A set of ``perfect'' survey data was produced to simulate fields that cover a
range of stellar densities. The fields correspond to galactic latitudes
of 90, 25, 10, 5, and 0
, respectively, at a longitude of 50
.
A good subpixel dithering scheme (as described above) was used. The Lorentzian
profile had a FWHM = 2.5 arcsec for all stars; no variation. A good dithering
pattern was introduced by stepping (0.15,42.23) pixels in X and Y. Also, there
was no subpixel response included, nor any dead pixels. Therefore, this is a
test of ``optimally observed'' survey data.
The results for the standard run are shown in Figures 3--7. In all five plots, a comparison of the numbers measured versus the input ``truth'' is shown. The top left plot shows the difference between the profile-fit and true magnitudes for each star matched; the top right has the difference between large-aperture photometry and truth. The bottom two plots show the difference between X and Y position as returned by Kamphot and the input truth. All are plotted as a function of input magnitude.
Results show that the best Kamphot photometry, for the bright stars over a
large range in density, has a 
mag. The bright star photometric
and positional accuracies returned by Kamphot are improved at moderate
densitites because of
more accurate offsets and due to an improved point-spread function (PSF);
more, higher S/N stars are available.
A re-reduction of the b=90
frames using a b=25
PSF shows small
improvement; thus it seems that the offsets are more important to both accurate
photometry and positions.
For most densities, the magnitude of stars for which Kamphot returns a
is between K = 14.5 and 15.0 . At b=0, the limit is 1-2
magnitudes higher due to incompleteness. Above K = 13 in that confused region,
the photometry results are consistent with what we see in the M92 data.