doi: 10.5028/jatm.2011.

03034111

Marcelo Bender Perotoni*

Technological Institute of Aeronautics
Numerical evaluation of an air-
São José dos Campos/SP – Brazil
marcelo.perotoni@ufabc.edu.br to-air missile radar cross section
Luiz Alberto Andrade signature at X-band
Institute of Aeronautics and Space
São José dos Campos/SP – Brazil Abstract: The remote detection of a vehicle requires that some kinds of its
andradelaa@iae.cta.br
emissions are tracked and detected. Usually, electromagnetic emissions
are used in the form of radar (electromagnetic waves in the range of
*author for correspondence radiofrequency and microwaves). Different types of antennas are used as
sensors, tailored to the signal frequency band and its polarization, as well as
to the target distance (higher gain antennas used for low amplitude signals).
For the specific case of radars, the use of computational methods to address
the electromagnetic signature (spatial pattern of the scattered energy from the
object) has become widespread, given the high costs and complex equipment
associated with these respective measurements. Therefore, the use of computer
simulation is ideally suited for creating a realistic database of targets and its
respective signatures. The same computer-created signatures database can
also be used for the thermal range, enabling a complete technology solution
for the signature and design of stealth vehicles, with reduced emissions.
Keywords: Radar cross section, Electromagnetic scattering, Computational
modeling.

INTRODUCTION
90
Radars have become a fundamental tool in the areas of
120 60
defense and homeland security, since the Second World
War (Grant, 2010). Since its inception, several new tools
were added, namely ultra wide band (UWB), processing
150 30
algorithms, digital signal processors, and so on (Kouemou,
2009; Skolnik, 1981).

Although modern techniques rely heavily on signal

processing (software) for increasing the detection 180 0
0 10 20 30
capabilities, the hardware is still a crucial issue. The reason
is that the frequencies used by radars normally require the
use of microwave instrumentation, which presents higher
costs and complexity when compared to low-frequency 210 330
circuitry and equipment.

The spatial waveform shape of the return radar signal

240 300
(echo) is the vehicle radar cross section (RCS) signature.
Every object has a specific signature, which helps to 270
identify what kind of structure is under analysis (for
Figure 1. Simulated result for a bistatic RCS of a glider (length
instance, determining whether it is a friend or foe). 14.6 m, wingspan 13.75 m). The incoming wave is
Figure 1 depicts the radar signature of a generic glider at represented by the arrow (500 MHz). The scale is
0.5 GHz, for a frontal incidence, taken from the examples shown in dBsm (CST, 2010).
supplied with the CST Microwave Studio® (CST, 2010)
package. The electromagnetic energy from the incoming or embarked somewhere. From Fig. 1, it is possible to
wave develops currents along the metallic surface, see that some directions have a higher energy density
which then re-radiate in several directions. That is the than others.
field which hits the receiver radar, located on the ground
The receiver signal can be captured and processed on the
same position where it was transmitted; this is the so-called
monostatic radar signature. Signatures are called bistatic
Received: 13/09/11
Accepted: 08/10/11 when the receivers can be spread in other directions. In

J. Aerosp.Technol. Manag., São José dos Campos, Vol.3, No.3, pp. 287-294, Sep. - Dec., 2011 287
 Perotoni M.B., Andrade L.A.

the case of Fig. 1, a bistatic scenario simulation is shown. Any metallic object illuminated by an incident
If a receiver was placed on the same position of the electromagnetic wave develops along its surface electric
transmitter (180º), it would be seen a target with 1.6 dBsm currents, which in turn re-radiate. The unknown to be
(or 1.45  m2). The unit used is that of a surface (square determined is the current density J(r), which is found as
meters), so that it has a relation to an analogous physical the solution of an integral equation. It is written as a matrix
area, which scatters the same energy. On the other hand, if equation, after the MoM discretization (Davidson, 2005),
the receiver is positioned at the 0 angle, the target will be in which MoM stands for method of moments. The solution
detected with a RCS of around 25 dBsm (316 m2). is achieved in an iterative approach, by methods such as
conjugate gradient, which uses approximately N2 operations
The measurement of the RCS signatures from real targets per iterations, with N equals to the number of unknowns.
(aircraft, tanks, vessels) is a complex and costly task.
For the case of an aircraft, it requires its placement on an The problem of a metallic object subjected to an incident
adequate area, which is normally wide (comparable to the electric field Ei(t) is represented by the electric field
aircraft size). In addition, the microwave instrumentation integral equation (EFIE) (Davidson, 2005):
has to be able to illuminate the object with enough energy in
order that the returned signal can be discriminated against
the environment noise floor. On top of that, measurements t µsG(r , r’) J(r’)dS’ = 4Ui tEi(r) (1)
kM 
done on the ground do not represent a true environment,
since during real flights there is no ground plane. In view
of these complexities, computer simulations have been where,
used to predict and analyze radar signatures. For instance,
the design of stealth vehicles (i.e. vehicles whose RCS t represents an unit tangent vector on the surface S;
signatures are very low when compared to their physical k the wave number;
size) relied on the computer analysis to get a geometric J(r’) the current density unknown;
shape able to scatter the incoming wave, in such a way the η the medium impedance; and
receiver signal is as small as possible (Grant, 2010). i the imaginary term.

