STMicroelectronics (NYSE: STM) is a world-class semiconductor manufacturer that provides customers with different electronic applications, has just announced the second version of STM32 microprocessors (MPUs) that feature an entirely new architecture that is built on their ecosystem and improving performance and security for applications that operate at the edge of industrial/IoT. “IoT edge” refers to the infrastructure for computing that supports Internet of Things devices and applications located at the extreme edges of networks. Sensors, actuators and other devices that gather information near the place where it’s collected along with routers and edge gateways and servers that analyze and process it are covered here. Edge computing seeks to cut down on the amount of latency, reduce bandwidth consumption and increase protection by processing information locally, instead of sending it to an centralized server to process. This allows for quicker, more efficient, and reliable decision-making for real-time applications that are essential to numerous consumer and industrial IoT applications.
Ricardo De Sa Earp, Executive Vice-President and General-Purpose Microcontrollers Sub-Group Group GM of STMicroelectronics stated, “Our new STM32MP2 Series devices further our investment in application processors by offering 64-bit cores, Edge AI acceleration, multimedia features, graphics processing, digital connectivity and digital connectivity – as well as embedded security features to meet emerging opportunities in secure Industry 4.0, IoT and rich user-interface applications.”
The flagship model of this new generation’s series, STM32MP25 comes with one or two Arm(r) Cortex(r) A35 CPUs that operate at 1.5GHz and an integrated 400 MHz Cortex(r) M33 processor that provides real-time processing capabilities. A NPU can provide 1.35 TOPS (tera-operations per second) of computing power for Edge AI acceleration applications like advanced machine vision and predictive maintenance. Supporting 32 bit DDR4 and LPDDR4 memory STM32MP25 delivers the long-term benefit of cost-effective designs. The STM32MP25 product line provides robust connectivity solutions that can be used in real-time Industrial applications. Data concentrators, gateways, communications equipment and real-time infrastructures, like PCIe peripherals. TSN support with two ports allows seamless switching, which is ideal for data transmission in real-time! PCIe and USB 3.0 peripherals, as well as CAN-FD peripherals make the range of products perfect. The networking and processing capabilities of the STM32MP25 range are combined to improve the detection of features and recognize them for use in security applications as well as industrial automation. One example: the MPU is able to take video from a five-million pixel sensor with 30 frames per second (fps) and then analyze the video using the Edge AI accelerator, and stream relevant video encoded using an HW encoders that include the detection metadata using Gigabit Ethernet to TSN using live streaming mode. A sophisticated 3-D graphic processing unit (GPU) that has 1080 px resolution to support video and graphics capabilities can provide users to experience rich experiences by enabling live-time Vulkan real-time graphics for Android applications. A decoder/encoder for 1080p video as well as multiple display connections, including the LVDS, the MIPI DSI 4-lane and the MIPI camera CSI-2 interfaces make connecting digital cameras and displays including raw-Bayer image sensors much easier as never before. State-of-the-art security features that support SESIP Level 3 certification include Arm’s TrustZone(r) architecture and resource isolation framework (RIF), as well as secure key storage, secure boot, unique device identification in one-time programmable memory, hardware cryptographic engines and DDR encryption/decryption on-the-fly.
DDR encryption/decryption is a security feature of microprocessors that encrypts and decrypts memory contents using Double Data Rate memory chips, also referred to as DDR memory. With this security feature implemented, the information saved in DDR memory is protected by real-time encryption. Then, it is encrypted when read out, thereby providing an additional layer of security from unauthorized access to the contents.
A Neural Processing Unit (NPU) is a microprocessor with a specific purpose specifically designed to speed up the learning and inference processes that are related to neural networks – an essential component for Artificial Intelligence applications. In contrast to general-purpose GPUs and CPUs NPUs are designed to perform specific tasks commonly used in deep learning algorithms, such as convolutional or matrix multiplication operations. By transferring computation from the GPU or CPU into an NPU AI software can deliver substantial throughput and power efficiency enhancements. NPUs are typically used in applications such as image recognition as well as natural language processing. autonomous vehicles, where the computational demands for deep-learning algorithms are extremely high.
These devices are made to work in the temperature range of -40 to 125degC. This makes the management of temperature and reliability more convenient for industrial environments. STM32MP2-series MPUs specifically designed for industrial use come with ST’s 10 year longevity warranty. They also come with various packaging options, like the 0.8mm pitch chip-scale packaging (TFBGA). This makes it easier to plan PCB design routing as well as low-cost designs that are four or fewer layers without the need for costly laser vias. Developers who use these new devices are able to benefit from the huge STM32MPU community that includes OpenSTLinux with an AI framework (X-Linux-AI) and Tools for development using STM32Cube. STM32Cube firmware runs either raw metal or real-time operating systems (RTOS) on its Cortex M33 embedded core. ST has begun to deliver samples of the STM32MP25 devices as well as evaluation boards to selected OEM customers. The production of large quantities of boards and chips is expected in the first half of 2024.
