Embedded applications with a higher number of logic gates are the main battlefields for FPGA positioning

According to Tony Massimini, senior analyst at Semico Reserch, sales of micro-logic components (microprocessors, microcontrollers, and DSPs) soared by 24.9% in 2010. However, this is because of the catastrophic recession in 2009. In a relatively stable sales environment, it is expected to grow by another 12% in 2011.

In terms of shipments, the micro-logic components will increase by 36.3% in 2010, which was a decrease of 10.4% in 2009. Micrologic turnover is dominated by the microprocessor and computing market. In terms of quantity, it mainly comes from the MCU. In 2010, MCU revenue increased by about 32%, and the number increased by about 40%. The growth covers all microcontroller parts - 8, 16 and 32 bits. This is faster than predicted by a year ago. This is due to the recovery and growth in some areas, such as new growth in the automotive industry, and the continuous growth of industrial control. Although industrial control has also declined in 2009, it is not like other markets. So bad.

Industrial MCU vendors value connectivity and peripheral development. Connectivity includes USB, RFID, low-power RF, ZigBee for control and automation. Green and energy saving are the biggest growth markets. Smart meters and smart grids are receiving more attention. More efficient motor control and factory control of home appliances require special low-power processes.

16-bit vendors offer higher processing speeds than 8-bit ones. Many 16-bit MCU vendors also sell 8-32-bit MCUs. In 2010, 8-bit MCUs cost about 85 cents. The average price of 32-bit MCUs is 3 to 23 US dollars. The reason why this price varies greatly is due to differences in packaging, performance, and memory configurations. MCU manufacturers are providing higher memory density and better peripherals to reduce external materials, sensors, and so on.

According to iSuppli's statistics, Renesas Electronics, which was merged by former Renesas Technology/NEC Electronics, is the largest embedded system company. TI is the second largest embedded system company. TI mainly benefits from DSP, MCU and MPU.

In the Chinese market, MCUs keep growing above the global average speed. In particular, local involvement in design is leading a new round of recovery and growth worldwide, while local Chinese designs are growing faster than the entire MCU system. However, in Figure 2 why does the "Chinese local participation in design ratio" curve show a downward trend between 2010 and 2012? Qiu Rongfeng, director of Renesas Electronics Greater China Region’s MCU Product Center, believes that China has occupied a large proportion of the entire world economy. At the beginning, China was basically the world’s largest factory. Exports are basically the biggest factor that makes the Chinese market well-done. However, from 2008 to 2009, China basically turned to domestic demand. High-speed development is at home, so we can see that in 2010 with the highest growth rate, China has a series of activities, such as home appliances to rural areas and smart grids. The rise of electricity meters, etc.; but normally, in addition to the internal digestion of the Chinese market, it also hopes that foreign exports will slowly increase; but in accordance with this idea, the two years is unlikely to enter the United States, Europe and Japan and other markets.

In the embedded software industry, embedded products have been continuously infiltrating into various industries in recent years, ranging from mobile phones, iPads, to base stations and space satellites. The embedded software plays an increasingly important role in it, and it has become the most important weight of different manufacturers from other manufacturers. According to incomplete estimates, the size of China's embedded software market in 2011 will reach around 460 billion yuan. Multi-core and platformization will become new trends in the development of new embedded software.

Software and System Shelves Constitute the Investment Focus of Chip Vendors In recent years, MCU vendors have provided more and richer tools and software development environments that are easy to use and save time. In the future, system designers will have more choices of applications: features, prices, low power consumption and more development tools.

Although the MCU integrates a richer periphery, process technology continues to evolve. In contrast, the cost of hardware is less likely to rise. Geo Lees, vice president of NXP and general manager of the global microcontroller product line, said that the company does not intend to generate excessive costs in terms of hardware, and its development focus will be more to software.

Zeng Jintao, Asia Pacific market manager for the Freescale Industrial and Multimarket Microcontroller Division, also believes that software is taking up a larger proportion of the total development cost. Zeng Jintao took the company's strategy as an example, saying that Freescale was the first to provide customers with a complete set of software support tools. Its software products include the free 8-bit Codewarrier tool and the real-time operating system MQX for 32-bit MCUs, along with many other software that enables customers to easily begin designing with Freescale MCUs. ADI DSP Asia Business Manager Lu Lei said that the company's software strategy is to provide customers with free optimized and tested high-performance software libraries, which can simplify customer development and shorten customer product development time. At the same time, Analog Devices provides open source Linux systems and drivers for its customers.

Mike Salas, director of MCU marketing at Silicon Labs, analyzed why software has become the focus of MCU vendors. He pointed out that with the continuous development of microcontrollers (MCUs) on the basis of their digital cores, by integrating more complex peripherals such as encryption/decryption modules, audio/video encoders and decoders, and the need for a complete protocol stack The wireless communication interface) MCU function is gradually expanding. The interaction of all these on-chip system functions also requires the same complicated firmware. The time and cost of developing and testing firmware is becoming a limiting factor affecting the launch of new products. To help reduce time-to-market, Silicon Labs offers a variety of firmware for its mixed-signal MCUs. This includes everything from simple peripherals such as ADCs to full application-level reference designs such as the Wireless M-Bus kit. Because many target applications of Silicon Labs MCU products need to communicate with a computer or application processor, the company provides communication function libraries and drivers as needed. All of this code allows embedded designers to focus on product development rather than sophisticated MCU or software protocols.

