This letter proposes an antenna affiliation switching and received signal selection in uplink for a distributed antenna systems (DAS). The proposed scheme is applied to a macrocell system in which each macrocell consists of multiple picocells. Conventional approach improves throughput by allowing antennas located at the boundary of a macrocell to switch their affiliation to an adjacent macrocell. However, no interference suppression by selecting received signals has been investigated. Therefore, this letter introduces a DAS system that controls received signal selection in addition to antenna affiliation switching in a DAS. Markov chain Monte Carlo methods and ε-greedy algorithms are applied for antenna affiliation switching and received signal selection. Numerical results obtained through computer simulation show that the proposed scheme improves the throughput by at most 0.30 bps/Hz/cell. Furthermore, the proposed scheme maintains a comparable fairness index compared to the conditional scheme.

Doherty power amplifiers (DPAs) are widely utilized in modern communication systems for their high efficiency, but they exhibit severe nonlinearities and memory effects when driven into saturation. Among various linearization techniques, digital predistortion (DPD) has become the mainstream solution to compensate for these impairments. However, traditional Generalized Memory Polynomial (GMP) models struggle to achieve high modeling accuracy for such highly nonlinear DPAs. To address this limitation, this paper proposes a novel DPD method based on joint magnitude and phase segmentation. By partitioning the complex signal plane into a polar grid, the proposed approach achieves a more precise fitting of the DPA’s strong nonlinear behaviors and memory effects. Moreover, a high-throughput hardware actuator tailored for the proposed method is designed based on a multi-branch Look-Up Table architecture and deployed on a Xilinx Zynq UltraScale+ MPSoC platform. Measurements on a 50W DPA driven by a 20MHz OFDM signal show that the method reduces the Normalized Mean Square Error to -35.07dB and the Adjacent Channel Power Ratio to -46.16dBc (lower) and -44.56dBc (upper), outperforming the conventional GMP baseline.

This letter explores the Tsetlin Machine (TM) as a universal, reconfigurable logic-centric framework for learnable channel decoding on edge devices. Using a Hamming(7, 4) code case study on an ESP32 microcontroller, we demonstrate a TM-based decoder capable of discovering decoding rules from data in just 24 seconds. To ensure robust convergence under hardware constraints, the on-device model doubles the scale of the offline model, yet maintains an ultra-low decoding latency of 37μs, translating to a real-time throughput of over 100kbps. Bit error rate simulations confirm that the TM decoder matches the performance of conventional hard-decision decoder. These findings highlight the inherent hardware-friendliness and scalability of TMs, offering a promising path for reconfigurable, self-learning decoders on resource-constrained platforms.

Although the use of carbon fiber reinforced plastic (CFRP) in automotive components has been expanding, its electrical properties in high-frequency band have not been fully investigated. In this letter, we constructed a microstrip structure using CFRP plate to examine electrical properties of the CFRP and identified its effective conductivity over a wide frequency range, including the GHz band. One advantage of this method is that the direction of the current along CFRP sample in the measurement system is the same as the direction of the return path of the common-mode current flowing through automotive components that adopt CFRP. In the letter, we first measured transmission characteristics of a microstrip structure with a CFRP plate as the signal line. By comparing the transmission loss with full-wave simulation, we were able to identify the effective conductivity of the CFRP is approximately 10,000S/m. Furthermore, we confirmed that applying CFRP as the return plane did not affect common-mode noise compared to the conventional metal planes.

Line-of-sight optical wireless communication has difficulty in consistently aligning a photodiode receiver with an LED transmitter. Currently, however, to overcome this drawback, a camera can accurately capture LED position, and the photodiode can be aligned with the LED. In this study, narrow field of view (FOV) photodiode is individually aligned with 5-wavelength LEDs including infrared LEDs. Space and wavelength division multiplexing (SDM/WDM) is implemented to achieve a higher transmission rate. Photodiode receivers with a narrow FOV of 1.57° achieved 5Mbit/s SDM/WDM transmission (2 SDM × 5 WDM × 0.5Mbit/s) based on the camera’s LED position detection.

The Nakagami-n (Rician) fading model is essential for characterizing line-of-sight (LOS) wireless channels in 5G and 6G networks. While the Nakagami-n model underpins LOS channel modeling in next-generation networks, the complexity of the modified Bessel function often necessitates the use of the Small Limit Argument Approximation (SLAA). However, the validity domain of this approximation has not been rigorously defined. This study quantifies the SLAA’s accuracy limits, specifically examining the interplay between fading severity (K), SNR (ρ), and diversity order (m). Our analysis reveals that the SLAA maintains structural fidelity (Root Mean Square Error (RMSE) <10^-3) only under near-Rayleigh conditions (K<0.04). Furthermore, we observe that approximation error exhibits a linear dependence on m, despite a marginal reduction at higher SNRs. Consequently, the SLAA is unsuitable for strong Rician (K≥1) or high-diversity systems, limiting its utility primarily to asymptotic analyses in weak LOS channels. These findings establish critical design guidelines for the safe application of SLAA in next-generation wireless system analysis.

The functional disaggregation of optical transport equipment is underway to create optical transport networks that offer faster time-to-market and optimized performance. In parallel, multi-vendor environments are being increasingly adopted by network operators. Even in the resulting disaggregated environments, network control should remain unified. To satisfy this requirement, we are developing a controller that incorporates an adapter function to absorb differences in control commands across multi-vendor devices. The adapter internally maintains control scripts written in an intermediate language and abstracts vendor-specific commands to hide inter-vendor differences. This paper proposes a method in which a large language model (LLM) assists in generating control scripts written in a domain-specific language (DSL). Compared with baseline approaches, the proposed method reduced the number of syntax/formatting errors in control scripts by 50%. Furthermore, key technical findings extracted from an evaluation and analysis are summarized, and prospects for further advancement of the proposed method are discussed.

We demonstrate 86.5-Tbps (801.5 Gbps × 108 channels) wavelength division multiplexing (WDM) transmission using the L- and U-bands over a 10.8-THz bandwidth across 70.4 km of field-deployed dispersion-shifted fiber (DSF). The key enabling technology is the wavelength-band conversion using periodically poled lithium niobate (PPLN) waveguides, which extends the usable bandwidth into the U-band where higher-chromatic dispersion reduces Kerr nonlinear impairments. Although U-band operation faces increased fiber loss, which is slightly exacerbated by bending, mechanical stress, and splicing in the field environment, these penalties are mitigated by the inter-channel stimulated Raman scattering (ISRS) gain derived from co-propagating L-band channels. The experimental results confirm the feasibility of bandwidth extension to the U-band and high-data-rate transmission over legacy DSF infrastructure.