FR2833116A1 - Wideband code division multiple access cellular radio telephone system having symbol set separator with parallel coding/scrambling coded sequences and transmitting parallel scrambled coded symbols. - Google Patents

Abstract

The cellular radio telephone transmitter (1) transmits symbol sets multiplexed in time intervals assigned to the user. There is a mechanism (11) separating each symbol set in an assembly of parallel symbols (do,k to di-1,k) and parallel units (12,o to 12,i-1) coding parallel relative symbols to the user using coding elements (Co,k to Ci-1,k) to the user. There are parallel means (13o0 to 13i-1) scrambling coded symbols with scrambled sequences, and parallel transmitters (14,0 to 15,0; 14,i-1 to 15,i-1) transmitting the parallel coded symbols.

Description

<Desc / Clms Page number 1>

CDMA transceiver for radiotelephone network
The present invention relates to radiocommunication links between transmitters and receivers in a Wide Band-Coded Multiple Division Access (W-CDMA) wideband and wideband cellular radiotelephony system, and more particularly in a mobile radio system. Universal Mobile Telecommunications System (UMTS) type in FDD (Frequency Division Duplex) mode.

 It is known that UMTS radio-mobile networks are based on a direct sequence spectrum spread technique by multiplying time-division multiplexed symbol sequences and respectively assigned to users by octagonal code element sequences. respectively allocated to users. Thus, for a given frequency, a base station successively transmits symbol by symbol the coded symbol sequences then scrambled at a bit rate of a few kbit / s.

 This serial broadcast of the symbols limits the bit rate for radio data links with the internet network or for videoconferencing.

 The aim of the invention is to increase the transmission rates in radiocommunication links, in particular downlink links, between base stations and mobile terminals in a W-CDMA type radiotelephone system.

<Desc / Clms Page number 2>

 For this purpose, a transmitter for a cellular radiotelephone system transmitting symbol sequences respectively multiplexed in time slots assigned to users, the symbols being encoded by sequences of code elements and scrambled by scrambling sequences, is characterized in that it comprises means for separating each series of symbols into sets of parallel symbols, I parallel means for coding the parallel symbols relating to a user respectively by I sequences of code elements associated with the user, I parallel means for scrambling I coded symbols relating to a user respectively by the scrambling sequences associated with the user, and I parallel means for simultaneously transmitting I coded and scrambled parallel symbols relating to a user, I being an integer greater than 1.

 Thanks to simultaneous transmissions of parallel coded and scrambled symbols, the invention increases the transmission rate in radiocommunications by a factor equal to the integer I for each user.

 The invention also relates to a receiver relating to a user for a cellular radiotelephone system having a single antenna receiving means for receiving a symbol signal transmitted by the transmitter according to the invention. The receiver is characterized in that it comprises I parallel filtering means respectively adapted to sequences of code elements associated with a user, I scrambling sequences associated with the user and I propagation channels between the I means parallel to transmit in the transmitter and the receiving means to deliver I parallel filtered symbols, and I

<Desc / Clms Page number 3>

 parallel means of equalization and decision to cancel the interference between the I parallel filtered symbols in order to restore the corresponding symbols relating to the user, I being an integer greater than 1.

 The receiver according to the invention uses only a conventional antenna, which does not modify the apparent structure of the mobile radio terminals.

 As will be seen hereinafter, the receiver of the invention seeks to eliminate the interference between the symbols received simultaneously due to the increase in bit rate achieved in the transmitter, by using cancellation algorithms. interfering not between symbols relating to several users, but between parallel symbols received simultaneously, relating to a user.

 According to a first preferred embodiment of the iterative interference cancellation type, each equalization and decision means comprises intersymbol interference calculation means for estimating the interference between 1-1 parallel filtered symbol samples other than the respective filtered symbol sample, means for subtracting from the respective filtered symbol the estimated inter-symbol interference to a corrected symbol, and threshold decision means for restoring the respective symbol according to the corrected symbol. Alternatively, a respective filtered symbol is subtracted from the estimated interference, and a second filtering means filtering the corrected symbol is added.

 According to a second embodiment of the multi-symbol detection type, the equalization and decision means for canceling the interference between the parallel filtered symbols apply a linear equalization algorithm by forcing to zero.

<Desc / Clms Page number 4>

 or according to the minimum mean squared error criterion on the parallel filtered symbols, or the equalization and decision means have a reverse decision equalization structure.

 Other features and advantages of the present invention will emerge more clearly on reading the following description of several preferred embodiments of the invention with reference to the corresponding appended drawings in which: FIG. 1 is a schematic block diagram of a CDMA transmitter according to the invention; FIG. 2 is a schematic block diagram of an interference canceling receiver according to the invention; FIG. 3 is a schematic block diagram of a variant of the interference cancellation receiver; FIG. 4 is a schematic block diagram of a multi-symbol detection receiver with linear transverse equalization according to the invention; and FIG. 5 is a schematic block diagram of a set of filter banks and circuit circuits. decision for transverse equalization with return of decisions in a multi-symbol detection receiver according to the invention.

