IEEE T R A N S A C T I 0 N S O N MICROWAVE THEORY AND TECHNIQUES A PUBLICATION OF THE IEEE MICROWAVE THEORY AND TECHNIQUES SOCIETY DECEMBER 1996 VOLUME 44 NUMBER 12A IETMAB (ISSN 0018-9480) [email protected] PART I OF TWO PARTS PAPERS Development of a general symmetrical condensed node for the TLM method - V. Trenkic ; C. Christopoulos ; T.M. Benson 2129 - 2135 Modeling and performance of a 100-element pHEMT grid amplifier – M.P. De Lisio ; S.W. Duncan ; Der-Wei Tu ; Cheh-Ming Liu ; A. Moussessian ; J.J. Rosenberg ; D.B. Rutledge 2136 - 2144 Hybrid method solution of scattering by conducting cylinders (TM case) - T. Roy ; T.K. Sarkar ; A.R. Djordjevic ; M. Salazar-Palma 2145 - 2151 Analytical dispersion analysis of loaded periodic circuits using the generalized scattering matrix - W.S. Best ; R.J. Riegert ; L.C. Goodrich 2152 - 2158 Dynamic shape of the depletion layer of a submillimeter-wave Schottky varactor - J.T. Louhi ; A.V. Raisanen 2159 - 2165 Order-recursive Gaussian elimination (ORGE) and efficient CAD of microwave circuits - P. Misra ; K. Naishadham 2166 - 2173 A coaxial 0.5-18 GHz near electric field measurement system for planar microwave circuits using integrated probes - T.P. Budka ; S.D. Waclawik ; G.M. Rebeiz 2174 - 2184 Numerical simulation of the power density distribution generated in a multimode cavity by using the method of lines technique to solve directly for the electric field - Huawei Zhao ; I. Turner ; Fa-Wang Liu 2185 - 2194 On the application of finite methods in time domain to anisotropic dielectric waveguides – S.G. Garcia ; T.M. Hung-Bao ; R.G. Martin ; B.G. Olmedo 2195 - 2206 Accurate and efficient circuit simulation with lumped-element FDTD technique - P. Ciampolini ; P. Mezzanotte ; L. Roselli ; R. Sorrentino 2207 - 2215 Generalized perfectly matched layer for the absorption of propagating and evanescent waves in lossless and lossy media Jiayuan Fang ; Zhonghua Wu 2216 - 2222 Analytical expansion of the dispersion relation for TLM condensed nodes - V. Trenkic ; C. Christopoulos ; T.M. Benson 2223 - 2230 Scattering of millimeter waves by metallic strip gratings on an optically plasma-induced semiconductor slab - K. Nishimura ; M. Tsutsumi 2231 - 2237 Electric screen Jauman absorber design algorithms - L.J. Du Toit ; J.H. Cloete 2238 - 2245 Analysis and linearization of a broadband microwave phase modulator using Volterra system approach - P. Celka ; M.J. Hasler ; A. Azizi 2246 - 2255 Application of a coupled-integral-equations technique to ridged waveguides - S. Amari ; J. Bornemann ; R. Vahldieck 2256 - 2264 Oscillator-type active-integrated antenna: the leaky-mode approach - Guang-Jong Chou ; C.-K.C. Tzuang 2265 - 2272 Intermodulation distortion in Kahn-technique transmitters - F.H. Raab 2273 - 2278 An analytical model for the photodetection mechanisms in high-electron mobility transistors – M.A. Romero ; M.A.G. Martinez ; P.R. Herczfeld 2279 - 2287 Stokes phenomenon in the development of microstrip Green's function and its ramifications - D. Chatterjee ; R.G. Plumb 2288 - 2290 A fast algorithm for computing field radiated by an insulated dipole antenna in dissipative medium - Lin-Kun Wu ; D. Wen-Feng Su ; Bin-Chyi Tseng 2290 - 2293 A fast integral equation technique for shielded planar circuits defined on nonuniform meshes – G.V. Eleftheriades ; J.R. Mosig ; M. Guglielmi 2293 - 2296 Analysis of electromagnetic boundary-value problems in inhomogenous media with the method of lines - A. Kornatz ; R. Pregla 2296 - 2299 New reciprocity theorems for chiral, nonactive, and bi-isotropic media - C. Monzon 2299 - 2301 (end) IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL. 44, NO. 12, DECEMBER 1996 2129 Development of a General Symmetrical Condensed Node for the TLM Method Vladica Trenkic, Member, IEEE, Christos Christopoulos, Member, IEEE, and Trevor M. Benson, Member, IEEE Abstruct- A general symmetrical condensed node (GSCN) modeling of the capacitance and inductance of the medium for the transmission line modeling (TLM) method, with six and preservation of impulse synchronism. In Section 111-A, the different link line characteristic impedances, six stubs, and six scattering matrix for the GSCN is derived from the equivalent lossy elements is described for the first time. It unifies all the network model [9], [lo], which formulates scattering in a currently available condensed nodes into a single formulation and provides the basis for the derivation of an infinite set of new node having