Designing a LED video wall control system — how to choose a controller and how to size it

1.The signal path — what happens between the content and the pixel

A LED video wall is not a monitor. It has no built-in scaler, it cannot interpret an HDMI signal by itself, and it does not distribute anything on its own. Every single pixel has to be told what brightness to show at what moment — and that work is done by the control chain. If any element of the chain is badly sized, you see it in the result: flicker, banding, lag, colour differences between cabinets. Most technical disputes are about one station of this chain, so it is worth going through all of them.

The content source is whatever produces the image: a media player, a PC, a camera, a video mixer or the output of a signal distributor. This is the only point where the resolution of the content is freely chosen, and it is here that it is decided whether the wall works at its native resolution or receives a scaled, softened picture. The practical rule: always produce content at the exact pixel size of the wall, not at 1920 × 1080 for someone to “blow up” later.

The media player or PC runs the playback software and passes the picture on from its graphics output. In an asynchronous system this role merges into the controller: the sending device stores and plays the content itself. In a synchronous system the player is a separate box, and the wall shows only what is on its output at that moment — if the machine freezes, the wall goes dark with it.

The sending controller (sending card, sending box, processor) is the heart of the chain. It takes in the video signal as HDMI, DisplayPort, SDI or a network protocol, carries out scaling and cropping, then divides the picture into as many data streams as it has network ports. It also holds the image and colour settings, the brightness, the gamma and the refresh timing — this is the device that accounts for most of the money and most of the later aggravation.

Cat6 cable or optical fibre carries the data from the controller to the wall. Over copper the safe run length is around 90–100 m, but above 80 m it is already worth switching to fibre: unshielded cable is sensitive to noise from the power supplies. Outdoors, only shielded, UV-resistant cable with watertight connector housings can be considered.

The receiving card sits in every cabinet. It takes its own slice out of the network data stream and handles grayscale processing, calibration correction and the actual driving of the LEDs. This card stores the cabinet configuration and the factory calibration data, which is why a cabinet whose receiving card has been replaced will show the wrong colour until the configuration and the calibration have been loaded back onto it.

The HUB card (adapter board) makes the physical connection between the receiving card and the modules: it carries the 16-pin module connectors, and the module power runs into it as well. The type of HUB (HUB75, HUB75E, HUB320) determines how many modules can be connected to one card and in what layout — this is the most common point of incompatibility when parts are bought from mixed sources. At the end of the chain stands the module, with its driver ICs.

2.Synchronous or asynchronous — which system belongs here

The difference between the two architectures in a single sentence: in an asynchronous system the content is stored in the controller and the wall plays it by itself, while in a synchronous system the wall mirrors a live video signal and has no content of its own. Everything else follows from that: how it is operated, how fault-tolerant it is, and what it costs.

The asynchronous system is cheaper and more self-sufficient. The controller has internal storage, content can be uploaded over 4G, Wi-Fi or the network, and from then on the wall runs with no computer attached. In return it can only manage a playlist: you cannot mix into it live, and the notion of latency does not even apply, because there is no live source. For an advertising sign, a shop entrance or an information wall, this is the right choice.

The synchronous system works on a live signal, so it is essential wherever the picture on the wall is being generated in real time: presentations, camera feeds, concert visuals, sports results, studio backdrops. There is a price for that: you need a reliable player or mixer alongside it, and the wall only works for as long as the source does. The two are not mutually exclusive — many professional controllers fall back to asynchronous playback if the synchronous source is lost.

3.Pixel maths, step by step

Sizing the control system stands or falls on a single quantity: how many pixels have to be transmitted per second. That gives you the number of ports, which gives you the controller type, which gives you the price. The calculation is simple, but it has to be done in order.

4.Three real walls, calculated end to end

Three typical orders, three lessons: in the first the cabinet grid works out neatly, in the second it does not, and in the third it turns out that even a huge wall can fit on a single port.

First case — 6 × 3 m, P3.91, outdoor rental wall, with 500 × 500 mm cabinets. One cabinet is 500 / 3.91 ≈ 128 × 128 pixels. The wall is 12 cabinets wide and 6 high, 72 cabinets in total. Its resolution is 1536 × 768 pixels, that is 1,179,648 pixels. Divided by 650,000 that is 1.81, so two ports is the minimum — in practice we split it across three ports, four cabinet columns per port, at a load of around 60%. A single four-port controller is enough.

