01 Structure of the Basic Sensor (Front-Illuminated) This type of sensor is commonly used in older devices, with its main advantages being simple manufacturing processes and low cost.

The photosensitive layer of the sensor is used to receive light, convert photons into electrons, and finally convert them into electrical signals (light intensity information) through processing such as readout circuits. In old-fashioned sensors, after passing through the microlens and the color filter, the incident light still needs to traverse the metal wiring layer before reaching the light-receiving surface (photodiode), as shown in the figure below. When light passes through the metal wiring layer, it may be blocked or scattered. Meanwhile, since light can only reach the photodiode through the paths reserved in the metal wiring layer, the sensor's microlenses require more compensation to achieve this. This results in the sensor being less friendly to CRA, also causing more severe vignetting and color shift at the edges, as well as relatively poor imaging performance in low-light conditions. For a long time, there were two optimization schemes for front-illuminated sensors: One is to use different materials to reduce the thickness of the metal wiring layer, avoiding light loss when incident light passes through the layer:

The other method is to increase the depth of the sensor's well to optimize its performance in low-light conditions.

02 Back-Illuminated Sensor

The emergence of the Back Illuminated CMOS Sensor (BSI) has brought about tremendous technological changes. It is characterized by swapping the positions of the metal wiring layer and the photosensitive layer, allowing light to directly enter the photodiode, with the metal wiring layer moved to the underside. Let's take a look at the schematic diagram from Sony:

With no obstruction from the wiring layer, light can directly enter the photodiode, greatly avoiding light loss before it reaches the photosensitive layer, improving light collection efficiency, and further enhancing imaging quality and low-light performance. 03 Stacked Sensor Finally, let's talk about the future development direction—the stacked sensor. From the schematic diagram alone, there seem to be no obvious optimization points, except that the pixel layer and the logic circuit layer are separated and stacked vertically.

In fact, the separate stacking brings enormous advantages. As can be seen from the schematic diagram below, the stacked sensor completely separates the photosensitive pixel layer from the signal processing layer through 3D stacking.

The photosensitive pixel layer only consists of photodiodes, microlenses and color filters, and this layer is dedicated purely to optimizing light sensing. Thanks to the stacked design, the circuit layer is no longer constrained by pixel space, enabling extensive integrated design. It can incorporate a large number of readout circuits, such as abundant parallel ADCs and even DRAM cache.

# English Translation A major advantage of these caches and parallel ADCs is that they can drastically reduce readout time (or read data after storing it in DRAM simultaneously), avoiding the rolling shutter effect caused by row-by-row readout. Of course, higher integration brings more benefits than that. High-speed readout also indirectly improves the sensor's dynamic range, reduces noise, and even supports more sophisticated on-chip computational photography. In summary, stacked sensors free the readout circuits from the pixel plane, offering greater design space and process flexibility, enabling more powerful functions and higher parallelism.

04 Summary Let's conclude briefly. The traditional front-illuminated sensor incurs light loss before reaching the photodiode due to the wiring layer. Back-illuminated sensor features flipped structure, with priority given to light reception as its advantage. Stacked sensor realizes layer separation, leaving ample room for circuit design. Enhanced integration facilitates parallel processing and mitigates the rolling shutter effect.