| | 3 | === Conclusions based on the initial ppImage implementation === |
| | 4 | |
| | 5 | 1. Individual row-by-row solutions are not stable between exposures. The following image shows a series of median solution images for OTA56/row 0 for the SAS footprint and for a series of similar science exposures taken after that date (highlighted frame is the first non-SAS footprint exposure). There are clear differences between the row-to-row patterns even within the short time frame of the SAS footprint. |
| | 6 | [[Image(median_0.0.jpg,1024px)]] |
| | 7 | |
| | 8 | 2. Applying only the PATTERN.MEDIAN to exposures (without DARK, FLAT, or PATTERN.CONTINUITY) yields a clear x-pixel direction gradient curve. This is effectively a result of solving for the DARK contribution on a per exposure basis. Applying DARK + FLAT removes this gradient on OTAs that are well behaved. This illustrates that the DARK model is not a significant problem for these OTAs. |
| | 9 | [[Image(science_gradient_0.0.jpg,800px)]] |
| | 10 | [[Image(dark_gradient_0.0.jpg,800px)]] |
| | 11 | [[Image(science_dark_mask_pattern_0.0.jpg,800px)]] |
| | 12 | |
| | 13 | 3. Applying only PATTERN.MEDIAN to a series of bias frames helps explain why the cell-column structures are not well removed by this correction. The larger scale ripples have a period that is not constrained to a cell. For example, in the highlighted bias below, I count ~11 ripples in the 4th (from left) column of cells. Whatever is driving these ripples is independent of the cell structure, and as such, these ripples do not reinforce in the PATTERN.MEDIAN calculation. |
| | 14 | [[Image(column_ripples_in_bias.jpg,1024px)]] |
