EE 461, Microwave Engineering Notes George W. Hanson 2008 2 Contents 1 Electromagnetic Theory 1.1 Introduction to Microwave Engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.2 Maxwell’s Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.3 Time-Harmonic Fields . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1.4 1.5 Wave Equations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Plane Wave Propagation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 9 9 10 13 14 2 Transmission Line Theory 2.1 The Lumped-Element Circuit Model for a Transmission Line 2.2 Field Analysis of Transmission Lines . . . . . . . . . . . . . . 2.3 The Terminated Lossless Transmission Line . . . . . . . . . . 2.4 The Smith Chart . . . . . . . . . . . . . . . . . . . . . . . . . 2.5 The Quarter-Wave Transformer . . . . . . . . . . . . . . . . . 2.6 Generator and Load Mismatches . . . . . . . . . . . . . . . . 2.7 Lossy Transmission Lines . . . . . . . . . . . . . . . . . . . . 2.8 Transient Transmission Lines . . . . . . . . . . . . . . . . . . 2.8.1 Waveforms and Spectral Analysis . . . . . . . . . . . . 2.8.2 Integrated Circuits and Ground Bounce . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 18 21 22 30 39 40 41 42 51 53 3 Transmission Lines and Waveguides 3.1 General Solutions for TEM, TE, and TM Modes 3.2 Parallel Plate Waveguide . . . . . . . . . . . . . . 3.3 Rectangular Waveguide . . . . . . . . . . . . . . 3.4 Circular Waveguide . . . . . . . . . . . . . . . . . 3.5 Coaxial Line . . . . . . . . . . . . . . . . . . . . 3.6 Surface Waves on a Grounded Dielectric Slab . . 3.7 Stripline . . . . . . . . . . . . . . . . . . . . . . . 3.8 Microstrip . . . . . . . . . . . . . . . . . . . . . . 3.9 The Transverse Resonance Technique . . . . . . . 3.10 Wave Velocities and Dispersion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57 57 59 64 65 65 67 70 71 73 73 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 Microwave Network Analysis 79 4.1 Impedance and Equivalent Voltage and Currents . . . . . . . . . . . . . . . . . . . . . . . . . 79 5 Impedance Matching and Tuning 5.1 Matching with Lumped Elements (L Networks) 5.1.1 Lumped Elements . . . . . . . . . . . . 5.2 Single-Stub Tuning . . . . . . . . . . . . . . . . 5.3 Double-Stub Tuning . . . . . . . . . . . . . . . 5.4 The Quarter-Wave Transformer . . . . . . . . . 5.5 The Theory of Small Re‡ections . . . . . . . . 5.6 Binomial Multisection Matching Transformers . 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 . 86 . 88 . 88 . 99 . 105 . 107 . 109 4 CONTENTS 5.7 5.8 5.9 Chebyshev Multisection Matching Transformers . . . . . . . . . . . . . . . . . . . . . . . . . . 115 Tapered Lines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 Bode-Fano Criteria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 6 Power Dividers and Directional Couplers 131 7 Electromagnetic Compatibility and Interference (EMS/EMI) 141 8 Microwave Filters 147 9 Appendix 149 9.1 Transient transmission line current . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149 CONTENTS 5 Preliminary Material “From a long view of the history of mankind - seen from, say, ten thousand years from now - there can be little doubt that the most signi…cant event of the 19th century will be judged as Maxwell’s discovery of the laws of electrodynamics” Richard P. Feynman 6 CONTENTS What is “Di¤erent”About High Frequencies? The short, and perhaps obvious, answer is wavelength. For any given object, one usually considers a frequency to be a “high” frequency if the wavelength ( = c=f ) is small compared to the size of the object. This is complicated by the fact that an object doesn’t have one “size” (a sphere, for instance, may be an exception). So if one considers electromagnetic scattering from a raindrop, and scattering from a large building, you would characterize di¤erent frequencies as being high frequencies. In a broad sense, through, many people consider high frequencies to be something like f > 100 MHz ( = 3 m in air). This is because we live in a world of characteristic sizes commiserate with this wavelength. For example, people are typically 5-6 feet tall (1.66-2 meters), cars are typically 8-12 feet long (2.66-4 meters), suburban buildings are typically 10-30 feet tall (3.33-10 meters), etc. An important point to remember is that electromagnetics is the physics of electrical engineering, and that the governing equations that describe all classical macroscopic electromagnetic phenomena are Maxwell’s equations. These are very complicated equations, but they can be simpli…ed in special cases. One such case is in the event of “low” frequencies— one can derive the classical circuit equations (Ohm’s law, Kirchho¤’s laws, etc.) as special, simple cases. These simpli…ed equations are often applicable in circuit analysis since typical circuit dimensions are very small compared to typical wavelengths in common usage. For example, consider the following discrete (not integrated) circuit) L= 5 cm R1 + + - Vout Vs R2 + - With Vs = V0 cos !t, is Vout = Case 1 f = 20 kHz ! = c f - R2 R1 +R2 Vs ? = 3 1010 cm=s 20 103 =s = 1; 500; 000 cm. L = 3:3 10 6 1: The circuit length is very short compared to a wavelength— low frequency approximations are applicable, 2 and Vout = R1R+R Vs : 2 Case 2 f = 6 GHz ! = c f = 3 1010 cm=s 6 109 =s = 5 cm. L = 1: The circuit length is equal to a wavelength— low frequency approximations are not applicable. CONTENTS 7 Case 3 f = 300 GHz ! = c f = 3 1010 cm=s 300 109 =s = 0:1 cm. L = 50 1: The circuit length is long compared to a wavelength— low frequency approximations are not applicable. In particular, for Cases 2 and 3, The “voltage” on the line will be a function of position along the line. The circuit “seen”by the source presents a length-dependent input impedance which must be carefully matched for e¢ cient power transfer. Practically, the circuit will not work since it will radiate energy into space. Furthermore, the electrically long distances the signal must travel will result insigni…cant dissipation to the use of imperfect conductors. Dispersion will degrade the signal before it reaches the output. To make it clear that the important thing is wavelength, consider a circuit of length L = 16:66x10 Case 4 f = 6 GHz ! = c f = 3 1010 cm=s 6 109 =s 6 cm. = 5 cm. L = 3:3 10 6 1: The circuit length is very short compared to a wavelength (same as Case 1)— low frequency approximations 2 are applicable, and Vout = R1R+R Vs . 2 8 CONTENTS Brief Chronology of Early Wireless Communications Developments Magnetic e¤ects known for much of recorded history. 1755: Benjamin Franklin began investigations which would lead to qualitative and quantitative ideas about electrostatics. 1769: Dr. John Robison began experiments which would lead to the inverse-square law (Coulomb’s law) governing electrostatics. Similarly by Henry Cavendish in 1773. Neither investigator widely publicized their work. 1785: Charles Augustin de Coulomb demonstrated the law of electric force, which is now called Coulomb’s law. 1820: Hans Christian Oersted found that a current carrying wire (electricity) could produce magnetism. 1821: Jean-Baptiste Biot and Félix Savart quantify the e¤ect discovered by Oersted. 1827: Georg Simon Ohm formulates what is now called Ohm’s law 1831: Michael Faraday demonstrated that a time changing magnetic …eld could produce an electric current. 1845 (or 1854): Gustav Robert Kirchho¤ formulated Kirchho¤’s laws. 1873: James Clerk Maxwell published the …rst uni…ed theory of electricity and magnetism. 1886: Heinrich Hertz assembled a “radio” system–a spark applied to a transmitting antenna caused a spark to be produced at a receiving antenna, posited near the antenna. 1901: Guglielmo Marconi developed a system to send signals from England to Newfoundland (' 3000 km) using electromagnetic waves. 1903: Marconi began regular transatlantic message service between England and stations in Nova Scotia and Cape Cod. Chapter 1 Electromagnetic Theory 1.1 Introduction to Microwave Engineering Read 1.2 Maxwell’s Equations Maxwell’s Equations— Di¤erential Form Classical macroscopic electromagnetic phenomena are governed by a set of vector equations known collectively as Maxwell’s equations. Maxwell’s equations in di¤erential form are r D(r; t) = r B(r; t) = r E(r; t) = r H(r; t) = e (r; t); m (r; t); (1.1) @ B(r; t) @t Jm (r; t); @ D(r; t) + Je (r; t); @t E is the electric …eld intensity (V=m) 2 D is the electric ‡ux density (C=m ) 2 B is the magnetic ‡ux density (Wb=m ) H is the magnetic …eld intensity (A=m) 3 e is the electric charge density (C=m ) 2 Je (= J in text) is the electric current density (A=m ) 3 m is the magnetic charge density (Wb=m ) 2 Jm (= M in text) is the magnetic current density (V=m ) V stands for volts, C for coulombs, Wb for webers, A for amperes, and m for meters. The equations are known, respectively, as Gauss’ law, the magnetic-source law or magnetic Gauss’ law, Faraday’s law, and Ampère’s law. 9 10 CHAPTER 1. ELECTROMAGNETIC THEORY 1.3 Time-Harmonic Fields Often the …elds of interest vary harmonically (sinusoidally) with time. Because Maxwell’s equations are linear, and assuming linear constitutive equations (typical below optical range), then time-harmonic sources ; J will maintain time-harmonic …elds E; D; B; H. Assume E(r; t) = E0 (r) cos(!t + E ); (1.2) then, for instance, n E(r; t) = E0 (r) Re ej(!t+ j!t = Re E(r)e where E(r) E0 (r)ej E E) ; o = Re E0 (r)ej!t ej E (1.3) is a complex phasor. Repeating for the other …eld quantities of interest, B(r; t) D(r; t) H(r; t) J(r; t) (r; t) = = = = = B0 (r) cos(!t + B ) D0 (r) cos(!t + D ) H0 (r) cos(!t + H ) J0 (r) cos(!t + J ) ) 0 (r) cos(!t + (1.4) we obtain the complex phasors B(r) D(r) H(r) J(r) (r) B0 (r)ej B D0 (r)ej D H0 (r)ej H J0 (r)ej J j : 0 (r)e (1.5) To obtain time-harmonic Maxwell’s equations, consider, for instance, Faraday’s law, r E(r; t) = @ B(r; t) @t Jm (r; t): (1.6) This can be rewritten as Re [r E(r) + j!B(r) + Jm (r)] ej!t = 0; (1.7) which must be true for all t. – When !t = 0; Re fr – When !t = =2; Im fr E(r) + j!B(r) + Jm (r)g = 0: E(r) + j!B(r) + Jm (r)g = 0: – If both the real part and the imaginary part of a complex number are zero, then the number itself is zero, and thus r E(r) = j!B(r) Jm (r): (1.8) 1.3. TIME-HARMONIC FIELDS 11 Repeating for all Maxwell’s equations and the continuity equations, we get the time-harmonic forms r D(r) = e (r); r B(r) = m (r); r E(r) = j!B(r) Jm (r); r H(r) = j!D(r) + Je (r); r Je(m) (r) = j! e(m) (r); (1.9) where all quantities are time-harmonic phasors. Note that in the time-harmonic case we obtain the convenient correspondence Z @=@t $ j! ( ) dt $ (j!) (1.10) 1 : Time-domain quantities can be recovered from the phasor quantities as, for instance, E(r; t) = Re E(r)ej!t : (1.11) Constitutive Equations For linear isotropic media D(r; !) = e(r;!) E(r; !); B(r; !) = e(r; !) H(r; !); (1.12) where e, e are the permittivity and permeability, respectively, of the medium. e = er "0 , e = er 0 0 is the permittivity of free space (' 8:85 10 0 is the permeability of free space (' 4 10 12 7 F=m) H=m) F stands for farads and H for henrys. For dimensional analysis, C = A s = F V and Wb = V s = H A, where s stands for seconds. Note that c0 = p 1 0 "0 = speed of light in free space (vacuum) Both e and e may be complex. The real parts of e, e are associated with polarization (electric and magnetic). The imaginary parts of e, e are due to polarization (molecular) loss, i.e., dipole friction and associated time-lag, Ohm’s Law e = e0 e = e0 je00 e0 : (1.13) Another relationship that is often useful for lossy media is the point form of Ohm’s law, Je (r; !) = Jm (r; !) = e (r; !)E(r; !); m (r; !)H(r; !); (1.14) 12 CHAPTER 1. ELECTROMAGNETIC THEORY e m (ohms 1 =m ) is the electrical conductivity of the medium (ohms=m ) is the magnetic conductivity of the medium V = A ohms: Complex Constitutive Parameters In (either) transform domain it becomes particularly easy to separate applied quantities from induced e¤ects. In (??) the …eld quantities represent the total …elds at a point in space. Assume that an impressed current density Jie (r) 6= 0 maintains E; D; H; B 6= 0, which, in turn, results in Jce (r) = e E(r) 6= 0; (1.15) where Jce is an induced conduction current density. The total electric current is i Je (r) = Je (r) + Jce ; (1.16) and Faraday’s law becomes r H(r) = j!e " E(r) + Jie (r) + = j! e " De…ning a new complex permittivity as j ! e " " leads to (1.18) e H(r) = j!" E(r) + Jie (r); r (1.17) E(r) + Jie (r): e j ! e E(r) (1.19) where we have separated the induced e¤ ects from the applied source. i Repeating for Jm (r) = Jm (r)+ m H(r) i = Jm (r) + Jcm and noting that r Jie(m) + j! leads to e j ! Assuming e ", e are real, the imaginary parts of ", i e(m) m = 0; (1.20) : (1.21) account for conduction loss. In general, one often writes " = ("0 =( 0 j"00 ) ; j 00 ) ; (1.22) j"00r ) "0 ; j 00r ) 0 : (1.23) or, in terms of the relative permittivity, " = ("0r = ( 0r where the imaginary parts account for all loss mechanisms (conductive and molecular).The following table lists some typical material parameters for dielectric media. 1.4. WAVE EQUATIONS 13 Material air glass wood gypsum board dry brick dry concrete fresh water sea water snow ice moist ground dry ground copper "0r 1.0006 3.8–8 1.5–2.1 2.8 4 4–6 81 81 1.2–1.5 3.2 20–30 3–6 1 "00r (1/ohm-m) <0.003 @ 3 GHz <0.07 @ 3 GHz 0.046 @ 60 GHz 0.05–0.1 @ 4,3 GHz 0.1–0.3 @ 3,60 GHz <0.006 @ 3 GHz 0.0029 @ 3GHz 0.01–0.001 1–6 0.000001 0.03–0.003 0.005–0.00001 5:7 107 Maxwell’s equations become r (" E(r)) = r ( H(r)) = r r E(r) = i e (r); i m (r); j! H(r) H(r) = j!" E(r) + (1.24) Jim (r); Jie (r); Often the superscript i is omitted in (1.24)–the interpretation of J depends on ; ". For example, a fairly general form is r r r r D(r) = e (r); B(r) = m (r); E(r) = j!B(r) Jm (r); B(r) = j! D(r) + Je (r); (1.25) where B = H and D = "E. – If ; " account only for polarization e¤ects, i.e., if ; " are real-valued, or if ; " are complex-valued where the imaginary parts are associated with polarization loss (dipole friction), then Jm ; Je are total currents. – If ; " contain the conductivities, then Jm ; Je are impressed currents. 1.4 Wave Equations When ! 6= 0 electric and magnetic quantities are coupled, allowing for wave phenomena. Vector Wave and Vector Helmholtz Equations for Electric and Magnetic Fields The independent Maxwell’s equations r r E(r) = j! H(r) Jm (r); H(r) = j!" E(r) + Je (r); (1.26) 14 CHAPTER 1. ELECTROMAGNETIC THEORY represent six scalar equations in six unknowns. However, they are coupled vector partial di¤erential equations. We want to uncouple the equations and obtain one equation which may be solved, assuming a homogenous medium. Taking the curl of r E(r) and of r r r r r E(r) H(r) H(r) leads to ! 2 " E(r) = j! Je (r) 2 ! " H(r) = r j!"Jm (r) + r Jm (r); (1.27) Je (r): These are the vector wave equations for the …elds. Noting that r r A = r (r A) r2 A, we also have r2 E(r) + ! 2 " E(r) = j! Je (r) + r r2 H(r) + ! 2 " H(r) = j!"Jm (r) r r e ; " r m Je (r)+ : Jm (r)+ (1.28) These are known as vector Helmholtz equations. 1.5 Plane Wave Propagation We want to consider the simplest solution of source-free wave equations— these will represent travelling plane waves which, under typical conditions, model realistic electromagnetic waves. Time-Harmonic Plane Waves in Free Space We …rst consider what kind of waves can exist in source-free homogeneous space characterized by "; . At any point in space where sources are absent the electric …eld satis…es r2 + k 2 E (r) = 0: (1.29) Assume we want to …nd a wave travelling along z, independent of x; y, and polarized in the x-coordinate. Assume bE (z) E=x The wave equation (1.29) becomes @2 + k 2 E (z) = 0 @z 2 which has solution E (z) = E0+ e jkz + E0 e+jkz : Plugging into Faraday’s law leads to the magnetic …eld as where = p =", and so the pair E0+ e jkz b E0+ e E = x jkz b H=y b H = y E0+ e E0 e+jkz ; + E0 e+jkz ; jkz E0 e+jkz ; 1.5. PLANE WAVE PROPAGATION 15 form a wave travelling along z. E ? H ?b z and the wave is called a TEM (Transverse ElectroMagnetic) wave. Phase and Attenuation Constants of a Uniform Plane Wave Assume k=( j ): (1.30) Then b E0+ e E = x is called the phase constant. b E0+ e = x j( j )z j z e z + E0 e+j( j )z + E0 e+j z e z (1.31) : is called the attenuation constant. 2 is real-valued and " = "0 Setting k 2 = ( j ) = ! 2 ", where imaginary parts, leads to v u u1 (!) = ! "0 (!)t 2 v u u1 p (!) = ! "0 (!)t 2 p s "00 (!) s "00 (!) 2 1+ 2 "0 (!) ! +1 2 1+ j"00 , and equating real and 2 "0 (!) (1.32) ! 1 : Time-Domain Waves, Phase Velocity, and Wavelength for Uniform Plane Waves In the time-domain, E (r; t) = b E0+ e Re x b = x E0+ z e j z e z + E0 e+j z e cos (!t z) + E0 e z z ej!t (1.33) cos (!t + z) For the …rst term, as t increases, the argument of the cosine remains unchanged if z increases correspondingly— thus the wave travels along the +z-direction. Correspondingly, the second term travels along the z-direction. The phase velocity (actually speed since it is a scalar) of the wave is found from d (!t dt ! z) = d Const = 0 dt dz dt = 0 (1.34) leading to vp = dz ! = : dt (1.35) 16 CHAPTER 1. ELECTROMAGNETIC THEORY The wavelength of the wave is the distance between adjacent wavefronts that produce the same value of the cosine function. If z1 and z2 are points on adjacent wavefronts, z1 = z2 2 or = z1 z2 = 2 (1.36) : (1.37) b-direction is propagating in the b Example 1.1 Assume a plane wave with electric …eld oriented in the x zdirection through dry soil at 1 MHz. At this frequency typical material properties are " = 3"0 and = 10 5 (ohms 1 =m ). Therefore, " = e " 3 j ! = e j 10 2 106 3"0 j 10 2 106 "0 5 5 (1.38) j0:179) "0 = "0 "0 = (3 j"00 such that (!) = ! = (!) v u u1 0 "t 2 0:0363 v u p u1 "0 t = ! 