740

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. AP-31, NO. 5 , SEPTEMBER 1983

Considerations for Millimeter Wave Printed Antennas

DAVID M . P O U R , MEMBER, IEEE

Absiruct4alculated data are presented on the performance of printed wave antennas, and cover the range of relative Permittivity from antenna elementson substrates which may be electrically thick,as would 2.55 to 12.8. The followingcharacteristics will be presented in he the case for printed antennas at millimeterwave frequeneies. Prioted Section 11. on polytetrafluoroethylene dipoles microstrip and patch antennas (PTFE), quartz, abd gallium arsenide substrate are considered. Data are 1) General element characteristics. given for resonant length, resoriant resistance, bahdwidth, loss dueto 2) Substrate properties. surface waves, loss due to dielectric heating, and mutual coupling. A h presented is an optimizatidn procedure for maximizing or minimizing 3) Resonant element length. power launched into surface waves from a multielemerit printed antenna 4) Resonant input resistance. array. The data are calculated by a moment method solution.

I. INTRODUCTION
HERE HAS BEEN rapid growth in printed antenna theory and technology during the last decade [i] , [2] . Most of this work was for antennas operating in theUHFto microwave frequencybands(300 MHz to 10 GHz), and characteristics of printedantennassuch aslow-cost, low-profile,conformability, and ease of manufacture were often found outweigh to the electrical disadvantages, such as narrow bandwidth lowand power capacity, for many applications. Currently there is increasing interest in millimeter wave systems and applications, such as aircraft-to-satellite communications and imaging array antennas [3], [4], as well as interest in completemonolithicsystems which combineantennaelements or arrays on the same substrate as the integrated =/IF front-end detector and amplifier circuits. Thus, printed antennas are being seriouslyconsidered for use at frequencies well above 30 GHz, whilepast applications were generally below 30 GHz. In these applications, substrates are often much thicker and have higher dielectric constants than at lower frequencies. It is the purpose ofthis papertopresentperformancedataforand discuss the applications of printedantennas on these types of substrates. As will be seen,the electrical performance ofthese antennas can be severely degraded, due to surface waves or mutual coupling. Bandwidthandinputimpedance are additionalproperties that are strongly affected by substrate thickness, often in a desirable way. Other factors such asdielectric loss and feeding techniques can also be significantly different at millimeter wave frequencies. This paper will consider two particular types printed of antennas: rectangular microstrip(patch)elements,andprinted dipoles. The intrinsic differences between these twoelements and their comparative electrical performances will be discussed. Three types substrate of materials will be considered: polytetrafluoroethylene (PTFE), quartz, and gallium arsenide. These materials represent substrates which may be used for millimeter

5) Bandwidth. 6) Losses due tosurface waves. 7) Losses due to dielectric.

Manuscript received November 18, 1982; revised March 15,1983. T i hs work was supported in partbyGrantsNAG1-163andNAGI-279 beLangley Retween the NationalAeronauticsandSpaceAdministration, search Center, Hampton, VA, and the University of Massachusetts, Amherst, MA. The author is with the Department of Electrical and Computer Engineering, University of Massachusetts, Amherst, MA 01003.

