Photometry

We performed photometry on both the original and deconvolved images of NGC 6352. The Stellar Photometry Software (SPS) developed by Janes and Heasley (1993) was used to perform PSF-fitting photometry on the original (blurry) data. For each master frame, a single spatially-constant PSF was generated from a large sample of bright stars scattered uniformly about the frame. This PSF was used to measure the magnitudes of all stars found using the SPS automatic star-finding algorithm. After this first pass, a few stars were added to the list from visual inspection, and the fits were repeated. The resulting color-magnitude diagram (CMD) is shown in Fig. 1a.

The photometric zero-points have been adjusted to correspond to the F555W and F785LP magnitudes determined from the restored images from CCD #7.

Aperture photometry was performed on all stars identified in the deconvolved master frames with aperture radii of 4 and 5 pixels for the F555W and F785LP images respectively. Due to the larger diffraction-limited core radius, more power in the first diffraction ring and a poorer match between the Tiny TIM and observed PSFs at the longer wavelength, the stellar cores in the restored F785LP filter images were slightly larger than in the F555W images, and we were able to use a smaller aperture radius in the F555W images to help minimize crowding effects. Aperture corrections were determined from the brightest uncrowded stars in each image subsection and applied to the magnitudes of all the measured stars. The corrections for the CCD #7 images were 0.059 and 0.036 mag, F555W and F785LP, respectively. For the CCD#8 images, the equivalent corrections were 0.044 and 0.080 mag.

The CMD for all stars whose nearest neighbor was greater than 5 pixels away is plotted Fig. 1b. Comparison with the CMD from the PSF fits to the unrestored images indicates that the main-sequence is better defined in the deconvolved data, especially at the faintest limit. Comparison of these results with those from our simulation (CFC) indicate that the precision of the NGC 6352 data is limited mainly by the mismatch between the actual PSF and that used in the deconvolution. We believe we could obtain even better photometric results from a PSF which matched the observed PSF more closely.

Fig. 2 illustrates the relative quality of the two different photometric reductions.

We have plotted histograms of the color difference between a star's position in the CMD and the mean main-sequence defined by a parabolic fit to the data below the level of the main-sequence turn-off. Note that our deconvolution procedure allowed for the spatial variation of the PSF while the PSF-fitting software did not. Four histograms are plotted, which represent the data for main-sequence stars of decreasing brightness in 1-magnitude bins. The dotted lines represent the scatter about the mean sequence determined from the PSF fits to the unrestored data, and the solid lines indicate the scatter in the aperture photometry on the restored images. Since the largest aperture used was 5 pixels in radius, only data from stars whose nearest neighbor was more than 5 pixels distant are included in the figures to minimize scatter in the diagram due to the use of aperture photometry in crowded fields. This scatter is obviously not due to the effects of the deconvolution algorithm but rather depends on the density of stars in the original image. Since the measurement aperture used was 5 pixels in radius, a nearest neighbor limit of 10 pixels would be necessary to completely remove the effects of crowding on the aperture photometry. In our experiments, we found that using a nearest neighbor criterion this strict removed an unacceptably large number of stars from our sample and did not produce a sequence noticably different from that shown in Fig. 2 otherwise.

With the profile fitting software, we were able to measure 727 stars fainter than 18th magnitude. Because of the higher signal-to-noise ratio in the deconvolved images and because the brightest stars and their environs were not measurable with SPS in the blurry data due to the large wings of the PSF, more stars were measured in both filters of the deconvolved data than in the original images, especially at the faint end. Thus, even after omitting stars with close neighbors from the histograms (about 15 percent of the total sample), there are a total of 834 stars represented in the histograms of the deconvolved data. Fig. 2 indicates that the scatter about the mean faint star sequence is much smaller for the aperture photometry of the deconvolved images than for the PSF fits to the original images. This effect is small at the brightest and faintest magnitudes but is particularly evident in the magnitude range 20 F555W 21.

Since NGC 6352 lies toward the Galactic center (l = 341, b = 7), even perfect photometry of the cluster images will produce CMDs with residual scatter due to the large field star contamination from the Galactic Bulge. Thus while some of the scatter in Figs. 1 and 2 is obviously attributable to photometric error, many of the larger deviations from the cluster sequence are likely due to this field contamination. Quantification of the extent of contamination in the CMD awaits analysis of our off-cluster field images. While a more extensive comparison between the photometric precision possible from the restored vs. unrestored cluster images in the absense of field star contamination would have been possible with simulated data, we believe the comparison between the results for actual data is more relevant to evaluation of the usefulness of the restoration procedure.