Why do photos of propellers look bent?
Most digital cameras do not capture the whole image at one instant — they scan it line by line from top to bottom, so anything moving fast is recorded at a slightly different position in every row.
Simple intuition
The plain reason, in everyday words
It is natural to assume a photograph captures a single frozen instant, like a snapshot of time. Most digital cameras do not work that way. The sensor is read out one horizontal row of pixels at a time, starting at the top and sweeping down, and the whole sweep takes a few thousandths to a few hundredths of a second. For anything slow, that is invisible. But a propeller blade might travel a long way across the frame in that time, so by the time the sensor reaches the bottom rows, the blade has moved. Each row records the blade in a slightly different place, and stacking those rows together produces a shape that never existed — a bent, sheared, or detached blade. The camera did not distort anything. It faithfully recorded a hundred different moments and stacked them.
The propeller really is bending under load.
Blades do flex, but by a tiny amount compared to what these photographs show. The dramatic curves and detached blades are entirely artefacts of readout timing.
A faster shutter speed fixes it.
Shutter speed sets how long each row is exposed; rolling shutter is caused by the offset between when rows *start*. A faster shutter reduces motion blur within each row and can actually make the shearing look sharper and more obvious.
It is a compression or software artefact.
It happens in raw sensor data before any encoding. You can reproduce it with compression disabled entirely.
Rolling shutter and the backwards-spinning-wheel effect are the same thing.
They are different. The wheel illusion is aliasing between frames and occurs on any camera; rolling shutter distortion happens within a single frame and requires sequential readout.
It is a clean lesson in the difference between what an instrument records and what was there, which matters well beyond photography. It also has practical consequences: it constrains flash photography, it is why drone and action footage needs specific stabilisation, and it is a known problem in machine vision and photogrammetry, where an uncorrected rolling shutter quietly corrupts measurements taken from moving platforms.
Who worked it out
The artefact long predates digital sensors. Focal-plane shutters in film cameras, which sweep a slit across the film, produce exactly the same distortion — the famous 1913 Lartigue photograph of a racing car with elliptical wheels and a backwards-leaning body is a mechanical rolling shutter at work.
What problem forced it
In film cameras the moving slit existed because it allowed much shorter effective exposures than a leaf shutter could achieve. In digital sensors, sequential readout exists because it lets many pixels share readout circuitry, making the sensor cheaper, smaller, and more sensitive.
How it changed since
CCD sensors dominated early digital imaging and were less prone to the effect; the industry's shift to CMOS in the 2000s made rolling shutter ubiquitous just as video on consumer devices exploded. The current direction is stacked sensors with fast parallel readout that shrink the offset towards insignificance, and global shutter CMOS designs that have finally become viable without an unacceptable sensitivity penalty.
Why wheels sometimes spin backwards on film
The other timing illusion in the same footage, and a genuinely different mechanism.
How image stabilisation works
The engineering that partly corrects this, and why it can only partly do so.
Written for Curio rather than collected from a forum — it is part of the curated corpus that ships with the platform. The references it draws on are listed under Sources.