overview on data reduction, calibration and analysis -...
TRANSCRIPT
Overview on data calibration, reduction and analysis
Spectral Lines
Andrea TarchiPhoto by G. Alvito
Outline
Outline• Calibration
Everything else matters To move or not to move? Through the mist Good ol’ Jansky!
Outline• Calibration
Everything else matters To move or not to move? Through the mist Good ol’ Jansky!
• Reduction Breaking the waves Summing up Would you buy this?
Outline• Calibration
Everything else matters To move or not to move? Through the mist Good ol’ Jansky!
• Reduction Breaking the waves Summing up Would you buy this?
• Analysis Bright enough One moment in time Spectral presences 3-D
CalibrationEverything Nothing else matters
CalibrationEverything else matters
We observe a point source with our (radio)telescope
The source has a certain power
We replace the telescope feed with matched load (resistor)
We adjust the load temperature until the power received = source power
The temperature reached is = to the Antenna Temperature
What do we measure?
• ON the source, we have:
Tmeas,ON(α,δ,Az,ZA)= Tsrc(α,δ,Az,ZA) + Telse
• OFF the source, we have:
Tmeas,OFF(α,δ,Az,ZA)= Telse
Where Telse=Tsys=TRX+Tgr(Az,ZA)+Tcel(α,δ,t) +TCMB+Tatm(ZA)
CalibrationEverything else matters
Tsrc + Telse Telse
OFFON
CalibrationEverything else matters
Tsrc + Telse Telse
OFFON
CalibrationEverything else matters
(Tsrc + Telse) - (Telse)
=
ON - OFF
CalibrationEverything else matters
(TON - TOFF)
[(Tsrc + Telse) - (Telse)]/ Telse
=
(ON – OFF)/OFF
CalibrationEverything else matters
This is what is typically provided during a line observation
(TON - TOFF) / TOFF
A spectrum in Tsys units
Conversion to K (or Jy) required
Telse=off = Tsys
Estimate required
A spectrum in Tsys units (counts, so far)
Conversion from counts to K (or Jy)
required
CalibrationEverything else matters
• Theory
• Noise Diodes
• Hot & Cold Loads
• Astronomical Measurements
See also next section
Noise diodes (frequency dependence; precision in diodes meaurements affects calibration accuracy)
Astronomical sources (quite accurate; knowledge in the flux cal. Intensity influences calibration accuracy)
Tsource aka TA of the source in K can be obtained
CalibrationEverything else matters
CalibrationTo move or not to move?
CalibrationTo move …?
Position Switching
ON
CalibrationTo move …?
Position Switching
ON OFF
CalibrationTo move …?
Position Switching
Pros• good baselines(typically)• Iittle a priori infos required
Cons• requires to move the telescope• source cannot be too extended
Calibration… or not to move?
Frequency Switching
ON OFF
ON-OFF
Calibration… or not to move?
Frequency Switching
Calibration… or not to move?
Frequency Switching
Pros• does not require to move the telescope• rapid ON-OFF switching (good efficiency)
Calibration… or not to move?
Frequency Switching
Pros• does not require to move the telescope• rapid ON-OFF switching (good efficiency)
Cons• frequency of the line has to be well known• requires time- and frequency- stable baselines
Calibration…or not to move?
Beam Switching
Pros• does not require to move the telescope
Cons• requires hardware• source cannot be too extended
Calibration…or not to move?
Beam Switching
Pros• does not require to move the telescope
Cons• requires hardware• source cannot be too extended
Calibration… or not to move?
Beam Switching with 2 beams (NODDING)Pros• does not require to move the telescope• ON and OFF at the same time
Cons• requires hardware• source cannot be too extended
CalibrationTo move or not to move?
Further methods
• Baseline fitting (simple but requires good baselines)
• Baseline fitting with average fit (for extended sources)
• Double position switching (to remove residual standing waves due to strong continuum)
CalibrationThrough the mist
CalibrationThrough the mist
When νobs ≥ 15 GHz, it is necessary to correct for atmospheric absorption/attenuation!
Path length through the atmosphere for the radiation from a celestial source
CalibrationThrough the mist
CalibrationThrough the mist
Tsys and AM can be determined
Tatm can be assumed ≈ to the air temperature (on ground level) or calculated through approximated relation
A number of measurements Least-square fit
τ (& T0) can be estimated
CalibrationThrough the mist
Average opacities measured at Effelsberg.
Zenith opacity applied as the default in the GBTIDL procedures for quick-look calibration of GBT data.
Warning: τ depends on the weather conditions; especially at high frequencies!
CalibrationThrough the mist
Sky-dip procedures and Water-Vapour Radiometers installed at the telescope sites and/or on the telescopes allow to derive (quasi) real-time τ estimates.
Left: Opacity at 22 GHz measured with the microwave radiometer (in the circle of the picture above) at the SRT site (Courtesy: F. Nasir, F. Buffa, G. Deiana, and collaborators)
CalibrationGood ol’ Jansky
CalibrationGood ol’ Jansky
CalibrationGood ol’ Jansky
When moving at higher or lower Elv, the dish loses sensitivity due to small-scale deformations, viz.the telescope Gain changes.
