Before the crates open
Two things have to be settled while everything is still in packing, and both are cheaper to fix there than after the wall is built. The first is count and model against the order: cabinets, modules, power supplies, receiving cards, spare modules and cabling.
The second is the batch. Every module on one wall should carry the same batch number, and that includes the spares. Batches differ slightly in binning, and a replacement from a different batch shows up as a block that sits a shade off its neighbours. The calibration file compensates for a lot of this, but the closer the match, the less correction has to do.
Photograph the labels before anything is unpacked and keep the batch number in the handover file. It is the one piece of information that is impossible to recover later and that every future spare order depends on.
The five frames, in this order
Do not judge a new wall by playing content on it. Content hides faults by giving the eye something else to follow. Run five flat fields instead, one at a time, and look for one specific thing in each.
Black first. Any emitter that stays lit on a black field is a stuck pixel, and it is easier to count them on black than on anything else. Then full white, which is where dead emitters, and any module that sits brighter or dimmer than its neighbours, become obvious. Then pure red, pure green and pure blue, which separate a dead emitter from one that has lost a single colour.
Walk the wall; do not judge it from the centre. Faults cluster at cabinet joins and at the ends of power runs, which are the two places a standing viewer looks least.
Seams and low grey, where faults hide
Two conditions account for most of what gets reported after handover, and both are invisible in a full-white photograph of a healthy wall.
The first is the physical seam. Module rows and columns have to sit on an even line, and a module that is a fraction proud or recessed catches light along one edge. Sight along the wall from a shallow angle with the field lit; a seam that is correct head-on usually stays correct at an angle, and one that is not shows up immediately.
The second is low grey. Bring the wall down to roughly ten or twenty per cent and look again. Colour shift between modules, banding across driver boundaries and a green or magenta cast in dark areas all live in the low end, and none of them appear on a bright white field. A wall that passes at full white and fails at ten per cent will fail again on any dark content it ever plays.
Measure instead of judging by eye
Three figures are worth taking with instruments, because the eye adjusts and the instrument does not. Brightness at the centre of the wall, measured with the room in its normal lighting state, should land in the band the specification sheet quoted. Power draw, taken with a clamp meter at full white, should sit near the peak figure: the indoor platform is quoted at 400 W/m² average and 800 W/m² peak, so a wall that reads far from those numbers has something to explain.
Refresh behaviour is the third. Point a camera at the wall and record at the shutter speeds the installation will actually be shot at. Scan lines that are invisible to the eye appear on video, and this is the only way to find them before the first production crew does.
On outdoor work add the ingress checks to the same pass: front and rear rated to IP65 under IEC 60529, drainage channels clear, and panel sealing intact at every cabinet joint. Water finds the one joint that was not closed.
Let the wall reach its working state first
A display measured cold is not the display that will run. Brightness and colour both move as the wall warms, so acceptance figures taken in the first minutes are not the figures the installation will hold. Bring the wall up to its normal working output and let it settle before any measurement is recorded.
The room matters as much. Indoor walls are usually accepted with the lights off, which is the easiest possible condition and the one the wall will never be judged in again. Accept it under the lighting state the space actually runs in, because that is where readability is won or lost.
If the project has a factory acceptance test, this is where it pays off. Faults found before shipping cost a module swap at the factory. The same fault found after the steel is in costs access equipment, and on a high or suspended wall that is most of the cost.
The files that have to change hands
Four documents should exist before anyone signs, and the first is the one most often missing.
The calibration file. This holds the per-module correction data that makes the wall uniform. Without it, a replacement module cannot be matched to its neighbours and the wall has to be re-calibrated as a whole. Confirm where it is stored, who holds it, and that it is included in the handover. A calibration file left on a commissioning laptop is a calibration file the project does not have.
Then the as-built drawings covering structure, power distribution and signal topology; a backup of the control system configuration, including screen layout and any scheduling rules; and the spare parts register with batch numbers recorded. On outdoor or hard-to-access installations, the maintenance and access method belongs in this set too, since it decides what a future repair actually costs.
What to write down before signing
Acceptance does not mean the wall is perfect. It means every open item is recorded with an owner and a date, and nothing is left to memory.
For each fault: where it is, which module or cabinet, what it looks like, and when it will be re-checked. A photograph next to a module identifier is worth more than a description, because the person returning to fix it may not be the person who found it.
Settle three things in writing at the same time. The warranty start date, which should be the day the installation is accepted. Check it against the ship date, because the two are often different by weeks. The response commitment for faults reported after handover. And the spare module quantity that stays on site. Standard warranty on this platform is two years, and the terms around labour, access and travel are worth reading separately from the period itself.
Putting this into practice: acceptance testing checklist
- Inspect before power Check crate count and model against the order, and confirm one batch number across the whole wall including spares. Photograph the labels before anything is unpacked; the batch number is impossible to recover later and sits in the handover file.
- Run five flat fields, not content Do not judge a new wall by playing content. Run black first (stuck pixels), then full white (dead emitters), then colour fields, and walk the wall. Faults cluster at cabinet joins and at the ends of power runs, where a standing viewer looks least.
- Measure three figures with instruments Brightness at the centre under normal room lighting, low-grey uniformity, and refresh behaviour recorded at the shutter speeds the installation will actually be shot at. The eye adjusts; the instrument does not.
- Collect four documents before sign-off The calibration file, the as-built drawings, a backup of the control configuration, and the acceptance report. Without the calibration file a replacement module cannot be matched to its neighbours.
Myths about acceptance testing checklist
“Switch it on with content and judge it as a whole.” — Content hides faults by giving the eye something else to follow. Most post-handover complaints are seam lines and low-grey shift, both invisible in a full-white photograph of a healthy wall.
“Acceptance can be done cold in the first minutes.” — Brightness and colour both move as the wall warms, so figures taken in the first minutes are not the figures the installation will hold. Accept it at its working state, not in the first minutes.
What to carry forward on acceptance testing checklist
Inspect in a fixed order so every fault gets a frame that reveals it: crates and batch numbers before power, then black, white, red, green and blue, then low grey at ten per cent, then seams from a shallow angle. Measure brightness, power against 400 W/m² average and 800 W/m² peak, and camera behaviour with instruments, with the wall warm and the room in its normal lighting state. Collect the calibration file, as-built drawings, control system backup and spare register before signing, and record every open item with an owner and a re-check date.