six different link-line impedances, three open- nodes, including nodes with improved numerical characteristics. circuit stubs, three short-circuit stubs, and six lossy elements The GSCN is derived in two ways: 1) from an equivalent network for modeling electric and magnetic losses. Applying particular model and 2) directly from Maxwell’s equations using central constraints to the GSCN, the scattering matrices for all existing differencing and averaging. The direct correspondence estab- condensed nodes [1]-(81 and their parameters are derived from lished between the GSCN TLM and a finite difference scheme for Maxwell’s equations provides a rigorous theoretical foundation the formulation of the GSCN presented here. for all available TLM methods with condensed nodes. A formal equivalence between the SCN TLM and the time- domain finite-difference (FD-TD) method was first established in [ll]. Recently, it was shown that the SCN TLM can I. INTRODUCTION T be derived directly from Maxwell’s equations applying the HE SYMMETRICAL condensed node (SCN) [ll has Method of Moments (MOM) [12]. The two above references, been the basis of the three-dimensional (3-D) trans- however, consider only the simple 12-port SCN, which cannot mission line modeling (TLM) method for many years. To model inhomogeneous and lossy media and is restricted only allow for the modeling of general lossless materials and to the cubic cells. More practical schemes, namely the stub- nonuniform grading of the mesh cells, the basic 12-port SCN is loaded SCN and the HSCN, were only very recently derived augmented by three open- and three short-circuit stubs [ll. In a directly from Maxwell’s equations, using central differencing development of the SCN referred to as the hybrid symmetrical and averaging [13]. A direct theoretical derivation of the Type condensed node (HSCN) [2], the characteristic impedances of I1 HSCN and the SSCN has not been presented in the literature the link lines are varied to account for mesh grading and to so far. model magnetic properties of the mesh, and three open-circuit In Section 111-B, we derive the GSCN directly from stubs are used to make up for any deficit in the capacitance. Maxwell’s equations, using principles established in [ 131. This A complementary HSCN, denoted as the Type I1 HSCN [3], derivation gives the same results for the field components in was recently developed using different link-line impedances the GSCN as obtained using the network model in Section to model electric properties and short-circuit stubs to correct 111-A, thus offering further evidence of the soundness of for any deficit in inductance. In a further recent development the method. Since the symmetrical super-condensed node is of the SCN, referred to as the symmetrical super-condensed derived from the GSCN by removing stubs, a field-based node (SSCN) [4]-[6], stubs are removed all together and all theoretical foundation to the SSCN is also established for the medium parameters are modeled by varying the characteristic first time. impedances of the link lines. Modifications of the SCN to account for electric and magnetic losses are described in [7] 11. GENERALT LM CONSTITUTIVREE LATIONS and [8] and can be readily applied to all other condensed nodes. In this paper, we present the development of a general The total capacitance and inductance of the block of space symmetrical condensed node (GSCN) that unifies all of the -w_ith linear dimensions Ax, Ay, AZ and material properties existing condensed node schemes [ 11-[8] into a single scheme Z,p defined as diagonal tensors and provides a template for the derivation of new nodes with improved propagation properties. A general formulation of the -- Erx 0 0 Prz 0 0 link and stub parameters is presented in Section I1 that can be used in connection with any TLM node. These parameters Prz are determined by a set of 12 equations that ensure proper modeled by a TLM node, in the i direction are given as [14] Manuscnpt received June 10, 1995; revised August 26, 1996. This work AjA k AjAIC c; was supported in part by the Engineenng and Physical Sciences Research = Eo&,,- Ai L; = POCLTZ- ai (2) Council, U.K. ingT, hUe naiuvtehrosirtsy aorfe Nwoittht inthgeh aDmep, aNrtGm7e n2tR oDf ENloetcttmncgahla amn,d U E.Kle.c tronic Engineer- where i,j ,I C E {x,y , Z} and i # j,I C. Equations (2) must hold Publisher Item Identifier S 0018-9480(96)08478-5. for any TLM node, constructed by an arbitrary combination 0018-9480/96$05.00 0 1996 IEEE 2130 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUES, VOL 44, NO 1 on these restrictions, we gi classification of the 3-D TLM condensed nodes use ime-domain schemes. 