Method of receiving CDMA signals with parallel interference suppression, and corresponding stage and receiver

Abstract

A method of parallel suppression of interference, and corresponding stage and receiver is disclosed according to the invention, parallel suppression of interference is carried out starting from the signals selected by a maximum likelihood criterion based on the calculation of a metric and the search for the smallest possible metric.

Description

The invention uses a Euclidean distance metric, afterwards referred to as the metric, of the form
(Σ({right arrow over (X)}−{right arrow over (Y)}))²
where {right arrow over (X)} and {right arrow over (Y)} represent two vectors. Such a metric measures, in a way, the distance between the two extreme points of the vectors. The smaller the metric is, the closer the vectors are.

The following four metrics, corresponding to the four formulated hypotheses, will therefore be calculated:
(M₁=(|Z_o(1)|−|Z_i(1)|)²+|Z_o(2)|−|Z_i(2)|))²
(M₂=(|Z_o(1)|−|Z_i(1)|)²+|Z_o(2)|+|Z_i(2)|))²
(M₃=(|Z_o(1)|+|Z_i(1)|)²+|Z_o(2)|−|Z_i(2)|))²
(M₄=(|Z_o(1)|+|Z_i(1)|)²+|Z_o(2)|+|Z_i(2)|))²

The smallest of these metrics corresponds to the configuration closest to the configuration at the output from the input stage and hence to the most likely configuration. If, for example, the smallest metric is the third one M₃, the most likely configuration will be: $\overrightarrow{Z_i}=\begin{array}{c}-\overrightarrow{Z_i}\left(1\right)\\ +\overrightarrow{Z_i}\left(2\right)\end{array}$

The means M will then supply the signals −|Z_i(1)| and +|Z_i(2) | and the two reconstitution circuits which follow it will spread these signals using the two appropriate pseudo-random sequences. The traditional means of parallel interference suppression will then receive the spread signals of maximum likelihood and will then be able to correct these signals in an optimum way.

In a general way, the means M for a stage of row i calculates the quantity $\sum _{\mathrm{block}}{\left(\overrightarrow{Z_o}-{\left(\overrightarrow{Z_i}\right)}_j\right)}^2$
where the summation is extended at least to the values that constitute the block of data within a time interval equal to N symbol durations.

When N =1, there are only K components to be processed (the case referred to as a single symbol block) and the number of hypotheses to be formulated is 2^K. With NK components, this number rises to 2^NK. To prevent too much complexity, N is limited to a few units, for example, less than 5.

FIG. 3 illustrates a complete receiver that comprises an input stage and an output stage, as for FIG. 1, with three parallel interference suppression steps with references V₁, V₂, V₃ conforming to what has just been described. The receiver further comprises the associated matched filters, F₁₁, . . . , F_1k, . . . , F_1K for the first, F₂₁, . . . , F_2k, . . . , F_2K for the second and F₃₁, . . . , F_3k, . . . , F_3K for the third.

In order to illustrate the variations in value taken by the metric as a function of the hypotheses made on the signs, we may consider the case of three users, each using pseudo-random sequences each with 63 elements or chips, the modulation employed being differential type modulation with quaternary phase modulation (DQPSK) with two channels per user, namely one channel in phase (called I) and one channel in phase quadrature (called Q). There are therefore 6 channels in parallel, or 2⁶=32 possible hypotheses on the signs of a single symbol block. These 32 hypotheses or configurations are labeled by their row in the diagram in FIG. 4, the row being shown on the x-axis, and the values taken by the metric being shown on the y-axis. Four different cases are shown corresponding to the four curves 51, 52, 53 and 54. The value of the metric is expressed in elements or chips. The scale is logarithmic. It can be clearly seen that for a certain configuration, the metric passes through a minimum. This configuration is that of maximum likelihood. It may also be observed that the minima are clearly evident and can therefore be easily exploited.

