CMOS Image Sensors: Pixels, Shutters & Designs

CMOS image sensors convert light into digital images using an array of pixels, each with its own amplifier, and can use either a rolling or global shutter. This article explains how pixels work, the differences between shutter types, and the main sensor designs used in cameras today.

Formula

The exposure value (EV) is a measure of the combination of aperture and shutter speed that gives a correct exposure. For a given scene luminance (L) in cd/m², ISO sensitivity (S), and exposure time (t) in seconds, the EV is:

EV = log₂((L × S) / (t × 100))

Where:

• L = scene luminance in candelas per square meter (cd/m²)
• S = ISO sensitivity (dimensionless)
• t = exposure time in seconds

The relative noise is approximated as the square root of exposure time, since noise in a pixel is proportional to the square root of the number of photons collected.

Worked Example

Using the default values: t = 0.01 s, S = 100 ISO, L = 1000 cd/m², a = f/2.8.

EV = log₂((1000 × 100) / (0.01 × 100)) = log₂(100,000 / 1) = log₂(100,000) ≈ 16.6

So the exposure value is about 16.6, which is a bright sunny day exposure. The relative noise is sqrt(0.01) = 0.1.

How Pixels Work

A CMOS pixel typically uses a photodiode that converts photons into electrons. The charge is then converted to a voltage and amplified by a transistor within the pixel. This is why CMOS sensors are called “active-pixel� sensors – each pixel has its own amplifier, unlike CCD sensors which transfer charge to a central amplifier.

The key components of a CMOS pixel are:

• Photodiode: absorbs light and generates electron-hole pairs
• Transfer gate: moves charge from the photodiode to the floating diffusion
• Floating diffusion: a capacitor that converts charge to voltage
• Source follower: a transistor that buffers the voltage
• Reset transistor: clears the charge before exposure
• Row select transistor: connects the pixel to the column readout line

The number of electrons a photodiode can hold is called full well capacity. Larger pixels (e.g., 4 µm vs 1 µm) can hold more electrons, which increases dynamic range and reduces noise at low light.

Shutter Types: Rolling vs Global

The shutter controls how long the sensor is exposed to light. There are two main types:

Rolling shutter exposes rows of the sensor sequentially, one after another. This is common in most consumer cameras because it is cheap and allows high readout speeds. However, it causes distortion when photographing fast-moving subjects – vertical lines appear slanted (the “jello effect�).

Global shutter exposes all pixels simultaneously, then reads them out. This eliminates motion distortion but requires more complex pixel design (e.g., adding a storage capacitor per pixel), which increases cost and reduces fill factor (the area of the pixel that actually detects light). Global shutters are used in industrial, scientific, and some high-end cinema cameras.

Some sensors use a mechanical shutter in front of the sensor to avoid rolling artifacts, but this is not a sensor design per se.

Sensor Designs: FSI, BSI, and Stacked

CMOS sensor architectures have evolved to improve light capture and speed:

• Front-illuminated (FSI): The wiring layer sits above the photodiode, reflecting some light. Cheaper but less efficient.
• Back-illuminated (BSI): The sensor is thinned and flipped so light hits the photodiode directly, avoiding the wiring. This improves quantum efficiency and is now standard in most smartphones and mirrorless cameras.
• Stacked (or back-side illuminated stacked): The sensor layer is separated from the logic layer and stacked vertically, connected by through-silicon vias. This allows faster readout, higher resolution, and features like global shutter or phase-detection autofocus on the same chip.

Another design is the organic photoconductive film sensor, which replaces silicon photodiodes with organic materials, but it is not yet mainstream.

