WCDMA-UTRA/FDD METHOD AND RECEIVER SYSTEM
The present invention refers to a channel estimation method and to a receiver for the downlink of wireless communication systems. In the last years, extensive research and standardization activities have been parallely carried out in Europe, Japan and North America on third-generation cellular systems.
Such systems, named UMTS (Universal Mobile Telecommunication Systems), are designed to offer a variety of wideband data services with bit-rate up to 2 Mbit/s. The terrestrial radio access, UTRA (UMTS Terrestrial Radio Access), is based on the WCDMA technology for the frequency bands allocated for services with FDD mode
(frequency-division-duplex).
The major source of impairment in digital wireless communication systems is represented by the phenomenon of multipath propagation, or multipath fading.
Wideband signals, in particular, are subject to a frequency-selective fading. The CDMA technology is based on the direct sequence spread spectrum (DS-SS) modulation technique, according to which a high chip-rate pseudo-random sequence, {c<° }, referred to as spreading sequence, modulates the user information data symbols {α<" } at the transmitter end . The modulated signal bandwidth is typically larger than the channel coherence bandwidth, so that DS-SS systems are generally associated to frequency-selective fading.
In principle, the spreading sequences exhibit an auto-correlation function which is an ideal Kronecker delta function and zero cross-correlation when they are just one chip out of sync. Hence, multipath signal components with relative delays larger than one chip turn out to be uncorrelated and, at the receiver, after despreading, they represent just independently attenuated multiple replicas of the transmitted information signal.
In the known technique, the most widely adopted receiver in the CDMA systems is the RAKE receiver or matched filter receiver, which exploits the inherent time-diversity of the multipath channel. A RAKE receiver separately tracks and conveniently recombines the multipath signal components through a bank of synchronous correlators, each of which decorrelates a different delayed version (of t, ) of the received signal. The output of the single correlator is then properly weighed by a factor , . In order to maximize the signal-to-noise ratio at the receiver output, the weights , are computed according to
the Maximum Ratio Combining criterion: if hi ~,) is the (I -1 )-th channel path amplitude and he ~l) its estimate, then
a , = h ϊ,' ~ l ) ' > l = --,L .
This solution requires initially an accurate estimate of the multipath channel delays and amplitudes to properly set the RAKE fingers to the positions of the strongest multipath components. Usually a circuit is used for each finger for the tracking of the optimum values of {t;} and {a,}.
The most widely used channel estimation method for the design of a RAKE receiver in the WCDMA-UTRA FDD systems is the pilot symbol-assisted method. Such method exploits the availability of a known sequence of pilot symbols that is periodically transmitted time-multiplexed with the information symbols in the user dedicated physical channel. A field of NPiiot bits is placed at the end of each slot. Hence, after the QPSK-coding and the spreading operation, a known sequence, xpf, of Lp= N PJI0, SF/2 QPSK chips is available every slot period. Due to the scrambling operation, that follows the spreading in these systems, the auto-correlation of xp/ is expected to be a Kronecker delta function and the sequence may be therefore used to sound the channel according to the correlation method. The accuracy of the estimate obtained with this method, however, is tightly related to the properties of auto-correlation of the sequence xpf. Unfortunately, in a UTRA FDD system the auto-correlation of xpf. can be considered to be impulsive only for relatively high values of the processing gain (SF>=64). For lower processing gains the autocorrelation does not resemble an ideal Kroneker delta function and the resulting estimate is no more satisfactory, with consequent degradation of the system performance.
The general scope of the present invention is to obviate to the above mentioned drawbacks, by producing a method of channel estimation at the receiver that firstly takes into account the non ideal properties of auto-correlation of xpf and then compensates them. Besides, a further purpose of the present invention is to produce a receiver realized according to such method while exhibiting a reduced complexity.
Toward this end, it has been thought to realize, according to the invention, a method for the estimation of the channel impulse response heq in a WCDMA transmission system.
