Sunteți pe pagina 1din 10

Per-Simon Kildal, Enrica Martini,

and Stefano Maci

W
e derive a general fundamental direc-
tivity limitation formula that applies to
nonsuperreactive antennas of any size
that fit within a minimum sphere of any
given radius rmin. The derivation is done by using a
new concept: the degrees of freedom (DoF) of the
field radiated by arbitrary sources within the mini-
mum sphere must be twice the maximum directiv-
ity. The formula converges to the known bound of
the directivity for large rmin. For small spheres, it
becomes equal to three, i.e., 4.8 dBi, which is the
directivity of the Huygens source. The transition
region between these two limiting cases is deter-
mined by counting the most significant spheri-
cal modes at the surface of the minimum sphere.
This is not trivial, because spherical modes have a
gradual cutoff when their order approaches krmin.
Therefore, we use a weighted summation where
the weighting factor is inversely proportional to the
radiation-Q of the modes. This extends the DoF
from a discrete to continuous function of the mini-
mum sphere radius. The final maximum directivity
is similar to a previously published heuristic limit.
The formulas are compared to results for measured
antennas with large directivity and superdirectivity.
image licensed by ingram publishing

Degrees of Freedom
and Maximum
Directivity of Antennas
A bound on maximum directivity of nonsuperreactive antennas.
Digital Object Identifier 10.1109/MAP.2017.2706659
Date of publication: 20 June 2017

16 1045-9243/17©2017IEEE August 2017 IEEE Antennas & Propagation Magazine


and the DoF over a sphere were
found independently and never
linked together. This article
uses (2) to determine the direc-
tivity limitation of nonsuperre-
active antennas of any size for
any size of the minimum sphere
encompassing the antenna. The
factor 2 in (2) is due to the exis-
tence of two orthogonal polar-
izations all over the sphere, i.e.,
it is due to the vector nature
of the field in contrast to the
scalar nature of the maximum
directivity. Thus, we can also
interpret (2) in the following
manner: “the maximum direc-
tivity of a system is a measure
of the degrees of freedom of
the power density radiated by
the system.” Based on this con-
cept, we determine the DoF
and, thereby, the ma x imum
directivity for nonsuperreactive
antennas of any size by count-
ing the number of radiating
spherical modes at the surface
of the minimum sphere. For
superreactive antennas [4], the
directivity can be, in principle,
higher than this limit. However,
as also stated in [4], the practi-
cal implementation of low-loss
superreactive antennas is very
difficult, and the bandwidth
of gain rapidly decreases with
increasing gain. The main con-
tribution in [4] is showing that
BACKGROUND the maximum possible radius of the significant reactive fields
The maximum available directivity of a large nonsuper directive of an antenna (or scatterer) is equal to the radius of the maxi-
aperture with area A is given by mum possible effective area of the antenna (or the radius of
the maximum possible total scattering cross section of a scat-
D max = 4rA/m 2. (1) terer). For resonant antennas (scatterers), this radius can be
much larger than the radius of the sphere circumscribing the
This formula is normally derived from the aperture efficiency
physical antenna (scatterer). The magnification factor between
integrals (see, e.g., [1, Sec. 7.3]) by using the Schwarz inequal-
the radius of the minimum sphere and the maximum radius of
ity [2]. For a circular aperture with radius rmin, (1) becomes
the significant reactive field is equal to the ratio between the
D max = ^ krminh2. However, in the absence of superreactive
free-space wavelength and the characteristic mode wavelength
sources, the N DoF of the field outside the surface of an elec-
of the radiator. This leads to a formula more general than (1)
trically large sphere with the same radius as the aperture is
for the maximum directivity of antennas, which, however, does
N DoF = 2 ^ krminh2 [3]. Therefore, for large nonsuperreactive
not violate (2).
circular apertures, we have
The maximum directivity in (1) is determined in [5] by ex­­
D max = N DoF /2. (2) panding the field over and outside the sphere in terms of spheri-
cal modes. However, the derivation has some limitations for
This very simple equation has not been published before, as small spheres with few modes due to the discreteness of the
far as we know. The formulas for the maximum directivity number of spherical modes and their gradual cutoff. Therefore,

