| | 1 | == DRAVG Issues & Concerns == |
| | 2 | |
| | 3 | == KPAT / DRAVG Priority Issues (+ Notes) == |
| | 4 | |
| | 5 | * latency |
| | 6 | * we have performed a [wiki:ThroughputDemo_20091104 throughput demonstration] which meets requirements. |
| | 7 | |
| | 8 | * magic masking |
| | 9 | * the latest large-scale run had a median streak count of 8. |
| | 10 | * we need to make an assessment of the current state of masking by masking bit (both warp and chip) |
| | 11 | * Heather has made an initial assessment of the masking using ~100 individual chips and warps. '''If this was done for 100 chip Files (not chip Runs) and equivalently for warps, then we need more statistics to say something useful.''' |
| | 12 | |
| | 13 | * stacking: noise correlation |
| | 14 | * we believe this has been addressed by Paul's recent work |
| | 15 | * '''we need an independent assessment if this is still an issue''' |
| | 16 | * Nigel is concerned about the photometry of psfs in the stacks |
| | 17 | * Nigel and Paul have both been investigating the situation. Some conclusions: |
| | 18 | 1. the latest version of the code (eg, stack_id 25532) results in scatter in psf magnitudes which are larger than the warp-warp photometry (0.04 vs 0.03). |
| | 19 | 2. the aperture magnitudes seem to be in better agreement, which says the pixels are probably OK |
| | 20 | 3. Paul started to use simtest to verify the quality of the pixel photometry |
| | 21 | |
| | 22 | * diff images: residuals and noise correlation |
| | 23 | * we have made some good progress on this with the dual convolution technique, and will generate a set of test images this week. |
| | 24 | * ''' this is critical''' : as of 2009/11/23 this has not been done |
| | 25 | * another significant source of residuals comes from poor astrometry |
| | 26 | * Gene has investigated the astrometry failures and improved the analysis (outlier clipping). Also, several cases of poor astrometry resulted from stars being falsely identified as cosmic rays and the wings being used in the fit instead. this should now be fixed in psphot. |
| | 27 | * we also also working on improving the quality of the detection analysis and the associated photometry in the diff images |
| | 28 | * Gene will be making this top priority again week of 2009/11/23 (this has been on the back burner while the astrometry issues were studied) |
| | 29 | |
| | 30 | * astrometry: some catastrophic failures (offsets>3") |
| | 31 | * some progress made in some extreme fields (M42). this may also result in improvements in other poor situations. otherwise under investigation. |
| | 32 | * Gene has included a number of fixes. there is a bit more work to make sure the warps do not use poor astrometry results. |
| | 33 | |
| | 34 | * xtalk artifacts. |
| | 35 | * for the near term, we will address this by masking the worst offenders in the camera masking stage. |
| | 36 | * still need to pin down the xtalk rules, but we can implement the xtalk we know. |
| | 37 | * as of 2009/11/23, this xtalk masking has been implemented by Chris, but with only limited testing to date. |
| | 38 | * Mark Huber says (preliminarily) it is looking '''good''' |
| | 39 | |
| | 40 | * other image artifacts. |
| | 41 | |
| | 42 | * photometric calibration |
| | 43 | * the zero point is now being calculated from the 'blue edge' of the ref - raw magnitude distribution to reduce the impact of halo stars. The error analysis is now done using bootstrap-resampling, but this can only tell us about the random errors, not the systematic errors from field-to-field. |
| | 44 | |
| | 45 | * star/galaxy separation |
| | 46 | * Eddie has made some plots of the quality of the current S/G flag: [wiki:PS1_rates]. The rules still need to be tweaked. |
