Site-specific propagation prediction for wireless in-building personal communication system design

TL;DR: Time delay comparison shows that the amplitudes of many significant multipath components are accurately predicted by this model, and the effective building material properties are derived for two dissimilar buildings based upon comparison of measured and predicted power delay profiles.

Abstract: The paper describes a geometrical optics based model to predict propagation within buildings for personal communication system (PCS) design. A ray tracing model for predicting propagation based on a building blueprint representation is presented for a transmitter and receiver located on the same floor inside a building. Measured and predicted propagation data are presented as power delay profiles that contain the amplitude and arrival time of individual multipath components. Measured and predicted power delay profiles are compared on a location-by-location basis to provide both a qualitative and a quantitative measure of the model accuracy. The concept of effective building material properties is developed, and the effective building material properties are derived for two dissimilar buildings based upon comparison of measured and predicted power delay profiles. Time delay comparison shows that the amplitudes of many significant multipath components are accurately predicted by this model. Path loss between a transmitter and receiver is predicted with a standard deviation of less than 5 dB over 45 locations in two different buildings. >

Summary

1.3 The Importance of Accurate Propagation Models

• In a mobile and portable radio environment, one communications terminal is allowed to roam throughout a coverage area with tetherless access to the communications system through a network of fixed base stations.
• This chapter discusses multipath propagation, its effects on transmitted signals, and methods used to measure and model the propagation.
• These effects are then considered in the development of a discrete multipath channel model.
• The definition of radio path loss between a transmitter and receiver is given, and a common statistical model for path loss as a function of distance is presented.

2.1 Multipath Propagation

• Ionospheric radio channels suffer from multipath propagation caused by scattering from millions of ionized particles in the ionosphere.
• The RF bandwidth of a probing signal determines the baseband time resolution, T,,, of propagation measurements.
• The instantaneous amplitude and phase can be continuously measured.
• The receiver performs a phasor sum of all E fields incident upon the receiver antenna.
• The propagation time delay is proportional to the path length traveled by the multipath component.

2.2 Multipath Channel Model

• The portion of the signal that remains when the shadow fading is removed from the received signal is termed the fast fading signal.
• Fast fading is caused by the rapidly varying phasor sum due to the phase changes of the individual multipath components incident upon the receiver antenna.
• This term is also commonly called the Rayleigh component since the probability distribution of the signal amplitude is often Rayleigh distributed.

2.3.2 Statistical Mean Path Loss Exponent

• Where n is the mean path loss exponent which indicates how fast mean path loss increases with distance, do is a reference distance, and d is the transmitter-receiver (T-R) separation distance.
• When plotted on a log-log scale, this power law relationship is a straight line.
• The mean path loss at a distance d in decibels is an algebraic equation defined as the path loss in decibels from the transmitter to the reference distance dy plus the additional path loss described by (2.3-1) in decibels.

PL (d)

• The selection of the reference distance dg is critical in the interpretation of path loss measurements.
• Urban mobile radio propagation measurements in [Se191b] show that a change in reference distance from 100 m to 1 km changes the perceived mean path loss exponent from 2.7 to 3.0.
• It is important to choose a reference distance that is appropriate for the propagation environment.
• The reference path loss PL(dp) is Site-Specific Propagation Prediction for Wireless February 7, 1993 In-Building Personal Communication System Design assumed to be due to free space propagation from the transmitter to dy.
• Measurements show this is valid assumption to within 1 dB nominally [Rap89a] .

4n (dp)

• Path loss is often modeled as a log-normal distribution about the mean power law described by (2.3-2) [Cox84] .
• That is, PL(d) = PL (d) +X, where X,, is a zero-mean lognormally distributed (normal in dB) random variable with standard deviation o in dB.
• In [Haw90] , [Haw91] , [Sei91a] , A statistical distance-dependent path loss model is useful for understanding the propagation of radio waves in buildings.
• Exhaustive measurements were required to obtain the data to determine the appropriate parameters for the models for these particular buildings.
• In addition, the mean path loss exponents can vary from less than two to greater than six depending upon the specific environment (see Section 3.1 Statistical Path Loss Models).

