Elevation-dependent Channel Model and Satellite Diversity

Transcription

1 Elevation-dependent Channel Model and Satellite Diversity for NGSO S-PCNs Hermann Bischl Markus Werner Erich Lutz German Aerospace Research Establishment (DLR) Institute for Communications Technology P.O. Box Wessling, Germany Tel: $ (2884,2826,2831) Fax: $ { Hermann. Bischl, Markus. Werner, Erich. dlr. de Abstract In this paper we propose an elevation-dependent Rice - Rayleigh/lognormal channel model for communications of personal and mobile users via non-geostationary satellites (NGSO S-PCNs). Based on this channel model, we discuss the performance of satellite diversity for NGSO S- PCNs. The parameters of the channel model are derived from channel measurements, which have been carried out for different elevation angles and in different environments. The measurements indicate that the channel behaviour in NGSO S-PCNs strongly depends on the elevation angle. In order to evaluate the performance of satellite diversity we present a method which allows the modelling of two statistically dependent satellite channels. The correlation between the channel states is also derived from channel measurements and depends on the azimuth separation of the two channels and on the elevation angles. We evaluate the performance of satellite diversity for LEO and ME0 systems (Globalstar and ICO) and different mobile user environments. For these systems, some crucial benefits and drawbacks of satellite diversity are discussed. It can be shown that the service availability can be significantly improved by satellite diversity. 1 Introduction The availability and quality of service in geostationary and non-geostationary satellite systems is crucially influenced by the particular characteristics of signal propagation in the link between the mobile or personal user and the satellite. Specifically in the LEO/MEO satellite scenario, the behaviour of the channel - and hence the parameters of any shadowing and fading processes for a channel model - are expected to be closely coupled with the varying elevation angle of the mobile user link. In this critical shadowing and fading environment, satellate diversity can act as an efficient countermeasure against QOS deterioration by exploiting multiple satellite visibil- ity and providing simultaneous communication via two or more visible satellites. Systems like Globalstar [l] and IC0 (former Inmarsat-P) [2] include the use of satellite diversity as fundamental system characteristic. From the channel modeling viewpoint, in the diversity case it is important to include possible correlations between the two channels, because this can be expected to influence the benefit of satellite diversity. In the following we first present an elevation-dependent analog narrowband model for a single channel, and then extend to a combined elevation-dependent and azimuthcorrelated analog model for the dual satellite diversity case. All parameters are derived from evaluation of measurement data. The models are then integrated into computer simulations for the performance evaluation of satellite diversity in terms of service availability improvement. Numerical results are given for two constellations: (a) Globalstar, with 48 LEO satellites at 1400 km in eight 52'-inclined orbits and (b) ICO, with 10 satellites at km in two 45'- inclined orbits. 2 Channel Model 2.1 Elevation-dependent narrowband model In order to investigate the characteristics of signal propagation in the mobile user link, a number of propagation measurements have been performed, and several channel models have been derived, describing the transmission path between a mobile/personal user and a GEO or non-geo satellite [3] - [8]. In this link, multipath fading occurs because the received signal does not only contain the transmitted signal but also echo components being reflected from objects in the surroundings. The received total power of the echoes mainly depends on the type of user environment (urban, suburban, rural, etc.) and on the antenna characteristic of the /96 $ IEEE 1038

2 user terminal. Antennas with wide-angle patterns tend to gather more echo power than directive antennas. Opposite to antennas mounted on top of a vehicle, handheld terminal antennas may pick-up strong specular reflections from the ground [5]. Variation of the received power with time is caused by movement of the user, of the (non-geostationary) satellite, or of reflecting objects. Shadowing of the satellite signal is caused by obstacles in the propagation path, such as buildings, bridges, and trees. The percentage of shadowed areas on the ground, as well as their geometrical structure strongly depends on the type of environment. For low satellite elevation the shadowed areas are larger than for high elevation. Especially for streets in urban and suburban areas, the percentage of signal shadowing also depends on the azimuth angle of the satellite. The fading process is switched between Rician fading, representing unshadowed areas with high received signal power (good channel state) and Rayleigh/lognormal fading, representing areas with low received signal power (bad channel state). The switching process is modeled by a twostate Markov chain (see Fig. 1). The Rician probability density of the momentary received power S in the good channel state is given by: PRice( S) =,,-@+%cl (2c6) (1) Here c is the direct-to-multipath signal power ratio (Rice-factor) and 10 is the modified Bessel function of order zero. When shadowing is present, it is assumed that no direct signal path exists and that the multipath fading has a Rayleigh characteristic with short-term mean received power SO. The probability density function of the received power conditioned in mean power SO is The slow shadowing process results in a time varying short-term mean received power So for which a lognormal distribution is assumed: PLN(S0) = 10 1 (10logSo - p)2. -ezp[- fialn10 So (3) Here p is the mean power level decrease (in db) and cr2 is the variance of the power level due to shadowing. Due to the movement of the non-geostationary satellites, the geometrical pattern of shadowed areas is changing with time. Similarly, the movement of a mobile/personal user translates the geometrical pattern of shadowed areas into a time-series of good and bad channel states. The mean durations in the good and bad state, respectively, depend on the type of environment, satellite elevation, and mobile user speed. Considering a mobile user with speed w, the mean extents (in meters) of shadowed and unshadowed areas, Eb Figure 1: Switching between good and bad channel sate and E,, translate into mean time intervals Db and D, that the channel stays in the bad or good state, respectively. For a transmission rate R, the mean state durations normalized to tlhe symbol dumtion and the transition probabilities Pbg and Pgb result in Fig. 2 shows elevation-dependent channel pzrameters of the Rice-Rayleigh/lognormal model for an urk an environment, which are derived from measurements. D,, c and p are increasing with the elevation angle, whereas Db and U are decreasing. I ny n O Elevation angle [degl Figure 2: Elevation-dependent channel parameters for urban environment,. 2.2 Correlation of two land rnobjile satellite channels Some of the considered satellite systems are based on the satellite diversity concept, which is in more detail discussed in the next section. Exploiting multiple satellite visibility on earth, the service avai1,ability (the perceni,age of time when the service is available) may substantially be improved. Of course, gain in service availability can only be achieved if the considered satellite channels behave different. Therefore, any dependency between t,he channels influences the benefit of satellite diversity. 1039

