Key Technology Research Of Dynamic Reconfigurable Logic Circuits Based On Cellular Nerual Networks

The traditional computer architecture is static.Each chip is set to perform only specific functions while designing the processor chip.It makes the existing computer hardware system lacks a certain degree of flexibility and controllability.By using the reconfigurable technology,the internal structure and function of the devices in the future computer could be changed dynamically according to the different needs of consumers?This is a new type of dynamic computer architecture.One of the key tecnoligies to bulid this new computing system is about the researches of reconfigurable computing and reconfigurable hardward design.In this dissertation,the main research goal is to establish a system of reconfigurable dynamic logic model and its circuit implementation based on Cellular Nerual Networks(CNN).Three types of CNN models are introduced: the uncoupled standard CNN model,the uncoupled heterogeneous CNN model and the coupled standard CNN model.The design ideas and implementation schemes of reconfigurable dynamic logic circuits based on these three CNN model are discussed respectively.On the premise of considering the engineering application,we propose the general guidelines for designing cell template parameters and the DE algorithm is used to optimize these parameters.In order to solve the problem that CNN implements Non-LSBF with multiple inputs,an improved CFC decomposition algorithm is proposed.On the other hand,the Non-LSBF can be implemented with fewer hardware resources by using the uncoupled heterogeneous CNN model.In this dissertation,a multi-input multi-output logic function circuit with reconfigurable function is also explored by using a coupled standard CNN model.This dissertation contains the following five aspects:(1)By analyzing the dynamics characteristics of the uncoupled standard CNN and its parameter constraints while converging to a stable binary state,we proposed a more concise CNN template parameter design principles which is suitable for actual circuit implementation.Based on these above,14 basic two input single output dynamic logic gates have been successfully constructed.We also design and verified a single cell circuit based on Op-Amp.Compared with the chaotic system,the reconfigurable computing unit based on CNN system is more easily VLSI implementation and has lower dependence on the initial value.(2)The ideas and methods for achieving larger scale reconfigurable combinational logic and sequential logic circuit module have been analyzed and expanded.It can reduce the hardware resource overhead of the circuit with improve the classic CFC decomposition algorithm.A reconfigurable D flip-flop based on the uncoupled standard CNN is designed.Single-edge D flip-flops with controllable clock edges and double-edge D flip-flops can be dynamically realized by adjusting the cell template parameters.(3)This dissertation proposed an uncoupled heterogeneous CNN to realize multi-input Non-LSBF.The data samples in high-dimensional space are processed with dimension projection to establish the corresponding discriminant equation expression.In this way,two types of uncoupled heterogeneous CNN model are constructed.With the uncoupled heterogenous CNN model can greatly simplify the system structure and reduce the hardware resources.(4)Due to the coupled CNN system has richer nonlinear characteristics.In this dissertation,a coupled three-cell CNN with external input is designed to construct a multi-input multi-output reconfigurable circuit.Firstly,the stability of this coupled three-cell CNN model is analyzed.Then,based on this model,a 3-bit associative memory circuit with reconfigurable functions is realized.Finally,the influence of the initial value of the system state variable xi on the realization of associative memory is discussed in depth.(5)By analyzing the basic constraints that CNN converges to binary steady state,we proposed a cell template parameters design principle that suitable for engineering implementation.This optimization problem can be translated into a nonlinear programming problem and adopt DE algorithm to solve.Select several sets of template parameters into the circuit to verify and analyze the simulation results.