ETSI EP BRAN 3ERI074a

29 January 1998
page 1

Source: Jonas Medbo, Ericsson Radio Systems

Title: Radio Wave Propagation Characteristics at 5 GHz with Mode-

lingSuggestions for HIPERLAN/2
Agenda Item:

Document for: Decision

Discussion X
Information X
 WG3 Temporary document XX

1 Introduction
The purpose with this report is to assess the current knowledge about propaga-
tion characteristics at 5 GHz by a literature survey. It is pointed out that the
knowledge of some critical parameters is poor for some type of environments.
For these environments extensive measurements are strongly recommended.
The use of ray tracing techniques, as an alternative, is not advisable since they
usually involve large uncertainties due to poor knowledge about electromag-
netic material parameters in the corresponding environments. Furthermore, the
difficulty to accurately reproduce the effect of diffuse scattering using these
techniques is well known.

Simple type of models, which agree statistically well with measurements, are
suggested for both channel simulations and path loss prediction.

2 Channel characteristics

2.1 Modeling
For simulation purpose a simple type of channel model described by a few mea-
surable parameters is adequate. Still, a sufficient level of complexity must be
kept allowing a realistic modeling of relevant channel characteristics. A model
based on a tapped delay line is suggested [1] (essentially the same as the Com-
Nets HIPERLAN/2 model [2]). The impulse response h is modeled as
N
h ( τ, t ) = ∑ ak ( t )δ [ τ – ( k – 1 )∆τ ] , (2.1)
k=1
where t is time, τ delay, ak are complex amplitudes, ∆τ is the tap spacing with
respect to time. For a given bandwidth, W, the channel is unambiguously deter-
mined if ∆t < 1 ⁄ W i.e. if Nyquist’s sampling theorem is fulfilled. For e.g.
W=25 MHz ∆τ should be less than 40 ns. N is given by the maximum excess
delay and ∆τ. For e.g. a maximum excess delay of 350 ns N ≥ 10 . The fading of
each tap is assumed to follow a Rayleigh probability distribution where differ-
ent taps are uncorrelated. The average power per tap a k is assumed to decline
exponentially with time i.e.

( k – 1 )∆τ
a k = A exp – ----------------------- , (2.2)
2Γ
where Γ is the expected rms delay spread and A is a normalization constant.
Some measurements indicate that also secondary clusters of taps, with exponen-
tial decline, might have to be taken into consideration. The variation of impulse
 WG3 Temporary document XX

response with respect to time (due to a k ( t ) ) is given by the corresponding Dop-

pler spectrum. A uniform distribution of angles of arrival is assumed for the
received signals. The corresponding Doppler spectrum for a moving receiver
with an omnidirectional antenna is given by
–1 ⁄ 2
D ( f ) = ------------------ 1 –  -----------
1 f 2
(2.3)
2π f max  f max

where f is the Doppler frequency, and f max = v ⁄ λ where v is the receiver

velocity and λ is the carrier wavelength. For directive antennas the shape of the
Doppler spectrum can be calculated using the corresponding antenna diagrams.
In order to allow modeling of Ricean channels a contribution that is constant in
amplitude and Doppler frequency may be added to the LOS tap.

The Rayleigh probability distribution of tap amplitudes can be realized by e.g. a

complex Gaussian random process where the time variation is given by Doppler
filtering. This means that a sequence of impulse responses, that are equidistant
in time, is generated. The distribution is transformed to the Doppler frequency
domain, by a DFT of the time sequence of each tap, and then filtered in accor-
dance with the desired Doppler spectrum. Finally the distribution is transformed
back to time domain by an inverse DFT.

In Fig. 1 is shown the comparison of a modeled and a measured power delay

profile from a typical indoor environment. The measurement was performed in
the frequency domain (201 frequencies between 1700 and 1900 MHz) and then
transformed to time domain by an inverse DFT using a Hanning window. The
modeled channel is pure Rayleigh with Γ set to 45 ns (Γ=expected rms delay
spread) and with a Gaussian noise contribution. The agreement between the
model and the measurement is very good for this particular case.
 WG3 Temporary document XX

0 a)
Power [dB] −20

−40

−60

−80

−100
0 200 400 600 800
Delay [ns]

0 b)
Power [dB]

−10

−20

−30

1700 1750 1800 1850 1900

Frequency [MHz]

Figure 1. Power delay profile, a), and frequency spectrum, b), measured (solid line) with omnidi-
rectional antennas in a typical semi-large open space indoor environment, and modeled (dashed
line).

