wiki:Summer2013/Simulations
Last modified 6 years ago Last modified on 04/22/2013 11:01:49 AM

Simulations for Summer2013

Stage 1a: Single-Epoch Model Investigations

  • Central raft only
  • i-band only (SEDs don't matter)
  • No CRs
  • Constant background (will be added in separately to improve performance; will not cause problems because we're only doing the center raft).
  • We'll work on eimages, using the processEImage.py task that Simon already wrote to create calexps. This lets us avoid ISR and amp combination. Only possible problem is that we may have to manually add some more metadata to the headers in processEImage.py.
  • Pointing at zenith
  • No blending:
    • Stars on a grid with small dithers (to avoid confusing astrometry.net).
    • Galaxies in between stars.
  • Stars should be < 18 mag, but below saturation limit.
  • Galaxies should be > 18 mag.
  • We probably don't need a separate image with a star at the position of each galaxy (the above should let us get PSF well enough already), but we'll ask if that changes.
  • Galaxies with Sersic indices of exactly 1,2,3,4 (1 and 4 most important)
  • Leave on diffraction spikes for now (should do no harm, may make analysis of shapelet approximation to PSF more realistic)
  • Turn off atmospheric layers that only affect the outer regions of bright stars (just an optimization(?), shouldn't affect these results)
  • No need to rotate galaxies to cancel shape noise.
  • Half-light radii between 0 and 3*FWHM, with realistic joint distribution of radius, flux, and ellipticity
  • Grid of 9 simulations: 3 seeing values x 3 exposure times (30s, 60s, 120s), all with the same population of galaxies

Stage 2a: Coadds

  • Input catalogs as above
  • Add small translational and rotational dithers
  • Normal LSST exposure times (30s to get realistic depth without snapcombine)
  • Realistic distribution of seeing
  • 32 visits
  • Also want a single exposure with the same depth as the coadd, and something like the average seeing

Stage 1b, 2b: More Realistic Galaxies

We wont need these images for a while, and we'll convene another meeting to discuss them before making them, in case desired parameters change.

  • mostly as above
  • multiple Sersic components
  • offset centers of components
  • non-integer Sersic indices

Support Software

One of our goals in this simulation design is to have a realistic PSF at each galaxy position, and realistic PSF variation across the field, without testing or otherwise relying on our ability to model the spatial variation in the PSFs. The above plans assumes we would write a simple custom star selector class that could just use the magnitude to select stars, and then use processCcd with no additional custom code (but with some unusual config parameters). We could also consider generating a separate set of sims in which there was a star with the same PSF at the position of each galaxy, but we currently feel this would require too much custom processing code relative to the current plan. UPDATE Andy B. may have updated the CatalogStarSelector? to do everything we need already; we should see if this does everything we need.