You are on page 1of 233
JAIN Iota RAYMOND) PIEUEBBERS (@e) ERC PRESS KARL 5S. KUNZ RAYMOND J. LUEBBERS CRC Press Boca Raton London New York Washington, D.C. Tinrary of Congress Cataloging Publication Dal Bowtewrss os ad ae. Pema tle-Mateaic 2 ined ani Later Raed Saracen nos ‘tna en ot mi ds Rm serene seca Senet aero caer oie y a a aa bre cose Bt ‘ete tk an ay a oedema in ny mr yay ec Any Fc, rng at rng yearn soe ESS hes etn et me er se ini Ce L280, Capote Bo Ram a 63 [aye ecicn d cpnar oe ‘Vis the CRC Press Web seat wwwcrepresst Aon ogo US cont it ‘ie i he Ue Sr Ames 7590 AUTHORS Karl S. Kunz, PLD, hus worked in industry, government, ad academia snc eevig his PD i plyscs NMSU in 1911, Now st Pn tty 8 Professor, he specializes in computational eapaerig in parce FDTD applications and developments for modeling erspace stems, atenes, ‘gia aye a pan fname Jot ow ses "His past ssc incade Lawreace Livermore National I shorty (LLL), TRW, Mision Research Corporation, BDM, and White Sands Mi sil Range He wasn charge o te Microwave and EMP Uroup at LLL with {esganchliy fr tri indoor vane range AF TRW be we 4 Tecoclogy ‘Seco Lede and an EMP Group Leader st MRC. Whi at BDM he was tit ‘ARES EMP Simultor Scentst At Wife Sands he worked a8 Dr, Kun i an active constant having worked with LLNL, Chening ‘Lt, Lockheed, BUA, and Dymaleton. He perfarms soe of ts werk at ‘tthe canting firm he founded ath ate, Kane Acces Ines ith his fates, Kaiser. Kun. While a enn Sth served as Departnent Head ‘nom 1986 to 1989 and hs enjoyed an activ esearch caer tought ht Raymond Lubbers, PhD. is Poise of Eerie! Engiceing tthe Peonsylania State Univer le seeived the BSED. degree an the Universi of Cincianat,and MLSE and PAD. dees fm he Ohi State ‘Univesiy where he was a member ofthe EleroSience Laboratory. He hss eon an ceetnca eaglacingfasty meter a Ohio Univers alt ‘Research Semis a the Locteed Plo Alo research abst. He as a0 1 Vising Professor at Toheku University, Seni, der the National ence oundation poy fa US 2apan Copeative Reba, Ur LiebOes has writen series of technical paper in several ares inclaing fqn ele tive sures, psometrial theory of eifiaction, and FDTD. He is Senior ‘Member of the Antena end Prpugton Society of te IEEE, an so the Board of Directors af the Aplin Compuinal Eactrmapines Society He hs chaired a number of IEEE and URSI conference season inling a0 LURSI Geneal Assembly session, ana reviews aces for several tech iowa PREFACE The finite difference time domain (FDTD) approach is rapidly becoming one of the most widely used computational methods in electromagnetics. There are several reasons for this, including the increased availability of low cost but powerful computers, and increasing interest in electromagnetic interactions with complicated geometries, which include penetrable dielectric and/or mag- of the portant pe: method. The fundamentals of FDTD can be grasped easily by undergraduate students, more easily than traditional frequency-domain approaches to electromagnetics. Yet FDTD is capable of computing electromagnetic interac- tions for problem geometries that are extremely difficult to analyze by other methods. It is this combination of simplicity and power that makes FDTD such a popular method. This also allows this book to serve a wide range of potential readers, It can be used to introduce undergraduate students to time domain electromagnetics, in which case only the first two parts of the book need be covered. It can be used in a graduate level course, in which case the mathematical basis of FDTD would be emphasized with topics selected from the sections on special capa- bilities and advanced applications at the discretion of the instructor. Finally, someone who wants to use FDTD to solve a particular problem in electromagnetics can use the book to learn FDTD basics and special capabili- ties necessary for their application. In addition, it can be seen, through the examples in this book, how FDTD has been applied to a variety of problems. The goals of the book are to provide the basic information necessary to apply FDTD to problems in electromagnetics, and to illustrate some of the types of problems that can be analyzed using it. The theoretical and mathemati- cal basis for much of FDTD is included, but the emphasis is on the practical aspects of applying FDTD. While equations are given in the text as needed to develop and understand the method and applications, complete detail is not included. For example, equations usually are not given for all vector components but only for one Tepresentative component. There are two reasons for this. One is to avoid a book filled with pages of almost but not quite identical equations. The other is that the book includes FORTRAN listings of a complete 3-D FDTD program based on the concepts presented in Part 1. This provides better documentation of the actual details of the method than equations included in the text. While a joint effort of the two authors, the primary responsibilities were divided between the authors by chapter. Karl Kunz wrote the majority of Chapters 1 and 2, all chapters in Part 2, the last section of Chapter 10, Chapter 12, all chapters in Part 5, and Appendix A, Raymond Luebbers wrote the majority of the remaining chapters, i.e, Chapter 3, Chapters 7 through 11 (except for the last section of Chapter 10), all chapters in Part 4, and Appendix B, Despite this division of labor we have attempted to make the book consistent throughout in both notation and philosophy. Insane ahs hemes many pol cons ok snmuny cyte mide. ekooaiga f hem ob ba some enue ao inpran 1 emanunmenioned Frienster effet Ka Kus so 5 ters owing is ok Severe mae ect ontibtns. Caper {o's clrely een he werk of ares P Huber, oh Beg ‘agro ae te conte Caples. oth ee pate soe ‘ho Se when hs wr ws dove, Orta who conibued ls SNOT topo nhs bck nie Dre iy and Dev ico Fence SAP cen ant Ke Kemga st Toboke Une. Sens TirunReate povided ty Cris Trcman an! Stan Kobi of Conor Univer n caoorion with Stn Mire and C. Lavoe of Davi ‘aan Lory ae ay ep “try New Mange vie cial fe rin mans SSE noer gests fr improve he Deh TOR ‘une mary pops red eases sp wei ster vite si par nen and eervewhnowelgment Vag ‘ii Rnd ot ow Haven on Capi Pre Bek, Alex Wo, en Taree Ptorse ane Ye, Ron Se Rh td, Rabe si he Sc Sabre Asch, Koni Seveya Tou Un Lid ‘icin tae Cn. ior por fee ropletsBo would fever presen tem Raymond Luebbers Univeral Park PA ‘August 1992 FOREWORD Any tok onthe Site direc tne emia FDTD) tcniue can caput te oof Gt soon ats published We hope ha i te it fone ones 0 TUL am wa treo iy Sue a Tae We Mt Spolgins tothe eater, eve. ae the er wv dy ern ives unde ext We hve ed to prove ome news of nesses Conte ith vane meme fee. "Theil element ae he pic of FT whic are ard the Maxwell quaons thy embody in dried om, th be foaion ‘sing cme fe er exesons whic Can ee ial) 2 tater pea geometry an he ase "bouskepingeeded ones the techni. Our inom sad fit ston cover hese mate td ‘Wechane net sos oe i eligi wel i i TT — tx dina cooping vege propagation ad roping ene ‘eaten ram ory srr Ihese ess sit he Operation the ace ene nahn fi fo the mot pr te sc eg he artical formusing of FDTD we enphar, the ater domo. $f omolaonir eyo mpienr se de, ose compe inthe equations employed than arate, te nl el ami od tif same pling py ih: Pa 3, ih ese ban penton, Stools avo ema wt ss espa meu or trong ie ex Inte scl raf ENT) eon ee whch ny he feo tie We tok te diet nonce what we hve bce peraly “On these sting Bw short a xpiion ate comes ith ny ‘eoionof FDTD srs rome work we der Pa He, weave pled Some a our best wor, extension 0 ease FDTD capable Sone se ‘er exe, earn vegies edge pede ‘natal malig capbietonae wo The resort capes wl totemmah, bare ea epect aber exec nro on wat oft tev Tiicin he god ewe ce will ese oe fy evens inthse mes ser renin ou bck which we woud sdvnced aplication and make of the exons fe per ton, ‘Wetnow tain ie ts wll come ld ews and pease ile i hore ners nny "hot of pe” and le be han tayhing ee wire sone of ie each fot eed how har gh Terard eg eh at FDTD har oe ued fo get ee ial we cet wi come of he thar fess ad som of the sheaves eval to FDTD —s mito cecy and inl A "IHS sng snp vero te cde copes ong My colenue, Profesor Raymond Loebbes, and Have had sting sine pong ts bo opti fo recognae sedan that [mach aie the plneering work he has coatibaled oti effot, mach of which spears inPars and We both ean activ FDTD research and wre worn 0 ‘Tithe many studeat an ends Professor Leters 0 ably acknowledges In the Preface; ter contnbuons have ben invaluable ‘Tou fad Dr. Rane Yor, who tae al his a 1966, we send oor rosa and ou haks eas art. Kunz ‘TABLE OF CONTENTS Chaplet 1. Inodut00 nn ss cer 2 References Chae 3 Referees nnn Caper 4 coer 3 Part 1: Fundamental Concepts Satced Meld POTD Ponti 21 Maxwell Cul Eauatons 22. Sepa Fed Formals 23. pertect Conductor FDTD Fermulaion 24 Perfect Conducior FDTD FORTRAN Code 25 Losy Material Formulation 27 FDTD Code Requirements and Archie. FDTD Basis. 3 Invoducion 33. Time Step Size for Stab 434 Speiying he Incident Fed 35 Buling an Objet i Yee Cel 366 Direct Compuaton of Toul Fs. 37° Radiation Boundary Condon art 2: Mane Appicatons Coupling Bs 43 Enero Pulse Response 44 Measiremen Fain. — 432 FDTD Male ofthe Aina 433 Comparison of Prediction an 44 Imerior Elecsomagntic Shielding 4.01 Proquenctes Above Aperture Cuil ‘5.2 rraqumnses Below Apenure Cut Waveguide apemure Couping vance by F. Amu and Ks Kun Si nod. Ackaowledpment <= Caper. Retrences chaer7 References Chaser. Referees chapter. Chapter 10. 52 Approach - 33 Riniha wns Sa Concanion Lossy Diceeie Searing. nati en 2 Imereti ofthe Seater Feld Metiod 63 _ Net Zane Sper Searing O31 Ikea Feld wn 632 Special Noes. 633 Predictions og ihuman boxy Ameo are Special Capes Far Zone Transformation. 7a ioducton 19 Thee Dimensional Transformation. 23 Two Dime fort ra Summary - Ti"mncton nce 13 82. Fin Order Debye Dapenion eal 43 pant der Drage Dispersion. sons 13k EA Seu Oter Daperive Maes S- Malpe Pls, : 6. Dileena: Methane 29 Stateed Fin Parlin See sen ot ror vo co 63 92 aot 8:3 Bequest Mati 18 94 Recne Elution oe 55 Demonataons conn us Subst Extn. 125 jon Ines = 18s (Copter 11, Nonlinear Loate and Material 1 Invoice, 2 Amenna with Nosiner Bade 121 Intoducto, 122 Types of Visa 129 ample nn 123.1 Model Fidei —— 1232 Movel Respons ~ 1233 Physical Proves isa 1234 nition Bung 2s 124 Resources and Cx of Visti no 237 Referens me Part 4: Advanced Applications (Chapter 13. Far Zone Scaring 131 Inoducion 122 Fundamental 133 Staeae Enos 134 Impedance Shee 105 Disuae to Outer Doanday nnn 384 136 Frequeney Dependent Materials med Rebeece ssissanencas aa 261 (Chapeer14. Aeanas. 68 1 loroducton ana 142 Impedance, Eine, and Gan 265 1143. Mooopole Astana on Conducting Bor 279 44 Shapd-Ead Waveguide Amana. “390 Releemedinnnnnnne nec Chapter 15. Gyrotropic Media... eee) 152. Magntzed Plasma a 153. Mageind Pera vm References 7 : an Chapter 16, Diereace Equations in Genera enn SBD 162 Diferencing Schemes for Hyperbolic Equations .330 163 Permutations on he Leypiog Technique. 333 Ren = ot Chap 7, Sabi, Dpenion, Aza sen $83 172 ‘Sti ne 336 13. Numeric Dispsion| a) UVa: Atty enone AS Reference ee - 346 CCxmper 18. Ouer Radian Bovary Conditions rn 3D 11 Inodcton Seen tT 182 Evolution af ORBCS 38 Mur Over Raia Bosany Co 30 "4 Ake ein of Ove Way Wave Equator ven 386 References ones = . 388 ‘Chaper 19,Aente Farman orn 399 19 Inoaicton.. = 3 102 oud Field Formation nnn 3S 193 Potenial Formulation 0 8A tmpet Schemet nso 3 1095 High Peqoeney Appensination 368 {95:1 Time Doman Smythe Kirchbolf Ajerure [Approxmaion and Maa ages fox High FDTD Frequency Modeling 368 196 Acoustic AnlogSala EqUYaent ences 365 T.6 Eguton of MoU aes 1969 Git La nnn 366 19.63 Conia Equa nnn 366 Referens ss a Anpe .Oe: Corin yt to Rede Dimes 369 Ineeducten 30 AUD ehlindialand Spbecl 370 Releenes Index 2D Cylidiea DIAGSDDAT. NZOUTSD.DAT’ an 435 Chapter 1 INTRODUCTION Of the four forces in nature — strong, weak, electromagnetic, and gravita- tional — the electromagnetic force is the most technologically pervasive. Of the three methods of predicting electromagnetic effects — experiment, analy- sis, and computation — computation is the newest and fastest-growing ap- proach. Of the many approaches to electromagnetic computation, including method of moments, finite difference time domain, finite element, geometric theory of diffraction, and physical optics, the finite difference time domain (FDTD) technique is applicable to the widest range of problems. This text covers the FDTD technique. Emphasis is placed on the separate field formalism, in which the incident field is specified analytically and only the scattered field is determined computationally. This approach is only slightly more complex in its basic implementation than the total field approach and more readily allows for the absorption of scattered fields at the limits of the problem space. The total field can be easily obtained from the combination of be the ccattered ay be found either approach, they are very similar in concept and capabilities. The FDTD technique treats transients (e.g., pulses) in the time domain, and it is applicable over the computationally difficult-to-predict resonance region in which a wave- length is comparable to the interaction object size. As a form of computational engineering, FDTD is part of a three-tier hierarchy consisting of: Computer Science . ‘Stresses the mathematics underlying algorithms as well as the structure and development of the algorithm Computer Engineering . Hardware based on and concerned with hardware architecture and capa- bilities including parallelism and fault tolerance Computational Engineering . Explores various engineering problems via numerical solutions to sys- tems of equations describing the phenomenon or process in question Computational engineering relies on computer science and engineering, but is not hardware, language, or operating system specific. It requires a computer Powerful enough to accomodate the problem in question, running within acceptable times and costs while producing the desired accuracy. Electromagnetic computational engineering encompasses the electromag- netic modeling, simulation, and analysis of the electromagnetic responses of complex systems to various electromagnetic stimuli. It provides an understand- ing of the system response that allows for the better design or modification of the system. ‘The Fine Dioronee Time Domain Method for Electroma ‘The FDTD eciqu offers many vantages aan lecomagnen mx sing, ston, nena Tapes ice Broadband response pefctons centre about he system rename intercon with an objet of any condctviy from that of» perfect enduro el Metal ota fw oF Zero condtvgy Trequcne) depandet coneitve parameters far modeling most Radar coe section (RCS) Surface response Current, pore dy Peveraiowineror coupling “Tesecaptice ae able for 6 ist of vere electromagnet stim covering 4 broad range of reuencies. Typical stil incase EMP terete pu) Ha igh power meron) ro ners The system epoca tinal ae guy ere. Thy cm Be mal large, organic organic, nan exoamspherie envionment 10 &| teranean on, Samples ofthe cvs yes of stems tha an be tees Aerosols ‘Sete ‘ira Homa Serine red antennas hate te ave ana at ayatome together ie that spill the weeelnuhs of nee td the charters system mensions ae usualy Iroducton 4 within an order of magnitude of each other. Thus, the broadband ‘Tepoee precios il pial encompass few sytem sone. yi technigrs employing the geomet teary of cfracton (TD) erasers (FO) ang it TD cnn ep pre pricions thi Pony md alow aol bogie pane peice Sons ofthe FDTD generated tine response predictions. Thus, resonance and Telow-rsonmce predic can be etended te extremely low freuen ‘Aleately, low fequeney versions of FDTD, in which the displacement ‘ents ignored and the equtons become due, can be used extend tow fequeey capable. shor, POTD san spa the cea eset ‘pion over more tun fur orders magnitude in frequency, and with ow sad Bek gee reso gun er of ep amputees becuse only nearest eghbi interactions ae involved. The im fPortnt variables are problem space sein ll requred to adel he eee {nd he numberof eve tps needed. These determine the compater ra tine {nd computation cost Les porta af he material types mele ad the ‘inher of reepore lone mentor. OF learn impute sompute ‘tional apbilty requirements are the ype of smal and he ype of response, ice iesses offre, which gure amdest mount of pospacesng ‘Over lon cll ca be acommdated on personal werkt fora {30 problem pace 100% 1003100 cels lage. At atypia! 10 FDTD cle per ‘wavelength, this space isa [0-vavelength cube. The limits of today's ‘aprcompeiers ae reached trough 10D millon celle with computation ‘ies onthe order of hore "The advange of PDTD can be summarized as its ability work wih ‘ide rangs of frequencies, simul, obj nvronment, reapense leans tnd computers To this Ist can be aed the advanage of computations! icency fr large problems in comparison with ter terniques soc she Iethod of moment epoca wha boat! els ae oie, Ft, the FDTD code, while inerenty volumetric, hs succesflly weated thin pines and hin wire arenas Its accuracy, using a sufciency of cel canbe rade a igh as dsed Conrad, eis cates oa few dence ‘seuracy can be made wth suprising fe ells Final, powerful visa tion tools are beng developed o enhance the users Understand of the ‘Sct physin wideyng the various prcenes tat FDTD can model, silt, and analyze. "The bsisf the FDTD cove ite two Maxwell cu equations in dervave| foam i the ine din These eget ae exyrened i inet ah by means of ental fine deencine. Only neureteiahbor ineracons ned be considered asthe else advanced temporally incre tne steps {ver tcl uf estan shape on expan bere, rel sages te posible, as ae reduced 2- and -D weatmen) 4 ‘he Finite Piference Time Domain Mth for Elecromasnetics Te hou sso te noe ata east sixkind electromagnetic compu onal problems exis Generation (power, deve, Hysons, et) tanec ines waves) Recepon/etectionadiation antennas) Covpingaetangpeneranon + Switchinghooieaiies Auta estan be etd sng FDTD. The it ea tes eis aster ignn he aston of ged pares ot Olan Foner tegen wcheqee imal many tne steps (on creo ie PD nos ae camping ames FDTD my te me compton! apc of ce ten wn tp egy Se co pe sea Th ec wn Sie gamcies rere che a en ai Sei Bcay ce connses renin, nti vn wens ito anon wien FDTD th met of oe, A CW nay igh todo moment fr exp, wil most ely Crate mg wna etm «nee econ ve wr stm ery pn Teh xt hate ofc ‘Tervedied tt exenaly an then compton with FDTD. ‘or reqs races ay fe py ciety eee itr renny esse ean uta eee by FDTD ae {Fino sine ne JT anand ntact by He seen Fundamental concep ‘Basie apicaons Spee eaabiites ‘Advanced appiations ‘Mamnemavea! bat of FD ad aera od “Te firs setion ets the mot base aspects of FDTD, providing he eer wid he fram od the bale rowdus fr FDTD operon. Along with the inrodutondesrbing FDTD's wit an aes of application, the to ‘Shapes of he fit section alow the reader 10 apply PDID 0 a host of robles ing the noodiapeve lary lsc TD cde Hed in Ape fin Be Chapter 2: Seatere Field TD Formulation Dicuss the dsretod Cen ile op ot Marl cu guns wp wih ‘Seopa etd forma bated. te egy we formate ony aca, which inthe limit of init conductivity became Ireodction 5 ‘eect conductors. The rudimeniary computer code requirements and Schtectre eede to soporte frmalzm ins operon ce are ‘resented ts well. Much mre detail about these es give ater pte formulations of Chapter 2. This includes imitations on cell and time step ste, specstying tenia ld and the abet oe analy tia ing the compater reruns require, ad applying om exer Roary ‘condi atthe exes ofthe FDTD computation pace. “The ec ution trate he bse apicton ofthe asc formation of [FDTD given ithe fist section: ‘as ist applied 19 an F-11 strato eaculatethe indeed surface ‘rents and charges fom a simulated EMP fl. This browpht together Sit he element represertave of FDTD modsing, ad for this son tn for some sense of history the modeling effort is icased in deta (Only exterior coins tented wih thi example snd to comple the lection of coupling, intro coupling modeling ofa snp shila eneute by an aperture and cotiing an inkir wis presente ‘The sponse examined above nd below sper cat with eon snd highly resnact behavior noted inthe wo regis A stongly Stesed point isthe