Welcome to the AMTA paper archive. Select a category, publication date or search by author.
(Note: Papers will always be listed by categories. To see ALL of the papers meeting your search criteria select the "AMTA Paper Archive" category after performing your search.)
The use of open-ended coaxial probes for dielectric characterization of breast tissue requires precise coupling between the probe and the skin surface, avoiding air gaps and excessive pressures. Since dielectric properties are very sensitive to these conditions, improper adaptation can compromise the precision and repeatability of the measurements, limiting their effectiveness in clinical applications such as tissue grading or pathology monitoring. This article evaluates the geometric performance of four open-ended coaxial probe tip configurations (flat, conical, beveled, and semispherical) with the objective of evaluating their fit to biological tissues. A custom-built measurement system operating between 1 and 8 GHz was employed to computationally simulate and analyze the S11 response for each probe geometry, assessing both precision and effective penetration depth. The results show that the conical and beveled probes provide shallow penetration depths (<2 mm), making them suitable for in vivo or ex vivo procedures, such as biopsies. The semispherical probe achieves deeper penetration (approximately 2 mm deeper than the flat probe), which is beneficial in surgical settings or before more complex imaging techniques, such as mammography. In terms of precision, the semispherical and conical probes demonstrated the best performance for breast tissue and tumor characterization, with an effectiveness rate exceeding 80%. These findings allow for the assignment of complementary functions to each probe type, improving their integration into clinical settings for rapid and reliable tissue diagnosis and assessment.
A new sparse sampling and compressive sensing based reconstruction and near-field imaging technique is introduced for the measurement of electrically large production test and diagnostics of nose-mounted commercial radomes. Simulated results are presented, where it is demonstrated that far- field results with an equivalent multi-path level of better than - 60dB can be obtained from circa 10% of the points required by a classical Nyquist equiangular spherical near-field acquisition scheme for the case of an electrically large, i.e. full size, commercial airliner nose-mounted radome enclosing an x-band weather radar. Furthermore, a new method for the rapid noninvasive nondestructive imaging and identification of defects within these radomes is presented that provides significantly clearer fault detection at a far earlier stage within the radome measurement campaign than has previously been possible.
This paper presents the design and simulation-based evaluation of a high-resolution millimeter-wave (mmWave) MIMO (Multiple-Input Multiple-Output) antenna array system for non-invasive medical diagnostics. The system is specifically optimized for applications such as early-stage tumor detection and soft tissue anomaly mapping, where high spatial resolution and tissue penetration are crucial. A 4×4 MIMO antenna array operating in the 28–40 GHz frequency band is proposed, leveraging the inherent advantages of mmWave frequencies— namely, shorter wavelengths for finer imaging resolution and wide bandwidth for enhanced contrast.The MIMO antenna array is designed using Rogers RT5880 substrate with a dielectric constant of 2.2 and a thickness of 0.787 mm to ensure minimal dielectric loss and mechanical stability. High-fidelity electromagnetic simulations were conducted using ANSYS HFSS to validate the antenna design. The resulting 3D radiation patterns confirm the beam directivity and uniform power distribution across all elements. The array was then integrated into a synthetic aperture radar (SAR)-based imaging model in MATLAB, where point- spread function (PSF) analysis revealed a lateral resolution of 3.2 mm and an axial resolution of 2.5 mm at 35 GHz. Imaging simulations on a multilayer human tissue-equivalent phantom model—comprising skin, fat, and muscle layers—demonstrated the system’s ability to resolve dielectric contrasts simulating benign and malignant tissue anomalies. The proposed MIMO antenna array enables real-time, contactless, and radiation-free imaging, positioning it as a cost-effective alternative to traditional imaging modalities such as X-ray or MRI for preliminary screening and continuous monitoring. The fully simulated results validate the concept’s feasibility and effectiveness for non-invasive medical diagnostics, particularly in point-of-care settings.
