FPGA's main application areas and industry chain composition

AI chips are mainly divided into CPU, GPU, FPGA, and ASIC. Among them, in the order of CPU, GPU, FPGA, and ASIC, the versatility is gradually reduced, but the computing efficiency is gradually improved.

FPGA is a product of further development on the basis of programmable devices such as PAL, GAL, CPLD, etc. It appears as a semi-custom circuit in the field of integrated circuits, which not only solves the deficiencies of customized circuits but also serves the original purpose. There is a disadvantage in that the number of programmable device gates is limited.

With its high programmable flexibility, short development cycle, and high parallel computing efficiency, FPGAs have been widely used in 5G and AI. Because many application scenarios of communication services need to be upgraded at any time, compared with FPGAs, ASICs are not flexible enough to keep up with the iterative updates of algorithms, so FPGAs are a better choice.

In the artificial intelligence market, the current demand for "training" is the most extensive, but since 2019, the demand for "inference" (including data centers and edge ends) will continue to explode rapidly. CPU-based traditional computing architecture cannot fully meet the needs of artificial intelligence high-performance parallel computing. FPGAs are low-power heterogeneous chips with short development cycles and flexible programming. Solutions in the field of artificial intelligence are currently evolving from software to software + chips. Fortunately, FPGAs are low-power heterogeneous chips with fast development cycles and flexible programming. Solutions in the field of artificial intelligence are currently evolving from software to software + chips.

FPGA application areas

1. Data acquisition and interface logic field

1. Application of FPGA in the field of data acquisition

Since most of the natural signals are analog signals, the general signal processing system must include the data acquisition function. The usual implementation method is to use an A/D converter to convert the analog signal into a digital signal, and then send it to the processor, such as using a single-chip microcomputer (MCU) or a digital signal processor (DSP) for calculation and processing.

For low-speed A/D and D/A converters, a standard SPI interface can be used to communicate with MCU or DSP. However, high-speed A/D and D/A conversion chips, such as video decoders or encoders, cannot directly interface with general-purpose MCUs or DSPs. In this case, FPGA can complete the glue logic function of data acquisition.

2. The Application of FPGA in the Field of Logic Interface

In actual product design, data communication with a PC is required in many cases. For example, sending the collected data to a PC for processing, or sending the processed results to a PC for display. The communication interface between the PC and the external system is relatively rich, such as ISA, PCI, PCI Express, PS/2, USB, etc.

 

Traditional design often requires dedicated interface chips, such as PCI interface chips. If you need more interfaces, you need more peripheral chips, and the volume and power consumption are relatively large. After adopting the FPGA solution, the interface logic can be implemented inside the FPGA, which greatly simplifies the design of the peripheral circuit.

 

In the design of modern electronic products, memory has been widely used, such as SDRAM, SRAM, Flash, etc. These memories have their own characteristics and use, and a reasonable choice of the memory type can achieve the best cost performance of the product. Because the function of FPGA can be designed completely by oneself, so can realize the controller of various storage interfaces.

3. Application of FPGA in the field of level interface

In addition to TTL and COMS interface levels, new level standards such as LVDS, HSTL, GTL/GTL+, and SSTL are gradually being adopted by many electronic products. For example, LCD screen drive interfaces are generally LVDS interfaces, digital I/Os are generally LVTTL levels, and DDR SDRAM levels are generally HSTL.

In such a mixed-level environment, if the traditional level conversion device is used to implement the interface, the circuit complexity will increase. Utilizing the feature of FPGA to support multi-level coexistence can greatly simplify design schemes and reduce design risks.

2. High-performance digital signal processing field

The fields of wireless communication, software radio, high-definition video editing, and processing put forward extremely high requirements on the amount of calculation required for signal processing. The traditional solution is generally to use multiple DSPs in parallel to form a multi-processor system to meet the demand.

However, the main problem brought by the multi-processor system is that the design complexity and system power consumption are greatly increased, and the system stability is affected. FPGA supports parallel computing, and density and performance are constantly improving, and it can replace traditional multi-DSP solutions in many fields.

For example, the realization of high-definition video coding algorithm H. 264. The DSP chip using TI's 1GHz main frequency requires 4 chips, while the StratixII EP2S130 chip using Altera only needs one to complete the same task. The FPGA implementation process is similar to the front-end design of the ASIC chip, which is conducive to the introduction of the back-end design of the chip.

3. Other application areas

In addition to some of the above application areas, FPGAs also have a wide range of applications in other areas.

   (1) The field of automotive electronics, such as gateway controllers/car PCs, telematics systems.

   (2) The military field, such as secure communications, radar and sonar, and electronic warfare.

   (3) Test and measurement fields, such as communication test and monitoring, semiconductor automatic test equipment, and general-purpose instruments.

   (4) Consumer products, such as monitors, projectors, digital TVs and set-top boxes, and home networks.

   (5) Medical field, such as software radio, electrotherapy, life science.

Comparison of FPGA industry development at home and abroad

At present, the leading foreign technology has reached the 7nm and 10nm levels, and the scale of 400-500 million devices can be realized. 5G wireless, data centers, automobiles, wireless communications, AI intelligence, industry, consumer electronics, medical science, etc., are becoming the main driving forces for the growth of the global FPGA market.

Since the end of the 1990s, domestic FPGAs have experienced an era from reverse design to forward design. Among the domestic FPGA products currently active in the market, mostly low- and medium-density products. For most of the domestic high-density FPGAs, their architecture is inseparable from the concept of LUT + wiring. In terms of specific products, the respective technologies, IPs, and even the corresponding application markets are also targeted. From this perspective, the technical level of domestic manufacturers in medium and high-density FPGAs cannot be compared with that of foreign countries, and both at the software/hardware level. There is still a certain gap.

The composition of the domestic FPGA industry chain

Like the chip semiconductor industry, the FPGA industry chain also follows the same processes and links.

In the equipment/materials link, related companies include North Huachuang, China Micro, Huafeng Measurement and Control, etc.;

In the EDA link, related companies include Huada Jiutian, Xinwei, Guangliwei, Ocas, etc.;

In the IP link, related companies include VeriSilicon, InsMicroelectronics, Cambrian, etc.;

In the IC design link, related companies include Unigroup, Gowin Semiconductor, Jingwei Qili, Lianjie Technology, Anlu Information, Fudan Micro, Zhiduojing, etc.;

In the IC manufacturing process, the main company is SMIC;

In the IC packaging and testing process, related companies include Changjiang Electronics Technology, Huatian Technology, and Tongfu Microelectronics.