FR2864382A1 - CDMA data decoding device for e.g. universal mobile telecommunication system, has interference cancellation units with decoders decoding error corrector code acting on error and estimation signals that are interlaced by interlacer - Google Patents

Abstract

The device has cascaded stages associated to decoding iterations, and having interference cancellation units (50). Each unit (50) has an adder (502) adding residual error and estimation signals to provide a resultant signal. An interlacer (507) interlaces the estimation and error signals. Each unit (50) has a channel decoder (504) for decoding an error corrector code acting on the resultant signal to provide a decoded signal to cancellation unit of a subsequent stage and an error correction signal. Independent claims are also included for the following: (A) a telecommunication system having a decoding device (B) a method of decoding CDMA code.

Description

Device and method for decoding CDMA data,

corresponding system.

  FIELD OF THE INVENTION The present invention relates to the field of telecommunications and more specifically to the decoding of data transmitted using a spread spectrum technique, in particular of the CDMA or Code Division Multiple Access (or CDMA) type. English Code Division Multiple Access).

2. Description of the prior art.

  In a telecommunication system using a CDMA technique, data may be transmitted in parallel by several independent transmitters using spreading codes to one or more receivers located in the same geographical area.

  Thus, Figure 1 illustrates a CDMA transmission system comprising K transmitters 11 to 1K.

  The transmitter 11 (respectively 12 and 1K) comprises: a source of bits generating a sequence of bits b1 (respectively b2 and bk); a channel encoder 110 (respectively 120 and 1KO) adapted to provide sequences of elements coded from the sequence b1 (respectively b2 and bk) equaling +1 or -1; an interleaver 111 (respectively 121 and 1 KI) adapted to interleave the sequence of coded elements provided by the encoder 110 (respectively 120 and 1KO); a spreading unit 112 (respectively 122 and 1K2) multiplying each of the coded and interleaved elements by a normalized spreading code if (respectively s2 and sK) which can vary with each symbol time; an amplifier 113 (respectively 123 and 1K3) amplifying the signal spread by a factor AI (respectively A2 and AK); and RF modulation and RF units (not shown in Figure 1).

  In summary, in a CDMA system as shown in FIG. 1, the transmitter k (with 1 <k <K) transmits a sequence of bits with an amplitude Ak For each transmitter, the bit sequence is coded (coding of channel) and then interlaced before the spreading operation. The spread, amplified and modulated symbols are then transmitted on a channel 13.

  The channel 13 may be modeled by an element 130 which combines all the signals transmitted by the emitters 11 to 1K, followed by a nano-centered Gaussian white noise adding unit 131 and of variance a2, to provide a noisy resulting signal. received by a receiver.

  The source rates of different users may be different. The size of the spreading code is such that the chip rate (a chip being a basic symbol after spreading, in French) is the same for all transmitters 1 to 1K Of course, more complex channels can also be modeled, especially fading channels due to multiple paths. In this case, the signal transmitted by each user propagates through its own transmission channel.

  The received signal r is given by the contribution of the set of K users 11 to 1K and gaussian white noise n.

  The receiver seeks from observation r, to find the information suites bk of each transmitter.

  A conventional detector according to the state of the art, associated with each transmitter, comprises a filter adapted to the spreading sequence sk of the emitter keme followed by the channel decoder corresponding to the encoder of the emitter.

  A more sophisticated solution is to use multidetection techniques (decorrelator, MMSE, MAP) as illustrated in the book of Verdu Multiuser Detection and published by Cambridge University Press, 1998, before using the channel decoder.

  According to another state-of-the-art technique, the decorrelator or MMSE detector is replaced by an iterative structure of the following type: SIC (Successive Interference Cancellation) (described by Rasmussen, Lim and Johansson in the article titled A matrix-algebraic approach to successive interference cancellation in CDMA (or a matrix-algebraic approach to successive interference cancellation in CDMA in French) and published in IEEE Transactions on Communications, 48 (1) pages 145 -151, in January 2000); or - Parallel Interference Cancellation (PIC) (described by Dongning, Rasmussen, and Lim in an article entitled Linear parallel interference cancellation in long-code CDMA multiuser detection (or cancellation of interference, parallel and linear in multi-user detection with long CDMA codes in French) and published in IEEE Journal on Selected Areas In Communications, 17 (12) pages 2074-2081 in December 1999).

