Just reading an interesting article on MCUs.
Highlighted what I deemed some pertinent comments / requirements / observations as indicated by some industry players where we can definitely assist.
What’s Next for the Microcontroller?
February 22, 2023
Robert Huntley
Although the microcontroller market is set for sustained growth, do MCU technical features and functions need to evolve to continue meeting customer requirements? Is the general-purpose MCU being replaced by application-specific versions?
The humble microcontroller, now more than 50 years old, represents a sizable chunk of the overall electronic-component industry. MCUs continue to dominate the embedded scene, and with good reason: They are flexible, configurable, and easy to program. With microcontrollers used in everything from laser printers to washing machines and heating thermostats to forklift trucks, MCU shipment data provides a reasonable indication of the state of the electronics industry.
Over the years, application-specific versions have developed to meet the requirements of use cases such as motor control, wireless connectivity and ultra-low power. Some MCUs feature highly configurable analog and digital blocks, borrowing architectural concepts more associated with FPGAs than MCUs. Others are marketed as general-purpose controllers, incorporating an array of fixed-function blocks—from A/D and D/A converters to serial connectivity, timers/counters, GPIO, and cryptographic accelerators—to suit a broad range of applications.
The microcontroller market exhibits continued growth
According to recent research by P&S Intelligence, the global microcontroller market accounted for US$18.80 billion in 2021 and is expected to reach US$43.61 billion in value by 2030, for a 9.8 % CAGR. The reasons for the projected growth are many, from the increasing use of machine learning in smart sensors to the dramatic increase in industrial automation systems.
Joe Thomsen, Microchip Technology
Although the microcontroller market is set for sustained growth, do MCU technical features and functions need to evolve to continue meeting customer requirements? Is the general-purpose MCU being replaced by application-specific versions?
According to Joe Thomsen, vice president of Microchip Technology’s 16-bit MCU Business Unit, the customers define the needs. “One of the things we do regularly is to evaluate what our customers are putting on their boards and what else is being implemented alongside the microcontroller,” said Thomsen. “Then we can determine how we can interface to those items more easily, more effectively, or [whether] we can actually integrate those features into the MCU itself.”
What technical innovations are happening?
Today’s MCUs are typically highly integrated devices with lots of functionality, intended to offer a single-chip solution for many designs. As customer needs and application use cases evolve, how is the MCU keeping up?
One focus is power. “Microcontrollers account for approximately 5% to 10% of the overall power consumption inside the vehicle, so we’re looking for further possibilities to reduce it,” said Ralf Koedel, vice president for microcontrollers at Infineon’s automotive division.
“Low power is one of the key items we need to look into to drive the MCU space further,” said Tim Burgess, senior director of the MCU Business Unit at Renesas. “The [MCU’s] low power is one of the major differentiators from microprocessors. As we decrease the process technology, it allows us to address low power by design. There’s always a conundrum with process technologies: When you go to a more advanced process node, you get better active consumption, but because the gate is very small, the leakage is significantly higher.”
Steven Tateosian, Infineon
Steven Tateosian, vice president, IoT, Compute and Wireless Business Unit of Infineon, approached the question from an industrial and consumer perspective, pointing to innovation’s role in raising MCU performance for a broader range of applications.
“What we started seeing in the last five years is more integration, such as multicore processors, mixed with DSPs and other accelerators,” he said. These additions are being made “without fundamentally changing things around [MCUs’] ease of use and power profiles, and the overall system cost advantages that the MCU brings over microprocessors.”
The versatility of low-power microcontrollers has made them
extremely popular for intelligent edge node applications, particularly those based on tinyML. Many voice-assistant–based applications rely on continuous cloud connectivity to conduct interference, only recognizing a trigger word or short phrase locally. However, this approach introduces latency and the risk of a device exploit. The need for local, deterministic decision-making is a priority.
Most microcontroller vendors focus on incorporating neural network accelerators into their MCUs, Koedel said, citing Infineon MCUs that integrate accelerators for automotive functions including graphical displays and ADAS radar processing.
Microchip’s Thomsen called
AI a game-changer for MCUs involved in real-time closed-loop control. “I think AI is probably the big revolutionary change for the MCU, and in a lot of cases, it’s going to be a revolutionary change for our customers’ applications,” he said.
