ESO
SINFONI Pipeline User Manual
Doc:
VLT-MAN-ESO-19500-3600
Issue:
Issue 1.0
Date:
Date 2005-10-19
Page:
79 of 88
10.2.2
Master dark and bad pixel map determination: si_rec_mdark
A set of input raw dark frames is stacked in a cube. An average with rejection (parameters bp_noise.low_rejection
and bp_noise.high_rejection), yelds a mean and a standard deviation, stdev. Pixels which deviate from the
mean more than a user defined factor (dark.threshold_sigma_factor) times the stdev are flagged as bad pixels.
This results in a bad (hot) pixel map which flags pixels with a high dark current.
A set of input raw dark frames is sorted according to DIT thus generating corresponding groups. Then an
average with rejection (controlled by parameters dark.low_rejection and dark.high_rejection) is computed
within each group of frames. This results in a master dark frame for each DIT.
On each possible pair of consecutive raw frames the read-out noise is determined in a region defined by
the parameters dark.qc_ron_xmin, dark.qc_ron_xmax, dark.qc_ron_ymin, dark.qc_ron_ymax, and using
dark.qc_ron_nsamp random samples each of size dark.qc_ron_hsize as described in 10.1.6. On the mas-
ter dark the fixed pattern noise is determined in two regions defined by the parameters dark.qc_fpn_xmin,
dark.qc_fpn_xmax, dark.qc_fpn_ymin, dark.qc_fpn_ymax, using dark.fpn_ron_nsamp random samples
each of size dark.fpn_ron_hsize as described in 10.1.7.
10.2.3
Master flat and threshold pixels (bad pixel map) determination: si_rec_mflat
The input flat field frames are stacked. An average with rejection (parameters lamp_flats.low_rejection
and lamp_flats.high_rejection) is computed to remove dynamic bad pixels (either cosmic rays or tran-
sient bad pixels). The mean lamp-off frame is subtracted from the mean lamp-on frame.
If lamp_flats.bad_ind==TRUE the intensity tilt of each column is removed (the fit of the pixel intensity
is subtracted from the pixel intensity) considering in this operation only pixels whose intensity differs
from the linear fit value by no more than lamp_flats.sigma_factor times the sigma of the pixel intensity.
To find the strong intensity deviations of bad pixels a threshold value must be found. For this reason, on
a rectangular region defined by parameters lamp_flats.llx, lamp_flats.lly, lamp_flats.urx,
lamp_flats.ury, the recipe computes a clean_mean of the intensity (lamp_flats.bad_low_rejection and
lamp_flats.bad_high_rejection) and its clean standard deviation clean_stdev (to have an estimate of the
noise variations in the flat field). The threshold value is given by the product clean_stdev*lamp_flats.factor.
If lamp_flats.thresh_index==TRUE the image corrected for the intensity tilt is further filtered, indicating
as bad pixels the ones which lie outside the intensity range [clean_mean-lamp_flats.mean_factor*clean_stdev].
Else no filter is applied. This results in a reference image. A median filter with a radius equal to
clean_stdev*lamp_flats.factor is applied for lamp_flats.iterations iterations to remove small clusters
of bad pixels. Finally pixels which have different values are promoted to bad pixels by comparing the
median filtered image with the reference image.
A master flat field is determined. If lamp_flats.interpol_ind==TRUE, using the input bad pixel mask,
and the slitlets position information, bad pixels are interpolated over a given radius lamp_flats.max_rad.
Then the intensity is normalised to that of the central pixel.
For QC purposes the fixed pattern noise is monitored on the resulting master flat field over two rectangular
regions defined by parameters lamp_flats.qc_fpn_xmin1, lamp_flats.qc_fpn_xmax1,
lamp_flats.qc_fpn_ymin1, lamp_flats.qc_fpn_ymax1, and lamp_flats.qc_fpn_xmin2,
