Fully Parallel and Reconfigurable Realization of DFT/IDFT using In-Memory Computing

This paper presents a fully parallel and customizable approach for implementing the Discrete Fourier transform (DFT) and inverse DFT (IDFT), which are essential in a wide range of applications such as quantum computing, image/video processing, and wireless communication systems. The proposed methodology leverages the in-memory computing (IMC) technique and utilizes an array of memristors to achieve highly efficient DFT/IDFT computations in a fully parallel manner, resulting in a significant increase in throughput. To further enhance the efficiency, a novel coefficient mapping scheme is introduced, exploiting the symmetry of DFT coefficients to reduce the required number of memristors and lower the hardware cost. Notably, the proposed design is capable of performing both DFT and IDFT operations using the same hardware architecture and supports arbitrary and variable DFT/IDFT sizes.