A Configurable Memristor-based Finite Impulse Response Filter

TL;DR

This paper proposes a novel, configurable analog FIR filter design using memristors that eliminates the need for A/D and D/A converters, combining the speed of hardware with the flexibility of software.

Contribution

It introduces a memristor-based FIR filter architecture with a new circuit for negative coefficients and a heuristic method for memristance calculation, enhancing configurability and accuracy.

Findings

Successful simulation validation in Cadence

Elimination of A/D and D/A converters in hardware implementation

Improved accuracy in setting filter coefficients

Abstract

There are two main methods to implement FIR filters: software and hardware. In the software method, an FIR filter can be implemented within the processor by programming; it uses too much memory and it is extremely time-consuming while it gives the design more configurability. In most hardware-based implementations of FIR filters, Analog-to-Digital (A/D) and Digital-to-Analog (D/A) converters are mandatory and increase the cost. The most important advantage of hardware implementation of a FIR filter is its higher speed compared to its software counterpart. In this work, considering the advantages of software and hardware approaches, a method to implement direct form FIR filters using analog components and memristors is proposed. Not only the A/D and D/A converters are omitted, but also using memristors avails configurability. A new circuit is presented to handle negative coefficients of the filter and memristance values are calculated using a heuristic method in order to achieve a better accuracy in setting coefficients. Moreover, an appropriate sample and delay topology is employed which overcomes the limitations of the previous research in implementation of high-order filters. Proper operation and usefulness of the proposed structures are all validated via simulation in Cadence.