2.4.1 Mean Excess Delay

• Notice that half of the energy outside the window arrives before the window and half arrives after the window.
• This parameter is a useful measure of the time dispersion of the The excess delay spread of a channel is defined as T,-tg where.
• To is the first arriving signal and 1, is the maximum delay at which a multipath component is within X dB of the strongest arriving multipath signal.
• The excess delay spread defines the extent of.

2.5 Multipath Propagation Measurements

• In [Rap89b] , it was shown that CW measurements averaged over space and wide band measurements averaged over time give equivalent path loss results when multipath component phases are independent and identically uniformly distributed over [0,27) or when multipath component amplitudes are uncorrelated.
• Hence, a simple CW measurement system (Bpr ~ 15 kHz) may be used to give path loss results that may be applied to wide band channels (Ber > Bc) as well when the wide band channel is averaged over space.
• This can be useful when there is not enough spectrum available for propagation experiments in a particular frequency band.
• When the data rate of a proposed communications system is less than the coherence bandwidth, a narrow band channel characterization Site-Specific Propagation Prediction for Wireless February 7, 1993 In-Building Personal Communication System Design is sufficient to determine system performance.
• When the data rate is larger than the coherence bandwidth, the structure of the channel impulse response must be known to determine the extent of intersymbol interference.

2.5.1 CW Measurements and Data Processing

• Propagation studies that examine the effects of carrier frequency, antenna height, antenna pattern, antenna polarization, and obstructing objects like floors and partitions [Haw91] , [Sei91a] , [Sei92a] on path loss and/or fading statistics can easily utilize the measurement system shown in Figure 2 .5-1. to develop path loss models and contours.
• The first variable attenuator prevents the amplifier from becoming saturated when the receiver is close to the transmitter.
• A digital oscilloscope records the squared magnitude of the impulse response (power delay profile).
• The term Gcay, converts the total integrated power to an equivalent peak power in a pulse of width T)p.

Site-Specific

• Then, path loss is given by It is important to point out the importance of receiver sensitivity and noise threshold in interpretation of measured channel parameters.
• This is similar to the radar probability of detection vs. probability of false alarm problem.
• Experience has shown that it is necessary to determine an appropriate threshold level for power delay profiles to avoid the inclusion of profiles that had a lost trigger or were corrupted with impulsive noise [Rap90a] , [Sei91b] , [Tel91].
• As a minimum requirement, visual inspection should be used to verify that the automatic procedure is working correctly.

2.5.3 Spread Spectrum Measurements

• For a code length of 2047 chips, the maximum dynamic display range is 33 dB.
• Allowing for some noise, a practical display range is about 30 dB.
• Longer chip sequences can provide larger dynamic display ranges.
• For pulse systems, the dynamic display range is limited by the number of quantized levels on a digital oscilloscope.
• Theoretically this value is 30 dB, however, amplifier noise usually reduces this to around 20 dB (see Figure 2 .4-1).

2.6 Summary

• This chapter presents a survey of the state of the art of statistical measurement-based propagation models in many different building environments.
• Models for path loss as a function of distance and median and maximum rms delay spreads are presented for different buildings.
• A statistical multipath channel model simulator that uses the statistical results of propagation measurements in a wide variety of buildings, SIRCIM, is described.
• This chapter concludes with promising site-specific path loss models in [Sei91a] and [Sei92a] that account for the shadowing caused by floors, walls, office partitions, and common objects found in many different types of buildings.
• In fact, the good agreement of these site-specific models with measurements led to the development of the ray tracing propagation model presented in Chapter 5 which is the primary contribution of this Ph.D. dissertation.

