KR102287791B1 - An Enhanced Jacobi Precoder for Downlink Massive MIMO Systems - Google Patents

Abstract

The present invention relates to a method of calculating a weight for increasing estimation accuracy by proposing a new form of the JC technique among approximation techniques for reducing the complexity of the linear ZF technique in a large-capacity MIMO system.
According to the present invention, it is possible to increase the convergence speed of JC through the preconditioning technique. Also, will the number of users increase? It is possible to secure the shortcomings of JC, in which the performance rapidly decreases. By reducing the number of iterations that satisfy the target BER, the overall complexity of the ZF precoding technique can be greatly reduced.
The JC method to which the weights of the present invention are applied has better error performance with the same computational complexity as the existing IM method capable of parallel execution. As a result, the JC technique proposed in the present invention requires almost the same complexity as the existing IM technique capable of parallel execution, and has the advantage of improving error performance compared to the existing technique. In addition, it can be seen through [Fig. 3] that a complexity lower than that of the ZF technique requiring a perfect inverse matrix is required.

Description

An Enhanced Jacobi Precoder for Downlink Massive MIMO Systems

The present invention relates to an efficient precoder based on the Jacobi algorithm and a method therefor in a large-capacity MIMO system, and more particularly, to an efficient precoder and method for improving error performance in a downlink large-capacity MIMO system.

MIMO systems are widely used in wireless communication. Massive MIMO, an advanced technology from MIMO, is a technology in which a base station uses hundreds of antennas and transmits data to many users. Thanks to the use of hundreds of antennas at the base station, the high-capacity MIMO system has higher spectral and energy efficiency than conventional systems. In a downlink system, in order to transmit data to multiple users, the base station must perform precoding in advance to remove Inter User Interference (IUI). Since the large-capacity MIMO channel property has asymptotic orthogonality in a single-cell environment, the zero forcing (ZF) technique has optimal performance.

However, in order to implement ZF, it is necessary to obtain a pseudo inverse matrix of the channel, and the computational complexity of this process is proportional to the cube of the number of users. Therefore, in a large-capacity MIMO system that provides services to many users, ZF requires high computational complexity. To solve this problem, a method for approximating a linear expression using IM (Iteration Method) has been developed, and it can satisfy the error performance equivalent to ZF with low complexity. Among them, the Jacobi (JC) method has low complexity, parallel processing, and performance comparable to ZF.

However, even in this JC technique, as the ratio of the number of users to the number of base station antennas increases, the error performance rapidly deteriorates, and when the ratio exceeds a certain ratio, signal estimation is seriously impossible.

Korea Patent Publication No. 10-0950245 (2010.03.23.)

An object of the present invention is to provide an efficient JC-based precoder in a large-capacity MIMO system by increasing the estimation accuracy of an approximation technique in preparation for deterioration in error performance as the number of users and the number of base station antennas exceeds a certain ratio.

The present invention proceeds by designing a precoder based on the JC technique, which is an iterative algorithm capable of parallel processing in order to reduce the high complexity of the linear ZF precoding technique based on the downlink large-capacity MIMO system.

In order to increase the convergence speed of JC, the condition number of the linear equation that is inversely proportional to the convergence speed of the IM (iteration method) and the precondition, a method of lowering the spectral radius of the iteration matrix multiplied by each IM (precondition) algorithm is used. Based on Richardson (RI), the precondition matrix is obtained through polynomial expansion (PE), multiplied by a linear expression, and then a linear approximation solution is obtained based on JC.

)의 추정 역행렬(W^-1)을 구하는 것을 목적으로 하며 PE를 이용하여 구한 프리컨디션 행렬은 아래 수학식으로 정의될 수 있다.In the present invention, the precondition matrix is a W matrix calculated based on the channel matrix (H) between the base station and the user (

) for the purpose of obtaining the estimated inverse matrix (W ^-1 ), and the precondition matrix obtained using PE may be defined by the following equation.

