WO2000064084A1

WO2000064084A1 - Method for reducing interference between users and use thereof in a radio access network

Info

Publication number: WO2000064084A1
Application number: PCT/FR2000/000958
Authority: WO
Inventors: Hikmet Sari
Original assignee: Alcatel
Priority date: 1999-04-15
Filing date: 2000-04-13
Publication date: 2000-10-26
Also published as: IL140262A0; EP1088414A1; AU3973600A

Abstract

Omnidirectional or sectorial emissions are provided in a network of radioelectric links radiated by base stations. User terminals are provided with directional antennae which can be pointed at said base stations. A CDMA mode is an encoding mode for the symbols that are to be emitted before modulation by a carrier. Some user terminals are located in unfavorable zones of said network where they receive parasitic emissions from other base stations that are aligned in the direction in which they are pointed. In order to avoid negative effects of said parasitic emissions, the signals directed at said user terminals are encoded with double-length coding sequences. A transmission quality factor amounting to as much as 3dB can be gained by means of said redundancy. The quality is further increased if the length is also increased.

Description

METHOD FOR REDUCING INTERFERENCE BETWEEN USERS AND USE IN A RADIO ACCESS NETWORK

The subject of the present invention is a method for reducing interference between users, and a use of this method in an access network of radio type. The field of the invention is preferably that of radio access networks using millimeter or sub-millimeter frequencies. These networks are designated in the literature under various names such as LMDS for Local Multipoint Distribution Systems (local distribution system with multiple fixed users). In these networks, a base station uses a sector antenna and the antenna of a user's terminal is a directional antenna, pointed towards the base station serving a cell in which this terminal is located. It is also possible that the antenna of the base station is not sectoral. These systems are presented in the article entitled "Broadband radio access to home and businesses: MMDS and LMDS" due to Hikmet SARI and published in the journal Computer Networks, vol 31, by Elsevier Science Holland in 1999, pages 379 to 393.

The purpose of networks of this type is to constitute an alternative to wired networks (pairs of copper wires, coaxial cables, or optical fibers) already in place in territories and which present performance limits or prohibitive infrastructure costs. The object of the invention is to solve interference problems which arise on the other hand in the use of networks of this type. In these multi-point radio access networks, the antenna of the base station can be omnidirectional, or sectoral. In the sectoral case, a cell of such a network is made up of several sectors. In a conventional implementation, a sector uses a frequency distinct from the other sectors of the same cell (possibly each sector uses a subset of frequencies, the subsets used in distinct sectors are disjoint). The principle of frequency allocation between neighboring cells aims to minimize the potential interference created in the cellular network.

Figure 1 shows an example of an LMDS network, rectangular with 90 ° sectors. References 1, 2, 3 and 4 designate four frequencies or frequency subsets used in this type of cellular network. In this type of network, each cell is made up of a square centered on a base station. These base stations are represented by circles. Each cell is divided into four sectors corresponding to the four sectoral antennas of the base station. The solid lines in FIG. 1 constitute the borders of the sectors belonging to the same cell, the dotted lines represent the borders between cells. Fixed user terminals are represented by dots. For example, points A, B and C represent terminals located in the same sector of the same cell.

One of these cells is shown in gray. Within a cell, each of the four sectors is labeled with a number which indicates the frequency, or the set of frequencies, which is (are) assigned to it. It can be seen that all the base stations use the same frequencies, but that the assignment of the frequencies to the various cells of the network is made so that there is no interference between adjacent cells.

Other geometries of cellular networks can be defined, using sectoral antennas of different angular opening, for example: 120 °, 60 °, 45 °. In all cases, the implementation of the network requires the use of several frequencies per cell, and generates interference between users of the system.

The signal to interference ratio (C / l) is a function of the user's position. The worst case corresponds to the extreme points located on the horizontal, vertical and diagonal lines of the cells for each base station. Indeed, an antenna associated with the terminals located at these points is directed not only in the direction of its own base station, but also in the direction of the base stations situated on the same horizontal (respectively vertical or diagonal). Given the frequency assignment mode, there is interference with a base station immediately following an adjacent base station, along the horizontal line (respectively vertical, diagonal). However, there is no interference with the adjacent base station.