 The embodiment described hereinafter in detail of a transmitter and receiver system according to the invention is presented in the context of a UMTS type of radio telephone system although it can be transposed to any W system. -CDMA. For the sake of simplicity, the processed signals will be

<Desc / Clms Page number 5>

 considered baseband upstream of filtering means, amplification and frequency transposition in the transmitter and downstream means corresponding to the latter in the receiver.

 The information symbols processed by the transmitter and the receiver for a UMTS type downlink from the transmitter in a base station to the receiver in a cellular mobile radiotelephone can be included as well in a DCH transport channel (Dedicated CHannel) dedicated to a single user, only to shared channels between several users for packet services exchanged with the Internet, such as the Forward Access Channel (FACH) and the Downlink Shared CHannel (DSCH) unidirectional transport channels. for the downlink. The transmitter and the receiver may be those of a radiotelephone and a base station in an uplink, in particular relative to the Common Physical Channel (CPCH).

 With reference to FIG. 1, a transmitter 1 according to the invention processes complex useful information symbols which are transmitted in series in particular by an amplification stage and distributed in temporally multiplexed time slots in a radio frame . A time interval contains a salvo (burst) assigned to a respective user and containing a determined number of complex symbols. For example, a time interval is composed of two data fields of identical length enclosing a training sequence (midamble) to estimate the propagation channel between the transmitter and the receiver. Typically, a time slot assigned to a

<Desc / Clms Page number 6>

 The user designated by k contains at least about 160 symbols.

phase 140 à 14,-, et un étage d'émission 150 à 15I-1. As shown in FIG. 1, the transmitter 1 essentially comprises, in connection with the invention, a series-parallel converter 11 and I parallel channels respectively connected to the outputs of the converter and each containing in cascade a respective spreading circuit 120 to 12 a respective scrambler 130 to 13, a modulator of

phase 140 to 14, -, and an emission stage 150 to 15I-1.

The number I of parallel channels in the transmitter is equal to the useful rate multiplication factor that the invention confers relative to a user.

d. f.. l {d k d k d k) d ki ème d'information utile {d,... d,... d) du k usager, avec 1 < k < K = 16 et avec l'entier supérieur strictement à 1 et inférieur au nombre prédéterminé de symboles par intervalle de temps (timeslot). En pratique, l'entier I est un sousmultiple du nombre de prédéterminé de symbole par intervalle de temps ; par exemple, pour un nombre de prédéterminé de symboles complexes égal à 16, le nombre de voie I peut être égal à 2, ou 4 ou 8.

k Il est à noter que chaque symbole complexe d 1 de durée de symbole Ts appliqué au convertisseur 11 traité jusqu'à l'entrée du modulateur de phase

respectif 14i est composé d'une partie réelle ai et d'une partie imaginaire Pi correspondant aux deux éléments binaires des voies en quadrature de phase I et Q d'un modulateur de phase à quatre états ; en d'autres termes, l'entrée du convertisseur 11 ainsi que chacune des I voies parallèles en sortie du convertisseur doivent être considérées comme des It is assumed that at the input of the serial parallel converter 11 are applied I complex symbols

df. l {dkdkdk) d ki th of useful information {d, ... d, ... d) of the k user, with 1 <k <K = 16 and with the integer greater than 1 and less than the predetermined number symbols per timelot. In practice, the integer I is a submultiple of the predetermined number of symbols per time slot; for example, for a predetermined number of complex symbols equal to 16, the number of lane I may be equal to 2, or 4 or 8.

It should be noted that each complex symbol symbol symbol symbol Ts 1 applied to the converter 11 processed up to the input of the phase modulator

respective 14i is composed of a real part ai and an imaginary part Pi corresponding to the two bits of the quadrature channels of phase I and Q of a four-state phase modulator; in other words, the input of the converter 11 as well as each of the I parallel paths at the output of the converter must be considered as

<Desc / Clms Page number 7>

paires de voies parallèles pour les parties réelles et imaginaires d'un symbole d'information d. ième Dans la i voie de diversité, avec 0 i I-

1, le symbole d'information dit subit une modulation 1 d'étalement de spectre dans le circuit d'étalement 12i en le multipliant par une séquence d'étalement

respective Ci (channelisation code) comprenant L i éléments de code prédéterminés (chips) ayant une période d'éléments binaires Tc selon la relation Ts = L. Tc. En accord avec la norme UMTS, les I séquences d'étalement Co à CI-1 attribuées selon l'invention au

kième 10 do,. b' k usager, au lieu d'une unique séquence attribuée à cet usager selon la technique antérieure, respectent également le caractère d'orthogonalité entre les séquences selon une structure arborescente des séquences d'étalement similaire à celle utilisée pour l'interface radio UMTS.

pairs of parallel paths for real and imaginary parts of an information symbol d. In the diversity path, with 0 i I-

1, the information symbol said undergoes a spread spectrum modulation 1 in the spreading circuit 12i by multiplying it by a spreading sequence

respective Ci (channelization code) comprising L i predetermined code elements (chips) having a bit period Tc according to the relationship Ts = L. Tc. In accordance with the UMTS standard, the CO-1 to CO-1 spreading sequences assigned according to the invention to

kth 10 do ,. b 'k user, instead of a single sequence assigned to this user according to the prior art, also respect the character of orthogonality between the sequences according to a tree structure of the spreading sequences similar to that used for the UMTS radio interface .

