HST/ACS observations of LCDCS-0829
Proposal ID: 10492
PI: Thomas Erben

I. The data

Version 1.1

This folder contains the reduced files:

A. Science and weight frames of the three filters
LCDCS-0829_F475W_sci.fits
LCDCS-0829_F475W_wht.fits
LCDCS-0829_F814W_sci.fits
LCDCS-0829_F814W_wht.fits
LCDCS-0829_F850LP_sci.fits
LCDCS-0829_F850LP_wht.fits
The weights give the effective exposure time for the corresponding pixel (depends on number of good pixels drizzeld to this position)

B. Masks which can be used as flag image to reject objects near the image borders or stellar diffraction spikes (which overlap with some arclets...)
LCDCS-0829_flagmask_border.fits
LCDCS-0829_flagmask_border_brightstars.fits

C. My beginner's attempt for a color image (wait for Mischa's image)
RXJ.tif
RXJ.jpg

When you work with the images be aware that the noise is correlated for nearby pixels. This is particularly important for defining S/N. See e.g. Casertano et al. (2000), AJ, 120:2747-2824 for a discussion.

D. Photometric zeropoints (added August 14, 2006)

Zeropoints are given at http://www.stsci.edu/hst/acs/analysis/zeropoints
In the AB system they are:
F475W 26.068
F814W 25.937
F850LP 24.862 
Note however that these zeropoints do NOT account for charge loss due to CTE degradation. In worst cases the systematic error can be up to ~0.1 mag, so if you need that precision, check the links below. Fitting formulas exists for point sources depending on chip position, object S/N, time, and background level (which needs to be determined from the flt images):

http://www.stsci.edu/hst/acs/documents/handbooks/DataHandbookv5/acs_Ch62.html#101906
http://www.stsci.edu/hst/acs/analysis/zeropoints
http://www.stsci.edu/hst/acs/documents/isrs/isr0503.pdf
http://www.stsci.edu/hst/acs/documents/isrs/isr0406.pdf

II. Data reduction

The data was processed using the MultiDrizzle (v.2.7.0) package (Koekemoer et al. 2002) with a number of add-ons.

II.A The standard reduction by MultiDrizzle include:
  1. Alignment of exposures according to image WCS
  2. Background subtraction
  3. Application of a static bad pixel mask
  4. 1st drizzling onto a distortion corrected output grid
  5. Rejection of Cosmic Rays via a median image and blotting back to input coordinate frame
  6. 2nd drizzling onto a distortion corrected output grid, taking cosmics into account

II.B Non-default MultiDrizzle parameters applied:
  1. Pixel scale for final image (II.A.6) 0.03''/pixel (original WFC: 0.05''/pixel) in order to decrease extra ellipticity noise/aliasing, with a kernel SQUARE (default)
  2. For the final drizzling (II.A.6) the input pixel were shrunken (pixfrac=0.9) in order to decrease noise correlations and yield a sharper psf
  3. I tried to optimise the MultiDrizzle sky subtraction (with little success, see below) and in the end did it outside of MultiDrizzle

II.C My add-ons:
  1. Manual masking of satellite tracks
  2. Refining relative shifts/rotations of exposures for one filter (typical accuracy: < 0.05 WFC pixels)
  3. Refining shifts such that the coadded images of the different filters are matched to the same grid (accuracy < 0.03 WFC pixels)
  4. Refining the bad pixel masks using several steps:
	-median image of the _flt files for hot/bad columns (detection with noao.imred package)
	-median of _cor files created by MultiDrizzle for hot pixels (detection with L.A.Cosmic)
	-inverted median of _cor files with previously masked pixels set to 0 for detection of cold columns (detection with noao.imred package)
	-manual rejection of some remaining warm columns
  5. Determination of background for each quadrant in the flt files with a mode of unmasked pixels (objects have also been masked out)


III. Discussion background subtraction

Already in the flt images different background levels for the image quadrants (each with a seperate readout amplifier) were visible.
These are probably due to slight problems in the bias subtraction. I sent an email to help@stsci.edu and got this as an answer from Anton Koekemoer. Therefore I directly modeled and subtracted the background in each flt image for each quadrant. 

IV. Remaining problems

In some regions several neighbouring columns had to be masked, which significantly reduced the S/N there. The effect is strongest seen in LCDCS-0829_F475W_sci.fits. It can also be seen in the color image as blue regions elongated in the readout direction.
I don't think there is much we can do about it, but if you have any idea, please let me know (tim@astro.uni-bonn.de)

For questions contact:
Tim Schrabback, tim@astro.uni-bonn.de