Data reduction for Optical Imaging and Spectroscopy

1
Data reduction for Optical Imaging and
Spectroscopy
Stefano Benetti
2
Focal reducer instruments

• Afosc (D, Al, B etc..)
• EFOSC2
• EMMI
• Dolores
• FORS1-2
• SOFI
• NICS
• ISAAC
• etc..

Simple but powerful imaging spec. almost
simultaneously
3
Optical mechanical scheme
4
stable instruments! (if the site is good)
5
and spectroscopy
Averages of 12 curves each scatter around zero
point 10 once normalised 3!
6
Calibrations are also stable If instrument
and/or single elements (slits/filters/grisms) have
NOT been removed/re-mounted from the
telescope/ Instrument ? you may use calibrations
(BIAS, FF, arcs) which are weeks-to-months old!
REMEMBER take out always the overscan the zero
point of the BIAS level may change quit a lot
also during a single night!
7
In summary
versatiles
powerfulspectroscopy of faint object
stable
Main problem

• Central light condensation (light reflected back
  from the detector
• 
  to the optics)
• prevent accurate photometry
• The effect can be as high as 5 in EFOSC-DFOSC
• 
  FORS1-2 as low as 0.7!

8
Data reduction the SN case
Asking an SN observation you get the unknown
(ToO)!!!
You have to deal with lots of configurations
ToO observations you reduce what you get
You need color equations and response curve
database
You get observations with no parall. angle used
what to do
Normally you dont observe an objet with ZD70o,
but for observing SN you must sometime do!
You observe trough clouds (rain?!?) and you must
be able to deal with it
and surely much more!
9
Data reduction general overscan, BIAS and FF
correction FIRST inspect visually all the files
to look for problems, defects, etc I suggest
the use of standard IRAF procedure
noao.imred.ccdred.ccdproc where for overscan I
normally fit a legendre with order 1 (a
constant!) to all images
10
Combine the biases with the task
noao.imred.ccdred.zerocombine
11
BE AWARE CCD windowing might produce slopes in
BIASes!
? If a box is used, ask BIAS with the same CCD
window!
12
sky FFs are better!
13
Imaging Flatfielding
Combine the FFs with the task
noao.imred.ccdred.flatcombine
14
with thin CCDs you get fringing in I
and you do not correct it with dome/sky
FF!!! because.
15
then how to deal with fringing?

• We need a master (super) flat! How to make it
• take I-band images of different fields (spread
  over one
• night) and corrected for bias and flatfield.
• flatcombined the images.
• compute the average value and subtracted it from
  the image.

How to correct?
Fringing is additive then it has to be
subtracted from the images after the pattern has
been scaled to the actual value of the fringing
in the image!
not known how variable is the pattern the
intensity might varies a lot, but the shape could
be quite stable, e.g. instrument has little
flexures (EFOSC-FORS-EMMI) problems with
CAFOS, DFOSC-like solve only FF with telescope
on target position! ? impossible for imaging!!!
16
Measuring SN magnitudes
Aperture photometry to be used as quick look
only PSF technique reliable up to mag16-17
NOT need to wait the
fading of the target Template subtraction gives
the better result (you
subtract the REAL background) but
you DO need to WAIT the fading (few
years)
17
Differences between methods
(Pastorello et al. 2005)
18
PSF method
We use snoopy (Cappellaro Patat-Ledan
software) based on daophot (noao.digiphot.daophot)
19
b) creation and check
20
c) Background definition
21
(No Transcript)
22
Template subtraction
The steps to follow are
23
Steps to follow

• Align the two images x,y translation and
  rotation
• Flux normalisation of template to science
  
• Degrade the image with better seeing!
  
• Subtraction
• If object is faint (peak signal3-5 bkg noise)
  
• better do the measure with PSF (not aperture)
• 
  

24
Photometric calibration
a) Photometric night
You have the color equations system U-u
KUSUx(U-B)AU z B-b KBSBx(B-V)AB z V-v
KVSVx(B-V)AV z R-r KRSRx(V-R)AR z I-i
KISIx(R-I)AI z where Ai atm. absorption
(table or fitting), z airmass
solve in Ki and Si through l.s. fitting of std
stars measurements having a big spread in
colors!!!
25
why large colours
26
Having Ki and Si ? solve previous system of
equations for SN and local sequence ?
UBVRI
b) Not photometric night
First you must have the local sequence calibrated

get color equations from archive ?solve system
for local sequence ? UBVRI

• You compare ? band and star m with true mag ?
  ?mi
• ? band take the average of ?mi
• You add ?mi to Ki of c.eq. ? solve the system
  again
• and get UBVRI of SN!!!

