EuCAP 2011 - Convened Papers

Beam-Waveguide Analysis Using

Complex Conical Beams
S. Skokic+, M. Casaletti#, S. Maci# and S. Sorensen*
+
Faculty of Electrical Engineering and Computing, Unska 3,
HR-10000 Zagreb, Croatia
sinisa.skokic@fer.hr
#
Department of Information Engineering University of Siena,
Via Roma, 56 - 53100- Siena, Italy
{casaletti, macis}@unisi.it
*
TICRA
Læderstræde 34, DK-1201 Copenhagen K, Denmark
sbs@ticra.com

Abstract— This paper presents a method of analysing beam the electro-magnetic wave presents significant challenges both
waveguide systems, based on Physical Optics and Complex from practical and analytical aspect.
Conical Beams. The analysis procedure for one reflector consists
of calculating the field spectrum in the auxiliary input/output
plane, expanding the field (spectrum) into complex conical
beams, propagating the beams to the subsequent reflector and
calculating the reflected fields in the next auxiliary plane via
Physical Optics. The method is demonstrated on a two-reflector
system. Optimal position of the auxiliary plane and efficient
calculation of the field spectrum are discussed. Huygens’ source
is employed as excitation, in order to generate physically correct
incident fields with a Gaussian taper. The results show good
agreement with the reference (pure Physical Optics) solution.

I. INTRODUCTION
The main characteristic of the quasi-optical frequency
range, spanning from the upper end of the microwave range
and all the way up to terahertz frequencies, is that the devices
interacting with the electromagnetic wave are much larger Fig.1: Beam waveguide system for conducting the electromagnetic wave from
the generator to the antenna.
than a wavelength, but still not large enough to be able to take
full advantage of the simplifications that optical systems and
devices benefit from. The term "simplifications" here refers The analysis of reflector antenna systems at millimetre-
both to the analysis of such systems and the electric and wave frequencies requires the development of new methods
magnetic properties of matter. For example, at optical particularly suited for that frequency and size range. The
frequencies there is almost no interaction between the analysis methods usually seek to express the radiated field in
electromagnetic wave and the molecules of the optical fibre terms of a relatively low number of wave objects of higher
core, resulting in very low losses of such structures. Similarly, complexity and in this way reduce the number of unknowns in
at microwave frequencies, electro-magnetic waves can be the system. The aim is to develop a modular process, whereby
conducted with minimal losses using waveguides, owing to the total reflected field from one reflector can be re-expanded
the very low losses in waveguide walls. into a new sum of wave objects of the same kind, and the
In the quasi-optical region, on the other hand, neither of whole procedure can be repeated in the same way for all
the two conditions is satisfied, and the losses in both metallic subsequent reflectors. Recently, we introduced a new beam
and dielectric guiding structures can be as high as 0.5 dB/cm formulation [2]-[3] and introduced the basic steps for
[1]. Therefore, the idea is to conduct the electromagnetic wave analysing reflection by a single antenna element [4]. In this
from the generator to the main radiating element through air. paper, we round up the analysis and demonstrate the
To ensure minimal leakage of energy (i.e. minimal losses), the application of the developed method to a system of two large
wave is directed in space and kept focused by a sequence of ellipsoidal reflectors illuminated by a Huygens’ source. The
curved, shaped metallic reflectors, called a beam waveguide, formulation and some preliminary results are shown in the
as shown in Fig. 1. However, such a solution for conducting following sections.

4106
 II. FORMULATION In the test cases, the first reflector was illuminated by a
The geometry of the analysed two-reflector system is Huygens’ source. The Huygens' source consists of an
shown in Fig. 2. For simplicity, both analysed antennas are electrical x-directed short dipole (a Hertzian dipole) and a y-
ellipsoidal reflectors. Another reason for working with directed short magnetic dipole, and produces a z-directed
ellipsoidal reflectors is that if the source is placed in the first beam (in local coordinate system). When this source is located
focal point of a reflector, upon reflection the reflected beam at a point with the complex position (x , y, z ) = (0, 0, jb ) , the
will focus in the plane containing the second focal point, beam it generates will have a Gaussian taper, similar to that of
which is optimal in the sense of keeping the beam tight. The real sources employed in quasi-optical reflector systems (e.g.
frequency is 3 GHz, i.e. λ = 0.1 m, and the semi-axes of both corrugated horn antennas). The far-field of a Huygens' source
ellipsoids are a = 5 m = 50λ, b = c = 4 m = 40λ. The first with a complex displacement b is
reflector is illuminated from its geometrical focus F1 by a G e − jkr kb cos θ
complex Huygens source with a beam taper of A = 10 dB and E ( r , θ, φ ) = E 0 e (1+cos θ)(cos φ θˆ−sin φ θˆ) .
r
a beam taper angle of θ = 6°, which corresponds to the beam
(1)
waist w0,inc = 0.326 m = 3.26λ. The incident beam is launched The advantage of a Huygens’ source over simple Gaussian
towards the apex of the first ellipsoid (point (0, 0, 4 m) on the beams lies in the fact that the field it generates is physically
z-axis). correct in both near field and far field, and is co-polar, in all
directions, to the linear x-polarisation as defined according to
z [m]
Ludwig's 3rd definition [8]. If a beam with a taper of A
decibels at an angle θt is desired, the necessary value of
complex coordinate b is calculated from (1) as
F1'
zwaist,2 ≡ F2' (
20 log (1 + cos θt ) 2 − A)
θinc zwaist,1 b = . (2)
20k (1 − cos θt ) log(e )
F1 F2

