Sie sind auf Seite 1von 11

EE245 WIRELESS DESIGN LABORATORY: 2.

4 GHZ INDOOR CHANNEL



1

AbstractThis paper presents the results of indoor wireless
channel measurements and characterization of an indoor
hallway at the 2.4 GHz ISM band (2.4-2.83GHz). Important
channel parameters like the propagation loss factor, Ricean K-
factor, delay spread, and coherence bandwidth are calculated
and discussed. The measurements were performed for different
antenna configurations in a hallway environment. All channel
parameters were derived from the power transfer function and
its associated channel impulse response. These estimated values
were also compared to a 2-D Finite Difference Time Domain
(FDTD) model, assuming perfectly conducting walls along the
hallway. From the simulation, power profiles in time were
obtained, from which path loss, delay spread and coherence
bandwidth were calculated. The path loss factor is found to be
lower than the free space attenuation. The mean value of the
Ricean K-factor is approximately independent of distance and
varies for each antenna configuration. The delay spread was
found to be about 10 ns for indoor sites, whereas the coherence
bandwidth was around 6 MHz.

I. INTRODUCTION
OST of the early research efforts in wireless
communication were made on individual radio links from
one transmitter to one receiver. The main problem was to send
signals over an unreliable radio link. The signal that arrives at
the receiver is greatly distorted due to time-varying and
frequency-selective fading. The radio waves are scattered from
reflecting obstacles and arrive at the receiver with different
delays, producing multi-path fading especially in indoor
communication with a dense architecture. These problems
have been largely handled after the introduction of digital
communication techniques.
In parallel to this evolution, the fixed Local Area Networks
(LAN) are becoming wireless. A Wireless LAN (WLAN) is
aimed to provide local wireless access to fixed network
architectures. WLAN developments are now driven by the
success of the Internet with its services and applications. For a

This paper is submitted as Final Report of the course EE245 Wireless
Design Laboratory at Stanford University coordinated by Professor D.
Leeson.
The authors are graduate students in the Electrical Engineering
Department, Stanford University, Stanford, CA 94305. They can be
reached by e-mail at: {eulffe, platinom, barretod}@stanford.edu.
while, the WLAN market did not bloom, probably because the
low data rate (2 Mbps) that was offered. In July 1990, the
Institute of Electrical and Electronics Engineers formed a
working group (IEEE 802.11) to establish a world wide standard
for WLANs. By that time, a number of Radio LAN products
had been developed in the ISM (Industrial, Scientific and
Medical) bands, 902-928 MHz, 2400-2483.5 MHz and 5725-5875
MHz. These frequency bands are unlicensed and can be used
for data transmissions if a number of rules are followed. The
efforts that the working group IEEE 802.11 made resulted in an
approved standard for WLANs in June 1997.
The standard consists of three different physical layer
specifications, one with infrared and two with radio
transmissions. In 1999 this Task group finished with the
specification for a new physical layer (802.11b) that increased
the bit rates from 1 and 2 Mbps to 5.5 and 11 Mbps. Most
recently, IEEE published the 802.11g standard that supports a
data rate of up to 54 Mbps.
The radio standards for WLANs operate in the 2.4 GHz
band, the only accepted ISM band available worldwide. In that
sense, a careful characterization in the WLAN signal
propagation channel is essential for determining optimum
methods to achieve the proposed data rates and quality of
service. This paper describes results of wireless propagation
measurements carried out at a center frequency of 2.45 GHz,
which may be useful in design of communication systems
operating in such band like those compliant with IEEE 802.11,
Bluetooth and HomeRF open wireless standard.

II. MEASUREMENT SYSTEM
A. Measurement Setup
The measurement system for the indoor wireless channel is
based in a swept frequency technique called channel sounder.
As shown in Figure 1, that system is built around a vector
network analyzer (VNA) Agilent 8753ES, which is capable of
measuring the S-parameters of any device under test (DUT)
connected between its two ports.
Two antennas are connected to the ports of the test set of
the VNA with the transmitting antenna associated to the first
port and the receiving to the second one. The connection of
the antennas was made using coaxial cables. The transmi tter
line (7.5 ft antenna cable and connectors) and the receiver line
Measuring the 2.4 GHz Band for Indoor Wireless
Communications
Enrique Ulffe, Manuel Platino and Daniel Barreto, Stanford University
M
EE245 WIRELESS DESIGN LABORATORY: 2.4 GHZ INDOOR CHANNEL

2
(3.75 ft antenna cable) have a loss of 1.55 dB each. The
connecting path between the receiver line and the VNA has a
loss of 3.6 dB (25 ft coaxial cable) or 17.2 dB (100 ft coaxial
cable). In this way, when a VNA is used as part of a channel
sounder, the DUT is the wireless channel, which includes
transmitting and receiving antennas, and the propagation
channel between them. Measuring the S
21
parameter, the
frequency response of the wireless channel is obtained in a 100
MHz band centered at 2.45 GHz (? 12.2 cm), taking 201 points
in the frequency scanning. The 201-points impulse response
from 0 to 300 ns was obtained through an inverse fast Fourier
transformation performed by the VNA with a frequency span
of 1500 MHz (1.5-3.0 GHz).

