I

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY. VOL. 40, NO. 3. AUGUST 1991 569

Experimental Evaluation of
Space/Frequency /Polarization
Diversity in the Indoor
Wireless Channel
Jean-Francois Lemieux, Mohamed S . El-Tanany, Member, ZEEE, and H. M. Hafez

Abstract-A series of experiments that were conducted to specifically at obtaining parameters pertaining to different
evaluate the relative merits of three types of diversities in the
diversity schemes.
indoor portable radio channels at 900 MHz is described. Un- ~ ~and cox i cox,
f [71 and ~M ~~ and Norris
~~ ~181 ~ ~ ,
modulated tones were used to probe the channel in a typical
university work environment. The statistics of the signal en-
used continuous wave (CW) signals to investigate the radio
velopes were analyzed to determine the correlation of the field coverage inside a building when the transmitter was placed
strengths when the signals are separated in space, frequency or outside. Alexander [9], [IO] was one of the first to report on
polarization. According to the results presented here, the perfor-
truly indoor propagation measurements. He used CW signals
Of a narrow-band system ( < 2oo kHz) at 900 MHz to probe the channel on one floor, where both the transmitter
greatly benefit from a frequency diversity scheme provided that
the separation between the two frequencies is greater than
and receiver were placed in locations inside the
approximately 5 MHz. Similarly, the advantage of space diver- buildi%. A h a n d e r and I1l1 used a two-branch
sity is found significant when the two antennas are spaced by space diversity receiver with selection combining to see
3/4 wavelength or more. It is also reported that co-polarized the possible improvements in the signal envelope. Patsiokas
and cross-polarized signals are nearly uncorrelated and their et al. 1121 performed indoor attenuation measurements at
relative field strengths differ by less than 2 dB in most locations.
several frequencies using cw There have also been
several reports on experimental data in papers devoted pri-
I. INTRODUCTION marily to system studies. In all these studies, however, there
has been little results that directly address the question of
T HE portable indoor radio
is being 'Onsidered
in the 900 MHz range diversity in terms of raw data. One can obtain the diversity
as a medium for future persona' parameters from the impulse response measurements, but that
communications systems [119 121. This consideration has requires a large volume of data and careful interpretation of
spurred a wide interest in the propagation characteristics of the results and the
the channel, and their impact on the design and performance In Section II a description of our experiment is given,
of low power portable radio links. followed by a presentation of the results in Section 111, and
A general analysis of channels involving mobile antennas, conclusions in Section Iv.
including diversity schemes, is given in [3]. Over the last few
years, however, several experiments were conducted by many 11. THEEXPERIMENT
researchers to measure the attenuation and fading characteris- A . Experiment Setup
tics of the indoor channel. These experiments can be classi-
A simplified block diagram of the setup is shown in Fig.
fied into two categories: wide-band and narrow-band. The two tones were generated by
In the transmitter (Fig,
purpose Of the wide-band measurements is to Obtain the
a 915 MHz signal with another low frequency signal.
complex-valued impu1se response Of the which rep- The range of the low frequency signal generator is 0.5 to 10
resents a complete characterization of the channel. Examples MHz, which allowed the frequency spacing between the
of the wide-band measurements are: Saleh and Valenzuela frequencies to change Over the range of to 2o MHz. In the
i4i, Devasirvatham i5i and Bultitude 161. The narrow-band
of space/polarization tests only one tone was required,
On the Other hand, are aimed at SO the mixer and the low frequency signal generator were
obtaining specific parameters which relate directly to the
disconnected,
design of a narrow-band system. They are considered to be a
The total maximum RF power at the amplifier output was
partial characterization of the channel. The work reported in 3o dBm, and the transmit antenna was a wavelength
this paper falls in the narrow-band category, and it is aimed monopole with a ground plane.
The receiver (Fig. l(b)) has two configurations, one for
Manuscript received June 20, 1990; revised December 3 , 1990.
The authors are with the Department of Systems and Computer Engineer. the frequency diversity tests and One for the space/polariza-
ing, Carleton University, Ottawa, ON, Canada KIS 5B6 tion tests. The difference between the two configurations is in
IEEE Log Number 9101618. the antenna connections. In the case of the frequency diver-

0018-9545/91$01.00 0 1 9 9 1 IEEE
570 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, V O L . 40, NO. 3, A U G U S T 1991

