NASA Technical Reports Server (NTRS) 19940018347: Transmission over EHF mobile satellite channels

6 Pages·0.41 MB·English

by  NASA Technical Reports Server (NTRS)

N94-22820
Transmission Over EHF Mobile Satellite Channels *
W. Zhuang! J.-Y. Chouinard, A. Yonga(_o_lu
Department of Electrical Engineering, University of Ottawa
161 Louis-Pasteur, Ontario, Canada K1N 6N5
Tel.: [phone], Fax: [phone]
Abstract cost-effective. However, increased use of L-band
frequencies leads to the important problem of ra-
Land mobile satellite communications at Ka-
dio spectrum congestion: because of the avail-
band (30/20 GHz) are attracting an increasing
ability of large bandwidths, Ka-band is attract-
interest among the researchers because of the fre-
ing more and more attention in the land mobile
quency band availability and the possibility of
satellite research field. Furthermore, at Ka-band,
small earth station designs. However, commu-
the user equipment (e.g. portable transceivers)
nications at the Ka-band pose significant chal-
can be significantly smaller and eventually less
lenges in the system designs due to severe chan-
expensive than the equipment used at L-band.
nel impairments. Because only very limited ex-
It is therefore expected that LMS systems op-
perimental data for mobile applications at Ka-
erating in the Ka-band will provide the service
band is available, this paper study the chan-
users with larger information capacities, a wider
nel characteristics based on experimental data
variety of services, and this at a low cost. Unfor-
at L-band (1.6/1.5 GHz) [1]-[3] and the use of
tunately, the Ka-band LMS channel is subjected
frequency scaling [4]. The land mobile satellite
to severe propagation impairments such as rain
communication channel at Ka-band is modelled
attenuation, scintillation, as well as large Doppler
as lognormal-Rayleigh fading channel. The first
frequency shifts, all of which must be taken into
and second-order statistics of the fading chan-
account while designing personal satellite com-
nel are studied. The performance of a coherent
munication systems.
BPSK system over the fading channel at L-band
The objective of this paper is to model and
and Ka-band is evaluated theoretically and vali-
analyze the EHF land mobile satellite channel,
dated by computer simulations. Conclusions on
and to evaluate the performance of a coherent
the communication channel characteristics and
BPSK system in the channel. Here, we consider
system performance at L-band and Ka-band are
the 30/20 GHz Ka-band channel with up links
presented.
at 30 GHz and down links at 20 GHz. Being
1. Introduction the critical link, only the down link propagation
channel is discussed. The remainder of this pa-
Land mobile satellite (LMS) services at L-band
per is organized as follows. The first-order statis-
frequencies have been investigated and developed
tical description of the LMS channel is discussed
throughout 1980s; system concepts and corre-
in Section 2. Section 3 presents the second-order
sponding technologies are being transferred to
statistics of the fading channel at Ka-band, which
the land mobile satellite communication industry
are compared with those at L-band. An upper
[5]. In order to be viable, future LMS telecom-
bound of the bit error rate of a coherent BPSK
munication systems, such as personal communi-
system over the fading channel is derived theo-
cation systems, will have to provide a large num-
retically in Section 4, and is validated by com-
ber of users with diversified services, while being
puter simulations for the cases of light, average
*This work is supported by a strategic grant and heavy shadowings. Finally, Section 5 con-
(STR-0100720) from the Natural Sciences and Engineer- cludes the present work.
ing Research Council (NSERC) ofCanada
?Currently with TR Labs, Edmonton, Canada.
499
2. First-Order Statistics of the LMS to be used for the channel at Ka-band with the
Channel model parameters being appropriately scaled. In
this paper, we concentrate our analysis on the
It is well known [1], [61that a land mobile ra- frequency non-selective fading channel.
dio channel can be modelled as a Rayleigh fad-
ing channel with its local mean, the line-of-sight Table 1: L-band channel model parameters.