Network Analyzer
Agilent 8753ES
Transmit Antenna
Receive Antenna
2.4-2.5 GHz
Port 1 Port 2
Floppy
Disk
PC - Processing

Fig. 1. Measurement setup.

The measurements were conducted connecting a standard 0
dBi dipole to port 1 (transmitter) and a receive antenna to port
2 and placing one in front of the other, perfectly aligned. The
receive antenna was chosen from a set consisting of another 0
dBi dipole, a 10-element 11 dBi Yagi, and a LHCP 18 dBi Helix,
in order to set different scenarios and measure the influence of
polarization and directivity of the receive antenna in the
characterization of the indoor wireless channel.



Fig. 2. Measuring the channel parameters.


A calibration was first done to measure the effect of auxiliary
elements as cables and adapters by connecting transmitting
and receiving circuits one to the other, skipping both
antennas. In order to ensure that the incident and transmitted
waves could be approximated as plane waves, the antennas
were placed 30.5 cm (1 ft) apart as point of reference, which is
larger enough than the far-field distance of 5.5 cm (r = 2d
2
/?,
where d = length of transmit dipole).



Fig. 3. Frequency Response Measurement.

The transmitter antenna was stationary while the receiver
was being moved along a linear path by means of a manual
positioning system scaled by the 1-ft square-carpeted floor.
Frequency response measurements (transfer functions) as the
sample showed in Figure 3 were taken every 5 ft along a 70 ft
linear path. Impulse response measurements were taken every
3 ft along a 15 ft linear path (fraction of the first 70 ft path). A
typical impulse response measurement (multipath delay profile)
is shown in Figure 4. At each position, a set of four
measurements spaced equally by a fraction of wavelength were
taken for further averaging and processing.



Fig.4. Impulse Response Measurement.
Main Path
Multipath
EE245 WIRELESS DESIGN LABORATORY: 2.4 GHZ INDOOR CHANNEL

3
B. Environment Description
The measurements were performed on a hallway in the
basement of the Packard Building (Electrical Engineering) of
Stanford University. The measurement environment was the
same for each antenna configuration as shown in Figure 5.



Fig. 5. Hallway environment showing carpeted floor and ceiling covered
with polystyrene tiles and the external metallic conduit.

The outer surfaces of the walls in the hallway are made of
plasterboard and a thin layer of paint. The inner part of the
structure is made of wood with metal pillar. The floors are
protected with a square-grid synthetic carpet. The ceilings are
covered with polystyrene tiles and include an external metallic
conduit and fluorescent lamps positioned across the corridor.
The Hallway contains doors on both sides.

13" 19"
49"
Tx
69.5"
38"
19"
6 ft
3 ft
Conduit
70 ft
Tx
6 ft
13"
19" 4 ft
5 ft
33 ft
4 ft 4 ft
9 f t
Front View
Top View
Packard Building
Basement

Fig. 6. Floor plan of the measurement environment.

The transmitter was mounted in a fixed position at a height
of 176.5 cm above floor level and at one end for the hallway
propagation measurements. The receiver at the other end and
at the same height was moved in steps (5-ft for frequency
response and 1-ft for impulse response) to different fixed
locations along the center line shown in Figure 6. Due to the
movement of people and equipment indoor propagation
channels are in general, time varying. Since the purpose of
these measurements was to investigate the wireless channel
spatial variability, it was essential to keep the channel
stationary during measurements. The experiments took place
during the night (1 am to 5 am) to ensure that the channel was
free of movement by people and equipment.

C. Antenna Design
The main objective of the project is to compare the
parameters of the wireless indoor channel at 2.45 GHz for
different transmit-receive antenna configurations. In that
sense, we fixed the transmit antenna (Dipole Vertically-
Polarized) and measured the impulse response and frequency
response of the channel for the following receive antenna
scenarios:
a. Dipole Vertically-Polarized
b. Yagi Vertically-Polarized
c. Helix Left-Hand-Circularly-Polarized
d. Yagi Horizontally-Polarized
Both transmit and receive antennas were mounted at the
same height above floor level and 125 cm from the ceiling. The
innovative mounting structure showed in Figure 5 was
designed based on portable and mobile coat hangers, which
allowed the mobility required for the different set of
measurements. Following is a brief description of the antennas
employed in this project.

1) Dipole: The Linearly Polarized half-wavelength Dipole
antennas used in the measurements were obtained from the
equipment of the EE245 Wireless Design Laboratory of
Stanford University. Both were designed and constructed
following the criteria of 0.95?/2 (about 5.80 cm for 2.45 GHz) as
the length of the radiating element. Therefore, each pole has a
length of 0.4755? (2.9 cm) such that the antenna can generate
an omnidirectional pattern with 0 dBi of gain. The transmit
dipole is shown in Figure 7.



