If edge AI is going to work for automotive and industrial applications, secure real-time processing is key. The safety of moving vehicles and the productivity of smart factory lines depend on the lowest possible latency between collecting data and acting on it.
At Electronica 2024 this week in Munich, Germany, Texas Instruments is launching two new microcontroller units (MCUs) that take the increasing need for real-time control to heart. The company said its TMS320F28P55x series of C2000 MCUs has an integrated edge AI hardware accelerator that enables smarter real-time control, with up to 99% fault detection accuracy. Meanwhile, the latest 64-bit C29 core in the new F29H85x series more than doubles the real-time control performance of existing generations, and will have an integrity level up to Automotive Safety Integrity Level (ASIL) D and SIL 3 safety ratings, TI said. It also supports real-time control applications that are specific to electric vehicles..
Of the TMS320F28P55x series, Vivek Singhal, vice president and general manager, Application Specific Microcontrollers, at TI said that applications that can take advantage of edge AI integrated onto an MCU via a neural processing unit (NPU) include functions “like arc fault detection In solar and energy storage systems and predictive maintenance in a robotic arm, applications where system safety and security is imperative now… Embedded processing is central to this real-time control capability.”
The NPU in the TMS320F28P55x series offloads the execution of the neural network model from the main CPU, achieving five to 10 times lower latency than software implementations to enable faster, more accurate decision-making, according to TI. The model running on the integrated NPU also learns and adapts to different environments through training, helping systems achieve greater than 99% fault detection accuracy to enable more informed decision-making at the edge.
Meanwhile, regarding the F29H85x series, Na Kong, product line manager, Automotive Application Specific Microcontrollers, at TI, said the C29 core’s performance advantages of the C28 core “are enabled by TI’s Very Long Instruction Word (VLIW) architecture that processes 2x to 3x more instruction at once, which is particularly useful for motor control and the digital power applications. It also delivers 5x faster FFT [Fast Fourier Transform] performance for better system diagnostics, tuning and arc detection. Core hardware interrupter prioritization enables a 4x faster interrupter response, which further reduces latency and supports real-time interruption.” It also supports 2x to 3x improvement in general purpose code for command processing and communication.
Specific to EVs, Kong said, “When designing subsystems like on-board chargers (OBC) and DC/DC converters for automotive customers, engineers mostly care about the efficiency, size, and cost. Integrating these two subsystems–OBC and DC/DC converters–into one box is a clear trend in the industry. As you can imagine, a more powerful real-time controller to control both functions would be needed.” The F29H85x series enables the integration and real-time control to deliver the efficiency, size, and cost benefits, she added.