DIA - Astro-WISE

Difference Image Analysis at
OAC
Groningen, 1st Dec 2004
AW-OAC team
• The DIA is a software written by P.R.Wozniak based on
the Alard & Lupton optimal PSF matching algorithm.
• DIA was originally created to search variable objects on
the OGLE-II bugle microlensing data
• The original version is not well documented (parameters
optimization is not easy) and it is optimized for OGLE data
• We present a modified version that includes:
–
–
–
–
New tool to prepare the images using the astrometry
New functionalities in the core (masks, external domains)
Python interface to send processes on a BEOWULF cluster
New tool to study the candidates (ascii catalogues, light curves,
frequency analysis, phase diagrams, stamps)
Groningen, 1st Dec 2004
AW-OAC team
Overview: Original DIA
Cross_regrid*
Registration and correction of input images
mstack
Creation of the reference image with best seeing
frames
getpsf
Global PSF on REF (used in getvar and phot steps)
Aga **
Creates difference images
Getvar **
Finds variable candidates
Phot **
Light Curve with PSF and aperture photometry on
the difference images
* Not used in the new version
** New functionalities added
Groningen, 1st Dec 2004
AW-OAC team
AGA step: Image Subtraction
• Once the reference frame ref.fits is created it is convoluted with a
kernel (spatially variable in general) in order to match as close as
possible each image. This convoluted image is then subtracted to the
current frame, producing a serie of subtracted images.
The Algorithm
• A list of objects is found on the REF (domains)
• A list of objects is found on the single image (domains)
• It matchs the domains and calculates the kernels
(3 Gaussians of costant widths multipied by polynomials)
• The best solution is taken to produce the difference image
Groningen, 1st Dec 2004
AW-OAC team
Getvar step: variable objects detection
Variable objects are detected using some preliminary variability measurements
based on the entire serie of difference images for a given field.
Final measurements are made only for these candidates.
The program starts by rejecting some fraction of the frames with the worst
seeing (in our case 10%)
The pixel is declared as variable if one of these two conditions are met:
1. There are at least 3 consecutive points departing at least 3σ from the base
line in the same direction (up or down), or
2. There are at least 10 points in total departing at least 4σ from the base line
in the same direction, not necessarily consecutive.
NB: The extension to other types of variables is straightforward!
Groningen, 1st Dec 2004
AW-OAC team
Phot Step: Photometry
For each variable object the program performs both profile and aperture
photometry on difference images keeping the centroid fixed
The format of the catalog is :
1. Flux – profile photometry
a psf
2. Flux error – profile photometry
3. Flux – aperture photometry
4. Flux error – aperture photometry
 ( f P / )

 (P /  )
2
i
i
 psf 
aap 
ri  rap
f
i
i
 ap 
2
i
2
i
1
i (Pi 2 /  i2 )