complexe S. dans le brouilleur 13i. i Ainsi dans l'émetteur 1, 1 séquences d'éléments k k k de code c à Cj et I séquences de brouillage S à Sk 0 Ob l 0 d S sont respectivement attribuées aux I voies de diversité parallèles. Each sequence of Q bits (chips) coming out of the spreading circuit 12k is multiplied element by element by a scrambling sequence

S. complex in the jammer 13i. Thus, in the transmitter 1, 1 sequences of elements kkk of code c to Cj and I scrambling sequences S to Sk 0 Ob 10 0 d S are respectively assigned to the I parallel diversity channels.

k k UMTS, les séquences d'éléments de code CQ à Cj 0 1-1 associées à l'usager k dans l'émetteur 1 sont identiques pour les I voies de diversité à une unique séquence d'éléments de code C associée à l'usager k et différente des autres séquences d'éléments de code However, so as not to modify the distribution of the code element sequences per radio cell covered by a database station for a given frequency, which are identical from one cell to another according to the standard

kk UMTS, the sequences of code elements CQ to Cj 0 1-1 associated with the user k in the transmitter 1 are identical for the I diversity channels to a single sequence of C code elements associated with the user k and different from other code element sequences

<Desc / Clms Page number 8>

respectivement associées aux autres usagers. En k k revanche, les séquences de brouillage SQ à Sj sont 0 1- différentes entre elles. Ces caractéristiques n'affectent pas avantageusement la capacité du réseau initial UMTS, puisque sans les séquences de brouillage, il faudrait une séquence d'étalement différente par voie de diversité ; puisque le nombre d'usagers potentiels est proportionnel au nombre de séquences d'étalement, le nombre d'usagers potentiels diminuerait avec le nombre de voies de diversité.

respectively associated with other users. In contrast, the SQ scrambling sequences at Sj are 0-different from each other. These characteristics do not advantageously affect the capacity of the initial UMTS network, since without the scrambling sequences, it would be necessary to have a different spreading sequence by way of diversity; since the number of potential users is proportional to the number of spreading sequences, the number of potential users would decrease with the number of diversity paths.

This is not the case thanks to a set of different scrambling sequences that can be used simultaneously in two cells sufficiently distant to avoid interference.

 The digital signal coming out of each jammer 13i is composed of complex baseband bits distributed over two parallel paths respectively corresponding to the real part and the imaginary part of the complex elements which are simultaneously applied to the inputs of the modulation modulator 14i. classical QPSK phase (Quadrature Phase Shift Keying) corresponding to four phase states such that {l, j, -1, -j} so that a mixer output of the modulator applies each complex element to the transmission stage 15j. The latter filters, amplifies and transposes the modulated signal to be transmitted by an antenna to a receiver according to the invention.

 In a variant, in order to reduce the number of antennas, the outputs of the diversity channels are combined in pairs so that two transmission stages have in common a transmitting antenna having crossed polarizations for transmitting two coded and scrambled parallel symbols. The antenna may be a patch antenna. So the outputs

<Desc / Clms Page number 9>

 stages 150 and 151 are connected to the terminals of a cross-polarized antenna law, ... the outputs of the stages 15i and 15i + 1 are connected to a cross-polarized antenna 16i, i + 1 '... and the outputs stages 161-2 and 161-1 are connected to the terminals of a cross-polarized antenna 16y 2 1 -1. The emitter 1 then contains I / 2 cross-polarized antennas respectively for I diversity channels per cell.

mb l l dk l. f'kième symboles complexes d. relatif a un k usager avec 1 des interférences intersymboles dues à la diversité de sources des antennes de l'émetteur 1. Pour annuler l'interférence entre les symboles qui a été créée, des algorithmes de détection multi-usager sont appliquées aux séquences d'étalement reçues par le récepteur. In the different receiver embodiments according to the invention there is provided a single conventional reception antenna AR connected to the input of a frequency conversion, filtering and amplification stage AND which produces a received signal r (t) in baseband sampled at the rate Tc = Ts / Q of the complex bits (chips) corresponding to the four phase states according to the QPSK phase modulation. For a given transmission frequency in a cell, the receiving antenna AR receives the chips with interference between the diversity channels of the transmitter, and therefore

mb ll dk l. f'th complex symbols d. relating to a user k with intersymbol interference due to the diversity of sources of the transmitter's antennas 1. To cancel the interference between the symbols that has been created, multi-user detection algorithms are applied to the sequences of spreading received by the receiver.

