Rolling Shutter Against Global

Generic engineering material, applicable to any module from any supplier.

For teams separating technical work from project administration, common ways time trackers can be manipulated provides additional workflow context.

A rolling shutter exposes and reads the sensor row by row. A global shutter exposes every pixel at the same instant and reads afterwards.

Where nothing moves, the two are indistinguishable. Where anything does, the difference is a geometry error rather than a matter of appearance.

What rolling shutter does

Skew. An object crossing the frame is at a different position when the last row is read than when the first was — so a vertical edge photographs as a slanted one.

Further background is available from LoRa Alliance.

Wobble, where the camera itself vibrates: the image distorts in a way that changes frame to frame.

Partial exposure, where illumination is pulsed. Some rows catch the pulse and others do not, producing bands across the image.

And distorted rotation, where a rotating part appears bent.

None is noise or blur. The image is sharp and the shape is wrong, which is why the problem survives every attempt to fix it with more light or a faster exposure.

Where it stops being cosmetic

Measurement. A dimension taken from a skewed image is wrong by an amount that depends on speed, which means the error varies between parts and cannot be calibrated out.

Position and alignment, for the same reason.

Reading codes on moving objects, where skew distorts the geometry the decoder relies on.

And any pulsed or strobed illumination, which is common in industrial installations and interacts badly.

Where the task is presence detection or a human looks at the picture, rolling shutter is frequently fine — and it is the majority of installations.

The honest trade

Global shutter sensors cost more, are available in fewer formats, and frequently have larger pixels for the same resolution — which means a larger sensor and a different lens calculation.

Rolling shutter sensors are cheaper, more widely available, and generally better in low light for a given size.

Which makes the decision a question about motion rather than about quality: if the subject and the camera are both still at the moment of capture, rolling shutter is the better engineering choice, not a compromise.

Making rolling shutter work

Three approaches where the sensor is fixed.

Stop the motion. Trigger capture when the part is stationary, which in indexed machinery is frequently possible and costs nothing.

Shorten the readout. Windowing to fewer rows reduces the time between first and last, which reduces skew in proportion.

And synchronise the illumination. A strobe firing while all rows are simultaneously exposed — the interval that exists in some sensors between the last row starting and the first finishing — behaves like a global shutter for the duration of the flash.

The third is the standard technique and it requires the sensor to support the mode, which is a datasheet question.

Estimating the skew

Worth doing before deciding, and the arithmetic is short.

Skew = object speed × readout time.

A part moving at 0.5 m/s, with a sensor whose readout takes 30 ms, displaces by 15 mm between the first row and the last — which across a 60 mm part is a quarter of its width.

The readout time is the figure to find, and it is not always the same as one over the frame rate: a sensor running at 30 frames per second may read out faster than 33 ms and idle between frames.

Ask for the row readout time or the total readout time. It is a specification question with a definite answer, and it converts a vague worry into a number you can compare against your tolerance.

Flicker from artificial light

A separate rolling-shutter effect and a common complaint.

Mains-powered lighting varies in brightness at twice the supply frequency. With a rolling shutter and a short exposure, different rows catch different points in that cycle — producing horizontal bands that drift up or down the image.

The remedy is choosing an exposure time that is a whole multiple of the light's period, which most sensors support as a flicker avoidance mode.

It needs to know the local supply frequency, which differs by region — and a system configured for one and deployed in the other bands visibly.

What a global shutter does not fix

Motion blur.

A global shutter captures every pixel at the same time and still captures them over a period. A subject moving during that exposure blurs exactly as it would otherwise.

Blur is an exposure-time problem and is solved by shortening the exposure, which requires light. Skew is a readout problem and is solved by the shutter type.

Confusing the two leads to buying a global shutter sensor for a blur problem, which is expensive and does not work.

Where it appears in a variant code

Usually it does not.

Shutter type is a property of the sensor, so it distinguishes module families rather than variants within one — and a family built on a rolling-shutter sensor offers no global-shutter option at any position of the code.

Which means this decision is made before the variant question, alongside format and resolution, and it is one of the few that cannot be revised by ordering a different suffix.

In one line

Ask whether anything moves at the moment of capture, and the answer decides the sensor family.

The short version

Rolling shutter exposes row by row; global shutter exposes everything at once
The result is skew, wobble under vibration, banding under pulsed light, and distorted rotation — all geometry errors rather than blur
It stops being cosmetic where the task is measurement, position, code reading on moving objects, or strobed illumination
Global shutter costs more, comes in fewer formats and is generally worse in low light for a given size
If subject and camera are still at capture, rolling shutter is the better engineering choice rather than a compromise
Where it must work with motion: trigger when stationary, window to fewer rows, or strobe during the simultaneous-exposure interval