IEEE Antennas & Propagation Magazine August 2017 17


it becomes difficult to determine the maximum directivity in a energy larger than the reactive energy at the radius rmin. It
unique manner. is well known that the spherical radial (Hankel) wave func-
The radiation Q-factor of the spherical modes plays an tion decays extremely rapidly with the radial distance until
important role in this article. Normally, it is used to deter- the index n is larger than its argument. This means that it
mine the fundamental bandwidth limitations of small anten- is very difficult to excite modes with a polar index n that is
nas versus the radius of the minimum sphere containing the higher than krmin with an antenna of a maximum dimension
antenna [6]–[8]. Originally, the Q factor quantifies the quality rmin. For this reason, the coefficients of the spherical mode
of a narrow-band resonator and represents the inverse relative expansion of the radiated field normally drop abruptly when
bandwidth of it. When dealing with antennas, however, the n becomes larger than krmin [6].
quality is better when the bandwidth is large, so it is gener- The same outcome can be achieved by applying a singular
ally accepted that a lower Q is better. In other words, the value decomposition to the radiation integral operator, which
antenna radiates more easily when it has a low Q. Inspired maps the currents over the minimum sphere onto the radiated
by this and because of the difficulty of exciting higher-order field [3], [10]–[12]. In this case, the relevant single values can be
spherical modes, we use the radiation-Q of spherical modes determined in a closed analytical form [12].
as an inverse weighting factor when counting the radiating If the actual location of the sources inside S is known,
modes originating from the minimum sphere. This allows then the number of modes required for the field represen-
us to estimate the DoF through a real number and thereby tation may be significantly lower than N DoF, but we need
extend the relation between D max and N DoF to the transition N DoF modes to create the maximum directivity in any pos-
region between small and large minimum-sphere radii. In sible direction and for any polarization.
other words, we use the radiation-Q as a measure of how dif-
ficult it is to excite higher-order spherical modes compared to GEOMETRICAL-INTERPRETATION OF DoF IN THE NEAR FIELD
lower-order spherical modes. The DoF given by (3) approximately corresponds to twice the
surface area of the minimum sphere divided by the surface area
RELATION BETWEEN NDoF AND Dmax FOR of a circular waveguide at the cutoff, i.e.,
LARGE ANTENNAS
= 1.18 # 2 ^ krminh2, (4)
4rr 2min
N DoF . 2 2
rr cwcut
THEORETICAL FUNDAMENT OF DoF
The DoF of the field radiated by arbitrary sources is the mini- where rcwcut = 1.841/k is the cutoff radius of the basic TE 11
mum number of independent scalar parameters needed to mode of a circular waveguide. Thus, we can locate approxi-
describe the field in a given domain [3], [10]–[12]. Here, we refer mately N DoF /2 circular cutoff-sized open-ended waveguides
to this domain as the entire space outside the surface S of the (OEWs) side by side over the whole surface of the minimum
minimum sphere surrounding the sources except for a small sphere, each supporting the two orthogonal TE 11 modes (see
region close to S, where very reactive fields originating from inset of Figure  1). Furthermore, the factor 1.18 in (4) is the
the source region within S may dominate. The DoF can also be inverse of 0.85, which is very close to the area fill factor (pack-
the number of scalar coefficients of any type of discrete field- ing density) 0.907 of identical circles on a plane. This fill factor
expansion that is needed for an accurate and nonredundant is smaller than the unity because there are some residual non-
description of the field. covered areas between the circles. The two numbers 0.85 and
Using spherical modes for the expansion leads to N DoF 0.907 are very similar, making the interpretation quite useful
essentially being a function of the electrical size of the source in practice. The discrepancy between them corresponds to a
distribution, i.e., the maximum dimension of the antenna. Out- directivity difference of 0.3 dB (obtained from 0.907/0.85). This
side the reactive region near S, and for large spheres, the DoF distribution of waveguides over the surface of the minimum
is, according to [3], approximately equal to twice the maximum sphere makes it possible to visualize the minimum sphere as a
number of spherical modes that are above the cutoff at the sur- golf ball with craters that are the apertures of the OEWs (see
face S of the minimum sphere, i.e., inset of Figure 1). In addition, the DoF is approximately twice
the number of such OEWs at the cutoff.
N DoF = 2 ^ krminh2, (3)
INTERPRETATION OF DoF IN THE FAR FIELD
where k is the wavenumber, rmin is the radius of S, and the A complementary interpretation can be done in the far field.
factor 2 is a result of the two orthogonal polarizations cor- The far-field sphere is now interpreted as a golf ball (the larger
responding to the transverse electric (TE) and transverse image in Figure 1), but in a completely different sense. The
magnetic (TM) modes. The latter means that we can always craters of the golf ball now represent the areas where each
find orthogonally polarized modes that create the same one of several adjacent narrow pencil beams intersects the
directivity. The spherical modes have a gradual cutoff, far-field surface. Each of these beams is due to a particular set
making the determination of the significant ones (i.e., the of excitations of the OEWs. Because of the two-dimensional
ones that are above the cutoff) uncertain. Therefore, (3) rotational symmetry of the sphere, the N DoF /2 OEWs can pro-
is obtained by counting those modes that have a radiated vide N DoF /2 nonoverlapping pencil beams. In particular, two

18 August 2017 IEEE Antennas & Propagation Magazine


nonoverlapping contiguous pencil beams can be obtained by
shifting the excitation set of the OEWs by one OEW in the cor-
responding direction. Therefore, each beam has a solid angle of
X = 8r/N DoF.

LINK BETWEEN NDoF AND Dmax Ω


The nonoverlapping beams with a solid angle of X = 8r/N DoF
are the narrowest possible beams obtained from the set of
OEWs at the cutoff. As such, each beam is associated with the
maximum directivity D max. By using the definition of directiv-
ity, we obtain D max = N DoF /2. S′
The individual pencil beams can also be explained as a
result of an equivalent aperture with a uniform field distri- 2rmin
bution equal to the cross section of the minimum sphere S
(Figure 1), since the directivity of such an aperture is
^4r/m 2 h (rrmin) 2 = (krmin) 2 = N DoF /2.

