Design study for a CDMA-based LEO satellite network: downlink system level parameters

Share Embed


Descrição do Produto

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 14, NO. 9, DECEMBER 1996

1796

Savo G. Glisic, Senior Member, IEEE, Jaakko J. Talvitie, Member, IEEE, Timo Kumpumaki, Matti Latva-aho, Student Member, IEEE, Jari H. Iinatti, Member, IEEE, and Torsti J. Poutanen

Abstract-The performance analysis of a new concept of a code-division multiple-access (CDMA) based low earth orbit (LEO) satellite network for mobile satellite communications is presented and discussed. The starting point was to analyze the feasibility of implementing multisatellite and multipath diversity reception in a CDMA network for LEO satellites. The results will be used to specify the design parameters for a system experimental test bed. Due to the extremely high Doppler, which is characteristic of LEO satellites, code acquisition is significantly simplified by using a continuous wave (CW) pilot carrier for Doppler estimation and compensation. The basic elements for the analysis presented in this paper are: the channel model, the pilot carrier frequency estimation for Doppler compensation, and multipath and multisatellite diversity combining.

I. INTRODUCTION HE modern trend in the analysis of mobile satellite communications is very much oriented toward the implementation of the code-division multiple-access (CDMA) concept. Project examples are summarized in Table I. One can see that most of the projects are considering low earth orbit (LEO) constellations due to a limited link power budget. Unfortunately, this type of satellite constellation generates another problem due to severe Doppler in carrier and chip frequency. In addition to this, due to low minimum elevation angles (see Table I), multipath and shadowing are also a problem that has to be carefully addressed. The existence of carrier and code Doppler result in a prolonged acquisition process or increased hardware complexity due to the need for a two-dimensional (delay and frequency) search of code synchronization. A number of papers have been published addressing this issue [ 11-[5]. In our system, we suggest Doppler compensation prior to code synchronization, and analyze conditions under which this solution is feasible. The approach is based on the utilization of a continuous wave (CW) pilot carrier, and the analysis, based on known frequency estimation methods [SI-[ 161, provides values for the minimum power of the pilot carrier needed to provide reliable Doppler

estimation. This is equivalent to the information of how much system capacity is to be sacrificed for the pilot carrier. The multipath and shadowing problem can be solved by using multipath and multisatellite diversity reception. In this analysis, we start with the model of the channel and then based on the channel delay profile we analyze how much improvements we can expect from using a multipath RAKE receiver. To combat the shadowing problem, we consider feasibility of using multisatellite diversity reception. The overall analysis demonstrates that CDMA for LEO satellites is a feasible approach, and detailed system parameters are presented along with discussion of the possible system improvements. In this initial analysis, we have not considered all aspects of receiver performance. Code acquisition, code tracking, and carrier frequency tracking and phase estimation are issues that we have not addressed in this paper. Also, the overall biterror rate (BER) performance of the receiver with multisatellite combining is not discussed. However, all these issues are currently being analyzed in detail. The paper is organized as follows: A general system model is presented i-n-Section 11, and the channel model is presented in Section 111. Pilot carrier frequency estimation is discussed in Section IV. Multipath diversity (RAKE) and multisatellite diversity reception are discussed in Section V. The results are presented and discussed in Section VI, and the conclusions are presented in Section VII.

11. MODELOF MULTISATELLITE CDMA DIVERSITYRECEPTION In this paper, we analyze the downlink and the mobile unit receiver. We assume a synchronous channel and the spectrum of the transmitted signal as presented in Fig. l(a). Analytically the transmitted downlink signal for the mth satellite can be expressed in time domain as S,(t)

=

SLm

+ s, + s,,

K

= Manuscript received May 1, 1995; revised October 4, 1995. This work was supported by the European Space Agency under Contract 10885/94fl\JLfl\JB and it was carried out in cooperation with Elektrobit Ltd., Oulu, Finland. S. G. Glisic, J. J. Talvitie, T. Kumpumaki, M. Latva-aho, and J. H. Iinatti are with the Telecommunication Laboratory, University of Oulu, Finland (email: [email protected]). T. J. Poutanen is with Elektrobit Ltd., Oulu, Finland. Publisher Item Identifier S 0733-8716(96)05668-5.

Re{C(k m) exp[(wot

+ (Pm)])

k=l

+ Re{exp[~(wo+ w c ) t + ~ m ]+}Pm(t)

(1)

where the first term represents K traffic channels, the second term the pilot carrier for Doppler estimation, and the third term a pilot channel signal for synchronization purposes.

0733-8716/96$05.00 0 1996 IEEE

GLISIC er a1 DESIGN STUDY FOR A CDMA-BASED LEO SATELLITE NETWORK

1797

4

4

RAKE

UNIT

FREQUENCY COMPENSATION DATA EXTRACTION

I RAKE UNIT

DOPPLER COMPENSATION FREQUENCY COMPENSATION DATA SUPPRESSION

RAKE UNIT

MULTISATELLITE DIVERSITY --f COMBINER

I

4

Fig. 2. General block diagram of the CDMA satellite receiver.

