Wideband Adaptive RF Protection
Abstract
Simultaneous-transmit-and-receive (STAR) or full-duplex (FD) transceivers are challenged by the tremendous self-interference from the transmitter to the receiver. This self-interference (i) is typically 130dB larger than the desired signal, demanding very high levels of cancellation (SIC), (ii) undergoes substantial frequency dispersion due to long delay spreads in the self-interference channel, (iii) and is subject to changes in the electromagnetic environment, requiring adaptive cancellation. Consequently, despite research efforts over the past five years or so, including pioneering efforts by CoSMIC lab under PI Prof. Krishnaswamy over several DARPA programs such as RF-FPGA, ACT and SPAR, current solutions achieve insufficient cancellation over narrow bandwidths and consume substantial DC power while significantly degrading receiver noise. This program leverages several breakthrough innovative concepts to propose a Modulated Ultra-wideband Self-Interference Cancellation (MUSIC) module that meets, and in several cases, exceeds the DARPA WARP metrics in both 0.1-1/1-6GHz bands, including: 1. Quasi-electrostatic (Switched-Capacitor) Delay Lines: that are wideband, highlyminiaturized, low-loss, fully-integrated in CMOS, and can support embedded voltage gain, enabling replication of the long delay spreads of the SI channel, allowing wide cancellation bandwidths, while relaxing the active gain requirements in the SIC path; 2. Active Noise-Cancelled Self-Interference Cancellation (NC-SIC): in which, the active RXside cancellation circuit is able to cancel +10-20dBm SI levels while cancelling its own noise and distortion; 3. Metamaterial/Switched-Resonator-Based Multi-port Delay Lines: that are compact, low loss, operate in the 0.1-1/1-6 GHz ranges, respectively, and enable transformative delaybandwidth (DBW) products for acoustics, providing coarse high-linearity delay for the first-lineof- defense cancellation; 4. Hybrid Analog-Digital Embedded Sensing and Machine Learning (ML)-based Control: consisting of a wideband channel estimator with direct signal-to-feature conversion and an embedded neural processor that runs an ML/optimization model, which dynamically adapts the coefficients of the multidomain FIR filter canceller in real-time.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Feb 24, 2021
- Source ID
- FA86502117012
Entities
People
- Harish Krishnaswamy
Organizations
- Air Force Research Laboratory
- Columbia University
- United States Air Force