基于遺傳算法的模擬電路優(yōu)化設(shè)計(jì)方法研究?jī)?yōu)秀畢業(yè)論文.pdf

工學(xué)博士學(xué)位論文 基于遺傳算法的模擬電路優(yōu)化設(shè)計(jì) 方法研究 于健海 哈爾濱工業(yè)大學(xué) 2010 年 4 月 國(guó)內(nèi)圖書(shū)分類號(hào)：tn401 學(xué)校代碼：10213 國(guó)際圖書(shū)分類號(hào)：621.3.049.774 密級(jí)：公開(kāi) 工學(xué)博士學(xué)位論文 基于遺傳算法的模擬電路優(yōu)化設(shè)計(jì) 方法研究 博 士 研 究 生：于健海 導(dǎo)師：毛志剛 教授 申請(qǐng)學(xué)位：工學(xué)博士 學(xué) 科 、 專 業(yè)：微電子學(xué)與固體電子學(xué) 所在單位：微電子科學(xué)與技術(shù)系 答辯日期：2010 年 4 月 授 予 學(xué) 位 單 位：哈爾濱工業(yè)大學(xué) classified index: tn401 u.d.c.: 621.3.049.774 dissertation for the doctoral degree in engineering research on the design method of optimization for analog circuits based on genetic algorithm candidate: yu jianhai supervisor: prof. mao zhigang academic degree applied for: doctor of engineering speciality: microelectronics and solid-state electronics affiliation: dept. of microelectronics science and technology date of defence: april, 2010 degree-conferring-institution: harbin institute of technology 摘 要 - i - 摘 要 作為可進(jìn)化硬件（evolvable hardware，ehw）的主要分支，電路進(jìn)化 設(shè)計(jì)已經(jīng)成為國(guó)際性的研究熱點(diǎn)。對(duì)于數(shù)字電路而言人們可以利用基于硬件 描述語(yǔ)言的邏輯綜合工具對(duì)其進(jìn)行快速可靠的綜合，可模擬電路仍然需要專 家進(jìn)行手工設(shè)計(jì)完成。所以對(duì)模擬電路進(jìn)行自動(dòng)化設(shè)計(jì)已成為電路設(shè)計(jì)中迫 切需要解決的問(wèn)題。而模擬電路參數(shù)優(yōu)化設(shè)計(jì)是現(xiàn)階段模擬電路設(shè)計(jì)的主要 難題和更現(xiàn)實(shí)的任務(wù)。 本文在系統(tǒng)地研究了模擬電路進(jìn)化設(shè)計(jì)的基本理論、關(guān)鍵技術(shù)和研究現(xiàn) 狀的基礎(chǔ)上，探討了優(yōu)化理論、優(yōu)化算法。從阻容元件組成的電路入手，并 擴(kuò)展到較為復(fù)雜的 cmos 模擬電路，對(duì)模擬電路的進(jìn)化設(shè)計(jì)、參數(shù)優(yōu)化從 理論和實(shí)驗(yàn)進(jìn)行了探索。 通過(guò)對(duì)遺傳算法進(jìn)行了深入地分析，并與其它優(yōu)化算法、優(yōu)化技術(shù)進(jìn)行 了比較，確立了遺傳算法為本論文研究方法的主要解決手段。在考慮了種群 多樣性的基礎(chǔ)上，對(duì)遺傳參數(shù)進(jìn)行了指數(shù)型調(diào)整。通過(guò)對(duì)數(shù)學(xué)函數(shù)的仿真， 驗(yàn)證了改進(jìn)后的遺傳算法可以解決全局優(yōu)化問(wèn)題，加快收斂速度。 通過(guò)對(duì)阻容濾波器單元電路進(jìn)行參數(shù)調(diào)整，驗(yàn)證了遺傳算法對(duì)模擬電路 優(yōu)化設(shè)計(jì)的有效性。通過(guò)對(duì)其結(jié)構(gòu)進(jìn)化研究，說(shuō)明了進(jìn)化后生成的結(jié)構(gòu)和經(jīng) 典結(jié)構(gòu)的相似性。分析了模擬電路結(jié)構(gòu)調(diào)整固有的缺陷和對(duì)電路進(jìn)化設(shè)計(jì)帶 來(lái)的困難，結(jié)合模擬電路發(fā)展方向和實(shí)際設(shè)計(jì)中的難點(diǎn)，確立了本文采用結(jié) 構(gòu)固定優(yōu)化電路參數(shù)的主要研究方法。 基于上述理論的研究，對(duì)運(yùn)算放大器進(jìn)行了優(yōu)化設(shè)計(jì)。針對(duì)具體電路的 進(jìn)化效果，對(duì)遺傳參數(shù)進(jìn)行了指數(shù)型自適應(yīng)調(diào)整；把模擬電路手工設(shè)計(jì)經(jīng)驗(yàn) 和改進(jìn)的遺傳算法相結(jié)合，通過(guò)經(jīng)驗(yàn)約束了編碼方案的搜索空間，簡(jiǎn)化了編 碼長(zhǎng)度；針對(duì) cmos 運(yùn)算放大器電路中的性能指標(biāo)繁多且相互制約的特 點(diǎn)，對(duì)適應(yīng)度函數(shù)進(jìn)行了分段線形調(diào)整，根據(jù)不同性能指標(biāo)的要求，在相同 結(jié)構(gòu)下，完成了不同用途的運(yùn)算放大器的優(yōu)化設(shè)計(jì)?；?hspice 仿真結(jié) 果，實(shí)現(xiàn)了對(duì) cmos 模擬運(yùn)算放大器參數(shù)優(yōu)化設(shè)計(jì)。給出了優(yōu)化后運(yùn)算放 大器的性能指標(biāo)、參數(shù)和仿真結(jié)果，并與其他方法優(yōu)化出的相同結(jié)構(gòu)運(yùn)算放 大器進(jìn)行了比較，說(shuō)明了本方法的有效性和優(yōu)越性。 針對(duì)帶隙基準(zhǔn)源電路，根據(jù)各性能指標(biāo)要求的差異，運(yùn)用了以次要目標(biāo) 作為邊界約束條件的方法，對(duì)其進(jìn)行了優(yōu)化設(shè)計(jì)。