Synh 1 BP filter square-law envelope detector, and sampled and stored on the
hard disk of a PC computer as channels 1 and 2.
The data acquisition system consisted of a dual channel
12-bit A/D converter with its utility software and a PC
spacejpolarization frequency
diversity diversity computer. The selection of timing parameters were dictated
by four constraints:
1) the sampling rate should be large compared to the
bandwidth of the envelope signal;
.5 MHz to IO MHz 2) the sampling on the two channels should be almost
+2 dBm
k instantaneous;
3) the length of each test run should be large enough to

U
O to 12' dB
step atten provide good statistics;

d
attenuator 4) hardware limitations should be observed; this includes
the required storage, and the interaction between the
114 wavelength controlling software and the A/D conversion.
mono-pole antenna
with ground plane Taking these factors into consideration the following parame-
Tx Amp1
ters were selected:
gain= 41 dB
1) each channel is sampled at a rate of 500 samples/s;
(a)
2) the time difference between two corresponding samples

v antennas
on each channel is 60 ps;
3) the time interval of each test run is 1 min.

B. Measurements Procedure
The measurements were performed in 20 locations on the
I diverstiy same floor. The floor plan is shown in Fig. 2 on which the 20
divider transmitter locations and the receiver location are indicated.
freq diversity
The drafting rooms shown in the plan are furnished each with
40 wooden drafting tables. The two laboratories have wooden
cabinets and tables, some equipment and a few metal cabi-

n
nets. There are six computer terminals in the terminal room.
The offices and study rooms have mostly wooden furniture.
E= 70 MHz xlal filter
The hallways are laid out in an H-shape. In the center, there

gain= 22 dB
gain= 22 dB lnqll
is a large skylight going right through to the floors below.
The windows in the lab, the skylight, some study rooms and
some outside windows have wire mesh in their center. The
walls are mainly made of plasterboard with wood studs and
some metal pillars. The floors are made of cement covered
with tiles. The ceilings are all cement with sound proofing
(b)
panels held by metal frames at about 50 cm from the ceiling.
Fig 1. (a) Transmitter. Frequency and space polarization. (b) Receiver. In the interspace there are many cables for computers, tele-
Frequency and space polarization diversity.
phones and power line wiring.
During the tests, the receiver was located in one of the
sity, only one antenna is required; the signal from that laboratories, while the transmitter was moved to the various
antenna was split using a power divider and fed to the two test locations. At each location, the exact antenna position
branches of the receiver. In the case of space/polarization was recorded by measuring the distance between the antenna
diversity two separate antennas were connected to the two and an easily recognizable building feature. With these mea-
branches of the receiver. The receive antennas were identical surements the location could be reported on a scaled floor
to the transmit antennas, and were mounted on a mechanism plan and the exact distance from the transmitter to the
that allowed us to change the distance between them and to receiver would then be computed.
change their orientation in the vertical plane. The received Upon bringing the transmitter to a test location and record-
signals in the two branches were mixed by separate local ing the antenna position, a short test was done to select the
oscillators such that the IF frequencies matched the center attenuator settings. If the signal level was not fully within the
frequencies of two identical crystal filters. The nominal cen- dynamic range of the receiver, the attenuator would be
ter frequency of the filters was 70 MHz, and the nominal 3 adjusted and another test run would be performed.
dB bandwidth was 25 kHz. During the data recording, the receive antenna was moved
The envelopes of the two IF signals were detected using a by hand along a circle cut inside the antenna mounting frame.
LEMIEUX et al.; DIVERSITY IN INDOOR WIRELESS CHANNEL 57 I

,
* - L-
17
16 19
45.00 I I

11 srudy r w m
0.00
100

Fig. 3. Attenuation versus distance. Example 1. Two tones separated by 1

MHz. All locations.
rem'nal

1-

Lh
mechanical
skylighl

laboraioq

study room
106

R -3 t_ 6 In fit chan. 1
__ .._.
.............
office
7J ,' I
Fig. 2. Floor plan of the measurement area and locations position. R 0.001
100 " " I 10'
' ' ' 102
distance (meters)
indicates the receiver's location.
Fig. 4. Attenuation versus distance. Example 2 . Single tone received by
two antennas 0.75 X (24.6 cm) apart. All locations.
The circle has a diameter of one wavelength of the lowest
-
frequency (about 33 cm). Moving the receive antenna is
necessary for two reasons: first, it averages out the received
field in the neighborhood of the selected location; and sec-
45.00 1 '
. '
8 .../'
,/;6
6
.,:..;/
Q ,
' ' '

