You are on page 1of 157

INFORMATION TO USERS

This reproduction was made from a copy of a document sent to us for microfilming.
While the most advanced technology has been used to photograph and reproduce
this document, the quality of the reproduction is heavily dependent upon the
quality of the material submitted.
The following explanation of techniques is provided to help clanfy markings or
notations which may appear on this reproduction.
1 The sign or "target" for pages apparently lacking from the document
photographed is "Missing Page(s)" If it was possible to obtain the missing
page(s) or section, they are spliced into the film along with adjacent pages. This
may have necessitated cutting through an image and duplicating adjacent pages
to assure complete continuity
2. When an image on the film is obliterated with a round black mark, it is an
indication of either blurred copy because of movement during exposure,
duplicate copy, or copynghted materials that should not have been filmed For
blurred pages, a good image of the page can be found in the adjacent frame If
copyrighted materials were deleted, a target note will appear listing the pages in
the adjacent frame
3 When a map, drawing or chart, etc., is part of the material bemg photographed,
a definite method of "sectioning" the material has been followed. It is
customary to begin filming at the upper left hand corner of a large sheet and to
continue from left to right in equal sections with small overlaps. If necessary,
sectionmg is continued agam—beginning below the first row and continuing on
until complete
4. For illustrations that cannot be satisfactorily reproduced by xerographic
means, photographic punts can be purchased at additional cost and inserted
into your xerographic copy These prints are available upon request from the
Dissertations Customer Services Department.
5 Some pages in any document may have indistinct print In all cases the best
available copy has been filmed.

University
Microfilms
International
300 N. Zeeb Road
Ann Arbor, Ml 48106

8310019

Trinh,TrangNha

ANALYSIS OF THE SUSPENDED H-WAVEGUIDE AND OTHER RELATED
DIELECTRIC WAVEGUIDE STRUCTURES

University of Illinois at Urbana-Champaign

University
Microfilms
I l i t e r II a t J 011 a l 300 N. Zeeb Road, Ann Arbor, MI 48106

PH.D. 1983

.

. 1983 Urbana. 1978 M.S.ANALYSIS OF THE SUSPENDED H-WAVEGUIDE AND OTHER RELATED DIELECTRIC WAVEGUIDE STRUCTURES BY TRANG NHA TRINH B. Illinois .S.. 1981 THESIS Submitted in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Electrical Engineering in the Graduate College of the University of Illinois at Urbana-Champaign. Wilkes College. University of Illinois.

. \AAMT* J .Director of TPhesis Research Head of Department Committee on Final Examination! $ l^X-eO^- f°.lJ.j^i^S* M ^ VAJ<- t Required for doctor's degree but not for master's Chairman J .UNIVERSITY OF ILLINOIS AT URBANA-CHAMPAIGN THE GRADUATE COLLEGE November 1982 W E HEREBY RECOMMEND T H A T T H E T H E S I S BY TRANG NHA TRINH ENTITLED ANALYSIS OF THE SUSPENDED H-WAVEGUIDE AND OTHER RELATED DIELECTRIC WAVEGUIDE STRUCTURES BE ACCEPTED IN PARTIAL F U L F I L L M E N T O F T H E REQUIREMENTS FOR T H E DEGREE O F DOCTOR OF PHILOSOPHY ) 2 .

D. N. to Professors S. His constructive criticisms and suggestions were very helpful and are greatly appreciated. to Professor J. for his invaluable guidance. and Becky Lazaro for typing the manuscript. I would also like to thank the members of the millimeter-wave group for their assistance.H. and to Dr. support and encouragement throughout the process of this work. Professor Raj Mittra. Special thanks are extended to Professor P. Lee and C. Liu for serving on the thesis committee.C.iii ACKNOWLEDGEMENTS The author wishes to express his deep gratitude to his advisor. Dyson for his technical assistance.W. . Deo for providing the much needed dielectric materials. Klock for his interest and helpful discussions. W.

3 3.2 Effective Permittivity and Permeability Approximations NUMERICAL AND EXPERIMENTAL RESULTS 3. 19 2..1 Cylindrical Dielectric Disc Resonators 102 5..1 Metal Loss 88 4.1 Suspended H-Waveguide 6 2.2 3.4 22 31 Normalized Propagation Constant of Single and Coupled Suspended H-Waveguides 31 Normalized Propagation Constant of Other Dielectric Waveguide Structures 50 Field Distribution of a Suspended H-Waveguide and Other Dielectric Structures of Rectangular Cross-Section 56 Conclusions 79 LOSS CHARACTERISTICS OF A SUSPENDED H-WAVEGUIDE 88 4.2 Dielectric Loss 93 APPLICATIONS 101 5.2 Dielectric Ring Resonators 106 5..iv TABLE OF CONTENTS Page INTRODUCTION 1 METHODS OF ANALYSIS 6 2.1.4 Horn Dielectric-Filled Trough Leaky-Wave Antennas 122 .1.3 Suspended H-Waveguide Leaky-Wave Antennas 114 5.2 Related Planar Dielectric Waveguides .1 3.1 Mode Matching 6 2.

V Page 6.+h 134 COUPLING INTEGRAL TERMS 137 REFERENCES 140 VITA 147 . APPENDIX B. 133 MATCHING OF TANGENTIAL FIELD COMPONENTS AT y . CONCLUSION APPENDIX A.

The microstrip lines also suffer high metal loss due to the high concentration of the current density on the metal surfaces and near the edges of the metal strips. H-guide [5]. such as image guide [3]. INTRODUCTION The renewed interest in the millimeter-wave spectrum is due primarily to the recent advances in the development of solid state devices up to 300 GHz. the ex- treme miniaturization of the center metal strip or the slot (in a slot-line) makes fabrication very difficult.[2]. Extremely tight surface tolerance requirements of metal waveguide components at millimeter-wave range contribute to the high cost of conventional systems.1 1. insular guide [4]. Attempts have been made to utilize microwave technology in the millimeter-wave frequency range but only with limited successes. However. in the upper millimeter-wave bands. Several types of dielectric waveguide structures are currently available as the guiding medium for millimeter-wave applications. Microstrips and other forms of printed circuits have been used successfully at the lower millimeter-wave range [1]. The dielectric sub- strate thickness must be very small to avoid higher-order modes. and trapped . The conventional metal waveguides become very lossy and more difficult to fabricate as the wavelengths become shorter. This has led to the investigation of alternate forms of low-los3 low-cost dielectric-based waveguide designs suitable for integrated circuit applications at the millimeterwave frequency range.

£.. This suspended H-waveguide is similar in many ways to other waveguide structures. these two insulating dielectric layers are metal-cladded dielectric sheets which are commercially available. For this same reason.. Ferrite applications. Compared with other dielectricbased guiding structures mentioned above. and e. the conductor loss is smaller than for the microstrip or image guide. most of the electromagnetic energy is carried in the guiding dielectric core. The metal enclosure also allows plenty of ground for dc biasing in solid state applications. (see Fig.2 image guide [6]. and therefore. It consists of a guiding dielectric core with dielectric constant s„ sandwiched between two low-loss dielectric layers of dielectric constants e. The dielectric arrangement is then shielded to obtain a closed structure.. the suspended H-waveguide enjoys a rigid mechanical support due to the metal enclosure. . Presently. hence. This allows the adjustment of the relative locations of the dielectric core and other components in the circuit without removing the top dielectric layer. In this thesis. In practice. Since E ? > e. the metal surfaces around the suspended H-waveguide become a natural enclosure. Note that the side walls can be removed if the separation between the dielectric core and these walls is large enough.. 1 ) . the radiation loss is probably less than for other waveguide structures due to the closed metal walls. another form of dielectric waveguide structure is presented—the suspended H-waveguide. Also. cross-talk among circuit elements can be eliminated. most of the com- mercial dielectric-base millimeter-wave ICs are shielded.

3 -a H 3 00 <u ai d (U a. tn 3 (U SO •H mwi .

A number of theoretical studies on the propagation properties of certain dielectric structures have been presented in the literature based on the effective permittivity and permeability [4]. accurate theoretical predictions of the field and propagation characteristics must be available. [15] proposed a similar approach using the transverse resonant procedures. However.1 describes the designs of some modified dielectric resonators which can be used in circuits that employ the suspended H-waveguide. due to the unique structural arrangement.[7]-[9]. Like any other oversized shielded waveguide structure. However. certain loss characteristics. higherorder modes may arise.4 such as the phase shifter or field displacement isolator. A rigorous mode-matching technique based on the field expansion in each subregion of the guide cross-section and the subsequent matching of the fields at the dielectric interfaces appeared recently in the literature [10]-[14]. Peng et al. dielectric resonators can be used to eliminate the unwanted modes. Section 5. In order to obtain reliable design information for the designs of both active and passive components derived from these waveguides. . These methods are simple and usually pro- vide reasonably accurate propagation information. and hence. This situation may not be desirable for many applications. they do not provide information on complete field distributions.can readily be integrated since the top and bottom metal plates can be used to support a dc magnetic field. Bonding between dielectric materials may not be needed since the metal enclosure will secure all dielectric layers tightly in place.

can be obtained directly from the formulation of the eigenvalue problem of the suspended H-waveguide. For generality.5 In this thesis. a coupled suspended H-waveguide is analyzed. . a similar mode matching technique is applied to the investigation of the suspended H-waveguide. All relevant charac- teristic equations for the propagation constant of the waveguide structure are given. Some components derived from the suspended H-waveguide are described in Chapter 5. which can then be easily reduced to a single H-waveguide. It will be shown later that the field and pro- pagation characteristics of other dielectric structures. The loss characteristics of this waveguide structure are included in Chapter 4. such as image guide and insular guide.

1. the thickness of the top and bottom layers and their respective dielectric constants e. and a perfect magnetic conducting (pmc) wall implies even mode propagation.1 METHODS OF ANALYSIS Mode Matching 2. The analysis using the mode matching technique can be applied by separating the cross-section into five subregions (I)-(V). the coupled line reduces to that of a single uncoupled structure. respectively. The analysis of the coupled lines can then be reduced to that shown in Figure 2b where a perfect conducting (pec) wall implies odd mode propagation. Suspended H-Waveguide The cross-section of a coupled suspended H-waveguide is shown in Figure 2a. even and odd modes can propagate in the coupled structure. It is well-known that an infinite set of discrete modes is supported by an enclosed metal structure. The locations of the top and bottom conductor walls of the guide are d and 1. The remainder are discrete higherorder modes some of which are propagating and the rest are below cut-off. The width and the height of the dielectric core are 2w and 2h.6 2. 2. .1. For a = b. The fields in each region can be appropriately expanded in terms of regional eigenfunctions which satisfy the boundary conditions at the conductor walls. only a small number of surface waves exist. For generality. Of these.and e.are different. Because of the symmetry of the structure about the x = -a plane.

.W O W 1 \ (pec) Wall b (b) Figure 2. (pec) Wall • A h «i 1 © (5) fpecl jpmcf © g -h -j •t -a ® 2 1 e0 e3 . €2>€1.7 r 2 W . (b) S t r u c t u r e adopted for a n a l y s i s . (a) Cross s e c t i o n of the coupled suspended H-waveguide.-2W iW-M (pec) Wall If €0 *2 2h L.* «—2'a- r €o • . I €3>€0 (pec) Wall Plane of Symmetry (a) .

Assuming that the propagation is in the z-direction.n1= l N' + I n=l C cosk^2'x + D sin k(2)x xn n (xn j Cn cos k(2)x xn + D' sin k(2)x xn n 2 2 sin k< >y / sin k< >h I yn J I yn J (2) cos Ikyn y (2) /cos k v "h I yn J (3a) . respectively. o i e n (lb) where e.joiu v x n. IIe and i n. V x n + V x 7 x H.8 The total fields are derived from the Hertzian potentials II and II. Region I (1 n > rd) x+a) cos ^ ( xn A (1) /sin k (1 >(h-d)l J sin kyn (y-d) I yn J (1) sin kxn (x+a) (2a) -n=0 I sin|k(1)(x+a) xm M I B m=0 m cos ym (1) /cos I ym cos kxm(x-fa) (2b) Region I I n<2> e N' . the field potential eigenfunctions can be expressed as follows (exp(-jk z) suppressed for clarity). which have a single x-directed component [16]. are referred to as longitudinal-section electric (LSE) mode and longitudinal-section magnetic (LSM) mode. e J o h H » juie e. is the relative dielectric constant of each region. The relationships between fields and potentials are: E » V X V X H (la) .

k (y+l) / s i n | k n yn Ly (4a) m (3) (-h+1) c o s k ym ( y + l ) / c o s k^> ym fc<3). ^ ' (x+a) cos k xm (4b) Region IV N' (4) c o s k xn ( x + a ) s r<4> = n-1 n (4) s i n k xn ( x + a ) in[k. I ym J (5b) Region V N' I cos n=l k (5) xn sin k(5)y (x-b) ) f + I' 5 s i n k< n>h| n .> " (x+a) [ xm M m«0 (3) (3) (-h+l) sin .fk l k ' " thi l s i n . i v" i H L ^ r 4 + Hm' s i n fk ( 4 ) h) "m fj:(4). h y / c o s kym J I y™ J cos k s i n | k ( 2 ) yJ | / s i n | k ( 2 ) h { ym ) { ym (3b) Region I I I 3) n< = ^ (x+a) cos k xn N I Kn ' n=0 rO) (3) s i n kxn ( x + a ) x i n |(k< k v 3. y ] • k^yll / Vex . s i n Ikf4*(x+a) (^ xm cos|k(4)(x+a) xm (4) c o s.„'y JSL sin|k<4>hl yn J coslk^ ' y . I ym .l cos k ' h I I y™ J. k ( 14 ) y . + G' n costk (4) h] yn J (5a) M* n.9 r(2) M1 I m«l M' + I m»l 2 E s i n fk< m >x 2 E* s i n k< >x m xm + F cos k ( 2 ) x m xm (2) + F1 cos k x m xm ~(2) 1 u .> COsfk^h) (6a) .

the transverse eigenvalues in regions I and III are given as follows.. J' m cos k<5>h| m ym fell ) '-RM (6b) These potentials have been chosen in such a way that the correspondent fields will satisfy all boundary conditions at the metal walls.4. and k^ ' and £' ' (i = 1. M and M' are infinite. The addition of the denominator terms in these potential functions greatly simplifies the final expressions of the eigenvalue equations obtained from the enforcing of the continuity conditions at the y = h and -h interfaces. To satisfy the boundary conditions at both side walls. . k<3> xm xm mil a + b (m + 2") —rr . odd modes (pec at x = -a) kxn™ .5) are the transverse wavenumbers for the LSM modes.. k x and k y (i = 1. in practice. the upper line in the brackets implies a pec wall at x = -a.1. while the lower line implies a pmc wall.4.3.5) are for the LSE x y modes. N'. one must limit these values to finite numbers for numerical computations. These odd (antisymmetric) and even (symmetric) modes are decoupled from each other and can exist independently in the coupled structure. An exact solution for the propagation constant and the field expressions can be obtained if N. In (2). (3) and (4). In (2)-(6).1.10 M' r(5) I sin xm m=l 5 cos k< >y ym . However. m = 0.2(7) £<».3.2.2.2. k<xn3> nil a + b even mode (pmc at x = -a) (n + I> i+b ' * -0.