Section 111 will discuss mutual coupling and surface wave effects in an array environment and present a procedure for the optimization of array efficiency (minimization or maximization of e = Prad/(Prad -k Psw), where P r a d is the desired radiated power and Psw is surface wave power). Although the results presented here are for printed dipole or microstrip patch elements, some of the trends and conclusions will apply to other types of printed antenna elements. All of the data presented in this paper were calculated using a moment method solution of a printed rectangular radiating element on a grounded dielectric slab. A detailed description of this method,withcomparisonsofcalculatedand measuredresults forinputandmutual impedances, has alreadyappeared in the literature [5] , so only a brief description will be given here. Important factors for antennas on electrically thick substrates include surface waves and mutual coupling [6]. The lowest order surface wave (TM,) has a zerocutofffrequency,andthus is excitedto some degree even on very thinsubstrates. AS the substrate becomes thicker, more surface wave modes can exist, and more power can be coupled into these waves. Mutual coupling between elements in arrays involves the transfer of power from one element to a nearbyelement via space waves (direct radiation) or by surface waves. Coupling levels greater than roughly 20-30 dB may have a deleteriouseffect on array performance, unless specificallyincludedin the design procedure. (This recently has been done printed for dipole arrays [ 7 ] , where the coupling coefficients were measured.) Thus, it is desirable for the theoretical solution to account for fields exterior to the radiating element, i.e., to account for surface wave power andmutual coupling. The cavity modeland transmissionline model cannot do this. In addition, neither of these models have yet successfully treated the printed dipole element, and neither are valid for antennas on thick substrates. The moment method solutionuses the rigorous dyadic Green's functionforthegrounded dielectric slab,and so includesthe exterior fieldsmaking calculationsfor surface wave excitation and mutual couplingpossible. Because of the general nature of the moment method formulation, printed dipoles as well as probe-fed or microstrip line-fed patches can be handled. Dielectric loss can be easily included by using a complex permittivity in the solution. The price paid for this versatility is a somewhat morecomplicatedsolution, primarily duetotheSommerfeld-

0018-926X/83/0900-0740$01.OO 0 1983 IEEE

POZAR: MILLIMETER WAVE PRINTEDANTENNAS

741

type integrals involved in details.

the Green's function. See

[5] for

11. PRINTED DIPOLE AND MICROSTRIP PATCH ANTENNA ELEMENTS

This section will discuss propertiesofprinted patch antenna elements, and present data their on characteristics. dipoles and electrical

A. General Element Characteristics

Printed dipole radiating elements have been extensively studied by Alexopoulos et ai. [8] , [ 9 ] , [ l o ] , via a similar moment center-fed printed method procedure. Fig. 1 shows a typical dipole element. In practice, the feed may take the form of a paralleltwo-wireline printed on the substrate. This feed line may carry RF er,ergy or, if a detector is placed at the dipole gap, I F energy. In either case, this type of feed is balanced with respect to the ground plane, which can be a serious disadvantage in some applications. The use of parallel microstrip feed lines to couple the radiating dipole, as in [7], can alleviate this difficulty at the expense of a more complicated feeding structure, possibly involving printedconductorsontwosubstrate levels. Advantages of the printed dipole are that it uses less substrate area compared topatchelements(particularlyimportantin arrays), andthat it can be used near its first or second resonances without deleterious higher order modeeffects. The microstrip patch antenna, also shown in Fig. 1 ?however, is inherently unbalanced with respect to the ground plane. The patch can be fed with a microstrip line, or witha probe from the bottom of the substrate. An unbalanced antenna element is probably advantageous when RF or IF circuitry is t o be combined with the antenna in a hybrid or monolithic configuration. Some disadvantages are thatthe rectangular patch uses more substrate area than the dipole, and that a probe-type feed may be difficult tofabricateonmonolithicsubstrates,or even on quartz substrates. A problem also exists with microstrip line feeds, since a microstrip line's characteristic impedance determines the feed line width, and is relatively constant with frequency. The size ofa resonant patch antenna, however, decreases with increasing frequency, so that a given microstrip feed line on a substrate has an effective upper frequency limit beyond which the resonant patch width would be less than the feedline width. Finally, should one realize the intrinsic differences in the electrical operationofprinted dipoles andpatches. The first resonance of a printeddipole, like a half-wave dipole in free space, is aseries-type resonance, while the first resonance of a patchantenna is a parallel-type(anti) resonance(Inreference to printeddipoles,theterms half-wave and full-wave refer to lengths in +e effective dielectric medium, and so are analogous to the operation of dipoles in free space.). This difference is a result of the field structure created in the vicinity of the element by the particular feed. For a dipole, the feed couples to the electric field component along the dipole axis, while a coax or microstrip line-fed patch antenna is coupled by the electric field component normal to the substrate.Printed dipoles andpatches,however, have similar current distributions, the thus radiation patterns are similar

"1
P
0
1

I I

I ----HALF-WAVE --MICROSTRIP

OlWLE
PATCH

m.

Fig. 1. Resonant lengths of a printed dipole and a microstrip patch versus d for cr = 2.55. W = 0.3 hg for the patch. Also shown are the printed dipole and patch geometries.

