Category Archives: LCD

Chip packaging technology has become a must-have in the semiconductor industry

Chips continue to be miniaturized, and the manufacturing process is advancing towards smaller 5nm and 3nm. Moore’s Law has been repeatedly passed to the end, and chip packaging technology is generally considered to be an important development direction of semiconductor technology in the next stage.

Dr. Yu Daquan, Distinguished Professor of Xiamen University and founder of Yuntian Semiconductor, once pointed out that as the development of Moore’s Law slows down, using advanced packaging technology to meet system miniaturization and multi-function has become a new engine for the development of the integrated circuit industry. Packaging technology came into being with the invention of integrated circuits, and its main functions are to complete power distribution, signal distribution, heat dissipation and physical protection. With the development of chip technology, packaging is constantly innovating, and the supply chain is facing a big test.

1. The heroes compete for advanced packaging

Advanced packaging technology can relatively easily achieve high-density integration of chips, miniaturization of volume, and lower cost, which is in line with the trend of high-end chips evolving towards smaller size, higher performance, and lower power consumption.

Especially high-density advanced packaging (HDAP) such as CoWoS (Chip On Wafer On Substrate), FOWLP (fan-out wafer level packaging) and WoW, which have shown great advantages in improving chip performance, have successfully attracted major mainstream chip manufacturers. The attention of packaging and testing, foundry and design manufacturers has begun to continue to invest in this field.

For example, CoWoS, which is a 2.5D packaging technology introduced by TSMC, is called wafer-level packaging. CoWoS is aimed at the high-end market, and the number of connections and package size are relatively large.

Since mass production of CoWoS began in 2012, TSMC has packaged multiple chips together through this packaging method of sharing substrates between chips, and the bare chips on the plane are interconnected through Silicon Interposer, which achieves small package size and high transmission speed. High, low power consumption, the effect of less pins. Also, FOWLP, a technology predicted to be the basis of next-generation compact, high-performance Electronic devices. According to Yole data, the total global output value of FOWLP is expected to exceed 2.3 billion US dollars in 2022, and the CAGR (compound annual growth rate) between 2019 and 2022 will be close to 20%.

It is reported that the first generation of fan-out packaging uses Infineon’s embedded wafer-level ball gate array (eWLB) technology, which was launched by Freescale (now NXP) in 2009. However, integrated fan-out packaging (InFO) was only produced by TSMC before this. TSMC’s InFO technology is the most striking example of high-density fan-out, and Samsung Electronics’ System LSI Division believes that it is this technology that led TSMC to grab Apple’s (Apple) A10 processor foundry orders. For this reason, Samsung Electro-Mechanics (Semco) stepped into the semiconductor packaging market and cooperated with Samsung Electronics to develop FOWLP technology, with a view to fully confronting TSMC in the new round of customer order competition.

According to Techsearch International, these HDAP technologies are driving the industry’s need for device-package co-design and new processes. At present, foundries have begun to use part of their limited production capacity for advanced packaging and testing, and traditional packaging and testing manufacturers are gradually upgrading to advanced packaging and testing. Although so far, foundries and packaging and testing manufacturers have not completely crossed businesses, and they are fighting independently in their respective fields. But in the future, the two sides will definitely enter more and more overlapping fields, and advanced packaging will become a battleground for military strategists.

 2. HDAP, it’s not easy to say I love you

However, there are still some challenges to achieve something similar to HDAP.

Data show that, first of all, HDAP is heterogeneous. Even if upstream EDA vendors have modified traditional tools to handle a variety of new technologies, these new technologies also require physical verification, such as design rule checking (DRC), layout and schematic comparison (LVS) and so on. In HDAP, the connection must be made through an interposer or some type of interconnection technology, in which case it affects the interconnection characteristics of the system, and these characteristics also affect each other, while affecting manufacturability characteristic design.

Second, the new HDAP technology requires the design team to work together to optimize the entire system, not just individual components. Increased engineering costs due to issues such as chip and package/interposer misalignment, connection errors between components and pads; manufacturing delays due to quality or errors in manufacturing data; and poor signal and power integrity performance, 2.5 Problems such as functional failures caused by unqualified thermal stability of D assembly.

Furthermore, HDAP also significantly increases the design complexity, needing to describe all interconnections from chip to substrate, from interposer to substrate, substrate to circuit board, and substrate to test board. This is very difficult to control in the traditional packaging industry. At present, many of them have to be integrated manually, with some piecemeal inspections.

