Future Trends of the MCU Design

CHIP

With the development and integration of AI and IoT, the design of MCUs has become more complex, gradually shifting from single-function microcontrollers to system-on-chips (SoCs) that integrate more functional features and stronger computing performance.

MCU is more intelligent (AI)

Since 2017, MCU manufacturers have tried to add AI functions to the MCU. For example, ST’s Project Orlando project is an experimental MCU ultra-low-power AI accelerator unit. Renesas released a programmable reconfigurable coprocessor DRP for MCUs in 2018. After three years of development, adding AI accelerators to MCUs is becoming mainstream. In application scenarios that require AI-related computing power, using dedicated AI accelerators is often more effective than improving processor performance.

In October 2020, Arm released the Ethos-U65 microNPU neural processing unit. This micro NPU is a very small NPU, only for embedded systems and IoT devices with limited size. NXP chose Ethos-U55 microNPU as the Cortex-M system. Ethos-U55 is specially designed for microcontrollers. It works with the Cortex-M processor and the system SRAM and Flash in the MCU. It provides the combination of performance and efficiency required by the MCU. But Ethos-U55 is not suitable for running complex ML applications on Cortex-A-based application processors.

AI MCU

From an application point of view, the main reason why AI accelerators and MCUs have become mainstream is that application scenarios that require AI are becoming more and more common. From the perspective of specific algorithms and models, it is focusing on a few models, such as machine vision (facial recognition, object recognition) and convolutional neural networks required in speech wake words, as well as in some more advanced speech recognition Recurrent Neural Network (RNN). Three years ago, the AI algorithm and application ecological prospects were not clear enough. So there was a dilemma between the two paths of specialization or generalization. Today, when applications and related algorithms are already concentrated, AI accelerators have become a clearer choice.

Here we have an example for the AI application: How Smart Traffic Management Transforming City’s Infrastructure?

Higher performance (Performance)

Continuously improving computing and processing performance is the unremitting pursuit of MCU design engineers and developers. In the past, we all followed the Arm to simultaneously enhance the performance of MCU chips. In recent years there has been a rival RISC-V. Here we sort out the performance improvement circuit diagrams of these two MCU microprocessor cores.

ARM Cortex is divided into 3 series, which are respectively aimed at 3 application areas of embedded (Microcontroller), real-time, and application. The corresponding cores are Cortex-M, Cortex-R, and Cortex-A series. The ARM Cortex-M processor is a series of upward-compatible, energy-efficient, easy-to-use processor cores, designed to help developers meet future embedded application needs, such as providing more functions at a lower cost and increasing connections, improve code reuse and improve energy efficiency. The Cortex-M series is optimized for cost- and power-sensitive MCUs and mixed-signal devices for terminal applications.

The Cortex-M series is based on the ARMv7-M architecture (for Cortex-M3 and Cortex-M4), while the lower Cortex-M0+ is based on the ARMv6-M architecture. The first Cortex-M processor was released in 2004. When some mainstream MCU suppliers chose this core to start mass production of MCU chips, the Cortex-M processor quickly became favored by the market. It can be said that Cortex-M is to 32-bit MCUs what 8051 is to 8-bit MCUs, and it has quickly become the industry standard microprocessor core. MCU suppliers develop their own based on this core and provide differentiated products in the market. For example, the Cortex-M series can be used as a softcore in an FPGA, but the more common usage is as an MCU that integrates memory, clock, and peripherals.

Lower power consumption (Power)

Consumer electronics, wearable devices, and other battery-powered IoT terminals all have strict requirements for low power consumption. System power consumption is one of the main considerations for IoT deployment. IoT devices in many application scenarios are battery-powered and require continuous use for more than 10 years. In many applications, the MCU is in a low-power sleep mode most of the time and is only occasionally woken up to read some data sent by the sensor, or process and transmit data.

The power consumption of the MCU subsystem includes two parts-the dynamic power consumption when the MCU is working, and the static power consumption related to the leakage current when the MCU is in the sleep state. Therefore, the total power consumption is affected by the working mode current, the sleep mode current, and the working mode duration. If the application is turned off most of the time, the sleep current is even more important than the operating current. A 32-bit MCU generally has a higher frequency and a correspondingly larger operating current, but its processing speed is also faster. It can save power by completing processing tasks faster and entering sleep mode faster.