TABLE I ELECTRICAL PROPERTIES O F FfICROWAVE SUBSTRATES tan 6 X-bmd

Substrate

er

F'TFE Qu
%As

2.55 3.78 12.8

0.001-0.003 0.001
0.002

were chosen for comparison in this paper. This choice was based on the fact that the permittivities range from 2.55 to 12.8, and that these materials are either in use today or are expected t o be used for millimeter wave antenna systems. Because of time and space considerations, not a l materials available today could be l compared here, but it is felt that other typical substrate materials have properties roughlyinrangeof thoseconsidered.Table I summarizes thetypical electrical properties of these three substrates. Polytetrafluoroethylene related and products Rexolite like and Duroid have been used extensively in the microwave band. Quartzsubstrates have very good dimensional stabilityand are often used in microwave integrated circuits. Gallium arsenide is probably the preferred substrate material for monolithic microwave integrated circuits.

C.Resonant Frequency
Clearly oneofthefirstconsiderations in the design of a printed antenna element is the length L of the element required for resonance. This length is a function of substrate thickness d and dielectric constant E, and, in the case of a microstrip patch, a function of the patch width W. Because the dielectric fills only part of the region surrounding the antenna element, the resonant lengthdoesnot scale with dielectric constantas l/&, as an antenna in a homogeneous medium would. Fig. 1 shows the required lengths for the first resonance of a printed dipole and a rectangular microstrip patch element versus substrate thickness d for a PTFE material. The patch width is W = 0.3 X, (X, is the free-space wavelength). The dipole length varies less than 6 percent for 0 < d < 0.5 b ,and is slightly longer than the patch length, which varies somewhat more with

[I1 >

PI.

B. Substrate Characteristics Three substratematerials-F'TFE, quartz, and gallium arsenide-

742

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. AP-31, NO. 5 , SEPTEMBER 1983

UJ
!
HALF-WAVE nlm! F

I I

WAVE-/-Z I

SURFACE 3A W .V 1 E S -

SURFACE WAVES-/

e !
0
1 2

dh /.

. 3

Fig. 2.

Resonant lengths

for half-wave and fun-wave printed dipoles = 3.78. versusd for

of the patch antenna is that it stops resonating for substrate thicknesses greater than 0.1 1 X,. With 1 increasing substrate thickness, the trend is t o an entirely inductive --- MICROSTRIP PATCH inputimpedancelocus. This effect occurs for both probe-type -.- FULL-WAVE DIPOLE andmicrostripline-type feeds. Thissituation is probably undesirable in most cases, and so the use of patches on thick substrates may not be practical unless some way of countering this inductive trend, by using acapacitive-gapcoupling from a feed line, for example, is used. The dotted continuation of the patch length curve in Fig. 1 is given only to show the length chosen for 4 the calculation of other data presented in Figs. 4, 6, and 8. An increase in patch width W can reduce the resonant length by a few percent; the length reduction greater for thicker substrates. is Fig.shows 2 the resonant lengths for half- and full-wave printed dipoles on a quartz (E,. = 3.78) substrate. (The usefulness I o ! 8 0 .05 of the full-wave dipole is discussed in Section 1 - . 1 D ) that the Note . 2 .25 ' d/x, I 5 length for the first resonance has decreased from the correspond- Fig. 3. Resonant lengths of half-wave and full-wave printed dipoles and a ing value for PTFE, but not by the factor d m , which microstrip patch versusd for E,. = 12.8. W = 0.15 A 0 for t h e patch. would be the case if scaling could be appliedwith dielectric constant. Fig. 3 shows therequired lengthsforhalf-and full-wave printed dipoles and a microstrip patch element onGaAs. Again, the patch element does not resonate for substrate thicknesses greater than 0.08 ho .
1

