You are on page 1of 9

2350 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 49, NO.

6, NOVEMBER 2000

Improving the Accuracy of Ray-Tracing Techniques


for Indoor Propagation Modeling
Kate A. Remley, Member, IEEE, Harry R. Anderson, Member, IEEE, and Andreas Weisshar, Senior Member, IEEE

Abstract—Problems with the use of ray-tracing techniques in when the wavelength is small compared to the geometrical
indoor propagation environments are identified, and a new set of dimensions of objects in the propagation environment, the
widely applicable diffraction coefficients is developed. The limita- so-called “high-frequency approximation.” Because of this
tions on the accuracy of the ray-tracing method in indoor propaga-
tion environments are first assessed. The effects of scatterers with inherent high-frequency approximation, the accuracy of the
dimensions approaching the wavelength of operation and of scat- ray-tracing method may depend on, among other things, the
terers with finite conductivity are considered. The accuracy of ray size of local scatterers as compared to the wavelength of
tracing is quantified by comparison to a full-wave simulation tech- operation [6], [7]. This is a particularly important issue in
nique, which combines the finite-difference time-domain method indoor propagation environments, where feature dimensions
with a spatial transformation technique, the Kirchhoff surface in-
tegral formulation. Simulation results demonstrate that when the may be on the order of the wavelength of operation, especially
magnitude and phase of the received signal components are prop- at the lower personal communications system frequencies of
erly accounted for, the ray-tracing solution may be accurate down 900 MHz ( cm) and 1800 MHz ( cm).
to a fraction of a wavelength. A new set of diffraction coefficients A second important aspect of indoor propagation modeling
is presented for calculations involving obstacles with finite conduc- is the accurate determination of received signal strength for
tivity. The new coefficients eliminate an artificial dip in the dif-
fracted field strength, which is often encountered when currently non-line-of-sight (NLOS) diffracted signals. NLOS conditions
available techniques are used. Validation is provided by compar- are frequently encountered in indoor applications such as
ison with full-wave simulations and measurements. Improved ac- wireless local-area network systems. Diffraction modeling
curacy in both the illuminated and shadowed regions is demon- in the ray-tracing method is well established for perfectly
strated. conducting wedges through the uniform theory of diffraction
Index Terms—Diffraction, finite-difference time-domain (UTD) [8], [9]. An extension of the UTD for wedges with
(FDTD) method, Kirchhoff surface integral, propagation mod- finite conductivity has been introduced for a limited range of
eling, ray tracing. applications [10]–[12]. However, many modern ray-tracing
programs use this extension for general calculation of diffracted
I. INTRODUCTION fields, often resulting in an artificial dip in received signal
strength [11], [13].

R AY TRACING [1]–[3] is a commonly used computational


method for site-specific prediction of the radio channel
characteristics of wireless communication systems. The ray-
This paper assesses the accuracy of the ray-tracing method
with regard to multipath conditions commonly found in indoor
environments, including those with scatterers having dimen-
tracing technique inherently provides time delay and angle of
sions approaching the wavelength of operation and those with
arrival information for multipath reception conditions. This is
scatterers of finite conductivity. Techniques are then demon-
important because the wide-bandwidth signals utilized in many
strated to significantly improve the accuracy of the ray-tracing
modern communication systems are particularly sensitive to the
method in situations where these scatterers are found.
detrimental effects of multipath distortion and dispersion, which
In Section II, the accuracy of the ray-tracing technique for
may produce a frequency-selective channel [4]. The ability to
characterizing propagation environments with electrically small
provide a high level of detail in the characterization of multipath
scatterers is assessed. This is accomplished by comparison to a
channels makes the ray-tracing technique particularly attractive
“standard,” that is, a solution of known accuracy. The standard
for indoor propagation environments. However, some aspects of
is generated in this case using a full-wave computational tech-
ray-tracing implementation must be considered carefully when
nique. Results are presented for propagation channel configu-
modeling propagation in indoor environments or the accuracy
rations including those in which both reflection and diffraction
of the method may suffer. This paper addresses some of these
play dominant roles, and those in which diffraction dominates.
issues.
Simulation results demonstrate that when the magnitude and
Ray tracing utilizes the principles of geometrical optics (GO)
phase of the received signal components are properly accounted
[1], [5], where propagation is assumed along discrete ray paths
for, the ray-tracing solution is accurate for feature sizes down to
from transmitter to receiver. In general, this assumption is valid
a fraction of a wavelength. A clear explanation for these results
Manuscript received December 21, 1999; revised June 5, 1999. is given in terms of ray path-length and its correspondence to
K. A. Remley is with the RF Technology Division of the National Institute of the Fresnel zone.
Standards and Technology, Boulder, CO 80303-3328 USA. In Section III, it is shown that a set of heuristic diffraction
H. R. Anderson is with EDX Engineering, Inc., Eugene, OR 97440 USA.
A. Weisshar is with Oregon State University, Corvallis, OR 97331 USA. coefficients commonly used in ray-tracing programs can give
Publisher Item Identifier S 0018-9545(00)10443-8. inaccurate results for certain propagation configurations when
0018–9545/00$10.00 © 2000 IEEE