CalibrationGood ol’ Jansky
When moving at higher or lower Elv, the dish loses sensitivity due to small-scale deformations, viz.the telescope Gain changes.
CalibrationGood ol’ Jansky
To convert from Antenna Temperature (K) to Flux Density (Jy)
Γ is typically determined though observations of know calibrators (and tabulated)
A correction factor, Cs, has to be applied if the source is not point-like
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
Atm. contributionRequires opt. depthand source Elv
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
Atm. contributionRequires opt. depthand source Elv
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
Atm. contributionRequires opt. depthand source Elv
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
Atm. contributionRequires opt. depthand source Elv
Telescope GainRequires gain curve and source Elv
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
Atm. contributionRequires opt. depthand source Elv
Telescope GainRequires gain curve and source Elv
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
Atm. contributionRequires opt. depthand source Elv
Telescope GainRequires gain curve and source Elv
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
Atm. contributionRequires opt. depthand source Elv
Telescope GainRequires gain curve and source Elv
Flux Gain (Sensitivity)Typically tabulated
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
T A[K] = counts·T cal[K]
T cal depends on freq
Atm. contributionRequires opt. depthand source Elv
Telescope GainRequires gain curve and source Elv
Flux Gain (Sensitivity)Typically tabulated
CalibrationGood ol’ Jansky
In summary, a good calibration procedure could be described by
For tricks and shortcuts…wait the hands-on actvities
T A[K] = counts·T cal[K]
T cal depends on freq
Atm. contributionRequires opt. depthand source Elv
Telescope GainRequires gain curve and source Elv
Flux Gain (Sensitivity)Typically tabulated
ReductionBreaking the waves
ReductionBreaking the waves
ReductionBreaking the waves
Spectrum: Well behavingSystem and source: CollaborativeObserver: Lucky
ReductionBreaking the waves
Spectrum: Well behavingSystem and source: CollaborativeObserver: Lucky
ReductionBreaking the waves
Spectrum: Well behavingSystem and source: CollaborativeObserver: Lucky
Baseline fit (and subtract) by a polynomial of order 1
ReductionBreaking the waves
Spectrum: Well behavingSystem and source: CollaborativeObserver: Lucky
Baseline fit (and subtract) by a polynomial of order 1
ReductionBreaking the waves
Often baselines are ‘corrupted’ by ripples (instrumental and structural origin) and standing waves (mainly originating from the reflective path between the receiver feed and the reflector).
Baseline fit and subtraction require a polynomial of order > 1
…but not higher than 5-7, otherwise line features can be destroyed or fake spectral lines created!
ReductionBreaking the waves
ReductionBreaking the waves
Spectrum: Reasonably well behavingSystem and source: Quite standardObserver: Still Lucky
ReductionBreaking the waves
Spectrum: Reasonably well behavingSystem and source: Quite standardObserver: Still Lucky
ReductionBreaking the waves
ReductionBreaking the waves
Spectrum: Badly behavedSystem and source: OddObserver: Probably Tarchi
ReductionBreaking the waves
Spectrum: Badly behavedSystem and source: OddObserver: Probably Tarchi
Remove scan!
ReductionBreaking the waves
Spectrum: Badly behavedSystem and source: OddObserver: Probably Tarchi
There are a priori ways to reduce the standing waves:• Focus modulation• Double position-switching observing mode
Remove scan!
ReductionBreaking the waves
Spectrum: Badly behavedSystem and source: OddObserver: Probably Tarchi
There are a priori ways to reduce the standing waves:• Focus modulation• Double position-switching observing mode
Remove scan!
• Shaping design (SRT)
ReductionBreaking the waves
Spectrum: Badly behavedSystem and source: OddObserver: Probably Tarchi
There are a priori ways to reduce the standing waves:• Focus modulation• Double position-switching observing mode
Remove scan!
• Shaping design (SRT)
ReductionBreaking the waves
Spectrum: Badly behavedSystem and source: OddObserver: Probably Tarchi
There are a priori ways to reduce the standing waves:• Focus modulation• Double position-switching observing mode
Remove scan!
• Shaping design (SRT)• Off-axis feed arm (GBT)
ReductionBreaking the waves
Spectrum: Badly behavedSystem and source: OddObserver: Probably Tarchi
There are a priori ways to reduce the standing waves:• Focus modulation• Double position-switching observing mode
Remove scan!
• Shaping design (SRT)• Off-axis feed arm (GBT)
ReductionBreaking the waves
Spikes and interferences
Our Interference-busters mobile unit
MobLab: a fully equipped mobile laboratory to seek radio interferences (Courtesy: F. Gaudiomonte, P. Bolli, F. Messina, and collaborators)
ReductionSumming up
ReductionSumming up
…..…..…..
+ many other more scans
Longer integration time gives a lower noise.
Observations are performed by obtaining several scans (on-off)In individual scans, there is no evident emssion…
…should we give up?