2 = Since k = b z kz = b z( p ! "002 1 + 02 + 1 " r (1.39) ! r "002 1 + 02 " 1 0:00108: j ), the wave has the form E (r) = E0 e jk r and in the time-domain The phase velocity of the wave is b E0 e E (r) = x vp = ! = b E0 e =x z j z cos (!t 2 106 = 1:73 0:0363 e z z) : 108 m/s (1.40) (1.41) (1.42) and the wavelength is = 2 = 2 = 173:1m. 0:0363 In travelling 1 km the amplitude will decrease from E0 to E0 e 20 log (0:339) = z = 0:339E0 , or by 9:396 dB. (1.43) Chapter 2 Transmission Line Theory Transmission lines are used to transfer electrical signals (information) or electrical power from one point to another in an electrical system. Transmission lines take a wide variety of forms, from simple wire pairs and cables to more complicated integrated structures for high-frequency applications. Some common transmission lines are: (a) (c) (b) w conducting strip ground plane ε,µ b (d) w ground planes ε,µ b conducting strip (e) 17 18 CHAPTER 2. TRANSMISSION LINE THEORY y b ε,µ x a z (f) where (a) parallel wires (two-wire line) (b) coaxial cable (c) parallel plates (d) microstrip (e) stripline (f ) rectangular waveguide The table below lists some characteristics of a few common transmission lines and waveguides. Structure Coaxial line Stripline microstrip line rectangular waveguide Frequency Range (GHz) < 50 < 10 < 100 < 300 Impedance Range (ohms) 10 100 10 100 10 100 100 500 Power Rating medium medium low high Ease of Device Mounting medium medium good good Low Cost Production medium good good poor In this chapter we will study the analysis of general transmission lines, mostly without considering speci…c physical structures. In the next chapter we will examine some speci…c geometries of transmission lines and waveguides. 2.1 The Lumped-Element Circuit Model for a Transmission Line To illustrate the analysis of transmission lines and transmission-line resonators, consider the generic twoconductor1 TEM transmission line depicted in the …gure below, where Vs and Is represent distributed sources. i(z,t) Vs(z,t) + - + v(z,t) Is(z,t) 1 Generally the term “conductor”is used in transmission-line analysis, and, in fact, the lines are usually conductive. However, other transmission systems, including those only involving dielectrics (e.g., optical …bers), can be modeled using these techniques. 2.1. THE LUMPED-ELEMENT CIRCUIT MODEL FOR A TRANSMISSION LINE 19 Generic two-conductor TEM transmission line. We assume that the transmission line has dimensions on the order of a wavelength or larger. In this case, the traditional circuit equations (v = iz, etc.), which come from Maxwells …eld equations specialized for the case when the physical dimensions of the structure are small compared to a wavelength (i.e., low frequency), are no longer valid. The low frequency approximations are valid when applied to a small ( z ) section of the structure. In this way, we treat transmission lines as many cascaded sections of electrically small circuits, and apply circuit models to each section. The lumped-element model for a small segment of the line is shown in the …gure below. i(z,t) R ∆z L ∆ z Vs(z,t) ∆ z i(z+dz,t) + - + G ∆z Is(z,t) ∆z C ∆z v(z+dz,t) - ∆z The circuit elements are R: series resistance per unit length for both conductors, ohms=m: (R = 0 for perfect conductors) L: series inductance per unit length for both conductors, H=m G: shunt conductance per unit length, S=m. (G = 0 for perfect insulators) C: shunt capacitance per unit length, F=m is : shunt current source per unit length, A=m vs : series voltage source per unit length, V=m The distributed sources may represent, for instance, distributed currents and voltages induced on the transmission line by an external source. If desired, a localized source can be modeled by vs = v0 (z z 0 ), and similarly for is . We will assume R; L; G; C 2 R. Applying Kirchho¤’s voltage law and current law to the circuit of Figure 2.2 yields, respectively, v(z; t) + vs (z; t) z L z R @i(z; t) @t z i(z; t) v(z + (2.1) z; t) = 0 and i(z; t) + is (z; t) z C Dividing these two equations by G z v(z + @v(z + z; t) z @t z and taking the limit as @v(z; t) = @z @i(z; t) = @z R i(z; t) G v(z; t) z; t) i(z + (2.2) z; t) = 0: z ! 0 lead to @i(z; t) + vs (z; t); @t @v(z; t) C + is (z; t); @t L (2.3) 20 CHAPTER 2. TRANSMISSION LINE THEORY and, upon assuming time-harmonic conditions dv(z) = dz di(z) = dz R i(z) j!L i(z) + vs (z); G v(z) j!C v(z) + is (z): (2.4) These two coupled …rst-order di¤erential equations can be easily decoupled by forming second-order di¤erential equations d2 v(z) dz 2 d2 i(z) dz 2 2 d vs (z) ; dz d is (z) (G + i!C) vs (z) + ; dz v(z) = 2 (R + i!L) is (z) + i(z) = (2.5) where 2 = (R + j!L) (G + j!C) ; (2.6) and = +j 2 C is called the propagation constant (1=m). The real and imaginary parts of the propagation constant are known as the attenuation constant ( ) and the phase constant ( ), respectively. Usually in microwave analysis one …rst considers the homogeneous equations d2 v(z) dz 2 d2 i(z) dz 2 2 v(z) = 0; 2 (2.7) i(z) = 0; corresponding to the absence of any source or load. General solutions are found as v(z) = v0+ e z + v0 e+ z ; i+ 0e z + i0 e+ z ; i(z) = (2.8) which represent voltage and current waves. Exploiting (2.4) leads to the relationship between voltage and current as 1 i(z) = v + e z v0 e+ z ; (2.9) Z0 0 where R + j!L Z0 (2.10) is called the characteristic impedance (ohms) of the transmission line. Time-domain form: Assuming sinusoidal time variation, v(z; t) = = With Re v (z) ej!t nh Re v0+ e ( +j )z + v0 e+( +j )z i (2.11) o ej!t : v0 = v0 ej (2.12) we have v(z; t) = v0+ cos !t z+ + + v0 cos !t + z + : (2.13) z The …rst term, corresponding to e , is a forward (+z-traveling) wave, while the second term, corresponding to e+ z , is a backward ( z-traveling) wave. Also, = 2 , vp = ! = f: (2.14) 2.2. FIELD ANALYSIS OF TRANSMISSION LINES 21 Lossless Lines: If the line is lossless (R = G = 0), then = +j =0+j ; r p 2 L = ! LC = ; Z0 = : C 2.2 (2.15) Field Analysis of Transmission Lines In these notes we won’t follow the derivations presented in the text, but only summarize some of the results. Note that the RLGC parameters for a given transmission line are speci…c to that line (depending on the geometric con…guration of conductors and dielectrics, and on material properties), and are determined by an electromagnetic …eld analysis of the structure. The circuit parameters are obtained as L = C = R = G = 2 Z H H dS; jI0 j S Z " (2.16) E E dS; 2 jV0 j S Z Rs H H dl; 2 jI0 j l Z !"00 E E dl; 2 jV0 j l where E is a linear function of V0 , H is a linear function of I0 , S = cross section surface of the line, l = conductor boundary, Rs = 1= ( s ) is the surface resistivity of the conductors, and V0 ; I0 are constants that cancel out with the numerators . Some representative results are provided in the table below. a w ε,µ a D ε,µ ε,µ d b L C R G ln ab 2 "0 b ln a Rs 1 1 2 a 00 b 2 !" b ln a 2 1 cosh "0 cosh 1 D 2a D ( 2a ) Rs a !"00 cosh 1 D ( 2a ) d w "0 w d 2Rs w !"00 w d Note that for TEM transmission lines comprised of perfect conductors, the exact EM …eld solution is in full agreement with the microwave model (with G; L; C appropriately de…ned). For lossy conductors the microwave model is no longer an exact solution, but usually provides a very good approximation. For non TEM transmission lines (we’ll discuss later, microstrip is one example), the microwave engineering model is an approximation of the rigorous EM …eld solution. 22 CHAPTER 2. TRANSMISSION LINE THEORY 2.3 The Terminated Lossless Transmission Line A terminated transmission line is depicted schematically in the …gure below. Zg Vg Z 0 ,v p + - ZL Z in z=-L z=0 The generator (modeled by Vg and Zg ) can represent, for example, the output of an IC chip, the output of an antenna, etc., and the load (modeled by ZL ) represents, for example, the input to an IC chip, antenna, television receiver, etc. The voltage and current on the line are v(z) = v0+ e i(z) = v0+ Z0 j z + v0 e+j v0 +j e Z0 j z e z ; z (2.17) : It is convenient to reform (2.17) as v(z) = v0+ e i(z) = v0+ Z0 j z + v0 +j e v0+ j z e v0 +j e v0+ z ; z (2.18) : De…ning a load re‡ection coe¢ cient = v0 v0+ j z + L we have v(z) = v0+ e i(z) = v0+ Z0 e +j z Le j z ; +j z Le (2.19) : More generally, de…ne (z) = v0 j2 e v0+ z ; (2.20) = v0 ; v0+ (2.21) = v0 j2 e v0+ where we have the special cases (0) = ( L) = L in L ; such that (z) = L ej2 z (2.22) 2.3. THE TERMINATED LOSSLESS TRANSMISSION LINE 23 leading to v(z) = v0+ e v0+ i(z) = Z0 e j z j z (1 + (z)) ; (1 (z)) : (2.23) The total impedance at any point is Z (z) = v (z) 1+ = Z0 i (z) 1 (z) ; (z) (2.24) and, conversely, (z) = Z (z) Z0 : Z (z) + Z0 (2.25) Important special cases: (0) = ( L) = L in ZL Z0 ; ZL + Z0 Zin Z0 = : Zin + Z0 (2.26) = (2.27) An important formula for Zin can be obtain as follows; Z (z) = v+ e v (z) = 0+ v0 i (z) Z0 (e + j z L ej z L ej z) (2.28) jZ0 tan ( z) ; jZL tan ( z) ZL + jZ0 tan ( L) = Z ( L) = Z0 : Z0 + jZL tan ( L) = Z0 Zin j z ZL Z0 One important use of Zin is that once the input impedance has been determined, the transmission line circuit can be considered to be Iin Zg + Vg + - Vin Z in - such that 8 < 1 Zin 1 V PL = Re fvin iin g = Re : Zin + Zg g 2 2 Time-average power ‡ow: Zin Zin +Zg Zin Vg 9 = ; : (2.29) 24 CHAPTER 2. TRANSMISSION LINE THEORY Pav = = = 1 Re fv (z) i (z)g 2 ( 1 v0+ Re v0+ e j z + L ej z e j z 2 Z0 8 9 > > + 2 < = 1 v0 2 j2 z j2 z Re 1 + L e e j j L > > | {z L } 2 Z0 : ; (2.30) L ej z ) pure im aginary = 2 v0+ 1 2 Z0 n 1 j 2 Lj o = P inc: P ref: in this case –not generally : When the line is not matched, all of the available power is not delivered to the load. The return loss is RL = 20 log j Lj : (2.31) Special cases: j j Lj = Lj = 0; 1; RL = 1; RL = 0 dB. (2.32) Standing waves and standing wave ratio: In the sinusoidal steady state, for a matched line ( v (z) = v0+ e For a mismatched line ( L 6= 0), j z = 0), L jv (z)j = v0+ = constant. ; v (z) = v0+ e j z 1+ L ej2 z : (2.33) (2.34) Let L =j Lj ej : (2.35) Then, jv (z)j = vmax vmin = v0+ 1 + j j(2 z+ ) Lj e v0+ (1 + j L j) ; v0+ = (1 j ; (2.36) L j) : The standing wave ratio (SWR), also known as voltage standing wave ratio (VSWR) is SW R = vmax 1+j = vmin 1 j Lj Lj : (2.37) Note that 1 SW R 1; and that SW R = 1 denotes a matched load. When will SW R = 1? When j L j = 1. This will occur for (a) a short circuit, ZL = 0, or (b) an open circuit, ZL = 1, or (c) a pure reactive load, ZL = jXL , XL 2 R. (2.38) 2.3. THE TERMINATED LOSSLESS TRANSMISSION LINE 25 The last case is due to the fact that a pure reactance cannot dissipate power, so everything must be re‡ected. Distance between successive minima: ej(2 zm + ) = ej(2 (zm +dmin )+ ) = ej(2 zm + ) ej2 ! 2 dmin = n ; n = 0; 2; 4; ::: so that dmin = n 2 = n 22 = dmin n = : 4 2 (2.40) To examine the e¤ects of a pure standing wave, consider a short-circuited line (ZL = 0; v (z) i (z) = v0+ e v0+ = Z0 j z e j z ej z + ej z (2.39) v0+ 2j sin ( z) ; = = v0+ Z0 L = 1). (2.41) 2 cos ( z) : In the time domain, v (z; t) i (z; t) = v0+ 2 sin ( z) sin (!t) ; = v0+ Z0 (2.42) 2 cos ( z) cos (!t) : 1 0.5 -7 -6 -5 -4 bz -3 -2 0 -1 -0.5 -1 sin ( z) sin (!t) vs. z for !t = 0; ; 2; 6: Observations: 1. Voltage is zero at the short and at multiples of =2, v = 0 at z=n ; n = 0; 1; 2; ::: 2. Voltage is a maximum at all points such that z=m ; 2 m = 1; 3; 5; ::: 3. Current is maximum at the short circuit and at all points where v = 0. At all points where v = vmax , i = 0. 4. The total energy in any length of line a multiple of a quarter wavelength is constant, merely interchanging between energy in the electric …eld (voltage) and energy in the magnetic …eld (current). This is similar to energy relations in a resonant LC circuit. 26 CHAPTER 2. TRANSMISSION LINE THEORY For j Then, Lj = 6 1, SW R < 1 and both a standing wave and a travelling wave exist on the line. Let v (z) = v0+ e j z + ej( z+ ) = v0+ e j z + ej( z+ ) = v0+ (1 + e j z j z + v0+ ej( = v0+ (1 )e | {z j z + 2 v0+ ej 2 cos {z | pure travelling wave = ej . (2.43) )e } L j z e z+ ) j z +e z+ pure standing wave 2 } 1.4 1.2 1 0.8 0.6 0.4 0.2 -7 -6 -5 -4 = 0:5; (1 bz -3 j z :5) e -2 0 -1 + cos ( z) vs. z It is easy to see that no real power ‡ow is associated with standing waves, since Pav = 1 1 Re fv (z) i (z)g = Re 2 2 2jv0+ sin z 2 v0+ cos ( z) Z0 = 0: (2.44) Example 2.1 A lossless, air-…lled transmission line having characteristic impedance 50 ohms is terminated with a load ZL . The line is 1:2 m long, and operated at a frequency of 900 MHz. Determine (a) SWR, (b) Zin , (c) RL for ZL = 100 ohms and ZL = 50 ohms. Solution: f = 900 MHz, ! L = = c 3 108 = = 0:333 m. f 900 106 1:2 = 3:6; 0:333 L= 2 L = 7:2 (2.45) (2.46) ZL = 100 : (a) L = SW R = ZL Z0 100 50 1 = = ZL + Z0 100 + 50 3 1 + j Lj =2 1 j Lj (2.47) (b) Zin = Z0 ZL + jZ0 tan ( L) = 49:10 Z0 + jZL tan ( L) j35:03 ohms (2.48) 2.3. THE TERMINATED LOSSLESS TRANSMISSION LINE 27 (c) RL = 20 log j Lj = 9:54 dB. (2.49) ZL = 50 : (a) L = SW R = ZL Z0 50 50 = =0 ZL + Z0 50 + 50 1 + j Lj =1 1 j Lj (2.50) (b) Zin = Z0 ZL + jZ0 tan ( L) = 50 ohms Z0 + jZL tan ( L) (2.51) (c) RL = 20 log j Lj = 1 dB. (2.52) Exercise 2.1 (in-class): Z 0 ,v p ZL Z in z=-L z=0 If L = =2 and ZL = 0 (short circuit), …nd (a) SW R, and (b) Zin . Exercise 2.2 (in-class): 1. The input to a television receiver presents an impedance of ZL = 60 + j75 ohms to a Z0 = 75 ohm coaxial cable. The cable has length 0:2 at the frequency of operation. Determine (a) SWR (b) L (c) Zin 2. Assume an antenna supplies a signal modeled by an open-circuit voltage of 5 V with source resistance 300 ohms to the transmission line described in (1). Determine the power delivered to the television. 3. A 300 ohm transmission line is short-circuited at the load-end. Determine Zin is the transmission line has length (a) l = =4, (b) l = =2; (c) l = : 28 CHAPTER 2. TRANSMISSION LINE THEORY Special cases of lossless terminated lines: I. ZL = 0 (short circuit) Zin = jZ0 tan L; (2.53) which is pure imaginary for any length line L: 4 2 0 1 2 3 z 4 5 6 7 -2 -4 tan( L) vs. Zin Zin = 0 at L = 0; = 1 at L = II. ZL = 1 (open circuit) Zin = L =2; ; 3 =2; ::: =4; 3 =4; 5 =4; ::: jZ0 cot L (2.54) (2.55) which is also pure imaginary for any length L: 4 2 0 1 2 3 z 4 5 6 7 -2 -4 cot( L) vs. Zin Zin L: = 0 at L = =4; 3 =4; 5 =4; ::: = 1 at L = 0; =2; ; 3 =2; ::: (2.56) III. L = =2, Zin = ZL (2.57) Transmission lines of length =2, or any integer multiple of =2, do not transform the load impedance –it is seen directly at the input. 2.3. THE TERMINATED LOSSLESS TRANSMISSION LINE 29 IV. L = =4; Zin = Z02 : ZL (2.58) This case is know as a quarter-wave transformer, and will be discusses in detail later. Determining v0+ : Although we often don’t need to explicitly determine v0+ , it is sometimes useful. To determine v0+ , consider the transmission line shown below. Zg Vg Z 0 ,v p + - ZL Z in z=-L z=0 We have v (z) v ( L) = v0+ e = j z v0+ ej L but v ( L) = vin = 1+ 1+ L L ej2 e z ; j2 L ; Zin Vg Zin + Zg (2.59) (2.60) Equating (2.59) and (2.60) leads to v0+ = Vg Zin Zin + Zg [ej 1 L + Le j L] : (2.61) Example 2.2 If Vg (t) = 2 cos 108 t + =4 , Zg = 1 ohm, Z0 = 50 ohms, L = 3 m, ZL = 75 ohms, and vp = 2:8 108 m/s, determine v0+ . Solution: = L = Zin 2 ! = 0:357; L = 1:07; vp ZL Z0 75 50 1 = = ZL + Z0 75 + 50 5 ZL + jZ0 tan ( L) Z0 + jZL tan ( L) 75 + j50 tan (1:07) = 50 50 + j75 tan (1:07) = 38:23 j13:42 ohms, = Z0 v0+ = Vg with Vg = 2 ej 4 leads to = Zin Zin + Zg [ej v0+ = 2:14 1 L + j0:23: Le j L] (2.62) (2.63) (2.64) (2.65) 30 CHAPTER 2. TRANSMISSION LINE THEORY Example 2.3 You …nd cable of unknown characteristic impedance having length 1:5 m. At a frequency at which you assume the cable is less than a quarter wavelength, you measure the open- and short-circuit input impedance to be oc Zin sc Zin = j54:6 ohms, = j103 ohms. What is Z0 ? Solution: Assume the cable is very low loss. oc sc Zin Zin Notice that you can also …nd ( jZ0 cot L) (jZ0 tan L) = Z02 ( j54:6) (j103) = 5623:8 p ) Z0 = 5623:8 = 75 ohms. = = as oc Zin sc Zin jZ0 cot L 2 = (cot L) jZ0 tan L j54:6 = 0:53 j103 cot L = 0:728 L = 0:941; = 0:628 1/m. = = ) ) 2.4 (2.66) (2.67) The Smith Chart The Smith chart is a graphical tool that serves as an aid in performing transmission line calculations. Although any such calculation can be easily performed on a computer, the Smith chart helps in developing intuition, and serves as a useful background upon which to plot microwave engineering data (from measurements or simulation) The Smith chart is really the portion j j 1 of the complex –plane, although the numbers on the Smith chart are impedance values. The mapping between the Z–plane and the –plane is developed as follows. Z Z0 = Z = r + jx = = 1+ 1 1 = 2 r (1 (2.68) 1+( 1 ( 2 i 2) + r + j i) r + j i) + 2j i : 2 r i Equating real and imaginary components, r= 2 2 r + i 2) + 2 r i 1 (1 ; x= 2j (1 i 2) + r 2; i (2.69) which can be rearranged to yield r ( 2 r 1+r 2 r 1) + i + 2 i 1 x 2 = = 1 2; (1 + r) 1 : x2 (2.70) (2.71) Recall that this is simply 1+ (2.72) 1 separated into real and imaginary components. Both (2.70) and (2.71) are circles in the complex –plane. Z = Z0 2.4. THE SMITH CHART 31 i. The …rst equation, (2.70), shows that the locus of constant resistance r in the r 1 circles with centers on the r axis at r = 1+r ; i = 0 with radii 1+r . r i plane is a family of Γi r=0 r=0.5 r=1 1 Γr ii. The second equation, (2.71), shows that the locus of constant reactance x in the of circles with centers on the r axis at r = 1; i = x1 with radii x1 : Γi r X=1 r=0 X=2 r=1 |Γ|<=1 . (1,0.5) Γr . (1,-0.5) X=-2 X=-1 Superimposing the previous two plots, and showing only the j j 1 portion, leads to i plane is a family 32 CHAPTER 2. TRANSMISSION LINE THEORY Γi r=0 X=2 X=0.7 |Γ|=1 Γr X=-0.7 X=-2 Continuing for the other values 0 r < 1; 1<x<1 (2.73) generates the Smith chart shown below. Note that on the Smith chart the numbers denote r and x values instead of r and i values. Re‡ection coe¢ cient values, along with other relevant information, is obtained from the scales provided at the bottom of the chart. iii. Since (z) = j2 z Le and L = ZL Z0 ZL +Z0 = ej , = constant, then (z) = ej ej2 z ; (2.74) such that moving along a uniform line toward the load end (i.e., increasing position z) is equivalent to rotating on a constant j j = circle (constant SW R circle) with increasing angle = 2 z in the counter-clockwise direction. Moving towards the source results in a clockwise rotation. Γi . . Γr Since 2 L + of L = =2. = 22 L + =4 L + , a full rotation about the Smith chart corresponds to a distance 2.4. THE SMITH CHART 33 34 CHAPTER 2. TRANSMISSION LINE THEORY iv. Since SW R = then a circle of constant 1+j j 1+ = 1 j j 1 (2.75) is also a circle of constant SW R. v. Since Z= 1+ 1 Y = = 1 + ej(2 1 ej(2 z+ ) z+ ) ; (2.76) then 1 = Z 1+ 1 = 1 ej(2 z+ 1 + ej(2 z+ ej(2 z+ + ) : ej(2 z+ + ) ) (2.77) ) Therefore, for a given point Z on the Smith chart, the corresponding admittance Y is obtained by rotating radians (one-half rotation) along a constant SW R circle. Thus, impedance (admittance) and be converted to admittance (impedance) by re‡ecting the impedance point through the origin. Γi Z . Γr Y Example 2.4 Assume L = 0:35 , ZL = 50 Solution: 1. Start on the Smith chart at Z = 1 . j200 ohms, Z0 = 50 ohms. Find L, Zin , in , and SW R. j4. 