J.T. Edward McDonnell1
The advent of the HIPERLAN standard in Europe and the emergence of the U-NI1 bands in the USA allow
the development of high bit-rate (>20Mbps) indoor wireless products at 5 GHz. The home is a fertile
application ground for such technology. This has prompted the need to investigate the domestic indoor
wireless propagation environment at 5 GHz. From measurements in houses, two empirical models were
proposed to describe.path loss variation with distance throughout a house. Both models are based on logdistance metric. One model is specific to individual rooms in a house whereas the other is a worst-case
generalisation for all‘roomson a floor.
The measurement equipment used was a HP 8665B (0.1-6000MHz) signal generator transmitting 17dBm at
5.2GHz (for HIPERLAN and U-NI1 compliance) from an quarter-wavelength monopole antenna and the
receiver again consisted of a quarter-wavelength antenna connected to an HP 8596E (9lcHz-12.8GH.z)
spectrum analyser. The measured path loss at a given point was obtained by averaging the received power
over an approximate 1 metre square at a height of 1 metre above the floor at that point.
Four homes were selected as representative examples of British domestic architecture ranging from Victorian
(pre-1900) to 1970s modem. In the following sections the results of the path loss measurements in these
properties will be discussed.
2.1 Modern SuburbanHouse
This house is on two levels with the bedrooms on the upper level. The external walls are made of standard
building bricks and on some internal walls in the lower level these bricks are left exposed as an interior
feature. The general layout of the ground floor is ‘open plan’ with a brick-built chimney in the middle of this
space. The transmitter was positioned 50cm above the floor on the ground floor. The first floor, which
contains the bedrooms, has plasterboard walls throughout. All the floors are made of chipboard that is
supported on wooden joists. Overall the sitting and dining areas, which are carpeted, were sparsely furnished.
Throughoutthe house the windows are single glazed with clear window glass.
From the measurements (figure 1) made on the ground floor it is apparent that even when there is an
unobstructed line-of-sight from transmitter to receiver the measured path loss is still worse than predicted by
the Friss equation. For example, even at a distance of 1 metre from the transmitter the path loss is 8dB worse
than predicted. These results point to the conclusion that the propagation environment is a complex one with
significant multipath effects. However, in areas which are shadowed by brick walls where the path loss would
be expected to be high, it is obvious that multipath is contributingenergy to mitigate the effect.
On the upper level of the house there is a marked uniformity of path loss values particularly in bedrooms 1,3,4
and the bathroom. The main propagation path to these measurement points is through the wooden floor, which
is known not to introduce any significant attenuation in signal strength [McD97]. The higher path losses in
bedroom 2 and part of bedroom 3 are attributableto signal propagation through brick on the ground floor.
’ Hewlett-Packard Laboratories,Filton Road, Bristol, BS12 6QZ. Email: [email protected]

0 1998The Institutionof ElectricalEngineers.
Printedand publishedby the IEE, Savoy Place, London WCPR OBL, UK

2.2 1930s Suburban terraced house
This house has two levels with an extension built from the lower level at the back of the house. The internal
and external walls are constructed from common red building bricks. Additionally, the front of the house is
faced with stone and all windows are double-glazed with clear glass. The extension containing the bathroom
on the ground floorjoins on to the original external wall of the house hence what has subsequentlybecome an
unusually thick internal wall. Floors are of pine boards supported on wooden joists. The transmitter was
placed 70 cm above the floor level at the telephone access point in the kitcheddining room.
This house and the house of section 2.1 are of approximately the same dimensions and the path loss
distribution (figure 2) is very similar in both cases despite the more enclosed structure of the ground floor of
the 1930s house. An interesting effect was observed in this house, namely, that there is a significantly higher
path loss in rooms that lie on the other side of the house from the transmitter. This effect which produces an
extra lOdB of loss on average can be explained by the presence of intervening brick walls in the signal
transmission path across the house. Overall this produces a parallel-vertical path-loss pattern in a house.