Individual scans
ReductionSumming up
…..…..…..
+ many other more scans
…Not yet!
Let’s try averaging all the scans
The signal to noise ratio (SNR) would increase by √N (N=number of scans)
Individual scans
ReductionSumming up
…..…..…..
+ many other more scans
…Not yet!
Let’s try averaging all the scans
The signal to noise ratio (SNR) would increase by √N (N=number of scans)
Individual scans
Sum of the scans
ReductionSumming up
…..…..…..
+ many other more scans
Individual scans
ReductionSumming up
…..…..…..
+ many other more scans
Individual scans
Smoothed (x4) sum of the scans
ReductionWould you buy this?
ReductionWould you buy this?
Gaussian or ‘normal’ distribution
ReductionWould you buy this?
Gaussian or ‘normal’ distribution
68% values drawn from a normal distribution are within one standard deviation (1 sigma) away from the mean.
ReductionWould you buy this?
Gaussian or ‘normal’ distribution
99.7% values drawn from a normal distribution are within three standard deviation (3 sigma) away from the mean.
ReductionWould you buy this?
3σ
ReductionWould you buy this?
3σ
5σ
ReductionWould you buy this?
ReductionWould you buy this?
Chances to publish your detection:
ReductionWould you buy this?
Chances to publish your detection:
• SNR > 5 σ (very likely)
ReductionWould you buy this?
Chances to publish your detection:
• SNR > 5 σ (very likely)
• SNR between 3 and 5 σ (likely)
ReductionWould you buy this?
Chances to publish your detection:
• SNR > 5 σ (very likely)
• SNR between 3 and 5 σ (likely)
• SNR ~ 3 σ (more difficult; the author should be very convincing and/or having an independent measurement for confirmation)
ReductionWould you buy this?
Chances to publish your detection:
• SNR > 5 σ (very likely)
• SNR between 3 and 5 σ (likely)
• SNR ~ 3 σ (more difficult; the author should be very convincing and/or having an independent measurement for confirmation)
• SNR < 3 σ (almost impossible)
ReductionWould you buy this?
Chances to publish your detection:
• SNR > 5 σ (very likely)
• SNR between 3 and 5 σ (likely)
• SNR ~ 3 σ (more difficult; the author should be very convincing and/or having an independent measurement for confirmation)
• SNR < 3 σ (almost impossible)
AnalysisOne moment in time
AnalysisOne moment in time
Moments are equally defined for T (in K)
AnalysisOne moment in time
Total Intensity
AnalysisOne moment in time
Mean Velocity
AnalysisOne moment in time
Dispersion
AnalysisBright enough
AnalysisBright enough
Speak
Δν
vpeak
Gaussian FitLine peak fluxLine FWHM (Δν= √8ln2·σ)Line peak velocity
AnalysisBright enough
If the distance of the emitting source is known the isotropic luminosity of the water maser can be derived:
• kilomaser or megamaser?
• star formation or AGN nature?
In two radio galaxies, water megamasers with Liso ~ 10000 Lsun has been found!!!
AnalysisBright enough
Column density can be derived from the line total intensity
If the dimension of the source (or its distance from us) is known, a gas mass estimate can be obtained
AnalysisBright enough
The velocity of the line provides information on the gas dynamics:
Frequency shift according to the non-relativistic Doppler equation in the optical and radio convention
νobs > νrest: the emitting gas is approching us toward the l.o.s.νobs < νrest: the emitting gas is receding from us toward the l.o.s.
For extragalactic objects, recessional velocity of the galaxy does matter
Recessional velocity of the galaxy
Red-shiftedBlue-shifted NGC
4258
AnalysisBright enough
AnalysisSpectral presences 3D
AnalysisSpectral presences 3-D
Contour map made bya regular grid of individual pointings
AnalysisSpectral presences 3-D
AnalysisSpectral presences 3-D
3-D structure
…but…
Not truly source 3-D
Data cube
AnalysisSpectral presences 3-D
Leiden-Argentine-Bonn (LAB) HI survey. Sweeping through the data cube;Galactic center projection. Kalberla et al.; Hartmann & Burton; Arnal et al.
AnalysisSpectral presences 3-D
Leiden-Argentine-Bonn (LAB) HI survey. Sweeping through the data cube;Galactic center projection. Kalberla et al.; Hartmann & Burton; Arnal et al.
Useful textbooks:• J.D. Kraus”Radio Astronomy”, 1986, Cygnus-Quasar Books • K. Rohlfs & T.L. Wilson “Tools of Radio Astronomy”, 1996, Springer-Verlag
This lecture has also taken inspiration from past lectures and talks by U.Back, D. Muders, K. O’Neil, and B. Winkel.
Useful textbooks:• J.D. Kraus”Radio Astronomy”, 1986, Cygnus-Quasar Books • K. Rohlfs & T.L. Wilson “Tools of Radio Astronomy”, 1996, Springer-Verlag
This lecture has also taken inspiration from past lectures and talks by U.Back, D. Muders, K. O’Neil, and B. Winkel.
Thank you…and…Happy