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 25, 2009 at 10:29 from IEEE Xplore. Restrictions apply.
REMLEY et al.: ACCURACY OF RAY-TRACING TECHNIQUES FOR INDOOR PROPAGATION MODELING 2351

scatterers are made of materials with finite conductivity. A


simple modification of the diffraction coefficients is shown to
greatly enhance the accuracy of ray-tracing in the character-
ization of both line-of-sight and non-line-of-sight diffracted
signals arising from objects with finite conductivity. Section IV
concludes this paper with a summary and conclusion.

II. ACCURACY OF THE RAY-TRACING METHOD FOR


ELECTRICALLY SMALL SCATTERERS
As stated in the introduction, ray tracing is a technique
that utilizes the high-frequency approximation in site-specific
characterization of electromagnetic wave propagation. In other
words, the size of local scatterers relative to the wavelength of
operation may be an important consideration in the accurate
implementation of the ray-tracing method. As feature size
decreases relative to the incident wavelength, limits on the
accuracy of GO emerge [6], [7]. This paper extends these dis-
cussions and relates them directly to requirements for accurate
ray tracing implementation.
In this section, the accuracy of ray tracing for local scatterer
sizes approaching the wavelength of operation is quantified. A
comparison is made between simulation results based on ray
tracing and those provided by a full-wave computational tech- Fig. 1. Typical spatial configuration used in simulations.
nique. The finite-difference time-domain (FDTD) method [14],
[15] provides a very accurate portrait of the received signal
for moderately sized grids [16]. It is, however, computation- equally spaced on either side of the wall, yielding diffraction
ally intensive and, therefore, generally not practical for site-spe- and specular reflection (Section II-B). Then, the receiver is
cific modeling and/or for use with large computational domains. positioned in the incident field shadow region, yielding a
For larger computational domains or to facilitate arbitrary re- diffraction field only (Section II-C).
ceiver placement, the FDTD method may be combined with a 2) The Ray-Tracing Method: The ray-tracing approach used
near-to-near or near-to-far field transformation technique such to find the scattered field in this experiment incorporates an ap-
as the Kirchhoff surface integral formulation [17], [18]. In this proximation called the relative phase method [19], [20]. The rel-
paper, the FDTD method and the combined FDTD/Kirchhoff ative phase method involves vectorially combining the contribu-
technique are used as standards for verification and validation tions from individual scattering elements on a single structure.
of the ray-tracing technique. Historically, this technique has been used to find the radar cross
section (RCS) of a complex scatterer when it is not feasible or
efficient to use a more accurate method, for example, the elec-
A. Description of the Comparative Simulations
tric field integral equation or the magnetic field integral equa-
Comparative simulations designed to illustrate the depen- tion method (e.g., [21] and [22]). The relative phase method is
dence of the ray-tracing method on the feature size of local described in more detail in the Appendix. The use of this tech-
scatters are described. First, the physical configuration of the nique in this paper provides a simple method for approximating
multipath environment used in the comparative simulations the received signal strength from a single, physically small scat-
is given. Then important implementation aspects for both the terer. The ray-tracing program MCS [23] is used to generate the
ray-tracing method and the full-wave method used in these “experimental” data.
simulations are discussed. 3) The FDTD Method: The implementation of the FDTD
1) The Simulation Environment: The simulated multipath method is constrained by the desire to determine the effects
environment consists of a vertical wall offset between transmit of scattering from the front face of the wall only. To mini-
and receive antennas of equal height, as shown in Fig. 1. The mize diffraction over the top of the wall, a grid cell size of
wall is of variable width. To assess the effects of scattering m and m is chosen.
from the vertical faces of the wall, reflections from the floor The cell size of 0.025 m corresponds to approximately 13 at
are neglected. The wall is made sufficiently high and deep to the carrier frequency of 900 MHz utilized in the simulations.
ensure that diffraction over the top and around the back may The overall grid size is 79 91 50. The large -dimension is
be neglected. With this configuration, up to four rays may be required to ensure that there is no diffraction around the back side
generated, corresponding to the direct path, a specular reflection, of the wall for the duration of the simulation. The computational
and two corner diffractions from the vertical edges of the wall. domain is terminated in perfectly-matched layer PML boundary
Two configurations are considered for assessment of scat- conditions [24]. Calibration between the FDTD simulation
tered ray interactions: First, the transmitter and receiver are and the ray-tracing simulation is accomplished by equating the

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 25, 2009 at 10:29 from IEEE Xplore. Restrictions apply.
2352 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 49, NO. 6, NOVEMBER 2000