(8) and (9) are given as follows: LSM modes LSE modes A kfxn4)(-w+a) k(4)(-w+a) xm B k kfp (w-b) xm W 2k<2>w xn X l k<2>/ek^) xn 2 xn x2 -1 xn°<w-b> 2k(2>v xm -k(2>/k<5> xm xm (10) . which are derived from (1). for -h < y <_ h yields different eigenvalue equations for k and k : pec wall at x = -a cos (A) W± cos (B) sin (W) + x 2 sln (B) cos (W) I (8) + x 5 sin (A) jx 3 cos(B) cos (W) + x 4 sin (B) sin (W)> = 0 pmc wall at x = -a X 5 cos (A) |x 3 cos (B) cos (W) + x 4 sin (B) sin (W)| (9) .11 Furthermore. Matching the tangential field components.4. across the interfaces at x • ±w. separate eigenvalue equations for k ' and k^ ' (i = 2.5) are obtained independently not only for LSE and LSM modes but also for the fields which are even and odd about the y => 0 plane.sin (A) jx x cos (B) sin (W) + x 2 sln (B) cos (W)1 =0 where all parameters in Eqs.

k 2 + k< 5 > 2 + k< 5 >' 0 z xn yn z xm ym (lie) Also. the .k 2 + k ^ 2 + k<X>2 O z x n y n z x m y r a (lla) sJc 2 = k 2 + k< 2 >' + k<2>2 = k 2 + k<2>2 + k^2>2 2 0 z xn yn z xm ym (lib) sJc 2 = k 2 + 4 3 > 2 + k<3>2 .k 2 + k< 4 > 2 + k< 4 >' 0 z xn yn z xm ym (lid) k 2 . £ (2) = £ (5) = ym ym ym (12a) yn £ (12b) ym There exist an infinite number of solutions for the eigenvalue equations (8) and (9). by a simple comparison when applying the continuity conditions.k 2 + k ( 1 > 2 + k ^ . *4 x5 k<2Vk(5> xm xm k<2>/k<4> i" 5 xm xm and .12 LSM modes X3 LSE modes 1 1 s2 k ( 5 ) / k ( 2 ) 2 xn xn X.k 2 + k ( 5 ) 2 + k ( 5 > 2 . it was found that the following relations hold: k (4) yn „ k (2) = k (5) yn yn = k £(4) . k 2 + k<3>2 + k^3>2 (lie) Elk l 2 3 0 z *n yn z xm ym k 2 .k 2 + k< 4 >' + k<4>* . If all solutions are taken into account.

IV and V are obtained. these amplitude coefficients are expressed in terms of those in region V. the following relations hold: ( n n (13a) n C n I' • n f ' D n I n (13b) n D' n I' n = K (G-I) G' n (13c) I' n where K c(2) xn (C-I) A LSM (4) k xn (5) cos kxn (w-b) (14a) . X The imaginary values repre- sent the modes that decay exponentially away from the dielectric core. For a reason which will become clear later. all unknown coefficients of regions II and IV now are represented by those in region V. IV and V form a complete set to represent an exact field distribution in these regions. For each mode.13 expressions for the potentials in regions II. In other words. For an LSM mode. the relations among the unknown coefficients of the field potentials in regions II. k (2) ~(2) and k are always real which implies a trans- verse standing-wave within the dielectric core.whereas k and £ X (i =• 4. Each solution constitutes an eigenmode which represents a particular eigenfunction.5) can either be real or imaginary.

< 2 > xn (2 cos 2kxn>w 4 c< > xn (14d) (4) sin kxn (-w+a) Likewise.K< " > m E' m J' m .14 . for the LSE modes. IV and V are t E m E J m J (15a) . the relationships among the unknown complex coefficients in regions II.(D-I) _ 4) cos k< (-w+a) xn 1 LSM e 2 .<G-I) 2 >w cos k< xn LSM 4) -cos k< (-w+a) (4) cos k (-w+a) << 2 > xn xn 2 xn ' 2 >w sin k< xn sin k^^-w+a) xn (5) cos|k (w-b) 1 xn (14c) (4) (-w+a) sin kxn A LSM = e r sin|2k(2)w 2 < (4) -cos kxn (-w+a) (4) cos kxn (-w+a) .w xn k 2 • cos [ xn [k< >w sin k ^ (-w+a) ) sin(k^(-w+a)) • sin xn (5) cos|k (w-b) 1 xn (14b) -cos [ k ^ (-w+a) [ xn (4) (-w+a) sin kxn .

< xm sin (4) *™ xm (w_b) (16c) cos kxm (-w+a) and (4) cos kxm (-w+a) • sin 2k (2) w J LSE 4) (-w+a) sin k< xm k sin fr.15 m m .^ xm J k+(4) xm • xm 2 >w Kos 2k< xm cos xm (16d) .<<o.(4) *xm (F-J) K m (5) (w-b) sin kxm J LSE (16a) l S xm ) <.K<F-J> m (15b) 1 F' m H m J m m = K< H " J ) m (15c) J' m H' m where (2> ^xm i 1 (E-J) K m A 4 LSE k< > xm sin :(2) xm (4) k.w + a ) • sin £< 2 >w xm cos|k(4)(-w+a) ' xm (4) cos *xm (-w+a) L ir<2) • cos k w xm + i (5) (w-b) sin kxm (16b) 4) -sin k< (-w+a) xm cos xm (H-J) K m LSE 2 >w sin k< xm -sin k (4) (-w+a) xm ]) ( 4) sin k v + xm (-w+a) xm 2 > • cos k< xm 4) . ™ (_w+a> k (2) .

K and L . The x-variation in each expansion function in regions I and III is chosen to be a sine or cosine type harmonic function to simultaneously satisfy the boundary conditions at both side walls.B. the upper line inside the brackets applies to the pec wall at x = -a (odd modes). the tangentials E . and the lower to the pmc wall (even modes).which were derived from Eqs. E and H must be X X z z all taken into consideration.. will be substituted into the remaining four equations. and (A-6). M = M') and also an equal number of LSM modes (i. J and J' become: n n m m .I. a system of homogeneous linear equations with unknown coefficients A. the resulting eight equations are shown in Appendix A. By appropriately multiplying these functions with either sin k ^ (x+a)lor cos k(i)(x+a) . x where k are given in (7). m To reduce the total number of equations. (A-2).3. Using Eqs.i = 1. A . n m n m (A-l). In order to satisfy the boundary conditions for all modes. I'. (l)-(6).e. and integrating over x from -a to b and X using the orthogonality properties of these expansion functions. Assuming an equal number of LSE modes in all regions (i. N = N'). B . These equations shall contain the x-dependence which is different for each region.K. (A-5).. the reduced system 2(N+M) homogeneous linear equations with unknown coefficients I .L.J n m n m n n m and J' is obtained.e. H .16 In (14)-(16). The total fields of the composite structure and the longitud-lnal propagation constant can now be obtained by matching the tangential field components at the plane y = ±h.I'.

E 1l o xp J* = 0 m (17) N weo I n-1 £ k k yn cot(k l yv )cot ( k vv )(h .d) ] V V ) h) R (v) yn ' n .d ) ] P (v) V.yv J n 2 E l 1 + k(1) xu M I \ m=l z xQ m k -k^.a y ) Q(IJ) + k c o t ( k h) U (u) m ym ym m 2 m=l k -k^' . O I' n XV <u) Sm(M) + (J 2 W X X .k< > £lk 1 o x m + J») = 0 m (18) .17 f -1 r N xv In(v) +-r Lk 2 - n=l 10 ^Z] xv Pn(v) + ^ X n> J (1) t a n fk (h-d) [ yy M k(1> m m-1 ym ym m m few ^ ^ " ( h .k 2 1 N ue y n=l k yn tan(k h) R (v) + yn n k O e-k^ 1 yv c o t f k ^ O i .

£<»' k c o t ( k h) U (y) _ ym ym m + ani o m=l X.k<f 3 o xv (J m . 1' 2 Lk . yv yv J k c o t ( k h) R (v) + .."^ LEJ.*2 .3 .0 m Pn(v) I1 n (20) .^ ym m e.k .. 3 yv I yv 2 e...k< 3 >' 3 o xv k<3>tan(lP>(-h+l)) M k m"! ym tan(k h) U (y) + .£<«' i 3 o where V = 0. k 2 .k / 3 o xv N "'• nil <3) N + we o n=l I k yn tan(k M + I k m=l m X)J JO M m S (y) + m yn h) R (v) n *.k 2 3 o m - ^ ' x J' m (19) e. N( 3 ) ' e_k . P n (v) yn yn n .k(3)cot|k(3)cot k(3)(-h+l)| v. 3 3 k< >tan(k< W)lJ yy t yy e.18 k (3) J T (v) + n»l \ P (v) xv 1 1 n n n e .J*) .2 ..N-1 (3) a Q (y) xy m xy -(e<-»«)) e_kv cot.

. k ^ x y x ~(i) and k^ are related to k according to (11).M-1 for pmc wall The upper value in the brackets applies to the pec wall at x = -a. while the lower line applies to the pmc wall.L are found within a constant multin m plicative factor and.1. a complete description of the total field in the guide via the potential functions.. and are shown in Appendix B.. P (v). Since the expansion functions of the potentials in the dielectric core region of the suspended H-waveguide are composed of both symmetric .. xor the y-axis [10]..U (y) are n m the coupling integral terms relating the regions [11].19 { 1. hence. 2... a set of amplitude coefficients A . Related Planar Dielectric Waveguides The system of homogeneous equations derived for the suspended H-wavegulde can directly be applied to other popular rectangular cross-section dielectric waveguide structures as shown in Figure 3. For each k .. The zeroes of the determinant of these 2(N+M) x 2(N+M) homogeneous equations are the longitudinal propagation constants k 's z of the composite waveguide structure. k^ . k ( .. It should be mentioned that the approach to formulate the eigenvalue problem for determining the propagation constant of the suspended H-waveguide described in this section can also be extended to other more complicated structures whose shapes and boundaries are rectilinear and parallel to either th»... In (17)-(20).2...M for pec wall 0..

(b) Image guide. Other dielectric waveguide structures: (a) Dielectric rectangular guide. (d) Partially-filled dielectric . (c) Insular guide.20 *0 Rectangular Guide (^i) (a) •Dielectric Strip (€1) WT^WP^W*. trough waveguide. *1 > € 0 Ground Plane (b) Guiding Layer (62) Dielectric Strip (^j) Ground Plane (c) Dielectric Guide Ground Plane (d) HP-285 Figure 3.

. By rearranging Eqs. the top dielectric layer (E. = 0 M. (The rectangular dielectric guide is equivalent to an image guide with twice the aspect ratio of the dielectric guide.e. and the top metal ground plane is moved further away from the dielectric core to minimize its effect on the propagation characteristics. (17)-(20) into a matrix form. only a portion of (17)-(20) is needed to calculate the propagation constants of the waveguide structures shown in Figures 3a. / 1 \ fl 1 n 1 i1 J „ m M. 3b. and 3d. M„ (21) I' n J* m All coefficients of the matrix in (21) are given m (17)-(20). it is easy to isolate the symmetric and nonsymmetric parts. whose physical ground planes coincide with the y-axis. i. b and d.) To analyze these structures.e. The amplitude coefficients with subscript n represent the LSM modes. A quick observation reveals that the equations that correspond to the coefficients of the submatrix in the first or third quadrant of expression (21) are indeed the eigenvalue equations of the structures shown in Figures 3a. M„ M.21 and antisymmetric functions with respect to the y-axis.) is replaced by free-space. and subscript m the LSE modes.. (22) .. i.

the final results may be in doubt. reasonably accurate results for the propagation constants of many complex structures are obtained [4]. the left side wall will be moved up against the dielectric core. Further discussion on this subject is given in a later section along with numerical results.[15]. A pmc wall at x « -w will generate a family of even modes in a guide whose total width is now 4w. Even though they do not provide field information. where only sufficiently accurate values of k are necessary. For many practical engineering applications. However. and unless enough terms are used.[12]. It can be reduced to that of an uncoupled guide simply by moving the left side wall (at x = -a) to infinity. the "effective" permittivity or permeability approximation or a combination of the two is quite useful. the fields can also be even or odd with respect to their centers. 2.2 Effective Permittivity and Permeability Approximations The mode matching technique described above will yield accurate results for both the field descriptions and the propagation constants if enough terms are taken into account.[17]. the computation can be minimized since the number of LSE and LSM eigenmodes necessary to satisfy the continuity conditions at interfaces of regions increases with increasing ratio (a+b)/2w. By doing this.since in a single guide.22 The analysis has so far been for a coupled structure. This procedure is quite involved. The techniques . in actual calculation. a pec wall at x = -w corresponds to a family of odd modes with respect to this new guide center. Likewise.

independently. 4b) if an appropriate "effective dielectric constant" can be found.e. f e *•< l h r K *. it would be appropriate to reintroduce the fundamental concepts again to facilitate their applications. However. i. The general applications of these approximations can be found in all references cited above. 1 + e (23) r *. Consider an H-wave (LSE-to-z) obliquely incident upon an infinite interface between two dielectric media I and II with relative dielectric constants and permeabilities e_. h 2 y2 . the effective parameter will be used. LSE (H-wave) and LSM (E-wave) modes. For abbreviation. y.23 described here are actually a modification of a one-term approximation of the mode matching technqiue. E.e. e». and y„.ulJ r **\k. They consider two types of modal fields.h e„ 2 V For an H-wave.. i. respectively (see Fig. The H-wave leads to the effective dielectric constant (or permittivity) method while the E-wave is associated with the effective permeability method [9].. 4a). when we refer to both the effective permittivity and the effective permeability and their mean values. The reflection (and transmission) coefficient of this obliquely incident case can be obtained from that of an equivalent "effective" normally incident one (see Fig. it is apparent that the "effective" concept is valid if the effective dielectric constants in (23) are replaced by: .

(b) Equivalent normal incident wave using the effective dielectric constant.24 *i 62 H-i f±2 n /E t (a) T Ht x £2 Ei I Er A 1r f±2 AE+ (b) •*-z Hi Ht HP-148 Figure 4. (a) TE obliquely incident wave. .

5a).) For an E-wave (TM) obliquely incident from medium I into II (see Fig.rL W : i k o 1 = 1. the effective dielectric constant (permittivity) must be used for the E-wave and the effective permeability for the H-wave. assumes that the geometrical discontinuities in a dielectric guide occur (alternatively) in one plane only. = (e. This.and LSM-modes.. in the y-direction. the effective parameter approach assumes that there is no coupling between the LSE. This approximate technique will now be applied to the suspended H-waveguide. the effective permeability will be utilized.2 (25) In short. Unlike the mode-matching technique. k . Since this is an LSM-mode. 6b) so that the characteristic equations for k slab. i.1.y. and the dielectric core region (II) is extended to infinity (see Fig.25 * i where k 1 y± i k 4 k o i .e. in effect..2 (24) is the free-space wavenumber. the equivalent normal incident wave (see Fig. Consider an LSM (TM-to-y) mode propagating in a coupled structure as shown in Figure 6a. The discontinuities in the y-direction are first ignored. can be obtained from this two-dimensional dielectric The characteristic equations for k X have already been derived . and kyi is the phase constant h k sin0. 5b) for the reflection (and transmission) coefficient is obtained if the relative permeability in each medium is replaced by *i •-Wi-.

26

(a)

tf

£1

S2

H-z

x
H," Ej

(b)

A

>

Hr
4

Ht/
A'

:—V

>

»

HP-149

Figure 5.

(a) TM obliquely incident wave.
(b)

Equivalent normal incident using the effective
permeability.

27

(pec) Wall

Mot^i
a

/peel
\pmcj

«o

' i,;

•G).

»'H?>WVi

•••x

(g)

M0»«3

1

1 1 . / .
(Pec'
Wall

• W O W

(a)
,1
' I*"

(Peel
\pmcf

'" ,,""
LI I

^(pec) V

,i'vC\

I

Mo V>2.->
*o iX'-f2.-V-

Mo
«0

J,"b "*>*?* fn

-a

W

-W

(b)
(pec) Wall

^L
CD

Mo^i
m^^m

-h-LLZ

<D

' • • » " ^ . « 2

-

i^_

Mo-«3
(pec) Wall

(0
Figure 6.

(a)

Cross-sectional structure adopted for the
effective parameter approximations.

(b), (c) Models for calculating kx and k ,
respectively.

28

in Eqs. (8) and (9). Once k

is obtained, the "effective" permeability

is defined according to (25).

(Note that the coordinate system in

Fig. 6 is different from that as shown in Figs. 4 and 5.)