The method and the receiver that have just been described assume, for the totally general case, that 2^NK hypotheses are formulated. The complexity of the method can naturally be reduced by reducing the block of data with K data (a single symbol block mentioned above). However this complexity can be further reduced, in the method of seeking the maximum likelihood, by only taking into account those signals for which the estimation is judged to have little reliability or to put it another way by excluding from the method those signals judged to be reliable. Assuming that Q signals are reliable, only K−Q signals will be retained for the calculation of the metrics, which corresponds to 2^K−Q hypotheses.

Means of measuring reliability are described and claimed in French patent application No. 98 09782 filed by the present applicant on the Jul. 30^th 1998.

However other criteria of reliability can be used, such as those which are described in patent U.S. Pat. No. 5,644,592.

FIG. 5 illustrates an embodiment of a stage simplified in this way. Compared with the stage in FIG. 2, stage V′ comprises reliability testing means T₁, . . . , T_k, . . . , T_K which receive the signals coming from the estimation circuits ES₁, . . . , ES_k, . . . , ES_K and which address either the circuit M for calculation of the metrics (branch marked NO) or the reconstruction circuits R₁, . . . , R_k, . . . , R_K (branch marked YES).

Naturally, several of these simplified stages can be cascaded, as for FIG. 3.

In another particular embodiment, the signals supplied by the matched filters can be linearly combined before they are addressed to the output stage. One can see in FIG. 6, the first weighting means P₀₁, . . . , P_0k, . . . , P_0K arranged at the output from the matched filters F₀₁, . . . , F_0k, . . . , F_0K of the input stage, weighting circuits P₁₁, . . . , P_1k, . . . , P_1K arranged at the output from the matched filters placed behind stage V′1 for parallel interference suppression and adders AD₁, . . . , AD_k, . . . , AD_K the inputs of which are connected to the weighting circuits and the output to the decision circuits D₁, . . . , D_k, . . . , D_K.

The weighting coefficients can be fixed or variable. Such a technique is described in U.S. Pat. No. 5,553,062.

One can also improve the reconstructions and estimations of the signals by using the reliability thresholds in order to reconstruct or not to reconstruct (or to only partially reconstruct) certain signals. Such a technique is described and claimed in the French patent application No. 98 03586 filed on the Mar. 24^th 1998 by the present applicant. A technique of this kind is also described in the patent U.S. Pat. No. 5,644,592. FIG. 7 illustrates this particular embodiment in the case of two simplified, (that is to say conforming to FIG. 5), parallel, interference suppression stages V′₁ and V′₂. One can see the first weighting circuits P₀₁, . . . , P_0k, . . . , P_0K, the second weighting circuits P₁₁, . . . , P_1k, . . . , P_1K and finally the third weighting circuits P₂₁, . . . , P_2k, . . . , P_2K.

The performance of a receiver according to the invention has been simulated by the applicant. To do this, certain hypotheses have been formulated for the pulse response of the propagation channel. Firstly, one can consider an ideal pulse response which would be formed by a single peak, which would correspond to an absence of multiple paths. However, one can also choose a more realistic hypothesis, illustrated in FIG. 8, where one can see a first amplitude peak 1 and three amplitude peaks respectively equal to 0.25, 0.12 and to 0.06 representing three secondary paths. The results of the simulation are shown in FIGS. 9 and 10 for these two hypotheses. In these Figures, the bit error rate is shown on the y-axis and the signal to noise ratio on the x-axis. The following reference numbers have been used for the curves.

• • 61, 71 corresponding to a traditional structure
  • 62, 72 corresponding to parallel interference suppression with one stage
  • 63, 73 corresponding to parallel interference suppression with two stages
  • 64, 74 corresponding to parallel interference suppression according to the invention (simplified version in the case of K=5 users)
  • 65, 75 corresponding to a theoretical parallel interference suppression for a single user in DQPSK modulation with a Gaussian channel.

It can be seen that the invention leads to a significant improvement in performance. In particular, a single stage of parallel interference suppression (in the simplified version) offers better performance than the traditional two stages.