2.2 Measurements
An increasing number propagation measurements at 5 GHz have been reported
in the recent years. The focus has been on wideband characteristics for office
type of environments. Reported measurements are rare for outdoor and large
open-space indoor environments. In order to establish the propagation charac-
teristics for these type environments further measurements are needed.

For office type of environments different measurements seem to agree well (See
Table 1). The upper limit of rms delay spread is about 50-60 ns for distances up
to 30 m while the typical value is 10-20 ns. Moreover, the assumption that the
power has an exponential decline agree very well with the measurements.
 WG3 Temporary document XX

Table 1 rms delay spread (ns) measured in office type of environments.

LOS NLOS

Ref. d < 10m 10m < d < 30m d < 10m 10m < d < 30m

ave max ave max ave max ave max

Pla93 [3] 401

Nob93 [4] 15 22 59

Dev90a [5] 502 1302

Jan92 [6] 24

Air96 [7] 16

Str95 [8] 11

Gue97 [9] 36 49

Haf97 [10] 25 40 12 30 40
103 113 83 113

Kiv97 [11] 25 50-60

1. Delay at 20 dB below main peak was measured. The rms value was
obtained assuming exponential decay.
2. Measured in a 100x50 m office environment i.e. d < 100 m.
3. Receiver 20 dB horn with 3 dB beamwidth of 15°.

For large open space indoor environments, like airport terminals, larger delay
spreads are expected. One measurement in a 130x100 m car-manufacturing
assembly-hall reports a 20 dB delay interval of 460 ns which corresponds to a
rms delay spread of 100 ns [3]. It is however unclear if this is a typical value
since no other similar measurements have been found in literature.

3 Path loss prediction

The predictive power of empirical models is essentially as good as that of
present physical models. For the case where the transmitter and the receiver are
located at the same floor the Keenan-Motley approach has been proven to give
accurate predictions. Expressed in dB the path loss L is given by

L ( d ) = L FS ( d ) + n w L w (3.1)
where LFS is the free space loss, i.e. for isotropic antennas
 WG3 Temporary document XX

L FS ( d m ) = 32.4 + 20 log10 f GHz + 20 log10 dm , (3.2)

and nw the number of walls that are penetrated. Reported rms prediction errors
are typically about 5 dB [12]. This approach can be further simplified by assum-
ing that the loss due to walls is a linear function (dB) of distance [5], i.e.

L ( d ) = L FS ( d ) + αd (3.3)
where α is a constant typically in the range 0.2 to 1 dB/m. For two explicit
buildings α was measured to be 0.47 and 0.23 dB/m where the standard devia-
tion of the measured path loss was 8.5 dB. A dependence of the standard devia-
tion, σ, on distance have been shown in [5]. It is given by

σ(d) = β [ P dB(d) – P dB(0.3m) ] (3.4)

where β is a constant and PdB is the received power in dB as predicted by the
model. For the two buildings mentioned above, β was determined to be 0.15 and
0.19 respectively. This linear loss modeling is suggested for office type of envi-
ronments. However, the fact that only a few measurements have been found in
literature seems to necessitate additional measurements.

0
Received power rel. 0.3m [dB]

−50

−100

−150
−1 0 1 2
10 10 10 10
Distance [m]

Figure 2. Received power in dB versus distance in meters. The straight solid line corresponds to
free space loss. The curved solid line corresponds to the linear loss model with α=0.5. Dashed lines
indicate bounds corresponding to two standard deviations for β=0.15.

If the receiver and transmitter are separated by one or more floors the path loss
behavior is different. Support for a model of the type

L ( d ) = L FS ( 1m ) + 10n log10 d + FAF , (3.5)

where the floor attenuation factor FAF depends on the number of penetrated
 WG3 Temporary document XX

floors, has been found for frequencies around 900 MHz [12]. It seems as if n has
only weak dependence on the number penetrated floors and therefore can be
fixed. The increase of FAF per floor decreases as the number of penetrated
floors increases. Typical values are n = 3 and for penetration of floor FAF =
13-16 dB, and for penetration of three floors FAF = 30 dB. In order to establish
realistic parameter values for this type of modeling at 5 GHz, measurements are
needed.

