A stacked sensor is a CMOS image sensor built from two or more bonded silicon layers, with the light-sensitive pixel layer on top and the readout, processing and sometimes memory circuits on separate layers beneath it. In photography, stacking matters mainly because it allows much faster sensor readout, which reduces rolling shutter distortion and enables high burst rates, blackout-free viewfinders and demanding video modes.
A conventional sensor places pixels and their supporting circuitry side by side on one chip, so space for circuits competes with space for pixels. Stacking moves the circuits underneath, where they can be larger, faster and manufactured in a different process optimized for logic rather than light capture.
How a stacked sensor is built
Stacked sensors are almost always back-side illuminated (BSI). The pixel wafer is thinned so light enters from the back, and its front side, where the wiring sits, is bonded face to face to a logic wafer. Connections between the layers are made with through-silicon vias or, in newer designs, direct copper-to-copper hybrid bonding at very fine pitch.
The lower layer carries the analog-to-digital converters (ADCs), timing logic and signal processing. Because it has its own area, designers can place many more ADCs working in parallel, which is the key to fast readout. Some sensors add a third layer of DRAM that buffers full frames almost instantly before they leave the sensor.
| Layer | Contents | Function |
|---|---|---|
| Pixel layer | Photodiodes, color filters, microlenses | Converts light into charge |
| Logic layer | ADCs, timing, signal processing | Reads and digitizes pixels quickly |
| Memory layer | DRAM in three-layer designs | Buffers frames at very high speed |
| Split pixel layers | Photodiodes and pixel transistors separated | Increases charge capacity in some designs |
The term covers designs with different degrees of stacking. Fully stacked sensors place the readout circuitry for the whole array underneath the pixels, while partially stacked sensors place fast circuits only along parts of the array, a compromise between speed and cost.
Why readout speed matters
Most image sensors use a rolling readout: rows are read one after another. If the readout of a full frame takes long, fast motion appears skewed, flash cannot sync with the electronic shutter and flickering lights cause bands. Readout time is the main specification that stacking improves.
| Sensor type | Typical readout time | Practical consequence |
|---|---|---|
| Conventional front-side or BSI | About 1/15-1/30 s | Visible skew with electronic shutter |
| Partially stacked | About 1/60-1/100 s | Usable electronic shutter for most action |
| Fully stacked | About 1/120-1/270 s | Electronic shutter replaces mechanical for sport |
| Global shutter | All pixels exposed at once | No rolling distortion, flash sync at any speed |
Fast readout also feeds autofocus and the viewfinder. A stacked sensor can deliver tens to over a hundred readouts per second for subject detection and phase-detection autofocus, keep the electronic viewfinder live during bursts and support high-resolution video with less rolling shutter wobble.
Stacked sensors compared with other sensor designs
Stacking is one step in a progression of CMOS designs. The table places it among its neighbors in the camera and sensor field.
| Design | Key idea | Main benefit |
|---|---|---|
| Front-side illuminated CMOS | Wiring above the photodiodes | Simple and inexpensive |
| BSI sensor | Light enters from the back, wiring below | Better light collection, especially small pixels |
| Stacked sensor | Separate bonded logic layer | Much faster readout and processing |
| Global shutter CMOS | Simultaneous exposure of all pixels | No rolling shutter effects |
| CCD | Charge shifted across the chip to one output | Historically uniform output, now largely obsolete |
Stacked and BSI are not alternatives but complementary steps: stacking almost requires BSI construction. Likewise, a global shutter sensor may itself be stacked, because storing and reading every pixel at once benefits from extra circuitry underneath.
Where stacked sensors are used
Stacked sensors first appeared in smartphones, where tiny pixels leave little room for circuits and on-sensor processing supports multi-frame computational photography. They then moved into compact cameras and professional mirrorless bodies.
- Sports and wildlife photography: silent electronic bursts of 20 to 120 frames per second with minimal distortion.
- Flagship mirrorless cameras: fast tracking autofocus and blackout-free viewing during continuous shooting.
- High-resolution video: 4K and 8K recording with reduced rolling shutter skew.
- Smartphones: rapid multi-frame capture for HDR and night modes in computational photography.
- Industrial and machine vision: high frame rates for inspection and measurement.
Trade-offs and misconceptions
Stacking improves speed but does not automatically improve every aspect of image quality. Cost, heat and dynamic range are the main trade-offs.
| Misconception | Reality | Practical takeaway |
|---|---|---|
| Stacked means better image quality | Speed improves; base ISO quality may not | Choose stacking for action, not for detail alone |
| Stacked equals global shutter | Most stacked sensors still read rows sequentially | Fast skew is reduced, not eliminated |
| All stacked sensors are equal | Readout speeds vary widely between models | Compare measured readout times |
| Stacking has no downsides | Higher cost, power and heat | Expect higher prices and battery use |
Some stacked full-frame sensors show slightly lower dynamic range at base ISO than slower sensors of similar generation, partly because fast parallel readout favors speed over the lowest possible read noise. For landscape or studio work, a slower, high-resolution conventional sensor can still be the better tool.
Explained without jargon
A camera sensor has two jobs: catching light and reading out what it caught. In an ordinary sensor, both jobs share one flat chip. A stacked sensor builds the chip in floors, with the light-catching part on top and the electronics underneath.
Think of a shop that moves its stockroom to the basement so the sales floor can be bigger and the staff can work faster. The result is a sensor that reads its picture very quickly, so fast-moving subjects look straight instead of bent, and the camera can shoot many photos every second.
A typical use case
A motorsport photographer covers a race from the pit wall, where cars pass at over 250 km/h. With a conventional sensor, the silent electronic shutter skews the cars and wheels into leaning shapes, so the photographer has relied on the noisy, wear-prone mechanical shutter at 10 frames per second.
After switching to a body with a fully stacked full-frame sensor, the photographer shoots with the electronic shutter at 30 frames per second and 1/2000 s. The readout of roughly 1/250 s keeps car bodies and wheels geometrically correct, and the viewfinder never blacks out, which makes panning easier to follow.
The photographer gains more frames at the decisive moment, eliminates shutter wear, and works silently in the media zone. The trade-off is a more expensive camera and faster battery use, which is accepted because speed is the deciding factor in this kind of work.