本論文使用穩懋所提供之 WIN 100-nm GaAs pHEMT PP10−20 以及WIN 120-nm GaN HEMT NP12−01 製程，設計毫米波之 Q 頻段可變增益放大器與 U/W 頻段功率放大器。 第一章緒論首先說明毫米波之功率放大器以及可變增益放大器，次者，本研究針對Q (33−50 GHz)， U (40−60 GHz) 與 W (75 −110 GHz) 頻段之重要性加以說明，最後闡述本研究之設計動機與預期規格。 第二章依據文獻理論並使用 WIN 120-nm GaN HEMT NP12−01 製程設計毫米波之 Q 頻段可變增益放大器。該電路採用了單級電流導引 (current-steering) 架構，小訊號量測結果在 40 GHz 的頻率以及控制偏壓為 11.8 V 時，增益 (gain)、輸入返回損耗 (input return loss， IRL)、輸出返回損耗 (output return loss， ORL) 分別為 4.5 dB， 13.9 dB， 7.5 dB。而可變增益放大器之重要參數量下，實現可調整增益範圍 (gain control range， ΔG) 為 11.9 dB 以及相位變化 (phase variation， Δφ) 小於 10.6∘。與其他文獻對比，此電路在Q 頻段取得中等水平，不過，相位部分仍須加強。與 GaN 製程相比，雖然文獻都是 VGLNA 的架構，但是與文獻 [1] 相比 ΔG 提升了 2 dB；與 CMOS 相比，此電路雖然相位變化比不上 CMOS，但是 在增益方面本次設計僅需單級電流導引架構即可比部分 CMOS 文獻提升約 4 dB 之 ΔG。 第三章將會以先前的設計經驗以及偵錯的結果使用穩懋所提供之 WIN 100-nm GaAs pHEMT PP10−20 製程並重新設計毫米波之 U 頻段功率放大器。該電路採用了單級 common source 架構，小訊號量測結果在 48 GHz 操作頻率時， 增益為 6.4 dB，而大訊號在中心頻率 48 GHz 下，實現大約 24 dBm 的飽和輸出功率 (saturated output power， Psat) 以及 22.4 dBm 的 1-dB 增益壓縮輸出 功率 (output 1-dB compression point， P1dB)，並保有 34.8% 的效率 (power-added efficiency， PAE) 。與其他文獻對比，此電路在 U 頻段實現1−dB下具有非常好的效率，與GaAs製程相比，由於功率放大器基本上都是由後級推動功率，所以可以比較出此電路 OP1dB 其實跟其他多級相近頻段之電路表現情況是相近的，且能在高功率下保持著與其他電路相近甚至大於 1∼10% 的效率。 第四章將嘗試設計在極高頻的情況下，使用穩懋所提供之 WIN 100-nm GaAs pHEMT PP10−20 製程設計毫米波之 W 頻段功率放大器。該電路採用了單級 common source 架構，小訊號量測結果在94 GHz 的操作頻率時，增益為 3.6 dB，而大訊號在中心頻率 94 GHz 下，實現 19 dBm 的 Psat 以及 18 dBm 的 P1dB，並同時在 A 類下保有 15% 的效率；在 92GHz 操作頻率時，大訊號量測結果 Psat、OP1dB、PAE 分別為 ∼ 20 dBm， 18.6 dBm， 25.2%。與其他文 獻相比，此電路在 W 頻段取得良好的性能，與 GaAs 製程相比，在 OP1dB 相差 ±3 dB 的情況下 PAE 高於其他相近頻段之電路大約 2 ∼8%；與 GaN 製程相比，雖然功率方面不如其製程，但本電路可以使用較小的面積，且能在該頻段具有高於某些晶片 2∼7% 的效率表現；與 CMOS 相比，此電路具有堪比 CMOS 小面積的優勢，且能具有比 CMOS 更高功率、效率的性能。至於 92GHz 的部分，雖然 輸出功率方面與 94GHz 相比接近，不過效率的部分卻是全部文獻中表現最好的，效率高達 25.2%。 綜合上述，本論文展示 GaAs 以及 GaN 製程於高頻功率放大器和可變增益放大器設計中的可行性與效能優勢，並針對電路架構、匹配網路、元件面積與耗能等限制進行優化，具體實現多款高效能功率放大器和可變增益放大器，而更多的結論總結以及最後分析將會在第五章進行討論。;This thesis presents the design of a millimeter-wave Q-band variable-gain amplifier (VGA) and U/W-band power amplifiers (PAs) based on the WIN 100-nm GaAs pHEMT PP10−20 and WIN 120-nm GaN HEMT NP12−01 processes provided by WIN Semiconductors. Chapter 1 introduces the fundamentals and applications of millimeter-wave power amplifiers and variable-gain amplifier. The significance of the Q (33−50 GHz)， U (40−60 GHz)， and W (75−110 GHz) frequency bands is then discussed， followed by the design motivations and target specifications of this research. Chapter 2 describes the design of a millimeter-wave Q-band VGA based on the WIN 120-nm GaN HEMT NP12−01 process and the theoretical considerations reported in previous studies. A single-stage current-steering architecture is employed to achieve gain control with a compact circuit configuration. According to the small-signal measurement results， at 40 GHz with a control bias voltage of 11.8V， the measured gain， input return loss (IRL)， and output return loss (ORL) are 4.5 dB， 13.9 dB， and 7.5 dB， respectively. In terms of the key performance metrics of the VGA， a gain control range (ΔG) of 11.9 dB and a phase variation (Δφ) of less than 10.6∘ are achieved. Relative to previously reported works， the proposed circuit demonstrates a moderate level of performance in the Q-band， while further improvement in phase stability remains desirable. Among the reported GaN based implementations， although most employ VGA or VGLNA architectures， the proposed design achieves a ΔG that is approximately 2 dB higher than that reported in [1]. Compared with CMOS-based implementations， the proposed design exhibits a larger phase variation; nevertheless， its gain control performance is competitive， achieving an approximately 4 dB improvement in ΔG over several CMOS designs while requiring only a single-stage current-steering architecture. Chapter 3 