NDoF FOR ARBITRARILY SIZED ANTENNAS


Figure 1. An illustration of a beam with a large directivity
and a main lobe covering a solid angle of X = 8r/N DoF on
NDoF AND Dmax FOR SMALL ANTENNAS the outer radiation sphere Sl. The beam originates from
When the minimum sphere diameter is comparable to or N DoF /2 independent uncoupled spherical modes over an
smaller than the wavelength, (3) is not valid. The maximum inner sphere of diameter 2rmin . The illustration corresponds
directivity of a source that is small in terms of wavelength to 2rmin being approximately 6m. There is also an indication
cannot be larger than the directivity of a Huygens source, of N DoF /2 circular discs over the surface S of the minimum
sphere. These correspond to the interpretation of the DoF
i.e., D max = 3, corresponding to 4.8 dBi (see, e.g., [1, Sec. in (4) as the number of circular waveguides at the cutoff that
4.5]). It is also well known that when the source is small in can fill the surface of S.
terms of wavelength, any field outside the minimum sphere
can be described by three orthogonal electric and three dominantly reactive to dominantly radiating. The radiated
orthogonal magnetic dipoles placed in the center of the and reactive powers are known to be proportional to R and
sphere. This means that the field has 6 DoF. Therefore, X, respectively. Therefore, the function Q n (kr) = X n /R n is
we have a quantity similar to the radiation-Q of the mode at radius
r. The functions Q n (kr) = X n /R n are shown in ­Figure  2(a)
N DoF = 6, D max = 3, (5) for the first ten polar indexes. The results are plotted against
kr (in logarithm scale). As expected, each mode cuts the axis
which also satisfies (2). We note that the field outside S can X/R = 1 approximately at kr = n [14], [15]. Therefore, modes
also be described in terms of six Huygens sources. As indi- of order
cated in [13], three orthogonal incremental electric currents
and three orthogonal incremental magnetic currents can be n = 6krmin@ (6)
linearly combined to provide a Huygens source pointing in
any direction. are reactive for r < rmin. Here, 6 $ @ denotes the floor function,
i.e., the function that gives the largest integer less than or equal
DISCRETE NDoF FOR ANY MINIMUM SPHERE to its argument. The curves in Figure 2(a) are valid for spherical
In the intermediate region between small and large minimum TE and TM modes, representing the two orthogonal polariza-
spheres, rigorous formulas do not exist for N DoF or D max . We tions of the outward propagating spherical modes.
now develop a general formula for the N DoF with the objective Each radial order n has 2n + 1 TE azimuthal spherical
to apply it in the section “Continuous N DoF Function for Any modes, related to variations of the forms sin m{ and cos m{,
Minimum Sphere” for estimating D max for any antenna size. for m = 0, 1, f, N [14], [15]. They are all orthogonal to each
Doing this, we presume that the factor 2 between N DoF and other and to the corresponding TM azimuthal modes. The
D max is valid for all sizes. total number of spherical modes from n = 1 to n = N there-
We base the formulation on counting those spherical fore becomes
modes over the minimum sphere for which the radiated power

N DoF = 2 / (2n + 1) = 2 6N 2 + 2N@; N = 6krmin@ .(7)


is larger than the reactive power. We do this by computing the N

complex wave impedances Z n = R n + jX n of the nth radial- n =1
indexed spherical modes to see whether the resistive part
R n is larger than the imaginary part X n, i.e., to determine It is only the order m = 1 of these modes that can con-
when the spherical modes change their nature from being tribute to the directivity along the z axis, because all the

IEEE Antennas & Propagation Magazine August 2017 19


modes with m ! 1 have zero value on the z axis. Therefore, CONTINUOUS NDoF FUNCTION FOR ANY MINIMUM SPHERE
the number of modes needed to generate a directive beam The definition in (8) gives a stepwise function that is not very
along the z axis of arbitrary linear polarization is 2n. Still, useful for intermediate-to-small minimum spheres. N DoF /2
N DoF is given by (7) if we want the directive beam to be should be related to D max, and D max must be a continuous
formed in any direction and for any ­polarization, which is the function of rmin because of the existing antenna literature.
background of the directivity formula in (2). This is also expected from a theoretical perspective, because
Introducing the condition in (6) into (7) and using (5) for the the spherical modes have a gradual cutoff. Therefore, we try to
small sphere case provides the stepwise function smooth the stepwise function in (8) by using the gradual cutoff
of the spherical modes. To that end, we introduce the following
6 for krmin 1 1 continuous effective N DoF:
N DoF = )
2 6(N) 2 + 2 (N)@ for krmin 2 1
, (8)

N DoF = 2 / ) (2n + 1) 3, (9)


N 1 + Q 1 (krmin)

n =1
1 + Q n (krmin)
where N = 6krmin@ . This stepwise function is illustrated as a
directivity using (2) in Figure 2(b) with some other curves that where N = 6krmin@ and Q n (kr) = X n /R n are the functions plot-
are discussed in the following sections. ted in Figure 2. The function in (9) is a continuous function of

1,000

100

10
Q = |X /R |

1 n=2 n=5 n = 10
n=1
0.1

0.01

0.001
2 3 4 5 6 7 8 9 2
1 10
kr
(a)

2
Large Sphere Limit
Q-..N = 10
100 9
8
7 Staircase Curve
6
5
(Discrete Mode Counting) Q-Smoothened,
4 N=5
NDoF/2

3
2

Q-Smoothened,
10 9 Heuristic Combination N=2
8
7
of Small and Large
6 Sphere Curves
5
4
Small Sphere Limit
3

2
3 4 5 6 7 8 9 2 3 4 5 6 7 8 9 2
1 10
krmin
(b)

Figure 2. (a) The ratio between the imaginary and real parts of the wave impedance of spherical modes of different order n.
(b) The DoF of the field by different approaches, showing N DoF /2 as representing the directivity.