 According to a first embodiment shown in FIG. 2, the receiver 2 is based on a cancellation of interference between the symbols generated by a parallelization of the information. All symbols are detected in parallel, as if they were

<Desc / Clms Page number 10>

 only then the interference in each of the parallel I-channels resulting from the parasitic detection of the symbols of the other 1-1 diversity channels is iteratively eliminated.

appliqué aux entrées de 1 filtres adaptés 200 à 20y dz Chaque filtre adapté 20i est adapté à la séquence k d'étalement respective ci, à la séquence de 1 brouillage respective Si et au canal de propagation ,, 1 de la i voie de diversité dont la fonction de transfert estimée sur quelques trajets entre la source 1-15i et le récepteur AR-2 est h. et présente i ainsi la réponse suivante : k k k ^k Pi (t) = ci (t). si (t) * h (t). The received baseband signal r (t) is

applied to the inputs of 1 adapted filters 200 to 20y dz Each matched filter 20i is adapted to the respective spreading sequence k, to the respective scrambling sequence Si and to the propagation channel, 1 of the diversity channel of which the estimated transfer function on a few paths between the source 1-15i and the receiver AR-2 is h. and thus presents the following answer: kkk ^ k Pi (t) = ci (t). if (t) * h (t).

k n Le symbole filtré y,' produit à l'instant nTs 1 par l'échantillonneur 21i est évalué par comparaison à des seuils dépendant des valeurs complexes, telles

k n que {l, j,-1,-j), en un symbole provisoire 8f, n dans un circuit de décision provisoire 22i. Un circuit de calcul d'interférence entre symboles 23i estime l'interférence entre le symbole provisoire

? k, n i. ième.--... --- ô. de la i voie de diversité avec les symboles 1 provisoires 8 à 8 et ≈à 8 dans les 1-1 autres voies, produits par les autres circuits de décision provisoire 220 à 23i-1 et 22i+l à 22I-1. At the output of the adapted filters, samplers 210 to 211-1 produce 1 filtered symbols parallel successively at times nTs punctuated at the symbol period Ts, n being an integer index increasing from a reference time.

kn The filtered symbol y, 'produced at time nTs 1 by the sampler 21i is evaluated by comparison with thresholds depending on the complex values, such as

kn that {l, j, -1, -j), into a provisional symbol 8f, n in a provisional decision circuit 22i. An intersymbol interference calculation circuit 23i estimates the interference between the provisional symbol

? k, n i. th ... - ... --- ô the diversity channel with provisional symbols 8 to 8 and 8 in the other 1-1 channels, produced by the other provisional decision circuits 220 to 23i-1 and 22i + 1 to 22I-1.

k n 23i est retranchée du symbole yf, n sortant du filtre i adapté 20i à travers l'échantillonneur 21i dans un The estimated interference coming out of the computing circuit

kn 23i is subtracted from the symbol yf, n coming out of the filter i adapted 20i through the sampler 21i in a

<Desc / Clms Page number 11>

soustracteur 24i afin d'annuler l'interférence entre k n le symbole yet les autres symboles parallèles. i ième L'interférence entre symboles IESI de la i voie en sortie du circuit de calcul 23i pour l'usager k à l'instant nTs s'exprime par la relation suivante :

1-1 k n n+1) TS k *k IESI =EJ pt-nTs). p (t-nTs). dt. jwi j=O

k n Le symbole corrigé yc k, n dénué de l'interférence 1 IESi entre symboles, sortant du soustracteur 24i, est comparé aux quatre valeurs complexes telles que {l, j,-1,-j} dans un circuit de décision 25i afin d'appliquer la valeur du symbole complexe restitué

à k, n à l'une des I entrés d'un convertisseur 1 parallèle-série 26 qui restitue une suite de symboles {d,... d.,... Ti} correspondant à la suite 0 de symboles appliqués à l'entrée du convertisseur série-parallèle 11 dans l'émetteur 1.

subtractor 24i to cancel the interference between kn the symbol yet the other parallel symbols. The intersymbol interference IESI of the i output channel of the computing circuit 23i for the user k at the instant nTs is expressed by the following relation:

1-1 kn n + 1) TS k * k IESI = EJ pt-nTs). p (t-nTs). dt. jwi j = O

The corrected symbol yc k, n devoid of inter-IESi inter-symbol interference, exiting subtractor 24i, is compared to the four complex values such as {l, j, -1, -j} in a decision circuit 25i in order to apply the value of the returned complex symbol

at k, n to one of the I inputs of a parallel-to-serial converter 26 which reproduces a sequence of symbols {d,...,...,... Ti} corresponding to the sequence 0 of symbols applied to the input of the serial-parallel converter 11 into the transmitter 1.

symboles et de soustraction dans les circuits 22i' 23i et 24i est itératif plusieurs fois, en pratique 2 à 3 fois pour chaque période de symbole Ts. Comme montré en traits pointillés à la figure 2, l'itération est réalisée en connectant la sortie du soustracteur 24i à l'entrée du circuit de décision provisoire 22 à la place de la sortie de l'échantillonneur 21i pendant deux à trois périodes successives grâce à l'adjonction de mémoires tampons à la sortie de l'échantillonneur 21i, ou en traitant plus rapidement les 2 ou 3 itérations successives entre les circuits 22i, 23i et 24i pendant une période de symbole Ts. Ces itérations successives par soustractions successives de l'interférence entre Alternatively, the provisional decision-making process for calculating the interference between

symbols and subtraction in the circuits 22i '23i and 24i is iterative several times, in practice 2 to 3 times for each symbol period Ts. As shown in dashed lines in FIG. 2, the iteration is carried out by connecting the output of the subtractor 24i to the input of the provisional decision circuit 22 instead of the output of the sampler 21i for two to three successive periods thanks to adding buffers at the output of the sampler 21i, or processing more rapidly the 2 or 3 successive iterations between the circuits 22i, 23i and 24i during a symbol period Ts. These successive iterations by successive subtractions of the interference between

<Desc / Clms Page number 12>

 IESi symbols recalculated 2 to 3 times in the 23i circuit refines the interference cancellation.