Fig. 2. Spatial configuration used for consideration of diffraction and specular


reflection. Transmitter and receiver are spaced equally from the center of the
wall. The edges of the wall move laterally in equal increments with respect to
transmitter and receiver.
(a)

(b)
Fig. 4. The (a) magnitude and (b) phase of the scattered field received signal
for the metal wall case shown in Fig. 3. The received signal strength is minimum
when the reflected and diffracted rays are 180 out of phase (reflector size !
Fig. 3. A comparison of the scattered field signal strength found using the 
0) and is maximum when they are in phase (reflector size 4), as shown by
ray-tracing method and the FDTD method for a metal wall. The ray-tracing the solid line in (b).
solution is calculated with and without the diffracted rays. Signal levels are
referenced to the direct ray.
Theoretically, as the width of the wall increases, the curves in
amplitude of the received direct path signal generated with each Fig. 3 should converge to the value of the specularly reflected
technique. signal. This is because the diffracted rays each have a signifi-
The excitation used in the FDTD method is a pulse-modu- cantly longer path length than the reflected ray, so their contri-
lated carrier with Gaussian pulse defined by bution will be negligible for wide wall widths.
Fig. 3 indicates that the ray-tracing technique under the rela-
(1) tive phase method approximation provides an accurate represen-
tation of the scattered field strength, even when the dimension of
where ns defines the pulse width, and ns is the scattering wall is a fraction of a wavelength. Since the spec-
a time offset. The pulse width has been selected to be narrow ularly reflected ray in the ray-tracing method is independent of
enough for accurate resolution of the scattered pulse when com- scatterer size, it can be concluded that as feature size is reduced
pared to the direct received pulse, yet wide enough to allow co- the contribution from the diffracted rays becomes increasingly
incident illumination of the wall and associated corners. Pulse important in providing an accurate total received field solution.
spreading of the scattered pulse due to the separate effects of Fig. 4(a) and (b) shows the components of the ray-tracing
reflection and diffraction is thus minimized. signal plotted on a linear scale. For smaller reflector sizes, the
magnitude of the two diffracted rays combined using the rela-
B. Case 1: Diffraction and Specular Reflection tive phase method is significant, as shown in Fig. 4(a). The solid
To assess the accuracy of ray tracing in simulations that in- curve in Fig. 4(b) shows that the diffracted and reflected signals
volve both reflected and diffracted rays, the transmitter and re- are out of phase for small wall sizes, so the total received signal
ceiver are located equal distances on either side of the wall, as is of small magnitude. Likewise, when the two signals are in
shown in Fig. 2. The width of the wall varies from 0.2 to phase, the total received signal is maximum. The point at which
10 . The wall is made of a metal with very high conductivity, the two rays (diffracted and reflected) are in phase (wall size of
which can be considered a perfect electrical conductor (PEC) approximately ) is the point where the extra path length (epl),
material. Results comparing the scattered field signal strength i.e., the difference in path length between the diffracted and re-
using the both the FDTD method and the ray-tracing method flected rays, is 2. This is also the edge of the first Fresnel
are shown in Fig. 3. The ray-tracing result is in good agreement zone [25], as shown in Fig. 5. The Fresnel zones describe the
with the full-wave method, especially for wider wall widths. spatial regions where signals add either constructively (e.g., for

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 25, 2009 at 10:29 from IEEE Xplore. Restrictions apply.
REMLEY et al.: ACCURACY OF RAY-TRACING TECHNIQUES FOR INDOOR PROPAGATION MODELING 2353

Fig. 5. Definition of the first Fresnel zone. An ellipsoid of revolution is formed


around the line connecting the image (T X ) of the transmitter (TX ) with the
specular reflection point (S ) and the receiver (RX ). The focal points are at TX
and RX . (Note that the center of the Fresnel zone would, in the example cited
in the text, correspond to S since R =R .)
Fig. 8. The location of source scatterers when the Kirchhoff surface integral
formulation is used for indoor propagation modeling.

Fig. 6. Spatial configuration used to assess accuracy in the diffraction region.

Fig. 9. Comparison of the FDTD simulation and the ray-tracing simulation for
a PEC wall of various sizes. RSB refers to the reflected field shadow boundary
and ISB refers to the incident field shadow boundary (defined in Fig. 10).

