MT3620

MT3620 features an Arm Cortex-A7 application processor operates up to 500MHz and includes large L1 cache and L2 cache and integrated SRAM for highly efficient operation over a wide range of potential applications. Two general purpose Arm Cortex-M4F I/O subsystems running at up to 200MHz support the requirements of the many on-chip peripherals including 5x UART/I2C/SPI, 2x I2S, 8x ADC, up to 12 PWM counters and up to 72x GPIO, allowing a diverse range of potential applications.

These two Cortex-M4F I/O subsystems are primarily intended to support real-time I/O processing but can also be used for general purpose computation and control. The Cortex-M4F cores may run any end-user-provided operating system or run a ‘bare metal app’ with no operating system. Flash memory to support the Cortex-A7 and both Cortex-M4F processors is integrated in the MT3620 package.

Outside of these three end-user accessible cores, MT3620 contains an isolated security subsystem with its own Arm Cortex-M4F core that handles secure boot and secure system operation. In addition, a 1x1 dual-band 802.11a/b/g/n Wi-Fi radio subsystem is controlled by a dedicated Andes N9 32-bit RISC core. This subsystem contains radio, baseband and MAC that is designed to allow high throughput applications with great power efficiency.

Operation of the MT3620 security features and Wi-Fi networking are isolated from, and run independently of, end user applications. Only hardware features supported by Azure Sphere are directly accessible to MT3620 end-users. As such, security features and Wi-Fi are only accessible via defined Azure Sphere APIs and are robust to programming errors in end-user applications regardless of whether these applications run on the Cortex-A7 or the user-accessible Cortex-M4F cores.

MT3620 provides an unprecedented level of security for connected devices. The secure solution provides device authentication and attestation, supports remote over-the-air software updates to maintain security in the face of evolving attacks, and also automates error logging and reporting. Please refer to Microsoft for more information.

NeuroPilot-Micro is a part of NeuroPilot and specifically designed for Tiny Machine Learning applications. It leverages NeuroPilot optimization tools and provides popular Neural Network frameworks for inference. NeuroPilot-Micro SDK supports TensorFlow Lite for microcontrollers (TFLm) and optimization APIs for model inference and optimization.

Further Reading >

• MT3620 Product Brief
• MT3620 Datasheet
• MT3620 Hardware User Guide
• MT3620 M4 User Manual
• MT3620 RTC (Power Down) Application Note
• MT3620 eFuse Content Guideline
• MT3620 M4 Hardware Functional Spec