The lights look perfect to your eye, and the camera disagrees. Bands crawl up the wall, the slow-motion shot strobes, white turns green halfway through a shot, and nobody on set can see any of it happening.
Nothing is broken. A camera samples light in short slices, and an LED dims by switching on and off very fast. When those two rhythms do not line up, the camera records a rhythm the eye cannot see. This is a timing problem with a handful of known fixes.
In this guide
An incandescent lamp is a hot piece of wire. Mains power in North America alternates 60 times a second, so the light output actually pulses 120 times a second — twice per cycle — but the filament cannot heat and cool that fast. Its thermal inertia smooths the pulses out, and the eye receives something effectively steady.
An LED has no filament and almost no thermal inertia. It responds to its drive current almost instantly, in microseconds. Whatever the driver does to the current, the light output follows exactly. That responsiveness is why LEDs are efficient, dimmable and colour-controllable — and it is also why they can emit light in rhythms that a camera will pick up and an eye never will.
The eye integrates light over roughly tens of milliseconds and is fairly forgiving about short gaps; a gap shorter than that is simply averaged away and felt as slightly dimmer. A camera shutter is far more specific. It opens for an exact interval, records whatever arrives during that interval, and closes. If the interval happens to catch a light that was off, the camera records a dark frame — even though the lamp was on for most of the second.
Pulse-width modulation dims an LED by switching its full current on and off rapidly, then relying on the eye to average the result. Run the LED on for 50% of each cycle and it looks half as bright. Run it on for 10% and it looks nearly off. The eye sees a smooth dim; the camera sees a square wave.
The rate of that switching — the PWM frequency — is the number that decides whether a fixture is safe on camera. Everything else in this article follows from it.
| PWM frequency | Cycle length | Ordinary video at 24–60 fps | High-speed / slow motion |
|---|---|---|---|
| 100 Hz | 10.00 ms | Risky — cycle can exceed the exposure | Unusable |
| 500 Hz | 2.00 ms | Work if shutter is long enough | Bands visible |
| 1 kHz | 1.00 ms | Usually acceptable | Bands likely |
| 2 kHz | 0.500 ms | Acceptable | Bands visible |
| 4 kHz | 0.250 ms | Clean | Marginal at high speed |
| 8 kHz | 0.125 ms | Clean | Acceptable at moderate speed |
| 16 kHz | 0.062 ms | Clean | Good up to roughly 240 fps |
The frequencies and cycle lengths in that table are exact reciprocals — 1 kHz is exactly 1 ms, 16 kHz is exactly 0.0625 ms — and they are worth learning because every calculation below uses them. Note the shape of the table: the useful range is not a gradient but a cliff. A fixture is fine until the exposure gets short enough to expose a single dark part of a cycle, and then it is not.
A camera records a frame by opening its shutter for a defined interval. On a video camera that interval is normally expressed as a shutter angle, inherited from film. A 180° shutter — the standard, and what you get if you never touch the setting — exposes the sensor for half of each frame period. So the exposure time is simply 1 ÷ (2 × frame rate).
At 24 fps a 180° shutter exposes for 20.833 ms. At 30 fps, 16.667 ms. At 60 fps, 8.333 ms. Those numbers set the whole problem: the exposure window is how long the light gets to be wrong before the camera notices.
| Frame rate | Shutter angle | Exposure time | PWM needed for 1 full cycle |
|---|---|---|---|
| 24 fps | 180° | 20.833 ms | 48 Hz |
| 30 fps | 180° | 16.667 ms | 60 Hz |
| 60 fps | 180° | 8.333 ms | 120 Hz |
| 120 fps | 180° | 4.167 ms | 240 Hz |
| 240 fps | 180° | 2.083 ms | 480 Hz |
| 480 fps | 180° | 1.042 ms | 960 Hz |
| — arithmetic check — | 90° at 30 fps | 8.333 ms | same as 60 fps at 180° |
The last column is the useful one, and it is easy to misread. It is not the frequency you need. It is the frequency at which the exposure window contains exactly one complete PWM cycle — the bare minimum, with no margin, at which the camera can still sample a full cycle. Give yourself a margin of roughly ten cycles inside the exposure and the requirement multiplies by ten: about 480 Hz for 24 fps, 600 Hz for 30 fps, 1,200 Hz for 60 fps, 2,400 Hz for 120 fps, and about 4,800 Hz for 240 fps.