For accurate receiver placement in the FDTD simulation, it


is useful to extend the FDTD domain via a spatial transforma-
tion technique such as the Kirchhoff surface integral formula-
tion. Details of this technique can be found in [18] and [26].
Receiver placement is then not limited to FDTD grid point lo-
cations. To implement the Kirchhoff surface integral formula-
tion in this case, the scatterer (wall) and transmitter are enclosed
Fig. 7. Top view of the simulation space for the combined FDTD/Kirchhoff
technique showing the diffracting wall and receiver placement. Fifty receivers
by the integration surface (Fig. 8). The field found using the
(circles) are spaced every 5.4 over the 270 arc. The transmitter location is Kirchhoff formulation at an arbitrary observation point will
also shown. be the same as would be generated by the original configuration
of transmitter and scatterer.
epl ) or destructively. These regions alternate every mul- Even with only one ray-tracing diffraction considered, results
tiple of 2 of the epl. However, the magnitude of the oscillatory in the shadow region are again comparable to results found using
effect decreases as the extra path lengths get larger, and thus the the FDTD/Kirchhoff method, with a mean squared error of ap-
effect is not noticeable for larger wall sizes in Fig. 4. proximately 5% over the range of the incident field shadow re-
gion. This result indicates that scatterer size does not necessarily
C. Case 2: Diffraction Only limit the accurate use of ray tracing.
To characterize the received signal in the shadow region, the In this section, the significance of the high-frequency approx-
simulation configuration shown in Fig. 6 is utilized. Observa- imation on the accuracy of ray tracing for electrically small scat-
tion points are placed in a 270 circular arc centered around the terers has been investigated. Another approximation commonly
diffracting corner. For the ray-tracing simulations, observation used in ray tracing involves a heuristic modification of equations
points are placed every 1 around the diffracting corner. For the used in the uniform theory of diffraction for modeling diffracted
full-wave simulations, 50 observation points are spaced every fields from dielectric wedges. An extension of this approximate
5.4 over the 270 arc (see Fig. 7). The observation points are technique is developed in the next section.
placed near to the corner at a radius of 0.75 m (2.25 ) to mini- Fig. 9 compares the received signal generated by the com-
mize diffraction around the back of the wall. bined FDTD/Kirchhoff technique for several different wall

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 25, 2009 at 10:29 from IEEE Xplore. Restrictions apply.
2354 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 49, NO. 6, NOVEMBER 2000

The obstacles (wedges) must have large interior angles, the


observation points must be near shadow boundaries (defined in
Fig. 10), and the observation angles must be greater than the
angles of incidence. In cases where these conditions are not
satisfied, a nonphysical dip in the diffracted field may result for
certain angles of observation [13].
In this section, the Luebbers formulation is modified by re-
defining the angular dependence of the diffraction coefficients
in a more physical way. This new set of diffraction coefficients,
described in detail in [18], enables more accurate prediction of
diffracted fields for a wider range of wedge interior angles and
for any observation angle. The new diffraction coefficients are
validated by comparison to results generated with the combined
FDTD/Kirchhoff method and to measurements.
Fig. 10. Geometry of the diffraction problem. The 90 (n =
3=2) wedge
problem is illustrated, with the “0” face and the “n” face denoted. A. Description of the Original Luebbers Formulation
To describe the Luebbers formulation for diffraction coef-
widths to the received signal generated by the ray-tracing ficients, a wedge with interior angle defined by (2 ) (see
method, since only one noninteracting diffraction is generated Fig. 10) is considered. Fresnel reflection coefficients, which cor-
by ray tracing in this case. The wall size is inconsequential in respond to the incidence face (“0” face) or to the diffraction face
the diffraction region. The magnitude of the diffracted signal (“ ” face), are incorporated into the two UTD diffraction coef-
found using the FDTD/Kirchhoff method in the incident field ficients associated with the RSB, i.e.,[2]
shadow region [for angles greater than that of the incident
field shadow boundary (ISB), see. Fig. 10] shows very little
dependence on the wall width until the wall size is a small
fraction of a wavelength. This behavior is in agreement with
results presented in [6] and [7].

III. DIFFRACTION MODELING FOR DIELECTRIC OBSTACLES


The uniform theory of diffraction [8] and [9] is the basis for
diffraction modeling in many ray-tracing programs. The UTD
accurately predicts diffracted fields arising from abrupt material
discontinuities for obstacles made of PEC material.
While there are well-established diffraction modeling
methodologies for PEC wedges, the development of a uniform
diffraction coefficient for wedges with finite conductivity is the (2)
subject of continuing research. A limited number of methods
have been proposed. An asymptotic solution for diffraction Here, the terms and are the Fresnel reflection coeffi-
from lossy wedges was originally developed by Maliuzhinets cients corresponding to perpendicular ( ) and parallel ( ) po-
in 1953 [27]. Subsequent improvements to this technique have larization, respectively. Superscript “0” refers to the illuminated
been proposed in [28]–[30], among others. This technique has wedge face (toward the incident field), and superscript refers
not been widely incorporated into ray-tracing programs due to to the diffraction face. is a Fresnel integral to correct for
the computational intensity involved [11]. the singularities at the shadow boundaries and is a normalized
A second method, proposed by Luebbers in 1984 [10], intro- distance. The two diffraction coefficient terms with argument
duces a heuristic modification to the UTD equations. Fresnel re- correspond to the RSB. The other two terms of (2)
flection coefficients corresponding to the dielectric material of correspond to the ISB through the relation .
the diffracting wedge are incorporated into the UTD coefficient As a demonstration of the nonphysicality that can result when
terms associated with the reflection field shadow boundary, de- the Luebbers formulation is used (erroneously) for arbitrary
noted RSB in Fig. 10. Further discussion on this work can be diffraction calculations, a ray-tracing simulation is carried out
found in [11], [12], and [31]and related work in [32]. As a utilizing an infinite wedge made from a lossy dielectric with
third solution, several numerical diffraction coefficients have S/m and . In these simulations, observation
also been proposed, e.g., in [33] and the references therein. points are located every 1 on a circular ring centered around
Because the heuristic model can be easily and efficiently the diffracting corner, similar to the configuration shown in
implemented in a computer program, this formulation is Fig. 7. The angle of incidence is 22 from either the zero face
currently used in many ray-tracing propagation programs, ( ) or the face ( ). The transmitter and
including [2], [23], [34], and [35]. As discussed by Luebbers observation point locations are m and
[10], [12], the accurate use of these heuristic diffraction coef- m from the diffraction corner, respectively, corresponding to
ficients is restricted to applications meeting certain conditions: and at the frequency of 900 MHz.

