Professional Documents
Culture Documents
Abstract—A novel model based on the bivariate inverse In the SD reception technique two or more base stations are
Gaussian (IG) distribution is presented for the prediction of the communicating with the satellite and the Earth-space path with
outage performance of a dual site diversity Earth-space system. In the least attenuation is selected, resulting to a selection com-
the present paper, the exceeded rain attenuation induced on each
Earth-space path is assumed that follows the IG distribution. The bining scheme [4]. SD actually takes advantage of the decor-
joint rain attenuation exceedance probability is calculated using relation of rainfall rate with the distance between the two radio
the long-term statistical parameters of rain attenuation on each paths and so the fact that the probability of simultaneous rain at-
individual slant path and adopting a spatial correlation coefficient tenuation occurring on two or more network routes is less than
of rain attenuation variables as a function of the separation dis- the probability of rain attenuation occurring on each individual
tance of the two Earth-stations. The Earth-Space path Database
of Study Group 3 (DBSG3) of International Telecommunication path.
Union (ITU) is employed in order to compare the performance The models for the prediction of the SD performance can be
of the new method with the ITU-R P.618-10 outage model. The classified into two categories: the empirical prediction models
numerical results from the comparison tests are very encouraging. and the physical-mathematical models. Two empirical methods
Index Terms—Fade mitigation techniques, radio propagation, for predicting the diversity gain are described in [5], [6]. The
rain attenuation, site diversity. model of Hodge [6] has been adopted by ITU in the recommen-
dation ITU-R. P. 618-10 [7] for the calculation of site diver-
sity gain (Section 2.2.4.2 of ITU R. P.618-10). Physical models
I. INTRODUCTION that predict the joint exceedance probability of rain attenuation
have been developed in [7]–[13]. The models that have been
M ODERN satellite communication systems operate at Ka presented in references [7]–[11] assume that the rain attenua-
and Q/V frequency bands in order to overcome the prob- tion on slant paths follow bivariate (for dual diversity scheme)
lems of spectrum congestion and to satisfy the high data rate or multivariate lognormal distribution. The ITU-R P. 618 model
demanding services. Tropospheric phenomena affect the prop- (in Section 2.2.4.1) [7] for the prediction of the outage proba-
agated electromagnetic waves and result in the degradation of bility due to rain attenuation with site diversity has been adopted
the quality of the link, for operating frequency above 10 GHz by ITU-R very recently. The difference between the models of
[1]. Among the tropospheric phenomena rainfall is the most se- Section 2.2.4.1 and Section 2.2.4.2 of the ITU-R. P. 618-10 Rec-
vere cause of attenuation of the signal and is the one that shows ommendation is that the first one refers to the joint exceedance
the most significant spatial inhomogeneity. Therefore, in order probability of rain attenuation for dual site systems, while the
to use higher frequencies fade mitigation techniques (FMTs) are second one is used in order to compute the site diversity gain
employed to deal with the rain attenuation fading [2]. using an empirical algorithm. Furthermore, the diversity gain
FMTs can be divided into three main categories [2]: Effective computed by the model of Section 2.2.4.2 has the limitation that
isotropic radiated power (EIRP) control techniques, adaptive the baseline length of the two Earth stations must be less than
transmission (adaptive modulation, adaptive coding and data 20 km.
rate reduction) techniques and the diversity techniques. Site di- In this paper a novel model for predicting the joint ex-
versity reception is the most efficient FMT from a technical per- ceedance probability of rain attenuation for a dual site diversity
spective view but unfortunately it is cost ineffective since it re- system is presented using the single and the bivariate inverse
quires at least two Earth stations installations along with a ter- Gaussian (IG) distribution [14]. The general unbalanced di-
restrial connection [3]. versity configuration is considered since different statistical
inputs for the single slant path distributions are used. The
IG distribution has been previously reported for the descrip-
Manuscript received December 05, 2011; revised February 01, 2012;
accepted March 26, 2012. Date of publication July 11, 2012; date of current tion of shadowing phenomena [15] and turbulence-induced
version August 30, 2012. This work was carried out under the framework of phenomena in free-space optical systems operating in weak
COST IC0802 and was supported by the NTUA THALES-MIMOSA project
turbulence conditions [16]. Also, in [17], the IG distribution is
funded by EU and GSRT.