(LOS) component, following a iognormal statisti-
L-band (1.5 GHz)
cal distribution. The channel corrupts the trans-
channel bo p
mitted signaling waveform by introducing a mul- light shadowing 0.158 0.115 0.115
tiplicative gain r(t) and phase shift 0(t), which average shadowing 0.126 -0.115 0.161
can be expressed as heavy shadowing 0.0631 -3.91 0.806
n(t) = r(t). exp[j0(t) l= (1)
The effect of multipath fading and shadow-
z(t). exp[jCL(t)] -Fg(t). exp[jCM(t)] (2) ing on the LMS system performance depends
on the system operation environments. In open
where the LOS and multipath phases (eL(t) and
and tree-shadowed areas, multipath fading re-
CM(t)) are uniformly distributed between 0 and
sults from both specular reflection and diffuse
27r. z(t) is a lognormally distributed random pro-
scattering from the terrain surrounding the mo-
cess representing the amplitude of the LOS sig-
bile earth station. With the increase of the RF
nal, g(t) is the envelope of the multipath com-
frequency, the amplitude of the reflection coeffi-
ponent and is Rayleigh distributed. The proba-
cient and the ground conductivity also increase;
bility density function (PDF) of the lognormally
therefore, more scattering at Ka-band may come
distributed LOS component is [7]
from small obstacles or terrain irregularities than
those at L-band [8]. On the other hand, the mo-
= 1 exp[_ (ln 2_#)2 ] (3) bile antenna radiation patterns at Ka-band are
expected to be narrow enough to eliminate the
problem to a great extent. As a result, in first
where p and v/-do are the mean value and stan-
dard deviation of the normally distributed ran- approximation, it may be considered that multi-
path fading does not impair system performance
dom process ln[z(t)]. The PDF of the received
at Ka-band more than that at L-band [4]. In
signal envelope r(t) is [1]
the following discussion, the power of multipath
r 1 component at Ka-band is taken to be the same
v as that at L-band.
f0 ° Z
With the RF frequency 30/20 GHz of Ka-band,
r2+ z2 . .rz..
exp[- (ln z2d-op)2 -e-go]l°( o)aZ (41 the shadowing on the LOS component may come
even from the leaves and the small branches of
where bo is the power of the multipath signal. the trees whose dimensions are comparable with
In Table 1, the parameters of the L-band (1.5 the wavelength (1.0-1.5 cm). Therefore, the LMS
GHz) mobile satellite channel are displayed [1]- systems at Ka-band are expected to suffer more
attenuation and diffusion due to shadowing in
[3]. The calculations are based on measurement
carried on in Ottawa, Canada. For the exper- comparison with those at L-band. To the authors
iment, the INMARSAT's MARECS A satellite knowledge, very limited experimental data at Ka-
was used. The elevation angle from Ottawa to band for mobile application s i_ avaiiab}e _at the
the satellite was about 20°. So far, the assess- present time in the open literature and existing
ment of the LMS model has been carried out only frequency scaling methods are not of straightfor-
at UHF and L'band. Due to lack of sufficient ex- -ward applicability. An attempt has been made t°
perimental LMS data at Ka-band, a comprehen- estimate Such a scaling factor using the Modified
sive prediction of the relevant fading parameters Exponential Decay (MED) model [9], developed
cannot be done. However, from the mechanism from static propagation measurements through
of the LMS modelling at L-band, it is reason- deciduous trees in Colorado, US. The attenua-
able to assume that the LMS model is suitable tion coefficient (dB/m) can be computed by the
50O
following equation [4]: signal loses approximately an additional 1.5 dB,
1.0 dB and 0.5 dB at the signal amplitude lev-
a _ 0.45f °'2s4 for 0 < dt < 14
els of -10 dB, -5 dB and 0 dB respectively. Fig-
a _ 1.33f°'2S4dt °'412 for 14 < dt <_400 (5) ure 3 shows the cumulative distribution of the
received L-band and Ka-band signal envelopes
where f is the frequency (in GHz) and dt is the
without multipath signal in the case of average
depth of trees (in m) intercepted by the LOS sig-
shadowing. The analytic values are obtained ac-
nal. The scaling factor is therefore:
cording to Eq. (3). From this figure, one observes
_20GHz _'_ 2"lal.SG'Hz (riB(cid:0)m). (6) that the signal at Ka-band loses an additional 0
to 3.5 dB in comparison to the L-band in the
It should be pointed out that such an estimate weak signal range (which is the most important
model is based on particular channel conditions, for fade margin calculation). Table 2 shows the
and that further modifications may be necessary amplitude level of the shadowed LOS components
when experimental data at Ka-band becomes at the cumulative probabilities of 70%, 80% and
available. 90%.
