STMicroelectronics recently announced that the STM32MP23x product line (covering STM32MP235, STM32MP233, and STM32MP231) has officially entered mass production and sales for the mass market. After launching the STM32MP25 series a year ago, the newly released STM32MP23x series focuses on cost sensitive industrial application scenarios, while retaining core features such as neural processing units (NPUs), heterogeneous architectures (Cortex-M33 and dual core Cortex-A35), support for Linux and real-time operating systems (RTOS), and high-performance network interfaces with time sensitive networks (TSN). In short, the STM32MP23x new series of products makes edge machine learning within reach, especially for applications that require only 16 bit DDR4/LPDDR4/DDR3L memory controllers and H.264 hardware decoding without video stream encoding.
STM32MP23x: Focus on key performance
Upgrade considerations
Taking an automated bottling plant as an example, its production process is complex, and the machines need to be coordinated by a controller to achieve precise operation through multiple sensors. Engineers are exploring new applications of machine learning, such as predictive maintenance, defect detection, and process optimization. Due to the fact that most of the existing systems in the factory run on Linux and technicians are generally familiar with the UNIX operating system, integrators urgently need a microprocessor (MPU) to upgrade and transform existing equipment.
Cost optimization strategy
STM32MP23x integrates two 1.5 GHz Cortex-A35 cores, allowing developers to build a low-power operating environment using Cortex-M33 cores while running Linux systems. This series of products has the same processor architecture and operating frequency as STM32MP25x, thus continuing the flexibility of STM32MP25x, but adopting a more cost-effective design solution. In practical applications, development teams often delay system upgrades due to cost issues, and STM32MP23x can effectively alleviate this pain point, shorten the system iteration cycle, and simplify the development process.
STM32MP23x: Optimizing Intelligence for Ubiquitous Edge Machine Learning
More possibilities for edge machine learning
In the era of edge machine learning, simplifying development is crucial because many companies recognize its value but often hesitate due to cost considerations. Taking the automated bottling production line as an example, a large number of cases have proven that the combination of motor sensors and machine learning can achieve predictive maintenance, significantly reducing overall operating costs. More and more AI systems can accurately detect product defects, helping factories reduce the flow of defective products into the market and improve customer satisfaction. However, many enterprises still have knowledge blind spots in the technological implementation of AI landing.
Not limited to NPU
Indeed, NPU is the core component of edge AI, but not all of it. STM32MP23x is equipped with a neural processing engine with 0.6 TOPS computing power, and also integrates a camera interface with Lite ISP to support fast processing of sensor images. In addition, it is equipped with a hardware decoding accelerator with 500 megapixels per second, as well as MIPI DSI and LVDS interfaces for human-machine interface. In contrast, competitor solutions often only provide NPU functionality. The STM32MP23x enables engineers to complete image acquisition, real-time processing, and neural network inference on a single MPU without the need for additional configuration of multiple MPUs or co processors.
STM32MP23x: Addressing the unique challenges of industrial applications
Efficient integration, twice the result with half the effort
Integration remains the core driving force for the development of microprocessors. Taking the intelligent factory of automated bottling line as an example, high integration can reduce system failure points and help build a more complete overall solution. For example, if the MPU responsible for running machine learning algorithms also provides computing power support for the employee interface on the display screen, it can significantly improve production efficiency and optimize workflow. This requires MPU to respond to the complex requirements of industrial applications with full stack capability, and the automated bottling line involves numerous subsystems that need to achieve efficient collaborative operation.
Reduce uncertainty and provide more solutions
The STM32MP23x inherits the core advantages of the STM32MP25x in industrial scenarios, including CAN-FD bus, SDIO 3.0 controller, USB Type-C power support, USB 2.0 host/device controller, and two Gigabit Ethernet modules supporting Time Sensitive Networking (TSN). Among them, the TSN module can construct a deterministic network to ensure the predictability of information transmission in scenarios such as motor control that require strict data synchronization accuracy and low latency. In addition, engineers can use TSN to implement packet priority management, plan emergency response mechanisms, and build redundant links to enhance the robustness of the system architecture. In summary, a high-performance microprocessor can become the core engine of factory intelligence.