If peripheral devices are allowed to communicate with each other and the sensors can be monitored without waking up the CPU, the total power consumption of the system can be greatly reduced. Silicon Labs’ EZR32 wireless MCU is a good example. It has a peripheral reflection system that allows peripherals to communicate with each other without waking up the CPU. Its low-power sensor interface can monitor up to 16 sensors. The combination of MCU and RF transceiver energy-saving technology makes wireless MCUs an ideal choice for battery-powered sensor nodes in IoT applications.

MCU is more secure (Security)

Beginning to focus on security issues is a manifestation of the development of a field to an advanced stage, because security technology input and output are often disproportionate, and in many cases, it also affects the development progress, but the lack of security is enough to cause a fatal blow. Integrating security technology from the MCU level, allowing developers downstream in the supply chain to integrate security into the development life cycle, itself is a relatively positive signal to the MCU ecology.

Almost every MCU manufacturer now adopts some security measures at the MCU level. For example, the security enhancements of Silicon Labs’ BG22 Bluetooth products include dedicated hardware-accelerated encryption, true random number generator, secure startup of the root of trust, startup engineering password verification, Secure Debug with Lock/Unlock, and Arm core support TrustZone.

A relatively typical security mechanism is a secure storage technology, that is, the bin of the plaintext is stored in ciphertext, and the program is automatically decrypted when the program is executed. The Key of each chip is also different. This is quite similar to the data protection measures of advanced mobile phone applications. There is also a typical memory partition isolation protection technology. This technology divides the Flash into multiple areas, and each area cannot access data and code mutually. This point is developed for different levels. For example, speech recognition manufacturers need to pour algorithms into MCUs, solution integrators need to integrate application scenarios, and terminal equipment manufacturers need to modify the man-machine interface. Accessing each other through API does not affect the development itself. At the same time, it is anti-copy, anti-tamper, and anti-erasing, and protects the core intellectual property rights under multi-user development and application.

Wireless connection (wireless)

The past few years have seen explosive growth in Internet of Things (IoT) devices and wireless connected products. The cost of electronic devices such as sensors and processors is gradually decreasing.  The blessing of wireless connectivity and AI performance has made many products more “smart”. And they can communicate with each other without manual intervention. However, successful AIoT products must meet the requirements of specific applications, such as low power consumption, long wireless connection ranges, and higher computing processing capabilities.

The core of this IoT system is a microcontroller (MCU), which is responsible for processing data and running a software stack that interfaces with the wireless transceiver device to achieve wireless connection and data transmission. The functional characteristics of MCUs and wireless devices depend on specific applications and system requirements. Intelligent sensor nodes integrate sensor functions and use 8-bit or 32-bit MCUs to run small radio frequency (RF) protocol stacks. These Internet of Things devices are usually battery-powered and connect to the gateway wirelessly. And the gateway performs heavier processing and data transmission.

MCU manufacturers can increase the competitiveness of their products by adding high-precision ADC or AI functions, but the integrated wireless connection function and support for various wireless protocols (such as WiFi, Bluetooth, Zigbee, etc.) will determine the application of MCU chips in the Internet of Things.

Wireless MCUs will become the standard processor chips in the AIoT era. International manufacturers such as Silicon Labs began to focus on the IoT application market a few years ago, and even divested other businesses to invest 100% in the Internet of Things.

The smaller size (Area)

To meet the needs of networking applications, MCU developers need to achieve the best balance between performance, power consumption, and size (PPA). Earlier, we briefly introduced high-performance and low-power designs. Now let’s discuss the design challenges of small size. The basic requirement of IoT terminal nodes is small size. The reason is that these devices are usually limited to a small base area. For example, the design of wearable devices, small size, and lightweight are the keys to gaining customer acceptance.

In the development history of small-size MCUs, the world’s smallest single-chip microcomputer launched by Microchip in 2004 is the PIC10F series in SOT-23-6 package. To this day, the chip is still in production and supply. It shows that the market attaches great importance to the small-size MCU demand.

Small package MCUs are ideal components for controlling IoT terminal node applications with limited size. Many MCUs have other features that can easily fit a very powerful design into a pin-constrained shape. Flexible pin assignments, autonomous operation, and intelligent peripheral interconnection devices are some examples of the advanced features of MCUs with small pin counts, which further enhance the adaptability of MCUs.

LEAVE A REPLY

Please enter your comment!
Please enter your name here