With the increasing number of channels in integrated radar MMICs, radar modules and networks, beamformer calibration techniques must adapt to the physical dimensions of these sensors. Typical far-field calibration requires measurements at the Fraunhofer distance. This ensures a maximum phase error of 22.5° over the aperture. However, literature shows that phase errors below 5° are required for acceptable side-lobe suppression. Compact Antenna Test Ranges (CATR) create virtual far-field conditions in limited space but unfortunately they introduce magnitude and phase errors in their measurement. A method for calibrating MIMO radars in CATR settings is presented using spatial averaging to reduce these errors systematically. Simulations with a 16-channel FMCW radar show maximum errors of below 0.25dB for magnitude and less than 2◦ for the phase with a single-digit number of spatial averages. Calibration with a 77-GHz MIMO radar sensor in the CATR confirms the technique’s ability to mitigate test zone non-idealities, improving radar imaging quality.
This paper presents a computational imaging method that uses a metasurface-based transmitting antenna to reconstruct high-resolution scattering signatures of arbitrarily shaped radar targets. Instead of relying on mechanical scanning or bulky antenna arrays, the proposed approach takes advantage of programmable phase distributions on a metasurface to generate diverse illumination patterns. These patterns help encode different scattering responses from the target, which are collected by a single receiving antenna. The scattered electric field is modeled as a linear matrix equation. This model includes the effects of each transmitting unit cell, the distribution of equivalent point scatterers on the target surface, and the propagation paths between the transmitter, the target, and the receiver. The result is a forward model that links randomized metasurface phase patterns to the measured backscattered field. To achieve high-resolution image recovery, a large number of unknowns—representing the complex amplitudes of point scatterers—must be estimated. Although multiple phase masks are used to increase measurement diversity, the number of measurements is still much smaller than the number of unknowns, making the system underdetermined. To solve this, we use a compressive sensing technique known as basis pursuit denoising (BPDN), which finds sparse solutions in the complex domain and enables accurate reconstruction despite the limited data. We verify the proposed method using numerical simulations on canonical radar targets. The reconstructed images show high similarity to the original targets, confirmed by quantitative comparisons such as structural similarity indices and error norms. These results demonstrate that the method can effectively extract scattering profiles using a compact, electronically reconfigurable antenna system. This work shows the potential of combining metasurface technology with compressive sensing to build efficient, lightweight radar imaging systems that do not require complex hardware setups.
We compare three techniques to improve sample efficiency when measuring near-field emissions from embedded computing systems on a planar region close to the device-under- test (DUT). The techniques are based on either an expansion as trigonometric polynomials or on assumptions about the mathematical regularity of functions that describe the measurands. These assumptions lead to a non-adaptive compressed sensing (CS) approach, an adaptive sparse grids (SG) approach, and an adaptive approach through a Gaussian process regression (GPR) surrogate model. We compare the techniques on a simulation of small computing devices that are measured by a near-field magnetic field probe on a planar region at different subsets of Nyquist grids at different heights above the device. The simulations show that observation height and observation area are important parameters for deciding an effective subsampling algorithm: when observation heights are lower, GPR and CS perform similarly in terms of sampling efficiency for the same estimation error. SG sampling outperforms Nyquist-style sampling, but requires more samples than CS or GPR.
In this work, we present C2MI (Covariance- Coupling Microwave Imaging), a second-order statistical formulation that enhances imaging under sparse measurements. C2MI models a second-order forward system, SL = M, where S comprises upper-triangular elements of covariance matrices from permuted and transformed sensing patterns, L is the vectorized upper triangle of the scene covariance matrix, and M contains variances of measurement vectors. We solve this underdetermined system via l1-regularized optimization with a structural constraint enforcing diagonal dominance: each off-diagonal Lij must not exceed its corresponding diagonal terms Lii and Ljj . This reflects the prior that strong inter-location coupling (captured by covariance off-diagonals) occurs only when their variances (diagonals) are high. The result is a sparse, interpretable, and structured estimate of scene covariance. Compared to l2 + total variation (TV) methods, C2MI accurately reconstructs sparse targets using fewer masks—for instance, recovering 14 active pixels in an 8×8 grid with only 20 frequency-sweep masks, versus 28+ for l2+TV. This improves efficiency, enabling real-time, portable, and low-cost microwave imaging. We validate C2MI in MATLAB simulations with metasurface antenna phase-gradient measurements, showing promise for security, biomedical, and non-invasive sensing applications.