This article presents a short overview of the numerical methods The primed r variable regards the source variable and
used in the microwave analysis. Monostatic and bistatic the unprimed r is the observation point variable. G is the
simulated signatures of a real short range, air-to-air missile are Green Function representing the problem, given by Eq. 2
presented. Comparisons with measurements are also shown. (Davidson, 2005):

G(r , r’) = (1- 12 ””r) |r - r’|

eik|r - r’|
METHODS k (2)

Numerical analysis
Analogously, a magnetic field integral equation (MFIE)
To solve the microwave range scattering problems, can also be written. When there are resonances due to
an appropriate solution of the Maxwell Equations is enclosed cavities, numerical problems arise. The solution
sought, subjected to the particular boundary conditions. is the use of a combination of both MFIE and EFIE, named
The numerical solution of those equations involves a Combined Field Integral Equation. That is the reason to
previous step, the discretization, where the object and avoid hollow structures; the missile or aircraft model is
its surrounding volume are sliced into small elements better simulated as made of a solid piece of metal, without
(forming the electromagnetic mesh). Then, the Maxwell hollow parts in their interiors. These equations can be
equations are applied to each of those small elements, numerically solved by the MoM, which basically finds the
whose fields/currents/voltages are determined. solution by a matrix inversion.

Volume meshes are commonly used when the object The Multi-Level Fast Multipole Method (MLFMM)
is electrically small, like most antennas. However, the (Song et al., 1997) is used to further reduce the problem
computation of large-scale models using volume mesh complexity, by making the MoM matrix sparse. It is
methods becomes intractable with even moderate hardware. achieved by the reduction of the coupling to only nearby
The reason is that the meshing of the hollow part of a missile elements through the use of small cubic volumes. Then, the
and the air area around it can be neglected, since the external problem has its number of operations reduced to Nlog(N),
shell is the main responsible for the scattering. For that kind enabling the computation of large scale problems. Figure 2
of application, surface mesh is used instead; only the external illustrates the idea by showing the coupling scheme
2D surface (sheet) is meshed. difference between the MoM and the MLFMM.

288 J. Aerosp.Technol. Manag., São José dos Campos, Vol.3, No.3, pp. 287-294, Sep. - Dec., 2011
Numerical evaluation of an air-to-air missile radar cross section signature at X-band

Figure 2. The left picture shows that in the MoM all elements are allowed to couple to each other, generating a very dense matrix. The
MLFMM, on the right, allows only certain elements to couple, resulting in a sparse matrix.

It is interesting to stress that some methods, like the RESULTS AND DISCUSSION
Finite-Difference-Time-Domain (FDTD), rely on
Bistatic simulations
mathematical operations that fundamentally are simple,
requiring subtractions and sums, but at expenses of The missile here analyzed is named Piranha, which was
large and constant memory accesses. On the other developed by a joint program between the Brazilian Air
hand, MoM and MLFMM require the inversion of a Force and the Navy. It is a short range, air-to-air unit, with
large matrix, mathematically and computationally an infrared seeker (Coelho, 2007). Figure 3 shows the
much more processor (CPU) intensive (Munteanu, model and its main dimensions. For 10 GHz, its electrical
Timm and Weiland, 2010). As the frequency increases size is 95 λ long and 22 λ wide.
even further, turning the electrical size of the problem
too large, then the alternative is the use of asymptotic
methods (Geometrical Optics – GO). A ray tracing
scheme computes the incoming and reflected rays
(Shooting and Bouncing Rays – SBR). Evidently,
there is a tradeoff between the precision and the 0.66 m
computational effort between the two approaches,
MoM/MLFMM, and asymptotic. It can be stated that, 2.85 m
as a general rule, the accuracy increases as we move
from Asymptotic to MLFMM and to MoM, whereas
the simulation speed that requires the random-access
memory (RAM) decreases. Figure 3. Picture showing the missile with its main dimensions.

