A stacked CMOS image sensor is a three-dimensional architecture that places the pixel array and logic circuitry on separate semiconductor substrates, then connects them vertically. The approach addresses a design challenge familiar to imaging-system developers: improving pixel performance while also advancing the processing, speed, and power characteristics of the logic that supports the sensor.
First developed for mobile-phone cameras, the architecture followed the adoption of back-illuminated sensors. Its significance extends beyond packaging: separating pixels from logic allows each part of the device to be developed using processes suited to its role, while vertical integration brings them together in one compact sensor assembly.
Why back-illumination was an important step
In a conventional front-illuminated sensor, metal wiring can obstruct some of the light approaching the pixels. A back-illuminated design receives light through the back side of the silicon, avoiding that obstruction. This helps small pixels capture light more efficiently and supports improved sensitivity and signal-to-noise performance, including clearer images in dark conditions.
Back-illumination introduces a mechanical consideration: the silicon must be thinned so light can enter from the rear. A supporting substrate is therefore used to provide mechanical strength. As imaging requirements expanded, however, the supporting structure presented an opportunity for a different architecture—one that could place a logic circuit substrate beneath the pixel chip.
How a stacked CMOS image sensor works
Rather than forming pixels and logic on the same silicon substrate, the stacked design optimizes them on separate substrates. The pixel chip sits above the logic circuit substrate, replacing the supporting substrate used in the back-illuminated arrangement. Through-silicon vias (TSVs) provide multiple electrical connections between the upper and lower chips.
These vertical connections allow signals and functions to pass between the pixel and logic portions of the device. The result is a three-dimensional structure in which the sensor’s light-capturing elements and its processing circuitry remain physically integrated, but can be developed with different process priorities.
What separate pixel and logic processes enable
Pixel and logic circuits have different design objectives. The pixel portion must capture light and preserve image quality. The logic portion supports signal processing and camera functions, and may benefit from advances in circuit integration and semiconductor processes. Developing both on one substrate can make it difficult to pursue these priorities independently.
Stacking separates those development paths. The source identifies three main advantages:
- Independent process optimization: Pixel and logic circuits can each be developed for their respective requirements, supporting high overall sensor performance.
- More integrated logic: Increased circuit integration on the logic chip can support new camera functions within the sensor assembly.
- Smaller chip size: The design was reported to reduce chip size by 30 to 40 percent compared with the earlier approach, supporting expansion of production volume.
These advantages are connected. Separating the circuits creates room to improve the logic without requiring the pixel technology to follow the same process path. At the same time, vertical integration keeps the two parts together in a compact device.
Imaging functions made possible by logic integration
The added logic capabilities supported functions that go beyond capturing an image. One example is high dynamic range (HDR) movie capture. HDR helps produce vivid colors in backlit scenes, where strong differences in brightness can make image detail difficult to preserve. The source contrasts video captured with and without this function to illustrate its effect.
Another example is image-plane phase-difference autofocus signal processing. Integrated processing for this function supports focus tracking. These capabilities show how sensor architecture can influence the features available to camera systems: the pixel array captures the scene, while logic integration enables additional processing within the sensor.
Manufacturing and system implications
For technical decision-makers, the development illustrates a broader semiconductor design principle: when two functions have different process needs, separating them physically can allow more targeted optimization. TSVs then provide the electrical links needed to combine the separate chips into a single three-dimensional device.
The reported chip-size reduction also had a production-scale implication. The stacked sensor was described as a product enabling mass production for consumer applications in the three-dimensional stacked semiconductor field. This connects device architecture not only with imaging performance and functionality, but also with the ability to produce the design at consumer-product volumes.
The approach does not eliminate the need to manage mechanical and electrical integration. It changes how those concerns are addressed: instead of using a support substrate beneath a back-illuminated pixel chip, the architecture uses a logic substrate and vertical electrical connections. That shift makes the relationship between packaging, interconnects, process choices, and sensor functions central to the design.
Conclusion: a platform for combining pixels and processing
Stacked CMOS image sensors evolved from back-illuminated designs by adding a separately optimized logic chip beneath the pixel chip. TSVs connect the two, enabling a compact structure that can combine pixel performance with more integrated processing. The resulting architecture supported functions such as HDR movie capture and focus-tracking signal processing, while also reducing chip size relative to the earlier approach. For automation and imaging-system professionals, the key lesson is how three-dimensional integration can bring distinct semiconductor functions together without requiring them to share the same process priorities.