CMOS, Stacked and Back-Illuminated Sensors: Understanding the Architecture

CMOS, back-illuminated (BSI) and stacked describe different aspects of image-sensor design. They are not three mutually exclusive sensor categories. CMOS describes the sensor technology; BSI describes how the photosensitive layer is arranged relative to wiring; stacked describes how pixel and logic circuitry are arranged across layers. A sensor can therefore be both BSI and stacked.

For a deeper look at the pixel and circuit arrangements, see our guide to digital sensor architecture.

CMOS is the underlying sensor technology

CMOS image sensors integrate photodiodes and readout circuitry using complementary metal-oxide-semiconductor processes. The architecture can include amplifiers, row and column selection, analogue-to-digital conversion and other functions. The exact arrangement varies by sensor generation and manufacturer.

Back-illuminated structure

In a front-illuminated design, wiring and circuit structures can obstruct some incoming light. A BSI structure places the photosensitive side so that light reaches the photodiode with less obstruction from front-side wiring. This can improve light collection, particularly for small pixels, but actual performance depends on photodiode design, microlenses, colour filters, process quality, noise and other factors. Sony describes both the light-collection advantage and the manufacturing challenges associated with BSI structures. See Sony’s BSI technology overview.

Stacked architecture

A stacked sensor separates the pixel section from some of the signal-processing circuitry and places the functional layers together. This can allow each layer to use a process suited to its role and can support additional logic or memory close to the pixel array. Stacking does not automatically mean BSI, although modern designs may combine both. Sony’s stacked-structure explanation describes the separation of pixel and circuit sections.

One practical implementation is described in our article on stacked sensor technology.

Compare the right attributes

Attribute Question to ask
CMOS process What readout and conversion functions are integrated?
Illumination structure Is the sensor front-illuminated or back-illuminated?
Layer arrangement Are pixel, logic or memory layers stacked?
Readout What are the measured readout times and rolling-shutter behaviour?
Image quality What are the measured quantum efficiency, read noise, dark current and dynamic range under stated conditions?
System cost What are the sensor, yield, packaging, power and downstream processing costs?

When does the architecture justify a price premium?

Do not infer a price premium from the architecture label alone. Compare exact sensors or camera systems using measured readout speed, noise, dynamic range, power consumption, frame rate and cost. Use the same test conditions and distinguish manufacturer specifications from independent measurements. A stacked design may be valuable for high-speed readout; BSI may help light collection; the system may combine both. The appropriate choice depends on the application.

For the effect of shutter architecture on motion capture, see our guide to global shutter cameras.

Conclusion

CMOS, BSI and stacked describe related but different design dimensions. BSI and stacked are not competing upgrades, and both can be used in the same sensor. Procurement decisions should be based on the complete sensor specification and measured system performance, not a simplified three-way label.