3.1 Statistical Path Loss Models

• This is likely due to the nature of the obstructions between transmitter and receiver in the different environments.
• In open-plan buildings, obstructions do not completely block the transmitter and receiver.
• In partitioned buildings, the partitions attenuate the signals, and hard partitions attenuate more than soft partitions.

3.2 RMS Delay Spread Results

• The measurements classified as residences in Table 3 .1-1 were conducted for indoor to outdoor radio paths, whereas the transmitter and receiver were both inside the building for the other measurements.
• Multipath components were reported to have significant energy out to excess delays of around 1 ps.
• Maximum observed rms delay spreads were on the order of 100 ns when a direct path existed between the transmitter and receiver and were as large as 420 ns when no direct path was present [Dev86] .

3.4 Small-Scale Narrow Band Fading Statistics

• Site-Specific Propagation Prediction for Wireless Models like SIRCIM are valuable for evaluating system performance on a statistical basis.
• Only limited information about the building is included.
• Site-specific path loss and delay spread models would allow more efficient system design and analysis.

3.6 Site-Specific Path Loss Models

• Large values are typical for data collected from different environments even within the same building.
• These parameters may be used in the model for a first-order prediction of mean signal strength when only T-R separation but no specific building information is known, but this model is clearly unsatisfactory for detailed site layout or capacity prediction.

3.6.1 Multi-Floor Path Loss Exponent [Sei92a]

• In multi-floored environments, equation (3.6-1) can be used to describe the mean path loss as a function of distance.
• Equation (3.6-1) is identical to equation (2.3-2) and emphasizes that the mean path loss exponent is a function of the number of floors between transmitter and receiver.
• The values of n(multi-floor) are given in Table 3.
• In the previous section, the path loss in multi-floored environments was predicted by a mean path loss exponent that was a function of the number of floors between transmitter and receiver.
• Alternatively, a constant Floor Attenuation Factor (dB), which is a function of the number of floors and building type, may be added to the mean path loss predicted by a path loss model which uses the same floor path loss exponent for the particular building type (equation (3.6-2)).

PL (d)=

• Note that the method used to determine the FAF was slightly different in [Sei92c] than in [Sei92a] .
• 1 [Sei92al of the difference between the measured and predicted path loss in [Sei92a] and [Sei92c], also known as "Office Building.
• The average Floor Attenuation Factors for an identical number of floors between the transmitter and receiver for the two buildings differed by 3-8 dB in [Sei92a] .
• All floors in the two office buildings were made of reinforced concrete.
• Office building 1 was built within the past ten years, and office building 2 was 20 to 30 years old.

3.6.4 Path Loss Contour Plots [Sei92a]

• The error contours for the West wing of the fifth floor of office building 1 in [Sei92a] and the contours of measured path loss for the same transmitter location and building wing are given in Figure 3 .6-4.
• The prediction error is less than 3 dB for about half of the area in this location.
• This may be partly explained in that the Attenuation Factors for the fifth floor in the West wing of office building 1 were the lowest measured and differ by about 0.4 dB from the Attenuation Factors used to predict path loss.
• Thus, on average, 0.4 dB more path loss is predicted than was actually observed for each partition and concrete wall in this wing.
• When many partitions and concrete walls are between the transmitter and receiver, the bias can be 4-5 dB.

Site-Specific Propagation Prediction for Wireless

• Figure 3 .6-5 shows the measured and error contours for the fourth floor West wing of office building lin [Sei92a] .
• February 7, 1993 In-Building Personal Communication System Design.
• The RCS model is a useful modeling technique for a first-order approximation for worst-case scattering, but accurate use of the model may be limited when buildings are close such that the far-field approximation is violated.
• The path lengths from the scatterer to the transmitter and receiver were assumed equal and could be determined from the time delay of the multipath component.
• Scattering in nonspecular directions was accounted for by non-coherent summation of the powers scattered from each integration patch (surface facet) on the surface of the scatterer.