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 25, 2009 at 10:29 from IEEE Xplore. Restrictions apply.
REMLEY et al.: ACCURACY OF RAY-TRACING TECHNIQUES FOR INDOOR PROPAGATION MODELING 2355

TABLE I
COMPARISON OF ANGLES INVOLVED IN THE CALCULATION OF
THE ORIGINAL LUEBBERS DIFFRACTION COEFFICIENTS AND THE NEW
DIFFRACTION COEFFICIENTS

(a)

(a)

(b)
Fig. 11. The ray-tracing field components vs. observation angle  for a
dielectric wedge with  =
0:1 S/m,  = 15:0, with (a) transmitter at
 = 22 and (b) transmitter at  = 248 . The dip in the diffracted ray is
nonphysical.

Results in Fig. 11(a) and (b) show the ray-tracing field com-
ponents over the entire range of observation point angles. In
the illuminated region, the interference pattern between the re-
flected ray (short-dashed line) and the direct ray (dash-dot) is
apparent in the total received signal (solid line). The reflected
ray has a constant value since the total path length is constant (b)
for all observation points.
Fig. 12. The ray-tracing field components vs. observation angle  for a
The dips in signal strength of the diffracted rays occurring in diffracting dielectric corner with  = 0:1 S/m,  = 15:0. (a)  = 22
the illuminated region ( RSB) in Fig. 11(a) and in the shad- [compare to Fig. 11(a)] and (b)  = 248 [compare to Fig. 11(b)].
owed region ( ISB) in Fig. 11(b) are erroneous. The dif-
fracted ray solution should decrease monotonically. Note that point and the RSB depends on both and , the new coeffi-
in Fig. 11(a), the inaccuracy may not be apparent since the di- cients, in general, contain both of these quantities.
rect and specularly reflected rays dominate the total received Four different cases for and are considered. They depend
signal in the illuminated region. To eliminate the inaccuracy on whether the angle of incidence is greater than or less than
and increase the versatility of the Luebbers diffraction coeffi- 180 (measured from the zero face), and whether the observa-
cients, a new set of diffraction coefficients has been developed tion point is in the illuminated region, where ,
by comparing ray-tracing results to diffraction field results cal- or in the shadow region. Table I compares angles and
culated using the combined FDTD/Kirchhoff method. Perpen- for the original and new diffraction coefficient formulations for
dicular polarization is considered without loss of generality. these four cases.
Simulation results incorporating the new set of diffraction
B. Development of the New Set of Diffraction Coefficients coefficients are presented in Fig. 12(a) and (b). The three
The basis for the modification of the original formulation is components of the ray-tracing solution are again shown. With
the determination of the angle at which the Fresnel coefficients the new diffraction coefficients, the diffracted rays always
and are calculated. In the original formulation [10], is decrease monotonically away from the shadow boundaries.
calculated at and is calculated at . For a Simulations have been carried out for many different angles of
wedge with an interior angle of 90 , commonly associated with incidence and for a variety of dielectric materials. All show the
objects in indoor propagation modeling, . In the same smooth, monotonically decreasing behavior. Simulations
new diffraction coefficient formulation, remains dependent for wedges with interior angles other than 90 demonstrate a
on . However, since the relationship between the observation similar level of improvement.

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 25, 2009 at 10:29 from IEEE Xplore. Restrictions apply.
2356 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 49, NO. 6, NOVEMBER 2000

(a)
Fig. 14. Comparison of the original Luebbers formulation (dashed), the
improved diffraction formulation (solid), and measured data from [13] (circles).
Here,  = 265 ,  = 0:01 S/m, and  = 4:0.

