Professional Documents
Culture Documents
unless more subsections than those normally required for and two simultaneous matrix equations for the column
driven elements are taken.In using the three-term theory matrices of complex coefficients ( A ( 2 )and
) {A(3)}:
the largest matrices t o be handledfor an N-element array
are of a dimension N X N and no convergence problems +
[ W ) ] { A ( 2 ) ) [@~(3)]{A(~)j
= - ( @ ~ ( l ) J A 2 ( l ) (12)
5
N hi
1i(2i')~2S(e,Zl) dz;
with
(3)
0 2 = j30
N
i=l
/
h i
-hi
I ; ( z < ) K 2 j ( O J x dz;.
l) (15)
with
Xi("(2) = sin/9O(hi- I z I) (8)
Xi'') ( z ) = COS / 3 -
~ COS &hi (9)
Xi(3) ( x ) = cos +/3& - cos +pohi (10)
m d AJl) = 0 for i # 2 (parasitic elements). Substitution
of (7) in (1) and use of certain approximate relations for
the integrals involved yield
.[l - cos (31) / Ices r+)- r+> cos
+ [1 - cos -
C)Il. 1 - sm - (20)
CHENG AND CHEN: SPACINGS FOR YAGI-UDA AFtRAYS 617
k P 2
k = 2
I h;
-I& i
SiCmn)
(2:) KkiP(Zk,Zi!) dzl
k # i, m = 1,2,3. (30)
The elenlents of [ ! @ d @ ) ] and [ + d c 3 ) ] in (19) are thesame
as those for [ \ k J m ) ] in (13) for m = 2 and 3. The elements We note that the additional term due to spacing pertur-
of [6(2)]and [ & ( 3 ) ] are the same as those for [+c2)] and bation is proportional to the difference (Adk - A&). As
[W] except when k = 2. For k = 2, we have a consequence, we write the new, perturbed matrices in
(12) and (13) as
= (0)- ip2id'(2) (26)
&(2) IpZi(2)
[@(m)]p = [@(m)] + [A@m)], nt = 2,3 (31)
&(3) = qrti(3) (0)- q r 2 & p 3 ) (27) [?Pdm)-Jp = [9?dm)] + [AfPdm)], m = 2,3 (32)
where f P 2 i ( 2 ) ( 0 ) and \k2i(3) (0) are { @2c1) )P = { +2(l) +
1 { Aa2") ] (33)
{ q az(1) 1 -!- { A*zd(') 1 (34)
ezi(")(0)
= 1 hi
-hi
dzi',
Si(m)(zi')Kzi(O,z() m = 2,3. (28)
{Ip#}P =
[@d@)ll
-1
ICFJ { Ad}
[PzIddi
(1 + jSoAdi sin e cos 4) *exp ( jSdi sin e cos 4).
i=l
Using the perturbed current coefficients in (39) and (40),
where [Pz] is an N X N square matrix defined in Appen- we obtain from (47)
..* 11. Similarly, a.nother N X N matrix [P3] can be
6 O N
defined (also in Appendix 11) from the right-hand side +j -
E'(e,#,) 'v E(@,,#,) exp ( jsdi sin g, cos 4)
of (42), and (41) and (42) become Bo i-I
[
[o(z)-J p
[*d")]
From (43),
( 3 q
[\Ed(3)ll
{AA(2)}
1 1{
{ A A ~
{a@))
=
CPZl
[p34
{Ad). (43)
+ 1tli(3)((e)
inversion.
array, a,nd
IA4 (45)
CaENG AND CHEN: SPACINGS FOR YAGI-UDA ARRAYS 619
(53)
{ B I }and [Cl] have previously been defined, respectively,
where Pi, is the t,ime-average input power. With spacing in (56) and (57) ;and [CZ]is a positive definite Hermitian
perturbation E becomes E', Pi, becomes Pin', and the matrix.