Claims

1. A method of receiving cdma signals with parallel interference suppression in which:

a composite signal (r(t)) is received comprising a plurality of k signals corresponding to information symbols which have been spread in frequency by k different pseudo-random sequences,

these k signals are correlated using said k sequences,

the corresponding k symbols are estimated,

the k signals correlated in frequency are reconstructed by despreading re-spreading said estimated symbols through using the corresponding pseudo-random sequences,

the contributions of the other remaining (K−1) signals are subtracted from a despread spread signal to provide k new signals, spread in frequency but cleared, at least in part of the interference,

2. A parallel interference suppression stage implementing the method according to claim 1, this stage comprising:

k inputs receiving signals correlated in frequency,

k means of for estimating (ES₁, . . . , ES_k, . . . , ES_k) k symbols corresponding to these k signals,

k means of for reconstructing (R₁, . . . , R_k, . . . , R_k) signals respread in frequency using the corresponding pseudo-random sequences,

means of for parallel interference suppression comprising k channels in parallel capable of subtracting from one despread spread signal, the contributions of the other remaining (K−1) despread respread signals,

k outputs supplying k signals spread in frequency, cleared, at least in part of the interference,

3. Stage according to claim 2, in which the means of calculating the metric placed between the estimating means and the reconstruction means comprise:

means of for formulating two hypotheses on the sign to be assigned to the amplitude of the signals supplied by the means of for estimation,

means of for calculating all the differences Z₀(k)−Z_i(k)_j, where Z₀(k) represents the signal at the output from the k^th matched filter of the input stage and Z_i(k)_j the signal at the output from the k^th matched filter of the stage of row i, the signal being allocated the sign corresponding to each hypothesis j,

means of for calculating the square of these differences, or (Z₀(k)−Z_i(k)_j)², means of for calculating the sum of these squares for all NK values of the signals, which leads, for each hypothesis (j), to the metric (M_j).

4. A receiver of cdma signals that implements the method of claim 1 and comprises:

a general input (E) suitable for receiving a composite signal (r(t)) formed from a plurality of k signals corresponding to information symbols which have been spread in frequency by k different pseudo-random sequences,

an input stage with k channels in parallel each comprising filters (F₀₁, . . . , F_0k, . . . , F_0K) to correlate in frequency the composite signal (r(t)) through one of the k pseudo-random sequences, this stage supplying k signals correlated in frequency,

at least one parallel interference suppression stage (V₁, V₂, . . . ,),

filter stages positioned between the parallel interference suppression stages and comprising k filters (F_1k, F_2k, . . . ) matched to the pseudo-random sequences,

an output circuit (S) comprising k decision circuits (D₁, . . . , D_k, . . . , D_k)

5. A receiver according to claim 4, in which the estimated signals have a certain reliability and in which the means (M) for formulating two hypotheses on the sign to be assigned to these signals means placed between the estimating means and the reconstruction means only take into account the signals with a reliability below a fixed threshold, the other remaining signals having a reliability above the threshold being used directly by the interference suppression means for parallel interference suppression.

6. A receiver according to claim 4, in which the input stage with its filters and each stage of matched filtering comprise means of weighting circuits (P_0k) to weight the outputs from the filters (F_0k), the output stage (S) comprising adders (AD_k) the inputs to which are connected to the weighting circuits (P_0k) and the output from which is connected to the decision circuits (D_k).

7. A receiver according to claim 4, in which each stage of filtering is followed by a weighting circuit (P_0k, P_1k, P_2k, . . . ) arranged between the output from the filtering stage and the input to the interference suppression stage, the weighting depending on the reliability of the estimation made in the stage.