 According to another variant of the first embodiment, the receiver 2a shown in FIG. 3 comprises in cascade in each of the diversity channels, a suitable filter 200 to 201-1, a sampler 210 to 211-1 and a provisional decision circuit 220. to 221-1, and an interference calculation circuit 230 to 241-1 connected to the negative input of a subtractor 240 to 241-1 arranged in the same manner as in the receiver 2.

1-1 k n k IESi = l 8.'p. (t-nTs). However, in the receiver 2a, the positive input i of the subtractor 24i in the diversity channel is not connected to the output of the sampler 21i, but is connected directly to the output of the stage ET which receives the signal received in baseband r (t). The interference IESi between symbols is thus directly deduced from the received signal r (t) in the subtractor 24i and is expressed by the following relation for the user k at the instant nTs:

1-1 knk IESi = l 8.'p. (Does nTs).

L'annulation des interférences entre symboles peut être également itérative, comme montré en traits pointillés à la figure 2. J = oj I

The cancellation of intersymbol interference can also be iterative, as shown in dashed lines in FIG.

chaque voie de diversité un filtre adapté 27i de k* fonction de transfert pf* (-t), comme le filtre 20i, i entre la sortie du soustracteur d'annulation d'interférence 24i et le circuit de décision à seuils 16il6 25i dont la sortie est reliée à la i entrée du convertisseur parallèle-série 26. The receiver 2a then comprises in addition to

each diversity channel has an adapted filter 27i of k * transfer function pf * (-t), such as the filter 20i, i between the output of the interference canceling subtractor 24i and the threshold decision circuit 16il6 25i whose output is connected to the input of the parallel-to-serial converter 26.

<Desc / Clms Page number 13>

 According to a second embodiment shown in FIG. 4, the multi-symbol detection receiver 3 detects all the symbols and cancels the intersymbol interference by matrix calculations in the parallel diversity channels at the output of the stage AND. The receiver 3 needs to introduce a guard interval W. Tc of a few code element periods, with W an integer greater than 1, typically also about 10, in order to cancel the interference between information symbols in the same diversity path introduced during transmission. Thus in the transmitter 1, it suffices to replace the period Ts by a period T such that: T = Ts + W. Tc, no information being transmitted during the guard interval W. Tc.

base r (t) à la séquence d'éléments de code respective k ci de l'usager k, à la séquence de brouillage S de l l l'usager k ainsi qu'à la fonction de transfert fI estimée sur quelques trajets du canal de propagation pour la ile voie de diversité. La réponse du filtre adapté 30i est encore la suivante :

p (t) = c (t). sk (t) k (t)
Les signaux filtrés discrets aux sorties du banc de filtres adaptés 300 à 30r-1 sont échantillonnés à la période T respectivement dans des échantillonneurs 311 à 31r-1 pour fournir un vecteur de symboles

k k, n k, n k, n T Yn ={Y(/,...y,... yi} où T dénote une transposition. As in the receivers 2 and 2a, each diversity channel comprises a matched filter 30i which correlates the code elements in each symbol period Ts of the received symbol signal.

base r (t) to the respective code element sequence k ci of the user k, to the scrambling sequence S of the user k as well as to the transfer function fI estimated over a few paths of the data channel. spread for the island way of diversity. The response of the adapted filter 30i is still the following:

p (t) = c (t). sk (t) k (t)
The discrete filtered signals at the outputs of the matched filterbank 300 to 30r-1 are sampled at the period T respectively in samplers 311 to 31r-1 to provide a symbol vector.

kk, nk, nk, n T Yn = {Y (/, ... y, ... yi} where T denotes a transposition.

<Desc / Clms Page number 14>

circuits de décision à seuils 330 à 33I-1 dans les I voies de diversité afin de produire des symboles âk à âk estimés do à dl-l qui sont sérialisés dans un convertisseur parallèle-série 34. The receiver 3 further comprises equalizers 320 to 32r-1 respectively followed by

threshold decision circuits 330 to 33I-1 in the diversity channels to produce estimated symbols kk to kk which are serialized in a parallel-to-serial converter 34.

 The introduction of the guard interval W. Tc eliminates the interference between the symbols in each of the diversity channels, which reduces the function of the equalizers 320 to 32I to a detection of I symbols in parallel and thus to an evaluation of the interference between I parallel symbols by matrix computations. Each equalizer 33i thus processes, at each period T, I parallel symbols, and not I symbols successively received during I periods T for a given user, according to the prior art, for example according to the article by SEITE and TARDIVEL, "Adaptive Aqualizers for Joint Detection in an Indoor CDMA Channel, "25-28 July 1995, IEEE, 45th Vehicular Technology Conference, Chicago, p. 484-488.

filtres 300 à 30r afin que chaque coefficient dans chaque égaliseur 32i annule un terme respectif de l'interférence des I symboles parallèles relativement k k

au symbole y.. Le vecteur de symboles Y est 1 multiplié par l'inverse d'une matrice d'intercorrélation rk telle que : According to a first variant of the second embodiment, the I equalizers 33i apply a linear transverse equalization algorithm by forcing to zero on the I filtered symbols emerging from the

filters 300 to 30r so that each coefficient in each equalizer 32i cancels a respective term of the interference of I parallel symbols relatively kk

to the symbol y. The vector of symbols Y is 1 multiplied by the inverse of an intercorrelation matrix rk such that:

<Desc / Clms Page number 15>

If/Tf 1f\/tr \ P... (PP)... (P) / \ If/ \ k k) Ilp 112 (pk j pk r'k (PO Pi 1... I-i pk, = (P )... P"... (P. P) ....... "......

