Rolling Shutter Propeller Bend Lab

Sensor & Rotation

Row readout time 20 µs

Sensor rows 120

Fan speed 900 RPM

Blade count 3

Shutter Type

Live Readouts

Frame readout T

2400 µs

Shutter mode

Rolling

Angular smear Δθ

0.0°

Tip angular speed

0 °/s

Scan progress

row 0 / 120

How it works

A CMOS sensor does not expose the whole frame at once. It resets and reads pixel rows one after another, so row r is sampled at a slightly later instant than row 0:

t(r) = t0 + r·τ            (rolling shutter)
t(r) = t0  for all r        (global shutter / CCD)
T = N_rows · τ              (total frame readout time)

Here the moving subject is a spinning fan instead of a panning scene, so the same row-by-row time lag distorts rotation rather than translation. Each blade is a straight arm at angle θ(t) = θ0 + ω·t. A sensor row at world-height y can only "see" a blade at the instant its arm actually crosses that height, so the reconstructed photo picks a different blade angle for every row:

θ(r) = θ0 + ω · t(r)
x(r) = y_row / tan(θ(r))     (row/blade intersection, rolling)
Δθ  = ω · T                  (total angle swept during one frame)

Because θ(r) keeps advancing while the sensor scans down the frame, a perfectly straight blade traces out a curved "sickle" shape in the photo — the classic bent-propeller / bent-fan-blade rolling-shutter artifact. A global-shutter sensor exposes every row at the same instant t0, so θ(r) is constant and every blade stays a straight radial line no matter how fast it spins.

Row readout time τ — real phone sensors run 5–30 µs/row; a slower readout means more angular smear per frame.
Sensor rows — more rows means a longer total readout time T = N·τ for one frame.
Fan speed — how fast the blades spin, in revolutions per minute.
Playback here runs 1000× slower than the real sensor (1 µs of sensor time shown as 1 ms on screen) purely so the row-by-row scan is visible — the formulas above use the real µs values.

Live sceneCaptured photo

Blade arm Current scan row