Second case — 4 × 2.25 m, P2.5, indoor, a framed structure built from 320 × 160 mm modules. One module is 128 × 64 pixels. Horizontally 4000 / 320 = 12.5 modules, vertically 2250 / 160 = 14.06 modules — neither is a whole number. In other words, that wall size cannot be built as drawn. Rounding down to 12 × 14 modules: 3840 × 2240 mm, that is 1536 × 896 pixels, 1,376,256 pixels in total. That is a load of 2.12 ports, so we divide it across three: four module columns per port, 512 × 896 = 458,752 pixels, a 71% load. The most important lesson here is not the port count but that the wall size is decided by the module grid, not by the drawing.

Third case — 10 × 5 m, P10, outdoor billboard, with 960 × 960 mm cabinets. One cabinet is 960 / 10 = 96 × 96 pixels. A grid ten cabinets wide and five high gives a 9.6 × 4.8 m wall with 960 × 480 pixels, which is a mere 460,800 pixels. That is 48 square metres of wall, and yet it fits on a single gigabit port at a 71% load. What belongs here is not a processor but a single-port asynchronous media player controller. On the other hand the wall is “only” 960 pixels wide, so the content has to be produced at that resolution too, with large lettering.

5.Refresh rate and the camera — why the wall flickers on video

A LED cannot light up “halfway”: it is either on or off. The driver IC produces grayscale by switching the LED on and off several thousand times a second, and delivering brightness through the proportion of on-time. This is what we call the refresh rate, and it has nothing to do with the frame rate of the video. Above 1,000 Hz the eye no longer perceives it, but the camera does, because its shutter only gathers light in a short window per frame.

At a shutter speed of 1/1000 s, the camera sees one millisecond of that flickering. On a 1,920 Hz wall that is barely two complete cycles, so the rows receive differing amounts of light: dark bands run across the recorded image. At 3,840 Hz one cycle is 260 microseconds, so the same 1 ms shutter already integrates close to four complete cycles and the banding effectively disappears. That is why the professional minimum for camera use is 3,840 Hz, and in an XR studio or for high-speed recording it is 7,680 Hz or above.

The scan rate (1/8, 1/16, 1/32) tells you what fraction of the module's rows are lit at any one time. At 1/16 scan the rows are powered for one sixteenth of the time, so sixteen times the instantaneous current would be needed for the same apparent brightness — which in practice means a module with a high scan division produces less light, but can be built with fewer driver ICs and is therefore cheaper. On outdoor walls above 5,000 nits, 1/2 or 1/4 scan is the norm, while indoors 1/16 or 1/32 is more than enough for 600–1,200 nits.

Scan and refresh are linked: with the same driver IC, a higher scan division leaves less time per row, so a high refresh rate is harder to achieve. Anyone asking for 3,840 Hz on a 1/32 indoor module needs a more modern driver IC — older types cannot manage it.

6.Bit depth, gamma and the dark parts of the image

Bit depth tells you in how many steps brightness can be graded. The incoming video is typically 8-bit (256 steps per channel), but good receiving cards expand this to 14–16 bits for internal processing. The benefit is not more colours but that enough steps remain for smooth gradients even when the brightness is turned down.

This is low grayscale — grayscale handling at low brightness — and it is what most visibly separates a good wall from a bad one. An outdoor wall runs at 10–20% brightness at night: if the controller is 8-bit, 25 to 50 of the 256 steps remain and a dark sky appears in bands. With 14-bit processing the same 10% still leaves several thousand steps.

Gamma is what maps the incoming numerical value onto actual brightness. The video standard is built around a gamma of roughly 2.2, but on a LED video wall it has to be matched to the viewing conditions: in a dark room a higher gamma gives a more contrasty image, while in strong daylight a flatter curve is needed. This is part of commissioning; it needs no new hardware.

The practical benefit of HDR support is that the controller understands HDR10 or HLG metadata and maps the wide dynamic range correctly onto the brightness of the wall. If the content is not HDR — and on advertising and information walls it almost never is — this feature brings nothing; in a studio or on a premium indoor wall it is a real difference.

7.Latency, genlock and several walls in sync

Latency is the time between the signal arriving and the pixel lighting up. The chain loses time in two places: the sending processor's scaling costs about one frame, and the receiving card's processing costs a further fraction of a frame. A well-sized system stays between one and two frames, which at 50 Hz is 20–40 ms.

