Beruflich Dokumente
Kultur Dokumente
HAPS
Links Performance
Flow
Simulation
Abstract
The High Altitude Platforms Stations (HAPS) represent one attractive option today for telecommunications applications. In the evaluation of these systems are necessary channel models to obtain the performance of them. In this paper, a statistical switched broadband channel model and a channel model based on time series generator are postulated and analyzed by simulation, representing several conditions related with the coverage and environmental conditions.
-
Keywords series.
I. INTRODUCTION
T HIS paper presents a statistical switched broadband channel model based on a transversal filter scheme with three states for digital transmission in HAPS link in L band. At the same time, when the system is operating in the Ka band is used a time series generator based for to characterize the weather conditions. Performance results were obtained by simulation.
II. HAPS (HIGH ALTITUDE PLATFORM STATIONS)
A. Systems based on HAPS In a system based on HAPS the platform is positioned above the ground for radio electric coverage area, at a height between 21 and 25 km. There are three areas of coverage: Urban (UAC), Suburban (SAC) and Rural (RAC) which are determined by the position of the receiver (the elevation angle and the distance from the
A. HAPS typical scenarios. Statistical definitions The first step toward a meaningful switched channel model is to identify typical scenarios. We will restrict that mentioned by the ITU-R in the Recommendation P.681-6 [4]. In a system based on HAPS, there are a LOS condition and multipath components in the Urban area. The LOS condition is due to the short distance from the platform to the receiver on the ground; the multipath component consists of many reflections from the obstacles in the area, each of them being independent and randomly phased. The envelope of this diffuse signal is characterized by means of a Rayleigh probability distribution. The sum of a constant envelope and a Rayleigh distributed diffuse signal results in an envelope Rice probability distributed [5] [6]. In the suburban area, the obstacles near to the receiver cause a signal shadowing and an attenuation of the direct signal. The attenuation of the direct signal varies due that some obstacles are moving (vehicles). The attenuation of the direct signal undergoes log-normal distribution [6].
B. The three states semi-Markov channel model As the environmental properties change, the received signal can not be represented by a model with constants parameters, therefore the channel could be modeled using a finite states Markov chain model. The transitions among states are determined by a matrix P, where each element Pij represents the probability that the channel changes from i to j state. For the channel model with three states A, B and C, is defined a transition matrix P
P=
,
subplatform point, SPP)[1]. A system based on HAPS will allow to obtain a better quality of the signal, due that most of the transmission time the system guarantees a LOS condition, diminishing the shadowing effect in comparison with the terrestrial system and also has a less delay with regard to the satellite systems [2] [3].
A stationary state vector ir could be calculated, using the properties of the Markov processes as
(1)
jose.cuevasgitesm.mx).
This work was partially supported by the European Commission (CAPANINA Project) and the Spanish State. J. L. Cuevas Ruiz is investigator-professor of the Electrical Engineering Dept., in the ITESM Campus Estado de Mexico, Mexico (phone: +(5 55) 58645555 ext 2476); fax: +(5 55) 5864 5651;e-mail:
J. A. Delgado-Penin, Author, Full-professor in the TCS-Department of Signal Theory and Communications, UPC, Barcelona, Spain. (+34 93 40 16815); (e-mail: delpengtsc.upc.edu).
re=1 (3) where I is the identity matrix, P is the transition matrix and e= [1 1 ... .]l [7]. Each element 2ti represents the percentage of the total time that the process remains in the i state. r= (7tA 7tB 7tC) (4) A semi-Markov process is a Markov chain where the time between changes of states are random and defined for some kind of distribution [8]. From a Markov process, a
ir(I-P)=0
(2)
1184
new semi-Markovian process may be defined and this process will be described by a new transition matrix r [8], where
r =
P.
i
A measure of delay spread which is also used is the rms delay spread c,. 1 (10) Rb <-
j-P
IJ
for
i.j
and
ri= 0 (5)
r
C. Fades distributions Some distributions for several kinds of channels are described for fade duration in the ITU-R Recommendation [4]: * A state. LOS condition. * B state. Slight shadowing. * C state. Total obstruction. For the A state, the duration follows an exponential distribution given by
7
a
a
0.
