The Off-Key Bard
Technology

Why Do LEDs Flicker Even When They Look Steady?

Or: why your camera sees flashing lights that your eyes don't. The lamp hasn't changed. You've just added a second timing system.

Or: Why Your Camera Sees Flashing Lights That Your Eyes Don't

Walk into a room lit by modern LEDs and everything looks normal. The light is steady, the room is bright, nothing seems to be switching. Then you point your phone camera at one of the bulbs and strange things appear. Dark bands crawl across the screen. The light seems to pulse. Part of the image is brighter than the rest. The same thing can happen when you film a TV, a monitor, or an illuminated sign.

Neither your eyes nor the camera is necessarily broken. Many electronic lighting systems produce light that varies over time, sometimes thousands of times per second. Your visual system blends that variation into a continuous glow. A camera, which gathers light in a different way, can expose it. Understanding why means looking at how an LED makes light, and at what the electronics around it are doing.

An LED Is Not a Tiny Filament

An incandescent bulb passes current through a thin filament until it glows hot. The filament has thermal mass, so when power drops it keeps glowing briefly while it cools. That sluggishness quietly smooths out whatever the electricity is doing.

An LED has no such cushion. Inside it is a semiconductor junction. When current flows, electrons and holes recombine and release energy as photons, a process called electroluminescence. There's nothing to heat or cool, so light output follows changes in current almost instantly. That speed is why LEDs work in displays, indicators, optical communication, and car lighting. It also means that if the current wobbles, the light wobbles with it.

And the current is rarely simple. A household bulb isn't an LED chip wired to the wall. Mains power is alternating current, while the LED needs carefully controlled conditions, because its voltage-current relationship is steeply nonlinear and uncontrolled current can damage it. Every LED lamp therefore contains a driver, circuitry that converts and regulates power. Two bulbs with similar emitters can behave very differently because their drivers differ. The LED may be capable of steady light. The driver decides whether it gets any.

Two Common Sources of Flicker

Ripple from the wall. U.S. mains alternates at 60 hertz, and 50 hertz is common elsewhere. A driver rectifies it, and if the result isn't smoothed enough, a ripple remains. With full-wave rectification, the ripple runs at twice the mains frequency: 120 hertz in the U.S. and 100 hertz in 50 hertz regions. Avnet's explainer on LED flicker says it directly: rectified mains would be 100 hertz in Europe or 120 hertz in the U.S., "right in the problem area." It notes that cheap, simple single-stage drivers with inadequate smoothing, sometimes skimping on capacitors, can have high ripple, with flicker around 30 percent common in such designs. Better drivers use capacitors, switching regulators, filtering, and feedback to hold the current steady, though price and size pull the other way. That doesn't mean every cheap bulb is bad or every expensive one perfect; it means the answer is hiding in the base of the bulb.

Pulse-width modulation for dimming. To dim an LED you can reduce its current, or you can switch it on and off rapidly and change the fraction of time it spends on. That fraction is the duty cycle. At 50 percent, the LED is lit half of each cycle; at 20 percent, a fifth. This is the same trick behind the computer fan control in an earlier post. During the "on" portions the LED can reach full instantaneous brightness, and the controller varies how long that lasts. To an observer, if the switching is fast enough, it looks like a lamp at half or a fifth of its brightness. But the LED may actually be turning completely off between pulses.

Other sources exist: poorly regulated drivers, electromagnetic interference, loose connections, incompatible dimmers, and the multiplexing used in displays. Treating all flicker as PWM is a mistake. The light can look similar while the cause is completely different.

Why Your Eyes Let It Slide

The retina and the neural pathways behind it respond to light with finite speed, and the brain blends information across short spans of time. When modulation is fast enough, individual changes fuse into steady light. This is flicker fusion. But there is no single frequency above which everyone sees a steady lamp. The threshold depends on brightness, contrast, how deeply the light varies, where it falls in your visual field, and the person. Peripheral vision can be more sensitive, and moving your eyes can reveal modulation you'd never see while staring.

Fusion also doesn't mean the flicker stopped. The signal is still there, like an audio tone above the range of your hearing, still a pressure wave and merely unheard. The light varies physically whether or not you consciously notice. Your perception is a construction, not a graph of the signal.

