IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL $2 NO. I, NOVEMBER 198 Exact, Closed-Form Expressions for Transient Fields in Homogeneously Filled Waveguides Steven L. Dvorak Abstract—It swell known that transient electromagnetic waves Jn waveguides ehibit dispersion. Exact, closed-form expressions, ‘which involve Bessel functions of the fist kind, have een derived for the impulse response of waveguide, but exact, closed form expressions for more complex pulses are absent from the literature: Tn this paper, it Is demonstrated that incomplete Lipechitz-Hankel integrals can be used to represent transient plses in homogeneously filled waveguides. A continuous wave Dabs is investigated in this paper, however, this technique cam ‘ko be applied to-a number of other transit waveforms. The Fesultng expressions are verified by numerically Integrating the pole distribujon multpied by the known impalse response. 1. FORMULATION OF THE TRANSIENT POTENTIAL ' this paper, a technique is developed which yields elosed- form expressions for transient electromagnetic fields in ibomogencously filed waveguides and lossy transmission lines. This general technique is applicable for transient sources which possess analytical Fourier transforms involving tran- seendental functions and “pole” terms ‘Other authors have obtained analytical expressions for tan- sient electromagnetic pulses in waveguides [1]-(6}, however, the results are either left in terns of inverse Laplace or Fourier transforms, or some kind of approximation is used to obtain approximate analytical results. For example, late time results are obtained in (1), [3] by using the method of steepest descent. In [4-16] the branch cut associated withthe slispersive propagation constant is removed by approximating the propagation constant by a Taylor series expansion about the fundamental frequency. In summary, a closed-form expression is available for the impulse response of a waveguide, howeve until now, no exact, closed-form expressions for transient electromagnetic pulses in waveguides have been derived, “The method whichis developed in this paper is applicable to homogeneously filled waveguides of arbitrary cross-section, However, forthe sake of simplicity, we investigate the ret angular waveguide shown in Fig. 1. We assume that the rectangular waveguide is homogeneously filled and bounded by perfect electric conducting walls on S. The electromagnetic felds inside the waveguide must satisfy Maxwell's equations everywhere in Vand the proper boundary conditions on S. However, instead of solving directly forthe fields, we fist find the potentials. Inthe frequency domain (e2* time convention), the transverse magnetic and transverse electric to 2 (TM. and T'B.) modes must satisfy the homogeneous, scalar wave Manuscript erived August 1, 1993: esd December 27, 1983 ‘Toe hor wth he Deptt of lel and Compr Engineering Fig. 1. Atmogeneomy ied rectangle warez equation {7} (7+ 6 a where 9 = de and sb =F; for the TM, and TE,_ modes, respectively. The tide serves a a reminder that @ is the frequency domain potential. We assume that the complex wavenumber is defined as, Jie? Fae) Q where lin(k) < 0 and 7 = «Je. ‘The solution for the wave equation has the genera form He y.2 0) = Cher kyyle* OY where the & denotes the forward and backward traveling waves, C* are the frequency dependent amplitude coefficients forthe forward and backward waveling modes, Y(k2 yu) represents the mode distbaton inthe tansverse pane, fy (2-12-12)! and Im(k,) <0. The wansvese dependence, ‘(ker yt). independent of frequency and real valued for all homogeneously led waveguides of arbitrary cross-section For example, the transverse dependence fo te TM modes in the rectangular waveguide in Fig. 1 is given by [7] e(hezshy) = cin) cin) whom be = ran by = mf m= 1,28, a mre 1 Bikes 30 kevin eTE, ads canbe epee O} where 1 = 0,1,2, 0 excepted ces and n= 0.1.2..00.305 on18-94809480600 © 1906 IEEEFourier transform techniques are now used to formulate an expression forthe transient potential. Whenever the eigeaval- tues are independent of frequency, the transient potential can be written as vA lenat) = Week “This potential is uniquely determined once C+ is spect We investigate two different ransen soures in the following I. PRoPAcanioy oF 4 CW PULSE In this section, we look atthe propagation of a continuous wave (CW) pulse inside homogeneously filed waveguide, In order to simplify the analysis, we only look atthe forward ‘propagating pulse. We assume that a transient source located at = 0 excites a single mode W*eU.04t) = Acoslunt+ §)[H()— H(t—T)W (ke kyu) 0 where A is a real-valued amplitude constant, 7 is the time width of the pulse, and 1 150 Hw) = {ti tea} ® js the Heaviside unit step function, Sources which have tore complied spatial dependencies can be handled by somming ove ll dhe possible modes inthe wavogide, The ‘modal coeiient is now obtained by enforcing the boundary conditions at = = 0 in the frequency domain eH) rhe itetahy F Ai fore IL » oe ore Ih Subsincing ths expresion into (6) yields Pein) = AMD for) ~ F(t) + HTH 7, -uy) eta} where Hon) = [a ay ‘Computation of the transient