20 August 2017 IEEE Antennas & Propagation Magazine


rmin, obtained by weighting each term inside the summation in a directivity of 1.8 dBi for any length up to 0.2m, and, for longer
(7) by the factor lengths, it approaches the directivity 2.2 dBi of the half-wave-
length dipole, which is also included. Practical small antennas
1 + Q 1 (krmin)
= Rn R1 + X1 . can have higher directivities than these values, but it is difficult
1 + Q n (krmin) R1 Rn + Xn
to find published data on this, because the focus in the design of
The physics behind this lies in the following qualitative argu- small antennas is on the bandwidth. The radiation patterns are
ment: we need to excite all N DoF /2 modes of different order normally quite irregular, and both the directivity and its direc-
equally to construct an antenna with the maximum directivity, tion normally change within their bandwidths. Therefore, we
and it is very difficult to excite the modes with high Q, i.e., those have no such examples in the table.
dominated by reactive power. This difficulty lies in the very We have gone through [18] and found a two-turn helix
narrow bandwidth of these high-Q excitations being present at antenna and two short backfire antennas with high directivity.
the same time as the excitations of the low-Q modes. Therefore, They have directivities that are between 2 and 1 dB lower than
the high-Q excitations are not stable over sufficient bandwidth the heuristic limiting curve; the latter number is also valid for
and can therefore not be excited in practice. The difficulty lies the short backfire antenna in [19]. The backfire antennas get
in exciting high-Q modes together with low-Q modes. high directivity by benefitting from strong multiple reflections
Figure 2(b) shows the continuous N DoF curves in (9) for between one large and one small reflecting plate. The short
three different values of N and that they have the characteristic backfire antennas in [20] and [21] have been optimized for large
of averaging with continuity the staircase curve in (8). For large directivity, and the resulting directivities are 0.8 and 0.14 dB
N, they also follow the asymptotic continuous curve in (3), and below the heuristic limit, respectively. However, the best one in
for small spheres, they approach three. Figure 2(b) also shows [21] has an impractical waveguide feeding structure that is not
the heuristic curve presented in the next section. included in the reported size.
The resonant reflector in [22] also uses multiple reflections
COMPARISON WITH HEURISTIC CURVE to increase the directivity, but here the multiple reflections are
Kildal et al. presented in [16] and [17] a formula that can be used weaker, and they appear between the paraboloidal reflector
for all antenna sizes obtained by combining the small and large and its feed. The resulting directivity is about 1.5 dB below the
sphere directivity formulas in a heuristic manner, i.e., limiting curve.
The Eleven antenna consists of two parallel log-periodic arrays
D max = ^ krminh2 + 3.(10) of dipoles over a small ground plane. It has a directivity of 11 dBi
at its lowest frequency, and this is very close to the limiting curve
This is also plotted in Figure 2(b), which shows that the Q when the ground plane is chosen as small as possible [23], [24].
-smoothened curves for large enough N are a bit lower than
(10) when 0.5 < krmin< 2.0 and a bit larger when krmin< 2. These SLOTS AND DIPOLES ON GROUND PLANES
correspond to maximum directivity deviations of 0.6 dB when A short cavity-backed slot radiates similar to a magnetic current
0.5 < krmin< 2.0 and −0.5 dB when krmin> 2. when it is located on a large ground plane, but only on one side
Thus, the heuristic formula (10) is more useful in practice of the ground plane. Therefore, the directivity in the vertical
than (9), which takes time to evaluate because it is based on the direction is twice as large as the directivity D = 3 = 4.8 dBi of
rather complex expressions for the fields of spherical modes. a short magnetic current in free space. However, this directiv-
ity degrades fast when the ground plane diameter is reduced
COMPARISON WITH DIRECTIVES OF KNOWN ANTENNAS because of the wide (uniform) E-plane radiation pattern. This
We choose to compare the presented models for the directiv- causes a lot of diffraction to the back side of the ground plane
ity bound with the directivities of some known antennas, in for small diameters and, thereby, reduced directivity.
particular, those which are claimed to have high directivity or A short vertical monopole on an infinite ground plane has
even to be superdirective. The results are gathered in Table 1 4.8 dBi directivity along the ground plane for the same reason
and Figure 3. The table shows some details of the dimensions as the slot. However, this also degrades fast with the ground
that are the basis for determining the radius rmin of the mini- plane size reduction.
mum sphere. We have chosen to present the curves in Figure 3 It is also possible to make small antennas on large ground
versus the diameter 2rmin of the sphere and the directivity in planes that have higher directivities than 4.8 dBi. For example,
decibels relative to an isotropic antenna to make them easier a small horizontal dipole at very small heights above a large
to use in practice. We have only included cases for which the ground plane has a theoretical directivity that approaches
conceptual antenna solutions have been verified with measure- 8.8 dBi (see, e.g., [1, Sec. 5.1.13]). This is large, but it has a very
ments; in a couple of cases, the published verifications were reactive impedance, so it is difficult to match. The directiv-
done for slightly different dimensions. ity is 3 dB lower, i.e., 5.8 dBi, if we replace the ground plane
by an opposite-fed dipole, coinciding with the location of the
SMALL- AND INTERMEDIATE-SIZED ANTENNAS dipole’s image when the ground plane was present, as seen from
We have included the directivity of the basic short dipole in Table 1. This directivity is then 0.9 dB larger than the heuristic
Table 1 and Figure 3 as a reference value. The short dipole has limit if the separation between the two dipoles is 0.1m. This