How a Camera Sees It

A camera doesn't blend. For video it records a sequence of frames, say 30 or 60 per second, and each frame collects light during an exposure interval. If the lamp's brightness varies during those exposures, a frame's brightness depends on which part of the lamp's cycle it happened to catch. One frame lands on a bright stretch, the next on a dim one, and the video seems to pulse even though the room looks stable to a person standing in it.

That's sampling again, and it connects to the microphone post. There, sampling a sound too slowly folded fast frequencies into false slow ones, aliasing. Video does the same with light. A lamp flickering at 120 hertz can appear to pulse slowly on camera because the lamp's rhythm and the camera's rhythm combine into a slow beat. Two fast processes make one slow, visible pattern.

Then comes the rolling shutter. Most phone sensors don't expose the whole frame at once. They capture rows in sequence, so each row records a slightly different moment. Under steady light that doesn't matter. Under a flickering LED, one band of rows may be exposed during a bright phase and another during a dim phase, and the picture fills with horizontal stripes. Depending on the timing relationship, the stripes sit still or drift. They aren't necessarily a defect. Each group of rows faithfully recorded its own moment, and the light was different at those moments. The image looks wrong because we expect a frame to be one instant of evenly lit reality, and with a rolling shutter it isn't.

Exposure time matters too. A short exposure catches a small slice of the cycle and exaggerates variation. A longer one averages over several cycles and smooths it, though longer exposures change motion blur and brightness balance. Many cameras offer anti-flicker modes or 50 hertz and 60 hertz settings that coordinate exposure with common lighting frequencies. They help with mains-related ripple, but modern drivers often use their own switching frequencies unrelated to the mains, so the regional setting is not a universal cure.

Signs, Screens, and Slow Motion

Scanning explains why LED signs look broken on video. Large displays often energize rows or groups in rapid succession instead of lighting every LED continuously. To your eyes the scan blends into a complete image. A rolling-shutter camera may catch different sections at different points of the sequence, producing dark bands or missing patches. The sign isn't failing. The camera is revealing a timing pattern your eyes normally blend.

Monitors raise a common mix-up. A refresh rate of 144 or 165 hertz says how often the image updates, not whether the backlight turns on and off that often. An LCD backlight may be steady, PWM-dimmed (some displays lean on PWM more at low brightness), or strobed, and OLED panels still need drive electronics that can modulate their light. You can't deduce flicker behavior from the refresh rate.

DC dimming, which lowers the drive current instead of pulsing the light, avoids some modulation, but color, efficiency, and uniformity shift with current. PWM is popular because it gives precise, repeatable control and integrates easily with digital electronics; manufacturers aren't using it because they want flicker. Higher modulation frequency helps, but depth and waveform matter as much: a lamp varying between 100 and 95 percent is very different from one swinging between 100 percent and darkness at the same frequency.

Slow-motion video is a useful demonstration. A high frame rate can show pulsing that ordinary video hides. But the same limits apply. A lamp modulated much faster than the recording can resolve may alias or look steady, and rolling shutter still complicates things. A phone camera is a clue, not a laboratory instrument. Bands tell you the light is interacting with the camera's timing. Their absence doesn't prove the light is flicker-free, because exposure, frame rate, processing, and anti-flicker features can hide or exaggerate modulation, and different phones will show different results under the same bulb.

When It Matters Beyond Your Phone

Camera trouble isn't just cosmetic. The U.S. Department of Energy's review of temporal light modulation groups the visible effects into three kinds. Direct flicker, seen with a steady gaze, typically occurs between about 3 and 90 hertz, with sensitivity peaking around 10 to 15 hertz, where modulation below 1 percent can be visible to an average observer. The stroboscopic effect needs a moving object in view and is generally recognized from about 80 to 2,000 hertz. The phantom array effect appears when your eyes sweep across a modulating source and can, in high-contrast conditions, be visible up to around 11 kilohertz.

The stroboscopic effect is easy to picture. Light a spinning fan with flashes whose timing interacts with its speed, and the blades can appear slow, frozen, or turning backward. The fan hasn't changed; the light is revealing it at particular positions. The phantom array is the dotted trail you may see when you flick your eyes past a rapidly modulated light, a line of separate flashes instead of a smooth streak. These show that flicker effects aren't limited to a stationary person watching a lamp pulse.