d. Aninterestingfeature

D.Resonant Resistance
Fig. 4 shows the input resistance of a half-wave printed dipole and a microstrippatchon a PTFEsubstrate versus thickness. As previously pointed out, the patch element does not strictly resonate for d > 0.11 A,; the patch resistance shown in Fig. 4 for d > 0.1 1 ho is the real part of the input impedance for a patch length of 0.270 ho. Since the printeddipole's first resonance is a series-type resonance, input the resistance is very small for small d, since electrically t i substrates imply highQ resohn nances. The microstrip patch, having aparallel-type resonance, shows a high input resistance for small d. Fig. 5 shows theresonant resistance of half-wave andfullwave printed dipoles and a microstrip patch on a GaAs substrate. Again, the patch resistance for d > 0.08 X, is taken as the real part of the input.impedance forL = 0.105 ho. The full-wave dipole has a parallel-type resonance, with high input resistance for small d, similar to a full-wave dipole in free space. This element has interestingadvantages in some applica-

half-wave printeddipoleandamicrostrip Fig. 4. Inputresistanceofa patch versus d for cy = 2.55. The patch is probe fed at a pointL / 4 from the edge, and W = 0.3 ho.

POZAR: MILLIMETER WAVE PRINTED ANTENNAS

I

743

Fig. 6 . Bandwidths of a half-wave printed dipole ana a microsmp patch = 2.55. W = 0.3 ho for the patch. The cavity model preversus d for is diction for the patch bandwidth also shown.
0
.05

d5

.25

dlh.

Fig. 5 .

Input resistance of half-wave and full-wave printed dipoles and a microstrip patch versus d for er = 12.8. The patch is probe fed at a point L/4 from the edge, and I= 0.15 ho.

tions[3] . First,its half-power beamwidth is significantly less than that ofa half-wave dipole. Second, if apairof full-wave dipoles are arranged b / 2 apart to form a subarray element, the E and H plane beamwidths will be about equal, andif a detector diode is placed in the center of the subarray and connected to the dipoles by a printed parallel line, as in [3] the X0/4 line length will yield an impedance inversion fromthe high input resistance of the full-wave dipoles to a low impedance for matching to the diode. All the dipoles are center fed, and all the patches are probe fed at a point L/4 from the (radiating) patch edge. Moving the feed position toward the end of the dipole or patch will increase the input resistance, at the first resonance.
~

or position does not affect the intrinsic patch bandwidth. The bandwidth increases rapidly with increasing substrate thickness, so that bandwidths of 10-20 percent can be obtained for substrate thicknesses in the range of d = 0.1 X. t o 0.2 ho. Also note that the bandwidth of a patch is significantly greater than that ofaprintect dipole, at leastover the range for which the patch actually resonates (d < 0.11 ho). These factsareconsistent with the antenna gain/bandwidth relation to antenna size, as discussed by Harrington [ 11] . The lowest achievable Q of an antenna is inversely related to antenna volume; since the patch antenna encompassesgreater a volumethandoestheprinted dipole, its Q can be lower than the Q of the dipole, hence the bandwidth can be greater. Also shown in Fig. 6 is a bandwidth calculation for the patch using the cavity model [ l ] . The cavity modelapproximation is seen to be useful forsubstratethicknesses d < 0.04 Xo. Fig. 7 shows the bandwidth of a printed dipole (half-wave) and a microstrip patch on a gallium arsenide substrate, versus d. Again the patch bandwidth is greater than the dipole bandwidth.

E. Bandwidth
Bandwidth is defined hereasthe half-power width of the equivalent circuitimpedance response. For a series-type resonance, as shown in [ 101 this bandwidth (BW) is 2R BW = dX I
I

F. Power Lost to Surface Waves

Both TE and TM surface waves can be excited on a grounded dielectric substrate. The cutoff frequency of these modes given is by D l ] .