Fig.7. Dipole 0 dBi antenna.
EE245 WIRELESS DESIGN LABORATORY: 2.4 GHZ INDOOR CHANNEL

4
2) Yagi: The Linearly Polarized 10-elements Yagi antenna
showed in Figure 8 was designed for 2.45 GHz and a bandwidth
of 35 MHz following the definitions in [4] and [5] as well as the
results of the DOS-based software ANTDL6WU by Gnter
Hock.


Fig.8. Yagi Antenna.

The antenna was created with copper material: 1 ft of 3/16
round tube for the boom and 1/8 round tube for the
elements. The length of the elements is shown in Table I.

TABLE I
LENGTH OF YAGI ANTENNA
ELEMENTS
Element Length (mm)
Director 8 50.54
Director 7 50.54
Director 6 50.98
Director 5 51.42
Director 4 51.86
Director 3 52.75
Director 2 53.87
Director 1 54.77
Driver 58.07
Reflector 63.48
We lineup the tip of the boom at 278.11 mm on a straight
edge ruler: 276.11 mm for the director 8 element plus 2mm to
space it from the tip of the square tube. Next, we marked on the
boom the values indicated in Table II which left approximately
26.69mm after the reflector element to mount the Yagi.

TABLE II
ARRANGEMENT OF YAGI ANTENNA
ELEMENTS
Element Reference
Director 8 276.11mm
Director 7 235.47mm
Director 6 196.55mm
Director 5 159.47mm
Director 4 124.97mm
Director 3 94.03mm
Director 2 67.51mm
Director 1 45.41mm
Driver 35.71mm
Reflector 13.00mm
Each of the elements was soldered to the boom at the
respective positions and fixed in place with epoxy superglue
which is an almost non-conductive material. Special care was
taken to center the elements to the boom. The driver was a ?/2
folded dipole. For impedance matching, the folded dipole was
directly connected to a 50O RG-8 coaxial cable. We did not
use any special matching system (like the ?/4 balun) since we
obtained appropriate values of SWR with the simple
configuration mentioned before. As can be seen in Figure 9,
the value of SWR at 2.45 GHz is about 1.8 whereas the minimum
was reached at 2.485 GHz. The elevation and azimuth Half
Power Beam Width (HPBW) were about 30 which are slightly
lower than the theoretical value of 35 and 32. In addition, the
obtained gain was 5.3 dBi, which is less than the designed
value of 11 dBi. The deviations can be explained by the fact
that we were not able to cut the elements at the exact lengths
indicated in Table I and to solder those elements in the precise
positions shown in Table II.



Fig.9. SWR for the Yagi Antenna in the range 2.4-2.5 GHz.. The scale is
0.5 per vertical division where point 2 indicates SWR = 1.12

3) Helix: This left-handed circularly polarized (LHCP)
antenna shown in Figure 10 is derived from information on
helix antennas in [4], [5] and was built by wrapping a 1/8
copper tube around a dielectric support of 45 cm made of 40mm
PVC plumbing pipe. In order to obtain a theoretical gain of 18
dBi, the spiral should consist of 17 turns equally spaced by 2.8
cm. The ground plane for the antenna is an 8 aluminum
endcap with flat base since we couldnt find a smaller diameter
to satisfy our design requirements.

C
?
= circumference of winding = 0.75? -1.33?
S
?
= axial length of one turn = 0.2126C
?
- 0.2867C
?
D = diameter of ground plane = 0.8? -1.1?
G
dB
= 11.8 + 10*log
10
(n*S
?
*C
?
2
) , where n = numbers of turns
HPBW = half power beam width =

S n C
o
*
52
(2.1)

EE245 WIRELESS DESIGN LABORATORY: 2.4 GHZ INDOOR CHANNEL

5

Fig.10. Helix Antenna.

To construct the antenna we used epoxy superglue each 4
revolutions to hold the copper tube around the tube, following
the spiral spacing. As we approached the base end, we didn't
glue down the last turn and did leave some excess wire
hanging off the end. In order to match the antenna from its
nominal 150 O to the cable 50 O impedance, we should have
used a 15-mm-wide triangular strip of copper (impedance
transformer) to connect the coaxial line to the copper spiral by
soldering the narrow pointy end of the strip to the wire and
match up where the lower corner at the large end will solder
neatly onto the stub of the BNC connector. However, we used
a smaller strip which affected the antenna performance and
increased its SWR. We also used epoxy to hold it the matching
system in place.



Fig.11. SWR for the Helix Antenna in the range 2.4-2.5 GHz.. The scale
is 0.5 per vertical division where point 2 indicates SWR = 2.37

Efficiency of the antenna was obtained by measuring the S
11

parameter in the VNA as can be seen in Figure 11.
Unfortunately, since we were not able to match the antenna
appropriately, we obtained values of SWR in the range 2.37-
3.37 which are not excellent but good enough for our project
purposes. Although the elevation and azimuth HPBW of this
antenna were about 20 which is slightly lower than the
theoretical value of 24, the antenna gain obtained (5.3 dBi)
was very different to the expected which could be explained by
the exceeded diameter of the ground plane as well as our poor
matching system.

