ES2288452B1 - SYSTEM AND METHOD OF RECEIVING A SATELLITE SIGNAL THROUGH A MODULAR SYSTOLIC ANTENNA ARRAY. - Google Patents

Abstract

System and method of receiving a satellite signal by means of an array of modular systolic antennas, by means of which an output signal is obtained from the array of antennas and (t), the array being formed by N identical Ai antennas, affecting each antenna Ai an input signal xi (t) and wherein said method comprises the use of a plurality of correlators (1) of the input signal.

Description

System and method of receiving a signal from satellite through an array of modular systolic antennas.

Field of the Invention

The present invention is encompassed within the field of satellite telecommunications, and more specifically, to Satellite reception systems. The present invention proposes the use of an optimized array of antennas as receiving system

Background of the invention

As a general rule, reception systems by satellite use a single adjustable antenna that serves as a sensor for the demodulation and decoding of the received signals. To the have only one antenna, the sensitivity of the system depends almost completely the size of the antenna used, and that for certain applications can become extremely large with the consequent maintenance and cost involved.

In order to reduce the costs of use of a single antenna, the use of a array of antennas with a smaller aperture and combine the signals at your exit to create a signal equivalent to the one received by a large aperture antenna.

Because the aperture of the array antennas is smaller, the quality of the signals at its output (measured in terms of SNR - Signal to Noise Ratio ) is considerably reduced, reaching values of -10 dB and making its combination an extremely complicated process.

Techniques are used to combine the signals correlation that calculate and correct signal mismatches at the exit of the sensors weighing them by a few pesos before perform the combination The calculation of weighting weights at from the calculation of the eigenvalues of the self-covariance matrix of the input signals you get good performance for very low SNR ratios

The use of antenna array is known according to a monolithic architecture, in which the array of antennas is Process as a whole. The use of a monolithic architecture it implies a series of problems that will be discussed later in detail. However, in the present invention a systolic architecture or topology for the array of arrays with a high number of antennas, the array of antennas as a whole, but the array of antennas is subdivided into basic correlation units that interconnect with each other concentrating all the inputs in a single output, solving the implementation problems posed by the architecture monolithic

Description of the invention

To alleviate the problems mentioned above, presents the method of receiving a satellite signal by array of antennas, object of the present invention. Through this method you get an output signal from the antenna array and (t), the array being formed by N antennas A_ {i} identical, with each antenna A_ {i} having an input signal x_ {i} (t). Said method comprises the following steps:

1st.-

Group the N input signals x_ {i} (t) in subgroups of correlator input signals c_ {i} (t);

2nd.-

Get for each of the subgroups of correlator input signals c_ {i} (t), using a different correlator for each subgroup, a signal of output of correlator z (t) by performing the following operation, z (t) = \ sum \ limits ^ {M} _ {i = 1} w * {i} c_ {i} (t), w_ {i} * being the conjugate complex of the weighting weights w_ {i} previously obtained for each corresponding correlator to correct the lags of their correlator input signals c_ {i} (t), and M being the corresponding number of entries in each correlator;

3rd.-

Group the previous signs of output of correlator z (t) of the different correlators in at least one new subgroup of correlator input signals c_ {i} (t);

4th.-

Get for each of the at least one new subgroup of correlator input signals c_ {i} (t), using a different correlator for each new subgroup, a correlator output signal z (t) performing the following operation, z (t) = \ sum \ limits ^ {M} _ {i} w ^ {*} _ {i} c_ {i} (t), where w_ {i} ^ {*} the conjugate complex of the weighting weights w_ {i} obtained previously for each corresponding correlator to correct the Offsets of its correlator input signals c_ {i} (t), and M being the corresponding number of entries in each runner

5th.-

In the case that in the previous stage there is more than one output signal of correlator z (t), repeat steps b-e; in case in the previous stage there is only one output signal from correlator z (t), obtain the output signal and (t) by y (t) = z (t), where z (t) is said only correlator output signal.

The calculation procedure for each unit Basic UBC correlation of weights and obtaining the signal of output and (t) is as follows:

1º.- All the samples are synchronously sampled input signals or antennas (N), from i = 1 to i = L, and a Record of L samples, from j = 1 to j = L:

one

       \ vskip1.000000 \ baselineskip
    

2º.- The optimal weights of the array are obtained, \ vec {v}, by calculating the eigenvalues of the matrix of autovariance of X ^ (k)}, where the calculation of the weights is performed by means of an algorithm of extraction of the dominant autovector like it can be the power method or the matrix-free.

3º.- Weights are normalized \ vec {v} with respect of the first sensor, obtaining the vector:

2

       \ vskip1.000000 \ baselineskip
    

4th.- The final weight is obtained by integrating compared to the previous one, with an integration constant \ Gamma = 0.9:

3

and you get the combined output as Y (k) = w H (k) X (k).

