Fully Integrated NVIDIA Jetson with Dual 10 GigE & Camera Link Interfaces
- High-speed FPGA-based Edge AI system: up to 26.8 Gb/s
- Integrated Dual 10 GigE Vision frame grabber and
Camera Link acquisition interfaces - NVIDIA Jetson Orin | Xavier NX SoM
- Altera Arria 10 FPGA for image acquisition and pre-processing
- Direct transfer of image buffers from FPGA to GPU memory
- Dimensions: 134 x 90 x 60 mm | Weight: 750 g
- Inline FPGA image processing options include:
- Real-time detection for object, event, or feature identification
- On-FPGA real-time compression: JPEG, Lossless, Quality+
- Real-time High Dynamic Range (HDR) correction
- Jetson processing options:
- Compression: H.264 | H.265
- RTSP output option
Jetson Dual Interface Frame Grabber (10 GigE & Camera Link)
The FantoVision20 is a rugged Jetson system with a fully integrated dual interface frame grabber. Based on the NVIDIA Jetson Orin NX / Xavier NX, it captures and processes from dual 10 GigE Vision and Camera Link (Deca | Full | Medium | Base | Dual Base) cameras simultaneously. It merges high-end image acquisition on an Arria 10™ FPGA with real-time GPU processing on Jetson, enabling low-latency recording, streaming, and Edge AI in an ultra-compact form factor.
Jetson Dual Interface Grabber Performance and Real-Time Image Processing
The FantoVision20 is available as a plug-and-play dual-interface system supporting 10GigE Vision and Camera Link frame grabbers, or as part of a complete Imaging & Vision solution with real-time FPGA preprocessing, image enhancement, and compression options.
The built-in FPGA can process image data during acquisition, before it reaches the Jetson module. This helps reduce bandwidth, storage, and host processing load while preserving low-latency performance for high-speed camera streams.
Optional FPGA processing includes Compression, HDR correction, Detection, and additional image enhancement modules.
Real-Time FPGA Processing Helps Enable:
- Reduced bandwidth for high-throughput image acquisition
- Low-latency processing directly in the acquisition flow
- Extended recording time through real-time compression
- Lower Jetson processing load by offloading selected tasks to the FPGA
- Custom image pipelines using Gidel IPs or user FPGA logic
For the full list of available image processing and enhancement options, please refer to the Options tab.
Flexible Development Tools
- Open architecture: split/chain processing between FPGA and GPU
- GPU development: CUDA | C/C++ and NVIDIA AI libraries on Jetson
- FPGA development: rapid pre‑processing deployment with ProcVision Suite
Integrated Vision Acquisition and Processing
Deploy a single compact node for acquisition, real-time pre-processing, Edge AI inference, and recording. This approach reduces cabling, latency, and overall system cost while improving reliability and maintainability.
For large-scale, synchronized camera deployments, Gidel’s InfiniVision multi-camera vision system enables synchronization and processing across 100+ sensors within a flexible acquisition framework.
For applications that require strict determinism, line-scan optimization, or guaranteed frame capture with fixed timing, ProcFG provides a dedicated deterministic image acquisition layer.
Both approaches integrate seamlessly into Gidel’s FPGA-based vision platforms and scale from single-node systems to multi-unit topologies.
Why Choose the FantoVision20?
- Compact NVIDIA Jetson system with built-in 10 GigE Vision and Camera Link FPGA frame grabber
- Up to 26.8 Gb/s aggregate input bandwidth
- FPGA-plus-GPU architecture for low-latency acquisition, processing, recording, streaming, & Edge AI
- Optional real-time FPGA processing, image enhancement and compression IPs
The FantoVision20 is the ideal choice when your workload lies on GigE Vision & Camera Link High Speed cameras in a compact, low-latency Jetson platform.