D. E-field simulation model
In order to compare the measurements with the ideal case of
a hallway with perfectly conducting walls, we developed a
numerical model based on the Finite Difference Time Domain
method, or FDTD. This method was originally proposed by
Yee [13] in 1966, when he described the basis of the FDTD
numerical technique for solving Maxwells curl equations
directly in the time domain on a space grid. In his proposed
method, he described how the space can be divided in a
regular grid, where each point has associated values of E and
H fields and the whole grid is updated every time step.
Tafflove and Brodwin later obtained the correct numerical
stability criterion for Yees algorithm, based on the appropriate
election of the grid spacing (x and y) and the time step t
[14, 15]. Later, in 1994, Berenger [16] introduced the Perfectly
Matched Layer (PML) boundary conditions to allow free space
propagation in a finite closed grid. In this paper we will
implement the Berenger PML boundary conditions, using the
field splitting method for y < 0 m and y > 22 m, while the path
walls were assumed yo be perfectly conductors at x = 2.4m to
the right of the hall and at different locations to the left,
depending on the layout profile seen in figures 6 and 13. We
are using the same set up as described in section B.



Fig.12. The FDTD grid. In this paper we are presenting a 2D simulation
based on the TM
z
mode, where we calculate the values of E
z
, H
x
and H
y

fields.

To determine the values of x, y and t we will use: x =
y = d/40 = 0.06 m and t = x / (2

v
p
) = 0.14 nsec, where d =
7.9 ft = 2.4 m, is the hall width at y = 0 m, and v
p
= 2.99.10
8

m/sec, is the propagation speed of the electromagnetic
radiation in free space. These values were chosen following
the criterion of stability developed by Taflove [14, 15, 17],
trying to use as much as possible our computational power of
memory and processing time. We will launch an incident field
separately from the total and scattered fields. This incident
field will be launched at x = 1.5 m and y = 0 m:
EE245 WIRELESS DESIGN LABORATORY: 2.4 GHZ INDOOR CHANNEL

6

( )

,
_

,
_



2
sec 6 . 0
sec 1 . 2
exp , 5 . 1 ) , , (
n
n t
y m x t y x Einc
(2.2)

The election of this type of wave to be launched in the
hallway was one based on the measurement setup. In effect, as
described in the previous section, we mentioned that for the
time domain measurements we used a bandwidth of 1.5 GHz. In
order to follow this set up as close as possible, we chose a
Gaussian pulse of the same bandwidth for the simulation.
We are using the Total Scattered fields method [18] for
this simulation, dividing the grid in two regions. For y > 10 cm
we will use total fields in the calculations and for 0 cm < y < 10
cm we will use only scattered fields. The incident field will be
run separately from the total and scattered fields, and will be
added to the scattered field at y > 10 cm. (Refer to Figure 13 for
the coordinate system used in the simulations)


Fig.13. Hallway layout, top view used for the numerical simulation. The
Point labeled Tx, correspond to the transmission antenna, while the
points Rxn, n = 16, are the locations of the receiving antenna, where
E
z
was recorded, at 20 ft, 30 ft, 40 ft, 50 ft, 60 ft, 70 ft respectively.

As said before, to do the simulation problem we will
implement the Berenger PML boundary conditions, with a PML
region 12 cells thick. This method requires an implementation
of an anisotropic medium to absorb the waves traveling away
from the grid, simulating an infinite long hall. In order to do this
me use a conductivity profile based on a 4
th
order polynomial
that goes from 0 to a maximum value
xmax
or
ymax
, depending
on the direction where we want to attenuate the wave (hence
the term anisotrpic). In this case the walls along the y =0 and
y=2.4 m axis are perfect conductors, therefore we only need to
calculate the PML boundaries for y < 0 m and y > 22 m. In order
to do this we must implement the difference equations to
separate the conductivity in the x and y axes. We start from the
expressions of the two dimensional TM
z
case [17]:


y
H
E
t
E
x
H
E
t
E
x
E
H
t
H
y
E
H
t
H
x
zy y
zy
y
zx x
zx
z
y x
y
z
x y
x





*
0
*
0
(2.3)

Discretizing the previous equation we get the following
update equations. These are the update equations we will use
here to calculate wave propagation. The indexes i and j are the
discrete representation of the grid, they are spatial indexes,
while n is the temporal index, representing the progression in
time by the time step t:


( )
) 2 (
2
2
2
2 / 1
*
0
, 1 ,
2 / 1
*
0
2 / 1
*
0
2 / 1
2 / 1 ,
2 / 1
2 / 1 ,
t y
t
E E
t
t
H H
j
y
n
j i
z
n
j i
z
j
y
j
y
n
j i
x
n
j i
x
+