The propagation of complex conical beams (CCB) has

been described in detail in previous communications [2]-[4].
In order to analyse a realistic system of reflectors, however,
x [m] there are additional problems that need solving. For instance,
one has to determine the optimal position for the input/output
plane for a given system geometry. By input/output plane we
field magnitude on spectrum in the
understand the plane in which the field is translated to CCB’s.
the first reflector auxiliary plane The goal is to do that where the beam is the most focused, so
as to reduce as much as possible the size of the sampling
surface in that plane. Another good consequence of such a
choice of input/output plane is that the computation of the
Physical Optics integral is the fastest there [5], since the
reflected field is most in-phase in the plane where it is focused.
Two possible choices for the input/output plane (“auxiliary
plane” in Fig. 2) are investigated in Sec. 3.
Another important aspect of the system analysis is the
field magnitude on the second efficient calculation of the electric field spectrum in the
reflector (PO + conical beams) input/output plane. Typically, a 2D Fast Fourier Transform is
Fig.2: Side view (top) and perspective view (bottom) of a system of two employed for that purpose. However, it should be noted that
ellipsoidal reflectors with a circular rim. Green lines denote the far-field the standard FFT algorithm samples the field on a rectangular
observation points. grid, and moreover returns the calculated spectral points also
on a rectangular grid. On the other hand, the CCB expansion
The analysis of a beam waveguide using wave-objects requires spectrum samples on a polar grid, which in return
consists of the following basic steps, which must be done for means that either an additional 2D interpolation or a
each reflector (the first reflector may not need step 1 below significant increase in the number of sampled points are
and the analysis of the last reflector obviously stops at point 3): necessary, the former translating into a somewhat lower
1) calculate field spectrum in the input plane accuracy of the algorithm, and the latter into an increase in
2) expand the incident field into complex conical beams calculation time. A better alternative to the classical FFT is
3) compute equivalent currents on the reflector surface the recently introduced Pseudo-Polar FFT algorithm (PPFFT)
using complex conical beams [9], which employs a series of 1-D FFT transformations tor
4) calculate reflected field via Physical Optics calculate the spectral data on a polar grid which is uniform in
radial direction and almost uniform in azimuthal direction.

4107
The third possible way of calculating the spectrum for the are actually geometrically symmetric around the focal point
CCB expansion is the use of inverse Near-Field to Far-Field F2 of the second reflector, which is considered optimal from
transformation (NF-FF) [10], [11]. Here, the reflected field the ray-analysis perspective (because F2 ≡ F1’). Therefore, the
samples are calculated on a spherical surface in the estimated two selected test scenarios allow for a comparison of optimal
far-field of the outgoing beam and the spectrum in the reflector positioning from the beam analysis viewpoint (first
input/output plane is then computed in closed-form. The two case) and the ray-optical analysis viewpoint (second case).
latter approaches (PPFFT and NF-FF) have been considered in Another modification introduced in the second test case is
the numerical verification. an increase in the second reflector surface r0′ = 15λ and a
Other important points of the algorithm include a criterion
for truncating the FFT-GPOF expansion used in the CCB small offset of the second reflector centre (Δx = 5λ; by default,
expansion and efficient handling of the complex geometry of the rim centre is set to coincide with the intersection of the
multiple reflector systems. The latter implies the definition of central ray and the reflector surface, to achieve the most
three coordinate systems at every reflection (global, reflector- symmetric field distribution on its surface). Comparing Figs. 2
based and beam-based) and the development of fast and and 3 (bottom images), one can see that in Fig. 3 the fields on
simple routines for converting field values and vectors the inner part of the second reflector are truncated more
between coordinate systems. abruptly than in Fig. 2. Consequently, for the second test case
one can expect to have more ripples in the radiation pattern in
the φ = 0° cut-plane.
z [m] The results are compared to the reference PO solution
obtained via GRASP SE and are presented in Figs. 4 thru 7. In
all plots, the reference axis for angle θ is the estimated
direction of propagation of the wave upon final reflection,
θinc obtained by ray analysis (shown in Fig. 3).
zwaist,1
F1 F2 ≡ F1' F2'
zwaist,2

x [m]

Fig.4: Radiation pattern (directivity) of the output beam after two reflections
(first test case); φlocal = 0°.

propagation trajectory center of the second reflector

of the central ray (offset by Δx = 5λ)

Fig.3: Side view (top) and perspective view (bottom) of a system of two
ellipsoidal reflectors (second test case).