In the following channel modelling and LMS
performance analysis at Ka-band, we use the L- Table 2: Amplitude level (in dB) of shadowed LOS
band parameters with the frequency scaling tech- components at different cumulative probabilities.
nique discussed above to calculate the Ka-band
Shadow- Pr. Pr. Pr.
channel parameters. The amplitude attenuation
ing RF 70% 80% 90%
of the lognormally distributed LOS component
light 1.5 GHz 0.49 0.23 -0.10
at L-band equals [4]
20 GHz 1.02 0.27 -0.63
A1.SGHz = exl"s (7) average 1.5 GHz -1.66 -2.07 -2.55
20 GHz -3.65 -4.53 -5.85
where X1.5 has a Gaussian distribution with
heavy 1.5 GHz -37.4 -39.4 -41.9
mean # = #1.5 and variance do = a_.5. At Ka-
20 GHz -77.5 -81.8 -88.3
band with RF equal to 20 GHz, the attenuation
of the received LOS signal power in dB W is 2.1
times of that at L-band (see Eq. 6), that is, 3. Second-Order Statistics of the LMS
Channel
A_ocHz(dBW) = 2.1 A2.SaH,(dBW)
2.1 Second-order statistics are used to describe
=_ A2OGHz = AI.5GHz
the time-dependent fading channel performance,
eX2o = C2.1 X1.5
which include level-crossing rate (LCR) and av-
::_ X2o = 2.1 Xi.s
erage fading duration (AFD). In the case of
#20 = 2.1 #LS Rayleigh multipath fading with lognormal shad-
=_a20 = 2.1al.S. (8) owing, the LCR normalized with respect to the
maximum Doppler frequency shift is [1]
Figure 1 is a software simulation model of the
LMS channel. In our simulations, the Gaussian LCR_(r = R) = LCR(r = n)/Fdm
noise after the low-pass (LP) filter is generated
by summing a large number of sine waves [6]. = _/2zr(1 - ,,_-2_,bob(ob(ol_+p22p)b+x4/_podb_J+_d-o°_)½o_,v_vfJ_'_ (9)
The bandwidth of the LP filters is B1 for the
where p is the correlation coefficient between the
shadowing component and B2 for the multipath
envelope r(_) and the envelope change rate ÷(t).
component.
The corresponding normalized AFD is
Figure 2 shows the cumulative distribution
of the received L-band and Ka-band signal en- AFD_(r = R) = AFD(r = R). Fdm
velopes over the fading channel with average
shadowing. The analytical results are obtained - v(,')d,-.(10)
from Eq. (4). In these figures, the signal am-
plitude level is relative to the LOS signal com- The analytic values and simulation results of
ponent without shadowing (0 dB). The Ka-band the second order characteristics (LCR and AFD)
,°, C_Z
for the average shadowing channel are shown in and if only the LOS component is considered,
Figures 4-5, where the amplitude level of the then
received signal is relative to the amplitude of
the LOS signal component without shadowing < 12 1 [¢_ exp[-_z2 (In z2-a/z0) 2]dz
(0 dB). In the computer simulations, assuming (12)
a vehicle speed of 43.2 km/h as an example, the
The upper bounds (analytical values) and sim-
maximum Doppler frequency shifts (which is the
ulation results of bit error rate for a coherent
bandwidth of the complex (_aussian noise filters
BPSK system at L-band and Ka-band are pre-
used to filter the process generating the multi-
sented in Figures 6-8. Figures 6-7 show the bit
path fading, i.e., B2 in Figure 1) are 60 Hz for
error rate due to amplitude fading of the LMS
the L-band signal and 800 Hz for the Ka-band
channel at L-band and Ka-band in the cases of
signal. The bandwidth of the filter for the shad-
light, average and heavy shadowing. With mul-
owing components (i.e., the B1 of Figure 1) is
tipath fading and at a bit error rate of 10-3, the
taken as 1/20 of B2, which is 3 Hz for L-band
system operating at Ka-band loses 0.75 dB, 3.5
and 40 Hz for Ka-band. The signal sampling
dB and 0.6 dB as compared to L-band in the
rate is 2400 Hz. Each simulation is performed
cas_ of :light, average and-heavy shado_ng re-
over 72,000 sampled data. The results show that
spectively. The-_OS signal component at Ka-
the maximum values of the normalized LCRs of
band suffers more additional attenuation (when
both L-band and Ka-band are very close to 12,
which means that the LCRs increase almost lin- compared to L-band) in average shadowing than
in light shadowing, which results in more system
early with the increase of the speed of vehicle or
loss at Ka-band in average shadowing (3.5 dB)
the increase of the RF frequency. The analytic
than in light shadowing (0.75 dB).
values are obtained with p = 0, i.e., no correla-
In the case of heavy shadowing, the LOS com-
tion between r(t) and ÷(t). In reality, there ex-
ists some correlation between r(t) and ÷(t) and Ponent is attenuated significantly, so that the
channel _can-be represented as a Rayleigh fad-
p may not be a constant, which may be further
ing channel (with only multipath components in
confirmed by the experimental data (see Figures
the received signal) at both L-band and Ka-band.
