Analog Building Blocks: VDIBA and CDBA Based Energy-Efficient High-Speed Memristor Emulator for Neuromorphic Applications

Abstract

In the field of neuromorphic computing, there is a growing need for high-frequency memristor emulators, especially for pattern recognition, image classification, and edge detection. A high-frequency memristor-based neural network can enhance synaptic weight updates and accelerate learning. This article presents an innovative memristor emulator circuit using CMOS-based building blocks: the Voltage Differencing Inverting Buffered Amplifier (VDIBA) and the Current Differencing Buffered Amplifier (CDBA). Our design achieves a maximum operating frequency of 60 MHz with a power consumption of only 2.25 mW. The memristor emulator is resistorless, electronically tunable, and functions in both grounded and floating configurations, as well as in incremental and decremental modes. We provide an analysis of transient behavior and voltage-current (V-I) characteristics, along with assessments of robustness and adaptability under various conditions. This memristor emulator is tailored for Adaptive Neural Networks (ANN) to mimic biological behavior and for Memristive Integrated-and-Fire (MIF) neuron circuits to replicate biological neurons, all developed using 180 nm CMOS technology. The proposed design has also been verified using ICs CA3080, LT1193, and AD844.