( m and m) and the received signal level


adjusted to better illustrate the dip effect, which becomes more
pronounced at larger distances. The improvement in the ray-
tracing results with is readily apparent, since the
dip in the diffraction region has been eliminated.
To further validate the new formulation, a comparison is
made with measured data and simulations presented in [13].
Diffraction from a right-angled brick wedge with conductivity
(b) of and relative permittivity of is considered.
Fig. 13. Comparison of the original Luebbers formulation, the improved
The transmitting antenna is an open-ended waveguide excited
formulation, and the combined FDTD/Kirchhoff method for  248 = at 30 GHz and located 1.5 m, or approximately 150 , from the
for (a) dielectric with  = 0:1;  = 15:0 and (b) dielectric with diffracting corner. The angle of incidence is 5 from the face of
 = 0:001;  = 3:4.
the wall, i.e., in this case. Measurements are taken at
a distance of 10 m. Results are presented in Fig. 14 for
C. Validation of the New Diffraction Coefficients the incident field shadow region. Comparison is made between
the measured results from [13] (circles), the original Luebbers
To validate the new set of diffraction coefficients, comparison
formulation calculated in [13] (dashed line), and the improved
is made to simulations utilizing the FDTD/Kirchhoff technique.
formulation (solid line). Good agreement is seen between the
The incident field shadow region is considered. To approximate
measurements and the new formulation.
an infinite wedge in the full-wave model, the transmitter and
The implementation of the new diffraction coefficients re-
observation points are placed close enough to the wedge to pre-
quires only a simple modification of the original formulation
vent diffraction around the back of the wedge for the duration
as shown in Table I. It remains computationally efficient when
of the simulation. Transmission through the dielectric material
compared to the original. Further, the proposed modification
is not significant, as determined by examination of mesh field
provides increased accuracy in both the diffraction and illumi-
plots from the FDTD simulations. The configuration shown in
nated regions, and it enables simulations over a wider range of
Fig. 7 is used, with 50 observation points located in a 270 cir-
angles than does the original formulation.
cular arc centered around the diffracting corner. The transmitter
and observation points are located at m and
m, respectively, away from the diffracting corner. These dis- IV. SUMMARY AND CONCLUSIONS
tances correspond to 2.4 and 2.25 at the carrier frequency of This paper opened with an investigation of the accuracy of
900 MHz. The Kirchhoff surface integral formulation is used to the ray-tracing method for local scatterer sizes approaching the
determine the field at the observation points. Note that the ob- wavelength of operation. Simulation results were presented for
servation points located within the Kirchhoff surface must be two different cases involving 1) diffraction and specular reflec-
disregarded. The FDTD/Kirchhoff solution is valid up to ap- tion and 2) diffraction only. The received signal predicted by the
proximately 260 . The FDTD grid cell dimensions are again ray-tracing method was compared to results generated using the
m( ), with to prevent FDTD and the combined FDTD/Kirchhoff methods. The use
diffraction over the top. of the combined FDTD/Kirchhoff technique provides a stan-
Fig. 13(a) and (b) compares the full-wave simulation results dard against which the ray-tracing technique can be evaluated
with the ray-tracing results in the diffraction region for and improvements developed. Simulation results demonstrated
. Two different materials are considered: Dielectric 1, with good accuracy of the ray-tracing method for scatterer sizes down
S/m, , and Dielectric 2, with S/m, to a fraction of a wavelength. An explanation for the ray-tracing
. The curves corresponding to the original formulation results was provided in terms of the path length difference be-
in Fig. 13(a) and (b) were calculated at ten times these distances tween the diffracted and reflected rays.

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 25, 2009 at 10:29 from IEEE Xplore. Restrictions apply.
REMLEY et al.: ACCURACY OF RAY-TRACING TECHNIQUES FOR INDOOR PROPAGATION MODELING 2357