.?6;4
ond, it simulates more realistically the actual use of the
hand-held equipment.
Each test was run for one minute. The test was repeated 0
..
U
.
+
.d'/
Q'd' -

cd
for five different frequency spacings, five different antenna
C
spacings and one polarization setting. All tests were run in
quick succession one after the other to capture the same
8cd15.00 -
Q) ... , channel 2
environment.
.... ,
UI
.
i'
cd
* ,
/
111. RESULTS
2
A. Attenuation Versus Distance
0.00 .
-
- /',
'_ ' lo' ' ' ' '
The first set of results concerns the attenuation rate in the
floor described in the previous section. Fig. 3-5 show three
examples of the relative attenuation versus distance. In each
of the three figures, the vertical axis represents the excess
attenuation relative to the attenuation at a reference distance Fig. 3 shows the measured attenuation at different loca-
of approximately 2 m. This distance was selected on the tions for two tones separated by l MHz (the two tones are
assumption that it represents the minimum distance between a labeled channels 1 and 2). As the figure illustrates there is no
transmitter and a receiver in an actual system. The horizontal significant statistical difference between the attenuation of the
axis represents the distance between the transmitter and the two tones. The regression lines of the two channels are
receiver on a logarithmic scale. (Each point on the graph is almost identical. It should be pointed out, however, that there
an average of 80 different readings taken at the same loca- is a correlation between the distance and the channel condi-
tion.) tions. Specifically, short distances implies the existence of a
572 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 40, NO. 3, AUGUST 1991

partial line-of-sight. The reader should refer to the floor plan

to identify the reception conditions of each distance. The 45.00
relation between the distance and channel conditions explains h

E-50.00
the fact that short distances have points below the regression
line while longer distances have points above the line. The 8-55.00
Q)
slope of the fitted lines is approximately 36 dB/decade. The
distance-power constant, a , is marked on each fitted line in -
~:-60.00
3
.e

Figs. 3-5. Y5.O0

Fig. 4 shows an example of the attenuation rate for signals 6-70.00
received by two antennas separated by 314 wavelength. -75.00
Again, the signals from the two antennas have near identical
-80.00
fitted lines, with nearly the same slope as the case of two
tones on the same antenna. -85.00 "'"'""''"""''""""""'~
1.00 1.50 2.00 2.50 3.00 3.50 4.00 4.50 5.00
In Fig. 5 channels 1 and 2 denote signals received by Time (seconds)
vertically and horizontally polarized antennas. The best fitted Fig. 6. Amplitude versus time. Example 1. Two tones separated by 1
MHz. All locations.
lines of the two channels show a difference of around 3 dB.
Although differences as much as 5 dB were observed when
the line-of-sight was partially present, they diminish as the
distance increases, which is an expected behavior. The atten- 45.00
h

uation of the cross-polarized signal (horizontal antenna) fol- E-50.00

lows nearly the same slope as the co-polarized signal (vertical
%.55.00
0)
antenna).
3-60.00
B. Frequency Diversity .e
4

Fig. 6-7 show two examples of the envelopes of two tones

<-70.00
as a function of time. In Fig. 6 the frequency spacing
between the two tones is 1 MHz, and as can be seen the two -75.00
envelopes exhibit a high degree of correlation. The frequency -80.00
spacing in Fig. 7 is 20 MHz and the two signals appear to be
-85.00 II"'
totally uncorrelated. 5.00 5.50
" "

6.00
" "

6.50
" "

7.00
" ~ "

7.50
" ~

8.00
"

8.50
" ~

9.00
" ' 1 1 " "

