An LED wall that looks perfect to the audience can fall apart the moment a camera points at it: rolling bands, shimmering moiré, banding in the blacks. None of it is mysterious — a sensor with a shutter samples the wall differently from an eye — but every one of these problems has to be designed out before the truck leaves, because none of them can be fixed at the mixer. This is the checklist working broadcast and xR crews run through.
Two different numbers get called “refresh”. The video feed runs at a frame rate — 50 or 59.94 or 60 Hz. The LEDs themselves are pulse-width modulated many times faster than that, and the datasheet figure for that is the refresh rate: 1,920 Hz on general-purpose panels, 3,840 Hz on camera-grade panels, 7,680 Hz and up on the high end. The eye never sees the difference. A camera shutter does.
A fast shutter is open for a small slice of each frame. If the wall’s PWM cycle is slow relative to that slice, the sensor catches different parts of the modulation across its scan — dark bands rolling through the picture, or a brightness flicker that changes with shutter angle. The working baseline for anything that will be shot is 3,840 Hz; faster shutters and high-speed work want more. The figure is on the datasheet, so it is checkable before a single panel is booked.
Even with high refresh, a wall that is free-running against the camera produces a slow beat: bands that drift through frame over seconds. The fix is to lock the processor’s output to the same reference the cameras use — genlock — so the wall’s frame timing and the shutter open at a fixed phase relationship. Broadcast processors accept a reference input for exactly this reason.
Run the wall at the camera’s frame rate, or an exact multiple of it, never at a nearby rate that “looks fine”. On multi-camera shows the whole chain shares one reference. Processors also offer shutter-matching modes that shift the wall’s output timing to sit inside the shutter window; useful, but no substitute for locking the rates in the first place.
Moiré is the interference pattern that appears when a camera sensor’s own grid beats against the LED pixel grid — a shimmering, colour-shifting texture that moves as the camera or focus moves. It is worst when the projected pixel spacing on the sensor lands near the sensor’s own pixel spacing, which is why it comes and goes with focal length, distance and focus.
The viewing-distance rule of thumb — one metre per millimetre of pitch — describes an eye, not a sensor. Walls that will be shot are specified finer than that rule alone suggests, and the tighter the framing, the finer the pitch. Virtual-production and xR stages, where the camera works close to the wall, commonly run in the 1.5–2.8 mm range; the ROE Black Pearl BP2 at 2.84 mm became the standard-bearer for that work, with the Ruby series at finer pitches following it. The practical mitigations, in order: more distance, a touch of defocus on the wall, a change of framing or angle, and finally a finer panel.
Cameras need a fraction of the brightness an audience does. Running a wall at 10–20% on set is normal — but many panels lose colour precision at low brightness, because dimming with PWM leaves fewer discrete steps in the dark end of the range. The result is banding in gradients and blocky shadows that the eye at full brightness never showed. High-bit-depth pipelines (10-bit and up) are the answer, and they cost bandwidth: NovaStar’s own datasheets halve port capacity at 10- and 12-bit, so a camera wall needs the processor sized for it.
Black is not black. An unlit LED panel is a grey object that reflects set lighting, and the camera sees that lift where the audience does not. Matte masks and louvred faces help; so does keeping key and fill off the wall surface. Match the wall’s white point to the camera’s white balance rather than to what looks right in the room.
Scan rate is the multiplexing ratio — how many rows of LEDs share one driver, written 1/8, 1/16, 1/32. Higher multiplex saves cost but each row is lit for a smaller fraction of the cycle, which lowers brightness headroom and makes the panel more fragile under fast shutters. Panels built for camera work run lower multiplex or compensate with much higher refresh; the datasheet lists both.
Camera positions are rarely on axis. Colour and brightness shift as the viewing angle opens, and different panels shift differently — the datasheet viewing angles are the starting point, but a wide-angle camera position deserves a look at the actual panel before the show.
Shoot the wall with the actual camera, lens and shutter settings at the planned distances before the show, and look at the recording, not the monitor. Put up full white, full black, a mid-grey gradient and the real content. Check for bands at every shutter angle the director will use, moiré at every focal length, banding in the gradient, seams and flatness between cabinets, reflections from set lighting, and — for xR — the latency between camera tracking and wall.
Content should be mapped pixel-for-pixel; scaling on the way in softens edges and can itself alias against the LED grid. Every item above is on the wall’s datasheet or provable with a five-minute test, which is why a camera wall that surprises anyone on show day is a planning failure, not bad luck.
RigPlan does this arithmetic for you. Enter the wall once and it derives panels and spares, rigging from your own stock, circuits, the data wiring plan and the full pack list — free, no sign-up.
Open the planner3,840 Hz is the working baseline for anything that will be shot; general-purpose 1,920 Hz panels can band under fast shutters, and high-speed work wants 7,680 Hz or more. This is the PWM refresh on the datasheet, not the 50/60 Hz frame rate of the feed.
Either the PWM refresh is too slow for the shutter, or the wall is free-running against the camera without genlock. Lock the processor to the camera reference, run the wall at the camera frame rate, and use a panel rated 3,840 Hz or better.
Finer than the eye rule alone suggests, and finer the closer the framing. IMAG and broadcast walls commonly sit around 1.5–2.6 mm; xR stages in the 1.5–2.8 mm range. Moiré depends on lens and distance as much as pitch, so a test shoot beats any rule.