correction
1. Define sky regions (either directly, or using sclipping techniques)
2. Fit bivariate B-spline (Dierckx 1993) as a
function of rectified coordinates
•
Essentially approximates an interpolating spline
along the wavelength coordinate, but with much
finer sampling than available in a single CCD row
3. Can generalize further and fit simultaneously to
all spectra in a frame. Thus get improved
resolution even if distortions are small.
Kelson, PASP, in press
LRIS Raw
Sky model
Background
subtracted
rms-smoothed,
divided by
noise: no
residuals!
Kelson, PASP, in press
NIRSPEC Raw
Sky model
Background
subtracted
rms-smoothed,
divided by
noise: no
residuals!
Kelson, PASP, in press
Wavelength calibration
Wavelength calibration I
Extract arc lamp spectrum
for each slit
IRAF: identify a few lines
and fit low-order function.
Then easy to use this fit to
find more lines and
improve quality of the fit.
Task reidentify to find arc
lines in other slits on same
image does not work well.
Usually have to do each
slit separately.
Not clear to me if this
uses trace information
from spectrum.
Wavelength calibration II
Kelson (2003) software
Automatically identify lines in all slits, and
computes pixel-wavelength transformation
Don’t know how it works, but it does! Can do
in minutes what used to take me days with
IRAF.
Flat fielding
Flat fielding
1. Remove the
“slit
function”:
variation in
sensitivity
along the slit
Needed to correct
for uneven
slits
Flat fielding
2. Remove the
“blaze”:
variation in
sensitivity in
dispersion
direction
Needed for flux
calibration,
unless star
observed in
every slit
Flat fielding
3. Remove
pixel-topixel
sensitivity
variations.
Usually
introduces a
lot of noise
Flux Calibration
1. Observe photometric standard through one (or
more) slits
2. Reduce normally, and flat field (remove “blaze”
function)
3. Divide by known spectral shape to get detector
response as function of wavelength.
Conclusions
IRAF
Kelson
Advantages
• Lots of documentation
• Most parameters are easily
understood and located
Disadvantages
• Wavelength calibration and
distortion corrections are
difficult and time consuming
• Cannot easily produce 2-D
calibrated images
• For well-behaved data,
wavelength calibrations and
distortion corrections are easy
• Potential for improved
background subtraction
• Allows easy production of 2dimensional reduced images
• Little interaction => fast
processing
• Very little documentation
• Non-trivial to install (uses
Python, VTK, other software)
For LDSS2 spectra, I find both give similar quality results

similar documents