  Figure 2 illustrates a CDMA signal receiver according to the state of the art. The receiver 20 comprises: an input 201 accepting a radio signal transmitted by the transmitters 10 to 1K described with reference to FIG. 1; an RF (Radio Frequency) unit 202 transposing the received signal into a baseband and adapting the signal to form an estimated signal r; a unit 203 for despreading the signal r and providing coded bit sequences corresponding to estimates of the output sequences of the coders of the transmitters 10 to 1K; a channel decoding unit 204 corresponding to the coders 10 to 1K and providing decoded data sequences b'1 to b'K, which, in the absence of transmission errors and of decoding errors, respectively correspond to the suites b1 to bK issued.

  According to the state of the art, a SIC structure is implemented in the unit 203, according to several cascaded structures in the form of several stages, each of the structures comprising an interference cancellation unit (or ICU of the English Interference Cancellation Unit) associated with one of the transmitters 10 to 1K and one floor.

  Figure 3 illustrates an ICU unit 30 corresponding to the m '' stage and kth transmitter.

  The ICU 30 accepts as input: an estimate bm-I, k provided by an ICU of the preceding stage m-1 and corresponding to the same emitter k or which is equal to 0 for the first stage; and a residual error signal em, k.

  The ICU 30 outputs: - an estimate bm, k; and a signal Aem k equal to the spread difference of the estimates for the user k provided by the current stage and the preceding stage (bm, k-bm l, k) .sk.

  The ICU 30 comprises: a multiplier 301 multiplying the residual error signal em k by the transpose skT of the spreading code sk (the multiplier thus despreading the signal em k), which amounts to filtering em k by the adapted filter to the sequence sk; an adder 302 adding the signal from the multiplier 301 and the estimate bm-I, k and providing the estimate bm k; and a multiplier 303 multiplying the signal derived from the multiplier 301 by the spreading code sk and supplying, as output, the signal Aem, k.

  The residual error signal em k is obtained by subtracting the residual error signal for the previous emitter em k_1 and the signal Aem, k_I. The first residual error signal e1 'is equal to the received signal r.

  More recently, CDMA turbo-type techniques have been proposed for jointly processing multi-detection and channel decoding, in particular: Varanasi and Guess proposed to decode (hard or hard estimation) and to immediately recode the part of the signal received corresponding to each user before subtracting this contribution from the received signal (in an article entitled Optimum decision feedback multiuser equalization with successive decoding achieves the total capacity of the gaussian multiple-access channel (or multi-user equalization with optimal decision feedback with successive decodings reaching the total capacity of the Gaussian multi-access channel in French) from the Conference Record of the Thirty-First Asilomar Conference on Signals, Systems & Computers, 2: 1405-1409, 2-5 Nov. 1997) report. The same operation is used on the residual signal to decode the information of the second user and so on until the last user.

  - Reed and Alexander proposed to use an adapted filter bank tracking (in parallel) different decoders before subtracting for each user the multiple access interference related to K-1 other users (in an article called Iterative Multiuser Detection using antenna arrays and FEC on Multipath channels "(or" Multiple-User Detection Using Antenna Matrices and Error-correcting Code in Multipath Channels "in French) from the IEEE magazine on selected areas in communications, 17 (12), pages 2082-2089, December 1999) - Wang and Poor proposed a multi-user detector consisting of parallel implementation of the MMSE filters associated with each user and tracking of the corresponding channel decoders (in an article entitled Iterative (Turbo) Soft Interference Cancellation and Decoding for Coded CDMA "(or" Interference cancellation, soft iterative (turbo) and decoding in ur of the CDMA coded "in French) published in the magazine IEEE Transactions on Communication, pages 1046-1061 in July 1999). These two elements iteratively exchange extrinsic information according to a decoder structure close to that illustrated with reference to FIG. 2, the output of the channel decoder banks being looped back to the input of the MMSE filterbanks; - Tarable, Montorsi and Benedetto proposed a simplification of the method presented by Wang and Poor (in an article entitled "A linear Front End for Iterative Soft Interference Cancellation and Decoding in Coded CDMA" (or "a front element for a soft cancellation and iterative interference and for decoding in an encoded CDMA system ", presented at the ICC conference, International Conference on Communications, June 2001) For the first iterations, a multi-user detector of the MMSE type is used followed by the decoders In the last iterations, the MMSE filter is replaced by a bank of suitable filters.

