Research And Application Of Binary And Ternary Logic Circuit Based On Memristor

Compared with traditional Complementary Metal Oxide Semiconductor(CMOS),memristor as an emerging micro/nano device,also known as Resistive Random-Access Memory(RRAM),which has a size reduction,High-speed rollover,low power consumption,non-volatile regulation,and CMOS compatibility.Under continuous voltage,the resistance of the memristor will change from low resistance to high resistance,or from high resistance to low resistance.When the voltage is stopped,the resistance value remains unchanged.Corresponding to the resistance state of a memristor and a logic signal,the memristor can be applied to a binary digital circuit or even a multi-value circuit.Memristor-based logic circuits will have Compute-in-Memory(CIM)characteristics,as well as reduce circuit area and improve system integration.Memristor,as one of the important devices in the post-Moore's law era,is considered to be one of the most promising candidates for the next generation of computing systems.From the perspective of optimizing the performance of logic circuits,this manuscript combines the characteristics of memristors to design and implement logic circuits based on memristors.The following is the main research content of the manuscript:1.The modeling and preparation of Memristor:Based on Hewlett-Packard(HP)'s research on memristor preparation,a resistive layer using HfO₂,ZnO,Al₂O₃ and other inorganic oxides as memristors was prepared.According to the device's I-V curve,threshold voltage,size,transition rate and other parameters,Verilog-A code is used to model and simulate.Device simulation was performed in Cadence,Mentor,and Synopsys EDA software to ensure that the device characteristics after modeling are consistent with the prepared device characteristics.2.Design of a binary full adder circuit based on memristor:The full adder circuit is a commonly used module and important link in arithmetic operations.With the sharp increase in the amount of processed data,higher requirements are imposed on the integration of the adder.In view of this,from the perspective of optimizing the area and delay of the full adder,a binary full adder circuit based on memristor is designed and implemented in this paper.The proposed circuit can reduce the large delay time and unnecessary power loss between data movement and memory.3.Design of ternary multiplier based on memristor:From the perspective of implementing the Boolean completion set,the complete set of bivariate and ternary univariate functions based on memristors and the realization of the complete set of multivariable functions are researched.A ternary univariate function complete set plus AND and OR operation can form a three-valued logical complete set.Combined with the switching signal theory,a ternary multiplier circuit based on memristor is designed.Compared with the traditional multiplier,the new multiplier has the advantages of high calculation efficiency,low power consumption,and simple structure.4.Design of a SIMON round function circuit based on memristor:Encryption circuits have developed well in recent years,but traditional CMOS circuits suffer from sudden power failures during operation,which will cause data loss.This paper presents a scheme of a SIMON circular function circuit based on a memristor,focusing on the structure of the memristor and the SIMON algorithm.Using memristive AND gate chains and memristive XOR chains to perform column mixing and encryption operations,respectively,and inverters were used to adjust the waveform.Finally,under TSMC 65nm CMOS process,Cadence software was used to complete the simulation verification to realize the power-off memory function.Memristor-based encryption chips will have higher integration density,lower operating voltage,lower power consumption,and faster speed.5.Implementation of memristor multi-layer perceptron network based on BP algorithm:To further improve the computing power and integration of memristors,the paper fabricates a 32×32memristor array and characterizes the device microstructure.The basic electrical synaptic characteristics of memristors such as essential synaptic functions of potentiation,and depression have been demonstrated,which provided the method for neural network simulation.Based on the above characteristics,the neural network uses MNIST(National Institute of Standards and Technology)as the database and combines BP algorithm.The ex-situ training method was used to realize handwritten digit recognition,and experimentally demonstrated the recognition accuracy rate reached 97.3%in 64 states and Conductance range 45u S-140u S.