The Next-Generation Very Large Array (ngVLA) is a future radio astronomy facility, building on the existing Very Large Array in New Mexico, which promises an order of magnitude improvement over current capabilities in terms of sensitivity, angular and spectral resolutions. With 263 antennas up to thousands of miles away, each one generating almost a terabit-per-second of data to be processed in real time, the design of a Central Signal Processor for the ngVLA project appears as a formidable technical challenge. We propose the SCalable, REconfigurable And Modular (SCREAM) design concept based on highly optimized building blocks that connect to each other to form the system, thus enabling economies of scale while minimizing hardware cost, size, and power. Functional flexibility is obtained through block reconfigurability. Its scalable architecture makes the SCREAM design a suitable candidate not only for the ngVLA, but also for any potential future interferometer facilities.

Radio frequency interference (RFI) is causing performance loss in microwave radiometry, radio astronomy, and satellite communications. As the number of interferers increases, the performance loss gets more severe and RFI excision becomes more difficult. In this regard, this paper introduces the multilinear algebra framework to the multi-interferer RFI (MI-RFI) excision research by proposing a multi-linear subspace estimation and projection (MLSEP) algorithm for single-input multiple-output (SIMO) systems suffering from MI-RFI. Having employed smoothed observation windows, a smoothed MLSEP (s-MLSEP) algorithm, which enhances MLSEP, is also proposed. MLSEP and s-MLSEP require the knowledge of the number of interferers and their respective channel order. Accordingly, a novel smoothed matrix-based joint number of interferers and channel order enumerator is proposed. Performance analyses corroborate that both MLSEP and s-MLSEP can excise all interferers when the perturbations get infinitesimally small. For such perturbations, the analyses also attest that s-MLSEP exhibit a faster convergence to a zero excision error than MLSEP which, in turn converges faster than a subspace projection algorithm. Despite its slight complexity, simulations and performance assessment on real-world data demonstrate that MLSEP outperforms projection-based RFI excision algorithms. Simulations also corroborate that s-MLSEP outperforms MLSEP as the smoothing factor gets smaller.

An algorithm to detect and estimate a linear mixture of signals corrupted by white Gaussian noise is presented. The number of signals is assumed to be unknown. The algorithm is based on a type of information theoretic criterion, capable of adjusting the probability of false alarm, and uses atomic decomposition refined by the expectation maximization method to efficiently compute the maximum likelihood estimate of the signal parameters. Signals are modeled as chirplets. The algorithm consistency and efficiency are shown by simulation.

Two detectors based on the GLRT (generalized maximum likelihood ratio test) are presented to detect a chirplet in coloured Gaussian noise for known and unknown covariance structures, respectively. They are applied to detect multiple chirplets by atomic decomposition, which is modified to account for the operating conditions of both detectors. Both detectors are CFAR with respect to the noise power. The second one is CFAR with respect to the covariance structure as well.

A post-processing detector for Atomic Decomposition in order to evaluate the quality of the extracted signal components is presented. It improves the signal representation and can work in a noisy environment. Its advantages over High Resolution Pursuit are investigated along with its application to the combination of two types of dictionaries to achieve high detection performance and high resolution.