If / Tf 1f \ / tr \ P ... (PP) ... (P) / \ If / \ kk) Ilp 112 (pk j pk r'k (PO Pi1 ... Ii pk, = (P ) ... P "... (P. P) ......." ......

et (pi, pj) = JL pi (t) p (t) dt, avec os (i, j) SI-1. (ppi-i) - (pi'pi-i) - i

and (pi, pj) = JL p (t) p (t) dt, with bone (i, j) SI-1.

^ k The vector of symbols {d.} produced by the 1 equalizers 320 to 321-1 through four-threshold decision circuits 330 to 331-1 such that {l, j, - 1, -j} is deduced from the following relation: {d} = sgn (ry), ikn where sgn is the sign operator.

 According to a second variant of the second embodiment, the I equalizers 320 to 321-1 apply a linear transverse equalization algorithm according to the minimum mean squared error criterion MMSE (Minimum Mean Squared Error) on the I filtered symbols leaving the filters 300 at 30j1.

h h. k, n l, ième d ff" d recherché. Soit tiP le j ieme des I coefficients de 1, J l'égaliseur 33i de l'usager k à l'instant nt, T la , d ff" k, n. ième 1. d k9 matrice carrée de coefficient t-. La l ligne de 1, J la matrice Tu continent à l'instant ne les n coefficients de l'égaliseur 32i qui produit le k n symbole sortant constituant un élément du 1 vecteur de symboles sortants : For this second variant, a compromise between the suppression of the interference between the I parallel symbols and the reduction of the power of a Gaussian noise at the output of each equalizer 32i is

h h. k, nl, th th dd fd "d searched, Let tiP be the j th of the I coefficients of 1, J the equalizer 33i of the user k at the moment nt, T la, d ff" k, n. th 1. d k9 square matrix of coefficient t-. The line of 1, J the matrix Tu continent at the moment does n the coefficients of the equalizer 32i which produces the kn outward symbol constituting an element of the 1 vector of outgoing symbols:

<Desc / Clms Page number 16>

k k, n k, n k, n. T n'LZQ''-'" [---/'"i-l ' k, n k, n-l k, n.-k-k k avec z = t yj'et Z = TY J=0 J=O

Les coefficients de l'égaliseur 32i sont choisis de façon à minimiser l'erreur quadratique moyenne : E (lz-d ').

kk, nk, nk, n. N'L '''--- --- --- --- --- --- --- avec avec avec avec avec avec avec avec avec avec avec avec avec avec avec avec avec avec avec avec avec avec avec avec

The coefficients of the equalizer 32i are chosen so as to minimize the mean squared error: E (lz-d ').

1 1.

k k, opt k, opt k, opt T-1 bk Topt = [t i, o t i, i.... t 1-1 1 k i k* k k k, n k* avec (D = E (Yn Yn) et, Si = E (dik,nYnk*), * désignant l'opérateur conjugué. We find that the optimal coefficients for the equalizer 32i of the user k satisfy the following matrix equation:

kk, opt k, opt k, opt T-1 bk Topt = [ti, oti, i .... t 1-1 1 kik * kkk, nk * with (D = E (Yn Yn) and, Si = E (dik, nYnk *), * denoting the conjugate operator.

dans les égaliseurs 320 à 32,-,, il s'agit de trouver la matrice T des coefficients optimaux sans avoir recours à l'inversion de matrice d'estimation (D k * Le pas de convergence A de ces algorithmes est le même et est tel que A < 2/Pe, où Pe dénote la puissance d'entrée à l'entrée de chaque égaliseur. Les

coefficients de la i ième ligne de la matrice T k sont n remis à jour selon la relation matricielle récursive :

rk,nk,n.rk, n-l, k, n-l.., k, n k, n., k*T 1, 0 1, 1-1 1, 0 1, 1-1 1 1 k n où d dénotent les symboles d'information d'une 1 séquence d'apprentissage transmise par l'émetteur 1. The inversion of the estimation matrix (Dk applied to the vector of the l parallel filtered symbols is advantageously avoided by using the stochastic gradient algorithm or the Kalman algorithm, the parallel algorithms are respectively implemented in the I equalizers 320 to zig
For the I gradient algorithms implemented

in the equalizers 320 to 32, - ,, it is a question of finding the matrix T of the optimal coefficients without having recourse to the inversion of matrix of estimation (D k * The step of convergence A of these algorithms is the same and is such that A <2 / Pe, where Pe denotes the input power at the input of each equalizer.

coefficients of the ith line of the matrix T k are n updated according to the recursive matrix relation:

rk, nk, n.rk, nl, k, nl .., k, nk, n., k * T 1, 0 1, 1-1 1, 0 1, 1-1 1 1 kn where d denote the symbols information of a 1 learning sequence transmitted by the transmitter 1.