One cycle means the exposure captures exactly one on-and-off pattern — mathematically the minimum, but it sits right at the edge. Real fixtures drift slightly, real frame rates are never exactly nominal, and real exposure timing has tolerances. Ten cycles inside the exposure absorbs all of that and averages the on-time properly, which is also what makes the brightness correct and not just flicker-free. This is why 480 Hz is a much better target than 48 Hz.
The calculations above would be the whole story on a camera with a global shutter, which exposes every pixel at once. Almost every consumer and prosumer camera, and most mirrorless bodies, use a rolling shutter instead: the sensor reads out line by line, from top to bottom, over a finite interval typically around 15–30 ms.
That means different rows of the image are exposed at different moments. If the light is changing during the readout, each row catches a different point in the PWM cycle, and the row records a different brightness. The result is horizontal bands — and because the pattern shifts slightly from frame to frame, the bands appear to crawl.
That chart also explains why higher PWM frequency is still the right fix even though it increases the band count. Twelve coarse bands across a frame are an obvious, objectionable pattern. Several hundred fine bands are individually too thin to resolve, and they average out into smooth light. The goal is to push the banding below the sensor’s resolving power, not to eliminate the underlying switching.
Bands show up most clearly on a plain flat surface filling the frame — a cyclorama, a wall, a seamless backdrop. They are almost invisible on a busy stage with moving performers and dark backgrounds. This is not a reason to accept a flickering fixture, but it does explain why two people can look at the same rig and disagree: one was shooting a wide shot of a lit wall, the other was shooting a close-up against black.
Slow motion is shot by running the camera at a high frame rate and playing the result back at a low one. A 120 fps capture played at 30 fps gives four times slow motion. But a 180° shutter at 120 fps exposes for only 4.167 ms — and the faster the capture, the shorter the exposure becomes.
Shorter exposure is exactly the condition that exposes PWM. The window that used to contain hundreds of cycles now contains a handful, and the flicker that was averaged away reappears as a visible brightness pulse in the slow-motion clip.
The practical rule that follows: decide the highest frame rate the shoot will use, then work backwards to the PWM frequency you need. Shooting a normal 30 fps interview and one 240 fps slow-motion insert means the rig has to satisfy the 240 fps case, which needs a PWM frequency roughly eight times higher than the interview alone would have demanded.
Flicker is only one of the ways LEDs misbehave on camera. A second class of problem appears without any banding at all: the colour changes as the fixture dims, or white drifts across the frame.
There are three common causes, and they have different fixes. Shallow dimming — where the fixture only reaches a modest dim depth before it stops or jumps — happens when the driver cannot hold a very short on-time stably. Colour drift at low levels happens when several emitters fall out of proportion; an RGB fixture relies on three or four separate dies dimming in step, and small differences become visible as a shift toward one colour. And a camera disagreeing with the eye is often a white-balance issue rather than a fixture issue at all: cameras and eyes apply different white references, and a colour temperature that looks right in the room can read differently on a screen.
The colour row deserves one extra note. A fixture rated for high colour fidelity is not automatically stable through its dim curve, and a fixture with excellent dimming is not automatically accurate in colour. They are separate engineering problems. When a shoot cares about both — skin tones on camera, dimmed low for a moody look — that combination is the one to test explicitly before the shoot day rather than discover on it.
The whole test takes five minutes, and it needs nothing but the camera you will actually shoot with. Do it once per fixture type and you will never be surprised on set again.
