References analyser with an HP85046A S-parameter kit.

The output of
1 FLIKUDA, ‘Equilibrium point analysis of ALOHA-type systems’.
A.:
the network analyser (port 1 of the S-parameter kit) was con-
Trans. IECE Japan, 1978,J6l-B, pp. 959-966 (in Japanese) nected to a 22dB amplifier (SHL 2000) through a 1.8m cable.
2 KLEINRCCK, L., and LAM, s. s.: ‘Packet switching in a multiaccess The output of the amplifier travelled through a 40m cable to
broadcast channel: Performance evaluation’, IEEE Trans., 1975, the transmitting antenna. The receiving port of the network
COM-23, pp; 410-423 analyser (port 2 of the S-parameter kit) was connected to the
3 TASAKA, s.: Performance analysis of multiple access protocols’ receiving antenna through a 10m cable.
(Cambridge, Mass: MIT Press, 1986) The calibration of this instrumentation was not mentioned
4 GELENB+ E., and MITRANI, I.: ‘Control policies in CSMA local area by Howard and Pahlavan.’ It can be performed by calibrating
networks: Ethernet controls’, Pe$ormance Evaluation Rev., 1982, out the cables and amplifier using the HP8753 calibration kit,
11, pp. 23S240
5 HARRUS G.: ‘A model for the basic block protocol of the Cam-
placing the two antennas in an anechoic chamber with enough
bridge ring’, IEEE Tram., 1985,SE-11, pp. 13lL136 separation between them such that near-field effects are negli-
gible, measuring the S,, and then dividing all measurements
of the S,, of the channel by this SZl.However, in the absence
of an anechoic chamber, we used a coaxial cable connection in
place of the two antennas. This is justified because the discone
antennas used have low VSWR (15dB maximum). However,
note that all raw measurements were stored.
The theory of operation is that S,, (the ratio of the received
FREQUENCY RESPONSE AND PATH LOSS to the transmitted voltages) as a function of the frequency is
MEASUREMENTS OF INDOOR CHANNEL the discrete frequency response of the channel between ports 1
and 2 of the S-parameter kit. The inverse FFT of this fre-
quency response is the bandlimited impulse response of the
Indexing terms: Radiowaue propagation, Measurement channel. S,, at a single frequency is the relative path loss.
Concerning the time invariance of the channel, the measure-
An improved method of operation of an HP network
analyser to perform measurements of the frequency response ments were collected in night with no other researchers in the
of the indoor RF propagation channel and the relative path building which allowed us to assume that the channel was
loss envelope and phase of a CW signal is described. An time invariant during the 1”s (minimum) to 800ms
improved postprocessing procedure is also described. Test (maximum) of the measurements. The time invariance was
results obtained for indoor measurements in the 1-2GHz confirmed further by ensuring that the results of ten measure-
band (2.3ns resolution after windowing) and 945 MHz path ments performed immediately after each other compared
loss measurements are presented as examples. favourably and by measuring the envelope of a CW signal and
making sure that it did not vary over extended periods of
time.
Introduction: Recently Pahlavan and Howard’ performed
measurements of the frequency response of the indoor RF Measurements, d a t a analysis and results: Measurements were
propagation channel using an H P network analyser. These performed to determine the propagation power law. This was
measurements are becoming increasingly important because of achieved by fixing the frequency of the network analyser at
attempts at introducing wireless PABXs. This Letter describes 945 MHz and moving the transmitting antenna at an approx-
an improved method of operation of the configuration pro- imately constant speed. This was performed for a variety of
posed by Pahlavan and Howard’ to perform measurements of channels. All measurements were made on the third floor of
the frequency response of the indoor RF propagation channel the NovAtel building in Calgary. Fig. 2 shows the result of
in an unprecedented 1 GHz band and of the relative path loss averaging all the measurements on a log-log scale. The error
envelope and phase of a 945MHz CW signal. Though the
setup proposed here is similar to that of Pahlavan and
Howard,’ crucial factors such as the type of antennas used,
the method of operation (the calibration) and data analysis
(windowing before taking the inverse FFT) are different.

Instrumentation and theory of operation: The setup used for

the measurement of the indoor RF propagation channel is
shown in Fig. 1. The system consists of an HP8753A network

-8 0 t
1 2 5 10 20 50 100

I HP 8753A
vector
distance. m 1298121

1- network onalyser Fig. 2 Propagation power law

S -parameter kit
bars shown indicate one standard deviation below and above
the mean. The measurements fitted well to a straight line
(Pearson’s correlation coefficient = 0.97) with a slope of
-38.35 which suggests that the power is proportional to l/r4.
In calculating the impulse response from the frequency
response, the data has to be windowed to minimise overlap of
neighbouring peaks. There are two opposing parameters in
choosing a window: the width of the mainlobe and the sup-
pression of the sidelobes.2 We chose to use a minimum three-
sample Blackman-Harris window which leads to a 6dB
mainlobe width of 2.3 bins, and a maximum sidelobe of
-67dB. Fig. 3 shows the results of taking the inverse FFT of
the averaged frequency response (zero padded to 2048 bins) in
the 1-2GHz band. The average was of ten frequency
Fig. 1 Measurement setup [298111 responses taken immediately after each other; each response
ELECTRONICS LETTERS 6rh June 1991 Vol. 27 No. 12 1021
had 1601 frequency steps and was performed in 800ms. This exponent in the environments measured is 4. A more extensive
leads to a power delay profile of 1601ns duration and 1ns series of measurements using this setup is currently being
carried out.