Will MCUs reach a limit where MPUs become more attractive?
With the MCU experiencing so much innovation and the number of use cases expanding, one wonders at what point the MCU will reach its limit and MPUs will become a more viable choice. The considerations for or against such a shift extend beyond technical specifications alone. Embedded engineering teams invest substantial time and money when selecting an MCU family for their designs, so they will want to stay with that architecture for as long as possible. Also, MCUs typically consume less power and are lower in cost than MPUs. MPUs are typically selected based on a software decision, the choice of interfaces, or purely for performance reasons, whereas MCU selection is more often related to hardware factors.
Bernd Westhoff, Renesas
For some MPU-based applications, there may well be a strong desire to migrate to an MCU, said Infineon’s Tateosian. “Some high-end smart thermostats with displays full of connectivity are MPU based, and some of them are high-end MCU based. The user won’t know the difference, but I can tell you the unit costs of those [two device types] are very different,” he said. “That’s a good example where some developers are willing to make the jump from an MPU to an MCU to save power and cost, and others see the software development effort as prohibitive.”
Bernd Westhoff,
director of IoT product marketing at Renesas, noted that there has always been a degree of overlap between MCU and MPU performance. “MPUs in the past were already at 200 and 400 MHz, and MCUs are easily catching up with that,” he said.
Westhoff also cited some fundamental differences between MCU and MPU developments. “MPU people expect to have Linux, not an RTOS, so you may have a heavy issue in the future to convince a Linux MPU person to become an RTOS MCU person and develop their software there.”
The general-purpose MCU is here to stay
Tim Burgess, Renesas
As MCUs benefit from more functionality, some inevitably become optimized for specific applications. Application-specific MCUs tend to focus on high-volume use cases, such as motor control. Could this trend continue so that the need for general-purpose microcontrollers declines?
“There are always going to be high-volume, low-cost solutions that ASICs [application-specific ICs] are going to take over,” said Microchip’s Thomsen. With shortening product development timescales, he added, the ability to select an MCU off the shelf today and start programming, even if it might be a bit more expensive, will meet a customer’s time-to-market requirements.
Renesas’ Burgess confirmed the continued need for general-purpose MCUs, observing, “It’s just not possible to tune or optimize application-specific processors for every single use case. There are thousands of different applications with so many different requirements, memories, packages, RAM, peripheral mixes, and when you get down to what you’d have to design for, there’s just not enough market to really justify it.”
Fifty years plus and still going strong
Ralf Koedel, Infineon
The microcontroller market continues to experience year-on-year growth thanks to technical innovations and an endless list of use cases. In the automotive market, for example, Infineon’s Koedel told EE Times Europe he doesn’t see an end to growth yet. “If you look into motorcycles, for example, the trend now in India is to go from combustion engines to electrification, with a lot more electronic content, enabling a lot more [MCU growth],” he said.
Of all the MCU use cases highlighted by the executives we contacted, it’s clear that machine learning-based applications will become more significant this decade. With its low-power attributes, an architecture optimized with neural network acceleration, and encryption functionality, the MCU is a suitable choice for this use case.
Robert Huntley
Robert Huntley is a contributor for EE Times Europe.
Tags:
Artificial Intelligence (AI),
Embedded,
ICs/Chips,
MCU,
Semiconductors
Industry races ahead of academia with the number of machine learning models released (32 vs 3 in 2022) 
The demand for AI professional skills is increasing in every single sector. 2% of job postings are now AI related 
The number of incidents of misuses of AI is increasing (think Midjourney’s more than real looking AI generated images of the pope rocking a white bomber jacket) 
Companies that have adopted AI continue to pull ahead with cost savings & revenue increases being the main drivers 
AI is both helping & harming the environment. While AI can be used to optimise energy usage, just BLOOM’s training run emitted 25 more carbon than a one-way trip from NYC to San Fransisco 
Policymakers are getting more and more interested in AI (although this is still lacking behind significantly) 
Citizen’s continue to distrust AI. Only 35% of Americans think positively about AI compared to 78% of Chinese What does this mean? 
AI is here to stay and its development will largely driven by the private sector to increase savings and reduce cost. 