nc
fc =

w0 -

dw

Iwo

4 d -

(2)

where 2 = R jX is the input impedance at the resonant frequency w o . For a parallel-type resonance (1) is used with R replaced by G and X replaced by B , where Y = G j f l is the input admittanceat resonance. Thisdefinition of bandwidth implies a standing wave ratio of about 2.4, for atransmission lineof characteristic impedance R or 1/G R. The derivative in (1) can be evaluated by calculating the input impedance two at frequencies near resonance and using a finite difference approximation. Fig. 6 shows the calculated bandwidths of a half-wave printed dipoleand a microstrippatch versus substratethicknessfor a PTFE substrate. The patch width is 0.3 X, and is fed by a probe at a point L/4 from the patch edge, although the feeding method

where c is the speed of light, and Iz = 0 , 1, 2,3, for the TMo, T M o mode TE,, TM2, TE, surface mode. that Note the has a zero cutoff frequency, so that it can be generated for any substrate thickness d. As the substrate becomes electrically thicker, more surface modes can exist and the coupling to the lower ordermodes can becomestronger.Forthinsubstrates (d < 0.01 X,) surface wave excitation is generally not important. Forthickersubstrates surface waves may have a detrimental effect printed on antenna performance. Surface wave power launched inan infinitely wide substratewouldnotcontribute to the main bean1 radiation and so can be treated as a loss mechanism. A radiation efficiency can then be defined as

e=

Prad

Prad + psw

(3)

744

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. AP-31, NO. 5 , SEPTEMBER 1983

-HALF-WAVE

r
DIP01E

MICROSTRIP PATCI i

Fig. 7.

4 % Bandwidths of a half-wave printed dipole and a microstrip patch versusd for er = 12.8. W = 0.3 X0 for the patch.
WAVE 1 SURFACE WAVES 2 SURFACE
I

.OS

.15

.2

.25

I
I t I

>?-

I
---HALF-WAVE DIPOLE -MICROSTRIP PATCH

I
I I

tp.9-

r=2.55

I
I
I

ifi iOsL
T
I

1
I
I
I I

0 ,

,
1

,
.2

% Fig. 8. Loss due to surfacewavesfor a half-waveprinted dipole and a microstrip patchversusd for er = 2.55. W = 0.3 ho for the patch.

.3

.4

where Prad is thepowerradiated via space wave (directmain beam power), and Psw is the power coupled into surface waves. Prad Psw is thenthetotalpower delivered totheprinted antennaelement. Dielectricloss is ignored here.Theeffectof afinite-sized substratewould be to diffractthe surface waves fromthesubstrate edges, possibly causing undesirable effects on sidelobe level, polarization, main or beam shape. Surface waves could also be diffractedbyorcoupledtofeed lines or components on the substrate. In the moment method formulation surface wave fields and space wave fields are easily separated from the Sommerfeld-type integralexpression for the total fields of an elementalcurrent source on a grounded dielectricslab-thesurface waves come fromthe residuesof thecontour integral.Since themoment methodimpedancematrixelements areexpressedin termsof integrals of the fields from the expansion modes, these elements can be broken up as

and the radiated power can written as be

Prad Re =

I:ZEtIm.

(6)

Zmn mn =Z

cad +

zw , sm n

(4)

where Z,, representsthematrixelement using the total field and Z E : and Z z n represent thedirectradiation (space wave) and surface wave contributions, respectively. Then, if In representsthecurrentonthe nth expansionmode,thetotalinput power can be written as
n m

Fig. 8 shows efficiency (3) versus substratethickness d for a half-wave printed dipole and a microstrip patch, for E , = 2.55. Observe that e + 1.0 as d + 0, sincesurface wave excitation is negligible for very thin substrates. As d gets larger the TMo surface mode becomes stronger, reducing e. However, theradiated power becomes greater as d increases, so that e levels off and starts to increase for d > 0.1 ho. At d = 0.2 ho, the next surface mode (TE,) starts to propagate, causing a slope discontinuity in e and adecrease in e as this mode becomes more strongly coupled. This type of slope discontinuity is also seen in related dielectric covered antenna problems [ 121 . An interestingfeature of Fig. 8 is the similaritybetween e for the dipole and the patch. Also, e does not depend on the feed location of the dipole or patch, or on the patch width W. Fig. 9 showsthe efficiency e for half-wave and full-wave dipoles and a microstrippatch on GaAs substrate.Notethat surface wave power accounts for over half ofthetotalinput power for d > 0.045 A,. G. Losses Due to Dielectric Power loss due to dielectric heating can be calculated by using the loss tangent complex and permittivity the for particular dielectric material. For the half-wave dipole(aseries-type reso-

