Commissioning, setup and technical acceptance — how to check you got what you paid for

1.Where acceptance begins — and where the supplier's responsibility ends

Accepting a LED video wall is not one signature but three separable milestones. Treat it as one and a single small fault forces you to withhold the whole completion, and the argument drags on for weeks. Treat them separately and the installer is paid on time for the parts that are good, while you can withhold payment precisely against the defective items.

Let the warranty start on the day of successful picture and functional acceptance, not on the day of delivery. Put this in writing: with a drawn-out commissioning burdened by a defect list, there can be two or three months between the two dates — exactly the amount that will later be missing from the warranty.

2.The installation sequence: from the first bolt to first switch-on

Installing a LED video wall is a deterministic sequence: every step is the precondition for the next. Where the order is broken — the modules clipped on before the frame has been brought into plane, for instance — putting it right afterwards always costs more than the half hour that was skipped.

3.Levelling: the gap and the step your eye spots before any instrument does

The two most common and permanently visible mechanical faults on a LED video wall are an uneven gap between modules and a height difference (a step) between adjacent modules. Both matter at the level of tenths of a millimetre: a 0.5 mm step draws a continuous dark or light line across the picture when seen at an angle, whatever content is running.

The check needs no instrument. Stand beside the wall so that your eye is close to the plane of the wall, and shine a strong torch along it almost parallel with that plane. This is raking light: every step casts a shadow and every gap draws a black line. Whatever shows up like this will also show in daylight, seen from the side.

For gap uniformity, put the wall on full white and look at it from the side, at 45 degrees. If any joint is wider or lets light through, it needs levelling or a module change. On an outdoor, ingress-protected wall this is not a cosmetic question: that is where the seal has its gap.

4.Test patterns and dead pixels: what twenty minutes tells you about the wall

Eight still images are enough to find every significant picture fault on the wall. Ask the installer to play them from their own player at native resolution with no scaling — not from a laptop desktop, which gives low-resolution compressed scaling and hides the faults.

With dead pixels, distinguish three cases. A dead pixel is always dark — it shows on a white image, not on black. A stuck pixel is always lit — it shows on a black image. In a partially faulty pixel one of the three chips is dead: the pixel is not dark but discoloured, for example a cyan dot in a white field where the red has failed. That is why all four base images have to be played; white alone is not enough.

The usual contractual threshold at acceptance sits between 0.001% and 0.0001%; serious manufacturers commit to the stricter figure for the ratio at handover and give a separate “no more than this during the first year of the warranty” figure. A bare percentage is misleading, though, so ask for three additions to the contract: two adjacent faulty pixels (a cluster) are unacceptable at any ratio; a stricter threshold applies to the middle third of the image area; and repair is by module replacement, not “at the next maintenance visit”.

5.Brightness, flicker, latency, power draw — four measurements that need no laboratory

Brightness. A light meter app on a phone will not tell you the wall's nits. The nit (cd/m²) is the luminance emitted by the surface, whereas a phone and a cheap lux meter measure illuminance (lux), which falls with the square of the distance and depends on the angle of incidence. Only a calibrated luminance meter gives an absolute value — if the quotation states a specific nit figure and the wall is visibly weak, a proper measurement is required. What anyone can do, though, is the relative measurement, and it catches most real faults: on a full white image, from the same distance, measure at nine points (four corners, four edge centres, one centre). If the extremes differ by more than 10%, the wall is not uniform: either the calibration is missing or there is voltage drop along the long 5 V branches.

Flicker. Take a slow-motion recording on a phone (120 or 240 frames per second) of the wall, both on full white and with moving content. If light and dark horizontal bands travel across the image, the refresh rate is low. For an indoor wall that will appear on camera the minimum is 1,920 Hz, photographable and filmable quality starts at 3,840 Hz, and broadcast studios typically expect something around 7,680 Hz. If the quotation promised 3,840 Hz but the slow-motion footage is striped, either the receiving card setting or the module's driver IC is not what was promised.

Latency. The simplest method: start a millisecond stopwatch on the source machine, mirror it to the wall, then photograph the monitor and the wall together at a fast shutter speed. The difference between the two readings is the total system latency. On a simple sending card chain it is 1–2 frames (20–40 ms at 50 Hz), and with a scaler or video processor typically 2–4 frames. At a live event where the wall repeats the stage camera feed, anything above 3 frames produces visible lip-sync drift.

Power draw. With a clamp meter, measure the current per phase on a full white image (this is the peak), then on a black image (this is the standby figure), and finally with the actual content. The difference between the peak and the catalogue maximum should not exceed 10–15%. If the measured peak is less than half the rated figure, you are almost certainly not measuring at 100% brightness — that has to be clarified, because when the brightness is turned up in summer the supply may later prove inadequate.

6.The acceptance test in 60 minutes

This schedule works if you have sent the test pattern pack to the installer in advance and there is a laptop, a torch, a clamp meter, an infrared thermometer and a phone on site. Start the wall cold so that the warm-up happens during the test.

7.Setting up for the location: brightness, gamma, colour temperature

The factory default setting exists to make the wall hit you between the eyes in the showroom. On site it is almost always wrong: too little by day, painfully much at night. Brightness is not a single number but four or five profiles, tied to the time of day and the weather.

Automatic brightness control. Never mount the light sensor facing the wall — it would measure the wall's own light and end up in positive feedback. Put it somewhere north-facing and shaded but with a view of the sky, preferably close to the wall so that it sees the same light as the viewer. Always supplement the automation with a manual override and a night-time upper limit, because an oncoming headlight or a snowfall can mislead the sensor in moments.

