Microprocessor (MPU) and microcontroller (MCU) manufacturers continue to address the growing IoT applications. These chips need to process large amounts of data from an increasing number of sensors while consuming low power consumption. To reduce power consumption, chipmakers are using techniques such as adaptive voltage scaling, power gating and multiple low-power operating modes.

The global market for IoT-connected devices is expected to reach about 38.6 billion units by 2025, up from 22 billion in 2018, according to market research firm StaTIsta. These connected devices now span multiple industries, from smartphones, smart appliances and home security systems to connected cars, smart cities and the Industrial Internet of Things.

As AI and IoT converge across many industries, the additional intelligence adds some challenges in terms of safety, reliability, performance and, of course, cost. These chips need to provide high-speed processing with enhanced performance while reducing power consumption. Some of these chipmakers also employ techniques such as advanced compression to reduce power consumption and machine learning (ML) models.

Below are examples of MPUs and MCUs for IoT and converged AI applications.

For a variety of interconnect applications, Microchip Technology Inc.’s PIC18-Q43 family of microcontrollers integrates more configurable Core Independent Peripherals (CIPs), offloading many software tasks to hardware, resulting in faster system performance and shorter time-to-market time. CPI provides greater design flexibility when creating custom hardware-based functions, allowing developers to more easily customize their specific design configurations. They are designed with extra capabilities to handle tasks without CPU intervention.

Configurable peripherals are interconnected for near-zero-latency sharing of data, logic inputs, or analog signals without requiring additional code to improve system response. Applications include a variety of real-time control and connectivity applications, including home appliances, security systems, motor and industrial controls, lighting, and the Internet of Things.

The CIP includes timers, simplified pulse-width modulation (PWM) outputs, CLC, analog-to-digital converter with computation (ADCC), and multiple serial communications, allowing developers to reduce development time and increase system performance. The CLC allows developers to customize functions such as waveform generation and timing measurements. According to Microchip, CIP can also implement the entire control loop in customizable on-chip hardware.

The PIC18-Q43 product family offers a variety of memory sizes, packages and price points.

Optimized for security and wireless communications, Renesas Electronics recently introduced the RX23W, a 32-bit MCU supporting Bluetooth 5.0 for IoT endpoint devices such as home appliances and healthcare equipment. The MCU also includes Trusted Secure IP from the Renesas RX MCU family to address Bluetooth security risks such as eavesdropping, tampering and viruses.

Based on Renesas Electronics’ RXv2 core, the RX23W achieves high performance of 4.33 Coremark/MHz and features improved floating-point unit (FPU) and DSP functions. The chip has a maximum clock frequency of 54 MHz. Optimized for system control and wireless communication, the RX23W offers full Bluetooth 5.0 low energy support, including long-range and mesh networking capabilities, and claims the industry’s lowest receive mode peak power consumption of 3 mA.

The RX23W also integrates a range of peripheral functions for IoT devices, including security, touch keys, USB and CAN functions. These features enable the RX23W to implement system control and Bluetooth wireless functions for IoT endpoint devices such as home appliances, healthcare equipment, and sports and fitness equipment on a single chip, Renesas said. Additionally, Bluetooth mesh capabilities make it an excellent choice for Industrial IoT devices collecting sensor data in factories or buildings.

The RX23W is now available in 7 × 7-mm 56-pin QFN and 5.5 × 5.5-mm 85-pin BGA packages with 512 KB of on-chip flash memory.

STMicroelectronics ultra-low-power STM32L5x2 MCUs are also designed to provide better protection for IoT connected devices, based on the Arm Cortex-M33 32-bit RISC core and Arm TrustZone hardware security. Trusted Computing authenticates devices connected to the network by creating a protected execution environment for network protection and sensitive code (encryption and key storage) to thwart attempts to compromise the device or software, while a second, independent execution environment allows running Untrusted code, the company said.

