| | 8 | I have been looking more at the difference image analysis, and have identified and fixed a couple of issues. |
| | 9 | |
| | 10 | The first problem was related to the normalization needed to match the flux in the two images, a prerequisite to matching the PSFs. This is essentially relative aperture photometry on both of the images. However, the apertures were defined to be the same for both images; as the images had very different image qualities, the apertures measured different fractions of the total flux, resulting in a bias. I have fixed this by defining separate apertures for each image, and at the same time I've made the aperture size a bit more robustly defined in terms of an effective curve of growth. |
| | 11 | |
| | 12 | The second issue was specifically related to dual convolution. In the dual convolution process, we are attempting to find the convolution kernels kA and kB which together match a pair of images A and B. This problem is over-determined: we can always add something to both kA and kB to increase the size of the convolved image. To minimize this, we have a penalty function which penalizes large kernels relative to small kernels (following Yuan & Akerlof). The penalty function involves the second moments. However, it was defined based on the second moments of the kernel set, resulting in the same values for both images. Instead, the image with the larger PSF should have larger penalties for the same sized kernels (ie, we should tend to smooth the smaller image). I've fixed this by adding the second moments of the images in quadrature to the kernel second moments; this seems to have the desired effect. |
| | 13 | |
| 39 | | |
| 40 | | |
| 41 | | |
| 42 | | |
| 43 | | |
| 44 | | |
| | 45 | * PSPS |
| | 46 | * Tracked down problem with data given to Jim last week: my fault (one frame loaded multiple times). Fixed. |
| | 47 | * Attempts to load to Maui with new DXLayer, various problems: |
| | 48 | * wrong paths in config |
| | 49 | * {{{Invalid format specification}}} error: investigated |
| | 50 | * wrong permissions on sandbox so could not debug there |
| 46 | | |
| 47 | | |
| 48 | | |
| 49 | | |
| 50 | | |
| 51 | | |
| 52 | | |
| 53 | | |
| 54 | | |
| 55 | | |
| 56 | | |
| 57 | | |
| 58 | | |
| | 52 | * IPP |
| | 53 | * Czar on Monday: |
| | 54 | * registration down, restarted |
| | 55 | * high load on ipp45: killed distribution, seemed to help |
| | 56 | * MD08 stack-stack diff. Tutorial from Heather |
| | 57 | * updates to czar wiki page, specifically, the reverts section |
| | 58 | * czartool: |
| | 59 | * revert status now written to Db from czartool so status on page is accurate at all times |
| | 60 | * added Nebulous class. roboczar now using this to log cluster disk use. |
| | 61 | * creating plot of cluster space over all hosts; added to czartool page |
| | 62 | * ippToPsps: |
| | 63 | * re-wrote exposure-getting SQL to only get exposures that are in the provided DVO Db and to pull out correct relevant smf files |
| | 64 | * added datastore product, DVO db and min and max object IDs to database table for auditing purposes |
