A seam is a brightness step, not a gap

People say the seam is visible and mean a line. What is actually on the screen is two cabinets whose LEDs sit 3 to 6 % apart in luminance, and the eye reads that step as an edge. The physical gap between cabinets on a 2 m sphere is under a millimetre: you cannot see it from three metres. A 4 % step you can see from eight.

Which is why tightening the hardware does not fix it, and why the fix belongs in the mapping and the calibration rather than in the spanner.

Bands are where it shows first

A sphere is built in latitude bands. The equator band carries the most cabinets, the poles the fewest, and each band comes off a different mould with a different cabinet count. Two consequences follow. The seam between two bands is where two receiving-card batches meet, so any difference in their programming shows along that circle. And the pole seams are seen almost face-on while the equator seams are seen at a glancing angle, so the same 4 % is invisible at the middle and obvious at the top.

This is the reason a client can point at the north pole and swear the whole screen is uneven. Half of what they are seeing is geometry, and it is supposed to be there.

Check the map before touching brightness

Put up a crosshatch test pattern and count modules from a known landmark — the north pole is easiest. If the seam line lands exactly on a module boundary, the geometry is right and you are chasing brightness. If it lands half a module off, or the content runs sideways on one band, the problem is the mapping: a cabinet placed one position out, or a band configured without the 90° rotation its modules need.

Rotation is the one people miss. Modules are mounted rotated on some bands to follow the curve; if that is not declared in the screen configuration, the band lights up correctly and shows its content sideways, and what looks like a seam is a pixel-order mismatch at the boundary.

Level in three passes: band, cabinet, low grey

Work in that order, in SmartLCT, with the room dark. Band brightness first, because that is the coarse step. Then individual cabinets inside the band. Then a 5 % grey field, which is where the eye is most sensitive and where an inexpensive colour meter is least useful: a cheap meter reads 5 % grey as noise, while your eyes read it as a patchwork.

Do not judge the result from a phone camera. A phone lifts the shadows and exaggerates exactly the step you are trying to remove, so you spend an hour correcting a seam nobody at the venue can see, and miss one that shows at the distance people actually stand.

Viewing angle does the rest

LEDs lose brightness off-axis. On a sphere every cabinet faces a different direction, so the surface is never uniformly bright to one viewer: the middle reads bright and the edges fall away. That falloff is correct and should stay. What should not be there is a step at a band boundary, because falloff is smooth and a step is not.

Judge at seat distance, at night, standing where the client will stand. We tell clients the same thing at handover: look from the far end of the room. If a line is still visible from there, it is ours to fix. If it disappears, it was never a defect.

Spares, and the part nobody plans for

Keep spare modules from the same production batch. A module from a different bin, fitted three years later, sits at a slightly different low-grey level and the repair becomes visible in exactly the conditions where seams show: dark room, low content.

We ship spares with the batch number written on the bag and keep the bin record for five years, which is longer than most sites keep the invoice. It is the difference between a repair and a patch.