Based On The Probability Calculation Of Fft

Modern communication systems have changed with the development of digitalsignal processing (DSP) technology to achieve high speed, wide band, and multimediainformation transmission. The performance improvement of DSP chips or units alsobenefits from the continuous advancement of integrated circuit technology. Today,integrated circuits have entered the deep submicron era. As device feature sizes andoperating voltages reduction with the continuous improvement of the integrated circuit,the performance of digital ICs become more and mor unreliable. Therefore, we have topay a huge cost in the terms of processing technology, design and applicationenvironment to ensure stable performance and accurate calculation for the deep sub-micrometer digital ICs.One of the characteristics of the digital signal processing is stochastic. In thetraditional communication system designs, this feature does not have to be a goodutilization. The computations in the traditional communication designs are accuratecalculation. In fact, it is enough for communication signal processing algorithms toguarantee the statistical properties of the calculation results. The stochastic calculationtransfer the traditional computation into stochastic domain, the results of stochasticcalculation can satisfy the stochastic properties requirements of system designs withthe algorithm design in stochastic domain. In stochastic computation, the valuebetween0and1is transferred the probability with which the bit is1in the random bitstream, and calculations in stochastic domain can be realized with simple gateoperations.In this dissertation, we proposed FFT algorithm structure based on stochasticcomputation,1024FFT algorithm are used as the case study. For the two mappingschemes from two’s complement system domain to stochastic domain, we design twoFFT structures. Compared with the traditional FFT implementation schemes, the performance proposed method is competitive. The design challenges in stochasticdomain are also analyzed in detail, to provide reference for stochastic computation inDSP system or other applications.