Fully-analog array signal processor using 3D aperture engineering

TL;DR

FASP heralds a paradigm shift in signal processing for super-resolution optics, advanced radar, and 6G communications and has the capability to achieve ~N times higher angular resolution than the Rayleigh diffraction limit.

Abstract

The rapid progress in radar and communication places increasing demands on low-latency and energy-efficiency array signal processing methods. There is an emerging direction of constructing analog computing processors for directly processing electromagnetic (EM) waves. However, the existing methods are constrained by 2D physical aperture and imprecise design process with inefficient computing architecture, resulting in limited sensing resolution and number of separated sources. Here, we present a fully-analog array signal processor (FASP) using 3D aperture engineering framework to perform super-resolution direction-of-arrival estimation, source number estimation, and multi-channel source separation in parallel for both coherent and incoherent sources. 3D aperture engineering is realized by constructing deep cascaded metasurface layers so that the diffractive propagation from oblique incident fields can be layer-wise modulated and piecewise encoded for perceiving EM fields far exceeding physical aperture limits. The multi-dimensional synthetic aperture (MSA) training is developed to characterize the metasurface modulation and optimize the neuro-augmented physical model for extending system aperture and generating high-order nonlinear angular response. FASP orthogonalizes the array response vectors of communication channels to map them into antenna detectors in the analog domain. The $N$-layer FASP has the capability to achieve ~N times higher angular resolution than the Rayleigh diffraction limit. Experiments further validate the source number estimation and independent channel separation of 10-target that can suppress radar jamming signals by ~20 dB and enhance channel communication capacity by 13.5 times at 36~41 GHz. FASP heralds a paradigm shift in signal processing for super-resolution optics, advanced radar, and 6G communications.

Figures (13)

• Figure 1: FASP using 3D aperture engineering. (A) The traditional electronic ASP architecture comprises an antenna array with a limited 2D aperture size, phase shifters, RF mixers, ADCs, and DSPs, which leads to high processing complexity. (B) SAR enlarges the aperture size and enhances the resolution through time-consuming mechanical scanning and virtual aperture synthesis processing. (C) FASP architecture utilizes deep cascaded metasurfaces to extend the perception aperture size with oblique incident waves and directly perform layer-wise modulation and coherent synthesis on the 3D extended aperture in the analog domain, thereby achieving array response vector orthogonalization and super-resolution sensing. (D) FASP performs array signal processing tasks at the speed of light, with high parallelism and low power consumption, can facilitate various wireless sensing and communication tasks.
• Figure 2: MSA training of FASP for super-resolution sensing. (A) Both input fields within and outside apertures are utilized to train the model to achieve light-speed aperture synthesis for ASP tasks. (B) MSA training method is accelerated with the neuro-augmented forward model that integrates mini-FCNNs to characterize the high-dimensional modulation of meta-atoms. Meta-atom geometries are updated by minimizing the loss function during training. (C) As the number of layers increases, both the amplitude and phase of output fields exhibit higher-order nonlinear responses to the incident angle. (D) FASP achieves scalable super-resolution for DOA estimation. Increasing the number of layers continuously improves the angular resolution while maintaining high energy efficiency and CSR. (E) Confusion matrix for two-target DOA estimation in simulation. (F) Full-wave simulation showing EM field distribution in the FASP. (G) A five-layer FASP comprising metasurface layers. (H) Experimental confusion matrix for two-target DOA estimation.
• Figure 3: ISCC architecture for parallel ASP tasks. (A) FASP selectively maps coherent and incoherent sources from different angles to the corresponding receiving antennas (top). The manufactured metasurfaces and a representative experimental output field of a single source is shown (bottom). (B) Experimental system utilizes AWG to generate independent source signals with FSW to analyze the processed signals. (C) Experimental confusion matrix evaluated on the single-source testing dataset. (D) MSA training enables the DOA estimation under broader wavelength range. (E) Experimental energy distribution matrix for 3$^\circ$ super-resolution DOA estimation of two sources under 45 angular interval permutations. (F) Experimental confusion matrix (left) and numerically evaluated confidence values (right) for source number estimation with $m$ varying from 1 to 10. (G) Eye diagrams and EVMs of QAM-16 and BPSK signals received by spectrum analyzer with and without FASP. (H) EVMs of received QAM-16 source signals from different angles (left) and at different symbol rates (right). FASP substantially improve the communication capacity.
• Figure 4: Radar Anti-Jamming with Multiple Separation Channels. (A) Time domain waveform and pulse compression results of the mixed LFM signal and the DFTJ (top). After FASP source separation, the SNR of the LFM signal is improved, and the DFTJ is effectively suppressed (bottom). (B) When the angular interval is increased from 3° to 6°, the jamming suppression ratio of the demixed LFM channel reaches 35 dB. (C) Before separation, the LFM signal is overwhelmed by the high-power NSJ (left); after separation, the LFM signal waveform and pulse compression spikes are detectable (right). (D) Waveform and spectrum of the mixture of the LFM, DFTJ, and single-tone signal when three signals are simultaneously transmitted (top). After multi-channel source separation, the LFM, DFTJ, and single-tone signal are successfully demixed in both the time and frequency domains (bottom).
• Figure S1: Mini-FCNN for predicting meta-atom modulations. (A) The mini-FCNN structure comprises a Fourier feature extraction layer and two fully-connected layers followed by ReLU activation. (B) Comparison between mini-FCNN predictions and CST simulations for meta-atom amplitude and phase responses. The high similarity between the modulation values across different geometric settings demonstrates the effectiveness of the proposed neural network for multi-dimensional modulation prediction. (C) The mean absolute errors of amplitude and phase modulation predictions under different frequencies using mini-FCNN.