potential relies on the eval- uation of the inverse Fourier transform given above. Fast Fourier transform (FPT) techniques ean be used to evaluate this imegral sumerically; however, additional insight can be gained by solving this integral analytically. Furthermore, the analytical solution for this integral is free from problems inherent to the numerical FFT (ie. aliasing). The «ansient potential associated withthe source distribu tion in (7) can also be obtained by integrating the transient source distribution multiplied by the analytical waveguide impulse response. However, the numerical integration masks Some of the physica insight which is available from a purely analytical solution and itis also relatively ineficien ‘The wave equation can also be solved directly using various Aiferential equation-based methods [8]. These purely numer- ical techniques require lage amounts of computer memory and computation time. However, they can be easily applied to inhomogencousl filled waveguides. ‘We uilize contour integration techniques to obtain an an: lytical solution for (10) in this paper. Before employing contour integration techniques, we need to investigate the singulaiies inthe integrand. The integrand of (11) possesses 1 simple pole at w = wip. I also contains a branch cut which is associated with the square oot in fz. When the waveguide is homogeneously filled, thereby yielding eigenvalues which are independent of frequency, the propagation constant can be rewriten as VPIR=1VF jE a1) ‘where wis the cutoff frequency forthe mode and w = 1/ fe is the phase velocity in @ good dielectric. Application of the quadratic formula yields k=l VGrane=o) «3 where the branch points are Focated at ky i+ V@aaP=1 ia oe sheen ces The branch cut, which is defined by Im(.) <0, is shown in Fig. 2 along with the original contour of integration, T., and the simple pole. For this definition of the branch cut, it ‘ean be shown that Ut, wy) = ~ (0) as) where + denotes the complex conjugate operation, This allows for the simplification of (10), )= ves Sete I(t.04) - AMMO Tup)}. 06) Fort < =/v, we close the contourat [| —» sc inthe lower half plane (ie., P< in Fig. 2). Since no singularities are enclosed by the contour Py -+T'< and the integral along I< equals zero, CCauchy’s residue theorem states that Muy) = 0: t< 2/0. a This result agrees with causality arguments Now we look atthe more complicated ease when t > =/v. This time the deformation is cared out in the upper hal plane. Unforuautely, the integrand in (11) #8 not ia form Which can be easly integrated along the branch cus in Fig, 2 However. this integral can be decomposed into a finite number of integrals which are in the form of incomplete Lipschitz-Hankel integrals (LIT) of the fist kind (8, 20), -< edu paineneesee ee ig Th.taaverst In onder forthe above integral representation to hold the iverson contour, Fy, must satisfy the conditions Re[(a Seo(a.c)28 Wie TRANSACTIONS ON MICROMAVE TIRORY AND TECHBIQUES, VOL £2, NO, 1, NOVEMBER 198 rao) Fig. 2. The confou f integration for 1( Rlemioa sex dtl Bk) 0 te plane Te prope aje*10] > 0 and Re[(u + je] > 0, This forces the pole 6 lie 10 the lft of the inversion contour. A typical inversion contour is shown in Fig. 3. A method for closing the contour is also illustrated in Fig, 3. The first step in aking the change of variables, B= 9/(2r), which yields soe ey) = He s/o? cao {fee ava} where de = YGF= TIFF and iy = sy ~ j/(2"). We wi assume that the losses are small enough so that ce $8 real ‘alusd. Next we tansform tothe ueplane (se (18) by using the change of variables les ow eo VE = TTA I should be noted that x > when 1 s/v, The new contour of integration, Py is shown in Fig, 4 for 7 = 8.0 10-9 Sim and = 5.0% 10"? Sf, The data in this example is forthe TE,o mode in a X-band rectangular waveguide. The branch cuts in Fig. 2 wansform to branch cuts Which emanate from the branch points at u = ++) and w= —j,and extend to w= +00 and w = ~joo, respectively (ee Fig. 4), These branch cuts are defined analytically by Re( Vu? + 1) > 0. Both of the contours in Fig. 4 wrap around the branch cuts and approach = :too on the right-hand-side ofthe branch cuts. Conductive losses were added to move the ‘contours away from the branch cuts. Increasing the amount of Toss moves the contour further from the branch cut. For the cease = 0, the portion of the original contour (see Fig. 2) |o] > ue transforms to the imaginary axis in the a-plane. Tn order to expresso as a function of u, we fist rewrite (20) asa quadratic equation and then solve the esuting quadratic equation, yielding en Fig. 4. A took tthe soto of ign for Tivo) in the wae Foe to diferent talus of The TE ode in and etal ie wed th exaipe Te parameter given by = 22.86 fom, b= 11.016 mm, fo = 100 GHz = toe stg. = 10.0 0% iad t= 50.38 To choose the correct sign in (21), we substitute (20) into 21) and find that to [o-2va=al] wil [ewe @ This eqtion i saised provided that we choose the upper signs in (21) and (22) singe the proper Riemann set in the ‘plane is defined by Rel Vat #1) > 0 ‘The res in (20) and (2) allow vs to write He~siyener f [ot aT ix] ce) veri du. 3)