According to this method, the coefficients of heq are estimated through a Least-Squares method. Besides, it has been realized an innovative architecture of WCDMA receiver with reduced complexity in which, according to the invention, the operation of despreading precedes the receive filter. In such receiver structure, despreading, channel estimation and receiving filtering are implemented through a tapped delay line with delays of Tc/2. Finally, it has been thought to realize, according to the invention, a receiver in a WCDMA-UTRA/FDD transmission system, comprising a reduced matched filter obtained by setting to zero a prearranged number of coefficients of the channel matched filter among those which, according to a Least-Squares estimate, have less significant amplitude.
Besides, the invention concerns a WCDMA-UTRA/FDD transmission system with the receiver realized according to the method described above . In order to clearly explain the innovative principles of the present invention and its advantages in comparison to the known technique, in what follows, a possible exemplifying system realization that applies such principles will be described with the help of the attached sketches. In the sketches: -figure 1 represents a simplified model of the known DS-CDMA transmission system. -figure 2 represents a receiver structure according to the invention;
-figure 3 represents a matched filter receiver as receive filter of figure 2. With reference to the figures, in figure 1 the simplified model of a complete UTRA-FDD transmission system is shown, where the spreading operation by c,(1) is represented within the symbol period [[mTs,(m + l)Ts ~\] by an interpolation filter with impulse response:
g ;](lTc) = c^ , l = 0,l,...., SF - l .
It should be noted the dependence on the m-th symbol period due to the use of long codes, that is, codes with a repetition period that is much larger than the data symbol period.
In the sequel, SF = TS/TC , i.e. the ratio between the symbol period and the chip period, represents the system processing gain, and T0 = Tc/2 .
The cascade of the transmitter pulse shaping filter, the channel filter and the receive front-end filter may be modeled by the equivalent channel filter with impulse response heq . The receive filter impulse response is denoted as gRc .
Finally, the despreading operation may be represented by a filter with impulse response:
where * denotes the complex conjugate value.
The model for the Rayleigh fading channel that is typically adopted in the literature is the wide-sense-stationary-uncorrelated scattering (WSSUS) model , according to which the signal at the channel output is given by the sum of delayed replicas of the transmitted signal weighed by time-varying complex Gaussian processes, which are uncorrelated and have zero mean. Moreover, the channel introduces a noise, that is denoted as w(t) at the output of the receive front-end filter in figure 1. As already mentioned, in such a transmission system the most widely used receiver in the known technique is the RAKE receiver, characterized by the drawbacks outlined above.
According to the invention, a receiver architecture is proposed, that differs from that illustrated in figure 1 for the order in which the receive filter, the despreading and the sampler are implemented. As shown in figure 2, according to the invention, in the receiver, the despreading filter with impulse response
precedes the receive filter gRc , whose output, directly sampled at the symbol rate, provides the decision variable.
This architecture implies an important reduction in computational complexity with respect to the architecture of figure 1 , since the rate of the signal processing in the receive filter is reduced of a factor Ts/Tc = SF .
In order to implement a conventional matched filter receiver, matched to the cascade gsp *heq , being g^ the filter similar to g we choose
Δ gRc nT0 ) = gMF nT0 ) = Kq ((0 ~ "T0 )•
In figure 3, the matched filter is implemented, in an innovative way, through a tapped delay line, with delays of Tc/2 (rather than of Tc as in the known technique), from which the name chip-matched filter (CMF). Each delayed input is weighed by the complex conjugate value of the estimated channel amplitude at that delay. Even though the filter coefficients are not exactly aligned with the strongest paths as they are in a RAKE receiver, the matched filter, contrary to what claimed in the known technique, succeds in collecting the overall multipath signal energy, provided that the signal processing at the receiver is performed at a rate which is at least twice the chip- rate. In fact, if the estimation is performed with period Tc/2 , a perfect reconstruction of the channel impulse response can be achieved.
On the other hand, such structure does not require an accurate estimate of the peaks' delays in the channel impulse response.