The ratio between the respective lengths of the paths carrying a useful signal and that carrying an interference signal is unfortunately small. For example, for sector 1 of the base station at the top left In Figure 1, the three most unfavorable points are indicated by A, B and C respectively. Suppose that 2D represents the distance between two neighboring base stations. A user at points B and C at the edge of the cell with his terminal pointed towards his base station (therefore at a distance D from it) interferes with the base station located at a distance 5D (D + 2D + 2D) . The terminal B shown in FIG. 2 is located at the cell boundary, on a horizontal straight line joining the base stations of the network together. Of course, the same diagram would be applicable at point C, located on a vertical line joining the base stations. In the diagonal case of point A, a Visera factor introduced over all the distances considered and the final result will remain unchanged.

Assuming that all base stations transmit the same power, the signal to interference ratio is equal to C / l = 10 log (5 ² ) or 14 dB. This description corresponds to a TDMA (Time Division Multiple Access) system having four frequency channels to cover an entire geographic area. If we now assume that there are just two channels available, the ratio C / l associated with the most unfavorable points becomes C / l = 10 log (3 ² ) or

9.5 dB. This can be easily shown by replacing 4 by 2 and 3 by 1 in Figures 1 and 2. It is then clearly seen that there is in this case interference between adjacent cells.

The choice of a frequency allocation architecture is decisive in this regard. An operator, in a given frequency range, is allocated a frequency band, by definition limited. In this band, it must provide uplink and downlink communications for each sector, taking into account that the base station transmits omnidirectionally or at least sectorally and that the terminals transmit directionally. The choice of such an omnidirectional transmission for a base station is linked to reasons of economy of equipment of the base stations.

The object of the invention is to remedy these drawbacks and to propose an access technique in which this interference is no longer a problem. Although the invention is intended to solve the problems of the downlink, nothing prevents it from also being used in the uplink. The invention therefore recommends using, for the downlink at least, a spread spectrum coding method of the CDMA type with the use of the same frequency in all the sectors. With such a type of coding, binary symbols transmitted of type +1 or -1 are coded by multiplication by a sequence of 2N coding bits +1 and -1 also. The result of this multiplication is a sequence of 2N binary signals +1 and -1 called chips. These chips are then used to modulate a carrier whose signal is radiated by the antenna of the base station. Since the time allocated for sending a symbol remains the same, all other things being equal, the spread spectrum factor resulting from this process is 2N. With orthogonal coding sequences, for example of the Walsh Hadamard WH type, the number of different sequences that can be used is 2N. And so a number of users likely to be simultaneously in this geographic location is 2N per cell if the radiation is omnidirectional and if it associates a sequence with each user.

In access networks with directional user antennas described above, interference to be addressed is only geometric. It mainly affects terminals receiving broadcasts from base stations aligned with respect to them. For example, a receiver at point B in FIG. 2 receives the transmissions from the base stations BTS1 and BTS2 since it points to the base station BTS1. As all the base stations transmit on the same carrier, an interference problem arises at point B. Since the base stations are not synchronized, the orthogonality of coding by coding sequences 2Ni (the coding sequence 2Ni is assigned to the downlink from BTS1 to B) is of no use to protect receiver B from the transmissions of stations BTS2 and BTS3.

According to the invention, it is chosen to transmit a message to user B from the base station BTS1, to code each symbol with a double sequence WH. This double sequence has a first and a second single sequence. In the continuation of this presentation, we will consider the term of simple sequence of length 2N, usable conventionally with a spread spectrum of factor 2N, and the term of double sequence of length 4N (or triple 6N, or quadruple 8N, or other) with also a spread of spectrum of factor 2N, usable in the invention. Such a solution effectively solves the problem because user B will receive the emissions (intended for it) from the base station BTS1 with a sequence 2Ni2Ni. The sequence 2Ni2Ni is in this case a double sequence of length 4N. At the time of decoding, the decoding of the chips received by the expected double 2Ni2Ni sequence will be 3dB greater than the decoding of the chips coded by a simple 2Ni sequence. A factor of 3 dB is thus added to the quality of the useful signal received. In some cases, this is sufficient to recognize the useful signal. If necessary, a symbol to be transmitted can be coded by a triple sequence, a quadruple sequence or any multiple sequence. In this case, there is a gradual gain in signal-to-noise ratio.

The subject of the invention is therefore a method of transmitting CDMA type messages between a base station and user terminals in which,

- for messages intended for certain user terminals, symbols of these messages are coded with a coding sequence of 2N bits to produce sequences of 2N chips, and

- the chips are emitted, characterized in that

- for other messages intended for certain other user terminals, symbols of these other messages are coded with a coding sequence of k2N bits to produce sequences of k2N chips, k being an integer greater than one.