perturbed gain becomes The objective of gain optimization by spacing pertur-
bation is t o find the small changes in theelement spacings
(54) such thak the array gain in a. given direction is increased
a.nd t o repeat the process until further increases in gain
From (52), are negligible. Hence, it is essential that
I Evo,do)12 = (E* + AE*) (E' + AE) AG(eo,do) = G'(eo,b) - G(@o,+o) (66)
= I E 1' + 2{Ad)T{B~]
+ {Ad]'[Cl]{Ad] be positive. Substitution of (53)- (62) in (66) yields
(55)
where
{BI)= Re (E'{D*}) (56)
and
CCll = {D}*{D}T. (57)
I n (56), Re ( E {D*) ) = real part of the product of E
a,nd the complex conjugate of the column matrix ID}. Note that thenegative sign in (68) for { B }in the numer-
The AT X N square matrix [C,] is positive semi-definite, ator of AG(Bo,dJo) in (67) implies that the array gain d
and, since {Ad] is a real mat,rix, [Cl] in the last term of decrease for an improper choice of { Ad}.
( 5 5 ) can be replaced by m e Cl]. Pi,,' in (54) is In order t o be certain that AG(&,&) slcill be positive,
Pin' = $ R e [Voz*IzP(O)] = Pi, + (Ad)T{B2) (58)
we make use of a known relation in the theory of matrices
[15], [19]. Applied to the present problem, the relation
where asserts that if m e Cz] is positive definite, then
. - -:Va Re [A2(')&(')(0)
p In
({BITIRe C2T1(B])({AdIT[Re CZ]{A~})
+ +
Z ( ~ ) A z ( ~ ) X(~0()1
A z ( ~ ) S (0) ~ ) (59) 2 ({Ad)T(B])2. (69)
and the kth element of the column matrix (B2}
is
I n (69), the equality sign holds when
{ A d ) = a[Re CZ]-~{B} (70)
For a.lossless array,theinput power equals the total
power ra,diated, a.nd Pin' can be mitten in an alternative where CY is a positive consta.nt. Hence, if t>hespacing
form : changes in { A d } are chosen, such that,
{Ad] = a[Re C2]-1(2{B~] - 60G{B2)) (71)
then
620 IEEE TRANSACTIONS
AND
ON ANTENNAS PROPAGATION, SEPTENBER 1973
/:h,,
TABLE I
GAINOPTIMIZATION YAGI-UDA
FOR SIX-ELEMENT ~ A Y
(PERTURBATION OF SPACINGS)
DECECMR
2h1 = 0.51X, 2h9 = 0.50% 283 = 2h4 = 2h5 = 2hs = 0.43X,
a = 0.003369X
,[<h,
Optimizedhray 0.250 0.336 0.398 0.310 0.407 11.81 -6 - 5 4 - 3 - 2 -I 0 I 2 3 4
A')?\,
10-2A/V
6 0
/ 0.15 , ) , jI
0.05
I
-6-5-4-3-2-1 0 I 2 3 4 5 -5-4-3-2 -I 0 I 2 3 4 -2-1 0 I 2 3 4 5
lO-'A/V
0.2 -
0.1 -
TABLE II
GAINOPTWATTIONFOR SIX-ELEMENT
YAGI-UDA
ARRAY
OF ALL ELEXENT
(PERTURBATTON SPACINGS)
Fig. 4. Normalized
patterns of six-element
Yagi-Uda
arrays
(Example 2).
The a in (71) should be sufficiently small to sat,isfy the
condition ( Adi) / d i << 1.
Equation (71) represents the requisite fmt perturba-
tion, with which AG and G' are determinedfrom (72) 0.310X. The director spacings are to be adjusted for gain
and (66) respectively. A second perturbation can then maximization (reflector spacing fixed).