The SBR method launches a dense grid of finite rays

The 3D complete missile model was imported from a
that hit the object, and later the multiple reflections are mechanical computer aided design file (CAD) (Catia,
computed, taking into consideration the geometry of the 1998) into the workspace of the electromagnetic solver.
problem. This method (Pike and Sabatier, 2002) extends The missile material is considered as being made out of a
the Physical Optics (PO) by defining the surface currents perfect electric conductor (without losses). The boundary
developed on the structure in terms of the computed conditions are set to open space. Figure 4 depicts the
fields. Therefore, it is also possible to map the surface surface mesh obtained along the missile surface. A good
current density, which is important for identifying hot quality mesh (i.e. with homogeneous elements, showing
spots in the aircraft. good aspect ratio and with similar sizes) helps getting a
better and faster simulation. The aspect ratio plays for the
The solver (numerical method used) was the I-solver surface mesh a vital role, meaning that the ratio between
(Integral Equation), based on the MoM/MLFMM the biggest and smallest component of the structure directly
method. The other solver that was employed was the impacts on the mesh quality. The ideal situation is when the
A-solver (Asymptotic), based on the Shooting and aspect ratio is close to one (largest and smallest dimensions
Bouncing Rays method. are similar).

J. Aerosp.Technol. Manag., São José dos Campos, Vol.3, No.3, pp. 287-294, Sep. - Dec., 2011 289
 Perotoni M.B., Andrade L.A.

the mechanical model contains a small detail, which is

approximately smaller than one tenth of the wavelength
(relative to the incoming plane wave), it can be removed
without gross loss of accuracy. The difficulties associated
with the mesh around a small screw or bolt (smaller than one
tenth of wavelength), for instance, do not pay off in terms
of final accuracy. It is therefore simpler if it is eliminated.

The first study regards the bistatic RCS response to a

X-band, 10 GHz signal (frequency where most onboard
radars operate). For that, the incoming plane wave is
assumed to be incident right on the frontal side of the
Figure 4. Surface mesh for the object. Zoomed in is a detailed
area near the stabilizers.
missile (90º in the Fig. 5), with the electric field aligned to
the missile longitudinal axis, as Fig. 5 suggests. The results
Surface cells are computationally more demanding than of the bistatic RCS is also shown in Fig. 5. It can be seen
hexahedral cells (used by time domain/finite-difference that the frontal RCS has a value of -9.6 dBSm. The RCS
time-domain (FDTD) or transmission line method (TLM) considers the absolute power received with the co and
solvers) (Munteanu, Timm and Weiland, 2010). It means cross polarizations included. It means that the absolute
that with hexahedral cells the structure is represented power involves both vertical and horizontal polarizations
by the use of small bricks (the mesh cells), whereas for – actually the square root of both squared components. A
the surface mesh, small planar triangles have to cover real world measurement will have to count on antennas
the structure surface, adjusted to complex details so that able to receive both polarizations. The simulation used the
they accurately represent fine details. Figure 4 shows the
MoM/MLFMM Method, with first order elements. They
surface mesh for the object. It can be seen that the mesh
are triangles with straight sides, and they enable a faster
elements are homogeneous (they have similar sizes), even
in the region of the stabilizers. It greatly improves the simulation in comparison to higher order elements.
convergence of the problem.
Another point of interest is the lateral incidence. Figure 6
Frequently, the mechanical model needs refinements in shows the excitation (the electric field here is orthogonal
order to enable a functioning surface mesh. Refinements to the missile axis), and the respective result. For this
here usually refers to simplifications. For instance, if situation, the lateral RCS has a value of 24 dBsm.

90
120 60

150 30

180 0
-25 -10 5 20

210 330

240 300
270

Figure 5. Computed RCS, units (dBsm). The scenario involves a bistatic response to a frontal excitation (90º in the figure).

290 J. Aerosp.Technol. Manag., São José dos Campos, Vol.3, No.3, pp. 287-294, Sep. - Dec., 2011
Numerical evaluation of an air-to-air missile radar cross section signature at X-band

90
120 60

150 30

180 0
18.91 29.46 40

210 330

240 300
270

Figure 6. Computed RCS, units (dBsm). The scenario involves a bistatic response to a lateral excitation (0º in the figure).