3.7 Summary

• The scattered fields can be determined from the surface currents that radiate in an unbounded homogeneous medium in the equivalent problem.
• The values of the surface currents are determined so that the total fields inside the body are zero.
• The scattering problem has been transformed into a radiation problem.
• Radiation and scattering are variations of the same process.
• As an example, the pattern radiated by an antenna is expressed by a Fourier transform that depends on the size of the antenna and the current distribution on the antenna or the field distribution in the antenna aperture.

4.3.2 Dielectric Bodies

• The diffraction coefficients in Section 4.3.1 Perfectly Conducting Surface were developed from the canonical scattering solution for perfectly conducting wedges.
• The canonical problem of scattering by a dielectric wedge is as yet unsolved.
• Hence, diffraction coefficients can not be derived directly from the solution.
• Instead, the diffraction coefficients are modified so that the continuity of the total field at the shadow boundaries is maintained.
• Each diffraction coefficient is made up of an incident diffracted term and a reflected diffracted term.

Finite Strip Model

• Consider a cylindrically spreading wave incident upon a finite width perfectly conducting strip.
• The geometry is shown in Figure 4 .3-4 where the variable a is first considered to extend to infinity.
• The cylindrical scattered fields consist of the geometrical optics reflected field from the surface of the strip and a diffraction contribution from each of the two edges.
• Each edge is modeled as an infinite half plane.
• For a wave polarized in the +x direction, the polarization is TE with respect to the two edges.

Infinite Strip Model

• The far-field conversion factor cannot be used when the observation point is close to the scatterer.
• As a model for the scattering from a perfectly conducting flat plate, the scattering is assumed to be the same as that from a perfectly conducting infinite strip with the same width as the plate.
• The finite length of the plate in the transverse direction is ignored, and spherical wave incidence and scattering are assumed.
• It is shown that both the near-field and the far-field model compare well with the physical optics solution in the specular direction when each model is used in the appropriate region of validity.
• In the far-field, the length of the plate in the transverse direction 'appears' small, and a far-field conversion factor in the finite strip model may be used to account for the finite length of the plate.

4.4 Physical Optics (PO)

• Equations (4.4-5) and (4.4-7) are the exact expressions for the scattered field due to a surface current and these equations are valid in the near field.
• The accuracy is determined by the accuracy of the physical optics assumption for the electric and magnetic currents on the surface of the plate in equations (4.4-1) and (4.4-2).

4.5.1 The Bistatic Radar Equation in Multipath Environments

• Time domain multipath propagation measurements are, in essence, a bistatic radar measurement of the environment.
• Thus, it seems reasonable that the impulse response of such a channel could be described in terms of a radar cross-section model.
• In most cases of practical concern, the RCS of a scattering object can vary by as much as 20-30 dB or more for small changes in frequency, angle of incidence, angle of reflection, wave polarization, and orientation of the object [Sko70] .
• The number of electromagnetic scattering problems that can be solved directly is small.

4.5.2 Incoherent Power Summation

• The radar cross-section model described above is useful for modeling the scattering in non-specular directions when the receiver is in the far-field of large scattering objects.
• This model is not applicable to close-in receiver locations, and can only be used for worst-case system design.
• A heuristic model that subdivides a surface into multiple facets such that the receiver is in the far-field of each facet can be used to extend the region of validity for the radar cross-section model.

4.6.2 Implementation and Complexity Issues

• Surface patches be smaller than A/2 by A/2 places a restriction on the minimum tessellation frequency that can be used.
• Hence, although each computation requires little computational effort, the total computation time required can be significant.
• The uniform geometrical theory of diffraction model appears to be more complex than the other scattering models.
• For each diffracting edge (or corner), the specific ray path must be found only once and only one set of computations are required.
• These approximations along with interpolation between the large and small argument polynomial estimates are implemented according to the algorithm in [Bal89] .