는 이완 매개 변수로, RI를 기반으로

(N_t는 기지국 안테나 수, K는 사용자 수), D는 W의 대각 행렬 이다.here,

is the relaxation parameter, based on RI

(N _t is the number of base station antennas, K is the number of users), D is a diagonal matrix of W.

According to the present invention, in order to increase the convergence speed of the approximation technique, if the JC technique is applied to the linear expression multiplied by the precondition matrix using the preconditioning method, the error performance can be increased by increasing the convergence speed compared to the conventional JC technique, and the user Performance can be achieved even when a certain ratio of the number and number of base station antennas is exceeded, and performance can be improved compared to other parallel processing IM methods.

The present invention reduces the number of repetitions of the JC technique required to obtain performance similar to that of the conventional ZF precoding, and has the advantage that the JC technique can be applied even in an environment with a large number of users.

[Figure 1] shows a downlink high-capacity MIMO system transmission/reception model.
[Figure 2] shows a simplified flowchart to help the accurate understanding of the process of the technique presented in the present invention.
[Fig. 3] shows the difference in complexity between the scheme presented in the present invention and the existing IM scheme in a table.
[Fig. 4] shows the precondition matrix supplemented by adding diagonal components to the first and second terms of the repetition matrix of the IM and the number of conditions in the matrix multiplied by W according to the number of users.
[Fig. 5] shows the spectrum radius of the repetition matrix multiplied by the IM according to the number of users.
[Fig. 6] and [Fig. 7] are graphs showing the difference in BER performance according to SNR between the existing IM techniques and the technique presented in the present invention for different modulation orders.

Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings so that those of ordinary skill in the art can easily carry out the present invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. And in order to clearly explain the present invention in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

With reference to the accompanying drawings, embodiments of the present invention will be described in detail so that those of ordinary skill in the art to which the present invention pertains can easily implement them.

1. Digital precoder unit

Since a plurality of transmitters and receivers exist in a MIMO system, Inter User Interference (IUI) occurs. In order to remove such interference, the transmitter performs precoding on the transmit signal through the precoder unit so that the signals do not interfere with each other. In the present invention, the transmitting unit is the base station 120 , and the receiving unit is the user 130 having a single antenna.

)1 illustrates a downlink high-capacity MIMO system transmission/reception model. K data streams 110 are transmitted to the base station 120 that transmits the signal, the base station 120 has N _t transmit antennas, and transmits the data streams to the K users 130 through the antennas. (

)

At this time, the transmission signal is precoded through the digital precoder of the base station. For precoding, it is assumed that the base station knows all the channel information (spatial channel) of each user.

The present invention relates to a process of generating and precoding a precondition matrix in a digital precoder installed in a base station.

1.1. Prior art ZF precoding technique

The received signal vector received by the user is expressed as Equation (1).

(K행,1열) 크기의 벡터로 나타나진다. 이때, P는 기지국의 송신 전력, H는

크기의 사용자의 채널 정보를 담고 있는 채널 행렬, s는

크기의 프리코딩 된 송신 신호 벡터, n은 평균이 0이고 분산이 1인

크기의AWGN(Additive White Gaussian Noise) 벡터를 나타낸다.The received signal vector (y) is

It is expressed as a vector of size (K rows, 1 column). At this time, P is the transmission power of the base station, H is

Channel matrix containing the user's channel information of size, s is

A precoded transmit signal vector of magnitude, n, with mean 0 and variance 1

Represents an AWGN (Additive White Gaussian Noise) vector of magnitude.

Signal interference can be removed by multiplying a transmission signal by a pseudo-inverse matrix by a channel matrix through Equation (1). Accordingly, the transmission signal vector precoded in the ZF technique can be expressed as Equation 2 below.

는 에르미트(Hermitian) 변환을 나타내는 것이다. 에르미트 변환행렬은, 기존의 행렬 H의 켤레 전치 행렬이다. 즉, 기존의 행렬 H의 각 원소에 대하여 켤레 값을 가지며, 본래의 행렬의 행과 열을 바꿔 전치 한 것이다. here,

is a Hermitian transformation. The Hermitian transformation matrix is a conjugate transpose matrix of the existing matrix H. That is, it has a conjugate value for each element of the existing matrix H, and transposes the rows and columns of the original matrix.