i
5. Background
i i
i
2
i
Where:
• Pi is PSF for pixel i
• fi flux for pixel i on
differece images
• fi,0 flux for pixel on
original images
• G is Gain and
 i2 
f i ,0
G
6. FWHW of the PSF profile
Groningen, 1st Dec 2004
AW-OAC team
New DIA Version: step 1 wcs2pix
• A C program that reads
astrometric informations from the
headers (CRVAL,CRPIX,CD) and
through the WCS library finds the
common part of the images.
• The WCS library is also used to
extract the object coordinates in
alfa e delta from the output files.
Groningen, 1st Dec 2004
AW-OAC team
New DIA Version: new functionalities
Added in Aga:
• Mask for each input image
• The kernel can be calculated using an external
list of objects
Added in Getvar:
• Ascii catalogs of variable objects
• Psf image is created and saved (QC check)
• VAR and ABS images are saved (see ISIS)
Added in Phot:
• Ascii light curves with phases
Groningen, 1st Dec 2004
AW-OAC team
The New DIA Version: how it works on the BEOWULF
Software
wcs2pix
prepare
Hardware
MASTER
storage
Network SWITCH
mstack
BeoRunner
getpsf
aga
Node 1
Node 2
getvar
phot
Node n
lc
Groningen, 1st Dec 2004
AW-OAC team
Step 1: cut all images using the astrometric solution
wcs2pix
prepare
MASTER
storage
Network SWITCH
mstack
BeoRunner
getpsf
aga
Node 1
Node 2
getvar
phot
Node n
lc
Groningen, 1st Dec 2004
AW-OAC team
Step 2: split the images in sub frames and copy them on nodes
wcs2pix
prepare
MASTER
storage
Network SWITCH
mstack
BeoRunner
getpsf
aga
Node 1
Node 2
getvar
phot
Node n
lc
Groningen, 1st Dec 2004
AW-OAC team
Step 3: mstack the subframes indipendently on the nodes
wcs2pix
prepare
MASTER
storage
Network SWITCH
mstack
BeoRunner
getpsf
aga
Node 1
Node 2
getvar
phot
Node n
lc
Groningen, 1st Dec 2004
AW-OAC team
Step 4: getpsf
wcs2pix
prepare
MASTER
storage
Network SWITCH
mstack
BeoRunner
getpsf
aga
Node 1
Node 2
getvar
phot
Node n
lc
Groningen, 1st Dec 2004
AW-OAC team
Step 5: aga
wcs2pix
prepare
MASTER
storage
Network SWITCH
mstack
BeoRunner
getpsf
aga
Node 1
Node 2
getvar
phot
Node n
lc
Groningen, 1st Dec 2004
AW-OAC team
Step 6: getvar
wcs2pix
prepare
MASTER
storage
Network SWITCH
mstack
BeoRunner
getpsf
aga
Node 1
Node 2
getvar
phot
Node n
lc
Groningen, 1st Dec 2004
AW-OAC team
Step 7: phot
wcs2pix
prepare
MASTER
storage
Network SWITCH
mstack
BeoRunner
getpsf
aga
Node 1
Node 2
getvar
phot
Node n
lc
Groningen, 1st Dec 2004
AW-OAC team
Step 8: lc
wcs2pix
prepare
MASTER
storage
Network SWITCH
mstack
BeoRunner
getpsf
aga
Node 1
Node 2
getvar
phot
Node n
lc
Groningen, 1st Dec 2004
AW-OAC team
LC Step: Analysis of the results
For each variable candidate the software produces a light curve in a file called
LC_NAMEFIELD_Xpixel_Ypix_alfa_delta-subframe.data
An automatic Fourier transform is done and an ascii file is created with frequency, power
and s/n ratio (power/standard deviation). This file is called as the LC file plus
Max=…_fre=…_sn=… in order to have the main informations directly in the filename.
The frequency of the max peak is used to add a phase column.
Groningen, 1st Dec 2004
AW-OAC team
The format of the LIGHT CURVE
1. Flux – profile photometry
2. Flux error – profile
photometry
3. Flux – apeture photometry
4. Flux error – aperture
photometry
a psf
 ( f P / )

 (P /  )
i
2
6. FWHM of the PSF Profile
i
i
 psf 
aap 
ri  rap
f
i
i
 ap 
2
i
2
i
1
i (Pi 2 /  i2 )

i
5. Background
i i
2
i
Where:
• Pi is PSF for pixel i
• fi flux for pixel i on
differece images
• fi,0 flux for pixel on
original images
• G is Gain and
 i2 
f i ,0
G
7. MJD-OBS
8. PHASE
Groningen, 1st Dec 2004
AW-OAC team
•
•
•
TESTs
69 Images VLT-FORS – GC 2kx2k B Band
46 Images WFI – Carina 8kx9k B Band
15 Images WFI – OACDF 7kx7k V Band
Beo0/1 – 2 Master + 8 nodes
69
2kx2k
wcs2pix
Beo2 – 1 Master + 16 nodes
15
7kx7k
69
2kx2k
1m15s
wcs2pix
15
7kx7k
46
8kx9k
1m
14m
prepare
2m
8m
prepare
1m
3m
21m
mstack
10s
37s
mstack
3s
10s
1m10s
getpsf
7s
12s
getpsf
2s
5s
10s
aga
4m20s
5m23s
aga
3m
3m50s
19m31s
Getvar*
1m11s
1m59
Getvar*
10s
14s
1m37s
Phot*
59s
1m32
Phot *
3s
10s
1m7s
Lc*
2m
1m
Lc*
8s
12s
1m
TOTAL
6m47s
19m58
TOTAL
4m26s
8m42s
59m35s
Groningen, 1st Dec 2004
* it depends from the threshold in Getvar
AW-OAC team
VLT-FORS Images
Groningen, 1st Dec 2004
AW-OAC team
VLT-FORS Image
Groningen, 1st Dec 2004
VLT-FORS difference Image
AW-OAC team
Light curves: VLT-FORS
Groningen, 1st Dec 2004
AW-OAC team
Light curves: VLT-FORS
Groningen, 1st Dec 2004
AW-OAC team
Light curves: WFI-Carina
Light curves: WFI-Carina
Groningen, 1st Dec 2004
AW-OAC team
Light curves: WFI-Carina
Groningen, 1st Dec 2004
AW-OAC team
Light curves: WFI-Carina
Light curves: WFI-Carina
Groningen, 1st Dec 2004
AW-OAC team
An object from the OACDF
Groningen, 1st Dec 2004
AW-OAC team
Other Light curves from FORS data
Other Light curves WFI-CARINA data
Groningen, 1st Dec 2004
AW-OAC team
Open Points:
• Improve the throughput in the preparation steps
• Photometry (PSF, aperture on original images)
and relative amplitudes.
• User Manual
• AW integration
Groningen, 1st Dec 2004
AW-OAC team