PHY5 – mmWave and THz communication
Wednesday, 5 June 2024, 16:00-17:30, room Darwin Hall
Session Chair: Sofie Pollin (KU Leuven, BE)
Performance Assessment of Irregular Array Configurations for Beyond 100-GHz Multi-User MIMO Systems
Yigit Ertugrul (Imec, Belgium); Kamil Yavuz Kapusuz (Ghent University-Imec, Belgium); Claude Desset (IMEC, Belgium); Sofie Pollin (KU Leuven, Belgium)
A comparison of thinned and clustered array configurations is presented where the former is generated by deactivating the radiating elements permanently while the latter is generated by utilizing a collection of subarrays. The performance of the clustered and thinned arrays is evaluated in the context of multi-user MIMO communications beyond 100 GHz frequencies. An optimization problem is described to find the optimal tiling or cluster configurations, i.e., arrangements of unit elements where the tradeoff between spectral efficiency (SE) and sidelobe levels (SLLs) can be tuned. It is shown that clustered arrays achieve higher SE and SLLs compared to their thinned array counterparts in the context of multi-user MIMO.
Sub-Terahertz Channel Gain Prediction for Scheduling of Over-The-Air Deep Learning
Rodney Martinez Alonso, Cel Thys and Sofie Pollin (KU Leuven, Belgium)
Enabling artificial intelligence native end-to-end systems in ultra-wideband sub-terahertz spectrum faces several challenges. The particularly complex channel variations and nonlinear behavior of analog components of the transceivers are major obstacles to the over-the-air adaptation of these systems. In this paper, we investigate an edge-based bidirectional long-short-term memory neural network capable of predicting the channel gain variations in Non-Line-of-Sight conditions. We aim to enable end-to-end autoencoders with a predictive model for scheduling the training phase when the power is above the receiver sensitivity and there are no large fading variations. Otherwise, the training of the end-to-end system will likely fail. With only 16 BiLSTM cells our model is capable of inferring the channel gain variations with a worst-case root mean squared error lower than 0.0547 (i.e., 1.1% compared to the normalized channel gain range). Also, with lower computational complexity, our model decreased the propagation of the error compared to traditional recurrent neural networks and deep-learning-based forecasting models.
Real World Field Trial for RIS-Aided Commercial 5G mmWave Wireless Communication
Ahmad Shokair and Ayoub Mohammed Toubal (Greenerwave, France); Guillaume Grao and Thibaut Rolland (Orange, France); Youssef Nasser (Greenerwave, France); Dinh-Thuy Phan-Huy (Orange, France); Geoffroy Lerosey (Greenerwave, France)
The performance and the functionality of the Reconfigurable Intelligent Surface (RIS) are evaluated in a real-world environment at mmWave frequency range. The ability of the device to redirect beams from a 5G gNB to a user equipment in shadowed areas are studied. The effect of the RIS beamforming on the received signal quality and the achieved data rate without creating additional interference in the surrounding area is also investigated. The experimental results show that the RIS managed to successfully operate as a coverage extender and enhancer for the 5G network, maintaining an interference-free environment, while maintaining low power consumption by operating on a battery.
Dual-Polarized Sub-THz Channel Measurements in D-Band in an Industrial Environment
Alper Schultze, Mathis Schmieder, Ramez Askar and Michael Peter (Fraunhofer Institute for Telecommunications, Heinrich Hertz Institute, Germany); Wilhelm Keusgen (Technische Universität Berlin, Germany); Taro Eichler (Rohde & Schwarz, Germany)
In this paper, we introduce a novel time domain (TD) correlation based channel sounder that operates in the D-band (110 GHz to 170 GHz) with which we performed dual-polarized sub-THz channel measurements in an industrial environment. The channel measurements cover all four channel polarization components: two co-polar channels and two cross-polar channels. This was achieved by changing the antenna polarization both at the transmitter (TX) and receiver (RX). To evaluate the impact of the antenna type, two different antennas at the RX side were used: On the one hand, a WR-06 waveguide probe antenna and on the other hand, a WR-06 waveguide pyramidal horn antenna with a higher gain and a narrower half-power beam width (HPBW). The collected channel impulse responses (CIRs) of 8 measurement positions were evaluated with regards to the path loss and the cross-polarization ratio (CPR). The measurements report, that the cross-polarization isolation (CPI) in average is around 30 dB and the average CPR is around −24.4 dB. Further, the antenna’s HPBW has a strong impact on the CPR leading to deviations of around 4 dB.
On the Efficacy of Fingerprint-Based mmWave Beamforming in NLOS Environments: Experimental Validation
Mohaned Chraiti (Sabanci University, Turkey); Ali Ghrayeb (Texas A&M University at Qatar, Qatar)
Fingerprint-based millimeter-wave (mmWave) beamforming is attracting growing interest due to its efficacy in reducing beam search/alignment time, and hence decreasing the channel estimation overhead to a negligible rate. The method relies on a pre-existing dataset containing potential high-gain beam directions, with location acting as a feature (fingerprint), to identify candidate beams. The fingerprint-based mmWave Beamforming is the inverse process of localization, assuming that the User Equipment (UE) has access to its position estimate. Accordingly, the UE considers the beams (directions and widths) in the dataset at proximate locations as candidates. The results in existing works, however, are often based on abstract models (often, the two-ray model), simulation results (typically based rays tracing simulator), and, in many cases, the outdoor environment with high probable Line-Of-Sight (LOS) link. In an effort to understand the extent and potential of such a technique, we have carried out a real-world experiment in an indoor office environment with high Non-LOS (NLOS) probability. We have trained a neural network model that provides the candidates’ beams given a UE location. Although the results show an average beamforming gain of 17 dB, there is a considerable gap with respect to the highest possible beamforming gain obtained through exhaustive search.