1
1
1
]
1

+


+
+
+
+
+
+
+



(2.4a)


( )
) 2 (
2
2
2
2 / 1
*
0
, , 1
2 / 1
*
0
2 / 1
*
0 2 / 1
, 2 / 1
2 / 1
, 2 / 1
t x
t
E E
t
t
H H
i
x
n
j i
z
n
j i
z
i
x
i
x n
j i
y
n
j i
y
+

+
+
1
1
]
1

+


+
+
+
+

+
+
+



(2.4b)


) 2 (
2
2
2
,
2 / 1
, 2 / 1
2 / 1
, 2 / 1
,
,
,
1
,
t x
t
H H
t
t
E E
i
x j i
n
j i
y
n
j i
y
i
x j i
i
x j i n
j i
zx
n
j i
zx
+


,
_

+
+
1
1
]
1

+


+

+
+
+



(2.4c)


( )
) 2 (
2
2
2
,
2 / 1
2 / 1 ,
2 / 1
2 / 1 ,
,
,
,
1
,
t y
t
H H
t
t
E E
j
y j i
n
j i
x
n
j i
x
j
y j i
j
y j i n
j i
zy
n
j i
zy
+

1
1
]
1

+


+

+
+
+



(2.4d)


1
,
1
,
1
,
+
+ +
+
n
j i
zy
n
j i
zx
n
j i
z
E E E (2.4e)

We see from these update equations that the conductivities
along x and y depend only on the respective indexes i and j, as
was expected. The electric and magnetic conductivities will
have the mentioned before, 4
th
order polynomial grading as:

[ ]
d
R m
d
y
y
y y
m
y

+

,
_

0
0
max max
2
) 0 ( ln ) 1 (
) (


(2.5)
EE245 WIRELESS DESIGN LABORATORY: 2.4 GHZ INDOOR CHANNEL

7
This expression is taken from [17], and was derived from
several experiments with the PML, under different conditions.
In this equation we used a 4
th
order polynomial, therefore m = 4.
We also used R(0) = e
-8
and d = 12.x = 0.72 m. With these
values we get
y max
= 0.0741 S/m. These were the values used
for this simulation. The main objective of this gradual increase
of conductivity, as Berenger specified, is to avoid reflections
from the absorbing layer back into the grid, and at the same
time attenuate as much as possible all the waves going into the
PML. To simplify the code, we will use the same above update
equations for all i and j, defining
x
|
i
=
y
|
j
0,
x
*
|
i+1/2
=

y
*
|
j+1/2
0 and |
i,j
=
0
, along the hallway, and using the
expression above in the PML regions, with |
i,j
=
0
in the rest of
the space. In order to implement the perfectly conductive walls,
we just force the scattered field to be the inverse of the
incident field at the walls locations, so we do not have to use

x
|
i
=
y
|
j
?, which is computationally impossible.



Fig.14. Recorded Electric Field magnitude at the receiving points Rxn, n
= 16, located at 20 ft, 30 ft, 40 ft, 50 ft, 60 ft, 70 ft respectively.

We also should mention that, to meet the PML conditions,
in the regions where
x
|
I
,
y
|
j
,
x
*
|
i+1/2
and
y
*
|
j+1/2
?are different
from zero, they must fulfill the Berenger conditions for zero
reflection [16]:


y y x x

0
0 *
0
0 *
(2.6)

Figure 14 shows plots of the total field distribution in time
for the hallway, at the locations shown in Figure 13. We must
say that the total field is calculated adding the incident field
and the scattered field for the region y < 10 cm, where we only
have scattered fields. These calculations can then be used to
estimate the measured channels parameters, using the same
method we will see in section III.

1) Path Loss
The maximum received electric field magnitude squared, |E
z
|
2

(the one corresponding to the main path) at each recorded
location Rxn is giving us a relative idea of the received power
P
r
magnitude at each location. We are expecting to have a |E
z
|
2

decreasing with distance d as (This expression is similar to
(3.1.), used in next section):

|E
z
|
2
(dB)
= K
o
(dB) n*10log
10
d (2.7)

Where n is the exponent of the power-distance relationship
and K
o
is a constant set by the amplitude of the transmitted
pulse. The results of this calculation are shown in Figure 15.


Fig 15. Path loss values obtained from simulat ion. The vertical scale is
dB
(V/m)
2
, not a strict power measure, except for a scale factor, K
o
,
assuming that the medium is isotropic.

2) Delay Spread
The calculation of the delay spread will be specified in the
next section, but we will present some results here from the
simulation, therefore it is good to keep in mind that section as
reference. Using expressions (3.4.) and (3.5.), we can obtain a
value of the rms delay spread,
S
, as defined in [3]. Note that
these two expressions use the received power profile, which
for the simulation was replaced by the electric field square
magnitude as a function of sampled time t
n
, |E
z
|
2
(t
n
), the one
shown in Figure 14. The results for our simulation are shown in
Figure 16.