IEEE Antennas & Propagation Magazine August 2017 21


22
Table 1. The different small antennas and antennas with directivities close to the heuristic limit.
Dimensions Obtained from Reference,
Converted in Wavelength, If Needed Diameter GP-Corrected Values Calculated Values
of the
Width (Dy) Height (DZ) Minimum Directivity
Width or (0 Means Vertical to GP Sphere in Computed Obtained Heuristic
Diameter Circualar (0 Means not Ground Plane Wavelength Directivity from Ref Directivity Difference
Antenna Type Reference (Dx) Shape**) Relevant) (GP) hs (D/m) (dBi) Minimum D/m (dBi) Limit (dBi) (dB)
Basic Antenna
Short dipole [1] 0.01 0 0 no GP 0.01 1.76 0.01 1.76 4.77 –3.01
Huygens source [1] 0.01 0.01 0 no GP 0.01 4.77 0.01 4.77 4.77 0.00
Halfwave dipole [1] 0 0 0.5 no GP 0.50 2.16 0.50 2.16 7.38 –5.22
Practical Antennas
Two-turn helical [18, p.574] 1.1 0 0 inherent 1.10 9.80 1.10 9.80 11.74 –1.94
Short backfire [18] 1.5 0 0 inherent 1.50 13.00 1.50 13.00 14.02 –1.02
Large short backfire [18, p.553] 2.2 0 0 inherent 2.20 15.50 2.20 15.50 17.06 –1.56
Dipole-fed short backfire [19] 2 0 0 inherent 2.00 15.20 2.00 15.20 16.28 –1.08
Short backfire with hard wall [20] 2 0 0 inherent 2.00 2.00 15.50 16.28 –0.78
Waveguide-fed short backfire [21] 2.51 0 0 inherent 2.51 18.00 2.51 18.00 18.14 –0.14
Resonant reflector (circular) [22] 3.1 0 0 3.10 18.27 3.10 18.27 19.91 –1.64

August 2017
Resonant reflector (circular) [22] 4.3 0 0 4.30 21.31 4.30 21.31 22.68 –1.37
Short horizontal dipole over GP [1, Fig. 5.15] 0.01 0 0.05 infinite GP* 0.10 5.80 0.10 5.80 4.91 0.89
Eleven antenna with finite GP [23]–[24] 0.7 0.7 0.15 inherent 1.00 11.00 1.00 11.00 11.10 –0.10
Folded monopole Yagi on GP [28] 0.02 0.04 0.24 infinite GP* 0.48 7.42 0.48 7.42 7.24 0.18
Electrically small Yagi [29] 0.065 0.095 0.095 infinite GP* 0.22 8.81 0.22 5.80 5.42 0.38
Magnetic dipole with director [30, Fig. 6] 2 0 0.52 inherent 2.07 9.2 2.07 9.2 16.55 –7.35
Same as above, simulated only [30, Fig. 6] 0 0 0.52 infinite GP* 1.04 8 1.04 8.00 11.36 –3.36
Superdirective Arrays
Two monopoles [26, Fig. 5] 0.2 0 0.25 infinite GP* 0.54 9.60 0.54 6.60 7.68 –1.08
Same as above, simulated only [26, Fig. 5] 0.05 0 0.25 infinite GP* 0.50 10.40 0.50 7.40 7.40 0.00
(Continued)

IEEE Antennas & Propagation Magazine


configuration is the background of the superdirective monopole
arrays in the next section.

Directivity Difference
Calculated Values

0.28
1.03
1.74
SUPERDIRECTIVE ANTENNAS

Limit (dBi) (dB)


Table 1. The different small antennas and antennas with directivities close to the heuristic limit (Continued).

Hansen has introduced the following conservative definition of

not valid
Heuristic superdirectivity [25]:
A useful operational definition of antenna superdirectivity

5.26
5.97
5.91
(formerly called supergain) is directivity higher than that
obtained with the same antenna configuration uniformly
Directivity
Obtained
from Ref

excited (constant amplitude and linear phase). Excessive

6.02
6.76
7.00
7.00
(dBi)

array superdirectivity inflicts major problems in low


­radiation resistance (hence low efficiency), sensitive exci-
Minimum D/m

tation and position tolerances, and narrow bandwidth.


GP-Corrected Values

Superdirectivity applies in principle to arrays of isotropic


­elements although, of course, actual antenna arrays are
*Infinite GP: The measured or simulated directivity is reduced here by 3 dB and the antenna height is doubled when calculating the diameter of the minimum sphere.
0.31
0.19

0.3

**Circular shape: The geometry of the antenna in the xy plane is circular, so the diameter is determined by its extent only. For rectangular shapes, we need to know
composed of nonisotropic elements.
This definition is of theoretical interest, but it also indicates
Wavelength Directivity
Computed

a series of practical problems. The theoretical directivities


10.00

6.76
7.00
(dBi)

reported in [25, Tables I and II] are much lower than the limit-
ing curves presented in this article, because the simulations in
[25] were performed using a one-dimensional array.
Minimum
Diameter

Sphere in

hs (D/m)

It is possible to achieve an endfire directivity of 5.8 dBi from


of the

0.31
0.19

0.3

two small dipoles in a closely spaced array when they are fed
180° out of phase. (This corresponds to the horizontal dipole
infinite GP*
infinite GP*
Ground Plane

over ground described in the “Short Horizontal Dipole over


inherent

periodic

GP” line under the “Antenna Type” column in Table 1.) By


adjusting the phase of one element slightly away from 180°, the
(GP)

array becomes superdirective and can theoretically achieve a


directivity of 7.2 dB over a very narrow bandwidth. These direc-
Vertical to GP
(0 Means not
Height (DZ)
Dimensions Obtained from Reference,

tivities are larger than the heuristic limit in (10). Such antennas
Relevant)
Converted in Wavelength, If Needed

0.043

small

have been experimentally verified by using vertical monopoles


0.22
0.15

mounted on a large ground plane and radiating along the


ground plane, but the reported measured directivity is, at most,
Width (Dy)

1.7 dB over the heuristic limit, as described in more detail at the


Shape**)
Diameter Circualar
Width or (0 Means

small

end of the next section. This higher directivity appears after the
both the extents in the x and y directions to determine the diameter in the xy plane.
0.19

monopoles have been miniaturized by bending them.

