| Summary: | 碩士 === 中華大學 === 電機工程學系 === 85 === Digital Signal Processing (DSP) has been used widely in the field of modern technologies. It appeared to revolutionize audio and video processing, telecommunication techniques, aviation control, and earthquake prediction. Moreover, due to the fast progress of VLSI, DSP is also getting more and more attention in academic and industry.
The design of a digital signal processor has as objectives of programmable and high-efficient. The programmable condition makes a processor easy to suit for any application, and the high-efficient condition makes it to meet the speed requirement. The design of a digital signal processor can be categorized as follows:
1. General-purpose digital signal processor (GPDSP): The hardware of a GPDSP can have simple addition, subtraction, multiplication and division only. The other complex functions are fulfilled by software. The advantage of GPDSP is that it can be re-programmable for any DSP applications, and its major disadvantage is that the execution time might be too long.
2. Dedicated-Hardware digital signal processor (DHDSP): The design of the processor is based on the requirement of the application. For each application, a special hardware is designed to meet its requirement. Since the function is performed by hardware, the execution speed is extremely high. The major disadvantage is that it can not be altered for any other applications.
Since both designs can not meet the basic requirements of a digital signal processor, it is the trend to design a processor in the mixed mode. The mixed-mode digital signal processor is based on the GPDSP, and some complex and useful functions, e.g. convolution, table look-up, and Fourier transform, are implemented by special hardware. This design can not only have the advantage of re-programmable for all applications, the complex function can also be speeded up by hardware.
Table lookup is the most important function for the DSP applications. Its applications includes data compression, solutions of complex functions, and constant lookup. In this thesis, we propose a high-efficient hardware design for table lookup. In the beginning, a special searching hardware is developped which can quickly locate the data. For each found data, we use a approximation technique (linear interpolation) to improve the performance of the result. Finally, a standard interface (DMA) is used such that the design can not only work independently, it can be used with other DSP processor to improve the overall performance. The experimental results show that the maximum operating clock rate of our design is 200MHz with reasonable area.
|
|---|
