Why FPGA is indispensable for breaking through Industry 4.0?

Image source: Photo Network

In the rapid development of the information age, the accelerated iteration speed of consumer products has led to a continuous increase in the cost of labor, materials, etc., and has also brought continuous pressure on factory production. This change in the direction of the manufacturing industry and the increase in demand for personalized products have put higher requirements on related companies. Based on this, the fourth industrial revolution with the Internet of Things and intelligent manufacturing as its core is quietly approaching.

In 2011, Germany first proposed the concept of "Industry 4.0", and the United States has also introduced similar concepts such as the Industrial Internet of Things. However, whether it is Industry 4.0 or the Industrial Internet of Things, the main overlapping parts are intelligence and the Internet of Things, which also shows that the global industry is accelerating its transformation.

Industry 4.0 is the future blueprint for the manufacturing industry developed by Germany and the industrial community, and is also a guiding light for industrial transformation and upgrading. Introducing the Internet into workshops and factories, allowing data to flow between various machines and equipment through the network, and maximizing the value of data is a great industrial revolution.

Compared with traditional industries, or compared with Industry 1.0, 2.0, and 3.0, the core of Industry 4.0 is the integration of physical information systems, with emerging technologies such as artificial intelligence, the Internet of Things, and cloud computing as the underlying pillars. By connecting production machines, equipment, and management systems, it indirectly integrates enterprises and value chain levels, realizes automated industrial management, and enters the era of intelligent manufacturing.

Industrial transformation process
(Photo source: Integrated stove network)

Of course, as the underlying technology continues to iterate and update, the market value of Industry 4.0 is also constantly exploding.

The latest market research from Markets and Markets shows that the global Industry 4.0 market size is expected to be US$71.7 billion in 2019, and by 2024, its market size will reach US$156.6 billion, with a compound growth rate of 16.9% during the period .

Driven by the strong market, current industrial applications mainly show the following trends: first, industrial networks are shifting to Ethernet; second, industrial system integration is improving, resource utilization is improving; third, industrial control has higher requirements for security. In addition, there are requirements such as latency, power consumption, and flexibility, which bring new challenges to traditional CPUs, ASICs and other processors.

In most application scenarios of Industry 4.0, new requirements are put forward for hardware acceleration of edge intelligence, requiring lower latency data processing capabilities. On the other hand, new equipment needs to have the ability to be reprogrammed to cope with the evolution of standards or protocols. In addition, the establishment of smart factories is inseparable from the application of big data processing and AI technology. These requirements force the SOC used in smart factories to introduce new architectures.

FPGA, which has excellent data processing and acceleration functions, has become a new choice for smart factories . FPGA is a digital chip, which is also called programmable logic device. Its internal structure is mainly composed of IOB (input and output module), CLB (configurable logic module), clock management module, embedded block ARM, rich wiring resources and other parts.

Diagram of the internal structure of an FPGA device
(Image source: Tencent Architect)

It can be seen that the internal wiring of FPGA can connect the smallest logic unit into a larger logic function unit, and the IOB connection completes the input and output of the signal. It is fundamentally different from the working method of CPU, GPU, DSP and other processors. It does not need to work through tedious steps such as obtaining instructions and decoding instructions. Instead, it uses the truth table as a device to draw a circuit diagram inside the chip, so its essence belongs to hardware development.

In addition, the biggest feature of FPGA is its reprogrammability and strong parallel computing capabilities . It is more flexible than dedicated chips ASIC, and its performance and real-time processing capabilities are better than CPU. However, it is difficult to compile and has a high threshold for application development, making it not suitable for large-scale applications (detailed comparison is shown in the figure below).

Comparison of four mainstream chips
(by Xinshiye)

Since FPGA can be easily reprogrammed according to new versions and can be reprogrammed on site, it is also very beneficial for industrial products with longer cycles. An industry insider pointed out: " FPGA technology does not require a minimum order quantity and has a longer lifespan , which is an important reason why it is unique. Many industrial customers who have used ASIC product designs for five years are now using FPGAs to replace ASICs ."

In addition, according to an engineer from the downstream application end, FPGA plays an irreplaceable role in the field of industrial control, especially when it comes to some motion control algorithms and image recognition algorithms.

In the industrial field that is accelerating transformation, FPGA is likely to replace ASIC and provide more flexible solutions, especially in application scenarios such as industrial communications, motor control, machine vision, edge computing, and industrial robots.