On an advertising sign or a shop entrance this figure is irrelevant. In a conference room it already matters: if the camera feed on the wall behind the speaker lags by half a second, it is distracting and makes autocue-style use impossible. At a live event, a picture drifting against the lighting and sound control ruins the rhythm of the show.

In a studio and in XR production, latency is not a matter of comfort but of frame accuracy: the camera and the wall have to show the same frame, otherwise the virtual background drifts against the real camera movement. Here the target is a measured and guaranteed latency below one frame, and for that you have to ask for the combined figure of the controller and the receiving card as a pair — the two values add up.

Where two or more LED video walls share a space with separate controllers, the walls refresh on separate timings. This is invisible to the eye but immediately obvious on camera: a band runs across one wall and not the other, or the two pictures differ by half a frame. The answer is synchronisation, which comes at two levels.

Frame lock means the controllers work to the same frame boundary, typically with one controller acting as master and the others locking to its signal. Many processors can synchronise directly with each other if they take their signal from the same source. This level is enough for two or three adjacent walls, provided it is not a studio.

Genlock is stricter than that: the controller locks to an external reference signal (tri-level sync or black burst), the same one the cameras and the mixer use. That puts the whole studio installation on a single common timing, and the wall's refresh sits at a fixed phase relative to the camera shutter. Most professional processors have a genlock BNC input, but the cheaper category does not — this specification has to be settled in advance, because it cannot be retrofitted.

8.Redundancy — when doubling up is worth it

There are three typical single points of failure: the network cable, the power supply and the controller itself. All three can be duplicated, and all three cost money, so the decision has to be weighed against the cost of an outage. On a roadside advertising sign, half a day of dark wall is annoying; a studio wall failing during a live broadcast is the end of the production.

Loop-wired Cat6 backup is the cheapest and most useful protection there is. A second cable runs from the last cabinet in the chain back to another port on the controller. If the chain breaks anywhere, the data arrives from the other direction and at most one cabinet drops out. There is almost no extra hardware involved, so it is worth doing on every wall where the ports allow it.

Dual power means two power supplies and two feeds in every cabinet, from separate phases or separate circuits. On rental walls and in studios this is standard; on a permanently installed advertising wall it is generally unnecessary. Duplication is only worth anything if the two feeds genuinely come from separate circuits — running two cables from the same miniature circuit breaker is self-deception.

With a backup controller (hot backup), two sending processors receive the same signal and both are wired to the receiving cards. The receiving cards watch the primary controller, and if its data stream stops they switch to the backup within a second or two. Justified in live production and in a studio, not on a shop window wall.

9.Calibration — keeping the wall a single colour

No two LEDs from the same production run light up exactly alike. Without calibration, therefore, even a brand-new wall is patchy: module boundaries show on a white image, and there are colour and brightness differences between cabinets. Calibration stores a correction coefficient per pixel or per module and pulls every pixel down to the level of the weakest.

Factory, pixel-level calibration is done on the production line: the modules are measured through with a camera and the correction data is written into the receiving card's memory. That data belongs to the cabinet, so it travels with it. If someone offers a wall with no calibration data, that is not a price advantage, it is a missing stage of work.

On-site camera calibration is carried out on the assembled wall, and it corrects what went wrong during construction: differences between cabinets from mixed production batches, replaced modules standing out, discolouration from ageing. The systems used for this are NovaStar NovaCLB, Brompton Hydra and Megapixel Omnical: they map the wall with a measurement camera and write per-pixel corrections back.

Recalibration is needed in three cases: after a module or cabinet has been replaced, if the wall starts to look patchy (outdoors after 2–4 years, because red, green and blue LEDs do not age at the same rate), and after the wall has been moved. On-site calibration is half a day to a day of work and appears as a separate line in the quotation — at contract stage, settle who recalibrates after a warranty module replacement, and at what price.

10.Compatibility traps when buying from mixed sources

The market is full of cheap “identical” parts, and control is the area where that comes back to bite hardest. The modules fit together physically, but the way they are driven is not standardised — and the difference only emerges at commissioning, when everything is already on the wall.

At handover, alongside physical acceptance of the wall, you also have to take receipt of the control documentation: the cabinet configuration file, a backup of the calibration data, the port mapping drawing, a backup of the controller settings and the software version. With those in hand, any competent engineer can repair the wall; without them you are tied to the supplier for the wall's whole service life. Those few files are worth more than the warranty period written into the contract. A controller and port mapping plan drawn up for a specific wall can also be requested through ledfalszaki.hu.