reComputer Robotics GMSL Board for Jetson Orin NX/Nano, MAX96712 Deserializer, 4Camera Coax Power/Data, 6Gbps/channel

Overview

The reComputer Robotics GMSL board is a Robotics GMSL board designed for the reComputer Robotics series based on Jetson Orin NX and Orin Nano. Both data and power are transmitted over coaxial cable. The board integrates the MAX96712 deserializer and is compatible with GMSL cameras using MAX96717 or MAX9295A serializers. It has native support in JetPack for multiple Sensing cameras and the Orbbec Gemini 335LG, and it can connect up to 4 cameras simultaneously.

Key Features

• Industry-standard deserializer: MAX96712
• Compatible with MAX96717/MAX9295A camera serializers
• Supports up to 6Gbps data rate per channel
• Data and power over a single coaxial cable
• Up to 4 simultaneous camera connections
• Ready-to-use camera models supported in JetPack (see table below)

Specifications

Product	reComputer Robotics GMSL Board
Deserializer	MAX96712
Operating Temperature	-20~60
Mechanical Dimension	86mm x 31mm x 18mm

Cameras Supported in JetPack

Manufactuer	Model	Resolution
Sensing	SG3S-ISC031C-GMSL2F	1920H*1536V
Sensing	SG2-AR0233C-5200-G2A	1920H*1080V
Sensing	SG2-IMX390C-5200-G2A	1920H*1080V
Sensing	SG8S-AR0820C-5300-G2A	3840H*2160V
Orbbec	Orbbec Gemini 335Lg	3D

Hardware Overview

• Compact add-on board for reComputer Robotics systems based on Jetson Orin NX/Nano.
• Four-channel coaxial interface visible on the board (channels labeled 03) for multi-camera connectivity.
• Installs inside the reComputer chassis using standoffs and board-to-board connectors as shown.

Application

• Robotics (Object detection and 3D reconstruction)
• BEV Detection

Documents

• Schematics

Part List

reComputer GMSL Board	x1
M3 Stud	x2
M3 Screw	x2

ECCN/HTS

HSCODE	8473309000
USHSCODE	8473309100
EUHSCODE	8471707000
COO	CHINA

Details

GMSL board for 4-Camera Robotics V1.0 with connectors and components

Multi-camera inputs use backbone, spatial cross-attention, and temporal self-attention to generate BEV for detection and segmentation, incorporating multi-view data, history BEV, and query mechanisms. (39 words)