ORBS and ORCS �omas Martin (Université Laval) Laurent Drissen (Université Laval) Laurie Rousseau-Nepton (Université Laval), Alexandre Alarie (Université Laval), Simon Prunet (CFHT)
ORBS and ORCS
�omas Martin (Université Laval)Laurent Drissen (Université Laval)
Laurie Rousseau-Nepton (Université Laval), Alexandre Alarie (Université Laval), Simon Prunet (CFHT)
4 mai 2016
Reduction and analysis of SITELLE's data
ORBS
ORBS
ORBS (outils de réduction binoculaire pour SpIOMM/SITELLE) is
a data reduction software for SITELLE.
2
ORBS
Data release 1
So what is ORBS ?
Reduction process
Combination
Phase correction
Flux calibration
Spectral calibration
Astrometry
Performances
The first data release !
• DR1 version released March 30, 2016 (1 month ago)• Since then all cubes from March mission have been fully
reduced and calibrated.
• Pure ’push-button’ software: no error on any of thereduction.
• All SV and March cubes have been checked and releasedto the PI’s (Wang, Spekkens, Martin, Bresolin, Lavoie,
Robert, Alarie, Simard, Joncas, Shara, Herczeg).
4
SITELLE’s Wiki
132.203.11.199/wiki-sitelle
5
132.203.11.199/wiki-sitelle
March 2016 reduction logs
132.203.11.199/wiki-sitelle
6
132.203.11.199/wiki-sitelle
ORBS
Data release 1
So what is ORBS ?
Reduction process
Combination
Phase correction
Flux calibration
Spectral calibration
Astrometry
Performances
What is ORBS ?
ORBCore module (23 000 lines)
ORCSData analysis tools(4 000 lines)
ORBS(13 000 lines)Data reduction module
• Reduction software forSITELLE and SpIOMM
• free-software (GNULicence)
• Written in Python• 40 000 lines of code• 5 years of development• Object-Oriented layered
architecture
• The one and only software capable of reducing SITELLE’sdata.
8
How can I know more ?
sourceforge.net/u/thomasorbs All the actively maintained
softwares are here !
9
sourceforge.net/u/thomasorbs
How can I know more ?
Browse the commits
10
How can I know more ?
Get informations on each release
11
How can I know even more ?
132.203.11.199/orbs-doc
12
132.203.11.199/orbs-doc
How can I know even more ?
Read the documentation: hundreds of pages of pure pleasure ;)
13
ORBS
Data release 1
So what is ORBS ?
Reduction process
Combination
Phase correction
Flux calibration
Spectral calibration
Astrometry
Performances
Why a reduction software ?
1 2
α
δ
λ
Calibrated spectral cubeWavelength, �ux, astrometry
Raw interferometric cubes2 sets of 400 - 1000 raw interferometric frames
~ 32 - 64 Go of raw data
FFT
In
Out
Because we need much more than a simple FFT to transform 2
interferometric cubes into 1 calibrated spectral cube.15
Why a reduction software ?
5 reduction steps:
1. Correct CCD images: bias, flat,
cosmic rays (more than 20 000 by
cubes, 5/s)
2. Align frames (guiding errors)
3. Merge (i.e. align) cubes (2
cameras = 2 data cubes)
4. Correct for phase
5. FFT
6. Calibrate data
A very sensitive guy
16
Why a reduction software ?
A not too bad raw interferogram
Reduced interferogram
Baseline distorsions kill Fourier !
Fourier transform is very sensitive to the quality of the input
signal. Any artifact on one sample creates a distorsion of the
whole spectrum.
17
Why a reduction software ?
Hα
[NII
]658
4
[NII
]656
3
Sky
Sky
Sky
Sky
Sky
Raw spectrum
Reduced spectrum
[SII
]671
7[S
II]6
731
• over√
(2) × more SNR• over 2 × more resolution• No deformation of the continuum (low frequency noise) 18
The pipeline in short
Combination
Transformation
Calibration
1 2
x
y
λ
α
δ
λ
C
Raw interferometric cubes
Calibrated spectral cubeWavelength, �ux, astrometry
Spectral cube
Combined interferometric cube
19
in less short. . .
Combination
Transformation
Calibration
1 2
C
Raw interferometric cubes
x
y
λ
α
δ
λ
Calibrated spectral cubeWavelength, �ux, astrometry
Spectral cube
Combined interferometric cube
P
1 2
α
δ
λS
1 2
α
δ
λL
Standard cube+ standard images Laser cube
Phase cube
• Preprocessed calibration data is needed:• standard star• laser source
• A phase cube must also be computed
20
ORBS
Data release 1
So what is ORBS ?