Effect of integration patch size

• In order to determine the minimum resolution required for integration using the physical optics method, the integration patch size was varied for a 10A by 10A square plate.
• For square patches with lengths less than A/2, there is no noticeable difference in scattering pattern as a function of angle for all incidence angles.
• Hence, when scattering is implemented by physical optics in site-specific propagation prediction programs, the size of the wavefront on the surface of the scatterer should be no larger than A/2 by A/2 for accurate computation of scattering patterns for both near and far field illuminance and observation distances.

Near Field vs. Far Field

• Since it is possible for receivers to be in the near field of scattering objects, the scattering pattern is examined as a function of observation distance for different angles of incidence.
• The incidence angle in the upper right plot is 30 degrees, and the incidence angles for the bottom two are 60 and 80 degrees.
• These patterns are the typical sin(x)/x shaped patterns for far-field scattering from a flat plate.
• Hence, the applicability of this model in the far-field is questionable at best.
• In the near field, Figure 4 .6-4 showed that the scattered power away from the specular direction is within about 5 dB of that predicted by the other methods, but the shape of the scattering pattern is not predicted.

Effects of Surface Roughness on Reflection Coefficient

• For scattering from dielectric surfaces, the scattered power depends on the Fresnel plane wave reflection coefficients.
• When a surface is not perfectly smooth, the coherent scattering in the specular direction is reduced.
• In the models presented here, the surfaces are assumed to be perfectly smooth.
• Site-Specific Propagation Prediction for Wireless February 7, 1993 In-Building Personal Communication System Design 0 0; 90 FIGURE 4.6-9.
• Reflection coefficient correction factor for slightly rough surfaces as a function of surface roughness in wavelengths and incidence angle.

4.7 Summary

• In-Building Personal Communication System Design within about 5 dB.
• For all distances, however, the non-coherent models do not predict the scattering pattern as a function of angle that is predicted by the uniform geometrical theory of diffraction.
• The UTD solution was shown to agree with measured radar cross-sections in [Ros66] .
• These objects can be used to model walls inside a building or the exterior walls of a building in a microcellular environment.
• The uniform geometrical theory of diffraction was shown to give accurate results with both the infinite strip and the finite strip model where the implementation is dependent upon the observation distance from the scattering object.

Site-Specific Propagation Prediction for Wireless February 7, 1993

• In-Building Personal Communication System Design 5 Site-Specific In-Building Propagation Prediction 5.0 Overview Buildings vary greatly in size, shape, and type of construction materials.
• As shown in Chapter 3, measured propagation statistics vary greatly from building to building and only broad conclusions related to the building type can be made (recall Table 3 .1-1).
• From the power delay profile, parameters such as path loss and time delay spread of indoor radio channels may be determined.
• Here, a geometrical optics model is used to predict the propagation of radio waves in buildings.
• This chapter presents the details of the ray tracing model.

5.1 Building Blueprint Representation

• The representation of buildings in a computer database, even different buildings represented within AutoCAD, can vary significantly from building to building.
• This section discusses a 'standard' representation of the major building features inside the AutoCAD database manager, and explains how the AutoCAD building representation is converted to a file compatible with the ray tracing prediction program.

5.1.1 Conversion of AutoCAD Data Format to Ray Tracing Input

• In order to implement site-specific propagation models, it is necessary to incorporate the site-specific building information into the propagation prediction tool.
• AutoCAD is used to represent the significant building features such as wall locations and building materials.
• It is straightforward to draw the wall locations on the Site-Specific Propagation Prediction for Wireless February 7, 1993 In-Building Personal Communication System Design BUILDING layer while the original drawing that was drawn on another layer is displayed on the screen.
• This representation includes 'thick' walls, stairs, doors, and features such as stairs that are external to the building.
• A database conversion program developed and described in detail in [Sch92] is used to convert the AutoCAD information to a format that can be input directly into the ray tracing program.

