Features
According to the requirements of digital signal processing, DSP chips generally have the following main features:
(1) One multiplication can be completed in one instruction cycle And one addition.
(2) The program and data space are separated, and instructions and data can be accessed at the same time.
(3) There is fast RAM on-chip, which can usually be accessed in two blocks at the same time through an independent data bus.
(4) Hardware support with low overhead or no overhead loop and jump.
(5) Fast interrupt handling and hardware I/O support.
(6) There are multiple hardware address generators that operate in a single cycle.
(7) Multiple operations can be performed in parallel.
(8) Support pipeline operation, so that operations such as fetching, decoding, and execution can be performed overlapped.
Compared with general-purpose microprocessors, other general-purpose functions of DSP chips are relatively weak.
Classification
DSP chips can be classified in the following three ways.
1. According to basic characteristics
This is classified according to the working clock and instruction type of the DSP chip. If at any clock frequency within a certain clock frequency range, the DSP chip can work normally, except for the change in the calculation speed, there is no performance degradation. This type of DSP chip is generally called a static DSP chip. For example, DSP chip of Japan OKI Electric Company, TMS320C2XX series chip of TI Company belong to this kind of category.
If there are two or more DSP chips, and their instruction sets and corresponding machine code machine pin structures are compatible with each other, this type of DSP chip is called a consistent DSP chip. For example, the TMS320C54X of TI of the United States falls into this category.
2. According to data format
This is classified according to the data format of the DSP chip. DSP chips whose data work in a fixed-point format are called fixed-point DSP chips, such as TI’s TMS320C1X/C2X, TMS320C2XX/C5X, TMS320C54X/C62XX series, AD’s ADSP21XX series, AT&T’s DSP16/16A, and Motolora’s MC56000 Wait. Floating-point DSP chips that work in floating-point format are called floating-point DSP chips, such as TMS320C3X/C4X/C8X from TI, ADSP21XXX series from AD, DSP32/32C from AT&T, MC96002 from Motolora, etc.
The floating-point formats used by different floating-point DSP chips are not exactly the same. Some DSP chips use custom floating-point formats, such as TMS320C3X, while some DSP chips use IEEE standard floating-point formats. , Such as Motorola's MC96002, FUJITSU's MB86232 and ZORAN's ZR35325, etc.
3. Divided by purpose
According to the purpose of DSP, it can be divided into general-purpose DSP chips and special-purpose DSP chips. General-purpose DSP chips are suitable for ordinary DSP applications. For example, a series of DSP chips of TI Company are general-purpose DSP chips. The dedicated DSP chip is designed for specific DSP operations, and is more suitable for special operations, such as digital filtering, convolution and FFT. For example, Motorola's DSP56200, Zoran's ZR34881, Inmos's IMSA100, etc. belong to the dedicated DSP chip .
Advantages and disadvantages
Advantages
Large-scale integration
Good stability and high precision
Programmable
High-speed performance
Embeddability
Convenient interface and integration
Disadvantages
High cost
High-frequency interference from high-frequency clocks
High power consumption, etc.
Applications
DSP chips are widely used in digital control and motion control The main applications include disk drive control, engine control, laser printer control, inkjet printer control, motor control, power system control, robot control, high-precision servo system control, CNC machine tools, etc.
The main applications for low-power, handheld devices and wireless terminals are: mobile phones, PDAs, GPS, digital radios, etc.
Digital signal processing digital filter
There are many practical types of digital filters, which can be roughly divided into finite impulse response type and infinite impulse response type. Hardware and software are available. Two ways to achieve. In hardware implementation, it is composed of adders, multipliers and other units, which is completely different from analog filters composed of resistors, inductors, and capacitors. The digital signal processing system is easy to be made of digital integrated circuits, showing the advantages of small size, high stability, and programmable control. Digital filters can also be implemented in software. The software realization method is to use a general-purpose digital computer to compile a program according to the design algorithm of the filter to perform digital filter calculation.