Measurement is messier than "flicker-free" suggests. Percent flicker compares maximum and minimum light levels, but ignores frequency; the flicker index accounts for waveform shape but not frequency. The ANSI/NEMA standard on temporal light artifacts, C82.77-6, uses a flicker measure for frequencies below about 80 hertz and a stroboscopic visibility measure for roughly 80 to 2,000 hertz, and states that the phantom array effect is out of its scope. The DOE review notes that no metric yet reliably predicts phantom array visibility and calls the IEEE 1789 recommendations controversial, too lax above 400 hertz. So a "flicker-free" label without a stated method, operating condition, and threshold is ambiguous. A product may look steady yet contain measurable high-frequency modulation, or behave differently when dimmed.

Health claims need care too. Avnet's explainer lists annoyance, eyestrain, and headaches, and says the most serious risks, migraines and seizures, are triggered in the 3 to 70 hertz range. Not all PWM lighting is dangerous, but for people with migraines or photosensitive conditions, well-controlled lighting is a reasonable choice.

Machines have a stricter problem than people. All About Circuits reports that modern traffic lights use PWM LEDs that switch too quickly for people to see, and short camera exposures can miss the pulses, distorting the captured signal. That "LED flicker" can affect driver-assistance systems. The mitigation is called LED flicker mitigation, which uses specialized exposure timing and sensor designs to capture rapidly pulsing lights more reliably, often by collecting light over longer effective exposure periods so the switching averages out. But longer exposure risks saturating the sensor and losing dynamic range, so sensor makers add per-pixel charge storage to keep both. The same logic applies to brake lights and to stadium lighting captured by high-speed broadcast cameras, where modulation that no spectator notices appears as banding in slow motion.

Dimmers, Smart Bulbs, and Failing Drivers

Dimming is where LED flicker most often shows up at home. Traditional wall dimmers were designed for incandescent loads and chop portions of the AC waveform. A filament's heat capacity smooths out the resulting power changes in its light. An LED responds quickly and depends on its driver, which has to interpret that altered waveform. Poor compatibility can cause flickering, buzzing, a limited dimming range, or flaky behavior, even when the bulb is perfect at full brightness. Avnet also points to low-frequency PWM in poor dimmers as a common culprit. If a bulb only misbehaves on a dimmer, check the bulb and dimmer specifications before blaming LEDs.

Smart bulbs add a microcontroller, a radio, and often separate color channels that may each use PWM, so a bulb that is steady at bright white may behave differently at a dim color.

A bulb that visibly pulses slowly or restarts is a different story from camera-only modulation; a degraded driver may be failing. Replace it rather than repairing mains-connected electronics.

How to Choose

Eliminating modulation entirely isn't realistic, and tiny measurable variations don't automatically mean visible or harmful effects. The goal is well-controlled output over the lamp's intended range. When shopping, people look at lumens, color temperature, wattage, and lifespan. Driver quality deserves a place on that list: stable current control, sensible dimmer compatibility, and, for filming or sensitive workspaces, credible measurements of temporal behavior. A good LED product is a complete electrical and optical system, not just a good semiconductor.

The Bard's Take

An LED can look perfectly steady while its light output is changing rapidly. LEDs respond almost instantly to current, and the electronics driving them may produce variation deliberately or by accident. Sometimes it's leftover ripple from turning household AC into regulated current. Sometimes it's PWM, where the light is switched rapidly to set its average brightness. A 50 percent duty cycle doesn't necessarily mean a lamp glowing at half intensity; it may mean a lamp that is lit half the time and dark the other half.

Your visual system fuses those rapid changes into a continuous glow. A camera sees differently. Its sensor collects light during specific exposure windows, often one row at a time, and when those windows interact with the light's modulation it records bands, pulses, and patterns your eyes never noticed. The light didn't change because you pointed a phone at it. You introduced a second timing system that can reveal what was already happening.

That matters beyond phone videos. Temporal light modulation touches visual comfort, makes moving objects look strange, complicates high-speed filming, and challenges cameras in cars and factories. Good lighting is more than enough lumens at a pleasant color. It's also control over how that light changes in time.

Electricity changes. Semiconductors respond. Light output rises and falls. A camera records alternating bands. And your eyes keep seeing a steady lamp. The LED may be flickering. Your visual system is just very good at making rapidly changing light look continuous.

Sources

← Back to the Desk — Technology