1) Stub-loaded nodes use the same chara for all link lines, which sets six co L,,/C,, = const. 2) Hybrid nodes use e that three extra conditions are given by L; = 0 or Ci = 0. The other three conditions are obtained by field components the previous three spec1 teristic admittanceslimpedanceso f the link and stub lines must be determined, and they ar as [15] Fig 1 3-D TLM symmewical condensed node (SCN) and stubs, and we refer to them as the general TLM constitutive relations. Consider the TLM symmetrical condensed node depicted in an 2-directed j-polarized link line and Yo,,Z ,,a re t Fig. 1. Let the distributed capacitance and inductance of an z- directed, j-polarized link line be denoted by indexes according to their direction and polarization as C,, and Lz,.T he total capacitance of an open-circuit stub and the total inductance of contributing to the cell’s capacitance and vely, in the a direction, are denoted as C; and Li. The general TLM constitutive relations (2) applied to the condensed node can be written as ajak + c,,aj + c: C~,A~C = EOEr,- (3) Qa In time-domain TLM, losses are modeled by inlroducing matched stubs, loaded to the nodes at the scattering points [ 1 Their presence does not affect the general system of The six equations defined in (3) and (4) by using all possible Given the effective electric and magnetic conductiv combinations of z,~,k E {x,y,z}, represent the basis for the and in the 2 direction, the parameters correct modeling of the medium using a generally graded TLM are defined as 171, [8] mesh with condensed nodes. They contain eighteen unknown parameters, namely C,, and L,, of the six link lines, C; of it stubs, and Li of the three short-circuit re are 12 degrees of freedom that can be lation of particular types of symmetrical condensed nodes. In time-domain TLM schemes, time synchronism must be (4) that the contributions t maintained in the mesh, i.e., the time step At must be the same throughout and therefore six more conditions are imposed in six lossy elements, and (5) With these extra conditions, six degrees of freedom still remain n solving (3)-(5). combination of stubs and can be readily shown that the link line and stub parameters of the existing 3-D time-domain condensed nodes can be obtained by imposing different constraints to (3)-(5). Based two different approaches, namely: 1) an equivalent network TRENKIC et al.: GENERAL SYMMETRICAL. CONDENSED NODE FOR THE TLM METHOD 2131 - - 16 17 18 1 0 0 i,, 2 0 -i,, 0 3 0 0 -i,, 4 if, 0 0 5 -ayz 0 0 6 0 i., 0 7 ", 0 0 8 -82, 0 0 9 0 i,, 0 10 0 -is, 0 11 0 0 i,, S= -12 0 0 -iv, 13 0 0 0 14 0 0 0 -15 o o o O e,, e,, e,, O o e,, 0- o I o O h, 0 0 0 16 0 0 0 f z - f z 0 f z - f , 0 0 0 0 1 0 0 0 j, 0 0 17 O - f y O 0 0 f, 0 0 f u - f y O 0 0 0 0 0 j, 0 -18 fz O - f , O 0 0 0 0 0 0 f , - f , O 0 0 0 0 j, 19 IC,, IC,, 0 0 0 0 0 0 IC,, 0 0 IC,, E, 0 0 0 0 0 20 0 0 IC,, IC,, 0 0 0 IC,, 0 0 B,, 0 0 I, 0 0 0 0 -21 0 0 0 0 IC,, t,, IC,, 0 0 4, 0 0 0 0 E, 0 0 0 22 0 0 0 m,-m,O m,-m,O 0 0 0 0 0 0 n, 0 0 23 0 - m , O 0 0 my 0 0 mu-m,O 0 0 0 0 '24 m, 0 - m , O 0 0 0 0 0 0 m,-m,O 0 0 Fig. 2. Scattering matrix of the general symmetrical condensed node (GSCN). (The first row and column are not part of the matrix. They give the port numbering for convenience.) model and 2) central differencing and averaging of Maxwell's {(~,z,y),(y,~,z),(z,y,~)}T.h e equivalent voltage in the i equations. direction, V,, is given by A. Derivation from the Network Model Scattering in a condensed node scheme is traditionally de- rived by imposing unitary and other conditions on its scattering matrix [l]. However, a network model introduced in [9] and and the equivalent current contributing to the magnetic field generalized and validated in [lo] shows that scattering in in the i direction, Ii, is given by condensed nodes can be conveniently described by scattering equations derived directly from the set of "equivalent" shunt and series circuits. For historical reasons and in order to allow comparisons with previous nodes, we also present the complete where (%.A k) E {(TY , z),( Y,z ,2),( z,x,Y )}. The voltages reflected to stubs and lossy elements are given as scattering matrix for the GSCN. For the same reasons, we keep the original node port numbering scheme [l], [15], although v; = v, - v;, (14) more