根據(jù)電路優(yōu)化的效果，在 哈爾濱工業(yè)大學(xué)工學(xué)博士學(xué)位論文 - ii - 運(yùn)用改進(jìn)的遺傳算法基礎(chǔ)上，采用了“先粗調(diào)，再細(xì)調(diào)”的二次優(yōu)化方法， 并對(duì)優(yōu)化效果進(jìn)行了比較。以溫度系數(shù)為主要目標(biāo)，完成了曲率校正帶隙基 準(zhǔn)源的優(yōu)化設(shè)計(jì)，給出了優(yōu)化后的參數(shù)，仿真結(jié)果，并與業(yè)界其它的曲率校 正帶隙基準(zhǔn)源進(jìn)行比較，驗(yàn)證了該方法的有效性和優(yōu)越性。 關(guān)鍵詞 遺傳算法；適應(yīng)度函數(shù)；模擬電路進(jìn)化設(shè)計(jì)；可進(jìn)化硬件；帶隙基 準(zhǔn) abstract - iii - abstract as one of the most important embranchment of evolvable hardware (ehw), evolutionary design of circuits has been the hotspot of worldwide research. logic synthesis tools based on hardware describe language (hdl) can be used to perform fast reliable synthesis for digital circuit, while as for analog circuit it still need analog experts to complete the design manually. so the automatic design of analog circuit has been the impendent problem to be solved in circuit design. while the optimization of analog circuit parameters is the difficult problem and more practical task at present. optimizing theory and optimizing algorithm is discussed in this thesis after studying the basic theory, key technique and research status in evolving design of analog circuit systematically. from circuit composed of resistors and capacitors to cmos circuit exploration has been processed in evolving design and parameters optimization of analog circuit through theory and experiment. the genetic algorithm (ga) has been analysed deeply and is compared with other optimizing algorithm in theory, and ga is introduced as the feasible means to solve the multi-object design in this thesis. it is proved that only by adaptively adjusting the genetic parameters of ga, the convergence problem can be solved preferably and also the convergence speed can be improved through the simulation of mathematical function. through adjusting the parameters and evolving the structures of filter composed of passive componets, the validity of optimization of analog circuit by genetic algorithm is proved. through the study of the improvement of structure, the similarity of the classical structure and the evolved structure is introduced. analyzing the difficulty of evolutionary design of circuits which is brought by the inherent limitation of the structure adjusting of analog circuit and combining the direction of analog circuits development with the difficulty of actual design of analog circuit, the researching way of optimizing the parameters of analog circuit based on the fixed structure is established. based on the above research, one operational amplifier is optimized. aiming at evolving effect of corresponding circuit, the genetic parameters are adjusted 哈爾濱工業(yè)大學(xué)工學(xué)博士學(xué)位論文 - iv - adaptively. the method which combines the manual experience of analog circuit with improved ga reduces the coding length with the help of design experience which restricts the searching space of coding method. aiming at variety and restriction