The results presented in Figs. 5 and 6 show that the detection and evasion, increasing the survivability rate
scattered power is higher to the situation where the of the plane under attack.
missile is illuminated laterally (-9.6 for frontal versus
24 dBsm for the lateral case). It is intuitive to see that Another result of the simulation is the identification
the physical area that intersects the incoming wave is of the hot spots, namely the particular points on the
larger for the lateral case, justifying the difference. A surface that concentrate the higher currents when
lateral illumination of an incoming missile is however illuminated by the plane wave. These currents are
preferred for an earlier incoming missile detection. responsible for scattering the energy back to the source.
For the case of an onboard radar which detected an Therefore, if the goal is to minimize the RCS towards
incoming missile, few seconds are left for the detection a stealth vehicle design, those hot spots need to be
and the evasion maneuver. Thus, few decibels of identified and eliminated. The alternative to eliminate
difference in the received signal (related to the RCS or diminish the current density on hot spots is by
parameter dBsm) can enlarge the period between the means of a geometric reshape or by using Radiation

A/m

0.01
0.00838
0.0071
0.006
0.00504
0.00422
0.0035
0.00289
0.00236
0.0019
0.0015
0.00116
0.000866
0.000611
0.000391
0.000201
0
Figure 7. Current distribution caused by a frontal (left) and lateral (right) incidence. The red areas are those where higher amplitudes
are developed due to the incoming plane wave. Since the lateral incidence has a higher overall RCS than the frontal case,
it also develops currents with larger amplitudes.

J. Aerosp.Technol. Manag., São José dos Campos, Vol.3, No.3, pp. 287-294, Sep. - Dec., 2011 291
 Perotoni M.B., Andrade L.A.

Absorbing Materials (RAM) (Grant, 2010). In Peixoto the missile (180º), in 90 steps. For each angle, the RCS
et al. (2011), measurements are presented, which show is computed in that single direction only. The results
that by covering a missile with RAM, the overall RCS also present a comparison between the MoM/MLFMM
is lowered. Figure 7 depicts the hot spots (visualized and the asymptotic solver.
as red areas) for the frontal (Fig. 5) and lateral (Fig. 6)
incidences. The incident plane wave has amplitude of A comparison between the results of both techniques
1 V/m. Since the frontal incidence has an overall lower is shown in Table 1 alongside with some experimental
RCS (the intersection area is much smaller than that results (Peixoto et al., 2011).
of the lateral case), the developed currents are also of
lower amplitude. It is worth mentioning that the measurement setup
showed a dynamic range limitation, i.e. too high-noise
floor (Peixoto et al., 2011). Therefore, only the higher
Monostatic simulations energy peaks of RCS were detected, like for the angles
of 180 and 90 in Table 1. It justifies the differences
A monostatic scenario involves the rotating of the seen for other angles, like 0. Another difference relies
transmitter around the target, i.e., the transmitter on the fact that the warhead is not metallic (it contains
and the receiver are located in the same point. the infrared seeker and other systems, so it needs to be
Computationally, it is a more challenging task, since transparent), whereas the computer model is completely
for every position the electromagnetic environment metallic. It imposes a severe difference especially for
is different, generating a different system matrix, too. the 0º incidence. Other significant difference is the fact
Figure 8 shows that the incident wave is swept from the that the measurement was done in an outdoor facility,
frontal direction (equivalent to 0º) to the rear side of with the presence of the ground, which for 10  GHz

90
120 60
AsymptoticSolver
MoM-MLFMM

150 30

180 0
-60 -40 -20 0 20

Figure 8. Illustration of the monostatic range of simulation and results for both MoM/MLFMM and asymptotic solvers. The
asymptotic solver used was the A-solver in CST Microwave Studio® (CST, 2010), using medium precision.

292 J. Aerosp.Technol. Manag., São José dos Campos, Vol.3, No.3, pp. 287-294, Sep. - Dec., 2011
Numerical evaluation of an air-to-air missile radar cross section signature at X-band