Digital signal processing Fourier transform
In 1965, JW Cooley and TW Tukey first proposed a fast algorithm for discrete Fourier transform, referred to as fast Fourier transform, expressed by FFT . Since the fast algorithm, the number of operations of the discrete Fourier transform has been greatly reduced, making the realization of digital signal processing possible. Fast Fourier Transform can also be used to perform a series of related fast operations, such as correlation, convolution, power spectrum and other operations. The fast Fourier transform can be made into a dedicated device, or it can be realized by software. Similar to fast Fourier transform, other forms of transform, such as Walsh transform, number theory transform, etc. can also have their fast algorithms.
Digital signal processing spectrum analysis
An analysis method to describe signal characteristics in the frequency domain, which can be used not only for deterministic signals, but also for random signals. The so-called deterministic signal can be represented by a predetermined time function, and its value at any time is definite; random signals do not have such characteristics, and its value at a certain time is random. Therefore, random signal processing can only be analyzed and processed based on random process theory and statistical methods. For example, statistics such as mean, mean square, variance, correlation function, power spectral density function, etc. are often used to describe the characteristics of random processes or random The characteristics of the signal.
In fact, most of the random processes that are often encountered are stationary random processes and ergodic. Therefore, the average of its sample function set can be determined according to the time average of a certain sample function. Although the stationary random signal itself is still uncertain, its correlation function is definite. When the mean value is zero, the Fourier transform or Z transform of its correlation function can just be expressed as the power spectrum density function of the random signal, which is generally referred to as the power spectrum. This feature is very important, so that fast transformation algorithms can be used for calculation and processing.
The data observed in practice is limited. This requires the use of some estimation methods to estimate the power spectrum of the entire signal based on limited measured data. For different requirements, such as reducing the deviation of spectral analysis, reducing the sensitivity to noise, and improving spectral resolution. Many different spectral estimation methods have been proposed. In linear estimation methods, there are periodogram method, correlation method and covariance method; in nonlinear estimation methods, there are maximum likelihood method, maximum entropy method, autoregressive moving average signal model method and so on. Spectrum analysis and spectrum estimation are still under research and development.
The application field of digital signal processing is very extensive. As far as the source of the acquired signal is concerned, there are communication signal processing, radar signal processing, remote sensing signal processing, control signal processing, biomedical signal processing, geophysical signal processing, vibration signal processing, etc. According to the characteristics of the processed signal, it can be divided into voice signal processing, image signal processing, one-dimensional signal processing and multi-dimensional signal processing.
Digital signal processing, voice signal processing
Voice signal processing is one of the important branches in signal processing. It includes the main aspects: speech recognition, language understanding, speech synthesis, speech enhancement, speech data compression, etc. Each application has its own special problems. Speech recognition is to extract the characteristic parameters of the speech signal to be recognized in real time, and match it with known speech samples, so as to determine the phoneme attributes of the speech signal to be recognized. Regarding speech recognition methods, there are statistical pattern speech recognition, structure and sentence pattern speech recognition. These methods can be used to obtain important parameters such as formant frequency, pitch, voice, and noise. Speech understanding is the theory and technology of natural language dialogue between humans and computers. Base. The main purpose of speech synthesis is to enable the computer to speak. For this reason, it is first necessary to study clearly the changes of speech characteristic parameters over time during pronunciation, and then use appropriate methods to simulate the process of pronunciation and synthesize it into language. Other related language processing issues also have their own characteristics. Voice signal processing is the basis for the development of intelligent computers and intelligent robots, and the basis for the manufacture of vocoders. Voice signal processing is a rapidly developing signal processing technology.
Digital signal processing image signal processing
The application of image signal processing has penetrated into various fields of science and technology. For example, image processing technology can be used to study the trajectory of particles, the structure of biological cells, the state of landforms, the analysis of meteorological clouds, the composition of stars in the universe, and so on. In the practical application of image processing, remote sensing image processing technology, tomographic imaging technology, computer vision technology and scene analysis technology have obtained great results. According to the application characteristics of image signal processing, processing technology can be roughly divided into image enhancement, restoration, segmentation, recognition, coding, and reconstruction. These processing technologies have their own characteristics and are developing rapidly.