The authors are with the Department of Electrical and Computer Engineering, used for modeling the conditional distribution of rain rate for
National Technical University of Athens, Athens GR-15780, Greece (e-mail: meteorological applications, due to the fact that IG distribution
thpanag@ece.ntua.gr).
is a skewed distribution. For the above reasons we have been
Color versions of one or more of the figures in this paper are available online
at http://ieeexplore.ieee.org. motivated to use the bivariate IG for the prediction of joint
Digital Object Identifier 10.1109/TAP.2012.2207073 exceedance rain attenuation statistics.
The proposed model is described in Section II. In Section III and is the correlation coefficient between the random variables
the DBSG3 [18] will be used in order to test the new model and and .
the results will be compared against the ITU-R. P. 618 phys- The joint CDF of the two variables and is given [19],
ical-mathematical model (in Section 2.2.4.1) [7]. The model in [20]:
the paragraph Section 2.2.4.1 of ITU-R. P. 618-10 Recommen-
dation has been tested inside ITU-R Study Group 3 with many
other models from the literature. Due to its good and reliable (6)
performance on the experimental data, it has been selected by
ITU-R and inserted in the ITU-R P.618-10 Recommendation. where
Finally, conclusions derived from this paper will be referred in
final Section IV.
(9)
(1)
(2)
(3) (10)
Assuming that two variables and which are greater The term unconditional that is used, means that there is no
than zero each of them follows the IG distribution with pa- separation between rainy and non-rainy periods for modeling
rameters and , and PDFs and , the rain attenuation distribution. The use of the unconditional
respectively, the bivariate PDF is constructed as follows [19], CCDF for rain attenuation modeling overcomes the problem of
[20]: the knowledge of probability of the existence of rain attenuation
. However, the probability of rain, that means that
rainfall rate is greater than zero can be de-
(4) rived for every location of the world using the ITU-R rainmaps
P. 837-5 [21]. This does not take into account the year varia-
with tions of the probability to rain in a single point. The value of
can also be used for the rain attenuation case with precau-
tion. An expression that relates the value with the probability
that rain attenuation exceeds zero dB value along
the slant path is needed. Summing up, using the unconditional
complementary cumulative density function the value and
(5)
the relationship between and are not required.
KOUROGIORGAS et al.: ON THE EARTH-SPACE SITE DIVERSITY MODELING: A NOVEL PHYSICAL-MATHEMATICAL OUTAGE PREDICTION MODEL 4393
(11)
Fig. 2. CCDF of rain attenuation for La Conception, Canada for GHz (12)
and with curve from DBSG3 database (o) and the IG distribution (solid
line) with parameters dB and dB.
where is the bivariate IG distribution (see ex-
pressions (4) and (5)). From [22] we have an expression that
The unconditional IG CCDF of rain attenuation approximates gives the joint exceedance probability (CCDF) in term of the
very well the rain attenuation distribution for probability levels joint CDF and the single CDFs:
less than 1%. So the ‘tail’ of a given rain distribution can be
modeled very well with IG model. Two examples are given here
in Figs. 1 and 2. In both figures the parameters and of the
CCDF of IG rain attenuation distribution (10) were computed
(13)
through a non linear regression fitting process with data points
of rain attenuation. Data derived from ITU-R. P. 618-10 for the As far as the correlation coefficient is concerned, we employ
city of Bangkok considering a link with operating frequency the Paraboni-Barbaliscia expression [11] that also has been used
equal to 20 GHz and elevation angle in Thailand in ITU-R physical-mathematical model [7]:
were used for the first Figure. Experimental data points from La
Conception in Canada from DBSG3 database with GHz (14)
and were used for Fig. 2. The computed parameters of
the IG distribution are dB and where is the baseline distance of the two Earth terminals in
dB for Fig. 1 and dB and dB km. If there are experimental data for the horizontal variation
for Fig. 2. of the rainfall medium for the specific region of the site diver-
From Figs. 1 and 2, it can be observed the fitting of rain sity system, the proposed model is flexible and the experimental
attenuation exceedance probability using IG distribution is very correlation coefficient is recommended for the model.
