Cabin Propagation Channel Model
A summary of the analysis of the radio propagation measurements in two aircraft bodies at GSM, UMTS and ISM bands is provided below. Results are divided in coverage analysis, large-scale model characterisation, wideband model characterisation in a tap delay form, frequency selectiviness analysis, and attenuation and shadowing study of several cabin objects.
No channel characterization for an
aircraft cabin wireless network deployment was available up to now. Hence,
results achieved during the WirelessCabin activities
are considered of high scientificical value.
COVERAGE
RESULTS AND ANALYSIS:
First
coverage conclusions emerging from the performance comparison between the use
of a conventional antenna and leaky line cable (see D1v1 „Project Review Report Year 1“)
pointed out the fact that the leaky line cable distributes the power uniformly
over the passenger’s area, without creating well-favoured areas as it
happens with conventional antennas.
|
(a) Transmitter antenna: patch antenna |
(b) Transmitter antenna: leaky line |
These results were confirmed by wideband
measurements and an example shows it in the figure on the left. Both plots show
the power received by an antenna moving along
the right aisle of the A330-200 cabin from the last row until the galley in
business class (roughly 29 m). A patch antenna on the transmitter side placed
above the front part of the right aisle ceiling was used in the above figure
and a leaky line cable over the ceiling of the whole right aisle was used on
the below figure. Both measurements were done at 2.45 GHz with a
transmitted power of 18 dBm. While the minimum
received power in the patch antenna scenario is below -50 dBm,
in the leaky line one is kept above ‑35 dBm,
more than 15 dB difference. The power in
figure (a) would further increase if the receiver moved further on to the
first rows, closer to the transmitter position, resulting in an even bigger
different coverage situation in the cabin.
Therefore, and without considering capacity restriction, a coaxial cable with leaky sections is the best solution to provide coverage in the whole passenger’s area of an aircraft, whereas a point antenna is a desirable solution to provide coverage in dedicated cabin areas, such as in business class.
LARGE-SCALE
CHANNEL CHARACTERISATION:
The large-scale channel characterisation
for conventional antennas consists in estimating the path loss exponent (
) and standard deviation (
in dB) of a zero-mean Gaussian distributed random variable (
also in dB) based
on the best-fit line of the measured data according to the following equation:
.
This was done for the two aircraft and all the frequencies were measurements were done; results are summarised in the next table. More than 50 samples were always used. As example, a figure with measured data and a regression curve between normalised received power at 2140 MHz versus distance in the A330-200 is showed.
|
|
From all results, it can be concluded that a
tunnel-like behaviour characterises the propagation inside a small-haul aircraft, |
|||||||||||||||||||||
|
while in a long-haul aircraft the
characterisation is more similar to the one of an indoor office environment. |
||||||||||||||||||||||
According to the equation describing
propagation from a leaky line antenna expressed by:
.
Large-scale channel characterisation for
this antenna provides the estimation of attenuation exponent () and longitudinal loss (
) of the cable. depends on the working frequency and is normally given by
the manufacturer, but it was also estimated to assess the reliability of the
method in calculating . Results
on the calculation of in both aircraft fitted with the
theoretical value with a relative error <6% for all frequencies except at
900 MHz, which is a resonant frequency of the cable used. =2.3 was obtained for both aircraft
and all frequencies except 900 MHz.
CHANNEL
IMPULSE RESPONSE CHARACTERISATION:
A tap delay impulse response characterisastions was carried out for a total of four
scenarios: 1- short-range ac with conventional antenna,
2- short-range ac with leaky line cable, 3- long-range ac with
conventional antenna, and 4- long-range ac with leaky line cable. In all cases
the impulse response can be expressed as the sum of 
echoes (or multipath components) having random amplitude
, delay 
, and phase 
:
,
where 
is the Dirac delta function. Characterisation of the in-cabin
channel impulse response was developed by determining the statistical
properties of the path variables 
, , , and in the four
scenarios mentioned before.
Due to the big amount of obstacles in the
in-cabin channel, the scatterers can be assumed as
uncorrelated, and therefore, were assumed to
be uniformly distributed in 
.