In this case,

*
U5 i s given by

V (X)

.(2)
x

" Uo " rn

2

(26)

The dielectric core region Is now replaced by an infinite slab with
*
an effective permeability y, (see Fig. 6c). If tangentials H and E
£*

X

are matched at the planes y = ±h, the characteristic equation for k

Z

is

obtained:
.(2)
k<1>(h-d)
- l k ( l ) cos

2

cos 2k< >h
y

(3)
sin k (-h+1)

.(2)
(1)
(3)
- e3,(3) sin k (h-d) cos|k (-h+l)

1y

y
+ sin 2k (2) h

e, sin k (1) (h-d)

\

Iy

+ - ^
e

2

y

(3)
sin k (-h+1)

iy

k(2)
(3)
fiX /r> cos|k (1) (h-d) cos k (-h+1)

k(1)k(3)
y

(27)

I y

y

where the transverse propagation constants k

, i = 1,2,3, are related

by
k 2 = E 1 k 2 + k< 1 )'
z
1 o
y

e 2 y* 2 (x) k 2 + k<2> = E 3 k 2 + k<3>

(28)

(2) _ _y~ k< 1 >(h-d) .(3) s i n ^ ^ . For LSE X (TE-to-x) mode. Using the effective permittivity approach. hence. k(» & L . k 'can be solved numerically. is described as follows: Fl8. the dis- persion characteristics of the original H-waveguide can then be calculated according to (28). „<« y x and the following r e l a t i o n : k2 .(3) cos I y 3) sin k^ (-h+1)j} sin 2k (2) h|jcos|k(1)(h-d) cos k(3)(-h+l) Iy k (2)2 (3) (D.e*(y) k2 .h + d ) ! sin k (-h+1) k 'k i y y y v }- (31) .k z The characteristic equations for k of these two modes are . the process is reversed.4 2 > 2 = k2 .k<5>2 The LSE (29) and LSWr modes can be obtained in the same manner by considering v LSE y e 2(x) : k — y x k . 6C (27) ^ .(2) sin k (1) (h-d) cos|k(3)(-h+l) 2 cos 2k< >h i y y y . the procedure for obtaining the longitudinal propagation constant k Flg.29 In Eq. (27). 6b H-WMW.. k and.k y z (30) LSMX : k Mo(y) 2 -»• k x .

e. geometrical mean. 3) e 2 > e^e-j The characteristic equations for the propagation constants derived here can be applied to the other guiding structures shown in Figure 3 simply by changing the separation of certain side-wall locations and the relative dielectric constants of certain regions.. Numerical values obtained by the effective parameter approximations and the more rigorous mode matching technique are presented in the next section.30 In summary. the effective parameter approximations are based on three basic assumptions: 1) Only one surface-wave type (either LSE or LSM) can propagate. 2) Each region in the guide can alternatively be treated as a two-dimensional problem. actual value of k z Usually. there is no geometrical discontinuity on one side while investigating the other. the (using mode-matching technique) lies somewhere between those obtained by the effective permeability and permittivity approximations. of these approximate values often produces a more accurate result. i..g. . An average. e.

another mode description must be used. The m and n subscripts are the number of maxima of the dominant electric field in the x. the guided modes in a partially dielectric-filled metal structure are neither pure TE nor TM. the suspended H-waveguide can be considered as an oversized waveguide that can theoretically support few higher-order modes at the upper millimeter-wave frequency range. 3. Any mode £. respectively. unlike the hollow metal waveguide. All calculations were done using the Vax 11/780 computer. y the field modes are described as E if the major transverse electric ran field is parallel to the y-axis and as E if the major transverse mn electric field is parallel to the x-axis. Some experimental results are also presented. This is the cut-off condition of any normal mode in an open dielectric structure. NUMERICAL AND EXPERIMENTAL RESULTS In this chapter. the range of the normalized propagation constant k /k Z is between 1 and v'eT. Therefore.1 Normalized Propagation Constant of Single and Coupled Suspended H-Waveguides In an open dielectric waveguide structure. o with a normalized propagation less than one is a leaky mode.and y-direction. where a large percentage of the . This phenomenon is not applicable to a partially dielectric- filled shielded structure whose propagation starts at k /k zero.31 3. the numerical computations of the propagation constant and field distribution of different waveguide structures are discussed. equal to However. In this thesis. Unlike the fundamental mode. Due to the high dielectric constant of the core and the large size of the metal shield.

= E_. As frequency increases.32 energy propagating in the guide is concentrated in the dielectric core. the top and bottom dielectric layers are the same. and a = b. 1 = d. i. Unless otherwise indicated. and modes in a single suspended H-waveguide. A thin absorbing film can also be placed at the inner side walls to damp out the modes with lateral expansion far beyond the dielectric core that are not permitted for normal propagation [5].e. x E.. these absorbing films should not have any effect on the propagation of the fundamental mode.. most of these higher-order modes can be eliminated by proper launching into and receiving from the guide. the higher-order modes are loosely bounded within the dielectric core causing most of the energy to be concentrated outside the core. £. however. In Figure 7. z o This suggests that the side walls have little or no effect on the propagation characteristics of the fundamental modes if the side-wall separation is large enough. Also shown is the effect of the total eigenmodes (equal number of LSM and LSE modes are used) on the convergence . and decays exponentially away from the core. Unless over-moded operation is desired. k /k remains virtually unchanged if the ratio b/w is larger than 4. many of these higher-order modes become tightly bound to the dielectric core as in the case of an open dielectric waveguide structure.5. For practical means. Because of the large separation between the core and the side walls. only the fundamental modes of two different polarizations will be investigated in great detail in this thesis. Figure 7 shows the influence of the side- v wall location on the propagation constants of the dominant Ei.. in this and many following figures.

--o - .3. o .14.33 1 2. = 9. E. b/w for different numbers of eigenmodes. (a) E ^ mode.4 | l | l | l | I -r " ' 1 . h/d = 0.25.2 O V. Normalized propagation constants vs. (b) E x ^ mode.o— ——°"""*""" 2. N EX - - 2. d/\ = 0.55. e.R (a) " 1 1 1 1 1 1 1 1 1 i Figure 7. w/h = 1. 2.0 / <_ 1.8.— .

For h/d less than 0. For the latter case. To obtain accurate results using the mode-matching approach. computation and the convergence. The For this reason. are very slow. z o In this case. the side-wall separation must be large enough. and many more eigenmodes must be used. such as perturbation techniques [20]. other methods. the field extends far beyond the dielectric core.. both techniques agree well except in the range where h/d is small.[13].may be numerically more efficient. the values of k /k for small h/d are not presented. therefore. the values of k /k of the z o E?.34 of the propagation constant. these line designations are used for the rest of this section. the results are almost indistinguishable. In Figs. 8 and 9. mode obtained by the effective parameter method are less accurate.05. it was found that either the effective permittivity or the effective permeability approximation alone usually provides a better result than their mean value. For N = 6 and N • 10. Unless otherwise indicated. In most cases. . A somewhat similar conclusion was also obtained in investigations of an open dielectric image line by other authors [11]. k /k The values of computed by the mode-matching technique are shown in solid lines whereas those calculated using the mean value of the effective parameter approximations are shown in dotted lines. For the fundamental E' mode. the phase characteristics are plotted as a function of h/d. which is the ratio of the dielectric core height to the total height of the suspended H-guides for different dielectric constants of the core and the normalized guide height.

\ = 0. £.0.0 h/d (b) v x Figure 8.»- •••-.s 1. .3.2C i.o . effective parameter: dashed line. (a) E 2 = '9. 1 = d. 1 .35 ».• h/d (a) l.a- 1...2 .4S 1.15. 1 r a.2- l. 1 a. e. a = b (-*»). = e» = 2.i<J/A0.8 (b) E2 = 3. 0. and E.75. Normalized propagation constants of Ei. h/d. Mode matching: solid o line.o .4 .modes of a single suspended H-waveguide vs. e. w/.

•• C a . h/d.8- 1.8- N 2.0 o. (a) 2w/XQ = 0.0G 2 .78 «*/* .s h/d (bl Figure 9.6 ' 1 '"" 0.0.0. (b) 2w/XQ = 0.0 d/»0.0M 1.4- 1.5- t. e x = ^ 3 = 2 ' 3 « 1 = d. Mode-matching: solid line. Normalized propagation constants of a single suspended H-waveguide vs.a 0.178 1.178 2.0 r 0 —' 0.0.165.0" N J* 1. effective parameter: dashed line.4 1 0. .a t.a «"d (•) 2.36 ».2 0.0 —i—|—i—|—i—r e.2 1 • 0. a = b (-"»).8 1.2- (.4 a.2255.3.

This is expected. therefore. The reverse is true when w/h becomes smaller. 10 and 11 for different values of s„ and the normalized guide height.5. For the E' mode. the ratio of h/d is arbitrarily chosen to be 0. The top and bottom . the effective parameter approximation becomes less reliable. less energy is concentrated in the dielectric core. The complete separation of LSE. For the E. k /k derived by the effective parameter z o approximation approaches the mode matching values. a crossover occurs for the propagation constants of the E^. and the omission of the "corner" regions in deriving the effective parameter technique is no longer y valid. As w/h increases.and LSM-modes is one of the basic assumptions in deriving the effective parameter technique and. For a higher dielectric constant of the core. the waveguide structure closely resembles a dielectric-slab loaded parallel guide whose LSE and LSM modes do not couple. particularly for the lower dielectric constant case and near cut-off.. In these figures.. since for large values of w/h. Figures 12 and 13 are the plots of the normalized propagation constant of a single suspended H-waveguide as a function of the ratio E2/EI for different normalized guide heights.modes.37 The effect of the aspect ratio w/h of the dielectric core on the propagation characteristics of an uncoupled suspended H-waveguide is shown in Figs. When w/h and e„ are small. the results obtained by this technique for large w/h should be closer to the actual values. mode. and E. the normalized propagation constants derived by both the effective parameter and the mode matching methods agree well.

0 I.8" O N.08- 0.0 I.00- 1.0 w/h (bi Figure 10.3 3.0- 2. Normalized propagation constants of a single suspended H-waveguide vs.60 3. e i = e 3 = 2 .0 w/h (a| 2.5 I I I I | I I I I | I I I I | I I I I | I I I I 1.8- 1. a = b (-*»). S 2.I I I I | I I I I j I I I I j I I I I j I I I I 0. Mode matching: solid line.78- +—+ 0. .S 3. (a) E 2 = 9.0.38 3.0- M 1.75. (b) e 2 = 3.0 2. effective parameter: dashed-line. 1 = d.78- o N 1.3 . w/h.8 1.281.8.0 2.S 2. 2.

75 d M 0 . E . (b) s 2 = 3.39 M >. a = b (-**>). effective parameter: dashed line. = e 3 = 2. Normalized propagation constants of a single suspended H-waveguide vs.I7S W/h (b) Figure 11.8. .3. w/h. (a) e 2 = 9.75.0 Cn" 3. N 3.0. Mode-matching: solid line. 1 = d.

Mode-matching: solid line. modes of a single suspended H-waveguide vs. Normalized propagation constants of the funaamental E y . and E?. . a = b (-*»). E 2 / E 1 ' E 1 = £-. 1 = d. effective parameter: dashed line.40 I » I I j I » I I { I I 1 I j I I » 1 I I I I I 2 3 4 5 6 e /e 2 i Figure 12.

Mode matching: solid line. . and E* modes of a single suspended H-waveguide vs.. a = b (-*»). effective parameter: dashed line. e 2 /E 1 «e 1 = e_.S o N Figure 13. 1 = d. Normalized propagation constants of the fundamental Ey.41 2.

many higher-order modes also exist in the guide.3. The insulating top and bottom dielectric layers are copper clad RT/Duroid with an £ = 2. mode. For this reason.. only the results obtained from the effective permittivity approximation are presented as the dotted lines since the effective permeability method inaccurately predicts a high cut-off of the dominant E-j. Figures 14 and 15 show the y x dispersion characteristics of the dominant Ei. modes converge when e« becomes large. In Figure 15. and E. the propagation v x constants of the Ei. and 1 = d. All measurements were done in the E-band. In this range.75).8) and stycast (E = 3. Experimental results are also presented in these figures. A somewhat similar behavior of the normalized constant is obtained by increasing frequency. .e. e. the dielectric core is made of alumina (e = 9. modes as a function of the normalized free-space wavenumber k d..42 dielectric layers are identical in this case. mode is very good. as shown in Figure 12. The frequency in Figure 12 is approximately 1. = e. The agreement between theoretical and experimental results of the E?. The configuration of the core becomes less important since the field strength at the dielectric boundaries is small. i. For large z~lz-\> most electromagnetic energy is tightly bound to the center of the dielectric core. and E.. A combination of a sliding short along the dielectric core and a reverse coupler at the input of the suspended H-waveguide has been used to record the standing wave pattern in the guide [21].4 times higher than that in Figure 13. Energy was launched from a metal waveguide.. The tapered transition from the suspended H-waveguide to the metal waveguide is shown in Figure 16.

the normalized free-space wavenumber.14 h/d . .3 ^ ^ ^ 8— ^ ^f&VS S— ~-~^0r / / X / 1 0— s»» i 0.2. .5 1 1 1 1 2.9. E 1 = E„.1.8.0 i «xp«r 1 m«n t a 1 1 1 | j a. 1 1 I 1 1 1 1 i. Normalized propagation constants of a single suspended H-waveguide vs. Mode matching: solid line. 1 = d.a k 1 J 1. effective parameter: dashed line. a = b (-**>) .5 .43 9 — w/h .0 od Figure 14. .4116 € 2 .8 € .

9€2- 3.0 2. the normalized freed.0 I 1 I I | I I 1 I { 1 I I I | I I I » | I I I I | 1 i II O. solid line. .a- a.75 €.s- 8. effective parameter: . Normalized propagation constants of a single suspended H-waveguide vs.44 2.0 1.S 1.8 0.2.5 3.3 l. £.5 2.0 Figure 15.S- i. e3» 1 Mode matching: dashed line. a = b (-*») space wavenumber.

Dielectric tapered transition from the suspended H-waveguide to the metal waveguide. .45 Side View Metal Waveguide Copper Clad Insular Layers / Metal Ground Plane Tapered Dielectric Core Top View Tapered Dielectric Guide Tapered Insular Layer Dielectric Guide HP-EiT Figure 16.

46
Thus far, only a single guide, which is symmetric both in the
x- and y-directions, has been studied,

Figure 17 shows the normalized
y

x

propagation constant of the fundamental Ei. and E... modes of a single
suspended H-waveguide whose top and bottom dielectric layers are not
identical.

In this case, the thickness of the top layer is varied

while that of the lower dielectric layer remains constant, k /k
X

is

o

plotted as a function of h/d for different dielectric constants s,
(e, = 2.3 in Fig. 17a and E. = 3.75 in Fig. 17b). For small values
of h/d (large separation), the top metal ground plane has virtually no
influence on the propagation characteristics of the guide. In these
figures, the effective parameter approximation accurately predicts
the behavior of the normalized propagation constants and can always
be used to check the existence of a particular mode before using the
mode-matching method.
The dispersion characteristics of a couple suspended H-waveguide
as a function of the normalized free-space wavenumber k d are shown in
Figs. 18 and 19 for different dielectric constants e~ of the core. As
frequency increases, the field strength outside each dielectric core
becomes smaller, and the propagation constants of even and odd modes
approach the value of an uncoupled structure as the influence of one
guide on the other becomesnegligible.

In these figures, subscripts

e and o imply a pmc wall and a pec wall at x = -a, respectively.

For

example, E?.. represents the even E^, mode (pmc wall at x = -a) while
E?,

represents the odd E^. mode (pec wall at x = -a).

script designation applies to the E.,, mode.

The same sub-

In Fig. 18, where the

47

2.78*

2.80-

h/l -

0.S

w/h -

I.a

,2.28N

2.00-

1.70-

I.S00.0

1

0.2

I • I

0.4

0.0

h/d
Ca 3
2.7C-

2.80-

&2- 0 . 8
«1- S.78
=3" * 3

h/l w/h -

0.S
1.2

,2.28-

*
2.00-

1.0

C b 3

Figure 17. Normalized propagation constants of the
y
x
fundamental E'

and E.. 1 modes of a single

suspended H-waveguide vs. h/d.
2.166.