According to the invention, in order to further simplify the receiver structure of figure 3, it is possible to set to zero all the coefficients except a prearranged number of them (e.g. 4) with the largest amplitude. Thus, a new receiver is obtained , here called reduced CMF (RCMF) . In order to get an estimate of the channel impulse response at Tc/2 , the Least Squares
(LS) method is used, as will be described below.
On the basis of its intrinsic computational complexity, the LS method does not lend itself to be used in an architecture of the RAKE type, but it has been found that it can be used with great advantage in a receiver architecture of the "matched-filter" type as that described above.
The problem of channel estimation is that of estimating the unknown parameters of the filter {heq} , given the two sets of observable variables: x^(nTc), n = Q,l,...,Lp - l and
r(iτ0). Note that heq has to be estimated at T0 and not at Tc. Toward this end, the poli-phase decomposition of h is applied. Let
h^=heq(2pT0);
N hlP)=h ((2p + l)T0), P = 0,...,- path -1.
the Least Squares (LS) method can be adopted to separately estimate the two impulse responses h
0 {p) and
at T
c. For this purpose, it is considered the filter with coefficients ^
( );p = 0,...,N
path -1] as model for the estimation filter. The number N
palh of coefficients to be estimated is determined by a trade-off between the need to solve the highest number of channel paths and the need to have a noise-free estimate. As known, in fact, for a given length of the pilot sequence, the signal-to-estimation error ratio increases by reducing N
path. You identify the estimation filter's input with x^
( (nT
c), the output with
r'(nTc)= h$x%((n-p)Te) p=0
and the desired response with d(nTc ) = r(nTc +IT0), with = 0 and / = 1 for the estimation
of h0 (p) and J^p) , respectively. The estimation error is defined as
According to the LS method the coefficients
are determined so as to minimize the cost function ε defined as
π=ι.
The limits /, and i
2 depend on the method used for windowing the input data. In our case they are fixed to , = N
path - 1 and i
2 = L
p -l . You note that the output of the estimation filter correspondent to the first N
path - I chip periods of the current slot pilot field, is affected by the last N
path - l values of the chips in the data field of the same slot. Consequently, to perform a LS estimate within a slot, only the last L
p - L
p - (N
palh - 1) chips of the pilot field are considered. Resolving for the coefficient
vector =
of the estimation filter, you have
h = φ-'Θ,
where Φ is the correlation matrix of the inputs and Θ is the vector of the cross- correlations between the inputs and the desired response.
Denoted with
the estimate of h
0 (p) and h[
p) , respectively, the estimate of the equivalent channel results to be
At this point, it is clear how the objectives have been achieved.
A receiver for the WCDMA-UTRA/FDD transmission system, in which a Least Square method is applied for the channel estimation, may compensate the non ideality of the auto-correlation of the employed chip pilot sequence, in the case of low processing gains (SF <64).
Moreover, a receiver structure is proposed, that includes a single despreader, matched to the intended user, followed by a reduced version of the matched filter, the reduced chip matched filter (RCMF), obtained by setting to zero a certain number of coefficients of the filter matched to the channel, between those that according to a Least Squares estimate result to have less significant amplitude. This receiver structure definitely outperforms a conventional matched filter, since in its design the coefficient estimates which are more affected by the noise effect are discarded. Moreover, for lower processing gains (SF <64), the RCMF performs significantly better than a RAKE
receiver with the same number of fingers. This is due the higher accuracy of the LS estimate as compared to the estimate obtained by means of the correlation method. As far as the number of coefficients to estimate is concerned, it has been found advantageous to constrain it to be less than 10 and, specifically, equal to 4. Indeed, by considering only the four most significant values of the LS-estimated channel amplitude, performance even superior to those achievable by a CMF can be obtained. Furthermore, the RCMF here proposed exhibits quite the same performance of the RAKE receiver for high processing gains (SF=128), the number of fingers being equal, and definitely superior performance for low processing gains (SF=16). For lower processing gains, in fact, the advantage offered by the LS estimation method becomes significant.
Obviously, the above description of a realization applying the innovative principles of the present invention is presented by way of example of such innovative principles and it should not be considered as a limitation of what is vindicated here.