where, for the Mi user, a_k is the real positive amplitude due to power control, bi f) is the z^'t complex QPSK data symbol, and s_k is the Walsh code. Here, for n = 0, 1, ..., N-l, sifri) = ±1, and 0 elsewhere. The period of the common complex scrambling code c(n) may extend over an entire frame of symbols. Due to the long PΝ code scrambling, {d ή)}„ is a sequence of uncorrelated complex signal elements, and due to the user-dependent power control, its amplitude distribution is unknown to the receiver. However, from the adaptation point of view, d(ή) represents the signal to be estimated.

The chips are given a transmission waveform p(t) of limited bandwidth. Thus the continuous time model for the multiple access baseband signal is:

where Tc is the chip interval duration.

B. RECEIVED SIGNAL

Due to multipath propagation, the received chip waveform at the wth receiver antenna is:

where γ_mι is the complex gain, and τ_t is the relative delay of the /th path of the multipath channel. Since, compared to the chip rate, the channel parameters are slowly time-varying, they may well be assumed constant over the time interval of interest. Thus the received signal at the th antenna is: r_m ( = ∑ d(n)h_m (l - nT_c )₊ η_m (t)

(4)

where η_m(f) is a process of white Gaussian background noise (AWGN) with two- sided power spectral density No/2. For a matrix representation, the continuous time waveform hJt-nTc) may be discretized into a vector_m(n). The infinite, M- dimensional received signal may be stacked into a vector

r =

(5)^lM

which is given by the matrix equation r = Hd + η_? (6)

where the vector d includes the transmitted multiuser chips

d = [- d{-\) d( ) d{\) ... , (7)

and the matrix H includes the received waveforms

H = [- h(-l) h(0) h(l) ...] , (8)

so that each of the column vectors of matrix H, for n = ..., -1, 0, 1, ...,

represents the received waveform conveying the information of the transmitted multiuser chip d(n). II. ADAPTIVERECEIVER

A. OVERALLRECEIVERSTRUCTURE

A block diagram of the structure of the receiver 10 is depicted in Figure 2. Signals r_t(t) through r_M(t) are received through at least one or a plurality of antennas 140_l3 ..., 140_M which are each coupled to a corresponding chip matched filter 150_l5 ..., 150_M. Each chip matched filter 150_1; ..., 150_M produces at least one sample per chip at its output. The chip sample sequences pass through delays 160_1; ..., 160_M to compensate for possible channel estimation delay.

In order to determine the mutually correlated chip estimates, the coherent RAKE receiver 170 includes a unit, designated Chip MRC 175, for perforating chip maximal ratio combining (MRC) over multipath components and antenna elements. The operation corresponds to channel matched filtering, and as such, minimizes the signal dimensionality with no loss of information prior to equalization.

Reference in this regard may be had to: I. Kaya, A. R. Nix and R. Benjamin, "Exploiting multipath activity using low complexity equalisation techniques for high speed wireless LANs," IEEE International Vehicular Technology Conference, VTC98, Ottawa, Canada, May 1998, pp. 1593-1597.

The mutually correlated chip estimates are adaptively separated by the adaptive separation filter 180. The output of the adaptive separation filter 180 is coupled to a correlator 190 which operates to obtain estimates for the desired users' data symbols by despreading the signal, i.e. by multiplying with a conjugated long scrambling code and user specific Walsh code, supplied by a code generator 200, and then integrating over the symbol period. The output of the correlator 190 is coupled to a deinterleaver 210 which in turn is coupled to a decoder 220. The deinterleaver 210 and the decoder 220 could be conventional in construction and operation.

Assuming perfect knowledge of the channel, the nth output element of the Chip MRC 175 is:

M L x(«) = h" («)r = ∑∑ ; r„ t);?(t - rc7; - r_z t_{? (10}) m=l 1=1

where ( )* denotes taking complex conjugate and ( )^H Hermitian transpose. In matrix form, the total output sequence x = [ ..., -x(-l), (0), x(l), ... ]^τ is x = H T^Λ r = Pd + H n η (ii) where the correlation matrix P = H^■H- Hw is denoted

In practice, instead of known γ_ml , coherent combining employs estimates a_{m /} ≡ γ_n for = 1, ... , M and / = 1, ... , L of the channel response provided by amplitude estimators 230_l5 ..., 230_L. These estimations may be based on dedicated pilot symbols or on a common pilot channel, received through antennas 140_l5 ..., 140_M.

From equation (11), it can readily be seen that by multiplying the vector x from the left with P^"1, chip decorrelation or zero-forcing equalization is performed. However, such an operation offers little advantage due to considerable noise enhancement.

B. ADAPTIVE SEPARATION

Due to multipath propagation, the combined chips are mutually correlated. The objective of the adaptive separation filter 180 is to remove this correlation. Since the correlation matrix P is of Toeplitz form, the separation is reduced into a filtering problem. As depicted in Figure 3, a symmetric filter v is used. Ideally, the filter is infinitely long, but in practice it may be truncated into a manageable length. A suitable filter length may be determined primarily by the channel delay spread. By denoting:

where 2 +1 is the filter length, the output of the separation filter 180 may be expressed as

z(ή) = v H ( )x(n) . (12)

Similarly to the above mentioned bootstrap algorithm described by Y. Bar-Ness and J. B. Punt, the adaptation is based on blind linear decorrelation. The adaptation step of the weights, foτf= 1, 2, ..., F , may be expressed as: v_f (n + l) = v_f (n) -

- f) , (13)

where μ(n) is the (possibly normalized) step-size parameter. Note that in the initial condition, Vy= 0, foτf- 1,2, ..., F, the overall receiver acts as a conventional RAKE receiver.