(b) 1/X

= 0.2

(a) 1/X

=

Mode matching:

solid line; effective parameter:

dashed line.

48

2.5Cj- 9.8
E
63- 2.3

h/d - 8.4S
w/h * 1.2
o/w - 1.0

2.0-

o

«

i.s-

1.0-

0.5«
T—i—r
0.0

0.5

T—r T — 1 — r
1.0

T—i—j—1—r T—r
2.0
1.5

y
Figure 18. Normalized propagation constants of even and odd E..., and
E*

modes of a coupled suspended H-waveguide vs. the

normalized free-space wavenumber.

1 = d, b/w •*• °°.

0.5 / / / h/d . and E. 2.1.6 .5 od Figure 19.1 modes of a coupled suspended H-waveguide vs.3.75 C| - e3-2.5— _ / S / / ' X ' E 11e // • 1 i i i i I i i i i1 0. b/w -*• °°.3 1.0 S / 1 1 1 1 1 1 1( 1. Normalized propagation constants of even and odd E^j. the normalized free-space wavenumber.2 a/w .5— MO*"* <-•» E y / +'" .45 w/h .0.0— IM / / E / 11o • EX 8.< ^ / " / 1.0 k i 1 1.0 . 1 = d.49 C^.1.

for an open structure. The . The dispersion curves of some lowest-order modes of an image guide (see Figure 3b) are shown in Figure 20.2 Normalized Propagation Constant of Other Dielectric Waveguide Structures It was mentioned previously that the method of analysis for the suspended H-waveguide could also be used for other planar dielectric structures of rectangular cross-section. 7 and 15 shows that the analysis of a totally shielded structure can be used for an open dielectric structure if the wall separation is large enough.50 dielectric constant of the core is much higher than that in the surrounding medium. Therefore. for clarity. 3. k /k obtained by the effective parameter s o approximation always approaches the mode matching's values at higher frequencies. However. therefore. all modes converge at much higher frequencies. only the dispersion characteristics derived by the mode matching method are presented. the normalized propagation constants of even and v x odd modes of both E'. in the normalized frequency range investigated here. 19). the propagation constant of any surface wave mode must be greater than the free-space wavenumber and less than the intrinsic phase constant of the dielectric. and E-1 modes converge rapidly as frequency increases. the difference between the two techniques is substantial. Information in Figs. (21) . The range of k /-k is between z o 1 and /e~ since. the propagation characteristics of many open dielectric waveguide structures can be obtained. the fields are weakly bound and.. Hence. For a lower dielectric constant of the core (see Fig. by using all or part of Eq.

3O s N 1.1 1.2. d/h -»• °°. solid line. . Normalized propagation constants of the fundamental E^. b/w -*• °>. Mode matching: effective parameter: dashed line.2- 1.51 1.5€ 2 .0- T P 0 kQh Figure 20. and X Zy.4- 1" x x 0)xp4»plmointal r*«ulfco) I") 1 1. the normalized free-space wavenumber. modes of an image guide vs.22 € 1.

leakywave antennas [23]. Another important waveguide structure is the rib dielectric waveguide. backward propagating waves may exist in the structure which can be used for reverse coupling applications. Since this waveguide structure has been extensively studied elsewhere [10]. The image-guide has found many applications in millimeter-wave integrated circuits. The partially loaded trough guides (see Fig. such as resonators [22]. The agreement between theoretical and experimental results is excellent throughout. only a typical dispersion curve is presented here (see Fig.. mode has a high cut-off. 21). isolators [24]. and.mode in the lower frequency range of this guide are similar to the characteristics of empty parallel metal plates whereas the Erj. If the dielectric constant of the dielectric is high enough. The propagation constant of the image line derived from the mode-matching method is obtained by recording the zeros of the determinant of (22). The dispersion characteristics of the E. in many cases. 3d) find applications in dielectric leaky-wave antennas [26]. The field mode descriptions of an image-guide are the same as those in a suspended H-waveguide. [11]. This waveguide is very similar to an insular guide (see Fig. 3c) . and phase shifter [25]. The dotted lines represent effective dielectric constant curves and the solid lines are the modematching values.[27]. can be used to reduce the radiation loss at a bend in an image-guide transmission line [6].52 theoretical results of both mode matching and of the effective parameter approximation are compared with the experimental results which were published by Solbach et al.

0 b/w .sa- 9.. effective parameter: dashed line. Mode- solid line. .5 2.0 1.0 k0h y Figure 21.50€ 2 . a partially loaded trough waveguide.1.53 1.2.47 1. matching: x and E.5 a " b 1.0 2.25 1 I I 1| II I I }II I I | I II Ij 1 I I I | 1 II I 8. Dispersion curves of the fundamental Ei.0 0. modes of d/h •+• °°.75- a.5 3.5 1.25- w/h -1..80- 0.

therefore. 7 and 17. The top metal ground plane and the side walls must be placed far from the dielectric core so that they have a negligible influence on the propagation of the field modes in the dielectric core region. Even the effective permittivity is not accurate in this case as the curves of the E?. If the dielectric constant of the core becomes higher. only the effective permittivity values are presented since the effective permeability wrongly exhibits a very high cut-off for the fundamental mode. 22. To analyze this structure using the same equations derived for the suspended H-waveguide.54 except that the dielectric guiding layer and the substrate are made from the same material. the values of k /k derived by the effective parameter approximation do not represent the true dispersion characteristics of the field modes in a rib dielectric waveguide. Most of the energy concentrates in the dielectric substrate layer. be ignored. The dispersion characteristics of a single dielectric strip are shown in Fig. Again. Consequently. the top layer of the suspended H-waveguide is replaced by free space. This assumption is not valid in a rib dielectric waveguide since most of the energy is actually confined to the substrate region. we have assumed that most energy is concentrated in the dielectric core. . The effect of the side-wall location on the propagation characteristics has been demonstrated in Figs. In deriving the effective parameter approximation. 22 show a substantial difference between the effective permittivity and the mode matching techniques. mode in Fig. The "corner" regions in the substrate could. The strip merely acts as a guide to control the lateral expansion of the field.

55 \ N 3.5 y Figure 22. . Dispersion curves of the fundamental Ei. effective parameter: dashed line. ^ and E ^ modes of a rib dielectric waveguide. Mode-matching: solid line.

however. More terms are necessary since a larger number of slowly varying eigenfunctions in the outer regions are required to satisfy the boundary conditions . 3. an infinite number of these eigenmodes (or terms) must be approximated by a finite series. To accurately represent a true open dielectric guide. Consequently. The mode matching technique can theoretically provide exact field distributions in the dielectric waveguide if an infinite number of eigenmodes are taken into account. a greater number of eigenmodes must be used to maintain the same degree of accuracy. the numerical accuracy can be limited by computer storage and computation time. when the waveguide is operated near cut-off. the effective parameter approximation values will approach the mode matching values as indicated in Figure 17.56 as in an insular guide. provide accurate information on the field distribution of the waveguide structure since the fields in certain regions are totally ignored. In general.3 Field Distribution of a Suspended H-Waveguide and Other Dielectric Structures of Rectangular Cross-Section It is well-known that the effective parameter approximation is adequate for calculations of the propagation constant. Hence. the theoretical metal enclosure must be enlarged to minimize its effect on the propagation and field characteristics of the guide. or a partially open structure such as the suspended H-waveguide when using the mode-matching technique. Due to the limitation of computation facilities.such as an image-guide or an insular guide. the fields extend far away from the dielectric core. It does not.

the modal expansion for the fields consists of ten modes. the x = 0 plane is equivalent to a pmc wall. For these figures. In this case. Unless otherwise indicated. 1 = d and £. the fields decay quickly away from the dielectric core. H and E z to zero at this plana. are equal x Far away from the dielectric core.. The top and the bottom dielectric layers are identical. The effect of the total number of eigenmodes on the field characteristics of a totally open or partially open dielectric waveguide will be discussed in this section. All the electric field components in the guide are normalized to the amplitude of the dominant electric field and all the magnetic field components to the amplitude of the dominant magnetic field. fewer eigenmodes are necessary. at x = 2w. only H y of H . = £„. On the other hand. equal numbers of the LSM and the LSE eigenmodes are used. distribution of H H . necessitating smaller wall separation when applying the modematching technique. e.e. mode.g. The plots x x and E of the E:f.. By comparing the amplitude of each field component at different y-locations. when the dielectric waveguide core has a higher dielectric constant or when the guide is operated far beyond cut-off. one can visualize the lateral distribution of that particular field. therefore. more energy is concentrated in the top and the bottom .57 across the dielectric interfaces with the rapidly varying fields in the core region. mode versus y/h in a single suspended H-waveguide for different x-locations in the guide are shown in Figures 23 to 26. H is very similar to E Since the field and since E is similar to and E are presented as the transverse fields. For an E^f. E . i.

7S Cj-2.4k0d-t.e- -0.0 Figure 23. Normalized calculated field distribution of H^ of the E ^ mode at different x-locations in a single suspended H-waveguide.0 2.a8.0e2-3.35 -a.8- -i.2 l-d N-10 c 3 -2.3 -0.3 d7h-2 w/h-1.2- x"2w -9. .58 0.

35 0.1 T—I—I—1——I—1—I—I——1—I—I—I—|—I—I—I—I—J 0.2UJ 0. . Normalized calculated field distribution of E of y the Ei.3 0. mode at different x-locations in a single suspended H-waveguide.59 0.5 2.3 k0d-1.0- -0.0 1.1 0.0 0.0 y/h Figure 24.3- e3-2.4c 2 -3.5 1.75 *\ -2.

.w— c2-3.60 w.2 l-d N-10 -0.2N >^x-w -0. Normalized f i e l d d i s t r i b u t i o n of H of the E ^ mode at d i f f e r e n t x .35 y/h Figure 25.75 d/h-2 et . 3 w/h-1.3 k0d-1.1x-2w -0.l o c a t i o n s in a s i n g l e suspended H-waveguide.2 .

61

0.2-

k0d-1.35

-0.1
0.0

T—n—r

T—i—i—r

0.5

i—i—i—r

T—i—i—r

1.0
y/h

1.5

Figure 26. Normalized field distribution of E

of the E*. mode

at different x-locations in a single suspended
H-waveguide.

2.0

62
dielectric insultating layers as indicated by an increasing amplitude
of H x for y > h as indicated in Figure 23.
Inspection of Figures 23 to 26 indicates that with ten eigenmodes,
the matching of the field components at the x

3

±w interfaces is

excellent except for the normal electric field component.

(For a

complete description of the waveguide parameters, see Figure 2.)
across the y =» ±h interfaces is more difficult since E

Matching

is a continuous

function of x at y = h , but a discontinuous function of x at y = h~.
The matching of the dominant field component H , however, is excellent
X

throughout.
The lateral field distributions at different y-locations and the
tests of their convergences for different numbers of eigenmodes are
shown in Figures 27 and 28. The general shapes of the distributions
in the insulating layers, e.g., y = 1.4 h, and in the dielectric
core region, e.g., y = 0.5 h, are very similar. This is a clear
justification for the effective parameter technique in assuming the
dielectric core as an infinite slab in deriving the transverse
characteristic equation of k . The plots of the fields for N = 7 and
N = 12 are almost indistinguishable.
of the H

Except at y = ±h, the matching

component at other locations is very good.

Figures 29 to 32 are plots of H , H , E

and E

at the x = 0.5w

plane in a single suspended H-waveguide whose top and bottom dielectric
layers are not identical.

The top layer has a dielectric constant

e, = 3.75 while the lower layer has e, = 2.3

The field components

in this structure are no longer symmetric in the y-direction.

Since

63

0.0y - 0.5h
-0.2-

-0.4-

62-3.75
6
i«2.3

d/h-2
w/h-1.2
l-d
o-b

S3-2.3

-0.8/ /

H

k 0 d-1.35
*

N-12
N-7
N«4

-0.8-

-1.8-

1
TTT M i l
0.0
1.0
0.5

Figure 27.

I I 1

1 I 1 1 I I I I I

1.5

2.0

I 11I

2.S

3.0

Normalized magnetic f i e l d d i s t r i b u t i o n s of the E ^
mode in the t r a n s v e r s e d i r e c t i o n at y = 0.5 h in
a s i n g l e suspended H-waveguide.

64 Figure 28.5 h in a single suspended H-waveguide. Normalized magnetic field distributions of the E* mode in the transverse direction at y = 0. .

8- Figure 29.65 0.5w in a single suspended H-waveguide whose top and bottom dielectric layers are dissimilar. . Normalized field distribution of H x of the E ^ mode in the vertical direction at x = 0.3 w/h-1.75 22-9.2- ej-3.2 d/h-2 l/h-2 o-b -0.0- -0.4- X X -0.8 £3-2.8- -8.

5w in a single suspended H-waveguide whose top and bottom dielectric layers are dissimilar.66 N Figure 30. . Normalized field distribution of H_ of the E^. mode z 11 in the vertical direction at x 0.

75 w/h-t .2- i—i—i—r -2 T—i—i—r -1 i—i—i—r T—i—i—r 0 y/h Figure 31.2 4-2.2e.5w X 0.1 -0.1 x-0. .3 l/h-2 8.5w in a single suspended H-waveguide whose top and bottom dielectric layers are dissimilar.-3.mode in the vertical direction at x = 0.67 0. Normalized field distribution of E of the E?.0 UJ -0.

8- T—i—i—r T—i—I—r -2 T—i—i—r T—i—r—r -i y/h Figure 32. .0* u -0.6-3.68 0.75 N 0. Normalized field distribution of E^ of the E ^ mode in the vertical direction at x = 0.5w in a single suspended H-waveguide whose top and bottom dielectric layers are dissimilar.

The plot of E at y = 0. Because of the symmetry of the waveguide structure at the x ° -a plane (see Figure 2). the fields in Figure 33 decay much faster away from the dielectric core since the dielectric constant of the core in the latter case is much higher. For the symmetric mode (or even mode). when a few more terms are added. The field distributions of a coupled suspended H-waveguide are similar to those for other coupled dielectric waveguide structures such as the insular guides or the inverted strips.. E and H are not presented since their behaviors closely resemble H and E .compared to Figure 27. and therefore. the field maxima are shifted slightly toward the top dielectric layer as clearly shown in Figures 29 and 30. i.5h clearly shows that more than four X eigenmodes are necessary to satisfy the boundary conditions at the x = ±w interfaces. the vertical plane at x = -a is replaced by an equivalent perfect magnetic conducting (pmc) wall.e. x x Except for the minor field component E .69 more energy is concentrated in the regions with higher dielectric constants. whereas for the case of the antisymmetric mode (or odd mode). However. It is interesting to note that. all x field components match very well at the y <=» ±h planes for N • 10. Typical field distributions in the horizontal direction are shown in Figures 33 and 34. more energy is concentrated in the core region. N = 10. As mentioned earlier. the fields are either symmetric or antisymmetric about that plane. respectively. the vertical plane is replaced by an equivalent perfect electric conducting . the matching of this normal electric field component is very good.

. Normalized magnetic f i e l d d i s t r i b u t i o n of the E ^ mode in the h o r i z o n t a l d i r e c t i o n a t y = 0.70 Figure 33.5h in a single suspended H-waveguide whose d i e l e c t r i c insulating l a y e r s are d i s s i m i l a r .

2- X til 0.5 x/w Figure 34.3 y-0.0 1.3- e« -3.0 0.8 e3-2.0- N-4 N«7 -N-10 -0.71 0.1 0.5 1.75 c2-9.5h in a single suspended H-waveguide whose dielectric insulating layers are dissimilar. .0 2.5 2.5h k0d-l.8 0. Normalized field distribution of E x of the E ^ mode in the horizontal direction at y = 0.25 0.1 1 l 1 1 | l 1 1 I | 1 I I l | 1 1 l 1 [ l l l l till 3.