2-7 of [1]). That explains the slight differences
The weight of the LOS component is dramati-
between the simulation results and the analytical
cally reduced compared to the cases of both light
results in Figures 4-5. With the channel param-
and average shadowings. The system perfor-
eters and the use of frequency scaling technique,
mance depends on the dominant multipath sig-
one finds that, in the case of average shadowing,
nal component. With the assumption that the
the signal level at Ka-band is slightly more at-
first-order statistics are the same for the multi-
tenuated than at L-band.
path fading at L and Ka bands, the additional
4. Performance Evaluation of a Coherent system loss at Ka-band in the heavy shadowing
BPSK System condition (0.6 dB) is less than that in the average
shadowing (3.5 dB), which is not the case when
It is assumed that a carrier recovery loop can only the LOS component is considered. Figure
track the carrier phase jitter due to the fading 8 shows the bit error rate of the system due to
channel. Therefore, only amplitude fading due average shadowing only (no multipath signal is
to the channel is taken into account. The upper taken into account) at L-band and Ka-band. It
bound of the average bit error rate is [10] is shown that at a bit error rate of 10-3, the Ka-
band system suffers an additional loss of 3.33 dB
1 1 a_v compared to L-band system.
P_ <
5. Concluding remarks
7 ]" A comparison study for the land mobile satel-
f0 °° 1 exp[_(ln2_P)2
lite communication system operating at Ka-band
exp[ 2(b0z+2 dz (11) and L-band hasbeen performed. The modelling
of LMS channel at Ka-band is based on chan-
nel experimental data at L-band and a frequency
502
scaling technique, because of lack of sufficient ex- [5] K. Dessouky, P. Estabrook, and T.C. Jedrey, "The
perimental data of the channel at Ka-band. Both ACTS Mobile Terminal", JPL SATCOM Quarterly,
first-order and second-order statistics of the LMS no.2, pp.l-9, July 1991.
channel are analyzed. The results show that: the [6] W.C. Jakes, Microwave Mobile Communication.
New York: Wiley 1974.
LOS signal component at Ka-band has 0.0 dB to
3.5 dB more attenuation in the presence of av- [?'] A. Papoulis, Probability, Random Variables, and
Stochastic Processes. Second Edition, McGraw-Hill
erage shadowing; the channel fades have a much
International Book Company, 1984.
faster rate at Ka-band compared with that at L-
[8] P. Beckmann and A. Spizzichino, The Scattering
band because the fading rate is approximately of Electromagnetic Waves b-_om Rough Surface.
proportional to the RF frequency, and the nor- Artech House, 1987.
malized average fading duration (AFD) increases [9] M.A. Weissberger, An Initial Summary of Models
correspondingly at Ka-band. An upper-bound for Predicting the Attenuation of Radio Waves by
Trees. U.S. Dept. of Defense, Report no. ESD-TR-
bit error rate of a coherent BPSK system in the
81-101.
LMS channel is derived, which is further vali-
[10] W. Zhuang, A. Yongac_oglu, J.-Y. Chouinard, D.
dated by computer simulations. With multipath
Makrakis, "Performance Analysis of EHF Land Mo-
fading and shadowing, at a bit error rate of 10-3, bile Satellite Systems," Technical Report No.l, De-
the system performance at Ka-band loses approx- part ment of Electrical Engineering, University of Ot-
imately 0.75 dB, 3.5 dB and 0.6 dB more than tawa, September 1992.
that at L-band; in the absence of multipath fad-
ing, the system performance at Ka-band loses ap-
proximately 3.33 dB more than that at L-band in 10,
the case of average shadowing.
5
References
[1] C. Loo, "A Statistical Model for a Land Mobile Satel-
lite Link", IEEE Trans. on Vehicular Technology,
vol. VT-34, pp.122-127, August 1985.