From equation (13) it can be shown that when the adaptation has reached the steady state (in the mean), the condition

- f))= 0 , f= l, 2, ..., F (14)

is satisfied. When the noise level is insignificant, this condition is also satisfied by a zero-forcing equalizer.

It should be noted that the steady-state filter weights cannot be solved from equation (14) without ambiguity, and that the algorithm does not offer global convergence. However, simulations have shown that the adaptation is stable if the step-size is kept reasonably small.

III. PERFORMANCE EVALUATION

In this section, the results of the computer simulations detennining the bit error rate (BER) performance of the adaptive chip separator, are presented, and compared to those of a conventional RAKE receiver. Consider a WCDMA downlink signal with K active, equal power users, QPSK data modulation and real orthogonal Walsh codes of length N = 4 or 32, along with a long complex Gold sequence for scrambling. A fixed step-size parameter is used for adaptation, and the separation filter length is set to 17 (F = 8). Error correction coding is excluded from the simulations.

The receivers are tested in a time-varying Rayleigh channel with delay spread 1 μs, with three resolvable paths of equal average powers, each independently fading according to a classical Doppler spectrum. A chip rate of 4 MHz with root raised cosine pulse shape filtering, a carrier frequency of 2 GHz, and a vehicle speed of 5 km/h are used in the simulations. Both single-antenna, and two-antenna receivers are simulated. In the two-antenna case, full diversity due to independent fading between the elements is assumed. Perfect estimates of the channel impulse response are given to the receivers. In Figure 4, the BER performance curves of single antenna receivers in the case of spreading factor N=32, with variable number of users, are depicted. The performance of the two-antenna receivers in the corresponding channel load situations are shown in Figure 5. Here, the x-axis indicates the E No per antenna element. The results for the single- and two-antenna receivers in the low spreading factor N=4 situation are given in Figures 6 and 7, correspondingly.

As can be seen, the chip separator offers considerable gain compared to RAKE especially in heavily loaded channels. The separator also seems to benefit more from the extra dimension given by the diversity antenna. In the single user case of low data rate and large spreading factor N=32, however, the detection performance is just slightly decreased due to adaptation jitter. It is worth noting though, that a high data rate user with N=4, clearly benefits from the separation even in the single user case.

The simulations show that the adaptive chip separator outperforms the conventional RAKE receiver, especially at high data rates when the spreading factor is small, or when the system is heavily loaded by multiple users.

It should be understood that the functions described herein may be implemented with discrete circuit elements, or as software routines that are executed by a suitable data processor. A combination of circuit elements and software routines may also be employed.

Thus, while the invention has been particularly shown and described with respect to preferred embodiments thereof, it will be understood by those skilled in the art that changes in form and details may be made therein without departing from the scope and spirit of the invention.

Claims

CLAIMSWhat is claimed is:

1. A receiver for use in a CDMA telecommunications system, comprising:

at least one antenna for receiving signals from a CDMA channel, said signals comprising a desired user signal;

combining circuitry, coupled to said antenna, for perforaiing chip waveform filtering and maximal ratio combining over multipath components and antenna elements to produce mutually correlated chip estimates of said received signals;

an adaptive separator, coupled to said combining circuitry, for adaptively separating said mutually correlated chip estimates; and

a correlator, coupled to said adaptive separator, for despreading an output of said adaptive separator to obtain an estimate for data symbols of said desired user signal.

2. The receiver of claim 1, further comprising estimating circuitry, coupled to said combining circuitry, for estimating a response of said channel, and wherein said combining circuitry utilizes said channel response estimate as a reference.

3. The receiver of claim 1, wherein said despreader further comprises circuitry for multiplying said output of said adaptive separator with a conjugated long scrambling code and a Walsh code, and then to integrate the result over a symbol period.

4. A method for receiving signals in a CDMA telecommunications system, said method comprising the steps of:

receiving signals from a CDMA channel, said signals comprising a desired user signal;

perforaiing chip waveform filtering and maximal ratio combining over multipath components and antenna elements to produce mutually correlated chip estimates of said received signals;

adaptively separating said mutually correlated chip estimates; and despreading said separated mutually correlated chip estimates to obtain an estimate for data symbols of said desired user signal.

5. The method of claim 4, further comprising the step of estimating a response of said channel, and wherein said step of performing chip waveform filtering and maximal ratio combining further comprises utilizing said channel response estimate as a reference.

6. The method of claim 4 wherein said step of despreading further comprises the steps of multiplying said adaptively separated mutually correlated chip estimates with a conjugated long scrambling code and a Walsh code, and then integrating the result over a symbol period.

7. A coherent RAKE receiver for use with received CDMA waveforms comprising a desired user signal, said receiver comprising:

combining circuitry, responsive to said received CDMA waveforms, for perforaiing chip waveform filtering and maximal ratio combining over multipath components and antenna elements to produce mutually correlated chip estimates of said received CDMA waveforms;