PAB
where the parameters /3 and y are function of the level of the shadowing and for d > l/3' . The duration for the others states follows a lognormal distribution valid for d > 0.1 m, with PB,C (D < d)= (1+ erf[(ln(d) - ln(a)) / J2]) 12 (7) where a is the standard deviation of ln(d), the mean value of ln(d) is ln(a) and the error function is defined in the ITU-R Recommendation P. 1057. The ITU-R Recommendation P.68 1 establishes the parameters and distribution for the duration of these states. These durations can be established in agreement with the characteristics of each state. Besides the transitions probabilities are defined for a semi-Markovian process with three states for different kinds of environments. Parameters above are shown in the Table No. 1.
D. Time delay spread In a mobile radio environment a single impulse originates many echoes of this impulse at the receiver. The received signal may be formulated as
PA(D< d) =1-/Jd-
(6)
PBA
PBC
PCA
PAC
0.65
0.65
0.42
0.35
0
0.35
0
0.58
0
PCB 1 1 1 The value of r is defined as the first moment of the power delay profile and is calculated by
Y ak Z-k
k p
_
EP(Zk )Zk
p k
and the rms delay spread is the square root of the second central moment and is defined by 7f jr 2 _(-)2 (12) where
E
-2
-
Za k
ZP(rk)
k
(1 1)
a27.2
k
Ya2
P(rk )k Z k
ZP(rk)
k
(13)
Hereby is possible that a channel is less selective in frequency if Tm is smaller than the time symbol T,. If the power delay profile is known then is possible to model it by means of a tapped delay line [10]. The output signal is given by
h(t, x) =
N(x)
i=l
(8)
N(X)
k=l
(14)
where ai are the amplitude of the ith signal with a delay ri; the variable N represents the number of echoes considered. An important parameter is the time delay spread interval which is measured from the first symbol to the last detectable echo, and is different for each kind of environment. The time delay spread would cause intersymbol interference (ISI) for digital transmission. To avoid ISI, the transmission rate Rb should not exceed the reciprocal value of the delay spread A if the receiver is at a standstill, [9].
<A
where a(t) is a complex Gaussian process, and Tk are the paths delays. For our analysis is considered the power delay profile reported in [11] for HAPS links, which was computed using parameters in the L band and with an elevation angle of 45. At same time, the model was evaluated using the power delay profile calculated for a HAPS link proposed in [11].
IV. BROADBAND CHANNEL MODEL In [11] are reported results related with HAPS performance and considering a not switched channel model. In this paper are considered switched channel models with two and three states. Some values of average
(9)
The channel autocorrelation function evaluated at t=O gives the average output power of the channel as a function of delay and this is referred as the power delay profile. The range of values of T for which the power delay profile is not zero is called the multipath delay spread, Tm.
receiver is in motion.
time delay spread for different environmental conditions are reported in the literature; typical values of 0.5 Rs for suburban and 3 Rs for urban areas were founds in [9]. In the model proposed by Dovis [1I1] are reported theoretical values for systems based on HAPS: 21 ns (rms delay spread) and 952.38 Khz (coherence bandwidth).
1185
In the approach of this paper are added some hypothesis of the reference [11] and [12] related with the tapped delay line model. Besides it was postulated and evaluated a system with a digital transmission rate of 200 Kbits/se (under these circumstances there is not ISI effect due to the channel). A. Three states channel model In agreement with tapped delay line model, a broadband switched channel model may be implemented by switching among three possible kinds of channel (Fig. No. 3). * Rice fading (A state). LOS and multipath without multiplicative fading. * Total obstruction condition (C state). * Obstruction of the direct component (B state). The transitions among three states, is defined in the Table No.1 and the parameters for the fading durations were established for the Suburban I environment indicated in the same Table, and we can see that the model just has transitions between adjacent states too. The transition matrix for a suburban environment I and II are the same, but the values of the parameters for the fading distribution change.
B. Channel model The channel model is shown in Fig. No. 1. In this case, the rain attenuation is simulated with the Time Series generator proposed by Fiebig [ 13].
V. CHANNEL MODEL BASED ON A TIME SERIES GENERATOR A. Multipath effect at high frequencies The multipath effect in a transmission at mm bands, with a fixed user, can be neglected [8]. This channel model will characterize a Fixed Wireless Access (FWA). A wave front is reflected by a surface if this one is considered to be smooth, whose ruggedness dimensions are relatively big compared with the wavelength. Is possible to establish that flat surfaces, with big ruggedness, compared with the wavelength, tends to be a reflectors. Surfaces with minor or comparable ruggedness, relative to the wavelength of the signal, cause dispersion of the energy. In consequence, is very important to define if a surface is smooth or not. For to determine the ruggedness level that a surface has, is used the Rayleigh expression (15); in agreement with this relation a surface could be considerate as a smooth if a ruggedness height h does not exceeds a critical height h,*; this factor is a function of the incidence angle 0 and of the wavelength 2, like is shown in (15). We can see that for frequencies in the Ka band, the critical height is very small (0.2 cm).