The plots of . However. The analysis of the total waveguide structure is. are plotted for different vertical planes in Figures 38 and 39.5h are shown in Figure 40. The matching of these components is very good even for N = 3. only the symmetric mode Ei. Since the general behavior of the fields in a coupled structure is well-known. The magnetic field plots in the hori- zontal direction at y = 0. reduced to that shown in Figure 2b. However. of the major magnetic component H x The field strength (or E ) is substantial at the pmc y wall as the maxima of the fields are shifted toward the center.5h plane. The investigation of the field distribution in many waveguide structures has been concentrated on the E ^ mode only. the metal side walls must be placed far enough from the dielectric core so that their influence on the propagation constant and the field characteristics is negligible. to obtain an accurate field distribution eight or more eigenmodes must be used. As many as ten eigenmodes are necessary to satisfy the boundary condition of the normal electric field component E faces. is presented here. In computing the propagation constants and the fields of an insular guide. there- fore. Figures 35 to 37 are the normalized distributions of four symbolic field components plotted at the y = 0. the matching process for H X at the x = ±w inter- and H converges more Z quickly. The dispersion characteristics of a coupled suspended H-waveguide have been shown in Figures 18 and 19. The transverse field distributions of a single insular guide (see Figure 3c). which is an open dielectric waveguide structure.72 (pec) wall.

.73 1.5- i—m—i——i—i—FT -2 -1 -rn—r~T i—i—i—r i i ii a x/w Figure 35. Normalized magnetic field distributions of the E:|. mode in a coupled suspended H-waveguide at y = 0.0- -0.5h.5- 0.0- 0..

---•.74 0. Normalized f i e l d d i s t r i b u t i o n of E of the z Eij.1 *».5h.3 i I I I i -2 jT ^N -^-^^V // // / // / / / / / S*y / / / y x\\J^^ £*"" -0.7 0. mode in a coupled suspended H-waveguide at y = 0.2 d/h=2 k0d-l.. ^«.2 1 / M—io 22=3. • _ >^-—' Y i i i 1 1 1 1 i 1 1 1 1 i i *I (i c/w Figure 36.2 y»0. i 3 .X "^ -0.75 N=6 e.25 w/h=1.=2.. «•" * \ E \ ""•^-^ N UJ 0-0 -~ "« \ > \ ^\ v-*^ V ^s \ .5h a/w=1..

5h. Normalized field d i s t r i b u t i o n of Ex of the E ^ l e mode in a coupled suspended H-waveguide a t y = 0.2d/h«2 w/h=l. . X k Q d=l.7 0.2 a/w=l.25 0. -8.2- i i i i -2 i i i r i i i i i i i r 1—I I i -l x/w Figure 37.8- UJ -0.75 0.

1 11 I I 1 I 1 1I 11 1 1 2 3 y/h Figure 38.2- e 1 »t e 2 . 3 -0. .1 ' 6755 / -0. Normalized field distribution of E x of the E ^ mode in the vertical direction for different x-locations in a single insular guide.5 l/h «3 k o 1 .76 0.I 1 I I I 1 I 1 I I I 1 1 1 I I I I 1 -3 -2 -1 8 .1 .2- w/h .4.3 75 e 3 -2.

4- -0.0N-8 x-2w -0. Normalized field distribution of H x of the E ^ mode in the vertical direction for different x-locations in a single insular guide.71-'' w/h«1.8k 0 l-1.2- -0.6.5 l/h«3 -0. .77 0.87SS -'•^ I 1 I I 1 | I 1 I 1 | I I I I | I 1 11 -3 -2 -1 0 I 1 I I | 1 I M | 1 I 1 1 2 3 y/h Figure 39.

5h.78 Figure 40. . Magnetic field distributions of the E ^ mode in a single insular guide in the horizontal direction at y = 0.

except for the normal electric field component. At the top and bottom metal ground planes (y « d. the longitudinal fields E and H Unlike the of the E?.E A X Z Z as the x XX major transverse electric field component. respectively. mode is also significant and is primarily due to this magnetic field component.. . mode. of the E«x n i mode. the fields at x = w + and x = w v always coincide. are not plotted. H z ponent as shown in Figure 47. at these locations. E ^ mode.4 Conclusions The propagation and field characteristics for two fundamental modes that have polarizations orthogonal to each other in many dielectric waveguide structures have been studied. For this reason. In z becomes the dominant magnetic field com- fact.5h. Figures 45 and 46 are the plots of the normalized electric and magnetic field components in the horizontal direction at y = 0. The matching at the y = 0.5h plane for N = 8 is excellent for all components. E and H are similar to H y y x and E in this guide and. As in the Ef. therefore.79 and H . 3. the tangential H is still very significant. the metal loss for the E. which has E E . mode are very substantial and are comparable to the corresponding major transverse fields (see Figure 45 to 46). which is the plot of the normalized magnetic field distributions at y =» d. The matching at x = w is very good since.H . The method of analysis derived for the suspended H-waveguide is general and can be . the matching at the horizontal interfaces y = ih is more difficult.1 ) . in a single suspended Hwaveguide are shown in Figures 41 to 44.

s^ = Eg.3 d/h-2 8.0 1.152 u 0.80 1.4- _xf2w **~--^ m X-0 0.«2.0x-wn e 2 -9.6- k0d-1.8- w/h-1.2x=w 0.8 e.2 l-d x-0\ 0. Normalized field distribution of E x of the E ^ mode in a single suspended H-waveguide.5 y/h Figure 41. .5 1.0- i—i—i—i—I—i—i—i—r 0.0 0.0 2.

81 8.28 Figure 42. Normalized field distribution of H x of the mode in a single suspended H-waveguide E^ e^ = £y .

5 1 1 i 1 1 1 ' 1.2- i 0. E.82 1.8- w/h-1.»— e 2 -9. Normalized field distribution of E g of the E ^ mode in a single suspended H-waveguide.. .8 i 1 ' ' 0. 1 1 1.IS2 N-8 - ^l 0.0 i 1 i 1 -"-^N.0 y/h Figure 43.4- 0.2 l-d VX-W* 9. = E.3 d/h-2 0.5 2.8 C.-2.8- k 0 d-1.

e^ = e^.83 N Figure 44. Normalized field distribution of H z of E ^ mode in a single suspended H-waveguide. .

0 I I I II I I 2.0 0. Normalized e l e c t r i c f i e l d components of the E^ mode in the h o r i z o n t a l d i r e c t i o n a t y = 0.84 0-0 | I I I I M i l 1I 1 I M M M 0.5 1.5 3. e i 3 V .0 x/w Figure 45.5h.5 2.0 1.

Normalized magnetic field components of the E ^ mode in the horizontal direction at y = 0.85 Figure 46.5h. .

Normalized magnetic f i e l d d i s t r i b u t i o n s of the E*. .5 I I | I II 1. e^ = E ^ . I I II 0. d/h-2 w/h-1 2 k 0 d-l 152 l-d N=«8 y-d H.5 2.86 e 2 -g 8 e t -2 3 11 H.0 | II 0.0 2.0 I | I I I I 1 I II M M 1. mode a t the top metal plane in a single suspended H-waveguide.5 3.0 x/w Figure 47.

can be used as the initial values for iterating in the mode matching technique. The theoretical and experimental results for many waveguide structures agree well throughout. Even though the effective parameter approximations do not provide very accurate results for the propagation constants. By doing so. For a high dielectric constant or at high normalized frequency. in turn. . the computation time that could be wasted on a random search by the mode-matching technique can be eliminated. both techniques agree well. in some cases they provide very useful information on the whereabouts and the existences of certain modes which.87 applied directly to other single and coupled planar structures of rectangular cross section.

the metal loss of conducting walls and the radiation loss due to bending. the metal loss is proportional to the square root of the operating frequency and the intensities of the tangential magnetic fields at the metal surfaces. only dielectric and metal loss characteristics of a suspended H-waveguide will be studied in this section.[30]-[32]. If the radius of curva- ture is sufficiently large.1 Metal Loss In general. have been previously investigated [22]. will not be discussed here. Recently. this metal loss has been reduced by adding a thin dielectric layer of lower dielectric constant between the dielectric core and the metal ground plane [4]. LOSS CHARACTERISTICS OF A SUSPENDED H-WAVEGUIDE The major sources of loss in most dielectric-based waveguide structures for millimeter-wave applications operating at frequencies beyond cutoff are the dielectric loss.[7]. Metal losses are therefore dominant as there is no separation between the dielectric core and the metal ground plane. since the fields decay rapidly in the . Therefore. 4. supports the E y A conventional image guide that modes has the dominant magnetic field H x parallel to the ground plane. a general technique for determining the radiation loss has been presented by several authors [28].[29] and. The dielectric and metal losses of many dielectric-based waveguide structures.88 4.such as the H-guide or the image guide.

it is assumed that the field configurations in the guide are not affected by the wall conductivity. are given by L = I R s { lHtaJ2dl T = I Re where (33) (34) (E x H ) • z da IH. and P_. .e. (l)-(6). m and (35) (36) Here.. a is the surface conductivity of metal. The field components in all regions are given in Eqs.89 dielectric layer. The attenuation in nepers per unit length can be calculated by the conventional power loss method. P P P. I = |H 1I + |H 1I + |H 1I at the metal surfaces 1 tan1 'x 'y 'z R t toyo \h s 2a m = surface resistance. mode since the parallel magnetic field components also decay rapidly in the insulating layers (see Figures 42 and 44). This principle also applies to the E-. i. Two low-loss dielectric layers are inserted between the dielectric core and the metal surfaces. The increase of dielectric loss due to the addition of the insulating dielectric layers will be discussed later. The proposed suspended H-waveguide utilizes the same principle to minimize the metal loss.. »2> s -^ where PT is the metal loss per unit length and P T is the transmitted power in the guide.

At lower frequencies.2.. the attenuation curves first increase then start to turn downward sharply as frequency increases. In the absence of the dielectric layers. the tangential magnetic fields at the metal planes are significant and higher metal losses occur.e. The surface current on the metal plane is therefore small. the attenuation curves for all other cases in Figure 48 decrease as frequency increases.g. the tangential magnetic fields decay rapidly toward the metal ground plane in the dielectric layers whose dielectric constants are lower than that of the core. As explained previously.75 in Figure 49. where E„ is much lower. h/d = 1. It is evident from these figures that the metal loss is much lower when two dielectric insulating layers are inserted. This is due to the fact that the dominant tangential magnetic field H is constant in the vertical direction throughout the guide. At higher frequencies (or higher dielectric constant of the . This feature can be beneficial when the guide is operated at the millimeter-wave frequency range . the dielectric constant of the core is 9. In Figure 48. The attenuation is roughly proportional to the square root of the frequency. the fields are weakly bound. Consequently. Except for h/d = 1. the attenuation increases sharply. In Figure 49. h/d=0. and the waveguide resembles an empty parallel plate waveguide. e. It is also important that the ratio h/d should not be too small.90 Figures 48 and 49 are the plots of the attenuation in decibels per meter of the E~\ mode in a single suspended H-waveguide whose top and bottom dielectric layers are identical. i. the fields are weakly bound..8 whereas e = 3. For small h/d.

.91 10-3 0. mode in decibels per meter due to the metal loss in a single suspended H-waveguide.0 Figure 48.17xl07. a . Attenuation of the E^.6.

8 aX^ N 1l t . m .92 Cg-3.»2.75 - /^h/d-1 e.0 i i i 2.— ———— \ ca ^. mode in decibels per meter due to the metal loss in a single suspended H-waveguide...17xl07.5 k i 1 1 1 1 2.0 1.6 l-d N-8 y 8. ^ ^ v ^ y fc 8.5 1 1 1 1 1. a = 6. 1 1 1 1 0.95 / £ .3 w/d-9. Attenuation of the E^.0 1 1 1 1 j 0.5 od Figure 49.

2. mode are very similar.6 curves. However.. mode and the E-. the fields become tightly bound to the dielectric core. the metal loss is significantly reduced as clearly indicated by the h/d = 0. .4 and 0.93 core). In general. The attenuation curves in decibels per unit length of the E?. Therefore. it is necessary to study the dielectric loss characteristics of the guide to obtain a complete overall loss figure for a new waveguide structure. the addition of a large volume of lossy dielectric materials may cause a higher dielectric loss. 0. 4. With the addition of the two dielectric insulating layers. the metal loss of the E. However. The intensities of the tangential magnetic fields at the metal planes are small resulting in smaller metal loss as indicated in Figure 49. field computation derived from the mode- matching technique for the power loss approach is very time-consuming. mode in a single suspended H-waveguide are plotted in Figure 50. the metal loss V x characteristics of the E{. with the exception that the loss in the E ~ mode is primarily due to the transverse magnetic field component H whereas the loss in the E-. The power loss technique can also be used to calculate the dielectric loss.. mode is mostly due to the longitudinal field component H . Near cutoff.2 Dielectric Loss The introduction of the dielectric insulating layers in a sus- pended H-waveguide significantly reduces the metal loss due to finite conductivity of the walls. mode is very high but decreases sharply as frequency increases.

00 I | I M 1. 8.2 I II 1.25 I I I II 1.50 I | II 1. Attenuation of the E* mode in decibels per meter due to the metal loss in a single suspended H-waveguide.2 10.17xl07.75 M | M 2.8 C.25 I I 2. a m = 6.50 Figure 50.6 l-d \ 6 03 7) •T.-2.00 I I j M 2.3 w/d-0. .- h/d-1 V E a 18-1 •v.94 l^-rr C 2 -9.

The characteristic equation for the propagation constant of a lossless suspended H-waveguide can be symbolically written as D(£1.£2(1 . In order to calculate o. is the dielectric loss tangent in the respective region. the left-hand side of (38) can be approximated by a Taylor's series in powers of jtti tan 5 ± .95 By introducing a complex dielectric constant to allow for a small loss in the dielectrics. The characteristic equation for the complex propagation constant k = k D(e1.E2. (21). the attenuation constant a.kz. are small.w) = 0 (38) where E^ = s^l . tan 6. Assuming that e.2. becomes .» the parameters in (37) are replaced by their complex counterparts..j tan 63) Tan <5.ja .j™. and a.£3. i = 1.e2»£o»k .j tan 52) (39) s 3 = E 3 ( 1 .j tan 5j_) E2 . associated with lossy dielectrics can be obtained directly by a perturbation technique which is based on the solution of the lossless structure.3.<o) = 0 (37) where D is the determinant of the system of linear and homogeneous equations shown in Eq.

Retaining only the first-ordered terms.D(e 1 . The computations should be terminated before this oscillation occurs. it was observed that the solutions converged quickly after a few iterations then began to oscillate. mode in a single.aj) + j + jotd jg- 3 g I z± tan S± j ^ .kz. Figure 51 shows the attenuation constants due to lossy dielectrics of the E-f.e 3> k . In the calculations of the partial derivatives. Eq.to) i°l i D(Elte2.e3.E3.e2.96 and jOj about the solution of a lossless structure [33]. More energy is stored in the substrate regions which have a much higher . consequently.e 2 . the fields are weakly bound to the low-loss dielectric core. suspended H-waveguide as a function of the normalized free-space wavenumber. It is clear that the dielectric insulating layers cause higher dielectric losses. At lower frequencies. (38) becomes D(e1.u)) =» 0 (40) z The first term of (40) is zero by virtue of (38).kz. I 1-1 a d £ i t a n 16 i3 |s i-w * 3k <41> z All the partial derivatives in (41) are evaluated numerically by the finite-difference technique [34] once the longitudinal propagation constant k z of the lossless structure has been determined using the mode-matching method.

suspended H-waveguide. Attenuation constant in decibels per meter of the E ^ mode in a single.50 2.25 d Figure 51.97 £ \ d8 0. .00 k O 2.