크기의 심볼(Symbol) 벡터, W는 추정행렬로,

크기의 행렬이며

로 정의되고,

는

로 송신 전력을 제한하는 스케일링 팩터(scailing factor)이며 tr(W^-1)의 값을 가진다.

는 기댓값이고,

는 행렬 대각 성분의 총 합을 나타낸다. x has mean 0 and variance 1

Symbol vector of size, W is the estimation matrix,

is a matrix of size

is defined as

Is

It is a scaling factor that limits the transmit power to , and has a value ^{of tr(W -1 ).}

is the expected value,

is the sum of the matrix diagonal components.

는 점근 표기법(big O)를 나타내며 알고리즘의 복잡도를 간소화하기 위해 사용된다.In the downlink, a transmission symbol (x) is precoded and transmitted as in Equation 2 using the channel defined as above for transmission and reception. However, in the case of a large-capacity MIMO system, as the number of acceptable users (K ) increases, the computational complexity of ^{the accurate estimation inverse matrix (W −1} ) required in the ZF technique is high as O(K ³ ). here,

denotes asymptotic notation (big O) and is used to simplify the complexity of the algorithm.

는 점근 표기법(big O)를 나타내며 알고리즘의 복잡도를 간소화하기 위해 사용된다.)In the downlink, a transmission symbol (x) is precoded as in Equation 2 using the channel defined as above for transmission and reception, and then transmitted. In the ZF precoding technique, it is necessary to obtain an estimation inverse matrix W ^-1 for precoding. As the number of users (K) increases, the computational complexity ^{of the accurate estimation inverse matrix (W -1} ) is high as O(K ^{3 ).} (here,

stands for asymptotic notation (big O) and is used to simplify the complexity of the algorithm.)

Therefore, in order to reduce the high complexity of the accurate inverse matrix operation, an iterative method (IM) technique that approximates a vector multiplied by a linear inverse matrix is used without obtaining an estimated inverse matrix.

를 유지하는 IM(iterative method)의 기법에서 공통적으로 표현되는 수식이다. x는

의 선형 방정식을 해결하기 위해 심볼(Symbol) 벡터(x)를 반복적인 알고리즘을 통해 근사하는 것이며, s_n는 송신 신호 벡터이며, n는 반복횟수를 나타내고, M은 비 특이 행렬이다. 이때, 상기 추정행렬

는 M 과 N으로 나타내질 수 있으며, M과 N은 분해 행렬이라하며, 분해 행렬을 설정하는 방법에 따라 IM(iterative method)기법이 달라진다.Equation 3 shows that the complexity of any iteration is

It is a formula commonly expressed in the IM (iterative method) technique that maintains . x is

In intended to cool through an iterative algorithm a symbol (Symbol) vectors (x) In order to solve the linear equations, _n s is the transmitted signal vector, n denotes the number of iterations, M is a non-specific matrix. In this case, the estimation matrix

can be expressed as M and N, where M and N are called decomposition matrices, and the iterative method (IM) technique is different depending on how the decomposition matrix is set.

When Equation 3 is _{expressed with respect to s n} , it is expressed as Equation 4 below.

은 반복적으로 곱해지는 반복 행렬로 정의한다. IM 기법들은

(K는 사용자 수)로 수렴하기 위해

의 조건을 만족해야 하며,

는 행렬의 스펙트럼 반지름(spectral radius)를 나타낸다. 실수 또는 복소수로 이루어진 행렬은 고유 값을 가지는데, 스펙트럼 반지름(spectral radius)란, 행렬의 고유 값의 최대값을 의미한다. 이때, 반복 행렬 B를

(I는 단위행렬)로 표현할 수 있다. here,

is defined as an iterative matrix that is iteratively multiplied. IM techniques are

(K is the number of users) to converge to

must satisfy the conditions of

denotes the spectral radius of the matrix. A matrix made of real or complex numbers has eigenvalues, and the spectral radius means the maximum value of the eigenvalues of the matrix. In this case, the iteration matrix B is

( I is the identity matrix).