-8senO
2 3.5 4 1.78 8 0.89 24 0.29 30 0.23 TABLE II: CRITICAL HEIGHTS FOR SEVERAL FREQUENCIES
(15)
For previously exposed, for a wavelength signal in Ka band, the most of the signal energy that hits on a surface will be dispersed, and the multipath effect can be
neglected.
Fig. No. 1. Channel model. Rain fading is understood as signal attenuation due to the clouds and precipitation where rain has a coherence time typically in the order of minutes. This time series generator was designed to simulate attenuation on a satellite system; we propose an approximation for to use this generator in a HAPS system. Basically, the principal difference between these systems is that in the satellite system is necessary to transmit with more power, for to achieve the same received power at the user that in the HAPS system. The attenuation level due to the rain is the same in both cases. C. Time series generator andfrequency scaling The implementation of the time series generator is based in a classification of the measure signal (power) in three kinds of the signal segments: * Almost constant (C). * Monotonously decreasing (D). * Monotonously increasing (U). In agreement with the analysis of the information that have been obtained from the measurements carried out by Fiebig, the attenuation level in a time t depends only of the attenuation in some time At seconds before and on the actual type of signal segment (C, D or U). Furthermore, the measured pdfs (probability density functions) of the likelihood P(y/x) for the segments C, D and U has a Gaussian-like shape, where P(y/x) is the likelihood that the attenuation level is y dB, conditioned that it has been x dB At seconds before. The implementation of this time series generator is based on the scheme shown in the Fig. No. 2. The channel model was implemented following the next steps: * A Gaussian random generator was implemented. At the beginning, in the initial conditions, were assigned a mean and a standard deviation (std) for a constant segment, for an attenuation of 1 dB (mean= 1, std=0.22). * The difference between two samples separates At seconds is calculated as [dif = r(t- At)-r(t)], to obtain the monotony of the shape follows the next criteria: * If the absolute value of dif is less or equal to 1, then is assumed that the function to be continued with the same trend (Constant C, Decreasing D or Increasing U). Or C for abs [r(t-1) -r(t) < I dB. * If dif is positive and bigger than 1, (if r(tAt)> r(t), y abs[r(t-At)- r(t)] >1), the
1186
* *
function will be decrease . D for [r(t- At)r(t)] > 1. If dif is negative and less than 1, (if r(t- At)< r(t), y abs[r(t- At)- r(t)]>1), the function will be increase. A for [r(t-1)-r(t)]<-] Then, a new mean and standard deviation will be assigned, in agreement with the kind of segment and the attenuation level; this is carried out measuring the attenuation level (dB), comparing the samples r(tAt)and r(t) and then the Gaussian generator take the new statistics parameters (mean and standard deviation) that are indicated in [7].
Time series Generator output
10
=-
-= = = = = = f == t= = = = = = = = X
10
.::.A state * B state tate Switched channel Switched channel + RS-CONV code
X 10
0-
10
EbN0(dB)
Perform ance
10
12
14
16
10
r(t)
10
Gaussian C
r(t- At )AMt
10
Fig. No. 2. Time series generator The values of the means and standard deviations for each attenuation level for the summer are given in [7]. In this reference there are reported values of the means and standard deviations for the spring too. The long term frequency scaling of attenuation allows extension of long term statistics at one frequency to a different frequency. The frequency scaling is widely used on the applications of FMT (Fading Mitigation Techniques), where there is a monitoring of a signal at a certain frequency and then use the frequency scaling to evaluate the attenuation at other frequency [14].
Lognmr
10
0.5
1.5
EbNO (dB)
2.5
3.5
4.5
Fi.Nov3 Sw o
The information given by Fiebig was obtained in a measurement campaign at 40 GHz, in Oberpfaffenhofen, Germany, with an elevation angle of 34.80. Considering this data, this evaluation is valid for a fixed user placed in the urban area.
1187