proposal for a self-calibrating and instrument-independent mos drs carlo izzo
TRANSCRIPT
Proposal for a self-calibrating and Proposal for a self-calibrating and instrument-independent MOS DRSinstrument-independent MOS DRS
Carlo IzzoCarlo Izzo
MOS arc lamp exposure: MOS arc lamp exposure: FORS2-MXU, GRIS_150I+27FORS2-MXU, GRIS_150I+27
Using first-guess models to Using first-guess models to find reference lines…find reference lines…
Earthquake!Earthquake!
High ExpectationsHigh Expectations Traditional data reduction techniques are based on first-Traditional data reduction techniques are based on first-
guess distortion models.guess distortion models.
• The instrument is stableThe instrument is stable, , together with its components together with its components (grisms, masks, filter wheels, etc.)(grisms, masks, filter wheels, etc.)
• The code can be kept simpleThe code can be kept simple, because the reference , because the reference patterns in the calibration exposures (flats, arcs) are safely patterns in the calibration exposures (flats, arcs) are safely identifiedidentified
• The procedures will be generalThe procedures will be general, because different , because different distortion models can be stored in appropriate configuration distortion models can be stored in appropriate configuration filesfiles
The hard realityThe hard reality
• The instrument is not stableThe instrument is not stable, changing in the short and , changing in the short and long time scales, requiring continuous maintenance work long time scales, requiring continuous maintenance work on the configuration fileson the configuration files
• The code cannot be kept simpleThe code cannot be kept simple, because the reference , because the reference patterns are not safely identified (e.g., due to unexpected patterns are not safely identified (e.g., due to unexpected contaminations, or to instrument instabilities)contaminations, or to instrument instabilities)
• The procedures cannot be kept generalThe procedures cannot be kept general, because the , because the ad hocad hoc solutions adding robustness to the DRS are typically solutions adding robustness to the DRS are typically instrument-dependentinstrument-dependent
VIMOSVIMOS
Interactive systemsInteractive systems
First-guess vs pattern-recognition
Distortion models parametersDistortion models parameters
DFO reports a problem…DFO reports a problem…
… … and the problem is “solved”and the problem is “solved”
… … but what’s the use of it?but what’s the use of it?
Using first-guess model…Using first-guess model…
• The pipeline may stop with a The pipeline may stop with a genericgeneric ““calibration failedcalibration failed””
• The pipeline may find a The pipeline may find a wrong solutionwrong solution
• The QC1 parameters may The QC1 parameters may show nothing strangeshow nothing strange
Using pattern-recognition…Using pattern-recognition…
• The pipeline always completes The pipeline always completes successfullysuccessfully
• The pipeline always finds the The pipeline always finds the right solutionright solution
• The QC1 parameters report The QC1 parameters report exactly what happenedexactly what happened
Do we need a physical model of the Do we need a physical model of the distortions?distortions?
YES!YES! A physical model of the optical distortions is necessary A physical model of the optical distortions is necessary for comparing the expected distortions with the observed for comparing the expected distortions with the observed ones ones (instrument health monitoring)(instrument health monitoring)
BUT:BUT: We should We should notnot use the model of the expected distortions use the model of the expected distortions as a first-guess as a first-guess (even if we may use it for fitting the data)(even if we may use it for fitting the data)
ALSO:ALSO: A A physicalphysical model of the instrument distortions is model of the instrument distortions is necessary for a meaningful instrument health monitoringnecessary for a meaningful instrument health monitoring
Fix the models, or fix the Fix the models, or fix the instrument?instrument?
• In principleIn principle, the instrument should be fixed., the instrument should be fixed.
• In practiceIn practice, it is often necessary to fix the models because:, it is often necessary to fix the models because:
• To fix the instrument is not always immediate (see for instance To fix the instrument is not always immediate (see for instance the light contaminations in FORS, or the flexures in VIMOS), and the light contaminations in FORS, or the flexures in VIMOS), and in the meantime we must keep reducing the datain the meantime we must keep reducing the data
• Sometimes the real optical distortions are “accepted”, even if Sometimes the real optical distortions are “accepted”, even if they are far from the instrument original designthey are far from the instrument original design
Using a pattern-recognition approach we Using a pattern-recognition approach we would not need to fix models anymore!would not need to fix models anymore!