3) Coherence Bandwidth
The results in time shown in Figure 14 can be used to
estimate the Coherence Bandwidth, B
C
. Performing a Fourier
Transform of this data, it is possible to obtain an estimate of
B
C
. Using as reference the correlation method described in next
section to calculate this parameter from the frequency
spectrum of the signal, we get the values shown in Figure 17.
EE245 WIRELESS DESIGN LABORATORY: 2.4 GHZ INDOOR CHANNEL

8


Fig.16. Obtained values of rms Delay Spread as a function of distance
from the simulated FDTD model.




Fig. 17 Obtained values of Coherence Bandwidth as a function of
distance from the simulated FDTD model.

III. CHANNEL PARAMETERS
A. Path Loss
A short duration pulse, with nanosecond duration, is used
by the VNA to probe the propagation channel. The duration of
the single pulse, inversely proportional to the bandwidth of the
transmission, determines the multipath resolution, i.e., the
minimum discernible path between individual multipath
components. The period of the periodic pulse signal
transmission determines the maximum measurable multipath
delay. Hence, successive multipath components with
differential delay greater than the width of the pulse and within
one period of the periodic pulse transmission can be measured
unambiguously. The single multipath profile of 300 ns
mentioned in section II.A and showed in Figure 4 can capture
two successive probing pulses which is enough for the
purposes of the measurement. Figure 4 also shows that the
response of the first probing pulse has decayed before the next
pulse arrives at the receive antenna which clearly indicates the
presence of path loss.


Fig.18. Path loss exponent for different antenna configurations.

The received power for each frequency of interest at each
point from the transmit antenna was extracted by averaging the
4 equally-spaced (by a fraction of wavelength) frequency
response data sets for each position. As we know, received
power (P
r
) decreases with the distance (d) as

P
r
= P
o
.d
-n
or P
r
(dB) = P
o
(dB) n*10log
10
d (3.1)

where n is the exponent of the power-distance relationship and
P
o
is a constant set by transmitted power and measured system
gain. n is also known as the path loss exponent.
In this project, the values of n were extracted from the slope
of the line that better fits the measured data of receiver power
with distance, both in logarithmic (dB) scale. Figure 18
presents a comparison of the path loss for the four
EE245 WIRELESS DESIGN LABORATORY: 2.4 GHZ INDOOR CHANNEL

9
measurement setup for the indoor environment at 2.45 GHz and
shows the reasonable well-fit log-log line with the respective
values of drop-off exponent.

B. Ricean K-factor
Fast fading or small scale fading in the signal is the result of
small variation in the spatial separation between the receiver
and transmitter which generates depth changes in the signal
amplitude and phase. In case a dominant Line-Of-Sight (LOS)
component is present, the amplitude distribution of fast fading
is characterized by the Ricean probability distribution function
(pdf). Otherwise, the Rayleigh pdf is the best characterization
of that small scale fading. In an indoor environment, a LOS
path between transmit and receive antenna is expected. The
Ricean K-factor is defined as the ratio between the LOS signal
power and the power of the multipath component [3]. The
Ricean pdf is given as [3]

'

<

,
_

+
0 0
0
.
) ( 2
2
2
2
2 2
r
r
r R
I e
r
r f
o
R r

(3.2)

where R denotes the peak amplitude of the dominant signal, s
2

is the power of the multipath components and I
o
(.) is the
modified Bessel function of the first kind and zero-order. The
Ricean K-factor is given as [3]
2
2
2
R
K (3.3)


Fig.19. Ricean K-factor for different antenna configurations.

The Ricean K-factor was extracted from the measured data
set by calculating the relationship between the power of the
LOS path (the first one arriving to the receive antenna) and the
average power of the multipath. Figure 19 shows a comparison
of the Ricean K-factor for the four scenarios under test. The
average values of the K-factor for each configuration are also
indicated in Figure 19. It can be seen that for most cases, the
K-factor is above 3 dB. From this, it can be concluded that the
Ricean distribution describes the channel better than the
Rayleigh distribution, which is a good approximation if K-
factor is < 3 dB. As can also be seen, the Ricean K-factor is
almost independent of the distance from the transmitter.

C. Delay Spread
Different reflections in the surrounding environment, will
introduce mltipath effects in the signal. This multipath effect
manifests as spreading in time of the received signal, compared
with the transmitted waveform. This characteristic is quantified
by the Delay Spread parameter. This parameter is extracted
from the measured power profile, as a function of each sampled
time instant t
n
, P(t
n
). Figure 4 shows a typical measurement of
P(t
n
) that we obtained during our experiment. As we find in [3],
the most commonly used time delay parameters in channel
characterization are the Mean Excess delay m

and the rms


Delay Spread,
S
.


n
n
n
n n
t P
t t P
m
) (
) (

(3.4)


( )

n
n
n
n n
S
t P
m t t P
) (
) (
2

(3.5)


Fig. 20. Rms Delay Spread for different antenna configurations.

