交叉耦合滤波器设计ppt课件.ppt
苏 涛西安电子科技大学，电子工程学院2008年春,现代微波电路和器件设计16、交叉耦合滤波器设计,Rapid Elliptic-Filter Design, 2002 Empowering Profitability, Ansoft Corp.,补充部分内容和讨论。,关注设计策略的过程、特点和问题。,Specifications to be met,fo = 400MHzBW* = 15MHzRipple = 0.08dBOut of band rejection 20dB* denotes ripple bandwidth,Design a 3D RF elliptic band-pass filter completely in software,Obtain circuit representation using known filter theory.,Link design parameters of 3D prototype filter to circuit parametersusing calibration projects representing only parts of 3D filter.,Simulate complete 3D filter in HFSSuntil specs are met.,Perform curve fitting in Ansoft Designer to obtaincorrections needed.,Prototype Low-Pass Elliptic Filter Response,Filter theory from low pass to bandpass,Low-pass to bandpassFo scaled to 400MHzBW adjusted to 15MHz,Bandpass filter response,Enhanced Circuit Model,Linear frequency dependence introduced in impedance inverters.New term small compared to constant term, but necessary for design in software.,What filter theory gives us,f1 = 400 MHzresonant freq. resonators 1&4f2 = 400 MHzresonant freq. resonators 2&3K12= 0.02894coupling constant 1-2 and 3-4K23= 0.02863coupling constant 2-3K14=-0.00942coupling constant 1-4QL= 29.69loaded Q, coupling to source & loadd12= 0d23= 0 frequency-INdependent couplingd14= 0 (for now),准椭圆函数和General Chebyshev函数综合,椭圆函数同样可以得到具有带外有限传输零点的响应，而且带内和带外都具有等波纹性质，但是左右对称。对于耦合谐振腔滤波器形式，可否采用椭圆函数综合？与General Chebyshev函数综合有何异同点？ Ansoft例子中直流响应非零的响应能够实现吗？合理吗？如何得到的耦合系数值？,3阶椭圆函数综合，经过带通变换后的形式为,4腔多耦合带通滤波器的增益函数为,逼近函数，3阶椭圆函数的增益函数分子和分母相差2阶 实现结构，4腔交叉耦合带通滤波器增益函数分子分母相差4阶，做修正,其中， 是3阶准椭圆函数,准椭圆函数的构造,显然的，修正后的准椭圆函数带内没有等波纹特性，需要做出等波纹修正，此处我们采用数值修正的方法。,取 导数为零的点，得到（1，1）内各点的最大 ，有,注：此处得到的是近似的等波纹特性。,3阶椭圆函数（虚线）与准椭圆函数（实线）带内曲线,3阶椭圆函数与准椭圆函数增益曲线,4阶椭圆函数修正得到准椭圆函数,4阶椭圆函数通过舍去带外零点修正得到准椭圆函数。此时带内最大值和最小值仍然为1和1，无需修正（但是，并不是每次波动都达到最大或最小值）。,4阶椭圆函数与准椭圆函数增益曲线,4阶椭圆函数（虚线）与准椭圆函数（实线）带内曲线,此时，低通原型直流增益非零，与Ansoft例子雷同。,CAD原理电路设计,Basic resonator,Cavity 2020200 mmMetal walls, air insideMetal cylinder R=5 mmCylinder stands on cavity floor, does not touch ceiling.Cylinder length is varied to tune the resonance.Irises in walls, not shown here, will provide coupling to other resonators.,Elliptic Filter, basic design,L1L2D,1,4,2,3,170-175mm,170-175mm,Disk distance0.6-0.7mm,Elliptic filter, basic design II,All iris heights 100mmW12 = W34,W14,W12W23,Calibration Projects,6 variables in 3D design.3 calibration projects used to relate geometry to circuit parameters.,确定2腔和3腔之间的耦合孔宽度,和腔内导体棒高度,确定1腔和4腔之间的耦合窗口宽度W14，耦合圆盘的直径D和腔内导体的高度L1,修正腔1和腔2之间耦合窗口的大小,1,2,3,Ansoft 滤波器设计实例,模型参数化,输出变量的设定,参数扫描和优化,1,1、建立 41*41*200mm 的滤波器腔体；2、中间十字型隔板，厚度1mm；3、耦合窗口大小立方体，高100mm，宽W23，厚度穿过隔板；4、隔板减去耦合窗口大小立方体；5、腔体减去隔板。,建立一个滤波器腔体的“内胆”（Ansoft默认外界面为PEC）,建立长度为L2的金属棒,在2腔和3腔的中心建立长度为L2的金属棒,设定工程类型、解和输出变量,1、设定工程类型为Eigen Mode；2、设定解3、设定输出变量f01和f02；4、设定输出变量f0和K23。,参数扫描,1、设定参数 W23 扫描，8mm2mm14mm；2、设定参数 L2 扫描，170mm5mm190mm。,参数优化,1、设定优化代价函数cost；2、设定优化。,PML的设定,有载Q值的计算,2,PML的设置,有载 Q 值计算,3,严格意义上讲，这两腔结构不同，不能应用对称结构的公式；但是，用该公式修正，比原有尺寸要好；,All,First filter eigen frequencies,Results from calibration projects used to model complete filter.First iteration not centered at 400MHz.,Corrected first filter,L12 and L23 adjusted to center response on 400MHz.Asymmetry in shoulder height due to frequency dependent coupling factors,Corrected first filter,F0 and BW very closeDo NOT see 3 equal ripples of 0.08dB!,Curve fitting,Curve fitting resultsOriginal targetsf1 = 399.93 MHz400 MHzf2 = 400.10 MHz400 MHzK12= 0.031220.02894K23= 0.02819 0.02863K14= -0.00930-0.00942QL= 28.4329.69d12= -0.0082d23=-0.0410 were zero in circuit d14= 0.0363,Ansoft Designer used to curve fit ideal prototype circuit to HFSS results.K12 too large W12 adjusted (causes F0 shift) L1, L2 adjusted,Second filter,Change the goals!,Curve fitting shows that we have d12=-0.0004d23=-0.0405d14= 0.0405Further improvement not possible with d12=d14=d23=0.New circuit parameters based on nonzero d-factors are needed.Desired points in theoretical filter characteristics and d-factors are known. Curve-fit used to find new goals for f1, f2, K12, K23, K14, and Q_L.,New TargetsOriginal targetsHFSS 2nd filterf1 = 399.85 MHz400 MHz400.11f2 = 400.14 MHz400 MHz400.16K12= 0.029810.028940.02924K23= 0.02876 0.028630.02837K14=-0.00864-0.00942-0.00942QL= 28.9729.6929.84based on d12=-0.0004, d23=-0.0405, d14=0.0405.,Filter response with d0,Circuit simulation results meet spec on F0, BW, ripple, and stop band.Asymmetry in stop band due to frequency dependent K-factors.Ripple is exactly 0.08dB and symmetric,Third filter,HFSS model adjusted to meet new goals.Ripple is off 0.01dB.BW is correct.,Fourth filter,Length of resonator changed slightly to make ripple symmetric.,First design saves time,ANSOFT,TRADITIONAL,TIME,Further designs save time AND money,ANSOFT,TRADITIONAL,MONEY SAVED,TIME SAVED,2.Design: 90 days 14 days + cost of software,76 days,1.Design: 45 days 7 days + cost of software,38 days,逼近函数和实现结构的数学表达式 等波纹函数 提取耦合矩阵 CAD设计策略 EM设计的修正和优化,