Looking for patternsLooking for patterns
The pattern: The pattern: wavelengthswavelengths• ……• 5400.5625400.562• 5460.7425460.742• 5764.4195764.419• 5769.5985769.598• 5790.6565790.656• 5852.4885852.488• 5875.6205875.620• 5881.9005881.900• 5944.8305944.830• ……
The data: The data: pixel positionspixel positions• ……• 1220.641220.64• 1253.231253.23• 1299.441299.44• 1304.071304.07• 1339.301339.30• 1400.331400.33• 1450.281450.28• 1457.321457.32• 1471.001471.00• 1496.211496.21• 1520.661520.66• 1522.441522.44• 1549.011549.01• ……
Looking for peaksLooking for peaks
________________________________________
Looking for peaksLooking for peaks
• Any local maximum identifies a peakAny local maximum identifies a peak• A peak positions is determined by parabolic interpolation of A peak positions is determined by parabolic interpolation of
the three nearby pixel valuesthe three nearby pixel values
Looking for peaksLooking for peaks
A simple case: FORS2-LSS A simple case: FORS2-LSS GRIS_1200RGRIS_1200R_ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
Identifying arc lamp linesIdentifying arc lamp lines
Wavelength calibrationWavelength calibration
Mean accuracy: 0.05 pixelMean accuracy: 0.05 pixel
Resampled spectrumResampled spectrum
Mean accuracy: 0.05 pixelMean accuracy: 0.05 pixel
Another Another case: VIMOS case: VIMOS GRIS_HRredGRIS_HRred
Wavelength Wavelength calibrationcalibration
Mean accuracy: 0.07 pixelMean accuracy: 0.07 pixel
Another case: Another case: FORS1-MOS GRIS_300VFORS1-MOS GRIS_300V
Wavelength calibrationWavelength calibration
Mean accuracy: 0.09 pixelMean accuracy: 0.09 pixel
Another case: FORS2-MXU Another case: FORS2-MXU GRIS_150IGRIS_150I
Wavelength calibrationWavelength calibration
Mean accuracy: 0.05 pixelMean accuracy: 0.05 pixel
Identifying the slitsIdentifying the slits
Select the reference wavelength: in this example,Select the reference wavelength: in this example, = 7000.00= 7000.00
Rectified imageRectified image
AccuracyAccuracy
AccuracyAccuracy
The accuracy of the extraction mask depends on many factors:The accuracy of the extraction mask depends on many factors:
• Number of fitted points,Number of fitted points,• Accuracy of peaks positions,Accuracy of peaks positions,• Appropriate choice of fitting models,Appropriate choice of fitting models,• Position along the spectral interval,Position along the spectral interval,
… … but, above all, but, above all,
• Correct identification of the detected peaks.Correct identification of the detected peaks.
Inaccuracy comes from misidentification!Inaccuracy comes from misidentification!
This system is flexibleThis system is flexible
• AnyAny MOS arc lamp exposure can be wavelength calibrated MOS arc lamp exposure can be wavelength calibrated (instrument-independency)(instrument-independency)
• This method can also be directly applied to the scientific This method can also be directly applied to the scientific exposures exposures (if the sky is visible and there are enough sky (if the sky is visible and there are enough sky lines)lines)
• This method may even be applied to intermediate products This method may even be applied to intermediate products from from anyany kind of spectroscopic data kind of spectroscopic data (not just MOS, but also (not just MOS, but also IFU, echelle, etc.).IFU, echelle, etc.).
GIRAFFE Medusa1_H525.8nmGIRAFFE Medusa1_H525.8nm
Mean accuracy: 0.10 pixelMean accuracy: 0.10 pixel
Other issuesOther issues
After the extraction mask is completely defined, the usual After the extraction mask is completely defined, the usual reduction steps can be applied:reduction steps can be applied:
• Object detection,Object detection,• Determination of the sky spectrum,Determination of the sky spectrum,• Optimal extraction,Optimal extraction,• Combining different spectra,Combining different spectra,• Error propagationError propagation