2. From the SW R legend (or constant SW R circle intersection with the positive-real axis), SW R = 20 ( SW R = 17:94 from the equation). 3. From the equation). legend and Smith chart angle, j j = 0:9, 6 L = 27 ( = 0:896 26:57 from the 4. To …nd Zin , rotate 0:35 toward the generator (clockwise) around a constant SW R circle ( 0:287 + 0:35 = 0:637 ). Zin = Z0 (0:13 + j1:17) = 6:5 + j58:5 ( Zin = 6:5206 + j57:67 from the equation). 5. in = 0:96 81:7 . 2.4. THE SMITH CHART 35 Note: all possible values of Z (z) are found on the constant SW R circle. Thus, it is very easy to visualize how Z varies along the line. Example 2.5 A transmission line is terminated in ZL . Measurements show that the standing wave minima are 102 cm. apart, and that the last minimum is 35 cm. from the load. The measured SW R is 2:4, Z0 = 250 ohms, and vp = c. Determine (a) frequency and (b) ZL . Solution: (a) 2 f = = 102 ) = 204cm., vp 1010 = 147:06MHz. 204 = 3 (b) vmin is l = 35 cm from the load. l= = 0:172. On Smith chart, draw SW R circle. Start at vmin and rotate ccw 0:172 . Zin = Z0 (1:16 j0:96) = 290 j240 ohms. 36 CHAPTER 2. TRANSMISSION LINE THEORY Example 2.6 A lossless 100 ohm transmission line is terminated in 200 + j200. The line is 0:375 long. Determine (a) L , (b) SW R, (c) Zin , and (d) the shortest length of line for which the impedance is purely resistive, and the value of the resistance. Solution: (a) Z L = 2 + j2. L = 0:626 30 = 0:62ej 6 . (b) SW R = 4:3 (c) Zin = Z0 (0:32 + j0:54) = 32 + j54 (d) l = 0:042 , R = Z0 4:3 = 430 ohms. 2.4. THE SMITH CHART 37 Exercise 2.3 (in-class) The input to a television receiver presents an impedance of ZL = 60 + j75 ohms to a Z0 = 75 ohm coaxial cable. The cable has length 0:3 at the frequency of operation. Determine, using the Smith chart, 1. SWR 2. L 3. Zin Smith Chart with lumped elements: 38 CHAPTER 2. TRANSMISSION LINE THEORY C B A XL=30 Z0 Xc=-200 Z0 ZL Z in 0.06λ Example 2.7 XL = 30 (Zind = j!L = jXL ), XC = 0.12λ 200 (Zcap = 1 j!C = 1 j !C = jXC ). Determine Zin . Solution: 1. First obtain ZA in the usual way–enter Smith chart at Z L = 2 + j1:5. Rotate 0:12 generator. Z A = 1 j1:3. toward the 2. Incorporate inductor L1 : (a) Option a: Z B = ZA + j30 50 =1 j0:7: (b) Option b: Note that adding reactance just means moving along a constant resistance contour to the desired point. The distance moved is j30 50 = j0:6. 3. Convert Z B to Y B . Y B = 0:67 + j0:47. 4. Incorporate capacitor C1 : (a) Option a: Y C = Y B + 1 j200 50 = 0:67 + j0:72. (b) Option b: Note that adding susceptance just means moving along a constant conductance contour 1 to the desired point. The distance moved is j200 = j0:25. 50 5. Either rotate 0:06 toward generator to get Yin and convert to Zin , or convert to Z C …rst then rotate 0:06 to get Zin . Zin = Z0 (0:44 j0:39) = 22:4 j18:97 ohms. 2.5. THE QUARTER-WAVE TRANSFORMER 39 Exercise 2.4 (in-class) Determine, by moving along contours on the Smith chart, the input impedance seen at the beginning of the line. 20+j10 ohms 50 ohms 0.15 2.5 200+j100 ohms λ The Quarter-Wave Transformer The =4 transformer is used to match two di¤erent real impedance values as shown below. 40 CHAPTER 2. TRANSMISSION LINE THEORY Z0 RL Z1 Z in λ/4 Zin = Z1 RL + jZ1 tan ( L) Z1 + jRL tan ( L) (2.78) but L = =2, Zin = Assume we want no re‡ection at the input ( in Z12 : RL (2.79) = 0, Zin = Z0 ). Choose Z1 = p RL Z0 . (2.80) To connect a real-valued load RL with a transmission linephaving Z0 6= RL , use a =4 length section of transmission line having characteristic impedance Z1 = RL Z0 . Standing waves do occur on the =4 section of line. Example 2.8 Design a quarter-wave transformer to match a 200 ohm load to a 50 ohm line. Z= p (50) (200) = 100 ohms. Quarter-wave transformers are very frequency sensitive. Why? Later we will study methods to broaden the usable frequency range by using cascaded sections of =4 lines. The multiple Re‡ection Viewpoint: Read 2.6 Generator and Load Mismatches In this section we consider the general case of a source-excited, loaded lossless line, and power delivered to the load. Zg Vg Z 0 ,v p + - ZL Z in z=-L z=0 2.7. LOSSY TRANSMISSION LINES PL = vin = 41 1 Re (vin iin ) , since the line is lossless. 2 Zin Vg vin Vg ; iin = = ; Zin + Zg Zin + Zg Zin (2.81) 2 noting that ZZ = jZj we get 2 PL = = 1 Zin 1 2 Re jVg j 2 Zin + Zg Zin 1 Rin 2 jVg j 2 2: 2 (Rin + Rg ) + (Xin + Xg ) (2.82) 1. Special case: load matched to line: ZL = Z0 , L = 0, SW R = 1, Zin = Z0 : PL = 1 Z0 2 jVg j : 2 2 (Z0 + Rg ) + Xg2 (2.83) 1 2 1 jVg j : 8 Z0 (2.84) Further, if Zg = Z0 , PL = 2. Special case: generator matched to a loaded line: Zin = Zg , in = 0. Rg 1 2 jVg j : 2 4 Rg2 + Xg2 PL = (2.85) Depending on conditions, either case may provide the larger power delivered to the load. What determines maximum power transfer from generator to load? As in circuit theory, conjugate matching. Zin = Zg (2.86) This will result in maximum power transfer to the load for a …xed Zg . The resulting power delivered to the load is 1 2 1 : (2.87) PL = jVg j 2 4Rg Note that PL is made large by making Rg small. 2.7 Lossy Transmission Lines Before considering lossless transmission lines we obtained (2.8) v(z) = v0+ e z + v0 e+ z ; 1 i(z) = v + e z v0 e+ Z0 0 = Z0 (2.88) z ; p + j = (R + j!L) (G + j!C); s R + j!L : G + j!C (2.89) 42 CHAPTER 2. TRANSMISSION LINE THEORY For lossy lines z e = e| {z z} j z e : (2.90) attenuation In this case, movements along the line no longer are represented by rotation along a constant SW R circle (for lossy lines j j = 6= constant). For the low-loss case we have ' 1 2 R + GZ0 ; Z0 p ' ! LC; (2.91) Z0 ' r L . C The distortionless line: The exact equation for of a lossy transmission line, p = (R + j!L) (G + j!C); (2.92) indicates that is a complicated function of frequency. If doesn’t have a linear relationship with frequency, then the phase velocity vp = != will be di¤erent at each frequency (vp = vp (!)). Di¤erent frequency components of a signal (as in a Fourier decomposition) will travel at di¤erent velocities, reaching the load at di¤erent times, resulting in dispersion of the signal (a form of distortion). This will be discussed later. One special case, mostly of academic interest, exists for a lossy line to be dispersion-free. If R G = L C then p = ! LC; (2.93) =R r C . L (2.94) The resulting line will not distort the signal, although attenuation will still occur. For a general lossy line power relations are Pin PL PLoss 2 v0+ n 1 j 2Z0 2 v0+ n = 1 j 2Z0 = Pin PL ; = where v0+ = Vg Zin Zin + Zg [e 2 in j 2 Lj o o e2 L ; ; 1 L + Le (2.95) L] (2.96) (compare with (2.61)). 2.8 Transient Transmission Lines (not in text) The time-domain transmission line (telegrapher) equations, (2.3), are @v(z; t) = @z @i(z; t) = @z R i(z; t) G v(z; t) @i(z; t) ; @t @v(z; t) C ; @t L (2.97) (2.98) 2.8. TRANSIENT TRANSMISSION LINES 43 where we have set the distributed sources vs ; is = 0. These …rst-order, coupled partial di¤erential equations are decoupled by applying @=@z to the …rst equation (2.97) and substituting into the second equation (2.98) to yield a second-order partial di¤erential equation for v. A similar equation is obtained by applying @=@z to (2.98) and substituting into (2.97), resulting in @2 @z 2 LC @2 @t2 (RC + GL) @ @t v (z; t) i (z; t) RG = 0: (2.99) For simplicity we will assume a lossless line (R = G = 0), resulting in @2 @z 2 LC @2 @t2 v (z; t) i (z; t) = 0: (2.100) We will solve for v (z; t) and use (2.97) to obtain i (z; t). In order to solve @2 @2 LC v (z; t) = 0 @z 2 @t2 (2.101) we use Laplace transforms. Recall L fv (z; t)g = V (z; s) ; @ v (z; t) = sV (z; s) v z; t = 0+ ; L @t @2 L v (z; t) = s2 V (z; s) sv z; t = 0+ @t2 (2.102) @ v (z; t) @t ; t=0+ where V (z; s) is the Laplace transform of v (z; t). Initial conditions are 1. v (z; t) across C cannot change instantaneously. Assume v (z; t < 0) = 0. Then, v (z; t = 0+ ) = 0. 2. i (z; t) ‡owing in L cannot change instantaneously. Then, @ v (z; t) @t 1 @ i (z; t) C @z = t=0+ = 0: (2.103) t=0+ Taking the Laplace transform of the wave equation (2.101) results in @2 @z 2 where 2 V (z; s) = 0 (2.104) p = s LC. The solution of (2.104) is V (z; s) = V + (s) e = V + (s) e z p (s) e+ +V s LCz +V z (2.105) p +s LCz (s) e where V (s) are constants of integration with respect to z. The solution v (z; t) is obtained via inverse Laplace transform, n o p p v (z; t) = L 1 V + (s) e s LCz + V (s) e+s LCz ; (2.106) and, using the time-shifting theorem L ff (t we obtain v (z; t) = v + t t0 )g = F (s) e p LCz + v st0 t+ (2.107) p LCz : (2.108) 44 CHAPTER 2. TRANSMISSION LINE THEORY p We can interpret the term LCz by noting that the propagation velocity (the velocity with which an imaginary observer must move to follow a point of constant amplitude on the wavefront) is obtained from v t p LCz p 1 LC = constant, p ) t LCz = constant, p d t ) LCz = 0; dt dz dt dz dt = (2.109) 0; (2.110) 1 = vp ; LC 1 . ) vp = p LC p = Therefore, v + (t z=vp ) | {z } v (z; t) = + forward travelling wave 1 v + (t Z0 (2.111) backward travelling wave The associated current can be obtained as i (z; t) = v (t + z=vp ) : | {z } z=vp ) v (t + z=vp ) (2.112) as shown in the appendix. In summary, we have v (z; t) = i (z; t) = v + (t z=vp ) + v (t + z=vp ) ; 1 v + (t z=vp ) v (t + z=vp ) ; Z0 (2.113) valid for an in…nite line. To analyze a …nite length, source-excited and loaded line we need to consider conditions at the source-end and load-end of the line. Terminated transient line: i(z,t) i(L,t) + + v(z,t) Z 0 ,v p v(L,t) - RL - With v (L; t) = = i (L; t) = = v + (L; t) + v (L; t) + vL (t) + vL (t) = vL (t) + i (L; t) + i (L; t) i+ L (t) + iL (t) = iL (t) (2.114) 2.8. TRANSIENT TRANSMISSION LINES 45 and Ohm’s law applied at the load, vL (t) = iL (t) RL ; (2.115) we obtain + vL (t) + vL (t) = = i+ L (t) + iL (t) RL 1 v + (t) vL (t) RL Z0 L (2.116) leading to vL = RL Z0 + v = RL + Z0 L + L vL (2.117) (compare with (2.26) for the phasor case). Launching waves on an in…nite line: Rg i(0,t) i(z,t) + Vg v(0,t) + - Z 0 ,v p v(z,t) - Ohm’s law at z = 0 results in v (0; t) = vS (t) = Vg (t) | {z } vs (t) for all t. In particular, vs 0; t = 0+ = Vg t = 0+ = Vg t = 0+ Rg i (0; t) | {z } (2.118) is (t) Rg is t = 0+ (2.119) + Rg vs (t = 0 ) ; Z0 leading to vs 0+ = Z0 Vg 0+ = v + 0; 0+ Z0 + Rg (2.120) which gives the initial amplitude of the wave launched on the line. For z; t 6= 0 we obtain v (z; t) = i (z; t) = Z0 Vg (t z=vp ) = v + (t Z0 + Rg v (z; t) = i+ (t z=vp ) : Z0 z=vp ) ; Complete transient response of a terminated, source-driven transmission line: bounce diagrams: (2.121) 46 CHAPTER 2. TRANSMISSION LINE THEORY Rg Vg Z 0 ,v p + - RL z=0 z=L z 0,0 z=L z 0,0 v+ t T ΓL v+ 2T ΓL Γg v+ 3T i T t - Γ i+ L ΓL Γg i + 2T 3T 2 ΓL Γg v+ 4T 2 -Γ Γ i+ L g 4T i(z,t) v(z,t) T = v (z; t) = i (z; t) = L = one-way transit time for wave transverse line. vp Z0 fVg (t Z0 + Rg + z=L + 2 L g Vg (t (t L Vg z=vp ) (t 2T + z=vp ) + 2 2 L g Vg 4T + z=vp ) + 1 fVg (t Z0 + Rg 2 L g Vg z=vp ) + L Vg (t 4T (t (t 2T z=vp ) z=vp ) + ::: ; 2T + z=vp ) + 2 2 L g Vg 4T + z=vp ) + (t L g Vg 4T L g Vg (t (2.122) 2T z=vp ) + ::: z=vp ) (2.123) Example 2.9 Consider a matched line (RL = Z0 ) excited by an ideal (Rg = 0) source. Determine v (z; t). Solution: Vg Z 0 ,v p + - RL z=0 z=L L ( g = 0; g = Rg Z0 = Rg + Z0 1 (2.124) is not needed since line is matched – no re‡ection comes back towards the source). v (z; t) i (z; t) Z0 Vg (t z=vp ) = Vg (t Z0 + 0 = Vg (t z=vp ) =Z0 = z=vp ) ; : 2.8. TRANSIENT TRANSMISSION LINES 47 Vg(t)/Vo 1 t To V(L,t)/Vo leading edge 1 trailing edge t T+To T=L/Vp Example 2.10 Assume that a step voltage is applied to a line mismatched at both ends. If RL = 3Z0 , Rg = 3Z0 , determine v (0; t) and v (L; t). Solution: Rg Vg Z 0 ,v p + - z=0 Vg (t) L v+ RL z=L = V0 u (t) ; RL Z0 3Z0 Z0 1 = = = = RL + Z0 3Z0 + Z0 2 Z0 V 0 = Vg t = 0+ = : Z0 + Rs 4 s; Then, v (z; t) = V0 1 1 u (t z=vp ) + u (t 2T + z=vp ) + u (t 2T 4 2 4 1 1 + u (t 4T + z=vp ) + u (t 4T z=vp ) + ::: ; 8 16 v (0; t) = = V0 1 1 u (t) + u (t 2T ) + u (t 2T ) 4 2 4 1 1 + u (t 4T ) + u (t 4T ) + ::: 8 16 V0 3 3 u (t) + u (t 2T ) + u (t 4T ) + ::: 4 4 16 z=vp ) (2.125) (2.126) (2.127) 48 CHAPTER 2. TRANSMISSION LINE THEORY v (L; t) = = V0 1 1 u (t T ) + u (t 2T + T ) + u (t 2T 4 2 4 1 1 + u (t 4T + T ) + u (t 4T T ) + ::: 8 16 V0 3 3 u (t T ) + u (t 3T ) + ::: 4 2 8 T) (2.128) (2.129) V(0,t)/(Vo/4) 3/16 3/4 1 t/T 1 2 3 4 5 6 V(L,t)/(Vo/4) 3/8 1 3/2 t/T 1 2 3 4 5 6 As a check, d.c. analysis yields v (t) = and V0 v (L; t = 1) = 4 ( 1+ 3Z0 V0 V0 = 3Z0 + 3Z0 2 1 2 1 + 1 2 2 + (2.130) 1 2 3 ) + ::: = V0 2 (2.131) which agrees with the plot. Example 2.11 Consider a pulse-excited line mismatched at the load. If RL = 2Z0 , Rg = Z0 , determine the voltage at the middle of the line. Solution: Rg Vg Z 0 ,v p + - z=0 RL z=L 2.8. TRANSIENT TRANSMISSION LINES 49 Vg(t)/Vo 1 t To v (z; t) = = v (L=2; t) = = L = g = 2Z0 Z0 1 = ; 2Z0 + Z0 3 0 Z0 fVg (t z=vp ) + L Vg (t 2T + z=vp )g Z0 + Z0 1 1 Vg (t z=vp ) + Vg (t 2T + z=vp ) 2 3 1 1 T Vg t + Vg (t 2T + T =2) 2 2 3 1 1 T 3T Vg t + Vg t : 2 2 3 2 V(L/2,t)/(1/2) To<T/2 1 1/3 t/(T/2) 1 (T/2) I. T0 < T =2 2 3 (T) (3T/2) 4 5 6 V(L/2,t)/(1/2) To=T/2 1 1/3 t/(T/2) 1 II. T0 = T =2 (T/2) 2 3 (T) (3T/2) 4 5 6 (2.132) (2.133) (2.134) (2.135) (2.136) 50 CHAPTER 2. TRANSMISSION LINE THEORY V(L/2,t)/(1/2) To=T 1 1/3 t/(T/2) 1 (T/2) 2 3 4 (T) (3T/2) 5 6 III. T0 = T V(L/2,t)/(1/2) To=2T 1 1/3 t/(T/2) 1 (T/2) 2 3 (T) (3T/2) 4 5 6 IV. T0 = 2T Exercise 2.5 (in-class) Two IC chips are mounted a printed circuit board. Pin 3 on one chip is connected to pin 5 of the other chip by a printed conducting trace, forming a 20 ohm transmission line. The transmission line has length l = 1 cm, and the velocity of signal propagation on the line is vp = 2:6 108 m/s. Pin 3 can be modeled as a voltage source which provides a 1:5 volt pulse starting at t = 0 and lasting for 1 ns, and having source resistance Rg = 10 ohms. Pin 5 can be modeled as providing a constant 30 ohm resistance. Plot the voltage waveform at pin 5 versus time. Example 2.12 Consider the charged-line pulse generator shown below. If Rg = Z0 , RL the voltage at z = 0. Solution: t=0 i(L,t) i(0,t) + = >> Z0 + v(0,t) Rg RL v(L,t) Z 0 ,v p + - Z0 Z0 determine - - z=0 z=L V0 Operation: resistor Rg is switched to close the circuit and interrupt the d.c. state. Output voltage v (0; t) is a pulse of width 2T These types of pulse generators are used to generate very high power, fast pulses. Initial state (t < 0): v (z; t) = V0 , i (z; t) = 0. For t > 0 v 0; t = 0+ = g = L = Rg Z0 = 0; Rg + Z0 RL Z0 ' 1: RL + Z0 V0 + v + = + Rg i 0; t = 0+ v ' v+ . Z0 ) 2v + = V0 (initial pulse amplitude). = Rg (2.137) (2.138) 2.8. TRANSIENT TRANSMISSION LINES v (z; t) = V0 |{z} 51 v + (t z=vp ) | {z } + d.c. voltage forw ard w ave (re‡ected on ce at RL ) v (z; t) = V0 v (0; t) = V0 = V0 V0 u (t z=vp ) u (t 2 2 1 1 1 u (t) u (t 2T ) 2 2 V0 fu (t) 2 u (t + v (t | 2T + z=vp ) {z } (2.139) backw ard w ave 2T + z=vp ) (2.140) 2T )g : v(0,t)/(Vo/2) 1 t 2T 2.8.1 Waveforms and Spectral Analysis The text (Pozar) deals with time-harmonic signals and associated analysis. However, a time-harmonic waveform can not be used to transmit information. In the analog domain, modulation is commonly used to impart information on a high-frequency carrier, which spreads the frequency content out from the single spectral component of the carrier. Time-harmonic analysis will be valid if the bandwidth of the modulated signal is su¢ ciently small. In the digital domain, one must consider pulses and pulse trains. In this section we consider the frequency-domain aspects of some digital signals, and the relation to time-harmonic analysis. Periodic Signals Periodic waveforms can be represented Fourier series. Let g (t) be periodic with period T , i.e., g (t) = g (t + T ). Then, 1 a0 X n2 t n2 t g (t) = + + bn sin an cos 2 T T n=1 where an = bn = 2 T 2 T Z T =2 g (t) cos n2 t dt; T g (t) sin n2 t dt; T T =2 Z T =2 T =2 for n = 0; 1; 2; :::, or in exponential form g (t) = 1 X cn ej n2 t T n= 1 where cn = 1 T Z T =2 g (t) e j n2T t dt; T =2 n = 0; 1; 2; ::: (the equality is in the sense of the Fourier series). For example, for a pulse train consisting of rectangular pulses having amplitude A, pulse width t0 , and period T , as shown below, 52 CHAPTER 2. TRANSMISSION LINE THEORY g(t) t0 A ... t T we have g (t) 1 X = cn ej n2 t T ; n= 1 cn = = Z 1 T 1 T = A Z T =2 j n2T g (t) e T =2 t0 j n2T Ae t t dt dt 0 t0 e T t0 T jn sin n Tt0 n t0 T (for this to be valid t0 < T =2, otherwise the limits of integration need to be adjusted). The quantity f0 = 1=T is called the fundamental frequency (or pulse repetition rate), in which case cn = At0 f0 e jn t0 f0 (sin n t0 f0 ) : n t0 f0 A plot of jcn j vs. n=T looks like the following. Cn c0 1 3 T T n/T 2 t0 1 t0 Although the spectrum is discrete, the envelope is drawn for convenience. Note that as T decreases (f0 increases), the density of spectral lines decreases, and the amplitude of the spectral lines increases, the shape of the envelope is determined by t0 , the width of a single pulse. Therefore, the spectral envelope of the pulse train, especially how far out in frequency it extends, is governed by the pulse width, t0 . Non-Periodic Signals If we consider a single pulse, we need to use Fourier transforms, Z 1 1 g (t) = G (!) ej!t d!; 2 1 ! = 2 f , where G (!) = Z 1 1 g (t) e j!t dt: 2.8. TRANSIENT TRANSMISSION LINES 53 For a single rectangular pulse, G (f ) = At0 which has zeros at f= sin ( t0 f ) ; t0 f m ; t0 m = 1; 2; 3; :::.These are exactly the locations of the zeros of the envelope of cn for a pulse train of period T . Comparing the spectrum of the pulse train and of the single pulse, sin ( t0 f ) ; t0 f (sin n t0 f0 ) jcn j = At0 f0 ; n t0 f0 jG (f )j = At0 one sees that the pulse width t0 e¤ects the amplitude of each spectrum in the same way. For a pulse train (such as a clock signal), the period T is not too important from a spectral analysis standpoint. It does a¤ect the density of the discrete spectral components, but not how fast the spectrum decreases with increasing frequency. From a spectral analysis standpoint, the pulse width is of paramount importance, for both the pulse and the pulse train. For a pulse train, the main importance of the period is logic timing. For a modulated analog signal, the important spectral region is near the carroer frequency, whereas for a digital pulse, the important spectral content is near the origin in the frequency domain. 