Professional Documents
Culture Documents
T HE
T HE
CRETE TIME / FREQUENCY COVARIANCE FUNCTION FOR USE IN THE RECEIVER STRUCTURE .
In the design and implementation of a receiver for wireless communications, it is important to have an understanding of the wireless radio channel, since it inuences the received signal signicantly. The derived channel model in this study is also used for verication. However, since a model of the wireless radio channel is a simplication, it does not necessarily follow that a receiver derived with the mind on one specic channel model works properly, when it is tested using a real wireless radio channel. The channel is modeled with the parameters from the LTE downlink SISO OFDM system based on the signal model derived in Chap. ??. The carrier frequenLTE CONSIDERATIONS
cies for downlink have been allocated in ten bands in the area 0.8 GHz 2.7 GHz [Dahlman et al.(2007), page 14]. A typical downlink is 4.5 MHz (5 MHz spectrum allocated), 15 kHz subcarrier spacing and based on the block sizes for the LTE downlink we chose 120 subcarriers and 56 OFDM symbols [Dahlman et al.(2007), page 59]. Realistically, the environment plays an important role in the behavior of the channel. The channel has different characteristics depending on if it is indoor, outdoor, urban, hilly ect.. A generic channel model based on a specic delay-Doppler power spectrum is derived. The model chosen is power delay prole of the COST 2071 typical urban model power delay prole and Jakes Doppler spectrum (also
1
The COST 207 model is an average from extensive collected data from different locations in Europe.
The model considers four different scenarios, each one characterized by a specic prole: typical urban
Mobile-station
Base-station
Fig 1.1: N ON - STATIC OUTDOOR URBAN ENVIRONMENT WITH MULTI - PATH FADING .
known as Clarkes spectrum). The advantage of this model is its simplicity and its generality. The channel model is rst described in continuos time. From this model a discrete time/frequency model is developed to describe the channel in each subcarrier. The channels between base-stations are assumed uncorrelated, which is a realistic assumption since they are not normally in physical proximity.
ponent (line of sight), Rician fading gives a better approximation2. As we only consider the small scale fading, we assume that the stochastic process behind the the time/frequency description of the channel is wide sense stationary (WSS), and that the auto-correlation for the delay/Doppler function is zero in all instances, where i and i , which means uncorrelated scattering (SU). The channel model
TAP DELAY LINE MODEL WSSUS CHANNEL
showing these characteristics is thus described as the WSSUS channel.We describe the channel as a tap delay line model, given by a time-variant channel impulse response [Tse and Viswanath(2005), page 41] and [Molisch(2005), page 121], which using a base-band representation is described as (1.1) h(t , ) =
Ntaps i=1
i (t )( i (t )) =
Ntaps i=1
where i (t ) is the ith tap and () is the Kronecker delta function, i (t ) is the timevariant ith delay, fc is the carrier frequency, i (t ) is the time-variant complex amplitude of the ith tap and Ntaps is the number of taps in the model. The amplitude of each tap i (t ) is modelled as a zero-mean complex Gaussian process, thus |h(t , )| is Rayleigh distributed. From [Bello(1963)] we use the notation to express the description of the channel in the different domains and the connection between them, see Fig. 1.2. The timevariant transfer function is the Fourier transform of the impulse response (1.1) from the delay domain to the frequency domain f (1.2) H (t , f ) =
Z
d h(t , ) exp( j2 f ).
A statistical description of the channel is obtained by the correlation function in time and frequency, given by the auto-correlation function of H (t , f ) over time t and frequency f [Molisch(2005)] (1.3) where
x
H (t , f )H (t + t , f + f )
f, t
= RH (t , f ),
denotes the expected value with respect to x and t and f is the shift
in time and frequency respectively. Notice, for the WSS channel the rst moment is constant. The time correlation function and the frequency correlation function of
2 Rician
h(t , ) Rh (t , )
WSS
US
H (t , f ) RH (t , f )
s(, ) SH (, )
WSSUS
CHANNEL
T HE
T HE ,
F OURIER
the channel are obtained from RH (t , f ) [Molisch(2005)] according to (1.4) (1.5) RH (t , 0) = RH (t ) RH (0, f ) = RH ( f ).