Related Products
View the full FantoVision system range
View the full PCIe GigE Vision frame grabber range
View the full PCIe Camera Link frame grabber range
General
- GigE Vision 2.1
- Camera Link 2.0
- 2 x 10 GigE cameras
- 4 x 5 GigE cameras
- 8 x 2.5 GigE cameras
- 15 x 1 GigE cameras
- Deca, Full, Medium, or Base
- Dual Base
- Optional PoCL support
- 2 x SFP+, fiber optic
- 2 x RJ45, copper
- 2 x SDR26
(mini-Camera Link))
- On-the -fly selective ROI acquisition
- Link Aggregation mode
- Chunk Data
- 1 x GbE
- 1 x USB 3.2 Gen 2 (single lane)
- 1 x USB 2.0
- 1 x HDMI Out
- 1 x Recovery
- 1 x Restart
- 1 x UART
- Up to 8TB
- Mono, Bayer, RGBA (8, 10, 12, 14 and 16 bits/color)
- RGB (8, 10 and 12 bits/color)
- Horizontal: 16 K pixels (64-bit)
- Vertical: 65 K lines
- For configurations supporting higher resolutions- Contact Us
For the full list, please refer to the Options tab.
For the full list, please refer to the Options tab.
- Area
- Line Scan
- 2 x RS422 deferential pairs
- 2 x Opto-Isolators
- 2 x High-Drive output
- 3 x LVTTL
- GPIO power out: 2 x 12V @ 1.0A aggregate
- CAN bus
- Customizable streaming & communication using NVIDIA tools
- RTSP streaming option
- Passive cooling
- Active cooling (Fan)
>650Kh (With Passive Cooling)
- CE
- FCC
- CCI
Processors:
- NVIDIA Jetson Orin NX
- NVIDIA Jetson Xavier NX
- NVIDIA Jetson Nano NX (Subject to minimum order quantity)
- JetPack 6.2.1
- Up to 8-core Arm Cortex-A78AE 2MB L2 + 4MB L3
- 1024 Core Ampere, with 32 Tensor Cores
- Up to 157 TOPS AI computation
- Up to 16GB @ 102.4 GB/s
- Video encode: 1 x 4K60, 3 x 4K30, 6 x 1080p60, 12 x 1080p30 (H.265), H.264, H.265, AV1
- Video decode: 1 x 8K30,2 x 4K60, 4 x 4K30, 9 x 1080p60, 18 x 1080p30 (H.265), H.264, H.265, VP9, AV1
- JetPack 5.1.3
- Up to 6-core NVIDIA Carmel (Arm v8.2) 6MB L2 + 4MB L3
- 384 Core Volta, with 48 Tensor Cores
- Up to 21 TOPS AI computation
- Up to 16GB @ 51.2 GB/s
- Video encode: 1 x 4K60, 2x 4K30, 4 x 1080p60, 8 x 1080p30 (H.265), H.264, H.265
- Video decode: 2x 4K60, 4 x 4K30, 8 x 1080p60, 16 x 1080p30 (H.265), H.264, H.265, VP9
- JetPack 5.1.3
- Up to 6-core NVIDIA Carmel (Arm v8.2) 6MB L2 + 4MB L3
- 384 Core Volta, with 48 Tensor Cores
- Up to 21 TOPS AI computation
- Up to 8GB @ 51.2 GB/s
- Video encode: 1 x 4K60, 2 x 4K30, 4 x 1080p60, 8 x 1080p30 (H.265), H.264, H.265
- Video decode: 2 x 4K60, 4 x 4K30, 8 x 1080p60, 16 x 1080p30 (H.265), H.264, H.265, VP9
- 160 GX
- 270 GX
- 660 GX
Environmental Conditions
- Commercial: 0 °C - 65 °C case temperature
- Industrial: -25 °C - 65 °C case temperature. (Limited FPGA options)
- IP53-compliant
- For configurations meeting higher IP standards, contact Gidel
FantoVision20 offers unmatched modularity: select the Jetson SoM, FPGA, storage, cooling, and beyond to fit your exact needs.
Looking to adapt the FantoVision20 Edge AI according to your Vision?
The FantoVision20 edge AI offers two powerful customization paths to meet specific application requirements:
- Pre-Configured FPGA Features
Gidel develops a range of pre-embedded modules tailored to your specifications, including real-time compression, image processing, and enhancement options.
For the full list of image processing and enhancements, please refer to the Options tab. - User-Level FPGA Customization
Leverage Gidel’s development tools and IP libraries to integrate your proprietary FPGA logic and extend the system’s capabilities for acquisition, image processing, and control.
The Gidel ProcVision Suite delivers a complete toolchain for advanced user-level FPGA customization of the data flow, image pipeline, image processing, and more, ensuring optimal performance for your vision or imaging application.
