EMASS ECS-DoT – Low Power Edge AI SoC Development Kit

EMASS ECS-DoT is a development kit for the ultra-low-power edge AI SoC ECS-DoT-22, featuring a 32-bit RISC-V core, dual deep learning accelerators and 4 MB of on-chip memory for battery-powered devices.

ECS-DoT at a Glance

The ECS-DoT-22 is designed as a sensor-edge SoC that processes audio, vision and sensor data directly where it is generated, without relying on cloud-based computation. Its event-driven architecture activates AI processing only when meaningful data is detected, enabling continuous monitoring at very low power.

Configuration and Features

• 32-bit RISC-V core with floating point support as control core
• Two dedicated deep learning accelerators delivering up to 30 GOPS
• 4 MB on-chip memory consisting of 2 MB SRAM and 2 MB MRAM, no external DRAM required
• Integrated decompression IP for models at roughly 1.3 bits per weight
• Standard I/O with I2C, I2S, UART, QSPI, CPI and 32 GPIOs
• Compact 5 x 5 mm QFN package for space-saving system design
• Active power consumption from 0.1 to 5 mW during inference

Typical Application Areas

The development kit supports evaluation of edge AI applications in the following areas:

• Wearables and hearables with always-on voice and motion detection
• Smart devices with local sensor fusion
• Industrial IoT systems and robotics, for example predictive maintenance
• Drones with strict weight and power constraints

Note for Development and Procurement

The kit supports bare-metal and RTOS development through an evaluation platform and the EMASS AI Toolkit for porting and optimizing custom AI models onto the deep learning accelerators. For availability, design-in support and project inquiries regarding the ECS-DoT development kit, use the contact page of Atlantik Elektronik.