lage umber of time pe needed to acutely ‘hracterze the highly resonant behavior ofthe cats induced on the Chapter 5: Waveguide Aperue Coupling. A wane wave propa Ing in e waveguide and cooled ta second waveguide vin Crear sper i eamind in thi ection. Atlee wheter FDTD can be ployed sucess to model waveguide behavior Its shown to work well and FDTD modeling Being rapidly extended int hie modeling + Chap 6: Losy Dice Seaerng, Cpe 2 develops the seatered {eid FDTD tomlin fo pret cect condos andy dle 0 = -a(V-B)/at > V- B= constant V(VxH = ab/a+3) > 0 = AV-D)a + VT + AV-D)/ = 3p / trom continty V-5+ap/—K = 0) + Uie.ny-p) = 0 <0 = constant 2fe-0)-a) vp where we have ued the vector idemity WV - VX A =O, Because the Fels and sures ae inal eto eo FOTO catcalans, tat nial ven=0 veD-p=0 ‘Tratefore. VB and (¥ Dp) aust be 210 fo all mes andthe ee ‘qutions ie silent for FDTD caleulatins. "ove tht lle he vergence equations are At prof he FDTD formal- {sm they canbe used a a est onthe peice eld response oth air forming D = cE and B= i from the pede Fels, te rsuing D and B most say te eng eotons 22 SEPARATE FIELD FORMALISM “Te won equtions 97 an? 8 anh ete to bt ta ik [DTD technige. Alcatel the fle can be expressed as Scared Feld OTD Formulation a Beem = pmo y pram 29) Wm Heme pins yen 210) "Tae tonal for he eprate ld approach hat the inset el cern seas canbe specified nayically Broughou he problem space whl the ‘catered lo ae found computationally and only the sere fil ee 0 important one. The scared els, eunaing om a scaring or iterion ‘ect can te moe ready absorbed than total eld by an oo ration ‘rsnfay coon pid at the pla space etemiten or fae. Tee «specially import in sitaton i which the seated field ae died sd feof och ower ampitoe than the tls. ‘On +more plop evel, thi sparatin alls ate spin the| ineaction proces. “Te seated wave rss on and within the interaction abet in response tothe ince Ri oat satily te appropiate ondary conn ono Within the meracion objec. These bounry conditions are the Maxwell {Suton emsles, hich ne ent of perfect condor eure Eo Pcie in the scr. Fo anything eer than prec onda the Seared fils depend on te const puters ofthe material, The ‘eatered els are suet othe Mawel equations or is media wenn he nein, while sie the main the ay the fe pace Masel ations ‘The incident alvayspropapatsin fe space (ren when passing tough tneitracton bjt seatee material) ands defined a eld it wold Iuisalways possible o combine the scared an incident fi to cbsin the {oa eld and with ital he insight he ol fel behavior provides. Alo ‘eal ela TD quationg a dese hey canbe btined fm he atred fel equations by stig the niente to zero and apn intial end "ons tothe setered (now also tol el. Tis issued Section 36 a toyed in Caper 14 hen total ane Fel ato we aly ules, Tocideo and scare lds mast satisfy the Maxell gutons indepen ‘ely (ve same ber nar mater). Te ieee il opted ete ‘ropsgating in fe space. Pre space could be generalize 1 ufo media (sh, for example) necessary. However, we shal ase fee space for ‘impli. Whi the incien ld travels tough espace trough he Probl space, theo id propagates in fe space onde ie seater tnd the media ofthe seater when its proppang within he seater. Inthe media ofthe seater the tt elds uy Vc Bm = mat ae oH! ey 1 The inte Pierce Time Domain Method for Blectromaeneis Vax! e26™ /a+ 08 am nie tein fds arin he ain ey fee epace cots Vx EM = = 3H aw Vx Hm = ea aw Vx (ps +E) = — palit +}. = neu on) en v5 (i eet) = ef + Hor Seared Fld ia he media Vag = ja for [iu-n jan are0°™] an Vc = ae" e+ oF + [le-s.)aem saree] (18) ie he seater in ie pe et ft iy Veet = 1 for wer defined values of permitivity and conti a hit locate, ‘Using this approach and eting an npr value M dooce Mth pe of lousy deci comained in the FDTD space, we can express the above oat muir fo his material a eee EGRLYOM) = esis siaMADM=DI)-DY * (EveanaT . br a FCRIZIM) = (ppaianjyaIGMAIM)*DT)>BE ~ (eroniaZ ESCTC(M) = EPS(M)/(EPSIM) SIGMA(M)*DT) = = ANCK(M) = SIGMA(a)* BE HSM) + SIGMA(M)* DT) EDEVCN(M)= DT*(EPS(M) ~ EPSO} (EPS(M) +SIGMA(M)*DT) ‘Using these definitions the FORTRAN expression fr Eatin 2.42 for Fin 24 The Finite iference Tome Doan Math for Borramagntce EXS(L.K) = EXS(J,K)* ESCTC(IDONE(.1.K)) = EINCC{IDONE(L,K))*EXIL.K) = FEVCN(IDONEIL.K))* DEXI.J.K) + (47411 )~ HS =1.K))* ECRLYUDONE(L.KI) = (V9(L1.K) -HYSIL L=1)* ECRLZ{IDONELLL.Y) 2a) In a lssy dslectic FDTD code the FORTRAN assignment (236) is executed for ells conning fe pce, (243) s executed forces containing tou alec an! wells wotang «pefet conduct, EXS(LIIO = “EXIM Te excited The coment of te IDONEILK) aay determines Svhich assignment execute fr EXS ina yrs (1K) Yor cel. Arays TIWOUs) and IDTAREGAS) wold sly deci te ata fering wth EYS and E7S respectively, Beeate no magaeticmatials ie preset, FORTRAN assignment 237i execute inal ells H magnetic Inneals ae o De conser, he map verso of (243) san be dxned fm (28) and arays IDFOURILS RY IDFIVEL.1.y and IDSIXG.K) can te wed to self te magnetic teria iterating wi HS, HYS, and ZS, respectvely 2.7 FDTD CODE REQUIREMENTS AND "ARCHITECTURE ‘We have derived the FDTD algorithm for six it component ina 3-D seslina spac that is composed of Yor cell We mast stl eerie wat ‘thtina emputtonal oppor segues and thee code requirement ‘able architec mus be given. ‘Lets fist define the code reuireents. There mast be a main computer roi th mis an ose ou shiv fe reainingsbeotine. Ths “ver seo trench tine calling he sbrotines ia the appropiate order "efoe time sping can begin tere most be a prot space defined, sncating pms sell ic te tp and cen fel. Constant ‘ipl that ned na be cemputed teach ine step may aso be eva tnd stored before time stepping tens. There must bea definition of the Saueier uw tt nt provides, whic consi of coding the information a {o which ell locaton onan materials ober than fe space: Monitor ots ‘rts locations a which responses ae examined most especie along with thetespane pe rage, sna el ower te afar sve taneforma ton ie bing wed te econ) ofthe far zone fells desied must be species ‘twin of cours the core ofthe cove, Most ofthe compu unig te wl be sent inte section, bot represcts smal fation ofthe lines of ede, Satered Feld FDTD Formulation as In aden tothe tine-sepped eld worovines we mest ako nave cue {nation body como ebntine that aah he cater eld he ‘utero poton ofthe problem space. spon cpl we mit ve «way of cuputig time step and then atthe end of the run “ump” it out as a listing oF as a file tor pourocesng. Ine output press may vole Wansanmaton ot the twa ne IT ie othe foe foe cton or etre These code reqement ae summunze a: + Diver 2 Problem spac sup EH field aleoitns + Outer raison boandary conion + Date ses + Far tone tansformation ‘hse queen ao cane presence sy dead! ‘Shine sce te Endl feld tne seine sonnel need obe rer ecu When step ae meet orton of be ‘Ft dmc cole ene on beet oe lapel a nie ca ow Ue pen Driver + Call problem space stp suo and he eect defen {Time steps over an index 8 Mie oom oe els et eames an weer am + Acc of selected ime steps calls data saver ofa zene subroutine 16 Sore data sampes + Attne completion ofl the time ups calls spp subroutines o ‘write oapr da Problem space setup Fe te the pntem pe re ‘Set the mbar fells in each dimension cel 898 ‘sing the cll dimersion 8x 9. + Gaeutes corsa matptes 26 The Fine Diference Tne Domain Method fr Eleciromagnesics, ‘Test object definition ‘Cals indvdoa Geld component nthe sll ame “app” ith tu itger variable IDONEIQK) fr B, i our example FORTRAN) indang te cmponton is det’ ow We EH also ier more complcand mati i i uely conveict wo 41 the defalt mera o free space We aray of “lags” can be read for a pepocessing check of the geomeny ant compaction of the obec. EM el artis "Cactthe respons f component fom itt ow pie tne value and tht ofthe peat igh eld guanifies (Es around He and He ‘round Es) according the ype of mail present a that component espace Lossy dielectric Pret condor ‘Owte radiation boundary condition "Abort eat perlite sated fel tthe trout prion of the problem space Data saver ce Srv scopnae Ula las ad td componces cen of the ‘amis inthe FDTD computation space in arays at hose tine eps Far nae transformation re laste anol i a agai cents on 2 lool sce Sorrounding the cbc and compas the carespanding Seated ot ‘ait fields in the far zoe ‘Te aciecture ofthe cae is quite staightforward given the modular rate of ech subroutine. FORTRAN COMMON blocks can be used ops faa berween stron, wilh he win Drive pgsam cesta ee [xsi flow chat ofthe cde sogears in Figure 2-2. A computer eae ising inode in Append B Scatered Field POTD Formulation ” svn noes REFERENCES 1 You ACS of nso of nil dy a ps ning Mel es pe ain EEE rho rp 1 ole i sects FEEE Trans EMC. 220), 2, 1988, Chapter 3 FDTD BASICS 3.1 INTRODUCTION In this chapter, the practical considerations of FDTD calculations are treated: . Cell size . Time step size . Incident field specification + Scattering object construction . Direct calculation of total fields . Outer radiation boundary condition . Resource requirements The choice of cell size is critical in applying FDTD. It must be small enough to permit accurate results at the highest frequency of interest, and yet be large enough to keep resource requirements manageable. Cell size is directly af- fected by the materials present. The greater the permittivity or conductivity, the shorter the wavelength at a given frequency and the smaller the cell size required. Once the cell size is selected, the maximum time step is determined by the Courant stability condition. Smaller time steps are permissible, but do not generally result in computational accuracy improvements except in special cases. A larger time step results in instability. When using the scattered field FDTD formulation the incident field must be analytically specified. An infinite variety of waveforms are possible, but experience has led to the Gaussian pulse as the incident waveform of choice. The exception to this is when frequency-dependent materials are included, in which case a smoothed cosine pulse has advantages, as de- scribed in Chapter 8. FDTD is capable of providing very large dynamic ranges, which can be in excess of 120 dB. Accuracies of 0.1 dB or less can also be obtained depending on the cell size, frequency, and shape of the objects being consid- ered. The scattering object can be “constructed” for the FDTD calculation using integer arrays for each field component. Different integer values indicate a different material and determine which FDTD field equations are used with what multiplying constants (depending on material) to update the field compo- nent, In the scheme described later in this chapter, free space FDTD equations correspond to a value of zero stored in the integer array, while perfectly conducting field equations are used for E field components corresponding to a value of 1. Numbers >I may be used for lossy dielectric and/or magnetic The Finite ifereae Time Domain Method for Flctromaenets mses Any shape nd meri can einclued within he constrains 0 Cie cowea and manu rqueny o nets Fr toblems which do ot ve an cet ed, antenna raiation cle ladoe, Tor example, an FDTD computer code based te steed field Section 36 “Whether total or catered fics are computed, Murs rst or cond onder snoring Dounce yore a elatvelyrefletion fee and ens ime ‘mented eminaio forthe POTD space, Other absorbing boundary con {Ton ar avallabe wich provid beter swoon with fewer cells eed tewwasn the wbject andthe over boundary Bat at the expense of aed “ompietity. These ate discussed in Chaps 14 and 18, alhlough for most FDTD uses, the Mur absorbing boundaries ar adequate and relatively ‘mp "Ate he wer determines the cell sie, 2 problem space are enough cencompss the catering Ob, plus space between the objet and the abso fg bouniary denned Al sunt of ine ep sufled ge ties fll characterization of the iteration of object and fel, ost Imporny ay resonant bebo, restated. From te numberof Ye ls ‘ely eximated. These eure include CPU tie (terined in partby te {eed ofthe compuer being sed) and the amount of solid state memory (AM) and extended strate copay (megane of ha die meer) ete fo the slaon and for soi he es ‘Upon completing hs chaper the reader shouldbe blew make FDTD cx inn al nian a iment 32 DETERMINING THE CELL SIZE ‘he tndamenn coms a ht he el ic tb me esha he J envleet for which accra sus are desired. The obviows ues on ih Vow much lest, and to hs mst beaded the question, “How orate o you wam he ein wb? Afi gused onc "10 ol fer wavcleuth meaninethat th side ofeach cll should be /10 0 lesa {be highs: equeny (shortest wavelength) of interest For some sitatons, Sen ava very acurue decrrain of taacseatering roses, 20 ‘smaller cells may be essary On the ote and reasonable results have ecw ound wth te 8 four ells per wavelength Ifthe ell siz is ade och slr a hs the Ny aig Lin, 2. 2, is aprosshe 00 ‘nel for tnenahe eas te obtained and seit lising is posbe for signal component above the Nyaus limit a so that if sme prin of the computational spare Is ile with pencrabie FDTD Basics, a ate emus se twee in hema o dete ema ‘Sts in tbe mov ate mck elroy fs sun sd prc ‘Socio wr beng coed if santo aries agnor {Pelee mb etc Poste meso eal wit ‘enna cf ea nthe de marae ‘riya htm”, Grape 8 Tenet wy sms em nae an oe ae ‘Srp of te ftdarbton From he Nygun sapling ten. he tt et spe ers per omg a non ee weigh pci decal tin mo ‘Sr pecs eee fae te oon Siro requencies wil pops at igh deren speds tough the fT acento Snjoton coer mat be hd wan scape vel wihcan bay ‘Scie yc te el oe eas oe smrsen th inportcacersso he rmbiengeoncry motte scarey mals Nil hs wil be me [Etat y aig tcl stalls hn 0 ds some Spe pommy fe omer unis a fo ear Be ‘Zones: Ansan ston we ane awh ce ‘ewe tckns fom 10 01203 Cel een Salo) wi ate ‘eos inet, Anthro flaw et eng fom sot ein wh eee of mesing «moo ‘Src wh etna cp yee sige crs Good wl ‘od sr aos may re ety eso aerate ino ach a vl modcg a sel gr wich pone ‘Sul peery brn Yow ofteomesie These onion ae ‘Onc tas en tein. he mer of eo model nega mate aout of espe wee Spt nd ‘erty nto on te yc crncl Asin spoil mani fc fom afer ded tan tore monn eo Spin fpr gn (RO) sd ek ‘gcrempuer, We ow crit es icmp of FD aan ‘talon cnr hcp ei ete eed sma gcc mom ND i obs 32 The Pte ifeonce Time Domain Method for Elecromasctics, 33 TIME STEP SIZE FOR STABILITY ‘once te el sie is determin he maxima si of th tine ep AE imetntls allows fon the Courant coon. To understand the basis fr the Courant contin, considera plane wave propagating ough an EDTD {mone me tp ay pi ia wave st ot pas tog mre fe rl rca dingo timestep POTD can propagt the wave on fom (nccel its nearest eghtors To deeine his tine ep contain we i ‘han wane deco that pane mare propagates mont apy bete| ‘eit pit oction, This direction wil be perpendicular to the ati planes ‘tthe FDTD pi Fora gdof dimension d (where d= 1.2, or, whale fide eu, Wei that with he mania els of propagation in mtn in th poem all the sbeed flight ee spc, a vats & oD for subiliy More gener fr 23. rectangular it var GES aF a Fapeience has inated that fr acta computations thet vale given byte ey i Gor) il poe acre esa ox fat Tac when the equalty hols, he dbcretiod wave most lotly lomxinaer be actual wave propgaion, and grid dispersion eors ae Hours, exceptions 1 his oc, Or situation in wich he ie ep ant be ruc fave (32) 8 ne he condi of he ater Thue gear tan eo ot oldsg sins (> ibe cles ret i ale han Core init Ts sel opal, {atc at acl th ime ep se et byte seed Hen ec pce. Ast vly in conducting material Wil be elle fre cac, he teste in an FDTD cleat tatincudes bath re space 155 Saductng rater wil be sch ht the Cour it wl ested ‘ener weve test neko nie ight conducting mt “ie may mie mich smaller FDTD cells than in suroundng fee space saree ies te ese ei he Cot ‘nts dacasion of ab Pd 113) and conduc Finns formating materials, Caper iy cimcerns for both onlnear Setons (Seston 113) mails FOTO basis 8 3A SPECIFYING THE INCIDENT FIELD ‘key advantage of wing the sotored fl formulation thatthe inde Sel is specifi anally. this section te practical considerations for doing this are pven fran incident plane wave. Sch a plane wave may be eoded fora senting caleslaton, for emp Oke ype fide lds fy peti fr tr poles aps, at te os shal The specie incident Fld wil be + Gaatian pb plane wave, a this provides a smo rol oi freuen content and is Simpler implement Proviso for abiruy incidence snl sao made int, conider specifying » general icident plane wv i the ine 80 ‘We assume spherical coordinate system with he origin coitedent ‘ith be oxgin ofthe FDTD Carsian stem. Fllowing normal wage 0 ‘messed fom the 2 ais and from the = nl Following te wl seater ing convention, we speity the diection tat the incident plane wave is sing om Oa, ing veo po om ei ne E-[, +E SHC Aes Re) 03) Safer jeer) ey wre Gand, a he peril corte ym ei etary i te impedance pc ithe sed of a tev fom te (State pt de FDTD cyto sp wich we ee ‘une he cid fel Tenn ft ma bain ie Sie wave or pu, for stance Ra tray eter sens For Sie ca pe mt poppet the FDTD pce ae th deny apes he catering eta he ae a i chen Soin “ecient te he pies oe Canes compe Incident fields as =e Ea Fy esd cos — Find Home oan sm Ey nh +B, cd cost cay copy cn yn CE na MM The Fite Diforoce Time Domain Method for Flctromsaneis ext conse specifying a particular fd component for a Gaussian pu Acmuniog tht he spl has been fund from tbe above expen let the fanetion ft be & Gaussian ule and express the x component ofthe sleeic fel ofthe plane wave as eff) wo uae rents He + RE 66 na mc ops eevee ey Ge or ie £2. wi time ep nt For ur plane wave incident rom the 89 direction we have 1) 05} cos sin + =1ay sing snd on 1K = 1zcoed Noe the 1 el oft in (3.7) corespoding othe ston inthe (3) Yee salle, componct Coresponding offs mathe inloded foreach eld conponen, Wernow define and and contin 50 1 provide a suitable pus ‘fant ofthe tin ep 20 we have coven. Altiovgh an iea! Gaasian ‘ls extends infinitely ime ours must be truncate ou ellos, and Ibceffects of this must be consider. We must aso spect th time duration they that than asta bandit ‘Tostan drei he drain of te Gabsian pulse, We select = 32, where Bi he umber of tne tps inte Gaussian ple fom he peak value Wo le Wamatin valve The plc wil exist fom = ~ D unl = = 2856 terovimaed aro ouside his rane, wih peak value a = BL "Te value at tuneton a ¢ = 0, Be drmind by ea as sce fom (9.5 de Onsen leo ocation wl have eae exp ABS) down from the maximum vale, We now need to determine cs tht his rucaton foes ot inode oan high equncies ino oo spctum, and yet does tan weston neon dtrnningveloe he cen altho cssetlly 2 “Tere sno Crt answer, bt aprotic solution so et change with tat wat te ania te ple i alveyseeducedy tc ame ‘ale We et = (4h). Thus, at tncation the pals is down by exp(-1), ‘ramos a0 8, Because a reasonable pol for esinge precision calelaton DO Rosier s oo i Son 025 050 | 076 100 185.80 is 120 (i significant esl dg) thi vac of eeu i action frequencies wel below dis evel. As a maximum «32-bit word length allows aos ite over 190 dB dynamic age ‘To late tert Oined with one vale we conse sination ‘fen used for ober examples in his book, 23 cubic cell with -om es. ‘Applying the Courant stblty condition we obin At of L924E- + OF (201924 ne, Using hein tp alongwith the paramere given ove yee the Gausin pls of Figure 1. The approximation othe puke is most iceable ea the peak value where th cre ofthe ata pulse spect ated wit sigh En sgt btwn th ice vali ‘The fast Fourier transformation (FFT) ofthis pulse, showa la Fie 32, ake it cleat the euncation of he le dl no rrodoe uname high {fequncin Attn cll per wavclengi ne woul pet ge scare 3 {fom our FDTD cautions for frequencies up 3 GH, ad from Fue 3 its clear that or Gaussian pus i proving relatively high signal levels ob "otis frequen. On the ote handy we may te consid alate ae instability if we have appreciable ener ia he ineident wave for wavelenes in which oa ce sz es than fur cells er wavelength. This caresponds to. frequency of 7.5 Cll al mt Fg 3-2 we ae that ot aps dey apron 130 Baths eaency—~ al enue movie sbi 36 The Finite Diferene Tine Domain Matha for Rlecromaznetics, i a4 8 6 7 8 9 10 Trequeney (GH) FIGURE 3. Foi om fe Uma sigs 1 mn nn i rm tl nama ma seriall rete Teese Seimenretrmimataaman watts oe seen eeeopeses eet loan eee eRe oes eras