Widespread application of DSP and FPGA in automotive electronics
1 Introduction
At the end of the 20th century, the wave of information revolution that emerged worldwide provided a golden opportunity for the breakthrough development of the automotive industry. The widespread application of information technology is the best way to solve the problems brought by automobiles, such as traffic congestion, traffic safety, environmental pollution, and energy depletion.
At the same time, with the development of automotive electronic technology, the proportion of electronic components in the cost of the whole vehicle has gradually increased. According to statistics, at present, electronic components have accounted for 20% to 30% of the total cost of automobiles produced in Europe and the United States, and automotive electronic components are also growing rapidly at a rate of 8.8% per year, especially the use of digital signal processor chips (DSP) will grow at a rate of 25% per year
It is estimated that by 2005, the market size of automotive electronic components will reach 17 billion US dollars. It can be seen that electronicization, integration, digitization, informatization, networking, intelligence, miniaturization and personalization have become and will continue to be important trends in the development of the automotive industry. This article only focuses on the digital application technology based on DSP and FPGA in automotive electronics.
DSP and FPGA technologies are widely used in many fields, and they are also widely used in the field of automotive electronics. Due to its strong real-time performance, it is possible to process voice in real time; because it realizes its functions through software programming oriented to chip structure instructions, it is possible to improve the original design or original function of the system by modifying the software without changing the hardware platform, which has great flexibility; and because DSP and FPGA chips are not specially designed for a certain function, they have a wide range of uses, large output, and can be very cheap. Therefore, the large-scale application of DSP and FPGA in automotive electronic systems will greatly promote the development of automotive electronic technology.
2 Comparison of the application of DSP and FPGA in automotive electronics
As a programmable very large scale integrated circuit (VLSI) device, DSP implements extended algorithms and digital signal processing functions through downloadable software or firmware. Its most typical use is to realize FIR filters and FFT algorithms. In hardware, the most basic structural unit of DSP is the multiplier-adder called MAC, which is usually integrated in the data channel, so that the instruction cycle time can be the same as the hardware arithmetic cycle time. In addition, the DSP chip also has several independent on-chip memories, ROM, RAM, parallel function units, phase-locked loops (PLLs), oscillators, several 8-bit or 16-bit buses, clock interrupt circuits, etc. In order to meet the requirements of wireless portable devices to store data without power, DSP chips also use technologies such as flash memory and ferroelectric memory. At present, most DSP chips use an improved Harvard structure, that is, the data bus and the address bus are separated from each other, so that instructions and data can be processed simultaneously, which improves processing efficiency. In addition, pipeline technology is also used to overlap the instruction time of steps such as fetching instructions, fetching operands, and executing instructions, greatly improving the operation speed.
FPGA refers to field programmable gate arrays.
Its basic functional modules are composed of n-input lookup tables, triggers for storing data, and multiplexers. In this way, as long as the data in it is set correctly, the lookup table can realize any Boolean function of the input by reading the data in it. The trigger is used to store data, such as the state information of a finite state machine. The multiplexer can select different combinations of input signals, and by connecting the lookup table and the trigger with programmable wiring resources, different combinational logic and sequential logic can be realized.
Due to the characteristics of the internal structure of FPGA, it can easily realize distributed algorithm structure, which is very beneficial for realizing high-speed digital signal processing in automotive electronics. Since the functional blocks implemented by FPGA devices can work simultaneously, parallel execution at the instruction level, bit level, pipeline level and even task level can be realized, which greatly speeds up the calculation speed.
The computing system implemented by FPGA can reach hundreds or even thousands of times that of existing general-purpose processors. In addition, since FPGA can be dynamically configured, the silicon area of the system is no longer a linear function of the number of wireless interfaces supported, so it is possible to integrate a system that supports all standards in a few or even one FPGA. However, since the existing FPGA development systems are almost all designed for ASIC prototype verification, these development systems are very efficient in saving engineering development time, but the utilization efficiency of FPGA resources is relatively poor. HDL language can greatly improve design capabilities, but the HDL design method has certain limitations in maximizing device performance, and cannot provide optimization and constraints for FPGA layout and routing.
3 Application of DSP and FPGA in Automotive Electronics
When it comes to the digitization of automotive electronics, we cannot help but think of the software radio technology that is currently very promising in wireless communication. Although it is aimed at wireless communication, the idea that software radio wants to achieve is the same as the goal pursued by automotive electronics for digital processing. Therefore, it is necessary to mention and adopt the implementation ideas and thoughts of this technology. The concept of software radio was first clearly proposed in May 1992 by Joe Mitola of MITRE. It is the product of the application of current computing technology, very large-scale integrated circuits and digital signal processing technology in wireless communication; the basic ideas and goals it pursues are: to construct an open, standardized and modular universal hardware platform, and to complete multiple functions such as working frequency band, modulation and demodulation type, data format, encryption mode, communication protocol, etc. with software to achieve highly flexible and open communication products. Therefore, for the research of automotive electronic digital products, the following main ideas of software radio can be fully absorbed: first, to free automotive electronic products from the constraints of hardware structure; second, it does not mean that hardware is not needed; third, automotive electronic products should be open and compatible. Open means open to use, open to production and open to research and development. Next, we will discuss the main applications of DSP and FPGA in automotive electronics based on the idea of software radio.