6. which results in lower dielectric loss.1) as Figures 48 and 51 indicated.8) is still lower than the total loss of a conventional H-guide (h/d .which con- sists of the dielectric loss and the metal loss. If a lower-loss copper- cladded dielectric sheet is available. . suspended H-waveguide as a function of the normalized free-space wavenumber. mode in a single. therefore. Figure 52 shows the attenuation constant due to lossy dielectrics in decibels per meter of the E. 6 and 0.6 and 0.. for given waveguide parameters. the dielectric loss curves of the E?. the dielectric loss of the suspended H-waveguide should be much lower. much less if two dielectric layers are inserted since the metal loss of a suspended H-waveguide is very low as compared to that for a conventional Hwaveguide. most of the electromagnetic energy is stored in the core. mode in a suspended H-waveguide are very close to the curve of a conventional H-guide (h/d = 1). 2 and 1 in Figure 51. Since the loss tangents of the substrates are much higher than the loss tangent of the dielectric core in this case..8. higher dielectric loss.. However. As frequency increases. Near cutoff.98 dielectric loss tangent and. For h/d a 0 . In Figures 50 and 52.g. therefore. the total loss of the E ^ mode is lowest when the ratio h/d is about 0. For h/d = 0. the dielectric loss increases monotonically as a function of frequency. the total (dielectric and metal) loss of the suspended H-waveguide (h/d = 0 . most of the dielectric loss is due to the substrates. is. e. the dielectric loss is extremely high but decreases sharply as frequency increases. tan 6-/tan 5„ = 6. The combined total loss.

^Td— ^ h/d. mode in a single.8 k o 1 1 2. suspended H-waveguide.2 2.0 1 2.99 29 8.4 I 1 I 1 j 1..1\ w \ ^h/d-9.2 -.0 t 1.1 1 1.3 C 2 -9.4 d Figure 52. Attenuation constant in decibels per meter of the E.-2.6 l-d <« 1 lan&i-6e-4 tandk-U-4 kani)g-6e-4 A 10 .6 1.8 C3-2.2 i 1. .3 15 w/d-8.

it appears that the dielectric losses dominate the total loss figure.100 In Figures 48 to 52. in practice. becomes lower [35]. . the surface conductivity at the millimeter wavelength metal losses become higher. However. and the The impact of the suspended H-waveguide should be more profound since the control of the metal losses becomes very Important at the millimeter wavelengths.

At the millimeter-wave frequency. insular guide and inverted strip guide. and leaky-wave antennas [42]. such as dielectric resonators [36]. . directional couplers [37]. Also. phase shifters [41].[40].[48]. many of those components have been introduced. Applications. on the other hand. usually consists of a common ground plane or a common substrate on which the guiding structure and associated components are located. For the image guide. a circuit involving many discrete components is usually very expensive and cumbersome. have been published in the literature. discrete components are connected together by many types of connectors to form a complete system. many components can be constructed cheaply and efficiently. By appropriate manipulations of the basic guiding structure. APPLICATIONS i Theenergence of the dielectric-based millimeter-wave integrated circuits requires the development of a totally new class of dielectric components. losses due to connectors are usually high.101 5. A dielectric-based millimeter-wave integrated circuit. suaded. In a conventional microwave circuit.[38]. The use of connectors in the circuit is strongly dis- A prime consideration from the structrual viewpoint in a millimeter-wave circuit is the adaptability of the guiding structure for other circuit components. Other components in which the basic guiding structure has been modified to perform a particular function are described in this section. ferrite isolators [39].

the dielectric disc is locked securely in place and no bonding material is needed. Simultaneous applications of the perfect and imperfect magnetic walls resulted in a better prediction of the resonant frequency [45]. Because of the metal enclosure.102 5. A circular dielectric disc of the same height as of the dielectric cars is sandwiched between two dielectric insulating layers. the structure adopted for analysis with familiar dimensional parameters is depicted as Figure 53b. [46] introduced a more complicated method based on the variational procedure and predicted the resonant frequencies within a 1 percent accuracy of the experimental data. The dielectric constant of the disc is much higher .1 Cylindrical Dielectric Disc Resonators Because of the unique arrangement of the dielectrics in a sus- pended H-wavegulde. The simple analytical technique described by Itoh is used here to analyze the resonant structure shown in Figure 53a. To conform to the waveguide parameters used in previous sections. Konishi et al. Itoh [47] applied the technique originally developed by Marcatili [18] for analyzing the propagation constants of a rectangular dielectric waveguide to accurately predict the resonant frequencies of the TE„. a cylindrical disc resonator can be constructed as described in Figure 53a. The dielectric constant of the disc can be different from that of the guiding dielectric core allowing the use of high-permittivity materials for high-Q applications. The traditional techniques for analyzing the cylindrical dielectric resonators have been based on the open-circuit boundary (magnetic wall) model [44]. ~ mode.

(a) Cylindrical dielectric disc resonator. (b) Structure adopted for analysis.103 Ground Plane X Substrate €1 Substrate €3 Ground Plane Dielectric Resonator £2 (a) Figure 53. .

2 E. and e^ = 1 and the substrate thicknesses t and 1 approach infinity.+ k 2 y pi e 9 k 2 = k<2> + k 2 . N (2) . the structure to be analyzed becomes the resonator In free-space in [44]-[46]. 3 o y pi . 3 and 4. (3) ^ .w A2sin kv 'y + A3co y • = • J (k .104 than that of the dielectric insulating layers. 2 o y pi (43) .k o y po . most of the electromagnetic energy is stored in the disc (region 2) and decays exponentially in regions 1.2 . Much less energy is concentrated in the corner (or the shaded) regions. (42c) sin k^3)(y+1) o cos ^ 3) pi (y+l) where Elk 2 1 o = k (1 >.2 k = k ' . Assuming no circumferential variation.2 . the vector potential that has a single component in the y-direction can be written as sin cos ( >-d> a .r) region 2 ov pi K (k r) region 4 (42b) region 3 (42d) pi y .r) o region 1 (42a) J (k . The corner regions will be ignored entirely..k = k + k . only a very small error is introduced if the matching of the tangential fields is applied to the boundaries of region 2 only. ° P° . When E. Consequently. In a high Q resonator.

are the unknown amplitude coefficients. (42a) and (42d) is for the TE5' mode y while the lower line is for the TM' mode. All field components can be derived from (42) [48].H = E = 0 for the TM mode. (J) y p <J) By matching the tangential fields at r = a.2. i = 1. respectively.3.are the relative dielectric constants of the respective regions: k = oi /E U = free-space wavenumber.(k a) J (k . ° omn omn respectively. -h £ y £ h and at y = ±h. the second index represents the order of the roots in . With no circumferential variation.(k .A3 and A. The first index is zero since there is no circumferen- tial variation. the following coupled characteristic equations are obtained.a) = 0 (44) v po os po ' 1 pi pi 1 po o pi ' and Equation (31) TMy mode: £ 2 k po K o (k po a > J l(kpia) + k pi K l ( k po a ) J o ( k P i a ) = ° <45> and Equation (27) The resonant frequencies of each mode are obtained by solving the respective set of two characteristic equations simultaneously.a) + k .whereas H . .A2. and J and k are the Bessel and the modified o o Hankel function of the second kind of order zero. E y = E p =» H. The upper line in the brackets in Eqs.. TE y mode: k K (k a) J. The mode designation for the TE and the TM modes are TE and TM .105 E-. A. 0 < r < a.K. • 0 for the TE mode.

mode plotted as a function of the geometrical dimensions are shown in Figure 54. most of the energy is concentrated in the disc. The numerical results predicted by the present method agree well with the experimental results published by Konishi et al. as described in Reference [46]. Figure 55 is a special case when the substrate's thickness approaches infinity and e. By comparing the resonant frequency curves in Figures 54 and 55. [46]. In the substrate regions. dominant mode instead of IE. The resonant frequencies of the TE-.i£3 = 1. the removal of the dielectric substrates has a minimal effect on the resonant frequences of the structure. and the third index indicates the number of field maxima of the dominant electric field in the y-direction... ^E011 is tne For example. Since the dielectric constant of the dielectric cylindrical disc is much higher than that of the substrate... the fields decay extremely fast. it is obvious that the removal of the dielectric substrates and the ground planes only causes a small change in the resonant frequencies of the structure. Consequently..2 Dielectric Ring Resonators This section describes two novel dielectric ring resonators that can be efficiently integrated into a dielectric-based MMW-IC to provide .106 the radial direction. 5. Figure 56 shows the curves of the resonant frequencies of the T M 0 1 1 mode in a cylindrical disc resonator as a function of the geometrical dimensions for two different values of the dielectric constants of the disc ( E 2 = 35 and 88).

Resonant frequencies of the TEQ»1 mode. .107 2.0 Figure 54.

-1 TE. 0 c CM Cg-SS 9..5 i—i—i—r 1.0 i—i—r 1. Resonant frequencies of the IE-.3- 0.8 .5* •xp«plm«ntal rwulfcs C461 8. mode. 2.5 h/a Figure 55.108 1.0 T—i—i—j—i—i—r 0.

mode.5 h/a Figure 56. '011 0..6 l-d TM. Resonant frequencies of the TM-.109 2.0 0. .-2.3 h/d-0.0- i—i—r 0.0 i — i — i — i — — i — i — i — r T—j—I—I—I—T t.06.5 2.0 1.

e' 2 ^) 2 21 (46) 1 . . With these assumptions. and (c) the resonators are perfectly matched. formulations: The following assumptions are made in our (a) all guides have identical characteristic impedances. Figures 57 and 58 show the top views of these resonators and the equivalent signal flow-graph schematics. They are particularly useful in a system that employs the suspended H-waveguide as the guiding medium. The metal side walls at ports 3 and 4 are short-circuit terminations. c = transmission and coupling c o e f f i c i e n t s of the d i r e c t i o n a l coupler Y = a + jk . The signal flow graph analysis [49] will be applied to the scattering equations to predict the frequency response of these resonators.tV2^ (47) where t . l'l i .tV ** and 2 -yt. the frequency responses of the resonator structure shown in Figure 57 can be approximated by the flow graph analysis as follows: t(l . c e ' S.110 filtering effects. The shaded regions in Figures 57a and 58a are the waveguide media. complex propagation constant in the half-ring z i = mean length of the half-ring (see Figure 57a). The resonant effect is due to the standing wave created by these terminations. such as the dielectric core of a suspended H-waveguide.. (b) the directional couplers have infinite directivity.

(b) Equivalent signal flow graph schematic. . (a) D i e l e c t r i c ring resonator.in 2W ovwv *0 i»'!S*' «^M © (pec) Wall (J) (a) (b) HP-»» Figure 57.

(a) Dielectric ring resonator. .-. . (b) Equivalent signal flow graph schematic. e I ! N L 9r niZ'jfi%: *o ©1 (pec) Wall (a) © t _p2r-* (b) HF-210 Figure 58.112 (pec) Wall ®?r . V . -r. .

were discussed in Chapter 4. the scattering parameters for this structure can be derived as . conditions of this filter occur when JS 21 | = 0. is comprised of the dielectric loss. For a nonsymmetric coupler. an effective coupling factor must be experimentally determined before the frequency response can be calculated. a. In this case. These losses For a symmetric coupler.k z. (46). The filter shown in Figure 58a is also analyzed by the signal flow graph technique.. which is the attenuation constant in the ring.odd *c and fc.113 kz is approximated by the propagation constant of an uncoupled straight guide.jsin^e""^ 2 (49) where k z. metal loss and radiation loss. (51) is the waveguide wavelength at the resonant frequency. n = 1. the transmission and coupling coefficients are given as [50].[51] t = cos «J> e" j * /2 (48) c .2. is the effective coupling length. it is usually impossible to predict t and c. The resonant From Eq.even . (50) t and c can be calculated either by the mode-matching or the effective parameter techniques. this corresponds to I £ n -&• where X . From Figure 58b...

behaves like a band-pass filter... 59a. a quick glance at the frequency response expressions reveals that S 2 1 has the characteristics of a bandreject filter. |s 2 . which corresponds to i * £ <« + |) Ag n = 1.2.cV 4 ^ (53) where Z is the length of the curved section on each arm of the resonator (see Figure 58a). . these resonators can be used either as a band-pass or as a band-reject filter. (54) For both resonator configurations..3 Suspended H-Waveguide Leaky-Wave Antennas Another possible application of the suspended H-waveguide is the leaky-wave antenna.. With an additional directional coupler at the input port. .c2e"4^ tV 2 ^ 1 .114 cd-fe" 4 ^) '211 'I'll (52) 1 . At resonance.. Since a complete investigation of this antenna is beyond the scope of this thesis.| » 0. while S.. The waveguide structure can be converted into a leaky-wave antenna (or a periodic band-reject filter) by edging periodic narrow slots on the top metal surface along the propagation direction. and can be flush-mounted. 5. The geometry of the leaky-wave antenna is shown in Figure This antenna is frequency scannable simply by varying the operating frequency.

115 (a) 5 — © " h. (a) Leaky-wave antenna. (b) Structure adopted for analysis.X (b) HP-Si* Figure 59.'dl® * • r €l u T-' ^S" - «3 *. .

.. the vector potential and a. the Hertzian potential and the field components of interest are e n E =v I •n<x. For the LSW mode. A dual formulation for an image guide leaky-wave antenna and a bandreject filter has been published recently [52].±2. Because of the periodicity of the structure.2. As a first step to solve this problem. The analysis is based on the spectral domain approach in which the Galerkin's procedure is applied in the Fourier transform domain.116 only the formulation of the eigenvalue problem for finding the complex propagation constants in the structure is presented. the dielectric core is transformed into an infinite slab in the x-direction via the effective parameter technique described in detail in Section 2. section to be analyzed is shown in Figure 59b.±l.. (58) . The cross Only the LSMy mode will be considered here even though the same procedure can be applied to the LSEy mode.1l field components must consist of all space harmonic terms. z d n = 0.ja + -^r.[53].y) 00 z^i n L** ex P(-J k z n z ) (55) ( x y ) eXpHk Z ' 2n > < 56 > 00 H where k e x * i I V x > y ) exp(-jk n=-°° z) (57) is the propagation constant of the n-th space harmonic and is related to that of the fundamental mode by k zn = k .

k 2o n ( 2 ) 2 = k2 x + k2 . h+d<y<h+d+l (61b) $ ( 3 ) = D sinh(n(3)(y-h)) + Ecosh(n ( 3 ) (y-h)) n n n . The appropriate solutions for the above differential equation are: 4>(1) .2.y .117 where k is the propagation constant of the fundamental mode.4 N(60) ' d I y From now on. i * 1. 0<y<h (61d) cosh(n(4)y) where n^ 1 ) 2 = kx2 + k 2zn .B . In the spectral domain. but the attenuation constant is the same for all harmonics.£ l k 2 zn 1o <62> .n(x.y) exp(jkxx) dx *n(k .y) - (59) where < j satisfied Tj n „2 * T n (k ) = 0 x. It should be mentioned that the propagation constant of the n-th harmonic is different from other harmonics. y>h+d+l (61a) — sinh(n(2)(y-h-d)) + C n cosh(n(2)(y-h-d-)).3.A exp(n (1) (y-h-d-D) n n $(2) . the potential as well as the field components are Fourier transformed via <j. and d is the periodic slot spacing. h<y<h+d (61c) ifi(4) = F . a is the attenuation constant of the structure. the symbol<v will be understood as being the Fourier transform of the quantity without it.

e..2^.e_k x zn 3o n n * v In (64).2 •= k + k ... i.. Since E (k ) is z x periodic.118 (3) 2 . Since the boundary conditions with respect to the x-coordinates are defined for all x's in the space domain.. at y = h : J at y = h+d : H ( 4 ) = H(3) x x g(4) = g(3) z z (63a) (63b) H ( 3 ) = H< 2) X (64a) X .(3) m g(2) = m> z gw 'z z at y .2^.2 .e. v> is the effective permeability of the TM-' mode in region 2.2 =k + k . respectively. z L/2 Ez (kx) = E (x) exp(jk x) dx -L/2 Z (66) X The z-dependence is not affected by the Fourier transform and has been excluded to obtain simplified expressions.h+d+1 : (64b) z E (2) = E ( 1 ) = E (k ) z z z X (65a) g ^ ) _ H < 2 ) = j (k ) (65b) X X Z X where E (k ) and J (k ) are the Fourier transforms of the unknown z x' z x' tangential electric field E (x) in the slots and the unknown current z J (x) on the metal portion.u e„k x zn H 2 o (4)2 . we have [54] .2 * .F are the n-th harmonics of the transformed potential in each n n region. A . i.. they will be defined for all k 's in the transform domain.

n ( 2 )e 2 ^UT n sinh E 3 TI ( 2 ) sinh(n(3)d)cosh(n(4)h) (n ( 3 ) d) sinh(n(3)h) + £2e3cosh(r/3)d)cosh(r/4)h) Before applying the Galerkin's procedure..n ( k x> I or E (k ) .z) 4Y . i. the basis functions can be chosen as .k ) n xn' zn' kznn (1) 1 - ln(1)fPl3inh(ri(2)A) + p cosh(n(2)A)] 2 n(2)[cosh(n(2)il) Vj+ sinh(n(2)a) P 2 ] J and p i P2 = e (4) (3) (4) i e 2 n cosh(n d) sinh(n h) + . Finally.k ) E (k .z) exp(jk «5 z) dz (68) Zil A W i s the width of each s l o t and E are the space harmonics of the zn surface slot field E .n x zn zn ' I}s_oo where the s G (k .z) = z x a E (k . For the LSMy mode. the last boundary condition (65b) 2 j XI X is enforced to obtain the following expression for J .m x' ' m=l (70) where a is the unknown constant. the unknown E (k ) is Z X expanded in terms of known basis functions.119 VV«> = n *z.k ) exp(-jk z) r J *• n x zn z. J <kJ = z x' (69) I G %(k .e. Using (67) and the boundary conditions (63a)z (65a). and the basis functions must be chosen m such that in the space domain they are non-zero only in the slots.n x' d e x (67) ^kznz> rW E (k . m z. the unknown harmonic amplitudes can be expressed in terms of space harmonic E ( k ) . M E x(k ..± z.

