Status: To Do
Fix Version/s: None
As noted in
DM-25125, there was a notable uptick in the number of DIA Sources in reprocessing the HiTS2015 dataset with ap_pipe between April 2020 (circa w_2020_15) and May/June 2020 (w_2020_20 or later).
I originally suspected the increase was due to an ISR bug (fixed in
DM-25149), but now I think it was due to a change in image differencing, because the calexp source counts did not change.
The new sources are prominently concentrated near CCD edges - did pixel masking change while the overall number of DIA Sources remained unchanged, perhaps? - but despite many new sources near edges, some do also appear smattered all across CCDs.
The most substantial difference imaging change in the relevant timeframe was the decorrelation algorithm (
DM-24371). In the ap_verify CI metrics, we noted a decrease in the total number of unassociated DIA Objects as well as changes in the ratio of DIA Sources to direct image sources for each CCD. However, the templates used by ap_verify CI changed to higher resolution at the same time, so it is difficult to disentangle. (Whereas the templates have not changed in my ap_pipe reruns since January 2020.)
This ticket is to actually investigate the significant (~8%) increase in DIA Sources.