Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-6830

Investigate effects of turning on the Brighter-Fatter correction for single-frame processing of HSC data

    Details

    • Type: Story
    • Status: Done
    • Resolution: Done
    • Fix Version/s: None
    • Component/s: ip_isr, obs_subaru
    • Labels:
      None

      Description

      In the process of comparing HSC vs. LSST stack single-frame processing runs, we have been running with the Brighter-Fatter correction (BFC) turned off. The reason for this to begin with was that is was not yet available on the LSST stack when we started these comparisons. We also want to isolate as many features as possible in order to confidently assess their individual effects The functionality was ported on DM-4837 with a default of doBrighterFatter=False. This issue is to continue the single-visit run comparisons (see e.g. DM-5301, DM-6490, DM-6491) with BFC turned on on both stacks.

      In particular, we are finding that slight differences in the reference stars selected for a given CCD can result in significantly different psf models. Also, it was noted in DM-4960 that LSST seems to select reference stars to a brighter cutoff than HSC. If a given field has a larger fraction of bright stars considered in the psf modeling, it is conceivable that it will be more significantly influenced by the BF effect, thus causing the large CCD-to-CCD variations seen in, e.g. DM-6490 (https://jira.lsstcorp.org/secure/attachment/28213/compareVisit-v1322-diff_base_PsfFlux-skyZp.png).

        Attachments

          Issue Links

            Activity

            Hide
            lauren Lauren MacArthur added a comment -

            I have run the HSC visit 19696 with doBrighterFatter=False & True on both the LSST and HSC stacks.

            The following plots the before and after comparison of the difference between the trace radii of the SdssShape vs. the PSF model shape at the position of the object. Only stars that were selected for the PSF modeling (i.e. calib_psfUsed=True) are plotted here:

            BEFORE:

            AFTER:

            The BF effect is clearly seen in the BEFORE, and the AFTER looks much improved. The small offset in the AFTER is a known issue with the current implementation (see, e.g. HSC-1189).

            This confirms that the implementation on the LSST side is working as advertised and it would be safe to set the config on obs_subaru to run with doBrighterFatter=True by default.

            As for the speculation above about improving the zeropoint differences between the current "HSC stack" vs. LSST, the following plots a comparison between the PSF mags of independent runs on the two stacks:

            NO BF CORRECTION:

            WITH BF CORRECTION:

            There is a definite improvement for some outlier CCDs (e.g. 46, 96), and milder improvements throughout, but not enough to account for all of the zeropoint differences and large CCD-to-CCD variations observed (also being investigated in DM-6490).

            Show
            lauren Lauren MacArthur added a comment - I have run the HSC visit 19696 with doBrighterFatter=False & True on both the LSST and HSC stacks. The following plots the before and after comparison of the difference between the trace radii of the SdssShape vs. the PSF model shape at the position of the object. Only stars that were selected for the PSF modeling (i.e. calib_psfUsed=True ) are plotted here: BEFORE: AFTER: The BF effect is clearly seen in the BEFORE, and the AFTER looks much improved. The small offset in the AFTER is a known issue with the current implementation (see, e.g. HSC-1189 ). This confirms that the implementation on the LSST side is working as advertised and it would be safe to set the config on obs_subaru to run with doBrighterFatter=True by default. As for the speculation above about improving the zeropoint differences between the current "HSC stack" vs. LSST, the following plots a comparison between the PSF mags of independent runs on the two stacks: NO BF CORRECTION: WITH BF CORRECTION: There is a definite improvement for some outlier CCDs (e.g. 46, 96), and milder improvements throughout, but not enough to account for all of the zeropoint differences and large CCD-to-CCD variations observed (also being investigated in DM-6490 ).
            Hide
            lauren Lauren MacArthur added a comment -

            Setting the doBrighterFatter config override to True in obs_subaru will be done as part of DM-6815.

            This investigation is now complete.

            Show
            lauren Lauren MacArthur added a comment - Setting the doBrighterFatter config override to True in obs_subaru will be done as part of DM-6815 . This investigation is now complete.

              People

              • Assignee:
                lauren Lauren MacArthur
                Reporter:
                lauren Lauren MacArthur
                Watchers:
                Bob Armstrong, John Swinbank, Lauren MacArthur
              • Votes:
                0 Vote for this issue
                Watchers:
                3 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved:

                  Summary Panel