Soeiiecarimahas mica rs ccna wronirsegcatersne stearate 5E rus, ies prvi pcan wth nae we mast Sian Hares ase ts Sete eatnin teeta remedy Peete Sob heb ce Set wee tow Skene aetaramernencae See RN een oars Svan Teme amae mise FDTD Basics a 35 BUILDING AN OBJECT IN YEE CELLS In Chater? an appmach fring integer ary spect the materiale Jn FDTD cells was presented. With hs prose, he DONE IDTHRE aay ae meant for speetying PAU and lene Matis. IDONE is ase 10 pec which marae intranet the fel component TWO forthe E, componente The DFOR IDSIX aay for magnetic matrls an conespona oe x,y, and 2 component of magnet el, respecte ‘Thea cospndences stand blow DONE Ey DrouR iormee inst W ‘The LIK subscript index denies in which Yee cel the materials located ‘The conten ofthe aay Wentfes whic water na in sais Fon example conten of © might specify fe space; 1 pefet conductor 2,2 fy keri witha pce prin an cot 3, oy ‘material ype tht exis inthe FDTD soace are set once andthe corespond- in maining tm he FDTD ape eur wate and =< g00Ne Seana Intercts with each eld componente specie separaey, modeling iso tropic materials with aga permit permeability sors is sraisht- forward in FDTD. . Ths Spare trys are wed he ver a conta inaependety the xa pcameetef tec sd mapotc material he Yon The placement [tie fld component in Yeu cll shown in Figure 33. Fr example, Seung an eemeat of ie IDUNE ary at some 14K canon 1s acy Tevtng dla mtr wor nace with he x component feet Feld located at posdon + [21K inthe FDTD space. Sting an element ofthe IDFOR aay some 1K Toetons actually loca magnetic mars ‘ators it ineracts than component of gre fl etd al positon 05 K+0.5. The spl difference between the IDONE and IDFOR array Tocaions ia det result ofthe field oes the Yee cl. This erent ‘Set nt eld loans ees “arsine on ncn” penis ‘hen buidling objets Smooth surfaces mut be approximated by stepped Yee { Tocatons, Eves an objet tha fis the Cartan coon system (or ‘Mtrpc nce) cont be peed erly ft cana of aera at fms both dilecic and magnetic properties because the loci and magnetic Fel! fctions ae fet spay. One remedy Tor Ws station 1 10 we rage als fhe peritivityo peresity inthe fy” eins whe ths offer ina factors Buti emedy only reoces rate han removes the sft. Mla “uiling” a objet tc acezay to ke in mid that oe en filing Ye cls with materia, but ate locating material atthe eld ations twin ae cell To lle tse us conser boing at 3% 3 cel ‘Seto pein he ny pla lrsed ms =KArin aD Yee cell space Tis FOR) [FIGURES Aten il mgr mil plat hme FORTRAN li ene pe Plate wit e compose oF mater type“ n whic th conse par ‘se fhe material ae et laches in the FIT copes ce Using the IDONE and IDTWO ays, his plas can be uit with he FORTRAN shown snogure 5-4. nes exampl, notte to sttemens that mst eluded tat the plate png formed Moe sting th Haye fo i te (83) wl ao sult in plate wih soon sides. To ulusrae as rumee conser “balding” the corespnding magnetic ‘mata Ifthe same FORTRAN logs ic unt toy Ho poe 4 ‘magnetic pate, the object generated would sotal be unconnected, sis tated in Figre 33. The core wat Bul the magne pl ithe folowing FORTRAN ent bo 10147 patna IRILNE4DRORILK)=11 HUNEDORVA I 10 CONTINUE, “Ts complication ia duet comes te beatin el ap ‘ens wii the Yee cl, ‘One ca define a correspondence between seting electri and magnetic {eh satin, 3 hatte ne wo Wo peer ee je en be ‘noid wo generate amasretc object. To sc this cose tha we hve set ie emit acl rats corey EXLIN), EOI ‘This determines one comer of» siclectric cae. If we wish to define the 40 The Pinte Difnence Time Domain Mth for Flrromaonrtics, cores of comesponding magnetic ebe offset one a cll in the 2 ‘econ we woud sete ction ote magnetic Hes HRI JK), Met), H71Ket) ax magnetic material using the IDFOR, IDFV, tnd IDSIX ars. This example ndiaes the following corespondence vee “building” leleene apo rage ve Dielectric object Magne object DONE - IDFORUS SK) IoTWoUL.K) DIVA) IDTHRENLIX) : SIKH) ‘niscaeaunseace apee witht cagsin te FORTRAN oe ery ‘Divi be pas nt peceang exam ae en pans new nn it “tow ick ar ts at ae aint sore exams tow pyle aa guecl chin FDUD pac” heelys a niet ane teraction ha of OTD eae Jp elnen'e tes gion il pve ow 1one mst int Bow FDTD ccs acres esse woe eet at emi the situation that we are computing the scattering by one of wna ers pan weve mt am 0 90 as eg ona with Se dpe ine requ eds ay eee ee ean, aa al TO ot TDS aay Tetons Se oc spe ne wd nots eso any oer vais wen SEER! AS hala tele ere ses an we ‘tie tts coe whch plane wave isnt fm he 8 sa ends aly eet ep Baca i iin eee cere we can openness sine ‘etc. Tis cued in Chae 1 nich sows Bat comet oaPe Sint ene FDTD pas comer 0 th Me ae acen tee FATT rc” for ame enna Eb) mama, For rancor “bing” a erty om SOUR Gee cscuan eDTDcele Afntwewontasse ude Se ees whe oar Yor ron whee Eel see scone Then we so may cone it th FDTD "eke SSSR sperern tse or fl sn nd te tae ae eee es pce Enprece incising kate el OR etm ety andcing veri resin at esc STREERD GS tact nat one four te Yee cl mens eae cai benicar te itn orn pert onda? cent raion orth elextenion tha whens plae espa nen oma pane pein concur trangia TSI Aa glow he ong magnet els re mani or pasies a“ 1a FDTD eaeulations the zero the H fed locations jus outside the conducting region. Thus, the FDTD surface” is apronimated a being halfway between these cece and the outer electric fied locations set to perfect conductor. ‘So josthow big is an FDTD object? tis generally a small action of acell Jeger te men emple em ep nd et FDTD oe ‘spproimate method, nd our physical reality ofthe plates i z physical reat ofthe plates only approximated Resuming wth general tpi of “biling" ojos a goo approach i to seneate squares (in 2-D or for hin objects in 3-D) orcubes of materia, where Seting a cae of material's notthe same as seting one Yee cll. Seting x cube Show inthe following FORTRAN, which ils ace of space ote Yoo IDK wih let material pe"MTYPES IDONE(LK=MTYPE, IDONEULK+DeMTYPE DONE 1)=RETYPE IDONE(3+1 K}-MTYPE IDTWOA.K)=MTYPE. ImTWoW igi0-MeTYrE IDTWO(t+J.K+1}=MTYPE IDTWOUL}+)-MTYPE IDTHRE(1 JK) MTYPE IDTHRE( Jo K)>MTYPE IDTHREAAI 1 oMTYPE ‘Ths approch to modting asoiabje sphere for example, io etemine it spat Yee el eae bin hope and tal 2 eld locaton othe marl of hese shown above Sein inal ld Tectons wills na Sphere at dos not have aoe face. This may ‘mete tpt x portent rss ote mtd erotic icing ene nia ht cmpoens a st at pa compte ce ‘Arte ot ep in an that rege ae «pla 1D ary does ot cane problems, bu rather maybe + we chine ‘Wtng ot t D aay analogs pining om ana tlh st sti ic walt ae tr Dy "ale wedi the clelaons Jost ais my ais tackaoand hy the pata cover is with rs, and nu paint ake one part fe foam neato St wean le FDTD atv ie ce Tisseton ofr with sone her aera ten ponents eton 42 —_The Fine Diference Time Doman Method fr Elecromasnetics vi some erent matenal by overwnting the HD aay. Ti appoHc | “nea athe advange Pailin he pony nig ed ‘hin the material region. Insite no it "Bg em ee Simpl objes can be ul by weting FORTRAN coe specifically, inthe Piste example stove. For mre complcate structures, Pasion powTans fray be aecvsry. These wil tae geometrical dts aleady ine aber fom a computer sided design (CAD) sofware package for example, and tran no Yes cel oestons, However, oe mus alvays keep iu te locaton to 2 material type is equivalent to specifying the material which interacts with specific eleceomugnetc eld component ata speci ocaten, 36 DIRECT COMPUTATION OF TOTAL FIELDS a tis task we sts ning the seated fed formation ia FT ‘oblems, at approach tat has many advantages. Some have already been ‘Seoussed in Chaper | and nd ter wile ilsaed in lar chapers Involving appliatons of FDTD. Matemacally the seater an ttl eld ‘uns regal! and yet nthe mito nites cel erate ‘Servs, However, dette approximations inbeeat inthe fie fer ‘Seino, aed a tal tne a il al sittin xin in which dir cleulation fhe tal elds may be simple o elmore secure esl ‘One of tne ives camping sll aca of Wl Gel sah a siding probleme where a plane wave acide in Which one ia concened ‘wthcalclting relatively ow lvel tales which may pene ashe’ repion of pace. This ca Un Ue iideat and seated els arent idemical in maine but opposite i sign. Both oa field and seatered fed Formulation re lined i dynamic range by the precision ofthe variables ‘ed ne compe esclainy, Howerer, elie ie fa, the incident fed popups exactly ouPh the FDTD space. while the ‘catered eld proaration i sujet gid dispersion ard ote ers. This ‘igh mor (mre precinct Ge oy pve the nident and seated ils rom canceling accuaey othe cater eld Formulation inplemeatton. The emo vanishes as frequency goes tard 20 fn rows 10 specie sie oly we pga equem i apromte. nthe tl Held formulation th ol Fl propagates trough the FOTD ‘mesh and ths gd spersion eos exist fr Boh the incident and seated ek. The hopes Ut tea hey we subject ote sae eo hey sy tendo cancel one another beter. However, some contin othe eters fel wil wal in direions trough the FDTD mesh fren frm the [operon diet of deme ds a seine wll tbe ag {he Same rid dispersion err. Ths, oe carat coum onl compensation ‘tpi diseion ern the ttl fel mode of calculation, | POTD Basics s ‘te stuatos in wich tl eld computiton maybe desirable oc de Aculin in empath nid al. Conse problem of opp ing FDTD toantema ration problems If we are dealing witha irae seawater ne gap ofthe ancl demi te teloas and tee consuming. AS shown in Capes 10 and 14 eliely ‘mpc o spect source nhs ep and dry corpuae he ol fds hat ‘el fr thi ore erring with he atena gomey. ‘When we wish compute in he fal eld made computer codes based oo seater eld onnultion respec ol id fomalaton "Geni can of the sated eld fonmatin, We marly epeciy a sre ‘tmplitde incident eld and then inser int he FORTRAN cde the necessary sour terms. or example, consider he cas in which we ste modeling + mitt FDTD grid We st the incident fi ample to zero, and aft ach update of te ed values, we execute a ie of FORTRAN specifying the source ‘elds his oetion. The process debe Chapter 1, an camps ae sven in Chapter 1, I summary. for most sutons, and especialy for satering problems, ‘ivr empuing he tse ld wig the sated eld Forlton Preferable, However ia some stations ect compattion of the tol fd maybe more accra simpler, and nthe cnc he scateted field FDTD ‘de emule ol Fi ode merely by sting th inci Feld pad 3.7 RADIATION BOUNDARY CONDITION ‘An ter ration boundary concn (ORBC) may no sways be neces cary when epplying FDTD. If he FDTD potcn spe is tome ty © condition tat an be implement rc ino eft difence eastions ff Chale 2 an ORBC i ot essary. or example fe ae ely dlecronapics yhenmcta aie «uae were: vem de ge kt fe athe walls iso, and we enlemcat hs by eine he TD ray soompi Tot atypia, we oe amin 1 mel aCe HE 6 situa in fe pace, sch ar scares contd ve ull Iie sated or aie els o poate boundless st ing eration oon Ufatatty tne PDT compro spe sy neces bounded, and when te sctered o radiated ids ave at the Boundary they wl be eed back nt he computa space unless we te pevealvc mane The nea emus vahe ppetren oa {o abso the scared or rie le when they ae a Hf he FDTD see that orate i ends ee at st tine arn Before eles om he ote baryon ma or the daa but for mst problems ths ttt vibes ae age 44 The Fine Dieence Tone Domain Method for Electromagnetics umber of cls ee6 1 mode 2 extent OF ee space suouning she ate or arn od eget edie wih cresting Smet onpute resource by action ofan ORBC. Tout nro a ORD ng aan FO plated wing the nal FTD equation of Cater 2 because sae the ‘ares naghbor ned components ees evan eee Ca ‘elring tare te he role ae an vale Th ml Ras ‘oki. wsina emg cmp jt be pen Reet openings othe Spore, inating te els he tard vein plane wave on the toundary by ooking the els os ihn he Sound wl at sa th wave pve Tie may flee shane for cong ha we tess of hse ORBC condos, + poplar and esl applied ORBC i preset, Theis corey called he Mr: toring Douay more rio framed oder Mar pending on er a he pen Iron asdf ema th ion th Boundary A fit oder conton Tens tuck oe sp time and noe pce one el acto: scant (mtn condos oc tak two stops ine sed nr cl atone Consider tht we ae ate x= ini of or FDTD computations space. Wwe decide ht ou ths plane wil oe Ey ant, fl compote ‘Using these eld componcas we an evalu the He aiferece crt terion needed up the H magoe eld compen t=. ‘Teor lle gor ld components wl valor pt ing he field componcts loads = 2 and Beyond a eat 1 he ‘musimum 1 dmetson clea th problem spaces which oeton We that agua spy an ORBC) However ene cannot vest he Band Feld omponeas a 20 wih he tal FDTD equation because he magnetic fede a= an eo ava ‘Ne uy plan dome her ih he Mar xproions et coer the E component late abt =0, =f and = (v2) The it okt Mur Gsmue fl compote is erin e1/2) = BNA K+I/2) + eS A pp 1/2)-BY0.581122)) ea) “The second onder ema for, a th boundary x = Os For Bases “ p(k 91/2) SMA ej 4/2) 48 OLR SOAK proj /2) 8p" (0.5 2172) +E (esloan 12) 630i 172) astea 2a) car + a8) (BO.i+ 1k 41/2) 2630.4 102) FEO MAST O RHO) earner ver tater + Sete ero wat (elon +312) (05 s1/9) Reo ju 1/9) FeLi) 7) ADP ik 17) FEE where we have extnded the expresion given in Refecace $9 jacude oncbial ell, ‘Considering the fist order Mur approximation, we see hat th cunt vale of, at Os euiate frm he previous nd ort vals as = Sead the Same yan postns. The second order estate uses previous valet from th reeling time eps, ad values a te aceny and yoo. ‘The equation needed 0 denn ster Fld components ot ther Lining surfaces ofthe FDTD space are eal deternine by moifiaton of (38) and 3.9), "These expression ar rsighforvaid io apply, but hae sane rival considers Fist, we mus be cucu when Stemining the Limits of out DTD space in tems of he ranges othe DO lop indices that determine shiek eid composts in wtih Yo sal ate he jbo spe a ‘These indices wil not be st property if me merely at all of them torn went sae limits Instead we mst consider the Yee cell geometry, and coh ofthe sn srs that limiter FDTD apace cacy termine the vals of the ary subcrits (Yee cell corns) of the elect Feld ‘ogonents onthe sce an set the DO lop ints ccd Ave conse lean sex er ORD reuse ‘als fom adjacent Yee ells it canoe used for ernie electri ld ‘ales hat ae adjcen! tothe intersection o wo af the ening plies, Even f scan rd Ma ing pid tt be ed components load aljacet othe edges of problem space. 46 TheFinite Difeence Time Domain Method for Electromagnetics (ne oer consceraon 4 thao determing te distance between the “tpt nd oso nay The ae om th ete ter say isToated te beer se abscrpton ofthe outward vein wats, This due Tomes waves becoming mae lke plane waves sey Wael ate io te ‘tween the objec andthe outer boundary is imted by computer memory. A ‘Soman cients mum of fn els betwee tbe ebject an! oH Feary Forse nore anil ogi epcialy high ‘cary i ended. Soe eagles ate sown ia Section 133 ‘Moving he ater bounty io close othe jet may cause stabs in ‘mee ofa problem for ate and ther callin, which ae excited bya Source within the space ahr than bya plane wae, 5 he ute ounday mest Imay be absorbed Hf he outer boundary i 0 close. Far example, spe aterng cludes 4 erespng” ave Ea propgates sound the spre and {ste egy The radiation fom hic crocpng wee can cxiy b oon he ‘rset backseat results (se Chapter 13), I the abserbing oer bound {Stoo let the sphere hs wave wll be tured and we senered eld ‘le wil be amet 38 RESOURCE REQUIREMENTS When oid ain of FDTD aac oem oe of he fitting hat mat cmt gen he comput ous Soo ae TOTD tnd wap of org ssn he bbl, Whe wee primal origin ine San ch sions Eine cone by fin deveinng hw lage te pblom geet Sor acoure! nun ute shnes aneeg frwach we te ‘eal Teme te bet se inwanelngb erie he abe ‘toes agin fr ming god sry th ne of chal Shed tom enh ofr neha ee kw at ge ery mcs) termine he mount of cmp ree eos “The mmbe of elo ove sn ads of he mambo ie eps ‘Sch ens “is con proves the beans fr exiting te cm rete ut fr pes pt bse et od ee Sr val fers te cel i on ey oe Seri. Fmd he lm py he we umber cel ne poem Shere deme Nha Doe ee Wes are ‘icsulnnganm anedin be ONTEGER* IDI aye witb Skeri némapree meal conor Then oem he Smet Stein ty te nn sp one FORTRAN ihe ae cane lon Por bases Storage = N x( ere we have elected the relatively small umber of sulin variables feeded fo sore tnpoary values inex DO loop, save real for Iter Processing rsp, a slr Fnctions and we af also egeting the emery nie wo tre the exesabeistrctons, This overead snesy rm lls inreassi wil come a sulle Faction ofthe total memory renued oan point operations reaued sing Operations = N > 6 components /cell 10 operations /componen xT nee Tis he ut sumer of te tps. Te ata numberof eueuaons {foreach component depends on the materi ype andeistence of te incident field ta particular ime tp Tere are alsa lapel tatemets which mus be etecued to determine hat iype of matali Space, peer coneuctr, ‘on asd pair tion ‘The number of tine sep Typically on he ere anesthe number faker 3 tie spe raver a sngle cell whe he time sep ee he (Coarat subi ovo. som estinate for Tis ‘Te10% JS! = number cells ona side ofthe pote space Combining he bove we find hat he oi umber of loting pint operations ‘sapprosintely sven by Operations = 10. Nx So 8 Ine Fine Divems Tine Doms Method for Bleromaeneties From his we se that the total ntber af fing pin operations ei ‘ropa othe number cls ne FDTD Hace aed wo the 43 pomer "Rens ror coer how te mer of fein pose operations sles wih fegurney. ‘The sie of the FDTD cll mast be scaled froporonal fo reheat atsin cack ie encom wl nae proportion regency. “Tis mcan the umber of ells inte (-D) problem space wil be proporican ID insqcney tothe td power, a he suet fang pi ations see nepoctot#0 eens ried tn the fourth power. This fourth ‘erscig with reuensy cmpares favorably © ter methods such she FP of moments Inston se mat 28 yeoble ses Umea Mecngt the FDTD meted sil ot equ fewer opetatos than Tovcaches nich requ the solo of a mati, especialy if ests over a tad of freuen are ee. ese (100 ext problems epace For hit amie of els anro= inuicy 30 Mocs of memory would be egies, withthe act amount in omewhat reste et storage of ter Vanables abd suction. Fos ‘tte uno machines aging rom coperemputrs 1052-5 PCS ‘Ac beable memory eed the maxima sunber of cel tht ean Be ‘eamodated ib ceresponding doeasd. For exampe, wit 10 Mints of Stun the bie space sx woul beets fram he above relation ‘Mana In sta experince U6 MBytes wll accrodse approxi {paely (72 cll indicating emery overheat fr mses and sila Tati ty pte about 308 ofthe meron deo state ant componens, Azan fr larger problem spaces with more cel this tvehead percentage Would be reed Rea Pe cemisr te compute tine nested for rpc problem. These etme ae only aprotimae, 8 the toa umber of te steps equi ‘Gpends on he goer being considered. Morte steps wl be neces” ‘errezonam gence fewer for ony geome with Rely camped aru (65 cell? problem with 1024 ie steps oe xu, prox aay 109» Hing ya perton ae egies & My “rms equi o tr the el components an TDT7? ays. Speeds “eae machines ane fr 1000 or more MFLORS (Naion owing ‘it Operatns For Sven for a supercomputer throng 149 50 MELOPS ‘erm for work ators oapprosinaely 2 MFLOPS for a 32-bit PC. The ‘an (CPU tines ar then estimate rm he above discussion as 175,28 min (tues IOMPLOPS wosh ston) and 4 ming respecsvly nace. with Src ma civic sere his problem space and far Zone fel {aconed,rnning tines on 10 MFLOPS work aon ana 3-480: (God 8 nad 310 nin, Obtaing rss For pobre of it ie i ewe al these mhines “The abn eens ofthe number of required operons arue tic OTD cabelas an nope the sight ove! of appv abun om Ries, » ent ing elie eyo co Honeve tbe FDTD coe tig conic regu sigaian mabe of ‘inl compton, os a nen alr es comlins are sumer fell ic with pesive marl (Chaper ‘Seem ny esti nou eee hp onthe esc apo od POTD cme ce Te ae estore pices wil pone» bas einaing rows ‘Se ra panei FOND cn REFERENCES ‘Stig es tie een nw enn BEET Me 2, Arn bin ae Ra. Ke Mie La a Peneton seg nm Gio RO PART 2: BASIC APPLICATIONS Chapter 4 COUPLING EFFECTS 4.1 INTRODUCTION While FDTD was introduced to the electromagnetics community by Yee in 1966, it was nearly 1 decade before it was used to any great extent for actual applications. There were several reasons for this, including the lack of the necessary computer hardware capabilities. What really got FDTD started was an application for which it was