[2] J. S. Butterworth, "Propagation Measurements for
Land Mobile Satellite System at 1545 MHz," Com- _E -10
munication Research Center, Department of Com-
munications, CRC Technical Note 723, August 1984.
-is
[3] C. Loo, "Digital Transmission Through a Land Mo-
bile Satellite Channel", IEEE Trans. on Communi-
cations, vol. COM-38, no.5, pp.693-697, May 1990.
"20.0 0.2 0.4 0.6 0.8 .0
[4] J. Fortuny, J. Benedicto, and M. Sforza, "Mobile
Prob (received slgnM emp|ltude :. ordinate)
Satellite Systems At Ku and Ka-Band", Proc. Sec-
ond European Conference on Satellite Communica- Figure _: Received signal envelope probability distri-
tions, pp.55-62, Liege, October 1991. bution for average shadowing (with multipath).
Figure I: Land mobile satellite channel model.
503
i0o
A_N (anai_c_J)
+ AWGN (simulation)
i°
A 1o" •.......... u_ (mml)'Ic_)
=,_ o Ugt_ (*irnuta_onl
_, .'_\ ..... Awag.(=-,_c=)
= "_ ;,_,\ • ^,.,,.-ogo¢.c.J,l,on)
I0" : It ',% • _ -- -- -- Heavy (enalyica#)
_tk oa+, *, "_
-5 _. ",% • _ • Heavy (_lation)
a_ \ o/',.,.?., , ,
1o.3 _, oe'-, ",, a_
-10 •i ° "-."- ;"
if,
I.,_ LIra-+banadn(andai)qlt_cal),.,=)
-15 Ira-band (simulazion) ,o, _ . o,,...:..\\
..............................
-20 . • . i . • . I . 10-5
o,0 0,2 .0_4 0,6 0,8 1,0 5 _0 I5 20 25 30 35 40 45 50
Prob (reCeived efg_it amplrtUde _ ordinate) Signal.to.Nolle ratio (dB)
Figure 3: Received signal envelope probability distri- Figure 6: BER of a coherent BPSK system in a LMS
bution for average shadowing (without multip+,th). channel at L-band.
lO1 IO0
AWGN (ar_ylcml)
• o o ÷ AWGN (M-nulation)
100 • 0 • 1o" I %. ........... Ught (ar,_ylcal)
tO" 1_ _,._ • Oght(timula_o_)
k "_'%%"_ ..... Average (aaa]ytical)
.,2 _ '..&_ _ • Average (rarraJlatioo)
+o2::., +. IO"2 '_. ".. %,.\ - - - mew(.r,.W,.,_)
"_ 10" 3 0 • 0"%• "_ _, • Heavy (sin_lado_)
10"41 "" • 0 • ..... , ,,,i,\
E = lO_ )']_ _.... ,p=\
I %+. .•..•2. -2\
"-o. •
10-7 .... ,.... I.... ..... =.... ..... ..... ..... 10" 5 • ",, _, _%
-30 -25 -20 -15 -I0 -5 0 5 10 10 15 20 25 30 35 40 45 50
91gnal amplitude level (dB) Signal-to-Nolle ratio (dB)
Figure 4: Normalized LCR in LMS channel at I,-band Figure 7: BER of a coherent BPSK system in a LMS
and Ka-band (average shadowing). channel at Ka-band.
106
100
105
........... L-band (anmlytca_ • I + AW_N (limulation)
104 0 L-band (_mula_on) "":"--+> - i........... L-t_fld (anafy_cal)
tO3 I-- Ks-band {anal)'_cal) +'_-*". "'. I...... m,-ba,',,:,¢ana,,.++.,)
g • Ka-band (simula_o_)
4( 102 i 10"; _o "4P
+'%11 "%
_e 101
N
11
E 10°
_.,oO,,:O F., 10"3 _ o '".., • , ,,.\.
t0-1
10-2
0 o 0
:::_:'::i ...................
10-3 '°'4 _ ° _ ".",
10-4 10.5 + ,
-30 -25 -20 °15 -10 -5 0 5 10 0 2 4 (_ 8 10 12 14 16 t8 20
Signal amplitude level (dB) Signal-to-Nolle ratio (dB)
Figure 5: Normalized AFD in LMS channel at L- Figure 8: Performance of a coherent BPSK system
band and Ka-band (average shadowing). over average shadowed LMS channel.
504