elegant numbering schemes described recently [ 171, [ 181 v,l;= LZ,, + v;, (15) yield a more symmetrical form of the scattering matrix. V,T, =v, (16) In the notation used here a voltage coming from the negative v;, side of the node (assuming the origin of coordinates at the cen- = R*,L (17) ter of the node) along an i-directed, j-polarized transmission where i E (2,y , z}. line is denoted as Kn,, whereas a voltage coming from the Using the scattering equations (10)-(17), we obtain the positive side of the same line is given as V,,, (i,j E {x,y , z} scattering matrix for the GSCN, given in Fig. 2. Because there and i # j). Voltages on the open- and short-circuit stubs are are no incident voltages from lossy elements, the matrix S V,, and V,,, whereas the voltages on electric and magnetic is written as a 24 x 18 matrix rather then a full 24 x 24 loss elements are V,, and Vm,. In all cases, voltages incident square matrix with zero columns 19-24. The elements of the on the node have superscript i while reflected voltages have scattering matrix S shown in Fig. 2 are superscript T. By following the principles established in [9] and [lo], the azj =Qj - b,, - 4 3 b,, = Q,Ck3 + complete scattering procedure in the GSCN can be described cZ3 = Q3 - b,, dZ3 - 1 4, = pkiz, by the following equations: fk =2(1 - Pk - uk) e,, = bk, yn,= v, f IkZ,, - y;, (10) 9, =2(1 - Qj - w,) it, = d,, ypj= v, F 42,, - yn, (11) h, = g, - 1 jk = 1 - fk where the upper and lower signs apply, respectively, for k,, = ea, 1, = QJ indexes (i,j,k)E {(~,Y,~),(Y,~,~),(~an,d~ (,iY,j),}k )E mk =2uk nk = -mk (18) 2132 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TEC QUES, VOL 44, NO. 12, DECEMBE with eliminating partitions of S related to the short-circuit stubs, the scattering matrix of the HSCN 121, [3] is derived. The (19) scattering matrix for the Type I1 can be derived from 5 ’ by setting yZ, = Yk,, Y (20) partitions related to the open-circ The scattering matrix for th GSCN by eliminating rowkolu (21) Yo, = Zsi, = 0. If a lossless medium is modeled, then the SSCN scattering matrix i y the submatrix Sln, with elements a,, = 1 - b, C2, and d,, = Lt3, as derived also for frequency-domain impedances [20],[ 21]. B. Derivation from Maxwell’s Equations where indexes i,j ,k take all possible combinations of x,y , z. We now derive the GSCN TLM scheme directly from Maxwell’s equations, by employing a procedure similar to that can be written in the following partitioned form, where each submatrix represents one of the matrices used in 1131. Maxwell’s equations, written in the Cartesian coordinate system, are given as outlined in Fig. 2 1 dE, dHk Pl(n12 x12) s(G1S2+xl?3l ) ’S(S1i2lnx 3) &Q&,,- dt 1~ d? - _adH_k’ - oe2E2 Six equations are contained in (2 is introduced as The mapping given by (27) will be The indexes In, os, ss, el, ml, indicating the physical purpose this subsection. It allows the rotation of dummy indexes of each submatrix, stand for link line, open-circuit stub, short- i,j ,5 in an arbitrary exp circuit stub, electric loss, and magnetic loss voltage ports, F(i,j,k ),F (j,k ,i ) and respectively. The superscripts of the submatrices define their After performing coor size. Some partitions can be removed from the matrix if stubs by or lossy elements are not . For example, if short-circuit stubs and lossy elements are not used in the node, all partitions with indexes ss, el and ml can be removed from S giving a 15 x 15 matrix. where k E {z, y, x}, Maxw By removing lossy elements from the node, setting G,, = 0 ten as and R,k = 0 in (21)-(24) and eliminating rows 19-24 in the scattering matrix S, it can be shown that the lossless GSCN conserves energy by confirming that STYS = Y 1191, where Y is a diagonal matrix with elements corresponding to the characteristic admittances of link lines and stubs. The matrix S has identical structure to the scattering matrix for the stub-loaded SCN [E].B y setting the characteristic impedances of link lines equal to the intrinsic impedance of the background medium, i.e., letting Z,, = 20,th e elements of the CSCN and the SCN scattering matrix become equal, as expected. If an homogeneous lossless medium is modeled on a uniform mesh, then stubs and lossy elements can be eliminated by setting Yo, = Z,, = G,, = R,k = 0 in (21)-(24) and the In this formulation, Y,Z partition SI, of the matrix S becomes equal to the original sidered as coefficients i 12-port SCN matrix [l] with the elements a,, = e,, = 0 and to