2.3 Victorian terraced house
This house is on three levels with the kitchen and dining room on the ground floor, the living room, study and
bathroom on the first floor and the bedrooms on the top floor. The external walls of the house are made of
stone and are 46 cm (18”)thick. All windows are single glazed with clear glass.
The transmitter was situated on the ground floor at the telephone access point and was positioned 1.2m above
floor level. The internal walls are for the most part made of brick with the exception of one plasterboard wall
on the first floor resulting from later alterations. The floors of the house are made of pine boards supported on
wooden joists.
The effects that were observed in the houses of sections 2.1 and 2.2 are similar to those found on the ground
floor of this Victorian house (figure 3). The higher path losses relative to the houses above observed in the
‘living room’ on the first floor can be attributed to the presence oftwo brick walls in the transmission path. It
is noticed that in the room directly above the transmitter, path losses are on average 5dB better than observed
in the other houses. Interestingly, the path losses in the ‘bathroom’ on the first floor are lower than expected
despite the intervening 18” thick wall. This effect demonstrates the mitigation of potentially severe path losses
by multipath signal propagation (in this case probably through the ceilingkloor and then through the bathroom
The path loss variation between the front and back of the house – the parallel-vertical pattern – is again
observed on all three floors of this house however it is more pronounced in this case.
2.4 Victorian apartment conversion
This apartment occupies one level of a converted four-storey Victorian town house. The external walls of the
house are substantial (3’7,’ thick) and made of stone. Many of the internal walls are of plasterboard as a result
of the conversion of the building into separate apartments. The two original internal walls which flank the
hallway are 30cm (12”) thick. The windows are all single glazed with clear glass. Again the transmitter was
situated at the telephone access point at a height of 30 cm above floor level.
The path losses (figure 4) in the room where the transmitter is situated are similar to those measured in the
other three houses. It is in the path losses in the kitchen area that marked differences arise with the other
houses. The path losses on the other side of the 3’7” thick wall are high. However, considering the thickness
of the wall, it might have been anticipated that the losses would have been more severe. It is in these situations
that the mitigating effects of multipath propagation are apparent.
Large door openings into the ‘hallway’ and plasterboard partitions in ‘bedroom 1 ’ contribute to give path
losses similar to those measured on the ground floor in the other houses. High path losses in the ‘bathroom’
and beside the wall adjoining the ‘shower room’ probably arise from shadowing caused by the thick wall
running alongside the ‘hallway’. The high path losses in ‘bedroom 2’ are most likely due to the two
intervening walls in the hallway which obscure the line-of-sight path to this room. Furthermore, this gives rise
to a path loss difference of between 15 and 17dB across the plasterboard wall between ‘bedroom 1’ and
‘bedroom 2’.
Many different models have been proposed to describe the propagation characteristics in the indoor
environment. These range from log-distance models [Rap961 including multiple-breakpoint models [Ake88],
to sophisticated ray-tracing techniques [Pah95]. Ideally, what is sought in this instance is a straight-forward
empirical model which encapsulates gross path loss characteristics which fit well to the measured data rather
than a compute-intensive but potentially more accurate model such as the ray-tracing one. The advantage of
this empirical approach is that all factors in the propagation environment are implicitly inherent in the derived
description. However, the empirical description may be appropriate only to those cases examined and its
generality needs to constantly checked against measurements in other environments.
The following log-distance model has been fitted to the path loss measurements:
PL(d)= PL(do)+lOnlog(-)
where PL(d) is the average path loss at distance d from the transmitter and n is the decay exponent. At a
reference distance of 1 metre from the transmitter the calculated path loss is 47dB at 5.2 GHz (Friss’
equation). Plotted below in figures 1-4 are the measured path losses versus the logarithm of distance from the
transmitter for each of the houses investigated. Superimposed on each scatter plot is the above path loss model
for increasing values of n.
