'Ration' IHS: European and American manufacturers include the top five industrial semiconductor production values ​​in 2017

1.IHS: European and American manufacturers include the top five industrial semiconductor production values ​​in 2017; 2. The reduction of benchmark defects is 诀窍 The increase in vehicle IC yield/reliability; 3. The mature ADAS technology The automotive sensing market is hot

1.IHS: European and American manufacturers include the top five industrial semiconductor production values ​​in 2017;

According to ISK Markit research, the global industrial semiconductor production value in 2017 was 49.1 billion US dollars, with an annual growth rate of 11.8%. Among them, the top five factories are all arranged by European and American operators, and Texas Instruments (TI) is the first place in the city. And ADI, Intel, Infineon and ST are divided into two to five. The agency estimates that the industrial semiconductor market will continue to grow by 2022. The annual compound growth rate (CAGR) is 7.1%.

IHS Markit pointed out that the recovery of the US economy and the strong demand in the Chinese market are the main sources of demand for the industrial equipment market in 2017. In addition, the temperature recovery in the European market has also brought strong momentum to semiconductor growth. Commercial revenues are showing an upward growth pattern. In addition, strategic acquisitions continue to be an important factor in shaping the overall industrial semiconductor market.

IHS Markit's definition of industrial electronics includes LED lighting, digital billboards, digital image monitoring, climate control, smart gauges, tractors, solar photovoltaic inverters, human-machine interfaces, and Medical electronics, etc. The semiconductors used in these devices include optical semiconductors, distributed power components, general-purpose analog components and microcontrollers (MCUs), etc.

In terms of the ranking of industrial semiconductor suppliers in 2017, Texas Instruments ranked the leading position in industrial semiconductors with a revenue of more than US$5 billion; after the acquisition of Linear Technology, Yardno was not only in the industry. The market's territory expanded even more, and the related product revenue reached 2.8 billion US dollars, and the trend jumped to the second place. Intel's revenue from the Internet of Things business continued to be double-digit growth, ranking third with a slight gap.

Infineon, ranked fourth, has market leadership in its power and energy sectors, including distributed power components and power management components in factory automation, traction, and solar, electric vehicles, power supplies, and related products. Revenue continues to grow strongly. The fifth-placed STMicroelectronics, whose industrial semiconductor revenues are derived from many applications such as factory and building automation, uses many of the company's MCUs, analog and distributed components.

2. The benchmark defect is reduced. The car IC yield/reliability is further improved.

The close relationship between yield and reliability of semiconductor ICs has been well studied and documented. The data in Figure 1 demonstrates this relationship. Similar results are available at the batch, wafer and chip levels. In short, the yield is high and the reliability is good. The correlation between yield and reliability is completely unexpected, because the type of defect that causes chip failure is the same as the type of defect that causes early reliability problems. The differences between defects affecting yield and reliability are mainly in their size and their position on the chip pattern.

The close relationship between yield and reliability of semiconductor ICs has been well studied and documented. The data in Figure 1 demonstrates this relationship. Similar results are available at the batch, wafer and chip levels. In short, the yield is high and the reliability is good. The correlation between yield and reliability is completely unexpected, because the type of defect that causes chip failure is the same as the type of defect that causes early reliability problems. The differences between defects affecting yield and reliability are mainly in their size and their position on the chip pattern.

Figure 1 The close relationship between IC component reliability and yield.

Therefore, reducing the number of defects affecting yield in the IC manufacturing process will increase the benchmark yield and improve the reliability of components in actual use. Recognizing this fact, the foundries serving the automotive market face two key factors. The first problem is economic: In order to improve reliability, it takes time, money and resources to increase yield, and what is the appropriate level of investment? The second question is technical: In order to raise the benchmark yield to the necessary level, what? Is the best way to reduce defects?

For OEMs that manufacture consumer electronics (mobile phones, tablets, etc.), "mature yield" is defined as a turning point in further investment in time and resources that does not necessarily increase yield. As the product matures, yield Stabilizing, usually reaching a high value but still well below 100%. Consumer product foundries will redistribute resources to the process and equipment for developing the next design node, or reduce costs to increase the profitability of their mature nodes. Ability, not to pursue higher yields, because doing so is more economical.

For automotive foundries, the economic decision to increase investment in order to increase yields has exceeded the typical marginal benefit decision. When reliability issues arise, automotive IC manufacturers may have to bear expensive and time-consuming failure analysis. And assume the economic responsibility for failure and product recovery during the warranty period of the product, as well as potential legal liability. Considering that the reliability requirements for automotive ICs are two to three orders of magnitude higher than consumer ICs, automotive foundries must achieve even more High benchmark yield level. This requires rethinking the meaning of "mature yield."