This paper investigates the use of self-supervised learning algorithms to reconstruct a high-quality image directly using data from microwave imaging systems. Specifically, we propose using neural fields, particularly the model entitled Instant Neural Graphics Primitives (InstantNGP) [1], chosen for its balance between speed and ability to capture fine details—to improve the spatial resolution of microwave image reconstructions. The performance of this approach is verified using both simulated and real experimental data of an Inverse Synthetic Aperture Radar (ISAR) operating in the C-band. The real and imaginary components of the complex return signal generated by this simulator were used as ground truth to compare to the predicted/synthesized return signal from the machine learning (ML) model. Both the neural field approach and the conventional reconstruction using matched filtering (MF) were used to produce reconstructed images, and a qualitative comparison between the images was performed. Preliminary results indicate that the ML approach produces reconstructed images with improved range and cross-range resolution compared to those created using a traditional matched filter technique.
Absorber treatment for an anechoic range is designed to attenuate the potential reflections from the walls, ceiling, and floor and to keep a certain level below the direct path between the range antenna (or probe) and the quiet zone (or minimum radiated sphere for spherical near-field ranges). There are, however, some antenna measurement systems where the range changes or moves as the data is acquired. In some cases, the probe moves around the antenna-under-test (AUT) along a section of circle supported by an arch or a gantry. In other ranges, the multiple probes are switched on and off; these probes are supported by an arch. Because the direction of the range moves with respect to the walls, ceiling, and floor, it is a bit more complex to arrive to an optimal absorber layout, as well as locating the preferred placements for the instrument rack, door, and vents in the range. In this paper, a higher-order-basis-function method of moments approach is used to model a gantry-supported probe as it moves around the location of the AUT. The power density at the walls as the probe moves is analyzed to arrive to an optimal absorber layout that will provide adequate levels of reflections for measuring an antenna. The paper looks at a gantry that moves from +135° to -135° with the AUT rotating 180° and for a gantry that moves from 0° to +135° with the AUT rotating 360°. The latter will require a smaller range with one of the walls closer to the location of the antenna under test. A series of recommendations based on the electrical size of the absorber at different areas of the range are provided.
The reduction of acquisition time in planar near field systems is a high interest topic when active arrays or multi beam antennas are measured. Different solutions have been provided in the last years: multi-probe measurements systems and the PlanarWide Mesh (PWM) methodology, which implements a non redundant sampling scheme that reduces the number of samples required for the far-field transformation, are two of the most well known techniques. This paper proposes the combination of both approaches to derive a multi-probe PWM grid which reduces the measurement times to the minimum. The method is based on treating the near-field to far-field transformation as an inverse source problem. The multi probe PWM is designed with a global optimization process which finds the best measurement locations of the probe array that guarantee a numerically stable inversion of the problem. A simulated measurement example with the VAST12 antenna is presented where the total number of samples is reduced by a factor of 100 using a 4×4 probe array
Mode filtering has been shown to be very effective in suppressing spurious reflections in antenna measurements. Specifically, it has been well documented that in the quasi-far-field, the two polarizations are decoupled, making it possible to apply standard cylindrical near-field theory on the amplitude and phase data acquired from a single polarization measurement on a great circle cut [1]. The method was further extended to allow data collected from an unequally spaced angular abscissa by formulating the solution as a pseudo-inversion of the Fourier matrix [2]. This formulation, however, can be prone to spectral leakage because of nonorthogonality of the Fourier basis on an irregularly sampled grid, especially when the positions deviate significantly from the regular grid [2]. In this paper, we propose to use Compressed Sensing (CS) to compute the Cylindrical Mode Coefficients (CMCs), which improves the signal to noise ratio, allowing more accurate recovery of the prominent modes. The CS recovery is tenable because with the coordinate translation of the measurement pattern to the rotation center, the Maximum Radial Extent (MRE) of the antenna under test is greatly reduced, making CMCs quite sparse in the mode domain. The novel application of CS presented in this paper further expands the generality of the mode filtering method, which is now applicable to under-sampled data (at below the Nyquist rate) acquired on positions that grossly deviate from the equally-spaced regular grid.
