A Thousand Reasons to Choose Analog Devices as your Intelligent Edge Partner

"Edge computing" is not a new concept. In fact, in the 1990s, Akamai launched the content delivery network (CDN). The idea was to introduce nodes geographically closer to end users to deliver cached content such as images and videos.

Similarly, artificial intelligence also has a long history of evolution. John McCarthy, a young professor at Dartmouth University, coined the term "artificial intelligence" while writing a proposal for a seminar.

But when two technologies collide to produce edge intelligence, a new term is born - intelligent edge. The edge in the intelligent edge refers to the boundary between the physical world and the digital world, such as various temperature sensors, acceleration sensors, acoustic sensors, etc. These sensors connect the real physical world and the invisible and intangible digital world. By superimposing intelligent features on the edge, nodes can be equipped with services such as advanced data analysis, scene awareness, real-time decision-making, self-organization and collaboration.

The term intelligent edge has appeared in the public domain more than once, but it is not common among traditional semiconductor companies. At this year's ADI Annual Investor Day, ADI CEO Vincent Roche proposed for the first time the corporate vision of "Empowering the Intelligent Edge" in addition to "ADI connecting the physical world and the digital world".

What is the smart edge used for?

The term intelligent edge is rapidly fermenting under the promotion of ADI and a series of other companies. For example, at CES 2023, ADI launched more than 30 demos focusing on the four major fields of automobiles, consumer electronics, digital healthcare, and industry to demonstrate the role of intelligent edge technology in continuously improving efficiency, security, and sustainability.

Also in Taiwan last month, ADI also held a seminar with the theme of "Intelligent Edge", focusing on the four aspects of Sensor Edge, Insight Edge, Cyber ​​Edge and Power, describing ADI's empowerment.

Why is ADI so focused on the intelligent edge? The important reason is that the scope of the intelligent edge is so broad that it is enough to change the future of the world. In a broad sense, what people come into contact with in daily life are edge devices. “Using the intelligent edge, factory operators can integrate information technology into operating processes to improve energy efficiency, production and uptime; automakers can promote the development of driver assistance systems from relatively basic functions to advanced autonomous driving systems; Medical institutions can even predict health problems and prevent potential risks, instead of waiting for patients to become ill or develop critical illness before treatment," said Xin Yi, senior engineer at ADI China Technical Support Center.

The edge has inherent advantages over the cloud. Through calculations closer to the data end, problems easily caused by the network such as bandwidth, delay, congestion, and security can be solved, while at the same time bringing huge advantages in power consumption and cost. "If there is a fire, take water from the sea. Although there is a lot of sea water, the fire will not be extinguished. Water far away will not save the fire nearby." Xin Yi interpreted the value of the edge with a sentence from "Han Feizi·Shuo Lin Shang".

Xin Yi said that in 2025, 75% of data will be generated and processed on the edge. Therefore, massive data requires the intervention of artificial intelligence for rapid judgment to fully realize the value of the data. For ADI's actual customer applications, it has far exceeded the fields that consumers are aware of, such as forest fire monitoring, collapse monitoring of roads, railways or dams, power grid security detection and other application scenarios, thereby achieving rapid alarm functions.

What layout does ADI have at the intelligent edge?

ADI has always been focused on building a bridge of communication between the physical world and the digital world. With the advent of the intelligent edge era, ADI still relies on a broad and in-depth platform strategy to enable broader empowerment, but it still does not break away from the physical and digital world. This mission of the Border Bridge.

As shown in the figure, ADI promotes the evolution of the intelligent edge through three levels: core technology, embedded digital and software, and solutions. Basic technologies include radio frequency, converters, amplifiers, interfaces, power supplies, and sensors. At the embedded level, they include processors, algorithms, communications, security, and AI/machine learning. At the solution level, they include VSM, 5G, wireless, etc. Vertical solutions such as BMS and motion control, as well as general platform solutions such as high-precision solutions, high-speed solutions, power management, etc.