A modern phone camera in slow-motion mode is an excellent flicker detector. Point it at the fixture, record at its highest frame rate, and scrub through the clip frame by frame. Bands and pulsing that are invisible in real time show up immediately. It is not a calibrated measurement, but it reliably distinguishes “this fixture is clean” from “this fixture will embarrass us on the hero shot”.
When flicker appears, there is a fixed order of remedies from cheapest to most expensive. Work down it rather than reaching for the last one first.
Fixtures built for this problem state the numbers directly. The ones below publish their dimming specification rather than leaving you to discover it on set.
STUDIO LUX FN400
$768
DUALTONE 200ZM
$241
DT 200W COB
$169
SPOTTY LED 150
$519
Because the eye and the camera sample light over completely different intervals. The eye integrates over tens of milliseconds and averages short gaps into a slightly dimmer impression, so switching that happens faster than that is invisible. A camera shutter opens for an exact, often much shorter interval and records whatever arrives in it. If that interval catches the light switched off, the camera records darkness where the eye saw steady light.
Pulse-width modulation dims an LED by switching its full current on and off very rapidly and letting the eye average the result. It is the standard method because it is efficient and holds colour stable across the dim range. The switching rate — the PWM frequency — is typically between a few hundred hertz and a few tens of kilohertz. Below a certain frequency relative to your shutter speed, the camera starts resolving the on and off states instead of averaging them.
Compare the PWM cycle length against your exposure time. As a bare minimum the exposure should contain one full cycle, which at 24 fps with a 180° shutter means about 48 Hz and at 240 fps about 480 Hz. That is the mathematical edge, not a working target. Allow roughly ten cycles inside the exposure for real-world margin: about 480 Hz at 24 fps, 1,200 Hz at 60 fps and 4,800 Hz at 240 fps. In practice, fixtures rated in the kilohertz range handle ordinary video comfortably.
That is rolling shutter. Most sensors read out line by line over an interval of roughly 15 to 30 milliseconds rather than exposing every pixel at once, so different rows sample different moments in the PWM cycle and record different brightness. The fix is the same as for other flicker — higher PWM frequency or a longer exposure — because a faster dimming rate pushes the bands closer together until they are too fine to see.
Slow motion is captured at a high frame rate, and a 180° shutter exposes for half a frame period. At 240 fps that is only 2.083 ms, eight times shorter than at 30 fps. A shorter window contains fewer PWM cycles, so the switching is sampled instead of averaged and appears as a visible pulse. Always test a fixture at the highest frame rate the shoot will use, not the one you happen to be shooting today.
Often, partly. Lengthening the shutter angle from 90° to 180° doubles the exposure window and suppresses most banding at no cost except motion blur, which is usually the desired look anyway. Adjusting the frame rate can stop bands from crawling. Neither fix repairs a fixture whose dimming frequency is genuinely too low for the shot — that needs a different fixture, or a different shot.
Set the camera to the highest frame rate and shortest shutter the shoot will use, point it at a plain surface lit by that fixture alone, then dim the fixture slowly through its whole range while watching the monitor. Repeat with the rest of the rig running, because interaction between fixtures is a real effect. A phone in slow-motion mode is a surprisingly effective detector if you do not have the camera to hand.
No — they are separate properties. CRI describes how faithfully the fixture renders colour. Flicker behaviour comes from the dimming driver. A fixture can have excellent colour fidelity and a dimmer that bands badly, or flawless flicker-free dimming and a colour response that shifts as it dims. If a shoot depends on both, the dimming specification and the colour specification each need checking independently.
This article explains the interaction between PWM dimming, camera shutter timing and rolling shutter readout. Exposure times and PWM cycle lengths are exact reciprocals and can be relied on; sensor readout times, band counts and the three readout figures used in the banding chart are representative values chosen to illustrate the relationship, not measurements of any specific camera — manufacturers rarely publish readout time, and it must be established for each body. The frame-rate and shutter-angle guidance describes general practice and should be verified on your own equipment before a shoot. Published figures for products sold on sanyilights.us are quoted from the current datasheet at the time of writing. Product availability and pricing are subject to change.
