Image Sensor
An image sensor converts a two-dimensional light pattern into a digital array of pixel values. CMOS sensors dominate today: each pixel has an amplifier on-chip (active-pixel sensor), enabling random access and low-noise rolling or global shutter readout.
An image sensor converts incoming light into electrical signals across an array of sensing elements. Charge collection, readout and conversion differ between CCD and CMOS architectures and between individual devices. Some CMOS designs use column conversion or correlated sampling, but those are not universal features of every image sensor.
In plain terms
A city of tiny light-collecting buckets: each pixel is a photodiode that accumulates charge proportional to the light that falls on it during the exposure, then a scanner reads every bucket's charge level and converts it to a number, forming an image.
Why designers use it
- Capture still and video frames in smartphones, DSLRs, and security cameras.
- Provide machine-vision input for barcode scanners, medical endoscopes, and industrial inspection.
- Enable facial recognition and gesture control in consumer devices.
- Detect astronomical objects in scientific telescopes and satellite instruments.
Best for
- Photography
- Machine vision
- Medical imaging
Key specifications
- Resolution: 0.3 MP (VGA) – 200 MP
- Pixel size: 0.6 µm – 5 µm
- Frame rate: 30 fps – 1,000 fps
- Dynamic range: 60 dB – 110 dB (HDR)
- Interface: MIPI CSI-2, parallel, LVDS, SPI
When not to use it
- When you only need presence/absence detection — a photodiode or PIR is far simpler.
- In extremely high-radiation environments without rad-hard variants — standard CMOS pixels latch up under ionising radiation.
Common mistakes
- Routing the clock and data lines from the MIPI CSI-2 interface without matched-length differential pairs, introducing phase skew that the receiver can't correct.
- Powering the pixel array and the I/O pads from the same supply, coupling readout noise into the analog signal path.
Where you will find it
- A smartphone rear camera uses a 50-megapixel CMOS image sensor with phase-detect autofocus pixels embedded in the array: dedicated pixel pairs measure the phase difference of the subject's light to compute focus distance faster than any contrast-detect algorithm, achieving lock in under 100 ms.
- A laparoscopy camera uses a 1/6-inch CMOS sensor at the tip of a 5 mm surgical instrument: the sensor's on-chip ADC and MIPI interface compress the image data to fit through the instrument's narrow electrical bundle, and a global shutter prevents the motion artefacts that a rolling shutter would produce as the surgeon moves the scope.
- The Perseverance Mars rover's hazard cameras use rad-hard CMOS image sensors to map the terrain in front of the wheels: each forward-facing stereo pair produces a disparity map that the autonomous navigation software uses to plan a safe path between rocks, with no human in the loop because the round-trip radio delay is up to 24 minutes.
A short history
Nokia Bell Labs' account credits Boyle and Smith with the CCD in 1969 and distinguishes Michael Tompsett's subsequent development of it for imaging in the 1970s. It also describes later contributions including Teranishi's pinned photodiode and Fossum's CMOS active-pixel work. Digital imaging developed through several distinct innovations rather than one instantly complete camera chip.
Good to know
- A light-sensitive device and a complete image-readout system are different engineering problems.
- Higher pixel count alone does not establish lower noise, better dynamic range or faster readout.
- Choose a sensor from its measured performance and interface requirements, not from a blanket claim that CCD or CMOS is always better.