of the parameters in cmos analog operational amplifier, different op-amp for different purpose could be optimized in one fixed structure according different performance specification. basing on simulation results of hspice, the optimization of cmos analog operational amplifier is achieved. the parameters and the simulation results of the optimized op-amp are presented, and the validity and advantage is proved after comparing with other op-amp which is optimized by other method in industry. as for bandgap reference, according to the different specifications of diverse conditions, the idea which looks on secondary targets as the boundary restrictions is proposed, and the problem about multi-objective normalization is solved. the secondary optimizing method about “coarse adjusting initially, meticulous adjusting successively” is proposed on the condition that improved genetic algorithm is used. the optimizing design about a piecewise nonlinear curvature corrected bandgap reference is realized aiming at the temperature coefficient and the optimized parameters are gived. the simulation results indicate the advantage and validity of the method comparing with other bandgap references which have reached the leading standard of industry. keywords genetic algorithm, fitness function, evolvable hardware, optimizing design of analog circuit, bandgap reference 目 錄 - v - 目 錄 摘 要. i abstract. iii 第 1 章 緒論. 1 1.1 課題的背景及其研究的意義. 1 1.1.1 可進(jìn)化硬件的簡(jiǎn)介 . 1 1.1.2 ehw的研究思路 . 2 1.1.3 電路進(jìn)化設(shè)計(jì) . 5 1.1.4 制約模擬集成電路自動(dòng)化設(shè)計(jì)的難點(diǎn). 5 1.2 課題相關(guān)技術(shù)的國(guó)內(nèi)外研究現(xiàn)狀. 7 1.2.1 模擬電路綜合工具 . 7 1.2.2 模擬電路進(jìn)化設(shè)計(jì)技術(shù) . 9 1.2.3 存在的不足和待深入研究的內(nèi)容. 12 1.3 最優(yōu)化理論. 13 1.3.1 最優(yōu)化問(wèn)題數(shù)學(xué)表達(dá) . 14 1.3.2 最優(yōu)化技術(shù)分類 . 15 1.3.3 最優(yōu)化算法的分類 . 17 1.4 課題的主要研究?jī)?nèi)容和論文結(jié)構(gòu). 21 第 2 章 遺傳算法的分析與改進(jìn). 23 2.1 基本遺傳算法. 23 2.1.1 基本遺傳算法的實(shí)現(xiàn) . 23 2.1.2 遺傳算法的特點(diǎn) . 24 2.2 遺傳算法用于求解多目標(biāo)問(wèn)題. 25 2.3 對(duì)遺傳算法的自適應(yīng)改進(jìn) . 27 2.3.1 基本算法改進(jìn)的必要性 . 27 2.3.2 改進(jìn)算法的設(shè)計(jì) . 28 2.4 改進(jìn)的遺傳算法與標(biāo)準(zhǔn)遺傳算法的性能的比較 . 37 2.5 本章小結(jié). 44 第 3 章 基于遺傳算法的阻容濾波器的進(jìn)化設(shè)計(jì) . 45 3.1 濾波器的分類及其主要性能指標(biāo). 45 3.1.1 濾波器的分類 . 45 哈爾濱工業(yè)大學(xué)工學(xué)博士學(xué)位論文 - - vi 3.1.2 無(wú)源濾波器的主要性能指標(biāo). 45 3.2 濾波器的參數(shù)優(yōu)化設(shè)計(jì)及算法比較. 46 3.2.1 濾波器的選擇 . 46 3.2.2 hspice簡(jiǎn)介及在本文中的應(yīng)用 . 47 3.2.3 運(yùn)用遺傳算法對(duì)阻容濾波器進(jìn)行參數(shù)調(diào)整步驟 . 48 3.2.4 編碼方案. 49 3.2.5 適應(yīng)度函數(shù)的編寫(xiě) . 49 3.2.6 遺傳參數(shù)的線性調(diào)整 . 51 3.2.7 實(shí)驗(yàn)結(jié)果與討論 . 51 3.3 基于自適應(yīng)遺傳算法進(jìn)化無(wú)源濾波器. 54 3.3.1 編碼方案. 54 3.3.2 適應(yīng)度函數(shù)的編寫(xiě) . 55 3.3.3 遺傳參數(shù)的調(diào)整 . 57 3.3.4 實(shí)驗(yàn)結(jié)果和討論 . 58 3.4 本章小結(jié). 63 第 4 章 基于自適應(yīng)遺傳算法的cmos運(yùn)算放大器參數(shù)優(yōu)化 . 64 4.1 運(yùn)算放大器. 64 4.1.1 運(yùn)算放大器簡(jiǎn)介 . 64 4.1.2 運(yùn)算放大器的設(shè)計(jì)及主要性能指標(biāo). 