Summary of outcomes from QUB tests and Paul's UK Visit

ISIS and ppsub appear to give similar results (with Paul's new version of ppsub with the major normalisation bug vix)

• if we choose the correct order of the convolution then ppsub gives reasonably

good subtractions, and dual convolution is not required. We saw good subtractions even with different psfs of FWHM~6 and 9 in the reference and input images (4 gaussian with the same order (2) as in the default parameters of ppSub. This gives 24 elements in the matrix and 71 degrees of freedom)

• if we do not use the convolved stack as the reference, we get results that we

are much more comfortable with, and can do our own photometry on the difference images much more confidently. We take Paul's point about the advantages of convolution but we see more reliable analysis, from standard photometry measurements with the unconvolved stacks.

• Our tests show there would be a huge gain (in quality and masking) if we implement the following strategy : create a High-Quality reference image per MDS per filter (good seeing, around 0.9")

this HQ-reference need only be one night of stable seeing, dithers combined without magic

always convolve HQ-reference to the new epoch (assume we will not get <0.9" often)

always sigma-clip and combine the 8-dithers at any epoch (*without* MAGIC) and give us the stack, we don't need the dithers

we will volunteer to pick the HQ-reference for each field and filter

for the HQ-reference, ideally on a good night we could aim for ~16 dithers to make sure we get full coverage of the edges

REMOVAL OF SPURIOUS SOURCES FROM DIFFERENCED CMF FILES

Using the 'BAD FLAGS' (see flag definitions here), the PSF_QF and the XY_MOMENTS values we have made some progress regarding the removal of spurious sources from the differenced image cmf files. Note that PSF_QF, or psf quality fit, represents the fraction of good pixels found within the aperture used to measure the source - therefore if this value is low then we know that a lot of the pixels have been masked. Below you will find a plot of all of the differenced sources found within one June 2009 MD08 exposure; red points are sources where a bad flag has been set against it, green points are source where no bad flag has been set. Also plotted are 7 of the best SN candidates that we at QUB found in the June 2009 data release (all epochs). It is clear that the SNe generally have low XY Moments (i.e. point source like) and a high PSF QF. Combined with the removal of the 'BAD FLAG' sources, we can use these two parameters to remove some spurious sources that contaminate the true sources.

Another statistic that shall be very useful for the removal of dipoles is the ratio of positive to negative flux within an aperture. Paul Price has told us that it is possible to measure this statistic (and any other suggested statistics?) an place them in the cmf files. This will help to remove even more spurious sources.

At the URL below you will find a differenced image 'exposure map' from MD08 taken from the June 2009 data release. If you click on any of the skycells you will find a list of all the sources detected in the difference image along with image stamps. It is possible to sort the sources by clicking on any of the column names and select subsets with the same flag by clicking on the flag descriptions:

​http://star.pst.qub.ac.uk/ps/exposure-maps/md08/16849/index.html