Sensor readout is the process by which a camera transfers the signal collected by each photosite to the analog-to-digital converters and then to the image processor, turning an exposure into digital data. In photography, readout speed and quality decide how fast a sensor can deliver a frame, how strongly moving subjects are distorted with an electronic shutter, and how much read noise is added to the image.
Readout is a separate step from exposure itself. Light is gathered during the exposure time, but the camera can only hand over a complete image once every row has been read, so the readout time often limits burst rates, video modes and the usefulness of the silent shutter.
How sensor readout works in CMOS and CCD sensors
In a CMOS sensor, each pixel converts its collected charge into a voltage with its own amplifier. The sensor then reads the image row by row: a row is selected, its pixel voltages are passed down the columns, and column-parallel analog-to-digital converters digitize the whole row at once before the next row follows.
Before conversion, a technique called correlated double sampling measures each pixel twice, at reset and after exposure, and subtracts the two values. This removes much of the reset noise and fixed offsets, which is a major reason why modern CMOS sensors reach read noise of only a few electrons.
A CCD sensor reads differently. Charge packets are shifted across the chip through registers to one or a few output amplifiers, so all pixels share the same conversion path. This gives very uniform images but slow readout and high power use, which is why CCDs have largely disappeared from consumer cameras.
The table below summarizes the stages of a typical CMOS readout chain.
| Stage | What happens | Effect on the image |
|---|---|---|
| Charge-to-voltage conversion | Pixel amplifier turns electrons into voltage | Sets conversion gain and part of read noise |
| Row selection | One row at a time is connected to the columns | Creates the time offset between rows |
| Correlated double sampling | Reset level is subtracted from the signal | Suppresses reset noise and offsets |
| Analog-to-digital conversion | Column converters digitize the row | Defines bit depth, typically 12-14 bits |
| Data output | Digital values stream to the processor | Interface bandwidth limits total speed |
Readout speed and rolling shutter
Because rows are read in sequence, the top and bottom of the frame are captured at slightly different moments when an electronic shutter is used. The total time from the first to the last row is the readout time or rolling shutter speed, and it determines how strongly fast motion is distorted.
Values vary widely with sensor design and resolution. The table gives typical full-frame ranges for still photography; exact figures depend on the model and the bit depth chosen.
| Sensor design | Typical readout time | Practical consequence |
|---|---|---|
| Conventional high-resolution CMOS | About 1/15 to 1/30 s | Visible skew with panning and moving subjects |
| Conventional lower-resolution CMOS | About 1/30 to 1/60 s | Usable silent shutter for slower subjects |
| Stacked CMOS | About 1/120 to 1/250 s | Fast bursts and little distortion |
| Global shutter CMOS | All pixels sampled at once | No rolling shutter skew at all |
| CCD | Slow, but exposure ends simultaneously | Global exposure with low frame rates |
Slow readout produces skewed verticals, bent propellers and banding under flickering lights, and it prevents electronic-shutter flash on most cameras. A stacked sensor attaches a logic and memory layer directly beneath the pixel layer, which multiplies the available bandwidth and shortens readout dramatically.
Readout modes and their trade-offs
Cameras do not always read every pixel at full precision. To reach high frame rates or video resolutions, they switch between readout modes that trade detail, noise or bit depth for speed.
| Readout mode | How it works | Trade-off |
|---|---|---|
| Full-resolution readout | Every pixel is read at full bit depth | Best quality, slowest speed |
| Reduced bit depth | Converters run at 12 bits instead of 14 | Faster, with slightly less shadow latitude |
| Line skipping | Only selected rows and columns are read | Fast, but causes aliasing and moiré |
| Pixel binning | Neighboring pixels are combined | Lower noise and resolution |
| Oversampling | Full readout downscaled in the processor | Sharp video, but more heat and skew |
Some sensors also offer dual conversion gain, switching to a high-gain readout at higher ISO settings. This lowers effective read noise in the shadows at the cost of reduced full-well capacity, which is one reason dynamic range curves show a step at a certain ISO value.
Sensor readout compared with related sensor concepts
Readout interacts with several neighboring concepts, but each describes a different property. The distinction helps when comparing camera specifications.
- BSI sensors improve light collection by moving wiring behind the photodiodes; they do not by themselves speed up readout.
- Stacked sensors primarily target readout speed and are often also backside illuminated.
- Megapixel count increases the amount of data per frame, so higher resolution usually means longer readout at the same bandwidth.
- Quad Bayer sensors can read binned groups of four pixels for faster, cleaner output or full resolution for detail.
Practical consequences and common pitfalls
Readout behavior should guide the choice between mechanical and electronic shutter. With a slow-readout sensor, the mechanical or electronic first-curtain shutter is safer for sports, aerial photography and scenes lit by LED or fluorescent sources.
- Using a silent electronic shutter for fast action on a slow-readout sensor, which bends moving subjects.
- Ignoring that burst modes may lower bit depth, reducing shadow recovery in post-processing.
- Assuming that a fast shutter speed prevents skew; exposure time and readout time are independent.
- Shooting under flickering artificial lights with an electronic shutter and getting horizontal bands across the frame.
Put simply
A camera sensor is like a large grid of tiny buckets that collect light. After the picture is taken, the camera has to empty the buckets and count what is inside. That counting step is the sensor readout.
Most sensors count one row at a time, from top to bottom. If the counting is slow and something moves quickly, the top and bottom of the picture show slightly different moments. Faster counting means less distortion and more pictures per second.
In practice
A motorsport photographer at a racetrack wants to shoot silently from a press position. With a high-resolution camera using a conventional sensor and a readout time of about 1/20 s, cars panned at 1/1000 s appear leaning forward and the wheels look oval, even though the exposure time is short.
Switching to the mechanical shutter removes the distortion, but limits the burst to about 10 frames per second. On the next assignment, the photographer uses a body with a stacked sensor and a readout of about 1/200 s. The electronic shutter now delivers 20 frames per second with no visible skew, and the cars keep straight, natural proportions throughout the sequence.