| | 27 | * Noise model fix: |
| | 28 | * Verified stack noise model: weighted average of covariance matrices, with weights set to inverse mean variance. |
| | 29 | * Shepherding magic verification data through fixed pipeline (magictest.20100423.%) |
| | 30 | * Investigating fix to pswarp to correct image and covariance scale using Jacobian rather than single value appropriate for entire FPA; needs more thought |
| | 31 | * New magic streaks have a wide distribution of ratios relative to the old magic streaks, with a median around a factor of a few, and extending out to more than 30. In terms of numbers, the median is about 500, compared to around 150. |
| | 32 | * Discovered that at least least one chip (XY57) has multiple cells (clustered around xy66) with bad gains (2--4 e/ADU, compared to ~1 for the rest of the cells) that is producing LOTS of streaks. |
| | 33 | * These bad gains appear to be present in data taken on sky April 19th. Sidik says we shouldn't trust the gains much at all since they are from a single x-ray measurement that was taken a long time ago (though there have been no voltage adjustments that would change the gains): FITS keyword comment is "est. gain from xray events c4923g0003X" |
| | 34 | * The exposure with the most streaks (2231 total) has most of the streaks on the edges of cells and masked regions on almost every skycell. These appear to be due to DetectStreaks not masking CONV.BAD (commented out)! |
| | 35 | * Gain problem disappears when masking CONV.BAD |
| | 36 | * Actually, many problems appear to be the use of an old version of DetectStreaks. Bathtub rings are now the main contaminants. |
| | 37 | * Blowing away these magic results to re-run |
| | 38 | |