Gidel can also integrate customer algorithms into the FPGA or develop custom FPGA logic according to specific application requirements.
FantoVision20: FPGA resources comparison
| Model | FV2016/164 | FV20272/27/27I | FV2066 |
|---|---|---|---|
| FPGA |
Arria 10 160 GX |
Arria 10 270 GX |
Arria 10 666 GX |
| FPGA-ALM | 61,000 | 101,620 | 250,540 |
| FPGA-M20K | 440 | 750 | 2,133 |
| FPGA-18*19 Multipliers | 312 | 1,660 | 3,556 |
|
|
2GB (Industrial) | 4GB | 2GB (Industrial) | 10GB | 4GB | 9GB |
|
|
Up to 12.8GB/s | Up to 24GB/s | Up to 15.2GB/s |
|
|
12.5Gb/s | 12.5Gb/s | 12.5Gb/s |
| Camera Link Support | Deca | Full | Medium | Base | Base | Base |
Grabbers SDK
Application Interfaces
- CameraConfig- For camera discovery and configuration.
- ggvcon- For setting GigE Vision network communication.
- ProcFG with supporting examples
- InfiniVision with supporting examples
- Gen<i>Cam GenTL producer libraries compatible with C/C++ compilers
- InitCam for developing user Gen<i>Cam camera configuration application
- GigE for developing camera network communication applications
Software Compatibility
- Ubuntu 20
- JetPack 6.2.1
- Ubuntu 20
- JetPack 5.1.3
| Name | Description | Type | Size | |
|---|---|---|---|---|
|
FantoVision20
|
Datasheet | 383.8 KB | ||
|
FantoVision20
|
Block Diagram | PNG | 154.5 KB | |
|
FantoVision20 Family
|
Getting Started User Guide | 2.0 MB |
Related Videos
FAQ
The FantoVision20 is a rugged, multi-interface Jetson system designed for hybrid vision environments. It can capture and process from 10 GigE Vision and Camera Link cameras simultaneously, supporting:
-
Up to 2 × 10 GigE (or 4 × 5G, 8 × 2.5G, 15 × 1G) cameras.
-
1 × 80-bit (Deca), Full, Medium, or Base Camera Link camera.
-
Dual Base Camera Link cameras.
-
Optional PoCL (Power over Camera Link) support.
The system utilizes an Altera FPGA dedicated to the acquisition and synchronization of both the GigE Vision and Camera Link layers. By offloading these intensive hardware tasks, the system ensures low latency and fast data transfer directly to the Jetson’s memory, keeping the Jetson CPU and GPU resources fully available for advanced Edge AI inference and analytics.
Yes. One of the core strengths of the FantoVision20 is optional inline FPGA processing for real-time compression, image enhancement, and detection directly in the acquisition flow. This allows selected tasks to run on the FPGA before the data reaches the Jetson module or host memory. For the full list of available options, please refer to the Options tab.
Virtually zero. Because the acquisition path, synchronization, and modular ISP/compression tasks are performed entirely on the internal FPGA, the Jetson CPU is freed from the heavy lifting of raw data handling. This enables you to dedicate the full power of the NVIDIA Orin NX / Xavier NX to high-level AI inference and complex application logic without performance bottlenecks.
The FantoVision20 provides a dual-development environment. For the GPU, you have full access to NVIDIA CUDA, C/C++, and AI libraries. For the FPGA acquisition and pre-processing layer, it is supported by Gidel’s ProcVision™ Suite, allowing for rapid deployment and customization of the imaging pipeline across both interfaces.
By utilizing Gidel’s InfiniVision™ architecture, the system provides a hardware-based, flexible, and deterministic infrastructure for synchronized acquisition across multiple sensors, even when mixing GigE Vision and Camera Link interfaces. This ensures zero-frame-loss performance and precise temporal alignment across all cameras—a critical requirement for accurate multi-view Edge AI inference.
Yes. The FantoVision20 is well suited for real-time recording and streaming applications. It can perform inline FPGA processing, including HDR correction, detection, and compression options such as JPEG, Lossless, or Quality+. This reduces the data load before it reaches the Jetson, supporting longer recording times, efficient streaming, and low-latency system operation.
All Gidel products come with a 12-month warranty, dedicated technical support, and regular firmware and driver updates to ensure high-performance reliability in demanding, long-term deployments.