In order to increase software strength, a group of application software talents are needed in the MCU design team. Zhao Yijun, general manager of Shanghai Pursunda Electronics Co., Ltd., said that Pursunda not only has a long-term team of engineers dedicated to application development, but also has an experienced chip design team. Such a combination makes Purdue Danone at the very beginning of chip design. Taking full account of the needs of application software design, the hardware functionality, performance, and software design requirements can be tightly integrated throughout the chip design process. In the application development process, engineers can also work closely with the design team to ensure that the software is effective and reliable.

System architects how to position precision In order to do a good job, some chip manufacturers' chief designers are not even chip design origins, but system application origins. Ramesh Kumar, global business manager of TI's multi-core DSP business unit, introduced the company's next-generation high-performance DSP, the TMS320C66x, and revealed how the company can do high-precision positioning of its products. For example, high-precision mission-critical applications - wireless base stations, medical and so on. In the research and development process of TI DSP, a certain proportion is put into software. This consideration comes from: Because you want to build a very good multi-core system or system-on-chip (SoC), you need very good system-level considerations. In TI's DSP team, there are quite a few people from system vendors such as Ericsson, Nosi, and even hired senior talents who have worked at communication companies such as Motorola to do TI's CTO (Chief Technology Officer) or chief architect. It is not because of how strong their chip design capabilities are, but rather that the key elements needed for the entire system architecture are very, very accurate. In this way, TI can fix the network processor and the transmission coprocessor in it. This is due to the system experts who come from the communications system companies, and provide the key elements of their challenges and needs, as well as features that have not been provided in the past, in TI's organizational structure. Although these communication system talents may not have developed a complete software package, TI will be able to develop more corresponding core software and build a multi-core system and multi-core DSP system into a highly competitive system product. .

In terms of WiMAX applications, TI has even hampered the entire WiMAX R&D team of a system vendor in North America. It is not that TI wants to develop any WiMAX products on its own. The key is to understand and understand the system and integrate traditional understanding and new requirements. To our multi-core DSP development.

The same is true in the medical field. TI has talented individuals who have worked with ultrasound or digital X-ray vendors as TI's system architecture experts because TI needs to be very focused on understanding what these application architectures do, so that content can be developed accordingly. The core IP library can provide very useful parts for customer development or our subsequent optimization.

It can be seen that MCU and DSP vendors are not only concerned with the chip architecture, but also with software and architecture systems to improve the ease of use of the product.

MCU, DSP, and FPGA infiltrate each other, and increasingly integrate MCUs with 8, 16, and 32-bit divisions. There are also differences between ARM and non-ARM. Currently, some vendors use the ARM Cortex-M4 core, hoping to enter the DSP field; DSP boss - TI Putting its own C2000 DSP series into the MCU business group; at the same time, FPGA manufacturers are also trying to penetrate the traditional DSP and embedded processor areas; some MCU vendors claim that their chips can replace FPGAs and DSPs. ? The application areas of the embedded processor are mutually infiltrated, and the chip functions are more and more integrated, so the boundaries become more and more blurred. What are their distinctive features?

TI's semiconductor business unit MCU business group is now not only responsible for its classic 16-bit MCU - MSP430, there are two years ago to acquire Luminary's ARM Cortex-M3-based 32-bit processor, but also its low-end DSP - C2000 DSP series Also allocated in [4]. So how do you coordinate the product line relationship when recommending a plan to your customers? According to Wu Jianhong, TI's MCU business development manager, “When introducing products to customers, we will definitely recommend the most suitable products.” Some software libraries such as Cortex-M3 and C2000 share. In this way, if some of the customer's products use the C2000, other products may use the Cortex-M3 MCU, which can use TI's software library. The author speculates that since TI has a powerful DSP function, TI did not use the ARM Cortex-M4 with DSP capabilities to authorize it, and focused on developing the ARM Cortex-M3 to enter the traditional 32-bit processing market; while the MSP430 series focused on the traditional 8-bit 8051 market.

NXP, Freescale, etc. use the ARM Cortex-M4 authorization. NXP also uses the ARM Cortex-M3 license. How to distinguish between M3 and M4? NXP's Geoff took the LPC4000 series launched at the end of last year as an example. The feature of the Cortex-M4 is that the DSP processing power is very powerful because the M4 has many free DSP library resources for users to use.

In addition, Geo also introduced the advantages of the M4 compared to the ARM9 core: In discussions with TV and set-top box customers, the company found that for high-performance cores, they are mainly implemented in hardware, which causes power consumption problems. If the MCU is always running at high speed, there is no problem; however, at low speed, the ARM9's low power consumption is not as good as that of the M4. Therefore, in some high-performance applications, M3 and M4 can be used on-chip to achieve high-performance MCU functions.

In June last year, Freescale announced cooperation with ARM to launch the Kinetis series based on the ARM Cortex-M4. Although Freescale has its own unique and rich architecture, such as the S08, ColdFire/ColdFire+, and Power architecture, Freescale believes that David Greenfield, senior director of product marketing for the largest special Alberta high-end FPGA product division of ARM MCUs, is getting started quickly. Kinetis extends to those who like to use the ARM architecture [5]. M4 not only is compatible with M3, but also has the function of DSP, which lays a fuss for future DSP applications.

MCUs and FPGAs

MCU companies such as STM (Semitel) [7] and NXP recently mentioned that some of their MCUs based on the M3/M4 architecture can replace FPGA functions; FPGA companies such as Altera, Xilinx and Microsemi (acquired by Microsemi in October 2010. Actel also sought to extend into the embedded space. What is the relationship between the two? NXP's Geoff said that the MCU's goal is not to completely replace the FPGA. For example, when a relatively simple application is under 3,000 gates, the MCU can implement applications similar to FPGA functions.

The FPGA company also admitted that embedded applications with higher logic gate counts are the main battlefields for FPGA positioning.