可同時穿透與反射的可重構式智慧表面及毫米波多天線模組之波束碼簿設計

Abstract

在未來B5G/6G的無線通訊系統中，波束成型技術至關重要。本論文中提出了兩項關於波束控制的新穎技術，其一為可同時穿透與反射的可重構式智慧表面，其二為毫米波多天線模組之碼簿設計演算法。 首先是可同時穿透與反射的可重構式智慧表面，其目的為改善一般可重構式智慧反射面的顯著缺點，將信號之覆蓋範圍延伸至智慧表面後方，以提供全方面的使用者服務。本論文結合穿透陣列及反射陣列的設計概念，提出了可同時反射及穿透之陣列單元。透過在自由空間中的量測架設對陣列單元進行驗證，其在28 GHz下的反射損失和插入損失分別為0.43 dB和1.26 dB，與模擬結果大致吻合，經過驗證該陣列單元可達到同時反射及透射之一位元相位切換且可獨立調整。接著，我們提出一種直流偏壓走線的設計來控制調控相位的二極體，透過模擬得到此種走線方式在波束掃描時將造成反射波束約1 dB的增益下降，而其穿透波束增益則降低約3 dB。而在驗證其波束掃描效能時，受限於製作成本，本論文僅以金屬帶線的有無來代替二極體的兩種操作狀態，以實現基於一位元相位切換之穿透及反射波束掃描，我們實作出具有400個陣列單元可同時穿透與反射的智慧表面，並在微波暗室中進行量測。反射波束和穿透波束的模擬增益值分別為24.2 dBi和22.3 dBi，而量測結果分別為18.08 dBi及20.04 dBi，其模擬波束在±45°的掃描角度範圍對應的損耗分別為2.5 dB和2.9 dB，而量測結果則為0.6 dB和5.0 dB，前述結果證實了我們所提出的基於一位元相位切換可同時穿透與反射之可重構式智慧表面的可行性。 第二部分為毫米波多天線模組之波束碼簿設計演算法。在多天線行動裝置中通常需要設計波束碼簿來優化裝置收發訊號的覆蓋範圍，然而波束碼簿的設計複雜度相當高，無法直接以窮舉法得到最佳化的結果，因此需要一套有效率的碼簿設計演算法，而在現代行動裝置密集度及通訊需求大幅提高的情況下，省電成為一項重要的指標，本論文中透過改善前人的演算法來優化全鏈碼簿及各種子鏈碼簿，接著我們提出一個基於動態規劃的演算法來結合全鏈碼簿及子鏈碼簿以達到省電及覆蓋效能的最佳平衡。我們的演算法可基於給定的省電效率下，得到最少的效能損失，相反地，也可在基於給定效能損失的可接受範圍下達到最佳的省電效率。與第三方演算法的結果比較，在給定相同的省電效率下，所提出之演算法可減少10%-19%的效能損失，而給定相同可接受效能損失的情況下，所提出之演算法可提升5%-17%的省電效率。此外，我們將所提出之演算法推廣至可適用於多頻率點，在實驗結果可發現如我們在低頻段與高頻段各選三個頻率點後，可得在給定省電效率下較第三方演算法可減少8%-30%的效能損失，以及在可接受的效能損失情況下8%-30%省電效率提升，而此演算法也可針對任意所關注之效能定義去做修改並達到最佳化設計，對於未來的實際應用層面更具優勢。

In the future B5G/6G communication systems, beamforming techniques will play a critical role. In this thesis, we propose two novel techniques regarding future beamforming applications: a simultaneously transmitting and reflecting reconfigurable intelligent surface (STAR-RIS) and the beam codebook design for millimeter-wave (mmW) multi-antenna modules. First, the goal of STAR-RIS is to overcome the main limitation of conventional reconfigurable intelligent surfaces (RISs). STAR-RIS can extend the coverage of signals to its backside to provide a more comprehensive service from all directions. In this thesis, we combine the design concepts of transmitarray and reflectarray and propose a novel unit cell of STAR-RIS. With the validation of the unit cell by over-the-air measurement, the reflection loss and transmission loss are 0.43 dB and 1.26 dB, respectively, which agree with our simulation results. By the measurement, we also verify the ability to achieve independent one-bit reconfigurability for the reflection and transmission phases. Then, we design the DC bias routing to control the on/off states of PIN diodes to realize the one-bit phase shifts in the unit cell. In beam steering applications, this DC bias routing design causes a less-than-1-dB reduction in gain for the reflected beam and a 3-dB reduction in gain for the transmitted beam in simulation. Due to the relatively high fabrication and bill-of-materials (BoM) costs, the two operating states of the PIN diodes are replaced by the presence or absence of a short metallic strip connecting the soldering pads to verify the beam steering performance of a 20x20 STAR-RIS without using PIN diodes. In our simulations, the peak gains of the reflected beam and transmitted beam are 24.2 dBi and 22.3 dBi, respectively, and the corresponding scanning losses when the main beam scans to 45 degrees are 2.5 dB for the reflected beam and 2.9 dB for the transmitted beam. Meanwhile, in our measurements, the peak gains are 18.08 dBi and 20.04 dBi for the broadside reflected beam and transmitted beam, respectively. Also, the corresponding scanning losses when the beam scans to ±45° are 0.6 dB for the reflected beam and 5.0 dB for the transmitted beam. These results demonstrate the feasibility of the STAR-RIS based on the proposed one-bit phase-agile unit cell. The second part is the beam codebook design for millimeter-wave multi-antenna modules. In mobile devices equipped with multi-antenna modules, design of the beam codebook is essential to optimize the coverage gain of transmitted/received signals. However, the complexity of beam codebook design is extremely high so that exhaustive search for the optimal solution is unattainable. Therefore, an efficient and effective algorithm is needed to address this problem. Aside from the algorithm performance, power saving is another indicator for beam codebook design in the higher demand of communications and higher density of devices. In this thesis, we first improve the performance of previous algorithms to obtain the optimal solutions for full-chain and sub-chain codebooks. Then, we propose an algorithm based on dynamic programming to achieve a better trade-off between coverage performance and power saving by the composition of full-chain and sub-chain codebooks. This proposed algorithm can achieve the least performance degradation based on a given power-saving ratio. Conversely, it can also achieve the optimal power-saving ratio given an acceptable range of performance degradation. In comparison to the third-party algorithm, the proposed algorithm achieves about 10%-19% less performance degradation than the benchmark algorithm under the same power-saving ratio. On the other hand, it achieves about 5%-17% higher power-saving ratio under the same acceptable performance degradation. Moreover, the proposed algorithm is adaptable to scenarios involving multiple frequency points. In experiments, three frequency points are selected spanning low and high bands. Results demonstrate that the proposed algorithm achieves 8%-30% less performance degradation while meeting specified power-saving ratios, and it also achieves 8%-30% higher power-saving ratios within acceptable gain reduction thresholds compared to the third-party algorithm. Additionally, our proposed algorithm is capable of optimizing the beam codebooks under various constraints, which represents a significant advantage for future practical applications.