As a new direction, HDAP has also begun to affect the design process of semiconductors. These new technologies enable partitioning of the overall design, like an interconnect with features on the outside of the chip that are very similar to the inside of the chip. Advanced packaging enables manufacturers to integrate different technologies (processes), but it also brings related challenges to traditional design tools.

On this basis, the traditional packaging design can no longer meet the changing needs of the market. How to efficiently complete the design and get it verified will bring new challenges to EDA tools. The market urgently needs new and more efficient processes, methods and design tools.

 3. Mentor’s weapon

Mentor is an EDA manufacturer that pays great attention to advanced packaging technology. Lincoln Lee, technical director of Mentor Asia Pacific, mentioned that as an EDA manufacturer, its products run through all aspects of design and packaging.

As early as more than 10 years ago (2007), Mentor saw potential opportunities in the packaging market and began to design solutions for leading customers. He emphasized that the semiconductor industry is developing rapidly, and advanced packaging is gradually becoming the main force. If it cannot quickly adapt to customer requirements, it will be left behind. It is a win-win for Mentor and its customers to make efforts in the field of advanced packaging.

In 2013, Mentor officially launched the Xpedition Package Integrator (XPI) high-density advanced packaging (HDAP) process, which is the industry’s first comprehensive solution for today’s advanced IC package design and verification.

According to the data, XPI products already have a high degree of integration, but based on the need for division of labor in the entire design process, Mentor splits the two functions of XPI into Xpedition Substrate Integrator tool and Xpedition Package Designer technology. The unique Xpedition Substrate Integrator (xSI) tool enables rapid definition and optimization of heterogeneous substrate package assemblies. New Xpedition Package Designer (xPD) technology implemented for physical packaging ensures that design signoff and verified data are synchronized. Caliber 3D Stack technology can perform complete Signoff DRC/LVS verification for various 2.5D and 3D stacked chip components.

In the process of continuous optimization, Xpedition can work with multiple people without splitting and avoiding multiple merging, so as to maximize team work efficiency. At present, Mentor is trying its best to solve the power consumption, heat dissipation and performance problems of multi-chip packaging. Lincoln explained that everyone wants to “squeeze” more functions into the same chip, but putting so many high-performance chips together will generate extremely high heat density. Therefore, thermal analysis is a very critical step.

At present, the number of companies making chips is decreasing, largely due to the high cost of advanced technology ranks, so advanced design tools are particularly important. Lincoln mentioned that Mentor’s advantage is that it has a very comprehensive process. The new solution can provide convenience for IC design manufacturers and meet their needs to a certain extent.

 4. Mentor and China

Now, many manufacturers in the Chinese mainland market have begun to pay attention to advanced packaging, especially packaging and testing companies. In recent years, overseas mergers and acquisitions have allowed Chinese packaging and testing companies to quickly acquire technology and markets, make up for some structural defects, and greatly promote the upward development of China’s packaging and testing industry.

According to data from the China Semiconductor Association, the scale of the packaging and testing market in mainland China has increased from 103.4 billion yuan in 2012 to 219.6 billion yuan in 2018. In the packaging and testing market in 2019, mainland China accounted for 28%, second only to Taiwan, China.

Lincoln pointed out that as early as 1989, Mentor had already entered mainland China. Although the market was mainly based on PCB board-level design at that time, the domestic IC industry in mainland China was still in its infancy, and the overall market size was not as large as it is now, Mentor did not underestimate the Chinese market and its future development potential at all. In addition, Mentor also cooperates with local governments, incubation platforms, universities and research institutes to reduce the cost of innovation and entrepreneurship, and actively supports future star enterprises.

Over the years, Mentor has witnessed the growing strength of China’s domestic IC industry, especially in the field of packaging and testing. Lincoln said that in recent years, China’s local advanced packaging and testing manufacturers have basically formed the industrialization capability of advanced packaging through independent research and development and mergers and acquisitions. However, in terms of the proportion of advanced packaging revenue to total revenue and the development of advanced packaging technologies such as high-density integration However, there is still a certain gap between China’s overall advanced packaging technology level and the international leading level.

