Camera Sensor Designs Explained: FSI, BSI, and Stacked CMOS

First: every digital camera sensor is a system

An image sensor contains millions of light-sensitive sites, commonly called pixels in camera specifications. During an exposure, each site converts arriving photons into an electrical signal. That signal must then be transferred, amplified, converted into digital data, read from the chip, and processed by the camera.

The sensor's physical layout affects how efficiently it gathers light and how quickly it can move data. It does not work alone. Microlenses, color filters, pixel size, readout circuits, analog-to-digital converters, processing, cooling, and firmware all influence the result.

FSI, BSI, and stacked therefore answer a narrow question: how is the sensor built? They do not answer the entire question: how good is the camera?

What is an FSI sensor?

Front-side illuminated, shortened to FSI, is the conventional CMOS structure. Incoming light reaches the front of the silicon, where wiring and transistor structures share space above or around the photodiode.

Modern FSI designs use microlenses to direct light toward the active area, so the wiring does not simply make the sensor unusable. Large pixels can still collect light effectively, and many excellent cameras have used front-illuminated sensors.

The challenge grows as pixels become smaller. More of each pixel's limited area is occupied by electronics, and light arriving at an angle can be harder to guide into the photodiode. This encouraged manufacturers to turn the structure around.

What is a BSI sensor?

Back-side illuminated, or BSI, does not mean a light shines from behind the camera. It means the sensor wafer is built and thinned so incoming light enters from the side opposite much of the wiring. The photodiode is closer to the light-facing surface.

With fewer structures obstructing the light path, a BSI design can increase the percentage of incoming light that reaches the photodiode. Engineers call this quantum efficiency: the effectiveness with which arriving photons become a useful electrical signal.

BSI can also accept light over a wider range of incident angles. That is valuable near the corners of a sensor, where rays from a lens may arrive obliquely, and it becomes increasingly important when pixels are very small.

What BSI can improve

Depending on the implementation, BSI can contribute to:

• higher light-collection efficiency;
• better sensitivity from small pixels;
• lower noise for a comparable exposure;
• more freedom in the wiring layout; and
• better performance with light arriving at steep angles.

These are engineering opportunities, not guaranteed camera-level outcomes. A manufacturer might spend the efficiency gain on smaller pixels and higher resolution rather than visibly cleaner high-ISO files. Another design may prioritize speed or compactness.

Why BSI does not automatically win

Noise comes from several sources. Photon shot noise is inherent in the arrival of light; read noise arises in the electronics; dark current grows with time and temperature; and downstream processing changes how noise looks. BSI mainly improves the light path. It does not erase every source of noise.

Sensor size and exposure also remain important. A photograph made with a larger entrance pupil or longer exposure can collect more light than one made with a smaller amount of light, regardless of the BSI badge.

Likewise, dynamic range depends on both highlight capacity and the usable shadow floor. BSI may help, but pixel capacity, readout mode, ISO, bit depth, and processing also matter.

What is a stacked CMOS sensor?

A stacked CMOS sensor divides functions across bonded chip layers. In a typical arrangement, the pixel section sits on one layer and a logic layer containing signal-processing circuitry sits beneath it. Some designs also incorporate fast memory.

A conventional BSI sensor still keeps the pixel and circuit sections within a more limited shared structure. Stacking lets engineers optimize the pixel and logic sections separately and gives the processing circuitry more area without making the sensor's face larger.

A stacked sensor is commonly back-illuminated, but the words describe different ideas: BSI describes which side receives light; stacked describes how functional layers are combined.

Why stacked sensors can read faster

A camera does not normally download the entire sensor at one instant. Most CMOS cameras scan rows sequentially. If the scan takes too long, a moving subject changes position between the first and last rows.

Extra high-speed circuitry, more parallel data paths, and on-sensor memory can shorten that readout. A faster sensor can enable:

• higher continuous-shooting rates;
• more autofocus and exposure calculations per second;
• less viewfinder blackout or interruption;
• higher-frame-rate video;
• faster electronic-shutter operation; and
• less rolling-shutter distortion.

The word can matters. Two stacked sensors can have different resolutions, readout times, memory, bit depths, and thermal limits.