presents the redesign of a millimeter-wave U-band PA using the WIN 100-nm GaAs pHEMT PP10−20 process provided by WIN Semiconductors. The design incorporates the experience gained from the previous implementation and the results obtained during the debugging process. A single-stage common-source architecture is adopted to simplify the circuit topology while maintaining the required output power and efficiency. At the operating frequency of 48 GHz， the measured small-signal gain is 6.4 dB. Under large-signal excitation at the same center frequency， the PA achieves a saturated output power (Psat) of approximately 24 dBm and an output power at the 1-dB gain compression point (P1dB) of 22.4 dBm， together with a power-added efficiency (PAE) of 34.8%. The comparison with previously published works indicates that the proposed PA provides excellent efficiency at the 1-dB compression point in the U-band. Since the output power of a PA is generally dominated by the final stage in a cascaded configuration， the P1dB of the proposed single-stage design is comparable to those of reported multi-stage PAs operating in similar frequency ranges. Furthermore， the proposed circuit maintains a comparable or even 1 − 10% higher efficiency under high power operation， demonstrating the effectiveness of the single-stage architecture in achieving a favorable trade off between output power and efficiency. Chapter 4 investigates the design of a millimeter-wave W-band PA operating at an extremely high frequency. The WIN 100-nm GaAs pHEMT PP10−20 process provided by WIN Semiconductors is employed， and a single-stage common-source architecture is adopted to reduce circuit complexity and chip area while maintaining competitive performance at the W-band. At an operating frequency of 94 GHz， the measured small-signal gain is 3.6 dB. Under large-signal operation at the same frequency， the PA delivers a Psat of 19 dBm and a P1dB of 18 dBm， while maintaining a PAE of 15% under Class −A operation. At 92 GHz， the measured large-signal performance further improves， with a Psat of approximately 20 dBm， a P1dB of 18.6 dBm， and a PAE of 25.2%. The performance comparison with previously reported works demonstrates that the proposed PA achieves competitive performance in the W-band. Relative to other GaAs-based designs， the proposed circuit exhibits approximately 2 − 8% higher PAE when the difference in P1dB is within ± 3 dB. Although GaN-based PAs generally provide higher output power， the proposed GaAs PA offers the advantages of a smaller chip area while maintaining competitive efficiency at the W band. In addition， its PAE exceeds that of several reported implementations by approximately 2 − 7%. Compared with CMOS-based designs， the proposed circuit provides a chip-area advantage comparable to thatof CMOS while achieving higher output power and efficiency. At 92 GHz， although the measured output power remains comparable to that obtained at 94 GHz， a significant improvement in efficiency is observed. In fact， the measured PAE of 25.2% represents the highest efficiency among the compared reported works， highlighting the potential of the proposed GaAs PA for high-efficiency operation in the W-band. In summary， this thesis demonstrates the feasibility and performance advantages of GaAs and GaN technologies for the design of high frequency power amplifiers and variable-gain amplifiers. Through the optimization of circuit architectures， matching networks， chip area， and power consumption， several high-performance PAs and a VGA are successfully implemented and experimentally characterized across the Q-，U-，and W-bands. The overall conclusions， detailed performance analysis， and future research directions are presented in Chapter 5.
