The fairy lights over the balcony look perfectly steady to everyone standing under them. Raise a phone and shoot ten seconds of video, and the string turns into a strobe act: dark bands roll up the lights like a slow shutter of shadow, though nothing in the room moved. Neither your eyes nor the camera is broken. The lights really are switching on and off, your visual system politely hides it, and the camera, reading the scene in slices, cannot. The whole effect is a disagreement between two ways of sampling the same blinking light, a cousin of the moment photographing a screen adds rainbows that were never there: the equipment keeps an honest ledger of a pattern your senses smoothed away.
Dimming by blinking
Many LED products do not dim by being half-lit; they dim by being fully lit a fraction of the time. A driver switches the LED completely on and completely off, thousands of times per second, and the percentage of time spent on, the duty cycle, sets the brightness you experience. The technique, pulse-width modulation, is cheap, precise, and invisible to most people most of the time. Lighting engineers do care about the invisible part, though: modulation rates that are too low have been flagged in a recommended practice from the IEEE precisely because some observers perceive or are affected by slow flicker, and federal lighting programs publish guidance on measuring it. In decorative light strings and inexpensive dimmable bulbs, the modulation rate is sometimes chosen for cost rather than comfort, which is why the strobe effect is commonest exactly there.
A camera reads like a typewriter
Most camera sensors do not capture a whole frame in one instant. They expose and read out row by row, top to bottom, so a single video frame is really a stack of thin time slices, each recorded a moment after the one above it. When the light source is continuous, the slices agree and the frame looks uniform. When the source pulses, each row lands somewhere in the on-off cycle: rows exposed during an ON pulse record a bright string, rows exposed during the gap record near-darkness, and the frame arrives striped. The geometric model of this row-by-row capture has been worked out in the computer-vision literature, where rolling shutter is treated as the root of whole families of distortion.
When two clocks disagree
The crawl comes from the arithmetic between the pulse rate and the frame rate. If every frame sampled the cycle at exactly the same phase, the bands would hold still; a slight mismatch means each frame samples a slightly later point in the cycle, so the stripes drift frame after frame, up or down depending on the sign of the mismatch. Slow-motion makes it worse, because higher frame rates sample the cycle more finely. This is the same family of aliasing that gives projector images their rainbow smear for some viewers, a color-wheel version of the eye being sampled too slowly, described in how a projector’s color breakup works. In each case two rhythmic systems, light and sampler, disagree about what counts as continuous.
Why your eyes are spared
Human vision has its own fusion threshold: past roughly sixty to ninety pulses per second, depending on the person, brightness, and viewing conditions, a flickering light merges into steady glow. Well-designed LED drivers pulse far above that range precisely so nobody will notice. Cameras have no such mercy; they freeze microseconds, so they will happily expose a pulse train your retina would never report. You can catch the disagreement cheaply: film the lights and scrub through the clip frame by frame, or pan the camera gently and watch the bands race. Indoors there is a bonus layer of culture shock, because lamps on different national power grids modulate at the local mains frequency, so a camera configured for one country’s grid can stripe the lights of another. Videographers learn the pairing deliberately, matching shutter settings to the region they shoot in, a small ritual of synchronizing two clocks nobody can see. The same mismatch ambushes video calls, when a webcam’s auto-exposure starts breathing in sympathy with the room’s lamps, and it once striped broadcast segments of LED stadium signage whose pulse rate happened to argue with television scan rates. Once you can name the mechanism, the artifacts stop looking like hauntings and start looking like schedules. People differ, too. Some individuals notice modulation that others fuse without effort, one reason the IEEE practice pushes modulation frequencies well above the comfortable minimum. Spotting which element in a system is misbehaving, the light, the sensor, or your own sampling eye, is exactly the diagnostic game Broken Slider plays with color controls.
The fix, when the bands bother you, is to change one of the two clocks: use lights that dim smoothly rather than cheaply, adjust the camera’s frame rate or shutter settings toward the local power frequency, or accept the stripes as a rare honest look at how your lights actually run.