2.8.2 Integrated Circuits and Ground Bounce Consider a typical integrated circuit (IC) logic device as shown below. Vcc A VA Q1 Vout Q2 B VB Gnd When Q1 is on and Q2 is o¤ the output voltage is high (Vout = VA = Vcc ), and when Q1 is o¤ and Q2 is on the output voltage is low (Vout = VB = 0). However, VA is not actually Vcc volts, and VB is not actually 0 volts, due to parasitic inductance. Both leads A and B are wires, and, like all wires, they have inductance. The inductance of a very short (electrically short) straight wire is very small, although perhaps not negligible depending on the application, leading to the model shown below. 54 CHAPTER 2. TRANSMISSION LINE THEORY Vcc Lwire Q1 Vout Q2 Lwire Gnd Since vind = Lind diind (t) ; dt (2.141) then if iind is large enough and/or dt is small enough, vind will be large and Vout (t) = Vcc vind (t) or Vout (t) = 0 + vind (t). This phenomena is called ground bounce (and Vcc bounce). If the noise margins of the circuit downstream are too tight, or the timing is too fast (all voltages approach their idealized values when the currents stop changing) a logic error may follow. To remedy this situation we may 1) slow down the circuit (not a viable choice), reduce the margins (often not a viable choice), or minimize Lind , which is usually the preferred method, if possible. The above described ground bounce occurs internal to a chip. Integrated circuits (ICs) have been traditionally formed on a silicon die, which is glued to a mechanical base (this is referred to as “packaging” the circuit). Small wires, called wirebonds, connect the die to the package’s external leads. Even through these wirebonds are typically electrically short, they have, like all wires, inductance (on the order of 2 nH for a typically wirebond). Therefore, any IC that has large transient currents can induce large, undesirable voltage transients on the wirebonds. Wirebonds connecting to ground often have large current transients, resulting in ground bounce as well. For example, if an IC chip has several hundred thousand gates, many of which are switching at the same time, and all connected to ground through the same wirebond, the switching current can become quite large, resulting in ground bounce of several volts. To make matters worse, the inductance of the ground and power planes is added to the inductance of the wirebonds, enhancing this deleterious e¤ect. Flip-chip technologies (which spread the ground connections over many solder bumps) are e¤ective at mitigating this e¤ect. 2.8. TRANSIENT TRANSMISSION LINES 55 Vcc Lwire+Lpower- plane Q1 Vout Q2 Lwire+Lground- plane Gnd Although ground bounce is not a “transmission line” e¤ect, it is particular to high frequencies, and relates to the electromagnetic properties of the circuit. Bypass capacitors are often used to remove the ground and power plane transients from the IC circuits. Since current can change “instantaneously”through a capacitor, the transient currents on the IC device don’t have to ‡ow through the ground and power planes, they ‡ow through the capacitor. Wide traces are used if possible (wide traces have lower inductance and larger capacitance). 56 CHAPTER 2. TRANSMISSION LINE THEORY Chapter 3 Transmission Lines and Waveguides In the previous chapter we studied transmission lines in an abstract sense, usually without referring to a particular physical structure. In this chapter we will study several types of transmission lines and waveguides commonly used in practice. As is evident from their physical structure, each transmission line and waveguide has particular advantages and disadvantages depending on the application of interest. Any waveguide or transmission line that is capable of supporting a TEM mode of propagation can be analyzed using the transmission line theory discussed in the previous chapter (for that mode). Waveguides and transmission lines that do not support a TEM mode can be approximately analyzed using the previously developed analysis methods. In general, the term “waveguide” is applied to structures consisting of a single conductor. In this case, the structure cannot support a TEM mode. Two-conductor structures are usually termed “transmission lines,” and can support a TEM mode (and higher-order modes). 3.1 General Solutions for TEM, TE, and TM Modes In this section we will study general solutions for propagation modes along structures invariant along the z axis (and therefore in…nitely long). The conductors will be initially assumed to be perfectly conducting; later brief mention will be made of attenuation calculations. It can be show from the theory of Fourier transforms Z 1 f (x; y; z)e j z dz; 1 Z 1 1 1 f (x; y; )ej z d : f (x; y; z) = F ff (x; y; )g = 2 1 f (x; y; ) = F ff (x; y; z)g = (3.1) (3.2) that …elds on a z invariant structure can be written as E (x; y; z) = H (x; y; z) = [e (x; y; ) + z ez (x; y; )] e [h (x; y; ) + z hz (x; y; )] e j z (3.3) j z where fe; hg represent the (as yet unknown) transverse …elds, fez ; hz g represent the (as yet unknown) longitudinal …elds, and is the (as yet unknown) propagation constant. The analysis in the text shows how to obtain these unknown quantities for a given physical structure. Here we will mostly be concerned with obtaining . Starting with the source-free curl equations r r E = j! H; H = j!"E 57 (3.4) 58 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES and assuming the form (3.3) leads to @Ez + j Ey = @y @Ez j Ex = @x @Ey @Ex = @x @y j! Hx ; (3.5) j! Hy ; (3.6) j! Hz ; (3.7) and @Hz + j Hy = j!"Ex ; @y @Hz j Hx = j!"Ey ; @x @Hy @Hx = j!"Ez : @x @y (3.8) (3.9) (3.10) The above equations may be solved for the transverse components Hx ; Hy ; Ex ; Ey in terms of the longitudinal components Ez ; Hz as Hx = Hy = Ex = Ey = @Ez @Hz j !" ; kc2 @y @x j @Hz @Ez !" ; + kc2 @x @y @Hz @Ez j +! ; kc2 @x @y @Ez @Hz j +! ; kc2 @y @x (3.11) (3.12) (3.13) (3.14) where kc2 = k2 = !2 " 2 = Often we write the above as = 2 (3.15) 2 2 p k2 2 : kc2 ; (3.16) and once kc is found then is determined (since k is known). We will study three types of waves that commonly occur on transmission line and waveguiding structures. 1. TEM waves A transverse electromagnetic (TEM) wave has no longitudinal components, i.e., Ez = Hz = 0. From (3.11)–(3.14) this would result in E = H = 0, and so no …eld would be present. However, a TEM wave can exist. The solution of this paradox is that for a TEM wave it must be true that kc = ) 0; (3.17) = k= ! p "= 2 p such that TEM waves act like the waves we have studied previously (in Chapter 2 we had = ! LC = 2 ). In the text it is shown that TEM …elds are the same as the static …elds (zero frequency) that exist on the transmission line, and satisfy Laplace’s equation. 3.2. PARALLEL PLATE WAVEGUIDE 59 2. TE waves It turns out that a transmission line must have two or more conductors for a TEM waves to exist. However, this is a necessary but not su¢ cient condition. The existence of two or more conductors does not guarantee that a TEM wave will exist; microstrip is an example of a two-conductor line that does not support a TEM wave. Another wave type is a TE (transverse electric) wave, with Ez = 0, Hz 6= 0. In this case all transverse …elds may be found from (3.11)–(3.14) using the …eld component Hz , which satis…es @2 @2 @2 + 2 + 2 + k 2 Hz = 0, or 2 @x @y @z 2 2 @ @ 2 + 2 + k 2 Hz = 0; @x2 @y @2 @2 ) + 2 + kc2 Hz = 0; 2 @x @y (3.18) which is solved subject to the appropriate boundary conditions for a given physical structure, although will skip this computation. The result of that analysis, through, leads to the determination of kc , and using p = k 2 kc2 ; (3.19) to the determination of . 3. TM waves TM waves have Ez = 6 0, Hz = 0, and the transverse …eld components are found from Ez using (3.11)– (3.14), where Ez is a solution of @2 @2 + + kc2 Ez = 0: (3.20) @x2 @y 2 As with TE waves, kc is determined in the course of solving (3.20). leading to . In summary, we have TEM waves TE waves TM waves 3.2 Ez = 0 Ez = 0 Ez 6= 0 Parallel Plate Waveguide Hz = 0 Hz 6= 0 Hz = 0 = k p = pk 2 kc2 . = k 2 kc2 A parallel plate waveguide consists of two conducting plates having width w, separated by spacing d, and …lled with material characterized by "; , as shown below. y ε,µ z x 60 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES 1. TEM waves The parallel plate structure supports a TEM mode, which is the dominant mode (i.e., the mode that exists at the lowest frequencies of operation). The TEM mode is the most important mode from a practical standpoint. The analysis in the text shows that for the TEM mode, E (x; y; z) = H (x; y; z) y = x V0 e d V0 j z ZT EM d e ; ( = k) ; j z ; and that d Z0 = ; w = (3.21) ZT EM = r = r " ; (3.22) " independent of frequency. β 1 k 2. TM modes Parallel plate waveguides also support TE and TM modes. For TM modes, Ex Ey Ez Hx where r = Hy = Hz = 0; j n = y e j nz ; An cos kc d n = An sin y e j nz ; d j!" n y e j nz ; = An cos kc d (3.23) n 2 n kc = . ; n = 0; 1; 2; :::; d d For n = 0 the TM0 mode is really the TEM mode. For n > 0 two situations occur: n (a) If k > n d n d then n = k2 is real-valued and, therefore, e j nz is a purely propagating mode. (b) If k < then n = j n is imaginary-valued and, therefore, e purely attenuating mode. Thus, two regimes exist: k > kc = n d (3.24) (propagation), and k < kc = n d j nz =e j( j n )z =e nz is a (attenuation). Note that this is fundamentally di¤erent that a wave propagating through a lossy medium, where we obtain e j ze z; (3.25) i.e., propagation and attenuation occurring at the same time. For the TMn mode considered here, the p wave either propagates (at higher frequencies, where k = ! " > nd ) or does not (at lower frequencies, p where k = ! " < nd ). If the wave is purely attenuating, it is called evanescent, or cuto¤ . 3.2. PARALLEL PLATE WAVEGUIDE 61 The frequency at which a cuto¤ (non-propagating) mode begins to propagate is called the cuto¤ frequency, and is determined by n n kc = ) fc = p . (3.26) d 2d " Real β k β Imag. β TEM TM1 TM2 k1 k k2 As frequency increases, more modes begin to propagate. This is usually undesirable. Why? Note that vp = but that the speed of light is What does vp k, vp (3.27) ! : k c= Therefore, since ! (3.28) c! c imply? Consider the TM1 mode. 1 Ez = r k2 2 d ; = A1 sin y e j 8 d 1 < j( y=d = A e {z 2j : | (3.29) 1z 1 z) } plane wave in -y,+z |e j( y=d+ {z 1 z) 9 = } ;: plane wave in +y,+z 62 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES y d +y,+z wave -y,+z wave θ z With ( 2 1 ky = kz = d 1 = k sin ; (3.30) = k cos 2 + ( =d) = k 2 ) the phase velocity of each plane wave is vpP W = ! = c: k (3.31) The phase velocity of each plane wave in the z direction is vp = ! 1 = ! k cos = c : cos (3.32) Since jcos j 1, then vp c. The phase velocity being larger than c is simply due to observing the wave along one direction (the z direction) when the wave is propagating along another direction (the direction). constant phase wavefronts L θ d With L cos = d, and observer at d sees the wave arrive in T seconds. The observer calculates the wave velocity to as vp = d=T , but the wave actually travelled a distance L in time T , such that c = L=T . Since d > L, vp > c. An example of this is a water wave hitting a shoreline at an oblique angle. The wave “break” travels up the coast faster than the wave is moving. We’ll see later that the energy velocity is always c. 3. TE waves For TEn modes, Ey Ex Hy Hz = Hx = 0; n j! = An sin y e j nz ; kc d j n = An sin y e j nz ; kc d n = An cos y e j nz ; d (3.33) 3.2. PARALLEL PLATE WAVEGUIDE where n For n = r 63 n d k2 2 ; n = 1; 2; 3; :::; kc = n d . (3.34) 1 two situations occur, as before: (a) If k > n d then n j is real-valued and, therefore, e nz is a purely propagating mode. (b) If k < nd then n = j n is imaginary-valued and, therefore, e purely attenuating mode. Thus, two regimes exist: k > kc = n d (propagation), and k < kc = n d j nz =e j( j n )z =e nz is a (attenuation). Note that for both the TMn and TEn modes, n = fc = r n d k2 2d n p " 2 ; (3.35) : In summary, for the parallel plate waveguide, TEM, TMn , and TEn modes exist. For the TEM mode, frequency). =k =! p ", and no cuto¤ exists (the TEM mode propagates down to zero q 2 n For the TMn and TEn modes, n = k 2 (n = 0; 1; 2; ::: for TM modes, and n = 1; 2; 3; ::: for d TE modes). Two regimes exist, below cuto¤ (k < nd , pure imaginary), and above cuto¤ (k > nd , n pure real). The dividing point is k = kc = nd , leading to fc = 2dp ". An in…nite number of modes exist, although at any …nite frequency a …nite number of modes are propagating. Are parallel plate waveguides used in practice? Sometimes, but more often an unintentional parallel plate waveguide is formed by a circuit designer’s attempt to shield a circuit from interference, or to eliminate undesired radiation (emissions) from a circuit. In this case, on must be aware of the modes of the waveguide, and especially the cuto¤ frequencies of the allowed modes of propagation. interference signal undesired emission IC chips coupling via surface wave ε initial circuit layout 64 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES undesired emission IC chips coupling via surface wave ε shielded circuit layout Or, perhaps, adding a top cover didn’t cause problems at an operating frequency f0 < fc;n . Later, operating frequency is increased and the circuit doesn’t function properly because f0 > fc;n .(for some n) and waveguide modes begin to propagate. One solution is to decrease d (to increase fc;n ), or to put in some lossy dielectric material to absorb the emissions (however, energy loss from the circuit still occurs, perhaps leading to circuit malfunction). 3.3 Rectangular Waveguide A rectangular waveguide is a hollow (or dielectric …lled) metal pipe with rectangular cross-section. It can propagate TE and TM modes, but not TEM modes. y b ε,µ x a z The table below provides a summary of results for rectangular waveguide. 3.4. CIRCULAR WAVEGUIDE TEmn mode p ! " q 2 2 (m =a) + (n =b) q 2 k 2 kc;mn k kc;mn mn c;mn g;mn vp;mn Ez Hz Ex Hx Hy Z TMmn mode p ! " q 2 2 (m =a) + (n =b) q 2 k 2 kc;mn 2 =kc;mn 2 = mn != mn 0 Amn cos (m x=a) cos (n y=b) e j mn z j! n j mn z k2 b Amn cos (m x=a) sin (n y=b) e 2 =kc;mn 2 = mn != mn Bmn sin (m x=a) sin (n y=b) e j mn z 0 m j mn z a Bmn cos (m x=a) sin (n y=b) e j 2 kc;mn j n 2 kc;mn b Bmn sin (m x=a) cos (n j!"n 2 kc;mn b Bmn sin (m x=a) cos (n j!"m k2 a Bmn cos (m x=a) sin (n c;mn j! m 2 kc;mn a Amn sin (m j m 2 kc;mn a Amn sin (m j n k2 b Amn cos (m Ey 65 c;mn j x=a) cos (n y=b) e mn z x=a) cos (n y=b) e j mn z x=a) sin (n y=b) e ZT E = k = mn j mn z c;mn ZT M = mn y=b) e j mn z y=b) e j mn z y=b) e j mn z =k TE modes: r k2 m a r 2 2 n b m; n = 0; 1; 2; :::; (3.36) The …rst (dominant) mode is the TE10 mode, assuming a b. TM modes: r m 2 n 2 k2 ; m; n = 1; 2; 3; :::; mn = a b r 1 m 2 n 2 fc;mn = + : p 2 " a b (3.37) mn = fc;mn = 2 1 p " The …rst mode to propagate is the TM11 mode. 3.4 Circular Waveguide Omit 3.5 Coaxial Line TEM modes: m a 2 + ; n b 2 : 66 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES V=V0 a ε V=0 b Z0 = E = H = b a ln 2 . V0 e j z; ln (b=a) r " V0 e ln (b=a) (3.38) = k; j z (3.39) : The common value of 50 ohm for the characteristic impedance of many transmission lines comes from some attributes of coaxial lines. – Power handling capacity: The maximum power capacity of a coaxial line is limited by voltage breakdown, and is given by a2 Ed2 Pmax = ln (b=a) (3.40) where Ed is the …eld strength at breakdown. Solving for the value of b=a that maximizes Pmax leads to a characteristic impedance of 30 ohms, as follows: Holding b …xed, d Pmax = da where a2 b a 1 + 2a ln a =0 (3.41) = Ed2 = . This results in ln (b=a) = 1=2. Then, 1 Z0 = 2 r " ln b a ' 60 1 2 = 30 ohms (3.42) assuming an air-…lled line (long ago most coax was air-…lled, since the available dielectric had too much loss). – Attenuation: The attenuation of a coaxial cable due to …nite conductivity of the conductors is c = The equation obtained by minimizing c Rs 2 ln 1 a 1 b . (3.43) is b ln a which has solution b a b a =1+ b = 3:591: a b a (3.44) (3.45) 3.6. SURFACE WAVES ON A GROUNDED DIELECTRIC SLAB 67 The corresponding characteristic impedance is r 1 b Z0 = ln ' 60 ln (3:591) = 76:71 ohms. 2 " a (3.46) Therefore, 50 ohms represents a compromise between maximum power handling capability (30 ohms) and minimum attenuation (77 ohms), (30 + 76:71) =2 = 53:35. 