4394 IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 60, NO. 9, SEPTEMBER 2012
Fig. 4. Experimental (+) and simulated (solid line) joint CCDF of rain
Fig. 3. Experimental data (+) and simulated (solid line) joint CCDF of rain
attenuation for two Earth stations with a distance 10.9 km, operating frequency
attenuation for two Earth stations with a distance 22.7 km, GHz and
11.4 GHz and elevation angle 33 located at Graz and Michelbachberg, in
located at La Conception and Vernet, in Canada.
Austria.
Concluding our methodology, we can calculate the outage The first one is the reduction of attenuation exceeded for
probability of a dual site diversity scheme using the following a given time percentage as this is shown in (15), [6], [12].
step-by-step procedure: The improvement factor is the division of the single site
• Calculate the IG rain attenuation parameters for every rain attenuation exceedance probability by the joint ex-
single slant path using either experimental data or by ceedance probability ( or equivalently the outage proba-
fitting on the ITU-R P.618 prediction curve. Then we will bility ) for a given attenuation level (16), [12]. These
have and . two measures which are essential for the design of dual site
• Calculate the long-term rain attenuation correlation coeffi- Earth-space systems can be calculated from the proposed model.
cient using expression (14) and
• Finally calculate the outage probability using the final (15)
close formula derived from expressions (13), (2), (6), (7)
and (9). where is the single-site rain attenuation exceeded for
Now we will present two characteristic comparison examples percentage of time and is the joint rain attenuation
of the proposed methodology with experimental data from the exceeded for percentage of time and
ITU-R Databank.
In Figs. 3 and 4 the joint rain attenuation exceedance proba- (16)
bility taken from DBSG3 database and calculated from the ex-
pressions (12)(14) are shown for two experiments. For Fig. 3
the first base station is located at La Conception in Canada and III. VALIDATION AND COMPARATIVE TEST OF THE MODEL
the second one at Vernet in Canada. Their distance is 22.7 km In this section, the proposed model will be validated with the
with GHz and . The parameters of the indi- Earth-space path experimental database of ITU-R, Study Group
vidual CCDF of IG distributions are dB, 3, DBSG3. The experimental joint rain attenuation exceedance
dB for the first slant path and dB, probability will be compared with the resulted one from the new
dB for the second slant path at Vernet. For model using bivariate IG distribution. The error will be com-
the second figure, the two base stations are located in Austria at puted for all the experiments for two cases: first, for every time
Graz and Michelbachberg with their distance equal to 10.9 km. percentage and second, for all time percentages. For the first
The operating frequency of the link is 11.4 GHz and the eleva- case the mean value, standard deviation and rms value of the
tion angle of both Earth stations is 33 . The parameters of the error were computed for every time percentage considering all
two IG distributions are dB, dB, the experiments, while for the second case the statistical param-
and dB, dB respectively. The error eters of the error were computed considering all the errors for all
between the predicted values of the joint exceedance probability data points, i.e., all time percentages and all experiments. The
and the measured values has been computed for both Figures. error is computed according to ITU-R. P. 311-13 error criterion
For Fig. 3, the mean value, standard deviation and rms value of [26]:
the error are 1.27%, 4.79% and 4.95%, respectively, while for
Fig. 4 they are 7.06%, 4.38% and 8.31%.
The advantage in terms of rain attenuation by deploying a
dual site diversity system instead of a single-site system can be
expressed by the terms diversity gain and improvement factor. (17)
KOUROGIORGAS et al.: ON THE EARTH-SPACE SITE DIVERSITY MODELING: A NOVEL PHYSICAL-MATHEMATICAL OUTAGE PREDICTION MODEL 4395
Fig. 6. Standard deviation of the error (%) for every time percentage for the
Fig. 5. Mean value of the error (%) for every time percentage for the proposed proposed model (o) and ITU-R. P. 618 (+).
model (o) and ITU-R. P. 618 (+).
recommendation for almost all time percentages. It is obvious [6] D. B. Hodge, “An improved model for diversity gain on earth-space
that the new model behaves better in terms of mean error for propagation paths,” Radio Sci., vol. 17, no. 6, pp. 1393–1399, 1982.