)
be performed using the new G' as the initial gain, andthe Since the driven element is a half-wave dipole, the
process repeated until further increases are negligible. It formulation outlined in (14)-(28) is used to determine
has beenfound that. this iterative procedure converges the current distributions in the array element.s, the radi-
rapidly a.fter a fewcycles. We t.hen have an unequally ated field, and the gain. The ga.in of t,he initia.1 array is
spaced Yagi-Uda a.rray which yields a maximum gain. found t o be S.06 (9.06 dB), which checks well with the
result obtained by Morris [14]. Now we keep the relative
positions of the reflect.or and the driven element fixed
VII. NUblERICaL EXAMPLES
(bzl = 0.250X) and perturb the positions of the directors
I n this sect,ionwe present the computed results of three in accordance with (70) in order to increase the gain. Six
examples which illustzate the effectiveness of increasing iterations yield the result,s shown in Table I. The opti-
the gain or directivity of Yagi-Uda arra.ys by spacing mized array is unequally spaced and has a gain of 11.81
pcrturba,tion. (10.72 dB), an increase of 46.5 percent (1.66 dB). The
E n m p E e 1: Six-element Yag-Uda array with a half- computed relative field intensities inthe direction of
wave driver (2h2 = 0.5OX; one reflector, 2hi = 0.51X; four maximum radiat,ion are 0.910 and 0.958, respect,ively,for
directors, 2h3 = 2h.j = 2h.5 = 2h6 = 0.43X; 0, = 0.003369X). the initial and op t,imized arrays. The normalized radiation
In the initial array, bB1= 0.25OX, ba2 = b43 = b s = bS5= patterns are shown in Fig. 2. It is seen t.hat the pattern
CHENG AND CHEN: SPACINGS FORYAGI-UDA ARRAYS 621
TABLE III
GAINOPTIMIZATION FOR TEN-WNT YAGI-UDA AILRAY
(PERTORBATION OF D ~ ~ E C TSPACINGS)
OR
2hl = 0.51X, 2hZ = 0.50X, 2hi = 0.43X (i = 3, 4, .**, lo), ~a = 0.003369X
bl/x bdX b,,/X bdX b6dX b16/h bm/X bpdX blO.O/X Gain
Initial Array 0.250 0.330 0,330 0.330 0.330 0.330 0.330 0.330 0.330 12.36
Optimized Array 0.250 0.319 0.357 0.326 0.400 0.343 0.320 0.355 0.397 16.20
for t.he opt,imum array has not only a narrower main VIII. CONCLUSIOK
beam but also lowersidelobes, a fact. which has been A method has been developed for the maximization of
noted previously [20]. t.he forward gain of a I'agi-Uda array by adjusting the
Example 2: Six-element,Yagi-Uda array 1it.h a half- interelement spa.cings. The effects of a finite element
wave driver (2h2 = 0.50X; one reflector, 2hl = 0.5lX; four radius and the mutual coupling beheen t.he array elements
directors, 2h3 = 2hl = 2h5 = 2$ = 0.43X; a = 0.003369h). are taken intoconsideration. A three-term expansion with
In the initial a,rray, b ~ 1= 0.280X, b32 = b.13 = b s = bG5= complexcoefficientsisused to approximate the current
0.310X. All element spacings a.re to be adjusted for gain distribution in the elements and to convert the governing
optimization. integral equations int.0 simultaneous algebraic equations.
The reflector spacing bP1in the initial srray is arbitrar- The array gain is ma.ximized by the repeated application
ily chosen to be 0.280X, and all ot,her element spacings are of a perturbation procedure which converges rapidly to
given as 0.310X. The gain of thisinitial array is 7.53 yield a set, of opt,imum, generally unequal, element spac-
(8.77 dB). Now all element spacings are adjusted simul- ings. 1llust.rative examples are given to show typical gain
taneously in t,he optimization procedure t,o increase t.he increases that a.re attaina.ble with this technique.
gain. The results are summarized in Table 11. The gain Although the formulat,ion using t,he three-term theory
of the optimized array is 11.85 (10.74 dB), an increase of a.ppears tedious, t.he end result is in a fairly simple form.
