Sampling and field considerations for PSF fitting photometry

 

Marcos Diaz - Sep/97

 

 

The task of performing PSF modeling and further fitting to stellar blends in crowded fields is faced by many science projects. The photometric accuracy in fields like the galactic bulge, Magellanic Clouds and clusters may be, in some cases, limited by deblending systematic errors. Some experiences indicate that there is always a price to pay in terms of S/N when one switches from aperture to profile fitting photometry, in order to deal with blends and seeing variations in dense fields. This extra source of noise is at least partially related to our poor knowledge of the PSF and the numerical manifold problem of deblending compact stellar groups.

Our tentative measure of the PSF of a current science exposure is empirical, often derived from "isolated" stars in the same field. The algorithmic description of the PSF is usually composed by an analytic core (a predefined function with 2 - 6 free parameters to be found from the data) plus a look-up table of deviations from the core. Both components require a minimum PSF sampling to be defined. The lack of symmetry and the presence of high-frequency spatial components in the PSF prevent the use of simple models while the undermodelling problems tend to be better undestood in well sampled images. Finally, some additional parametrization may be included if the PSF characteristics vary across the science field in some "well behaved" way.

The graph shows the error ratio Q between the magnitude estimates using the PSF fitting algorithms available in DAOPHOT II and the corresponding aperture photometry errors as a function of the image sampling. The data were synthesized by the IRAF/ARTDATA package assuming a Salpeter luminosity function and uniform random distribution. The field richness correspond to a "moderately crowded galactic bulge exposure". The image core is modeled by a elliptical gaussian while the input data PSF is a symmetric gaussian. All the properties of the PSF was based on ~30 fiducial stars evenly distributed over the frame. The sampling breakpoint is well defined at 2.8-3.0 pix/FWHM. For even more crowded fields breakpoint slightly shift towards higher values.

 

 

F ratio

Detector / Wavelength (µ)

Sampling (pix) (r0 = 30cm)

Sampling (pix) (r0 = 16cm)

f/16

40µ PICNIC / 1.8

2.1

3.0

f/16

27µ ALLADIN / 1.8

3.1

4.5

f/12.5

13.5µ EEV / 0.5

6.7

9.7

f/12.5

22µ EEV / 0.5

4.1

6.0

f/10

13.5µ EEV / 0.5

5.4

7.8

f/10

22µ EEV / 0.5

3.3

4.8