Technical Analysis of

The Fuji Photofilm Conference Digest on Super CCD

November, 1999

Fuji Photofilm had a presentation about their new Super CCD in the conference of Institute of Image Information and Television Engineers of Japan (ITE) on November 24^th, 1999. This is the analysis for the technical digest that consists of six text pages in Japanese and 15 figures in English.

Summary of The Paper:

The paper was written form the camera system side of view. The integration density of pixel on a CCD chip was almost reached at the technical limit of the current semiconductor-processing rule. Thus the different approach from the present way had been required to increase the CCD resolution. If the pixels were arranged in the way of interleaving in adjacent two lines, the spatial frequency response would be diagonal other than square for the square pixel arrangement. The spatial frequency for the object in the real world was supposed to be higher in horizontal and vertical directions. Therefore, the diagonal frequency response for the interleaved pixel arrangement was fit to the real world, they were claiming in the paper.

Analysis
1. Device Structure

In the coming years, if the number of pixels on a CCD die is becoming larger, the pixel size will be smaller, and in accordeance, the ration of a photo diode area in a pixel will be smaller and the sensitivity will be also reduced. Hence the systematical approach to increase the resolution will be necessary for the future CCD. In the technical paper, Fuji is comparing the Supper CCD with a conventional interleave IT CCD while a progressive CCD were cited in the press release or in their Web page.

Three types of CCD structures are shown in Fig.1 and the comparison is shown in the below table.

Table 1. The Comparison of Three Types of CCD
Type	Interleave IT CCD	Progressive CCD	Supper CCD
Photo Diode Shape	Rectangle	Rectangle	Octagonal
Photo Diode Area	Smaller	Smaller	Larger
Poly-Silicon Layer	2 Layers	3 Layers	2 Layers
V-Drive Phase	4 Phase	3 Phase	4 Phase
Mechnical Shutter	Required	Not Necessary	Not Necessary

The merits of Supper CCD are

1) Larger photo diode: Higher senitivity than IT CCDs

2) Two layer polysilicon electrodes: the same as the interleave CCD or easier than the progressive CCD of three layers

3) No mechnical shutter required for DSC use.

2. Merit of Octagonal Photo Diode

The lens aperture usually has a shape close to a circle, while the photo diode is rectangle. When the aperture is wide in a dark environment, the focus point is not exactly on the photo diode through on-chip-lens and some part of the incoming light is out of the rectangle photo sensing area. Therefore, the sensitivity is reduced. This type of light out of the photo diode is called “kicked out light” in Japanese. If the photo diode shape is closer to a circle like octagonal, the kicked out light can be smaller than the rectangle's one. Fig.2 shows this phenomenon.

Fig. 2 Kicked Out Light

3. Spatioal Frequency Response and MTF

Fig3. Shows the spatial frequency responses of Supper CCD and the square pixel arrangement CCD. The later is not shown in the paper but is added by the reporter.

Fig.3 Spatial Frequecy Responses

Two advantagies of the diagonal pixel arrnangement are explained in the paper. One is that the objects in the real world has higher frequencies in vertical and horizontal directions as shown in Fig.4, which are the power spectrum in the frequency domain of 500 images including natural scenes. The other is that the human eyes have higher sensitivities in the frequencies of vertical and horizontal directions than the 45 and 135 degree directions as shown in Fig.5.

Based on their analysis, the diagonal shape frequency response meets the nature of the image scene and the human visual system. Only the weak point is green’s less Nyquist frequency in vertical and horizontal, conpared with the square arrangement of Bayer.

4. Resolution

It was mentioned by the presentator that the vertical and horizontal resolutions were increased by 1.4 times. This is understandable, when seeing Fig.3. But it was not 1.6 times as announced in the news papers and Fuji’s Web page, and the magnitude of 1.6 overpassed the Nyquist limit increment of 1.4. This is logically irony.

5. Dynamic Range

The paper was written in the camera system stand point and no information about the device itslef in the presentaion and digest paper. However, the worlds, “twice dynamic range could be achievable with 1.3M Supper CCD, compared with 2.0M conventional CCD”, were described in the paper as a conclusion. This was also announced in the news and Web page, but no technical evidence was not explained. We have to wait untill the device side technical presentation in ISSCC, February 2000.

Conclusions as This Report

In the conclusion of the paper, it is written that1.3M pixel Supper CCD would have the equivalent resolution as of 2.0M pixel conventional CCD, that clearly implied 1.6 times higher resolution. I believe the magnetude of 1.6 experimentally observed in the actual comparison of the conventional CCD and Supper CCD using a resolution chart.

From the camera system side of view, no major demerits were explained and were seemed to be exsited in the paper. However, no information about the structure inside of the device. I am supposing that the channel stopper which separating the pixel inside the device should be zig-zag and the overflow drain should have a vertical structure. We are not sure the conbination of this kind of channel stopper with vertical overflow drain could sufficiently work.

Two issues are predicted from the structure; one is a blueming/smear problem and the other is a manufacturing yealding. The further device side technical information is being waited.

End of Report

TERLのTopへ。

　Nori-sanのHPへ。