conveniently represent b,, = d,, = 112. sions. They are chosen in Similarly, by setting Z,, = Z,, and ZSk = 0, for all grees of freedom demand combinations of indexes 2, j,k E {x,y , z}, i # j,k and (7)-(9), these coefficients TRENKIC et al.: GENERAL SYMMETRICAL CONDENSED NODE FOR THE TLM METHOD 2133 the link and stub admittanceshmpedances of the GSCN. In the where the upper and lower signs, respectively, correspond to derivation of the SCN presented in [12] these coefficients were the incident and reflected voltages. not explicitly introduced since there were no stubs and all the A similar transformation which establishes a correspon- link-line impedances were identical. dence between the electrical and magnetic field components Following [ 131 we introduce mixed space-time coordinates at the cell center and the incident and reflected voltages on as the open and short stubs is given as &=i+Er;& =i-t" nr(l/a)K(c) = 2 . nv:;" for k E {x,y,z} and rewrite (30) as n~(l/2)L(C)Z,z = T 2 * n q " . (33) By substituting transformations (32) and (33) into difference 2 equations obtained from (3 1) we derive +- aK yo2 -+GeZV,=O 2 at z,i ari + - + Rm;I; 0. These two equations represent, respectively, the charge and 2 7at the flux conservation laws in the i-direction [22]. By using central-differencing at point (n,p ,q , T) [13], where Another set of finite difference equations can be obtained + n is a time and p, q, r are space coordinates, a set of finite from (30) by using central differencing at point (n difference equations can be obtained from (31). The space 1/2, P, 4, coordinates of the cell's boundaries and the cell's center are referred to as (Pf i,q,r)= (x*'> (PAf = (I/*> ;, (p,4,rf =(z*) (P,W)= (e). The terms associated with the mixed coordinates (k are differenced as, for example whereas terms associated with vk are differenced as, for example + - [n+(l/a)V,(k-) n+(l/z)lj(k-)Zkzl. The terms associated with the time coordinate t^ are differenced as, for example av, at^ - - n+(l/Z)K(C) - n-(1/2)I/(C). Similarly as in [13], we now introduce the variable transfor- mation in order to establish relationships between the electric and magnetic field components at the cell boundaries and the incident and reflected voltages at the cell center: 2134 IEEE TRANSACTIONS ON MICROWAVE THEORY A By substituting (37)-(39) into (36) we obtain which after using (33) gi at the time step n as + qz+ +2 “et) [n+l%(C)- n%(C)I + + + n+lTpz) = Ykz(n+lV;nzn +lVZpz) q z ( n + l ~ n z These expressions are equivalent to (14) and (15) obtained from the network model. Because there are no incident pulses “reflected” to these ele -- - [n V‘ jnk - n V‘ gpk -nv;n3- tnVlpP+3n v,T, Therefore, the complete ering equations is + Em, . ~L(c)] given by (45)-(48), which + for the GSCN already de -[n+lynk - n+lTJ;.& - n+lVlng n+lVZpj + presented in Fig. 2. n+lV,,I. (41) Finally, substituting (34) and (35) into (40) and (41) we obtain voltages and currents at the center of the node at time The formulation of the g step n as given here is derived both and from Maxwell’s equations. T in [lo] that the network mode of decoupled series and shunt pplying the charge which are equivalent to (12) and (13) derived from the network model. The scattering matrix for the GSCN can be derived by averaging appropriate field components at point (n,p , q, T) condensed nodes into a sin in the mixed coordinate system. The procedure is similar to can be derived from the that used in [13]. For example, by averaging the component additional conditions to the + (% IgZ kz) with respect to qk we obtain + 2[nV,(c) nIg(C)Zkz] foundation to all nodes contained in the formulation of the + -- [n-(l/z)V,(k+) n-(1,2)J3(k+)Zkz] GSCN. This gives for the first time a fie + + of the symmetrical super-condensed nod [n+(l/z)V,(k-) n+(1/2)4(k-)ZkZl. (44) be derived from the GSCN by removing all stubs and Substituting (32) into (44) we derive a reflected voltage at the the theoretical foundation for any new GSCN-ba time step n as node scheme with an arbitrary number of stubs an link lines. Following the discussion about accuracy given in + Vlnz= K(.) .r,(C)Zkz - VZpz (45) [13], one may conclude th where x(c) and I,(c) are defined from (42) and (43). Si&- larly, by averaging the component (K IjZkz)w ith respect - equations gives clear insight in field components and the trans model. All field compone = K(c)- r;(C)Zkz- Vlnz. (46) of a cell at the time mo transformation (28) and ons for all link lines can be derived