In this paper, the full-wave computational electromagnetic simulation of the production test, measurement, and calibration of a 5G, 24 elements, C-band, active, planar array antenna together with a representative open-ended rectangular waveguide probe with, and without, absorber collar were evaluated using a large computing cluster and a proprietary full-wave solver. In this way, various components within the measurement could be carefully and precisely examined providing a framework for further measurement optimization. Particular attention has been paid to the presence of the standing waves in the simulated near-field measurement. This is a crucial feature of most practical measurements, but is omitted from the vast majority of simulations due to the computational effort required to evaluate it, and which is absent from the standard near-field theory. Here, the presence and impact of this phenomenon has been carefully examined with a range of intensive simulations being harnessed to quantify their impact, as well as enabling various methods for their minimization to be explored in a convenient and highly controlled fashion.
This paper proposes a fast human exposure Absorbed Power Density assessment approach, based on a combination of over-the-air radiative field measurements and fullwave electromagnetic simulations. This so-called augmented OTA technique relies on the computation of an equivalent source or digital twin, which reproduces the radiation properties of the device under test. At short separation distances, the interaction between the human model and the device is however not negligible. A novel solution to model the influence of multiple reflections is introduced, where the inside of the equivalent source box is filled with a perfect electric conductor, thereby creating a reflective digital twin model. Simulation results demonstrate the relevance of this approach for enabling accurate absorbed power density evaluations.
Extrapolation method is regarded as one of the most accurate methods for obtaining the absolute far-field gain of an antenna. This paper will compare the efficacy of several data processing techniques for calibrating low frequency antennas with long ring down time. Traditionally, measurement data are preprocessed to remove ripples from multipath reflections before a curving fitting is applied. We will first investigate two traditional data processing techniques. The first technique is to apply time domain gating to the vector response vs. frequency data at each separation distance. Then the gated data as a function of distance is fitted to the polynomial equation. The second technique is spectrum domain filtering. The vector response as a function of distance is transformed to k domain at each frequency. A band pass filter is applied in k domain to keep only the direct antenna response. In this study, we propose a new approach - the magnitudes of the antenna response as a function of distance including the ripples is fitted to a more complete generalized antenna response equation with the antenna-to-antenna multiple reflection terms included. This paper will compare the three techniques using a set of measurement data on double-ridged waveguide horn antennas in a fully anechoic extrapolation range.
The success and efficiency of many classical iterative plane-to-plane based phase retrieval algorithms is to a large extent dependent upon the fidelity of the initializing, i.e. guiding, phase estimation [1], [2]. This is especially so when using these techniques to recover the phase of active electronically scanned array antennas such as those employed within beam-steering mm-wave Massive MIMO antenna systems intended for 5G New Radio applications where the performance of the algorithm, and its ability to not become trapped within one of the (possibly many) local minima, is particularly dependent upon the quality of the initializing guess where access to a phase reference is not always convenient, or even possible. Many traditional phase recovery iterative Fourier methods employ simulation or passive measurement supported phase initialization [1], however this information is not always readily available, or in the measurement may require a destructive, invasive, examination of the device under test (DUT). In this work we address this issue by presenting a proof of concept which employs a machine learning based neural network [3] to estimate the initializing phase function based on the assessment of the measured amplitude only near-field pattern. Here, we show that there is sufficient information contained within the difference between the two near-field amplitude only scans to be able to determine the antenna beam steering characteristics. A simplified beam steering case with electronic scanning in one, or more, scanning axes is demonstrated and verifies the power of the novel method, as well as illustrating its inherent resilience to noise within the amplitude only measurements, and verification of the robustness of the approach thereby extending the range of measurement applications for which this class of iterative Fourier algorithms may be successfully deployed [4].
This paper proposes a range-Doppler imaging method based on FFT-MUSIC method for FMCW radar systems. With the growing significance of vehicle and human motion recognition in automotive radar, the accuracy of conventional deep learning network-based recognition methods is reduced because it depends only on distance, speed, and angle information provided by conventional radars. Therefore, various types of imaging radar methods have recently been proposed. Among them, the range- Doppler imaging algorithm is widely used. This algorithm can simultaneously analyze both distance and velocity characteristics of a vehicle or person. However, conventional range-Doppler imaging based on the FFT algorithm has limited resolution, which cannot obtain detailed information on the target. Although the FFT algorithm is widely used in many applications, its lowresolution characteristics can limit its ability to provide detailed information. In particular, improving velocity resolution often requires the extraction of a significant amount of data. To address this issue, a range-Doppler imaging method based on FFT-MUSIC is proposed in this paper. This technique has been simulated using Remcom’s WaveFarer® software package. The proposed algorithm is effectively able to distinguish between two moving vehicles in several cases in which the ranges and velocities are too close to be resolved by conventional FFT methods. We can observe that the proposed algorithm enhances the velocity resolution by approximately twice as much as the conventional algorithm. Additionally, in indoor environments, the proposed algorithm provides a detailed representation of the indoor multipath, outperforming conventional algorithms. The high-resolution radar imaging offered by the proposed method will enable improved target recognition and thus enhance overall performance in practical applications.