Figure 2 highlights the difference between the mature yield of consumer products and automotive OEMs. Any type of fab will increase the yield curve, so almost all of the systemic impact yields have been resolved. The yield loss is mainly caused by random defects in the process equipment or the environment. At this time, the consumer product foundry may consider the yield and reliability to be "good enough" and take the appropriate approach. However, in the automotive industry, generation The factory uses a continuous improvement strategy to push up the yield curve. By reducing the incidence of defects affecting yield, automotive foundries can also reduce potential reliability defects, thereby optimizing their profits and reducing risk.

The automotive supply chain (from OEMs to Tier 1 suppliers to IC manufacturers) is forming a mindset that “every defect is important” and a strategy of pursuing zero defects. They recognize that when potential defects leave After the foundry, it finds and solves the cost of each step forward in the supply chain by 10 times. Therefore, the current method of over-reliance on electrical testing needs to be replaced by the lowest cost strategy, that is, the potential failure in the foundry Stop. Only a methodical implementation of the plan to reduce defects, the foundry can achieve zero defect targets, and can be strictly audited by car manufacturers.

In addition to robust online defect control capabilities, some of the ways that car purchasing managers want to see to reduce defects include:

Continuous Improvement Program (CIP) to reduce baseline defects

. Best equipment workflow

Bad Equipment Improvement Program

Continue to reduce baseline defects

The line defect strategy is the basis for any strict reduction of the baseline defect plan. To successfully detect yield and reliability defects affecting its design rules and component types, the foundry line defect strategy must include appropriate process control equipment and appropriate Inspection sampling plan. The defect detection system used must have the required defect sensitivity, be well maintained and up to specification, and use carefully adjusted inspection procedures. The inspection sampling must be sufficient for the process steps to quickly detect the process or equipment. In addition, there should be sufficient detection capacity to support accelerated anomaly detection, root cause differentiation and risk WIP tracking control plans. With these elements, automotive foundries should be able to achieve a successful baseline defect reduction program. The plan can demonstrate an improvement in yield trends over time, providing further improvement goals and equating industry best practices.

One of the biggest challenges in a baseline defect reduction plan is to answer: Where does the defect come from? The answer is often not so simple. Sometimes, defects are detected after multiple process steps. Sometimes, only after the wafer passes through the other After the process and "decoration" of the defect, it becomes apparent, which means that the defect is more obvious in the detection system. The device monitoring strategy helps solve the problem of the origin of the defect.

In equipment monitoring/device certification (TMTQ) applications, a wafer of wafers is first tested to run in a designated process equipment (or reaction chamber) and then retested (Figure 3). Any new defects must be due to the specified process equipment. The results are clear; there is no doubt about the root cause of the defect. The car foundry pursuing zero defect standards recognizes the benefits of the equipment monitoring strategy: through sensitive testing procedures , Appropriate Control Limits and Out of Control Action Plan (OCAP), can reveal random yield losses from each process equipment and resolve them.

Figure 3 After the “pre-check” detects the reference data of the wafer, the wafer can be used to cycle some or all of the process equipment steps. The “post-test” reveals the defects added to the process equipment.

In addition, as shown in Figure 4, the newly added defects of the process equipment are plotted over time, which provides a record of sustainable improvement that can be audited and used to set future defect reduction targets. The foundry can Classification of defects that occur on each device, and generates a database that can be used as a reference for failure analysis of field failures. This method requires very frequent device certification (at least once a day), usually with the best device workflow discussed below or Bad equipment improvement plan is used together.

Figure 4 Continuously improve the cleanliness of the equipment over time. The root cause of the problem is clear, and the defect reduction target can be set objectively quarterly or monthly. In addition, comparing the defects of the two process equipment can show which machine Cleaner. This helps guide equipment maintenance activities and locks the cause of discrepancies between devices.

AWF/bad equipment improvement plans have their own advantages

The best equipment workflow is another strategy used by foundries to meet the zero defect standards required by the automotive industry. With the best equipment workflow or automotive workflow (AWF), wafers for automotive ICs are only at the fab. Running in the best process equipment. This requires the fab to know the best machine for any custom process. To reliably determine which machine is the best, the foundry uses the online and equipment to monitor the detected data and then only those Machines are used in automotive workflows. Limiting automotive wafers to a single device at each process step can result in longer cycle times. However, compared to process flows with higher defect rates that can lead to reliability issues This approach is still preferred for automotive wafers. Coupled with a methodical continuous improvement program, most foundries can usually achieve multiple AWF-compliant equipment in each step by setting a quarterly defect reduction goal.

Because this method is difficult to scale, the best equipment workflow is best suited for small-scale WIP-based OEMs. For foundries that produce automotive products in large quantities, priority should be given to more streamlined continuous improvement programs, as shown below. The method of improving bad equipment.