uniquely well suited: determining the electro- magnetic energy which would interact with and penetrate into an “almost” closed conducting object, such as an aircraft, due to an incident pulse of electromagnetic energy. This pulse may be due to natural phenomenon such as lightning, or it may be due to a nuclear detonation. Because the problem dealt with transient fields interacting with (to a good approximation) perfect conduc- tors at relatively low frequencies (but not too low), it was ideally suited to FDTD. This problem was divided into two parts. The first dealt with determining the exterior surface charges and currents which would be excited on an object (again, perhaps an aircraft) due to the incident pulse. The second part dealt with how electromagnetic pulsed energy might penetrate into a region that was partially enclosed by conducting materials designed to shield it from the incident energy. This problem was often idealized to a representative geom- etry, since modeling actual electronic equipment, cable bundles, etc. was too challenging for the computer power then available. In Chapter 2 we developed the governing finite difference equations for lossy material media. A simplified and specialized case was the treatment of a perfect conductor. In this chapter we show how this capability can be used to treat exterior response and interior shielding with pulse excitation. In addi- tion some more recent results for shielding effectiveness at frequencies below the cutoff frequency of the shield aperture are presented. These illustrate real- world applications of FDTD that often require only modest resources, but typically show relatively complex behavior. Under exterior response we will consider FDTD calculations of aircraft exterior charge and current response to an incident electromagnetic pulse (EMP). This application was an early impetus to FDTD development. A nonoperational F-11] aircraft was EMP tested in one of the EMP simulators in the U.S., the horizontally polarized dipole (HPD) facility, to obtain system responses that could in principle be extrapolated to true threat levels. We show here how well FDTD was able to predict the measured simulated EMP re- ‘Sponses for this aircraft test. In the second broad area of this chapter, interior coupling, we are con- cerned with shielding of an interior component, in particular a wire, from an incident EMP. We divide this topic into two subsections based on whether 1 The Fine Biference Time Domi Method or Elcromagnati the frequencies of interest ae above or Below the cat frequency ofthe Sed ape FDTD position of shelting involves extremely rvomsmt tehavior that sees lage expendes of computer tne for accra ‘Characterization. Ths offers ye another challenge for the FDTD method ess we provide some backoundon he phenomenon of EMPin general and FDTD atempts to pret and undesund i, 42 ELECTROMAGNETIC PULSE “The EMP sca wi hiatal Dt an inde args cures carson aia hat int coupe te nr ‘earth hy can cae lars oemporry ages in mission Shc) susjsens: nee sun ten ents er Sef in may nuns mt he hardened aa een ese 10 Be tclar weap eft know a8 EMP ‘Once cle Yh nd ier ed HEM grid EE, acest Cope len race ace toon te xe ‘ache th per amore Teer the het reset fa [Rpscnting wits nec, ate cee sete ard he sey {Seva f the ur ction an i est of, an spa abou te as mgt fi, EMP ithe sl of aleve raion fom tee te 'cncet thane lr Fld ego he re of 304i ae posible The ple maybe appointed by a ele expnetial fhe fo Rice ee) amas tne One er 0 man wis fe met flr 1 je Tue tong spec conten apt afew meyahe a Sopra cetnt rand 100 ME ; Thelen cern ang oes east ines wre onthe orf jaa ite ght be aoa 50s. More ‘Sle nrk wih ep est! notes hes cane ho 2 015. ie Lighng ean ey eer wt nal peat ue of 1090 fndaceunaly pine Meh HHOO0 A and per” Deb has Felsen on hich ret more svere, EMP er ihn. Wi Teta charmerang econ: devs tr re meer el ‘movorcaupling pemeny comers snd developing the ol ch 8 FDTD) tomodel i ner couping west hs question cart te detiatly anes, Liphg des ccs own aaa ie Us. inn dos wow FP se avery seus dee Tn iter cave uch ofthe early work om seeing peti problems concent on ever sta espns suc a shin cre ad age. ‘Swe cxarny peed espe hgh rly ha ea clei els cold be messed wt ace mound capacitance senses Sia bree to sua charge, one magn ee sng a itm nia aan creme Soul eed er cures aaping Fee 5s ‘te decomposed into ally and tcunferetallyNowing, onthe itera {ormteroe responses and eacuonc device fale. Awang the sucess of 3 ‘much more ambitious interior coupling program te transer fncton could some debate o be ongoing. What was taken asa starting pit in sear work was tate lnger the exer response Ie moe IKely Cleseonis damage or apes woul occur fede, Conny enh cold he leamed and wus lame in thse early eflons in which FDTD played an important oe 43 EXTERIOR PULSE RESPONSE. steal chare and cen esponses a sulted EMP. The EMP simultc Seas the HPD faity frat athe Kitng Ai Fore Base in ATbugurae, FRM. The hey suc ins apiaton wet the inate Fishing the irra the aeaft model, ad the computatiol resources available fer he efor. These tuts ily altered Toon are preset in all modeling ‘What may not mays be obvious is the underlying physic. For example, the enerorrespnse of an aera resoant but mich mor damped hn he ‘esaant scope fn cae thn le ncn. Av aula Shr me record wil suffice fran exter strat response preicton 3 ‘ompaed tote record neice of «thn poe tena, he aneana ‘Bac nd a avy de an of wine iw vevily pues Uy an Srerture, ever for large pete, te wireline eons 0 mich tore resonant than the bare antenen, Extremely long records in tes of umber oftime ep equ ar then ceded Rather spl sighs provided by te amenntne features and the distinction between exter and inion ‘expose have proved very wseful in atpating he length of he response and thee tie empterresotes eget peit he eps sin FDTD. Many oter ara exeroe sponse sais have teen performed y the hors and oer investigates. Of pacar ies was an F106 aera ‘lied by NASA fo tie sneered of High ele om: alsa I Teauted modeling ighaingstachet and detachment chumels. owes ering ae exing the raft atthe nse and wi. The ucraighing ‘ewan able wo provide enembeexeroresporbe pet, response predictions were ao atempled with intl sues. The problems countered arose ro aly undetessnaing the ine duran an esoures rede acre hari te iro reps 4431 MEASUREMENT FACILITY. ‘The HOP filiy iste in Altec, NM. Tae to ne ‘us EMP simulates icing the verclypolaized pole (VPD) fi. 56 The Fine Diftcnce Tine Doman Method for Electromagnetic, vanced esench EMP sinultor (ARES), and TRESTLE te wort’ lags {Ehuramume epneata ave sea oye abe ro 1S inte consrntion The HDD fary (Rw 41 hal aop a wis Atrayedon ses oop 3 al ais cetr where ere isa Tear MV igang pe dpeng on be model employed. The caret caged ty ptr nn the anemlke ams ofthe lop pres the nominly cull polarized E ls deed of te Smal. “Tels racine byte HPD sebltor ae gute complex. The ec atts opposinel ritalin tetra elow he radon th mainte facy wher ae aa ype pe. Ax ‘lott wt dee fn escent and wed in the FOTD model A fe space magne eld (I= 377) was ascied with the eect ely The round wasimoeled 9 conducting sre hat podced tne varying ‘nkevon cfc = 1» (079) et nc The hei eld mal ‘Ryredced resombly wel he easred Geld dat inthe volume enclosing {Pe nref when laced eer dey etow he usr o along the man xs ke ey “This mov! i ot account fr he eld variations arising fom ple shot voshatvaaton vik could be costal Teper exed 1010 90% ‘Sumestr vated tween fo, wh © nein nomial Vaions F108 or mre ne sect cone! betwen [010 30 le were common. Conpasns etme esiteens nd pedis o aco hese Sooe Above 50M te variation were rp ugh hat he experi an coul not be rele upon 432 FDTD MODEL OF THE AIRCRAFT. "The aca wed inthe measurements Was 2 sonoperationl F111 with much fife component snluang te engine moved is etenor ire war, however, ct Tie FDTD model wid SD. retangua els Couotn eg not y= a a pe ate asf 25: es Th lo swna oe ag Vit alin oft ae ant Te oe cl aie tensed Tay el ne panto tee one ee ees pen pel er Gb noel) Seon eae ote Eee ay cei few ea eae e pe cee Spe ue goles op thor be oe a ene Saati igen ec lar bany Sens rt eck Savina Marve ro aero aes "hea fas we dyn te! len ext fs of esate epg ES Ne ‘pata ou mar en cf be eon oe Rog oupe i ee ta fowas wate ae ee ow dty olan cae canis wee fae ko By etapa ecg Shona neannete eee eocpe ed nmnce occa wg wea fay Scent word ry ns be Te tee FT madcicrain laa eet cere Soeur foranfDTDx timely cy SAUieiaet acetone muta ee "heal ms fr hs oes wee a we at contin wie thee ema wo hy a echelon. A a co 3 oe ae ‘xchat ne ec apse a 8 ‘The Fine Diference Time Domain Method for Elecromagnencs 1.3 were observed i ead mel tbe. Using 4 eq a cell spe se the upper Hint of accure moeing ils 75 ME 3 the hihes Irequency securely modeled 43. COMPARISON OF PREDICTIONS AND MEASUREMENTS [Lets proced to compare te FDTD preictins wih mesuremens. Ast ‘of compares o he dts peice whe Geld ad tal tes ad ‘tual seat measurements are presen hee forte ara Iocated 30 fat along the zai o the fc, wih the faselage center 2 m above the ‘ron plane. ne tet loco were enpye ute geo, Co eal scythe ter were lex demanding When i cones 0 ‘ompirion with te coe, The aiera oselage is rested parlor pepe ‘Sur 1 the MPD oops hat teary mes te lee el is pale we Frelage the wings mreqetvely Vanna sensor lection about the aia ‘Sface Figure 43) were emplojed. The maget loop and epaive 3p Semors. derivative cuypr ofthe charge apd caem respon, sel, Grin and J, where the dervativecurents 1 and Jae ail adieu ferential, respectively, wee left unchanged. The pede responses Were Geena so mar cect companion as posse: “he comparione)» ample af which ate shown here (Fires 441 4-1D. range agreement fom excl opr 1s god excellent when atone {espe voied and canbe por whens weak rexpnse uvae. Tis ‘Shoracteroston of repomes ra on esting he aera fsa 3 thin hatfwaveengh dipole when the El! puri oe fselge, nd the Comping Bie FIGURES T3810} EP Ss FIGURE 44 Ta, Qa, “The Fine Difereme Tine Donan Meta for lecronmgntce 03" anorassee) igus 49. 1150 J, (0) Bowe Coupling Bec a Toe (os) FIGURE 40, 1150, Q (0, oe verse $y Frauen (nt) Dems ast erie ts Habe ings o behave he same when Eis paral othe wings. Noe dt circum ‘he rall nthe lowes rexonance oral wavelength response ae wousisome Based on ecord lenis and dyna ange of he isramenttin the dats are computed only over a fequeney range of 59 0 Mie Even with al hes tains its evident hat thi cy version of FDTD as an elective tol fr presiting exterior scat response w EMP. 44 INTERIOR ELECTROMAGNETIC SHIELDING ‘excluded fom a cavity By the cavity wal, which lo Keep the problem 62 The init Difvonce Tine Domain Method for leromagnctics, ween ae pect yt a ne pening or apt. ein te important aspect of cavities with apertures isthe sheng provided. Ch trzing the sieldng provided bya sell nd pet the ebyectve oF ‘haere he highly veonan tre of he nero response ofan inkerior mental consieratons, A measut ofthe shicking effectiveness of «condi. Ing shel forming sity peed by an aperture nthe response a an menor Inside eel onmlizedo the caret inducedon he wre witout the sel? ‘Aterately the uae of te fequeey domain cure npraed vee sien ane se Boe ach wire eonance the ital canbe approximated by AA@Af whee ‘Rite isthe reomunce peak ample and Af he resonance with, Equa fem) a Farias een wes ae i ne dorsi coment sgoaed ‘Ore van dtingish between two shielding eps, fequecis above peruse cus wheein petre ct he feguoney at which he wave {Grek equals the apr ccumference, and fequecies telow were sf, Above cute antenna ke resonances have Qs somewhat pear {han i cespending wale mie, mile ew euch Qe ae moc "We fst teat the higher frosueney regi at which FDTD works with ‘havsesingsilding effectiveness i employed here. We ten Ue Ie Far tno dic fow feqcncy rpm, where he ebavior ofthe iateror wie Creme) resonant an foes excecdgy lo TDD runs for sera horton The ssn definition of shiing efectvenes isemployed oe os the signatures ofthe resonances (Qs) ar oilers. ‘Shicking eftects nether fepme ne found wo Ue ft fom iitne and mech ofthe dium hn hw FDTD modi has moe our nde Tanding failing. We wil ee ht ver lage esoues mast be expended to ceusely character scang ene 441 FREQUENCIES ABOVE APERTURE CUTOFF Lt us fist consider selling a estes above apa ut. The OTD moleling pected re n this ic and its experimental alton’ teas perfonned by on of he autor (Kane) whi a the Lawrence Livenore Rational Laberatnes (LLNL). generic coupling tx objet all Be Fectiminny Liner eres ehjet (PLUTO) ais ans ine Erling response measured at the Tratsen Range Facility. FDTD, i 3 “eo Substantaiy aptued sing Apa, wan a rhe Cooling Bees 6 ou 442 rummy Lem une he HO 44.1 1Cany Shedig Mode! Deseriton Beto furher conse the FUT cauabos Its dese he om PLUTO (Figure 4-12) nas a nominaly 1m tl (8.1 cmt be precise) tum stn cnr 2m deter I He mot one or ‘fe Livermore Trncot Rage Fut so that etl age we Tomed.Variou size apes 00 ace ple” were meured on the surace ot he cyst next fe Boor A numb o apes eine he ‘nn pas inns wat mnt syed cpt ‘model war 75 13 cm This was conden a aig ce esas of he ‘niet age aeruresize inking exer and inne and ects he ore "Yost Only Fle incon woasie onthe perth F lel othe nero wie were weated computationally. Two interior cavity fete were pysealy employed and computa mole tay 229 nd Toei needs 225 emo pu the ty otha {ire sboned at Bu nds was preset, ow oly partly pun he cay 0 ina ne shored toed and open he ote mas preset The Wes ‘ould be sat at any Oe ne aon Eigue413) A fie tone ‘ere compat model forth 225 em eavily and the cena wire Inction Test Pin (TP) 2 vas computationally mold fer he 30m “The wires were conputanly mole wing he eaonsip ‘hong he wire ran, Thi 6a "tick" ul cll wire computational model tne Fine Dijeveme Time Domain Method or Flcromaonai whee se Se te tat OUR 414 Fe eres de of LO, sempre tng ct SeERe bt psr rhe tovcompeneenRefacl erasy Sera tinct Eocene meaner eas Counting Bees 6 4412 Transient Range Test Facity "The PLUTO experimen object mst in tr be laminate by incident pulsed electromagnetic eld (EMF). This il was provied experimentally By the Trarsten Range Facility at Lawrence Livermore Labo. The measurements wee taken. The existing monocone was being replaced by an Sppraded monocone with a machined “row pese™ a the Werte and ith "eh sorprted cic sde pan The newer monacne called the EMP ‘rgnering research rmidiectoal uiatr (EMPEROR) allowed operation Spo at as 18 GH, But was no et aval. Along wie was subsuues ite pce "Themdtotnge nen extey ell behned fel for shot times. The dsadvanage was an approximately 12 dB over field sengh. An indoor fc (ony ater surounded by an elecomagnetic absorber). i ‘these comitins had to be limited to a 20-ns duration As seen later, this data ‘tucson obscured some very important response fetus, Source ano with a damped sinus tempal behavior E=(B Gand |e|= 70 Vin 7 /24m) sinfa{osns op #8"» wee @ 1s me elevator angle. Ins ume-comun computsonat 1 Nels imei a ry septs rel of the en ler oe nds aproimately twice as re as te experimental ime rai el I Isao coresponaingy lager te Requency domain. the water futon feemed between thes to sce, the evened ant FDTD roel wae lied 10 al the predicted dita alowing a det comparison of PLUTO ‘easements an predictions. The upper fsuency lit was set bythe pulser [EMPEROR in pace used an clermugncic absorbing shroud und a swept coningous wave (CW) source. They provided frequency domain measure ‘none to 18 Gis in ampbade and phase. These measurements coud Be Inverse warsformed to ye ime demain data in eer configu, thin wite er EMPEROR, it was no feasible | irl mente crane oa nro wire. Cmeerialyaalable cutet robes suitable for his ype of wie measurement, soc she TeAonin CT "espn cnly up 10 1 GHz. To maximize frequency response an equivalent ‘int chu was pls Aenea modeling aneetewo apled 4 Cowxa cable entering he base of PLUTO with only it center contr {xendng beyond the foo. Using time domain refetometry (TDR) inpt Impedanse meaeuremeat and he mearued ata sale pons the we eres IGURE 615 isis heh ont oe st ETE responses fora wt bitsy terminated at its base within PLUTO could be [BtGoed or the compansons made here die wie was assumed 0 be shored to he foe 4413 Data Hvatation 100 ae petting 19) wa mae forthe 2.5 cmavity, withthe wre the center of te einer a TP 2: the wire was shorted at bah ods “Fac nn war 496 une sep ong an ook 4 ow « VAX 117780. Th ie oman pede epee was tncated and overlayed with the measured dn (igure 416), FFTs (Figure 4-1) of these 20-08 long data records were Rit andthe spree ws god. We ct at he ine tthe 100 8 | tecod, with what apest o Be ~64 eyes atte loves resonance, showed Te decy. Not surprisingly, hen, an overlay of he FFTs of the 100 ns rrefiged da and 20s tested prod data (Fue 18) sow sei eeiternces withthe 10 eon ving much sharper peaks with ‘ghee amps, What we esmed ler ad what dsc inthe section Babs a lpemue cot tate hres resnan bad a Qin the thsrand [Je rote atime eco ore hike T0000 Tong or the beer pt of Conling Eft: o RGURE 17 Foie mr ny eee sce ame of 25 st speed pd eh ce fogs) ‘nouns ne nein ben) 500000 ine swith be PLUTO cpanel employ end tie exceeing BAN) & VAX T1780 “Te tap peak ft 00 epi canbe ssid with ‘ ea ne inte wih ro ype |. Wire or antenatike males conespoding to the TEM modes ofthe outa geomeny wi equeney even by fone gS whee Zhe wire ep cc the ped of ight the amen. at (Gael) gives ot onder harmonic in heeping wih the sey 68 ine Finite Devens Tene Dorin Method for Elecromagntire esas toa en de wi ct smear Recah foremersmmarnclemat 12 Cavity ees enexpuning tthe uansverse magnetic (TM) modes of ‘hecaviy picked up by the wie acting aa cavity probe and ven by cre Ris the cavity rads ad yi te mt otf the equation J) re sa dell examination of two fie 10-8 sows 1 ie Semaescutor sya te bsee te Ce Tre Hemcsy whe 22cm te open athe ced hed ot bt. Hore 20 ord as ete ne ‘SLR ate 2) wre cen ner Aspe mae 2 Oe 2 ih ea sal wine oie sig sca fay in ies ial 0 is ame wa. Cuphing Feces ° recut (om FIGURE 420 Fie eel 522 Sci wees 30cm cn P20 the 200-n data record, me would hve ben very suprised to in an even ‘igberQat seeming variance with th asfonmof Figure 420, The 20-n¢ “Staevu pvidee = 17T = 5 MIU inthe waft aa 300 Me, ‘2 100isall at can bereaved. The bigherQ comes rm iting te data inthe tine domain, an approach we rely on heavily in the next section. 