.m x z x z. the Galerkin's procedure is applied by taking the inner product of J and E for different values of s and sum over z z.M (74) The fact that the left-hand side equals zero comes directly from the Parseval's theorem.2. respectively.kzn' ) exp(-jk ^v J zn z) n™-oo m B l I (72) Jn(kx. J and E are zero in the .m (x.120 E z.m x zn x m m=l n=-»J s = 1.n(kx .4 y „£.. The following matrix equation is obtained.. HI G (k . over a unit cell.z) » cos f(2m-l)n L J X rnQ |2(m-l)n COS | L ^ L ~ 2 -X .m..z) .n x zn z. 9 z (71) W — 2 z From (68).k ) E (k .kzn) Ez. since in the space domain.n E (k .m..z). E is given by z. I amGn (kx . m=1.m.k ) dk } a = 0 n x zn z.2 .m ( k x' z ) ex P<J k z n z > dz cos(kxL/2) sin(k znW/2) ^(2m-l) k zn (kxL)2-((2m-l)n)2 (kznW)2-(2(m-l)n)2 (73) Now. the transform of the unknown current becomes: M Jz (kx. (69) and (70)..k ) E (k .k ) = ± z.z) exp(-jkznz) where E and J are the amplitude coefficients of the n-th space z..m. s all n.s.z) and J (k . |j W ~ 2~ — .2.n n harmonic of E (k .m.n x zn d rW E z.

The range of |n| in these expressions can be as small as 2 [52]. the radiation pattern in the plane that contains the longitudinal axis of the antenna can be computed readily. I n=-oo' W z n * E " z. The radiation pattern in the transverse plane can be approximated by that of a slot antenna.n(kx'kzn) dkx . there appears to be little or no contribution to the expression beyond a certain value of k .l.0 (75) The solution for the complex propagation constant of the fundamental mode can be obtained by a numerical procedure which searches the roots of (74) within some numerical tolerance [52]. the integration need be carried out only along half of the k axis. Since the amplitude ratio of each element in the slot arrays is determined by a. the complex propagation constant is then obtained from the following eigenvalue equation.121 complementary regions. The . It accounts for the radiation loss and is strongly influenced by the size of the slots. The real part of the propagation constant controls the main beam direction while the attenuation constant a determines the beamwidth and aperture efficiency. A nontrivial solution for a exists only if m the determinant of (74) equals zero which is also the eigenvalue equation that determines the phase constant of the dominant space harmonic. and the integration can be truncated accordingly. If only one term of the slot field distribution is taken into account. Since all constituents of (74) and (75) are symmetric with respect to real k . Moreover. according to Reference 52.

In the transverse (or H-) plane. With this arrangement. These perturbations cause the guided-wave energy to radiate off the dielectric guiding structure.122 / angular direction of the leaky beam measured from broadside in the y-z plane is approximated by 6 n =• sin"1(Re(kzn/ko)) (76) This equation can be used as a tool to determine the relationship between the main beam direction and the periodic spacing d of the slot. For antenna applications. Unlike the periodic slots of the antenna described in the previous section. the periodic perturbations in this antenna structure are metallic strips which are painted directly on the dielectric guide. the antenna behaves like a linear array in the longitudinal (or E-) plane which is the plane containing the longitudinal axis of the antenna. the structure's parameters are usually designed such that only the n • -1 space harmonic will radiate. A complete discussion of this antenna structure can be found in Reference [26]. the radiation pattern resembles that of a horn.4 Horn Dielectric-Filled Leaky-Wave Antennas The antenna described in this section is designed for integration with the millimeter-wave circuits that employ the image guide or the trapped image guide as the guiding medium. In this structure. . 5. Each of the strips behaves like an element of a linear array. Some excerpts from that reference are presented here. the dielectric guide with periodic perturbations on top is embedded in a rectangular trough with metal flare attached along each side as shown in Figure 60.

.123 Metal Flare Metal Ground Plane Dielectric Leaky-Wave Antenna -J U3. Horn dielectric-filled trough leaky-wave antenna.4 mm Figure 60.

the radiation pattern is not symmetric about the main beam direction. Experiments show that the rate of energy radiated along the dielectric leaky-wave antenna structure is strongly influenced by the width of the perturbation strips on the top surface of the dielectric guide. due probably to the large mismatch at the first strip. Therefore. if the metal strips are too wide (>0. But equally important. If the variation of the strip width is correct.124 At each strip. the effective aperature is very small. all the surface energy . the sidelobes are very high. For very narrow strip widths. A suitable modification of the geometry in the leaky-wave region of the antenna should be made to obtain a slowly varying leaky rate to increase the conversion efficiency [56]. It is well-known that if the attenuation rate of the traveling wave in the z-direction due to the energy leakage at the leaky-wave region of the antenna is a constant (neper/m). the radiation from each element is so small that a very long antenna has to be constructed to radiate all energy in an effective manner. which implies a larger aperture. the guided energy should be completely radiated by these strips so that there is no energy left at the antenna truncation which would radiate endfire. is not optimum [55]. consequently. a portion of the guided-wave energy will propagate through the entire antenna length. it is desirable that the rate of energy radiated be small. On the other hand. the bulk of radiated energy is produced by the 8 first few strips and. Thus. the energy transfer from the leaky beam to the surface wave.5 \ ) . Also. and vice versa. for large strip widths.

(76). For a broadside array. is the guided wavelength. m < 18 W m (77) 0.±2. The antenna is frequency scannable if |x /X + nXQ/d|<l.015(m-l)) X . X is the free-space wavelength. the real part of the propagation constant of the structure is approximated by that of an unperturbed structure. For periodic strip spacing. which for clarity. the structure's parameters are usually designed so that only the n = -1 space harmonic will radiate.. and n is the index of the space harmonic (0. The width of the m-th strip is found from the following empirical relation: (0. is given again as 8 n = sin _1 (X 0 /X g + nXQ/d) (78) where d is the perturbation spacing between the centers of two adjacent strips.+l.4 X where X . d should be made equal to the guided wavelength X • Figure 62 .15 + 0. the angular direction 6' of the leakywave beam measured from broadside is approximated by Eq.). m > 18 For slight perturbation. and all the energy will be radiated through this harmonic. This can be done by tapering the size of the periodic scatterers [55] as shown in Figure 61. only the widths of these metallic strips were experimentally investigated..125 will convert into a symmetric leaky beam (and vice versa). Since the mechanisms of perturbation and radiation are not clear.. For the antenna applications.

H-plane yz-plarw E-pkme HP-2S0 Figure 61. .126 Dielectric Silvered Strips •*-z xy-plane. Tapering the width of the metallic strips and the coordinate system.

4mm -4 €rs2.127 10 8 • 6 4 THEORY EXPERIMENT d = 2.85 mm 2 0 -2 a x b = 3. .47 -6 d = 2.73 mm -8.4X 1. 72 FREQUENCY (GHz) Figure 62. Computed and measured main beam directions.

With H as the major magnetic field component in the dielectric guide.4 x 1. the metal flares are attached to both sides of the trough to improve the radiation characteristics in this plane. The length of the flare horn was Figure 63 shows the comparison between the measured relative gain of the constructed antenna as a function of the flare angle . Under these guidelines and constraints. The entire angular scanning range can be shifted more positively or negatively by changing the perturbation spacing between the adjacent strips. The trough depth was arbitrarily chosen to be 3. It was observed that as long as the trough was not too shallow. a dielectric waveguide of dimensions 3. the current element on each metal strip is z-oriented. The depth of the trough is such that the propagation constant of the structure can be determined using the dielactric-loaded trough guide model.[58].4 mm was constructed.128 shows the measured and computed main beam directions as a function of the frequency. and the principal aperture field of the metal flares is E z . The dielectric guide was designed to fit snugly against the side walls of the metal trough thus eliminating the need for adhesive materials which were used to bond the dielectric guide to the ground plane. chosen to be 4 cm. the radiation patterns of the structure were almost independent of the trough depth. Assuming the structure is infinitely long in the z-direction allows the metal flares to be modeled as an H-plane sectoral horn [57]. The height of the dielectric guide is then chosen to provide single-mode operation.4 mm to simplify the construction. In the transverse (or H-) plane.

70 . Relative gains of the constructed antenna and an H-plane sectoral horn versus flare angle.129 0 10 20 30 40 50 60 a (degrees) Figure 63.

The halfpower beamwldtha (HPBW) in the E. the radiation patterns in both the E. . The consequent experimental beam angle deviated slightly from the normal direction. 13°. The antenna was designed for broadside radiation. For the strip width distribution chosen according to (77) and the flare angle selected from Figure 63 for maximum gain. The number of the silver strips on top of the dielectric guide was 32. The main beam level of the antenna Is relatively stable throughout the operating frequency range except at around 84 GHz where it falls about 2 dB below the reference level (see Figure 65). This is interesting since the strip width determines the rate of energy radiated off the guide but does not change the angle of radiation significantly. respectively.and H-planes are 4°. The results of this figure have confirmed the assumption that the H-plane sectoral horn is a reasonable model to predict the flare angle for maximum gain in the H-plane. The side- Since there is no endf ire radiation detected.130 and that of an H-plane sectoral horn.and H-planes of the horn dielectric-filled trough leakywave antenna were measured and plotted in Figure 64. it is assumed that most of the radiated energy is produced by the strip elements.and H-planes and was measured to be 26 dB. The overall gain of the antenna is the product of the gains in the E. lobe levels were at least 25 dB below the main lobe.

Radiation patterns in the E.131 Gain*26 dB E-Plane* H -Plane • -90 -45 8' 0 (degrees) 45 90 HP-2B8 Figure 64.and H-planes of the constructed antenna. .

. Main beam level versus frequency.132 00 -J UJ > UJ -I < Ui CO < 2 FREQUENCY (GHz) Figure 65.

CONCLUSIONS In this thesis. The overall lowloss figure and a superior structural layout of this waveguide promise wider applications in the millimeter-wave integrated circuits. A new dielectric waveguide configuration—the suspended H-waveguide—has been Introduced along with a detailed loss analysis.133 6. . A theoretical formulation of the eigenvalue equation based on the spectral domain approach for determining the propagation constant of a periodic slot waveguide array has also been presented. Some essential integrable components. such as dielectric resonators and leaky-wave antennas. the field and propagation characteristics of many dielectric waveguide structures have been analyzed. Both the mode matching and the effective parameter techniques described in this thesis are very general and can be applied directly to many planar dielectric waveguide structures of rectangular cross section. The theoretical and experimental results agree well throughout. have been studied.

m v ym ' m ym nr m=l (A-3) .134 APPENDIX A MATCHING OF TANGENTIAL FIELD COMPONENTS AT y = +h Eight equations resulting from matching the tangential field components E .n M c n . o ym h.IS V1 o . H .xm ' cos|k(1)(x+a) xm m=l ± x )(J + J ^n.m i< mm ^m N (1) I n k kxn (A-2) -sin k (1) (x+a) xn s )) M A + z n=0 n N' I ym B m I U)U C(D kU. and H at the planes y • ±h are X X z z -I v-* n»0 <1>21 Elk 2 1 o °s | k ^ (x+a) j] xn '•i 1 o •(OH xn e. • m-0 ° *» ' (1) cos kxm (x+a) (1) cos kxn (x+a) tan (1) (x+a) sin kxm = y i n=l 8<t>(1)(x) k —Zjr— tt + I ') z 8x n n M' V uu k Y ^ ( x ) -tan(k h) J + cot(k h) J») *•. E .£<»' xm m=0 » n-1 l 1 ° xn J (A-l) n } (x+a) sin kfi xm M' k^>2 Bm .

m xL- (J m + J m> (A-4) N .135 c o s 1| k ( 1 ) ( x + a ) xn (1) I cotfk (h-d) I yn y weoe .n (x) \ Bm - M« • < cot <V> X » " tan(k h v yn n > V + Am-1 tiPj*) k z ~ i ln.n n n (A-5) sin|k^3)(x+a) L = I cos|k^}(x+a) i} $i (x)(J m .1 yn k~ n=0 sinlk^^x+a) xn + M I I*•m-0 cos|k(1)(x+a) xm kk^> z xm -sin|k(1)(x+a) I xm W <i)< I weo e.. (x)(-I + I ' ) T e..J') T h. 2 2 _ . 1 yn T<t> e.m m m 2 (1) e.kxm ' m=l (A-6) . .k ~ xm m=0 *• 3 0 K n (i) <). k 2 .k 1 o .<«' 'xn I n e.k (3 >" 3 o xn I n=0 (3) sin kxn ( x + a ) n=l "> M ^ (x+a) cos kxn f .k £.

are the x-variation components of the Hertzian potentials in respective regions (see Eqs.m I *•nmi m-0 ° ym' cos ^ (xmx + a ) 9*(i)(x) k — ^ (-1 + I') z 3x n n' :(i) ^o k ym O x ) ( t a n ( k y m h ) J m + cot < k y m h > Jm> (A 7) " k cos xn>(x+a) ? we e.h + D l K.5.n^ m n<=l h) I + tan(k h) I') + LV k — S f S yn ' n yn n' .k(3) n =o ° 3 y n sin cos ( k ^ (x+a)] M + y kk m=0 ^ ( . and (5)).136 -sin fk<3)(x+a) N (3) y k k K z xn n=0 n cos ki?(x+a) xn N' r~(3-> VJ.J1) ra nr v(J (A-8) The upper line in the brackets indicates the pec wall at x = -a while the lower line applies for a pmc wall. e. . tan k (-h+l) ym M + I m™U sin £ ( 3 ) (x+a) M xm .. z 3x J ' m=l . i = 2. (4).4. +u'<x) o 1 yn e. is the relative dielectric constant. i = 2. The $ (x) and $} ( x ) .4.k Y CD.5. • cot k^(x+a) (3) L z x m (cot(k v -sin[k^(x+a) N' = y uie e.y wy V> '• . (3).

k< > xm i 8 <») m I1 xm m 1 xm m 2 * (5) I .e 2 ) v*> .k< >] xn *•> • [< . (VyH)/ K ^ > i.(v)' + fe6 29 ko2 .(G-I) R (v) = Kn^ " i .) (v) • +K {<«> <»-» I j M } + I 4 ( v ) n (v) + e ^ c z) Tn(v) = i -1 • c ^ v»> i^ < .k«> l Kn ^ xn fr> I.v*> . + kxm 4 U H J) .( M . I"1 C *' iI 2„(.137 APPENDIX B COUPLING INTEGRAL TERMS 2 Pn (V) .>> = Kf m m ~ E J) a5L(U)m+ K^ o" F J) L(y) + K^ m " / L(y) + L(y) o and LSM modes Ix(v) ck ( 4 ) ( x + a ) —w cos xn -a t sin k ( 4 ) ( x + a ) xn \ ^ i } ( x ) dx .M • <°-" 2 2 I.fko2 .*r > ¥* (5) + kxn I_(v) 8 £(4)' V"> • ( • 2 - K xm m H ' J) V " > + V2 o 2 2 .

b ) J «v^ ( x ) dx w r (4) sin kxn ( x + a ) w -a I6(v) (4) cos kxn ( x + a ) rw (2) 2) sin k x ) i | / ( x ) dx v xn ' v -w fW I?(v) = I8(v) w where f (2) 1 2 ) ^ ( x ) dx cos k ^ ' x (5) sin kxn ( x .(v) = I s i n (k< xn -w b y^ I 5 <v) t (5) 1) cos kxn ( x . (x) dx r v w 2 >xl •r v< « (x) dx I.138 I2<v> = -w (2) cos kxn x (1) ij.\ cos k ( 1 ) ( x + a ) (_ XV ^ 2 ) ( x ) dx .b ) cos | k ^ (x+a) ^ 1} (x) (1) sin Ik xv ( x + a ) ')) s i n ( k ^ (x+a) ^ 2 ) ) i|/ 2 ) (x) dx XV (x) .