[7] Propagation Data and Prediction Methods Required for the Design
every time percentage and for the overall error, as this is shown of Earth-Space Communication Systems, ITU-R. P. 618-10, Geneva,
in Table I. The mean error of the proposed model is % 2009.
and for ITU-R. P.618 is %. [8] E. Matricciani, “Prediction of site diversity performance in satellite
communications systems affected by rain attenuation: Extension of the
As for the standard deviation, it is observed that ITU-R. two-layer model,” Eur. Trans. Telecommun., vol. 5, no. 3, pp. 27–36,
P.618 model behave better in almost every time percentage (see 1994.
Fig. 6). However, in overall the proposed model is statistically [9] J. D. Kanellopoulos, S. G. Koukoulas, N. J. Koliopoulos, C. N.
Capsalis, and S. G. Ventouras, “Rain attenuation problems affecting
slightly more stable than ITU-R. P. 618 as from Table I the the performance of microwave communication systems,” Ann.
standard deviation of the error of the proposed model is 29.19% Telecommun., vol. 45, no. 7–8, pp. 437–451, 1990.
[10] A. D. Panagopoulos, P.-D. M. Arapoglou, G. E. Chatzarakis, J. D.
and for ITU-R P.618 30.25%. Furthermore, the performance of Kanellopoulos, and P. G. Cottis, “A new formula for the prediction
a model is evaluated via the rms value of the error. From Fig. 7 of the site diversity improvement factor,” Int. J. Infrared Millimeter
it can be referred that the proposed model has a lower rms Waves, vol. 25, no. 12, pp. 1781–1789, 2004.
[11] A. Paraboni and F. Barbaliscia, “Multiple site attenuation prediction
error than ITU-R. P. 618 for every time percentage apart from models based on the rainfall structures (mes- or synoptic scales) for
0.001%–0.003% time percentages. The overall rms error for advanced TLC or broadcasting systems,” presented at the XXVII URSI
the proposed model is 29.2% and for the recommendation of General Assembly, Maastricht, 2002.
[12] A. V. Bosisio and C. Riva, “A novel method for the statistical predic-
ITU 34.37%, as this is shown in Table I. Therefore, the overall tion of rain attenuation in site diversity systems: theory and compara-
performance of the proposed model is better comparing to the tive testing against experimental data,” Int. J. Satellite Commun., vol.
ITU-R. P. 618. We can conclude that the prediction made by the 16, pp. 47–52, 1998.
[13] L. Luini and C. Capsoni, “MultiEXCELL: A new rainfall model for the
new model based on bivariate IG distribution is accurate and analysis of the millimetre wave propagation through the atmosphere,”
closer to the experimental data than the prediction of ITU-R. in Proc. 3rd Eur. Conf. on Antennas and Propagation (EuCAP), Berlin,
P. 618. Germany, Mar. 23–27, 2009, pp. 1946–1950.
[14] R. S. Chikkara and J. L. Folks, The Inverse Gaussian Distribution:
Theory, Methodology And Applications, M. Dekker, Ed. Boca Raton,
FL: CRC Press, 1989.
IV. CONCLUSION [15] N. D. Chatzidiamantis, H. G. Sandalidis, G. K. Karagiannidis, and S.
A. Kotsopoulos, “On the inverse Gaussian shadowing,” presented at
A new simple method was presented for computing the the Int. Conf. on Communications and Information Technol., 2011.
outage probability of rain attenuation for dual site diversity [16] N. D. Chatzidiamantis, H. G. Sandalidis, G. K. Karagiannidis, and M.
systems. This method is a novel mathematical approach based Matthaiou, “Inverse Gaussian modeling of turbulence-induced fading
in free-space optical systems,” J. Lightwave Technol., vol. 29, no. 10,
on the bivariate Inverse Gaussian distribution. The main as- pp. 1590–1596, 2011.
sumptions are rain attenuation on each individual path of the [17] B. Kedem, L. Chiu, and Z. Karni, “An analysis of the threshold method
diversity system follows the Inverse Gaussian distribution and for measuring area-average rainfall,” J. Appl. Meteor., vol. 29, 1990.