The invention will be better understood on reading the description which follows and on examining the figures which accompany it. These are presented for information only and in no way limit the invention. The figures show:

Figure 1: an already commented representation of a rectangular cellular network with 90 ° sectors;

Figure 2: an already commented representation of an interference scenario in the network of Figure 1; FIG. 3: a representation according to the invention of an assignment of orthogonal coding sequences of the CDMA type inside a cell;

Figure 4: a representation of a preferred generation of a signal encoded with a CDMA technique;

Figure 5: a representation of interference in an uplink, from a user to a base station; Figure 6: a representation of interference in a downlink from a base station to a user;

Figure n ° 7: a representation of gray areas corresponding to strong interference (descending direction); Figures 8a and 8b: a temporal representation of different coding modes of the invention.

The networks described above show the limits of TDMA cellular networks. For a network with four frequency bands (occupied band 4W), in the worst case, the signal to interference ratio C / l is worth 14dB. For a network with two frequency bands (occupied band 2W): the C / l ratio is 9.5 dB. It is proposed with the invention, using a different access technology, to improve these limits, for example by obtaining, for an occupied band 2W, a C / l ratio better than 9.5 dB. For this purpose, in the invention it is possible to use the same frequency in all the sectors using the CDMA access technique. A variant of this technique is proposed which makes it possible to significantly increase the C / l ratio associated with the most unfavorable positions.

As before, we consider in an example a cellular network whose geometry reproduces that described in FIG. 1, but we replace the four carrier frequencies (or groups of frequencies) 1 to 4 by the same carrier frequency coded by disjoint subsets S1 and S2 of orthogonal decoding sequences (for example Walsh-Hadamard sequences) of length 2N. This is shown in FIG. 3. If the sectorization comprises more than four sectors, for example six or eight, it is made so that disjoint subsets of orthogonal sequences are assigned to two adjacent sectors in the cell. The number 2N is retained here to allow a simple comparison, in terms of signal to interference ratio, between the solution of the invention and that of the state of the art. But this is not an obligation. This number could even be odd although for other reasons it will preferably be even.

The assignment of the subsets S1 and S2 to the four sectors of a cell is carried out as follows. Two sectors diagonal from one another in a cell are served by the same subset of sequences

S1 (or S2). Two contiguous sectors, i.e. neighboring but in two adjacent cells, are served by two subsets of sequences S1 (or S2) identical. For reasons of symmetry, the number of sequences in each subset will be N. But this is not an obligation. If necessary, some adjustments to the general sequence distribution plan may be required. This assignment mode ensures the absence of interference between neighboring sectors in a cell, thanks to the orthogonality between any of the sequences of S1 and S2.

Preferably, all of the cells in the network, and, inside each cell, all of the sectors, use the same frequency band of width 2W. In this bandwidth it is allowed to use a spreading factor equal to 2N. The number of users per sector is limited to N (ie a total of 4N users per cell) by assigning to each user a particular sequence taken from one of the subsets of N sequences S1 or S2. The preferred characteristics of the cellular network are therefore the following:

- spectral occupancy: 2W

- spreading factor: 2N

- number of sequences assigned to a sector: N - two adjacent sectors use subsets of disjoint sequences, each of cardinal N

- maximum number of users in each sector: N

It remains to minimize the interference between users of neighboring cells. All neighboring cells using the same frequency band, and the orthogonality being lost due to the non-synchronization of the base stations, a user in one cell interferes with all the users of the other cells. But the interference is deterministic and is not uniform. To make it uniform in a known manner, additional coding is implemented, carried out using separate PN sequences in each cell.

The signal of each user is therefore first spread by a sequence of length 2N and is then multiplied by a sequence PN, also of length 2N, without additional spectral spreading. The multiplication is made chip from the Walsh-Hadamard coder by bits of the PN sequence. This last operation, shown in Figure 4, is intended to separate signals from different cells. There is thus no interference between users of the same cell because their respective codes are orthogonal. But a given user is interfering with users from all other cells. The interference is not zero because the different conceivable PN sequences are not orthogonal to each other.

In the uplink, from a user terminal to a base station, the base stations receive interference from a small number of users located in the fields of their sector. This small number, shown schematically in the shaded areas in FIG. 5, is the number of users whose antennas are oriented towards the base station in the center of FIG. 5. The total level of interference is therefore low, and equal for all the users.