in this manner and they are found to be in agreement with the + scattering equations (10) and (11) derived from the network n - 1/2, n l/2,. . . , by and the excitation in the G model. Averaging V, and I, with r ect to t^ at the point (n,p,q , r) The derivation of the gives + 2. nAz(c) = n-(l/z)A,(c) n+(l/z)A,(c) A E W,II TRENKIC et al.: GENERAL SYMMETRICAL CONDENSED NODE FOR THE TLM METHOD 2135 the SSCN, requires the least computer storage and is the most [20] D. P. Johns, “An improved node for frequency-domam TLM-The efficient but its implementation is more complicated due to ‘Distributed Node’,’’ Electron. Lett., vol. 30, no. 6, pp. 500-502, Mar. 1994. the presence of link lines with different impedances. The only [21] P. Berrim and K. Wu, “A new frequency domain symmetrical condensed reason why one should combine the complexity of such a TLM node,” IEEE Microwave Guided Wave Lett., vol. 4, no. 6, p__p . scheme with the extra storage of stubs is to achieve better 180-182, June 1994. J-22-1 J . L. Herring and C. Christoooulos, “The auulication of different propagation characteristics. The possibilities opened up by meshing techIni ques to EMC problems,” in 9tIh_ A nnu. Rev. Prog. in the GSCN formulation have been exploited to develop and Applied Comp. in EM, NPS Monterey, CA, 1993, pp. 155-762. [23] V. Trenkic, C. Christopoulos, and T. M. Benson, “Advanced implement the matched SCN (MSCN) [23] and the adaptable node formulations in TLM-The matched symmetrical condensed SCN (ASCN) [24]. These nodes use combinations of link node (MSCN),” in Proc. 12th Annu. Rev. Prog. in Applied Comp. and stub parameters which offer improved accuracy and mini- Electromagn., Monterey, CA, Mar. 18-22, 1996, vol. I, pp. 246-253. [24] -, “Advanced node formulations in TLM-The adaptable symmet- mized dispersion error in modeling inhomogeneous microwave rical condensed node,” IEEE Trans. Microwave Theory Tech., vol. 44, circuits [23], [24]. no. 12, pt. 11, pp. 2473-2478, Dec. 1996. REFERENCES P. B. Johns, “A symmetrical condensed node for the TLM method,” Vladica ’Ikenkic (M’96) was born in Aleksinac, Yu- IEEE Trans. Microwave Theory Tech., vol. MTT-35, no. 4, pp. 370-377, goslavia, in 1968. He received the Dipl.Ing. degree Apr. 1987. in electncal engineering with computer science from R. A. Scaramuzza and A. J. Lowery, “Hybrid symmetrical condensed the University of NiS, NiS, Yugoslavia, in 1992 and node for TLM method,” Electron. Lett., vol. 26, no. 23, pp. 1947-1949, the Ph.D. degree from the University of Nottingham, Nov. 1990. Nottingham, U.K., in 1995. P. Bemni and K. Wu, “A pair of hybrid symmetrical condensed TLM Since 1992, he has been a Research Assistant, nodes,” IEEE Microwave Guided Wave Lett., vol. 4, no. 7, pp. 244-246, Department of Electrical and Electronic Engineer- July 1994. ing, University of Nottingham. His research interests V. Trenkic, C. Christopoulos, and T. M. Benson, “New symmetrical include numerical modeling using the transmission super-condensed node for the TLM method,” Electron. Lett., vol. 30, line modeling method and its implementation to no. 4, pp. 329-330, Feb. 1994. electromagnetic compatibility and microwave heating problems. -, “Generally graded TLM mesh using the symmetrical super- Dr. Trenkic received the IEE Electronics Letters Premium award in 1995. condensed node,” Electron. Lett., vol. 30, no. 10, pp. 795-797, May 1994. -, “Theory of the symmetrical super-condensed node for the TLM method,” IEEE Trans. Microwave Theory Tech., vol. 43, no. 6, pp. 1342-1348, June 1995. P. Naylor and R. A. Desai, “New three dimensional symmetrical Christos Christopoulos (M’92) was born in Patras, condensed lossy node for solution of electromagnetic wave problems Greece, in 1946. He received the Diploma in elec- by TLM,” Electron. Lett., vol. 26, no. 7, pp 492493, Mar 1990. trical and mechanical engineering from the National F. J. German, G. K. Gothard, and L. S. Riggs, “Modeling of materials Technical University of Athens, Athens, Greece, in with electric and magnetic losses with the symmetrical condensed TLM 1969 and the MSc. and D.Phi1. degrees from the method,” Electron. Lett., vol. 26, no. 16, pp. 1307-1308, Aug. 1990. University of Sussex, Sussex, U.K., in 1970 and P. Naylor and R. Ait-Sadi, “Simple method for determining 3-D TLM 1975, respectively. nodal scattenng in nonscalar problems,” Electron. Lett., vol. 28, no. 25, In 1974, he joined the Arc Research Project pp. 2353-2354, Dec. 1992. of the University of