README: SM3A ERO observation of Abell 2218

This directory contains the drizzled images created for the SM3A ERO
obervation of Abell 2218.  The images are:

 Image File    Drizzle Weight File    Filter    Exposure Time (s)
F814W_mos.fits   F814W_mosw.fits      F814W       12000
F606W_mos.fits   F606W_mosw.fits      F606W       10000
F450W_mos.fits   F450W_mosw.fits      F450W       12000

All individual exposures were 1000s long, with two exposures
taken per orbit.

The mosaic images contained in this directory were created
using onlychips 2 and 3 of the WFPC2 images. The purpose of 
the EROs was to rapidly demonstrate the success of the HST 
servicing mission; as these two chips contained most of the 
objects of visual and scientific interest in these datasets, 
only they were included in the mosaic.   

The observations were dithered using a combination of the
standard two and four-point WFPC2 dither patterns on top
of larger throws, for a total displacement of the telescope
by more than 5" in both the X and Y axes of the chips.

The dark created for the NGC 2392 observations was also used to 
process this data; however, in addition, a further bad pixel mask 
was created by combining the F450W data in place without removing
the dithers so that only features intrinsic to chip remained.
Cosmic rays and other image defects were located and 
masked using the techniqe described in the Fruchter and Hook
drizzling paper (see link in parent directory). 
The pixel mask containing both chip defects and cosmic rays
was applied to weight files which reflect the sensitivity of the chips
in each filter as a function of position.  The final
resulting weights reflect the relative sensitivity across the
images fairly accurately, but cannot be used to predict the absolute
variance of the final image.

The drizzled images have a pixel scale one-half that of the 
original WFPC2 images, or about 0."05.   The full set of Drizzle
parameters used is recorded in the image headers.  The user of these 
images should remember that because the noise in drizzled images is 
correlated, the true image noise over an area of N (>>1)  pixels is 
greater than root(N) times the image r.m.s.  In particular, with the 
drizzle parameters used  here, scale=0.5 and pixfrac=0.6, the final 
the true image noise is about 1.66 times greater than that predicted by 
root(N).

Note for users wishing to work on individual images:

The dark calibration image used to process this data has been placed
in the archive.  It is now the "recommended" darkfile for these datasets,
and will be used in calibration-on-the-fly processing by the archive
of these datasets.