Chips Ahoy: Analog-Digital Chiplets for O(n) FFT Multi-Beam Apertures
Abstract
Fast Fourier Transforms (FFTs) are a class of fast algorithms for efficiently realizing n-point discrete Fourier transforms (DFTs) and its inverse at low multiplication complexity. FFTs famously reduce the O(n^2) matrix-vector multiplication complexity of the DFT-vector product down to O(n log n) using sparse matrix factorization - a major reduction for large n. Precise lower bounds on FFT complexity as a function of DFT size n are known. Reducing complexity beyond FFT lower bounds is provably impossible. However, this limitation arises due to an assumption of exact computation. We use controlled compromises in computing DFTs to achieve the world s fastest and smallest-SWaP FFTs. The approximated FFTs are realized as analog and digital chiplets and integrated into a single package. The system in package (SiP) is used for wideband multi-beam RF apertures relevant for Navy applications. A prototype aperture consisting of a grid of beams of dimension 128x32 is proposed as a test platformfor the SiP implementation. The aperture will operate inthe 1-5 GHz band with an instantaneous bandwidth of 500 MHz.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Jun 13, 2024
- Source ID
- N000142412369
Entities
People
- Habarakada Madanayake
Organizations
- Florida International University
- Office of Naval Research
- United States Navy