The JC technique is a system of linear equation solving that guarantees the convergence of the iterative method in a system of linear equations composed of a strong diagonal dominant matrix. In the JC technique, a specific matrix is expressed by decomposing it into a diagonal matrix and a matrix other than the diagonal matrix. As described above, the IM technique varies according to the set values of the decomposition matrices M and N. The decomposition matrices of the JC technique are (M _JC ) and (N _JC ), and are expressed as Equation 5. In the JC technique, in order to obtain a transmission signal vector, the decomposition matrix is expressed as Equation 6, and the following Equation 6 is repeatedly performed. save

In addition, D represented in Equation 5 is a diagonal matrix of W, L is a strictly lower triangular of W, and U is a strictly upper triangular matrix of W.

Equation 6 is a conventional JC technique, where s _n is a transmission signal vector when the n-th repetition is performed, s _n-1 is a transmission signal vector when the n-1 repetition is performed, and x is a transmission symbol.

The conventional JC technique has low complexity and parallel processing is possible, but as the number of users increases, communication deteriorates and algorithm performance deteriorates.

2. Invention of the present invention

Accordingly, the present invention proposes a new type of JC technique that satisfies the convergence condition and increases the convergence speed of the approximation technique in order to alleviate the rapid performance degradation as the number of users increases. [Figure 2] shows a flowchart of the present invention. As shown in [Fig. 2], the proposed JC technique is largely composed of steps of obtaining a precoding matrix through a polynomial expansion (PE) technique (220), and applying the precoding matrix to the JC technique (230).

2.1. Estimation matrix generator

를 생성하는 단계를 수행하는 장치이다. When the channel of each user is known ( 210 ), the estimation matrix unit generates a channel matrix H generated based on the channel, and based on the channel matrix, the estimation matrix

It is a device that performs the steps of generating

2.2. decomposition matrix generator

As described above, the decomposition matrix generator is a device that performs the decomposition matrix generation step of setting the decomposition matrices M, N of the estimation matrix W in order to determine the linearized IM technique regardless of the number of users.

In the present invention, the precondition matrix is obtained using a polynomial expansion (PE) method, which is a method for obtaining a similar estimated inverse matrix. The estimation matrix W follows the above-described prior art method.

The pseudo inverse matrix obtained according to the PE method is called the estimated inverse matrix, and is expressed as [Equation 7] below.

을 만족하게 선택해야 한다. In order for the right side of Equation 7 to converge on the estimated inverse matrix, M, N,

should be selected satisfactorily.

은 반복적으로 곱해지는 반복 행렬로 정의하면,

의 조건을 만족해야 하며,

는 행렬의 스펙트럼 반지름(spectral radius)를 나타낸다.As seen in Equation 4,

is defined as an iterative matrix that is repeatedly multiplied,

must satisfy the conditions of

denotes the spectral radius of the matrix.

In the present invention, M and N are selected based on the RI (Richardson) iteration method to increase the convergence speed. M,N based on the RI (Richardson) iteration method is expressed as [Equation 8] below.

는 이완 매개 변수(relaxation parameter)로 RI(Richardson) 반복법 기반에서는

로 설정한다. here,

is a relaxation parameter, and based on the RI (Richardson) iteration method,

set to

2.3. Precondition matrix generator

The precondition matrix generator is a device that performs the precondition matrix generation step of generating a precondition matrix for precoding a transmission signal.

In the present invention, using [Equation 8], the precondition matrix is proposed as shown in [Equation 9] below. By applying [Equation 8] to [Equation 7], the precondition matrix P _c can be obtained as in [Equation 9].