POZAR: MILLIMETER WAVE PRINTED ANTENNAS

745

,=12.8

-HALF-WAVE DIPOLE --_ MICROSTRIP PATCH -.- FULL-WAVE DIPOLE

.3 SURFACE
WAVES
I

r tan s=.003
-I I

tanb.003

I I
I

-HALF-WAVE DIPOLE
0
MICROSTRIP PATCH Er=2.55

I
I
I

I
I
1

.05

1
4J-a

.1s

.2

.2 5

.os

Fig. 10.
1

a 0

.15

.2

.25

Loss due to dielectricforahalf-wave dipole and a microstrip patch versusd for E, = 2.55. W = 0.3 X0 for the patch.

Fig. 9. Loss due to surface waves for half-wave and full-wave printed d& poles and a microstrip patch versus d for E , = 12.8. W = 0.15 hg for t h e patch.

nance), for example, the radiation efficiency based on dielectric loss can be calculated as q =-, Rr (7) Rr +- RI where R, is the radiation resistance at the input terminals and Rl is the loss resistance. Rr and RI can be found from two calculations of input impedance; one with tan 6 = 0, and one with tan 6 # 0. The radiation resistance is R, = Re(&) for tan 6 = 0 , and the loss resistance is found from R, + R I = Re @in) with tan 6 # 0. This is an accurate procedure for small losses. For full-wave dipoles or microstrip patches (anti-resonances), the efficiency is calculated using conductances in (7). Notethat efficiency as defined by (7) does not includepower loss to surface waves (althoughit does includeheating loss fromsurface wave fields). Efficiency based on dielectric loss for a half-wave dipole and a microstrip patch versus d is shown inFig. 10 for E, = 2.55. Thelengths of the dipole andpatcharechosenforresonance, and the patch width is again W = 0.3 A. Loss tangents of 0.001 , , and 0.003 were used. It is seen that the patch efficiency is greater than the dipole efficiency, that and efficiency improves rapidly as substrate thickness increases. Bothoftheseeffects can be explainedby notingthat,for a given powerlevel, the fieldsare moreconcentrated for thin substrates or narrow antennaelements,thus more power is lost to dielectric heating than in cases of thicker substrates orwider elements.
111. PRLNTED ANTENNA ELEMENTS IN AN ARRAY ENVIRONMENT

Fig. 11. E-planeandH-planemutualcouplingmagnitudebetween wave dipoles and between microstrip patchesversusd for E ,

half= 3.78.

This section will discuss some aspects of the printed antenna element in an array configuration-in particular, mutual coupling between array elements and array effects surface wave power. on

A . Mutual Coupling
When printed antenna elements are in an array environment, large mutual coupling levels can degradesidelobe levels, main beam shape, and possibly cause array blindness. Foreknowledge ofthecouplingbetweenarrayelementsandproper inclusion

into array the design procedure can minimize these effects ~ 7 1 ~131. , The calculation of mutualcoupling as two-porttransfer a impedance the by moment method is described in [SI. The method yields magnitudes as well as phase,andcomparisons with measured data for patches are shown in [5].Mutual coupling between parallel and collinear half-wave dipoles versus separation is shown in [9]. Calculations for full-wave dipoles have also been made, with the result that full-wave dipole coupling is about 10 dB less thanthe half-wave dipolecouplingforboth parallel and collinear configurations. (These data are not shown here for lack of space.) Fig. 11 shows the coupling between parallel half-wave dipoles and patchesand collinear half-wave dipoles andpatcheson a quartz substrate versus substrate thickness. elements The are resonant, and the spacing between elements is 0.5 io. For thin substrates the coupling levels are very low but increaserapidly with increasing thicknessandthentend to oscillate forthicknesses greater than about 0.5 Xo. The dominant coupling mechanism for the parallel configuration is via space wave fields; since these fields are stronger in the broadside than in the endfire directions, coupling the levels between parallel dipolesarefairly large for close spacings, but

746

IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. AP-31, NO. 5 , SEPTEMBER 1983

drop off quickly as the spacing increases. Surface waves are launched in the dipoles endfire direction and so have most effect for thecollinear configuration.