Fig. l(b). This signal can be presented as Fig. 1. Signal spectra. (a) Transmitted signal for synchronous channel (pilot carrier located to the spectral null), (b) received signal for synchronous channel at the input of the demodulator, and (c) received signal for synchronous channel at the input of the demodulator with prenotching.

(3) m

m

where f c m is the channel impulse response for the transmission between satellite m and the receiver. The overall system performance would be further improved Parameter C ( k , r n ) is a complex signal that can be ex- if for LEO satellites the position of the different transmitters pressed as in the orbit could be such that the differential Doppler, A f o (difference in carrier Doppler) is larger than the bit rate R. For C(k1 m) = Cr(k m) 3Cz(k, m) any other satellite constellation, where A f o for each pair of = Sm(Ckmrdkmr jckmzdkmz) ( 2 ) satellites is not larger than R, an additional frequency offset where S, (real) stands for the “satellite code” which is a long could be introduced in the pilot carrier of each satellite. An code used to separate signals coming from different satellites, additional characteristic of the downlink signal structure is that Ck, (complex) stands for the kth user “channelization code” all users are chip, symbol, and frame synchronous. used to separate different users within the same satellite, and Instead of using a pilot tone, information about the Doppler d k m (complex) is the convolutionally encoded information of can be distributed in the network from the central node by the lcth user of the mth satellite. C k , may be either the Walsh using an frequency shift keying (FSK) signal with a level function or the Gold sequence. In general, different satellites higher than the level of the CDMA signal. This approach may have different sets of users in the traffic channel, but all fequires complex coordination within the network. Undei these cunditions, the receiver structure for three satellites is shown of them have the specific user we are interested in. Let’s say user IC = 1 is common for all satellites, i.e., in Fig. 2. The operation of the receiver can be described as d l , = d l , b’m. Because all other signals will appear as follows. The first step is to detect the Doppler shift for each satellite. interference from the point of view of user 1, it is irrelevant Independent of whether this information is sent by the network whether or not d k , = d k , tlm and b’k. For notation simplicity, in the form of an FSK or M-ary FSK signal, or whether we will assume that this condition holds. The channelization codes c k are the same for each satellite, i.e., Ckm = ck,‘dm. a pilot carrier is used, the detector may be based on fast Fourier transform (FFT). This block is designated in Fig. 2 The structure of the satellite (long) code is the same for each as the “frequency compensation data extraction” block. Three satellite but initial states can be different, S, = S ( t - A,) separate frequency downconversions are performed by using where A, is a predetermined time shift. This is not necessary, local carriers f o = f ~ , ,where f D , is the estimated Doppler and due to different delays of the signals coming from different for the mth satellite (in our study m = 1,2,3). satellites with the same initial state, S, = S , Vm can be Prior to frequency downconversion, frequency compensaused. With these simplifications, (1) can be used with C = tion data (FSK signal or CW pilot) can be suppressed by using CT- $2 = S(Ckrdlir j C k z d k z ) . an adaptive narrowband interference canceller based on the A CW pilot is placed in between two CDMA signals (in least mean square (LMS) algorithm [6], [7]. If the Doppler the spectral null, see Fig. 1) so that zero or rather a low level range is narrow compared with the signal bandwidth, a simple of interference between the traffic channels and a pilot tone passive notch filter can be used for these purposes. In general, should be expected. In order to be able to operate with a low this can be analytically represented as s , = sL* fin, where f r n level of the pilot, the CDMA signal should be prenotched is the impulse response of the receiver notch filter. The third so that the estimation of the pilot signal frequency will not be option is not to suppress this signal but to use only the inherent affected by the variation of the number of users in the network. processing gain of the system. This will slightly reduce the This will be represented as smt = s k t * f t n , where f t n system capacity. is the impulse response of the prenotch filter and * stands Each frequency downconversion will produce a sum of three for convolution. A multiple satellite signal spectrum, received signals, one of which will be with essentially no Doppler, by the mobile unit for a synchronous channel is presented in and the other two which will contain a residual Doppler.

+

+

+

+

IEEE JOURNAL ON SELECTED AREAS IN COMMUNICATIONS, VOL. 14, NO. 9, DECEMBER 1996

1798

satellite-to~ground

terrestrial

I

I

I

Q;

LOS

MP

Fig. 3.

Overall structure of the channel model.

If the residual Doppler is larger than the bit rate, these two signals will be additionally suppressed in the correlation process so that, in further processing, only one signal from a corresponding satellite will be dominant and signals from the other satellites will be suppressed. The additional separation between two signals from different satellites is based on different delays so that the long code will be able to separate such signals. In general, further processing is based on a RAKE receiver. Once a sample of the decision variable is formed at the output of a RAKE unit, this sample is combined in the multisatellite combiner with the outputs of the other RAKE units to receive the final decision about the bit being received. Prior to combining, the differential delays between the satellites have to be estimated so that the samples of the same bit are combined for the final decision. This function is formally represented as a separate block. In practice, this means proper synchronization of the long code in each RAKE unit.