Reduction process
Combination
Phase correction
Flux calibration
Spectral calibration
Astrometry
Performances
2 output ports: to get all the incoming light
Beam splitter+ Compensator plate
Fixed Mirror
Moving mirror
Detector
Source Problem50% of the light goes
back to the source
Classical interferometer(1 output port)
SITELLE's interferometer(2 output ports)
Solution
Detector 1
Detector 2
Source
Moving mirror
Beam splitter+ Compensator plate
Fixed Mirror
• 1 output port = half the light !
22
2 output ports: to correct for transmission variations
Step index
Tran
smiss
ion
21:05 23:25 21:45 23:00Day 1 Day 2
Transmission uncertainty < 5e-4 (0.05%)
• Sky transmission variations modulates the interferograms• Sum of light on both ports = relative transmission.
23
2 output ports: to remove non-modulated scattered light
Combined interferogram
Background rising (up to 15 times its normal value)
Normal background intensity
First night Second night
Camera 1 interferogram
Before
A�er Background peak has disappeared !
• Non-modulated scattered light must be removed !• Combination automatically removes it.
24
ORBS
Data release 1
So what is ORBS ?
Reduction process
Combination
Phase correction
Flux calibration
Spectral calibration
Astrometry
Performances
What is the phase ?
• The output of a Fourier transform is a complex vector• A complex number can be represented by amplitude/phase
instead of real/imaginary
Amplitud
e
Phase
Re
Im
26
Effects of phase correction on the spectrum
Without phase correction real spectrum and imaginary spectrum
are mixed
• No phase correction =⇒ spectrum =√
Re2 + Im2
• With phase correction: spectrum = Re
Wavelength (nm)
Power spectrumReal partImaginary part
27
Effects of phase correction on the spectrum
Wavelength (nm)
+σ-σ 0
Phase corrected spectrum (sinc)
Gaussian noise distribution
+σ-σ 0
Power spectrum (sinc2)
Squared Gaussian noise distribution
Using power spectrum is bad
• Noise is√
2 higher
• Noise is folded (N 2)• Resolution is degraded
28
ORBS
Data release 1
So what is ORBS ?
Reduction process
Combination
Phase correction
Flux calibration
Spectral calibration
Astrometry
Performances
Absolute flux error
[NII] λ6584Hα -6.88 (±2.83) %
-10.87 (±3.02) %Measured �uxStandard �ux
Flux
in e
rg/c
m2 /
s
Wavelength in nm
Hα
HeI
λ66
78
[SII
] �67
17,3
1
[NII
] λ65
84
[NII
] λ65
63
FitData
Integrated spectrum
0.17
0.32
0.47
0.61
0.76
0.91
[NII]λ6584/Hα ratioBONUS !
Comparison with M1-71 spectrophotometry (Wright et al. 2005)
(the 3% error comes from Wright’s data)30
Absolute flux error
SN1 SN2 SN3
4000 4500 5000 5500 6000 6500 7000
20
40
60
80
100In
tegr
ated
�ux
[10-
14 e
rg/s
/cm
2/A
]
Wavelength [A]
CALIFA's integrated spectrumCALIFA's imagery calibrationSITELLE's integrated spectrum
Note: SITELLE's spectra have been convoluted to respect PPAK's resolutionNote: A correction factor of 0.65 has been applied to consider PPAK's �lling factor
Comparison with NGC628 integrated spectrum (CALIFA survey,
Sanchez et al. 2010) 31
Absolute flux error
Object
NGC3344 Hα vs. SpIOMM -4% ±2%(Rousseau-Nepton et al.) Hα + [NII]λ6584 vs. SpIOMM -4% ±3%M1-71 Hα vs Wright (2005) -7% ±3%
[NII]λ6584 vs Wright (2005) -11% ±3%NGC628 SN1 vs. CALIFA -6% ±6%
SN2 vs. CALIFA -7% ±6%SN3 vs. CALIFA -9% ±6%
HETDEX field Lyα flux of ∼ 20 high redshift -5%±7%(Drissen et al.) galaxies
• General bias of -5% (bad estimation of the modulationefficiency).
• Once corrected an uncertainty of ± 5% is a conservativeestimation.
32
Relative flux error
• Phase correction flux error is � 1%• Relative flux error dominated by flat field error ∼ 1% on
80% of the field
33
ORBS
Data release 1
So what is ORBS ?
Reduction process
Combination
Phase correction
Flux calibration
Spectral calibration
Astrometry
Performances
Good news
With a Fourier transformed spectrum
• Spectral calibration depends only on the zero point.• Relative wavelength error of the lines in a single spectrum
is impossible.
Zero point is the only uncertaintyWavelength (nm)
35
Absolute velocity error
• Absolute wavelength calibration is based on a HeNe [email protected] nm
• But Nominal wavelength is not well known + drift with time• Measured absolute error is around 70 km/s since the
beginning of the year ( =⇒ HeNe ' 543.62 nm).• Can be corrected via sky lines with a precision < 0.5 km/s
Wavelength (nm)
36
Relative velocity error
Relative error is the pixel-to-pixel variation of the zero point
IC348 (Herczeg et al.)UGC7899 (Spekkens et al.) PG1216+069 (Wang et al.)