One example is in the RF high-speed market, ADI has launched the MxFE analog front end, which provides intelligent integration of RF analog-to-digital converters (ADCs), RF digital-to-analog converters (DACs), on-chip digital signal processing, and clocks/PLLs to enable multiple Chip sync. This eliminates discrete data converters and traditional RF front-end components to reduce size, weight, power consumption and cost. Integrated digital signal processing supports the move to software-defined radio platforms.

Also in radio frequency applications, ADI has launched a 5G radio unit platform that complies with O-RAN specifications, including all core functions required for O-RAN compatible 5G radio units, including baseband ASICs, software-defined transceivers, signal processing and power supply. This advanced radio platform is designed to significantly increase performance and improve form factor to meet the severe power consumption and cost challenges that next-generation networks will face, while shortening customer design cycles.

MCUs are the key to the intelligent edge

With a broad and deep layout, ADI Intelligent Edge will empower more partners. Among them, MCU is obviously the most representative product of edge intelligence.

A single-chip MCU can integrate multiple functions such as acquisition, conversion, processing, calculation, decision-making, output control, etc. It is the smallest "brain" unit in electronic systems and covers thousands of markets. In fact, according to IC Insights statistics, global MCU shipments reached 31.2 billion units in 2021.

Xin Yi emphasized that when ADI announced the acquisition of Maxim in 2020, it attached great importance to Maxim's digital products, including MCUs. "Previously, ADI also had digital chips, but they were mainly aimed at professional applications. Maxim's MCUs are mainly general-purpose, including security, low-power consumption, and AI MCUs. Therefore, the integration of the two parties has achieved complementary advantages."

Xin Yi said that since 1995, ADI's MCU product shipments have exceeded 1 billion pieces, and it has continued to innovate and change in accordance with market demand. For example, before 2002, the company mainly used the 8051 core. In 2004, it began to launch its own MAXQ core series MCUs. After 2012, it launched Arm core products. After 2020, it launched smart edge MCUs. These milestones marked the leadership of ADI. The determination of the MCU trend.

In terms of general-purpose microcontrollers, ADI's shipments are not at the forefront, but its distinctive features have gained market recognition in many fields. "Low power consumption, excellent interface, new communication, comprehensive evaluation, easy development, and strong security" are Xin Yi's evaluation of ADI MCU. In particular, it has more than 30 years of experience in IP protection and root of trust, making it very suitable for security-sensitive applications.

Independently develop CNN IP and embark on the first train of intelligent edge

Xin Yi said that in fact, Maxim began to develop CNN IP in 2013, and launched the industry sensation MAX7800X series of intelligent edge MCUs in 2020. It does not require networking and supports battery power supply, making it very suitable for AIoT applications.

How powerful is the MAX7800X? Let’s take stock.

MAX78000 block diagram

As shown in the figure, MAX78000 has several major features:

In terms of computing, the first is the dual-core heterogeneous structure, which includes the Arm Cortex-M4F master, which is responsible for application execution and management, and the RISC-V core, which is responsible for DMA. Data is transferred to CNN through RISC-V, and M4 can sleep. , thus saving more energy consumption.

Another main feature is the self-developed CNN IP. Since it is a hardware-accelerated IP and does not require MCU intervention at all, it is faster, simpler to operate and more efficient than software-based AI execution.

This CNN has 64 parallel processors, supports 432KB weights, and supports up to 3.5M weights.

CNN has 512KB of independent storage space and does not need to occupy main memory. The memory is located around the processor, so it can significantly reduce the delay in the data transfer process. This product uses flow control technology, which can perform line-by-line recognition of convolutions and support larger image recognition in a limited space, thereby reducing costs.

MAX78000 layout, CNN accelerator accounts for the vast majority

Other features include integrated SIMO power, eliminating the need for an additional PMIC, and support for security encryption including AES, TRNG and Unique ID. In terms of clock source, except for an external 32kHz crystal oscillator, the rest are integrated internally.

In terms of peripherals, it includes rich interfaces such as ADC, UART, I2S, I2C, SPI, etc., and also supports special buses such as camera parallel port and 1-Wire.

The MAX78000 MCU+CNN accelerator solution achieves a hundredfold speed improvement compared to traditional MCUs, or only requires 1% of the power consumption level, and its cost (official website is about US$10) is between MCU and FPGA.