The Bad Equipment Improvement Program is the opposite of the best equipment workflow because it can proactively address the worst process equipment in any given process step. The foundries that have the greatest success in reducing baseline defects often use poor equipment to improve their plans. They first down the worst device in each process step and adjust the device until it exceeds the average of the rest of the devices in the same group. They repeat the process over and over until all devices in the same group match Minimum Standard. An effective bad equipment improvement program requires the factory to have a well-organized equipment monitoring strategy to certify each process equipment at each step. At least one certification process is required to complete each day on each equipment. Be sure to collect enough data to have ANOVA or Kruskal-Wallis analysis determine the best and worst equipment in each group. A bad equipment improvement program will schedule downtime for process equipment and is known to upgrade the entire fab to the car. One of the fastest ways to standardize. By improving yield and reliability, the strategy finally mentions Effective productivity and profitability of automotive foundries.

(The author of this article is KLA Senior Director and Chief Scientist) New Electronics

3. The ADAS technology is mature. The car sensing market is hot.

In recent years, countries have been pushing ASAS into safety regulations, and related technologies have matured. Therefore, the demand for 3D sensors for all types of smart cars is growing.

Smart cars are the future potential of the communication manufacturers and traditional car manufacturers to compete in the AIOT market. At present, the advanced driver assistance system (ADAS) technology has matured, and the next stage of international manufacturers will work hard to monitor the road conditions of the L3 level self-driving system (Table 1), and the “environmental awareness” capability is a basic requirement, and the vehicle 3D sensing device is an indispensable component. The smart car's 3D sensor is different in distance and use, and mainly includes:

Ultrasound radar (Ultrasound):

Short detection distance (<6M), 用于侧撞警示及停车辅助系统.

. Car camera:

Medium detection distance (<100M), 主要用来辨识路标与障碍物, 但易受浓雾, 强光, 大雨等天气影响.

Optical Radar (LiDAR):

Referred to as Guangda, long detection distance (150M), high precision, can quickly establish a 3D geographic information model of the surrounding environment.

Millimeter wave radar (mmWave Radar):

The detection distance is long (100~250M), which can identify obstacles and is less affected by the environment (night or bad weather), but the accuracy is limited.

ADAS regulations drive the car after sensing

In recent years, countries have been pushing ASAS into safety regulations, and it is estimated that the demand for various 3D sensing devices will be driven in the short term (Table 2). For example, in 2018, the United States will force new cars to install RVC and LDW, and mainland China will incorporate ADAS into safety regulations. 2020 In the year, the United States, the European Union, Japan's new car will be forced to install AEB, while mainland China will include FCW, AEB, LDW, PDS into the safety score.

However, in order to improve the reliability of automotive environment sensing, manufacturers usually integrate more than two types of sensors to obtain more accurate data, improve the environment around the car, the ability of pedestrians, reduce the probability of accidents, and achieve autonomous driving or The advanced auxiliary driving function of the smart car.

Car camera development is maturing

In general, short-range ultrasonic radar technology is quite mature, almost the standard for existing new cars. And car cameras must have better durability, high photosensitivity, high dynamics, and The new car must be equipped with at least 5 cameras (one for the front-view depth of field lens, two for the left and right side-view lenses, and two for the front and rear myopia lenses), so the potential business opportunities should not be underestimated.

Wide range of millimeter wave radar detection 77GHz is the development focus

Millimeter wave radar has a wide range of detection, ranging from medium range (24 GHz) to long range (77 GHz). At present, the focus of large-scale development is at 77 GHz, such as Infineon, NXP and ST. Among them, the millimeter wave chip (MMIC) is an important component, mainly transmitting and receiving microwave signals. The process is mainly silicon germanium (SiGe), and the cost is lower than that of gallium arsenide (GaAs). It is expected that the new car will be equipped with a long-range radar in the future. The medium-range radar with four corner positions is also expected to have the same demand. The CMOS process has lower production costs, but the current technology is not yet mature, and it is the development direction of the next stage MMIC.

LiDAR is developing in mass production L3 self-driving is expected to accelerate touchdown

Lida is a distance measurement based on the principle of laser light TOF. It is not affected by the ambient brightness. It can be used to sense the surrounding environment to establish a 3D geographic information model. It is one of the necessary sensors for automatic driving. Traditional mechanical light Because of its large size and high cost, independent equipment is difficult to integrate into the design of the car body. Therefore, both Dachang and the startup have developed a miniaturized Solid-State LiDAR, which integrates optical scanning and sensing components or a single CMOS chip. All solid state light), in order to achieve the needs of small size and mass production economy.

At present, ADAS emphasizes auxiliary functions, and can only achieve speed or direction local control. To achieve advanced driving of L3~L5, LiDAR can complement the data collection capability gap of other sensors in complex environments, with simultaneous positioning and map construction. (SLAM) realizes real-time navigation function, and looks good in the future.

Figure 1. The smart car is equipped with many sensing components.

(The author of this article worked at the MIC of the CM) New Electronics

2016 GoodChinaBrand | ICP: 12011751 | China Exports