5.2.1 Background

• The solution for the source ray directions is adapted from the theory of geodesic domes [Ken76] , [Pug76] , [Wen79] .
• An icosahedron is inscribed inside the unit sphere.
• Figure 5 .2-3a and b show two views of a regular icosahedron.
• The dotted line is the difference in ray separation for a given source ray and its own nearest neighbors.
• Thus, this method of launching the source rays provides wavefronts that completely subdivide the surface of the unit sphere with nearly equal shape and area.

5.2.5 Identification of Specular Rays

• Adjacent rays that are received by overlapping reception spheres are eliminated during the processing of the raw ray tracing output.
• If nearly uniform separation is not maintained, the test ray will not be The reception sphere provides a method for identifying specular components.
• The size of the reception sphere is determined by the separation between source rays and the distance travelled by the received ray.
• The propagation distance is fixed, but the tessellation frequency can be changed to directly affect the error introduced by using a brute force ray tracing method that incorporates a reception sphere.
• This error is not the error caused by the difference between the measured and predicted propagation, but it is the error introduced by the implementation of the propagation model.

5.3 Diffraction

• The implementation of diffraction in the computer program is separate from the brute force ray tracing.
• Since there is no recursion, the diffracted ray paths may be found by a straightforward search for all paths that satisfy the correct geometry for a diffracted ray.
• Diffracting corners are modeled as dielectric wedges.
• Diffracted rays are found for all combinations of transmitter and receiver that have a direct path to the diffracting wedge.

5.4 Processing Raw Ray Tracing Program Output

• Once the data is separated by receiver location, a .sort file is created which contains the received multipath components sorted by increasing excess delay.
• The .sort files are created and multiple representations of identical multipath components that may have been captured by overlapping reception spheres are eliminated.
• The first line contains the ray number, the receiver number, the ray type, the excess delay in nanoseconds, the received field strength relative to the value measured at one meter over free space, the departure and arrival angles, and the two ray reflection history strings.
• The second line contains the ray incidence angle history.

6.1 Comparison Criteria and Error Function Definition

• In Whittemore Hall, the dielectric constant that minimizes the squared error functions is €,=4.4 at both 1.3 GHz and 4.0 GHz.
• In Norris Hall, the best-fit effective dielectric constant is ¢,=7.4.
• The resultant Fresnel reflection and transmission coefficients as a function of incidence angle are shown in Figure 6 .2-1.
• The Fresnel reflection and transmission coefficients are not sensitive to small changes in the dielectric constant so that this ray tracing method may be used with confidence even when the optimal effective building materials are In-Building Personal Communication System Design.

6.3 Comparison of Measured and Predicted Power Delay Profiles

• This section provides a qualitative comparison of measured and predicted power delay profiles.
• The measured and predicted power delay profiles for different measurement locations at the two different frequencies and two different buildings are compared and discussed for both line-of-sight and obstructed measurement locations.
• The uniform geometrical theory of diffraction was included in the model.
• For the measurement and prediction locations presented here, the diffracted rays do not contribute significantly to the received power delay profiles.
• Hence, it is included in the model for completeness.

6.3.1 Whittemore Hall -1.3 GHz

• The fundamental theory for site-specific propagation modeling was outlined in Chapter 4.
• The theory of geometrical optics was presented for application in an automated ray tracing program.
• Scattering and diffraction from flat plates were computed using the uniform geometrical theory of diffraction, physical optics, and a heuristic incoherent scattering model based on radar cross-section.
• The applicability of each model to site-specific propagation prediction was discussed.
• Propagation parameters such as path loss, rms delay spread, and delay interval can be computed from the power delay profiles.