The reception system of a signal by satellite by array of antennas that implements the method above includes at least:

A plurality of A_ {i} antennas;

A plurality of basic units of UBC correlation;

And where said basic correlation units configure a plurality of interconnected and scaled arrays each.

Brief description of the drawings

Then it goes on to describe very brief a series of drawings that help to better understand the invention and that expressly relate to an embodiment of said invention presented as a non-limiting example of is.

Figure 1 shows, according to the state of the art, the monolithic architecture used for arrays of antennas

Figures 2A and 2B show, according to the state of the art, the results obtained in a simulation for a monolithic architecture according to different algorithms used.

Figure 3 shows, according to the state of the art, the treatment of signals in a monolithic architecture.

Figure 4 represents the scheme of an example of a basic correlation unit of four inputs and one output, used in the systolic architecture object of the present invention Figure 4B shows a simulation of the combination gain for said basic unit of correlation.

The combination gain obtained in a simulation using the Power Method algorithm for different configurations of the basic correlation units is shown in Figure 5.

Different figures are shown in Figures 6A and 6B proposed systolic topologies for an array of 16 antennas inputs and 64 entries, using basic units of correlation of four inputs and one output.

Figure 7 compares the earnings of combination obtained by simulation in monolithic architectures of 16 and 64 antenna arrays with 16 systolic architectures and 64 antennas of figures 6A and 6B.

Description of a preferred embodiment of the invention

As shown in Figure 1, it is considered an array formed by N identical antennas that receive a signal coming from satellite and contaminated by Gaussian noise so Independent. In the first instance, it is assumed that all digitization processes and signal processors use the same source of synchronism. In Figure 1 \ theta_ {d} ago reference to the angle of arrival of the signal.

If x_ {i} (t) are the incident signals in each of the sensors or antennas A_ {i} and w_ {i} are the weights calculated by a correlator 1, the signal of output and (t) is obtained as:

4

as shown in Figure 1 through the logical operators product (x) and sum (+). Where w_ {i} * indicates the conjugate complex of weights w_ {i}.

Assuming that the input signals can be break down like

5

where s_ {i} (t), l_ {i} (t) and n_ {i} (t) correspond to the components of wanted signal, interfering and noise respectively, the signal of output can be obtained using matrix notation how:

6

and the signal ratio / (noise interference), SINR ( Signal-to-Interference-plus-Noise Ratio ) to the array output is calculated as:

7

where P_ {d}, P_ {l} and P_ {n} are the output powers of the array corresponding to the Desired signal components, interference and noise respectively of the input signals. These signals being defined as the average value of the squares of the output signals corresponding:

8

The optimal weights W_ {OPT} will be those that Maximize the SINR relationship. To calculate the optimal weights We assume that the array works in a narrow band. In these circumstances:

9

       \ vskip1.000000 \ baselineskip
    

where U_ {d} is the pointing vector ( Steering vector ):

10

Where \ tau_ {di} is the delay of the front of wave of the desired signal to reach the sensor or antenna A_ {i} with respect to the sensor or antenna A_ {1} and s (t) the component of desired signal.

In these circumstances, the output signal of the array and its power level are worth:

eleven

Where R is the self-covariance matrix of the desired signals.

Given that the matrix R can be break down like

12

Where \ Lambda = diag (\ lambda_ {1}, \ lambda_ {2}, ..., \ lambda_ {N}) containing the ordered eigenvalues \ lambda_ {1} \ geq \ lambda_ {2} \ geq ... \ geq \ lambda_ {N}, and where V = [\ vec {v} _ {1} \ vec {v} _ {2} ...
\ vec {v} _ {N}] {T} are R. orthonormal autovectors.

In turn, the array of autovectors V can be decomposed into V = [V_ {s} V_ {n}], where the columns of V_ {s} correspond to the eigenvalues of R of greater eigenvalue (Subspace signal) and where the other V_ {n} autovectors correspond to noise subspace:

13

Decomposing W and recalculating:

14

Then the weights that maximize the power of array combination output correspond to the autovector of the greater self-value of the self-covariance matrix. The procedure To find the dominant self-value is as follows:

?

Get the array of self-covariance R of the input signals.

?

Solve the problem of eigenvalues R \ vec {v} = \ lambda \ vec {v} and sort the eigenvalues \ lambda_ {1} \ geq \ lambda_ {2} \ geq ... \ geq \ lambda_ {N}

?

Make W_ {opt} = \ vec {v} 1

However, the procedure outlined is computationally expensive because you get all the eigenvalues while only the greatest of all is needed they.

