CMOS, short for complementary metal-oxide-semiconductor, is the image sensor technology used in nearly all current digital cameras, in which each pixel converts light into charge and contains its own amplifying transistors, and the signal is digitized on the same chip. In photography the term usually refers to the active pixel sensor built with this manufacturing process, as opposed to the older CCD design.
CMOS sensors replaced CCDs in most cameras because they combine low power consumption, fast readout and on-chip processing with image quality that now matches or exceeds the older technology. Variants such as BSI and stacked sensors are refinements of the same basic architecture.
How a CMOS image sensor records light
Every pixel of a CMOS sensor contains a photodiode that converts incoming photons into electrons through the photoelectric effect. During the exposure the electrons accumulate in the diode, and the amount of collected charge is proportional to the light that reached that pixel. A color filter array, usually the Bayer pattern of red, green and blue filters, makes each pixel sensitive to one color band.
In the common four-transistor pixel, a transfer gate moves the charge from a pinned photodiode to a small floating diffusion node, where it becomes a voltage. A source follower amplifier inside the pixel buffers that voltage, and a row select transistor connects it to a column line. The sensor reads the reset level and the signal level and subtracts them, a process called correlated double sampling that removes much of the reset noise.
The column voltages then pass to analog-to-digital converters, typically one or more per column, which produce 12-bit or 14-bit values. Gain applied before conversion is what the camera uses to implement much of its ISO setting.
| Stage | Component | Function |
|---|---|---|
| Light collection | Microlens and color filter | Directs light and selects a color band |
| Conversion | Photodiode | Turns photons into stored electrons |
| Charge to voltage | Transfer gate and floating diffusion | Converts charge into a measurable voltage |
| Buffering | In-pixel source follower | Drives the column line |
| Digitization | Column-parallel converters | Produces 12-bit or 14-bit values |
Key performance parameters
The image quality of a CMOS sensor depends on a handful of measurable properties. Quantum efficiency describes the share of photons that generate electrons, full well capacity sets the maximum charge before clipping, and read noise sets the floor below which shadow detail is lost. The ratio between full well and read noise defines the engineering dynamic range.
| Parameter | Typical range | What it affects |
|---|---|---|
| Quantum efficiency | Roughly 50 to 80 percent at peak | Light sensitivity and noise |
| Full well capacity | Several thousand to tens of thousands of electrons | Highlight headroom |
| Read noise | Roughly 1 to 5 electrons | Shadow noise and low-light quality |
| Bit depth | 12 or 14 bits | Tonal resolution of raw files |
| Full sensor readout time | From under 5 ms to about 30 ms | Rolling shutter distortion |
These values vary with pixel pitch, sensor generation and design choices. Larger pixels usually hold more charge, while smaller pixels pack more resolution. Many sensors now use dual conversion gain, switching each pixel to a higher-gain mode at elevated ISO to lower read noise in dark scenes.
CMOS compared with CCD and its own variants
The defining difference from CCD lies in where the signal becomes a voltage. A CCD shifts charge packets across the chip to a single output amplifier, while a CMOS sensor converts charge inside each pixel and reads rows in parallel. This makes CMOS faster and more power efficient, and it allows processing circuits on the same die.
| Sensor type | Defining feature | Main advantage | Main limitation |
|---|---|---|---|
| Front-side illuminated CMOS | Wiring above the photodiodes | Simple and inexpensive | Wiring blocks some light |
| BSI CMOS | Photodiodes face the lens directly | Higher efficiency, better corners | More complex manufacturing |
| Stacked CMOS | Logic and memory on a separate layer | Very fast readout | Higher cost and heat |
| CCD | Charge shifted to one output node | Uniform response, easy global shutter | Slow, power hungry |
A BSI sensor moves the wiring behind the photodiode, which improves light collection especially at small pixel sizes. A stacked sensor bonds the pixel layer to a separate logic layer that carries converters and memory, which shortens sensor readout dramatically and enables high burst rates and electronic shutters with little distortion.
Readout and electronic shutter behavior
Most CMOS sensors use a rolling shutter, reading the image row by row. If the subject or camera moves during that interval, vertical lines lean and propellers bend, and artificial lights that flicker can produce horizontal banding. The severity depends directly on the total readout time.
A global shutter CMOS sensor stores every pixel’s charge simultaneously and then reads it out, eliminating skew. It requires an extra storage element per pixel, which historically cost some dynamic range and base sensitivity. Stacked designs with very short readout narrow the gap, so the choice between rolling and global shutter is increasingly about subject speed and lighting rather than overall image quality.
| Symptom | Cause | Practical response |
|---|---|---|
| Leaning verticals in pans | Slow rolling readout | Use mechanical shutter or a faster sensor |
| Banding under LED light | Flicker during readout | Match shutter speed to mains frequency |
| Bent fast-moving objects | Subject moves between rows | Global shutter or mechanical shutter |
| Reduced bit depth in bursts | Faster but coarser conversion | Use slower drive mode for maximum quality |
Limitations and common misunderstandings
CMOS sensors show several characteristic weaknesses. Fixed pattern noise arises from small differences between pixel amplifiers and column converters, which appears as faint stripes in heavily lifted shadows. Long exposures add thermal noise from dark current, which grows with temperature and exposure time.
- CMOS is not a quality level. It names a technology, and sensors built with it range from small phone chips to medium format.
- Megapixels do not define performance. Sensor area, quantum efficiency and read noise matter more for low-light results.
- Higher ISO does not make the sensor more sensitive. It increases gain applied to the same collected charge.
Explained without jargon
A camera sensor is a grid of millions of tiny light buckets. During a photo, each bucket fills with electrical charge in proportion to the light that falls on it. In a CMOS sensor, every bucket has its own small measuring circuit, so the camera can read many of them at once and very quickly.
That speed is why phones and modern cameras can shoot fast bursts and video. The main trade-off is that rows are read one after another, so very fast motion can look slightly tilted unless the sensor is especially quick.
A typical use case
A motorsport photographer shoots cars passing at high speed with a camera built around a stacked CMOS sensor. Using the silent electronic shutter at 1/2000 s and 20 frames per second, the photographer captures the exact moment a car clips the apex without the noise or vibration of a mechanical shutter.
Because the sensor reads the full frame in a few milliseconds, the wheels stay round and the barriers behind the car stay straight even while panning. On an older body with a slower rolling readout, the same settings produced visibly slanted backgrounds. Readout speed, not resolution, decided whether the images were usable.