Digital signal processing vibration signal processing
The analysis and processing technology of mechanical vibration signals has been applied to the research and in production. The basic principle of vibration signal processing is to add an exciting force to the test body as an input signal. Monitor the output signal at the measuring point. The ratio of the output signal to the input signal is called the transfer function (or transfer function) of the system constituted by the test body. The so-called modal parameter identification is carried out according to the obtained transfer function, and the main parameters such as the modal stiffness and modal damping of the system are calculated. In this way, a mathematical model of the system is established. Then the dynamic optimization design of the structure can be made. These tasks can all be carried out using digital processors. This analysis and processing method is generally called modal analysis. In essence, it is a special method used in signal processing in vibration engineering.
Digital signal processing, geophysical processing
In order to explore deep underground oil and gas and other mineral deposits, seismic exploration methods are usually used to detect stratigraphic structure and lithology. The basic principle of this method is to apply an artificial shock at a selected location. For example, an explosion method is used to generate a vibration wave that propagates underground, and a reflected wave is generated when it encounters the boundary of the stratum. Place a column at a certain distance from the vibration source. The susceptor receives the reflected wave that reaches the ground. Judge the depth and structure of the formation from the delay time and intensity of the reflected wave. The seismic record received by the susceptor is more complicated and needs to be processed before geological interpretation can be carried out. There are many processing methods, including deconvolution method, homomorphic filtering method, etc. This is a problem that is still under study.
Digital signal processing and biomedical processing
Signal processing in biomedicine is mainly used to assist the research of basic biomedical theories and for diagnostic examination and monitoring. For example, it is used for basic theoretical research in cytology, neurology, cardiovascular science, genetics, etc. The human brain nervous system is composed of about 10 billion nerve cells, which is a very complex and huge information processing system. In this processing system, the transmission and processing of information are carried out in parallel, and has special functions. Even if one part of the system is obstructed, other parts can still work. This is something that computers cannot do. Therefore, the research on the information processing model of the human brain has become an important subject of basic theoretical research. In addition, the study of nerve cell models, the study of chromosomal functions, etc., can all be carried out with the help of the principles and techniques of signal processing.
Signal processing is used in more successful cases of diagnostic examination, such as automatic analysis system of brain electricity or electrocardiogram, tomographic imaging technology, etc. Tomography technology is a major invention in the field of diagnostics. The basic principle of X-ray tomography is that X-rays pass through the observed object to form a two-dimensional projection of the object. After the receiver receives it, it can be restored or reconstructed to calculate the two-dimensional projections in a series of different directions, and the tomographic information of the entity can be obtained after arithmetic processing, so that the tomographic image can be obtained on the large screen. The application of signal processing in biomedicine is in a stage of rapid development.
Digital signal processing has many other uses, such as radar signal processing, geoscience signal processing, etc. Although they have their own special requirements, the basic technology used is roughly the same. In these aspects, digital signal processing technology plays a major role.
Related terms
·Instruction cycle: the time required to execute an instruction
·TMS320VC5402-100, 100MHz, 10ns
·MAC time: the time of one multiplication plus one addition
·FFT execution time: the time required to run an N-point FFT program
·MIPS: the execution per second Millions of instructions, TMS320VC5402-100, 100MIPS
·MOPS: That is, millions of operations per second are executed
·MFLOPS: That is, millions of floating-point operations are executed per second
p>
·BOPS: One billion operations per second
Common chip
1)Power supply: TPS73HD3xx, TPS7333, TPS56100, PT64xx
2)Flash: AM29F400, AM29LV400, SST39VF400
3)SRAM: CY7C1021, CY7C1009, CY7C1049
4)FIFO: CY7C425, CY7C42x5
5)Dualport: CY7C136, CY7C133, CY7C1342
6) SBSRAM: CY7C1329, CY7C1339
7)SDRAM: HY57V651620BTC
8)CPLD: CY37000 Series, CY38000 series, CY39000 series
9)PCI: PCI2040, CY7C09449
10)USB: AN21xx, CY7C68xxx
11)Codec: TLV320AIC23, TLV320AIC10
12)A/D,D/A : ADS7805, TLV2543.