There are alternative algorithms that only calculate the autovector of the dominant autovalue and therefore considerably reduce the computational load. A relationship of The best known algorithms is as follows:

fifteen

Where N is the number of antennas and L is the number of signal points used where K is the number of iterations

The results, obtained in simulations, using the Power Method and Matrix-Free algorithms are similar. A detail of the results of combination gain and loss for an array of 2, 4 and 8 sensors are those presented in Figures 2A and 2B, respectively, where the conventional algorithm is represented in dashed lines and the Power Method algorithm is shown in continuous lines. .

There are other algorithms ( Lagrange Mutiplier Method, Linear Lagrange Multipler Method, Linear Modified Conjugate Gradient Method ) that, not being as well known computationally, are very efficient although they have a high practical implementation difficulty in FPGAs ( Field Programmable Gate Array ).

The calculation of the weights obtained through the eigenvalues procedure meets the specifications requested, since no knowledge of the type of signal or array topology used, in addition to being insensitive to calibration mismatches of phase of the sensors or antennas.

A summary of the main advantages and drawbacks of calculating weights for eigenvalues is the next:

-

Advantages:

?

Maximize the SNR at the exit of array

?

They get the optimal weights regardless of the topology of the array.

?

They work with SNR very low.

-

Disadvantages:

?

Unable to cancel interference

?

They have a computational load very high

Keep in mind that each of the dominant eigenvalues of the self-covariance matrix corresponds to an input signal source, so if a calculation of all eigenvalues could be identified and tuned the array to each of the signals of interest.

Array tuning could be done. introducing the eigenvectors of the dominant eigenvalues corresponding to the directions of the signals of interest.

The procedure for calculating the weights of Weighting w, of an array depends on its architecture. Are defined Two types of architectures: monolithic and systolic.

In monolithic architecture, already known in the state of the art, the array is processed as a whole, as it is shown in Figure 3.

In monolithic architecture the calculation of eigenvalues are performed by processing the signals of all sensors at the same time, for this the correlator 1 must be sized in memory and computational load.

If X = [x_ {1} x_ {2} x_ {3} ... x_ {N}] T {are the input signals and W = [w_ {1} w_ {2} w_ {3} ... w_ {N}] T the weights obtained by the correlator 1, the output signal is obtained as Y = W * T} X.

The implementation of the algorithms of eigenvalues with monolithic architecture has the following problems:

?

Difficulty in sizing and interconnection of processors.

?

The broker is made ad-hoc for a given topology.

?

 Computing capacity and necessary memory grows exponentially with the number tickets.

?

It is necessary to use hardware multiprocessor to distribute the computation of algorithms.

Due to the exposed problems, the use of an architecture that has been called systolic is proposed. In the systolic architecture the array is subdivided into Basic Correlation Units (UBC) that interconnect with each other to concentrate all the inputs in a single output. The characteristics of the UBC are the following:

?

Autonomous

?

Nxl configuration

?

They are implemented in a single FPGA or DSP (Digital Signal Processor).

Although formally the number of entries of a UBC can take any value, from a practical point of view the configurations (2x1, 4x1 and 8x1) are those used because They can be easily implemented in a single FPGA / DSP.

       \ newpage
    

A detail of a UBC in 4x1 configuration and of its combination gain is that represented in Figures 4A and 4B, respectively.

It is interesting to note that for SNR <-15 dB better results are obtained in the combination gain using the calculation methods of the dominant autovalue (shown with circles) than with the conventional method that calculates all eigenvalues (shown with crosses).

The combination gain results have been obtained for a QPSK ( Quadrature Phase Shift Keying ) signal with a raised cosine filter factor of 0.6 with an integration time of 1600 symbols and an oversampling of 5 samples per symbol.

The combination gain for the case of UBC in 2x1 configuration (shown with circles), 4x1 (shown with asterisks) and 8x1 (shown with crosses) is shown in Figure 5, using the Power Method algorithm.

Examples of implementation of arrays 16 and 64 entries with 4x1 UBCs are shown in Figures 6A and 6B, respectively. The array of 16 sensors is built with in one 16x4x1 two-layer systolic configuration and 64 with a 64x16x4x1 configuration of three layers in which in each UBC_ {k, n} the subscript k identifies the layer number and the subscript n the UBC within it.

The profits of the systolic arrays of the proposed configurations with respect to the monolithic ones shown in Figure 7 (for simulations with 1600 symbols and taking 5 samples per symbol). The dashed line shows the monolithic architecture, while online continues the systolic architecture (in circles 16 entries, with crosses 64 tickets).

From the graphs it can be seen that with the monolithic architecture achieves greater gain for SNRs below -10 dB but at the cost of greatly increasing the complexity of the hardware and software of the array, however, with the proposed systolic topology the desired gains are achieved for signals of interest (SNR = -10) with a reasonable cost.

Using the proposed systolic architecture The following technical advantages are obtained:

?