B. A m y Efficiency-an Optimization Procedure

Section 11-F discussed printed antenna element efficiency e, based on power lost to surface waves. This section will discuss whathappens t o e whenelements are combined in anarray, where it will be seen that e can be increased or decreased, depending on the element excitations, from the efficiency of an isolated element. An example will be given fortwo collinear half-wave dipoles spaced ho/2 apart on a quartz substrate. Microstrip patches can be treated in the same manner, and the procedure can beapplied to arrays with more than two elements. 1 Since all mutual coupling terms between array elements would I 0 , be included, (5) and (6) apply to the total input power and total 0 .05 1 .15 2 25 3 d..hradiated power, respectively, for an array of printed elements. efficiency e for two collinear half-wave dipoles The overall array efficiency, based on power lost to surface Fig. 12. Optimized h0/2 apart versus d for E,. = 3.78. L is chosen from Fig. 2 for resonance. waves, can then be written in matrix form as
I I

the eigenvalue problem,

[Al[VPl = e [ B l [ V P l ,
where [Z] is the total (square) impedance matrix, [&ad] is thecontributiontotheradiatedfield, [I] is a column vector of expansion mode currents, and the superscript t denotes transpose. If we let [R] = Re [Z] and [&ad] = Re [ & d l , then (8) can be written as

(13)

[&ad] [s] and [B] = [S*] [ R ] [SI are where [A] = [S*] Hermitian2 X 2 matrices. A similar procedurehas beenused for free-space array optimization [14]. If the dipoles are identical then, by symmetry, it can be shown that [A] and [B] have the form

To illustrate the method an example of two printed dipoles, on each dipole, will be discussed. withthreeexpansionmodes The [Z] and [R]matrices are then 6 X 6 , and [I] is a six-element column vector. If the expansion modes are numbered consecutively down each center-fed dipole, the terminal currents for the fKst andsecond dipoles will be I , and I,, respectively. If VI is the(port) voltageapplied to dipole1and V2 is the(port) voltage applied to dipole 2, then

the with a ] , a 2 , b l , b2 real. The eigenvectors, representing feed voltages for optimum e , are then either even or odd:

The corresponding eigenvalues are then the efficiencies resulting from the above excitations:

= Nowlet [Il0] [I] for V I = 1, V, = 0 andlet [Io ] = [I] for VI = 0, V2 = 1. Then by superposition the dipole currents caused by excitation port voltages V I and V2 can be written as

[ f l = [SI [VPI

(1 1)

port voltage vector and [I = S where [VI is a two-element [ [ I l o ] , [ I o l ]] is a 6 X 2 matrix. The array efficiency can then be expressed in terms of the port voltages as

Note that (12) expresses the performance index e as a ratio of two quadratic forms. Thus, (12) can be optimized by solving

Generally the even modeproducesmaximum e while theodd mode produces minimume . The optimized efficiency e for two collinear half-wave dipoles &,/2 apart versus substrate thickness (er = 3.78) is shown in Fig. 12, for even and odd mode excitations. Also shown is the efficiency of an isolated dipole. As can be seen, the efficiency can be improved by as much as 30 percent for even mode excitation. A similar calculation for two parallel dipoles results in a 10 percent improvement for even mode excitation. For the data shown here, maximum improvement occurs cophasal for excitation of the array elements-a very practical result for broadside arrays. Odd-mode excitation generally produces a reduced efficiency,whichmeansmorepower isbeing coupled to surface waves-a result which may be of interest for surface wave antennas. In Fig. 12 the e, and eo curves cross at about d = 0.19 ho. The change in efficiency for printed antenna elements in an array can be partially explained in terms of the phasings of the

POZAR: MILLIMETER WAVE PRINTED ANTENNAS

747

surface wave fields. Surface waves, launched endfire from each dipole, are significantly out of phase (because of the ho/2 dipole spacing) and tend to cancel. It is hypothesized that an element spacing exists such that maximum cancellation occurs ande -+ 1, at least forsubstrate thicknesseswhere onlyone surface wave mode exists. This situation would probably occur not when more than one surface mode is present, since the different phase constants would preclude totalcancellation.