SUPERDIRECTIVE MONOPOLE ARRAYS


small
0.08
0.16
0.17
(Dx)

The two-dipole endfire arrays described previously have been


studied numerically and verified by measurements of mono-
poles over a large ground plane in [26] and [27]. The ground
[27, Fig.11]
[27, Fig.5]

plane makes the monopoles work as dipoles, but the directiv-


Reference

ity is 3 dB higher. Therefore, we have reported in Table 1 the


Two folded cylindrical helix dipole [33]

[32]

equivalent dipole. The full dipoles would be needed in a practi-


cal antenna design, but the monopole makes it simpler to pro-
vide an experimental setup without having to consider baluns
Two monopoles with parasites

and cables that disturb the measurements. We have chosen to


Two bent-wire monopoles

include in Table 1 and Figure 3 the verified straight monopole


The embedded element

design in [26, Fig. 5], which has a directivity 1.1 dB below the


Large Dense Arrays

heuristic limit. The theoretical value of the same is exactly at the


limiting curve.
We have also included the bent-wire monopole array from
Antenna Type

[27, Fig. 8]. This is much smaller than the previous straight-wire
array, and, hence, it also has a directivity that is 1.7 dB above
the heuristic limit. Similarly, the bent-wire superdirective para-
sitic array in [27, Fig. 11] has a 1 dB higher directivity than the

IEEE Antennas & Propagation Magazine August 2017 23


limiting curve. Thus, these miniaturized solutions have high In [33], an electrically small, two-element Yagi antenna is
directivity relative to their sizes and 1.7 dB above the limiting presented. Low- Q, folded cylindrical helix dipoles are used for
heuristic curve. Therefore, our formulation does not hold for the driver and the reflector element for enhanced bandwidth
superresonant ­antennas. performance. The antenna has krmin = 0.95, and the measured
realized gain in the endfire direction is found to be 6.19 dB with
SUPERDIRECTIVE YAGI-TYPE ANTENNAS a significant bandwidth of 2.82% (a 1-dB-drop-in-realized gain).
The old-style Yagi antennas are normally considered to be This corresponds to a value 0.28 dB above the heuristic limit.
superdirective, although not according to Hansen’s definition
previously given, because they have only one element excited. CONCLUSIONS
Also, they only have directivities close to our limiting curve We have formulated a bound for the directivity of nonsuper-
when they are short, because they only exploit extents in one reactive antennas of arbitrary dimensions being half the DoF
direction, i.e., endfire. The directivity of  the closely spaced of the field. This is based on the observation that, in the
folded Yagi antenna in [28] is 0.1 dB above the  limit, and the asymptotic limit for small and large antennas, there is a factor
electrically small Yagi in [29] is 0.4 dB above the limit. Both of 2 difference between the directivity limit and the DoF. We
have been verified by measurements but only for monopole- argue that this factor 2 must hold for any antenna size. The
Yagis over a large ground plane, so it is not ­certain that they DoF is related to the number of radiating spherical modes
work when realized with dipoles due to practical ­factors such as distributed over the minimum sphere. When searching for a
the required balun and the mechanical support structure. way of counting these modes, we have faced the difficulty that
The antenna in [30] is claimed to be superdirective. It has the DoF formula is associated with natural numbers, while we
a directivity vertical to the large ground that is theoretically have searched for a continuous bound for the directivity as a
11 dBi over an infinite ground plane and measured 9.2 dBi with function of the antenna size. This difficulty has been overcome
a finite ground plane. The performance is based on an image in by observing that the spherical modes for a given order suffer
the ground plane, which reduces the directivity by 3 dB if the from a gradual cutoff versus radius rather than an abrupt one.
ground plane is replaced by an actual symmetric copy of the Therefore, we have introduced a weighting factor when count-
geometry of the antenna. The directivity of this antenna without ing the number of radiating modes, which is inversely propor-
the ground plane is 0.3 dB below the heuristic limit. tional to the radiation-Q of the individual modes. The final
weighted counting expression becomes
a continuous function versus krmin that
asymptotically converges to the known
25
value for large rmin and to the correct
value 6 for small rmin.
Heuristic Formula (Solid Red Curve)
The radiation Q is a measure of how
Resonant Reflector, Large

20 difficult it is to make a certain spherical


Waveguide-Fed Short Backfire mode radiate. A narrow bandwidth, as
Resonant Reflector, Small

an inverse of Q, implies, therefore, that


Short Backfire with Hard Walls
Directivity (dBi)

