FR2812127A1 - METHOD OF FORMING ZEROS BY AN ARRAY OF ANTENNAS - Google Patents

Abstract

The invention relates to a process for forming zeros by an array of antennas (A1, A2,..., AM). Each antenna (A j ) is associated with phase information and gain information. The method comprises, for each antenna, a phase quantization step and a gain quantization step making it possible to define least significant phase bits and least significant gain bits, and a phase minimization step. the total power received by the antenna array (A1, A2,..., AM) by modification of at least one least significant gain and/or phase bit. The step for minimizing the total power received comprises a microcanonical annealing step.

Description

METHOD OF FORMING ZEROS BY AN ARRAY OF ANTENNAS

  TECHNICAL FIELD AND PRIOR ART The invention relates to a method for forming zeros by an array of antennas. The invention applies, for example, in fields such as satellite telecommunications, cellular telephony or even local networks

wireless.

  An antenna array is generally designed to receive signals from one or more specified coverage areas. A process for forming zeros makes it possible to modify the reception law of the antenna network in phase and / or in amplitude in order to place reception zeros for the signals located outside the

useful cover.

  Several methods are known to the man of

art for the formation of zeros.

  The document "Phase-only Adaptive Nulling with a Genetic Algorithm", Randy L. Haupt, IEEE Transactions on Antennas and Propagation, vol.45, N 6, June 1997 discloses a method for adjusting the phases of the signals received so as to place several deep zeros in the direction of interference and to minimize disturbances on the useful part of the diagram

network influence.

  The document "Experimental Adaptive Nulling with a Genetic Algorithm", Randy L. Haupt, H. Southall, Microwave Journal, January 1999, discloses a device for the formation of zeros by phase adjustment and

antenna gains.

  These two documents relate to a method of forming zeros consisting in adjusting the phases of the signals and / or the gains of the antennas, previously quantified and coded on a certain number of bits, so as to minimize the total power at the output of the network. In the absence of disturbances, such an approach leads to minimizing the useful signal. In order to avoid this minimization, the method is only implemented when the ratio of the useful signal to the signal made up of disturbances and noise is low. Only the least significant bits of the phases and amplitudes are modified. To determine the least significant bits which minimize the output power, the methods use a genetic algorithm, namely an algorithm which simulates the laws of genetics. Genetic methods have many disadvantages. They constitute an empirical approach which does not ensure, with a limited speed, the

  asymptotic convergence towards optimal solutions.

  Values must be given to parameters such as the size of the population concerned, the number of generations, the crossing rate, the mutation rate, the number of executions necessary to obtain good solutions, etc. A a large number of potential solutions, called individuals, must be kept in mind. The memory size occupied by these solutions can then become significant, which is to be avoided for on-board antenna networks such as the networks used in satellite communications. In addition, in order to sort the various potential solutions kept in memory, a large

  number of power measurements is necessary.

  The invention does not have the drawbacks

mentioned above.

  Presentation of the invention In fact, the invention relates to a method of forming at least one zero by an array of antennas, each antenna being associated with phase information and gain information, the method comprising, for each antenna, a phase quantification step and / or a gain quantification step making it possible to define least significant phase bits and least significant gain bits, and a step of minimizing the total power received by the network antennas by modifying at least one least significant bit of gain and / or phase, the step of minimizing the total power received implementing a combinatorial optimization method. The combinatorial optimization process

  includes a step of microcanonical annealing.

  According to an additional characteristic of the invention, the microcanonical annealing step comprises an iterative Creutz procedure which comprises the random selection of an antenna from the network, the random selection of at least one bit from among the least significant phase bits and / or gain associated with the selected antenna, the inversion of the least significant bit selected and the measurement of the total power received by the antenna array in the configuration

thus selected.

  According to another additional characteristic of the invention, the step of, implementing the Creutz procedure comprises N successive random selections of a "demon" among D "demons" stored, each "demon" stored being associated with a energy value, and a value of a test variable that counts the number N of random selections made

  is increased by one at each selection.

  According to another additional characteristic of the invention, the energy value associated with a demon is compared with the variation of the power induced by the inversion of the least significant bit so that, if said variation is less than the energy associated with the demon, the energy associated with the demon is corrected and the value of the least significant bit confirmed and, if not, a

new daemon is selected.

  According to another additional characteristic of the invention, the phase and gain information of each antenna is memorized if the total power measured has the value of a value which is the smallest

  value measured since the start of the process.

  According to another additional characteristic of the invention, when the number of selections reaches the value N, the energy value associated with each of the demons is reduced, the value of the test variable is reset to zero and a new sequence of selections

  successive random of a demon is performed.

