T-Chip Group
Intelligent hardware website
Innovate intelligent hardware and solutions.
Server Business Website
Server products and efficient solutions.
Consumer Business Website
StationPC NAS products and services.

CAM-1126BQ38 Motherboard

AI-ISP+AOV3.0: Redefining Visual Perception

· AI Vision Processor RV1126B

· Supports Multimodal Models

· Built-in 8-Megapixel AI-ISP

· Intelligent Encoding Engine

· Multi-Camera Stitching ＆ Video Stabilization

· Hardware-Level Security System

· 38mm × 38mm Mini Size

· Development Materials

Manual Wiki Download

Qwen

Gemma

Phi

MiniCPM

CAM-1126BQ38
CAM-1126BQ38 Mini Motherboard

Powered by RV1126B AI vision processor (3 TOPS NPU), this compact module (38mm×38mm) supports 2B multimodal models and deep learning. It features a 12M AI-ISP for excellent day/night imaging, 4K video processing, multi-camera AI stitching, 6-axis digital stabilization, and multi-source fusion (RGB/IR/ToF). With AOA low-power sound detection and built-in Chinese cryptographic (SM2/SM3/SM4) security, it is ideal for AIoT applications like facial recognition, smart security, and industrial vision.

AI Processor RV1126B

Powered by the RV1126B, a high-performance 64-bit quad-core AI vision processor, it is built on ARM Cortex-A53 cores and integrates NEON SIMD and FPU, with a clock speed of up to 1.6GHz. Its performance exceeds that of comparable chips by over 2x, delivering robust support for intensive computing and multitasking.

Supports Multimodal Models

Built-in 3 TOPS NPU supports running LLMs and multimodal models up to 2B parameters (e.g., Qwen, Gemma2-2B, Phi2, MiniCPM). Featuring advanced optimizations like quantization and Transformer acceleration, it delivers robust on-device AI capabilities for image understanding and cross-modal fusion in lightweight scenarios.

Built-in 8-Megapixel AI-ISP

Equipped with a new-generation hardware-based ISP supporting up to 12MP, complete with accelerators for HDR, 3A, LSC, noise reduction, and other key algorithms. Building upon this foundation, an AI-ISP (up to 8MP), enabling superior spatial noise reduction and enhanced image processing capabilities.

Intelligent Encoding Engine

RV1126B features a smart encoding engine that supports 8MP at 45fps. Its dynamic bitrate optimization cuts bandwidth usage by 50% versus standard CBR, doubling recording time with equal storage while maintaining image quality. This reduces storage and transmission costs, ideal for video surveillance and data intensive applications.

Multi-Camera AI Stitching

It combines AI-powered multi-camera stitching, 6-DOF video stabilization, and multi-source fusion to support 2×6MP real-time panoramic stitching with seamless wide-angle output. The system integrates visible light and infrared imaging to deliver stable, detailed visuals day and night, ensuring an immersive viewing experience in complex lighting.

Hardware-Level Security System

RV1126B integrates hardware-level security with support for SM2/SM3/SM4 cryptographic algorithms. Featuring TrustZone isolation and a key management system, it provides full-chain protection for data and AI models, meeting stringent security requirements for intelligent devices.

38mm × 38mm Mini Size

The motherboard measures just 38mm × 38mm × 11.5mm. Its ultra-compact design allows seamless integration into smart cameras, surveillance cameras, and other space-constrained devices, freeing up valuable internal space for core components and significantly improving overall integration and space efficiency.

Development Materials

Provides reference designs and comprehensive technical documentation for the baseboard, enabling users to efficiently conduct secondary development and rapidly build self-controlled products.
AI-ISP+AOV3.0: Redefining Visual Perception

RV1126B integrates a dedicated 8MP AI-ISP that operates independently from the NPU, saving power and bandwidth. It supports AI Remosaic for high-quality imaging in both daylight and low-light conditions. The chip also includes AOV 3.0 for low-power audio event wake-up with 1mW standby, enabling continuous monitoring. Combined with hardware 6-DOF stabilization and binocular stitching, it delivers stable, seamless wide-angle video, meeting the demands of security and automotive applications.

