To effectively harness the capabilities of the NVIDIA and Raspberry Pi 4 platforms for embedded vision applications, The Imaging Source provides a range of MIPI CSI-2 board cameras. However, many embedded machine vision systems—especially those using multiple cameras—require longer cable lengths than what standard MIPI CSI-2 setups can achieve (typically limited to around 20 cm).

To address this limitation, the FPD-Link III bridge enables extended cable runs up to 15 meters. This technology transmits image data through a compact coaxial cable with a diameter of just Ø 2.8 mm at speeds of up to 4.16 Gbps. Notably, it handles not only video signals but also control commands and power supply simultaneously. The bandwidth supports demanding applications, such as transferring high-resolution images from a 5-megapixel camera running at 30 frames per second.

Developers have the flexibility to choose between MIPI CSI-2 cameras or FPD-Link III variants, with options for either standard enclosures or IP67-rated housing designed for harsh environments. These compact cameras are available in monochrome and color configurations, utilizing advanced CMOS image sensors from Sony and On Semiconductor (both global shutter and rolling shutter types), and offer resolutions ranging from 0.3 MP to 8.3 MP.

Whether an application prioritizes high resolution, real-time frame rates, or seamless integration with embedded platforms, The Imaging Source provides a wide selection of camera modules tailored for these needs.

Why Opt for These Interfaces?

The key differentiator lies in the hardware features of embedded platforms like NVIDIA Jetson Nano™ and Raspberry Pi 4. These systems incorporate hardware-accelerated image signal processors (ISP) that are directly connected to their MIPI CSI-2 or FPD-Link III interfaces. This design eliminates unnecessary data conversions and latencies, ensuring optimal performance for vision tasks.

The Imaging Source’s cameras are engineered to deliver raw image data directly to these interfaces, minimizing processing overhead. Their streamlined functionality also makes them cost-effective while still supporting advanced features like hardware-accelerated de-bayering, color correction, white balance adjustments, lens calibration, and image compression (including formats such as H.264/H.256).

Expansion Options

For multi-camera setups or AI-driven applications, developers can utilize carrier boards that support up to six camera inputs on platforms like NVIDIA Jetson Xavier NX or AGX Xavier™. These systems are ideal for running parallel neural networks for tasks such as image segmentation and object detection—especially with the full software ecosystem backing NVIDIA’s hardware solutions.

The Imaging Source also ensures compatibility by providing tailored drivers that integrate seamlessly into frameworks like NVIDIA JetPack, enabling efficient data transfer to pre-trained AI models. Similarly, Raspberry Pi users benefit from its onboard ISP capabilities when handling high-resolution H.264 encoding and network transmission via Ethernet or Wi-Fi.

Community Support

Both platforms are backed by extensive open-source communities and maker resources, making it easy for developers to find help through forums or online training programs offered by NVIDIA and other contributors.

*Note: The Imaging Source plans to release products compatible with NVIDIA’s Xavier NX and AGX Xavier in Q1 2021. For product updates or inquiries, please contact us directly.

This article was originally published in the September 2020 edition of Markt&Technik magazine under the title ‘FPD-Link III ergänzt MIPI CSI-2’.

Last Updated: 2025-09-05 00:20:59