3.1 Automotive voice signal processing based on DSP and FPGA
Voice processing in automotive electronic products mainly involves digital processing of voice, voice encoding and decoding, voice compression and voice recognition. One of the more popular automotive electronic products abroad is the voice recognition system, which has potential application prospects, including voice-controlled telephones, voice-operated navigation, voice-controlled selection of radio channels, anti-theft voice identification, etc. For example, an application based on a hidden Markov model (HMM) that is independent of the speaker and recognizes 100 instructions. According to the literature, the size of the acoustic HMM model will be. For timely processing including subdivision/windowing of input voice samples, MFCC extraction, probability calculation and Viterbi search, the DSP’s computing capacity is generally required to be 100 million multiplication and addition (MAC) operations. For the recognition of continuous voice signals, better digital signal processing speed and larger storage space are required.
Since the speech recognition system needs to process and sample the sound in real time, it needs a lot of calculations. If 20% of their computing resources are allocated to 10 million MAC speech recognition applications, the processor needs to have 50 million MAC capabilities. Therefore, DSP and FPGA must be used to complete its tasks. The processing speed of DSP and FPGA plays a decisive role in the complexity and performance of speech signal processing application systems. The implementation of high-speed DSP and FPGA can realize modern speech processing and recognition technologies such as channel adaptation and sound domain adaptation. In theory, the faster the DSP and FPGA processing speed, the better the application performance of automotive speech processing and recognition products.
With the increasing diversification of applications, DSP and FPGA have evolved from no longer an independent chip to a component core. This allows designers to select the appropriate core and dedicated logic to “glue” together to form a dedicated DSP and FPGA solution to meet the needs of signal processing. At present, there are also chips that integrate DSP cores and ASIC microcontrollers. Automotive electronic systems use general-purpose DSPs and FPGAs to implement speech synthesis and error correction coding. Speech synthesis, speech compression and coding are the earliest and most widely used applications of DSP. Vector encoders are used to compress speech signals into channels with limited bandwidth.
3.2 Automotive image signal processing based on DSP and FPGA
Digital image processing and analysis technology is a relatively mature two-dimensional signal processing technology. It has been widely used in communications, biomedicine, industrial detection and military, and of course, a large amount of image processing will be involved in automotive electronics. Image processing in automotive electronics mainly includes motion image processing and still image processing. At present, many industries have opened the global positioning system (GPS) in cars. In addition to transmitting its own location coordinate information, the vehicle-mounted GPS system also needs to transmit image information of its environment, such as the scene of the rescued wounded, emergency disaster relief scene images, etc.
At the same time, the traffic monitoring images of each traffic intersection need to be transmitted back to the traffic command center, and image signal processing is also required. For this kind of automotive motion image, the main features are: First, multi-rate compression. Due to the time-varying characteristics of wireless channels, the effective bandwidth, transmission mode and data rate of the system often change constantly; accordingly, it is necessary to adopt a multi-rate compression method to flexibly adapt to such changes in channel bandwidth.
Second, the compression ratio is large.
For example, the data volume of NTSC TV images is about 167Mb/s, and it must be compressed by about 200 to 6000 times to meet the requirements of transmission bandwidth. Third, motion compensation of moving images. Due to its own relative motion, moving images will have Doppler frequency shift problems. For high-speed moving cars, this frequency shift is often not negligible, and the obtained images must be motion compensated.
In recent years, with the rapid development of microelectronics technology and the improvement of chip manufacturing technology, DSPs and FPGAs have continued to emerge. The signal processing system of a chassis or even a cabinet in the past can now be completely completed by a single-chip DSP or FPGA, and the system design will also transition from the past PCB board design to the design of VLSI and UVLSI (very large-scale integrated circuit) chips. At the same time, due to the extensive adoption of DSP and FPGA technology, digital image processing has also undergone major changes in hardware structure. It has developed from a basic serial structure to a parallel processing structure, from a single-chip DSP or FPGA processor to a multi-DSP or FPGA processor system, or a high-speed processing system with array DSP and FPGA. With the development of society and economy, and people’s increasing requirements for the real-time performance of digital image processing systems, the application scope of digital image processing systems based on DSP and FPGA in automotive electronic products will become wider and wider, such as in-vehicle conference television, in-vehicle videophone, in-vehicle machine vision, etc.
3.3 Automotive Adaptive Real-time Processing Based on DSP and FPGA
The clock delay of FPGA can reach nanoseconds, combined with the parallel processing mode of DSP and FPGA, so DSP and FPGA are very suitable for ultra-high-speed and real-time signal processing. As mentioned above, due to the characteristics of the internal structure of FPGA, it can easily realize a distributed algorithm structure, which is very beneficial to the realization of high-speed digital signal processing in automotive electronics. Because automotive electronic products usually require a large number of filtering operations, and these filtering functions often require a large number of multiplication and accumulation operations, and through the distributed arithmetic structure, FPGA can effectively realize multiplication and accumulation operations.