  The various techniques of the state of the art have the disadvantage of being relatively complex to implement and / or to provide poor performance (case of the standard receiver which consists in putting suitable filters followed by decoder banks) or not optimal (receivers do not always converge to the performance associated with a single-user system).

  Another disadvantage of these methods of the state of the art is that they require knowledge of the matrix of intercorrelations of the spreading codes (KxK dimension matrix), which leads to complex calculations, especially when the codes of spreading (which is the case in communications according to UMTS (Universal Mobile Telecommunication System).

  Certain methods illustrated above (in particular the method proposed by Wang and Poor) require, in addition, complex calculations of matrix inversion type.

  3. Presentation of the invention The invention in its various aspects is intended to overcome these disadvantages of the prior art.

  More specifically, an object of the invention is to provide a CDMA decoder and a corresponding method, relatively simple to implement.

  The invention also aims to provide a CDMA decoding technique particularly well suited to multi-user reception and having good performance.

  For this purpose, the invention proposes a device for decoding at least one signal spread by a spreading code representative of source data coded by an error correction code, the device comprising at least one stage associated with an iteration of decoding, each of the stages comprising at least one interference cancellation unit, each of the interference canceling units accepting, as input, a first estimation signal and a first error signal and providing, as output, a second estimation signal for an interference canceling unit of a next stage; and - a first error correction signal which, in combination with the first error signal, forms a second error signal supplying an interference canceling unit of the same stage or an interference canceling unit of the same. next floor; the device being remarkable in that each of the interference cancellation units comprises: means for adding a signal representative of the first error signal and the first estimation signal, providing a first resultant signal; error correction code decoding means acting on a signal representative of the first resulting signal to provide a first decoded signal, the second estimation signal being representative of the first decoded signal.

  Thus, each of the interference cancellation units comprises error correction code decoding means, which are therefore implemented at each iteration.

  Preferably but not exclusively, the interference canceling units are arranged in the same stage and the second error signal supplies the next interference canceling unit of the same stage or, if the cancellation unit of Current interference is the last of the considered stage, the first interference canceling unit of the next stage.

  The second error signal obtained by combining the first error correction signal with the first error signal is preferably calculated by removing the first error correction signal from the first error signal.

  Moreover, the invention is compatible with an implementation of the device which comprises several cascaded stages, each of the stages corresponding to a single iteration or which comprises a reduced number of stages, two stages that can be reused for several iterations associated with a plurality of stages. decoding the same data.

  According to one particular characteristic, the device is remarkable in that, in each of the stages, each of the interference cancellation units is adapted to reduce or cancel the interference associated with one of the spread signals.

  Thus, within the same stage, the cancellation units are preferably cascaded, each of the units being associated with a spread signal therefore corresponds to a transmitter transmitting data according to its or its own spreading codes.

  According to one particular characteristic, the device is remarkable in that each of the interference cancellation units comprises despreading means of the first error signal producing a first despread error signal, the addition means adding a representative signal. the first despread error signal and the first estimate signal.

  Thus, the implementation of the device is relatively easy, especially in the case of a Gaussian channel.

  According to one particular characteristic, the device is remarkable in that the interference cancellation unit comprises: means for estimating the transmission channel of the source data; and adapted filtering means taking into account the spreading code and the channel estimation, and applied to the first error signal to produce a first filtered error signal, the adding means adding a signal representative of the first filtered error signal and the first estimation signal, the adding means adding a signal representative of the first filtered error signal and the first estimation signal.