Acknowledgments: Discussions with Dr.C. Tannous, Mr. A.

Angus and Mr. R. Davies of AGT R&D and Mr. D. Tholl,
Mr. A. McGirr and Dr. S. Kazeminijad of NovAtel Commu-
nications Ltd. are gratefully acknowledged.

H. ZAGHLOUL 20th March 1991

R&D
ACT Limited
19E 411 1st St.
SE Cnlgnry, Alberta T2G 4Y5,Canada
G. MORRISON
M. FAlTOUCHE
6
0 50 100 1% 200 250 300 The Alberta Telecommnnications Research Center
and Depnrtment of Electrical Engineering
time, n s m The Uniwrsity of Cnlgnry
a
Calgary, Alberta T2N IN4, Canada

O l En, References
1 PAHLAVAN, K., and HOWARD, s. J.: ‘Frequency domain measure-
ments of indoor radio channels’, Electron. Lett., 1989, 25, pp.
-40 16451647
2 F. I.: ‘On the use of windows for harmonic analysis with
-60 the discrete Fourier transform’, lEEE Proc., 1978, 66, pp. 51-83
(Attention is drawn to a suspested typographical error. Table 1 of
the paper suggests that the 60dB mainlobe width for the
minimum three-sample Blackman-Harris window is 1.81 whereas
E our simulations and calculations suggest that it is 2.28)
-100

-120$-...: ..:... ‘ . . : . . 1. . : - - -300

-I&
0 50 100 150 200 250
time.ns
b
Fig. 3 Impulse response with line of sight unobstructed
a Unwindowed
b Windowed WIDE-ANGLE FIBRE OPTIC IMAGING
ARRAYS FOR VISION AND DISPLAY
bins. Fig. 3n shows the unwindowed results and Fig. 36 shows SYSTEM APPLICATIONS
the windowed results. Note the precursor and the reduced
dynamic range in Fig. 3a due to the high sidelobes of the
rectangular window. The resolving power of this technique is Indexincl t e r m : Optical fibres, Optics, Fibre optics
shown by comparing Fig. 36 and Fig. 4 which is a power
delay profile collected in the same environment as Fig. 3 b Experimental arrays of plastic nonimaging optic beam
expanders and concentrators have been fabricated and
except that the two antennas in the setup leading to Fig. 3 b attached to multimode fibre bundles for both remote image
had an unobstructed line of sight whereas in the setup of Fig. pickup and modulator display applications. Each expander
4 the line of sight was obstructed. It is noticeable that the first or concentrator acts as a pixel with spatial and angular
arrival with no line of sight present (Fig. 4) is weaker than the resolution adjustable during its design. Wide-angle com-
second arrival whereas in Fig. 36 the first arrival is the strong- pound eye designs are possible.
est.
In conclusion, this Letter presents an improved method of
operation and postprocessing of a setup to measure the fre- Introduction: Coherent fibre optic bundles have long been
quency response of indoor RF propagation channels. Mea- used to convey pattern information from one end (input) to
surements of the impulse response of the indoor propagation the other end (output). The output end can be viewed directly
channel at unprecedented resolutions were demonstrated. The by the human eye or can be put in immediate contact with a
Letter also presents measurements of the propagation power focal plane imager such as a CCD image array. Fibrescopes
law. The results indicate that the propagation power law have been used for visual inspection of remote or inaccessible
places, such as borescopes for gun barrels and the inside of
tubing and endoscopes for medical inspection. Fibrescopes
can include an illuminating system for the area to be viewed,
making them very useful in forensic and military applications.
Most coherent fibre bundles have fixed numerical aperture
and thus do not make the spatial and angular characteristics
of the output light different from those of the input. Conven-
tional bundles of plastic fibres have no built in magnification,
so that they can magnify only by separating the fibre ends,’
leaving dead zones between them. Bundles of glass fibres can
be coherently expanded to achieve true magnification that is,
however, subject to aberrations. In addition, these glass
bundles are heavy, inflexible, and difficult to fabricate into
large arrays.
1-20-)
0 50
:.:
100
:: . . . : . ... .-
150
.
200
.:
250
: I :6.
300
We report the augmentation of fibre bundles with non-
imaging optic coupler^^^^ to enhance input or output. Specifi-
tlme,ns 129814 cally, we have fabricated prototypes of two different devices: a
Fig. 4 Windowed impulse response with line of sight obstructed wide-angle vision system analogous to the compound eye of
1022 ELECTRONICS LETTERS 6th June 1991 Vol. 27 No. 12