Distribution
of number of multipath components
A summary of 1 3 distribution in
for the four scenarios is presented in the table below. The first phenomena
that can be observed is that, in conventional antenna scenarios for both
aircrafts, there is a twofold trend for as the receiver
moves away from the transmitter: (i) first increases because
of both the increment of scatters and the decrement of the highest peak of the
impulse response which implies a decrease of the threshold used when counting multipath components and hence more echoes are considered;
(ii) from a certain point on, decreases since
echoes earlier considered become so weak that are masked by noise and
therefore, not counted. The figure with the measurement data and the fitting
curves for the A319 is also shown below
|
conv. ant. |
Nakagami 5 areas, depending on d initial increase of N
2 |
Gaussian 6 areas, depending on d initial increase of N
4 |
||||
|
leaky
line |
Gaussian one single area
|
Gaussian one single area
|
||||
|
a/c type |
short-range
a/c |
long-range
a/c |
did not display any trend with respect to neither distance
nor position of the receiver for a given row in the leaky line measurements;
the same Gaussian characterisation was obtained for both aircraft. It is worth
pointing out that this distribution only makes sense at integer values and
those lying in the range from 9 to 24 (certainly avoiding negative values).
Inspection of the results for all scenarios suggests that the use of the leaky
feeder makes the multipath behaviour in terms of
number of echoes more uniform over the cabin.
Distribution
of path gains
The characterisation of path gains, 
, was done in two steps: (i) first, the mean amplitude 
for the k-bin was
studied and determined by an exponential decay (
), and then (ii) the variation over
this mean value 
was modelled in
terms of a probability function.
In most cases 
and 
were found to be
function of the Tx-Rx distance, or even lateral dependece of the exponential decrease was observed caused
by shadowing of the stowage bins, like the areas identified in the left figure
(scenario 3).
Lognormal, Nakagami,
lognormal and Weibull distributions characterise 
for scenarios 1
to 4 respectively. For further details, refer to deliverable D11.
Distribution
of arrival time sequence
Three approaches known in literature for
arrival time characterisation were followed (a pure Poisson arrivals process, a
modified Poisson process called 
developed by
Suzuki, and a cluster-type process presented in
’87 by Saleh and Valenzuela), but none of them described
satisfactorily the channel impulse response in many cabin positions. Therefore,
it was decided to characterise the multipath
components arrival time based based on an empirical
probability approach.
FREQUENCY
SELECTIVNESS ANALYSIS:
A frequency selectiveness study has
been also done in terms of mean excess delay (
), root mean square delay spread (
) and the coherence bandwidth (
) in the same scenarios mentioned in
the channel impulse response characterisation.
|
|
|
|
|
sc. 1 |
63.7 |
2.5 |
|
sc. 2 |
45.8 |
3.47 |
|
sc. 3 |
91 |
1.75 |
|
sc. 4 |
60.2 |
2.64 |
This table summarises the most restrictive according to the maximum calculated in
each scenario. By comparing these values with the bandwidths of the systems under
scope, it is concluded that the use of RAKE receivers receivers
or other equalization techniques would improve the performance of WLAN for use
in aircraft, while the performance of BluetoothTM
will not be highly affected thanks to the frequency hoping scheme.
ATTENUATION,
SHADOWING AND FAST FADING EFFECTS CAUSED BY CABIN OBSTACLES:
Wideband
meauserement were also investigated in scenarios
where either the receiver unit or something around it moved. With these
measurements, thanks to the combination of video synchronisation and received
signal, it was estimated the attenuation effect that cabin objects might cause.
These values should be taken into account in link budget calculations for
in-cabin communications. In particular, the following objects/effects were
studied: stowage bin attenuation,
lavatory’s door attenuation, back-seat table attenuation, lavatory
compartment shadowing, seat shadowing, separation walls shadowing, stowage bin shadowing, and fast fading effect caused by a
metallic trolley. These moving
scenarios can be spitted in three general situations: (i)
changing from almost LOS to NLOS (shadowing), (ii) adding an extra
obstacle when already in NLOS (attenuation), and (iii) multipath
fast fading.
|
|
|
|
(a) Receiver placed on a seat
with back-seat table folded |
|
|
|
|
|
(b) Receiver placed on a seat
with back-seat table unfolded |
|
As an example, this figure summarises the back-seat table attenuation measurement. The
figure consists of two video snapshots and two plots of total received power. A
vertical blue line in the received power plots marks the instant of time
associated to the video snap-shot. A mean difference of 2.5 dB can be observed
between both situations.
From
the analysis of all the results it can be concluded that when designing an
in-cabin radio netowork a fading margin of at
least 9 dB (10-12 dB to be on the save side) must be taken into
account in the link budget calculation.