With the new STM32L5 series of MCUs with clock frequencies up to 110 MHz, ST enables designers to include or exclude every I/O, peripheral or flash or SRAM area from TrustZone protection. This allows for complete isolation of sensitive workloads for maximum security, ST said.

In addition, TrustZone is designed to support secure boot, special read and write protection of integrated SRAM and flash memory, and encryption acceleration, including AES 128/256-bit key hardware acceleration, public key acceleration (PKA) and AES-128 on-the-fly decryption (OTFDEC), with for protecting external code or data. Additional features include active tamper detection and secure firmware installation. Together, these safety features provide PSA Certified Level 2 certification.

The STM32L5 series also offers ultra-low power consumption thanks to added technologies such as adaptive voltage scaling, real-time acceleration, power gating and multiple low-power operating modes. These technologies enable the MCU to deliver high performance and long runtime, whether the device is powered by a coin cell battery or energy harvesting, ST said.

The switch mode buck regulator can also be dynamically powered up or down when the VDD voltage is high enough to improve low power performance. The ULPMark scores for ultra-low-power efficiency based on real-world benchmarks developed by EEMBC are: 370 ULPMark-CoreProfile and 54 ULPMark-PeripheralProfile (1.8 V).

Additional MCU features include 512 KB of dual-bank flash memory that allows read and write operations and supports error-correcting code (ECC) with diagnostics, 256 KB of SRAM, and supports high-speed external memory including single, dual, quad, or octal SPI and Hyperbus Flash or SRAM, and interfaces for SRAM, PSRAM, NOR, NAND or FRAM.

Digital peripherals include USB full speed with a dedicated power supply, allowing customers to maintain USB communication even when the system is powered at 1.8 V, and a UCPD controller that complies with the USB Type-C Rev. 1.2 and USB Power Delivery Rev. 3.0 specifications. Smart analog features include an analog-to-digital converter (ADC), two power-gated digital-to-analog converters (DACs), two ultra-low-power comparators, and two operational amplifiers with external or internal follower routing and programming gain amplifier (PGA) capability.

The STM32L5 series offers its own STM32CubeL5 one-stop software package, which includes hardware abstraction layer and low-level drivers, FreeRTOS, Trusted Firmware-M (TF-M), Secure Boot and Secure Firmware Update (SBSFU), USB-PD device drivers programs, MbedTLS and MbedCrypto, FatFS file system and touch-sensing drivers.

The STM32L5x2 MCUs are ideal for Industrial IoT applications, including metering, health (human or machine) monitoring, and mobile point-of-sale. The STM32L5x2 MCUs are available in standard temperature grades (-40°C to 85°C) or high-temperature grades from -40°C to 125°C for consumer and commercial applications.

Fusing AI and IoT On its AI platform, Arm recently introduced its Cortex-M55 processor and Ethos-U55 Neural Processing Unit (NPU), touted as the industry’s first microNPU for Cortex-M. For demanding ML applications, the Cortex-M55 can be paired with the Ethos-U55 microNPU, which together deliver a 480x ML performance boost compared to existing Cortex-M processors.

The Cortex-M55 is known as the most powerful Cortex-M processor for artificial intelligence, and it is also the first Arm Helium vector processing technology based on the Armv8.1-M architecture, which can provide more energy-efficient DSP and ML performance. Compared with previous generations of Cortex-M, the Cortex-M55 has 15 times higher ML performance, 5 times higher DSP performance, and higher efficiency.

A new feature of the Cortex-M processor, Arm custom instructions, will be available to extend the processor’s capabilities to optimize specific workloads, the company said.

The Ethos-U55 is highly configurable and designed for machine learning inference in area-constrained embedded and IoT devices. According to the company, it offers advanced compression techniques to save power and significantly reduce the size of ML models, enabling the execution of neural networks that previously only ran on large systems.

These processors work with Arm TrustZone to ensure easier integration of security into a complete system-on-chip.

