can be attributed to the mutual orientation of

communication is in practice constrained to

the antennas, which are not normally co-

`touch' range and hence limits its usefulness.

polarized and in the boresight direction of each

PAN systems, in which user touch or

other. Occasionally this leads to very large (up

handshake,  is used  for  low data  rate

to 25 dB) discrepancies between the data and

communication and has been demonstrated, [4],

the free space curve. The complexity of

at 10MHz. At higher frequencies, propagation

determining the relative orientation of the

has been characterised at ISM and other bands,

antennas along the on-body propagation path

for example, 400, 900 and 2450MHz, [5], to

renders it impractical to attempt a more

produce  simple  analytical  models  for

accurate theoretical prediction model, as the

stationary bodies. The measurements were

radiation pattern contribution and polarization

taken with dipoles oriented vertically to the

mismatch could not be measured sufficiently

body surface, and were taken over an angle of

accurately.

180º around the body. The propagation mode

is assumed to be a creeping wave which

The data points represented by circles in Fig. 4

exhibits an exponential decay of power, [6].

correspond to the scenarios when there was no

The expressions are of two slope form with a

LOS between the antennas, and the dominant

breakpoint part way around the body. Beyond

propagation path corresponded to a creeping

the breakpoint angle, the two rays are assumed

wave propagating around the body. Generally

to interfere, resulting in smaller decay but

speaking, the distance variation of such a wave

increased variability. The decay coefficient at

is fairly complex and depends on a number of

2.4 GHz corresponds approximately to 2

parameters. For example, the distance variation

dB/cm, assuming that the body radius is 20 cm.

is different on the free space sections of the

Significant interference takes place beyond the

propagation  path  and  on  the  sections

breakpoint at 400 and 900 MHz, indicating the

conformal to the body. However, because the

presence  of  multipath  effects,  although

human body has a complex geometry, which is

multipath is seen to be much smaller at 2.4

highly variable between individuals, tracing

GHz.

propagation paths around it is difficult and

often inaccurate. Nevertheless, the non-LOS

Measurements of channel loss, for many body

data points do appear to follow a trend shown

postures and antenna positions, have recently

as a straight line in Fig. 4. This trend

been made at 2.45GHz, [7, 8]. Fig 1 shows the

corresponds to an exponential attenuation

results collated to allow separation into classes

according to the following linear regression

of on-body propagation path gain - distance

formula:

variation. Two types of path gain variation

with the distance can be roughly distinguished

G  P [dB]  0.36d[cm]  35

(2)

based on the propagation scenario, namely,

variation in the illuminated region and in the

where d is the antenna separation distance

shadow region. The data points for the

propagation scenarios in which a line of sight

represented in centimeters and GP is the path

(LOS) was present, (receiver in the illuminated

gain in dB. The actual data points are spread

region), are shown in Fig. 4 by `+' signs. The

around this line with a standard deviation of

path gain of the antennas in free space in the

5.6 dB.

ground plane direction relative to each other,

given by the Friis formula, is also shown for

A number of propagation scenarios could not

comparison as a solid line. Fig. 4 shows that

be clearly identified as either LOS or non-LOS

most of the LOS data points follow the free

because the receiving antenna was very close

space curve but are generally below it with the

to the shadow boundary. Therefore, with a

mean difference between the measured and

slight change of body posture it could become

theoretical values of 5.1 dB. Thus, the path

either obstructed (completely or partially) or

gain in the illuminated region at 2.45 GHz is

unobstructed by a body part such as the trunk

given by the formula:

or an arm. These transition data points are

shown in Fig. 4 as x marks. Some of them

seem to follow the same trend as the LOS data

G  P [dB]  5.33  20 log  10 d [cm]

(1)

while others are clustered together with the

non-LOS data points.

The standard deviation of the difference

between the measured LOS values and those

given by this formula is 4.2 dB. This deviation