Liverpool, Liverpool, U.K., V. Trenkic, C. Christopoulos, and T. M. Benson, “Simple and elegant and spent two years working on vacuum arcs and formulation of scattering in TLM nodes,” Electron. Lett., vol. 29, no. breakdown while on attachment to the UKAEA 18, pp. 1651-1652, Sept. 1993. Culham Laboratories. In 1976, he joined the University of Durham, U.K., 2. Chen, M. N. Ney, and W. J. R. Hoefer, “A new finite-difference as a Senior Demonstrator in Electncal Engineenng Science. In October 1978, time-domain formulation and its equivalence with the TLM symmetncal he joined the Department of Electncal and Electronic Engineering, University condensed node,” IEEE Trans. Microwave Theory Tech., vol. 39, no. 12, of Nottingham, Nottingham, U.K., where he is now Professor of Electrical pp. 2160-2169, Dec. 1991 Engineering. His research interests are in electrical discharges and plasmas, M. Kmmpholz and P. Russer, “A field theorehcal derivation of TLM,” electromagnetic compatibility,e lectromagnetics, and protection and simulation IEEE Trans. Microwave Theory Tech., vol. 42, no. 9, pp. 1660-1668, of power networks. Sept. 1994. Dr. Christopoulos received the IEE Snell Premium and IEE Electronics H. Jm and R. Vahldieck, “Direct derivations of TLM symmetrical con- Letters Premum awards in 1995 densed node and hybrid symmetrical condensed node from Maxwell’s equations using centered differencing and averaging,” IEEE Trans. Microwave Theory Tech., vol. 42, no. 12, pt. 2, pp. 2554-2561, Dec. 1994. D. A. Al-Mukhtar and J. E. Sitch. “Transmission-line matrix method Trevor M. Benson (M’96) was bom in 1958. He with irregularly graded space,” IEE Proc., pt. H, vol. 128, no. 6, p.p.. received the first-class honors degree in physics and 299-305,-Dec: G81. C. Christoooulos, The TI ransmission-Line Modeling (TLM) Method. the Clark Prize in experimental physics from the University of Sheffield, Sheffield, U K., in 1979 and Piscataway: NJ: IEEE Press, 1995. S. Akhtarzad and P. B. Johns, “Generalized elements for TLM method the Ph D degree from the same university in 1982. of numencal analysis,” Proc. IEE, vol. 122, no. 12, pp 1349-1352, After spending over six years as a Lecturer at Dec. 1975. University College, Cardiff, Wales, he joined the P. P. M. So and W. J. R Hoefer, “A new look at the 3D condensed University of Nottmgham, Nottmgham, U.K., as a node TLM scattering,” in IEEE Int. Microwave Symp. Dig., Atlanta, Ga, Senior Lecturer in Electncal and Electronic Engi- June 1993, pp. 1443-1446. neering in 1989, was promoted to the post of Reader V. Trenkic, T. M. Benson, and C. Christopoulos, “Dispersion analysis in 1994, and Professor of Optoelectronics in 1996. of a TLM mesh using a new scattenng matnx formulation,” IEEE His current research interests include experimental and numencal studies of Microwave Guided Wave Lett., vol. 5, no. 3, pp. 79-80, Mar. 1995. electromagnetic fields and waves with particular emphasis on propagation in R. E. Collin, Foundations for Microwave Engineering, 2nd ed. New ophcal waveguides and electromagnetic compatibility. York: McGraw-Hill, 1992. Dr. Benson received the IEE Electronics Letters Premium Award in 1995 2136 IEEE TRANSACTIONS ON MICROWAVE THEORY AND TECHNIQUE ling and Performance of a lement pHEMT Grid Ampli Michael P. De Lisio, Member, IEEE, Scott W. Duncan, Member, IEEE, Der-Wei Tu, Cheh-Ming Liu, Alina Moussessian, James J. Rosenberg, and David B. Rutledg Abstruct- A 100-element hybrid grid amplifier has been fab- ricated. The active devices in the grid are custom-made pseudo- ransistor (pHEMT) differential- for gain analysis and compare rid includes stabilizing resistors the grid has a peak gain of 10 a gain of 12 dB when tuned for The maximum andwidth is 15% at 9 GHz. The m~ni~unmois e figure is 3 dB. The maximum saturated output ljlput &=am power is 3.7 W, with a peak power-added efficiency of 12%. are a significant improvement over previous grid Output Beam ed on heterojunction bipolar transistors (HBT’s). E I. INTRODUCTION UASI-OPTICAL amplifiers combine the output powers of many solid-state devices in free space, eliminating the losses associated with waveguide or transmission- Fig 1 A gnd amplifier A honzontally polarized input beam is incident line combiners. The first quasi-optical amplifier was a 25- from the left The output beam is vertically polarized and is radiated to the nght The polarizers independently tune the output and input circuits element grid amplifier [l]. A grid amplifier is an