Gamma. The default is 2.2, which suits video content and normal ambient light. In a dark interior, 2.4–2.8 gives a deeper, more contrasty image. Gamma becomes critical at low brightness: below 20% the grayscale of many walls collapses and fine gradients merge into a single block. Better controllers offer a dedicated function for this — in NovaStar systems, the mode that maintains high grayscale depth at low brightness — along with 14–16 bit internal processing. If the wall runs at night, have this function switched on at acceptance and check the result with the grayscale ramp test pattern.

Colour temperature. The factory default is 6500 K (D65), a cool, bluish white. On an outdoor daytime wall that is correct. Indoors, in warm-lit surroundings — among wood, brick and 3000 K lamps — that white sticks out as blue; setting it to 5000–5500 K lets the wall settle into the space, at the cost of a few per cent of maximum brightness. If there are several walls next to each other or in one space, give them all the same colour temperature and gamma, otherwise the difference shows immediately.

8.Content at native resolution: bitrate, codec, type size

The resolution of a LED video wall is the number of its physical pixels, nothing else. A 4.0 × 2.25 m P3.9 wall is 1025 × 577 pixels, so in practice 1024 × 576. Send it a 1920 × 1080 video and the processor scales it down: every second pixel of information is lost, thin lines disappear and small type blurs into a grey smudge. That is why material “rendered in Full HD” looks worse on the wall than something made at exactly 1024 × 576.

The rule, then: produce the content pixel-for-pixel at the wall's native resolution. If that means an unusual aspect ratio — a narrow strip instead of 16:9, for example — the graphics have to be designed for it, rather than a standard aspect ratio being forced in.

Codec and bitrate. The native resolution is typically well below Full HD, so bitrate is rarely the bottleneck: 8–25 Mbit/s with H.264 High profile is more than enough, and half that with H.265. What really matters: the frame rate must match the system's (25 or 50 fps in Europe, not 29.97 or 30), the video must be constant frame rate — material recorded on a phone with a variable frame rate stutters — and the container must be handled natively by the player. At acceptance, test playback with your own live content as well, not only with the installer's demo material.

The table rests on two rules of thumb. One: the comfortable minimum viewing distance in metres is roughly equal to the pixel pitch in millimetres — any closer and the pixel structure becomes visible. The other: a capital letter is comfortably legible when its height is about one two-hundred-and-fiftieth of the viewing distance.

The minimum measured in pixels is independent of that, and it is absolute: a simple sans-serif capital is illegible below 8–10 pixels in height, because there are not enough pixels to draw the shape of the letter. With a serif face, a thin stroke weight or lower case, the practical minimum is more like 12–14 pixels. Diacritics are a trap of their own: the double acute on a Hungarian Ő or Ű, or a German umlaut, cannot be resolved from two pixels, so text in an accented language blurs where plain English would still work.

9.Handover documentation: what to demand before you sign

The value of the handover documentation becomes clear when, two years later, you have to get the wall repaired by someone else. Without the configuration file and the passwords, the repair starts with someone reverse-engineering the system — more hours of work than the fault finding itself. Ask for it electronically, and save a copy to your own storage as well.

Two items are missing most often: the calibration data and the passwords. Without the calibration file the wall goes patchy the moment anyone performs a factory reset — and at that point there is nothing to be done without the installer. Without the passwords you can neither change content nor turn to another provider. Make these two items a condition of the completion certificate, not a polite request.

10.Warranty, spare parts, training and day-to-day operation

The warranty on a LED video wall typically covers the modules, the receiving cards, the power supplies and the controllers, for between two and five years. In practice, though, what decides matters is not the duration but what is excluded and how repairs are carried out. Typical exclusions: overvoltage and lightning strike, water ingress through inadequate sealing, unqualified intervention, sustained brightness or temperature above the permitted level, and the fitting of parts not supplied by the vendor.

“Module-level replacement” means that in the event of a failure the faulty LED or driver IC is not unsoldered; the whole module is swapped for one from stock. That is the correct procedure — it is fast and it does not spoil the uniformity of the wall. What has to be clarified: where the replacement module comes from (local stock, or from China with a four to six week lead time), who pays for the call-out and the access equipment, and what happens if the replacement module is no longer from the same production run.

That last point is the most insidious risk. There are small colour differences between production batches (bins) of LED chips, and module types change every two or three years: a later replacement module will show even if it is technically identical. That is why it is worth buying spare modules equivalent to 3–5% of the wall from the same shipment at the time of purchase, and storing them on site — the cheapest insurance you can buy alongside a LED video wall.

Training. On-site training is part of acceptance: at least an hour, for at least two people — because one person goes on holiday. Have a short written summary of the training produced in your own language, and record it on video with a phone; that will be worth more later than any manual.

Day-to-day operation. Name one person responsible and one deputy — if everyone is responsible, nobody is. Do not circulate the passwords by email; put them in a password manager, and restrict remote access to those who need it: a LED controller reachable from the internet with factory passwords is a genuine attack surface. Scheduled switch-off and switch-on does not only save electricity, it also slows the brightness depreciation of the LEDs and the ageing of the power supplies. Finally, keep a simple operating log: what was replaced and adjusted, and when — in the third year, those few lines are what will tell you whether the wall is ageing evenly.

Acceptance is not a matter of distrust but of common interest: a good installer welcomes a detailed report, because it protects them too against unfounded complaints later. If the supplier makes difficulties about carrying out the tests described here, that is information in itself. And if you need expert help with acceptance or setup, ledfalszaki.hu is available for that as well.