The development of a new set of diffraction coefficients was [13] C. Demeterscu, C. C. Constantinou, and M. J. Mehler, “Scattering by a
then presented. These new coefficients are adapted from a set of right-angled lossy dielectric wedge,” Proc. Inst. Elect. Eng. Microwave
Antennas Propagat., vol. 144, pp. 392–396, Oct. 1997.
heuristically modified UTD coefficients. The original heuristic [14] K. S. Yee, “Numerical solution of initial boundary value problems in-
coefficients were developed for use in a limited set of diffracted volving Maxwell’s equations in isotropic media,” IEEE Trans. Antennas
field modeling applications involving scatterers with finite con- Propagat., vol. AP-14, pp. 302–307, Dec. 1966.
[15] A. Taflove, Computational Electrodynamics: The Finite-Difference
ductivity. They have inherent inaccuracies when used otherwise. Time-Domain Method. Boston, MA: Artech House, 1995.
The new set of coefficients eliminates the inaccuracies and thus [16] , Computational Electrodynamics: The Finite-Difference Time-Do-
enables application of this ray-tracing formulation to a wider main Method. Boston, MA: Artech House, 1995, pp. 40–46; 93–106.
[17] J. De Moerloose and D. De Zutter, “Surface integral representation radi-
range of propagation modeling problems. The results presented ation boundary condition for the FDTD method,” IEEE Trans. Antennas
in this work show that ray-tracing may be used with confidence Propagat., vol. 41, pp. 890–896, July 1993.
for accurate characterization of multipath environments such [18] C. A. Remley, “Time domain modeling of electromagnetic radiation
with application to ultrafast electronic and wireless communication sys-
as those found in indoor propagation channels, including those tems,” Ph.D. dissertation, Dept. of Electrical and Computer Engineering,
with dielectric obstacles. Oregon State Univ., Mar. 1999.
[19] J. W. Crispin and K. M. Siegal, Eds., Methods of Radar Cross-Sec-
tion Analysis. New York: Academic, 1968.
APPENDIX [20] J. W. Crispin and A. L. Maffett, “Radar cross-section estimation for com-
THE RELATIVE PHASE METHOD plex shapes,” Proc. IEEE, vol. 53, pp. 972–982, Aug. 1965.
[21] E. F. Knott, J. F. Shaeffer, and M. T. Tuley, Radar Cross Sec-
Utilizing the relative phase method, the radar cross section tion. Dedham, MA: Artech House, 1985, pp. 90–112.
due to scattering elements on a body is given by [22] W. L. Stutzman and G. A. Thiele, Antenna Theory and Design, 2nd
ed. New York: Wiley, 1998, pp. 427–492.
[23] EDX Engineering, Microcell Communication Simulator Software, 1997.
[24] J. P. Berenger, “A perfectly matched layer for the absorption of electro-
(3) magnetic waves,” J. Computat. Phys., vol. 114, pp. 185–200, 1994.
[25] A. J. Giger, Low-Angle Microwave Propagation: Physics and Mod-
eling. Boston, MA: Artech House, 1991, pp. 99–124.
where is the RCS of the th component and is the relative [26] K. A. Remley, A. Weisshaar, S. M. Goodnick, and V. K. Tripathi,
phase angle of the th component. The RCS is related to the “Characterization of near- and far-field radiation from ultrafast elec-
tronic systems,” IEEE Trans. Microwave Theory Tech., vol. 45, no. 12,
electromagnetic fields by pp. 407–412, 1998.
[27] G. D. Maliuzhinets, “Excitation, reflection, and emission of surface
waves from a wedge with given face impedances,” Sov. Akad. Phys.,
(4) vol. 3, pp. 752–755, 1958.
[28] D. E. Eliades, “High-frequency scattering from the edges of impedance
discontinuities on a flat plane,” IEEE Trans. Antennas Propagat., vol.
where , are the scattered electric and magnetic fields, AP-31, pp. 16–33, July 1983.
respectively, and , are the incident fields. [29] A. V. Osipov, “General solution of a class of diffraction problems,” J.
Phys. A Math. Gen., vol. 27, pp. 27–32, 1994.
[30] C. Demeterscu, C. C. Constantinou, and M. J. Mehler, “Corner and
REFERENCES rooftop diffraction in radiowave propagation prediction tools: A review,”
[1] W. L. Stutzman and G. A. Thiele, Antenna Theory and Design, 2nd in Proc. 48th IEEE Vehicular Technology Conf., May 1998, pp. 515–519.
ed. New York: Wiley, 1998, pp. 545–620. [31] R. J. Luebbers, “Propagation prediction for hilly terrain using GTD
[2] H. R. Anderson, “A ray-tracing propagation model for digital broad- wedge diffraction,” IEEE Trans. Antennas Propagat., vol. AP-32, pp.
cast systems in urban areas,” IEEE Trans. Broadcasting, vol. 39, pp. 951–955, Sept. 1984.
309–317, Sept. 1993. [32] W. D. Burnside and K. W. Burgener, “High frequency scattering by a
[3] K. R. Schaubach, N. J. Davis IV, and T. S. Rappaport, “A ray tracing thin lossless dielectric wedge,” IEEE Trans. Antennas Propagat., vol.
method for predicting path loss and delay spread in microcellular envi- AP-31, pp. 104–110, Jan. 1983.
ronments,” in Proc. IEEE Vehicular Technology Conf., May 1992, pp. [33] G. Stratis, V. Anantha, and A. Taflove, “Numerical calculation of
932–935. diffraction coefficients of generic conducting and dielectric wedges
[4] H. R. Anderson, “Building corner diffraction measurements and predic- using FDTD,” IEEE Trans. Antennas Propagat., vol. 45, pp. 1525–1529,
tions using UTD,” IEEE Trans. Antennas Propagat., vol. 46, Feb. 1998. Oct. 1997.
[5] M. Born and E. Wolf, Principles of Optics, 2nd ed. Oxford: Pergamon [34] G. A. J. van Dooren and M. H. A. J. Herben, “Field strength prediction
Press, 1964, pp. 375–386. behind lossy dielectric obstacles by using the UTD,” Electron. Lett., vol.
[6] M. I. Herman and J. L. Volakis, “High frequency scattering by a double 29, pp. 1016–1018, May 1993.
impedance wedge,” IEEE Trans. Antennas Propagat., vol. 36, pp. [35] K. Rizk, J. Wagen, and F. Gardiol, “Two-dimensional ray-tracing mod-
664–678, May 1988. eling for propagation prediction in microcellular environments,” IEEE
[7] J. R. Mautz and R. F. Harrington, “Radiation and scattering from large Trans. Veh. Technol., vol. 46, pp. 508–517, May 1997.
polygonal cylinders, transverse electric fields,” IEEE Trans. Antennas
Propagat., vol. AP-24, pp. 469–477, July 1976.
[8] J. B. Keller, “Geometrical theory of diffraction,” J. Opt. Soc. Amer. A,
vol. 52, pp. 116–130, Feb. 1962.
[9] R. G. Kouyoumjian and P. H. Pathak, “A uniform geometrical theory of Kate A. Remley (S’92–M’99) was born in Ann
diffraction for an edge in a perfectly conducting surface,” Proc. IEEE, Arbor, MI, in December 1959. She received the B.S.
vol. 62, pp. 1448–1461, Nov. 1974. degree (magna cum laude), the M.S. degree, and the
[10] R. J. Luebbers, “Finite conductivity uniform UTD versus knife diffrac- Ph.D. degree in electrical and computer engineering
tion prediction of propagation path loss,” IEEE Trans. Antennas Prop- from Oregon State University, Corvallis, in 1993,
agat., vol. AP-32, pp. 70–76, Jan. 1984. 1995, and 1999, respectively.
[11] , “Comparison of lossy wedge diffraction coefficients with applica- She was a Broadcast Engineer in Eugene, OR,
tion to mixed path propagation loss prediction,” IEEE Trans. Antennas between 1983 and 1992. She joined the Radio-Fre-
Propagat., vol. 36, pp. 1031–1034, July 1988. quency Technology Division of the National Institute
[12] , “A heuristic UTD slope diffraction coefficient for rough lossy of Standards and Technology, Boulder, CO, in 1999,
wedges,” IEEE Trans. Antennas Propagat., vol. 37, pp. 206–211, Feb. where she develops metrology for characterization
1989. of nonlinear devices used in wireless communication systems.