.| sin k xm ™<x"b> 'w kv-t/ (x+a) xy dx (1) cos £ (x+a) I xy In these equations. <//1}(x) by ^ ( x ) . i8(y) . . and the lower line to pmc wall (even mode) at x • -a.I„(y) are obtained by replacing k^ ' in I.(v). k xn and k xm are the solutions of the eigenvalue equations (8) and (9). the upper line in { } applies to pec wall (odd mode) at x = -a.g....Ig with k ^ .(y). ... k ( 1 ) and tc(1) are given in (7).139 LSE modes I.. e.K (F " J) are the coupling xv xy n m terms described in Eqs. K* C _ I ) . ..... (14) and (16).

Microwave Theory Tech." in Symp. Griemsman and L. W. MTT-23." IEEE Trans. MTT-23. no. Schneider. pp." IEEE Trans.." 1973 IEEE MTT-S Intl. "Millimeter-wave integrated circuits. MTT-28. Birenbaum. . "A low-loss H-guide for millimeter wavelengths. Itoh and B. Microwave Theory Tech. no. November 1976. Microwave Theory Tech. Itoh. pp. October 1978. 788-794. 10.. Knox. no. 1959. Knox and P. [2] J.M..V. vol. Boulder. no. pp. MTT-23." in Symp." IEEE Trans..O. "New waveguide structures for millimeter-wave and optical integrated circuits. vol.140 REFERENCES [1] M.W. New York: Polytechnic Press. Microwave Theory Tech. Adelseck. pp.M. "Inverted strip dielectric waveguide for millimeterwave integrated circuits. pp.. 10. pp. [6] T. 543-562." IEEE Trans. October 1975. vol. 751-754. June 4-6. McLevige.. 11. Co. 11. vol. Digest. "Dielectric waveguide microwave intregrated circuits An overview. "Trapped image guide for millimeterwave circuits. [5] J. Toulios. "Integrated circuits for the millimeter through optical frequency range. on Submillimeter Waves. Paul and Y-W. MTT-24. 1973. [7] T. [3] R. 806-814.A. vol. pp. 821-827. Itoh and R. November 1976. New York: [4] Polytechnic Press. Mittra. Microwave Theory Tech. T.' IEEE Trans. Chang. Microwave Symp. 1970. 497-516. 1433-1436. [8] R. on Submillimeter Waves. "Millimeter-wave image guide integrated passive devices. December 1980.

1. Microwave Theory Tech. M"fT-29. Microwave Theory Tech. "Effective gradedindex guides for millimeter-wave applications. A." IEEE Trans. Microwaves. Microwave Theory Tech. Mendenez. "Electromagnetic fields and the phase constants of dielectric image lines. 51-55. New York: Academic Press. Opt. ch. R.. . Schlosser and H. Yang and N. March 1979.D. Thesis. Jamnejad. [11] K. University of Illinois. Deo. 843-854." IEE J. [10] W. 36-43. pp. vol. 3. Urbana." IEEE Trans." in Advances in Microwaves. [16] R. pp. vol. "Partially filled waveguides and surface waveguides of rectangular cross section. and Acoust. no.. New York: McGraw-Hill. September 1981. Deo. 1960. Field theory of guided waves.141 [9] R. Y. pp. 1966. April 1978. no. pp. Department of Electrical Engineering. 9. no. P. T. 2. 9. Collin. Unger. vol. Crombach. Oliner. September 1981. [13] N. [14] U. 319-387..G. Wolff. pp. 6. vol. Hou and V.. Part I . MTT-28.E. MTT-29.L. Microwave Theory Tech. [15] S." Ph. Peng and A. MTT-26. " Guidance and leakage properties of a class of open dielectric waveguides..C. Mittra. "Theoretical and experimental study of millimeter-wave integrated circuits. pp. Solback and I. January 1980. 1980. 870-874. "Analysis of single and coupled rectangular dielectric waveguides. Mittra. "Analysis of open dielectric waveguides using mode matching technique and variational methods." IEEE Trans.Mathematical Formulations." IEEE Trans. 4. no. 266-274. no. [12] R. vol.

. 10. [19] J. University of Illinois at Urbana-Champaign. pp. vol. vol." Coordinated Science Laboratory Report. Time harmonic electromagnetic fields. 7. 7. E.L. 647653." Master's Thesis. "Hybrid millimeter-wave integrated circuits. Microwave Theory Tech. [24] B. 1961. R-813. September 1969. Tech. . ch. 1969. Department of Electrical Engineering. 48. Shindo and T.F. [23] R.J. Jacobs. "Dielectric rectangular. no. Tech. June 1980. 7. EC0M-74-0577-F. pp. "Analysis of fin-line at millimeter wavelengths. 48.. Wilson." U." Bell Syst. 2133-2160. Levin and J.E. MTT-28..W. [18] E." Bell Syst. 6. May 1978. Yang. 747-751. Freibergs and K. vol. October 1978.S. "A circular-harmonic computer analysis of rectangular dielectric waveguides.. vol. pp. 1982." IEEE Trans.142 [17] P. [20] R. Army Electronics Command. no. Harrington. "A new method for the analysis of dielectric waveguides for millimeter-wave and optical integrated circuits. waveguide and directional coupler for integrated optics. Horn. "Low-loss rectangular dielectric image line of millimeter-wave integrated circuits. Marcatlli. New York: McGraw-Hill. J. MTT-26. University of Illinois at Urbana-Champaign." IEEE Trans. Goell. Itanani.E. [21] J. no. July 1975.E. "Electronic modulated beam-steerable silicon waveguide array antenna. Research and Development Technical Report No. [22] S. no. J. pp. Kietzer. Klohn. Microwave Theory Tech. 2071-2102. H.A.

143
[25]

H. Jacobs and M.M. Chrepta, "Electronic phase shifter for
millimeter-wave semiconductor dielectric integrated circuits,"
IEEE Trans. Microwave Theory Tech., vol. MTT-22, no. 4, pp.
411-417, April 1974.

[26]

T.N. Trinh, R. Mittra and R.J. Paleta, "Horn image guide leakywave antenna," IEEE Trans. Microwave Theory Tech., vol. MTT-29,
no. 12, pp. 1310-1314, December 1981.

[27]

M. Desai and R. Mittra, "A method for reducing radiation losses
at bends in open dielectric structures." 1980 IEEE MTT-S
Intl. Microwave Symp. Digest, Washington, D.C., pp. 212-213,
May 1980.

[28]

E.A.J. Marcatlli and S.E. Miller, "Improved relations describing
directional control in electromagnetic waveguidance," Bell Syst.
Tech. J., vol. 48, no. 7, pp'. 2161-2188, September 1969.

[29]

D. Marcuse, "Bending loss of the asymmetric slab waveguide,"
Bell Syst. Tech. J., vol. 50, pp. 2551-2563, 1971.

[30]

F.J. Tischer, "Properties of the H-guide at microwave and millimeter-wave regions," Proc. IEE, Suppl. 13, pp. 47-53, 1959.

[31]

V.R. Bui and R.R.J. Gagne, "Dielectric losses in an H-planeloaded rectangular guide," IEEE Trans. Microwave Theory Tech.
(short papers), vol. MTT-20, pp. 621-623, September 1972.

[32]

P.P. Toulios and R.M. Knox, "Rectangular dielectric image lines
for millimeter-wave integrated circuits," Proc. of Western
Electronics Show and Convention, Los Angeles, California, vol.
14, pp. 3/3-1 to 10, August 25-28, 1970.

144
[33]

S.K. Chatterjee and R. Chatterjee, "Dielectric loaded waveguides A review of theoretical solutions," Radio Electron. Eng., vol.
30, pp. 259-288, November 1965.

[34]

M. Abramovitz and I.A. Stegun, Handbook of Mathematical Functions,
U.S. Dept. of Commerce, Nat. Bur. of Stand., 10th ed., 1972.

[35]

F.J. Tischer, "Excess conduction losses at millimeter wavelengths,"
IEEE Trans. Microwave Theory Tech., vol. MTT-20, no. 11,
pp. 853-858, November 1976.

[36]

T. Itanami and S. Shindo, "Channel dropping filter for millimeterwave integrated circuits," IEEE Trans. Microwave Theory Tech.,
vol. MTT-26, no. 10, pp. 759-764.

[37]

K. Solbach, "The calculation and the measurement of the coupling
properties of dielectric image lines of rectangular cross section,"
IEEE Trans. Microwave Theory Tech., vol. MTT-27, no. 1, pp. 54-58,
January 1979.

[38]

T.N. Trinh and R. Mittra, "Coupling characteristics of planar
dielectric waveguides of rectangular cross section," IEEE Trans.
Microwave Theory Tech., vol. MTT-29, no. 9, pp. 875-880,
September 1981.

[39]

V.P. Nanda, "A new form of ferrite device for millimeter-wave
integrated circuits," IEEE Trans. Microwave Theory Tech., vol.
MTT-24, no. 11, pp. 876-879, November 1976.

[40]

K. Araki and T. Itoh, "Analysis of periodic ferrite slab waveguides
by means of improved perturbation method," IEEE Trans. Microwave
Theory Tech., vol. MTT-29, no. 9, pp. 911-916, September 1981.

145
[41]

H. Jacobs and M, Chrepta, "Electronic phase shifter for millimeterwave semiconductor dielectric integrated circuits," IEEE Trans.
Microwave Theory Tech., vol. MTT-22, no. 4, pp. 411-417, April 1974.

[42]

T. Itoh, "Applications of gratings in a dielectric waveguide for
leaky-wave antennas and band-reject filters," IEEE Trans.
Microwave Theory Tech., vol. MTT-25, no. 12, pp. 1134-1138,
December 1977.

[43]

S. Ray, R. Mittra, T.N. Trinh and R.J. Paleta, "Recent developments
in the millimeter wave antennas," in 1981 Proc. Antenna Applications
Symp., Univ. of Illinois, September 1981.

[44]

S.B. Cohn, "Microwave bandpass filters containing high-Q dielectric
resonators," IEEE Trans. Microwave Theory Tech., vol. MTT-16,
pp. 218-299, April 1968.

[45]

Y. Garrault and P. Guillon, "Higher accuracy for the resonance
frequencies of dielectric resonators," Electron. Lett., vol. 12,
no. 18, pp. 475-476, September 1976.

[46]

Y. Konishi, N. Hoshino and Y. Utsumi, "Resonant frequency of a
TE0,,. dielectric resonator," IEEE Trans. Microwave Theory Tech.,
olo
vol. MTT-24, no. 2, pp. 112-114, February 1976.

[47]

T. Itoh and R.S. Rudokas, "New method for computing the resonant
frequencies of dielectric resonators," IEEE Trans. Microwave
Theory Tech., vol. MTT-25, no. 1, pp. 52-54, January 1977.

[48]

R.F. Harrington, Time Harmonic Electromagnetic Fields. New York:
McGraw-Hill, 1961, ch. 5.

20. Collin and F. Ogusu.146 [49] S. ch. New Jersey: Prentice-Hall. [57] W. 1961. Levy. E. Adam.." in Cross-Field Microwave Devices. Topical Meet. "Directional Coupler. Ed. [52] R. 3. pp. [50] R." in Digest. no. 115-209. Zucker. Tamir and H." IEEE Trans. 1. on Integrated Optics-Guided Waves. New York: McGraw-Hill. Microwave Theory Tech. Antenna Propagat. 1966. Tomiyasu and S. 2. ch. vol. "The transvar directional coupler. ch. July 1981. 1. vol. 4. . 1972)." Proc. pp. Antenna Engineering. New York: McGraw-Hill.L. "A spectral domain approach for computing the radiation characteristics of a leaky-wave antenna for millimeter-waves. "Propagation properties of a planar dielectric waveguide with periodic metallic strips. 6. 1968. 652-654. MTT-29. Antenna Theory. Materials and Devices. 16. New York: [51] Academic Press.L. pp. Schelkunoff and H. Weeks.F. vol.. Okress. ch. 1969. NV. AP-29. New Yorl- Academic Press. 1. Fris. Tech. 1953. ch. Arnaud. [55] T. pp. [56] R. 922-926. K. "Unified theory of optical-beam couplers. 16-21.A. January 1981. Mittra and R. 19. vol. Bertoni.B. [58] S." IEEE Trans. no. New York: Wiley. 1969. no.E.T. [54] J. Papers. 41. 1952. "Periodic structures. Kastner. Microwave Theory and Applications." in Advances in Microwaves. Cohn. [53] K. 2. IRE. (Las Vegas. Antennas: Theory and Practice.J. vol.

His publications are: T.G. University of Illinois. Microwave Symp. he has been a research assistant at the Electromagnetics Laboratory.S.J.J. Microwave Theory Tech. R. "Horn image-guide leaky-'. Malherbe and R. in 1978. Mittra. 205-207. degree in electrical engineering from the University of Illinois. vol. His research interests are in millimeter-wave integrated circuits and antennas. May 1980. Trinh and R. He received the B. Vietnam. December 1981. Urbana.. Trinh." IEEE MTT-S Intl.N." Proceedings of the 1981 Antenna Applications Symp. September 1981. MTT-29. S. T.N.'ave antenna." IEEE Trans. T. University of Illinois. Wilkes Barre. Paleta.L." IEEE Trans.D. Trinh and R.N. Trlnh is a student member of the Institute of Electrical and Electronics Engineers and Sigma Xi. "Coupling characteristics of planar dielectric waveguides of rectangular cross-section. on November 28. pp. Trinh.S. "Recent developments in millimeter-wave antennas. Mr. Mittra and R. and the M. MTT-29.N.. 875-880. R. September 1981. degree.A. in 1981. T. Since 1978.147 VITA Trang Nha Trlnh was born in Nha Trang.. Paleta. Ray. where he has been studying toward the Ph. 1133-1136. Microwave Theory Tech. pp. IL. Mittra. J. Digest. 1956. degree in electrical engineering from Wilkes College. "A metal-to-dielectric waveguide transition with applications to millimeter-wave integrated circuits. pp. PA. vol. Mittra. .

Mittra. "Field profile in a single-mode curved dielectric waveguide of rectangular cross-section. . 1137-1140. December 1981.148 T." IEEE Trans. Trinh and R. pp. Microwave Theory Tech. MTT-29.N. vol..