27
After that you have to homogenize all
measurements taken with different instruments
28
FFs in spectroscopy
watch the blue!
29
Steep gradient
lots of FFs to improve S/N!!!
30
How to get rid of FF continuum?
with task noao.imred.ctioslit.response
31
Be aware to properly fit the continuum ?
interpolate only higher frequencies!
TNGDolores LR-R
32
Result of division
33
zone with reflections should be avoided! To be
specified in ToO observations!
34
The importance of the background subtraction also
in spectroscopy
or How to write two papers onan inexistent Ha
35
SN 1990M
Polcaro Viotti 1991 Hydrogen in a Ia!
Della Valle, Benetti et al 1996 No Hydrogen!!!
If background removed!
36
Importance of background
bkg well removed
bkg
Ha
bkg not well removed
Ha
Fitted background Optimised extraction
37
Extraction of the spectrum noao.imred.ctioslit.
apall
A recent example the classification of SN
2005H (mag15.9 offsets 0.3E 1.8N)
how difficult is to find the object and define
bkg even with a relatively bright SN.
be aware always of what is doing trace!!!
38
In fact, here is a nice spectrum.
btw, they didnt align the grism very carefully
we start apall
39
and background
40
checked the trace!
41
Result
not a SN!?!?!?!
42
we went back to have a closer look to the place
where the SN should be
43
Back to the spectrum
44
Apall again, but mashed rows were the contrast
stellar tracing to galaxy bkg is higher
Here is where to define the SN position
45
And the bkg
46
Finally we got the spectrum of a nice SNII!!!!
47
If target is faint and grism well align better
NOT trace spect.!
48
result (avg of 2)!!!
49
Spectrum in wavelength noao.imred.ctioslit.id
entify noao.imred.ctioslit.refspectra
noao.imred.ctioslit.dispcor
goal correlate pixels with wavelength with an
arc spectrum
Godness expected with a 300gr/mm (2-3 Å/px)
grism range4500Å 20-30 used line ? rmslt1Å
if 4-5 degree polynomial
50
Suggestions
Each instrument/grism needs a table (wavelength
of lines may change with resolution) ? keep
vast archive of arc tables!
Often with ToO you dont get what you ask ? be
prepared to calibrate spectra with skylines
(see above, especially in the red!)
Alternatively use previous solutions ? anchor
zero point with skylines
Use bright skylines (e.g. OI 5577, 6300, etc)
to get spectral resolution
51
Spectrum in flux noao.imred.ctioslit.standard n
oao.imred.ctioslit.sensfunc noao.imred.ctioslit.ca
librate
(see previous slides for godness (10 - lt3) and
stability (yes) of response cur.)
I prefer to use standards from Hamy (1992, 1994)
or Oke 1990 useful link http//www.eso.org/observ
ing/standards/spectra/
Check the header parameters! (exptime and airmass)
52
Check the spectrophotometry
Convolving passbands to a spectrum a nice IRAF
task is stsdas.hst_calib.synphot. calcphot
53
spectrophotometry vs. photometry
If everything went well ? ?U?B ?V?R?I0
If night not photometric ? ?U?B ?V?R?I?0
If you get ?U??B??V??R??I you are in trouble!!!!
Did you use parallactic both for target and
standard? probably not!
54
What is parallactic angle?
To apply if dzgt30o
See Filippenko 1982, PASP 94, 715
55
its amplitude
At z1.1 ? point source at 6500Å is displaced by
0.75 at 3500Å At z1.3 ? point source at 6500Å
is displaced by 1.40 at 3500Å
as function of H.A. and d
56
How losses may change derived parameters
57
spectrophotometry vs. photometry
If everything went well ? ?U?B ?V?R?I0
If night not photometric ? ?U?B ?V?R?I?0
If you get ?U??B??V??R??I you are in trouble!!!!
Did you use parallactic both for target and
standard? probably not! probably because it was
a ToO
So, please, use (ask) always p.a. or few minutes
trough a wide (5-10) slit!
My suggestion is NOT to use this spectrum for
measuring line fluxes, only positions
(wavelengths)!
If you desperately need it, I may teach you
(privately) a way to massage it!!!!
58
Modelling atmospheric absorptions
To be done AFTER normalization to photometry!
You need a flux calibrated spectrum of a hot
star obtained at a similar airmass!
You divide the zapped spectrum to the unzapped one
59
You get
60
?
Then you apply to SN spectrum before (SNIa
z0.30)
?
?
?
61
After correction
?
?
?
?
you may adjust the abs. intensity with
(atm.fitsk)/(1k) and kgt0lt0
62
Conclusions
What I would like you keep from this long lesson
is that when starting a data reduction of a SN
you must be prepared to face always new unknowns
and challenges.
Moreover I would suggest NOT to stop to the first
result you get, but keep trying to see if you get
a better result (S/N), this is True also for an
expert reducer. Curiosity is our main drive in
science, lets be curious also when reducing!!!