Unlike the uplink, the interference experienced by a user in the downlink depends on their position within a sector. Let us first examine the case of users located at points A, B and C in Figure 6. User A is located at the edge of the sector, on a diagonal line passing through its base station. The antenna of this user is pointed towards his base station which is at a distance ÏD. This antenna is also directed to another base station at a distance 3V2E>. If the sector associated with this other base station includes K users, then the ratio C / l is given by

C / l = 10 log [(2N / K) 3 ² ] (1) Indeed, for a standard useful signal at 1, the interference created by a user

1 2 unique amounts to _j , where the terms 3 and 2N respectively express the ratio of the squares of the distances and the gain provided by the PN sequences (the correlation coefficient of two PN sequences of length 2N judiciously chosen being equal to 1 / 2N). In the presence of K users, the ratio C / l is therefore well expressed by (1). It decreases with K. For K = N, we have: C / l = 101og (2 x 3 ² ) = 12.5 dB The user's antenna is also pointed at other base stations, also located on the diagonal , at distances (2M + 1) ÏD, with M> 1, but, the distance being greater, the corresponding interference will not be predominant.

User B is located at the edge of the sector, on the horizontal line passing through the base station. User C is also at the edge of the sector, on the vertical line passing through the base station. The antenna of one or other of these users is pointed towards its base station which is at a distance D. This antenna is also directed towards another base station at a distance 3χD, and it receives interference from of two sectors of the cell centered around this other base station. If these two sectors include K-ι and K users respectively, then the ratio C / l is given by:

For Ki = K ₂ = N, we have C / l = 101og (3 ² ) = 9.5 dB

It can be noted that these C / l ratios of 12.5 dB and 9.5 dB are valid, in each sector, only in the vicinity, on the one hand of point A, on the other hand of points B or C. In fact, in outside the shaded areas (see figure n ° 7), the C / l ratio will be much higher than these values.

For example, a receiver at a point A ', located on the diagonal midway between the end of the sector and the base station, that is to say at the edge of the gray area, receives the signal much more strongly. from its base station as an interfering signal from another base station. The ratio C / l (in the case where K = N) is then equal to:

For points B 'or C located on the horizontal or vertical midway between the end of the sector and the base station, i.e. at the limit where K = N) is equal to:

Consequently, we can therefore distinguish, for the ratio C / l, three distinct zones in a given sector:

- an area of strong interference, in the vicinity of points B and C where the ratio C / l takes a minimum value of 9.5 dB, equal to the value obtained for a system of the same 2W band using TDMA; - a slightly more moderate interference zone, in the vicinity of point A, where the C / l ratio can drop to 12.5 dB;

- finally the rest of the sector, where the ratio C / l will always be greater than 14 dB (or even greater than 17 dB between A 'and the base station). Since the user antennas are directional (angle of the opening on the order of 2 ° to 4 °), this area of lower C / l ratio will in fact cover the smallest part of the sector, which opens the way to the possibility improve the C / l ratios of the zones with a lower C / l ratio, at a reasonable additional coding cost. In the invention, several objectives can be set:

1st objective: improve the C / l ratio of zones B and C (the worst-off),

2nd objective: improve even more clearly the C / l ratio of zones B and C and in parallel also improve (but more modestly) the C / l ratio of zone A (which would otherwise become the worst-off)

The idea of the invention is to assign to the users most disadvantaged by their geographical position within a sector a coding process allowing them to benefit from a gain compared to the other users, in order to compensate for their handicap a priori. To this end, the delimitation of zones A, B or C defined in FIGS. 5 or 7 only serves to illustrate this difference in treatment as a function of the geographical position of the user. The invention remains perfectly applicable to zones the limits of which would be different, for example a zone B extending from point B to a point situated beyond or on the contrary below point B 'on the horizontal line joining B and the associated base station.

The ratio C / l is greater than 12.5 dB at any point outside of zones B and C. In the invention, without modifying the duration of the chips, and therefore the spectral occupancy, a coding sequence of double length is associated , 4N, or triple or quadruple, instead of a simple length of 2N, for calls descending from a base station with disadvantaged users in zones B and C. This reduces by 3 dB for doubling, or by 4, 8 dB for tripling or 6 dB for quadrupling, for these disadvantaged users, interference from a signal transmitted by another adjacent base station and intended for other users. In practice, we will see in Figures 8a and 8b how in a way preferred each symbol to be transmitted to such a disadvantaged user undergoes at least two successive codings by two simple coding sequences of length 2N to thereby form a double sequence of length 4N, and that this coding operation produces symbols of 4N chips. Sequences of length 4N must be orthogonal to sequences of length 2N of other users and must be orthogonal to each other. The bits of these trains of successive 4N chips are then processed in a transmitter like the bits of the sequences of 2N chips.