Application Scenarios

It is widely applicable in various industry sectors such as facial recognition, access control gates, intelligent security, smart network cameras, smart peepholes/doorbells, and dash cams.

Facial Recognition
Facial Recognition
Access Control Gates
Access Control Gates
Intelligent Security
Intelligent Security
Smart Network Cameras
Smart Network Cameras
Smart Peepholes/Doorbells
Smart Peepholes/Doorbells
Dash Cams
Dash Cams
Interfaces
Specifications

CAM-1126BQ38

CAM-1126BJQ38

Basic Specifications

SOC

Rockchip RV1126B

Rockchip RV1126BJ

CPU

Quad-core 64-bit ARM Cortex-A53, integrated with NEON and FPU, maximum main frequency of 1.6GHz

NPU

3TOPS NPU, supporting INT4/INT8/INT16/FP16 mixed-precision computing

ISP

Built-in 12MP ISP with multiple algorithm accelerators including HDR, 3A, LSC, 3DNR, 2DNR, sharpening, defogging, fisheye correction, gamma correction and feature point detection. Integrated 8MP AI-ISP enables day-night dual-mode adaptation with AI Remosaic technology; supports AOA (Always On Audio), 6-DOF digital anti-shake, binocular AI dynamic stitching and more.

Encoding/ Decoding

Encoding: 12M@30fps H.264 Decoding: 4K@30fps H.264

RAM

LPDDR4/LPDDR4X (1GB/2GB/4GB Optional)

Storage

eMMC (8GB/16GB/64GB Optional)

Storage expansion

1 × TF Card

Power

DC 12V (Connected via 8P-1.25mm Wafer connector)

Power consumption

Normal: 1.2W(12V/100mA), Max: 3.6W(12V/300mA), Min(Sleep): 0.48W(12V/40mA)

OS

Debian12, Buildroot+QT

Software support

Supports private deployment of lightweight large language models and multimodal large models with up to 2B parameters under the Transformer architecture, such as small-sized large AI models including the Qwen series, Gemma2-2B, Phi2, InternLM2, MiniCPM series, TinyLLAMA, and RWKV7. Supports deep learning frameworks such as TensorFlow, TensorFlow Lite, PyTorch, Caffe, and ONNX.

Size

38.09mm × 38.11mm × 11.54mm

Weight

≈20g

Environment

Operating Temperature: -20℃ ~ 60℃, Storage Humidity: 10%～90%RH (non-condensing)

Operating Temperature: -40℃ ~ 85℃, Storage Humidity: 10%～90%RH (non-condensing)

Interface Specifications

Ethernet

1 × 100M Ethernet (Led out via 8P-1.25mm Wafer connector), supporting TSO (TCP Segmentation Offload) and USO (UDP Segmentation Offload) network acceleration

WiFi

1 WiFi module on board, supporting 2.4GHz/5GHz dual-band WiFi6 (802.11a/b/g/n/ac/ax)

Video input

1 × MIPI-CSI DPHY (Supports MIPI V1.2 version; 1×4Lanes or 2×2Lanes, 40Pin-0.5mm)

USB

1 × USB2.0 (Led out via 24Pin-0.5mm FPC connector), 1 × Type-C (Device)

Button

1 × Reset, 1 × MaskRom

Antenna

1 × WiFi antenna

Other interfaces

1 × Expansion Interface (24Pin-0.5mm FPC connector, led out: 5V, 1×USB2.0, 1×ADC, 1×I2C, 1×UART, 10×GPIO), 1 × RS485 (Led out via 8P-1.25mm Wafer connector), 1 × RTC Battery Holder (2Pin-1.25mm), 1 × Debug (3Pin-1.25mm), 1 × Recovery (2Pin-1.25mm)
Customization

Firefly team, with over 20 years of experience in product design, research and development, and production, provides you with services such as hardware, software, complete machine customization, and OEM server.

Hardware Solution

Software Solution

Turnkey Solution