January 23, 2021
Recently, the new auto industry has moved frequently. Tesla's market value has surpassed the market value of other traditional auto giants in just one year, and China's new car three outstanding has also encountered the market. In the semiconductor industry, Nvidia, Qualcomm, Apple and other giants are also actively deploying the smart car industry, and this article will analyze the new markets and opportunities that the smart car industry brings to semiconductor chips.
Smart cars may repeat the story of smart phones
The smart car industry began to become popular four years ago, and as the products of a new generation of car companies represented by Tesla began mass production and entered mainstream applications, the portrait of the next generation of cars began to become clear.
Different from the speculation that the next generation of cars will be driven by autonomous driving, it seems that the real driving force of the next generation of cars will be intelligence, and autonomous driving will only be a component of intelligence. The so-called smart car, the process is similar to the previous smart phone-in the process of smart phone, the technical process is to introduce a standardized hardware platform and operating system, and allow developers to implement various apps in the operating system to improve The user experience greatly expands the monotonous user experience of the original feature phone that mainly used to call and send text messages.
We believe that smart cars may also follow a similar path. Through the unification of the underlying hardware platform interface and operating system, smart cars will provide users with a new user experience, and this user experience may be upgraded through software. Last week, automotive media "Automotive News Europe" conducted an exclusive interview with Nvidia CEO Huang Jianxun. During the interview, Huang Jianxun believed that the future of smart cars will have a user experience that can be upgraded with software, thereby bringing a new business based on value-added services Model, which further confirms our point of view.
Just as the intelligentization of mobile phones brought new markets to the semiconductor industry in the past, smart cars will also bring new markets and opportunities to the semiconductor industry.
Chips needed for smart car multimedia
If you ask what is the first impressive change that the smartphone gave users, I believe most users will answer the touch screen, because the user interface is often a key element that can quickly grab the user's heart. We believe that this is also true for smart cars, so our first optimistic opportunity is the chip market for smart cars and user interaction.
At present, the user interface of the car is still not optimized enough, or it is completely stuck in the old-fashioned button design with the dashboard, or just copy the touch screen of the smartphone. We believe that smart cars will have innovations in both the user input interface and the user output interface, and these innovations are opportunities for semiconductor companies.
First of all, in terms of user input, smart cars need to be able to provide users with a non-interference input method, that is, users can complete the interaction with the smart car without leaving the steering wheel while driving. At present, the mainstream technology path in this field is to use voice control technology, that is, based on voice recognition technologies like Amazon Alexa or Baidu Xiaodu, extract relevant information from the user's voice input and complete corresponding commands. Obviously, this requires the installation of microphones in the car. More importantly, in order to achieve a better user experience and recognition accuracy, multiple microphones may be required to form a microphone array. This is because the environmental noise is often large during car driving. In order to allow the voice recognition system to distinguish the user’s voice from the environmental noise, one method used is to use multiple microphones to form a microphone array, and use beamforming technology to Improve the signal-to-noise ratio. In addition, considering the car's tolerance to stability and extreme noise, a smart car may need to install multiple microphones that can meet vehicle regulations and have a large dynamic range to implement a voice input interactive system. Therefore, the smart car market is a new opportunity for microphone chip companies, whether it is traditional companies such as Knowles or emerging companies such as Vesper. In addition, for user voice, smart cars need to be able to complete voice recognition and related operations within a short delay, so a powerful computing chip is needed to complete the processing. We will analyze the opportunities in this area in detail later in this article. .
In terms of user output, the most promising technological path at present is to use HUD technology to project the dashboard and other information onto the front windshield, and use an AR-like approach to achieve an immersive driving experience so that the user’s line of sight The information provided by the smart car can be received at the same time without leaving the road.
In terms of HUD, its core growth market lies in projection-related chips. At present, there are already many semiconductor companies striving to enter the relevant market. For example, TI's DLP (Digital Light Processing, digital light processing) chip has great potential in the HUD field. The technology used by DLP is a micro-mirror array, which uses digital signals to control the characteristics of the mirror surface in the mirror array, so as to achieve precise optical path control and projection effects. However, its main disadvantage is high cost. In addition to DLP mirror arrays, another well-known technical path is laser scanning technology, which uses lasers to project one pixel at a time, and uses MEMS to achieve rapid scanning of the pixels to complete image projection. In the field of HUD projection, the speed of related chip technology innovation is relatively fast. In the next few years, it is expected to see the development and competition among these different technologies. It is still difficult to say which technology will become the most mainstream, but the market’s The prospects are beyond doubt.
Sensors empower the core experience of smart cars
If the user interaction system is the gateway to smart cars, then assisted driving and even smart driving are the core experience of smart cars. In this field, we believe that autonomous driving is still some distance away from mass popularization, but various levels of assisted driving (such as semi-automatic lane changes, fixed-distance cruises, etc.) have been rolled out on a large scale, and it is expected to be in the next few years. Upgrade steadily. Whether it is autonomous driving or assisted driving, a new set of sensor systems is needed behind it.
