From several typical scenarios, we can see the high-performance "product power" of ON Semiconductor in enabling edge intelligent applications

At present, edge intelligence has potential application value in many fields, such as realizing intelligent manufacturing in the field of industrial automation to improve the efficiency and maintainability of factory equipment, realizing advanced driving assistance and road environment monitoring in the field of intelligent transportation, and realizing remote health monitoring in the field of healthcare. As a leader in the global semiconductor industry, ON Semiconductor also relies on its deep technical accumulation in image sensors, low-power Bluetooth MCUs and hearing aid SoC product design to help terminal systems achieve smarter decisions and lay the foundation for the widespread application of edge intelligence.

Edge intelligent terminal systems should be able to process and analyze data in real time in order to respond quickly to the environment and user needs. Taking image sensors as an example, with the rapid development of technology, modern image sensors are required to have more intelligent features. They are not only capturers of visual information, but also front-end executors of intelligent analysis and decision-making, which is an intuitive reflection of the trend of edge intelligence.

Although edge intelligence has great potential, it also faces a series of technical challenges. In the application fields of smart wearables, smart homes and even emerging AI, the visual system needs to provide higher resolution, understanding and judgment capabilities at the lowest possible cost, size and power consumption. ON Semiconductor's image sensor technology occupies a leading position in the global automotive and industrial markets. Its core competitiveness lies in the deep optimization of intelligent perception capabilities. The Hyperlux LP series sensors have ultra-low power consumption and support built-in motion detection functions. They can quickly wake up the system when a moving object is detected, further optimizing the power consumption of the system. The internal stacked architecture design can minimize the product size, and the smallest model is as small as a grain of rice.

Taking the AR0822 sensor as an example, it has built-in high dynamic range fusion algorithm and motion object capture algorithm, which can greatly reduce the consumption of system resources while ensuring image quality, and supports a variety of multiple exposure synthesis linear fitting functions - DLO (Digital Lateral Overflow) and SCMAX (Smooth Combination Max) intelligent fitting. This mode reduces the noise in the critical brightness area during multiple exposure synthesis, achieves 120dB image data output, effectively reduces the receiving data and processing time of the back-end processor, and improves the presentation of image details. In addition, the AR0822 also has enhanced near-infrared sensitivity and pixel merging (binning)/windowing output (windowing) and other sophisticated camera functions.

Furthermore, the design of image sensors that combine deep learning and neural network technology is leading a new wave of intelligent perception. These sensors can directly perform complex target recognition, classification and even prediction tasks on the edge by integrating or closely cooperating with dedicated AI processing units. In order to output scene information more accurately and quickly in more complex and diverse environments, ON Semiconductor's image sensors will integrate higher resolution, faster speed, embed more intelligent algorithms and even deep algorithms, as well as detection of non-visible light bands, etc. in the future, bringing more exquisite and detailed images to edge intelligence.

Due to the extremely high real-time requirements of edge smart hardware, Bluetooth low energy (BLE) technology has become one of the most popular electronic product connection technologies, and is widely used in consumer electronics, industry, automobiles, healthcare, computers, smart buildings and other fields, with an extremely amazing market development space. The Bluetooth low energy 5.2 wireless microcontroller RSL10 and the latest RSL15 low-power Bluetooth chip launched by ON Semiconductor use advanced semiconductor processes and dual-core architecture to ensure that applications with high real-time requirements can complete related calculations at the terminal level, avoiding the delay caused by data transmission to the cloud for processing. This design concept not only optimizes the overall energy efficiency of the system, but also ensures the immediacy of data processing and the autonomy of the system.

Low-power Bluetooth MCU solutions make full use of the characteristics of the Bluetooth standard, such as higher data transmission rates, longer transmission distances, and broadcast data expansion functions, making them an ideal choice for IoT devices, especially those battery-powered smart devices. They greatly enrich the communication capabilities and application scenarios of edge devices, including equipment asset monitoring, precise positioning services in telemedicine scenarios, etc. While maintaining long-term operation, they can quickly respond to user commands or environmental changes, perform data collection, simple analysis, and even decision-making tasks without frequent interaction with the cloud, thereby greatly reducing power consumption and extending the device's working cycle.

Another typical application case is the advanced micro AFE CEM102 recently released by ON Semiconductor, which can measure electrochemical information and ampere current with high precision. It is designed to be used with the RSL15 Bluetooth 5.2 certified wireless microcontroller. Compared with the individual solutions, this combined solution has higher accuracy, lower noise and lower power consumption, which can simplify the bill of materials and improve configuration flexibility, ultimately releasing more development resources. More importantly, the flexibility of this solution makes it suitable not only for sensors based on electrochemical measurements, but also for a variety of sensors that need to accurately measure small currents, allowing designers to develop more accurate, lower power, and more compact edge smart devices for sensing applications, such as wearable medical monitoring solutions to further improve the user experience and truly push intelligent decision-making to the edge of the device.

The wave of edge intelligence has also swept the medical market, especially with the aging of the population and the increasing demand for intelligent diagnosis and treatment experience. The design of personalized medical equipment such as hearing aids is no longer a simple audio amplification component, but needs to become more professional and intelligent, thus evolving into a micro-computing platform that integrates advanced digital signal processing, artificial intelligence algorithms and low-power management. By adopting advanced AI algorithms, hearing aids can best analyze the surrounding sounds in real time, intelligently identify and enhance voice signals, and effectively suppress background noise, so that the wearer can enjoy a clear and natural conversation experience even in noisy environments. This intelligent processing capability is completed directly inside the hearing aid without relying on external cloud services, which not only ensures the immediacy of data processing, but also protects the privacy and security of users, fully demonstrating the dual value of edge intelligence in improving user experience and protecting personal privacy.

With more than 30 years of experience in hearing aid chip design, ON Semiconductor is a leading hearing aid chip supplier in the industry. It has created a series of advanced professional digital hearing aid/OTC auxiliary hearing solutions, including Ezairo 7160, Ezairo 8300/8310, J10/J20 low-power Bluetooth wireless OTC and other platforms. In response to the personalized and intelligent industry needs, ON Semiconductor's hearing aid solutions keep pace with the times, from the early 130nm to the current 22nm process, from dual-core to 6-core, to ensure that the solution has been greatly improved in terms of performance, power consumption and latency. For example, in terms of voice delay, ON Semiconductor's mainstream solution can achieve less than 3ms. In addition, due to the development of Bluetooth low-power technology, wireless hearing aid solutions with Bluetooth functions are becoming more and more popular. For example, J10/Ezairo7160 is a typical wireless hearing aid solution.

J10 OTC hearing aid platform functional diagram

Ezairo 8300/8310 is more adaptable to the functional requirements of hearing aids in the future. The ADC bit number of Ezairo8300/8310 is higher. On the basis of conventional processing, it has been expanded to a 6-core solution, and the processing power has been more than doubled. It has a built-in NNA neural network accelerator to meet the needs of AI offline computing. It can perform local processing such as voice wake-up, volume adjustment, and basic parameter adjustment in a low-power state. It can even realize automatic adaptation functions through deep learning algorithms based on the user's hearing curve and usage, combined with the user's hearing aid usage habits. In addition, the traditional environmental scene classification function is realized by specific algorithms. If there is a neural network accelerator, the environmental classification algorithm will be more flexible and can achieve more accurate environmental scene recognition and switching. The introduction of AI functions can improve the automatic switching of different application scenarios, and add an automatic detection voice array, which can better allow users to receive valuable voice without being disturbed by environmental noise.