111. CHANNELMODEL In this section, we propose a wideband land mobile satellite channel model that is based on preliminary results from a wideband measurement campaign. The model is a statistical (synthetic) model, with part of the parameters being deterministic functions of the satellite orbit geometry. The model is a general-purpose model in the sense that it is adaptable to different system configurations, i.e., different satellite constellations and user terminal types. Multisatellite reception with up to three visible satellites is included in the model as three separate satellite links. Each satellite link model consists of a satellite-to-ground part (deterministic) and a terrestrial part (statistical). The proposed model is restricted to the down-link only and includes the propagation channel and antenna effects. The model parameters are functions of the satellite constellation (LEO), the user terminal type (hand-held, portable and vehicle-mounted), and the elevation angle. Furthermore, two environment types, rural and urban, are considered. An overall block diagram of the channel model is shown in Fig. 3.

Fig. 4. Detailed block diagram of the terrestrial part.

The satellite-to-ground part is modeled by a delay 17;., an overall attenuation L, (including both free-space loss, and antenna pattern effects), and a Doppler shift W D ~ The ~ . modeling of the terrestrial part of propagation is based on a combination of currently available narrowband models [ 171, [18] and a wideband tapped delay line model. The approach is to extend the narrowband models to wideband models by replacing the Rayleigh component in the narrowband models with a tapped delay line model. The terrestrial part structure includes Mt taps, divided to the line-of-sight (LOS) signal and an overall multipath (MP) component. A general block diagram of the LOS and MP blocks of the terrestrial part -is shown in Fig. 4, where the coefficients gi ( t ) are uncorrelated complex Gaussian processes. Note that the first tap ( i = 0) is reserved for the LOS path, and the Mt - 1 remaining taps for the multipath part (MP). The general structure shown in Fig. 4 is used in slightly different configurations for three separate terrestrial environment cases: rural, urban unshadowed (urban good state), and urban shadowed (urban bad state). The real-valued coefficients C1 and C2 have different meanings and values in these configurations.

A. Rural Channel In the rural narrowband case, the model of Loo [17] is adopted. An additive combination of Rayleigh and log-normal fading is therefore assumed in this case. The probability density function (pdf) p ( r ) of the received signal envelope is given [17] by r

t (4)

where IO is the zeroth order modified Bessel function of the first kind. The parameters for the model are the mean ,ULN and the standard deviation C T L ,of ~ the log-normal fading, and the average power of the Rayleigh fading 20;. The model corresponding to (4) is presented in Fig. 5 , with a real-valued

GLISIC et al.: DESIGN STUDY FOR A CDMA-BASED LEO SATELLITE NETWORK

1799

3 Fig. 5. Terrestrial channel model by Loo [17].

log-normal process g L N ( t ) and a complex-valued Rayleigh process g R ( t ) . The model can be considered to consist of Rician fading, where the Rice-factor (LOS to multipath power ratio) is modulated by a log-normal process. Expanding the narrowband model of Fig. 5 to the wideband structure shown in Fig. 4, we replace C, by g L N ( t ) and C1 by GnLp,where the parameter G, controls the multipath power. Requiring that the overall average power of the MP part is equal to unity, we have that G,, =

a.

B. Urban Channel The terrestrial part in the urban case is modeled after Lutz et al. [18], as shown in Fig. 6. The model consists of a combination of Rician fading (left branch in Fig. 6) and lognormally shadowed Rayleigh fading (right branch in Fig. 6). In this case, note that the log-normal and Rayleigh processes are multiplicative, not additive-which is different from the Loo model. The Lutz model is characterized by four parameters; the mean ~ . L Nand standard deviation OLN of the log-normal distribution, the Rice-factor e, and the time-share of shadowing A (giving the average share of time spent in the shadowed state). The parameter A is defined using the average durations of the good and the bad states ( D g and Db, respectively, as shown in Fig. 6) as [18]

The overall pdf of the received signal power is defined in [ 181 by combining the Rician and the Rayleighflog-normal densities, weighted by 1 - A and A, respectively. Expanding the narrowband model of Fig. 6 to the wideband structure shown in Fig. 4, we replace in the good state Cz by unity and C1 by c-ll'. In the bad state, we replace Cz by zero and C1 by g L N ( t ) . For simplicity, it is again required that the overall average power of the MP part is equal to unity.

C. The Multipath Part From the above, it is clear that the MP part is the same in all model configurations. Each resolvable multipath signal, i.e., each tap in the model, is by itself a sum of several unresolved, reflected and scattered signal components, and is therefore subject to severe fading. This unresolved multipath is included in the time-variant path coefficients y; ( t ) .The coefficients are

Fig. 6. Terrestrial channel model by Lutz et al. [IS].

assumed to be complex zero-mean Gaussian signals, resulting in a Rayleigh-distributed amplitude. The fading rate of the path coefficients is determined by the bandwidth (total Doppler spread) of the signals g Z ( t ) . The power spectra of the path coefficients depend on the type of environment considered. For an urban area, the spectra are all of the classical (Jakes) type [18]. For a received unmodulated carrier this doppler spectrum is defined as [I91

4 K7 )1 2 4

s ( f )=

2KfDm

(II.

for

-

fDm

Lihat lebih banyak...

Comentários

Copyright © 2017 DADOSPDF Inc.