  Thus, the invention also makes it possible to process complex channels, for example, with multipaths, while providing good performance.

  In this way, the invention is particularly well suited to mobile telecommunication networks, in particular of the UMTS or HSDPA type.

  According to one particular characteristic, the device is remarkable in that it comprises normalization means of the first resultant signal for forming the signal representative of the first resulting signal supplying the error correction code decoding means.

  Thus, the invention makes it possible to exploit the properties of a turbo type decoding, by standardizing, for example, the data as a function of the variance of the Gaussian noise present on the transmission channel and the amplitude of the signal received.

  According to one particular characteristic, the device is remarkable in that the interference cancellation unit comprises: subtraction means of the first estimation signal to the second estimation signal, the subtraction means providing a second resulting signal ; and means for spreading the signal representative of the resulting second signal to form a spread-out result signal.

  Thus, the cancellation unit makes it possible to determine the second resulting signal, the latter being able to be used to estimate the error affecting a signal corresponding to a transmitter.

  According to one particular characteristic, the device is remarkable in that the spreading means form the first error correction signal.

  In this way, the determination of the first error correction signal is particularly simple to carry out, the error correction thus obtained being sufficiently reliable for transmission on a Gaussian channel and / or with little interference between the signals. issued by different users.

  According to one particular characteristic, the device is remarkable in that it comprises means for filtering the resulting signal spread to form the first error correction signal.

  In this way, the error correction can take into account the characteristics of the transmission channel, the filtering being preferentially adapted to the latter.

  According to one particular characteristic, the device is remarkable in that it comprises means for interleaving and / or deinterleaving data.

  Thus, the device makes it possible to interlace and / or deinterlace in particular the data corresponding to the estimation or error signals in order to optimize the performances associated with a turbodecoding type structure.

  The invention also relates to a system comprising: means for transmitting at least one signal spread by a spreading code representative of source data coded by an error correction code; means for receiving the spread signal and the decoding device adapted to decode the spread signal, according to the invention.

  In addition, the invention relates to a method for decoding at least one signal spread by a spreading code representative of source data coded by an error correction code, the method comprising at least one decoding iteration, each of the iterations. implementing at least one interference cancellation accepting, as input, a first estimation signal and a first error signal and outputting: a second estimation signal for a cancellation unit of a next iteration; and a first error correction signal which, when combined with the first error signal, forms a second error signal feeding an interference cancellation in the same iteration or an interference canceling of the next iteration; the method being remarkable in that each of the iterations comprises: - a step of adding a signal representative of the first error signal and the first estimation signal, providing a first resultant signal; an error correction code decoding step acting on a signal representative of the first resulting signal to provide a first decoded signal, the second estimation signal being representative of the first decoded signal.

  4. Detailed description of the invention.

  Other features and advantages of the invention will emerge more clearly on reading the following description of a preferred embodiment, given as a simple illustrative and nonlimiting example, and the appended drawings, among which: FIG. 1 presents a multi-transmitter CDMA transmission system, known per se; FIG. 2 illustrates a CDMA receiver, known per se, capable of decoding signals transmitted by several users of the system of FIG. 1; FIG. 3 describes an interference cancellation unit (or ICU) of the English Interference Cancellation Unit implemented in the CDMA receiver of FIG. 2; FIG. 4 illustrates a CDMA receiver according to the invention according to a particular embodiment and capable of decoding signals transmitted by several users of the system of FIG. 1; Figure 5 illustrates an interference cancellation unit and implemented in the receiver of Figure 4; Figures 6 and 7 show decoding results of a CDMA receiver implementing devices according to Figures 4 and 5; and FIG. 8 illustrates a CDMA receiver, according to a variant in accordance with the invention, capable of decoding signals transmitted by several users of the system of FIG. 1, where each of the transmitted signals propagates through its own transmission channel, and including Rake means.