3 In [lo] a concept for modelling two statistically dependent satellite channels was developed. To this end, a combined Markov model was derived which includes the correlation between the two channels. It was shown that a negative correlation between the channels increases service availability, and vice versa. For the definition of the correlation coefficient we consider the amplitude ci(t) of channel i = 1,2 as a stochastic process which is 0 for the shadowed channel state and 1 for line-of-sight condition: 18C 0 Ci(t) = 0 bad channel state 1 good channel state. (5) 270 The mean value and variance of the channel amplitude are: Figure 3: Azimuth correlation of shadowing in urban environment. With this, the correlation coefficient can be defined as time average According to (8), the correlation coefficient can be evaluated from pairs of (time-synchronized) channel measurements with regard to the same mobile terminal and generally depends on the user environment, as well as the elevation and azimuth angles of the channels. As shown in [9], the dependency on the azimuth angles may approximately be described as a function of their difference Ap. With this simplifying limitation, p can be estimated from circular measurements at constant elevation angles [5] or from fish-eye photos [6], [7] for a single fixed user position, according to Here, p depends on the user environment, the chosen pair of elevation angles, and the azimuth separation Ap. An example for the azimuth correlation of shadowing in an urban area is given in Fig. 3. The correlation decreases with increasing azimuth separation and is smaller if the satellites have different elevation angles. The dependency of the correlation factor (and of service quality) on azimuth separation Ap indicates that the statistics of Ap are an important characteristic of a nongeostationary satellite constellation. Fig. 4 shows the cumulative probability density function (CDF) of Ap for two example constellations, Globalstar and ICO. For both constellations, the probability for Ap < 30 is negligible. Therefore, satellite diversity should be effective I Alimuth [deg] Figure 4: Probability density function of Acp for the Globalstar and IC0 constellations. It is clear that the combined shadowing behaviour of two mobile satellite channels can be modeled by a fourstate Markov chain, describing the possible combinations of good and bad states of channels 1 and 2, respectively, Fig. 5. Positively correlated channels tend to occupy equal states, i.e., compared to the four-state model for uncorrelated channels, the probabilities of states 0 and 3 are higher, and the probabilities of states 1 and 2 are lower. Accordingly, variables x, y, v, and w can be introduced which increase the transition probabilities leading to states 0 and 3 and decrease the transition probabilities leading to states 1 and 2. For negatively correlated channels, the opposite modifications are required [ 101. The useful range of the variables 2, y, v, and w is limited by the requirement 0 5 pzj 5 1, i,j = 0 these ranges, the variables x, y, v, and w can be chosen freely. However, in order to have a single parameter for adjusting the correlation coefficient p, we couple 2, y, U, and w through a scaling coefficient c, 0 5 c 5 1, such that they move simultaneously within their ranges. This procedure reduces the degree of freedom contained in the general four-state model, however, it will not restrict the 1040

