Design And Efficient Implementation Of The Fast Filter Bank

As a general form of fast Fourier transform(FFT),the f ast filter bank(FFB)has the characteristics of high side-lobe attenuation and low complexity.It also has adjacent channel merging characteristics that are not available in other filter banks.Hence,FFB has broad application prospects in software radio technologies such as digital channelization,spectrum analysis,and multi-carrier modulation.However,the lack of research on FFB hardware implementation has hindered the practical application of FFB.Based on this,this paper has carried out research on the design and efficient implementation of FFB and application innovation.The main work of this paper is as follow:Firstly,the structural characteristics of the FFT are analyzed from the perspective of the filter bank,and the amplitude-frequency response characteristics of the FFT are improved,so that the structure of the FFB is obtained.According to the structural characteristics of FFB,the unified expression of FFB is derived.Combined with the characteristics of the half-band filter,the general expression of FFB is further deduced,and the characteristics of FFB are deeply analyzed.The cor rectness of the above derivation is proved by an example.Secondly,the FFB is analyzed from the perspective of frequency response masking(FRM),and the general design criteria of the FFB are derived.Under the premise of determining the attenuation of the stop-band of each stage of the prototype filter,according to the global constraints of FFB,an optimal design method for the number of FFB channels and the width of the transition is proposed.The proposed method can design FFB with minimum complexity and minimum group delay,and proves the effectiveness of the method through simulation.subsequently,the bottom-up design method is used to study the field programmable gate array(FPGA)implementation structure of FFB.The serial and full-parallel design ideas are used to design the hardware implementation schemes of the sub-filter modules and the FFB tree structure,and the performance comparison between the two design ideas is carried out.On this basis,considering the actual hardware limitations,an optimal configuration structure design method is proposed to achieve a balance between hardware resource consumption and system data throughput.Finally,a new nonparametric spectrum analysis algorithm is proposed based on FFB.The algorithm improves the spectral analysis performance of the FFT by replacing a part of the butterfly structure of the FFT with multiple parallel FFBs,and obtains a higher precision spectrum.In this paper,the algorithm is analyzed in detail and the hardware structure of the algorithm is given.Experiments show that compared with the existing spectrum analysis algorithms,the FFB-based spectrum analysis algorithm requires more samples,but its performance against inter-spectral interference is better.Furthermore,the multiplier resources required to implement the FFB-based spectrum analysis algorithm are approximately equal to the FFT.