The display is the part that behaves

An LED module is a piece of silicon and plastic with no moving parts. In a kinetic screen it is a passenger. The mechanism moves it, the cable feeds it, and both of those have a wear life measured in cycles.

In the field the failure order is boringly consistent: cable first, home switch second, brake third, gearbox fourth. The LED driver board comes last, and usually because water got in through a seal the mechanism worked loose.

An outdoor flower-shaped LED sculpture whose twelve petals open on mechanical hinges
Twelve petals on hinges, eighty-four different boards behind them. The mechanism is the simple part. The spare-parts list is what takes the meeting.

Cables die by bending, not by current

A cable chain has a rated life, typically five to ten million cycles, and that rating assumes you respect the minimum bend radius, usually seven to eight times the cable diameter. Route a bundle at half that radius because it looked neater and the rating is gone.

Do the arithmetic before you accept the number. Ten moves an hour, twelve hours a day, is 43,800 cycles a year. A five million cycle chain is a hundred years of that, and yet chains fail in year two, almost always at a radius or a fixed clamp where the cable was never allowed to relax.

Two positions or infinite: the decision that sets the budget

If the shape only has states, closed and open, or three set positions, then render one pixel map per state and switch maps with the mechanism. This is the cheapest and sharpest option, and it covers most kinetic walls and opening flowers.

Continuous motion is a different project. Every frame needs geometry mapping in real time, the media server spec goes up a tier, the content budget doubles, and any latency between the mechanism and the video shows up as a visible lag between a panel and its picture.

The centre of gravity moves, so the torque does too

Push a column of modules half a metre forward and the load stops being a wall and becomes a cantilever. Motor torque has to be sized for the worst position, not the rest position, and the frame has to take the bending moment when half the panels are out.

Outdoors it gets worse. A closed screen might present 0.5 kN per square metre of wind area. Open the petals and the same unit presents two to three times that, on arms that are now levers. A mechanism rated for the closed state is not the same machine when it is open, and the wind load case has to be written for both.

Nobody budgets the homing routine

Cut the power halfway through a move and the controller has no idea where the mechanism is. On restart it has to find a reference. Absolute encoders cost money per axis. Limit switches plus a hard stop cost less, and every axis gets jogged into its stop on first power-up.

That first hour is slow and slightly nerve-racking. Each arm goes to its stop alone, the offset is written into the controller, and the sequence is saved. Save the file, copy it to a USB stick, and tape the stick inside the cabinet. The second power cut is the one that finds out whether anyone did.

Safety is a design item, not a sticker

Panels that close have pinch points at hand height, which is exactly where visitors put their hands. The workable combination is a light curtain or a pressure edge on the closing side, an E-stop chain that kills motion and not just video, and a two-speed profile that travels fast and covers the last 50 mm slowly.

Add an audible warning two seconds before movement. It sounds like a small thing until the first time a child walks into the gap.

Service access when the shape is closed

A kinetic screen spends most of its life in one position, and that position is usually the one that blocks access to the modules behind it. Build a service state into the controller: one button that opens a specific bay and holds it there under a lock.

Then check the spare. A module at three metres, reached from a platform, with a mechanism that must not move while a hand is inside, is a two-person job. Price the spare parts around that, not around a flat wall.

The week that decides whether it ships on time

Five hundred dry cycles before any video runs. Eight hours of thermal soak afterwards, because gearboxes warm up and clearances change. Brake release timing measured on every axis. Then sync: run a vertical bar across the whole surface at slow speed and watch whether the picture and the panels arrive together.

If that week is not in the schedule, it happens during the client's opening event instead. The pixel pitch will not be what anyone remembers about the project.