Traditional industrial communication mainly considers the data transmission and exchange of the device connection layer within a small range. The transmission protocols between industrial Ethernets are different and cannot be fully compatible. The Industrial Internet of Things is different from external communication. It is a very closed network with very strict requirements on the real-time, deterministic, security and reliability of data. In the era of Industry 4.0, all transmission protocols are required to be the same.

SOPC Industrial Ethernet Controller Hardware Structure
(Drawn by Xinshiye, Information Source: Weeqoo)

In the process of industrial transformation, FPGA may become an iterative means for the rapid development of the industry. In this process, different communication protocol standards will coexist for a long time, and data conversion between devices requires a very flexible conversion medium. FPGA is expected to become the first choice.

At present, the development of motor control is becoming more and more diversified and complex. The inherent programmability and parallel processing capabilities of FPGA chips are very suitable for motor control of high-end products, especially in the industrial field . FPGA performs parallel processing in hardware and does not occupy CPU resources, which can maximize system performance. At present, many FPGA manufacturers will integrate some optimization algorithms into FPGA, which can realize multiple motor control through one platform based on FPGA devices, greatly improving energy efficiency.

In terms of machine vision, the current mainstream industrial camera solution is still a general software solution based on X86, that is, the front-end camera collects images, and the image processing part is completed in the PC or PLC through software algorithms. However, due to the limited computing performance of the CPU, for visual algorithms with high computational complexity (such as target recognition and tracking based on SIFT feature points), its processing speed is difficult to meet the real-time requirements of the system. The parallel computing power of the GPU is very high, which can greatly improve the computing speed, but its high power consumption and large size are also its obvious disadvantages. On the whole, with the increasing integration of FPGA, the actual scale can be larger and the power consumption is lower. FPGA-based embedded vision systems will become an important direction for the development of machine vision.

Compared with cloud computing with centralized storage and processing, edge computing can provide more obvious efficiency and flexibility for industrial control. FPGA is more suitable for edge computing scenarios than other heterogeneous processors. Compared with GPU, FPGA can greatly optimize bandwidth and does not occupy CPU resources. FPGA also has natural advantages in the low latency and stability required by the industrial field. On the one hand, FPGA integrates a large number of caches and external DRAM interfaces, which reduces the interaction with the CPU during the calculation process. On the other hand, FPGA can implement MIMD (parallel instructions and parallel data) design to realize the transmission and interaction of intermediate data between pipelines, reduce dependence on cache, and reduce data transmission delay.

In terms of industrial robot applications, the rich wiring resources, embedded dedicated IP, basic programmable logic units, etc. inside FPGA enable it to handle different types of tasks at the same time, achieving greater flexibility while meeting the timing requirements of some high-speed interfaces. These are all necessary conditions for industrial robots to realize control functions.

In addition to the above five application scenarios, FPGA is also used in industrial fields such as industrial cloud and display control . It can be seen that the flexibility and excellent parallel computing performance provided by FPGA can help the industrial field to greatly reduce costs and power consumption while achieving intelligent and high-performance capabilities. These are all good opportunities for the future development of FPGA in the industrial field.

Industrial transformation and upgrading has opened up new virgin territory for FPGA applications. According to MRFR statistics, the global FPGA market size was about US$6.9 billion in 2019, of which the industrial sector accounted for about 12%, with a market size of about US$830 million. It is expected that by 2025, the global FPGA application share in the industrial market will increase to 19%, and the market size will exceed US$2 billion.

According to the latest research data from Guanyan Tianxia, ​​the size of China's FPGA market in 2019 was approximately 18.75 billion yuan, of which industrial market applications accounted for approximately 28% , with a market size of approximately 5.25 billion yuan. It is estimated that by 2025, the market size of FPGA in China's industrial field will reach 10 billion yuan.

Map made by Xinshiye
(data source: Guanyan Tianxia)

From the perspective of industry development trends, the demand for FPGAs in the Chinese industrial market is slightly higher than the global demand, which has also brought new opportunities for the development of local FPGAs. However, in the Chinese FPGA market, Xilinx and Altera have long been a duopoly, with market shares of 52% and 28% respectively. From technology to intellectual property rights, domestic FPGA manufacturers face considerable challenges.