In [Equation 9], P _c is a precondition matrix, W is an estimation matrix, I is an identity matrix, and D is a diagonal matrix of W. In the present invention, in ^{order not to exceed the computational complexity O(K 3} ), PE is expanded to the first order term, and in order to more closely approximate the inverse of the W matrix within a complexity lower than ZF, in the present invention, in the present invention, diagonal dominant in a large MIMO environment. Using the property of W, a quadratic part that complements only the diagonal component was added.

2.4. Estimation matrix transformation part

In the present invention, the estimation matrix transformation unit is an apparatus for performing the estimation matrix transformation step of transforming the existing estimation matrix W in order to reduce the computational complexity.

의 양변에 프리컨디션 행렬을 곱한다. 곱해진 선형 방정식은 아래 [수학식 10] 과 같이 표현 된다.After calculating the precondition matrix as in [Equation 9], to increase the convergence speed of the JC algorithm by reducing the condition number of the linear equation, the linear equation

Multiply both sides by the precondition matrix. The multiplied linear equation is expressed as [Equation 10] below.

Since [Equation 10] calculates matrix-matrix multiplication, the computational complexity becomes O(K ³ ). Therefore, in the present invention, the expression is modified to reduce the computational complexity. First, in order to apply the JC method, [Equation 10] is transformed as shown in [Equation 11] below.

는 변형된 추정행렬이라 하며,

,

는

를 의미한다. [수학식 11]에 JC를 적용하기 위해 변형된 추정행렬

를 아래 [수학식 12]와 같이 분해한다.here,

is called the transformed estimation matrix,

,

Is

means Estimation matrix transformed to apply JC to [Equation 11]

is decomposed as in [Equation 12] below.

는

의 대각 행렬(diagonal matrix),

는

의 엄격한 하삼각 행렬(strict lower matrix),

는

의 엄격한 상삼각 행렬(strict uppper matrix)이다. 즉, 추정행렬 W에 본원 발명에서의 프리컨디션 행렬을 곱한

이며, [수학식 12]이다.here,

Is

The diagonal matrix of

Is

Strict lower triangular matrix of

Is

is a strict upper triangular matrix of . That is, the estimation matrix W is multiplied by the precondition matrix in the present invention.

and [Equation 12].

2.5. Deformation estimation matrix JC method application part

The transformed estimation matrix JC technique applying unit is a device for performing the transforming estimation matrix JC technique application step of applying the transformed estimation matrix generated by the estimation matrix transforming unit to the JC technique.

와 변형된

를 JC 기법에 적용하여, 하기 수학식 13과 같이 나타낼 수 있다. The modified estimation matrix

and transformed

By applying to the JC technique, it can be expressed as Equation 13 below.

의 모든 성분을 구해야만 하기 때문에 계산 복잡도는 O(K³)을 요구하게 된다. 본 발명에서는 이러한 복잡도를 줄이기 위해,

의 필요한 성분들만 구하는 방법으로 [수학식 13]을 아래 [수학식 14]처럼 변형 한다.If we solve the expression of the JC method as in [Equation 13],

Since we have to find all components of , the computational complexity requires O(K ³ ). In the present invention, in order to reduce this complexity,

[Equation 13] is transformed as shown in [Equation 14] below in a way to obtain only the necessary components of .

의 성분 중 엄격한 하삼각 행렬(

) 성분 엄격한 상 삼각 행렬(

) 성분 는 구할 필요 없이,

의 대각 행렬 성분

만을 구하면 되기 때문에, 계산 복잡도는 O(K³)에서 O(K²)으로 감소된다. 초기값

은 아래 [수학식 15]와 같이 표현한다.If the expression is transformed into [Equation 14] as above,

Strict lower triangular matrix (

) component strict phase triangular matrix (

), there is no need to obtain

diagonal matrix component of

, the computational complexity is ^{reduced from O(K 3} ) to O(K ² ). initial value

is expressed as [Equation 15] below.

The above method applies the preconditioning method to JC to increase the convergence speed. It has higher BER (bit error rate) performance than the existing IM (Iterative method) method that supports parallel processing even within a similar complexity, and it can be estimated in an environment with a large number of users. Overcome the shortcomings of JC that cannot.