This paper proposes a range-Doppler imaging method based on FFT-MUSIC method for FMCW radar systems. With the growing significance of vehicle and human motion recognition in automotive radar, the accuracy of conventional deep learning network-based recognition methods is reduced because it depends only on distance, speed, and angle information provided by conventional radars. Therefore, various types of imaging radar methods have recently been proposed. Among them, the range- Doppler imaging algorithm is widely used. This algorithm can simultaneously analyze both distance and velocity characteristics of a vehicle or person. However, conventional range-Doppler imaging based on the FFT algorithm has limited resolution, which cannot obtain detailed information on the target. Although the FFT algorithm is widely used in many applications, its lowresolution characteristics can limit its ability to provide detailed information. In particular, improving velocity resolution often requires the extraction of a significant amount of data. To address this issue, a range-Doppler imaging method based on FFT-MUSIC is proposed in this paper. This technique has been simulated using Remcom’s WaveFarer® software package. The proposed algorithm is effectively able to distinguish between two moving vehicles in several cases in which the ranges and velocities are too close to be resolved by conventional FFT methods. We can observe that the proposed algorithm enhances the velocity resolution by approximately twice as much as the conventional algorithm. Additionally, in indoor environments, the proposed algorithm provides a detailed representation of the indoor multipath, outperforming conventional algorithms. The high-resolution radar imaging offered by the proposed method will enable improved target recognition and thus enhance overall performance in practical applications.
Absorber fences have been used on compact ranges since their first implementations. The purpose of this fence is to hide the feed positioner and reduce the direct coupling between the feed and the device under test (DUT). A known problem caused by such a fence is that it diffracts the plane wave generated by the reflector, creating an interfering ripple on the illumination of the DUT in the quiet zone. Traditionally, fences have serrated edges to direct this diffracted signal away from the quiet zone. However, this redirection is not always achievable or even repeatable from one facility to the next. Often low frequency requirements drive absorber physical size, leading to very large absorbing surfaces that cannot be optimized to reduce this interfering signal. In this paper, the fence design presented in a recent publication [1] is further optimized by modifying its shape and absorbing material parameters. The performance of this new design is compared with traditional fences.
The spherical wave expansion-based transmission formula allows to accurately evaluate the coupling (or S21 parameter) between a transmitting and a receiving antenna. Its use as tool for probe corrected spherical near-field to far-field transformation is well accepted and documented. On the other hand, its direct use in the evaluation of antenna measurement performance has been exploited only in recent years. In this paper we will show how measurement performances predicted with the transmission formula compare with actual measurements. Taking as examples relatively complex antenna measurement systems like spherical near-field, plane wave generators and CATR, we will focus on the prediction of the accuracy of the measured radiation patterns, also including the approximation of reflections from the test environments, and on the evaluation of link budgets.
This communication provides an effective two-steps strategy to compensate for known 3-D probe positioning errors occurring in the non-redundant (NR) cylindrical near-to-far-field (NTFF) transformations. As first step, a phase correction, here denoted as cylindrical wave correction, is employed to perform the correction of the positioning errors relevant to the deviations of the measured NF samples from the nominal scanning cylinder. Then, an iterative procedure will be applied to retrieve the NF samples at the points specified by the adopted sampling representation from those obtained at the previous step and affected by 2-D positioning errors. Finally, after properly reconstructing the correctly distributed cylindrical samples, the data necessary to apply the classical cylindrical NTFF transformation can be restored in accurate way by employing a 2-D optimal sampling interpolation (OSI) formula. It should be noticed as, to derive the NR sampling representation, as well as the OSI scheme, it is necessary to provide a proper modeling of the antenna under test. This modeling has been got by shaping the source with a prolate spheroid. Numerical tests will show the capability of the procedure to compensate these 3-D positioning errors.