ESO
SINFONI Pipeline User Manual
Doc:
VLT-MAN-ESO-19500-3600
Issue:
Issue 1.0
Date:
Date 2005-10-19
Page:
72 of 88
The recipe finally determines several lists of line positions:
– a line list of row indeces for the line positions;
– a clean list of well separated line positions;
– a list of wavelengths corresponding to the raw positions;
– a list of wavelengths corresponding to the clean raw positions.
Those lists allow to associate each emission line in each spectrum with an approximate pixel position and
the exact wavelength.
10.1.9
Dispersion relation and wavelength map determination
A Levenberg-Marquardt fit of a Gaussian profile of given fwhm is performed considering the raw posi-
tions which lie in regions defined by boxes of half-size half_width, each centered at the line positions
determined by the algorithm described in 10.1.8. The fit is restricted to the lines whose amplitude is
greater than min_amplitude with respect to the background. This fit allows to determine each line posi-
tion with sub pixel accuracy.
Detector defects may lead to erroneous line detections. To prevent these misidentifications the fit is
limited to lines which lie no more than pixel_tolerance pixels from the corresponding position, obtained
assuming a linear dispersion guess_disp1 model. Possible outliers are flagged by setting the line-position
parameter of the fit to zero.
For each image column, a polynomial fit of degree n_a_coefficients and coefficients
+
'
is performed,
so as to determine the dispersion relation between the listed wavelength values
!
and the Gauss-fitted
positions
>
_
+/
!
for each image column using the singular value decomposition method.
!
$
_
_
+
'
>
_
+/
!
Data points which lie beyond max_residuals pixels from the corresponding fit value are rejected. In
the fit each data point is weighted by a factor proportional to the product of the guess dispersion value
guess_disp1 and the error associated with each position point from the previous Gaussian fit.
Then the positions of the slitlet edges are determined. Offset coefficients
+
'
resulting from the poly-
nomial fit for adjacent columns are compared. When they differ by more than pixel_dist*guess_disp1,
it is assumed that this is due to the crossing of a slitlet’s edge. Here pixel_dist indicates the minimal
distance in pixels between adjacent slitlets measured along the wavelength direction. This provides a first
approximation of the slitlets’ edges.
A clean average of the coefficients is determined by rejecting the extreme 10% low and high values and
performing a kappa-sigma clipping, where the constant kappa is set by sigma_factor.
Then a singular value decomposition fit of the polynomial coefficients is performed across the spatial
extent of each slitlet using a polynomial of low degree n_b_coefficients and the smoothed polynomial
coefficients
+
'
6
3
are computed anew.