Sky lines velocity difference (in km/s)
km/s
• < 5 km/s of relative velocity error on 90% of the FOV• In most of the cubes a sky lines map can be created with
ORCS to reduce the relative error to < 1 km/s
37
ORBS
Data release 1
So what is ORBS ?
Reduction process
Combination
Phase correction
Flux calibration
Spectral calibration
Astrometry
Performances
WCS
WCS comparison with USNO-B1 catalog
• WCS error < 1.5” (5 pixels) in 90% of the field
39
ORBS
Data release 1
So what is ORBS ?
Reduction process
Combination
Phase correction
Flux calibration
Spectral calibration
Astrometry
Performances
ORBS Performances
Fully Automated, robust and easy to use
• All cubes are reduced without any human intervention• Launched via a very simple command line
Fast
• Parallelized at 90%• Process a 1000 steps cube in less than 7 hours• Most cubes are processed in less than 3 hours (
ORBS Performances
Scalable
• Process 68 Go of raw data with less than 23 Go of RAM(max performances)
• Needed RAM can go under 8 Go with reduced performances• Can be scaled to be used on any regular computer
Free, well-coded and maintainable
• Respect free software guidelines• Written in Python• Object-Oriented• Fully commented and documented
42
ORCS + Viewers
ORCS + Viewers
ORCS
Viewers
ORCS
ORCS (outils de réduction de cubes spectraux) is a fit engine for
fast analysis of spectral cubes.
44
What is ORCS ?
ORCS: a fit engine
• A wide variety of models:• emission/absorption spectrum• continuum• filter (to fit lines on the border of the filter)• possibility of adding model grids (e.g. stellar population)
• Multiple constraints: co-varying velocity / FWHM /amplitude =⇒ enhance the precision of the fit by using allthe information.
45
Typical output
• Each fitted parameter can be mapped.
Velocity�t error
FWHM�t error
Amplitude�t error
Amplitude FWHM Velocity
1 emission line = 6 maps
46
Hα emission line from ORCS on IC443 (Alarie et al.)
6:18:00.0 55.0 50.0 45.0 40.0 35.0 30.0 25.0 17:20.0 15.0
54:0
0.0
53:0
0.0
52:0
0.0
51:0
0.0
22:5
0:00
.049
:00.
048
:00.
047
:00.
046
:00.
045
:00.
044
:00.
0
Right ascensionD
eclin
atio
n24
40
55
71
86
101
117
132
148
163
179
6:18:00.0 55.0 50.0 45.0 40.0 35.0 30.0 25.0 17:20.0 15.0
54:0
0.0
53:0
0.0
52:0
0.0
51:0
0.0
22:5
0:00
.049
:00.
048
:00.
047
:00.
046
:00.
045
:00.
044
:00.
0
Right ascension
Dec
linat
ion
1.45e-17
5.35e-17
9.26e-17
1.32e-16
1.71e-16
2.10e-16
2.49e-16
2.88e-16
3.28e-16
3.67e-16
4.06e-16
Velocity (in km/s)Flux (in erg/cm2/s/A)
47
Other features
• sky lines fit• automatic correction of heliocentric velocity• sky removal
48
Sky lines + HII region lines fit
Flux
(in
ergs
/cm
2 /s)
[NII
] λ65
84
Hα
[NII
] λ65
63
[SII
] λ67
31[S
II] λ
6717
DataFitResidual
Wavelength (in nm)
Absolute velocity calibration with a precision of 500 m/s !49
Sky lines + 2 HII regions lines
2 HII components at : 49.35 [±1.3] km/s and 3.33 [±3.33] km/s
Wavelength (in nm) Wavelength (in nm)
Flux
(in
ergs
/cm
2 /s)
[NII
] λ65
84
Hα
[NII
] λ65
63
[SII
] λ67
31
[SII
] λ67
17
A bubble in IC348 ?
50
ORCS + Viewers
ORCS
Viewers
2D viewer
Visualize your data cube Extract a spectrum and �t it !
Fit resultsFit parameters
52
3D viewer
videos/3dviewer.avi
53
videos/3dviewer.avi
Conclusion
To conclude
• First data release• All data has been fully reduced with ORBS without any
human intervention.
• ORBS has proved to be functional, fast, precise and robust.• We have discovered small but obvious calibration biases
(easy to correct)• Next data release in a few months:
• flux error < 2-3%• absolute+relative wavelength error < 1 km/s
• ORCS is a precision tool especially developped forSITELLE’s data
• ORCS and viewers now under strong development butready to be used by the community.
• Releasing tools and working with data will be the best way toimprove them. 54
ORBSData release 1So what is ORBS ?Reduction processCombinationPhase correctionFlux calibrationSpectral calibrationAstrometryPerformances
ORCS + ViewersORCSViewers
Conclusion