66 4.1.3 運(yùn)算放大器優(yōu)化設(shè)計(jì)的發(fā)展現(xiàn)狀. 67 4.2 針對(duì)主要指標(biāo)運(yùn)算放大器參數(shù)優(yōu)化. 67 4.2.1 編碼的設(shè)計(jì)方案 . 68 4.2.2 適應(yīng)度函數(shù)的編寫(xiě) . 69 4.2.3 試驗(yàn)結(jié)果及其討論 . 70 4.3 基于自適應(yīng)遺傳算法的cmos放大器的參數(shù)優(yōu)化方法 . 73 4.3.1 編碼的設(shè)計(jì)方案 . 73 4.3.2 適應(yīng)度評(píng)估方法 . 74 4.3.3 遺傳參數(shù)調(diào)整策略 . 76 4.3.4 實(shí)驗(yàn)與討論. 79 4.4 本章小結(jié). 85 第 5 章 基于自適應(yīng)遺傳算法的二階補(bǔ)償帶隙基準(zhǔn)源設(shè)計(jì) . 86 5.1 帶隙基準(zhǔn)源的基本原理、性能指標(biāo). 86 5.1.1 帶隙基準(zhǔn)源的基本原理 . 86 目 錄 - vii - 5.1.2 帶隙基準(zhǔn)源的主要性能指標(biāo). 87 5.2 分段曲率校正帶隙基準(zhǔn)源 . 88 5.2.1 原理. 88 5.2.2 電路實(shí)現(xiàn). 90 5.3 算法的實(shí)現(xiàn). 91 5.3.1 編碼設(shè)計(jì)方案 . 91 5.3.2 適應(yīng)度函數(shù)的編寫(xiě) . 93 5.3.3 遺傳參數(shù)的調(diào)整 . 94 5.4 實(shí)驗(yàn)與討論. 95 5.5 本章小結(jié). 103 結(jié) 論. 104 參考文獻(xiàn). 106 攻讀學(xué)位期間發(fā)表的學(xué)術(shù)論文 .117 哈爾濱工業(yè)大學(xué)博士學(xué)位論文原創(chuàng)性聲明.118 致 謝.119 個(gè)人簡(jiǎn)歷. 121 哈爾濱工業(yè)大學(xué)工學(xué)博士學(xué)位論文 - - viii content abstract(in chinese) . abstract(in english). chapter 1 introduction.1 1.1 background an purpose.1 1.1.1 brief introduction of ehw.1 1.1.2 the researching thought of ehw.2 1.1.3 circuit evolution design .5 1.1.4 difficulty of analog integrate circuit design automation.5 1.2 researches of domestic and overseas corelative topic on ehw.7 1.2.1 analog circuit synthesis tools.7 1.2.2 technic of evolution design on analog circuit.9 1.2.3 existing shortcomings and the research to study.12 1.3 optimization theory.13 1.3.1 math expression of optimization method .14 1.3.2 analysis of optimization technic .15 1.3.3 claasificationof optimization algorithms 17 1.4 contents and outline of this dissertation .21 chapter 2 comparison of the optimization algorithm and improvement of the genetic algorithm.23 2.1 simple genetic algorithm .23 2.1.1 realization of the algorithm.23 2.1.2 the characteristic of ga.24 2.2 solving the multi-objective optimization based on genetic algorithm.25 2.3 self-adjusting of genetic algorithm.27 2.3.1 necessity of the improvement of simple genetic algorithm.27 2.3.2 design of improved genetic algorithm .28 2.4 performance comparison between improved genetic algorithm and simple genetic algorithm.37 2.5 summary .44 content - ix - chapter 3 evolutional design of the filter composed of resistors and capacitors based on genetic algorithm.45 3.1 classification of the filter and its main performance specification .45 3.1.1 classification of the filter.45 3.1.2 main performance specification of passive filter .45 3.2 parameters optimization of the filter and comparison of the algorithm.46 3.2.1 selection of the filter.46 3.2.2 introduction of hspice and purpose in this dissertation .47 3.2.3 the process of parameter adjustment of the filter based on genetic algorithm .48 3.2.4 coding project.49 3.2.5 composing of the fitness function.49 3.2.6 linear adjustment of the genetic parameters.51 3.2.7 results and