The key features of the MATPHOT algorithm for precise and accurate stellar photometry and astrometry using discrete point spread functions (PSFs) are described. A discrete PSF is a sampled version of a continuous PSF, which describes the two-dimensional probability distribution of photons from a point source (star) just above the detector. The shape information about the photon scattering pattern of a discrete PSF is typically encoded using a numerical table (matrix) or an FITS (Flexible Image Transport System) image file. Discrete PSFs are shifted within an observational model using a 21-pixel-wide damped sinc function, and position-partial derivatives are computed using a five-point numerical differentiation formula. Precise and accurate stellar photometry and astrometry are achieved with undersampled CCD (charge-coupled device) observations by using supersampled discrete PSFs that are sampled two, three or more times more finely than the observational data. The precision and accuracy of the MATPHOT algorithm is demonstrated by using the c-language mpd code to analyse simulated CCD stellar observations; measured performance is compared with a theoretical performance model. Detailed analysis of simulated Next Generation Space Telescope observations demonstrate that millipixel relative astrometry and mmag photometric precision is achievable with complicated space-based discrete PSFs.

Online MNRAS abstract
Online MNRAS article (if a subscriber to MNRAS)
PDF version of article (... if not)
astro-ph/0505455 ( PDF) astro-ph preprint

The National Center for Supercomputer Applications (NCSA) and the Ohio Supercomputer Center (OSC) sponsored the Reconfigurable Systems Summer Institute which was held July 11--13, 2005 at the Beckman Institute for Advanced Science and Technology on the campus of the University of Illinois at Urbana-Champaign.

Volodymyr Kindratenko, a senior research scientist at NCSA, gave a presentation
First-hand experience on porting MATPHOT code to SRC platform ( PPT) ( PDF)
based on the MATPHOT code for stellar photometry and astrometry with discrete point spread functions.

Dr. Kindratenko analyzed the MATHOT code and determined that the current implementation spends approximately 90% of the total computation time calculating the two-dimensional damped sinc interpolation of PSF models. Kindratenko ported the C implementation of the interpolation algorithm to an SRC MAPstation which is based on SRC's patented MAP processor that has 2 Xilinx Field Programmable Gate Arrays ( FPGAs).

---

American Astronomical Society 205th Meeting poster #153.09
2005 January 13, San Diego, CA

The MATPHOT Algorithm for Accurate and Precise Stellar Photometry and Astrometry Using Discrete Point Spread Functions

AAS abstract
ADS abstract

---

Mathematical Challenges in Astronomical Imaging
Institute for Pure and Applied Mathematics
2004 January 27, UCLA, Los Angeles, CA

PDF version of the AI2004 poster

Mathematical Challenges of using Point Spread Function Analysis Algorithms in Astronomical Imaging
Ken Mighell (National Optical Astronomical Observatory)
PDF version of the presentation

---

MATPHOT release: 2005OCT28

Retrieve the MX source code file:

MX source code gzipped tar ball:
mx_200511091448.tgz
(^- click the right mouse button on this link)

Build MX (tested on Apple OS X Tiger and RedHat Linux 7.2):

Unpack the tar ball by typing the following command:
zcat mx_200511091448.tgz | tar -xvf -

Go down to the new mx directory:
cd mx

Build MX:
make

CHECK: If the file bin/mx_mpd exists, you have built MATPHOT!

Now try out MATPHOT:

Go down to the src/mpd directory:
cd src/mpd

Type the following command:

demos/ngst
and you will see MATPHOT analysis of 10 simulated
Next Generation Space Telescope observations.

Now start the ds9 astronomical image display
by typing the following command:

ds9

We can see the MATPHOT fitting process live using ds9
by typing the following command:

demos/ngst

Data Model Residual

We can simulate 20,000 NGST observations
by typing the following commands:

demos/ngst_20000 > ngstx2 &

tail -f ngstx2

This will take a while...
and will go faster if ds9 is not used to visualize the fitting process of all 20,000 stars :-)

When finished, use custom SuperMongo macros to look at the results
by typing the following command:

sm/go ngstx

The resultant plot should look like the following plot:

The grey lines are the predicted median values; the wide grey bands show the predicted top fence range of the box-and-whiskers plots; and the dashed grey lines show the predicted 5-sigma limits. This plot shows that millipixel relative astrometry and millimag photometric accuracy is achievable with very complicated space-based discrete Point Spread Functions. See the AAS 205 poster for more details.

---

More to come...

---

Kenneth Mighell
Associate Scientist  
Kitt Peak National Observatory
National Optical Astronomy Observatory
 
EMAIL: mighell@noao.edu
 MAIL: P.O. Box 26732, Tucson, AZ  85726-6732
FEDEX: 950 N. Cherry Ave., Tucson, AZ  85719
PHONE: (520) 318-8391
  FAX: (520) 318-8360
  URL: http://www.noao.edu/staff/mighell/

---

Last updated: 2005 October 28 @ 17:06 MST

---