4 Summary and conclusions

It has been shown that a realistic description of propagation at 5 GHz, for repre-
sentative indoor environments, can be achieved by simple models. The range of
parameter values of the suggested channel model is well established by litera-
ture for office type of environments. A type of path loss model (supported by
[5]) for office environments with transmitter and receiver at the same floor have
been suggested.

It remains to determine the representative range of channel model parameters

for outdoor and large open-space indoor environments. This is important since
significantly larger delay spreads are expected for these type of environments
(supported by [3]). Directive antennas are of particular interest since they might
allow a reduction of delay spread to the same level as for office environments
(supported by [10]). A literature search indicates that measurements are needed.

Concerning path loss, the suggested modeling is probably adequate. The uncer-
tainty about the range of values for model parameters is, however, very large
since only a few measurements at 5 GHz have been found in literature. Hence,
an extensive measurement campaign seems to be required.
 WG3 Temporary document XX

5 References
[1] K. Pahlavan, A. H. Levesque, ‘‘Wireless Information Networks’’, John
Wiley & Sons, 1995.
[2] ETSI EP BRAN Temporary document WG3TD73, ‘‘A simple/accurate
Propagation Model for the 5.2 GHz Indoor Radio Channel’’, S.
Mangold et al.
[3] Plattner-A; Prediger-N; Herzig-W, ‘‘Indoor and outdoor propagation
measurements at 5 and 60 GHz for radio LAN application’’, 1993 IEEE
MTT-S International Microwave Symposium Digest (Cat.
No.93CH3277-1), Atlanta, GA, USA, 14-18 June 1993, p.853-6 vol.2.
[4] P. Nobles et al., ‘‘Propagation Measurments in an Indoor Radio Envi-
ronment at 2,5 and 17 GHz’’, IEE Colloquium on ‘High Bit Rate UHF/
SHF Channel Sounders - Technology and Measurement’ (Digest
No.1993/233), London, UK, 3 Dec. 1993. Sponsors: IEE. In: p.4/1-6,
1993.
[5] D.M.J. Devasirvatham, C. Banerjee, M.J. Krain & D.A. Rappaport,
‘‘Multi-Frequency Radiowave Propagation Measurements in the Porta-
ble Radio Environment’’, in Proc. Second IEEE Int. Symp. Personal,
Indoor and Mobile Radio Commun., London, England, Sept. 1991, pp.
98-103., COST 231 TD(90)14
[6] G.J.M. Janssen and R. Prasad, ‘‘Propagation Measurements in an
indoor Radio Environment at 2.4GHz, 4.75GHz and 11.5GHz’’, in Pro-
ceedings of the 42nd IEEE Vehicular Tech. conf., Denver, Colo-
rado,May 11-13 1992.
[7] J. Airs, ‘‘Comparative Indoor RF Channel Soundings at 2, 5 & 17
GHz’’, Wireless Personal Communications 3: 353-363, 1996.
[8] Street-A-M; Moss-J-G-O; Edwards-D-J; Mehler-M-J, ‘‘Indoor propa-
gation measurements at 5 GHz’’, IEE Colloquium on 'Propagation in
Buildings' (Digest No.1995/134), London, UK, 14 June 1995. Spon-
sors: IEE. In: p.2/1-6, 1995.
[9] Guerin-S; Guo-Y-J; Barton-S-K, ‘‘Indoor propagation measurements at
5 GHz for HIPERLAN’’, Tenth International Conference on Antennas
and Propagation (Conf. Publ. No.436), vol.2, Edinburgh, UK, 14-17
April 1997. Sponsors: IEE. In: p.306-10 vol.2, 1997.
[10]P. Hafezi, D. Wedge, M.A. Beach, M. Lawton, ‘‘Propagation Measure-
ments at 5.2 GHz in Commercial and Domestic Environments’’, in
Proc. of Int. Symp. on Pers., Indoor and Mob. Radio Comm.,
PIMRC'97, Helsinki, Finland, 1997, pp. 509-513.
 WG3 Temporary document XX

[11]J. Kivinen, P. Vainikainen, ‘‘Wideband Indoor Propagation Measure-

ments at 5.3 GHz’’, COST 259 TD(97)54.
[12]Seidel, S. Y. and Rappoport, T. S., ‘‘914 MHz Path Loss Prediction
Models for Indoor Wireless Communications in Multifloor Buildings,’’
IEEE Transactions on Antennas and Propagation, Vol. 40, No. 2, Febru-
ary, 1992, pps. 207 - 217.