  The general principle of the invention is based on the introduction of channel decoding and transmission channel estimation means in a structure of cancellation of successive interference or SIC (while, according to the state of the art , the channel decoding operations and transmission channel estimates are made, if necessary, outside a SIC structure). More precisely, according to the invention, a SIC structure uses interference cancellation units (or ICUs) comprising means for decoding a series of data representative of a symbol estimate to be decoded corrected by a signal. error, estimated data being generated at the output of the decoding means.

  FIG. 4 shows an embodiment of a SIC structure 40 according to a particular embodiment of the invention, comprising M stages 41 to 4M cascaded. The SIC structure is adapted to decode a signal emitted by the system presented with reference to FIG.

  Each of the floors 41 to 4M itself comprises K interference cancellation units ICUkm, a unit noted ICUkm corresponding to the k'th unit of the mth floor of the SIC structure 40.

  The unit ICUkm accepts as input: a soft estimate (or soft English) bm_1 k provided by the ICUkm 1 of the previous stage m-1 and corresponding to the same transmitter k or which is equal to 0 for the first stage ( bo k = 0); and a residual error signal em, kE The ICUkm unit outputs: a soft estimate bm k; and a signal Aeyn k equal to the spread difference of the estimates for the user k provided by the current stage and the previous stage (Aem k = (bm k bm I, k) .Sk).

  The first residual error signal el, l is equal to the received signal r. The residual error signal em k is obtained by subtracting the residual error signal for the previous transmitter em, k-1 and the signal Aem, k-1 for k between 2 and K (2 s k s K). The residual error signal em 1 (for m strictly greater than 1) corresponding to the input of the m '' stage of the structure 40 is obtained by subtraction of the residual error signal for the previous transmitter em_1 K and the signal Aem_1 K provided by the previous stage in structure 40.

  FIG. 5 illustrates a unit 50 corresponding to one of the ICUkm units of the structure 40, the latter having a similar structure.

  The residual error signal ern k is despread in a correlator 500 multiplying the signal em k by the spreading code skT corresponding to the signal emitted by the transmitter 1k (it is a suitable filter).

  The despread signal is then deinterlaced by a deinterleaver 501 11k 1, inverse of the interleaver Hk used in the corresponding transmitter 1k. The interleavers 11k and 17k- 'are arbitrary and are, for example, of the pseudo-random type or as defined by the UMTS or HSPDA (High Speed Downlink Packet Access) standards.

  The soft estimate bm-1, k corresponding to the same transmitter 1k calculated at the previous iteration (or initialized to 0 for m equal to 1) is then added to the residual error signal despread and deinterlaced in an adder 502.

  The signal ym k thus obtained is then multiplied, in a multiplier 503, by a coefficient equal to twice an amplitude Ak divided by the variance of the Gaussian noise Vm k (representing the residual multiple access interference plus the additive noise) .

  Here, the signal Ym, k is written as ym, k Ak.bk + Vm, k or Vm, k represents residual multiple access interference plus additive noise. The values of Ak and of var (vm k) are calculated in a unit 509 accepting in particular input ym k and operating in the following manner.

  The noise Vm k is approximated by a centered Gaussian random variable whose variance is given by: = 1 A2p2k (1 bm,.) + E A2p2k (1-bm 1, i) + Q2 i <k i> k

T

  where pi k = Sk ESi represents the cross-correlation between two sequences sk and si.

  We show that the extrinsic information of bk at iteration m is given by: P ym, k bk = +1] \ gym'k.Ak Âm tbk 1 = log rr = PLym k bk = -1 varf l vm This extrinsic information serves as input to a DCk channel decoder 504 corresponding to the coder 1k0 implemented in the transmitter 1k.

  At the first iteration corresponding to m equal to 1, only the pilot symbols are used to estimate Ak and var (vmk).

  var (v \ m, k) For the following iterations (m> 1), this estimation jointly uses the pilot symbols emitted by the corresponding transmitter 1k with the data and the soft values (or soft values in English) of the estimated data, bm -1, k, at iteration (m-1) for the same transmitter 1k.