On this basis, Mentor is continuously supporting the development of Chinese manufacturers in the field of advanced packaging, helping Chinese manufacturers improve performance and power consumption. At the same time, Lincoln also mentioned that among chips of the same level, heterogeneous integration can help optimize the performance of products using the Mentor HDAP design environment.

 V. Summary

In 2020, the battle surrounding advanced packaging continues to escalate, and advanced chip manufacturers are constantly increasing their efforts to explore a broader space for chip innovation. Although the core details of these technical methods are different, everyone’s strategy is to continuously increase chip density and realize more complex and flexible system-on-chips to meet customers’ increasingly rich application needs.

As the manufacturing process approaches the limit and the cost increases infinitely, Mentor will play an increasingly important and indispensable role with its rich experience in this field.

The Links:   2MBI400TC-060-01 SKIIP 32NAB125T12

Fourth batch of this year! 11 illegal social organization websites were shut down according to law

Recently, the Ministry of Civil Affairs, the Central Cyberspace Administration of China, and the Ministry of Industry and Information Technology have thoroughly implemented the decisions and deployments of the CPC Central Committee. They have been strict, expeditious, and targeted, and have shut down the fourth batch of 11 illegal social organization websites and their new media in 2021. account, and the associated web pages have been cleared.

Since the special campaign to further crack down on illegal social organizations on March 20 this year, the Ministry of Civil Affairs, in conjunction with relevant departments, has taken action against illegal social organizations that have been banned in accordance with the Cybersecurity Law of the People’s Republic of China, the Measures for the Administration of Internet Information Services and other relevant laws and regulations. The website was investigated, and 32 illegal social organization websites that were still operating and their WeChat, Weibo and other new media accounts were shut down in batches, further consolidating the results of offline crackdown and rectification, and consolidating the closed loop of online and offline governance.

The shutdown involves 11 illegal social organizations that have been banned, including the China Association for Continuing Education of Teachers, the Chinese People’s Association of Writers and Artists, the China Wisdom Education Federation, and the China Pet Industry Alliance.

The Ministry of Civil Affairs will continue to maintain a high-pressure attack situation, strengthen network monitoring and investigation, conduct simultaneous online and offline investigations, and resolutely eradicate the breeding ground for illegal social organizations. For the sponsors of illegal social organizations who are of a vile nature and persist after repeated teachings, they will also be submitted to the Ministry of Industry and Information Technology to be included in the blacklist of illegal Internet sites (sponsors) in accordance with the law. All sectors of society are welcome to log on to the Chinese Social Organization Government Affairs Service Platform (www.chinanpo.gov.cn), and provide information on suspected illegal social organization activities, key leaders and key personnel through the “Complaints and Reports” column.

Relevant institutions and the public are reminded again that when cooperating with social organizations or participating in their activities, they must pay attention to verifying their identities. The registration information of national social organizations can be inquired online through channels such as the China Social Organization Government Affairs Service Platform, China Social Organization Dynamic Government Affairs WeChat (WeChat: chinanpogov) and other channels.

List of 11 illegal social organizations whose websites have been shut down in the fourth batch in 2021:

1. China Teachers’ Continuing Education Association

2. Chinese People’s Association of Writers and Artists

3. China Smart Education Federation

4. China Pet Industry Alliance

5. China Packaging Alliance

6. China Art College Entrance Examination Education Alliance

7. National Real Estate CIO Alliance

8. National Campus Literary Alliance

9. China Overseas Degree and Student Information Center

10. National Qualification Evaluation Center for China Cleaning Service Industry

11. Silk Road International Public Welfare Museum

The Links:   NL6448BC33-49 LQ064V3DG04

Maserati’s first hybrid car is about to be released, officially opening the era of electrification

Fiat Chrysler Automobiles said the first charged Maserati Ghibli, which will be a hybrid, will be available in 2020. In addition, the first new Maserati sports car will be launched in 2020, which will be Maserati’s first all-new model since 2015.

Previously, the Maserati brand announced its new development plan in manufacturing, electrification and autonomous driving technology, and will officially open the electrification era in 2020. Among them, in the production plan of the first hybrid and pure electric models, Maserati will produce an electric sports car and a new SUV.

In 2020, Maserati will release its first hybrid car, an electric version of the Ghibli sedan. In the future, all new Maserati vehicles will be equipped with hybrid or pure electric drive systems, and all electric models will fully integrate Maserati’s new-generation electric drive technology, providing unique driving modes, long cruising range, and fast charging functions.