[18] ITU-R. Databank DBSG3 2010 [Online]. Available: http://www.itu.
a well-accepted correlation coefficient is employed. Using this int/publ/R-SOFT-SG3/en
method the balanced and the unbalanced joint exceedance prob- [19] E. K. Al-Hussaini and N. S. Abd-El-Hakim, “Bivariate inverse
ability of rain attenuation can be calculated. The performance Gaussian distribution,” Ann. Inst. Stat. Math., vol. 33, no. 1, pp.
57–66, 1981.
of the proposed model was validated using DBSG3 database [20] B. Seshardi, The Inverse Gaussian Distribution. Oxford: Clarendon
and compared to the ITU-R. P. 618 model. Computing the error Press, 1993.
of the two models, the behavior of the proposed method has [21] Characteristics of Precipitation for Propagation Modeling. Geneva,
2007, ITU-R. P. 837-5.
been shown to be better comparing to the one of ITU-R. P.618 [22] A. Papoulis and S. U. Pillai, Probability, Random Variables and Sto-
model. chastic Processes. New York: McGraw-Hill, 2002.
[23] T. F. Coleman and Y. Li, “On the convergence of interior-reflective
Newton methods for nonlinear minimization subject to bounds,” Math.
ACKNOWLEDGMENT Program., vol. 67, no. 2, pp. 189–224, 1994.
[24] T. F. Coleman and Y. Li, “An interior, trust region approach for
The authors are grateful to all the Propagation Experts that nonlinear minimization subject to bounds,” SIAM J. Opt., vol. 6, pp.
418–445, 1996.
made the experiments that were included in the ITU-R Data- [25] MATLAB [Online]. Available: http://www.mathworks.com/products/
bank. matlab/index.html
[26] ITU-R 311-13, Acquisition, Presentation and Analysis of Data in
Studies of Tropospheric Propagation. Geneva, 2009.
REFERENCES
[1] R. K. Crane, Propagation Handbook for Wireless Communication Charilaos I. Kourogiorgas (S’12) was born in
System Design. Boca Raton, FL: CRC Press, 2003. Athens, Greece, on July 6, 1985. He received the
[2] A. D. Panagopoulos, P.-D. M. Arapoglou, and P. G. Cottis, “Satellite Diploma Engineering in electrical and computer
communications at Ku, Ka and V bands: Propagation impairments and engineering in 2009 from the National Technical
mitigation techniques,” IEEE Commun. Surveys Tuts., 2004. University of Athens (NTUA), Athens, where he is
[3] COST 255 Final Report “Radiowave Propagation Modeling for Satcom currently working towards the Ph.D. degree.
Services at Ku Band and Above,” ESA Publication Division, 2002. From October 2009 to June 2011, he was with
[4] P. D. Arapoglou, K. P. Liolis, M. Bertinelli, A. D. Panagopoulos, P. G. Department of Electromagnetism and Radar, Office
Cottis, and R. De Gaudenzi, “MIMO over satellite: A review,” IEEE National d’Études et Recherches Aérospatiales
Commun. Surveys Tuts., Mar. 2011. (ONERA), Toulouse, France. He is participating in
[5] A. D. Panagopoulos, P.-D. M. Arapoglou, G. E. Chatzarakis, J. D. COST Actions IC0802 and IC1004 and ESA funded
Kanellopoulos, and P. G. Cottis, “Long-term rain attenuation proba- SATNEX III. His research interests are channel modeling for satellite and
bility and site diversity gain prediction formulas,” IEEE Trans. An- terrestrial communication systems.
tennas Propag., vol. 53, no. 7, pp. 2307–2313, 2005. Mr. Kourogiorgas is a member of Technical Chamber of Greece.
KOUROGIORGAS et al.: ON THE EARTH-SPACE SITE DIVERSITY MODELING: A NOVEL PHYSICAL-MATHEMATICAL OUTAGE PREDICTION MODEL 4397