The CASSIS spectra will be updated if improvements in the algorithm
warrant a new release. We are therefore encouraging users to
provide feedback on the
quality of the spectra especially.
Foreword on what is meant by "publication ready"
The CASSIS spectra are supposed to be "publication ready" and "ready to use". Note that this doesn't mean there's no more work to do to achieve a scientific result. It means that users can publish the data because it corresponds to what was observed. However, it is still necessary to know how the spectrum was calculated. For instance, the slit orientations can be different between several observations of the same object, or the pointing can also be different (mispointings because of bad coordinates, mispointings because of bad acquisition. Also, it should be noticed that the SL and LL apertures have different sizes, and the orientation between the 2 apertures is close to 90 degrees. This means that for extended sources, the region probed in SL or LL is not the same.
CASSIS provides the extraction of the spectra as they were observed. Users are then free to apply some post-processing steps, but these steps are entirely motivated by scientific choices. This is the reason why these steps cannot be performed automatically in archives like CASSIS.
Primary changes in v7 (current version)
- New product as of July 14th, 2017. The default tapered column extraction integrates the flux in a spatial window that scales with the source spatial extent as inferred from optimal extraction (comparison between PSF and observed profile). The correction due to the light lost outside the slits is the point source correction by default. The pipeline also includes the (infinitely) extended source lightloss correction as well as an experimental partially-extended correction. The latter uses the spatial extent as inferred from optimal extraction, assumes the the source is circular, and broadens the 2D STINY_TIM PSF accordingly. The lightloss correction is then calculated for each wavelength, providing in effect a partially-extended flux calibration. An illustration can be seen for NGC595, an HII region in M33. The lightloss correction method can be chosen on the right panel. If selecting partially-extended source, the SL and LL spectra have different shapes (slopes) and they now stitch well. In some other cases, a residual stitching factor can still be necessary, but the overall spectrum shape is much more realistic. This product is an optional download product. The default spectrum has not changed, and there's just this new option when comparing extraction & background subtraction methods.
- Background subtraction. The local background that is removed for order-background images (see v6 changes below) is now also removed for nod-subtracted images. This change affects only very faint sources (i.e., fainter than the background difference between the on-source and off-source background) for which the nod-subtraction is chosen as the best background subtraction method.
- Tapered column extraction. For the tapered column extraction of nod-subtracted images, the spectrum can be uncertain is the source is extended since the window in which the flux is integrated could include part of the "negative" spatial peak (due to the nod image difference). In LR7, the spatial extent threshold - above which the use of tapered column extraction for nod-sutracted images is not reliable anymore - has been fine tuned from 1.75 times the PSF width to 1.5 times the PSF FWHM.
- Website. The slit orientation plot, which allows overlaying the SL and LL slits on archival images for any observation, was rotated with an erroneous angle. This is now fixed.
- Warning: The total uncertainty reported (column 3 in the spectral data file) is the sum of the uncertainty on the extraction (column 4) and the uncertainty related to differences between nod spectra (column 5). The errors should in fact be added in quadrature, which can be done by reading the columns for each uncertainty individually (columns 4 and 5) and combine them in quadrature.
Primary changes in v6
Primary changes in v5 as compared to v4 presented in the 2011 CASSIS paper
- Background subtraction. The other order images can be subtracted to remove rogue pixel as well as very extended emission such as the zodiacal dust emission. Since the other order images correspond to a spatial position significantly far from the source position (i.e., on the order of a full slit length), this can result in a local residual background emission in order-subtracted images. This sometimes led to some contamination in v5 spectra, especially for faint sources. The local background is now removed for order-subtracted images, by modeling a flat large-scale emission across the slit. The same local background is subtracted for the tapered column extraction and optimal extraction. In the case of tapered column extraction though, the local background is not removed if the source is very extended (since in this case the "local background" is likely originating from the source itself).
- Default extraction method. When the detection level is too low and no accurate spatial extent can be determined, optimal extraction is now chosen as the default spectrum as it is the method providing the best signal-to-noise ratio. The spatial extent is also now better determined.
- Image and spectra combination. When the dispersion of coordinates over the exposures is large, individual exposure images cannot be combined anymore and have to be extracted separately instead. The dispersion threshold has been increased in v6. Fixed a bug when multiple spectra for a given module/order/nod have to be combined (which happens only when the images could not be combined because of a significant dispersion of the pointing over the exposures).
- Default spectrum. The default spectrum is now trimmed (overlaps between spectral orders are removed).
