Proposal for a self-calibrating and instrument-independent MOS DRS

1
Proposal for a self-calibrating and
instrument-independent MOS DRS

• Carlo Izzo

2
MOS arc lamp exposure FORS2-MXU, GRIS_150I27
3
Using first-guess models to find reference lines
4
Earthquake!
5
High Expectations

• Traditional data reduction techniques are
  based on first-guess distortion models.
• The instrument is stable, together with its
  components (grisms, masks, filter wheels, etc.)
• The code can be kept simple, because the
  reference patterns in the calibration exposures
  (flats, arcs) are safely identified
• The procedures will be general, because different
  distortion models can be stored in appropriate
  configuration files

6
The hard reality

• The instrument is not stable, changing in the
  short and long time scales, requiring continuous
  maintenance work on the configuration files
• The code cannot be kept simple, because the
  reference patterns are not safely identified
  (e.g., due to unexpected contaminations, or to
  instrument instabilities)
• The procedures cannot be kept general, because
  the ad hoc solutions adding robustness to the DRS
  are typically instrument-dependent

7
VIMOS
8
Interactive systems
9
First-guess vs pattern-recognition
10
Distortion models parameters
11
DFO reports a problem
12
and the problem is solved
13
but whats the use of it?

• Using first-guess model
• The pipeline may stop with a generic calibration
  failed
• The pipeline may find a wrong solution
• The QC1 parameters may show nothing strange

• Using pattern-recognition
• The pipeline always completes successfully
• The pipeline always finds the right solution
• The QC1 parameters report exactly what happened

14
Do we need a physical model of the distortions?

• YES! A physical model of the optical distortions
  is necessary for comparing the expected
  distortions with the observed ones (instrument
  health monitoring)
• BUT We should not use the model of the expected
  distortions as a first-guess (even if we may use
  it for fitting the data)
• ALSO A physical model of the instrument
  distortions is necessary for a meaningful
  instrument health monitoring

15
Fix the models, or fix the instrument?

• In principle, the instrument should be fixed.
• In practice, it is often necessary to fix the
  models because
• To fix the instrument is not always immediate
  (see for instance the light contaminations in
  FORS, or the flexures in VIMOS), and in the
  meantime we must keep reducing the data
• Sometimes the real optical distortions are
  accepted, even if they are far from the
  instrument original design
• Using a pattern-recognition approach we would
  not need to fix models anymore!

16
Looking for patterns

• The pattern wavelengths
• 5400.562
• 5460.742
• 5764.419
• 5769.598
• 5790.656
• 5852.488
• 5875.620
• 5881.900
• 5944.830

• The data pixel positions
• 1220.64
• 1253.23
• 1299.44
• 1304.07
• 1339.30
• 1400.33
• 1450.28
• 1457.32
• 1471.00
• 1496.21
• 1520.66
• 1522.44
• 1549.01

17
Looking for peaks

• ________________________________________

18
Looking for peaks

• Any local maximum identifies a peak
• A peak positions is determined by parabolic
  interpolation of the three nearby pixel values

19
Looking for peaks
20
A simple case FORS2-LSS GRIS_1200R

• _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _
  _ _

21
Identifying arc lamp lines
22
Wavelength calibration

• Mean accuracy 0.05 pixel

23
Resampled spectrum

• Mean accuracy 0.05 pixel

24
Another case VIMOS GRIS_HRred
25
Wavelength calibration
Mean accuracy 0.07 pixel
26
Another case FORS1-MOS GRIS_300V
27
Wavelength calibration
Mean accuracy 0.09 pixel
28
Another case FORS2-MXU GRIS_150I
29
Wavelength calibration

• Mean accuracy 0.05 pixel

30
Identifying the slits

• Select the reference wavelength in this
  example, l 7000.00

31
Rectified image
32
Accuracy
33
Accuracy

• The accuracy of the extraction mask depends on
  many factors
• Number of fitted points,
• Accuracy of peaks positions,
• Appropriate choice of fitting models,
• Position along the spectral interval,
• but, above all,
• Correct identification of the detected peaks.
• Inaccuracy comes from misidentification!

34
This system is flexible

• Any MOS arc lamp exposure can be wavelength
  calibrated (instrument-independency)
• This method can also be directly applied to the
  scientific exposures (if the sky is visible and
  there are enough sky lines)
• This method may even be applied to intermediate
  products from any kind of spectroscopic data (not
  just MOS, but also IFU, echelle, etc.).

35
GIRAFFE Medusa1_H525.8nm
Mean accuracy 0.10 pixel
36
Other issues

• After the extraction mask is completely
  defined, the usual reduction steps can be
  applied
• Object detection,
• Determination of the sky spectrum,
• Optimal extraction,
• Combining different spectra,
• Error propagation

37
THE END