Rolling shutter is a readout effect

With a typical electronic shutter, rows begin and end exposure at slightly different times. A fast-moving vertical object can lean, a rotating propeller can bend, and a quick pan can skew the whole scene. Flickering LEDs may create bands because different rows sample different phases of the light.

A faster stacked readout reduces the time difference and therefore the distortion. It does not necessarily eliminate it. A true global shutter exposes all pixels for the same time interval, but global-shutter sensors have their own design tradeoffs and should not be confused with every stacked sensor.

Mechanical shutter versus electronic shutter

A mechanical focal-plane shutter also exposes the frame with a traveling slit at fast shutter speeds, but the curtains usually cross the sensor much faster than a conventional electronic readout. A fast stacked sensor can make the electronic option practical for subjects that would distort on a slower sensor.

Electronic shutter can be silent and vibration-free, but flash compatibility, flicker behavior, RAW bit depth, dynamic range, burst speed, and noise can change by camera and mode. Check the manual instead of assuming the fastest advertised burst preserves every other feature.

For the mechanical parts and usage count, see Camera Shutter Count Explained.

Partially stacked and two-layer pixel designs

Sensor terminology continues to evolve. A manufacturer may move only certain circuits to a separate layer, leading reviewers to use terms such as partially stacked. Newer two-layer transistor-pixel structures divide photodiodes and pixel transistors within the pixel chip itself, in addition to the underlying logic layer.

These designs can create more room for the photodiode or transistor and improve saturation capacity, noise, speed, or cost in different combinations. Marketing names are not always standardized across companies, so the measured behavior matters more than the label.

Does a stacked sensor improve still-image quality?

Its clearest advantage is usually speed, not a universal increase in base-ISO image quality. Faster readout can nevertheless improve the photographs you can successfully make: autofocus receives more frequent data, electronic-shutter distortion is lower, and a fast burst may capture the decisive moment.

There can also be tradeoffs. More circuitry and high data rates create heat and power demands. In some designs or operating modes, the fastest readout uses lower bit depth or other compromises. A non-stacked sensor may deliver comparable or better tonal quality for slow, controlled subjects at a lower price.

Which photographers benefit most?

Sports, wildlife, and action

Fast readout, frequent autofocus updates, and blackout-free bursts can materially improve tracking and timing. This is where a stacked sensor often earns its cost.

Event and performance photography

A quiet electronic shutter can be valuable, but photographers must test stage LEDs and venue lighting for bands. Speed helps; it does not make flicker disappear.

Video

Reduced skew during pans and subject movement is useful. Also check resolution, oversampling, crop, overheating limits, recording formats, and stabilization.

Landscape, architecture, and studio

When subjects are controlled or stationary, maximum burst speed and electronic readout may matter less. Resolution, lens choice, base-ISO dynamic range, tethering, and price can be stronger priorities.

How to compare cameras without getting lost in labels

1. Identify the problem. Do you need silent shooting, lower rolling shutter, cleaner shadows, more resolution, or a less expensive body?
2. Check the exact mode. Readout can change between full-resolution stills, compressed RAW, electronic shutter, mechanical shutter, and video resolutions.
3. Look for measured readout time. A sensor scan measured in milliseconds is more informative for distortion than the word “stacked.”
4. Compare normalized images. Judge noise and detail at the same output size, not only at 100% pixel view.
5. Read the restrictions. Verify flash, anti-flicker, bit depth, autofocus, and lens compatibility at the desired burst rate.
6. Evaluate the complete system. Grip, viewfinder, battery, lenses, storage, and workflow affect every assignment.

Common misconceptions

• “Back-illuminated” means the sensor has a backlight. It describes the direction from which light enters the silicon structure.
• BSI always has less noise. Exposure, pixel design, read electronics, temperature, and processing also determine noise.
• Every stacked sensor is a global shutter. Most still use a rolling electronic readout, only faster.
• Stacking automatically adds dynamic range. It creates design freedom; the final capacity depends on the implementation and mode.
• A high burst number proves fast full-sensor readout. The burst may use a crop, reduced bit depth, JPEG, or pre-capture mode.

Sources and further reading

The physical distinction between back-illuminated and stacked structures was checked against Sony Semiconductor Solutions' official image-sensor technology overview and stacked-structure reference. Sony's two-layer transistor pixel announcement was used for the newer layered design. Electronic-shutter behavior was cross-checked with Sony's official electronic front-curtain explanation.