The sensor system required for this intelligent driving will include cameras, millimeter-wave radar and even LiDAR and other sensors, and we believe that the largest market increase in the next few years is the high-precision millimeter-wave radar chip. At present, traditional millimeter-wave radar chips have been widely used in blind spot monitoring, reversing and other fields, so millimeter-wave radar has been technically recognized by the mainstream. The traditional millimeter-wave radar is mainly point detection and ranging, while the next generation of high-precision millimeter-wave radar needs to be able to scan an area and achieve ranging (rather than focus on the ranging of a point), and act as a camera An important addition. Technically, this requires millimeter-wave radar to complete beamforming and scan the beam in the ranging area, which increases the complexity of the radar system by an order of magnitude. For millimeter-wave radar chips, the next generation of high-precision millimeter-wave radars requires multiple millimeter-wave transceivers and related digital signal processing modules to be integrated on the chip to implement a small SoC, which requires more work than traditional millimeter-wave radars. A lot higher.
At present, the high-precision millimeter-wave radar chip market is still on the rise, and many chip companies have layouts in this field, and Chinese companies such as Gatland also have a place in this market. The market for high-precision millimeter-wave radar is relatively large, and products have certain thresholds but are not unattainable. Therefore, we expect more domestic and foreign companies to enter this field in the next few years as the market heats up.
Chip opportunities for computing platforms
Perhaps the most critical technical difference between smart cars and traditional cars is that there is a powerful central processing system that can run a general-purpose operating system and achieve an upgradeable user experience, while the aforementioned multimedia, sensors, and other systems will be compatible with this central processing system. Combine organically. For example, user input in a multimedia system requires the central processing system to provide sufficient computing power to complete neural network calculations, and the HUD in a multimedia system requires the central processing system to complete the rendering algorithm to convert the information to be displayed into the picture to be projected. The central processing system does sensor fusion based on the input of the sensor system and takes corresponding actions at any time, such as steering wheel correction, acceleration and deceleration, etc.
We believe that the future smart car central processing system chip will be a complex SoC, which has completed multiple input and output processing. First, it needs a strong processor core to run the operating system and some core driving logic. In addition to the processor core, a large amount of computing power is needed to complete functions such as assisted driving and user interaction (we believe that these two types of functions will be completed by a large number of neural networks in the future, and neural network calculations require a lot of computing power). In addition, for some specific functions (such as rendering), dedicated modules will be placed on the SoC to complete efficient processing.
In the field of smart car central processing units, similar to smart phones, we expect there will be two ecosystems: open and closed. The open ecosystem is basically led by a certain manufacturer and forms a standardized software and hardware alliance. There can be multiple software and hardware solutions and platforms under the same standard, similar to the Android ecosystem in smartphones; while the closed ecosystem is based on operation The system to the hardware are all in the hands of one manufacturer, similar to the Apple iOS ecosystem.
At present, Qualcomm, Nvidia and other chip companies are already actively deploying smart car central processing chips. We believe that their products are most likely to target the open ecosystem, because for chip companies, the shipments generated by the open ecosystem are often far greater than the closed ecosystem. . For example, Qualcomm's Snapdragon series of automotive platforms use Snapdragon SoC to provide processing capabilities, which are more suitable for the processing capabilities of this generation of smart cars. I believe that in the future, as the mainstream of various modules (such as HUD, etc.) in smart cars will gradually join More dedicated processing modules; while Nvidia is more aggressively starting from the computing power side, using GPU to provide powerful computing power for smart cars (especially assisted and autonomous driving), which is expected to provide empowerment for the next generation of smart cars. One of the two sets out to meet current customer needs, and the other is more forward-looking. We expect that at some point in the future, the two will converge to a closer solution or become complementary parts of the same system.
In terms of closed ecology, next-generation smart car companies know best what their pain points are, so they can combine car products and operating systems to make targeted chip designs to empower the core features of their products. For example, Tesla has made a large investment in the central processing chip of smart cars. The target product is a chip with high computing power to support autonomous driving, because this field is Tesla's ambitious and huge investment bet. We expect Apple will also design its own chips for automotive products. We believe that compared with the smart phone market ten years ago, the current smart car market is likely to have more closed ecosystems. The main reason is that the semiconductor industry foundry industry has further matured and the threshold has been further reduced. Therefore, smart car companies will have more The possibility of self-developed chips, so as to realize the empowerment of core differences.
The design of the smart car central processing unit will largely determine the specific characteristics and capabilities of the smart car, and even the shape and business model of the smart car. It is expected that many important giants will enter the market in the next few years, and our HOREXS will continue Keep close attention to meet the market needs to manufacture the chip packaging substrates the market needs.