Full text

Site-Specific
Propagation
Prediction
for
Wireless
In-Building
Personal
Communication
System
Design
by
Scott
Y.
Seidel
Dissertation
submitted
to
the
Faculty
of
the
Virginia
Polytechnic
Institute
and
State
University
in
partial
fulfillment
of
the
requirements
for
the
degree
of
Doctor
of
Philosophy
in
Electrical
Engineering
PROYED:
Theodore
S.
Rappaport,
Chairman
Charles
W.
Bostian
GAry
S.
Brown
Werner
E.
Kohler
Brian
D.
Woerner
February
7,
1993

S4oQ.
er
Se

Site-Specific
Propagation
Prediction
for
Wireless
February
7,
1993
In-Building
Personal
Communication
System
Design
Site-Specific
Propagation
Prediction
for
Wireless
In-Building
Personal
Communication
System
Design
by
Scott
Yates
Seidel
Theodore
S.
Rappaport,
chairman
Electrical
Engineering
(ABSTRACT)
buildings
for
personal
communication
system
(PCS)
design.
Background
on
the
growth
of
nn
I
eee
wireless
communications
is
given,
and
the
importance
of
accurate
propagation
models
is
discussed.
The
peculiarities
of
propagation
in
mobile
and
portable
radio
environments,
particularly
multipath
propagation
and
its
effects
on
transmitted
signals,
are
described.
Current
in-building
propagation
models
are
presented
and
the
progression
from
statistical
in
the
presence
of
scattering
bodies
are
merged
with
a
site-specific
description
of
the
prop-
agation
environment
to
improve
upon
the
accuracy
of
existing
propagation
models.
A
geometrical
optics
ray
tracing
model
for
predicting
propagation
based
on
a
building
blueprint
representation
is
developed
for
a
transmitter
and
receiver
located
on
the
same
floor
inside
a
building.
The
measured
and
predicted
propagation
data
are
presented
ethnehatei
as
power
delay
profiles
that
contain
the
amplitude
and
arrival
time
of
individual
multipath
components.
Measured
and
predicted
power
delay
profiles
are
compared
on
a
location-by-
location
basis
to
provide
both
a
qualitative
and
a
quantitative
measure
of
the
model
accu-
racy.
The
concept
of
effective
building
material
properties
is
developed.and
the
effective
building
material
properties
are
derived
for
two
dissimilar
buildings
based
upon
compar-
ing
measured
and
predicted
power
delay
profiles.
Time
delay
comparison
shows
that
the
amplitudes
of
many
significant
multipath
components
are
accurately
predicted
by
this
model.
Path
.
loss
between
a
transmitter
and
receiver
is
predicted-with
a
standard
deviation
of
less
than
5
dB.
Ideas
for
improving
the
accuracy
and
expanding
the
applicability
of
the
models
applied
here
to
wireless
in-building
propagation
prediction
are
suggested.

Site-Specific
Propagation
Prediction
for
Wireless
February
7,
1993
In-Building
Personal
Communication
System
Design
Foreword
This
work
has
developed
in
response
to
the
growth
of
wireless
communications
systems.
The
success
of
cellular
radio
indicates
that
ubiquitous
personal
communications
can
become
a
reality.
As
part
of
a
complete
Personal
Communications
System
(PCS),
wireless
coverage
will
be
required
inside
buildings.
The
propagation
conditions
severely
affect
the
quality
of
communications
over
a
radio
link.
These
propagation
conditions
are
highly
dependent
upon
the
location
of
major
objects
such
as
walls
and
partitions
inside
buildings.
To
date,
in-building
propagation
models
have
incorporated
only
a
limited
amount
of
site-specific
information
concerning
the
propagation
environment.
Yet,
it
is
the
specific
geometry
that
influences
the
propagation.
Hence,
the
major
contribution
of
this
disserta-
tion
is
that
the
propagation
characteristics
may
be
accurately
predicted
by
incorporating
information
about
the
location,
size,
and
electrical
properties
of
major
building
features.
Two
different
contributions
to
site-specific
propagation
prediction
are
presented.
First,
a
soft-partition
and
concrete
wall
attenuation
factor
path
loss
model
for
a
single
floor
system
is
given.
Also,
a
floor
attenuation
factor
path
loss
model
for
multi-floored
office
buildings
that
contain
large
open
areas
where
individual
offices
are
cubicles
separated
by
cloth-cov-
ered
soft
partitions
is
presented.
The
second
model
is
a
more
general
ray
tracing
method
that
can
be
used
to
predict
power
delay
profiles
for
transmitters
and
receivers
located
on
the
same
floor
within
a
building.
The
power
delay
profile
contains
both
path
loss
and time
delay
information.
Innovative
contributions
to
ray
tracing
include
a
novel
way
to
determine
the
source
ray
directions
in
three
dimensions,
the
development
of
a
partially
automated
ray
tracing
com-
puter
code
that
includes
the
use
of
a
standard
computer
aided
design
(CAD)
program
for
the
building
database,
and
an
optimization
routine
for
determining
effective
building
mate-
rial
electrical
properties.
This
work
presents
the
first
ray
tracing
for
propagation
prediction
where
measured
and
predicted
power
delay
profiles
are
compared
on
a
location-by-loca-
tion
basis
as
a
function
of
time
delay.
An
error
curve
as
a
function
of
excess
delay
is
defined,
and
the
area
under
the
squared
error
curve
is
minimized
by
varying
the
reflection
Page
iii

