The tuning of the active deblending parameters is discussed in the Ken Marsh's page. Figures presented here provide some statistical results on active deblending. Figures 1 to 5 contains all scans from the night 971116n.

I. Analysis
a. Initial chi2, improved chi2:

Figure 1 is a color-magnitude diagram obtained for actively deblended sources. The Y axis corresponds to a reconstruction of the original magnitude by adding the flux of the deblended components. These two figures are generated for a different chi2 minimum improvement (dchi) during the deblending process.
For Figure 1a and  Figure 1b, dchi=0.5 and dchi=1.0, respectively. Diamonds correspond to successful deblends, crosses to inconsistent deblends. The inconsistency refers to the case when the secondary source for each band cannot be cross-identified by the bandmerge process.
 

Figure 1a	Figure 1b

 

Figure 2a represent deblended H-Ks vs the initial chi2 (before deblending) for sources with dchi=0.5. It shows that inconsistent deblends (crosses) occur for all chi2 values. Figure 2b gives the dchi distribution of the successfully deblended sources.
 

Figure 2a	Figure 2b

 

The optimal value for the dchi parameter is derived from the Figure 3. Between dchi=0.5 to 1.0 the fraction successful deblends increase to 88% to 94% and remains at this fraction for larger dchi.
 

Figure 3

 

b. Color-color diagrams:

Color-color diagrams are shown in Figures 4:
4a) version 2 of the pipeline, before the active deblending implementation, for sources that corresponds to Figure 4c. Extremely blue objects were identified as the result of an inadequate design of the passive deblending algorithm in Ks (see Laurent's page).
4b) version 3 of the pipeline, dchi=0.5, successful deblends only.
4c) dchi=1.0, successful deblends only. One sigma error bars are indicated for objects far from the normal main sequence star locus.
4d) version 3 with the active deblending turned OFF for sources corresponding to those represented in Figure 4c. Extremely blue objects of the v2 pipeline (Figure 4a) do not exist in the v3 pipeline with improved PSFs.
4e) version 3 with the active deblending turned ON but with remerged magnitudes by adding fluxes from each deblended component.
4f) version 3 for sources actively deblended in less than 3 bands. For non-deblended bands, the large aperture photometry gives the upper limit.
 

Figure 4a	Figure 4b	Figure 4c
Figure 4d	Figure 4e	Figure 4f

 

Figure 5 represents the density color-color diagram for deblended sources (same as Figure 4c) and non-deblended sources (i.e. blend flag = 1 in all bands). Contours are normalized to the number of objects in each plot and begin at a value of 2 with a step of 2 (units are arbitrary). The lowest contour is in agreement for the two plots but the maxima are 21 and 9 for the deblended and non-deblended sources, respectively. The statistic for the deblended sources is still low (1228 objects) but this figure suggests that the stellar population is different with an excess of sources around J-H=0.2 H-K=0.05 that would correspond to F5 type stars.
 

Figure 5

 

c. Relative numbers of deblended sources

Figures 6 (dchi=1.0) compare the number of deblended sources (active successful, active inconsistent, passive) with the total number of deblended source in JHKs simultaneously (6a), the total number of sources (6b), the total number of sources deblended in at least one band (6c). Figures 6d give the fraction of deblended sources with respect to the total number of sources in a scan.
Each point corresponds to a science scan, 6 full nights are used in these figures (971116n, 980319s, 980403n, 990523n, 990723s, 990923s) which cover fields from high Galactic latitudes to the Baade's window close to the Galactic center.
Plots generated with the new version of the pipeline (April, 5th : new nl.prophot? parms and new galworks with fits header fix).
 

Figure 6a	Figure 6b	Figure 6c	Figure 6d

 

II.  Optimized version of prophot (4/20/01)

Figure 7a is the new version of Figure 4c (same night: 971116n). Figures 7b and 7c are for the primary and secondary components and the active deblending, respectively. Figures 7d and 7e are for the 6 nights (971116n, 980319s, 980403n, 990523n, 990723s, 990923s) and include very high density fields.
 

Figure 7a	Figure 7b	Figure 7c
	Figure 7d	Figure 7e

Number of deblended sources for two different versions of prophot (night 971116n):

Date	march, 28	april, 20
Active	1865	180
Passive	383	399

The evolution of color-color diagrams for different delta(mag) between the primary and the secondary components, or for different separation distances between the two components is represented in these two animations: delta(mag), distance separation.
6 nights are used (971116n, 980319s, 980403n, 990523n, 990723s, 990923s).
 

Figures 8 are the new version of figures 6. The number of active deblends has strongly decreased and is now close to the number of passive deblends. Arcs for low density fields (N<1000 stars/scan) in Figure 8d are hyperbolae that correspond to a small integer number of deblends.
 

Figure 8a	Figure 8b	Figure 8c	Figure 8d