Time (seconds)
The degree of decorrelation between the received en-
Fig. 7. Amplitude versus time. Example 2. Two tones separated by 20
velopes can be best examined by evaluating the probability of MHz. All locations.
simultaneous "fading." If the envelope of the frequency f 1
is denoted by s( f 1) and the envelope of the frequency f 2 is
denoted by s( f 2), then, the probability of simultaneous
fading is defined as: Pr[s( f 1) < x and s( f 2) < XI; where x
is an arbitrary level. Fig. 8 illustrates the cumulative proba-
bility of simultaneous fading events for different frequency
spacings. The figure also includes the cumulative probability
of fading on one of the two individual channels. Since the
average power level changes slightly from one test to another
and from one frequency to another in the same test, it was
necessary to normalize all envelopes to their average power.
This normalization allowed us to plot all the results on the
same graph.
Theoretically, if the two channels are uncorrelated, the
probability of simultaneous fading (as defined above) will be -31-29-27-2523-21-19-17-15-13-11-9 -7 -5 -3 -1 1 3 5 7 9 11
Normalized Average Rcccivcd Power (dB)
the product of the two individual fading probabilities. Assum-
Fig. 8. Cumulative probability of simultaneous fading versus normalized
ing that the two normalized channels are uncorrelated the average received power. Two tones separated by 1, 5 , 10, 15 and 20 MHz.
theoretical cumulative probability of simultaneous fading is
plotted in Fig. 8 along with the experimental curves. 9. The figure indicates that a 314 wavelength separation
It is clear from the figure that a 5 MHz separation between between the two antennas results in uncorrelated signals and
the two tones is needed for the two signals to be substantially brings the probability of simultaneous fading close to its
uncorrelated. It can also be concluded that a separation theoretical limit. This confirms results obtained in [ll] by
greater than 10 MHz does not affect the results significantly. Motley and Alexander.

C. Space Diversity D. Polarization Diversity

The procedure discussed for the frequency diversity tests is In the case of polarization diversity, the signal levels on
repeated for space diversity and the results are shown in Fig. the two antennas are inherently different. In this case the
 ~

LEMIEUX er al. : DIVERSITY IN INDOOR WIRELESS CHANNEL 573

-31-2P27-2523-21-19-17-15-13-11-9 -7 -5 -3 -1 1 3 5 7 9 11
Normalized Avcragc Reccivcd Power (dB)
Fig. 9. Cumulative probability of simultaneous fading versus normalized
average received power. Single tone received by ‘two antennas separated by
0.75, 1, 1.5, 1.66, a n d 2 A.

..................
2 uncorr. chan.
2 cross-polar.11111.
_.-.-.-.
10-5 ‘ 1 ’ 1 ’ 1 , ~ ” ‘ ~ ‘ ~ “ “ ~ ‘ 1 ’ ~ ’ ~ ’ ’ ’ ’ (~ I 1’ ’ ~
-31-29-27-2523-21-19-17-15-13-11-9-7 -5 -3 -1 1 3 5 7 9 1 1
Namalizcd A v m g e Rcccived Power (dB)

Fig. 10. Cumulative probability of simultaneous fading versus normalized

average received power. Single tone received by two cross-polarized anten-
nas.

average power level has not been normalized. The results of the antennas is of the order of a 3/4 wavelength. Uncorre-
the polarization diversity tests are shown in Fig. 10, and lated signals can also be obtained if the two signals are
from that figure one may conclude that polarization diversity carried on two frequencies separated by at least 5-10 MHz.
appears to be a very efficient way to combat fading. It should
be pointed out, however, that since there are two physically REFERENCES
separate antennas involved they had to be separated in space. A. J. Motley, “Advanced cordless telecommunications services,’’
IEEE J. Select. Areas Commun., vol. SAC-5, pp. 774-182, June
The actual distance between the centers of the two antennas 1987.
was about 3 cm. This distance, however small, contributed to D. C. Cox, “Universal portable radio communications,” IEEE
the decorrelation between the two received signals. Trans. Veh. Technol., vol. VT-34, pp. 117-121, Aug. 1985.
W. C. Jakes Jr., Ed., Microwave Mobile Communications. New
York: Wiley, 1974.
IV. CONCLUSION A. A. M. Saleh and R. A. Valenzuela, “A statistical model for indoor
In this paper the results of an experimental investigation multipath propagation,” IEEE J. Select. Areas Commun., vol.
SAC-5, pp. 128-137, Feb. 1987.
into the potential advantages of several types of diversities D. M. J. Devasirvatham, “Time delay spread and signal level mea-
were presented. CW signals in the 900 MHz band were used surements of 850 MHz radio waves in building environments,” IEEE
to probe the channel and obtain direct correlation statistics Trans. Antennas Propagat., vol. AP-34, pp. 1300-1305, Nov.
1986.
pertaining to the diversity schemes. Polarization diversity R. J. C. Bultitude, “Measurements, characterization and modeling of
seems to offer a good choice for system designers since it Indoor 800/900 Hz channels for digital communications,” IEEE
does not require extra bandwidth and the spacing between the Commun. magazine, vol. 25, pp. 5-12, June 1987.
H. H. Hoffman and D. C. Cox, “Attenuation of 900 MHz radio
two antennas could be small. The difference in power be- waves propagating into a metal building,” IEEE Trans. Antennas
tween the co-polarized and cross-polarized signals is typically Propagat., vol. AP-30, pp. 808-811, July 1982.
3 dB unless the line-of-sight is present. Signals received with D. C . Cox, R. R. Murray, and A. W. Norris, “Measurements of 800
MHz radio transmission into buildings with metallic walls,” Bell
two antennas become decorrelated when the distance between Syst. Tech. J . , vol. 62, no. 9, pp. 2695-2717, Nov. 1983.
514 IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY, VOL. 40, NO. 3, AUGUST 1991

