Is driverless driving safer than human driving? Where are the opportunities for domestic automotive chips?

The 4th IC Alumni Forum organized by the Integrated Circuit Branch of the Shanghai Jiao Tong University Alumni Association, which was held concurrently with IC China 2019, was divided into two themes. The first half was on "Artificial Intelligence and AI Chips", while the second half focused on the topic of automotive electronics.

Xie Zhifeng, vice president of the Integrated Circuit Branch of the Shanghai Jiao Tong University Alumni Association and founder of Aixin Education, hosted the forum. He said that Shanghai Jiao Tong University has made many achievements in artificial intelligence chips, and that autonomous vehicles will also use artificial intelligence technology. Xie Zhifeng then gave an example to illustrate that autonomous driving is safer than human driving, and said that he is looking forward to the future of autonomous driving.

The "New Four Modernizations" of Automobiles Start with "Chips"

The theme of the speech by Cheng Jinglei, Chairman of Jiazhou Automotive Technology (Shanghai) Co., Ltd. and President of Zhongde Capital Partners, was "The New Four Modernizations of Automobiles Start from the Chip". Cheng Jinglei interpreted the chip industry from the perspective of automobile demand, covering the overall requirements of the "new four modernizations" of automobiles, the current status, dynamics and development trends of automotive chips.

Cheng Jinglei, Chairman of Jiazhou Automotive Technology (Shanghai) Co., Ltd. and Partner and President of Zhongde Capital

Today, the development of science and technology has driven tremendous changes in business models in the automotive industry, which rely on four major aspects: 1. Electrification as the basis; 2. Networking as the condition; 3. Intelligence as the key; 4. Sharing as the trend.

The question now is, how to do it?

Cheng Jinglei said that there is no solution that meets the required requirements. Even though the speech on AI mentioned that there are more than 200 companies in this field, when will it be used in automobiles? Although some AI companies have released chips, they are still in the demo stage.

From the perspective of automotive electronics, electronic components currently account for 30% of the cost of a car, and this figure may rise to 50% in the future. The areas covered by automotive electronics are also very broad, ranging from powertrain to body digitalization, energy management, and safety.

So what are the requirements for the chip?

Cars are a comprehensive component of the Internet of Things, and their requirements for power consumption, computing power, security, and cost are higher than those for consumer chips. Cheng Jinglei said that in the current autonomous driving demos around the world, even at the L3 level, the chip power consumption cannot meet the size requirements, not to mention the computing power of the L4/L5 level.

Automotive chips can be summarized into three directions: perception, decision-making and behavior control; further divided into six categories: sensing, radio frequency, display, power and drive, storage, and logic. To analyze the status and competitive advantages of domestic manufacturers, we must analyze them one by one from the six perspectives and cannot make a hasty judgment.

Why is it so difficult to make automotive chips? Why is it that from mechanization to electrification of automobiles, only a few companies such as NXP, TI, and Renesas can provide automotive-grade chips? Why can't other companies enter this market? The industry chain behind this is very complicated, from basic research and development to chip design, component design, and OEMs. Although there are so-called pure chip design companies, in the automotive industry, in order to consider the integrity of the car, the ODM form is usually adopted. The OEM, component manufacturers, and semiconductor technology companies collaborate to innovate and ensure the development of the industry.

In addition, automotive chips have a strong demand for miniaturization and integration. More than 200 computing chips are integrated in the car, so they must be miniaturized. Today, high-speed computing is enough for general companies, but it is far from enough for cars. To solve L4/L5 level autonomous driving, the number of floating-point calculations must be exponentially higher.

Regarding the issue of running internal vehicle codes, the 4,000 types of codes in self-driving cars must run on different chip architectures, and must also have a common instruction set and system-level platform architecture.

Regarding automotive chip standards, the international testing standards for automotive chips are still relatively traditional and are based on electronic devices. In Japan, the United States, and Europe, automotive and electrified chip standard systems are still being promoted. Currently, Shanghai Jiaotong University's artificial intelligence has research in this area, and some domestic colleges and universities still have a lot of room for improvement. The development of domestic vehicle autonomous driving and chip technology needs to achieve differentiated competition with international competitors, which will be of great benefit to the development of domestic automotive chips.