We are therefore in a situation where users are served by codings with a single simple coding sequence of length 2N per symbol, while disadvantaged users benefit from a double sequence of length 4N per symbol. If necessary, a triple or quadruple coding sequence could be allocated to code the same symbol of the messages to be transmitted for these disadvantaged users. Everything happens as if we added redundancy by repeating the symbol for certain messages only. The messages for which such redundancy is added are messages intended for geographically identified users, the disadvantaged nature of the communication of which is known, and the quality of which can be improved by identifying the coding sequences allocated to them, and by a doubling (tripling ...) of the coding of the symbols of their message by these sequences.

In compensation, as this operation divides by two (by three, by four, ...) the information rate (the duration of the chips remains identical with the same spreading factor, so as not to increase the spectral occupancy ), it is preferable to associate two coding sequences of length 4N with the user in question in order to keep his information rate constant (and therefore offer this user the same service as other users). The quality gain of 3 dB is therefore not clear, since the number of sequences associated by the base station will increase.

Suppose that the number of users with double sequences of length 4N is, by sector, equal to m. The total number of simple sequences is then (Nm) + 2m = N + m. Indeed Nm users have a simple sequence with a coding length of 2N, and m users have a double sequence with a coding length of 2 x 2N. The real gain in quality is:

NOT

G = 10 log 2-

N + mJ for m = Ν / 20 G = 2.8 dB => C / l _{(B θ} u _C) = 12.3 dB for m = N / 8 G = 2.5 dB => C / l _(B or _C ) = 12 dB for m = N / 4 G = 2.0 dB => C / l _(B o _UC ) = 11.5 dB

The net gain is all the better as the number of users requiring a double length sequence is reduced, so that surfaces B and C have a reduced area. The invention is perfectly applicable to fixed radio access systems, where the directivity of the antennas of user terminals is important.

An example of generation of elongated sequences from a simple length sequence, for N = 4 and m = 1, can be as follows: 1 1 1 11 111 if 111111111111111111 - two sequences of

11111111-1-1-1-1-1-1-1-1 length 4N = 16 are associated with a user 1 -11 -11 -11 -1 s2 - three users have

11-1-111-1-1 s3 a sequence of

- | _1_111-1-11 s4 length 2N = 8 Figures 8a and 8b show a preferred implementation of the method of the invention for a disadvantaged user. This transmits a message comprising symbols ai, a2, a3, a4 ... At each symbol time Ts, with a conventional CDMA coding, we would normally assign a sequence of length 2N, here 2N is 8. So the duration Ts is equal to 2N times the time Te of a chip. In the invention, the symbol ai assigns a coding sequence of length 4N (of length 16 here). The coding sequence naturally lasts longer than the symbol time Ts of the symbol ai since the number of chips transmitted is double without changing the frequency occupancy of the carrier. Two solutions are possible. Or, FIG. 8a, we accept to penalize the user and we reduce his useful bit rate by two (or three or four depending on the type of multiple sequence used). In this case, the real symbol time becomes double (triple, quadruple ...) of the basic symbol time. In practice, to keep an existing transmitter architecture, the easiest way is to code twice immediately a symbol ai to be transmitted in sequence 2Ni retained for a user. In reception, the simplest is to consider a double decoding length of 4N. In this case, the decoder uses a 4N sequence, for example that which is the first indicated above for the user if.

Either, Figure 8b, we do not accept this penalty. In this case, two coding chains are used, and two double coding sequences (of length 4N each) simultaneously. For example, the two double sequences are the two provided above for the user if. Thus, the symbol ai is coded by the beginning of its double sequence, by a first string. Then one begins to code with a second chain the symbol a2 by the beginning of its double sequence, while the symbol ai is still being coded by the first chain and by the second part of its double sequence. Each double sequence has a length 4N and a duration 4N times Te. When the first chain has completed the development of the chips relating to the symbol ai, it immediately begins, under the same conditions, the development of the chips of the symbol a3, while the symbol a2 is still being coded by the second chain. And so on. If triple or quadruple length coding sequences are chosen, preferably there will be three strings or four coding chains working simultaneously. It can be seen that by doing so, two symbols (at least) are emitted simultaneously, at least in part. In this case, a suitable receiver must have two decoding channels (three channels, four channels). These chains are put into service simultaneously to alternately decode symbols ai, a2, a3, a4 of a message transmitted to this user.