As described in section II.A, we performed frequency and
time domain measurements for the four different antenna
setups. From the time domain measurements, using these
expressions to calculate
S
, we get the results shown in Figure
20.

EE245 WIRELESS DESIGN LABORATORY: 2.4 GHZ INDOOR CHANNEL

10
D. Coherence Bandwidth (B
c
)
We measure the coherence bandwidth to be able to get an
estimate on the frequency range for which the channel can be
assumed to be flat. To be able to do this, we measured the
Frequency Response (Figure 3) of the channel and calculated
the normalized auto-correlation function (Figure 21), a similar
method as the one utilized in [6].
0
0.2
0.4
0.6
0.8
1
0 20 40 60 80 100
Frequency (MHz)
N
o
r
m
a
l
i
z
e
d

A
u
t
o
c
o
r
r
e
l
a
t
i
o
n

Fig. 21. Normalized Autocorrelation function for Frequency Response

We then defined the Coherence Bandwidth as the average
frequency spacing over different samples for which the auto-
correlation of the Frequency Response data was above a
certain threshold (i.e. 0.9, 0.7 and 0.5). Figure 22 shows the
results obtained for a 0.7 correlation threshold at different
distances. As we can see, B
c
tends to decrease as the distance
increases. This phenomenon could be explained by the fact
that with longer distances, the relative power of the non line of
sight paths compared LOS path increases, accentuating the
effect of multipath and reducing B
c
.
The Cumulative Distribution Function (Figure 23) of the
different measurements clearly show that the worse
performance was on the Dipole-YagiCross setup. This is
probably because the LOS path under this setup is severely
attenuated due to the antenna polarization, thus the multipath
effect is more severe. Furthermore, it could even be argued that
if the antennas are perfectly cross polarized, we would obtain a
Rayleigh distribution in the fading characteristics rather than
the more desirable Rician distribution.

TABLE III
AVERAGE COHERENCE BANDWIDTH (MHZ)
Correlation 0.9 0.7 0.5
Dipoles 3.12 12.34 24.81
Dipole -Yagi 6.62 16.78 26.68
Dipole Yagi Cross 2.61 9.14 18.91
Dipole - Helix 4.00 13.51 26.51

The performance of the two-dipole setup was also worse
than the Dipole-Yagi and Dipole-Helix setups since we can
expect more multipath effects due to the lack of directivity in
either the receive or transmit antennas. Finally, it is clear that
under LOS conditions, it is advantageous to utilize directive
antennas to minimize the multipath effects, increasing the
Coherence Bandwidth.
Corr > 0.7
0
5
10
15
20
25
30
35
0 10 20 30 40 50 60 70
Distance (feet)
C
o
h
e
r
e
n
c
e

B
a
n
d
w
i
d
t
h

(
M
H
z
)
Dipoles Dipole-Yagi
Dipole-Yagi-Cross Dipole-Helix
Corr > 0.9
0
2
4
6
8
10
12
0 20 40 60 80
Distance (feet)
C
o
h
e
r
e
n
c
e

B
a
n
d
w
i
d
t
h

(
M
H
z
)
corr > 0.5
0
10
20
30
40
50
0 20 40 60
Distance (feet)
C
o
h
e
r
e
n
c
e

B
a
n
d
w
i
d
t
h

(
M
H
z
)

Fig. 22. Coherence Bandwidth vs. Distance for a 0.7, 0.9 and 0.5
correlation threshold.

Table III summarizes the measurement results under the
different setups. Depending on the configuration and on the
correlation threshold, we estimated the Coherence Bandwidth
of the channel to be between 3 and 27 MHz for our
experiments. Under a more conservative stand we would
estimate that the Coherence Bandwidth is between 3 and 7
MHz. Thus, the general believe that B
c
= 1/
s
is off by a factor
of ten in our measurements, since we estimated the delay
spread to around 8nsec. Based on our results, this relationship
would be closer to B
c
1/(10
s
).

EE245 WIRELESS DESIGN LABORATORY: 2.4 GHZ INDOOR CHANNEL

11
0
0.2
0.4
0.6
0.8
1
1.2
0 10 20 30 40
X=Coherence Bandwidth
P
r
o
b
a
b
i
l
i
t
y

x
<
=
X
Dipoles Dipole-Yagi
Dipole-YagiCross Dipole-Helix

Fig. 23. Cumulative Distribution Function of measured Coherence
Bandwidth Values (Corr > 0.5)