4414 Validation of FOTD Resuls at Frequencies Above Aperture Catt ‘Tie al equcny dain cop have aay ben sn aT 2 fora 72 Sm ire shore ar tan ens Sint comparsons oc TP I, 34.208 forthe tame geometry show generally similar good agrament. The code greta wihin Od inpomer amplitude and win 10% a mashing the tequency of te resonant peaks. We concluded hat FDTD is capable of making reasonably acum intro coupling predictions for modesty com plex mel objet en hs ety sage of develugcat me 8 tha Freqsency domain code shouldbe used withextreme care ifthe area sed fal. We nr thar Qs on the order of several hundred imply extremely clese Tiagciny spacing teictents fv at eaperinet if We repre in be sccuraely haracterize, Now welow Ox nthe hosun ar posible with the type of geometry examined ere and thal even FDTD mst tern fog ime ally eae te da ‘4445 Shielding EMectiveness Characterization Above Aperture Cutaf! ‘When the ite este eons for shee a siete wire te ‘tied ne eaves dain ier eng smoothed over moet feauency ‘rnd sing th auto spectral dens algorith the tos ner uityabove jst et (Figure +21). Th a fe the cme it eed ‘rvides ite or shielding abo aperure ut Below aperare cua he ‘a generated inthe sfo for wires shred a hel sete fo soe fo 1 Doras Matha for Blvromagnatie sceuatly charcesie the sponse Below apere cut wee te ly Knot slike pont egies lng te eco orc 442 FREQUENCIES BELOW APERTURE CUTOHY "Spel condvting sel abut uty with wml pene and an ica ne wit ay ing ite eminatons wer used inthe previous section Ti model cectomoptete shielding, Exprincts show ite slang above ‘Germine the heldng. These experiment id ot have suficint resolution to azquately resolve the shin behavior Below apertre cu where me iota FDTD seed ere with pt milo tie steps to accurately taracterize| sive esomices blow aster eaft-The reve obtained were ite ei and are bilysuomarized bere. Only a modest shielding eet ob- {eve below cul fr perf conductng Ses with varying size pers Teun of te, wie esate foc apy a0 toe egteny decreases, The O ofthe wire eonances als increases rail wit eresing freqonc)-Tese effect are neal feigning the energy in each wire mance vay igh sno fn te anette wie repo 6 he mur sine vars, This reduction oly eases modesty ‘ith smal ‘perue ie 4221 FDTD Geometry Taplice fhe PLUTO geometry td in the LLNL measurements ode etindis ue eponaue section of chapters conser etn ar cavity odsed by 28 88 FDTD cells Figure 4.22) was wed Wath ach cell dimension 25 cy this bjt has oughly he same dimesio 8 ne LUND tna Th elon tin large a eon aed For the provi FDTD raodlof PLUTO, This cll ican inenor cavity mension init he Coming Bes n CURE 432, coupeg ectanglar cavity made to nly he is four resonances fan eral wit ‘This she region in equncy fom abou he aperre ca Frequency on ow torte sperure aes we cose, namely 849% anda eels, Tete fer enngh meoan tad pert ci hari te se Frege) ampite envelope hehavio and the resonant with a est 3p- [rouumatyA model win more cells on he onde ote $2 52> 8k cls ‘eta mol PUTO TN se me, lhe alls aoe Sev ‘dy, bt could ot berate salable sources. a VAX 8580 Rn nes nts mice for monte es ner apron 40m CL "The wie was centered i the cavity. sorted othe base ofthe csi and sm ear ep a Re cn, png ares so of he py ne tm. The ankle resonances peed st sprosimately 125 178, 1003, an 879 Mtl, as expected trom the om Hep oe we Sd 6 {ermine Th wie enh we ray ts ne ws ta of PUTO "epening on PLUTO configuration, acouning forthe lower resonant re 12 tne fine Diveme Fine Buna Met for Ektromagnaie quencies despite the roughly equivalent cov sections Apress 20% 10 EINSS27 Som. and 10> 309 forthe etal cavity mere iar ne tole lager sper wed with PLUTO. The rectangular cavity model ‘Salgured to give resprses sit t0 what woul have beh een with PUTO nade measurement pen race fy ane! = the code mulling BLT mi for afew millon tine ses ‘Our de is wed a perfectly conducting meal nd then alarm. We asconered hat te nelson of te copsatity of unin ad only = ein efary even forte sae ape To more cll sete eet SeRRSinig we inteused the sm Toss of he sted unt ey eae thea ns for he #2 xl pete a he west onan, Ta SEE shen o eas aponimaely 10° 9s eo 6 fran, Ashi Glos uch lrg and unease clang i the renaning scuson nly tren the perc conductor ests 44422 Incident Feld Desrition Decne ns was shag sy. en de sided an anise seoren a nined he pose depend ld amit removed [SIINGE spe soe th espa cun be aul othe dig fel “Sur appeach nso uoe 1 san waver sini that Seseibes sn Cooper, which decays penately 0.0 ofits peak amplitude befor being wuncated Traneauon elects re therefore onthe cer of 120, Ui down ra peak response aplaes ure yee tht te aacnedc tie ace! smn he Cex waveform or pulses dow ‘y's early xual actor athe Nyquist frequency. Any alae signal isthe insignia 44423 Response Predictions ‘Carer apnea he as fhe wre was eel for he ane apr ong The riporme as charcterze a he requeney domain after Fours ravafrming the bine domain responses 2s te resonance amplitude ‘Roy ounce wah ad Q and romance energy Uo) ca te FON RIS lonanes ie Hghosfeqveney ecnance glad Bee ‘ictessonds wo approximately 16 FDTD cll 9 a wavelength, easonably “SESER ate upper mi egency former cup) reps na Rees, enc tme oman ar on esrpotted peak ami At =O Shey te ean energy Uy These wales were ound fom fitered (sine ‘AeEhpue equeney dori Window te domain responses hat et oy Ce revo rough a ‘Ouran was st cise on the 8% 4 cell spere which asthe faster dey ate orl he resonances, since ashe pst petit rae edn uate hoagh hr apr fay we made rs Ergun, 18K 256K, 312K, and finaly meg tine steps. Ths progesinn| Eocene mere oe ino longer me ecru ay apes Cosine Ec n satu te ie tna ne ove ey incase nly convrped sae aut aa mich ie ie zy, bt a mich higher expetdand wih ams higher Qorequvaleny seh ane eects 12) Tncimazes na menor yency resonances that did not require the ve recor sh accuracy (Figure 4-25), = Fog scar eats Nowy etal aso a nd UY" we fd wing hate (a) A,cwPat, ans = Ae * da, 1A Phe Fite Pigerence Time main Mth for Flcromaonati Chine bees FIGURE 438 Els te gros for Bt ape. am : When sh amar aperares we ound hatte west fewwency | esonance hd ot decayed ch afer Ilion tne ses. Neves we Sout ot energies Um 7 A) wing he lve eatin) ha we | ae teas Uoiee weir oan he NST a pe oe a ta et ar temans fa se ee ninth ei tnd teh i abet esi. se errata rata pommel via nse fee) SMe ee eta snes we me ee a rh pee et al re ee acc gumann Seah me toma sae tm tsa ee teopon Ene ones eb ape cnt tse Ca sent asters ae a Set ee on Ae ea cane nat ve Nets or te ia coe Oe alto ct acl te massif re ee Oe ans apy ao we H Cee ee air ey ine aa a eed ud tatrnlt napi ea eae (ens 20 ho Sms ae ad ee i een hee cn sy el seca ae om caiey es TABLES Berg: Taste ane 16 The Bite Difrence Tie Domain Method for Flectomagaanic rrequency (ix) 1 WILION TIME-STEPS AGURE 436 FFT magus rane or en 4s wells inteion“", Clery more els ineroepemetrsand rep “Save vats need be cvained wing the ight obane er fo roe flly characterize sili ffeciveness REFERENCES 1 ae HS. ls EMP acon, Pipes Tne nd Ref ‘aan Rous Wehapen oe {unt Cand kr, Uy Rtn te ons ne el coer 3018 pn remem om, re aye ary. Ages [Kang Sat Lees ements in ea ‘ren Conan ees so een IEE Tos M2, Beit 8, xd Tne Foie Lite Ce 3DXL) {tha a Gr NA, ey Seth con ime HE a Bh | | Seth carne ep ‘Reta Laren Lite Na Laan Eom Meu real} eae ai GTi saeco name ie ‘Seopa ett Tan oe 1 Chapter 5 7” WAVEGUIDE APERTURE COUPLING (Article by P. Alinikula and K. S. Kunz) 5.1 INTRODUCTION Waveguide coupling through an aperture is a fundamental electromagnetic problem. It has been approximated reasonably well with Bethe small hole theory,' and with some more recent modifications of it? However, these analytic solutions do not apply to arbitrary shaped complex waveguide cou- pling structures. Instead, numerical methods must be used FDTD is an appropriate technique for time-domain analysis of passive microwave and millimeter wave structures. The strength of FDTD lies in its capability to model any volumetric structure with what are typically rectangu- lar cells. No requirements for symmetry or smooth surfaces are needed. Better modeling is, of course, achieved by using more cells. Additionally. the FDTD. technique, as shown in Chapters 2 and 7, can handle complex materials. Examples are lossy and anisotropic dielectrics as well as ferromagnetic mate- rials. Some modeling of nonlinear materials has also been performed, with results presented in Chapter 11 Recently, FDTD has been used in modeling microstrip structures,** and in predicting aperture coupling for a shielded wire.? These results suggest that FDTD can be successfully applied to waveguide coupling problems. To show this we will model coupled waveguides using FDTD and demonstrate both propagation and coupling capabilities. The forward and backward couplings are predicted for two geometries: a single circular aperture and a dual circular aperture. The single aperture coupler simulations are performed using different mesh dimensions and different simulation durations to observe the sensitivity of the model and by that means determine a suitable setup for FDTD waveguide simulations, The dual aperture geometry, however, is selected to be the same as in Reference 2 in order to be able to compare the calculated results with measurements, 5.2 APPROACH In this section the representation of the waveguide geometry with FDTD. cells and the method used to launch a waveguide mode in an FDTD waveguide calculation are both discussed. The problem geometry, shown in Figure 5-1 consists of four sides that are defined to be perfectly conducting, i.e., the tangential electric field components on these planes are forced to be zero. The remaining two sides, the ends of the waveguides, are terminated using Mur’s first order approximate absorption condition’ (discussed in Chapter 4) to absorb the incident waveguide mode fields, thus approximating matched waveguide terminations. This condition produces some reflection for waves that are not normal to the boundary. In a “ ‘he Fie Diornce Time Doon Mth fo ectomagetce HOURE £4 Cog! mingie ae ape el Sant show, ssncpide problem, however fhe se mine Ie mama a the ist ‘nde Mur Conon i shown Yo be suit a Feast a frequencies above ‘uot "The meat platshawees the te wave defined ae infil hin bu tected it astm approximate thickness of Ballo the cell. Circular gees re apron with square ces Sever ier cel lamers era sand Tos ere wie. Naturally, he atl ehape fe mode mone Accurately when more cls are usd. The distance of the apeure fro the Source launcet was selected otha he whole ple ean fi side te Eute ‘Components inthe backard and forward propagtondretons were sled toteone fourth ofthe aperture diameter away rom he edge fhe aerre, coupling coefficient was cilcuiied a the ratio of electric fel atthe est iecation inside te upper waveune ao he octet sd an unset "Two difrem apenre geometries hat were analyzed are shown in Figure 45:1 The fistcosts of single cemered coupling apr wrth cclar shape nase iige sis. The ameter fhe pero 6018 whe the Froud side dimension of he waveguide. A the waveguides havea 2:1 rato exween the sie Thoe feet mesh dimensions Were wed: 2% 1099, Santer and AP 6 108 in je and an ietons, wey “The eomesponding nts of the apertrs wee 8,12 and 6 el. The ‘on ctr hs two parle! ela coupling oper tha ae lgy ‘The diameter ofthe apertures is (3 cell. This eometry was constructed 10 be ‘Menlo that i Refetence 2 none compre the ess "Th ide waveguide ade ide nde mano ing a eee wal hin he source locaton oth he oe propa non one ‘drt. Atranvee lect el sources allowed io spat oly field has a sinsoidal distribution across the ple resulting ina Wave that approximates the TE, mode. Teoporlly the incident field appeas ia the ‘The sino vaaon Is used to nize the DC power abd SIS the spectral cama ofthe pe upeat ia feguency. However, a ues pulse sone eneay is forced side te waveauie a eabencis Blow cu, This ‘tegy i doy 01 minimum y Using elaely lng Gaston envelope 5.3 RESULTS Stale sere cole simulans showed that ae 8192 He eps me ‘emaning ener in poping waves et ell ha ie raion 1 negligible eet forthe fst Fourie tasformed (FFT) response. Te fr ‘and id ack ward coupling cotter $2 1 x se geome wh "cell hole darter are town in pues 428 and ‘Te sent ofthe model mesh dmesons ws examine by comp 84 The Fine Diference Tine Noman Meta for Flrramoanetis rome a very a. The resus inst that 32 cls sufficient for cling We boa sie of the waveguide and 12 cels forthe apertre ‘amc, especely. “Te trata hockwad coupling coeticeats fr a coupler ith two pert shown ia Faure 5. The expert Forward coupling rss etc a ey love mh oth expeient Except forthe bund ees. the sation site fall ine the measurement accuracy range: Mores Incevy.it was obered nate dev fumadackard soupling) we Inger han AR Sine cin hk wit iim of sroialy 2B arecviy 5.4 CONCLUSION Simple wsveeuide coping problems caa Re analyze sing FDTD. With sell modsled geomet aviary tide he measurement Tor ANE CaN [cached Although es singe problers canbe apeonated reasonably Well wih altel expressions, the resis show re poeta of FDTD w elmore contested sea poem heraver enh fe ‘ic and meal screws andslab aswell as waveguide junctions can be aly Inthe same nner nd ithe same rescues. Wareenide Arete Connie as ACKNOWLEDGMENT ‘Tis shpter adits gues originally appear as an aisle byP- Albu and KS. Kunt in/EEE Microwave and Guided Wave Ltrs. VOL 1, NOB, REFERENCES 1, Rae Thea ie by mal es Pe Ret 6S 3 [Bee UH AR EH, Mhurba ‘ted Koa. oi 4 Kani K, Sh Doandbaeirs RJ Lot sn lag tts Chapter 6 LOSSY DIELECTRIC SCATTERING 6.1 INTRODUCTION In Chapters 4 and 5 we showed results obtained using perfect conductor implementations of scattered field FDTD and compared them to measure- ments. In this chapter results obtained using scattered field FDTD extended to lossy dielectrics will be shown. Before presenting these results we restate the finite difference equations for a material media developed in Chapter 2. For the results presented in this chapter these equations do not incorporate any fre- quency dependence in the constitutive parameters, €, 41. and 6. as discussed in Chapter 8. The materials considered are frequency independent and linear. This formulation can nonetheless treat, at least approximately, most materials com- monly encountered in electromagnetic problems. In order to demonstrate the capabilities of the scattered field formulation applied to lossy dielectrics we treat two lossy dielectric code applications: Mie sphere scattering and electromagnetic (EM) penetration into the human body. The first application demonstrates the validity of the approach. The computationally generated dielectric sphere results! overlay the analytic and accepted Mie series response predictions very accurately. The second applica- tion* indicates the geometric and material level of complexity that can be successfully modeled using this approach, even with Jimited computer re- sources, A review of the lossy dielectric equations is followed by a physical inter- pretation of the new terms arising from a lossy dielectric material scatterer in Place of a perfect conductor, The applications are given next. 6.2 INTERPRETATION OF THE SCATTERED FIELD METHOD In this section we discuss the physical interpretation of the scattered field approach to FDTD calculations. The equations have already been presented in Chapter 2 and are repeated here for convenience: aust =H) aH 1g pen) H at te se os a ae (PX) 87 88 The nite Difeence Tame Domain Method for Eleromagnetes For simpicty we have remove the manneic conductivity" 8 we ae inerved only i ony electrce sn ths chap. derenced for. 8 {howe iy Chaper 2, tone exo (ning Ean components ascnerge) an ean Jeu Fou) (eeu. ((€=toty Versa } en)" ne (Ls Ky Wr a), ay eon HULKY"! ALK DY Yat) Hy JK! =H KI) (8) oa “cust ns vf Whi these unions may provide ite itl insight they canbe intr ced on te bsiof ocident,Setered, anole. AS delined in Caper Bie ince i Fi which eit nthe abso fhe tee th Tol il isthe el inthe presence ofthe seater, and the scatered el Ihe iferencebeenen these two such tot he seated eld ued 10 te Tetus fis consier a perfect conductor, Wii he perfect conductor the seared cet eld ual othe pepaivea thence elec ed ue ‘etic Field enced et the surface ofthe perfect conductor propagates ay from he sure (in ee space) and when aed he ni eld reduces the onl Gls inching al seating und shadowing eos Fo osy ils steed ie ae crete thoughout the seater. ‘Tyesesctered ils popugte trough the losy dielectric reghn 80 ta affect the ol il ote Foe hi oan we refer fo he diference “Shon ht py tote ly elec as embeding voto Bouniry (Sa paar ve ew seated ies as does the term Sexe The frst wo terns are symmetie if we reine magnetic conductivity f : (25222 waar! We cose ato employ as chapter, and fs euson wo necesary Treiher =e ¢ 1 iy the rms dssppeat, a expected ress hat the materi ao longer delecine or mage The tens so appear tte iit Ro does wt vy nse stat EL ~ Hin tition =P ive hive a Tose elect andthe eas are feat those of fee space [hcp a and} apes n ple of a ye ce porns Ie tons ow Et mel sed a exarpe ee The em in which depends ono involves Fy, not E!, and thus des not appear as do the em forthe ter constutve paar or a ne ‘aati becomes more rapid. Taking 2 Fourier component of E with fe {ene we ee that eae Spy serving the well understod Behavion Isr i more elect (magne) than conduct 'A somewhat different insight is obained by eting ether € oF go (0 ELK) = BULK) BULKY ws El which mpi E20 OTR for) not eld wi be finds he oss ier, ora igh, but ot inf he al eld ny very showy ones he Body, moving wih he characterise velo 99 The Finite Difevence Tone Domain Method jor Electromagnets 6. NEAR ZONE SPHERE SCATTERING “To irate the actaey and capabiics of the eared fd FDTD fomuaton ee os ace we detrine Ye nea el steed by 2 {ile ape and compar ten he ext Me soto. To accomplish {hiss EDD syproumaton os exert pete wan enews fs eye phon igure Whos Inca in ame Tw diferent ales Sie wen enpyed co 2eande~9ey Fora spereof= (32) 5225 Siang sas ce prose ound og he ye seer ec oy gue hh he deter se. The ‘Sits oe ected yd oe cv each repo. Tel size wat selec She ge dared = 2a was on 4 INCIDENT FIELD. he ince ield was afm plane wave o he orm Prat Rte- testi) ty=(i-e "00 = et st ton so tha behates 2 tpfeson with “emnothed? ening ee rising it snotty 10 os. Nove tha ee wavefons Fist cones he sphere a aor he coordinate sytem an spe geometry sw in Tabi acta. NOTES ® ‘The results presented here were obtained with an earlier version of the lossy cen FDTD cole! which was slighty ferent from the form derived in Gaye twa exponential dileeeing shee sown teow) ewe ecco and sided he moporeencivity * a 0 rede for lossy letic materials HLLKy =H aaKyre afte") “aaa? Big Kr eta 1g ers aelYa += Yoll)) Fa Fe) F(R HH ZaiK) (64) y(n ayn rer yee aft). CUT esany «HEH “ ‘(VG =YU-) (usa mI) olak)-2K-0) J Tis eee reduces 1 form early the same as derived in Chap 2 when handle large values of 0. However, the ode Becomes unstable inthis form for ‘OaUE> I, while de necly iensicalfomulation of Chaper 2 emai sabe ‘Ot only inate ithe une oan arerveron ofthe ascbine oer ‘ation Bou conition tan fs eer Mr, Te ie was eerie Fancy Menno icecream eat the end he pe. (6.3. PREDICTIONS ne amie ress wed or companion’ were fo hy iver ner rrsforming dhe sosoidly excited scared el esonse Hl given Dy ‘Station’ an weighed bythe Fourie tans ofthe nde! Feld FO) The made dna cr ieee tae en rft}s [RoHioledo 66 Figures 62a hough h conpare the FDTD and Rayleigh-Mie analytic ‘olution inthe time domain fore = efor he tangential let and magretc {ld tthe four pins om the apres in Figure 61, Agree cele, especially cosiering that eaively few els were shed i he cerpuans Figures 63 throgh hare fe € = 2. Except oe see igh ‘equ oes agin nemo Tis mn’ ve omnes by appropriate fering or atematively and somewhat preferably by wing 2 Seca es "hot" poe 64 HUMAN BODY ABSORPTION Acxemely challenging problem's prediction othe penetration of let ‘magae eu (MES) fo a uran Body The tre ofthe Bay ge ‘enlisted ashe contve paramere vary with poston Sele the result scssed ere? FDTD has been extesvey used forts problem. ‘poyngmae power computer and more etal el tbody 92 ne Pinte Daference Lime Domain Meo for Elecromagnenes Ne FIGURE 61, THREDE ml ta el hee hen Flan ebro pis ith ecenttfor including the equency dispersive effects of hunan tses thing the mets oF Chap 8 “hi etion ilsttes tat analy accurse res were obaied for ths problem sing seatered field FDTD ith very ied computer resources In seo couse eon of 3 Puan Boy was ger ing ces WS fem in sie ins problem ace of W07R6 lle The ely we Domogeneous with material properties of average sue (wo thirds muscle). Time ay moet sep a were ae oe ae ‘The second model is eoclosed in subvoume ofthe fst model (gst 64, te second mae! renders the tape of he bay moe Fal) (gre $a eam nla trina ge 6a the) ht pose wh the le body slong wih the appropriate consiaive parameters (Table 61), aon repo oh ha ae Frequencies, incorporates appoximarey the presence of te wae body be- ‘ase of te statapem. AC hipe reqences the responses he nero Losey Dietecri Scatering 9% The Pine Digeonce Time Domain Method fr ircromacnetcs, Relative Diktri Constant and Conductivity of Various Loss eee Scanering o rertt tt Predictions wete ade ofthe stantaneusseific bsation aes SAR) according 0 sani~ |B) Ba) whee p 5 the density i Byrn’ of he saat with the cl, Frequency’ domain information sas obtained by applying the FFT to cach ‘sector component and Espey Ine poralzedSAR(O) wasfomed from sqaning the spite fhe elon oa ld afer semen or the cet Feld spect. o* ‘The Finite Diference Tne Doman Meta fr Electromagnetics Lossy Decne Seating ” [Normalized SAR (w) othe est and second rans mere compre (Figure 67) sen the second a a he ee rege noe of he a rm (8 1 = 095 Sim). Reasonable agreement between the cures seen ove the Frequency rage at which he re expose to oer, The SAR (predicted iesured foal values a 380 Me (Figure 68) with very geod saree, 'Whea the inhomogeneous dete hattan made! ee forthe ses 0 she same SAR (0) pret for on to bck saraon (Figure 69) alles eusidrbly trem the homogeneous model. The retin near the frat Surface i hel de tothe reeace ofthe stun in he laboogenesos model of ron 2: When ie fo side “can” i made of te inemogencous ‘noel of on 2 constable SAR increase in he un i observed capared {othe homogeneous an est (Figue I). The enhanced SAR inthe ing ‘maybe of iene, epi fn Ride hype in Cancer team. Move al resi can be obtained withthe sz ‘resed memory ani speed of moder comptes, ad ised reas se 100 The Fine Difference Time Domain Mth jor Electromagnetics Y — | Xe sa ‘ero ect |S cont cr pT ry SOM oy Ditecre Scaring mo ‘Eig deg 161 toc i eh en 88 ‘ef o References Sand 6. However thee cry ress inch even ‘idea resources seated fall FDTD scape of proving nig io problems exerts lew aalyee using ule a, REFERENCES: ‘elo dee acts IEEE Tro EC, 222),20, 1980 ‘Soop Rf faet MOR Fant Koer K pmm of etorce ‘Shi Shere ie pie Maren Pe 90 ‘nating EM sepa naman as, PEE Tro Bien. 3618 197 Stans Gand ytd Tate hs se he fetes en PART 3: SPECIAL CAPABILITIES Chapter 7 105 FAR ZONE TRANSFORMATION 7A. INTRODUCTION All of the applications and examples given in earlier chapters have involved only the electromagnetic fields within or adjacent to the scattering or radiating object such that the fields are contained within the FDTD calculation space. FDTD can also be applied to analyze scattering from radar targets or radiation from antennas, with the desired results being the far zone scattered or radiated fields that lie outside the FDTD space. Near zone to far zone transformations applicable to time harmonic fields are well known. Early FDTD calculations of fa Ids us a S. Because of this the FDTD far zone results were obtained at only one frequency per FDTD calculation run, and usually obtained with sinusoidal time variation excitation? The procedure for these single frequency far zone calculations is straight- forward, With a sinusoidal time-harmonic source specified, the FDTD calcu- lations are stepped through time until steady-state conditions are reached. The complex time-harmonic electric and magnetic currents flowing on a closed surface surrounding the object are then obtained. This involves very little computer storage, being four complex tangential fields (two electric and two magnetic) or surface currents for each Yee cell face on the closed surface. If these complex fields or currents are written to disk, then in postprocessing the far zone radiated or scattered fields can be calculated in any direction. This a good method to apply when far zone radiation or scattering patterns are desired at only a single frequency. In order to obtain far zone results at multiple frequencies a hybrid approach is available which uses pulsed excitation for the FDTD calculations, but supplies frequency domain far zone fields. For each frequency of interest a running discrete Fourier transform (DFT) of the time harmonic tangential fields (surface currents) on a closed surface surrounding the FDTD geometry is updated at each time step. The running DFTs provide the complex fre- quency domain currents for any number of frequencies using pulse excitation for the FDTD calculation. This is more efficient than using time harmonic excitation, which requires a separate FDTD calculation for each frequency of interest. It requires no more computer storage (per frequency) for the complex surface currents than the frequency domain far zone transform described above, and like it, provides frequency domain far zone fields at any far zone angle. If far zone results are desired at a few frequencies then the running DFT approach seems to be the optimum choice. The running DFT method requires more computational effort than the time domain far zone approach described in the following, however, if far zone results are desired at more than a few frequencies, and does not efficiently 106 The Fn Diferone Tie Doman Meh for Electroma rar gt Se ne ee 3g emt lin Sol eB a Scion megs WS ageaeonera “een nan tesa Tamla Res monty if essing m FD en seein sr tae Far Zone Transformation a nthe reminder of thie ehapler running summation Yate ar Zone ‘ranean fr bth 2a -D wil be ven Whe for most ectremnag transformation is me suaghiforward in 3-D. Fo this reason the 3-D wan Fonmaton vl be given is, tolled by mt for 2-D. For th wanstarma idle used. However the ransfonttins sven i tis hacer cou be Pio for piston to ther coordinate stems desirable the far tone transformation developed ee a also sven im Chapters 1 ahd [or catering sn antenna apc 7.2. THREE-DIMENSIONAL TRANSFORMATION ‘The FDTD approach given in this Book ished on seater Fld loan, with mie age ne a eh ome [DID appecach in his book 1 well sited for absining far tne catered ‘shen he arennas being fed mus operat in aoa field mode ees hte ‘Sno iden edo exc the problem, Ins snot asouce ofa aay ‘eo etd fomulaon cane wsedeasl for computation of antenna aiation by mumercly stung he eden eld ample fo 20 an aang drt 11 for seme examen, nie folowing discussions ths caper 9 unerstod haal me| Aer ld ani ae weed ele scot te FISD tae ‘ofthis bok, and ht foramen (ation calclation the nin ld Set to zr so hat the sctered Fels computed by the FDTD cade wil be “Tobin the 3 tne domain far one tasforiton we begin with | ‘equeney domain near zane a zone ransformaton sound a Ramet al* 1 8 the local surface unt nomal. then vector ime armani eave Seater surface ements J, (0) 8% 4 [a snd (0) "8 EO) ext ‘nthe es, whore Ho) and E(o) atthe setered magic and lsc Fela te wrt Referee defines tie Hanne vector potenti N() aL.) 38 2, (oho) a a Ho) =[ Mo) 02) 108 The Finite Diference Tne Dorin Method for Electromaunets wth j= Jt. ast wavenumber. jas the unit eco t the fr zane fel point athe veto te sours pot of tegration and Sa he closed trae sarunding the eset, Ate Ging (a) a (0) the ane harmonic far zoe lec fields ate eal obuaned fom Reference 6 Eg = jexpkR) (- Ny +L 208) wn = jeseiAR) (Ny +Ly)228) a ‘ede fl pot, ad the wavelength a the Feeney of ies Teer develo th coesponing tine drain a 20 asformation| rien thee freouency denendence s17.1)10(7 Tosimply th see Founettanfommation ects we fs deine ie harmonic eto penis W (ole joer 1) Nase) as Ujol= jor 28) Lewy (sae) 06 Reconniing that k= 2c =a and cho where isthe equeney nd the Speed of ab the coresponding etn fr We far ze cet ‘se pave no equeney epee wd ote wu bye =n us uw Now coniering 71) an (7.2), we ecognie tha he Jo mls in (7.5) 2nd) ourespond aie avave apd he expen at ni ing po tor steely) creeps ene Tas ea ely ftvese tutor W(0) and Uta af (75 and (7.6) ante une dom eto potential wire se Sabet A= Ree} ay La area st a)e-Ree'} om 0) fx Hand M,(t)=-i x EC) ate the ted etic Und agit acted sracé cures on ie ctose sua "noun he seater Recner thee een dependence i 7.7) and (7.8) they apply inthe wine dain as well and ane used with 19) (710) 10 Convert ene dan vest pvenay ar ae elesie He ar come sranormanon ww ‘Toppy (77) ough (710) .9 FDTD ealelatons we must inemprt the coect Empat and spi quunizaton. Fis, 2 sed surface mst Be rectangular oe such hat all sures ofthe bot are Vee cll ces. To make {he proves of obtiing the comet surface creas spe a possible, et ‘els at locations carespondng tothe cemer of each Yee el face whichis tom the bo sure Ino lt the proses af determining the tangents fli: and surface cent ts consider the contusion froma single FOTD cl face Sehich ison the surfer of the 8 inepration box TRe compet far Zone the tangential ek components a he ceierof each cll fae Ingo the Fnepation surface a each tine se. Kx, where oJ. and Kare imgers Because we are using the Yer unit cell, the field quite ae a peste at any patclr suc of he cell To ‘Serine the tangent vale: athe cote fel fasts bh a on the imepeton sae SF vals frm ferent cll Wl be spatially ‘erage Ths isnecesiytoakun aici asic) theelece ad Sompenst for av well The time dross in (79) and (710) wil Be Sppmnimates as finite diferenes "When ealaing (79) an (710) one Ra choice of tring he sce current (or equivalently the angel Belson he stegraton sac) vs time ore fr ene W and vectors, Because or mt tsi iif sly spor sorage ml be euired to rete tine istry ofthe nace cee Thanh ofthe far zone W and U veces, he surface crenata ie step ‘ile alse ad thee contin othe ature ne aay by diane) {arsine ever potentials W and U wil be determined a tre Some storage col be saved if he thn and phi components of W and U were sterined at each tie sepa sik Caresian veto components of W apd tbe narod a> pny Lu oral ent F anac Because the aroun of stage needed save te Sx components of W apd isa very sal faction of he storage needed fer he FDTD cells hems, ‘ich tie stp and evalu (77) an (78) afterall tne eps have been ‘opted ‘Ciniesjoing ipa ase te 1 ape and RA ve dey fn te referece poi to he fa Ze Feld pois suppressed otha the Factor Lad the ume ay “RIG in (7.8) and (7.10 are cd when these Toillsrat the aproach we will consider the catered ell component nce oman rae att cee parr of Eval rom el IJK atl rom jc cele Lt ay sre that he 0 The Fine Dierence Tome Doman Meth for Elcsromanetis ‘noma vctr = and that the ,K Yeu cll i wa th nyaion sr Face thon atte cmer of hs ce face E'= AT J¥ 1K)» EIU CUE {Cotsiing the rape ll coments will ire averaging four te). Tis lei Fe compnent wll produce a seared magic M,=-JxE=ES any sothton tissue comttates to Uony-Oursrac ea of imation fede a.and we approximate Eas constant over his surface To keep rack Spatial ference athe center of el... ell near he cetera te space ul ov idea cnvenn reference pons. The vetr rom he reltene ell, cna) ASHE HOH, y KK ee ‘The al o's ade Bea he nea suace of the LK cl face iret consi ne al celine ition a fom he el ene. ‘negra sure "Now us consider how to apposite the ine dive in (7.10) sing od fie freee Wertine it (10 by (#6 thon eve {G.0 ino (7.10). we ten crate he tine derivative in (7.10) as centered Fine dilerence we fn tht El combate oUt es t= (a) and v flonan tye]= anager ~ may cy uyfin-waisr-(-Ale]=anadet -EE"Yiass8 gy Inorerto easement efor els ube empty uae adh seat eset cae sate eps a “Pooroed ino He cares stra bean bx o lative te del. Aho ew me srg acon mye eure or tear oe este rch be mmr te py fel ed om pro he ‘Caer le later sy ene fr oe ied pot wb aan trad nun, We mus baru mt ava aso pense whch tig str frie arpeggios “rn tevnd he dino hams ed ete dW pot ate, ties deni hs of he evn ans ted se the comme ng fps es we dele Rh tne fa fe Mice elle pt one aon sie cbc ote oe ed Par Zone Teansermaion m {it then fo the ctangelar prepped nration suc the tng of time which must be accomodated fo he ar one sete des fq R sistas mas) Practically speaking, withthe refrac cell athe catr of he FDTD space, itis simple al sate to ake bh , ad Ry a hal the eat Saga oui mays pve. Ate qunang t= ma. Msg tons will he required for earh U7 and W vectne componont, where “ ‘Te ay se wl M opp il ot slay ge an coma carp tana ae ca cad SPOS cache Ran HIE wit mh apne ne sop (Stung 1 Ca tit) a Mel oe tug ren pany cet nets ne SSuniniann esaneatepat pee ane ‘Set apa Tor ump nag ar soso ee ‘wpe Fao totus ren hegan the comoles FFT a net aun wr Fr ca ne ran Bose tm ane (1 10 ma ple ot ea a feccrwae ore aoe wre eth eae et) ‘Sons sim ne Ph gee onto ee ins {ined sane pl fee ier fs MEDIO Emwmy oman sca was a pa ee ‘the FDTD sai weferenc point. Fr example to compare pase ern thera unas cag sie PSTD ne Dep wigs me vento Ce met oe meaner pe ej gaa ms nae eine oc tee ‘he O1Dsjue aaa ana fe etic ing oie ems on i nt soma to mene until Todos ust remo c a oe aie eat ed ce titi po es en ens lee Mins toestoncanapapectch estes ie M12 The Finite Difference Tie Dorin Method or Bectromagneis as) and GINT isthe greatest integer fonction. (For magnetic fells al an Bison 0 We argument of (7.19) fo compensate torte ne ee) Acting? 13}and 1) Epcommbesostors 90 ot the Aime nev por tot and egatively over he following tie eva ‘hice hat U, ine rage esos =, ad 1 a8 ‘Ucutdon we conse U tne sage locaton mt san the tne prod fem” ttm +3 amet the coon ie sgn ee Umer a}=daeet |) Joss) aw vufmas}-ararey [35-m)}ea) cam) uiernaje-nanet [See-s]lomay gay ‘hc ahove example gives the eto the te quate vector potential fom te Ecoponent ae ceter ofthe face of cell 1K which exon sonod surface Sa ha ot outa nore =. The sonpiee ee Sects apd Uf al ie ope cm he ing By summine the Conibnns fon ie aber angental cei and mgt eld components Slt cemer of sue fae ei 1J.K 4d hen ke aes tw al of {hcl ceo This hn repeated fx each ime tp. Aer al fhe ime nes ae commited, the U ahd W veto potas are changed fom {Canaan syvicaleorinats, and ine ave coma a ane ke fess ‘Steonined wing eto (07 ond (78 ‘Noe that dcteoining fa 208 ress a a ouber of bisaic sane angce during the sane FTO computation Wout ite ingame saea and sou nt ht» signet rmmuttoal ade. Tnerderlatte he application of the 3-D tise dma nea 1 zone wasformaton, ne wie eaten ty ape cmon ate fo Far Zone Transformation 3 hnain wane scared fen te frequeay domain RCS wil tes be Semin Exams of far tte scterhg efor ples Compo of tier materi. for other scteng geet, and of eset far 7 tala rom anna ae given in Cape (teri pane yaaa x i els abe the cy plane shown in gre 7-1 The pate ot cere tine dmenson eae uated recon rom te Me secon ret ‘election rece at x comer f the poem space, his paces the ‘tors pertrmanc tre second oder Mur asortang. oun. The te ep s 0192 ne The incident pine wave i Gassan phe 1 Vim in impli win e= pulse nih of 2.6, The hlegraton sifce 2 Peugeot he hs dels aco thier ures, sate clase sow the ain gre ts gad pe wove mc n= 99 ‘ans fr zane cos poaied etre lect lie sren in Fg 2 The ess are Fre tnomne and avy he Fore ane fe indent ue an wt in ale th seeing cw srt, ath te ss shonin igure 73 Ress ote vate etho nena shown for Eumparion, sources oe aeeces Pewee he tno mths the lore FIT ration hry eons he ire ie a he FDTD plac (appimayYetl= tems oppened tities thin semen mene ne 73 TWO-DIMENSIONAL TRANSFORMATION Inthe proviut action 4 mat for anfaming ota zone FDTD wasn results deny othe far sane in .D wan presred and demonstrat To his Section te eoteonding 2D trast i dtemiacd® The 2-D gexety sed to model he space. Aer devin the 2D wasormation, some sale rests for 2 far ene searing By apse conducting infnely Ione Tn the 3-D derivation Ie Fequeney domain ar zone tapsirmation ca tions were Fourie rartormed tthe tne domain a served thesis for teanfonning near sone FDTD Fld to the far sone dict in these ‘domain. ede o simplify the derivation ofthe 2-D transformation cur proach wl bo compare the frequency domain far one vector pont fgotions for beth 2 ani'3 By ant'tysoaparing an obi te ane ‘reoded to conver the 3-D tne dona fa zone trasfrnation to Fncion In2-D. Because the application we ae primary concered wit calcula tid ne hme Dyerence time Daman Meta for tivcromagnencs nw, uk ra Gan a pobre pe wre om 4S § ‘or Lone trongommation FGUHE 7 fr see ple bcc os ss ee freely capability. OF cous in these 2D calculations ony the Field components (eter wansveseeecne (1b) oe 1M panzatin ws ase) wll ext ‘We agai suzound he seater with coved suface Sand coer hat quater tangent lene and magnetic time namic sutace carens Ima exist on this srtace The 2D vector potentials which corepond he SSD patti deine (78) and (7.6) re Alo) ie cxf 9%} 1 (o}erfio'cod- 6) a6" 21 F(o)= [2° ear{ PJ (a) en orcole 6") 4° cr20) where and arethe coorats ofthe source point fing, nd pad 6 the coordinates of the far zoe id pn. The corespending far 200e Faded lea obsned fom ‘The FinteDigerence ime Domain Meth for Eectromagnencs EjeomA +R aw aon +, on, (ne can hen easly convert © RCS by apying inthe 3.Dease, where Eig the Founer unstom fete Eo of (7.7) The comesponding2-D seating with defined as {oul ) a FT) os whee Ey the Fourier anton of eae fo (128) (7.28) far ane fields cannot be conveniently applied in the 2-D case, To understand thereon fo ths consi tht we want obtain he asent els seatereo yma lngsucture The Gurnon hs wane ester Blea be onder of magne loners ine than the draton of te pulsed plane Wave which echo The mathematical spect hat ams uso hs comp Tn order to evalate the Fourier nso of 7.21) and (7.22) dre in thetime domain the factor requis convolution operation To aud thi our appoacs mil Beto med the 3D rue iten eevouly 0 provide epesemtive but ot physically observable an te Seseof magi Inga annintely lng eater ould be pialycosuctd, or teas posta) 20 sine domain far zone fields These il be the Rel ‘alacd bya unit length ot the seater corespondig othe definition of the seterng wih Tis fr Zoe trinsic field can then be conerted tothe tat fr tn stny sate ney domain lds and wed for wie ond ‘catering wih caleations assuming pulsed excitation ofthe ime domain FDTD computation. Ths conversion wl involv a simple mulipicaon i lution. Should ce atu dine domain far zone fields be required sey can then be obtained by an adtonal Fourier transform ofthese ests back 1 Imorderio comer ou previous 2. resahsto2-D, we compte th two 8 tun a oping (7.7 a 78 han 70 ee 0).08), (023) and (720 comspond exactly. ad wo ase lve Mtunanng 1) out Rn Mh et FFM i the 1 ended and normalize oat of -D fat 20 fel calettions whi he 1) “Placer sumlaly nrain ot of 2-0 clans, 20 compensation eded hoe eit Final, consider (75) and (7.6 ¥.(121)and (2.2 The aoa dimes sion otiegatonn (73) and 6 compensated ty computing ees ete by i wi ote emer cererponing the stag wth Deni engi fined in (1.28) Ti comenponds in (75) and (1.8 having fos vaaien sn neratng the ¥ vane ove unt ance, The ex Consiering te ering fats, ti ey determine ain the eguency ‘kn, the relationship tence far ano levine cds bend em 2 SD {sr nt anton foro lenght the seer wins? writin se the 2D tani a sone fils B= FE Eo a2 With these sul the 3-D transi tie domi far 29 tansfrm even 2D as follows: ‘excediathe2-D problem, For example, ora TE, computation ony H, Bad the cmresponding surface cents arene. 2. Cakula he ejestatie ane free ne nnn io ing the sD etd deserted eae. bt ora 2D inetaion surface which closes the eater. Let, the cordate uni el divesion nthe Sune tein (7.20), fn ete. equ | ete) Resp isin ower ht is eld is powell observable Ke epee the seatered el ade fom a uit length of he saterer in he tine 3. Fourier transform the est of step 21 and nutty the rest the fazer ac jt (727. Ts rel the ay tte regen Jo a2 far ee wh ten De sed (7.20) CE ‘ater wih a anton of een 4 Tl the time domain 2D Far zone fel is dese, can be bse by 3 ‘snhonalFaunersansormanen ot the esse coane (3) tack 9 tine domain, M18 The Finite Dierene Tine Domain Method jor Blecromanes tn ner desta the aoaiies af the atone approach oon teamion tar une fells in 2D. sestering wis. Hequeney for a el [tec eating yl of aus 02S mare calla Bosh TE and TM Pariations te comer. The clare Sch squares in a $0 S09 cl [rebiom sce Ona 32-6486 hsed PC renin at approximately 1 MFLOPS, en pen eu sinnn 2 Tie 22 att ecm, The Mur second der absorbing Youndary was ved. rr th polarization tbe ceo plane wave was a Gassan pulse which reveled inthe etn, achat as sleds 2D te tps ‘ie eulae ort ley sho the response ot al ime tp ae ‘rca eras The tie domain elt Ta Zoe seated ed forthe Tat plataton shown in Fare 7 The satering with apie ad ice tie rom these els by Foun asfoing them and ving Bye Fourier ofthe icin ple sow in Fgues 75 and fan pres wet wane exacts. ee Fy eampling iin ue ate cetrf the liner. The covtesponding ‘euler TE polarization ae shown i Figures 7-7 trough 79, This he ‘heut polation for apotmarng sol sale wih a ass DTT cles he spree in ath nagnitoe an hase suite goo. A Statler an wot else was sue bin hese res se 05m ‘is corespo at the 10GHr upper fcqaeey hain he pl 929 cells per evelenah 74 SUMMARY An ficient ne domain eat ze to ft zone wantin for FDTD compotion has een present The apo st keep arming ws eae eRe far soe te domain vce tetas de 10 the angel ‘Gece and magetic Hd on lose surface sutunding the care at Ech tine ep. Atte end ofthe compuaton Mee veto pours 6 ‘SS o tne domi far one fre For catering eatin. nay fa Snes dccton an be iced fin during one FDTD compu {anal rn, For anena radon ears, ansea aston ony oan dications ean fin compte ring ane PTD cae. ‘ulti na wane adaonparer However. ifradation pate st any Seles and only one frequency ate fog, the approach Gesre Rekccnee Il fog lots etuny st owe fin I ration faery a sty angles nd at aimed umber of fequecis are desire eh {tering DPT approach of Releenes 3 and wi tak less compete Alicia tha etme domain me dered his caper. ‘he 3D far none vansfrmaton piven shove det yrodces he phys cally stoervable ansem far zoe fe Teed nest els can eon se iteady uate fequoncy domain ees FET The Pm domain [ir nme wansfematon ives the fits adted by 2 unit length of the ‘omer These Hes cathe uso dteomine Wideband Seaterng 9h ar Lone trangormation ww PPC Cylinder, TM, 500x500 FDTD space an if 7 Peo: enemas rth oan ar FO TD we PEC sine, Ma {90n300 FDTD space i [= ‘rrequency ( FOURE 15. Sten ws spine stan a ane ne Sn nt ont ah ct stn PDPD ene OS 120 The Finite Difference Time Domain Meth or Electromagnenics PEC Cvlinder TM. 500x500 FDTD space 1g Math Phase greet) Ie meg mip PRC Cylinder. TR, 50Ne800 FOTN space i | z a . ime ons) FIGURE 73. Pemne rece ld tne TE pie ce re ed Fl sien pts an See use FORD we 03 Far Zone Tramformaton a PEC Cinder, TEs, 500500 FOTD space Scattering Wats (43/e) Frequency (i) PRC Cylinder, TEx 200x000 FDTD space adit 28 meters 5 } Frequency (is) FIGURE 1 Scania with yon baie ete doin ansomet 122° The Fine Difference Time Domai Method jor betromaghenes afer woaton oF aston with the apie miyng ittorpitenn he ext cua wandered fr bath 2: and 3D ft ‘Soe uienasns by canon of ide hand scaring om in Sic tong elias ad ples thro pad plane wave exciton nthe {ine uma Using pe extn sgtcany mae ethan co Truuency tan! of icin Coie nthe capaiy copa FDTD tus or eee) depeen mater, idan est fa 00s tng gt nog aun eens ara san be ote Sen efit REFERENCES ‘Sng inp He Tw ty C3098, (Src conc dimes es pun opie wens IEEE >: ado CNet fl ar Nae, Quy Rescate ih res sce men «Fane za, Ctl yon he Ents ee «Rn, Ss Watney, and Van Dutt Fran! Bares o Communion NEES} ova om Se tr Stes) Rese nt, sce repro fal geo, Clb, Of + Lite any eect oma un Crate 8 FREQUENCY-DEPENDENT MATERIALS. 