A rig that looks flawless from the seats can be unusable on a recording, and the reason is that eyes and sensors integrate light differently. The eye averages over a comparatively long window and forgives rapid modulation. A camera samples, and anything that switches faster than it samples gets recorded as pattern rather than as brightness.
Rows are not exposed at the same moment
Most modern sensors read out progressively: the top row begins its exposure, then the next, and so on down the frame. If the light is constant this is invisible. If the light is pulsing several hundred or several thousand times a second, different rows catch different parts of the pulse train, and the frame records horizontal bands of bright and dark.
Those bands rarely sit still. Unless the modulation and the frame rate happen to be in a whole number relationship, the pattern lands slightly differently on each successive frame and appears to roll up or down the picture. The rolling is the diagnostic: a static gradient is usually a lighting design, a crawling one is almost always a timing beat.
Where the pulsing comes from
Three sources account for most of it. Pulse width modulated LED drivers switch the diode fully on and off, and the switching frequency is a design decision that varies enormously between fixtures. Discharge sources on magnetic ballasts pulse at twice the mains frequency, which is one hundred or one hundred and twenty times a second depending on the country. And video walls and screens have their own refresh and scan behaviour, which beats against a camera in the same way and adds moire on top of it.
Tungsten is the exception that explains the rule. A filament is pulsed by the mains just as thoroughly, and its thermal inertia averages the pulses into steady light before they ever reach the sensor. The absence of flicker in the tungsten era was a property of hot metal, not of good engineering practice.
The remedies, in the order they are worth trying
Raise the modulation frequency. Many theatrical fixtures allow the drive frequency to be set, and moving it well above the rates a camera samples at removes the interaction entirely. This is the fix that costs nothing and is most often simply not switched on.
Dim by current rather than by pulsing where the fixture supports it, because a source that is not switching has nothing to beat against. The cost is the small colour shift described in the article on dimming.
Match the shutter to the mains where the flicker is mains derived. Shutter intervals that are whole multiples of the supply's cycle collect the same amount of light every frame. This is why crews working in different regions use different default shutter settings, and why a setting carried across a border stops working.
Change the fixture. Some drivers cannot be moved out of the way, and no camera setting rescues them.
High frame rates make everything worse
Slow motion shortens the exposure window dramatically, so modulation that was safely averaged at twenty five or thirty frames becomes plainly visible at two hundred. A rig signed off during a normal rehearsal can fail on the one shot the production most wanted, and the only reliable approach is to test at the frame rate that will actually be used, through the camera that will actually be used.
The console adds a slower flicker of its own
Beneath the fixture's own switching sits a second rate: how often the control system sends a new value. A full universe refreshes about forty four times a second, which was ample when a filament smoothed the gaps and is marginal for a fast electronic fixture recorded at speed.
The symptom is different from banding and is easy to misread. Instead of stripes, a smooth movement or fade appears to advance in small discrete increments, because the light really is holding each value until the next packet arrives. On a recording this reads as judder in the fade rather than as a pattern across the frame, and no shutter setting removes it. Fewer slots per universe, or a networked control path carrying the same rig across several universes, is what actually changes it.
Flicker is also a health question, not only a picture one
Modulation below roughly a hundred and something hertz has been studied for effects on people rather than on sensors, and a recommended practice published by the Institute of Electrical and Electronics Engineers addresses acceptable modulation depth against frequency for exactly that reason. It is a professional standards document rather than a product claim, which is why it is worth naming. Fixtures chosen for camera performance usually satisfy it comfortably; very cheap drivers frequently do not.
What we cannot verify
Drive frequencies are published by fixture manufacturers when they are published at all, often as a single figure that applies at one dim level and not others, since some drivers change strategy at the bottom of the range. We reproduce no such figure. Claims that a fixture is flicker free are made by the party selling it and mean nothing without a stated frame rate, shutter interval and dim level. The only assessment worth acting on is a test recording made in the venue.
The short version
- Eyes average rapid modulation; sensors sample it and record pattern.
- Progressive readout exposes rows in sequence, so pulsing prints as bands.
- Bands that crawl are a timing beat; static gradients usually are not.
- Raising drive frequency, or dimming by current, removes the interaction.
- Mains derived flicker is handled by matching shutter to supply frequency.
- High frame rates expose modulation that passed at normal speeds.