This work presents the design, fabrication and characterization of two Radio-Frequency Identification (RFID)-based antenna sensor tags for wireless, near-real-time cold-weather temperature sensing. The designed temperature sensors operate by tracking the antenna’s resonant frequency across the North American UHF RFID band (902–928 MHz), and may be integrated with any commercially available RFID chip to realize either lower cost or greater read range. The designs are implemented using standard printed circuit board (PCB) processes and commercially available passive components. Due to the limited sensing bandwidth, the two sensors employ different sensing elements to achieve different temperature sensing ranges and corresponding accuracies. Design A utilizes a temperature sensitive substrate and Design B uses a temperature sensitive capacitor. Results for Design A attain a measured temperature sensitivity of 93.8 kHz/°C, a broad temperature range of −30 °C to +30 °C, electrical size of ka = 0.4 at 915 MHz and a maximum estimated read range of 7.4 m. Design B achieves state of the art temperature sensitivity corresponding to 710 kHz/°C, a narrow temperature range of −15 °C to +15 °C, electrical size ka = 0.22 at 915 MHz and a lower bound for maximum estimated read range of 6.4 m. The two sensors exhibit good temperature accuracies of ±1.37 °C and ±0.911 °C for designs A and B respectively.
A frequency-reconfigurable filtering antenna with monopole radiation patterns is proposed using a fusion-based design paradigm, whereby the filtering function is directly integrated into the radiating structure, while co-optimization of the antenna structures enables frequency reconfiguration using a single bias voltage. The driven monopole is loaded with a stacked circular patch to generate a passband-following radiation null at the upper band edge, thereby producing a reconfigurable low-pass filtering response under a single-bias control. Additionally, short-ended, varactor-loaded parasitic strips are co-optimized with the driven monopole to generate a frequency-reconfigurable radiation null at the lower band edge while maintaining omnidirectional radiation patterns across the operating frequency range. These structures exhibit similar frequency-tuning behaviour over the capacitance range of the varactor, which enables coordinated tuning of the passband and both radiation nulls under single-bias control. Measurements confirm a 23.4% center frequency tuning range and a peak realized gain of approximately 4 dBi with an omnidirectional pattern. Simulated gain suppression exceeds 11.1 dB and 9.5 dB in the lower and upper stopbands at the highest and middle frequency states, respectively, while measurements show a minor degradation due to fabrication and component tolerances.
A versatile approach for the synthesis of dense phased array (DPA) antennas, characterized by sub-half-wavelength inter-element spacings, is here presented. The approach is able to fit user-defined power pattern masks, while fulfilling additional geometrical and/or electrical constraints on the geometry of the array aperture and/or on the array excitations. The synthesis is based on an inverse source (IS) formulation and exploits the weakly-radiating directions of the radiation operator, which causes the ill-conditioning and the non-uniqueness of the IS problem at hand. More in detail, the unknown element excitations of the DPA are expressed as the linear combination of a minimum-norm or radiating (RA) term and a suitable weekly-radiating (WR) component, which are combined into a single set of complex excitations, enabling their implementation through a conventional beamforming architecture. The former, computed via the truncated singular value decomposition (SVD) of the array radiation operator, is devoted to generate a far-field power pattern that fulfills user-defined pattern masks. The other one allows one to fit additional geometrical and/or electrical constraints on the geometry of the array aperture and/or on the beam-forming network (BFN) when determined with a customized global optimization strategy. A set of numerical examples, concerned with various array arrangements and additional design targets, is reported to prove the effectiveness of the proposed approach.
Spatially varying metasurfaces embedded in closed cavities destroy the modal separability that underlies conventional cavity Green-function formulations. This paper develops a dyadic Green’s function (DGF) for a perfectly conducting rectangular cavity loaded by an internal, infinitesimally thin, spatially varying metasurface described by scalar electric and magnetic generalized sheet transition conditions (GSTCs). The cavity is partitioned into two PEC-terminated subcavities and represented by diagonal TE/TM Dirichlet-to-Neumann operators. Projection of the GSTCs onto the transverse cavity basis produces susceptibility-overlap matrices whose off-diagonal entries explicitly quantify metasurface-induced intermodal coupling. The resulting environmental operator yields the compact DGF form GMS = GD +PM−1B, allowing a single factorization to be reused for arbitrary electric-dipole positions and orientations. For harmonic susceptibility profiles, analytical selection rules directly map the metasurface spatial spectrum to allowed TE/TM modal transitions. One- and two-dimensional numerical sweeps verify these rules with forbidden-mode leakage on the order of 10−14. Spatial modulation increases the separation of a representative TE11/TM11 pair from 140.6 to 307.0 MHz and produces genuine hybrid eigenfields. For a common prescribed dipole moment and weak-loss regularization, a P = 2 profile gives a 3.97-fold ratio of peak electric-field amplitudes relative to the uniform-sheet reference; this measures local field concentration, not a demonstrated received-power or system-performance gain. The formulation is validated by exact empty-cavity and uniform-sheet limits, modal convergence, reciprocity residuals below 8.1×10−16, and an independent physical-space GSTC collocation solver. The proposed framework provides a reusable analytical–numerical tool to predict modal routing, resonant spectral control, and field redistribution in idealized metasurface-programmable RF enclosures, without volumetric meshing of the canonical model or source-specific refactorization for a fixed sheet.
Accurate indoor path loss prediction requires reliable performance not only at new receiver locations, but also when the transmitter itself is not part of the training data. This work evaluates empirical, large-scale (LS), and residual random forest (RF) models under a hierarchical multi-transmitter framework comprising known-transmitter spatial generalization, zero-shot transmitter transfer, TX–region double holdout, and few-shot TX–frequency calibration. Measurements from four transmitter locations at 0.7, 3.5, and 7 GHz yielded 2,536 quantifiable observations. With represented transmitters, the basic RF with absolute coordinates achieved the lowest macro RMSE of 2.993 dB, compared with 4.140 dB for LS. Under the stricter zero-shot TX–region holdout, LS obtained the lowest RMSE at 7.585 dB. Few-shot calibration produced the largest recovery: at k=8 positions per frequency, LS improved to 4.757 dB and the geometry-aware RF with absolute coordinates reached 4.483 dB. Its paired hierarchical difference relative to LS was -0.274 dB with a 95% confidence interval of [-0.674, 0.096] dB, so the advantage was not conclusive. These results indicate that TX-specific calibration can reduce transfer error more than model-family changes alone.
