Global shutter vs rolling shutter

How each shutter exposes the sensor

The shutter mode describes how a CMOS sensor starts and stops integrating light across its pixel array. A global shutter exposes every pixel at the same instant and then transfers the accumulated charge to shielded storage simultaneously, so the entire frame represents one common moment in time. A rolling shutter, by contrast, resets and reads the array one row (or small group of rows) at a time, sweeping from the top of the sensor to the bottom. Each row therefore has its own exposure window, slightly offset in time from its neighbours.

The consequence is subtle but important: in a rolling-shutter image the top of the frame was captured a few hundred microseconds to a few milliseconds earlier than the bottom, depending on the readout speed and resolution. For a perfectly still scene this is invisible. The moment anything in the field of view moves during that row-by-row sweep, the different capture times become visible as geometric error.

Most older and many high-resolution CMOS sensors are rolling shutter because the architecture is simpler and uses the pixel area more efficiently. Global-shutter pixels need extra in-pixel storage and transistors, which historically reduced fill factor and raised cost, though modern designs have largely overcome this.

Motion distortion and skew with rolling shutter

Because each row is captured at a different instant, a moving object is sampled at a slightly different position in every row. A vertical edge moving horizontally is rendered as a slanted edge (skew); a fast object can appear stretched, compressed or wobbly (the classic 'jello' effect). If illumination is pulsed or strobed during readout, only the rows that happened to be integrating during the flash receive light, producing bright and dark bands across the frame.

For industrial vision these artefacts are not cosmetic, they are measurement errors. Edge positions shift, so dimensional gauging is biased; barcodes and printed characters shear, hurting decode and OCR reliability; and any feature used for alignment or robot guidance lands in the wrong place. The faster the motion or the higher the resolution (longer readout time), the worse the distortion becomes.

You cannot fully correct rolling-shutter skew in software for a general scene, because the true object position during readout is unknown and may be non-uniform. Reducing exposure time limits motion blur but does not remove the row-to-row time offset that causes skew. The reliable fix is to remove the motion during capture or to use a global shutter.

When rolling shutter is acceptable

Rolling shutter is a perfectly good choice when the scene is static at the moment of capture, or when motion is slow enough that the displacement across the readout time is far below your measurement tolerance (a useful rule of thumb is well under one resolved pixel of movement during readout). Typical examples include indexed stations where the part is fully stopped and settled before triggering, microscopy, web or document scanning of stationary material, and general inspection where parts are presented one at a time and held still.

Rolling-shutter sensors also bring genuine advantages worth weighing. They are commonly available at higher resolutions and smaller pixel pitches, often with lower read noise and excellent low-light performance, and they tend to cost less per pixel. For a static, high-detail inspection such as reading fine print or detecting small cosmetic defects on a stopped part, a rolling-shutter sensor can deliver more usable detail for the money.

If you must image moving parts with a rolling shutter, the workaround is to freeze the scene optically rather than electronically: a very short, bright light pulse fired while the whole array is simultaneously integrating can effectively define the exposure. This only works if the flash occurs during the brief window when all rows overlap, which constrains timing tightly and is generally less robust than simply choosing a global-shutter camera.

• Part is fully stopped and mechanically settled before the trigger
• Continuous but very slow motion, with sub-pixel displacement across readout
• No strobed or pulsed lighting synchronised to a moving subject
• Detail, resolution or cost per pixel matters more than motion handling
• You can guarantee stable, distortion-free timing for every capture

Why global shutter is mandatory with motion

Whenever the object and camera move relative to each other at the moment of capture, global shutter is the correct and usually the only defensible choice. Because all pixels integrate over the same time window, the image is geometrically faithful: straight edges stay straight, dimensions are unbiased, and codes and characters keep their shape regardless of speed. This is essential for conveyor and indexing-line inspection, pick-and-place and robot guidance, flying-trigger applications, and any high-speed sorting where parts never stop.

Global shutter also pairs cleanly with strobed lighting. You can fire a short, high-intensity flash during the common exposure window to freeze motion and dominate ambient light, knowing every row sees the same pulse. This combination, a global-shutter sensor plus a brief synchronised flash, is the standard recipe for sharp, repeatable images of fast-moving product.

For metrology and gauging specifically, geometric integrity is non-negotiable. A rolling-shutter skew of even a fraction of a pixel translates directly into a measurement bias, so global shutter is the baseline expectation for any inspection that reports dimensions or feeds closed-loop control.

Exposure, noise, PRNU and cost trade-offs

Global-shutter pixels carry extra in-pixel storage nodes, which adds two engineering concerns. The first is shutter efficiency (sometimes expressed as parasitic light sensitivity): the storage node can pick up a small amount of stray light during readout, which limits performance in very bright or long-readout conditions. The second is that the additional transistors historically reduced fill factor, lowering sensitivity unless compensated by microlenses or backside illumination.

Rolling-shutter designs avoid the storage node, so they often achieve lower read noise, higher quantum efficiency and finer pixel pitch at a given price. Both architectures still exhibit fixed-pattern non-uniformities, photo-response non-uniformity (PRNU) and dark-signal non-uniformity, which good cameras correct through factory calibration; these are not a reason to prefer one shutter over the other. The practical cost picture is that a global-shutter sensor of comparable resolution typically still commands some price premium, but far less than a decade ago.

When you size exposure, remember the two effects are different. Motion blur (smear within a single pixel) depends on exposure time and is controlled the same way in both shutter modes. Rolling-shutter skew depends on total readout time across the frame and is independent of exposure, which is why shortening the exposure cannot rescue a rolling-shutter image of a fast-moving object.

Modern CMOS global shutter and how to choose

Current global-shutter CMOS sensors have closed most of the historic gap. Backside illumination, improved microlenses and refined pixel design have raised fill factor and quantum efficiency while pushing shutter efficiency to very high levels, and resolutions now extend well into the range once reserved for rolling-shutter parts. Industrial families from brands such as Basler, iRAYPLE and Omron / Sentech (STC) offer global-shutter models across GigE Vision, USB3 Vision and CoaXPress interfaces, so the shutter choice rarely forces a particular bus.

The selection logic is straightforward. If there is any relative motion at capture, default to global shutter; if the scene is reliably static and you want maximum resolution, sensitivity or value, rolling shutter is a sound and economical option. Confirm the data you actually need (smallest defect resolved with roughly 3 to 5 pixels), the line speed, the trigger and strobe scheme, and whether your measurement reports geometry, then let those drive the shutter decision rather than choosing afterwards.

If you are unsure how fast your parts really move relative to the readout time, or you are weighing a borderline slow-motion case, it is worth validating with a representative trial or asking for help before committing to optics, lighting and mechanics around the wrong sensor.