Improve N-band background subtraction.

11_0-critical_algorithms.tex

@@ -219,16 +219,24 @@ \subsubsection{LM bands}
 The best approach to determine the flux scaling factor of each dither exposure is yet to be determined (e.g.\ by median scaling or by a 2D/PCA approach based on~\cite{Hunziker2018}).
 The pipeline would also support sky subtraction using separate sky images for out-of-field dithering modes, i.e.\ when the observed field is crowded or full of extended emission.
 
-\subsubsection{N bands}
-To ensure proper background sampling at the longer wavelengths, small ($< 1000 \arcsec$) and fast ($\approx 10 \mathrm{Hz}$) chopping offsets of the field with respect to the N IMG detector are expected to be performed by the METIS- internal chopper mirror.
+\subsubsection{N bands}\label{sssec:nbandsbackgroundsubtracion}
+To ensure proper background sampling at the longer wavelengths, small ($< 1000 \arcsec$) and fast ($\approx 10 \mathrm{Hz}$) chopping offsets of the field with respect to the N IMG detector are expected to be performed by the METIS-internal chopper mirror.
 The chopping residuals from the previous step will be corrected by nodding the telescope to a different position and repeating the same chopping procedure.
-The chopping and nodding sequence is expected to follow what is currently being done with VISIR.
+%The chopping and nodding sequence is expected to follow what is currently being done with VISIR.
+A 3-point parallel nodding sequence is followed, which should give a higher \ac{SNR} than the 2 point pattern used in e.g.~VISIR (cf.~\cite{METIS-operational_concept}).
+The
+
 The pipeline will determine the sky background from the chopped and nodded images and produce background maps without the sources and background-subtracted versions of the maps with the sources.
 For each nodding cycle the subsequent chop-cycle frames are subtracted from each other (mean of on-source frames minus mean of off-source frames) resulting in a single chop-difference frame, i.e. nod half-cycle frame.
 Then for subsequent nodding cycles the nod half-cycle frames are subtracted from each other resulting in a nod-difference frame.
 The mean of all nod-difference frames is the final background-subtracted image.
 Additionally, the optimal extraction (PSF-weighting) will be used to subtract the positive/negative source images from the final background-subtract image (2/3/4 depending on the relative chopping/nodding directions) to produce a single stacked image of the sources of interest.
 
+There might be extensive spatial modulation (dithering) within a nod half cycle.
+That is, the 3-point chop pattern is repeated a few times, then the slightly move the central position is moved slightly after which the 3-point chop pattern is repeated.
+This is done several times in a nod-half cycle, and then repeated in the same pattern in the other nod position after the nod is performed.
+This is to mitigate the (many) bad/hot pixels in the GeoSnap array, therefore chop-cycle frames will be subtracted such that the bad pixels can be worked around.
+
 \subsection{Wavelength calibration and distortion correction}\label{ssec:criticalwavelengthanddistortion}
 The wavelength calibration and distortion correction critical algorithm is split in a LSS part (Section~\ref{ssec:criticalwavelengthanddistortionlss}) and an IFU part (Section~\ref{ssec:criticalwavelengthanddistortionifu}).
 

Overview_IMG_LM_N.tex

@@ -1,14 +1,15 @@
 \subsection{Imaging in LM and N}
 \label{ssec:overview_lm_imaging}
 
-\textbf{Note: The pipeline layout has been modified compared to the
-  PDR design in order to achieve better modularity. Basic reduction
-  and background subtraction have been split into two recipes that now
-  are applied to both standard calibration and science data. ADI recipes have been added since PDR however integration of \ac{HCI}
-  into this workflow requires more work: \ac{HCI} images will be treated
-  the same way at least through basic reduction, possibly through
-  background subtraction. \ac{ADI} combination may require a separate
-  recipe, at least for some \ac{HCI} configurations.}
+% HB 20230731: This note is not really necessary I think.
+%\textbf{Note: The pipeline layout has been modified compared to the
+%  PDR design in order to achieve better modularity. Basic reduction
+%  and background subtraction have been split into two recipes that now
+%  are applied to both standard calibration and science data. ADI recipes have been added since PDR however integration of \ac{HCI}
+%  into this workflow requires more work: \ac{HCI} images will be treated
+%  the same way at least through basic reduction, possibly through
+%  background subtraction. \ac{ADI} combination may require a separate
+%  recipe, at least for some \ac{HCI} configurations.}
 
 The purpose of the pipeline is to correct or remove contributions from
 the instrument, telescope, and atmosphere and produce science-grade
@@ -28,13 +29,13 @@ \subsection{Imaging in LM and N}
 way the data have to be reduced.
 
 The GeoSnap detector has more stable gain than AQUARIUS detector,
-which was still in the baseline at PDR.  Chopping is still necessary,
-albeit at a lower frequency of a few Hz, and the standard chop/nod
-technique will be employed for background subtraction.  As the dark
+which was still in the baseline at PDR\@.  Chopping is still necessary, albeit
+at a lower frequency of a few Hz, and a chop/nod technique, which meets the
+specific ELT requirements (Section~\ref{sssec:nbandsbackgroundsubtracion}),
+will be employed  for background subtraction.  As the dark
 signal is automatically removed when the exposures from the different
 chop and nod positions are combined no master dark is required for the
-reduction of science data. Flat fielding may be possible, pending
-further investigation of the detector stability
+reduction of science data. The GeoSnap data is also flat fielded.
 
 Observations and reduction of LM band data with the HAWAII2RG detector
 can proceed as in the near infrared. After dark subtraction and