III. NUMERICAL RESULTS

For numerical tests, we considered two scenarios. The first
test case, illustrated in Fig. 2, was created such that the two
reflectors were symmetrical with respect to the actual focus of
the reflected beam (point zwaist,1 in Fig.2). In the second test
case, illustrated in Fig. 3, the second part of the two-reflector
Fig.5: Radiation pattern (directivity) of the output beam after two reflections.
system was altered, and the z-axis apex of the second ellipsoid (first test case); φlocal = 90°.
was set to the point (6 m, 0, 4 m). It can be noted that with
such a placement of the second ellipsoid, the two ellipsoids

4108
 IV. CONCLUSION
A method of analysing beam waveguide systems based on a
combination of Physical Optics and Complex Conical Beams
has been presented and a two-reflector system has been
analysed. A Huygens’ source was employed as excitation, in
order to generate physically correct incident fields with a
Gaussian taper. Two methods of computing the spectrum in
the plane where the field is expanded into complex conical
beams have been compared, yielding results of similar
accuracy. Overall, the results are comparable to the referent
Physical Optics results, while the computation times have
Fig.6: Radiation pattern (directivity) of the output beam after two reflections been reduced by a factor of ten, proving the applicability of
(second test case); φlocal = 0°. the developed method.

REFERENCES
[1] P. F. Goldsmith, “Quasi-Optical Techniques,” Proc. IEEE, vol. 80, pp.
1729-1747, Nov. 1992.
[2] S. Skokic, M. Casaletti, S.Maci and S. Sørensen “Complex Conical
Beam Expansion for the Analysis of Beam Waveguides”, Proc. 3rd
European Conference on Antennas and Propagation (EUCAP 2009),
Berlin, Germany, 2009.
[3] S. Skokic, M. Casaletti, S.Maci and S. Sørensen “Complex Conical
Beams for Aperture Field representations”, IEEE Trans. Antennas &
Propagation, accepted for publication (2010).
[4] M. Casaletti, S. Skokic, S.Maci and S. Sørensen “Beam Expansion in
Multi-reflector Quasi-Optical Systems”, Proc. 4th European Conference
on Antennas and Propagation (EUCAP 2010), Barcelona, Spain, 2009.
[5] T. Bondo and S. Sørensen “ Physical Optics Analysis of Beam
Waveguides Using Auxiliary Planes”, IEEE Trans. Antennas
Propagat., vol. 53, No. 3, Mar. 2005, pp. 1062-1068
[6] N. J. McEwan and P. F. Goldsmith, “Gaussian Beam Techniques for
Illuminating Reflector Antennas”, IEEE Trans. Antennas &
Fig.7: Radiation pattern (directivity) of the output beam after two reflections. Propagation, vol. 37, No. 3, pp. 297-304, 1989
(second test case); φlocal = 90°. [7] H-T. Chou and P. H. Pathak, “Uniform Asymptotic Solution for
Electromagnetic Reflection and Diffraction of an Arbitrary Gaussian
Beam by a Smooth Surface with an Edge”, Radio Science, vol. 32, no.
The agreement is seen to be very good within a dynamic 4, pp. 1319-1336, 1997.
range of 40 dB, proving the applicability of this analysis [8] K. Pontoppidan (Ed.), GRASP9 Technical Description, TICRA
method. As expected, for the second test case the field Engineering Consultants, Copenhagen, Denmark, 2005.
[9] A. Averbuch, R. R. Coifman, D. L. Donoho, M. Elad and M. Israeli,
distribution has fewer ripples in the φlocal = 90° cut-plane (Fig. "Fast and Accurate Polar Fourier Transform", Appl. Comput. Harmon.
7) than in the φlocal = 0° cut-plane (Fig. 6). The curve in Fig. 7 Anal., vol. 21, p.p.145–167, 2006.
is also smoother than the corresponding field distribution of [10] C. A. Balanis, Antenna Theory – Analysis and Design, 2nd edition, John
Wiley and Sons, NY, USA, 1997.
Fig. 5 (for the first test case), owing to the larger radius of the [11] J. J. H. Wang, "An Examination of the Theory and Practices of Planar
second reflector. Near-Field Measurement", IEEE Trans. Antennas & Propagation, vol.
We also note that the two alternative approaches to 36, no. 6, pp.746-753, 1988.
spectrum calculation, described in Sec. 3, yield results of
similar accuracy. Both approaches have been found to be
significantly faster (approximately 10 times) than the pure
PO-based analysis.

4109