A professional with the online name "Qingshuang Yimeng" told Xinshiye: "Domestic FPGAs are still in their infancy. When companies judge the performance of FPGA devices, the focus is not on how many LUTs the FPGA device has. This kind of hardware stacking template design is not difficult. The matching EDA software and corresponding IP are the important parameters for selecting FPGA performance . For example, how much utilization the LUT can achieve, whether the clock network can be well adapted, whether the IP corresponding to the device is stable, how is the compatibility, etc., these aspects are the key points that domestic FPGAs need to improve."

Faced with so many challenges, how can domestic FPGA manufacturers break through and seize the dividends of FPGA in the process of industrial transformation and upgrading?

Lu Zhe, marketing director of Unigroup Tongchuang, a leading domestic FPGA manufacturer, said : "The industrial market is the second largest application market for FPGAs after the communication market. It has a wide coverage, a long life cycle, and high requirements for real-time processing and flexibility. Traditional industrial market applications are mainly concentrated in the fields of IO unit control, industrial communication, motor drive, servo motor control, human-machine interface, security monitoring, etc. Driven by the wave of Industry 4.0 and artificial intelligence, emerging industrial needs are emerging, such as new industrial communication protocols, sensor fusion, machine vision, edge computing, industrial cloud computing acceleration, etc., which are all new application directions for FPGAs. Unigroup Tongchuang has launched small CPLDs and small and medium-sized FPGA products for the industrial field . It has mature solutions in industrial-grade IO expansion, servo motor drive and control, EtherCAT communication, image and video processing, security monitoring, high-definition fine-pitch LED control cards, and high-definition TV display control. It has been successfully introduced and mass-produced and shipped by many well-known domestic equipment manufacturers. In view of the emerging market directions in the future, we are also actively planning and deploying higher-end FPGA products and system platform solutions to meet the diversified needs of the industrial market."

Image source: Photo Network

Of course, domestic FPGAs are not without opportunities, and their advantages are also very obvious. According to an application engineer, "The FPGA development process may encounter many difficulties, such as code migration, resource optimization, timing optimization, etc., so when they choose domestic FPGAs, they first pay attention to the strength of domestic manufacturers, including product performance, quality, software capabilities, and long-term continuous and stable supply capabilities. Secondly, they pay attention to technical support capabilities and solutions, as well as whether the local team can respond to customer needs in a timely manner and assist customers in completing projects. "

I still remember that when I attended the Munich Electronics Show last year, the head of a downstream enterprise said: "We still hope that domestic FPGAs can achieve better results and establish an independent, complete and controllable industrial chain." They will also try to use and support new products launched by domestic FPGA manufacturers.

In addition, according to Lu Zhe, marketing director of Unigroup Tongchuang, in addition to the successful introduction and mass production and delivery of Unigroup Tongchuang's products by system manufacturers in the fields of communications, industry and consumption, Unigroup Tongchuang has also begun to gradually establish and drive the application and development ecology of downstream products and solutions , and establish cooperative relationships with professional partners in many industries, including the launch of ARM Cortex-M1 soft core solutions, EtherCAT master station solutions, professional low-cost evaluation and learning development board kits, and IP solutions for various vertical fields. At the same time, the company is also actively promoting strategic cooperation with professional solution providers in emerging markets such as artificial intelligence and data center accelerators, promoting university plans to cultivate the domestic FPGA development ecology, expanding the industry-university-research cooperation model with universities, and working with customers and partners to promote innovation based on domestic FPGA solutions, and gradually gaining recognition from the domestic FPGA application ecology. At the same time, the company also looks forward to establishing a win-win relationship with more partners.

At present, domestic FPGAs are making breakthroughs in core application areas (such as communications, industry, etc.), and simply pursuing the scale of LUTs is no longer the focus of FPGA development in the future. " Future FPGAs will integrate more and more heterogeneous units, which will have high requirements for achieving powerful computing performance, ultra-high storage bandwidth and I/O processing capabilities on advanced nodes, and will also bring challenges of higher costs and power consumption. " said an industry insider.

Seeking opportunities amid challenges and breaking through blockades, domestic FPGAs must build a complete ecosystem if they want to gain a foothold, including EDA software, rich IP libraries, materials, equipment, etc. Only by building a complete ecosystem can we flexibly respond to different application scenarios and market environments.

The accelerated industrial transformation has given domestic FPGAs unlimited possibilities, but it has also brought more challenges. Next, we expect more excellent FPGA industry chain companies to emerge in China, breaking the monopoly of foreign manufacturers in high-end applications from the entire industry chain including design, manufacturing, packaging and testing.