array of closely-spaced differential pairs of transistors. Fig. 1 shows the approach. A horizontally polarized input beam excites rf currents on the input leads of the grid. This drives the report on the modeling and performance of a 100-element transistor pair in the differential mode. Currents on the output X-band grid amplifier using pHEMT leads produce a vertically polarized output beam. Metal-strip photograph of the assembled grid is sho polarizers provide independent tuning of the input and output circuits. Other types of quasi-optical amplifiers using patch 11. DEVICED ESCIUPTI antennas [2]-[4], back-to-back integrated horn antennas [SI, [6], folded slot antennas [7],a nd probe antennas [8] have The differential-pair c been demonstrated. The largest number of devices have been Martin Laboratories. incorporated in a 100-element HBT grid amplifier [9]. GaAs/InGaAs/GaAs pHEMT’s. The total gat Recently, pHEMT technology has developed rapidly. transistor is 100 pm distributed amo Millimeter-wave pHEMT’s are capable of high gain, high gates were defined by an output powers, and low noise figure, making them the device and a bi-level resist. Th structure is a of choice for high-frequency applications [lo]-[ 121. Here we double-doped heteros Manuscnpt received April 10, 1995, revised September 13, 1995. ‘Ihs details about the devices can be found in [lo]. research was supported by the Army Research Office and Lockheed Martln The differential pair chip layout is Laboratories. M P De Lisio held NSF and AASERT fellowships M P. De Lisio is with the Department of Electncal Engineering, Umversity sources of two pHEMT’s are tied to of Hawaii-Mwoa, Honolulu, HI 96822 USA pair. Unlike the HBT’s in previous g S W Duncan is with Sanders, a Lockheed Marti terminal can be extern 03061 USA D -W. Tu is with the Laboratory for Physical this gate control bias to Maryland, College Park, MD 20783 USA. because the gate draws very little bias current. The 0.5-pF C -M Liu is with the Rockwell Science Center, Rockwell Intemahonal capacitor and 2-kR resistor are feedback elements to stabilize Corp , Thousand Oaks, CA 91385 USA A. Moussessian and D. B. Rutledge are with the Department of Electncal the pHEMT’s. The two 100-0 source resistors are intended to Engineering, California Institute of Technology, Pasadena, CA 91 125 USA reduce common-mode gain. These resistors should not effect J. J. Rosenberg is with the Department of Engineenng, Harvey Mudd the differential-mode the overall College, Claremont, CA 91711 USA. Publisher Item Identifier S 0018-9480(96)08479-7. efficiency of the grid 001 8-9480/96$05 00 0 1996 IEEE De LIS10 et al.: MODELING AND PERFORMANCE OF A 100-ELEMENT pHEMPT GRID AMPLIFER 2131 Fig. 2. Photograph of the amplifier grid. The grid is a 10 x 10 array of pHEMT differential pairs. The active area of the grid is 7.3 cm on a side. 7.3 mm Fig. 4. The grid amplifier unit cell. The width of the input and output leads is I 0.4 mm. The width of the meandering bias lines is 0.2 mm. Arrows indicate Fig. 3. The differential pair chip layout. the directions of rf currents. 111. GRIDA MPLIFIERM ODELLING is re-radiated from the vertical drain leads. Bias to the drain An important advantage of grid amplifiers is that the unit and source is provided by the thin meandering lines. Gate bias cell primarily determines the driving-point impedances seen is provided along the horizontal gate leads. The resistors in by the device, while the power scales with the grid area. This the gate leads suppress common-mode oscillations and will be allows one to optimize for gain and noise figure by the choice discussed later. of the unit cell and to independently select the grid size to meet The assembled grid amplifier tuned for 10 GHz is shown the total power requirement. Previous grid amplifiers [l], [9] in Fig. 5(a). The grid and polarizers are fabricated on Rogers were designed empirically. We have developed a model for Duroid boards with a relative dielectric constant of 2.2. The the grid amplifier that predicts its performance [13], [14]. output tuner is a Duroid board with E, = 10.5. A simple The unit cell is shown in Fig. 4. The cell size is 7.3 mm transmission-line model for the grid amplifier is shown in on a side. The input beam is coupled to the gates of the Fig. 5(b). For convenience, one-half of the unit cell is used in transistors through the horizontal gate leads. The output beam the analysis, with the result that the characteristic impedances