The first double sequence comprises a repetition of the same single sequence (with eight 1 here) while the second double sequence comprises a single sequence (the same as for the first double sequence) and another single sequence complementary to the first single sequence . In this way the two double sequences orthogonal to each other. In theory, it would be possible to constitute the double sequences from any single sequence, as long as the double sequences obtained are orthogonal to each other. However, since such double sequences are made up of single sequences, it should not unnecessarily neutralize simple orthogonal sequences between them and whose number is limited (to 2N) and thereby limits the number of users in the cell.

It will be noted that the doubling, tripling or other of these chains is not particularly penalizing in terms of hardware because the coding and decoding are operations processed by processing processors already contained in the base station and the user terminals. .

In the context of FIG. 8b, one could also transmit the two symbols ai and a2 completely simultaneously on the two chains, and continue by transmitting the symbols a3 and a4 and so on.

To obtain higher gains, by allocating to the users located in the zones of strong interference sequences of even greater length, for example 8N, and in order to respect the constraint of not increasing the spectral occupancy, the number of sequences to assign to these users is also higher (equal to 4 for sequences of length 8N). We could for example use the following strategy:

- use of length 8N sequences for users located in zones B and C: gross gain of 6 dB; - use of 4N length sequences for users located in zone A: gross gain of 3 dB;

The following example illustrates this assignment:

- Total number of users: N = 4

- Number of users in zone B or C: rαi = 1 - Number of users in zone A: m ₂ = 1

Either a user in zone B or C with four sequences of length 8N developed from the sequence 11111111 which becomes on the one hand 11111111 11111111 which itself becomes 11111111 11111111 11111111 11111111, and 11111111 11111111 -1-1-1-1 -1-1-1-1 -1-1-1-1-1-1-1-1, and on the other hand 11111111 -1 -1 -1 -1 -1 -1 -1 -1 -1 which becomes 11111111 -1-1-1-1-1-1-1-1 11111111 -1-1 -1-1 -1-1-1-1, and 11111111 -1-1-1-1-1-1-1 -1 -1-1-1-1-1-1-1-1 11111111.

We also see here only a single sequence is used to produce all the quadruple sequences. The simple sequence is only combined with its complementary image. Thus preferably a simple sequence is concatenated with a repetition of this simple sequence or with a complementary simple sequence.

Or a user in zone A with two sequences of length 4N developed from the sequence 1-11-11-11-1 which becomes on the one hand 1-11-11-11-1 1-11-11-11- 1 and on the other hand 1-11-11-11-1 -11-11-11-11.

Let two users have a sequence of length 2N (a) 11-1- 111-1-1 and (b) 1-1-111-1-11. For the general case, the net gain can be evaluated by considering the total number of users (N), the number of users in zone B or C (mi) and the number of users in zone A (m ₂ ):

The total number of sequences: T = (N - mi - m ₂ ) + 4χrτiι + 2χm ₂

= N + 3 ^χ m-ι + 1 m ₂ In zone B or C, the net gain amounts to:

(i) factor 4 expresses the lengthening by a factor of 4 of the sequences assigned to users in zones B or C;

N (ii) the factor expresses the relationship between the

N + 3m, + - total number of single sequences assigned taking into account the assignment of several simple sequences to users in certain areas and the number of simple sequences that would be assigned if all users had a single sequence

In zone A, the net gain amounts to Gnet = 10, or:

(i) factor 2 expresses the extension by factor 2 of the simple sequences assigned to users in zone A;

NOT

(ii) the factor: see above

This realization leads to the following numerical examples: (a) m1 = Ν / 8 and m2 = N / 8

For zone B or C, the ratio C / l in the absence of elongation of the sequences is 9.5 dB, the gross gain due to the lengthening of the sequences is 6 dB, the decrease in gain due to the increase in the number of sequences is 1.8 dB. The minimum C / l ratio in zone B or C is 13.7 dB For zone A, the C / l ratio in the absence of lengthening of the sequences is 12.5 dB, the gross gain due to the lengthening of the sequences is by 3 dB, and the decrease in gain due to the increase in the number of sequences is 1.8 dB. The minimum C / l ratio in zone A is therefore 13.7 dB (b) m1 = N / 20 and m2 = N / 20