57.3 percent (1.97 dB). The matrix equations need not be reformulated for differ-
The real and imaginary pa,rts of t.he currents in the ent a.rrays once they have been obt.ained.The formulation
elements of t.he optimized array withunequal spacings itself is on firm grounds and has been expounded in many
are plot,ted in Fig. 3, which includes t,he effects of mutual research a.rticles and several books. As far a.s its applica-
coupling and finite dipole radius. It is interest,ing to find tion to t,he present problem is concerned, the only numer-
that, thedirector spacing for t,he optimized array is 0.2501, ically tedious part is the evaluation of definite integrals
which confirmswith what hasbeen foundby other investi- of the type given in (24), (25), (A-7), and (A-8). The
gators [lS]. The normalized radiation patterns for both method of momentswith subsectioning cannot conven-
the initial and t,he optimized arrays are given in Fig. 4. ient.ly be used here because the critica.1 dependenceof the
Again, the pat.t>ernfor t,he optimized a.rray has a narrower currents in the parasitic elements on mutual coupling
mainbeamas well as lowersidelobes. The computed demands h e subsectmiorlingand the consequent manipu-
relative field intensities in the direction of ma.ximum lat,ion of complex matrices of very large dimensions.
radia,t,ionare 0.920 and 0.976, respectively, for t,he initial The largest matrices encountered in the spacing per-
a.nd optimized arrays. turbation technique using the t,hree-term theory are of a
Example 3: Ten-element Yag-Uda a.rray xith a half- dimension X X N for an LY-element array. Theconvergent
wa.ve driver (2ha = 0.5OX; one reflector, 2hl = 0.5lX; eight iterat.ive procedure yields the opt.imum spacings for maxi-
directors, 2hi = 0.43X, i = 3,4,.. -,lo;a = 0.003369h). In mum gain nit.hout the needfor a haphazard trial-and-
the init,ial array, bPl = 0.250X, bB = b43 = = b10,9 =
error approach or for interpreting a vast dat,a collection.
0.3101. The director spacings are t.0 be adjusted for gain
maximization. I
APPENDIX
Wit,h t.en elements in a Yagi-Uda array, it mould be Expressions for 4 d r n ) 4kk2(1),
, $ d m ) , and $ k B d ( l ) in (35)-
impract,icalto use the moment method wit.h subsectioning (38) :
for numerical solut-ion. However, only 10 X 10 matrices
a.re involved in t,he present formulation. The results for [ ; ; y l L k ) - $kid'(%) COSPOh.k, k # i
=
the opt,imized array are summarized inTable 111. The
calculated gain for t.he array with eight equa.llyspaced k = i
directors is 12.36 (10.92 dB) which checks very closely (-4-1)
n-ith the result of hIorris [14]. The gain of the optimized
array is 16.20 (12.10 dB, an increase of 31 percent (1.18 $kZ(') (h.k) - (1 - 6k2) $ k ~ a ' ( ' ) COS p&, k #2
+&) =
dB). Even for this example, the t,otalcomput,ingtime for
seven iterat,ions on an IBM 370/155 computer t,ook only k = 2
about 5 min. (A-2)
622 ANTENNAS
IEEEON
TRANSACTIONS AND PROPAGATION, SEPTEMBER 1973
(A-7)
and
IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATIOR, VOL. -21, NO. 5, 6231973
SEPTEMBER
Maruzen Co., 1954, D. K. Cheng and C. A. Chen, “Optimum element spacings for
c31 W. Walkinshaw, ‘‘Theoreticaltreatment of short Yagi aerials,” Yagi-Uda arrays,” Syracuse Univ.,Syracuse,N.Y.,Tech.
J . Inst. Ekc. Eng., vol. 93, pt. IIIA, no. 3, pp. 598-614, 1946. Rep. TR-72-9, Nov. 1972.
c41 D. G. Reid, “The gain of an idealized Yagi array,” J . Inst. D. K. Cheng, “Optimization techniques for antenna arrays,”
Eke. Eng., vol. 93, pt. IIIA,no. 3, pp. 564-566, 1946. Proc. IEEE, vol. 59, pp.. 1664-1674, Dec. 1971.
C51 R. M. Fishenden and E. R. Wiblin, “Design of Yagi aerials,” D. L. Sengupta, “On d o r m and linearly tapered long Yagi
Proe. Inst. Elee. Eng., vol. 96, pp. 5-12, Mar. 1949. antennas,” IRE Trans. Antennas Propagat., vol. AP-11, pp.
C6 1 H. W.Ehrenspeck and H. Poehler, “A new method for obtain- 11-17, Jan. 1960.