Figure 1 corresponds to the histogram of an image of darkness of a CCCKAF-400. It represents the number of pixels having a given value. One notices the presence of 3 appreciably equidistant bumps, and of which, which more is, the heights are in a constant report/ratio compared to the preceding one. An examination on a large scale shows the existence of a fourth bump which satisfies these two criteria.
Figure 1. Histogram
On the following figures, the curve of figure 1 has undergoes a horizontal adjustment of +94 ADU, and a scale factor of 0.07 and it is posted in yellow superimposed on the original curve. A light offset in Y was applied locally to the curve of origin to hold account of the exponential component which is superimposed on the 4bumps. These appear show well the proposal advanced at the beginning, confirming one not and one ratio constant between two successive bumps of the histogram.
Figure 2. Bump 2 (red) and bump 1 (relocated + scale)
Figure 3. Bump 3 (red) and bump 2 (relocated + scale)
Figure 4. Bump 4 (red) and bump 3 (relocated + scale)
This relation between the bumps must make it possible toinclude/understand the origin of these abnormal pixels. It implies initially that it acts of a
phenomenon on very small scale (sub micronic).
For a reason to be determined, very small sites, distributed randomly on the surface of the CCC, generate a certain well defined number of electrons per unit oftime for a given temperature. An analysis made in addition goes up that they are always the same sites.
The distribution randomly is proven by the fact that, on a given surface (in fact a pixel of 9 microns with dimensions), there is a probability 1/N of having 1 site, 1/(N*N) to have 2of them, 1/(N*N*N) to have 3 etc of them...
The smoothness of the peaks (with the noise of measurement near) shows that it acts of a perfectly definite phenomenon; ALL the sites generate exactly the same number of electrons.
The most probable cause is to be sought in the development of the silicon bar. Other studies, which never mentioned the phenomenon described here, showed that, according to the supplier of the Silicon bar, for the same process applied by elsewhere, the behavior of the hot pixels was not the same one, with very important reports/ratios of about 4.
Same measurements were made on 4 CCC different from origin KODAK, produced at several years of interval, and one finds the same phenomenon, but, which more is, with practically them SAME ratios.
However, on the KAF-1600, the peaks do not appear as symmetrical as on the KAF-400. An observation of the distributions makes it possible to include/understand the origin of the skew, which results from a selection in manufacture: indeed, for the KAF-1600 (1536 * 1024), KODAK defines, for the classification of this CCC, a central zone and a peripheral zone, the rate of defects, for a given class, being élévé more in the peripheral zone. However, the detailed analysis shows that indeed there is less of hot pixels in this central zone, in a significant way. So that such a fluctuation of probability can be highlighted, it should be supposed that an important percentage of CCC leave to the dustbin, what would explain why the price grows much more quickly than surface (a KAF-6400, (3072 * 2048) costs approximately 100 times more expensive than a KAF-400 (768 * 512)
This was translated from a French Web site http://www.geocities.com/jcpelle/histogra.htm