FantoVision40

Video, Machine Vision and AI Inferences on the Edge
Gidel’s FantoVision 40™ is a pioneering compact computer enabling image
acquisition and processing from 4 x 10GigE or 4 x CoaXPress 2.1 cameras.
The FantoVision’s innovative architecture merges high-end image
acquisition with real-time image processing and/or compression using
Nvidia Jetson™ embedded computer with optional pre-processing/compression
on Intel Arria 10™ FPGA. The Jetson boasts up to 100 TOPS AI
compute capability using Nvidia’s comprehensive libraries. The GPU and
FPGA interconnect via 4-lane PCIe Gen 3. With up to 2 Tera Byte+ SSD, the
system can perform demanding real-time processing, compression, and
recording. The FPGA is fortified with up to 10 GB DDR4@200 Gb/s.

Open Customizable Image Processing
The FantoVision is also distinct in its open architecture enabling embedded
AI/image processing on GPU and FPGA. Software engineers can program
their algorithms on GPU using CUDA C/C++ and NVIDIA’s AI libraries.
In addition, developing and deploying optional pre-processing block
on FPGA is simple and fast using Gidel’s novel ProcVision™ Suite.

Scalable Solution
The FantoVision opens the way for new compact, cost-effective, scalable
vision and imaging solutions for high-bandwidth, low-latency applications.
Multi-FantoVision units can be interconnected to provide unique
and scalable topologies. Using Gidel’s InfiniVision™ open frame grabber
flow, 100+ sensors can be synchronized and processed simultaneously.

Key Features:
• 4 x 10GigE-Vision or up to 4 x 12.5 G
• 4 x CoaXPress 2.1
• Computer: Nvidia Jetson Xavier NX
• (Option for Jetson TX2/Orin NX)
• FPGA: Arria 10 – 160/270/660
• Small body: 134 x 90 x 60 mm3 (5.28” x 3.54” x 2.36”)
• GPU-FPGA interconnectivity:
• PCIe Gen 3 x4
• Image processing:
• On Jetson – supported by Nvidia JetPack SDK
• On FPGA – supported by Gidel ProcVision suite
• FPGA interfaces: 8 x RS422, 4 x Opto-Isolator, 4 x Output Drivers (30V/0.5A), 6 x GPIO 3.3V bi-dir (5V tolerant), JTAG
• GPIO power out: 2 x 12V (1A)
• Host interfaces: RS232, 1GbE, USB 3.1/2.0, HDMI, UART, Recovery, Restart
• Jetson computer key performance:
• Up to 100 TOPS AI computation
• Up to 16 GB LPDDR5 @ 102.4 GB/s
• FPGA resources:
• 160K/270K/660K LEs
• 2 GB–10 GB DDR4 @ up to 25.6 GB/s
• Up to 2,133 M20Ks
• Up to 3,374 18 x 19 multipliers
• Up to 16 I/O PLLs
• Max. power consumption : 15-45 W (dependent on system configuration)
• NVMe 100 GB - 2 TB SSD
• Passive or active cooling