  The maximum likelihood criterion is then used, which leads to the determination of the amplitude of the signal received from the transmitter 1k at the iteration m according to the relation: T 1. 2 y (l) - (i) im , k m-1, k ^ = 0 Am, k = T 1 2 b (i) i = p m-1, k where T is the size of the frame to be decoded and bm) -1, k for i ranging from 0 to T-1 are the symbols (drivers and data) of the frame transmitted by the transmitter 1k, which is to be decoded.

  For the data, b (i) _1, k corresponds to the output of the channel decoder (before amplification) at the previous m-1 iteration.

  y (m, k for i ranging from 0 to T-1 corresponds to the symbols observed at the iteration m for the transmitter 1k.

  The variance var (vm k) is estimated according to the relation: var (v) = 1 T 1 y (i) bel bel) 2 m, k T m m, km, km -1, k) = 0 The signal? , m (bk) thus normalized (Am (bk) = ym, kE2Am, k / var (vm k)) is then decoded by the channel decoder D Ck 504 which provides the decimal logarithm of the posterior likelihood ratio (conditionally to all observation) of the set of bits (for both the information bits and the parity bits): LLR (bk = Log This report is then transformed into a soft estimate (or soft) of binary elements bm k: bm, k E bk ym, k = tanh (0, 5LLR bk ym, k A multiplier 505 then makes it possible to calculate the soft estimate of bk at the iteration m according to the relation: bm, k - 4m The ICU 50 also includes a subtractor 506 which makes the difference of the soft estimates bm, k-bm-I, kE This difference is interleaved by a interleaver 507 17k.The interlaced difference is itself spread by a multip linker 508 which multiplies this difference by the code sk, the result being equal to Aem k.

  The result Deys k thus obtained is subtracted from the residual signal em k to obtain the new residual signal em, k + 1 corresponding to the next transmitter (if k <K) or to obtain the new residual signal em + 1.1 corresponding to the first emitter at the next iteration (em, K + 1 = em + 1,1) FIGS. 6 and 7 show the decoding results of the receiver 40 implementing the unit 50. More specifically, FIGS. 6 and 7 respectively present the bit error rate 61 (TEB or BER of the English Bit Error Rate) and the frame error rate 71 (TET or FER of the English Frame Error Rate) as a function of the signal-to-noise ratio 60 ( Eb / No) expressed in dB with a number of transmitters K equal to 31, a spreading factor also equal to 31 (with Gold codes) and frame sizes of 640 bits.

  The curves 62 and 72 respectively illustrate the BER and TET when a conventional detector known per se (that is to say a detector with filtering adapted to the spreading sequence followed by the channel decoder as illustrated in FIG. ) is used.

  The curves 63, 64 and 65 show the BER after decoding by the receiver 40 after respectively one, two and three iterations. It can be seen that, as of the second iteration, the BER is very close (less than 0.1 dB) to the result obtained with a single transmitter (no interference between transmitted signals) illustrated by the curve 66. The gain relative to a detector conventional is of the order of 0.5 dB to obtain a BER of 10-2 and is even more important for lower BER.

  Likewise, the curves 73, 74 and 75 show the TET after decoding by the receiver 40 after respectively one, two and three iterations. It can be seen that, as soon as the second iteration, the TET is very close (less than 0.1 dB) to the result obtained with a single transmitter (no interference between transmitted signals) illustrated by the curve 76. The gain with respect to a conventional detector is of the order of 0.4dB to obtain a TEB equal to 0.5.10 "2. This gain is all the more important that the TET is low.

  In the case where the propagation channel of the transmitter 1k has a multipath impulse response of the form: Lk fl ck (t) = ckl.lt t tk1) l = 1 where Lk is the number of paths of the channel, ck l and -cm are respectively the complex gain and the delay of the leme path of the signal emitted by the transmitter 1k (this is the case of UMTS, for example, in the uplink direction, terminal to base station), ICUkm is generalized to the structure 80 illustrated with reference to FIG. 8. The structure 80 includes an estimate ck m (t) of the transmission channel of the transmitter 1k at the iteration m. Structure 80 applies to many multichannel channel transmission contexts and, in particular, to the uplink context of the UMTS-FDD system.

