Diapositiva 1

1
De-biasing interferometric visibilities in
AMBER/VLTI data of weak sources
G. Li Causi1, S. Antoniucci21, E. Tatulli3
1 INAF-Osservatorio Astronomico di Roma, Via di
Frascati 33, I-00040 Monteporzio Catone 2
Universita degli Studi di Roma Tor Vergata,
via della Ricerca Scientifica 1, I-00133 Roma 3
INAF-Osservatorio Astrofisico di Arcetri, Largo
E. Fermi 5, I-50125 Firenze
FRINGES
FRINGES ?
Calib (HR2379_B02, 1.75mas, K2.24)
Target (ZCMa A, K4.2)
21th Jan 2006, UT1-UT2-UT4, AMBER MR _at_ 2.16µm,
50ms integration, seeing0.6proposal
76.C-0817 by Nisini B., Antoniucci S., Li Causi
G. et al.
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
2

• Are there FRINGES on the target?
• Hypotheses
• ...FRINGES ARE NOT FORMED - ...the target has
  INTRINSIC NULL VISIBILITY (i.e. it is very
  large) - ...FIBERS ARE NOT CENTERED on the
  target - ...fringes do not form because of
  MALFUNCTIONS (vibrations, etc.)
• ...FRINGES ARE FORMED - ...but they are UNDER
  THE NOISE because the FLUX IS TOO LOW - ...but
  the EYE CANNOT SEE THEM - ...but they are HIDDEN
  BY ARTIFACTS

? FRINGES ARE THERE, as we can see by means of an
enhanced elaboration (Fourier filtering on
deviations from local median well see better in
the following)
...but, if we do amdlib Extraction
(amdlibComputeSpectralCalibration,
amdlibComputeP2vm, amdlibExtractVis b 1000),we
obtain mean calibrated vis of 1.14 /- 0.11
? meaningless ! 0.81 /- 0.09 0.58 /-
0.09 Why?
FRINGES !
Target (ZCMa A) enhanced
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
3
Let us have a look at the Fourier power spectrum
of the interference channels
spurious peaks
Targetphotometric channel
Target (ZCMa A)average power spectrum
Calibratoraverage power spectrum
Darkaverage power spectrum
There are spurious peaks, which means spurious
fringes in the image frames, even in the Dark and
even in the masked DK channel !
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
4
We can see that the enhancement of one
photometric channel of a dark, where we dont
expect fringes at all, shows lots of fringes,
with the same period and inclination as those in
the Target
Dark P2 channel - original
Dark P2 channel - enhanced
Hence ? the fringes found in the Target are
spurious !!
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
5

• Removing the artifact
• Such spurious fringes are possibly caused by
  electromagnetic interferences on the detector
  itself.
• Perhaps they could be removed by hardware,
  however current and past observations are
  corrupted.
• So we have to remove them by software.
• NOTE the artifact is not removed by the normal
  Dark frame subtraction because the fringes have
  random phases.
• Our approach
• compute averaged power spectrum from all the Dark
  frames
• identify artifact peaks in the frequency plane
  andmake a binary mask for each channel
• Compute power spectrum of each target frame and
  replace realistic power estimation to the masked
  regions
• reconstruct corrected frames by inverse FT
• re-write the AMBER fits files
• redo all the process for any fits of the
  observation,including WAVE_3T, P2VM, Calib,
  Target, Sky and Dark themselves,re-computing the
  mask from their respective Dark frames
• After this pre-processing, launch the amdlib
  standard processing on the corrected files.

Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
6
1. Compute averaged power spectrum from all the
Dark frames
Dark of Calib, Sky and Targetinterf and phot
channels
Dark of Calib, Sky, TargetDK channel
P2VM Darkinterf and photchannels
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
7
2. Make a binary mask for each channel
MASK for Dark, Calib, Sky and Targetinterf and
phot channels
MASK for Dark, Calib, Sky, TargetDK channel
P2VM Darkinterf and photchannels
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
8
3. Replace realistic power estimation to the
masked regions
baseline 1 SIGNAL
Target (ZCMa A)
Peaks are replaced by sigma-clipped average on
each column
Calib Target (ZCMa A)
Dark
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
9
4. Reconstruct corrected frames by inverse FT
Target
? Compare enhancements
spurious fringes
real fringes
Target (ZCMa A) - original
Target (ZCMa A) - corrected
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
10
4. Check the residuals on Dark P2 channel
? Compare enhancements
spurious fringes
no residual spurious fringes
Dark P2 channel - original
Dark P2 channel - corrected
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
11
Now we launch the standard amdlib processing on
the corrected files
amdlib Extraction (amdlibComputeSpectralCalibratio
n, amdlibComputeP2vm, amdlibExtractVis b
1000),we obtain mean calibrated vis of 0.81 /-
0.11 ? now less than unity! 0.87 /-
0.09 0.63 /- 0.09
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
12
We also see that the dispersion of instrumental
visibility among the files is reduced
V2
V2
file
file
Calib - original
Calib - corrected
V2
V2
file
file
Target - original
Target - corrected
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
13
We applied the same pre-processing also to an
observation of FU-ORI (magK5.15, calib HD42807
magK4.85) and we found a similar improvement in
the SNR and visibility dispersion. If we put all
the observations together we see how the artifact
damage increase with decreasing coherent flux,
with different behaviours for the first baseline
and the other two.
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
14
Software availability The software that we have
developed to perform the described
pre-processing, which corrects the detector
fringe artifact, is available at the following
web address http//www.mporzio.astro.it/licausi/
ADC/
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
15

• SOME TECHNICAL NOTES
• Our filtering must be applied to frames already
  corrected for bad-pixels in fact bad pixels
  translate into noticeable power offsets at all
  frequencies.Due to the fact that we replace the
  power but we cannot restore the phase, these
  offsets prevent to remove the artifact fringes
  because the offsets power themselves are assigned
  to the artifact phases.Thus our method only
  works when the median power is near the noise
  level, that is when bad pixels are removed.In
  practive, we perform a sigma-filtering on the
  frames before to apply the correction.Another
  way could be to apply our program on the
  amdlibRawCalibrate output, i.e. the frames
  already corrected by bad pixels, flat, dark and
  sky, then to run amdlibExtractVis on the
  corrected fits using a dummy unity flat with no
  dark and no sky.
• Our analysis of the extracted visibilities, of
  both the original and corrected files, shows that
  the real visibility dispersion across the files
  is underestimated by ExtractVis with default
  parameters, while the e THEORIC parameter
  provides a correct error estimate.We have
  checked this on the dark files, before and after
  correction of detector artifact the default
  extraction provides errors not compatible with
  the expected null visibility, while the e
  THEORIC setting provides a V compatible with zero.

Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
16
Do we get scientific results?

• Our corrected and calibrated visibilities still
  do not seem to be scientifically meaningful (but
  weve not used frame selection yet)
• baseline 1, the shortest, shows a V less than the
  other baselines, which is not easily modellable
• the value of V2 for baseline 1 is not compatible
  with observations of Keck Interferometer,
  published in the literature (Monnier et al. 2005,
  ApJ 624, 832).

Keck
b1
b2
b3
Gianluca Li Causi INAF, Rome Astronomical
Observatory - licausi_at_oa-roma.inaf.it - Grenoble,
30th Nov 2006
17
(No Transcript)