[Introduction]STMicroelectronics, the world’s leading microelectromechanical systems (MEMS) manufacturer serving multiple Electronic applications, announced the launch of the Intelligent Sensor Processing Unit (ISPU). The new product integrates a digital signal processor (DSP) suitable for running AI algorithms and a MEMS sensor on the same chip.
● Integrate signal processors and artificial intelligence algorithms on MEMS sensors
● Significant space and power savings while introducing local decision-making capabilities
On February 23, 2022, STMicroelectronics (ST; NYSE: STM), the world’s leading microelectromechanical systems (MEMS) manufacturer serving multiple electronic applications, announced the launch of the Intelligent Sensor Processing Unit (ISPU). ). The new product integrates a digital signal processor (DSP) suitable for running AI algorithms and a MEMS sensor on the same chip.
In addition to being smaller and reducing power consumption by up to 80% compared to system-in-package products, the sensor and AI fusion solution brings electronic decision-making capabilities to the edge of the application. In the field of edge applications, smart sensors can promote the advent of the Onlife era, spawning innovative products with sensing, processing and execution functions, and realizing the integration of technology and the real world.
In the age of Onlife, accepting a life of continuous assistance provided by connected technologies, enjoying natural, transparent human-computer interaction and seamless transitions, without perceiving the difference between online and offline. The STMicroelectronics ISPU processor can migrate intelligent processing functions to the sensor, supporting the “no longer away from the edge, but into the edge” lifestyle, and promote the advent of the Onlife era.
ST’s ISPUs offer substantial advantages in power consumption, packaging, performance and price. Proprietary ultra-low-power DSPs allow algorithms to be written in the C language familiar to many engineers, and also allow quantized AI sensors to support full-bit-to-bit-precision neural networks. In tasks such as activity recognition and anomaly detection, this feature ensures applications with excellent sensing accuracy and energy efficiency by analyzing inertial data.
“While technically challenging, integrating a sensor with an ISPU on the same silicon chip really elevates a sensor-based system from an online experience to an onlife experience,” said Andrea Onetti, executive vice president of STMicroelectronics’ MEMS sub-division. Products can improve sensor functionality by reducing the amount of data transmitted and speeding up decision-making, while storing data locally can enhance privacy protection. The new products also reduce size and power consumption, helping to reduce system costs. In addition, using commercial AI models Writing ISPU algorithms is easy and basically supports all major AI tools.”
This ST’s proprietary DSP that can write algorithms in C language is an enhanced 32-bit reduced instruction set computer (RISC), which can expand the system during the chip design phase, adding special instructions and hardware. The processor provides a full-precision floating-point unit with a fast four-stage pipeline, supports 16-bit variable-length instructions, and includes a single-cycle 16-bit multiplier. Interrupt response is four cycles. This ISPU-integrated smart sensor is housed in a 3mm x 2.5mm x 0.83mm standard package and is pin-compatible with ST’s previous generation products, making it easy for customers to upgrade quickly.
Integrating the sensor and ISPU on a single chip is also a good way to save power. ST’s power consumption calculations show that in sensor fusion applications, the power consumption of the new product is one-fifth to one-sixth that of the system-in-package; in RUN mode, the power consumption is one-half that of the system-in-package to one third.