- Spatial extent. Better determination of the spatial extent, a parameter used to decide the best extraction method between optimal for point sources and tapered column for partially-extended sources. The new spatial extent determination method is more robust and provides more accurate values when a significant background emission is present in the spatial profile.
Things to know & caveats
- Extraction method. CASSIS now chooses the best extraction method between optimal extraction and tapered column extraction, based on the source spatial extent. The default spectrum shown and the default products reflect this choice. The alternative method can still be accessed through the product menu.
- Error bars. v5 provides the
total error (RMS+systematic+calibration) by default. The individual
error types are provided as extra columns.
- S18.18.0 Calibration. v5
introduces the latest and final calibration files S18.18.
- Spike detection. Rejection
parameters and thresholds for spike detection in spectra and images were updated and improved.
- Image and spectra combination.
The combination of pixels now uses a resistant mean with outlier
rejection in v5 rather than a combination of the median and
error-weighted average in v4.
- Defringing Defringing in LL1 is
applied only when the S/N is larger than a given ratio (currently >5 in
both v4 and v5). In some cases, the S/N could not be determined
resulting in no defringind applied. This is now fixed.
- Shift between nod spectra. When
a serendipitous (contaminating) source is found in the slit, the local
background (modelled by a 0-order polynomial) subtraction is ignored
because it might introduce systematic errors. In some cases, only 1
nod is affected by this problem, while for the other one a local
background subtraction is performed. This implies that a shift can be
observed between the 2 nod spectra. This problem is now solved.
- Contaminating sources. The
presence of contaminating sources in the slit and in the background
image(s) is better constrained and less conservative than in v4. The
parameters were adjusted so that a contamination is identified as such only when
it affects significantly the intended source spectral extraction.
- Ignoring nod spectrum with a low detection
level. For each background subtraction method (by-order, by-nod,
in situ), the CASSIS algorithm compares the detection level of each
nod. If one nod is significantly worse than the other it is ignored
for the nod spectra combination. In v4, the same detection level was used for
all backgroung methods. This bug is corrected.
- Tapered column extraction.
Tapered column extractions in v4 are presented with the best
background subtraction found based on diagnostics from the optimal extraction (e.g., presence of contaminating sources).
This can cause a
problem for tapered column extraction of extended sources since the
by-nod subtraction will remove part of the source itself. In addition,
the tapered column extraction width scales with the source extent, so
the extraction aperture will include the "negative" source
corresponding to the source in the other nod position. v5 spectra now
assumes a by-order subtraction by default for tapered column extractions (or in situ subtraction ifa contaminating source is present).
- Uncertainties. The new version of the SMART/AdOpt package is used
for optimal extraction (v8.2.7). This version corrects for a bug in the flux uncertainty determination. The uncertainties were underestimated in v4, sometimes by several factor.
- Plots on the webpage. Plots had
some display bugs for the log axes. The detailed plots showing nod
spectra didn't have the 2nd x-axis (observed wavelength).
- SL/LL jump. It is common to
observe a jump between the SL spectrum and the LL spectrum, at around 14um. This is
observed in partially extended sources and is a result of the different
entrance slit dimensions across the spectrometer apertures.
The CASSIS spectra are unstitched. CASSIS provides by default the best extraction
method between optimal (for point sources) and tapered column (for partially-extended sources).
However, even when using tapered column extraction, it is not always possible to recover in the small
SL aperture the emission observed in the large LL aperture. For this reason, jumps are still often
observed in extended sources. Users have the possibility of scaling SL up to match the LL spectrum. We also provide an experimental flux calibration for extended sources, assuming a given source geometry. The resulting spectra are available on request only.
- Residual fringes. The LL1
defringing is a complex process that might require different fitting
parameters in some cases. A set of parameters was chosen for CASSIS to
accomodate most of the fringing patterns. Some residuals can
remain. CASSIS provides the fringed spectra as optional products if
the users desire to defringe the spectra themselves.
- Spikes. Although a particular
attention is given by CASSIS to the removal of bad pixels and spikes
in the spectra, it is possible that some spectra show spikes. For this
reason, it is advised to check the various versions of the spectra
(individual nod spectra, different background subtraction method,
tapered column extraction vs. optimal extraction).
- Coordinates The coordinates
found in the headers (RA_SM, DEC_SM) correspond to the extracted
source's coordinates. In the case of combined spectra (i.e., when 1 spectrum is provided per
module or per AORKEY) these coordinates might not be
accurate. We advise users to use the (RA_FOV, DEC_FOV) keywords in the
headers which correspond to the coordinates of the FOV.