50 ohms is also typically realized using reasonable material dimensions. Coaxial lines also support higher-order modes (TE and TM modes) (the cuto¤ frequency for the …rst higher-order modes can be obtained approximately as fco ' p c "r (a + b) (3.47) We will not consider them here, but it is important to note that above some frequency these modes will begin to propagate. Multi-mode propagation is usually to be avoided, since more than one mode carries power, potentially leading to increased dispersion (as will as input–output coupling e¢ ciency decreases). 3.6 Surface Waves on a Grounded Dielectric Slab A grounded dielectric slab is a dielectric sheet on a ground plane, and forms the backbone of printed circuit technology. x surface wave ε0 d εr z This waveguide can support TE and TM surface waves, that are bound to the vicinity of the surface (they decay exponentially vertically). These waveguide can be used to carry signals in the horizontal direction, and are, indeed, used for this purpose at very high frequencies (typically optical frequencies). At lower microwave frequencies, and considering typical slab thickness values used in printed circuit technology, the surface waves are not strongly bound to the surface, and so are not typically used as intentional waveguides. However, circuits printed on a ground slab will excite surface waves that may interfere with other circuits printed on the same substrate. circuit 1 circuit 2 coupling via surface wave ε 1. TM modes: The TM modes have …eld components Ez ; Ex ; and Hy . The propagation constant the solution of q q q 2 2 2 tan d = "r er k02 er k02 k02 is determined by (3.48) 68 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES and the cuto¤ frequencies from fc = nc p 2d "r 1 ; n = 0; 1; 2; ::: (3.49) The n = 0 mode (TM0 ) propagates down to zero frequency (although exactly at ! = 0 the …elds vanish). Thus, this mode is always present, and often leads to undesirable coupling. In general, the higher the frequency the more energy carried by surface waves, resulting in higher coupling. 2. TE modes: The TM modes have …eld components Hz ; Hx ; and Ey . The propagation constant the solution of q q q 2 2 2 cot d = er k02 er k02 k02 is determined by (3.50) and the cuto¤ frequencies from fc = (2n 1) c p ; 4d "r 1 n = 1; 2; 3; ::: (3.51) All TE modes have a low-frequency cuto¤. As discussed previously, surface waves can lead to signi…cant coupling between circuit elements. For example, consider a grounded slab of thickness d and permittivity "r "0 , and an in…nite line source carrying current I = I0 cos (!t) ; (3.52) as shown in the …gure below. x P ε0 d εr z 3.6. SURFACE WAVES ON A GROUNDED DIELECTRIC SLAB 69 This source produces only TE electromagnetic …elds. For both the source and the observation point P (x; z) placed on the dielectric interface, the total …eld is given by Z 1 p2 2 1 k0 x j z F ( )e e d (3.53) Ey = 2 1 where p2 2 p2 2 k1 coth k d p2 2 p 2 12 2 k2 + k coth k I0 0 1 1d p F( )= j!"0 2 2 k2 p with k1 = "r k0 . It can be shown that the …eld can be decomposed into three parts: p 1+ p k02 (3.54) (1) due to the direct radiation from the source (in the absence of the grounded slab), (2) due to interaction of the source and the grounded slab (e.g., multiple partial re‡ections), and 1. Ey 2. Ey 3. 2 (3 Ey due to surface waves excited in the slab. The origins of these …eld components are shown in the …gure below. x P E (1) E (2) ε0 d εr E The three …eld components are, for x = 0; kz e jkjzj Ey(1) ' E1 p ; jkzj (3) z 1, e jkjzj Ey(2) ' E2 q ; 3 jkzj Ey(3) ' E3 e j jzj (3.55) where E1;2;3 are constants. Therefore, the surface wave is the dominant contribution to the …eld, and does not decay with distance from the source (for an in…nite source). For "r = 2:25 and d = 1 cm, the TE surface waves are given as shown below. 1.5 TE1 1.4 β/ k0 TE3 1.3 TE5 1.2 1.1 1.0 0 10 20 30 40 f (GHz) 50 60 70 70 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES At approximately 6:7 GHz, 20:12 GHz, and 33:5 GHz, the …rst three TE surface waves begin to propagate, respectively. After those frequencies, a plot of jEy j vs. frequency shows spikes corresponding to surface wave excitation. 5 E y (a.u.) 4 3 2 1 0 3.7 0 10 20 f (GHz) 30 40 Stripline Stripline is formed by two ground planes with a conducting strip between the planes. w ground planes ε,µ b conducting strip Stripline supports a TEM mode. The analysis of the TEM (and the higher-order TE and TM modes) is di¢ cult; some results for the TEM modes are provided below. =k=! p " since TEM Z0 ' 30 b ; p "r we + 0:441b we b = w b 0; 0:35 w 2 b (3.56) (3.57) ; w b w b > 0:35; : < 0:35 To obtain strip width w for a given Z0 , w b = x = x; 0:85 30 p "r Z0 p 0:6 0:441: p e Z < 120; pr 0 ; x; er Z0 > 120 (3.58) 3.8. MICROSTRIP 3.8 71 Microstrip Microstrip transmission line is one of the most common types of transmission line extensively used in printedcircuit technology (printed using photolithographic techniques). It consists of a conductor of width w printed on a grounded dielectric slab. w conducting strip ground plane ε, µ b Microstrip does not support a TEM mode. In fact, microstrip doesn’t support TE or TM modes either, only hybrid modes (those with all six …eld components). However, the dominant mode is like, at su¢ ciently low frequencies, a TEM mode, prompting a quasi-TEM analysis. The details are included in the text, with some results summarized below. p p = k0 "e = ! 0 "0 "e ; (3.59) c ! =p ; vp = "e g where "e = = p 0 "e ; " r + 1 "r 1 1 q + 2 2 1+ If given line dimensions we obtain Z0 as ( 60 8b w ; 12b w 1 < "e < "r : w 4b ; 120 ; p w "e [ w d +1:393+0:667 ln( d +1:444)] p Z0 = "e ln + w b w b 1 1 ; and for a given characteristic impedance and dielectric constant, the required w=b ratio is ( 8eA w w b < 2; e2A h 2; i = "r 1 2 0:61 w b B 1 ln (2B 1) + 2"r ln (B 1) + 0:39 ; b > 2; "r (3.60) (3.61) (3.62) where A = B = r Z0 "r + 1 "r 1 + 60 2 "r + 1 377 p : 2Z0 "r 0:23 + 0:11 "r ; (3.63) Example 3.1 Design a microstrip line on a "r = 2:25 substrate of thickness b = 0:2 cm to have a characteristic impedance of 50 ohms. Solution: From the above equations, w=b = 3:037, so w = 0:6074 cm. Dispersion: 72 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES The previous formulas for Z0 and "e are valid for low frequencies. The e¤ects of frequency can be approximately accounted for using "ef f (f ) Z0 (f ) where F = p 4bf "r c p p 2 "r "e p + " ; e 1 + 4F 3=2 r "e "ef f (f ) 1 = Z0 ; "e 1 "ef f (f ) = 1 h w i2 + 1 + 2 log10 1 + 2 b 1 (3.64) (3.65) Plot of "ef f (f ) vs. f . "ef f (0) = "e . Losses: Losses on microstrip consist of three components: conductor loss, dielectric loss, and radiation loss. Usually conductor loss dominants over dielectric loss (at least for high-quality microwave substrates). Radiation loss include energy converted into surface waves, and energy radiated into space. Both e¤ects are most signi…cant at circuit discontinuities, and are minimized when b= 0 0:01 (e.g., at 2:4 GHz, 0 = 12:5 cm, and most practical substrates are electrically very thin). Approximate conductor and dielectric loss formulas as c where Rs = p ! 0 =2 ' Rs Np/m, wZ0 (3.66) is the surface resistivity of the conductor, and d = k0 "r ("e 1) tan p 2 "e ("r 1) Np/m, (3.67) where tan = Im f"g = Re f"g. In summary, we have considered several types of transmission lines and waveguiding structures. Some, like stripline, coaxial line, and parallel plate waveguides, support REM modes, as well as TE and TM modes. Others, like rectangular waveguide and grounded slab waveguides, only support TE and TM modes. Still others, like microstrip, only support hybrid modes. If a structure supports a TEM mode and is lossless, then that mode (there may be others too) can be modeled using the transmission line theory developed in Chapter 2. For these modes p p = ! " = ! LC; (3.68) p and Z0 = L=C follows from the …eld analysis in Section 2:2. 3.9. THE TRANSVERSE RESONANCE TECHNIQUE 73 If a structure does not support a TEM mode, then no unique de…nition of Z0 exists, although often only the ratio of impedance values is necessary. To examine actual high-frequency e¤ects on microstrip transmission lines, consider a line 10 cm long, 1:73 cm wide, on a d = 0:5 cm grounded dielectric slab having "r = 2:2, such that Z0 ' 50 ohms (by approximate analysis) at 5 GHz. The transmission parameter S21 in dB, as computed by Ansoft Designer SV, is approximately 0 dB for 1 f 10 GHz (it decreases from 0 dB at 1 GHz to 0:2 dB at 10 GHz. To compare with a more accurate analysis, the same structure was analyzed by Ansoft Ensemble SV, which performs an accurate (and time consuming) electromagnetic analysis. The line is fed and terminated by 50 ohm probes. The results is shown below. 10 cm microstrip line, probe fed and terminated. Full-wave electromagnetic analysis. w=1.7262 cm, d=0.5 cm, εr=2.2 0 S 21 (dB) -10 -20 -30 -40 1 2 3 4 5 6 7 8 9 10 f (GHz) In large part, the decrease in S21 at high frequency is due to the probes, which can’t be analyzed by the simple circuit-based analysis. For comparison, the simulation in Designer was virtually instantaneous, while the more accurate Ensemble simulation took about 20 minutes. 3.9 The Transverse Resonance Technique Skip 3.10 Wave Velocities and Dispersion p For sinusoidal steady-state situations we have c = 1= ", the speed of light in the material characterized by ; ", and vp = != . For a TEM wave vp = c, whereas for TE and TM modes these velocities di¤er (and generally vp c). In this case there is another velocity that is useful (and physically signi…cant) to describe the propagation of signals. Group Velocity: The group velocity, vg , is the velocity at which a narrow-band signal propagates. Considering a signal to be made up of many Fourier components at di¤erent frequencies, one can de…ne a phase velocity for each wave component. If is not a linear function of !, then vp = != is di¤erent for each Fourier component. As a result, the wave “breaks up” as it propagates down the line, resulting in dispersion. No single phase velocity describes the wave. The concept of group velocity overcomes this problem for narrow-band signals. 74 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES Let f (t) F (!) : : time-domain signal, frequency domain (Fourier transform) signal f (t) $ F (!) where the Fourier transform pair is F (!) f (t) = = Z 1 2 1 f (t) e 1 Z 1 j!t dt; (3.69) F (!) ej!t d!: 1 Modeling the system (in this case, the transmission line) as having transfer function Z (!), then Fin (ω) Fo (ω) Z (ω) transfer function of transmission line Fo (!) = Z (!) Fin (!) : (3.70) Question: What is the transfer function Z (!)? Consider a lossless, matched line, Z 0 ,v p ZL Z in z=-L where ZL = Z0 . In this case v (z) = v0+ e z=0 j z , such that v ( L) = v0+ ej v (0) = v0+ ; and so vout = Z (!) = e vin L ; j (!)L (3.71) : Therefore, the transfer function of a lossless, matched line of length L is e j (!)L . Question: What is Fin (!)? Fin (!) is the Fourier components of the signal fin (t). Example 3.2 If fin (t) is a rectangular pulse of width 2T0 , (3.72) 3.10. WAVE VELOCITIES AND DISPERSION 75 f(t) 1 t 2To Fin (!) = = Z Z 1 fin (t) e 1 2T0 e j!t j!t dt (3.73) j!T0 dt = 2T0 e 0 sin (!T0 ) : (!T0 ) 1 0.8 0.6 0.4 0.2 -10 0 -5 5 x 10 -0.2 Plot of sin( x) = ( x) vs. x. Therefore, the time-domain output is fo (t) = = = = If (!) = ! (linear in !; = p Z 1 2 Z 1 2 Z 1 2 Z 1 2 1 1 1 1 1 1 1 Fo (!) ej!t d! (3.74) Z (!) Fin (!) ej!t d! e j (!)L Fin (!) ej!t d! Fin (!) ej(!t (!)L) d!: 1 " for TEM waves), then Z 1 1 fo (t) = Fin (!) ej!(t 2 1 = fin (t L) , L) d! (3.75) which is an exact replica of fin (t) shifter by L to account for time delay in passing through the network. p p For example, if = ", then L = "L = L=c = T , the one-way transit time. 76 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES Therefore, lossless TEM lines are dispersionless. If (!) is not linear in !, (as for TE and TM waves), then Z 1 1 Fin (!) ej(!t fo (t) = 2 1 (!)L) d! =? (3.76) and fo (t) is not, in general, simply a replica of fin (t). If (!) is not linear in !, the dispersion will occur. To consider group velocity, consider a narrow-band signal s (t) = f (t) cos (! 0 t) = Re f (t) ej!0 t (3.77) as in an amplitude modulated waveform with carrier frequency ! 0 . the signal f (t) represents some information that is modulated with a high frequency carrier for transmission. Assume that the highest frequency component of f (t) is ! m , and that ! m ! 0 . The various Fourier transforms are Z 1 F (!) = f (t) e j!t dt; (3.78) 1 Z 1 Z 1 S (!) = s (t) e j!t dt = f (t) ej!0 t e j!t dt 1 1 Z 1 j(! 0 !)t = f (t) e dt 1 = F (! !0 ) : F (ω ) ω S (ω ) −ω 0 ω0 ω Now, consider the modulated signal to be the input to a transmission line or waveguide having transfer function Z (!) = e j z . Sin (!) = F (! ! 0 ) ; So (!) = F (! ! 0 ) e j z ; Z 1 1 so (t) = Re So (!) ej!t d! 2 1 Z 1 1 = Re F (! ! 0 ) e j z ej!t d! 2 1 Z !0 +!m 1 = Re F (! ! 0 ) ej(!t z) d!: 2 !0 !m (3.79) 3.10. WAVE VELOCITIES AND DISPERSION If ! m ! 0 , expand 77 (!) about ! 0 , (!) = (! 0 ) + ' + 0 0 0 d d! (! (! 1 d2 2 d! 2 !0 ) + !=! 0 (! 2 ! 0 ) + ::: (3.80) !=! 0 !0 ) : Then, so (t) = = = = 1 Re 2 Z ! 0 +! m ! 0 ) ej(!t F (! z) d! (3.81) !0 !m Z !0 +!m 0 1 F (! ! 0 ) ej (!t 0 z 0 (! !0 )z) d! Re 2 ! ! Z !0 m m 0 1 Re F (y) ej ((y+!0 )t 0 z 0 yz) dy (c.o.v. y = ! 2 !m Z !m 0 1 F (y) ej (t 0 z)y dy Re ej(!0 t 0 z) 2 !m !0 ) which is so (t) = n Re ej(!0 t 0 0z = f t 0 z) f t cos (! 0 t 0 0z o (3.82) 0 z) which is a time-shifted replica of the input fin (t). 0 The velocity of the envelope f t 0 z is the group velocity, 0 0z t d t dz 0 0z = constant, = 1 vg = (3.83) 0; 0 0; ) vg = 1 0 0 = d d! 1 : !=! 0 Waveguide velocities: p = k2 kc2 = r ! c 2 vp = ! c2 c ; v = = : g c2 ! k0 ! k0 c = ; vg < c < vp ; d d! = vp vg = kc2 ; k0 c c = c2 : k0 (3.84) 78 CHAPTER 3. TRANSMISSION LINES AND WAVEGUIDES Chapter 4 Microwave Network Analysis 4.1 Impedance and Equivalent Voltage and Currents Voltage and current can be uniquely de…ned for TEM transmission lines H + E H - v= Z ( ) E dl (4.1) (+) independent of path l from one conductor to another, and I I= H dl; (4.2) C+ where C + is any closed path enclosing the (+) conductor. With Z0 = v I we see that Z0 is uniquely de…ned. Voltage and current cannot be uniquely de…ned for a non-TEM line. As an example, consider a rectangular waveguide. 79 (4.3) 80 CHAPTER 4. MICROWAVE NETWORK ANALYSIS y b Ey1;0 = A sin where ZT1;0 E =k = 1;0 . x a 0 x e a j z Hx1;0 = ; Ey1;0 ZT1;0 E ; (4.4) Then, for the T E10 at z = 0, v= Z ( ) E dl = (+) Z b A sin 0 x x dy = Ab sin a a (4.5) and we see that voltage depends on position x. At x = 0 we obtain v = 0, and at x = a=2 we obtain v = Ab. Therefore, voltage associated with the T E10 mode is not unique. What can be done for non-TEM lines? Useful results are often obtained from the following considerations: 1. De…ne voltage and current for a particular mode such that voltage is proportional to the transverse electric …eld and current is proportional to the transverse magnetic …eld. 2. Voltages and currents should be de…ned so that their product gives the power ‡ow (which is uniquely de…ned) of the mode. 3. The ratio of voltage to current for a single travelling wave should be chosen equal to the characteristic impedance of the line. Not that this value is itself arbitrary. Proceeding with the rectangular waveguide example, let vc be the voltage evaluated along the center of the guide (at x = a=2); vc = Ab. Let Iz = Z a 0 A Hx dx = 1;0 ZT E Z0 = Z a 0 sin x a A dx = 2 : a ZT1;0 E vc Ab b 1;0 = a A = Z : Iz 2 a TE 2 Z 1;0 (4.6) (4.7) TE This is known as the voltage-current de…nition of Z0 . Other de…nitions: P ; Iz2 v2 Z0 = c ; P Z0 = power-current, power-voltage. (4.8) 4.1. IMPEDANCE AND EQUIVALENT VOLTAGE AND CURRENTS 81 Since 1 2 P = Z S Re fE H g dS (4.9) 2 ab jAj ; 4 ZT1;0 E = the power-current de…nition yields 2 ab jAj 4 Z 1;0 TE Z0 = 2 a A 1;0 ZT E 2 = 1b 4a 2 2 ZT1;0 E (4.10) and the power voltage yields 2 Z0 = (Ab) 2 ab jAj 4 Z 1;0 TE b = 4 ZT1;0 : a E (4.11) Note that each of the three de…nitions results in a di¤erent value for Z0 , but that all de…nitions result in the form b 1;0 Z ; (4.12) Z0 = a TE where is a constant. Thus, methods that only depend on the ratios of impedances will work (and give the same answer) regardless of how Z0 is de…ned. If one needs to match a waveguide (or other non-TEM line) to a structure that has a unique impedance (i.e., a TEM line), then the usual approach is to use the de…nition that gives the best agreement between experiment and theory. Often, for non-TEM lines, and for TEM lines as well, one needs to consider the electromagnetic …eld analysis of the structure (at least to some extent). Consider example 4.2 in the text, which shows the side view of a waveguide junction. ε0 y =b εr ε0 TE10 y =0 z z=0 a = "r = such that 3:485 cm, b = 1:580 cm, 2:56; f = 4:5 GHz, r a = = r k02 2 = 27:5 m a 1 for the air-…lled guide, TE10 mode, and d "r k02 2 a = 120:89 m 1 for the dielectric …lled guide. Only the TE10 mode propagates in either region. 82 CHAPTER 4. MICROWAVE NETWORK ANALYSIS Γ Z0a Z0b z=0 Z0a = k0 0 = 1292:1 ohms, a Z0d = = the same result for k d Z0d Z0d = 293:9 ohms, Z0a = + Z0a 0:629; would be obtained for any of the three de…nitions of Z0 . So, this is enough analysis. However, consider the juncture of two waveguides having di¤erent cross-sections but where b=a remains constant. Assuming the T E10 mode propagates in both sections, Z0 is the same in each section! This would imply = 0 which doesn’t make sense. So, a one-mode transmission line analysis is not enough here. Physically, the geometrical change at z = 0 excites an in…nite number of higher-order modes. Those that propagate (i.e., those that are above cuto¤) will travel away from the discontinuity. Below cuto¤ modes will decay away from the z = 0, but will store energy in the vicinity of z = 0. This stored energy can be a very important part of the analysis. In the end, usually a transmission line mode can be obtained, with lumped reactances modeling the stored energy e¤ects (see text Section 4.6). 