On the other hand, the large amount of complex mathematical operations required can be completed by DSP or ASIC composed of DSP core. In automotive electronic products, special attention is paid to the size, weight and power consumption of the products; in terms of data transmission, the large amount of data generated by the digitization of sound signals in automotive electronic systems must rely on high-performance DSP and FPGA to reduce the requirements of storage space and transmission bandwidth, and it is necessary to perform adaptive real-time processing of tasks such as encoding, decoding, color space conversion, echo cancellation, filtering, error correction, multiplexing, bit stream protocol processing of video signals and audio signals, which is often impossible to complete without DSP and FPGA.
Control theory processing is a difficult and key issue in automotive electronics
. The open-loop, dead-loop, optimal, and adaptive control systems established by classical and modern control theories are used to achieve the optimal control of the car. To establish these control systems, first identify a certain system of the car, such as the ignition advance angle optimization control system, establish the mathematical model of the system, and then use the corresponding control method to optimize the control.
However, the structure of the engine itself is relatively complex, and there are many factors affecting ignition. It is difficult to theoretically derive the mathematical model under the optimized ignition state. Therefore, generally, the best ignition advance angle under various working conditions is found by experimental methods, and then stored in an external memory with increased capacity based on DSP and FPGA or DSP and FPGA array; this can avoid the use of computers.
During the control process, the system detects the working conditions of the engine (such as engine speed, power, etc.) in real time, and uses the table lookup method to find the best ignition advance angle under the working condition, and then controls the ignition after correction. Compared with the traditional computer-based control method, this greatly reduces the volume on the one hand; on the other hand, it is more real-time and flexible.
Suspension electronic control refers to a control system that can adaptively process the vehicle’s roll, front and rear tilt, and automatically adjust the damping force of the shock absorber after the computer detects the signal of the steering and braking conditions. It can prevent tilting and improve the ground adhesion of the wheels. Ultrasonic height sensors are used to control the height of the vehicle body, air springs are used to adjust the elastic system, and grating detectors are used to measure the steering angle, etc. The emergence and development of DSP and FPGA have made the control of each system centralized, forming an intelligent control system for the whole vehicle.
“Intelligent transportation system” is the common pursuit direction of the future automobile and transportation industries. It will include intelligent highways and intelligent car systems. It combines advanced highway information processing technology and radar anti-collision technology, connects highways and cars into a whole, can greatly improve car flow, and significantly reduce the incidence of traffic accidents. Therefore, products related to automobile intelligence have been highly valued by automobile manufacturers. The intelligent transportation system can provide the driver with the shortest distance and the best driving route that can bypass the relatively concentrated vehicle density according to the target information provided by the driver. “Safety first” is always the first choice for users to buy cars. The currently popular automotive millimeter-wave adaptive anti-collision radar is developed to solve the large number of traffic accidents caused by collisions on highways. Because the relative speed between cars on highways is very high, the extraction of radar echo signal frequency difference must be real-time. Therefore, the extraction and processing of radar echo signal frequency difference, as well as the feedback control processing of the adaptive anti-collision control system, are often implemented using DSP or FPGA.
4 Development Outlook
Looking at the major achievements of automobile technology in recent decades, most of them are breakthroughs in the application of electronic technology. Electronic technology has become an important factor in the development of the automobile industry.
The emergence of DSP and FPGA has brought revolutionary changes to automotive products and automotive electronic technology. The use of DSP and FPGA in the world’s automotive industry has surged, from the previous single-chip DSP or FPGA processor to multiple DSP or FPGA processors, or high-speed processors of DSP and FPGA arrays. Automotive electronic products based on DSP and FPGA can meet the needs of future automotive development. In the era of coexistence of multiple models, the general hardware platform built with DSP and FPGA as the core can achieve this compatibility through different software loading methods. With the continuous development of automotive electronic technology in the future, the speed of DSP and FPGA will continue to increase. As far as DSP is concerned, it is currently developing rapidly.
The main trends are: implementing multiple MACs, more registers, wider program buses and data buses, and higher operating frequencies in a single-chip DSP; from a structural perspective, using SIMD and MIMD, using ultra-long instructions, etc. As far as FPGA is concerned, due to the adoption of sub-micron technology, it is faster and has more gates. At present, Lucent and XILINX both have products with more than 100,000 gates, and integrate some new functions, such as System on Chip, Programming on System, etc., making it more flexible.
my country’s research on automotive electronic systems is not deep enough. Anti-lock braking systems, airbags, automatic transmissions and diesel engine electronic control systems have only been explored in some universities and enterprises and have not entered the practical stage. High-tech represented by automotive electronic technology is the “bottleneck” of the development of my country’s automotive industry. In view of this situation, my country’s research on automotive electronic technology should not only focus on energy saving, environmental protection and safety of automobiles, strive to master their core technologies as soon as possible and narrow the gap with developed countries, but also take emerging technologies such as vehicle-mounted communications and high-speed real-time signal processing technology as breakthroughs, rely on the results of national information technology research, develop advanced vehicle-mounted computing and information processing products, drive the progress of the entire automotive electronic technology, and improve the level of electrification of my country’s automobiles.
