The ML-enabled edge MCUs available in three design tiers

A new family of microcontrollers optimized for machine learning (ML) applications at the edge claims to enable real-time command and response, eliminating the need for cloud connections while substituting high-performance microprocessors.

Infineon Technologies has unveiled the next generation of PSOC microcontrollers that are AI-enabled for real-time responsiveness in connected home devices, wearables, and industrial applications. The new PSOC Edge E8 series of MCUs—E81, E83, and E84—facilitates compute responsive AI while balancing performance and power requirements and providing embedded security for Internet of Thing (IoT), consumer, and industrial applications.

Figure 1 The new edge MCUs enable developers to quickly move from concept to product and facilitate ML-enabled IoT, consumer, and industrial applications. Source: Infineon

The PSOC Edge E81 utilizes the Arm Helium DSP technology and Infineon’s NNLite Neural Network (NN) accelerator. It uses a combination of Cortex-M55 plus DSP for the high-performance domain and Cortex-M33 and DSP for the low-power domain. E81 microcontrollers are primarily targeted at cost-effective design solutions.

The PSOC Edge E83 and E84 microcontrollers, while offering the same combination for high-performance and low-power domains, also use the Arm Ethos-U55 micro-NPU processor and provide a 480x improvement in ML performance compared to existing Cortex-M systems. At the same time, E83 and E84 use the NNlite accelerator for ML applications in the low-power compute domain.

The microcontroller trio

Steve Tateosian, senior VP of industrial MCUs for IoT, wireless and compute business at Infineon, spoke to EDN before the release of PSOC Edge E8 series MCUs. He said that the ML-enabled edge MCU classification aims to facilitate the right product for the right application at the right price point. He quoted a thermostat as an example to explain how these MCU tiers work.

With an E81 microcontroller, a basic thermostat may have an LCD doing cloud-based natural language recognition. On the other hand, a mid-range thermostat may want to recognize voice locally by implementing natural language on device itself, thus removing cloud from the equation altogether. That’s an E83 microcontroller.

Finally, for Nest-like high-end devices, designers can add features like gesture and motion control as well as low-power graphics display—up to 1028×768—for a rich graphical user interface (GUI). “All three devices support voice/audio sensing for activation and control, while the E83 and E84 MCUs deliver increased capabilities for advanced HMI implementations, including ML-based wake-up, vision-based position detection, and face/object recognition,” said Tateosian.

Figure 2 Three ML-enabled PSOC edge MCUs aim to facilitate the right product for the right application at the right price point. Source: Infineon

“Designers can create a cost-effective solution with E81, but if they want to add a stronger ML acceleration hardware, they move to E83,” he added. “They can use E84 if they want to add graphics support.”

Design support services

All three edge MCUs support extensive peripheral sets, on-chip memory, robust hardware security features and a variety of connectivity options including USB HS/FS with PHY CAN, Ethernet, WiFi 6, BTBLE, and Matter. “The PSOC Edge E8 series MCUs feature a rich peripheral mix with many options in terms of in-memory as well as external memory support,” Tateosian said.

When designing ML applications on edge devices, engineers must be conscious of the amount of code in general,” he added “So, the amount of memory as well as the type of memory located on the MCU are critical.” These MCUs offer an elegant solution in terms of on-chip RAM encompassing SRAM and RRAM content.

Hardware design support includes an evaluation base board with Arduino expansion header, sensor suite, BLE connectivity for provisioning and Wi-Fi for smartphone, and cloud connectivity. On the software side, the new PSOC Edge E8 series MCUs are compatible with the earlier versions of PSOC for edge MCUs to ensure that design engineers can reuse their software investments.

Moreover, Infineon’s ModusToolbox software platform provides a collection of development tools, libraries, and embedded runtime assets to complement the development experience. It also integrates Imagimob Studio, which Infineon acquired through its purchase of the Swedish firm last year. It delivers end-to-end ML development capability spanning from data to model deployment.

Infineon will demonstrate the capabilities of this MCU series for AI and ML applications at Embedded World in Nuremberg from 9 to 11 April 2024.

<!–
googletag.cmd.push(function() { googletag.display(‘div-gpt-ad-native’); });
–>

The post The ML-enabled edge MCUs available in three design tiers appeared first on EDN.

The ML-enabled edge MCUs available in three design tiers

Become a member

Become a member

Become a subscriber

Become a sponsor