4 Constellation Simulation mutual visibility <a Is determined by geometry --- mutually visible single channel model Channel 7 channel above fade margin Figure 6: Approach for satellite diversity simulati(3n and performance evaluation in terms of service avail%bility. 4 Numerical R,esults Figure 5: Four-state Markov model for two land mobile satellite channels. range of possible values for the correlation coefficient. The scaling factor c is determined by the correlation coefficient of the channels to be modeled [lo]. 3 Satellite Diversity Satellite systems offering single coverage of the service area may rely on extensive link margins to overcome not too heavy shadowing (tree shadowing, e.g.). Other systems such as Globalstar or IC0 essentially provide double coverage of the earth. This feature enables the application of satellite diversity, i.e., the simultaneous communication with a user via two or more satellites. If one of these satellites is shadowed, there is some chance for another satellite being still in view to the user and maintaining the service. In this way, satellite diversity can substantially improve service availability (the percentage of time when the service is available). Fig. 6 illustrates these considerations in the case of dual satellite diversity and at the same time shows the two levels of simulation performed during performance evaluation: On the constellation level, mere geometrical visibility conditions for any given mobile terminal MT and its serving gateway GW are determined. Depending on the actual geometrical situation (O,l, or 2 satellites mutually visible) a channel simulation is performed on the base of the corresponding channel model as introduced above. With two satellites visible, pure selection diversity is assumed. We performed extensive simulations for one GW at (-100' W / 40' N) and 16 UPS equally distributed in a circle around the GW. The minimum allowed elevation angle was 5' for tlhe GW and 20' for the MT. li'urthermore, we assumed that the GW always selects the two satellites with the the highest elevation angle. Figs. 7 and 8 show the service availability for the LEO-satellite system Globalstar (48 Satellites, 1400 km orbit altitude) and the MEO-satellite system ICCl (10 Satellites, km orbit altitude), respectively. For both systems w: assumed a link margin of ;I db. The channel parameter3 were taken from measurements. For the city environment the channel parameters are shown in Figs. 2 and 3. In both satellite systems we have a significant improvement in service availability through satellite diversity. The effects are more distinct in the Globalstar system than in the IC0 system, whereas the ab,solute availability figures in,he diversity case are nearly the same in both systems. 4 reason for that are the better visibility statistics for Clobalsar. It turned out from simulations that a user has two (mutually) visible satellites at 80% of the time in the Globalstar system and at 510% of the time in the IC0 slstem. Comparing the highvvay environment with the city environment one might conclude that reasonable satellite PCN seems to be very well possible in highway or rural areas, whereas in urban areas the satellite will of course not be more than a complementary or a backup solution for terrestrial systems. Furthermore, we can say that the simulation results depend strongly on the channel model, which has to be chosen carefully. As Figs. 7 and 8 show, the evaluated service availability for satellite diversity in an urban environment is ;significantly reduced, if the evaluation is based on a digital channel model (good channel state, bad channel state) iinstead of the analog Rice-Rayleigh/lognormal model. 1041

5 e----& I c--l- c----* ,...., the difference of the azimuth angle and the elevation angles of the satellites. Simulation results show that service availability can be significantly increased by satellite diversity. Reasonable satellite PCN seems to be very well possible in highway or rural areas, whereas in urban areas the satellite will of course not be more than a complementary solution for terrestrial systems. - City, diversity, analog channel model City, diversity, digital channel model City, no diversrty, analog channel model lo00 UP distance from GW [km] Figure 7: Service availability versus UP distance from GW for the Globalstar system. : l *-----* ----_* i Highway, diversity, analog channel model City, diversity, analog channel model City, diversity, digital channel model --c City, no diverslty, analog channel model I I I I UP distance from GW [km] Figure 8: Service availability versus UP distance from GW for the IC0 system. 5 Conclusions An elevation-dependent channel model for NGSO S-PCNs has been derived from channel measurements. The model is a Rice-Rayleigh/lognormal model with elevationdependent channel state durations D, and Db, Rice-factor c, mean power level decrease p and standard deviation U. Based on this channel model satellite diversity for LEO (Globalstar) and ME0 (ICO) systems has been investigated by extensive computer simulations. In this context, a correlation between the land mobile satellite channels was introduced with a correlation coefficient depending on References [l] R. A. Wiedeman, A. J. Viterbi, The Globalstar mobile satellite system for worldwide personal communications, in Proc. 3rd Int. Mobile Sat. Conf. (IMSC 93), Pasadena, CA, June [a] Inmarsat-P The world in your hand. Inmarsat brochure, Summer [3] C. Loo, Measurements and models of a land mobile satellite channel and their applications to MSK signals, IEEE Trans. Vehic. Technol. VT-35 (1987), pp [4] E. Lutz et al., The land mobile satellite channel - recording, statistics and channel model. IEEE Trans. Vehic. Technol. VT-40 (1991), pp [5] A. Jahn and E. Lutz, DLR channel measurement programme for low earth orbit satellite systems in Proc. 3rd Znt. Conf on Universal Personal Communiactions (ICUPC 941, San Diego, USA, Sep./Oct [6] A. Jahn, Measurement programme for generic satellite channels. Final Report, Inmarsat Purchase Order P004001, Nov [7] A. Jahn, Propagation data and channel model for LMS systems. Final Report, ESA Purchase Order , Jan [8] A. Jahn et al., Narrow and wideband channel characterization for land mobile satellite systems: Experimental results at L-band, in Proc. 4th Int. Mobile Sat. Conf (IMSC 95), Ottawa, Canada, June [9] P. P. Robet, B. G. Evans and A. Ekman, Land mobile satellite communication channel model for simultaneous transmission from a land mobile terminal via two separate satellites, International Journal of Satellite Communications, vol. 10 (1992), pp [lo] E. Lutz, A Markov model for correlated land mobile satellite channels. Submitted to International Journal of Satellite Communications. [11] H. Bischl, M. Werner and E. Lutz, On satellite diversity and mobile user environment for NGSO S-PCNs, in Proc. 16th AIAA Int. Comm. Sat. Syst. Conf. (ICSSC 961, Washington, DC, Feb