Mosaics from a WSDS-Processed Simulation (vsn 1.0)

Below we present co-adds created from 70 simulated framesets initially processed with Version 1.0 of the WSDS pipeline. These frames are from Ned Wright's May 2008 simulation and cover a region of ~8 square degrees near ecliptic coordinates λ=0, β=+30°. The simulation used sources from the 2MASS catalogs and included cosmic rays (both single spike and streaked hits), image persistence (latents), and diffraction-spike artifacts.

Broadly speaking, the following WSDS modules/subsystems were exercised: executive, simple ingest, instrumental calibration (ICAL), source detection (MDET) and photometry (WPHOT), single frame position reconstruction (SFPRex), and Frame Co-addition: Bmatch for background (offset) matching; AWOD for temporal outlier detection and masking; AWAIC for frame co-addition.

All co-adds below have the dimensions of a proposed WISE Atlas Image: ~1.56° x 1.56° and a pixel size of 1.375 arcsec. In-scan is in the vertical direction with Ecliptic North pointing down. Sorry! - this is how it would look like from Australia. In addition to outliers detected and flagged with AWOD, input pixels tagged as saturated and (simulated as) dead were also flagged and omitted from co-adds.

  1. Band-1 depth-of-coverage map 
     The color bar corresponds to the approximate depth (effective number 
     of frames). Accounts for flagged pixels.

  2. Band-1 intensity co-add
     The blue pixels are NaNs. These crept in because of (1) many saturated 
     input pixels were flagged in one spot (e.g., in the core of the bright 
     star at lower left), and (2) NaNs in the input uncertainty frames since
     these are used as inverse variance weights during co-addition. Their 
     source has been traced and they will be dealt with in a future version.

  3. Band-1 uncertainty (1-sigma) co-add 
     Compared to the coverage map in 1, 1/√N scaling is in force. 

  4. Band-1 outlier map
     White indicates that an outlier was detected in at least one 
     of the frame pixels in the stack at that location. Regions within
     yellow squares are zoomed-in in 5 and 6 below. Latent's dominate 
     the outlier count together with their diffraction spikes. Note 
     that frame convergence towards the pole means that frames from
     adjacent scans will be slighted rotated with respect to each other.
     This means that diffraction spikes will be mis-registered on the sky 
     since their geometry is fixed in frame-hardware coordinates. This 
     slight mis-registration means that outliers will result along 
     the diffraction spike arms.

  5. Band-1 zoom-in
     This is a zoom-in of the large yellow square shown in 4.
        Top left: outlier map;
        Top right: co-added region with outliers omitted; 
        Bottom left: depth-of-coverage map with outliers omitted; 
        Bottom right: co-added region with outliers included;
     Circles represent matching outliers: mainly latents and cosmic ray
     streaks. The yellow arrow points to a source that's blended with a
     cosmic ray streak (bottom co-add). This source is left untouched after 
     outlier flagging (top co-add).

  6. Band-1 input frames (corresponding to 5.)
     This montage shows all the input frames overlapping more-or-less with 
     the region shown in 5 (aligned in WCS coordinates). Corresponding
     outlier features are circled in green, and the red boxes contain
     the same source just to guide the eye. Also note the increased noise 
     compared to the co-adds in 5.

  7. Band-1 frame masks (corresponding to 6.)
     These are the frame pixel-masks corresponding to the montage in 6 with
     the same outlier features circled. Note that in addition to outliers
     detected with AWOD (shown in white), these masks also store
     information on the hardware state of each pixel, e.g., dead pixels 
     as simulated and flagged upstream (shown in light blue).

  8. Band-1 diffraction spike zoom-in
     This is a zoom-in of the small yellow square shown in 4. Why was a
     streak of outliers detected? This montage shows all the input frames
     overlapping at that location in the co-add, aligned in WCS coordinates.
     It turns out that the spike is only present in 3 of the 13 frames.
     So, it appears this is either a "latent spike", or one that appears only
     when the parent star (that causes the spike) falls on a frame. I suspect 
     the latter.

  9. Band-2 depth-of-coverage map
     Similar product to 1 above.

 10. Band-2 intensity co-add
     Similar product to 2 above.

 11. Band-2 uncertainty (1-sigma) co-add
     Similar product to 3 above.

 12. Band-2 outlier map
     Similar product to 4 above.

 13. Band-3 depth-of-coverage map
     Similar product to 1 above.

 14. Band-3 intensity co-add
     Similar product to 2 above.

 15. Band-3 uncertainty (1-sigma) co-add
     Similar product to 3 above.

 16. Band-3 outlier map
     Similar product to 4 above. Note the propensity of outliers! These are 
     actually simulated dead pixels (responsivity = 0) in the band-3 frames.
     They are picked-up by AWOD since the band-3 background is relatively high
     and all the dead pixels (with virtually no signal) stand-out as outliers.

 17. Band-4 depth-of-coverage map
     Similar product to 1 above.

 18. Band-4 intensity co-add
     Similar product to 2 above.

 19. Band-4 uncertainty (1-sigma) co-add
     Similar product to 3 above.

 20. Band-4 outlier map
     Similar product to 4 above.



Last update - 28 July 2008
F. Masci - IPAC