Why a sphere needs this more than a flat wall
On a flat video wall, a dead module is annoying and replaceable: you pull one cabinet off the front and push a new one in. On a sphere the modules are curved, often cut to a specific latitude band, and sometimes only fit one position. A spare has to have been built at the same time, from the same batch, with the same calibration.
That is the whole reason to test at the factory. Faults found there cost a module swap on a bench. Faults found on site cost a crate, a flight, a crane, and a fortnight of an installation team standing around.
Start with a full white field, and leave it up
Full white at 100% is the honest test, because it shows everything: dead pixels as black dots, weak modules as grey patches, colour shift as pink or green zones, and any cabling fault as a dark band. Let the sphere run white for 20 to 30 minutes before you judge it, because a cold module and a warm one are not the same brightness and you will chase ghosts.
Then look at it from 2 to 3 metres, and from the distance the end user will actually stand. A dark dot you cannot see at 3 m is not a defect worth rejecting a shipment over; a dot visible from 3 m on a white field is.
Then the patterns that catch what white misses
A grey ramp from 0 to 100% catches banding and low-brightness nonlinearity — the sort of thing that ruins a night-time scene and looks fine on white. A 10% grey field catches modules that cannot hold a low level and drop out entirely. Colour bars catch dead colour channels, which white hides because the other two channels fill in.
We run all three in one pass, photographed at each stage. A sphere that passes white but fails 10% grey will look blotchy the first time someone plays a dark video, and by then it is in a mall.
The seam needs its own test, and its own angle
Sphere seams hide at the back when you look straight on. Put a horizontal line pattern up and walk around the ball, because a seam that is fine at eye level can show a 1 mm step at 30 degrees above. The seam also moves: it is the joint that opens as the frame settles, which is why a sphere that looked perfect at the factory can show a line six months later.
So the factory check is not just 'is the seam tight now' — it is 'is the seam adjustable in situ'. Ask how the joint is closed: shims, slots, or nothing. If the answer is nothing, the seam has nowhere to go when it moves.
Four hours of soak, not four minutes of looking
Thermal faults do not appear in the first five minutes. We run the sphere at full white for four hours with the back closed up, which is the worst case for heat, then re-check the same patterns. A power supply that is marginal, a fan that is undersized or a module with a bad solder joint usually shows up in that window.
If a supply fails in hour three on the bench, it was going to fail in week two in a lobby. That single test has saved more site visits than any other in our process.
Signing off when you cannot be there
Most of our buyers never see the factory. That is fine, but the video call has to be structured. Ask for a fixed camera on a tripod, not a phone walked around by hand; ask for the pattern to be shown at each stage with a timestamp; ask for a close-up of any module the camera operator thinks looks off.
Then ask for one specific thing: a slow pan around the whole equator at close range on white, uncut, three minutes or more. It is boring to watch and it is the only way to see every module. A five-second clip of the sphere spinning is marketing, not a test.
The list we run, briefly
Twenty-minute warm-up on white; white at 100% inspected from 3 m; grey ramp 0–100%; 10% grey field; colour bars; horizontal line pattern walked around the seam; four-hour soak at full white; re-check of all patterns; dead pixel count logged against the agreed limit.
The agreed limit matters. Put a number in the PO — three dead pixels per million is a common commercial figure — and test against it, so nobody is arguing about 'acceptable' after the crate is open.

