A 3.4 μm pixel pitch global shutter cmos image sensor with dual in-pixel charge domain memory – iopscience gas kansas city


To realize the said high-performance GS CIS, while several methods have been proposed, we adopted and reported a dual in-pixel charge domain memories (CDMEMs) type GS CIS 1) and an in-pixel CDMEM electricity sources uk type GS CIS 2– 5) in previous research. The CDMEM type GS CISs are widely used, and various proposals have been reported. 1– 13) Figure 3 summarizes the comparison with other type GS CISs. The CDMEM type GS is advantageous in low kTC noise compared to a floating diffusion (FD) based type GS, 14) since the correlated double sampling operation can be performed. In addition, in high FWC and sensitivity compared to a voltage domain memory (VDMEM) type GS, 15– 17) since there are fewer pixel components. Furthermore, in high-definition and low cost structure compared to a pixel-level multi-layer stacked type, 17, 18) because of simple circuit structure. On the other hand, the CDMEM type is disadvantageous in low FWC and low sensitivity compared to the FD based type and the pixel-level multi-layer stacked type, since there are more pixel components in the same layer with CDMEM. In addition, in high parasitic light sensitivity (PLS) compared to the VDMEM type and the pixel-level multi-layer stacked type, since the CDMEM is close to a photodiode (PD) and aperture electricity experiments for high school of the light shield (LS) structure. Therefore, we adopt a light guide (LG) structure 2– 4) and devised the dual CDMEMs layout (e.g. symmetrical oblique transfer structure) to eliminate disadvantages. In addition, in order to improve total image quality, to improve saturation is desirable as well as sensitivity improvement. Therefore, we adopt an already reported multiple-accumulation shutter (M-ACS) technique. 2– 5)

In Fig. 4(a), first, during the period t1– t2 is the PD accumulation period. The Nth frame signal charges are accumulated in the list of electricity usage by appliances respective PD(N). Next, at time t2, the signal charges are globally transferred to the in-pixel memory MEM with the pulse GS for turning on the GS charge transfer transistors of all the pixels. Thereafter, during the period t2 to t3 is the MEM holding and readout period. The signal charges are sequentially transferred and readout from the MEMs of respective rows to the FDs and the column signal lines [MEM(N)]. In Fig. 4(a), the charge accumulation electricity measurements units period and the charge holding period of the same frame are represented by the same suffix and same color (online) [e.g. PD(N) and MEM(N) are represented by same suffix N and same blue color (online)]. This manner also applies to the following similar timing diagram figures. In the period MEM(N), at the same time, the ( N + 1)th frame signal charges are accumulated in the respective PDs [PD( N + 1)].

As another example of constrains, we describe the synthesis of multiple images in Fig. 2. The image synthesis of multiple frames acquired with different accumulation times is often used as a method to expand the SN ratio and dynamic range of images. 19– 30) Figure 5 shows a conventional multiple-frame HDR readout timing diagram. As described in Fig. 5, images of different accumulation times (e.g. short accumulation period and long accumulation period) are obtained as images of discontinuous and different frames. Therefore, a time difference occurs in a plurality of images, and when a moving object is photographed, a jerkiness image or a double image occurs as shown in Fig. 2(a) or 2(b).

Figure 6(a) shows a top view diagram and Fig. 6(b) shows a cross section diagram of our proposed two CDMEMs GS CIS. The two CDMEMs are overall covered with LS structure to avoid holding signal charge changing due to the influence of parasitic light. In order to ensure gas problem in babies the sensitivity, the LS aperture size for the PD is not changed as in the case of one CDMEM GS CIS. 2) Furthermore, in order to suppress the variation of two CDMEMs, CDMEMs (MEM), the GS transistors, and the rolling shutter charge transfer (TX) structures for FD are arranged symmetrically for PD. By placing CDMEMs in the diagonal direction of PD, FD and TXs can be placed between CDMEMs and the FD capacitance also can be reduced. As a result, pixel readout noise can be reduced. To guide p gaskell the incident light for the PD, we adopted a high refractive index material based LG structure. In addition, for small LS aperture for the PD, we adopted the low height structure based on the three copper wiring layers, the LG structure have a gentle slope, 2– 4) and the dual lens structure (the inner-lens and the top-microlens. 2– 4, 10, 11) The top and bottom diameter of the gentle slope LG structure are 2.4 μm and 1.1 μm, and the LS aperture size for the PD is 1.3 μm square, respectively.

Figure 11 shows captured images by conventional single-accumulation HDR procedure with one CDMEM and gas national average 2009 Fig. 12 shows captured images by suggested alternately multiple-accumulation HDR procedure with dual CDMEMs. In Fig. 12, the short exposure is divided in 5 periods and the long exposure is divided in 4 periods, and are performed alternately. In Figs. 11 and 12, total exposure time is identical. The short exposure is 115 μs (23 μs × 5) and long exposure is 14.7 ms (3.68 ms × 4, 115 μs × 128). In particular, in the short exposure image, the synthesis result of moving subjects are unnatural in Fig. 11. As an algorithm of HDR synthesis, the short exposure data is cited for the saturated part data with long exposure. Since the corresponding part with short exposure, however, a mismatch occurs. On the other hand, the synthesis result of moving subjects are natural by scattered short exposure in Fig. 12. According to the calculation from the ratio of accumulation period (128 times), the dynamic range improvement is about 42 dB. On the other hand, we allocated about 8 dB overlap for the stitching point to be hardly noticeable. As a result, the gasbuddy near me dynamic range increases from about 77 dB in high saturation binning readout procedure with the multiple-accumulation to 111 dB in alternately multiple-accumulation HDR readout procedure.