NXP Semiconductors’ i.MX RT600 crossover MCU family is designed for ultra-low-power, secure edge applications, including audio, voice, and machine learning, bridging the high-performance and gap between integrations. (The i.MX RT1170 is the winner of EP’s 2019 Product of the Year Award.)

The expansion builds on the company’s ML offerings, including the recently announced i.MX 8M Plus applications processor with a dedicated NPU. This is the first device in the i.MX family to integrate a dedicated NPU for advanced machine learning inference at the industrial and IoT edge. It also packs an independent real-time subsystem, dual-camera ISP, high-performance DSP, and 3D GPU for edge applications.

The i.MX RT600 multi-core crossover processor family features an Arm Cortex-M33 running up to 300 MHz and an optional Cadence Tensilica HiFi 4 audio/voice digital signal processor (DSP) running up to 600 MHz with four MACS and Hardware-based transcendental and activation functions.

Built on a 28-nm FD-SOI process optimized for active and leakage power, the i.MX RT600 supports a high-performance core with 4.5 MB of on-chip low-leakage SRAM configured for synchronous zero-wait-state access, making it suitable for real-time Perform audio/speech, ML, and neural network-based applications.

The crossover MCU also features EdgeLock, NXP’s advanced embedded security technology and ML support using the eIQ for Glow neural network compiler.

Security features include secure boot with an immutable hardware “root of trust,” SRAM Physically Unclonable Function (PUF)-based unique key storage, certificate-based secure debug authentication, AES-256 and SHA2-256 acceleration, and The DICE security standard implements cloud-to-edge communications. The chip also includes an optional fuse-based root key storage mechanism for secure boot and cryptographic operations, and a public key infrastructure (PKI) or asymmetric encryption that provides dedicated asymmetric accelerator.

The crossover processor includes an audio/voice subsystem supporting up to 8 DMIC channels, hardware for Voice Activity Detection (VAD), and up to 8 I 2 S peripherals. Other peripherals include SDIO for wireless communication, Hi-Speed ​​USB with on-chip PHY, 12-bit ADC with temperature sensor, and multiple serial interfaces including 50-Mbits/s SPI, I3C, and six configurable serial interfaces (USART, SPI, I2C or I2S) with individual FIFO and DMA service request support.

NXP plans to implement the Ethos U-55 in its Cortex-M-based microcontrollers, crossover MCUs and real-time subsystems in application processors targeting resource-constrained industrial and IoT edge devices.

According to NXP, the highly configurable Ethos-U55 machine learning accelerator is paired with the Cortex-M core to achieve a small size and more than 30 times higher inference performance compared to high-performance MCUs.

Eta Compute Inc.’s ECM3532 Neural Sensor Processor (NSP), billed as the first AI multi-core processor for embedded sensor applications, features the company’s patented Continuous Voltage Frequency Scaling (CVFS) and consistently delivers as low as 100 μW active power consumption. About the application. The ECM3532 multi-core NSP combines MCU and DSP with CVFS to optimize execution for best efficiency, making it suitable for IoT sensor nodes.

Eta Compute’s NSP is designed for always-on image and sensor applications, offering complete software and hardware products. The platform delivers artificial intelligence to edge devices and turns sensor data into actionable information for applications such as voice, activity, gesture, sound, image, temperature, pressure and biometrics. The platform addresses challenges in edge computing, including faster response times, improved security, and improved accuracy.

The standalone AI platform includes a multi-core processor that includes flash memory, SRAM, I/O, peripherals, and a machine learning software development platform. CVFS significantly improves the performance and efficiency of edge devices. The self-timing CVFS architecture automatically and continuously adjusts internal clock rates and supply voltages to maximize energy efficiency for a given workload. The ECM3532 is available in a 5 × 5 mm, 81-ball BGA package.

Reviewing editor: Peng Jing

Leave a Reply

Your email address will not be published.