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 25, 2009 at 10:29 from IEEE Xplore. Restrictions apply.
2358 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 49, NO. 6, NOVEMBER 2000

Harry R. Anderson (M’79) received the B.S.E.E. Andreas Weisshaar (S’90–M’91–SM’98) received
degree from the University of California, Santa Bar- the Dipl.-Ing. degree in electrical engineering from
bara, in 1972, the M.S.E.E. degree from Oregon State the University of Stuttgart, Germany, in 1987 and the
University, Corvallis, in 1985, and the Ph.D. degree in M.S. and Ph.D. degrees in electrical and computer en-
electrical engineering from the University of Bristol, gineering from Oregon State University, Corvallis, in
U.K., in 1994. 1986 and 1991, respectively.
From 1972 to 1980, he was with Hammett & Since 1991, he has been on the Faculty of the
Edison, Inc. a consulting firm, where he worked Department of Electrical and Computer Engineering
on designing a variety of communication systems at Oregon State University, where he is currently an
including AM, FM, and TV broadcast, paging, Associate Professor. His current research activities
and microwave, and developing novel computer are in the areas of computer-aided design of RF/mi-
algorithms for automating propagation prediction and other engineering crowave circuits and components, interconnects, electronic packaging, and
studies. From 1980 to 1981, he was with Harris Corporation, where he wireless communications. He has authored or coauthored more than 90 journal
was responsible for designing and supervising the construction of several and conference papers in these and related areas.
large broadcast communication projects in the United States, Africa, South
America, and Asia. In 1981, he joined the consulting firm of H.R. Anderson
& Associates, where his work included data communications for utility load
management and system automation, two-way MMDS and microwave system
design, and developing advanced physical propagation models for VHF,
UHF, and microwave frequencies. In 1985, he founded EDX Engineering,
Inc., a software company that develops wireless system planning tools for
communication technologies operating in the 30 MHz to 100 GHz frequency
range. At EDX, he pioneered the use of ray-tracing propagation models for
broadband wireless system design. He is currently President and CEO at EDX.
His research now focuses on broadband fixed wireless systems, including
propagation and channel models, system capacity using adaptive antennas and
power control, optimum channel allocation schemes, and site-specific system
resource allocation.
Dr. Anderson has published more than 40 papers in IEEE TRANSACTIONS and
other journals, including two papers that have won IEEE TRANSACTIONS Best
Paper of the Year awards. He is a Registered Professional Engineer in California
and Oregon.

Authorized licensed use limited to: Aalborg Universitetsbibliotek. Downloaded on September 25, 2009 at 10:29 from IEEE Xplore. Restrictions apply.

You might also like