15 high-order spherical modes cannot be


Short Backfire

excited in a stable way, as they become


very sensitive to geometric tolerances,
Short Backfire

Eleven Antenna
10 and this bandwidth becomes narrower,
The three red dots are
inversely proportional to rmin.
Two-Turn Helix

different superdirective
Folded Yagi The new continuous bound for the
monopole arrays.
Electrically Small Yagi directivity is very close to the heuristic
5
Huygens Source 4.8 dBi formula in [16] and [17] with a maxi-
Short Halfwave Dipole mum deviation of 0.7 dB. The compari-
Dipole son with known antenna directivities
0 shows that the maximum directiv-
2 3 4 5 6 7 89 2 3 4 5 6 7 89 2 3 4 5
0.01 0.1 1 ity formula (both the DoF-based for-
Diameter of Smallest Surrounding Sphere Including Ground Plane (λ ) mula and the heuristic one) provides
a representative practical limitation.
Figure 3. The fundamental directivity limitations of antennas presented as a function Short backfire-type antennas includ-
of the diameter of the smallest sphere that can enclose the antenna, including its ing resonant reflector antennas show
ground plane (if any). The dotted steep red line is the asymptote for large antennas, directivities very close to the maximum
the dotted horizontal red line is the asymptote for small antennas, and the solid red limit. The directivity of two parallel
line is the heuristic combination of these to a curve valid for any antenna size. The
black squares are the directivities of some basic and known practical antennas. The dipoles over a small ground plane is
red circles represent superdirective monopole-array antenna solutions that have also quite close to the limit. Small Yagi
been verified by measurements. antennas can be optimized for slightly

