Genlock & frame sync — the invisible workflow behind XR

The invisible metronome of digital reality

When the human eye looks at an LED wall, it sees a smooth, consistent surface. However, as soon as a camera is placed in between – whether in a virtual production (VP) environment or a hybrid corporate studio – this reality often disintegrates into digital fragments. Horizontal lines move through the image, colours flicker in wave patterns, and the depth of the space appears unstable. These phenomena are not faults of the LED modules themselves, but symptoms of a missing beat. In the world of extended reality (XR), the synchronisation between camera scanning (shutter), the image output of the graphics card (GPU), and the refresh rate of the LED wall is the most critical variable. Without a precise Genlock signal, the illusion of immersion collapses immediately.

The mechanics of failure: Why pans stutter

In traditional broadcast environments, it was often sufficient to roughly match frame rates. In the XR sector, where Unreal Engine calculates scenes in real-time and distributes them across hundreds of square metres of LED surface, requirements rise exponentially. A conventional monitor often runs at 60 Hz, while a film camera might record at 24 or 25 fps. Without explicit coupling, the camera begins a new exposure while the LED wall is still in the process of vertically building the image of the previous frame. The result is the dreaded "tearing".

Genlock (generator locking) ensures that all devices in the network use the same time measure. It defines exactly the starting point of every frame. When we at Lumexo plan systems, we do not view Genlock as an optional feature, but as the backbone of the entire visual infrastructure. It is the difference between a static background image and a living world with correct perspective.

Technical standards: From Blackburst to PTP

Historically, Genlock was an analogue signal – known as blackburst. In modern, IP-based workflows, the landscape has changed. Today, we primarily talk about SMPTE ST 2110-10 and the Precision Time Protocol (PTP) according to IEEE 1588.

PTP (Precision Time Protocol)

In a digital XR network, a "grandmaster clock" device takes the lead. Via PTP, time signals in the microsecond range are distributed over the Ethernet network. This is crucial when multiple media servers (e.g., disguise vx4 or render nodes with NVIDIA RTX A6000) calculate a common image segment. Without PTP, the transitions between the individual LED controllers could be minimally shifted, which becomes visible at the edges of the LED cabinets as a vertical offset.

The role of the LED processor

This is where devices like the Brompton Tessera SX40 or the NovaStar MX40 Pro come into play. These processors act as interpreters. They receive the Genlock signal from the camera or the sync generator and ensure that the LED modules emit their light pulses exactly within this window. The Brompton SX40, for example, offers features such as "ShutterSync". This allows the user to adapt the LED refresh cycle exactly to the camera shutter angle, instead of having to tether the camera to the wall.

Frame-Sync: The interplay of GPU and panel

Frame-Sync goes beyond pure timing. It describes the ability of the render engine to provide the next frame exactly when the LED wall is ready to display it. In a professional XR pipe, this is often solved via NVIDIA Quadro Sync II cards. These cards connect the GPUs of several render servers at the hardware level to ensure that every server outputs the same frame at the exact same time.

When a camera tracking system (such as Mo-Sys StarTracker or OptiTrack) reports the position of the camera in space to the Unreal Engine, the rendered image must reflect this data point. If the latency here is too high or frame-sync is missing, the background "swims" behind the camera foreground. This is referred to as "drift". This effect destroys the parallax and thus the spatial depth.

Phase offsets and shutter speed

A critical aspect in practice is the "phase offset". Even if the camera and LED wall run at the same frequency (e.g., 50 Hz), the wall might begin its update while the camera shutter is already halfway open. By shifting the phase in the LED processor, this moment can be adjusted millisecond by millisecond until the camera image is absolutely clean.

Practical example: Corporate XR studio in Vienna

A recent project illustrates the complexity. A financial service provider installed a 120 m² LED backdrop based on Alfalite Modularpix Pro with a pixel pitch of 1.9 mm for quarterly analyst calls.

The setting:

• LED area: 15m x 4m Alfalite LED.
• Processing: 2x Brompton Tessera SX40 in redundancy mode.
• Cameras: 3x Sony HDC-3500 on remote-controlled Vinten tripods.
• Render engine: Unreal Engine 5.3 on 3x workstations with NVIDIA RTX 6000 Ada Generation.
• Sync centre: An Evertz 5601MSC High Definition Master Clock Translator.

In the test phase, it became apparent that a slight flickering occurred in the grey-scale areas during fast pans of the automatic cameras. The solution lay in fine-tuning the Genlock signal via the Evertz generator, which provided both the Sony cameras and the Brompton processors with a tri-level sync. By activating "Frame Multiplication" on the Brompton processor, the LED wall could be operated at a significantly higher internal rate, which completely eliminated the flickering for the camera sensors, while the Genlock guaranteed absolute phase stability.

What we see in practice

When supporting projects, we repeatedly identify recurring challenges that often only become a problem during live operation:

1. Underestimated network load: PTP signals are sensitive to jitter in the network. If sync signals are routed over the same VLAN as large asset transfers, the sync can become unstable.
2. Mixed operation of generations: Attempting to integrate older LED controllers without a dedicated Genlock input into a modern XR setup almost always leads to compromises in the camera shutter angle, which reduces image quality.
3. Cable runs and signal delay: With 12G-SDI signals and large distances, signal propagation delays must be taken into account. Every millimetre of cable path means a loss of time.
4. Driver discrepancies: Often frame-sync fails not because of the hardware, but due to incompatible NVIDIA driver versions between the individual render nodes, which prevent the G-Sync handshake.
5. Ignoring rolling shutter characteristics: Every camera reads the sensor differently. Without the ability to calibrate the phase offset on the controller (as with the NovaStar MX40 Pro) visually on the control monitor, consistent image matching is pure gambling.

The future: GhostFrame and multi-camera sync

An exciting outlook is provided by GhostFrame technology. This involves displaying multiple images (sub-frames) on the LED wall within the same 1/50th second interval. Through clever timing, Camera A sees background 1, while Camera B sees background 2, while the human eye on-site perceives a combination or a neutral image. This requires a depth of synchronisation that goes far beyond classic Genlock and shows where the journey of visual infrastructure is heading: the complete decoupling of real and digital visual axes.

Recommendation from Lumexo

For a future-proof XR or virtual production environment, we recommend the following key points:

• Rely on hardware sync: Never rely on software-based sync. Use dedicated master clocks (e.g., from Evertz or Brainstorm) and ensure that all end devices (camera, GPU, LED processor) have a physical sync-in.
• Choose high-end processing: In XR environments, processors such as the Brompton Tessera SX40 or the NovaStar COEX series (MX series) are without alternative, as they provide the necessary tools for phase shifting and shutter synchronisation.
• Plan the network for PTP: If you rely on IP-based synchronisation, use switches that support the PTP-V2 protocol at the hardware level (boundary clocks) to minimise jitter.
• Validation through measurement: Use tools such as "sync check" or high-speed reference recordings to objectively measure the offset between LED refresh and camera scanning, instead of relying on the naked eye.