<Desc / Clms Page number 17>

matrice Topt des coefficients optimaux sans avoir - 1 recours à l'inversion de matrice d'estimation C,.

For the Kalman algorithms implemented in the 320 to 32I-1 equalizers, it is necessary to find the

matrix Topt of the optimal coefficients without having recourse to the estimation matrix inversion C ,.

cy-la puissance du bruit additif blanc gaussien, et k, n d. les symboles d'information d'une séquence d'apprentissage transmise par l'émetteur 1. They are given by the following iteration: k, nk, n ,. k, nl .. k, nl. pk, k, nk, nk * T zip 0 '- 1, 1-11, 0- "i, Il ni'i Yn) kkk kT with P = Pn-i (U-KY) the covariance matrix of error, U is the matrix unit of order I, kk * Kk p-lyi the gain of Kalman, Z yKipK yK + y p-iY

cy-the power of Gaussian white additive noise, and k, n d. the information symbols of a training sequence transmitted by the transmitter 1.

d^k k k k d = sgn (Zn) = sgn (T opt Y il) Lorsque les I algorithmes ont convergés, la matrice T-opt des coefficients optimaux pour le
2-1 récepteur 3 de l'usager k est égale à (rk + # U) et correspond à l'égalisation par forçage à zéro selon la première variante, au bruit additif près. Finally, the decision circuits 330 to 331-1 produce, after convergence of the algorithms, the vector of the estimated symbols:

d ^ kkkkd = sgn (Zn) = sgn (T opt Y il) When the I algorithms have converged, the T-opt matrix of the optimal coefficients for the
2-1 receiver 3 of the user k is equal to (rk + # U) and corresponds to the equalization by forcing to zero according to the first variant, to the additive noise near.

 According to a second embodiment, the set of decision equalizers 320-330 to 321-1-331-1 has a decision feedback equalization (DFE) structure according to the vector notation diagram shown in FIG. Figure 5 in which the dependence on the index k of the user has been removed to lighten the writes. The set of decision circuit equalizers comprises a bank of input filters 35 respectively connected to positive inputs of I subtractors 36, 1 circuits of

<Desc / Clms Page number 18>

 decision 37 respectively connected to the outputs of the subtractors 35, and a bank of I return filters 38 respectively between the outputs of the decision circuits and the negative inputs of the subtractors 35.

 Since the propagation channel is known to the receiver, the I components of the symbol vector Yn at the output of the samplers 310 to 311-1 are ordered according to their decreasing powers, the first component being the most energy-carrying and so on. This ranking according to the order of decreasing powers is important to ensure proper operation of the receiver.

réponse \fi et produit le vecteur de symboles filtrés :

o, o-Oj-0, 1-1'yo o......... The symbol vector Yn is applied by the samplers 310 to 311-1 to the input filterbank 35 which constitutes a whitening filter.

answer \ fi and produce the vector of filtered symbols:

o, o-Oj-0, 1-1'yo o .........

Zn = Yn = 0 0 i, i-i, I-1 Yi 0 ...

0 ...... 0 i-ij-i yi-i.

0 4to o '0............ 0' qO O 0....... ".., 0 I, 0...'"... Like the matrix of the input filters, the matrix representing the return filter bank 38 is triangular:

0 4to o '0 ............ 0' qO O 0 ....... ".., 0 I, 0 ... '" ...

0 / I-1, 0 --- n-lj-2 O

<Desc / Clms Page number 19>

1 Dn = sgn (Sn) où Sn dénote le résultat des soustractions dans les soustracteurs 36, soit :

^ "il n n T Sn= (Zn -RDn) = (\I'Yn - RDn) =[sO,... si'... s 1-1] i 1-1 1-1 i ^ = 1 ln soit . ly- = Y,-rlDn 1 1, J J 1, J J j=O j=O ième A la i ligne de cette matrice \} fI

correspond aux coefficients du filtre d'entrée de la fizz voie de diversité. La notation r est ième identique pour le filtre de retour de la i voie de diversité. The decision-making process in decision-making circuits 37 is

1 Dn = sgn (Sn) where Sn denotes the result of subtractions in subtractors 36, that is:

^ "it nn T Sn = (Zn -RDn) = (\ I'Yn - RDn) = [sO, ... if '... s 1-1] i 1-1 1-1 i ^ = 1 ln let l = Y, -rlDn 1 1, JJ 1, JJ j = O j = O th On the line of this matrix

corresponds to the coefficients of the input filter of the diversity route fizz. The notation r is the same for the return filter of the diversity channel.

