【作者】 刘赟；

【导师】 石艳玲；

【作者基本信息】 华东师范大学 ， 微电子学与固体电子学， 2005， 硕士

【摘要】 近年来,硅基片上螺旋电感作为射频芯片的重要器件成为了国内外学术界和业界的研究热点。片上电感可应用于移动通信系统的多个模块中,是实现单片硅基射频集成电路（Si—RFIC）的关键。与传统的分立电感相比,片上电感具有噪声小、功耗低、成本低和易集成的优点。本文主要研究了的硅基片上电感的电感值算法和品质因子Q的优化,设计实现了新颖的线宽、间距变化电感结构并自主提出了片上电感的分圈迭代算法。 在广泛的文献调研基础上,本文用电磁场理论详细分析了片上电感的损耗机理。在用电磁场仿真软件HFSS8.0对不同形状和结构参数的电感作了大量分类模拟讨论后,本文提出了线宽、间距渐变的电感结构以提高电感Q值。这种基于电感结构的Q值优化法相对于其它Q值优化方法更易实现,且不增加任何工艺成本。对所设计的优化电感样品制备测试后,测试结果为优化电感L=5.8nH,f_SR=5.1GHz,其Q_max=11.5比相同外径未优化电感的Q提高了6.5%,其频率特性与未优化电感基本相同。且在2GHz频率范围内,该电感的Q值特性与小外径的未优化电感相同。这表明采用本文提出的优化结构可以在保持频率特性不变的情况下,获得更高Q值、更大电感值的片上电感。 为了对上述优化电感进行计算,本文结合了Greenhouse分段叠代和Jenei整体平均两种经典算法,自主提出了分圈迭代平均算法,实现了线宽间距渐变结构的电感值计算以及电感值和版图参数的互逆运算。该算法基于电感版图参数并结合具体工艺参数,可计算多边形电感和不规则结构的电感值,同时可以将电感值转换为相应的版图参数。测试结果表明该算法不仅运算效率高,精度高,相对误差在9.4%以内,而且降低了片上电感的设计难度从而缩短了电感设计周期。 在此基础上,本文研究了片上电感在集成LC电路中的应用,用无源低通滤波器的特性来表征电感的实际性能。采用网络综合法设计了巴特沃思LC低通滤波器（LPF）,并用分圈迭代算法计算了电感元件的版图参数。利用实际的工艺条件进行多次流片后,成功制备了四阶LC—LPF。所制备的截止频率为5GHz的四阶LC—LPF特性曲线与模拟结果大致相同,50MHz时插入损耗为-0.65dB,半功率点-3dB在4.7GHz,偏离了理论值0.3GHz。这表明用分圈迭代算法所设计的电感版图参数是正确的,可以在电路设计中实际应用。 总之,实验结果显示本文优化电感结构的设计和分圈迭代算法是合理准确的,工艺实现是可行的,说明了此种结构和算法应用于射频集成电路的可行性,对片上电感的设计起到了一定的理论指导作用。更多还原

【Abstract】 Recently, silicon-based on-chip spiral inductor which can be widely used in wireless telecommunication has received more and more international attention as it is a key element to realize monolithic silicon-radio frequency integrated circuit (Si-RFIC). There are many aspects to research on-chip inductor and the research in this paper is on the computation of inductance and optimization of quality factor (Q) of inductor.Based on referring to many papers, we analyze in detail the energy loss mechanism of the inductor by electromagnetic theory and discuss several proposed approaches to optimizing the Q in substrate perspective. To optimize the Q in structure perspective, we put forward a novel inductor with variable width and space after a large amount of simulation by electromagnetic solver HFSS8.0. The proposed optimizing method is relatively easy to implement without any additional process cost. And based on Greenhouse algorithm, we compute the inductance of the novel inductor by physics -based approximation algorithm which calculates the inductance by summing every turn of the inductor and realizes the conversion between electrical parameter and geometrical parameters. The algorithm is efficient and scalable and also provides good accuracy for conventional inductor with errors of about 7%. After the optimization and computation of the on-chip inductor, we design a Butterworth low-pass filter (LPF) to explore the integration performance of the inductor.After a lot of experiments, monolithic on-chip inductors and LC filters are successfully fabricated on silicon with high resistivity in CMOS process. Measurement results indicate that the novel inductor with Q improved by 6.5% performs the same frequency characteristics as the conventional inductor. In addition, the error between the inductance calculated by the proposed physics-based approximation algorithm and measured inductance is within 9.4%, which shows the physics-based approximation algorithm is enough accurate for design. Moreover, the fourth-pole LPF cuts off at 4.7GHz which is close to the designed value at 5GHz.In conclusion, all the experiment results show that the proposed optimizing structure and physics-based approximation algorithm are reasonable and IC designers will benefit from them.更多还原