4.1. IMPEDANCE AND EQUIVALENT VOLTAGE AND CURRENTS 83 84 CHAPTER 4. MICROWAVE NETWORK ANALYSIS Chapter 5 Impedance Matching and Tuning The purpose of a matching network is to provide a good match between a given transmission line and a given load. In this chapter we will consider ways to design matching networks with consideration to matching network bandwidth, physical implementation of the network, and the ability to adjust the network. Several types Zg Vg matching Z 0 ,v p + - ZL network Z in Transformer Matching (not in text book) A transformer can be used to match a real-valued load RL to a line having characteristic impedance Z0 by appropriate choice of the turns ratio. Zg n1:n2 i1 i2 + + Vg V1 + V2 ZL - Z1 Z2 - - Assume Zg = Z0 and ZL = RL . Assuming an ideal transformer, V2 n2 = ; V1 n1 i2 n1 = : i1 n2 (5.1) We want Z1 = Z0 and Z2 = RL , where Z1 = V1 =i1 , Z2 = V2 =i2 . Then Z1 = V1 = i1 n1 n2 V 2 n2 n1 i2 = n1 n2 85 2 V2 = i2 n1 n2 2 Z2 : (5.2) 86 CHAPTER 5. IMPEDANCE MATCHING AND TUNING Recalling that we want Z1 = Z0 and Z2 = RL we obtain Z1 = Z0 = 2 n1 n2 Z2 = and so Z0 = n1 n2 2 n1 n2 RL ; (5.3) 2 RL : (5.4) So, given Z0 and RL , a transformer with turns ratio n1 = n2 r Z0 RL (5.5) provides a perfect impedance match. Example 5.1 Match a 300 ohm line to a 75 ohm load. Solution: Choose r r Z0 300 n1 = 2: = = n2 RL 75 (5.6) This technique works very well at low frequencies, where a transformer can be constructed to operate in a somewhat ideal fashion. The common 300 ohm–to–75 ohm adapter used in many television systems utilizes this technique. The match is frequency independent over the band of frequencies for which the transformer exhibits nearly ideal behavior. 5.1 Matching with Lumped Elements (L Networks) jX (a) Z Z jB 0 L A jX (b) jB Z L A By correct choice of X and B, a perfect match from Z0 to complex-valued ZL can be achieved (at one frequency). If ZL =Z0 is inside the unity circle on the Smith chart, choose design (a), else choose design (b). As in all matching network designs, the goal is to reach the center of the Smith chart. The procedure is best explained with an example. 5.1. MATCHING WITH LUMPED ELEMENTS (L NETWORKS) 87 Example 5.2 Match a 50 ohm line to a 100 + j100 ohm load using an L network. Solution: ZL =Z0 = (100 + j100) =50 = 2 + j2; choose design (a). From the Smith chart, B = 0:19 ) inductor, 1 0:19 = = !L 50 ) !L = 263:16; Binductor X Xcap = 1:7 ) capacitor, 1 = = 1:7 (50) !C ) !C = 0:0118: (5.7) (5.8) Equations are presented in the text to solve this problem as well. For design (a), with ZL = RL + jXL , p p 2 + X2 XL RL =Z0 RL Z0 RL L B = ; (5.9) 2 2 RL + X L 1 XL Z0 Z0 X = + : B RL BRL 88 CHAPTER 5. IMPEDANCE MATCHING AND TUNING For the previous example, (5.9) yields B X = = 0:0038 ) !L = 263:159; 86:6 ) !C = 0:0115: Example 5.3 Using the calculated values of !L and !C, pick L and C for a perfect match at f = 30 GHz, and plot the re‡ection coe¢ cient vs. frequency. 1.0 0.8 ρ 0.6 0.4 0.2 0.0 0 5 10 15 20 25 30 35 40 45 f (GHz) 5.1.1 Lumped Elements Lumped (R; L, and C) elements can be realized at high frequencies if the size of the element is small compared to . Unfortunately, their performance is usually quite non-ideal. For instance, elements may (and often do) exhibit stray/parasitic capacitance, inductance, and resistance e¤ects. Elements often have a non-negligible fringing …eld, and may resonant at certain frequencies (depending on the physical structure of the element). 5.2 Single-Stub Tuning (single lumped element also) 5.2. SINGLE-STUB TUNING 89 jX Z Z 0 1 Z L L Given Z0 and (complex) ZL , the goal is to choose X and (Z1 ; L) such that a perfect match is obtained at a given frequency. Often Z1 is chosen as Z0 for convenience. The procedure will be illustrated by example. Example 5.4 Series matching: jX Z Z 0 Z Z in 1 Z L L A Assume Z0 = Z1 = 50 ohms, ZL = 25 + j30 ohms. Then, Z L = ZL =Z0 = 0:5 + j0:6. We want Zin = 50 ohms. 1. Choose L such that Re Z A = 1. 2. Choose X such that Im fZin = ZA + jXg = 0 (at which point we are at the center of the Smith chart. Proceeding as above, 1. Start at Z L on Smith chart and rotate to either (a) Point E, Z A = 1 + j1:1 and L = 0:065 , or (b) Point F, Z A = 1 j1:1 and L = 0:235 . (a) From Point E, add a series capacitor such that X Cap = 1:1 ) 1 1 = !Ccap 50 1:1 ) !Ccap = 0:01818; or, (b) From Point F, add a series inductor such that X ind = 1:1 ) !Lind 1 = 1:1 ) !Lind = 55: 50 Therefore, two possible designs are !Ccap = 0:01818; L = 0:065 and !Lind = 55; L = 0:235 : The …rst design is better, since L is shorter. Note that 1(a) ! 2(a), and 1(b) ! 2(b). 90 CHAPTER 5. IMPEDANCE MATCHING AND TUNING If, for example, f = 2 GHz ( = 15 cm), !Ccap !Lind = 0:01818 ) Ccap = 1:45 pF, L = 0:975 cm, = 55 ) Lind = 4:38 nH, L = 3:53 cm. 1.0 Inductor 0.8 0.6 ρ Capacitor 0.4 0.2 0.0 1.0 1.5 2.0 2.5 3.0 f (GHz) 3.5 4.0 4.5 5.0 5.2. SINGLE-STUB TUNING 91 92 CHAPTER 5. IMPEDANCE MATCHING AND TUNING Design using non-lumped elements: short-circuited or open-circuited stubs: Usually short-circuited stubs are preferable to open-circuited stubs, since they don’t radiate. If using a stub rather than a lumped element, the design proceeds as described above as far as the determination of X and L. To generate a certain X, one chooses a stub having the appropriate length Ls using the Smith chart. For example, to generate X = 1:1 (assuming Z0;stub = Z0 = 50 ohms) for a short-circuited stub, start at Z = 0 on the Smith chart (corresponding to a short circuit) and rotate toward the generator to X = 1:1 (Point 1). The corresponding length is Ls = 0:368 (Ls = 5:52 cm at 2 GHz). To generate X = 1:1 (assuming Z0;stub = Z0 = 50 ohms) for a short-circuited stub, start at Z = 0 on the Smith chart and rotate toward the generator to X = 1:1 (Point 2). The corresponding length is Ls = 0:132 (Ls = 1:98 cm at 2 GHz). 5.2. SINGLE-STUB TUNING 93 There are four possible combinations if we assume Z0;stub = Z0 and if open-circuited stubs are acceptable. The design resulting in the shortest stub length should generally be chosen, everything else being equal. Stubs with Z0;stub 6= Z0 : If we choose stubs having characteristic impedances di¤erent than Z0 of the line, we can often obtain shorter stubs. For example, if the stub needs to provide X = 1:1, and Z0;stub = 100 ohms, then for a short-circuited stub 1:1 (50) = 0:55 ) Ls = 0:08 ( = 1:2 cm at 2 GHz). 100 High characteristic impedance short-circuited stubs are good for implementing inductive reactance. Try to implement X = 1:1 with Z0;stub = 10 ohms and a shorted line: 1:1 (50) = 10 5:5 ) Ls = 0:279 ( = 4:128 cm at 2 GHz). Low characteristic impedance short-circuited stubs are good for implementing capacitive reactance. 94 CHAPTER 5. IMPEDANCE MATCHING AND TUNING All of the previous steps were for series stubs. Another approach is to use parallel (shunt) stubs, or shunt lumped elements. Z Z jB 0 Y in YA 1 Z L L The procedure is similar to that for series stubs, except you begin at Y L rather than Z L . Note: The procedure using lumped elements and stubs is identical through identifying the required value of X (or B). After that point, we either determine !Lind or !Ccap , or for a stub design we determine the required stub length. For all of the designs, it is convenient to use the Smith chart. The Smith chart also leads to better insight into the problem, compared to using design equations. Example 5.5 A lossless 200 ohm line is connected to a 100 j150 ohm load. Design 1. a single short-circuited shunt stub matching network (Z0;stub = 200 ohms), 5.2. SINGLE-STUB TUNING 95 2. a single open-circuited shunt stub matching network (Z0;stub = 200 ohms), and 3. a lumped element shunt matching network. For the stub designs, use the shortest stub lengths possible. Solution: Ls Z Z 0 Z 0 Y in Z 0 L L 1. ZL = YL = 100 j150 = 0:5 200 0:62 + j0:92: j0:75; Enter the Smith chart at Y L and rotate along the constant SW R circle towards the generator until you intersect the G = 1 circle. Y A = 1 + j1:3 Y B = 1 j1:3 ) LA = 0:036 ) LB = 0:194 stub needs to provide j1:3 stub needs to provide +j1:3 ) LA ) LB s = 0:105 s = 0:396 Choose Design A, LA = 0:036 ; LA s = 0:105 . 96 CHAPTER 5. IMPEDANCE MATCHING AND TUNING 2. Same as (1) up through Y A = 1 + j1:3 ) LA = 0:036 stub needs to provide j1:3 Y B = 1 j1:3 ) LB = 0:194 stub needs to provide +j1:3 B For an open-circuited stub, LA s = 0:354 and Ls = 0:146 . Choose Design B, LB = 0:194 ; LB s = 0:146 . 5.2. SINGLE-STUB TUNING 97 Ls Z Z 0 Z 0 Y in ZL = 100 Z 0 L j150 ohms, Z0 = 200 ohms. 3. Same as (1) up through Y A = 1 + j1:3 Y B = 1 j1:3 ) LA = 0:036 ) LB = 0:194 In this case the lumped element must be chosen to supply B A = j1:3 B B = +j1:3 j 1:3 )choose inductor, !Lind = 200 ; )choose capacitor, !Ccap = !Lind = 153:85 !Ccap = 0:0065: 1:3 200 ; L 98 CHAPTER 5. IMPEDANCE MATCHING AND TUNING To compare designs, assume f0 = 900 MHz ( = 33:33 cm). Then, 1. short-circuited stubs: LA = 0:036 = 1:20 cm LA s = 0:105 = 3:50 cm LB = 0:194 = 6:466 cm LB s = 0:396 = 13:20 cm 2. open-circuit stubs: LA = 0:036 = 1:20 cm LA s = 0:354 = 11:80 cm LB = 0:194 = 6:466 cm LB s = 0:146 = 4:867 cm 3. lumped elements LA = 0:036 = 1:20 cm Lind = 2 153:85 900 106 = 27:21 nH LB = 0:194 = 6:466 cm Ccap = 2 0:0065 900 106 = 1:15 pF Short-Circuit Stub 1.0 0.8 0.6 ρ B 0.4 0.2 0.0 0.6 A 0.7 0.8 0.9 1.0 1.1 1.2 1.1 1.2 f (GHz) Open-Circuit Stub 1.0 0.8 B 0.6 ρ A 0.4 0.2 0.0 0.6 0.7 0.8 0.9 f (GHz) 1.0 5.3. DOUBLE-STUB TUNING 99 Lumped Elements 1.0 0.8 ρ 0.6 B 0.4 0.2 A 0.0 0.6 0.7 0.8 0.9 1.0 1.1 1.2 f (GHz) 5.3 Double-Stub Tuning (double lumped element also) Single-stub matching can match any load impedance, but (L; Ls ) must be chosen correctly. Since it is often hard to have a stub that can be moved along a transmission line (if, say, ZL or operating frequency changed), it would be useful to …x the distance from the load to the stub(s), and simply choose the correct stub length. The double-stub tuner allows this to be done, and changes in ZL or operating frequency can be accommodated by changing Ls1 and Ls2 , with L1;2 …xed. Note: the text book places the …rst stub over the load impedance. The method shown here is more ‡exible, since sometimes the load terminals are not easily accessible. L s2 Z L s1 Z 0 A Y in 0 Z L B L2 L1 Procedure: 1. Mark Y L on the Smith chart and draw constant SW R circle. 2. Draw G = 1 circle rotated by L2 (Smith chart on the next page shows L2 = =8). 3. Move from Y L a distance L1 on the constant SW R circle (toward generator) to get to Y B + . 4. Add susceptance (moving on constant G circle) to get to intersection with rotated G = 1 circle, Point Y B . The added susceptance comes form choosing Ls1 correctly. 5. Draw a new constant SW R circle using Point Y B . Rotate on new SW R circle a distance L2 (you will intersect the non-rotated G = 1 circle at Point Y A+ ). 6. Add susceptance (moving along G = 1 circle to get to the center of the Smith chart, Y A . 100 CHAPTER 5. IMPEDANCE MATCHING AND TUNING Example 5.6 For the double stub tuner shown below, …nd the shortest lengths of Ls1 and Ls2 for a match if ZL = 400 + j200 ohms and Z0 = 200 ohms. Assume L1 = 3 =16 and L2 = =8. Solution: L s2 Z L s1 Z 0 A Y in 1. Enter Smith chart at Y L . B L2 ZL = YL = 0 L1 400 + j200 = 2 + j1; 200 1 = 0:4 j0:2: ZL Z L 5.3. DOUBLE-STUB TUNING 101 2. Rotate on constant SW R circle 3 =16. 3. Add susceptance Y = j0:925 ti intersect the rotated G = 1 circle. 4. Rotate on new SW R circle to intersect non-rotated G = 1 circle (this is the same as de-rotating the point on the rotated G = 1 circle). 5. Add susceptance Y = j0:17 to get to the center of the Smith chart. Now we know we need to add Y = j0:925 at Point B, and Y = values by choosing Ls1 and Ls2 correctly. j0:17 at Point A. We obtain these 6. Rotate from Y = 1 Z = 0 to obtain Y = j0:925: ) Ls1 = 0:131 : 7. Rotate from Y = 1 Z = 0 to obtain Y = j0:17: ) Ls2 = 0:223 : Note: We could have used lumped inductors to provide the Y values. Also, at Point 2 on the Smith chart we could have added susceptance to intersect the rotated G = 1 circle at a di¤ erent point. 102 CHAPTER 5. IMPEDANCE MATCHING AND TUNING 1.0 0.8 ρ 0.6 0.4 0.2 0.0 20 22 24 26 28 30 32 34 36 38 40 f (GHz) Example 5.7 A 100 + j100 ohm load is to be matched to a 50 ohm line using a double stub tuner as shown below. Determine the stub lengths to provide a patch, and …nd the forbidden zone of impedance/admittance values. L1 = 0:4 , L2 = 3 =8. 5.3. DOUBLE-STUB TUNING 103 Solution: L s2 Z L s1 Z 0 A L2 = YL = Z L B Y in ZL 0 L1 100 + j100 = 2 + j2; 50 0:25 j0:25: The stub having length Ls1 needs to provide Y = j1:3 Ls2 needs to provide Y = j0:6. From the Smith chart, Ls1 Ls2 = = 1:4 + B = 0:1459 ; 0:086 : :1 ) B = 1:3 . The stub having length 104 CHAPTER 5. IMPEDANCE MATCHING AND TUNING 1.0 0.8 ρ 0.6 0.4 0.2 0.0 20 22 24 26 28 30 f (GHz) 32 34 36 38 40 5.4. THE QUARTER-WAVE TRANSFORMER 5.4 105 The Quarter-Wave Transformer In this section we will consider the bandwidth performance of a single quarter-wave transformer. In subsequent sections we will determine how to design multi-section quarter-wave transforms that can be used to achieve a broadband match. Z0 RL Z1 Z in λ/4 p We know that if Z1 = Z0 RL the above quarter-wave transformer provides a perfect match at the frequency where L = =4. To determine the behavior of the transformer at other frequencies, consider the expression RL + jZ1 tan L Zin = Z1 : Z1 + jRL tan L Let = L = electrical length, such that 2 0 L= (5.10) 4 such that 2 0 = (5.11) L= 2 0 4 at the design frequency f0 (c = 0 f0 ). Then, = in (using Z1 = p (5.12) Z0 RL ). By further manipulations, it is shown that j If we assume f ' f0 , then in j =h 1 1 + 4Z0 RL = (RL ' =2 and sec2 j where Zin Z0 RL Z0 p = Zin + Z0 RL + Z0 + j2 tan Z0 RL in j = 2 Z0 ) sec2 1, resulting in jRL Z0 j p jcos j = 2 Z0 RL i1=2 : jcos j (5.14) is a constant. If RL does not vary with frequency, |Γ| Γm θ=β L θm π/2 π− θ m π (5.13) 106 CHAPTER 5. IMPEDANCE MATCHING AND TUNING where 2 = L= 0 4 0 = 2 | f = : 2 f0 {z } (5.15) assum ing TEM m o de If we have a value bandwidth of m for the maximum re‡ection coe¢ cient that can be tolerated, then we obtain a usable (5.16) m : 2 The approximation f ' f0 was made to see where the characteristic “vee”shape in the above plot comes from. We can determine a more useful expression for bandwidth by equating m with the exact expression for j in j, =2 j mj = 1 j h mj m m i1=2 ; (5.17) m 2 : m m, cos Since sec2 Z0 ) 4Z0 RL 1+ = Solving for cos 2 1 + 4Z0 RL = (RL 2 2 sec (RL Z0 ) p 2 Z0 RL 1 1+ (RL Z0 ) cos = 2 1 m = m = p 1 cos m 1 p 2 RL Z0 ; 2 jRL Z0 j m ! p 2 RL Z0 m p : 2 jRL Z0 j 1 m (5.18) = ( =2) (fm =f0 ), the frequency at the lower band edge is fm = 2 m f0 : (5.19) If we de…ne fractional bandwidth as f 2 (f0 fm ) = =2 f0 f0 2fm =2 f0 we obtain f f0 =2 4 m 4 =2 cos 1 p 1 m 4 m (5.20) ! p 2 RL Z0 : 2 jRL Z0 j m (5.21) The above analysis assumes TEM lines, and ignores the e¤ects of reactance associated with discontinuities of the transmission line when there is a step change in line width. The latter e¤ect can be compensated for by making a small adjustment in the length of the matching section. Example 5.8 Design a quarter-wave transformer to match a 75 ohm line to a 30 ohm load at 1 GHz. Determine the percent bandwidth for which SW R 2. Solution: Since at f0 = 1 GHz 0 = 30 cm, L = Z1 For SW R = 2, = 30 = 7:5 cm, 4 (75) (30) = 47:43 ohms. 0 4 p m = = SW R 1 1 = SW R + 1 3 5.5. THE THEORY OF SMALL REFLECTIONS 107 and so f f0 = 2 = 2 = 2 ! p 2 RL Z0 p cos 2 jRL Z0 j 1 m 0 1 p 2 (75) (30) A 4 1=3 cos 1 @ q 75j 2 j30 1 (1=3) 4 m 1 4 1 cos (0:7454) = 2 4 0:7297 = 1:071; or 107% bandwidth. The means the band extends from f0 f = 0:465 GHz 2 f0 + f = 1:535 GHz. 2 to Note: for SW R < 1:1, as desired SW R #, For a …xed m, m m = 0:0476 and f =f0 = 0:128 (12:8%), or from 0:936 GHz to 1:064 GHz. So, #, and BW #. the larger the ratio RL =Z0 (or Z0 =RL ), the smaller the bandwidth, as shown below. |Γ| ZL = 4, 0.25 Z0 ZL = 2, 0.5 Z0 1 5.5 The Theory of Small Re‡ections The theory of small re‡ections will facilitate the development of approximate formulas for the design of multi-section quarter-wave transformers, which will be shown to be relatively broadband. I. Single Section Transformer: 108 CHAPTER 5. IMPEDANCE MATCHING AND TUNING T 21 T 12 Z1 Z2 Γ Γ1 RL Γ2 Γ3 T 21 1 Γ1 e e -j θ -j θ Γ3 T 12 Γ2 e e -j θ -j θ Γ3 T 12 Γ2 where have 1 = T21 = Z2 Z1 ; 2 = 1; Z2 + Z1 1 + 1 ; T12 = 1 + 2 ; is the overall re‡ection coe¢ cient and i, = 3 ZL Z2 ; ZL + Z2 (5.22) i = 1; 2; 3, are the interfacial re‡ection coe¢ cients. We = 1 + T12 T21 3e j2 = 1 + T12 T21 3e j2 + T12 T21 1 X 2 j4 3 2e n n j2n 2 3e + ::: (5.23) : n=0 Using the geometric series result 1 X xn = n=0 leads to x for jxj < 1 T12 T21 3 e 1 2 3e j2 + e 1 3 : 1 + 1 3 e j2 = 1 = If the impedance do not di¤er greatly, then and so 1 1 1 + and ' 1 3 + j2 j2 : (5.25) j2 are small, and so 3e (5.24) 1 3 is very small compared to 1 (5.26) The above expression shows that is dominated by the …rst re‡ection from the initial discontinuity between Z1 and Z2 , 1 , and the …rst re‡ection from the discontinuity between Z2 and ZL . II. Multi-Section Transformers: 5.6. BINOMIAL MULTISECTION MATCHING TRANSFORMERS Now consider the N 109 section transform shown below, θ θ Z1 Z0 θ ... Z2 ZN ZL Γ Γ0 Γ1 Γ2 ΓN-1 ΓN where we assume that ZL is real-valued. We have 0 = Z1 Z0 ; Z1 + Z0 n = Zn+1 Zn ; Zn+1 + Zn Assume that Zn increases or decreases monotonically. Then n; n n N n = ZL ZN : ZL + ZN (5.27) is real-valued and has the same sign for all > 0 for ZL =Z0 > 1; < 0 for ZL =Z0 < 1; (5.28) and ( )' 0 + 1e j2 + 2e j4 + ::: + Ne j2N (5.29) from considerations of the single-section transformer. In some cases it is desirable to simplify further. Assume that the transformer is made symmetrical, in the sense 0 = N , 1 = N 1 , etc. (this does not imply that the Zn ’s are symmetrical). Then, n h i h i jN ( ) ' e jN + e jN + 1 ej(N 2) + e j(N 2) + 2 ej(N 4) + e j(N 4) 0 e ) N 1 ej + e j ; N odd 2 ; (5.30) + ::: + N ; N even 2 leading to ( ) ' 2e jN +::: + f 0 cos N + N 1 2 1 2 N 2 cos ; ; 1 cos (N N odd N even 2) + ) 2 cos (N 4) + ::: + n ; cos (N 2n) (5.31) which is seen to be a …nite (truncated) Fourier cosine series for ( ). Since a Fourier cosine series can represent any smooth function , we can synthesize any desired re‡ection coe¢ cient response as a function of frequency by proper choice of the n values, and by using enough sections (some limitations are imposed by the Bode-Fano criteria as discussed later). Two choices of desirable functions to implement are presented next, in each case leading to a method to choose the n values in (5.29) (or (5.30) or (5.31)). 