The Key Technologies Of Low Power FIR Digital Filter Design

Finite impulse response(FIR) filters with high stability and linear phase characteristics,are widely used in telecommunications equipment and control systems. With the rapiddevelopment of large scale integrated circuits, the frequency of signals and the data containedare increasing, which makes the need urgent for high performance FIR. In the filters design,better performance can be achieved if key modules consume less power and resource. Therefore,the key modules optimization of FIR filter is an important issue in the field of signalprocessing.This thesis discusses the advantages and disadvantages of the various implementationmethods, and underlines how to choose the best architecture in different occasions. With thedevelopment of FPGA based on lookup table(LUT), the FIR filter implemented with memoryarchitecture has been widely used. The memory-based architecture can be implemented bydistributed algorithms(DA) and LUT-based multiplier. Based on the two methods, the thesismakes a thorough study on the principles and implementation of high-performance FIR. Themain contents of this thesis include the following aspects:(1) DA-based FIR filter design includes determination of parameters and coefficients,module partition and the final simulation. Basing on the traditional architecture of DA,optimization design by removing redundant coefficients is proposed and each module isdesigned specifically, then RTL-level coding with Verilog is carried out, finally the simulationis performed with EDA tools. The experimental results show that the power consumption andthe area of the FIR filter have been effectively reduced.(2) FIR filters are implemented by LUT-based multiplier. Two improved LUT architectures namedd-LUTand b-LUTare proposed based on the traditional APC-LUT. The d-LUTandb-LUThave no address coding unit with simple control circuit, which make the proposed FIR achievesbetter performance on power and area. To implement LUT whose inputs are large width，the LUTdecomposition method is deplored.