Shooting
an LED wall

Moiré is a sampling problem, not a screen fault

A wall is a grid of emitters. A sensor is a grid of photosites. Point one at the other and you are sampling a regular pattern with another regular pattern — and when the two spacings land close to a simple ratio, the difference between them shows up as slow bands that crawl when anything moves. That is aliasing, and the wall is innocent.

It follows that anything changing the ratio helps, and anything changing the content does not. Repainting the graphic will not fix moiré. Moving the camera might.

What changes the ratio

Move	What it does	Cost
Change distance	Alters how many panel pixels land on each photosite. Usually the fastest fix: a metre either way often clears it.	Framing changes.
Change focal length and re-frame from a new spot	Same effect, different route — same shot, different sampling.	Perspective changes.
Open the aperture	Shallower depth of field, so the wall falls slightly out of focus and the grid is optically blurred before it is sampled. This is the classic fix and it looks intentional.	Exposure, and the wall goes soft.
Come off axis	Rotates the sensor grid relative to the panel grid. A few degrees is often enough.	Angle on the wall, and LED brightness falls off axis.
Defocus slightly	Blunt but effective on a background wall nobody is meant to read.	Obvious if there is text on it.

The other banding, which is not moiré

Horizontal bands that roll through the frame are the panel's refresh, not its pixel grid. An LED wall does not hold a steady light level: it strobes, at a scan rate that may be thousands of hertz, and a camera shutter that is not a whole multiple of that period catches some strobes and misses others. The fixes are different:

• Shutter angle or speed — walk it until the bands stop. 1/50 and 1/60 are the usual landing points, and a 180° shutter at the frame rate you are shooting is the first thing to try.
• The wall's own scan and refresh settings — a processor set to a higher refresh, or to a camera-friendly scan mode, fixes it at source. This is the right answer if the wall will be on camera all day, and it is a five-minute job for whoever owns the processor.
• Genlock if the chain supports it.

Tell the two apart by moving the camera: moiré from the pixel grid moves with the camera, scan banding does not.

What the wall does to your exposure

• It is a light source, not a surface. It does not take key light; it throws it. Expose for the talent and the wall will usually be hot, which is why walls are run at 30–60% brightness for camera rather than the full output they are sold on.
• Black is not black on camera. A wall's floor is a dark grey that a sensor lifts further, so a "black" background often reads as a smoky panel with visible seams. Check it on the camera, not with your eyes.
• Colour shifts off axis. LEDs are directional; a camera 40° off centre does not see the colour the front row sees. Grade to the camera's position if the camera is the audience.
• Watch for a green or magenta cast that only appears on camera. Sensors and LEDs disagree about spectrum, and the fix is a white balance shot off the wall itself, not off a card in front of it.

Staging: the decisions made before anyone lifts a lens

• Height and rake. If the wall is above head height, everyone is looking up and the camera is looking up with them — which puts the panels' worst viewing angle in the shot. Lower is nearly always better on camera.
• Distance to the nearest person. Inside about 3.4 × pitch metres the pixels are individually visible. The calculator gives the number for a given wall; if the front row is inside it, that wall is scenery, not a screen.
• Ambient spill. Stage wash landing on the panels raises their black level far more than it raises anything else in shot. Flag the wash off the wall before reaching for the processor.
• Where the seams fall. Panel joins are invisible in motion and obvious on a flat colour, so put flat colour where the seams are not: a gradient, a texture or a slow move hides a join that a plain background advertises.
• Content at the wall's own raster. Not 1080p scaled up. Scaling is what softens the one-pixel detail moiré loves and text needs.

On the day, in this order

1. Prove the pixel map before anything else. Put up a numbered panel map at the wall's exact raster. A swapped tile shows the wrong number, a rotated one shows the corner marker in the wrong corner, and a dead receiver card shows nothing. Five minutes here saves an argument at rehearsal.
2. Check 1:1 mapping. A one-pixel checkerboard should look like a crisp checkerboard, not a soft grey field. If it is soft, something in the chain is scaling and every later judgement is being made through it.
3. Set black level. Dark-detail steps, with the house lights at show state, looking at the camera's picture rather than the wall.
4. Check uniformity. Flat fields at 25, 50 and 100 per cent. Mismatched batches and dead pixels show on flat colour and nowhere else.
5. Then move the camera for moiré: distance first, aperture second, angle third.
6. Then walk the shutter for scan banding.
7. White balance off the wall and grade to the camera's position.
8. Only now put the show content up. Every earlier step is quicker to diagnose on a pattern than on a picture somebody has opinions about.

Patterns for each of those steps

Free, generated at your wall's exact size, nothing to sign up for: panel map, 1:1 moiré tests, dark detail, uniformity, boresight, crosshatch, SMPTE bars, PLUGE, convergence, multiburst and safe areas. If you would rather have a fixed 1920×1080 set to keep on a stick, there is a kit.

Deckboy plays them, and the show, from one cue list — free and open source for Windows, macOS and Linux, at whatever raster the wall actually is, with a moving frame counter and latency clock for the faults a still cannot show.