  According to another additional characteristic of the invention, the energy values associated with the

  different "demons are chosen identical.

  The method of the invention modifies the phases and / or the quantized gains on the basis of the measurements of the total power received by the antenna array and advantageously uses a limited number of bits of

phase and gain of each antenna.

Brief description of the figures

  Other characteristics and advantages of the invention will appear on reading a preferred embodiment of the invention described with the aid of the attached figures, in which: - Figure 1 shows the method of forming zero according to the mode preferred embodiment of the invention, - Figure 2 shows a block diagram of the device for implementing the method of

  zero formation according to the invention.

  Detailed description of modes of implementation of

  the invention FIG. 1 represents the process for forming zeros according to the preferred embodiment of the invention. The process includes 3 main steps: - an initialization step; - a stage of implementation of the Creutz procedure;

- an annealing step.

  The initialization step (step 1) includes 3

main basic steps.

  As mentioned previously, the gain and phase shift associated with each antenna are quantified. So, a phase shift is coded by a word

  of P bits and a gain by a word of G bits.

  A first elementary step of the initialization step consists in defining a number of least significant gain and / or phase bits that the method will be able to modify. A number Np of least significant bits of the word of P bits is thus chosen, as is a

  number NG of least significant bits of the word of G bits.

  By way of nonlimiting example, Np and NG can respectively be equal to 3 and 2 for P and G

equal to 8.

  We denote by Cod the coding word associated with the modifiable bits. Cod is thus, for example, a vector consisting of the succession of values of the Np bits of

phase and NG gain bits.

  A second elementary step consists in defining initial values of parameters which will be used during the method according to the invention. Thus, D real numbers, commonly called "demons", associated with respective energies Ecl, Ec2, ..., ECD are stored, an integer N, for example N = 100xNpxNG, is defined and stored, a real number has belonged at the interval [0, 1 [and close to 1, for example a = 0.96, is defined and stored, an integer mem of initial value 0 is defined is stored, a whole number memopt close to 0, for example memopt = N / 1000, is defined and memorized, and a test integer initially at 0 is defined and memorized. As will appear below, these different parameters represent quantities necessary for implementing the method according to the invention. A third elementary step consists in initializing a measurement of the total power Pout received by the antenna array. The corresponding coding word Cod is created and stored. A Popt parameter is defined as the minimum output power measured. Codopt coding corresponding to the parameter

  Popt is also defined and stored.

  The implementation step of the Creutz procedure includes a displacement step (step 2) which includes the following steps: - increment of the test parameter of a unit (test = test + 1) - random selection of an antenna of the network and at least one least significant bit of phase and / or at least one least significant bit of gain to be modified, - inversion of the selected least significant bit (s) (bit = 1 - bit), - measurement of the total power Pn of the antenna array with the setting thus obtained,

  - calculation of the power variation AP = Pn - Pout.

  - random choice of a "demon" d selected from

the D memorized.

  If the inversion of the bit (s) mentioned above

  above leads to a power variation less than ECd (step 3, AP <ECd), the following operations are implemented (step 4): storage of the new setting in the coding word Cod, - the energy value Ecd of <"daemon"> d is corrected by the AP value (Ecd = Ecd AP), - the mem parameter is incremented by one (mem =

mem + 1).

  If Pn <Popt (step 8), then the setting of the least significant bits found is the best and it is

  memorizes (step 5). It comes Codopt = Cod and Popt = Pn.

  Steps 8 and 5 are optional.

  If the inversion of the bit does not lead to a negative power variation, the setting of the least significant bits is stored in the Cod parameter, provided that Pn is less than Ecd, and the following operations are carried out: - the mem parameter is incremented by one unit (mem = mem + 1), - the energy value Ecd of the "daemon" d is corrected by

the AP value (Ecd: = Ecd - AP).

  As long as the number of settings tested in the second step (Creutz procedure) is less than the value N (step 6), the process returns to the start of the Creutz procedure. When the number N is reached, the process proceeds to the annealing step (step 7). The

  test variable is then reset (test = 0).

  If the number of settings memorized (mem parameter) during the Creutz procedure is greater than the memopt limit value, then the following operations are implemented (step 7): - the mem parameter is reset, - the energy values of the "demons" are decreased, for example by multiplying each of them by the value a, - the process returns to the beginning of the Creutz procedure. If the number of settings stored (mem parameter) during the Creutz procedure is less than or equal to the memopt limit value (step 9), the process

  returns to the start of the Creutz procedure.

