company blog about Ambarella CV28AQ Series CV28AQ26-A0-RH CVflow AI SoC For Intelligent Automotive Application

Shenzhen Mingjiada Electronics Co., Ltd. supplies the Ambarella CV28AQ series CV28AQ26-A0-RH CVflow AI SoC, designed specifically for smart automotive applications.

The Ambarella CV28AQ26-A0-RH is an automotive-grade CVflow AI system-on-chip (SoC) and a core model within the CV28AQ series, engineered specifically for smart automotive visual perception scenarios. It integrates high-performance image processing, video encoding and decoding, and deep learning AI acceleration capabilities. With a 10nm advanced process, ultra-low power consumption and high integration as its core advantages, it is the ideal main control chip for Level 2 and below driver assistance systems, smart cockpits and in-vehicle imaging systems.

I. Core Positioning and Automotive-Grade Reliability

The CV28AQ26-A0-RH is designed for mass-production-level intelligent vehicle applications and strictly adheres to automotive electronics reliability standards:

AEC-Q100 Grade-2 automotive certification, with an operating temperature range of -40°C to 105°C, making it suitable for the harsh temperature conditions found in vehicles;

10nm low-power CMOS process, with typical power consumption <1.5W, balancing performance and energy consumption whilst reducing the complexity of in-vehicle thermal management design;

The entire chip series is already in mass production for passenger vehicle projects, with SDK software compatible with the CV2x series (CV2FS/CV22AQ/CV25AQ, etc.), simplifying algorithm porting and project iteration.

II. Architectural Core: CVflow AI Engine and Multi-core Processing

1. CVflow® Deep Learning Acceleration Engine

As the core computing unit of the chip, it is specifically optimised for in-vehicle vision algorithms:

Supports CNN/DNN neural network inference acceleration, suitable for algorithms such as object detection, lane line recognition, pedestrian/vehicle classification, and driver status monitoring;

Compatible with mainstream frameworks such as Caffe, PyTorch, TensorFlow and ONNX, and provides a CNN toolkit to enable rapid algorithm deployment and optimisation;

Single-chip computing power meets the real-time inference requirements of multiple in-vehicle scenarios, supporting parallel processing of data from multiple cameras.

2. Multi-core CPU and Storage Configuration

Equipped with a dual-core ARM Cortex-A53 processor to ensure efficient system control, data scheduling and application-layer algorithm execution;

16-bit DDR interface supporting LPDDR4x/LPDDR4 (1.296 GHz, up to 1 GB), meeting the caching requirements for high-resolution image data;

Supports NAND Flash (up to 128 GB) and three SD card controllers, suitable for in-vehicle video storage and local data backup scenarios.

III. Image Processing and Video Capabilities: The ‘Ultra-HD Eye’ of In-Vehicle Vision

1. High-Performance ISP (Image Signal Processor)

6MP30 (180MPixel/s) video processing performance, supporting 6-megapixel @30fps HD image input;

Integrated multi-exposure interlaced HDR and 3D motion-compensated time-domain filtering (MCTF), restoring clear details in strong light, backlight and low-light scenarios, suitable for complex road conditions such as tunnels, night-time and rain or snow;

Hardware-level fisheye/ultra-wide-angle distortion correction and electronic image stabilisation (EIS); supports three independent sensor inputs (SLVS/MIPI CSI-2/LVMOS), suitable for panoramic imaging and multi-camera electronic rear-view mirror solutions.

2. Efficient Video Codec

Hardware-integrated dual-format HEVC (H.265)/AVC (H.264) encoder, supporting 6MP@30fps HD encoding, balancing image quality and compression efficiency;

Flexible multi-stream output: capable of simultaneously encoding dual streams of “main channel HD + auxiliary channel SD”, used respectively for local storage and remote transmission via Wi-Fi or Bluetooth;

Supports CBR/VBR bitrate control, catering to the bandwidth requirements of scenarios such as in-vehicle dashcams and ADAS video transmission.

IV. Comprehensive Interfaces and Security Features

1. Dedicated In-Vehicle Peripheral Interfaces

Video Output: MIPI DSI/MIPI CSI-2/FPD/Analogue output, compatible with in-vehicle displays and auxiliary cameras;

Communication Interfaces: USB 2.0, Ethernet AVB (10/100/1000M), UART, SPI, I2C, I2S, for interfacing with in-vehicle ECUs, radar, and Bluetooth/Wi-Fi modules;

Control Interfaces: RTC, PWM, GPIO, ADC, meeting the power management and sensor control requirements of in-vehicle systems.

2. Hardware-Level Information Security

Supports ARM TrustZone, secure boot, OTP (One-Time Programmable) memory, and DRAM encryption to prevent firmware tampering and data leakage;

The Rock Information Security Solution has been mass-produced in multiple in-vehicle projects, meeting the data security and functional safety requirements of smart vehicles.

V. Typical Application Scenarios for the CV28AQ26-A0-RH

Advanced Driver Assistance Systems (ADAS): Front-facing monocular camera, enabling Lane Departure Warning, Forward Collision Warning, Pedestrian Detection and Traffic Sign Recognition;

Electronic Rear-view Mirror (CMS): Replaces traditional optical rear-view mirrors, providing high-resolution real-time imagery, and supports Blind Spot Detection (BSD) and Lane Change Assist;

Smart Cockpit Systems: Driver Monitoring System (DMS) and Passenger Monitoring System (CMS), identifying states such as fatigue, distraction and absence from the driver’s seat to enhance cockpit safety;

In-Vehicle Dashcam (DVR): 6MP HD recording + HDR + distortion correction, supporting multi-channel synchronous recording to meet in-vehicle video evidence requirements;

Surround View Monitor (AVM): Stitching from 3–4 fisheye cameras to achieve a 360° view, assisting with parking and navigating narrow roads.

VI. Summary of the CV28AQ26-A0-RH’s Advantages

The CV28AQ26-A0-RH offers “optimised computing power, controllable power consumption, automotive-grade reliability and high integration” as its core competitive strengths. Compared to other in-vehicle SoCs in its class, it combines Ambarella’s technical expertise in ISP image quality, video encoding and CVflow AI acceleration, whilst providing proven mass-production experience and comprehensive SDK support. This enables customers to rapidly implement low-cost, highly stable smart automotive vision solutions.