  To change from the ICUkm 50 to the ICUkm 80, the despreading function is replaced by a Rake filter (filter adapted to the spreading sequence sk convoluted with the transfer function in the ICUkm unit) and the function of spreading by the spreading function convolved with ck (t). The ICUkm 80 therefore has elements in common with the ICUkm 50 illustrated with reference to FIG. 5. These common elements being similar to those of the ICUkm 50, the latter bear the same references and are not described further.

  The soft estimate bm-I, k is first interleaved by an interleaver 11k 807 used in the corresponding transmitter 1k (and similar to the interleaver 507).

  The interleaved soft estimate brn_I k is then multiplied by a spreading code sk corresponding to the signal emitted by the emitter 1k in a multiplier 808 to form an interlaced and spread soft estimate.

  The bm-1, k interleaved and spread data are filtered by a channel filter 801 estimated at the previous stage (impulse response given by Ck m_I (t)). For m equal to 1, Ck o (t) is zero for any value of k between 1 and K (1 s k s K). The channel filter Ck m (t) is a linear filter whose impulse response is Ck m (t) which is an estimate at the iteration m, of the impulse response of the transmission channel of the transmitter 1k.

  The bm_1 k data interleaved, spread and then filtered are then added to the residual error signal em k in an adder 806 to form a signal rm, k.

  The signal rm k and the data ((m) 1, k) O i T are used to update the estimator of the impulse response of the transmitter 1k denoted Ck, m (t) in a unit 802. .

  The impulse response Ck m (t) is obtained by minimizing the mean squared error: T-1 2 i * (i) min rn k - Ck, m bm -1, k i0 The minimum is obtained by direct matrix inversion or iteratively. The quantities (b- (Mi) - 1, k) 0 s i s T 1 represent the soft decisions of the previous iteration or, for the pilot symbols, their exact values.

  A Rake filter 803 whose coefficients are provided by the unit 802 is applied to the residual error signal eYY1 k. The output of the filter 803 is normalized by a factor equal to: Ck ((m) 2 1 = 1, l where Lk is the number of paths of the channel corresponding to the transmitter 1k and where Ck lm represents the complex gain of the 1st path the signal emitted by the transmitter 1k and estimated at the iteration m.

  The output of the Rake filter is then deinterlaced by a deinterleaver 501 IIk 1, inverse of the interleaver IIk before being added to the soft estimate bm-1, k in an adder 502 making it possible to obtain a signal ym, kE As previously, the signal ym k is written as ym k = Ak.bk + vyn k and the values of Ak and the variance of vynk (var (vmk)) are calculated in a unit 509.

  The signal ym k is then multiplied, in a multiplier 503, by a coefficient equal to twice an amplitude Ak divided by the variance of the Gaussian noise Vm, k The signal?, M (bk) thus normalized (Xm (bk) = 2 klvar (vm k) is then decoded by a decoder DCk channel 504 to obtain an estimate bm k before being multiplied by Âm k in a multiplier 505 for the determination of the soft estimate bm k at iteration m.

  A subtractor 506 then calculates the difference of the soft estimates bm, k-bm.l, kE This difference is interleaved by an interleaver 507 IIk and the interleaved difference is itself spread by a multiplier 508 which multiplies this difference by the code sk.

  The output of the multiplier 508 is filtered by the estimated channel of Ck, m (t) in a filter 805 which provides a result equal to Aem. Of course, the invention is not limited to the embodiments mentioned above.

  In particular, the skilled person can make any variant in the structure of interference cancellation units (cascaded units or iterative structure with re-use of the same means during a next iteration).

  In addition, the channel coding can be arbitrary (convolutional code, en bloc, ...). Likewise, the interleaver (direct Ilk.or inverse Ilk-1) are also arbitrary (for example, random interleaver or conform to UMTS or HSDPA standards).

  Note that the invention is not limited to the CDMA decoder but extends to any system implementing CDMA communications, including mobile radio communication systems (eg UMTS or HSDPA).

