The drawing shows a 12 m wall with a gentle bow in the middle, three segments, keynote stage. The rental quote comes back with 24 cabinets. On site, the centre row is one cabinet short and the crew solves it by widening the black border at both ends, which is the moment the client notices the screen does not reach the set.
Nothing unusual happened. The quote was built on the chord length of the wall. The screen has to be built on the arc.
The 2.6% nobody puts in the quote
Take a 12 m chord with a 1.2 m rise at the centre - a typical "gentle curve" requested for a keynote backdrop.
- Radius: R = (c^2 / 4 + s^2) / (2s) = (36 + 1.44) / 2.4 = 15.6 m
- Half-angle: asin(6 / 15.6) = 22.6 degrees, so the wall turns 45.2 degrees in total
- Arc length: R x theta(rad) = 15.6 x 0.789 = 12.31 m
A 500 mm cabinet gets you 24 cabinets on the chord and 25 on the arc. One cabinet, at the exact point where the audience's eye sits. Every curved wall has this gap between the plan view and the build, and it is always resolved by either a bigger black border or an extra column of cabinets nobody ordered.
If you are quoting, quote the arc. If you are approving, ask which number the count came from.
Minimum radius is a hinge decision, not a product label
Every supplier says "curved cabinet". What matters is the index step - how many degrees each cabinet-to-cabinet joint can take, and whether it locks there.
A 500 mm cabinet indexed at 5 degrees per joint gives a minimum radius of 0.5 / (2 x sin 2.5 degrees) = about 5.7 m. Step it to 10 degrees and you can get down to about 2.9 m, but now each cabinet is a flat facet and, at 4-6 m viewing distance, the client can count them. That is the real trade: softness of curve versus how close people sit.
Two ways out, both with a cost:
- Narrower cabinets. 250 mm cabinets at 5 degree steps get you to about 2.9 m with a smoother surface. Twice the cabinets per metre, twice the joints, twice the tolerance stack.
- A fixed-curve cabinet. Lovely geometry, and it only builds one radius. If your inventory already carries two radii, a third is a purchase, not a rental.
The AIRD3.91-H is a 500 x 500 x 65 mm P3.91 rental cabinet built for curved builds at 1000 cd. Before you quote it against a 3 m radius design, read the index step on the datasheet and do the 0.5 / (2 sin theta/2) sum yourself. It is a two-line calculation and it decides whether the design is buildable at all.
Where you put the two breaks matters more than the curve
The request is almost always "three segments": a curved centre, two flat wings at the ends. Two decisions follow immediately, and both belong to the set designer, not the LED supplier.
Put the break on a fold. If the flat wing meets the curve mid-gradient, the eye reads a crease in the picture. If it meets at a hard vertical graphic edge - a column, a title bar, a black band - the joint disappears.
Budget two extra data runs and two extra ports. Three segments is not one wall with bends in it. It is three surfaces, usually fed from three directions. That means sending-card ports, spare runs, and a processor that can address three canvases. Ask who is paying for those before the rig plan is signed.
The geometry that always fails is the "one big flat canvas, let the processor bend it" approach. A curve built from flat facets is a polygon, not an arc. A 16:9 render mapped across it is compressed at the outer facets and stretched through the centre. The content does not look broken - it looks slightly wrong, which is worse, because nobody can name the problem in the review.
The tolerance stack is the thing that actually gets delivered
25 cabinets, 24 joints. If each joint is within +/-0.5 mm - a normal build tolerance - the worst case is 12 mm of cumulative drift across the wall. You will not see it in a photo. You will see it on camera at an angle, as a wavy line along the top edge, and you will see it in the seam where the wing meets the curve.
Three things that hold the line, in order of cost:
- Level the top edge with a laser, not a tape. On curved builds the top rail is where drift concentrates.
- Pre-build the curve flat on the deck. Assemble, check the joint angles against the drawing, then fly it. Re-indexing a joint at 6 m in the air is a two-hour job with a scissor lift.
- Keep the outer facets free of text. They are at the worst viewing angle on the wall. Logos and lower-thirds belong on the flat wings.
Rigging: the centre of the wall leans forward
A curved wall is not a straight wall with a bend. Its mass sits forward of the truss line, and the centre of the arc is the furthest forward point. On a flown build that means the pick-up points at the centre carry a forward moment, and the top rear braces have to be set out to correct it rather than just square.
The symptom of getting this wrong is the bottom row lifting 3-5 mm off the deck at the centre, or a visible gap at the deck line once the wall is flown. It is also why the deck plan needs the curve drawn on it - the base line of a curved wall is an arc, not a straight line, and the stage set has to be built to the arc.
The decision, honestly
For a one-night keynote, three segments is often the wrong answer twice over: custom content, custom steel, and a load-out that takes an extra hour. The comparison that matters:
| Option | Cabinets (12 m) | Content | Load-out | Looks like |
|---|---|---|---|---|
| Flat wall, 2 angled wings | 24 + 2x6 | Reuse the flat master | Fastest | 70% of the curve, from the front row |
| 3-segment curve, indexed | 25 + 2x6 | Re-rendered per segment | +1 hour, +1 lift | The design as drawn |
| Full smooth arc, 250 mm | 50 + wings | Re-rendered, tighter mapping | +2 hours | Best on camera |
If the client is filming and the wall is the hero, pay for the curve. If it is a two-camera stream with a speaker in front of it, the flat wall with angled wings reads identically on screen and goes back in the case an hour earlier.
The question worth asking in the production meeting is not whether the arc is beautiful. It is whether anyone will be standing far enough back to see it.
Three-segment curve, or flat with angled wings - which would you build for a keynote?
I work with MuenLED, a Shenzhen LED manufacturer. The cabinet figures above are from the published AIRD3.91-H specification; the geometry and tolerance stack come from quoting and building these walls.