For zone B or C, the ratio C / l in the absence of lengthening the sequences is 9.5 dB, the gross gain due to the lengthening of the sequences is 6 dB, and the decrease in gain due to the increase in the number of sequences 0.8 dB. The minimum C / l ratio in zone B or C is 14.7 dB. For zone A, the C / l ratio in the absence of lengthening of the sequences is 12.5 dB, the gross gain due to the lengthening of the sequences is 3 dB, and the decrease in gain due to the increase in the number of sequences is 0.8 dB. The minimum C / l ratio in zone A is 14.7 dB.

An evaluation of the m / N ratios can be carried out by calculating the ratio between the areas of the gray areas of a sector and the sector itself (assuming the user terminals uniformly distributed within a sector). This ratio depends on the directivity of the users' antenna. If the users are distributed homogeneously inside a sector, example (b) corresponds to the case where the whole (Zone B + Zone C) covers 5% of the surface of a sector, the zone Also at 5% and the "not grayed out" area 90%.

The above considerations have been developed within the framework of an ideal rectangular cellular network. They could also be applied to other cellular network topologies, for example with a hexagonal pattern, where the base stations use 120 ° sectoral antennas. In a real system, established on the ground, the geometry of the cells will not be as regular, because it will have to take into account the topography of the ground and the existence of constructions.

In terms of interference, they can in some cases be partially reduced by choosing the location and height of the base stations, within the limits of the possibilities offered by the land and the constructions.

On the other hand, the user terminals are in practice too numerous to benefit from these engineering techniques. In addition, their installation site is basically determined by the domicile of the user, and not by engineering parameters. Unlike base stations, their installation may not require teams that fully understand site engineering techniques.

It is impossible in practice to significantly minimize interference by the use of conventional engineering techniques.

Consequently, the method described above is of great interest: its implementation is very simple in a so-called fixed service system since it suffices to allocate the extended sequences to user stations whose geographical position corresponds to the areas of high interference. It is also possible, by deducing an interference map from the nature of the terrain and the constructions, to predict the level of interference that a user located in a given location will undergo and therefore to assign a priori the type of sequence suitable for his case, even in the case of a real network. . Digitized models of real terrain have already been established for French territory for example and can be used for this type of calculation.

In the case where the user station is transportable, the system could also be used by means of, for example, a location device associated with these stations.

Optionally, the detection conditions can be determined empirically and the user stations can be provided with a switching device which allows to choose or not a redundant mode or a normal mode.

Claims

1 - Method for transmitting CDMA type messages between a base station and user terminals in which, - for messages intended for certain user terminals, symbols of these messages are coded with a coding sequence of 2N bits to produce sequences of 2N chips, and

- the chips are transmitted, characterized in that - for other messages intended for certain other user terminals, symbols of these other messages are coded with a coding sequence of k2N bits to produce sequences of k2N chips, k being an integer larger than one.

2 - Method according to claim 1, characterized in that - for these other messages, at least two symbols are transmitted simultaneously.

3 - Method according to claim 2, characterized in that

- for these other messages, k symbols are transmitted simultaneously.

4 - Method according to one of claims 1 to 3, characterized in that

- a radiation cell of a base station is divided into sectors,

- the same carrier frequency is used for all the sectors of the sectorized cell, - coding sequences are distributed into subsets (S1, S2),

- different subsets are assigned to user terminals which are located in neighboring or contiguous sectors.

5 - Method according to one of claims 1 to 4, characterized in that - different base stations of a cellular system transmit chips on the same carrier frequency and with the same bandwidth.

6 - Method according to one of claims 1 to 5, characterized in that - the symbols or the chips are coded by bit sequences random (PN).

7 - Method according to one of claims 1 to 6, characterized in that

- to constitute a k2N coding sequence, a simple sequence is concatenated with a repetition of this simple sequence and or with a simple complementary sequence.

8 - Method according to one of claims 1 to 7, characterized in that

- in a user terminal, k decoding chains are put into service simultaneously in order to decode in parallel k symbols of a message transmitted to this user.

9 - Method according to one of claims 1 to 8, characterized in that

- in a user terminal, a symbol is decoded with a decoding sequence of length k2N.