In terms of autonomous driving technology, all new Maserati models and upgraded versions of current models in the future can achieve different levels of autonomous driving, including L2-level Maserati high-speed driving assistance systems and L3-level autonomous driving systems that are close to fully autonomous driving.

The Links:   NL6448BC26-09C VS-26MB120A

Functional safety and information security of automotive AI: new topics require new solutions!

Determining the link between the impact of random hardware failures at the edge and machine learning models is a critical link necessary to build trust in the real world. Without building trust, we cannot use machine learning and AI to implement the safety-related functions required by self-driving cars.

Whether in formal or informal discussions, NXP employees often talk about the interaction of functional safety and information security. When it comes to cars and artificial intelligence, the topic can get more complicated.

In last year’s blog post, Future Challenges: How to Secure AI, we mentioned that NXP is laying a solid foundation for safe AI, such as NXP’s white papers on algorithmic ethics and Auto eIQ.

We are not alone in the quest for functional safety and information safety in automotive AI. In Germany, the German Federal Office for Information Security (BSI) and ZF (ZF), together with the German certification body TUV Nord, are exploring how to test AI in cars. The German Research Center for Artificial Intelligence (DFKI) and TUV SUD are also working on automotive performance testing of AI systems.

While the exact requirements for these tests are unclear, we will build on what we already have.

Learn how NXP can help you comply with standards and achieve system-level functional safety design, click here >>

How the car of the future will use machine learning and artificial intelligence

On the inference side, we believe that the cars of the future will leverage AI and machine learning to provide safety-related functions. That’s why we’re working with ANITI and ONERA to study the impact of random hardware failures in AI-based systems. This work is on top of the state-of-the-art established by functional safety standards such as ISO 26262, which explicitly exclude machine learning.


Functional Safety Block Diagram

A key area of ​​this collaboration is fault injection. Fault injection techniques can be used for the following 3 purposes:

1 Verify the effectiveness of the security mechanism
2 Prove the robustness of a particular design to random hardware failures
3 Verify in the field that the equipment can detect foreseeable failures

Determining the link between the impact of random hardware failures at the edge and machine learning models is a critical link necessary to build trust in the real world. Without building trust, we cannot use machine learning and AI to implement the safety-related functions required by self-driving cars.

How to build trust with functional safety?

As far as building trust is concerned, we know that in cars running AI safety features in the near future, fully following machine learning models will not be enough to keep occupants safe. In this case, the importance of the hardware itself shifts to the quality of the model.

It is common practice to evaluate metrics related to the average performance of a model. The typical method is to use the accuracy rate to define the rate of correct classification. In target detection (an essential function of radar), mAP is usually chosen as the indicator. In addition to the correct classification of objects, mAP also covers the quality of the derived bounding boxes.

However, we believe this is not enough to build trust in the model. These metrics may simply fail to distinguish strange errors from understandable human errors simply because the shape of the detected object is rare.

Also, if a strange error occurs, you need to understand the cause of the error in order to fix it. Even if the prediction is correct, it may be based on incorrect biases that one would like to avoid.

How to understand AI algorithms?

In order to solve these problems, we need appropriate methods to open the “black box” of AI algorithms so that humans can understand these predictions. Grad-CAM is an interesting example of this neural network approach. A 2019 paper by Ramprasaath R. Selvaraju et al. explains how to multiply a feature map by its gradient to determine which parts of the input are most important for the model’s predictions. This provides users with a very valuable tool to better understand what the model has learned and, in turn, make predictions.

This also increases trust in the model and helps developers spot model shortcomings in the training set and then address them.

The authors give an example where two models were trained to identify doctors and nurses. Grad-CAM provides an overlay (similar to a heatmap) that tells people which parts of the picture to use to make judgments. In the first model (“biased”), the model uses facial features to identify. This is something we don’t want to see, given the bias in the training set (more photos of female nurses and more photos of male doctors), the model may judge that the photo is “nurse”. After analysis and retraining, the second model can make decisions based on other elements. Grad-CAM is used to verify that these elements (eg stethoscope) are required.

This example can easily be extrapolated to safety-related functions. Grad-CAM ensures unbiased training of driver fatigue markers in cases detected by driver monitoring systems.

