New CMOS-matrix expands the possibilities of shooting moving objects

This translation is an adaptation of an article published by Canon engineers in Japanese Journal of Applied Physics Japanese Journal of Applied Physic .

The use of photosensitive matrices in phototechnics made it possible to move away from the use of a mechanical shutter and its variations. This had a positive effect: the absence of vibrations at the time of the shutter release and the ability to significantly increase the speed of shooting, without complicating the design. But the transfer of photographic equipment to a new level brought new problems that are associated with high-speed shooting.


')
To understand the essence of the difficulties, it is necessary to disassemble the principle of operation of photosensitive matrices. Speaking of them in the plural, we mean the matrix, made by different technologies. There are similarities and fundamental differences in their work. Let's start with the general features. Any photosensitive array consists of a set of photodiodes that convert the luminous flux incident on them into an electrical signal. The difference lies in the way the signals are accumulated and read: the exposure of the image is not determined by the time the shutter opens, but by the time between zeroing the charge of the matrix and the moment the information is read from it.

In the CCD matrix, the signal is read line by line, and such a shutter is called a running or rolling shutter. During line-by-line reading, a fast-moving object has time to change its position, so distortions appear in the image. And the greater the speed of the object, the greater the distortion in the picture.

Partially this problem is solved in CMOS-matrices, which have recently become an alternative to CCD-matrices. Here, signal reading is possible from any fragment of the matrix and in any order. This not only increases the speed of data exchange, but also allows you to get random access to individual pixels.

In fact, the CMOS-matrix is ​​an integrated microcircuit, where each pixel forms a separate cell and has its own harness that converts the charge of the photodiode into voltage directly in the pixel itself. In general, the cell consists of:

• photodiode;
• electronic shutter;
• a capacitor accumulating charge from the photodiode;
• signal booster;
• bus read lines;
• bus signal transmission processor;
• reset signal lines.

During shooting, an image is formed by synthesizing several frames. On the one hand, this gives depth and saturation to the image, but on the other hand, when you shake or shoot moving objects, the image quality decreases. This is expressed in blurring, “double” image or the effect of a running shutter. The reason is the alternation of the processes of exposure and reading. We take conditionally during the exposure time t. Then at time t, the first frame is captured. In the period t + t, the data of this frame is read. Then, after resetting the matrix, the next frame is executed. Thus, the gap between frames is t. This situation is similar to the rolling-gate algorithm.

One solution to this problem was proposed by our developers, and it was as follows. In a conventional CMOS cell, a single capacitor with a strapping is used that performs the function of a memory element, so at any time during the shooting the cell is either in the state of charge of this capacitor (exposure) or discharge (reading). In the cell of our development, two memory elements are used. Due to this, two processes can occur simultaneously. After shooting the first frame, while reading data from one memory element, the next frame is immediately exposed to the second memory element. This ensures continuous recording and image stability.

However, the meaning of this invention is not limited to the continuity of shooting. In fact, we got several different modes of operation of the CMOS-sensor. It all depends on the pixel reading procedure.

• When reading with a high frame rate, the pixel saturation can occur due to either multiple saturation of the photodiode, or one-time saturation of the memory element. At the same time, image clarity is combined with saturation.
• In the high saturation shooting mode, the two storage elements are simultaneously filled and read. This reduces the frequency of reading, which as a bonus gives a reduction in total power consumption.

The possibility of multiple accumulation is used when performing a series of exposures, for example, when alternating short and long. At the same time, the storage elements alternate: a signal of short exposures accumulates on one, and long exposures on the other. When compared with a CMOS matrix with one storage element and a total exposure equal to a series of 5 short and 4 long exposures, the improvement in dynamic range is about 42 dB.

Increasing the pixel piping detail results in increased spurious noise. To reduce its effect, the cell elements are located diagonally symmetrically relative to the photodiode. From the influence of the light flux, they are protected by a light screen. Only a photodiode left an aperture of 1.3 microns. Focusing of the light incident on the photodiode is carried out using a double lens unit and a light guide. In the block between the lenses is a color filter in accordance with the Bayer pattern. For the fiber used material with a high refractive index. Due to this, the fiber in the form of an inverted cone has a small height corresponding to three layers of copper wiring. The upper diameter of the fiber is 2.4 μm, and the lower diameter is 1.1 μm.

A single pixel of the matrix, according to the Bayer pattern, consists of a pair of pixels with double memory cells. A single pixel block includes:

• 2 photodiode;
• 4 storage elements (capacitor);
• 13 transistors.

The total size of the matrix is ​​2676 N × 2200 V, which is almost 5.9 megapixels.
The comparative table shows the characteristics of the various read modes of the developed CMOS-matrix with dual intra-pixel memory and the usual matrix, which has comparable indicators.


In fact, the developed CMOS image sensor with a pixel pitch of 3.4 μm with dual intra-pixel memory has about 5.3 effective megapixels and a dynamic range of more than 110 dB when exposed in one frame with alternate multiple accumulations. This mode is especially suitable for shooting moving objects and can be used in movie cameras, machine vision devices, automobiles, aerial photography and in surveillance cameras.