8. INTRODUCTION ante een ce a oh sgt FID tt runt fr eRe hate yg Rae oo prepreeertinee opera hc ten d eel oman ier cand Seed rnp nae Wists er pce gol onan asa osha al pen we ot [Sciacca nto owen invc forme pease oe TOD can thane sce a rb a eran epee mes vem fr pation En vl ahs ae fey Siena cope coments cma etre he teeny amine ie nay } Penmarrjocntoiie( 2 vele-ieat ay omples eave peomiiviy tha inlay the combined effec of rel and ‘vale, which could be specifi in FDTD cakultions. Cea he Frogncydependons andthe me would te dagrsv, he diet Frequency components of the elecroagetic Held would wave at diferent ‘ne eavency dependence ofthe media mt he deserted Constant el vais of conductivity and peritvy an 1). However depenence than this ad the fair of thi chap conceal with ‘xeosions of FDTD tat allow secure wie freien Bad with comput ton of wana lectremagti el thee mua Foote atl the Feauency-dependeat mater! parameters have Bech deemiaed say. tly we fo otersthey may be eerie fom measured date, Wb out Sluis willbe tance on vt hil pe ei at be treated using he same anproach 124 The Fite Diference Time Domuie Method for Electromagnetics ommpue tie ad srae thn with constant conitative preter an if ‘els ta single fegueney or vera naow teguecy and are al ha ae Seo, thenconctans parte sh hs ard Homever fhe fequenty ‘epee of the aerial ay ff the resus ver he and of equi etre, te ng fc depend FT, cn re case Sf tepuate FDTD eal and changing the estan constitutive para er ales fe cach frequency range of meres. ‘heen! duro of Fegan dependent mi ide he om af this bok, andthe interested rear ieee Yo texts concerned with the Sj sachs hose in Referees Ian 2 fora more extensive eaten We wil bein wit the sist case that of 4 material whose complex permit deserted by an equation with oe fist oder pol. Such mat Peritiiy the Deve caution. Fe these matral the freauene depen ene ofthe copes relive permit is described by Elapae’ jee 4 oe rl) (62) snore isthe tai pein at eo fequensy. ithe nite equncy [rmiye is de lana ame, sn 0) 2 the fegoncy doesn Sosceply Am extremely commen repreeriative material of ths spe is ‘te, th pia parameters (they change with temperate and pesare) wien = Tat! y= 9.4 1D. Tae corepanding comple relative reminivy av funtion of feaency ia shown in Figure 81. Clay the Feritivity of water ver his egueny ange sot wel approximated by = Eine conan ae "Wat hc rvation, ft ds consider ow we can extend FDTD to this sivaton At lens to ifr porsche can Be plied In one he elton {hip netmeen Dat Ex exynensonthe eve domain sy aie uation. “This eqoaton then aprroximatel 2 diltece equation and ups ach hie sto along wih the E and H les. This approach considered in ‘weer eo expres te relatos between and Ein th tine domain ‘tha coolio integral Jee Aslan a BW) =eceutle) Frequency Dependent Materials, ns were thee doin suscep, isthe Fore wanfonm of 0) ad zr elds are asad for negative dime In previous chars th ssuanyion an ben at) and fen za ae (8) Hu oes familar D = €6E, whet i evan 10 € ‘Now we consider how ie might include the fll fects of (83) in our FDTD callie To imply this we wil temporarily work nl cn bu later inthe chaper ores ate extended to sestered el formu feining, with = nas usual = ‘ime mea the the legeaton becomes i part uation tha (8) ‘Seguivalen De =e.coE" rey Lere "| lay «s) 126 he ne Dierence time Doman Method for Electromagnetics "The valu of Da th nest ime se is We now posed to use (5) and (6) 1 eliminate D rom the derive DID equations. Io smpity the alge nvoved we come ony ED “ve, tthe onencion Dc aihtoraar Forth Lee we inde ‘only he Dy, Ean H, eld components and assume propagation i the #3 Sections, Te it x'=¥ Ax we have Hom the Matwell UH uation pela Pit +1) -Hr oes) «8 ‘The conductity 6 has bees eluded tor generality in modetng mater Sahih unite dst) water, hve oti cen mem Feeney ‘Wernced to ciminate the D, terms from (8.7) i det be abet save for Bo" Using (83) and (6) oe fd tat Deg) —ee-fep")-E30 et ey E EP 8) wo an Tea om nd wheritsseumed he conte parameter vse tt material aflecting the elece eld at spat loan x= 188 “Tsim obtain he upateequaton fr # We OUP (8) 25 PDH =(evea eon!) eat EG) e9E BPO") am a aso define a aw Frequency Dependent Mrerils ro Now substituting 8.10) 11 et 7) sei oF, win Sree 1 _¥ emu am Peay he FDTD update equation forthe magretic i is unchanged eo hat 1 are considering oly guns dependent ihc materia, For frequen dependent agnstc teins sma derivation would be spied ‘tect Mawel guint ohne cores pa enn ‘Wien proceed wth he next step, which sto determine) fre Debye areal wee) din acelin gen n( 6.2) Fm sn his ei ou nyete 6 wun 20 susp faction i ple in Figure 82, which dees te Dia ela function of tie ingested for Ent 3), The eres fteetne ix density D isaly resto staan vale tea decays Paponerily 1 ae debe by (81). ‘Aor pon oe made In conten wih (8.14) is thatthe impulse responeis causal 7s 2 for ezlive meso thal he tril does sorespond othe ial befor it oceure Thi Fanaa cae ‘onthe fequeney domain suscep fenton ya) that it mat ave 2 ‘tsa Furr transfor. This contain is equvaen the more commonly encountered Vramers Keon relaontiptetnce tall we 128 The Fine Difference Time Donan Meh or Electromagnetics “| zanl \ | — 0 ‘Tate (plcoveconts) tainly mre intiv, ven Kichaté Feynman amie that edd 0 coms lol wndertand he Cachy theower ‘Having found x0) we can preset to evaluate the constants reed #0 ‘pda (12) a each ume step. Applying (89) 1 (8.14) we bin Pe bculet feet] ws a) a(t [lee Ie ao proved one (17) ing (15) ant (16) we ice ta i ‘order o evaluate the convolution suration in (812) ety we mst save Sil he pas vales of te electnc el, at least far enough Back in ime 10 “Song tthe time mie for 9 dey to sal enough value 0 hat the contibtons from ltr imes are smal enh onl. ation © the compoter storage necesny, fe competion: reqated to evar te noaton mst he mate steak fine rpm each EDT ell Feat he Smnmaton in (8-12) directly means ta oly elavly small suber of DID cel canbe toe ma eomputation fr reqcncy-ependem mae aka computer time reed fo update tse lee ells wil be Frequency Dependent Materials bs ‘eave free nee. mi be fase 1 we FDTD with sia ine ‘ication an un aay computations ovr te frequency band of ies, ‘Shangng in constutnepnancee wns wi oc However, theexpnental tie dependence of will low uso avo his iret evaluation, We ean place ith recursive evasion realign lonseate this we wil define an accualation variable "in pace of the oor Serra an were Ya singles! varale To show how this can he apie eer Fm(nayt = Ray! ea ad for = 2 Eetroyart stay" +eha’ ain From (8.16) we ely obi eatin agree Bae 20 hich is deo the special nature of th xpos Faction th shiing tin time corresponds to muting bya constant proportional the tine shi Tring (>) a 18) 810) 9s be ded SLi +e ap! aeag! ce ty! wn From which it ftlows tht This est allows ws 10 evaluate the sunmation in (8.12) very effin component locaton, we Reed soe only the sng vate Isat of 130 The Finite Diforence Tre Domain Med for Electromagnetics ‘mating se atin al hse penalty thousands of rms in the sues ton st cach time sop only the simple pate equation of (822) need be "To demonstrate FOTD calulations vin this metho consider he -D _oblem of pulsed pine ave traveling in vac mal incident om TOO cel is we. wit he interface lcd at ell $00, Each cell tas 2 Ieagh of 87m pdt ime sep stake tthe Courant iit 8 t= (CHULZD gs, React na Gating ple wave Te pulse ill hac pital with of 296 cel hetween the 0D! ample nation pois, Connect equtons for nal ales of he flea he fi tne te or ‘icng tls Gatean pulses at propagate onl ne pune sae thm, where x1 AN ae tat ty wa woto=to enfcied—i—)4,))) aan ‘i fori sii an eo omer, wth 100 the aoycnsen amp fib peak soe felis 6013 ann 198 Th nue af the pues shown ia Figure 83 ri Mu stg oy ed oe | whe ort pues refetng From te boundary at cell 100 did ot rer the Inerae before the tne spate Spoil plo feat fd ve. poston forthe inerice and popaates inthe water are shown in Figures B-1 w 86 alr inceaing number of time ps, or compurion tsa elation canbe done with FDTD, ht sing constant ales of peitivy and conductivity, If we, for example, decide the the conve parameter t 20 Gl, ten we Find from (82) tat €| 22343 im. The complex petty of amen nth thee vals of € aad ‘is shown in gue £7 ch on comparton wih Figure 1 5 clay mot very god apron The FDTD cael forte pe popup “rape uing hese constant vale of € and Ger 60 ie steps bows in Figure 84. Compara with Fire 86 ote the Tog tal on he eieted bed tots onsara conduct azarae, a he lin tw Split ofthe pulse propagating tough the water "The acurcy of the une convolution etd I wae by compet Frequeney Dependent Materials m1 3 ~. “ash wl Frequeneruist omptaions were conned unl te transes the vaewam-mediu ine {see wore icipated This regard approxima 00 te ney Fr ‘ecanive convolution method bu 2000 time steps wit he ws! Constant “ake FOLD tobe iar alo eel Co {aking the Foie ansfom of he eet fd mente vacuum cl 500 ttn eae anna by the Foran tence ple {it pace thong tcl The led i cbc by stating ‘cide ie withthe ati eon replaced wih vaca) fo the tal ei ote ae shown in igtes 8-40 6) incl 500 ea ime sep. ‘The esting resin cofcet sapdb slew in Fige 89, ey withthe xt solution bined ro anisole meds. The ecu comvoaon ssw ace ane wih etal, he the 8.3 FIRST ORDER DRUDE DISPERSION nis secton we wl deal with slightly mre complicated mati with ‘ompex permit esrb hy the Brose dspeson rlatn'= Eeyore oe ‘ofiv, 8) (825) 13 The Fine Diference Time Domain Methdjor Elecrommgatcs ——|——_‘reqeney Bepeten materia 136 The Finite Difrence Time Domain Method for Electromagnets Mater” w/o Convolution oz CURE Er eps bin meer a ) eee Frequedey cts) ‘Soe and cinco FT compo a, Frequency Depenten Materials us eres the ruin panna requency ad vi the elsn feaueey. ‘ne donot inte cont terme, we ase =O then he ealing ewer 35) has to poles one t= Oa he tera =. The Foi ransom of ths case mus be medio cout fr” In Reference 8 we desl th this situation by wig the tine domain sassy function xid= Efe] Uq) an which tas Fare asf eqs © 20) of (8.26) excep at = 0, wre "liters by 2 Boi, whee a the De de cn Raber han [receding section i simpler ete conductivity tr tans in 12) Tosimpliy the suscep ation We can espn the compespeitivity 2 Conn) an 82 wee aon toee, x(0)e 2 won Comping (28 ad (829 ee il csp am 136 The Finite Diference Time Domain Method forBlecromagnencs ws) Having made thes asim we compe (831) (8.2) Seven a thevar we the Beye set oh prevering ei for Me materia i We te ass) avn vs FET han te cnet of (2.22). Aa we eho trek. the FDTD wade auton obaned ns way erresponds xt 1 faroinned wth eo condut an the sasecpity fopetion a 8.27), IS showin Refococe Thus, we ee tha fr our purposes a Dre dispersive materi s merely Debye materi witha nonzero conductivity. However, the frequen) Alpenlcne of the eung soon print gt diferent than for 1 Tebve material becuse the sgn ofthe exponential sscepbliy ter in {ibis pegate. To Masa hs, cose the example af an stoped fron Out exami pnsn ana laa ngueney of 28:7 GH2 = (29) Thu a colon fegueey v, of 20% 1 The complex permit for his Hasna, inci the effets ofthe conductivity fs shown in Figure 10. “Te ume domain snc uiiy, the Pour anor oF (2.31) shown in Figue 8.11 From Figure 8-10 we se tha the tal pat of the complex periiviy cnanges sigh neat he plans. Below segues este reat fof the permit fe nate, the plasma behaves somewhat Tike a Srrvepuide felon cof in tat elcromagnetic waves do not propaga [Reon e plas imc siamegint meses psa hough te ‘Bsn altgh mith some ss wich decreases wih eeasing aency The pavns Beves moe ke fe space. Hs high frequency mit. While FDTD win conan costae prance canbe appli wo Debye medias tn single eqn fr Dre materi theres some que at which te eztv el pr ofthe permit will cause ne ofthe FDTD mullying Constants fo be singuly,a hieyieney inal FIT canes be wc ‘White ees omvoton em added, however, FDTD isappliable over Abeer Requency range ofthe Drade permit variation, ss demons Frequency Dependent Moteiat 1 Plasma Aetatie Poem tuvity Frequeacy ans) ‘This demonstration wl esa 0 that given for wae ia the previous ‘enleulate both reflection and transmis for aplasia “lah 1S thick Tepper pen ae, The FDTD ct ve = heh $0 andre space nthe remaining cel fit der Mr Soba boundary is se on bath ends fhe pole space absorb he elected and ‘rasmites eis The FDTD ume eps wken aba the Courant iat 5 at ide 0.125 ps as calculations with A at the Count it tended to be ‘stable for hs materia ‘ue em AME Hee ¢ mney conn, an en Frequency that ake exceedingly Tong times vo dsp, much ike He long “a one retiete pute forthe noncnvotveswater”shwn in ge8 resting FDTD calculations at ow tequeacis, especialy forth wasted ‘els as mey nave very How amples a ow tequencies. To ave this has a ener content at 2 frequency. Equations fr the ntl valves of the elect and magnet feds nthe compat space ate t= Orr & {Ganson ple (2), 1.24) can be nied fr the Cavan dete rine wing 13 ine hme bugerence tome Domain Mena or Blectramagnencs 3097801000 180 FIGURE 1, Tne oma ny en pm Bai) Bol 24-4) es) Li) 24 i BEY @30) here he GD substi frst Gassan dsivtv nd te G subscript othe ‘Gausian pulse Ean els given in (8.23) and (6.28. For hs example the “Gausinn diate ule fan half nih j= 6 xl The spectra ft ‘oesponding Gussan derivative pie showa a Fgue 8-2. The incident ‘Gatsian derivative pulse athe it ime step sown in Figure 8-1, andthe ‘esl fl ne FDTD sonata 1000 se Ap tue shown in Fisuts 14 and 15. The pla i obviously mee bihly Sispersive than war, withthe transite and ected ples adally dit fee in shape th fe niet ple Tn order demnstate he acu of cursive convolution FDTD when ‘pled wo is igh dispersive media the reflection and transmission oe ‘eas forte plana slab wilt computa equeny domain ing, Fourier ransforns ofthe FDTD time domain esata was done fr water the previos section, and compared tote exact fequeney domain sluio, ecw pune paling, 900 tine tye wete alae one o allow Be Frequency Dependent Materials ww rans response sete to 0. This earespomsappoviately wo the Fall fie Scale of the suscep forth aun in Figure 811. The vfecton ect mariage and pve ae shonin Figures B16 abd 617 The reflection coeticien esults show quite accurate agrement withthe exact Soltion over a dynamic range of over 30 dB going stnhly through the andthe eal pat of he perv ces si. The tarsmission ie ets ‘rienced by the fbn sare above the plas egueey are as ite (Resttly conftest FDTD rea, “The wansmison coefficient el, shown in Fcues B18 an 19, are curate oer a >70 dynanic range with he reversing slope of the pase ‘tthe aninision coeffi! tow fequnee competed cur 8.4 SECOND ORDER DISPERSIVE MATERIALS Up to this point the comes fequency dos permitiy has heen escibed by either a Tst ode pol oa ft Ode ole ps conc te ats euler in mdr pe wan. Wel hs isadequate wo mode! lat as sprotinately we ange of mater seme uteri ei one or more soon ter ples oat deste tet ‘Sniyes pensuiy. latin secon we wl extend the preva recurve onlin appreach 10 sale second order po, and inthe following Section o multiple pls witha demotion given frame with wo Seven wider pin This genes sul ele our FLD elon 0 ural model ay fegueney dependent material The plete ‘er pole fo conse i the Lenz fonm and we wil begin wih thi “The empen permit fr sequent Spent mate WH one second eet Leet pole can be deserted cole sle,- wn tet pid soi ey an epg ee We tin the comesonding ie domain suscep fanetion 7.0 by cosiering the Founervansorm pa 507" sift )UIO ae 38) Cmparng these equations we se that (3) wil provide he comet exes Sor 7 provi nt 0 ‘The Finite Difference Tne Domain Method jor tetromagrecs Prequeny-Dependem Materials ‘Freguetey (oe) rice 2, egy ps fe enn ai pe ~ | 7 i) Fel (ite lasts Sab Praca tab fj 0 \\ x0 | \ | g 2 Noa hh of a HI / ~ NTT val | j \ \ | vacuum | plasma vacuum 222f sac plasma vacuum an saan os en PGUNE Ee esi be in eva pe Ve. Ter ir ie ee i roan | 184 tne hme Lerence tome oman Method or Blectramagneics a9) 5 The time dependence of 7 in (0s ot a foo for which he coe- sponding deci eonvouton can be aplated veusvely- However, if me ‘ein complex me doin sce ehoeemtutay aan, 49 joerc] aan shar the “donee compen quits and Re isthe ral operator,» recuniveconvolton canbe implement. Applying (8.9) and (811190840) really proues| a) and sth he impor of (843) bing the elton mg ay) ‘rhs als ei ea of he snr ow dine Frequency Dependent Materials us pene oe cb vale ie i i ng fast oder poles using EBay soem 646) Finally, me obtain he quantes needed update the eect il (8.13) by taking the eal prs of tes sani ay (nas) wit me sumenation of 8.98) pastes recurney ang (85), ‘Wepa toethats event eter ple nin dem scepebty he onolution of which withthe elec Feld canbe eats ecusively movied ate recursive aculaon compe Other tan hs complek mei a suscepti fein wih + werd order Pee‘ mone “ite t evaluate reurvely than with ist ner poke le te Lorentz second order pt has a Fur tasform given by Sich af magrtized plasmas an fers have Foute anf that aso vole me Foner tans pr Salas + Jo) WO -hernlgyJ Ue ae ea) in addon wo (8.38). In combination hese two Fourie anf pis can be tied 0 yield complex time dans sscepeity function for move Shmpiatedpenaiites that Bove cfnm 146 The Finite Difference Time Domain Mth or Electromagnetics qt) e080) +7 ,9] U0 {-inkiormt 30 “The only extension rm he shar tay (8.400 elaced by i ‘Sone exanpics of FDTD caleultons formate wit more complicated Sood one dpeson ae med i Caper 13, whi als aly wih he ane nie acum elt for FDTD cleans fr 3 mati 1 ee eed rr Let ewe i eter sa weg tan esos ma os er hwo seco pres 8.5 MULTIPLE POLES Woke mt tae acon ein os egy een ffeqacy bod of inert, some materia, afc! det, and opal eats forexanpe, have mie complctcd feqcor Debs For bese Trl nrc tar ene ple may be needed in he peritiviy Section 0 Tere behavior Bee we ae dealing wi incr mater, comb inp efe f exh pase mere ange seu eet rice and pve av exam 0 material wih wo second ode oes. Othe Cebit of poles cn be cose i sar mane "orca sopose th send Gd Lorentz pols deri she complex permit 28 co 2 Go cers 8)S Sani at th the comin ht for ere now os the resonant fen forse ph oles is the damping Soft te pn pte Cle cont che thee of {UhoTiae previous secon necessary noe mute ples inthe FOTD rs ete sown. gan 9 Wee) i Frequen Dependent Materials 7 Parameters forthe To Lorentz Poles peony rene | Psotharthese equations apply teach pole seperate, Eston (7) and (248) ae moied =| 352) ei : ae) won wing 46) “Te demonstration will gin vole akculaion of he asi fel for 2 pled pine nave mnie tana been va ‘he dispersive medium, bu fr thisexamol he medium i deseibed by to secon ore Lorentz poles Refering (51) with P= 2, the constants which Table #1. The comfucty taken a 0, The feqney don cme relaiveperuavi orths median shown ip Fgure 820. The onesqond ime time domain