It is not a straightforward exercise to find a unifying model to describe all the path losses in such a disparate
collection of houses. However, certain generalisations are proposed for the properties examined:
In the room in which the transmitter is located, provided there is an unobstructed line-of-sight (LOS) path
to the receiver, then n =2 (the path loss exponent)
The further away a room is from the transmitter, then the higher the path loss exponent
Within a given room there is generally an increase in path loss with distance from the transmitter
There is more homogeneity in the path loss within a room than there is between rooms
In an adjoining room to the transmitter and on the same level of the house providing there is a LOS path
then n = 2 to 3
In an adjoining room to the transmitter and on the same level of the house where the internal walls are
made of brick then providing there is an obstructed path (OBS)the path loss exponent is n = 4 to 5
On the upper floor of houses where the internal walls are made of brick then the room directly above the
room where the transmitter is located has n = 4 to 6
On the upper floor of houses where the internal walls are made of brick then the room not directly above
the room where the transmitter is located has n = 6 to 7
On the upper floor of houses where the internal walls are made of plasterboard then n = 3.5 to 4.5.
— y =47+50*log(x)
—– y=47+4o*log(x)
—- y =47+20*l@x)
— – y=47+30*log(~)
E BeQOom3(ld)
= EMn”l(lev2)
v Kitchen(lev1)
o Sittingareaoev1)
0 Diningarea(lev 1)
A Bedroom2(1d)
Figure 1: 1970ssuburban house
2 5
Tx-rx separation (metre)
Figure 2: 1930s suburban terraced house
1″ /
. stuctv(k@
0 utilim(levl)
90 D Diningroom(lev1)
1 2 5 10
Figure 3: Victorian terracedhouse
2 5
Figure 4: Apartment in a Victorian house
If a less involved rule-of-thumb were required to cover the floor on which the transmitter is located then
instantiating the previous model gives an average worst-case path loss at distance d:
PL(d)= 57+3010g(d)
where the path loss exponent is 3 and an extra lOdB is added in to take into consideration the maximum
spread of power arising from the arrangement of rooms, building material attenuation and random shadow
Another path loss prediction model that was investigated was the free-space plus wall/floor attenuation model.
In this model, the line-of-sight free-space path loss is combined with the attenuation through any intervening
walls/floors in this LOS path. Using an attenuation of 14 dB/single-brick wall [McD97], this model
consistently over-estimates the path loss. Again, assuming that the attenuation through stone is similar to that
through concrete blocks (3.4dB/cm) [McD97], then the model over-estimates path losses in the ‘Victorian
apartment’. For these reasons this model was disregarded. This discrepancy between the measured and
predicted path losses arises because the effect of multiple propagation paths to a given measurement point is
not taken into consideration.
The path losses at 5.2 GHz were measured in four typical houses. From these measurements two path loss
models are proposed both based on a log-distance model. The advantage of this empirical approach is that all
factors in the propagation environment are implicitly inherent in the derived description. However, it is
cautioned that the empirical description may be appropriate only to those houses examined and its generality
needs to checked against measurements in other environments. The lowest path loss exponent is 2 (free space)
in the room where the transmitter is located. This exponent increases to between 6 and 7 in rooms on the
upper floor at the other end of a house with brick internal walls. More modem houses with plasterboard
internal walls have lower path loss exponents on the upper floor of between 3.5 and 4.5. Finally, if a rule-ofthumb is needed to get a quick estimate of the path loss on the same floor as the transmitter is located then the
following is suggested: PL(d)= 57+30log(d) where PL is the path loss at distance d from the transmitter.
[Ake88] Akerberg, D., “Properties of a TDMA Picocellular Office Communication System”, IEEE
Globecom, pp. 1343-1349, December 1988.
[McD97] McDonnell, J.T. Edward, “The Attenuation of Building Materials at 5 GHz”, Hewlett-Packard
internal report, June 1997.
[Pah95] Pahlavan, K., Levesque, A., “Wireless Information Networks”, J. Wiley & Sons, Inc., New York,
[Rap961Rappaport T., “Wireless Communications”, Prentice Hall, New Jersey, 1996.

Leave a Reply

Your email address will not be published. Required fields are marked *