Date: 17 Sep 2004 1100 PDT

Attendees: Sean, Mark L., Lexi, Bill, Dan, Mark D., Jason, Matt

Agenda:
   1) Update on progress in developing background estimators
   2) New business

Update on background estimators: No progress has been made on improving background estimators. Sean tried several canned IDL routines including tri_surf, interpolate and maskinterp. maskinterp did not work at all; the other routines used non-local information in performing the background estimate which resulted in a poor fit to the IRAC data. tri_surf and the thin-plate spline routing grid_tps were extremely processor intensive.

Use of uncorrected images: After much discussion, the consensus of the group was that future investigations of correcting muxbleed should be performed on uncorrected data as the current correction may distort the functional form of the artifact. However, the creation of BCDs without the muxbleed correction is non-trivial as the data must be run through the offline pipeline. Using the offline pipeline requires a significant overhead in identifying and transferring the correction calibration files. As a result, we are restricting our initial test cases to uncorrected versions of the 30 Doradus ERO and EliasN1 GOODS verification field AORs. Once we have determined a better and robust correction, then the muxbleed module can be turned off in the online pipeline.

Muxbleed in GOODS ultra deep field: Lexi discussed his analysis of muxbleed in the GOODS ultra deep data. Lexi has registered the muxbleed for multiple BCDs to examine the variation of muxbleed with position on the array. To the eye, there is very little variation in the muxbleed pattern. The coadd of multiple BCDs shows that the muxbleed extends for some extent and is significant when coadding many frames. The GOODS team can use the rotation of the field to mitigate muxbleed.

Mark D. pointed out the need to account for the effects of flat-fielding the data when attempting fits of muxbleed. We will continue to work with flat fielded data for the present as the variations due to the flat are small compared to our inability to estimate the background and find a functional form to reasonably fit the artifacts in the data. Mark also pointed out the need to have a good estimate of the PSF when trying to fit the muxbleed close to the source. At present, we only try to fit the muxbleed at a distance of 8 or more pixels from the peak of the PSF.

Action items:

1) Lexi to send SSC 2d images of muxbleed vs. output

2) Sean to send muxbleed corrected and uncorrected images to Lexi

3) Bill and Sean to work on muxbleed vs. fowler test

Next meeting: TBD