5.6 Binomial Multisection Matching Transformers A desirable function to implement for matching purposes is given by g( ) = A 1+e j2 N : (5.32) 110 CHAPTER 5. IMPEDANCE MATCHING AND TUNING The absolute value is j jg ( )j = jAj e = jAj e ej + e jN ej + e = jAj ej + e = N j N 2N jAj jcos j ; N j (5.33) N j N = jAj j2 cos j which is plotted below. (n) The function (5.32) has the property that at = =2, jf ( )j = 0, n = 1; 2; :::; N 1, where g (n) indicates the nth derivative of g. Therefore, the response is as ‡at as possible near the center frequency f0 (i.e., at = 0 = =2). Such a response is known as maximally ‡at. So far, (5.32) is merely a function with a nice ‡atness property. To obtain a multisection transformer that behaves like (5.32), we equate the actual (approximate) input re‡ection coe¢ cient of the multisection transformer, (5.29) (or (5.30) or (5.31)), and (5.32), in order to determine what n values will lead to the response (5.32). Since (5.29) is in the form of a series, in order to equate (5.29) (or (5.30) or (5.31)) and (5.32) we use the binomial expansion N X N (1 + x) = CnN xn ; (5.34) n=0 where the binomial coe¢ cients are CnN = (N N! ; n)!n! (5.35) to convert (5.32) to series form, leading to g( ) = A 1+e = A N X j2 CnN e N j2n (5.36) : (5.37) n=0 Equating (5.37) with (5.29) leads to g( ) = A N X n=0 CnN e j2n = ( )' 0 + 1e j2 + 2e j4 + ::: + Ne j2N : (5.38) 5.6. BINOMIAL MULTISECTION MATCHING TRANSFORMERS 111 The constant A can be determined by letting f ! 0 ( = L = 0), ZL Z0 ; ZL + Z0 ZL Z0 ) A=2 N : ZL + Z0 g (0) = A2N = (0) = (5.39) (5.40) Equating nth terms in (5.38) results in n = ACnN = 2 ZL Z0 N C ; ZL + Z0 n N (5.41) the sought-after formula for choosing n values to yield a response that is equal to the maximally-‡at N response (5.32)1 . Since CnN = CN n , then n = N n and the transformer turns out to be symmetrical. The …nal step is determining the desired impedance values. Since 0 = Z1 Z0 ; Z1 + Z0 n = Zn+1 Zn ; Zn+1 + Zn N = ZL ZN ZL + ZN (5.42) then, for example, n = Zn+1 Zn =2 Zn+1 + Zn N ZL Z0 N C ZL + Z0 n (5.43) which can be solved for Zn+1 in terms of Zn as 1+2 N ZL Z0 N ZL +Z0 Cn 1 N ZL Z0 C N ZL +Z0 n Zn+1 = Zn 2 : (5.44) One can start with n = 0 (Z0 ) to …nd Z1 , etc. Interestingly, while (5.44) follows exactly from (5.43), making the following approximation to (5.43) leads to better results. Since x 1 2 + x+1 3 x 1 '2 x+1 ln (x) = 2 for x ' 1, then x 1 x+1 3 + ::: (5.45) Zn+1 =Zn 1 1 Zn+1 Zn+1 Zn = ' ln Zn+1 + Zn Zn+1 =Zn + 1 2 Zn (5.46) for Zn+1 ' Zn (this assumption was used in developing the theory of small re‡ections, and so the approximation in (5.46) is consistent with the approximations leading to (5.29), leading to a self-consistent formula that provides accurate results). Therefore, making the same approximations on both sides of (5.43) leads to n ' Zn+1 Zn = ln Zn+1 = ln 1 Zn+1 ln '2 2 Zn ZL 2 N CnN ln ; Z0 ln Zn + 2 N N ZL Z0 N C '2 ZL + Z0 n CnN ln N CnN 1 ZL ln ; 2 Z0 (5.47) (5.48) ZL ; Z0 (5.49) resulting in Zn+1 = eln Zn +2 1 Actually, N N Cn ln ZL Z0 since (5.29) is an approximate formula, then the resulting response is approximately that of (5.32). (5.50) 112 CHAPTER 5. IMPEDANCE MATCHING AND TUNING which is generally valid for obtain 1 2 Z0 ZL 2Z0 . Bandwidth can be determined as follows. From (5.32) we j when 0 m mj = 2N jAj jcos N = 2 jAj cos N N mj (5.51) m =2. Therefore, m = cos 1 = cos 1 " " 1 N j mj jAj 2N 1 2 j # # 1 N mj jAj (5.52) : From (5.20) we have f f0 2 (f0 = fm ) 4 =2 m " 4 1 j mj cos 1 2 jAj (5.53) f0 = 2 where # (5.54) 1 ZL ln 2 Z0 N A=2 1 N (5.55) Example 5.9 Design a three section binomial transformer to match a 50 ohm load to a 100 ohm line. Calculate the bandwidth for j m j = 0:05. Solution: = eln Zn +2 Zn+1 ln Zn +2 = e and C03 = 3! = 1; 3!0! C13 = N 3 N Cn ln 3 Cn ln 3! = 3; 2!1! ZL Z0 50 100 (5.56) ; C23 = (5.57) 3! = 3. 1!2! Therefore, Z1 = eln Z0 +2 3 C03 ln 50 100 Z2 ln Z1 +2 3 C13 ln 50 100 ln Z2 +2 3 C23 ln 50 100 Z3 = e = e π/2 100 91.7 = 91:70 ohms, (5.58) = 70:71 ohms, (5.59) = 54:53 ohms. (5.60) π/2 π/2 70.71 54.53 Γ0 Γ1 Γ2 Γ3 -0.043 -0.129 -0.129 -0.043 Z L = 50 5.6. BINOMIAL MULTISECTION MATCHING TRANSFORMERS 113 To compute the bandwidth, f 4 =2 f0 1 cos " j 1 2 mj 1 N jAj # ; (5.61) and, with N A=2 and m 1 ZL ln =2 2 Z0 31 2 ln 50 = 100 0:0433 = 0:05, f f0 =2 4 cos 1 " 1 2 A 1+e 1 3 0:05 0:0433 j2 (f ) N # = 0:703; 70:3% vs. f =f0 : 0.30 0.25 ρ 0.20 0.15 0.10 0.05 0.00 10 14 18 22 26 30 34 38 42 46 50 f (GHz) Example 5.10 Design an N section binomial transformer that needs to operate from 1:2 GHz to 2:8 GHz with j m j 0:05. Assume ZL = 100 ohms, Z0 = 50 ohms. Solution: In order to determine the Zn values we need to …rst determine how many sections (N ) are 114 CHAPTER 5. IMPEDANCE MATCHING AND TUNING required. f 2 (f0 fm ) 2 (2 1:2) = = = 0:8; f0 f0 2 1 ZL 1 100 A = 2 N ln = 2 N ln = 2 N (0:3466) 2 Z0 2 50 " 1 # f 4 j mj N 1 1 =2 cos f0 2 jAj " 1 # N 4 0:05 1 1 =2 = 0:8: cos 2 2 N (0:3466 (5.62) (5.63) (5.64) (5.65) Solving for N we obtain N = 3:64 ) N = 4. With N determined, the design proceeds as in the previous example. C04 = 1; C14 = 4; C24 = 6; Z1 = eln Z0 +2 4 C04 ln 100 50 Z2 ln Z1 +2 4 C14 ln 100 50 ln Z2 +2 4 C24 ln 100 50 ln Z3 +2 4 C34 ln 100 50 Z3 Z4 50 = e = e = e π/2 π/2 π/2 52.2 62.1 80.5 C34 = 4; = 52:214; = 62:093; = 80:525; = 95:76: π/2 95.8 Z L = 100 0.30 0.25 ρ 0.20 0.15 0.10 0.05 0.00 10 14 18 22 26 30 34 f (GHz) Note: Pascal’s triangle is also useful for evaluating CnN , 38 42 46 50 5.7. CHEBYSHEV MULTISECTION MATCHING TRANSFORMERS n 0 1 2 3 4 5.7 115 CnN 1 1 1 1 2 1 3 3 1 3 4 1 6 4 1 Chebyshev Multisection Matching Transformers Another function with desirable properties to implement in a multisection transformer is a Chebyshev polynomial. The Chebyshev polynomials are given by T0 (x) = 1; T1 (x) = x; (5.66) T2 (x) = 2x2 3 T3 (x) = 4x 1; 3x; 4 T4 (x) = 8x 8x2 + 1; .. . Tn (x) = 2xTn 1 (x) Tn 2 (x) : Another, equivalent form can be given where the Chebyshev polynomials look like trigonometric functions (although they aren’t). Letting x = cos , jxj < 1, then it can be shown that Tn (cos ) = cos (n ), and Tn (x) = cos n cos 1 Tn (x) = cosh n cosh x ; 1 x ; jxj < 1; (5.67) jxj > 1. The two main properties of Chebyshev polynomials are the following: 1. For 1 x 1, jTn (x)j 1 and Tn oscillates between map this range into the passband of the transformer. 1; Tn (x) has n zeros between 2. For jxj > 1, jTn (x)j > 1. We will map this range to outside the passband. 1 0.5 -1 0 -0.5 0.5 x -0.5 -1 cos n cos 1 (x) vs. x, n = 1; 2; 3; 4. 1 1. We will 116 CHAPTER 5. IMPEDANCE MATCHING AND TUNING |Γ| Γm θ=β L θm π/2 π− θ m x=+1 π x=-1 With x = cos , as goes from 0 to , x goes from +1 to 1. Force x = +1 at = m . To implement this we replace cos with cos = cos m = sec m cos : = = and x = 1 at (5.68) m cos cos Tn and when m, cos cos m Tn (cos ) ! Tn (then when = m = Tn (1) = 1; (5.69) m m, cos ( cos Tn m) = Tn ( 1) = 1:) (5.70) m Therefore, T1 (x) = x; ) T1 (sec 2 T2 (x) = 2x = sec T3 (x) = 4x3 m = sec 4 T4 (x) = 8x m m cos ) = 2 (sec m cos ; (1 + cos 2 ) 2 1 3 3 (sec m cos ) m cos ) 1; m cos ) = 4 (sec (cos 3 + 3 cos ) 3 sec m m cos ) m cos ) + 1; cos ; 2 8x + 1; ) T4 (sec = sec4 cos ) = sec 3x; ) T3 (sec 3 m 1; ) T2 (sec 2 (5.71) m m 4 cos ) = 8 (sec m cos ) 8 (sec 4 sec2 (cos 4 + 4 cos 2 + 3) m 2 (cos 2 + 1) + 1: Procedure: Equate the actual multisection re‡ection coe¢ cient (5.31) to the N th order Chebyshev polynomial Ae jN TN (sec m cos ) for a given N . That is, ( ) ' 2e jN +::: + f 0 N 1 2 cos N + cos ; N ; 1 2 2 1 cos (N N odd N even 2) + ) = Ae 2 cos (N jN 4) + ::: + TN (sec m n cos ) : cos (N 2n) (5.72) The constant A is again determined by letting f = 0; ZL Z0 = ATN (sec ZL + Z0 ZL Z0 1 )A= ZL + Z0 TN (sec (0) = m) ; m) (5.73) ' 1 ZL 1 ln 2 Z0 TN (sec m) . 5.7. CHEBYSHEV MULTISECTION MATCHING TRANSFORMERS 117 Since max TN (x) = 1 for jxj < 1, then A is also the maximum allowable re‡ection coe¢ cient magnitude in the passband, 1 ZL 1 ln : (5.74) m =A= 2 Z0 TN (sec m ) Bandwidth: Using (5.74), TN (sec m) = 1 m and, since sec m = 1= cos 1 ZL ; ln 2 Z0 (5.75) 1 we use m TN (sec m) 1 = cosh N cosh ) sec m = cosh sec = m 1 cosh N 1 1 ZL ; ln Z0 m 2 1 1 ZL ln Z0 m 2 1 (5.76) such that f f0 = = 2 (f0 4 fm ) 4 =2 m f0 4 1 sec 1 cosh cosh N jN Ae TN (sec m (5.77) 1 1 m 1 ZL ln 2 Z0 : cos ) Example 5.11 Design a three section Chebyshev transformer to match a 100 ohm load to a 50 ohm line, with m = 0:05. Solution: A= sec m m = 0:05; 1 1 cosh 1 N m 1 1 = cosh cosh 1 3 0:05 = cosh = 1:4075; m 1 ZL ln 2 Z0 1 100 ln 2 50 = 0:7806: To determine the Zn values, from (5.72) we have ( ) ' 2e j3 f 0 cos 3 + 1 cos g = Ae ) 2 f 0 cos 3 + 1 cos g = AT3 (sec = A sec3 m (cos 3 + 3 cos ) 3 sec j3 T3 (sec m cos ) : m cos m cos ) (5.78) (5.79) (5.80) 118 CHAPTER 5. IMPEDANCE MATCHING AND TUNING Equating similar terms, cos 3 : 2 = A sec3 0 cos : 2 ) m; 3 = 3A sec 0 = 0:0697; 1 (5.81) sec m ; = 0:1036: m 1 From symmetry, 3 = 0; π/2 Z0 Z1 Γ0 Γ1 2 = 1: π/2 π/2 Z2 Z3 Γ2 (5.82) ZL Γ3 To determine the Zn values, note the relationship n = Zn+1 Zn 1+ $ Zn+1 = Zn Zn+1 + Zn 1 n n such that 1+ 1 1+ = Z1 1 1+ = Z2 1 Z1 = Z0 Z2 Z3 0 = 57:49; 0 1 = 70:77; 1 2 = 86:97: 2 The bandwidth is f f0 = = 2 (f0 2 fm ) 4 =2 m f0 4 0:7806 = 1:006; 100:6% (compare with 71% for the binomial transformer with N = 3. (5.83) 5.7. CHEBYSHEV MULTISECTION MATCHING TRANSFORMERS 119 0.30 0.25 ρ 0.20 0.15 0.10 0.05 0.00 10 14 18 22 26 30 34 38 42 46 50 46 50 f (GHz) 0.30 0.25 Chebyshev, N=3 0.20 ρ Binomial, N=3 0.15 0.10 0.05 0.00 10 14 18 22 26 30 34 38 42 f (GHz) Note: If given m and f =f0 , determine required N then recalculate m using N . Example 5.12 For the previous example of matching a 100 ohm load to a 50 ohm line with m = 0:05 using Chebyshev polynomials, design an appropriate microstrip implementation. Assume "r = 9:8, f0 = 1 GHz, and that the load is an in…nite length of 100 ohm line. Solution: ( 8eA w ; w b < 2; e2Ah 2 i = (5.84) "r 1 2 0:61 w b B 1 ln (2B 1) + 2"r ln (B 1) + 0:39 ; "r b > 2; where r Z0 "r + 1 "r 1 A= + 60 2 "r + 1 377 B= p : 2Z0 "r 0:23 + For Z1 Z2 Z3 Z0 = = = = 57:49; 70:77; 86:97; 50; ZL = 100 0:11 "r ; (5.85) 120 CHAPTER 5. IMPEDANCE MATCHING AND TUNING we obtain w0 b w1 b = 0:978; = 0:721; wL = 0:137; b w2 = 0:427; b w3 = 0:227 b Since c vp = p = f; "e c 1 ) =p : "e f (5.86) With "e = we obtain 1 4 "r + 1 "r 1 1 q + 2 2 1+ = 2:95 cm; 50 57.49 2 4 = 3:01 cm; 70.77 86.97 3 4 (5.87) 12d w = 3:06 cm 100 ... ... λ1 λ2 λ3 4 4 4 Note that higher "r leads to smaller and therefore shorter transformers. The Chebyshev transformer is called an equal ripple transformer. It optimizes bandwidth at the expense of passband ripple. Note: Many other transformers are possible, simply by equating the desired function for ( ) with (5.29). 5.8 Tapered Lines An alternative to using …nite impedance steps such as in the binomial or Chebyshev cases is to use a continuous taper. ZL Z(z) Z0 z=0 z=L Considering the continuously tapered line to be made up of a number of incremental section of length z, with impedance change Z (z) from one section to the next, 5.8. TAPERED LINES 121 ∆Γ Z+∆Z Z ∆z and incremental re‡ection coe¢ cient is = As Z ! 0, we can replace (Z + (Z + Z) Z Z Z = ' . Z) + Z 2Z + Z 2Z (5.88) ; Z by d ; dZ such that dZ : 2Z This can be manipulated into a convenient expression as follows. i h d ln Z(z) Z0 1 dZ (z) d = =2 ; dz Z hdz i dz d = (5.89) (5.90) Z Form the integral summation of If Z (z) is known, then d 1 d ln Z0 = dz 2 dz (all partial re‡ections) and add the approximate phase shifts, Z Z 1 L d e j2 z ln = dz: 2 z=0 dz Z0 (5.91) can be found. 1. Exponential Taper: Z (z) = Z0 eaz ; 0 < z < L: (5.92) We want Z (0) = Z0 ; (5.93) aL Z (L) = Z0 e = ZL : Therefore, 1 ZL ZL ) a = ln : Z0 L Z0 eaL = (5.94) Then, = = = 1 2 1 2 1 2 Z L e j2 z e j2 z e j2 z 0 Z L 0 Z L 0 1 1 ZL = ln 2 L Z0 1 ZL = ln e 2 Z0 Z d Z ln dz dz Z0 d ln (eaz ) dz dz 1 ZL ln dz L Z0 L e j2 z dz 0 j L sin L L : (5.95) 122 CHAPTER 5. IMPEDANCE MATCHING AND TUNING When L = 0, performance. sin L L = 1. From the plot, we see that (2 L > L > ) L > 2) for best 2. Triangular Taper: Another common method is to assume a triangular taper for d ln (Z=Z0 ) =dz, ( ZL 4z d ln ZZ0 0 z L=2; L2 ln Z0 ; : = ZL 4 z 4 ln ; L=2 z L dz L L2 Z0 This results in Z (z) = ( 2 Z0 e2(z=L) ln(ZL =Z0 ) ; 2 2 Z0 e(4z=L 2z =L 1) ln(ZL =Z0 ) ; 0 z L=2 L=2; : z L (5.96) (5.97) Integrating the triangular expression leads to = 1 e 2 j L ln ZL sin ( L=2) Z0 ( L=2) 2 : (5.98) 5.8. TAPERED LINES 123 For L > 2 the peaks of the triangular taper are lower than the corresponding peaks of the exponential 2 taper because of the (sin x=x) factor. However, the …rst null occurs at 2 , whereas the …rst null appears at for the exponential taper. 3. Klopfenstein Taper: The Klopfenstein taper has been shown to be optimum in the sense that the re‡ection coe¢ cient is lowest over the passband, for a given taper length greater than some critical value. Alternatively, upon speci…cation of m , the Klopfenstein taper yields the shortest length L. = 0e j = 0e j cos L q 2 ( L) A2 qcosh A 2 cos A2 ( L) L cosh A ; L > A (passband), ; L < A; (5.99) where 0 = 1 ZL ZL Z0 ' ln ; ZL + Z0 2 Z0 1 A = cosh 0 (5.100) : m The impedance taper must generally be calculated numerically from ln Z (z) = 1 0 ln (Z0 ZL ) + A2 (2z=L 2 cosh A where (x; A) = ( x; A) = Z 0 and where I1 (x) is the modi…ed Bessel function. x 1; A) ; p I1 A 1 y 2 p dy; A 1 y2 0 z jx 1j L; (5.101) (5.102) Example 5.13 Design a triangular taper, an exponential taper, and a Klopfenstein taper (with m = 0:02) to match a 50 ohm load to a 100 ohm line. Plot the impedance variations and resulting re‡ection coe¢ cient magnitudes vs. L. Solution: (a) Triangular Taper: For the triangular taper ( 2 Z0 e2(z=L) ln(ZL =Z0 ) ; Z (z) = 2 2 Z0 e(4z=L 2z =L 1) ln(ZL =Z0 ) ; = 1 e 2 j L ln ZL sin ( L=2) Z0 ( L=2) 0 z L=2 L=2; ; z L (5.103) 2 : (b) Exponential Taper: For the exponential taper Z (z) = Z0 eaz ; 0<z<L 1 ZL 1 a = ln = 0:693 ; L Z0 L 1 ZL j L sin L = ln e : 2 Z0 L (5.104) 124 CHAPTER 5. IMPEDANCE MATCHING AND TUNING (c) Klopfenstein Taper: For the Klopfenstein taper 0 ' 1 ZL ln = 0:346 2 Z0 A = cosh 1 0 (5.105) = 3:543 = 1:13 ; m Z is evaluated numerically and in the passband ( L > A = 1:13 ) = 0e j cos L q 2 ( L) cosh A A2 : (5.106) Example 5.14 Design an exponentially tapered matching transformer to match a 100 ohm load to a 50 ohm line. Plot j j vs. L, and …nd the length of the matching section (at the center frequency) required to obtain j j 0:05 over a 100% bandwidth. How many sections would be required if a Chebyshev matching transformer were used to achieve the same speci…cations? 5.8. TAPERED LINES 125 Solution: Z (z) = a = Z0 eaz ; 0<z<L 1 ZL 1 ln = 0:693 ; L Z0 L z Z (z) = 50e0:693 L 1 ZL j L sin L ln e 2 Z0 L sin L 1 100 sin L ln = 0:3466 : 2 50 L L ) j j = = For 100% bandwidth, f f0 = 4 2 m ) fm = = 1 f0 ; 2 2 (f0 fm ) f0 = m = 1; 4 and so 1 f0 2 3 f0 2 f represents a 100% bandwidth. G (x) = 0:3466 Starting at L = 1:72 (where j j L sin(x) x vs. x= . 0:05) 1:72 ) 2 L 1:72 2 1:72 ) L For f : : 1 3 f0 to f0 ; 2 2 2 0 to 0:666 0: Therefore, L is needed. For example, if f0 = 30 GHz ( 0 0:86 (2 = :01 m), 0) = 1:72 0 = 0:86 : 126 CHAPTER 5. IMPEDANCE MATCHING AND TUNING 0.10 0.08 ρ 0.06 0.04 0.02 0.00 20 22 24 26 28 30 32 34 36 38 40 f (GHz) For the Chebyshev, cosh N= 1 1 cosh cosh = 1 2 m 1 cosh 1 ln ZZL0 1 0:05 1 m) 1 100 2 ln 50 (5.107) (sec sec = 2:97; 4 ) N = 3: Since a three section Chebyshev transformer is 3 has length L = 1:72 0 ). Note that for the Klopfenstein taper, 0 = A = 0 =4 = 0:75 0 it is smaller then the exponential taper (which ZL Z0 = 0:333; ZL + Z0 cosh 1 0 = 2:5846: m Since the passband is de…ned as L > A, and the maximum ripple in the passband 0:05, then for our value of A the ripple will be at most 0:05 for L > A = 2:5846; L 2 Since max =2 > 2:5846; L > 2:5646 ) L > 0:411 : 0; L > 0:411 (2 0) = 0:8227 0 would give a 100% bandwidth. Note that this is quite a bit shorter than the exponential taper, but still longer than the Chebyshev taper. 5.9 Bode-Fano Criteria The Bode-Fano criteria provides information about the theoretical limits that constrain the performance of an impedance matching network. 5.9. BODE-FANO CRITERIA 127 Zg matching Z 0 ,v p + Vg - ZL network Z in Assume that the matching network is lossless, and ZL is generally complex. 1. Can we achieve a perfect match (zero re‡ection) over a speci…ed (nonzero) bandwidth? 2. If not, how good can we do? What is the trade-o¤ between m and bandwidth? The bode-Fano criteria gives the optimum result that can ideally be achieved. As an example, consider the following. matching Γ(ω) network The Bode-Fano criteria is, for this case, Z 1 ln 0 1 d! j (!)j RC : (5.108) Assume that we desire the response to be as shown below. |Γ| 1 Γm ω ∆ω Then Z 0 1 ln 1 d! = j (!)j ) Z ln ! ! 