Site-Specific
Propagation
Prediction
for
Wireless
February
7,
1993
In-Building
Personal
Communication
System
Design
coefficients
of
the
building
materials.
Important
propagation
channel
statistics
that
include
path
loss,
rms
delay
spread,
and
delay
interval
are
compared
on
a
location-by-location
basis
to
quantify
the
model
accuracy.
A
thorough
literature
review
is
included
throughout
this
work.
First,
a
background
for
understanding
propagation
in
mobile
and
portable
radio
environments
is
discussed.
Measurement-based
propagation
models
are
compiled
from
the
literature.
These works
start
from
statistical
descriptions
of
how
fast
mean
path
loss
increases
with
distance
and
median
and
maximum
rms
delay
spread,
and
lead
into
site-specific
path
loss
models
in
multi-floored
and
soft-partitioned
environments
developed
by
the
author.
In
order
to
develop
more
accurate
propagation
models,
electromagnetic
theory
concerning
the
inter-
action
of
radio
waves
with
scattering
objects
is
presented.
This
includes
the
theory
of
geo-
metrical
optics,
implemented
by
the
automated
propagation
prediction
tool.
The
chapter
on
scattering
covers
fundamental
aspects
of
a
propagation
mechanism
that
are
often
misunderstood.
A
description
of
physical
optics
and
the
geometrical
theory
of
diffraction
are
given,
and
these
electromagnetics
models
are
applied
to
the
scattering
from
a
smooth
rectangular
flat
plate.
These
models
can
be
used
to
compute
the
scattering
from
interior
and
exterior
building
walls
in
either
an
indoor
or
an
outdoor
microcellular
radio
environment.
These
models
are
compared
and
contrasted
with
a
heuristic
model
for
scattering
from
similar
surfaces
given
in
[Sch92].
Limitations
for
the
applicability
of
each
model
in
a
site-specific
propagation
prediction
tool
are
given.
This
dissertation
presents
a
thorough
background
of
the
mobile
and
portable
radio
propagation
environment.
Using
knowledge
of
this
environment,
theoretical
electromag-
netic
scattering
models
are
merged
with
site-specific
information
about
the
physical
prop-
agation
environment
to
predict
the
propagation
channel
characteristics
as
a
function
of
location.
This
represents
a
major
shift
in
the
development
and
application
of
propagation
models.
Previously
developed
models
for
propagation
in
buildings
have
only
incorporated
limited
amounts
of
site-specific
information,
and
these
models
then
rely
on
statistical
results
of
radio
propagation
measurements.
With
the
advent
of
increased
computational
power
and
more
efficient
coding
algorithms,
the
models
developed
in
this
dissertation
can
Page
iv