  According to a variant of the invention, the initial values of the energies Ecl, Ec2, ..., EcD are chosen to be identical and sufficiently high. It is then possible that most of the codings tested during the implementation of the Creutz procedure are memorized. The microcanonical annealing step has many advantages. The configuration of the process is simple and robust. It follows that the performances obtained are not very sensitive to small variations in the parameters. Furthermore, it is possible to limit the optimization process over time. Thus, the duration of the convergence process can be reduced by

decreasing N or a.

  FIG. 2 represents a block diagram of the device for implementing the method of

  zero formation according to the invention.

  The device comprises an antenna block B, a receiver R, a power measurement circuit MP and a test and optimization module T. The antenna block B comprises an array of M elementary antennas A1, A2 ,. .., AM. The signal from each elementary antenna Aj (j = l, 2, ..., M) is corrected: it is phase shifted and amplified. Thus, information coded by a word of P bits represents the phase shift of the antenna and information coded by a word of G bits represents the gain of the antenna. The word of P bits and the word of G bits are respectively transmitted to a correction circuit of

  phase PHj and to a gain correction circuit AGj.

  The phase correction circuit PHj and the gain correction circuit AGj also receive the coding word Cod from the test and optimization module T. The corrected code from the phase correction circuit PHj consists of 'a word of P bits whose Np bits of least significant have for values the values of the Np bits of the word Cod. Similarly, the corrected code coming from the gain correction circuit AGj consists of a word of G bits of which the NG least significant bits have for

  values the values of the NG bits of the word Cod.

  The corrected phase code from the PHj circuit and the corrected gain code from the AGj circuit are transmitted to a signal reconstruction circuit MDj

antenna.

  All of the reconstituted signals from the various devices MDj (j = l, 2, ..., M) are summed in an adder E. The signal from the adder 1 is transmitted to a receiver R. A power measurement circuit MP receives a signal from the receiver R in order to measure the signal strength from the adder E. The measured power information from the circuit MP is supplied to the test and optimization module T. The test and optimization module T includes the various elements making it possible to implement the method according to the invention for determining the word Cod to be applied to the antenna array. It follows that the test and optimization module T includes memory circuits and a calculation unit for implementing the Creutz procedure and annealing it.

  microcanonics as described in Figure 1.

Claims (7)

  1. A method of forming at least one zero by an array of antennas (Ai, A2, ..., AM), each antenna (Aj) being associated with phase information and gain information, the method comprising , for each antenna, a phase quantization step (P) and / or a gain quantification step (G) making it possible to define least significant phase bits (Np) and least significant gain bits (NG) ), and a step of minimizing the total power (Pn) received by the antenna array by modification of at least one least significant bit of gain and / or phase, the step of minimizing the total power received implementing a combinatorial optimization method, characterized in that the combinatorial optimization method comprises an annealing step   microcanonics (2, 3, 4, 5, 6, 7, 8, 9).   2. Method according to claim 1, characterized in that the step of microcanonical annealing (2, 3, 4, 5, 6, 7, 8, 9) comprises an iterative Creutz procedure (2, 3, 4) which comprises the random selection of a network antenna, the random selection of at least one bit among the low phase and / or gain bits (Np, NG) associated with the selected antenna, the inversion of the selected low weight and the measurement of the total power (Pn) received by the antenna array in the configuration thus selected.   3. Method according to claim 1 or 2, characterized in that the step of implementing the Creutz procedure (2) comprises N successive random selections of a "demon" (d) from D stored "demons", each memorized "demon" being associated with an energy value (Ecl, Ec2, ..., ECD), and in that a value of a test variable (test) which counts the number N of random selections made   is increased by one with each selection.   4. Method according to claim 3, characterized in that the energy value associated with a demon (d) is compared to the variation of the power induced by the inversion of the least significant bit so that, if said variation is lower than the energy associated with the demon (d), the energy associated with the demon (d) is corrected (4) and the value of the least significant bit confirmed and, if not, a new demon (d) is selected.   5. Method according to claim 4, characterized in that the phase and gain information of each antenna are memorized (5) if the total power measured has a value which is the smallest value measured since the start of the process. 6. Method according to any one of   claims 3 to 5, characterized in that, when the   number of selections reaches the value N, the energy value associated with each of the daemons (d) is reduced, the value of the test variable (test) is reset and a new sequence of selections   successive random of a demon (d) is performed.   7. Method according to any one of   claims 3 to 6, characterized in that the values   of energy (Ec1, EC2, ..., ECD) associated with the different "demons" are chosen identical.