IV. CONCLUSIONS
Results confirm the impact of the different antenna
configurations on the channel parameters. Antenna directivity
improves multipath characteristics of the wireless channel.
Circularly polarized antennas additionally show more stable
performance since there is no dependence on the relative angle
between transmit and receive antennas. In general, as received
power increases the channel performs better showing higher
coherence bandwidth and lower delay spread values.
The path loss factor is found to be lower (n = 1.7-1.8) than
the free space attenuation in a hallway environment. However,
comparing with the results of the simulation model, the
characteristics of the hallway differ from the ideal conductive
waveguide.
Dominant Line-Of-Sight path is expected in indoor
environments like hallways. The mean value of the Ricean K-
factor is approximately independent of distance and varies
from 16 dB to 30 dB for different antenna configurations. These
high values confirm Ricean Fading characteristics.
Comparing the measured results of Delay Spread and
Coherence Bandwidth, we find that the relationship between
Bc and s s is not exactly inverse but B
C
1/(10 s
s
). This is also
found in the numerical simulations, even thought the values of
simulated Delay Spread are higher than the measured ones,
and the values of simulated Coherence Bandwidth smaller than
the ones measured. Based on our measurements, a
conservative estimate of the Coherence Bandwidth of the
channel is between 3 and 7 MHz based on the antenna
configuration.
The numerical simulation shows that the ideal, perfectly
conducting walls is not a faithful representation of the walls in
the hallway. This conclusion is derived mainly from the path
loss factor obtained for the simulation (n 1), which is the one
we would obtain modeling the hall as a waveguide. In this case
the energy is concentrated in the y direction, allowing for a
higher delay spread, and a decrease of power as a function of
distance d, as d
-1
, instead of d
-2
, which is usually the case in
free space.
REFERENCES
[1] T.A. Wysocki and H. Zepernick, Characterization of the Indoor
Radio Propagation Channel at 2.4 GHz , Journal of
Telecommunications and Information Technology, 2000, vol. 3,
no. 4, pp. 84-90.
[2] J. S. Davis and J. M. Linnartz, Indoor Propagation at 2.4 GHz in
Wireless Communications Demo Edition 1997, unpublished.
[3] T. S. Rappaport, Wireless Communications: Principles & Practice.
New Jersey: Prentice Hall, 1996.
[4] W. L. Stutzman and G.A. Thiele, Antenna Theory and Design. New
York: John Wiley & Sons, 2nd ed., 1998.
[5] American Radio Relay League, The ARRL Antenna Handbook. R.D.
Straw Ed., 19th ed.., 2000
[6] I. Cuias and M. Garca Snchez, Measuring, Modeling, and
Characterizing of Indoor Radio Channel at 5.8 GHz, IEEE Trans.
Vehicular Technology, vol. 50, no. 2, pp. 526535, March 2001.
[7] M. Win, R.A. Scholtz, and M. A. Barnes, Ultra-Wide Bandwidth
Signal Propagation for Indoor Wireless Communications, in IEEE
Int. Conf. Communications, pp. 5660, June 1997.
[8] A. Bohdanowicz, G. J. Janssen, and S. Pietrzyk, Wideband Indoor
and Outddor Multipath Channel Measurements at 17 GHz,
unpublished.
[9] J. Purwaha et al, Wide-band Channel Measurements at 60 GHz in
Indoor Environments, unpublished.
[10] A.G. Simarou and M.O. Al-Nuaimi, Wideband Propagation
Measurements for Indoor Radio Channels at 62.4 GHz,
unpublished.
[11] D. Cox and R. P. Leck, Correlation Bandwidth and Delay Spread
Multipath Propagation Characteristics for 910 MHz Urban Mobile
radio Channels, IEEE Trans. Communications, vol. 23, no. 11,
pp. 12711280, November 1975.
[12] D. Cox, 910 MHz Urban Mobile Radio Propagation: Multipath
Characteristics in New York City, IEEE Trans. Communications,
vol. 21, no. 11, pp. 11871194, November 1973.
[13] Yee, K. S., Numerical solution of initial boundary value problems
involving Maxwells equations in isotropic media, IEEETtrans.
Antennas and Propagation, Vol. 14, 1966, pp. 302-307.
[14] Taflove, A., and M. E. Brodwin, Numerical solution of steady
state electromagnetic scattering problems using the time-dependent
Maxwells equations, IEEE Trans. Microwave Theory and
Techniques, Vol. 23, 1975, pp. 623-630.
[15] Taflove, A., and M. E. Brodwin, Computation of the
electromagnetic fields and induced tempereatures within a model of
the microwave-irradiated human eye, IEEE Trans. Microwave
Theory and Techniques, Vol. 23, 1975, pp. 888-896.
[16] Berenger, J. P., A perfectly matched layer for the absorption of
electromagnetic waves, J. Computational Physics, Vol. 114, 1994,
pp. 185-200.
[17] A. Taflove and S. C. Hagness, Computational electrodynamics: the
finite-difference time-domain method.- 2
nd
ed., Artech House,
2000.
[18] Umashankar, K. R. and A. Taflove, A novel method to analyze
electromagnetic scattering of complex objects, IEEE Trans.
Electromagnetic Compatibility, Vol. 24, 1982, pp. 397-405.

Das könnte Ihnen auch gefallen