Table 1. Comparison between the performances with two Table 2. Comparison between the performances of the two
different solvers and measurements. different solvers.
Solvers/ RCS (dBsm) Comparison Asymptotic MLFMM
Angle Experimental Simulation time 4 minutes 9 hours
MLFMM Asymptotic
(Peixoto et al., 2011)
Peak RAM used 48 Mbytes 2.8 GBytes
0º -9.62 -3.73 -22.0
30º -15.7 -23.5 CONCLUSIONS
60º -8.22 -14.2 A study concerning the X-band RCS signature of a real
90º 18.4 17.3 17.0 air-to-air missile is presented. Two numerical techniques to
perform the computer simulation are shown, alongside with
120º -8.14 -22.7 results compared to measurements performed in an outdoor
150º -14.6 -16.0 facility. The results showed a reasonable similarity with
measurements, considering that the real world measurement
180º -3.59 1.59 0.00 setup and the missile were not completely similar to the
virtual representation. Since RCS measurements require
a complex and sophisticated setup, which is not always
may behave like a kind of ground plane, whereas available, prediction techniques based on simulation can
the simulation was done in a perfect non-reflective be implemented in order to complement the real world
environment (Peixoto et al., 2011). measurements. The requirements for computer prediction
are the software package and a moderately equipped
Comparisons with measurements involve a difficult task, hardware, alongside with mechanical models of the objects
regarding an accurate object model (not easily available that are free from unnecessary details, but they are also
with correct materials and geometrical details), as well accurate in terms of dimensions and shapes.
as a correct representation of the measurement setup.
Therefore, computer simulations should be faced as a
complement to measurements, enabling a somewhat REFERENCES
easier and less costly alternative.
Catia, 1998, version 5, Retrieved in Oct, 27th 2011, from
The faster simulation of the asymptotic solver results www.3ds.com.
in some angles showing larger differences compared
with the I-solver (for instance 60º and 120º). Further Coelho, L. V., 2007, “Missile Approach Warning System
refinements in the asymptotic solver, like requiring and its application in defense aircraft”, Conference SIGE
higher precision, might better approximate those IX, São José dos Campos, Brazil, in Portuguese.
results in regard of comparisons with the MLFMM.
It is usually assumed that both solvers show similar CST Microwave Studio, version 2010, Retrieved in Oct,
results for regions where a high RCS value is present 27th 2011, from www.cst.com.
(main lobes), whereas minor lobes or nulls may present
considerable differences. The use of more than one Davidson, D.B., 2005, “Computational Electromagnetics
solver comes into play whenever it is necessary to for RF and Microwave Engineering”, Cambridge
cross-check results from within virtual simulations, i.e., University Press, Cambridge, England.
when measurements are not available. If two different
numerical methods with two different mesh types give Grant, R., 2010, “The radar game: understanding stealth
results that are similar (though not absolutely equal), and survivability”, Mitchell Institute Press, Arlington, the
the user can then achieve a certain degree of confidence US.
on the simulation.
Huang, Y., Boyle, K., 2008, “Antennas: from theory to
A comparison showing the performance in terms of practice”, Ed. Wiley, West Sussex, the UK.
required RAM memory and time is shown in Table  2.
The computer used was a Quad Core Opteron, Kouemou, G., 2009, “Radar Technology”, Ed. In-Teh,
2.51 GHz, with 64 GBytes RAM. It is noticeable Vukovar, Croatia.
that the asymptotic solver presents an advantage by
trading the speed for precision (Sadiku, 2001; Huang Munteanu I., Timm M., Weiland T., 2010, “It’s about time”,
and Boyle, 2008). IEEE Microwave Magazine, Vol. 11, No. 2, p. 60-9.

J. Aerosp.Technol. Manag., São José dos Campos, Vol.3, No.3, pp. 287-294, Sep. - Dec., 2011 293
 Perotoni M.B., Andrade L.A.

Peixoto, G.G., Alves, M.A., Orlando, A.J.F., Rezende, Sadiku, M.N., 2001, “Numerical Techniques in
M.C., 2011, “Measurements in an Outdoor Facility and Electromagnetics”, 2nd Edition, CRC Press.
Numerical Simulation of the Radar Cross Section of
Targets at 10 GHz”, Journal of Aerospace Technology Skolnik, M.I., 1981, “Introduction to Radar Systems”,
and Management, São José dos Campos, Vol. 3, No. 1, 2nd Edition, Ed. Mc Graw Hill.
p. 73-8.
Song, J., Lu, C.C., Chu, W.C., 1997, “Multilevel Fast
Pike, E.R., Sabatier P.C., “Scattering”, 2002, Ed. Multipole Algorithmfor Electromagnetic Scattering by
Academic Press, ISBN 0-12-613760-9, London, the Large Complex Objects”, IEEE Transactions on Antennas
UK. and Propagation, Vol. 45, No. 10, p.1488-93.

294 J. Aerosp.Technol. Manag., São José dos Campos, Vol.3, No.3, pp. 287-294, Sep. - Dec., 2011