Averring across many channels given many different locations it is assumed that (1.6) RH (t , f ) = RH (t )RH ( f ).
The assumption is not true for one instantaneous channel, since the time and frequency correlation functions here cannot be assumed independent. Because of the US assumption the second moment of the scattering function is (1.7) s (1 , 1 )s(2 , 2 )
,
= SH (1 , 1 )( 1 )( 1 ).
The power spectrum is also obtained from the correlation function in (1.3), see Fig. 1.2, by the Fourier transform in the time direction and the inverse Fourier transform in the frequency direction of the correlation function [Durgin(2003)] (1.8) SH (, ) =
Z
d f exp( j2 f )
dt RH (t , f ) exp( j2t ).
Receiver
OF THE CHANNEL WITH NO LINE OF SIGHT BETWEEN TRANSMITTER SCATTERS ARE EQUALLY DISTRIBUTED IN ALL DIRECTIONS AND
T HE
T HE SIGNAL FROM ALL DIRECTIONS HAS ON AVERAGE THE T HIS MODEL OF THE D OPPLER SPECTRUM IS GIVEN BY
(1.12).
It follows from the factorization in (1.6) that SH (, ) also factorize: SH (, ) = SH ()SH ().
by the spectrum. Jakes spectrum is computed from the denition of the Doppler shift. The Doppler shift is caused by the velocity of the receiver (1.9) where c0 = 3 108 frequency3
m s
fc cos() Hz, c0
and is the azimuth of arrival of the radio wave. The Doppler shifts
from and is the same, we thus limit the probability density function (pdf) to p() = U (0, ). Thus for each frequency in the Doppler dispersion is given (1.10) f = fc 1 || cos () , c0
where = || cos(). Transforming this uniform distribution into the Doppler domain we get (1.11) p() = d U (0, ). d
If we consider the Doppler spectrum as the pdf in (1.11) and by realizing that the maximum Doppler shift max is caused by = 0 or = , we thus by simple differentiation and insertion nd the average energy from each Doppler shift 1 max max , 2 1 ( max ) max SH () = (1.12) 0 elsewhere, max 4 .
P OWER DELAY PROFILE
The spectrum can be interpreted such that innite many contributions, which are very small, come from 0o and 180o , which creates two singularities in max and The power delay prole (pdp) SH () is generated by a set of delayed paths. For
the tap delay line model, each cluster has an exponential prole. The pdp is typically different depending on the environment. To simplify the model, SH () is approximated by an exponential decaying function [Molisch(2005), page 121], the physical explanation of this phenomenon are still not solved, however this has been veried by extensive measurements campaign [Molisch(2005)]. For the indoor scenario, models explaining have been made [Steinb ock(2008)], where it was shown that this behavior can be explained by reverberation effects in the environment. A simple description of the pdp prole is given as
3 For 4 This
0 1 2 Delay [s] 3 4 5 6 7
30 40 50 Power [dB] 60 10 0 10
Power [dB]
max
D OPPLER FREQUENCY IS = 1 s.
(1.13)
SH () =
exp
0 0 > ,
where is the delay spread. The delay spread is dened as the square root of the second moment of the pdp. The delay-Doppler power spectrum SH (, ) is then dened from the disjoint specters (1.12) and (1.13) as exp( ) 0 and max max 2 max 1( max ) SH (, ) = (1.14) 0 else.
D ELAY-D OPPLER POWER
SPECTRUM
The power delay prole (pdp) is depicted in Fig. 1.4. From the power spectrum SH (, ), we nd the autocorrelation function RH (t , f ). The problem is considered
disjoint, so rst the autocorrelation function RH (t ) is derived from the Doppler spectrum (1.12) by the inverse Fourier transform [Tse and Viswanath(2005)] (1.15) RH (t ) =
Z
d SH () exp( j2t ).