The topology is expandable. The Introduction of new sensors basically consists of adding e interconnect new UBCs, without having to modify the operation of none of the pre-existing modules.

?

The architecture is insensitive to the physical situation of the sensors and does not need calibration. The array weights are obtained by the calculation procedure of eigenvalues

The procedure for calculating weights for A UBC unit is described below. Be the signal x_ {i} (t) the incident signal on the UBC sensor i of L sensors, the procedure for calculating their weights and obtaining of the output signal and (t) is as follows:

1.- All the samples are synchronously sampled input signals or antennas (N), from i = 1 to i = L, and a record of L samples, from j = 1 to j = L.

16

2.- The optimal weights of the array are obtained, \ vec {v}, by calculating the eigenvalues of the matrix of autovariance of X ^ (k)}. Weights calculation is done by means of one of the exposed algorithms (Power Method or Matrix-Free).

3.- Weights \ vec {v} are normalized of the first sensor.

17

4.- The final weight is obtained by integrating with respect to the previous one with the integration constant \ Gamma = 0.9

18

5.- The combined output is obtained as:

19

To get the array factor of an array systolic of K layers, it is necessary to obtain the weights of the array monolithic equivalent.

twenty

The equivalent array is calculated by obtaining the G_ {k} transfer functions of each layer k. The function of G_ {k} transfer is defined as the grouping of the weights of the UBC of each layer in a single vector. Each function of transfer G_ {k} combines the input signals of layer k in a small number of departures. The succession of layers gives the array an inverted pyramid structure that ends in a single exit. The equivalent array transfer function is will calculate as:

twenty-one

Claims (6)

1. Method of receiving a satellite signal by means of an array of antennas, by means of which an output signal is obtained from the array of antennas and (t), the array being formed by N identical A_ {i} antennas, affecting each antenna A_ {i} an input signal x_ {i} (t) characterized in that it comprises the following steps:

group the N input signals x_ {i} (t) in subgroups of signals of correlator input c_ {i} (t);

get for each of the subgroups of correlator input signals c_ {i} (t), using a different correlator for each subgroup, a correlator output signal z (t) effecting the next operation, z (t) = \ sum \ limits ^ {M} _ {i = 1} w_ {i} ^ {*} c_ {i} (t), w_ {i} ^ {*} being the conjugate complex of the weighting weights w_ {i} previously obtained for each corresponding correlator to correct the lags of their correlator input signals c_ {i} (t), and M being the corresponding number of entries in each correlator;

group the previous output signals of correlator z (t) of the different correlators in at least one new subset of signals from correlator input c_ {i} (t);

get for each of the at least one new subset of input signals from correlator c_ {i} (t), using a different correlator for each new subgroup, a correlator output signal z (t) performing the following operation, z (t) = \ sum \ limits ^ {M} _ {i} w ^ {*} _ {i} c_ {i} (t), where w_ {i} ^ {*} the conjugate complex of the weighting weights w_ {i} obtained previously for each corresponding correlator to correct the Offsets of its correlator input signals c_ {i} (t), and M being the corresponding number of entries in each runner

if in the previous stage there is more than one correlator output signal z (t), repeat steps b-e; if in the previous stage there is a single correlator output signal z (t), obtain the output signal and (t) by y (t) = z (t), where z (t) is said single signal of correlator output

2. Method of receiving a satellite signal by means of antenna array, according to claim one, characterized in that the calculation procedure for each basic unit of UBC correlation of the weights and obtaining the output signal and (t) is as follows:

they are sampled from synchronously all input signals or antennas (N), from i = 1 to i = L, and a record of L samples is made, from j = 1 to j = L:

22

you get the optimal weights of the array, \ vec {v}, by calculating the eigenvalues of the autovariance matrix of X (k);

normalize the weights \ vec {v} with respect to the first sensor, obtaining the vector

2. 3

you get the final weight integrating with respect to the previous one:

100

and you get the combined output as Y (K) = W H (k)} X (k).

3. Method of receiving a satellite signal by means of an array of antennas, according to claim two, characterized in that the calculation of the weights is carried out by means of an algorithm for extracting the dominant autovector from the self-covariance matrix. 4. Method of receiving a satellite signal by means of an array of antennas, according to claim 3, characterized in that the algorithm for extracting the dominant autovector from the self-covariance matrix is one selected from: - power method method ; - matrix-free method. 5. Method of receiving a satellite signal by means of an array of antennas, according to claim two, characterized in that the final weight is obtained by integrating with respect to the previous one with an integration constant γ = 0.9. 6. System for receiving a satellite signal by means of an array of antennas that implements a method according to claims 1 to 5, characterized in that it comprises at least:

a plurality of A_ {i} antennas;

a plurality of basic units of UBC correlation;

where said basic correlation units configure a plurality of arrays interconnected and scaled to each other.