[ ;1-1 J ^ n n 1-1 =sgn z~ di = sgn z-jd" L j=o J

Les coefficients et r des filtres d'entrée et de retour 35 et 38 sont calculés de préférence selon l'algorithme du gradient stochastique, bien qu'un calcul selon l'algorithme de Kalman soit possible

mais plus complexe. La mise à jour des coefficients 1 1 'P, rn des filtres d'entrée et de retour 35 et 38 de la i voie de diversité résulte des relations matricielles récursives de l'algorithme de gradient :

The seme symbol is thus decided according to the relation:

[; 1-1 J ^ nn 1-1 = sgn z ~ di = sgn z-jd "L j = o J

The coefficients and r of the input and return filters 35 and 38 are preferably calculated according to the stochastic gradient algorithm, although a calculation according to the Kalman algorithm is possible.

but more complex. The updating of the coefficients 1 1 'P, rn of the input and return filters 35 and 38 of the i diversity channel results from the recursive matrix relations of the gradient algorithm:

n où di dénotent encore les symboles d'information d'une séquence d'apprentissage transmise par l'émetteur 1.

n where di still denote the information symbols of a training sequence transmitted by the transmitter 1.

Claims (13)

 separating each sequence of symbols into sets of k I parallel symbols (C 1, D 1-1), I parallel means (120, 12, -,) for encoding I parallel symbols relating to a user respectively by I sequences k k of elements of code (C, Cj) associated with the user, I parallel means (130, 1i-1) for scrambling l coded symbols relating to a user respectively kk by I scrambling sequences (So, SI-1) associated with the user , and I parallel means (140-150, 14I-1-15I-1) for simultaneously transmitting I coded and scrambled parallel symbols relating to a user, I being an integer greater than 1.

 1-Transmitter (1) for a cellular radiotelephone system transmitting symbol sequences respectively multiplexed in time slots assigned to users, the symbols being coded by sequences of code elements and scrambled by scrambling sequences, characterized in that it comprises means (11) for 2-transmitter according to claim 1, in kk the sequences of code elements (CQ, II associated with a user are identical to a single sequence of code elements associated with the user and different from other sequences of code elements respectively associated with other users.

3-transmitter according to claim 1 to 2, wherein two parallel means for transmitting (14i-15i'14i + 1-15i + 1) comprise in common a transmitting antenna (16i, i + 1) having crossed polarizations respectively to transmit two coded parallel symbols scrambled. <Desc / Clms Page number 21>  4-Receiver relating to a user for a cellular radiotelephone system having a single antenna receiving means (AR, ET) for receiving a signal (r (t)) of symbols transmitted by the transmitter according to any one of the claims 1 to 3,

 characterized in that it comprises I parallel filtering means (200-210'201-1-211-1 300-310'301-1k 31y-1) adapted to I code element sequences (Co, Cj) associated with the user, I kk scrambling sequences (So, SI-1) associated with the user and I propagation channels between I parallel means for transmitting (140-150, 14I-1-15I-1) in the transmitter (1) and the reception means (AR, ET) in order to deliver I

 Parallel filtered symbols (y, n yf), and the equalizing and deciding means 0 1-1 (220-250, 22-25, 32o-33o, 32j-33) to cancel the interference between the symbols filtered parallel to return the corresponding symbols relating to the user, I being an integer greater than 1.  The receiver according to claim 4, wherein each equalizing and deciding means comprises intersymbol interference calculating means (23i) for estimating interference (IESi) between 1-1 parallel filtered symbol samples.

 kn other than the respective filtered symbol (y. 'n), a means (24i) for subtracting from the respective filtered symbol the estimated intersymbol interference in a corrected symbol kn (yc'), and a decision means at 1 thresholds (25i) for restoring the respective symbol (d- ') according to the corrected symbol. 1 <Desc / Clms Page number 22> The receiver according to claim 4, wherein each equalization and decision means comprises intersymbol interference calculation means (23i) for estimating interference (IESi) between 1-1 parallel filtered symbol samples. other than a respective filtered symbol (y '), means (24i) for subtracting from the received signal (r (t)) the estimated inter-symbol interference to a corrected symbol, a second filtering means (27i) adapted to the sequence of coding elements, the scrambling sequence and the propagation channel associated with the user for filtering the corrected symbol produced by the means for subtracting, and threshold decision means (25i) for restoring the respective symbol

 dz according to the filtered corrected symbol. 1 The receiver according to claim 5 or 6, wherein the process of interference between

 symbols in the calculating means (23i) and in the means for subtracting (24i) is iterative.  8-receiver according to any one of claims 5 to 7, wherein each means of equalization and decision comprises a means of

 threshold decision (22i) between the respective filtering means (20i - 21i) and the respective symbol interference calculating means (23i).  9-Receiver according to claim 4, wherein the equalization and decision means (320-330, 32I-1-33j-i) apply a zero-forcing linear equalization algorithm on the I parallel filtered symbols . <Desc / Clms Page number 23>  Receiver according to claim 4, wherein the equalization and decision means (320-330, 32-33y) apply a linear equalization algorithm according to the minimum mean squared error criterion on the filtered I symbols. parallel.  11-A receiver according to claim 10, wherein the equalization and decision means use a gradient algorithm or a Kalman algorithm to avoid the inversion of an estimation matrix applied to the parallel filtered symbols.  The receiver according to claim 4, wherein the equalization and decision means (35-38) have a feedback return transverse equalization structure.  13-Receiver according to claim 12, wherein the equalization and decision means comprise an input filter bank (35) and a return filter bank (38) whose coefficients are calculated according to a gradient algorithm or a Kalman algorithm.