8. An apparatus, comprising:

a receiver to receive a composite signal (r(t)) comprising a plurality of k signals corresponding to information symbols which have been spread in frequency by k different pseudo-random sequences,

said receiver further to correlate said k signals using said k sequences,

an estimation means to estimate the corresponding k symbols,

a reconstruction means to reconstruct the k signals correlated in frequency by re-spreading said estimated symbols using the corresponding pseudo-random sequences,

wherein the contributions of the remaining (K−1) signals are to be subtracted from a spread signal to provide k new signals, spread in frequency but cleared, at least in part of the interference, wherein:

all the possible hypotheses on the signs of the NK correlated signals are to be formulated, where N is a whole number greater than zero,

for each hypothesis, the distance metric is to be calculated between the group of correlated signals undergoing processing and the corresponding signals before processing,

the hypothesis for which the metric is smallest is to be retained, being the hypothesis which has a maximum likelihood,

only those signals corresponding to this maximum likelihood hypothesis are to be reconstructed.

9. An apparatus, comprising:

at least one spread-spectrum parallel interference suppression processing stage to receive a received signal comprising k spread-spectrum signals spread with k codes, to despread and demodulate the k spread-spectrum signals in parallel to obtain k extracted symbols, and, for each of the k extracted symbols, to generate a new spread-spectrum signal using the corresponding one of the k codes, and further to obtain an estimate of each of the k spread-spectrum signals by subtracting a quantity based on the remaining K−1 new spread-spectrum signals from a previous input signal;

wherein said processing stage includes:

at least one estimator to provide amplitude estimates corresponding to the k extracted symbols; and

at least one metric calculator to calculate at least one metric based on the amplitude estimates and to choose a set of k maximum likelihood symbols based on the at least one metric, the set of k maximum likelihood symbols to be used as the k extracted symbols to generate the new spread-spectrum signals.

10. The apparatus according to claim 9, further comprising:

a set of k parallel despreaders to receive the received signal and to provide k parallel inputs to a first one of said at least one processing stage.

11. The apparatus according to claim 9, wherein said at least one metric comprises a Euclidean distance metric.

12. The apparatus according to claim 9, further comprising:

at least one reliability testing device to test of each of the k extracted symbols to determine that a number, Q, of the k extracted symbols are reliable, wherein said metric calculator is to calculate at least one metric only on the k-Q extracted symbols not determined to be reliable.

13. The apparatus according to claim 9, wherein said processing stage further comprises:

at least one weighting circuit to weight at least one of the extracted symbols and a corresponding at least one of the set of k maximum likelihood symbols; and

at least one adder to add the resulting corresponding weighted symbols to obtain at least one of the k extracted symbols to be used to generate the new spread-spectrum signals.

14. A method, comprising:

processing a received composite code-division multiple-access (cdma) signal, comprising k spread-spectrum signals spread using k codes, said processing comprising despreading and demodulating the k spread-spectrum signals in parallel to obtain k extracted symbols, and, for each of the k extracted symbols, generating a new spread-spectrum signal using the corresponding one of the k codes, and said processing further comprising obtaining an estimate of each of the k spread-spectrum signals by subtracting a quantity based on the remaining K−1 new spread-spectrum signals from a previous input signal;

wherein said processing includes:

providing amplitude estimates corresponding to the k extracted symbols; and

calculating at least one metric based on the amplitude estimates and choosing a set of k maximum likelihood symbols based on the at least one metric, the set of k maximum likelihood symbols to be used as the k extracted symbols to generate the new spread-spectrum signals.

15. The method according to claim 14, wherein said at least one metric comprise a Euclidean distance metric.

16. The method according to claim 14, wherein said method comprises:

performing said processing at least twice, each processing an initial processing utilizing results generated by the previous processing.

17. The method according to claim 14, wherein said method further comprises:

determining a reliability of each of the k extracted symbols to determine that a number, Q, of the k extracted symbols are reliable, and calculating at least one metric only on the k-Q extracted symbols not determined to be reliable.

18. The method according to claim 14, wherein said processing further comprises:

weighting at least one of the extracted symbols and a corresponding at least one of the set of k maximum likelihood symbols and adding the resulting weighted symbols to obtain at least one of the k extracted symbols to be used to generate the new spread-spectrum signals.