Selecting a Programming Scheme for Memristor Elements

Abstract

Introduction. An array of memristive elements can be used in prospective neural computing systems as a programmable resistance (analog multiplication factor) when performing operations of analog vector multiplication, discrete in time. To form the required resistance, the memristor should be subjected to a programming procedure. This article discusses conventional programming schemes and proposes a new versatile programming scheme for memristor elements.Aim. To identify or develop an optimal programming scheme for memristors by analyzing the advantages and disadvantages of existing methods.Materials and methods. The programming procedure can be carried out using either SET or RESET, depending on a different direction of movement according to the volt-ampere characteristic of the memory and its transfer to a particular state. The programming process is controlled in the LTspice circuit modeling program.Results. Typical programming schemes of memristors were analyzed; advantages and disadvantages of existing methods were revealed. A new versatile circuit based on a variable resistor was proposed. The circuit was simulated both under a fixed resistance of the variable resistor and when varying the memristor resistance values within their permissible range.Conclusion. In comparison with the RESET mode, the SET programming mode provides for a greater linearity of variations in the memristor resistance. The use of a circuit based on a variable resistor and a bipolar voltage source allows programming of any type and eliminates the need for recommutation of the memristor. The simulation results confirm the feasibility of the proposed method. The proposed circuit can be complemented not only with a comparator, but also with an ADC. This will provide the possibility of selecting various means for measuring the memristor resistance both during programming and for the purpose of monitoring the memristor resistance at the end of the procedure.

Keywords

• memristor
• programmable resistance
• analog neural calculations
• programming scheme
• artificial intelligence