How to Choose a Sensor

When selecting a camera or sensor, consider:

• Pixel size: larger pixels generally mean better low-light performance and dynamic range, but lower resolution for a given sensor size.
• Sensor size: full-frame (36×24 mm), APS-C (about 23.5×15.6 mm), Micro Four Thirds (17.3×13 mm), and 1-inch (13.2×8.8 mm) are common. Larger sensors collect more total light.
• Shutter type: if you photograph fast action, a global shutter or a mechanical shutter is better than rolling.
• Readout speed: important for video and burst shooting. Stacked sensors have the fastest readouts.
• Quantum efficiency: the percentage of photons converted to electrons. BSI sensors have higher QE than FSI.

For scientific or industrial use, you might prioritize global shutter and low noise over resolution. For general photography, a BSI stacked sensor with a rolling shutter is a good balance.

Common Problems and Misconceptions

Noise: All sensors have read noise (from the amplifier) and shot noise (from the random arrival of photons). Increasing ISO amplifies the signal but also the noise, so it is not “adding� noise but making it more visible.

Hot pixels: Defective pixels that always appear bright. They are usually mapped out by the camera’s firmware.

Rolling shutter artifacts: Many people think a fast shutter speed eliminates rolling shutter, but it does not – the distortion depends on the time between row exposures, not the exposure time.

Megapixel myth: More megapixels do not automatically mean better quality. If the sensor size is fixed, more pixels means smaller pixels, which can reduce low-light performance.

Global shutter vs mechanical: A mechanical shutter still exposes the whole sensor simultaneously (when the curtain is fully open), but it has moving parts that wear out and can cause vibration.

Costs and Trade-offs

Sensor cost is driven by die size, manufacturing process, and yield. Full-frame sensors are much more expensive than APS-C because the larger die area reduces the number of chips per wafer. Stacked sensors require advanced packaging, adding cost. Global shutter sensors have lower fill factor and thus lower sensitivity, which is why they are not used in consumer cameras where low light is common.

For most users, a BSI rolling-shutter sensor is the best value. Global shutter is worth the premium only if you need to capture fast motion without distortion, such as in machine vision or sports analysis.

Frequently Asked Questions

What is the difference between CMOS and CCD sensors? CCD sensors transfer charge across the chip to a single amplifier, giving very uniform output but slower readout and higher power consumption. CMOS sensors amplify each pixel individually, allowing faster readout, lower power, and integrated features like autofocus, but historically had more fixed-pattern noise.

Why do smartphone photos have a “jello� effect on fast-moving objects? That is rolling shutter distortion. The sensor reads rows sequentially, so the top of the image is captured slightly before the bottom. When the subject moves during that time, it appears slanted or curved.

Can a global shutter be used in a smartphone? Yes, some recent smartphone sensors (like Sony’s IMX902) use a global shutter, but they are still niche. The trade-off is lower sensitivity and higher cost, which is why most phones stick with rolling shutter.

What does “back-illuminated� mean? In a back-illuminated sensor, the silicon layer is thinned and the photodiode is exposed to light from the back side, away from the wiring. This increases the amount of light that reaches the photodiode, improving sensitivity and reducing noise.

How does pixel size affect image quality? Larger pixels collect more photons, which increases the signal-to-noise ratio. This means less visible noise in low light and higher dynamic range. However, larger pixels mean fewer pixels for a given sensor size, so there is a trade-off between resolution and light sensitivity.

Is more megapixels always better? No. If the sensor size stays the same, more megapixels means smaller pixels, which can reduce low-light performance and dynamic range. The best choice depends on your use case: if you print large or crop heavily, more megapixels help; if you shoot in low light, larger pixels are better.

Related

• PHP RAG Pipeline: Build Answers That Cite Documents
• GST Authorities and ITC: Must They Verify the Supply?
• Fake Number Calls: Is Someone Spoofing Your Number?

in t | Exposure time | s | 0.01 | 0
in s | Sensor sensitivity | ISO | 100 | 0
in l | Scene luminance | cd/m² | 1000 | 0
in a | Aperture f-number | - | 2.8 | 0
out e | Exposure value (EV) | - | log2((l * s) / (t * 100)) | 1
out n | Relative noise | - | sqrt(t) | 3