Bright Star Parameter Tuning

Bright stars present a number of problems to both point source and extended source extraction. These objects create "halos" (low surface brightness emission extending well beyond the PSF), diffraction spikes, horizontal stripes, glints and persistence residual ghosts. For more info on the horrors of bright stars, see GALWORKS Bright Star Cleansing .

This memo is concerned with the specific task of tuning the bright star blanking parameters, including the "confusion" radius, diffraction spike length and the horizontal stripe blank criterion. The procedure is as follows: collect coadds containing bright stars (i.e., K < 10 mag), measure the confusion radius & diffraction spike length (per band) and rate the necessity of horizontal stripe blanking (per band). The plots below summarize the results thus far. These results are preliminary. As more data is analyzed, the plots will be updated.

To date (970923) over 150 stars have been examined in detail. The stars are all located at mid galactic latitude regions (glat from 30 to 50 degrees). They range in brightness between 2 and 10 mag (in Read 1) at JHK (of course, saturation begins around 4 or 5 mag, so the brightest stars have very uncertain R1 mags -- the plots reflect this). See the following image for an example of an evil bright star (4th mag or so).

• JHK coadd images of a bright star field (K = 4)

The confusion radius for each object was found by closely examining each coadd (J, H & K) and finding the point at which the "halo" emission dissolves into the background noise. This method sides on the conservative side, in the sense that for galaxy extraction we want to avoid ALL light associated with bright stars. Similarly, for the diffraction spikes the coadds were carefully examined and the spike length measured for each bright star. Again, we want the spike light to fade well into the background noise. Finally, the horizontal stripes were carefully examined. For bright stars there are always three stripes, one centered on the star (referred to as the "primary" stripe), and the other stripes located ~286 pixels above and below the bright star (referred to as "secondary" stripes). For fainter stars, only the stripe centered on the star remains, and for the faintest stars this stripe dissappears altogether. For each star we rated the "appearance" of each stripe (0 == no appearance, 1 == appearance). The histograms are still prone to small sample statistics, but as more coadds are examined the values should stabilize. The following images show examples of a bright stars with the confusion radius marked (red circle) and the spike length denoted by green crosses. First the first example, the confusion radius is about 100 pixels for J and H, and 90 pix for K. The spike length is close to 150 pixels. For the second example, a moderately bright star, the confusion radius is about 30 pix. The final example shows a relatively faint bright star (K = 8.4); the confusion radius is now about 20 pix.

• JHK coadd images of a bright star (J=3.6, K = 4.6) with confusion radius marked (red) and spikes lengths denoted by green crosses;

• JHK coadd images of a bright star (J=7.3, K = 5.8) with confusion radius marked (red) and spikes lengths denoted by green crosses; note the horizontal stripe (primary).

• JHK coadd images of a bright star (J=9.2, K = 8.4) with confusion radius marked (red) and spikes lengths denoted by green crosses; note the horizontal stripe (primary).

Results

The confusion radii, spike lengths and horizontal stripe ratings are summarized in the following plots. We also show a fit to the data, comprised of a power series. The horizontal stripe analysis is summarized by a "rating". By this we mean the number of stars (per mag bin) in which a horizontal stripe is observed, normalized by the total number of sources. Thus, for the brightest stars, these stripes are always observed, so they have a appearance rating of 1.0. For the faintest stars, these stripes begin to dissappear and the rating approaches zero.

• Confusion Radius with power series fit

• Diffraction Spike Length with power series fit

• Horizontal Stripe Rating
  white solid line == primary stripe (centered on star)
  yellow dashed line == secondary stripes

Bright Stars at Low Galactic Latitude

A number of scans through the galactic plane were processed and the bright stars analyzed accordingly. The fields cover glats from 4 to 7 degrees, so the source density is very high (approx 9000 srcs/deg*2 brighter than K=14). The following plots show the results for bright stars. For comparison, the low stellar density results are overlayed on the plots.

• Confusion Radius
  low source density == white points
  high source density == purple points

The confusion radius is essentially the same for the low and high stellar density cases.

• Diffraction Spike Length low source density == white points
  high source density == purple points

At the faint end, K > 6, the diffraction spike lengths for the high stellar density case are systematically smaller than that of the low stellar density case. This effect is due to the increased confusion noise for the high stellar density case.

• Horizontal Stripe Rating
  white solid line == primary stripe for low source density
  white dashed line == secondary stripes for low source density
  purple solid line == primary stripe for high source density
  purple dashed line == secondary stripes for high source density
  

Similar to the diffraction spike lengths, the horizontal stripes are systematically less pronounced in the high stellar density case (compared to the low stellar density case). This effect is due to the increased confusion noise for the high stellar density images.