[9] S . E. Alexander, “Radio propagation within building at 900 MHz,” Mohamed S. El-Tanany (S’80-M’81) received
Electron. Lett., vol. 18, no. 21, pp. 913-914, Oct. 1982. the B.Sc. and M.Sc. degrees in electrical engineer-
[IO] S . E. Alexander and G. Pugliese, “Cordless communications within ing from Cairo University, Giza, Egypt, in 1974
buildings: Results of measurements at 900 MHz and 60 MHz,” and 1978, respectively, and the Ph.D. degree in
British Telecom. Technol. J . , vol. 1, no. 1, pp. 99-105, July 1983. electrical engineering from Carleton University,
[ I l l A. J. Motley and S . E. Alexander, “Diversity advantage for cordless Ottawa, ON, Canada, in 1982.
telephones,” Electron. Lett., vol. 19, no. 14, pp. 530-533, July Between 1982 and 1985, he was employed with
1983. Miller Communications Systems, Kanata, ON,
[12] S . J. Patsiokas, B. K. Johnson, and J. L. Dailing, “Propagation of Canada, where he worked on several research and
Radio Signals inside Buildings at 150, 450 and 850 MHz,” in R o c . development projects related to data transmission
Veh. Technol. Conf., pp. 66-71, May 1986. over the mobile satellite and aeronautical channels.
He joined the Department of Systems and Computer Engineering at Carleton
University in 1985, where he is currently an Associate Professor. At
Carlton, he has been actively involved in several research projects related to
mobile and portable radio systems. His current interests are in the research
and development of data transmission subsystems for the mobile radio and
satellite channels, and the analysis of the synchronization algorithms of such
Jean-Francois Lemieux received the bachelor of subsystems.
engineering (electrical) degree from the Royal Mil-
itary College of Canada and the master’s degree in
electrical engineering from Carleton University, H. M. Hafez received the B.Sc. and M.Sc. degree
Ottawa, ON, Canada, in 1990. in electrical engineering from the University of
He joined the Canadian Forces in 1977. He was Alexandria, Alexandria, Egypt, in 1971 and 1974,
commissioned as a Second Lieutenant in 1982 and respectively, and the Ph.D. from Carleton Univer-
served in Lahr, Germany, as a Signal Officer for a sity, Ottawa, ON, Canada, in 1980.
Signal Squadron and a Tank Regiment from 1983 Between 1980 and 1981 he was with the Federal
to 1986. He then spent a year in Kingston, ON, Department of Communications, Canada, as a con-
Canada, as an Instructor at the School of Commu- sultant in systems engineering in the area of mobile
nication and Electronics. His master’s thesis dealt with diversity measure- communications and spectrum management. In
ments for indoor radio communications. Until the Spring of 1991, he worked 1981 he joined the Department of Systems and
as a Requirement Officer for the Directorate of Communications and Elec- Computer Engineering, Carleton University, where
tronics Systems Requirements. In April 1991 he was promoted to the rank of he is currently an Associate Professor. He has acted as a consultant to
Major and assigned to the Low Level Air Defence Project as the Communi- several industrial and government organizations, including Bell Northern
cations Engineer. Research, Bell Canada, NovaTel and MDI.