24 August 2017 IEEE Antennas & Propagation Magazine


higher directivities than the presented maximum, which [5] R. F. Harrington, “Antenna excitation for maximum gain,” IEEE Trans.
Antennas Propagat., vol. AP-13, no. 6, pp. 896–903, Nov. 1965.
is expected. [6] R. F. Harrington, “On gain and beamwidth of directional antennas,” IRE
Superdirective two-element arrays of vertical monopoles Trans. Antennas Propag., pp. 219–225, July 1958.
have experimentally verified directivities that are up to 2 dB [7] L. J. Chu, “Physical limitations of omni-directional antennas,” J. Appl. Phys.,
vol. 199, pp. 1163–1175, 1948.
higher than the heuristic maximum when the effect of the [8] R. E. Collin and S. Rothschild, “Evaluation of antenna Q,” IEEE Trans.
ground plane is removed. This removal means that the mono- Antennas Propag., vol. 12, no. 1, pp. 23–27, 1964.
poles are considered to be dipoles in free space. The excitations [9] D. Nyberg, P.-S. Kildal, and J. Carlsson, “Effects of intrinsic radiation Q on
mismatch factor of three types of small antennas: single-resonance, gradual-
of the two dipoles to obtain the superdirectivity are also very transition and cascaded-resonance types,” IET Microwave Antennas Propag.,
sensitive to phase errors. vol. 4, no. 1, pp. 83–90, Jan. 2010.
We are missing practical small antenna examples with high [10] O. Bucci, “Computational complexity in the solution of large antenna and scat-
tering problems,” Radio Sci., vol. 40, pp. RS6S16, 2005. doi:10.1029/2004RS003196.
directivities close to the heuristic limit described by the 4.8 dBi [11] E. Martini and S. Maci, “Generation of complex source point expansions
directivity of the Huygens source because most publications on from radiation integrals,” Progress Electromagnetic Res., vol. 152, pp. 17–31, 2015.
small antennas focus on radiation efficiency and bandwidth of [12] E. Martini and S. Maci, “A closed-form conversion from spherical-wave-
to complex-point-source-expansion,” Radio Sci., vol. 46, pp. RS0E22, 2011.
the input matching. When radiation patterns are presented, they doi:10.1029/2011RS004665.
normally do not give the directivity, because both the directivity [13] P.-S. Kildal and X. Chen, “Fundamental Limitations on Small Multi-Beam
and its direction normally vary significantly within the ­bandwidth. Antennas for MIMO Systems,” in Proc. 8th European Conf. Antennas and Prop-
agation (EuCAP 2014), The Hague, The Netherlands, Apr. 2014. pp. 329–331.
It is important to note that the heuristic limit is a practi- [14] R. F. Harrington, Time-Harmonic Electromagnetic Fields. New York:
cal one that has been well rooted through the current article McGraw-Hill, 1961.
in spherical mode theory. It is difficult, but not theoretically [15] C. A. Balanis, Advanced Engineering Electromagnetics. New York Wiley, 1989.
[16] P. S. Kildal, “Fundamental Directivity and Efficiency Limitations of Single-
impossible, to realize antennas with higher directivity than the and Multi-Port Antennas,” Proc. 2nd European Conf. Antennas and Propaga-
limit, in particular, if the diameter of the minimum surround- tion (EuCAP) 2007, Edinburgh, Scotland.
ing sphere is smaller than 1 m. For larger antennas, the possible [17] P.-S. Kildal and S. R. Best, “Further investigations of fundamental directiv-
ity limitations of small antennas with and without ground planes,” in Proc. IEEE
improvement is very small, and the bandwidth of the improve- Int. Symp. Antennas and Propagation (IEEE AP-S), San Diego, CA, July 2008.
ment reduce inversely proportional to krmin. [18] K. Fujimoto and J. R. James, Mobile Antenna Systems Handbook, 2nd ed.
Norwood, MA: Artech House, 2001.
[19] H. W. Ehrenspeck, “The short-backfire antenna,” Proc. IEEE, vol. 53, no.
ACKNOWLEDGMENTS 8, pp. 1138–1140, Aug. 1965.
The first version of this article [16] was presented in a special [20] M. G. Bray and E. Lier, “High efficiency short backfire antenna using elec-
session at the 2017 European Conference on Antennas and tromagnetically hard walls,” IEEE Antennas Wireless Propag. Lett., vol. 14, pp.
1614–1617, Apr. 2015.
Propagation, “J.R. James Memorial Session Small, Mobile, [21] D. P. Gray and L. Shafai, “Optimisation of large diameter short backfire
and Printed Antennas.” The article created some discussion antenna by cavity juncture curvature,” Electron. Lett., vol. 37, no. 11, pp. 675–676,
with Steven R. Best about the validity of the maximum 4.8 dBi May 24, 2001.
[22] P.-S. Kildal, “A small dipole-fed resonant reflector antenna with high
directivity of small antennas. However, he later confirmed this efficiency, low cross polarization and low sidelobes,” IEEE Trans. Antennas
and further contributed with more examples published in [17]. Propag., vol. AP-33, no. 12, pp. 1386–1391, Dec. 1985.
He has also helped to provide additional references on super- [23] R. Olsson, P.-S. Kildal, and S. Weinreb, “The Eleven antenna: A compact
low-profile decade bandwidth dual polarized feed for reflector antennas,” IEEE
directive antennas for this article, for which we sincerely thank Trans. Antennas Propag., vol. 54, no. 2, pt. 1, pp. 368–375, Feb. 2006.
him. This work has been supported in part by the European [24] J. Yang, M. Pantaleev, P.-S. Kildal, B. Klein, Y. Karandikar, L. Helldner, N.
Research Council (ERC) via an advanced investigator grant Wadefalk, and C. Beaudoin, “Cryogenic 2–13 GHz Eleven feed for reflector anten-
nas in future wideband radio telescopes,” IEEE Trans. Antennas Propag., vol. 59,
ERC-2012-ADG_20120216 to P.-S. Kildal and by the Swed- no. 6, pp. 1918–1934, June 2011.
ish Governmental Agency for Innovation Systems (VINNOVA) [25] R. C. Hansen, “Fundamental limitations in antennas,” Proc. IEEE, vol. 69,
within the VINN Excellence Center Chase at Chalmers. no. 2, pp. 170–182, Feb. 1981.
[26] E. E. Altshuler, T. H. O’Donnell A. D. Yaghjian, and S. R. Best, “A mono-
pole superdirective array,” IEEE Trans. Antennas Propag., vol. 53, no. 8, pp.
AUTHOR INFORMATION 2653–2661, Aug. 2005.
Per-Simon Kildal was with the Signals and Systems Department [27] A. D. Yaghjian, T. H. O’Donnell, E. E. Altshuler, and S. R. Best, “Electri-
cally small supergain end-fire arrays,” Radio Sci., vol. 43, p. RS3002, 2008. doi:
at Chalmers University of Technology, Gothenburg, Sweden. He 10.1029/2007RS003747.
passed away on 21 April 2016. He was a Fellow of the IEEE. [28] S. Lim and H. Ling, “Design of a closely spaced, folded Yagi antenna,”
Enrica Martini (martini@dii.unisi.it) is with the University IEEE Antennas Wireless Propag. Lett., vol. 5, no. 1, pp. 302–305, 2006.
[29] S. Lim and H. Ling, “Design of electrically small Yagi,” Electron. Lett., vol.
of Siena, Italy. She is a Senior Member of the IEEE. 43, no. 5, Mar. 1, 2007.
Stefano Maci (macis@dii.unisi.it) is with the University of [30] O. S. Kim, S. Pivnenko, and O. Breinbjerg, “Superdirective magnetic dipole
Siena, Italy. He is a Fellow of the IEEE. array as a first-order probe for spherical near-field antenna measurements,”
IEEE Trans. Antennas Propag., vol. 60, no. 10, pp. 4670–4676, Oct. 2012.
[31] P. W. Hannan, “The element-gain paradox for a phased-array antenna,”
REFERENCES IEEE Trans. Antennas Propag., vol. AP-12, no. 7, pp. 423–433, July 1964.
[1] P.-S. Kildal, Foundations of Antenna Engineering: A Unified Approach for [32] P.-S. Kildal, A. Vosoogh and S. Maci, “Fundamental directivity limitations
Line-of-Sight and Multipath. Kildal Antenn AB: Gothenburg, Sweden, 2015. of dense array antennas: A numerical study using hannan’s embedded element
[2] W. Rudin, Principles of Mathematical Analysis. New York: McGraw-Hill, 1964. efficiency,” IEEE Antennas Wireless Propag. Lett., Aug. 2015.
[3] O. Bucci and G. Franceschetti, “On the degrees of freedom of scattered [33] I.-J. Yoon and H. Ling, “An electrically small Yagi antenna with enhanced
fields,” IEEE Trans. Antennas Propag., vol. 37, no. 7, pp. 918–926, 1989. bandwidth characteristics using folded cylindrical helix dipoles,” Microwave
[4] A. D. Yaghjian, “Sampling criteria for resonant antennas and scatterers,” J. Opt.Tech. Lett., vol. 53, no. 6, June 2011.
Appl. Physics, vol. 79, no. 10, pp. 7474–7482, 1996. 

IEEE Antennas & Propagation Magazine August 2017 25

S-ar putea să vă placă și