【FPGA + Nvidia/Sophon GPU + AI】Autonomous Driving Multi-core Heterogeneous Implementation: 16-channel Automotive Camera Real-time AI Analysis Solution
Designed based on Xilinx ZYNQ Ultrascale+ MPSoC series FPGA chips, this solution is applied in autonomous driving fields such as unmanned driving, low-speed special vehicles, data acquisition vehicles, and in-vehicle simulation testing systems.
Autonomous Driving: 16-channel Automotive Camera PCIe Capture Card Solution.
-
·16-channel Cameras
Supports up to 16 channels of GMSL1/2 camera input
-
·8MP Cameras
Supports up to 8 channels of 8MP resolution cameras
-
·PCIe 3.0 Bus
Supports PCIe 3.0 x8, x4, x1 modes, compatible with PCIe 2.0, with a maximum theoretical bandwidth of 8GB/s
-
·Multi-platform Support
Intel CPU platforms, AMD CPU platforms, NVIDIA Jetson platforms, Sophon BM1684X platforms


Product Overview
-
·Designed based on Xilinx ZYNQ Ultrascale+ MPSoC series FPGA chips
-
·Supports multi-channel GMSL1/2 camera input, up to 16 channels, and up to 8MP cameras
-
·Supports PCIe 3.0 bus
-
·Supports 2 channels of 10/100/1000M Ethernet interfaces
-
·Supports H.264/H.265 encoding and decoding
-
·Supports programmability, allowing custom algorithms to be added
-
·Supports V4L2 calls
-
·Supports low latency
-
·Supports external trigger signal input and board synchronization signal output
Product Advantages
-
1. Supports multi-channel GMSL1/2 camera input
Supports up to 16 channels of 1920*1080@30 resolution cameras
Supports up to 8 channels of 8MP resolution cameras
-
2. PCIe 3.0 Bus
Supports PCIe 3.0 x8, x4, x1 modes, compatible with PCIe 2.0, with a maximum theoretical bandwidth of 8GB/s
-
3. Cable Length Support
Up to 40 meters (3Gbps) in GMSL1 mode
Up to 20 meters (6Gbps) in GMSL2 mode
-
4. Automotive-grade Connectors
Uses Amphenol Z Code FAKRA automotive-grade coaxial connectors
-
5. Synchronous Acquisition
Supports external trigger signal input for synchronous triggering of multiple boards, high-level trigger, trigger level supports 3.3V-5V
-
6. Trigger Mode
Supports board synchronization signal output for synchronous triggering of multiple cameras (cameras must support trigger function)
-
7. Software Support
Supports Linux operating system, using V4L2 framework driver
Supports AXI interface for peripheral expansion, and peripherals using MSI interrupts
Supports standard IIC and UART devices
Supports video formats: YUV2, RGB888
Memory modes support USERPTR and MMAP
Supports camera configuration and query operations
Provides a 16-channel camera display host PC DEMO
Key Parameters
Mechanical Structure
Board Dimensions
Standard full-height size
Video Interface
Amphenol Z Code FAKRA automotive-grade coaxial connector
Trigger Interface
SMA connector
Video Parameters
Bus
GMSL1/GMSL2
PoC Power Supply
Max 1A@12V per channel
Number of Channels
Max 16 camera inputs
Input Resolution
1920*1080@30fps, supports up to 16 channels
4096×2160@25fps, supports up to 8 channels
Video Format
YUV2, RGB888
Software Development
Host Operating System
Tested Linux kernel version: 5.4.0-91-generic
Software Support
Supports Linux operating system, using V4L2 framework driver Supports AXI interface for peripheral expansion, and peripherals using MSI interrupts Supports standard IIC and UART devices Memory modes support USERPTR and MMAP Supports camera configuration and query operations via IIC bus Supports common operations: VIDIOC_DQBUF VIDIOC_QBUF VIDIOC_STREAMOFF VIDIOC_STREAMON VIDIOC_REQBUFS VIDIOC_QUERYBUF VIDIOC_QUERYCAP VIDIOC_QUERYCTRL VIDIOC_G_PARM VIDIOC_S_PARM VIDIOC_G_FMT VIDIOC_S_FMT VIDIOC_G_CTRL VIDIOC_S_CTRL
Temperature Range
Operating Temperature
-40℃ - +70℃
Operating Humidity
10% - 90%
Storage Temperature
-40℃ - +85℃
Storage Humidity
10% - 90%
Application Areas
Unmanned Driving
Used for camera video acquisition and transmission in unmanned driving systems

Low-speed Special Vehicles and Data Acquisition Vehicles
Used for camera video acquisition and transmission in data acquisition vehicles

In-vehicle Simulation Testing Systems
Used for video restoration after camera access in ADAS and autonomous driving testing systems

Multi-platform Support
Intel CPU Platforms
Tested with Intel 12th Gen Core i5-12400 processor
AMD CPU Platforms
Tested with AMD Ryzen 9 5900X processor
NVIDIA Jetson Platforms
Tested with Xavier, Orin architectures
Sophon Platforms
Sophon BM1684X platform