suse fn (38), wilh ens la es included a.) = 2.004%. shown ia Faure 82 For the -D FDTD calculation 1500 spi cels are wsed, ith he vaca: eda tidal 200. fit ner Mab tun in Inet tcl while the 1000 cl tepion connie the depres provided adeqste ime dla othr pulses refleting from he wry a el ante arth sare forthe water example Seton #2. with x= 37S Hy and at = 0.125 pr= ie, where the speed of gh in acim “The excite asin pe plane wane Te use fas he ane ‘ed forthe wate example and its peetum is showe a Figure 83 ‘Spal plas felt ld vost forthe plea it eft fo he lnc al poppe ie Serene mein ae snows Fp 2 108-25 ater increasing ner ote ope The ote pre ate ofthe medi leary inate ‘ie wevray ofthe FDTD resis ae again shown by comping te ‘election cont eauncy fom the wate ds. The FDTD com Pregueny Dependent Mererils wo ne rman Methed forElecramaeis 150 The Fine Difference Time Domain Meth or Becromagnecs | os | rn ha deeb eee 100 V austaaber) teetie Field V/e) medium Postion | (cet Number) beste Fie 07 {st pm her 00 ETD ue ep Frequency Dependent Materials a oa mado Frevueaev ail ‘ed cnc FT compa eh ans were conned wt he rans a th vac oben medio nerace ‘ete dcp Thc nied poi 900 tie peo 280 pth ‘hen compared io Fgue 821 ceesponds easonay wel ith ch dustion ‘fn sabcepty response. Compaen ofthe reletion cof at {complseinthe sare nar cf the screams Th eng ‘elation coeficietsapitde snd phase sre sown in Figures 826 and 27 and spay ence agreement wth the ext slat, 86 DIFFERENTIAL METHOD ‘The prosous sections of hs chr psc he ouive eee ypoach for efficient cluding frequency dependent mately in FDTD ‘aeutaons However, an allenatve meth exis thatespeses therein 8 convolationinepral®® This section watines the diferent meh and ona it withthe recursive cemlution meta {Let uc uso for an erample a fequeny depends mata wth singe secon order Lovet pole. Then, fam (97) we can exes the relationship ‘etweenD and Et he feency doa &s 152 The Fine Dierence tome Vomain Meth fer Eesvamannice renee Ve ~, i . \ B [= ne | wot rr oe eC GURE 7, rho esi eu Lan eau comps Dee) Fo) «(6 cree canton scone tte ne oman ay dierent ert gnaing ys eqivalen time devatve at a= JO) 35 im rane Tey he ileal int SES ner ie 201 28,e. 78 eH web 28, Bo SE te sono a fie ference eatin ata! «26,21 +2)0"" 40" lola’ - 25.1 2)0" +4." ojate, 98,4742 JE" 20 sfepate, «28,4 +e.) Fegucnsy Depot Mais 159 ani up at eich sme step in ation to (8.13) and (7 (wth 2e0| Fee Ths at each ine sep (856) woud be update, th re ad ‘0 uate as with the esaingH fie sed fnd the ex ne vale of Din tire kal} gow fase compar the rleultion ein ov (86) othe for(8.16, the recursive convolution update appears te relatiel simple anda require Storage of fewer mutupicave constant, een tough i egies comes [ter trae for previous values ofthe Fils To apply the diferent equ ‘or apron, nore eal (8.36) lor a maenal with order Mig sing M previa vlie Pant ML, atom previgs sales of Eo ‘be sored relative wo nonlispersive FDTD ealclaons Comdering fourth ‘over ssperson, comespoasig to the example given previously with two ‘eon ter enon ple, the dione wea wos eguiesoage af ‘even aoa el variates, while the recuse convolution apc e- {qurea storage of wo conplet vale. These two complex variables corre ‘Srl oF varies in computer storage; otha thr i ery ely 8 2:1 svings in storage by using the recursive convolution mato fer is cae [Aste order ofthe dispersion incretes sro sppogches 2:1, with the Aiea maths in gensra oging 2M el hacer varies whe the recusivecomolsio method reques M2 comple bcksore varies, ‘ie eguvalen of M eal variables. "The eplanstion for ths ilfernce Heo th wo pach eh the rem equaon approach egies nding te lax Sey D bere fd Ing, while he econ convolution meta does ne. Ts the diferent ‘guton metod yee addon information ie vale of Dat cach tne Sep), bu atthe expense of atonal ealeulaion and storage. This aon !nfomacon regarding D may be weal n some sivas for example, “ealing wth noize mci ad ach st the eal ue ‘on preach maybe preferable Imay alo be tht to th ares mer ot sorage locations wed (nd therefore Increase iformaion retuned) the ‘ieenta squaion approach ny te mors scat the sas aber oF 184 The Fine Difference Time Domain Method for Eectromagnetes tle ee od ‘The recurve area appears sme wo Implement and ‘res los atonal compte storage per FDTD ce 8.7 SCATTERED FIELD FORMULATION Ine preceding ection ot ms nates ecrsiv tain pro toning fsuecy-dopnten mere FDTD bas been peated or “imply these derivations and simple I-D examples wer caused using tea ells, Ar we have dneased previo apes, fr any pois Scored eld EDT cael prof Even ascatered fel ode, See ion is owe ttl fk in cel cotaning dispersive mattis, This vk welt oss toe ig es accra crn eae am Digest hourver, hate atvanae of sii “his ction derives ie acter il for of he recurve consolation meta We toh psn Cpe a inp ete Dad Etat ool seroniion Th exprsing this elation ship asin (83), (2.16) Becomes gen ee) ae sole" ee) a] em eee ps (i sa 39) where" denotes convoeton, Following the procedure in Chapter 2, 6 vet a0) It we apply the sme proces to this equation as was dane in Section 8.2, he ful wl bean etn eqovaen to (12) Buen ufing ewe ‘Songonct ofthe seine fal nad he al ea Frequens Depentem Materas ts aera es) weve mea ero fied) H (son) ‘The scusve evaluation of the surmation ofthe sete fill wit the seep tne Intron the it sie of (861) 5 pefomed asd Cael peel ith shaper fort eld th change reted. Most ‘the reining terms re evalsted in a stightforward mamer Howevee test ern (1), he devine of he convolution ofthe cde field ‘the swept fart resus ome addon tre We meat ke to evaluate ts ia elosed fo, both to ede computer te ad increase ‘curacy. wets can be done foe Gat pul snpler oso [lshif we comers lfernt fmf cde plasma coin ule ‘We now digress fom our implementation ofthe seatered field form of recursive conolutin FDTD to conser this pulse “Tho indent el at spc ring moth orne ple i given by aay 1/32, and = 287. where Tis teenth ofthe pl and she delayed ine the eles ed ton son {5 gienin(34. Compre this pus oth Gass ule deseiedin Seton 34 Fr como predate pid he sth Pube, ie. T= 28. A Gausian ple with B= 32 and determined in 156 The Finte Difference Time Domain Method for Electromagnecs atsoutes ys tintin 7 Eton ron coop s Gain pale . Suton forte sep of A100 178 pee shown ia Figure 828, The sooth Seslbate sow m gure has pri T comesponding to B= 2. This ‘Shue orB was decried sue hate Fourie wari pea of he eo sane same amelie tos fortene. Tae spectra the two pues ‘Geran Fgue 829 withthe same normalization cs clear tha the ‘Shon cosine pulse des not hve as wie acominuow feguney Eada te Seat tarpon pb atin, aia most tuations the Gaussian pe (Riper Hower the sooth cosine pulse equation i more realy mrad anally wih he sarcepety neon and weatlize Werf “Wane simplify the following derivation le us conser a general su cept fonction aeaerun) as sche A, amt, aye complex accommodate second ode ols If his ‘eine cen ne real prof the flowing est is taken, For multiple poles ‘summation over pds i ese ‘We nom pred neve soled foreman aso Frequency Dependent Materials 1s ° ee 10 — 0 ~ Frequency cate)“ ™" Sting fr he ince it rr (8 6) fre esti fnetion fom) we ae a, [bean 32 ° + BL 3 facentaie-nessn Slee aoled woo) Tr evaating (45) we mast separate or es Ineaaing 5) sino ee aves depending on + erreo alr x0)-0 (566) 188 The Fnte Diference Time Domain Method for Blcromagnetics, + fw oeret a Bere aden +3 NEL feo) cofu.s}-a,00.snlgas}esie™'] 6 tay vlan) Finally, because the lower its te stegrals (99) change fo {TO Bt the fine drain fhe incident pulse, a because the sine and Cosine terms become either I or when €=T, we cbt for T <7 ofaak Set] cess) Rlere ala, eo ‘agin eid hat hl ptf (857) an 68 0B ake i he Sect oon nt th oes for end dmb nsesnd oer tule nd a suman over ponders formule pols “\ bot commento the urea vt of (86) shoud be made Depo edrtiot panhe corenonng Yeo tte ee the eal pat oh) cage ic ey Ee Ug atcom tha orion hs the case en he thonmannrs tfc] et goat und inthe celaon cf ‘Rp Bese hie wiper cates mesa ove ao STaltprech zo mpi cress, Te pysl ean for ts {Rin eu een of he exponent se rl eponing ly ‘siete FDTD conoutton inssee.Alae vale be Tepes {hage cough meray overflow) a nda that dapeive {Gewese compo nay be oie tl a es at for at torent nw ore tii pies bese he mtr an ‘Ceo odled wing const sand ter oon ie te mayor ne FUL tongues sersy Race oe ea Pe tation he covaon eg ees oa ma ‘aon f te sets land ont amp meno cle pn iat rte ner cramp consid sn Sonton the ie al ft compton changed such tit both Are moti TON sh te ena To he niente SSresponoly 100) nes tng Th af ep weld re tn os Igo gro eign an overow a compat However ia ths caton he upper fequeney Unit of he PTD res Frequency Depender Materials 19) Y — erwativeo Convoution sev T= Sesceptiity veo Tames"? Brae cpercntery tee Chee nae a a renee ‘ales of permit and conductivity willy seert sal te icity fang eerie comelton The pa hha ina’ atthe eure conn ay or med tobe voted wal ‘etm or sceered dmc ist te els S185) n 88 yan xanga Into cramp ena the cava ofthe so coin indent pls shown in Pure 28 en tas actnwihy8 3 10 andy > 10° ‘tine in 63) The samt av eloped ip 30 alongwith be ascepabiityFuncion. an shows the pede Imray left coneaon proces sive convolution FDTD ection fr» pled plane nave ident on ‘acount fac ieee vine seater Tilo, or 0 ve the smth eosine pave apron in unpltea ataon e ‘Gas pulse wed fer he revi alto, the peak amphi st 100 Vin a he ae wilh pare 1. Aterom tne sp he water fe 200 an 0 ene see th seared eet el pion Ss show niger 3 al 8.5, Conmarig © igus 88a, 160 The Fite Difference Time Domain Method or betrays wm \I vacuum. | \ rater Too Zoo 300 400 540 8700600" B00 1000 Position (Cal Number) [GURE £31. Ses es pe asp ep ie ae ional eee ie transite pls of -100 Vi amplitude tht prop ce hough he water a the spd of Light in vacum, hs pus exact ‘Retnmumencl nj oumel fe inidet plea itchoal tha Fest oh two diferent aorcaches wil be identical teas win he omer pression of he calulations. Te econ cutie eakalatins ae imple san for he a eld cor evn the niet il the vacoum-waer interface is determined ‘Siijaeay. alo, the ested felis equal tothe scatered field at te ‘Mace soma to sutra ficient from total le necesary The Is foe eect oft reueey is ident oat shown in Figure 8.9 and ino epee Adding the aJotennl rer Ue FFD pate sqetion (2.61) carly snc the competion ude Tate comping in wl Fl How (ee hor neo be evant ony forthe duration ofthe ice use SiSlahsy ama numper of me ys relaie theta amber in a se {problem Forel. forthe pls used in Figure 8-30 with = 22 he psod JF eowcnpnds to 8 tine ses wie apical FDTD calvin may vole ESSE hound time sep Ome he nee ple as ated throng any ‘seat pom location, RD is evaluated instead, and it ony equi ication of single exponential function, because the remainge ot te (lation can tc cast dove ow single merical constant. The be we “Sironat ers in also ne be evaluated only rng the duration Mine Tnidon pulse Theor he aout of adatonal computation nt ‘otra met th eo heat 2% least might apes at iat lane, and for stat i nc, and for many applications, ep Cia eaten pls he aan oe seated el rata eighth egh etna ere iypel sastins hore on 2 sl fan fe FOTD csi ied ith dpesive mars. A fr ster tom see coos of fens pnt dof Chapter 13 REFERENCES een cd ee te a cea i a | Eanes Sah SE aeiin IS ak tema Soin emacs HEE Tas nr. Canpe 81, 1a The Finite Difrence Tine Daman Meth fr Electromagnetics, ‘oop Me Op tL, 40,5, Sip ye nent ret ri Hb apes use oon sen a pegs fo chops ‘SURFACE IMPEDANCE, 9.1 INTRODUCTION Series impedance boundary cenitos a enplaye Wy cimine the ‘nveral volume of losy dielectric objecs fom scaring elelatons In sppicatons of FDTD they can be ery inpran bec the alo cimite the ed to se small cella male meen Uy alter wavelngt ‘ending media, throughout he solution volume In ths haste we FDTD ‘implementations of the surface impedance how condition ae presented ‘Ons implementation neces he fequency dependence of the sac pelece boundary contin, while the other iw disperse stare moe lance boundary coon that is applicable over a very large fequeney ‘one (-D) and two dimesions(2:D), Extension nthe dimensions) shouldbe stsighifrward. To simply the derivation we have considered sraignforvard aplication ofthe methods of Chaper 2 Because the set fields can be bine a any time dig a scattered eld FDTD compu tion, atematvely for Feld vy lying suis Inpednee cond ons on could se oll even terrain enn ho FIST pace were computing scatters es, ‘When FDTD is icaiy ypc w analyze cecomagnee tld (LAE) mations wth osy det obec both he Gc extra a at those inside ie objet arecalelted However, orsome epics aly theextrr fick ar of increas Beate he wivelegth nie hese ans 4s uch smaller thn che fee space wavelength, acirtely comparing the era ds may reir a much finer spt rd within the objec han eer i This gral interes se compte eres regu Che material hs elas hgh Ie. the lc dn pte ery | ino he materi, nd ony he els ery close o he arface fect the external ‘sop: iano effects te nor elds onthe elds «an be anoosimated by 2 star rene aye (IBC) Ta und condition clininats the nc fo cust the els neal he ‘ogc. This rears he ber Oc nthe FDI ston pe tly by tinting cel within the omy dice ata sling ager te "be used inthe enero repion witout the ned to dvde the FDTD pei Steen regions mis erent el ies ne steps OF historical interes. stare impstnrefnnary sanions wee i posed by Leonovich inthe 1940s and were rigorously devcloped by Sein 1960 paps. Dung te past $0 year, researcher hve aplicd surface impedance concen the funy dan nee tere te sealering problems. Tune domain surface peas cones receved 16 164 The Fine Difference Tine Domain Method for Blecromaxneics tit atenon unl eel Toche! hae iad fice imped t= cee ae tcp eqn tine Sonat slain, but pesened nied SiR don est Based onthe apo ee ty Toke Fisyand Ture dengan FDTD sre pedance iia that given fer ur id notes ecu evaaton a te reed coal ‘Ts mikes ter plementation ofan FUTE STDC wei My aed Sout hae mpomontt an SIBC inthe POTD mead, which does lode cursive evaluation of he convoin. Howe ce, their mp ‘Ron hs amine savant eave wth given Pete xa he ee te os tr er ue atin method Ue ‘seSoithercerte candy alos tangents chonge On era ‘ear halemesaton dees mt eget he spare ant ot hic Wt mtd given Ro nd on than eeene 6 the exponen fi cet Go not chang with changes erator COM, "These veins re included ths cape. Beene a ete ey pete neti 10 he Cound condiin, even Tough when apie inte te domain ot ely a8 ingen Howeve Be tem warlace ete 30 ey we tated te eae for iplementing an SIBC nthe FDTD method is esac the computor Tescurcerogrements for mole bigs Tian any trae ote le fling dscusin we estima te ‘Sica durton As FDTD i noma applied penetnble obec, asl Ruhl mt be cen ose ie Del me ep a a sree a] ‘en cehoe #, eth cmplx relative permit ofthe material and anda The oeclengt isd he mater and in fee ace, respectively. Te cm lex parton) for ky ie > ae @ shoei he aan Feeney The comple aie peritivity is dtr nin wing (8.2) 8 03) Sujace mpecance 108 {he mae is» good concn over all fequenis finest then the sonstitve parameter say the condion ge 04) ‘Therefore, @, can be ppronimat San, os) lon espace and perhaps eset contr exit then spline he SIBC woul alow uso deine the el dens ase on he fe pace 96 ‘Thus, by comparing 9.1) apd 96) we see tha the cell dension would be vacesed a proption 1 ,| By applying the SIBC, This would allow fet luge cls the FDTD vn iy desea sy regurement by 3 fcr of Wel hese the dimension fhe EDIT compton (21.02. oF 823, ‘There would lo be considerable sings in computation tne, Since the ‘il cmenstons are ager he tne te cap ne made larger It we Hue AE so mat ne tne sep se 1s increase in proportion 4 Ima a de ive eg be response wile the same wheter he SIBC is nied then the va umber ‘tine ses wile rduce in pogrton the cress nt Snce he ‘Snipa ti necessary to aponmately proportions! ome nM ber of FOTD cel ties the tal nike af tne dep (re decom in Seton 3.8), a estimate ofthe coopuaonsl savings factor S by which he ‘scuaon at he Coutts 166 The Fine Diference Tie Domain Method for Elcromayneis “Thc ete nels he incensed ine which may Be nesesay 1 36 cyte) Chapter 10 185 SUBCELLULAR EXTENSIONS 10.1 INTRODUCTION The computational cost of the FDTD technique scales directly with the number of cells. Often the object being analyzed has important structural features, thin wires, or narrow slots (for example) that are very small in at least ‘one dimension as compared to the main body of the scattering or coupling object. Large cells that allow the main body to be accurately rendered are inadequate when it comes to the “small” structure features. Reducing the cell size throughout the FDTD computational space is one method for dealing with this situation, but it is computationally expensive, and a method that may not even be practical if computer resources are inadequate. Two ways around this problem are discussed here. The first approach uses large FDTD cells throughout the computation space, but approximates the small geometry elements by modifying the equations for the large cells that contain them. For example, a surface impedance like that considered in the previous chapter, may be used to include material layers thinner than the FDTD cells. Another variation involves special equations for calculating the fields in the vicinity of wires thinner than the FDTD cell size. Effects of lumped circuit elements which are contained within one FDTD cell may also be included by modifying the field equations for that cell. Development of this approach often involves application of Maxwell's equations in integral rather than differential form, but the finite difference equations can be obtained easily from the integral form of Maxwell's equations, and they are more easily modified to include geometry variations within an FDTD cell. The second approach is a replacement of the region about the structured feature of interest with a finer grid, what has been referred to as the expansion technique. This approach is effective in that a finer grid is only employed where needed, The effort lies in combining in a physically reasonable way the two geometries: the first or coarse model with the second or finer and spatially smaller model. The discussion in Chapter 6 of human body modeling utilizes this approach. In this chapter the expansion technique is presented from the standpoint of modeling interior coupling through a relatively small aperture into an aircraft, 10.2 INTEGRATION CONTOURS In modeling structures that are small as compared to the FDTD cell size (thin wires, for example), a useful approach is to apply the integral form of Maxwell’s equations in the derivation of the FDTD equations rather than the differential form. The integrations are carried out so as to include the sub-cell 180 tne tne Dugerence tome Doman Mead for Elcom Hz(1.J,K) _| eta ) / /® (1+1,,K) Ey(1,K) Ex (I,K) omer and Hl effets, and he coresponding FDTD update equation i Then obtaied nth secton he undaentl approach given and nthe neXt ‘ction applicts hin wire, The connpt applying nepaionrorre in FDTD was developed ty Taflove and Umashanka References 1 and 2 Tolls the approch corsiger aa He component of magne elt and the stirling levine fil component that ae shown in Fp 1-1 The tual Yee ntton sedan thei locations re separated by x,y, and ‘he usual We now app the Maxwell integral equaon con tothe fe in Figue 10-1, the inepral contour being a square with sides Fusing tothe or ne el components Wie asarptce a {he kde ar ifr long coh ie ofthe comur, we obtain the rl [EEL A). )-E (Ry =n danas 102) fe now vite trough hy cay and earange ems lily, teres ‘te sua FDTD upd equation BALM) (LIHLA) , EARNER) See tay Ez(1W,K) __[ a0 / / Hy (14K) Hy(F1,,K) Hx (11d-1,K) where we ecogiz thw ems on he ight hands asthe finite ifeence form af the ssl operation, sch nhs sur derston Regan Chap 2. It we apply the oer Morwell neal equation {Hae S 1B Be fob. B aay othe feds in Figure 102i similar fasion we obtain (809-10), (408), 0-1 se Ze, (L.K)anay +08, (.L.KIbxay 03) ® which as expected, reduces 0 the ul FDTD eguaton 2.8). PEALEN) 96.1.4) BAQJ-LK)-H, ay ‘The point of this exercise i to sow that we can obtain the FDTD diference equations by using eter he fleet orth nr! forms of be Maxell ‘oon ale othe Feld pstons in the Va

You might also like