#############################################################################
### Written by Charles Romero
###
#############################################################################

If you have not already done so, please read Romero et al. 2015:
http://adsabs.harvard.edu/abs/2015ApJ...807..121R
and Romero et al. 2017:
http://adsabs.harvard.edu/abs/2017ApJ...838...86R
as these papers give a fairly comprehensive explanation of how the data were
created and how analysis was performed with the data.

#############################################################################

Here, I briefly add some practical information for the files:

#########    DATA   #########
The Data fits, Noise fits, SNR fits, and PtSrc fits files are all gridded with
the same astrometry. The center of the map is given by the X-ray (ACCEPT)
coordinates, and should also be chosen as the reference pixel within the fits
header. The pixel size is 1", and the total map is therefore 10' on a side.

Given the MUSTANG transfer function, it most likely that only the central
arcminute will be of interest/utility; however, for a few clusters, some
structure beyond the central arcminute can be seen.

#########    NOISE   #########
The noise files are created from gain-flipped TODs, and therefore contain
correlated noise. A more complete description of correlated noise (that is
portable and may potentially be made available to the public) is desired,
but does not currently exist. Still, as found in Romero et al. 2015, and
Romero et al. 2017, the noise is quite uncorrelated on scales less than 9"
(the beam FWHM). The weight map (contained in the fits as another extension)
has the RMS of the associated pixel, assuming white noise. If multiple
noise instances are desired, we recommend generating noise maps from the
weight map.

#########     SNR    #########
The documentation for how the SNR maps are calculated will be reported
separately. [No link is provided for SNR README in hopes that this comes.]
Already, a description was given in Romero et al. 2015 and Romero et al. 2017.
The important point to note is that the SNR maps are smoothed by a 10"
FWHM Gaussian kernel. In doing so, correlations within the noise
manifest themselves. That is, we test the sigma distribution of noise maps.
With our noise map, if we multiply it by the square root of our weight map,
then we have a sigma (SNR) map, where the distribution should be Gaussian,
centered on zero, with a width (sigma) of 1. However, as we smooth by larger
kernels, this width increases (for 10" FWHM, it is usually a little less than 2).
Therefore, in our SNR maps, we have corrected for this factor.

#########    PTSRC   #########
The point source files are the scaled point sources that were fit in Romero
et al. 2017. These point sources have been filtered by the transfer function,
and therefore have peaks that appear slightly below the values reported in
Romero et al. 2017. You may subtract these point source maps directly from
the data, or, if desired, you can rescale a point source map before
subtracting it (i.e. you have a better estimate of the point source flux).
This would preserve the shape of our beam and the filtering appears to still
be linear, even up to point source flux densities seen in MACS 1423.

#############################################################################