REFERENCES
K. R. Carver andJ. W. Mink, Microstrip antenna technology, IEEE Trans. Antennas Propagat., vol. AP-29, pp. 2-24, Jan. 1981. [2] R. J. Mailloux, J. F. McIlvenna, and N. P. Kernweis, Microstrip array technology, IEEE Trans. Antennas Propagat.. vol. AP-29, pp. 25-37, Jan.-1981. [3] K. S . Yngvesson, T. L. Korzeniowski, R . H. Mathews, P. T. Parrish, and T. C. L. G . Sollner,Planw millimeter wave antennas with application to monolithic receivers, Proc. SPIE, vol. 337, (Millimeter Wave Technology), 1982. [4] D. B.Rutledge and M. S . Muha, Imaging antenna arrays, IEEE Trans. Antennas Propogat., vol. AP-30, pp. 535-540, July J982. [5] D. M. Pozar,1nput.impedance and mutual coupling of rectangular microstrip antennas, IEEE Trans. Antennas Propagat., vol. AP-30, pp. 1191-1196, NOV. 1982.. [6] J. R. James, P. S . Hall, ahd C. Wood, Microstrip Antennas: Theory andDesign. Stevenage, U.K.: Peter Peregrinus, 1981, pp. 51-64. [7] R. S. Elliot, and G . I. Stem, The design of microstrip dipole arrays including mutual.coupling, Part 1: Theory; Part 11: experiment, IEEE Trans. Antennas Propaga!.. vol. AP-29, pp. 757-765, Sept. 1981. [8] I. E. Rana. and M. G.Aexopoulos, Current distribution and input impedance of printed dipoles, IEEE Trans. Antennas Propagat., vol. AP-29, pp. 99-105, Jan. 1981. [9] N: G . Alexopoulos and I. E. Rana, Mutual impedance computation between printed dipoles, IEEE Trans. Antennas Propagat., vol. AP29, pp. lO&lll, Jan. 1981. [lo] N. G . Alexopoulos,P. B. Katehi, and D. B. Rutledge, Substrate optimization for integrated circuit antennas, IEEE Trans. Microwave Theory Tech., vol. MTT-31, pp. 55C-557, July 1983. Electromagnetic Fields. New [ l l ] R. F. Harrington, Time-Harmonic York, McCraw-Hill, 1961. 1121 M. C. Bailey and C. T. Swift, Input admittnce of a circular waveguide aperture covered by a dielectric slab, IEEE Trans. Antennas Propagat., vol. AP-16, pp. 386-391, July 1968. . [13] M. C. Bai!ey and F. G . Parks, Design of microstrip disk antenna arrays, NASA Tech. Mem. 78631, Feb. 1978. [ 141 R. F. Hamington, Field Compwation by Moment Methods. New York, MacMillan, 1968. [I51 E. H . Newman, P. Bohley, and C. H. Walter, Two methods for the measurement of antenna efficiency, IEEE Trans. A n t e k s Propagat., vol. AP-23, pp. 457461, July 1975.
,

[I]

IV. CONCLUSION
This paper has presented data on resonant length, input resistance, bandwidth, surface wave power, dielectric loss, and mutual coupling printed for dipoles and microstrip antennas. patch Emphasishas been on theeffect of thesubstrate,whichmay have a dielectric high constant may and not be electrically thin atmillimeter wave frequencies. Many other configurations of array elements could be studied for mutual coupling effects and surface wave power optimization. It is conceivable that for an array with many elements surface wave power can be made negligible for particular element spacings and excitations. A very important need is t o verify moreofthetheoretical calculationswithmeasurements,particularlyforprinteddipole input resistance and surface wave losses for dipoles and patches. Some type of modified Wheeler cap method [ 151 may work for the latter measurement. ACKNOWLEDGMENT The author would like to thank Professor Yngvesson of the University of Massachusetts for his interest in the implications of this work, and Professor Schaubert of the University of Massachusetts for a critical review of the manuscript.

David M Pozar (S74-M80); for a photograph and biography please see . page 350 of the May 1982 issue of this TRANSACTIONS.