Design and Implementation of Recursive Filters using Processing In Memory

Infinite Impulse Response (IIR) filters play a crucial role across diverse domains including signal processing applications and wireless communication systems. The recursive nature of IIR filters presents significant challenges, including high computational complexity and memory requirements, particularly in the hardware implementation of high-order filters. This paper presents a novel approach for efficient hardware implementation of IIR filters, which enables high-throughput, fully parallel, and reconfigurable realization of recursive filters. The key idea is to employ the processing-in-memory (PIM) technique, in which the filter coefficients are programmed into memristive devices, enabling them to perform essential multiply-accumulate operations efficiently. This method not only improves processing efficiency but also highlights the flexibility achieved by harnessing the unique capabilities of PIM to program filter coefficients within memristors. Furthermore, the proposed scheme supports arbitrary filter frequency responses and variable filter lengths.