| | 7 | |
| | 8 | * ran into some problems with the big parallel dvomerge tests. local I/O hangups (possibly due to other operations on the same machine) caused the table writes to fail in some cases. New code in dvomerge mostly kept this from being a problem: a backup copy was kept that could be used for recovery. But it was a bit tricky to tell which dvomerge succeeded where. I added more tracking in the table headers of prior merges, and also enabled a replace option in dvomerge. A few files on ipp053 failed in a way which was not caught by the file system, breaking 3 or 4 tables. Fortunately, these could be recovered from the snap-shot generated 2012.05.03 (and re-merged with WISE and SuperCOS). Bottom line of the tests: large parallel dvomerge operations can succeed, but we need to carefully check for failed tables and re-merge them if needed. One nice feature is the dvomerge recovery method (in which previously merged tables are automatically detected) worked great. |
| | 9 | |
| | 10 | * discovered a problem with the WISE public release tables: about 8000 objects had an undocumented value for the photometric quality. I had to re-create the WISE ddvo database, and use the -replace option with dvomerge. I sent a bug report to Roc Cutri, who has since added that field to the documentation. |
| | 11 | |
| | 12 | * I created a script to query the big DVO database and assemble g-i, i-[J,H,K] color-color diagrams, and then fit for the stellar locus in this diagram. The goal was to check on the synthetic photometry comparison issues found earlier. For better or worse, the large effect that I had previously suspected was from 2MASS was not present. I tracked it back to the influence of USNO-B on the synthetic photometry. Despite the low weight, the large lever arm of the USNO-B photometry meant that the extinction was driven largely by that photometry, pushing around the predicted grizy mags. The large-scale pattern visible in that comparison (PS1_i_ vs Synth_i_) corresponds in fact to the POSS plate footprints. There ''is'' an effect from 2MASS calibrations visible in the g-i,i-H and g-i,i-K plots, but it is at a much lower level than I earlier expected. Those plots suggest primarily K-band calibration issues at the 2-4% level, not too inconsistent with the 2MASS claims. |
