Uploaded image for project: 'Data Management'
  1. Data Management
  2. DM-17429

PSFs on coadds are narrower than in model

    Details

    • Type: Story
    • Status: Done
    • Resolution: Done
    • Fix Version/s: None
    • Component/s: obs_subaru
    • Labels:
    • Story Points:
      1
    • Epic Link:
    • Sprint:
      DRP S19-5
    • Team:
      Data Release Production

      Description

      It has been observed that the PSF as measured on the coadd is narrower than the PSF model says it should be. This has found to be true on the warps as well, so we believe it's an artifact of warping. Bob Armstrong is experimenting with higher-order Lanczos kernels.

        Attachments

          Issue Links

            Activity

            Hide
            rhl Robert Lupton added a comment -

            Let's run the RC candidate and look at the timings. This change seems to be an impovement, so let's make it (if it has no bad consequences) without more testing.

            Show
            rhl Robert Lupton added a comment - Let's run the RC candidate and look at the timings. This change seems to be an impovement, so let's make it (if it has no bad consequences) without more testing.
            Hide
            rhl Robert Lupton added a comment -

            Yusra AlSayyad and Hisanori Furusawa agree that the timing is acceptable.

            We are waiting on the weak-lensing analysis of i-band coadds before deciding what to do in the next release.

            Show
            rhl Robert Lupton added a comment - Yusra AlSayyad and Hisanori Furusawa agree that the timing is acceptable. We are waiting on the weak-lensing analysis of i-band coadds before deciding what to do in the next release.
            Hide
            rhl Robert Lupton added a comment -

            How should we close this out?  My understanding was that we decided to go with Lanczos5, and that NOAJ would run S18A i-band with this setting; this has been done.  The WL team must sign off on the results, and decide what we use for the S19A, and in practice this means Bob Armstrong and Hisanori Furusawa

            Show
            rhl Robert Lupton added a comment - How should we close this out?  My understanding was that we decided to go with Lanczos5, and that NOAJ would run S18A i-band with this setting; this has been done.  The WL team must sign off on the results, and decide what we use for the S19A, and in practice this means Bob Armstrong  and Hisanori Furusawa
            Hide
            yusra Yusra AlSayyad added a comment -
            Show
            yusra Yusra AlSayyad added a comment - Paul Price will you take a look at https://github.com/lsst/obs_subaru/pull/196
            Hide
            price Paul Price added a comment -

            First commit needs a lot more detail (what's wrong with the current default, why is lanczos5 better, why not lanczos7, etc.). Otherwise, good.

            Show
            price Paul Price added a comment - First commit needs a lot more detail (what's wrong with the current default, why is lanczos5 better, why not lanczos7, etc.). Otherwise, good.
            Hide
            yusra Yusra AlSayyad added a comment -

            Thank you. Will do.

            If you're visiting this page from the future in search of why we changed the config, behold Bob Armstrong's plots showing the difference between the measured size of point sources and size of the PSF model as a function of FWHM.

            "Coadd" Is the original from S18A warped with the Lanczos3 kernel. It's positive which means that point sources are Narrower than the PSF models. Increasing the lanczos order to 5 moves it to ~ -0.001, which means the coadd point sources are slightly larger than the PSF models. (This is what I would expect given uncertainties in astrometry). Increasing the Lanczos7 makes the point sources even larger compared to the PSF models. We don't understand why this is, and are choosing lanczos5 for its empirical improvement, while we do further investigation

            Show
            yusra Yusra AlSayyad added a comment - Thank you. Will do. If you're visiting this page from the future in search of why we changed the config, behold Bob Armstrong 's plots showing the difference between the measured size of point sources and size of the PSF model as a function of FWHM. "Coadd" Is the original from S18A warped with the Lanczos3 kernel. It's positive which means that point sources are Narrower than the PSF models. Increasing the lanczos order to 5 moves it to ~ -0.001, which means the coadd point sources are slightly larger than the PSF models. (This is what I would expect given uncertainties in astrometry). Increasing the Lanczos7 makes the point sources even larger compared to the PSF models. We don't understand why this is, and are choosing lanczos5 for its empirical improvement, while we do further investigation
            Hide
            yusra Yusra AlSayyad added a comment -

            With the commit message edited and Jenkins happy, I plan to merge in an hour or so. 

            Show
            yusra Yusra AlSayyad added a comment - With the commit message edited and Jenkins happy, I plan to merge in an hour or so. 
            Hide
            yusra Yusra AlSayyad added a comment -

            Changed myself to assignee +1 SP to reflect only the config validation/change to the LSST stack. That estimate does not include Bob Armstrong's analysis which was 99% of the effort on this ticket.

            Show
            yusra Yusra AlSayyad added a comment - Changed myself to assignee +1 SP to reflect only the config validation/change to the LSST stack. That estimate does not include Bob Armstrong 's analysis which was 99% of the effort on this ticket.

              People

              • Assignee:
                yusra Yusra AlSayyad
                Reporter:
                price Paul Price
                Reviewers:
                Paul Price
                Watchers:
                Jim Bosch, Paul Price, Robert Lupton, Yusra AlSayyad
              • Votes:
                0 Vote for this issue
                Watchers:
                4 Start watching this issue

                Dates

                • Created:
                  Updated:
                  Resolved:

                  Summary Panel