  Similarly, the invention is not limited to BPSK (Binary Phase Shift Keying or Modulation by Binary Phase Displacement in French) but extends to any PSK (Phase Shift Keying) modulation. ) (including QPSK (Quaterly PSK) used in, for example, UMTS) or QAM (Quadrature Amplitude Modulation) (including 16-QAM) used in HSDPA systems).

  In addition, the invention is not limited to the criterion of the minimization of the mean squared error but also extends to any other criterion for estimating the impulse response of the transmission channel.

  It will be noted that the invention is not limited to a purely material implantation but that it can also be implemented in the form of a sequence of instructions of a computer program or any form mixing a material part and a part software. In the case where the invention is partially or totally implemented in software form, the corresponding instruction sequence can be stored in a removable storage means (such as for example a floppy disk, a CD-ROM or a DVD-ROM) or no, this storage means being partially or completely readable by a computer or a microprocessor.

Claims (11)

  1. Device for decoding at least one signal spread by a spreading code representative of source data coded by an error correction code, said device comprising at least one stage (41, 42, 4M) associated with an iteration of decoding, each of said stages comprising at least one interference cancellation unit (50), each of said interference canceling units accepting, as input, a first estimation signal and a first error signal and providing, at the output,: a second estimation signal intended for an interference canceling unit of a following stage; and a first error correction signal which, in combination with the first error signal, forms a second error signal supplying an interference canceling unit of the same stage or an interference cancellation unit of the same next floor; characterized in that each of said interference cancellation units comprises: means (502) for adding a signal representative of said first error signal and said first estimation signal, providing a first resultant signal; error correction code decoding means (504) acting on a signal representative of said first resultant signal to provide a first decoded signal, said second estimation signal being representative of said first decoded signal.   2. Device according to claim 1, characterized in that, in each of said stages, each of said interference canceling units is adapted to reduce or cancel the interference associated with one of said spread signals.   3. Device according to any one of claims 1 and 2, characterized in that each of said interference cancellation units comprises despreading means (500) of said first error signal producing a first despread error signal, said adding means (502) adding a signal representative of said first despread error signal and said first estimate signal.   4. Device according to any one of claims 1 and 2, characterized in that said interference cancellation unit comprises: - means for estimating the transmission channel of said source data; and - adapted filtering means taking into account said spreading code and said channel estimate, and applied to said first error signal to produce a first filtered error signal, said adding means adding a signal representative of said first filtered error signal and said first estimation signal, said adding means adding a signal representative of said first filtered error signal and said first estimation signal.   5. Device according to any one of claims 1 to 4, characterized in that it comprises normalization means (503, 509) of said first resultant signal to form said signal representative of said first resultant signal supplying said code decoding means error corrector.   6. Device according to any one of claims 1 to 5, characterized in that said interference cancellation unit comprises: subtraction means (506) of said first estimation signal to said second estimation signal, said means subtraction providing a second resulting signal; and spreading means (508) of the signal representative of said second resultant signal to form a spreading result signal.   7. Device according to claim 6, characterized in that said spreading means form said first error correction signal.   8. Device according to claim 6, characterized in that it comprises means for filtering said resulting signal spread to form said first error correction signal.   9. Device according to any one of claims 1 to 8, characterized in that it comprises means for interleaving (507) and / or deinterleaving (501) data.   10. System comprising: means for transmitting at least one signal spread by a spreading code representative of source data coded by an error correction code; means for receiving (40) said spread signal and said decoding device according to any one of claims 1 to 9 adapted to decode said spread signal.   11. A method of decoding at least one signal spread by a spreading code representative of source data coded by an error correction code, said method comprising at least one decoding iteration, each of said iterations using at least one interference canceling accepting, at the input, a first estimation signal and a first error signal and outputting: a second estimation signal for a cancellation unit of a next iteration; and - a first error correction signal which, when combined with the first error signal, forms a second error signal feeding an interference cancellation in the same iteration or an interference canceling of the next iteration; characterized in that each of said iterations comprises: a step of adding a signal representative of said first error signal and said first estimation signal, providing a first resultant signal; an error correction code decoding step acting on a signal representative of said first resultant signal to provide a first decoded signal, said second estimation signal being representative of said first decoded signal.