NXP continues to invest in AI solutions for functional safety and information security. By building trust in training data and ensuring flawless execution of ML models, even in the event of random hardware failures, NXP is committed to a bright, secure future.

author of this article

Andres Barrilado, functional safety assessor at the NXP Center. Andres has worked as a security architect for radar front-end devices and as an applications engineer for automotive sensors. He lives in Toulouse and enjoys traveling, running and exploring unknown environments with his wife in his spare time.

Wil Michiels, Security Architect at NXP Semiconductors, focuses on security innovations that enhance security and trustworthiness in machine learning. He enjoys delving into model confidentiality, adversarial examples, privacy, and parsability.

The Links:   CM150E3U-12H 7MBP50RA060 6MBI25J-120

Dalian Dapinjia Group launches AI image recognition and vehicle recognition solution based on NXP i.MX8QM

On May 13, 2021, Dalian General Holdings, a leading semiconductor component distributor dedicated to the Asia-Pacific market, announced that its subsidiary Pinjia has launched an AI image recognition and vehicle recognition solution based on NXP’s (NXP) i.MX8QM.

Today’s society is gradually developing into a multimedia-centric economic system that is highly dependent on data and automation. As an important part of the system, the automobile industry is also undergoing an unprecedented intelligent upgrade with the advancement of many technologies. As autonomous driving and assisted driving technologies become more mature, how to help customers develop AI applications has become a new topic.

The AI ​​image recognition and vehicle recognition solution based on NXP i.MX8QM launched by Dalian Dapinjia adopts the eIQ 2.0 software development environment, integrates a number of different algorithms and provides corresponding APIs for customers to develop and use. The core chip of this solution, i.MX8QM, can stably handle complex and heavy resource consumption such as image recognition, machine learning, and data computing.

Figure 1 – Block diagram of Dalian Dapinjia’s AI image recognition and vehicle recognition solution based on NXP i.MX8QM

The i.MX8QM chip is a GPU equipped with 4-core A53, 2-core A72 and 2 built-in GC7000XSVX. It adopts NXP’s advanced technology, not only has a flexible and fast startup mechanism, but also provides Display failover function. And this chip has passed ISO26262 and ASIL-B certification, so it can better ensure the safety level on the vehicle system.

This solution is developed based on NXP’s native BSP 5.4.24_2.1.0, adding Python elements, and can use GPU/NPU to improve the computing efficiency of AI-type neural networks, making the application scenarios more complete. Not only that, this solution is also equipped with an integrated development environment for users to quickly enter the development field, and also provides open source learning modules such as currently popular OpenCV, TensorFlow, TensorFlowLite, caffe, etc., which can help users save development costs and development costs. time.

Figure 2-Scenario application diagram of Dalian Dapinjia’s AI image recognition and vehicle recognition solution based on NXP i.MX8QM

Core technical advantages:

Ÿ Automotive Gade, ASIL-B;

Ÿ 16x Vec4-Shader GPU, 32 compute units OpenGL® ES 3.2 and Vulkan® support Tessellation and Geometry Shading;

Ÿ 2xArm A72 core + 4 A53 core;

Ÿ MIPI CSI can connect two high-definition cameras at the same time;

Ÿ Pinjia provides cross-platform (PC to I.MX) ML (Machine Learning) applications.

Program Specifications:

Ÿ Python 3.7;

Ÿ TensorFlow 2.1;

Ÿ TensorFlowLite 2.1;

Ÿ OpenCV 4.2.0;

Ÿ ArmNN 19.08.

The Links:   BKO-NC1122-H03 FP75R12KT3

What did SMIC say in response to the U.S. government’s blacklisting of two Chinese companies?

C114 News on December 4th (Yan Yi) SMIC issued an announcement saying that this morning, it was concerned about the news that the company was added to the list of Chinese military-related enterprises on the website of the US Department of Defense. The company is evaluating the impact, please investors Be aware of investment risks. Since then, SMIC has been temporarily suspended in Hong Kong.

It is reported that on Thursday, local time in the United States, the US Department of Defense issued an announcement saying that it will blacklist four Chinese companies including SMIC. In June and August this year, the U.S. Department of Defense successively included 31 companies including Huawei, Hikvision, and China Communications Construction Group Co., Ltd. on the list.

Under the National Defense Authorization Act, the President of the United States has the authority to use the International Emergency Economic Powers Act (IEEPA), including sanctions, against these companies’ operations in the United States.

The Links:   M150XN05 V5 LQ150X1LG91