BMS for new energy vehicles

Hou Yizhen, senior manager of the electronics and electrical forward-looking technology development department of CATL, gave a speech entitled "BMS for New Energy Vehicles". According to Hou Yizhen's on-site introduction, CATL was founded in 2011 and mainly deals with battery cells and battery management related to electric vehicles. "In recent years, CATL has cooperated with almost all vehicle manufacturers, with customers all over the world. Currently, 800,000 vehicles are equipped with our BMS system, and the entire BMS team has 460 people."

Hou Yizhen, Senior Manager of Electronic and Electrical Advanced Technology Development, R&D Department, Contemporary Amperex Technology Co., Ltd. (CATL)

Hou Yizhen said: "In the whole vehicle, in addition to the computing requirements of autonomous driving, BMS is the second most demanding product. Battery cells are very complex, and the current battery management system still cannot manage all battery cells very accurately. The next generation of BMS will have more complex algorithms for management and calculation, and may use very high-end products (processors) on the market today."

BMS chips are the chips most likely to be replaced by domestic ones. In Hou Yizheng's impression, there are no Chinese ones among the hundreds of thousands of chips in use. The millions of battery management chips purchased by CATL this year are almost all from American manufacturers.

According to Hou Yizhen, CATL will purchase 300 million automotive-grade battery sample chips and 30 million high-end automotive MCUs in the next 10 years, with a purchase amount of more than US$1.5 billion. He raised a question: "When can we use domestic chips to get rid of the tight supply chain situation?"

Hou Yizhen then put forward several difficulties and suggestions for the localization of battery management system chips: First, in terms of design, the next generation of high-performance battery cell sampling chips and high-performance processors must meet the requirements of the functional safety architecture; secondly, in terms of process, the BCD semiconductor process must meet the requirements of high operating voltage (>100V) and high-precision digital-to-analog conversion (±1mV) under complex electromagnetic environments and working conditions; next is the quality issue, which needs to meet the quality requirements of automotive-grade semiconductor products and have a quality assurance system from raw material processing to chip production, packaging and testing; finally, cost, while meeting functional and performance requirements and achieving low-cost goals, it is necessary to balance functional requirements, design, process, production and other requirements.

We hope that domestic chips can prioritize quality, iterate quickly, and eventually gradually complete domestic substitution.

Everything is connected, and the world is smart

The theme of the speech by Sui Ronghua, CEO of Huihan Microelectronics Co., Ltd. (hereinafter referred to as "Huihan"), was "Everything is Connected, Smart Travel in the World". At the beginning of the speech, Sui Ronghua briefly introduced Huihan's business: "Huihan is a leading domestic provider of Internet of Vehicles and Internet of Things solutions. With autonomous driving and smart travel as the ultimate goal, and intelligence and networking as the direction, it creates the 'digital building blocks' of autonomous driving and smart travel."

Sui Ronghua, CEO of Flaircomm Microelectronics Co., Ltd.

Sui Ronghua believes that the automobile industry is "pessimistic in the short term, but cautiously optimistic in the long term." From the long-term perspective of the international automobile industry, when the number of cars per thousand people reached around 150-200, the major automobile consuming countries saw a shift in the growth rate of automobile sales, from a double-digit growth rate to a growth rate of around 0-5%. We judge that the Chinese automobile industry will most likely continue this pattern.

From a short-term perspective, there are multiple reasons for the negative growth in the automobile industry since the second quarter of 2018, some of which have gradually improved since the Spring Festival of 2019. We judge that the industry's sales growth rate will most likely gradually recover from the third quarter of 2019, and the current market's industry expectations are too pessimistic.

The automobile industry is the jewel in the crown of human industrial civilization. Autonomous driving is the ultimate dream of the automobile industry, which integrates the most cutting-edge information technologies such as artificial intelligence, 5G, chips, sensors, and algorithms. However, the process of realizing autonomous driving is still long, and many technologies need to be continuously polished and improved in the process, but they can be split and industrialized in advance. This is also the background of Flaircomm's autonomous driving "digital building blocks" strategy.