By solving the integral, we get the second moment of the time transfer function
0 5
RH (t )| [dB]
10 15 20 25 30 0.05 0.04 0.03 0.02 0.01 0 0.01 0.02 0.03 0.04 0.05
Fig 1.5: T HE
QUENCY OF PAGE
AUTOCORRELATION FUNCTION
RH (t ) FOR A MAXIMUM D OPPLER FRE DEFINE THE COHERENCE TIME AS THE TIME WHERE [T SE AND V ISWANATH (2005), T HE
COHERENCE BANDWIDTH IS PROPORTIONAL
Time [s]
THE B ESSEL FUNCTION HAS DECREASED FROM 1 TO 0.05 THIS CASE Tc = 3.97 ms . TO THE MAXIMUM D OPPLER SHIFT.
39], IN
(1.16)
RH (t ) = J0 (2t max),
where J0 () is the Bessel function of the zeroth order, given by (1.17) J0 (2t max ) = 1
Z
0
The auto-correlation function is depicted in Fig. 1.5. The discrete function RH [n]
C ORRELATION FUNCTION
OVER
is obtained letting Ns n = t , where Ns is the time between two OFDM symbols, thus (1.18) RH [n] = J0 (2Ns nmax ).
OFDM SYMBOLS
By the Fourier transform of the power delay prole SH (), we nd the time correlation function RH ( f ) of the channel as RH ( f ) = (1.19) 1 exp( j2 f ) 0 1 RH ( f ) = , 1 + j2 f d exp
Z
which has the characteristic of a low-pass lter, see Fig. 1.6. The width of the lowC ORRELATION FUNCTION
OVER SUBCARRIERS
pass lter is proportional to the delay spread . The auto-correlation function for the subcarriers l is thus a discrete function given by (1.20) RH [l ] = 1 , 1 + j2Lsp l
Z
0
where Lsp is the subcarrier spacing. The time/frequency correlation function is (1.21) RH [n, l ] = 1 1 + j2Lsp l d exp ( j2Ns nmax cos ) .
| [RH ( f )] | [dB]
10
15 5
Frequency [MHz]
Thus in the selected channel model we end up with a description, where the channel is characterized by two parameters; the delay spread and the maximum Doppler frequency max , notice that this is not equal to the Doppler spread. For the COST 207 typical urban channel model (which has the considered exponential prole) the delay spread is approx. 1 s (30 dB on 7 s, [Molisch(2005)]). The approximate coherence bandwidth is thus Bc
1 5
Fig. 1.6. For the general case a linear relationship exist between the delay spread and the coherence bandwidth, from [Fleury(1996)] the uncertainty relation between the coherence bandwidth and the delay spread was stated. (1.22) Bc 1 . 2
Over the coherence bandwidth the channel is assumed constant5. Thus across a small number of neighboring subcarriers with a spacing of 15 kHz, the channel frequency response can still be assumed at. However over the whole bandwidth of 1.8 MHz the channel is frequency selective.
5 The
10
11
References
[Bello(1963)] Bello, P. (1963). Characterization of Randomly Time-Variant Linear Channels. IEEE Transactions on Communications, 11(4), 360393. [Dahlman et al.(2007)] Dahlman, E., Parkvall, S., Sk old, J., and Beming, P. (2007). 3G Evolution: HSPA and LTE for Mobile Broadband. Elsevier. [Durgin(2003)] Durgin, G. D. (2003). Space-Time Wireless Channels. Prentice Hall, rst edition. [Fleury(1996)] Fleury, B. (1996). An uncertainty relation for WSS processes and its application to WSSUS systems. IEEE Transactions on Communications, 44(12), 16321634. [Jakes(1974)] Jakes, W., editor (1974). Microwave Mobile Communications. Wiley Interscience. [Molisch(2005)] Molisch, A. (2005). Wireless Communications. Wiley, second edition. [Steinb ock(2008)] Steinb ock, G. (2008). Modeling of the Specular and Diffuse Multipath Components of the Indoor Radio Channel. Master thesis, Aalborg University. [Tse and Viswanath(2005)] Tse, D. and Viswanath, P. (2005). Fundamentals of Wireless Communication. Cambridge university press.