1 j mj d! = RC jln mj ! ln : 1 j mj RC ; (5.109) 128 CHAPTER 5. IMPEDANCE MATCHING AND TUNING Note that as m varies from 0 to 1, RCjln mj varies from 0 to 1. So, : 0 ! 1; ! : 0 ! 1: (5.110) m For a …xed RC, broader bandwidth can only be achieved at the expense of a higher re‡ection coe¢ cient in the passband. Also, the passband m cannot be zero unless ! = 0. Thus, a perfect match cannot be maintained except at possibly a …nite number of frequencies. As R and/or C increase, ! decreases. |Γ| ω |Γ| ω Example 5.15 A parallel RC load having R = 50 ohm and C = 3 pF is to be matched to a 30 ohm line from 1 to 3 GHz. What is the best m that can be obtained? Assume a square-wave pro…le. Solution: |Γ| 1 Γm 2π(1) Z 0 2π(3) Z 2 (3 109 ) 1 1 d! = ln d! j (!)j j mj 2 (1 109 ) 1 2 (2) ln 20:94 ) m 0:1888: j mj 1 ln ω 20:94 109 ; (5.111) 5.9. BODE-FANO CRITERIA 129 Alternatively, to match from 1 to 10 GHz, 2 (9) ln 1 j mj 20:94 ) m 0:691: 130 CHAPTER 5. IMPEDANCE MATCHING AND TUNING Chapter 6 Power Dividers and Directional Couplers (Here we will cover Sections 7.4, 7.2, and 7.3) Directional couplers are passive microwave components that provide power division or power combining. They are usually represented in the following way: 1 input 2 through 4 isolated 3 coupled The main idea is that power is input into Port 1. Some fraction of the input power goes through to Port 2 (the “through port”), while the rest goes to Port 3 (the “coupled port”). Ideally, no power is transferred to Port 4 (the “isolated port”). C = 10 log10 D = 10 log10 I = 10 log10 P1 , coupling factor in dB, P3 P3 , directivity in dB, P4 P1 , isolation in dB. P4 (6.1) The coupling factor indicates the fraction of input power coupled to Port 3. Directivity indicates the coupler’s ability to isolate forward (Port 3) and backward (Port 4) waves. Note that I = C + D. An ideal directional coupler would have I = D = 1 (P4 = 0). The device is reciprocal in the sense that power can be applied to Port 2, in which case Port 1 becomes the through port, Port 4 becomes the coupled port, and Port 3 the isolated port. Also, if power is applied to both Ports 2 and 3, power is combined and emerges from Port 1. Bethe Hole Coupler: 131 132 CHAPTER 6. POWER DIVIDERS AND DIRECTIONAL COUPLERS The principle of operation of this coupler, and of all couplers, is that two separate waves can be combined in-phase at the desired output port, and can be combined out-of-phase at the desired isolated port. The details of the single-hole coupler are provided in the text. It can be shown that, for the TE10 mode, the skewed coupler design leads to r 2 4k02 r03 C = 20 log10 ; = k02 : (6.2) 3ab a One can thus design a coupler to achieve a certain coupling. This will only hold at one frequency, and is very frequency sensitive. To decrease sensitivity, multi-hole couplers can be design, analogous to multi-section quarter-wave transformers. Two-hole coupler: Here we use parallel guides. The apertures are small, and placed g =4 apart. Most of the input wave is transferred to the through port. Aperture 1, referenced at 0 phase, radiates a backward (B1 ) and forward (F1 ) wave into the upper guide. Aperture 2 does the same, B2 and F2 , respectively. At the reference point, B1 ej z z=0 = B1 ; F1 e j z z=0 = F1 (6.3) 133 due to Aperture 1. At this same reference point, the backward wave from Aperture 2 will be B2 ej g =2 = B2 ej = B2 : (6.4) If jB1 j ' jB2 j, these waves will cancel, such that no power comes out the isolated port. Does any power come out of the coupled port? Yes: at the location of Aperture 2 we have F1 e j g =4 + F2 e j g =4 = (F1 + F2 ) ( j) ; (6.5) and so the two forward waves add in-phase. It is important to note that for exact cancellation of the backward waves, jB1 j must equal jB2 j. This will not be true (why?), and so we never obtain perfect isolation. The two-hole coupler is less frequency dependent than the one-hole coupler, even though the spacing will be g =4 at only one frequency. Example 6.1 Assume C = 3 dB with in…nite directivity. Determine the power dissipated at ports 2,3, and 4. Solution: = power At Port 2: 1 2 p 2Z0 Z0 2Z0 + Z0 1 1 = 2 2 at Port 3 2 and at Port 4 with 1 36 watts back out to Port 1. N + 1 hole couplers: p 1 : 9 1 4 = watts, 18 9 1 watt, 2 1 2 = 1 1 = watts, 2 36 134 CHAPTER 6. POWER DIVIDERS AND DIRECTIONAL COUPLERS All apertures are spaced d = g =4 apart. The amplitude of the wave incident on the input port is A, as is the amplitude at the output port of the through wave (of course, this is not self-consistent, nevertheless it yields reasonable results)! Set the phase reference to be at the …rst aperture. At the coupled port, N X F = Ae j N d Fn (6.6) n=0 (remember we have N + 1 apertures), and at the output of the isolated port, B=A N X j2 nd Bn e : (6.7) n=0 Then, C= 20 log10 D= 20 log10 = C F = A 20 log10 B = F 20 log10 N X Fn dB, (6.8) n=0 20 log10 N X Bn e PN Bn e j2 PN n=0 Fn nd n=0 j2 nd dB. n=0 Assume that all aperture are round holes with identical positions, s, relative to the edge of the guide, with rn being the radius of the nth aperture. It can be shown that Fn = kf rn3 ; Bn = kb rn3 ; (6.9) where kf;b are dependent on aperture shape, and are slowly varying functions of frequency. Therefore, C= D= 20 log10 jkf j C 20 log10 N X rn3 dB, (6.10) n=0 20 log10 jkb j 20 log10 | N X rn3 e j2 nd n=0 {z highly frequency sensitive dB. } In the design of multi-section quarter wave transformers we considered binomial and Chebyshev methods to choose the impedances Zn . Here, the design involves choosing the hole radii using similar methods. Binomial Response: Pick rn3 proportional to the binomial coe¢ cients, rn3 = kCnN (6.11) 135 (recall CnN = N! ): (N n)!n! (6.12) The proportionality constant k can be determined from C= = = N X 20 log10 jkf j 20 log10 20 log10 jkf j 20 log10 20 log10 jkf j 20 log10 jkj rn3 (6.13) n=0 N X kCnN n=0 20 log10 N X CnN : n=0 From this, and with a desired value of C and a known kf ; N , one can determine k. Note: " 2k02 jkf j = sin2 3ab 1;0 " 2k02 jkb j = sin2 3ab 1;0 2 s a 2 1;0 k02 2 s 2 1;0 + 2 a k0 2 sin 2 sin s + a s a 2 2 2 1;0 a cos 2 s a cos 2 s a 2 2 2 1;0 a !# !# ; (6.14) : (6.15) A Chebyshev response can also be obtained. Example 6.2 Design a …ve-hole coupler with a binomial response. The center frequency is 10 GHz, and the required coupling is 18 dB. The physical structure is a rectangular waveguide with round coupling apertures centered across the broad common wall of the guides. Solution: For an X band waveguide, a = 0:02286 m, b = 0:01016 m (from the textbook appendix). For a 5 hole coupler, N = 4. From (6.14)-(6.15), 105 ; jkf j = 1:113 jkb j = 1:705 106 : Then, C = ) jkj = = 20 log10 jkf j 20 log10 jkj 20 log10 N X n=0 20 log10 jkj = C + 20 log10 jkf j + 20 log10 log101 7:071 1 20 10 8 C + 20 log10 jkf j + 20 log10 : With rn3 = kCn4 , r0 r1 r2 r3 r4 = = = = = 4:135 6:564 7:514 6:564 4:135 mm, mm, mm, mm = r1 ; mm = r0 : CnN ; N X CnN ; n=0 N X n=0 CnN !! 136 CHAPTER 6. POWER DIVIDERS AND DIRECTIONAL COUPLERS C= 20 log10 jkf j 20 log10 jkj 20 log10 C 20 log10 jkb j 20 log10 CnN n=0 = 18 dB at f = f0 ; D= N X N X rn3 e j2 nd n=0 = 283 dB at f = f0 . Note that C is very insensitive to frequency (only via kf , which is a slowly-varying function with respect to frequency; upon taking log10 jkf j we obtain a fairly insensitive function of frequency for C. However, D is very frequency sensitive due to the summation term. 100 5 Hole Coupler 80 D 60 40 C 20 0 4 6 8 10 12 14 16 14 16 f (GHz) For a 3 hole coupler (r0 = 6:564; r1 = 8:27; r2 = 6:564 = r0 ), 100 3 Hole Coupler 80 D 60 40 C 20 0 4 6 8 10 f (GHz) and for a 15 hole coupler 12 137 100 15 Hole Coupler 80 D 60 40 C 20 0 4 6 8 10 12 14 16 f (GHz) Directional couplers can also be implemented with transmission lines, although the theory will not be developed here. Hybrid Couplers: 138 CHAPTER 6. POWER DIVIDERS AND DIRECTIONAL COUPLERS Hybrid couplers are special cases of directional couplers where C = 3 dB. 1. Quadrature hybrid: There is a 90 phase shift between Ports 2 and 3 when fed at Port 1. 2. Magic-T hybrid: There is a 180 phase shift between Ports 2 and 3 when fed at Port 1. Power Dividers: Power dividers are similar to direction couplers, but are usually three-port devices. The idea is still to split input power into two parts. We will brie‡y study a lossless divider. The resistive divider and Wilkinson divider are also very common. All of these dividers are T-junction dividers; some physical implementations are shown below. 139 1 ou tp ut A lossless divider is depicted below. Z1 input + Z0 Vo jB - Yin Z2 2 ut tp ou The susceptance B accounts for the physical discontinuity in the waveguide or transmission line. We have 1 1 Yin = jB + + ; (6.16) Z1 Z2 and we want Yin = 1 : Z0 (6.17) If we ignore B then 1 1 1 + = : Z1 Z2 Z0 (6.18) If Z1 and Z2 satisfy this requirement then we divide the power with no re‡ection. For example, Z1 = Z2 = 2Z0 will work, resulting in a 3 dB divider. If the divider segments feed lines with Z0 characteristic impedance, quarter-wave transformers can be used to provide a good impedance match. 140 CHAPTER 6. POWER DIVIDERS AND DIRECTIONAL COUPLERS Some drawbacks to this type of divider are that there is no isolation between output ports, and looking into an output port one sees an impedance mismatch. To determine impedance values for a given power split, the following procedure is used: Pin = 1 v02 2 Z0 (6.19) since the divider is match looking into the input, and P1 = If I want P1 = xPin and P2 = (1 1 v02 ; 2 Z1 P2 = 1 v02 : 2 Z2 (6.20) x) Pin , then 1 v02 1 v2 Z0 = xPin = x 0 ; ! x = 2 Z1 2 Z0 Z1 1 v02 1 v02 = (1 x) Pin = (1 x) ;! 1 P2 = 2 Z2 2 Z0 P1 = (6.21) x= Z0 : Z2 Example 6.3 Determine the normalized output impedances for a 3 : 1 power split. Solution: We want 3 1 Pin ; P2 = Pin ; 4 4 3 Z0 4 )x= = ) Z1 = Z0 ; 4 Z1 3 1 Z0 x= = ! Z2 = 4Z0 : 4 Z2 P1 = 1 (6.22) Chapter 7 Electromagnetic Compatibility and Interference (EMS/EMI) From "Introduction to Electromagnetic Compatibility" by Clayton R. Paul (Wiley: 1992) Radiated Emissions The FCC regulates "unintentional radio-frequency devices". It is illegal to sell or advertise for sale any such products until their radiated and conducted emissions have been measured and found to be in compliance. Class A devices are those that are marketed for use in a commercial, industrial or business environments. Class B devices are those that are marketed for residential use. Class B limits are more stringent than Class A limits FCC Emission Limits for Class B Digital Devices, Radiated Emissions (at 3 meters): Frequency (MHz) uV/m dB(uV/m) 30 - 88 100 40 88 - 216 150 44 216 - 960 200 46 > 960 500 54 Hertzian dipole model of a radiating transmission line: The far …eld (jrj , Lw ) from a wire of length Lw carrying a constant current I0 along the z axis is E(r) = b (I0 ) (Lw ) (j! ) E (r) = jI0 Lw 2 10 7 f e e jkr 4 r sin (7.1) jkr r : This assumes current is constant (not a bad assumption for Lw ) Note that the measurement position r = 3 meters may not be in the far …eld 141 142CHAPTER 7. ELECTROMAGNETIC COMPATIBILITY AND INTERFERENCE (EMS/EMI) For two wires (see …gure), E (r) = jLw 2 10 7 f jk(r+ ) e I1 + I2 r+ jk(r e ) ; r (7.2) where b b y r = sin sin s = sin sin ; 2 (7.3) (7.4) since cos (90 ) = sin r z s/2 ∆ y I2 I1 s so that E (r) = jLw 2 10 Di¤erential mode currents (I0 = I1 = 7 f , sin k 2s sin sin 7 f 7 I0 jLw 2 10 7 f f e jk e r e : (7.5) ejk (7.6) jkr (2j sin (k )) r jkr r s sin k sin sin 2 (7.7) : (7.8) ' k 2s sin sin , 2 = 1:32 10 14 1:32 10 14 E dm (r) = = jkr e E dm (r) = I0 Lw (2 ) 10 Field is maximum for + I2 ejk I2 ) = I0 Lw 4 10 Assuming that s jk I1 e r E dm (r) = I0 jLw 2 10 = jkr e r 7p "f 2 I0 Lw f 2 s e e jkr r s sin sin (7.9) sin sin ; (7.10) jkr r I0 Lw f 2 s sin jsin j (7.11) = 90 , E dm;max (r) = 1:32 10 r 14 I0 Lw f 2 s: (7.12) 143 Maximum radiation occurs in the plane of the wires, broadside to the wire axes. Radiation can be minimized at a certain frequency and position by – reducing the current level – reducing the wire length and/or wire separation (overall, reducing the loop area L s) – use ferrite beads (adds series inductance to attenuate high-frequency harmonics) Common mode currents (I0 = I1 = I2 ) Assuming that s E cm (r) = I0 jLw 2 10 7 = I0 jLw 2 10 7 = jI0 Lw 4 10 7 , cos k 2s sin sin f f f e jkr e r e + ejk (7.13) jkr (2 cos (k )) r e jk jkr r (7.14) s cos k sin sin 2 ; (7.15) '1 E cm (r) = 1:26 10 6 1:26 10 6 jE cm (r)j = r jI0 Lw f e jkr (7.16) r I0 Lw f (7.17) Common mode radiation is insensitive to rotating the cable/wires Radiation can be minimized at a certain frequency and position by – reducing the current level – reducing the wire length – use a common-mode choke (see …gure) – use ferrite beads (adds series inductance to attenuate high-frequency harmonics) Example: Lw = 1 m, I0 = 20 mA, s = 1 mm, f = 50 MHz, r = 3 E dm;max (r) = 1:32 10 14 20 10 3 (1) 50 3 = 220 V/m = 47 dB V/m 106 2 1 10 3 (7.18) (7.19) (above the limit) jE cm (r)j = = 1:26 10 6 10 6 I0 Lw f r 1:26 20 3 = 420; 000 V/m (7.20) 10 (far above the limit) Q. What I would produce 100 V/m? A. For di¤erential mode, 9 mA, for common mode, 4:8 A 3 (1) 50 106 (7.21) (7.22) 144CHAPTER 7. ELECTROMAGNETIC COMPATIBILITY AND INTERFERENCE (EMS/EMI) Raise frequency to 1 GHz: E dm;max (r) = 1:32 14 10 20 10 3 (1) 1 3 = 88; 000 V/m = 99 dB V/m 2 109 1 10 3 (7.23) (7.24) (way over) jE cm (r)j = = 1:26 10 6 10 6 I0 Lw f r 1:26 (7.25) 20 10 3 = 8; 400; 000 V/m !!! 3 109 (1) 1 (7.26) (7.27) Q. What I would produce 500 V/m? A. For di¤erential mode, 114 A, for common mode, 1:1 A Susceptibility Models Circuit Model: i(z,t) Vs(z,t) i(z+dz,t) + - R ∆z L ∆ z ∆z + ∆z Is(z,t) G ∆z C ∆z v(z+dz,t) - ∆z d2 v(z) dz 2 d2 i(z) dz 2 2 v(z) = 2 i(z) = d vs (z) ; dz d is (z) (G + i!C) vs (z) + ; dz (R + i!L) is (z) + (7.28) where 2 = (R + j!L) (G + j!C) ; (7.29) and = +j 2 C is called the propagation constant (1=m). The real and imaginary parts of the propagation constant are known as the attenuation constant ( ) and the phase constant ( ), respectively. An incident electromagnetic wave induces voltages and currents on the transmission line, which are represented by Vs and Is . Faraday’s law can be used to determine Vs . The induced emf around a loop l with surface S is I Z emf = E(r) dl = j! B(r) n dS: (7.30) l S b. The distributed voltage Assume the y axis is vertical, so that the x axis is into the paper, and n = x source for two wires separated by distance s is Z z+ z=2 Z s Z s V (z) ' j! Bxi (y; z) dydz ' j! z Bxi (y; z) dy (7.31) z V (z) = Vs (z) = z z=2 j! 0 Z 0 0 s Hxi (y; z) dy: 145 Since the wire separation s is much less than , Z s Hxi (y; z) dy ' Vs (z) = j! j! sHxi (z) (7.32) 0 An incident, vertically polarized electric …eld will induce a voltage between the wires, causing a displacement current through the capacitance between the wires (we ignore here conduction current due to a lossy insulator). Using Ohm’s law, Z s I (z) Eyi (y; z) dy = I (z) Zc = V (z) = (7.33) j!C 0 Z s Is (z) = j!C Eyi (y; z) dy ' j!CsEyi (z): 0 Rather than solve (7.28), we can simply things further by assuming that L one section of line. Replacing z by Lw and setting R = G = 0, VL = = , so that we only need RL Lw (Vs + Rs Is ) L LC + j! (Rs RL C + L) j!RL Lw s Hxi + Rs CEyi 2 ! RL LC + j! (Rs RL C + L) !2 R RL + Rs RL + Rs Note that the area of the loop formed by the two wires, A = Lw If frequency is su¢ ciently low, (7.34) (7.35) s is important. RL Lw (Vs + Rs Is ) RL + R s j!ARL Hxi + Rs CEyi : = RL + Rs VL = (7.36) Example: Assume that the plane wave jky E=b zE z e ; b H=x Ez e jky (7.37) is incident on a two-wire transmission, with Ez = 1 V/m and ! = 30 MHz. The wires have center-to-center spacing D = 1:22 cm, radius a = 0:1 cm (this is the standard 300 ohm twin-lead line), and length Lw = 1 m. In this case, D "0 ; C= L = 0 cosh 1 ; (7.38) D 2a cosh 1 2a and for this line L = 1 H and C = 11:2 pF. For this incident wave there is only a voltage source, Vs = j! sHxi = such that Hxi = 4 Assume RL = 1000 , Rs = 10 7 10 j1:22 mV 1 = 3:33 377 10 (7.39) 9 : (7.40) . Then, VL = j!ARL RL + Rs Hxi + 0 = j1:21 mV. (7.41) If, instead, the incident plane wave was b Ey e E=y then jkx ; H=b z Rs CEyi = 1:12 and VL = 10 Ey e 10 j!ARL 0 + Rs CEyi = RL + Rs jkx ; (7.42) (7.43) j0:04 mV. (7.44) 146CHAPTER 7. ELECTROMAGNETIC COMPATIBILITY AND INTERFERENCE (EMS/EMI) Chapter 8 Microwave Filters 147 148 CHAPTER 8. MICROWAVE FILTERS Chapter 9 Appendix 9.1 Transient transmission line current We obtained the transient transmission line voltage as (2.111), v (z; t) = v + (t z=vp ) + v (t + z=vp ) : (9.1) Using @v(z; t) = @z L @i(z; t) @t (9.2) we obtain @v (z; t) @z @v (t z=vp ) @ (t z=vp ) @ (t z=vp ) @z 1 0 v vp @i(z; t) L ; @t = = = (chain rule) (9.3) leading to @i(z; t) @t 1 v +0 (t z=vp ) Lvp 1 ) i (z; t) = v + (t Lvp = v 0 (t + z=vp ) ; z=vp ) (9.4) v (t + z=vp ) + c (z) (9.5) where c (z) is a constant of integration. To determine c (z), use @i(z; t) = @z C @v(z; t) : @t (9.6) From (9.5) and using (9.3) we have @ 1 i (z; t) = @z Lvp2 = C v +0 (t z=vp ) v 0 (t + z=vp ) + @ c (z) @z (9.7) @v(z; t) : @t With @v (z; t) @v (t z=vp ) @ (t z=vp ) = @t @ (t z=vp ) @t = v 0 (1) 149 (9.8) 150 CHAPTER 9. APPENDIX we have @ i (z; t) = @z = 1 v +0 (t Lvp2 C z=vp ) + v @v(z; t) = @t C v +0 (t 0 (t + z=vp ) + z=vp ) + v 0 @ c (z) @z (9.9) (t + z=vp ) : Therefore, such that | 1 + C v +0 (t Lvp2 {z } z=vp ) + v 0 (t + z=vp ) + @ c (z) = 0; @z (9.10) 0 @ c (z) = 0 ) c (z) = constant. (9.11) @z c (z) = c provides at most a d.c. current that provides a d.c. o¤set to the q time-dependent current. We will L 1 p = Z0 , the desired expression ignore this o¤set (set c = 0) to arrive at, upon noting that Lvp = L LC = C i (z; t) = 1 v + (t Z0 z=vp ) v (t + z=vp ) : (9.12)