Experimental and statistical studies of x-band transhorizon radio links over the sea.
This thesis is concerned with the statistical studies of microwave signals received
from beyond the radio horizon. A measurement campaign has been conducted in
collaboration with Rutherford Appleton Laboratories and CNET (§) in France in
order to collect propagation data across the English channel. The transmitter sites
were located at Lannion and Cap d'Antifer in the North French coast and the
receiver sites at HighcIiffe and Portsmouth in the South of England.
Several original findings have been obtained by the research. For the two types of
path considered of 155 and 250 km respectively, the cumulative distributions of
transmission loss level L display two separate regions; one represents the
troposcatter region and the other one represents the anomalous region. The
separation between the two regions occurs between 10 and 20 dB below the free
space path loss irrespective of the type of path considered. The two regions appear
to be well modelled by two distinct normal distributions with their particular mean
and standard deviation. The cumulative distributions of the transmission loss derived
from measured data were also compared with prediction models available from
Ii tera ture.
The duration D of exceedance or enhancement above signal level threshold was
investigated in terms of the joint statistics of transmission loss and duration of
exceedances Prob (D' ~ DIL' s L). Two different distributions were analysed and
modelled; the number distribution Pn and the time distribution PI. P,dDIL)
represents the fractional number of cases where durations have been greater than
D once the path loss is less than L, whereas PdDIL) 5ives the relative time with path
loss less than L due to all durations greater than or equal to D.
It is found that both Pn and PI are well modelled by log-normal distributions for all
types of propagation conditions and these distributions are very skewed. For PI the
(§) Centre Nationale D'Etudes des Tclecomunications, Paris and Lannion Laboratories
mode computed can range from fractions to several minutes whereas the median can
range from many tens of minutes towards the troposcatter levels down to a few
minutes close to the free space loss threshold of L = 0 dBf. The logarithm of the
median duration of Pt was found to be linearly related to L (dBt). As regards the
standard deviation, an (lnD) and at(lnD) are very close in agreement with the theory
and range from 1 to 4 when D is in minutes. They were also found to be linearly
related to L (dBD.
The research has shown that there is a strong seasonal dependence of the path loss
and the monthly attenuation may vary over a wide range throughout the year, with
variation being greatest on the longer path. The research has also shown that during
a 24 hours period there is, in general, a rise in the probability of exceeding a given
level at around 1000 GMT and this probability reaches its peak at around 1900 GMT
to then begin to decrease at about 2000 GMT. This pattern is much more pronounced
in the summer months than in the winter months. The daily variation is also more
pronounced for the shorter paths than for the longer paths. The time between 11 no
to 1900 GMT appears to be the strongest period of potential interference.
The research has included a preliminary analysis of the correlation between signal
level strength and the local surface meteorological parameters: temperature, water
vapour pressure and atmospheric pressure. As expected good correlation with
temperature was observed particularly with weekly averaged maximum daily values.
However; atmospheric pressure is the most dominant parameter on signal
enhancements. Finally, short term fading of clear air troposcatter signals was
investigated and the power spectral density analyses show significant components up
to about 5 Hz.