The computational complexity of JC presented in the present invention can be analyzed through [Equation 14].

의 대각성분

를 구하는 부분, P_c를 포함하여 JC알고리즘을 진행하는 부분들이 추가된다.In calculating the complexity, only the number of multiplication operations is considered, and the part that obtains the precondition matrix (P ) in the existing JC method,

diagonal component of

The parts that proceed with the JC algorithm are added, including the part to find P _{c .}

For low computational complexity, the precondition matrix (P ) is obtained by changing [Equation 9] as shown in [Equation 16] below.

가 요구된다.

를 구하는 복잡도는 행렬- 행렬 곱의 대각성분만 구하면 됨으로 K²가 요구된다. 초기값

은 처음 프리컨디션 행렬(P) 와 송신 벡터

를 곱한 후,

와

를 곱하여

의 계산 복잡도가 요구된다. [수학식 14]의 한번 마다 필요한 계산 복잡도는 차례로 W,P,

가 차례로 계산되는 벡터와 곱해짐으로 계산 복잡도 2K²+K가 요구된다. 마지막으로 구해진 s⁽ⁿ⁾을 ZF 프리코딩 절차인 H ^H s ⁽ⁿ⁾ 를 곱해주는 계산 복잡도는 N_tK가 요구된다. The complexity of finding the precondition matrix (P) is

is required

^{K 2} is required because only the diagonal component of the matrix-matrix product needs to be calculated for the complexity. initial value

is the initial precondition matrix (P) and the transmit vector

After multiplying by

Wow

multiply by

of computational complexity is required. The computational complexity required for each one of [Equation 14] is W, P,

is multiplied by a vector that is computed in turn, requiring a ^{computational complexity of 2K 2 +K.} The computational complexity of multiplying the finally obtained s ⁽ⁿ⁾ by the ZF precoding procedure, H ^H s ⁽ⁿ⁾ _{, requires N t} K.

Therefore, n iterations of the JC technique presented in the present invention require a computational complexity of (2n+3)K ² +(n+4)K+N _t K .

[Fig. 3] is a diagram showing the amount of multiplication operation required to calculate a precoded transmit vector in the conventional IM techniques and the JC technique proposed in the present invention.

[FIG. 4] The precondition matrix obtained up to the first term using PE, the precondition matrix supplemented by adding a diagonal component to the second term, and the number of conditions of the matrix multiplied by the gram matrix (W), respectively, are shown according to the number of users. The number of base station antennas (N _t ) is fixed to 100. Each matrix is expressed as [Equation 17] below.

The method of obtaining each condition number is expressed as [Equation 18] below.

It can be seen from [Fig. 4] that the number of conditions obtained by multiplying the gram matrix and the precondition matrix supplemented by adding a diagonal component to the quadratic term using PE (Polynomial Expansion) is smaller.

[Fig. 5] The spectrum radius of the repetition matrix (B) multiplied by the IM is shown according to the number of users. In [Fig. 5], it can be seen that the spectral radius does not significantly increase even when the number of users of the JC technique proposed in the present invention increases. In [FIG. 5], the number of base station antennas (N _t ) is fixed to 100. The repetition matrix of each IM is expressed as the following Equation [Equation 19]. In the following [Equation 19], B _JC is the iteration matrix of the conventional JC technique, B _RI is the iteration matrix of the RI (Richardson) iteration _{method, B DJ} is the iteration matrix of the damped JC technique, and B _PJC is the JC technique proposed in the present invention. is a repeating matrix of

In [Equation 19], the iteration matrix of JC (Jacobi) and RI (Richardson) iteration methods and damped Jacobi (DJC) iteration matrices were squared in order to compare at a similar complexity to the preconditioned Jacobi presented in the present invention. The spectral radius of each repetition matrix was obtained using the Frobenius norm of each repetition matrix.

<Example>

6 and 7 are diagrams comparing the performance of the technique of the present invention and the conventional technique according to the number of users, respectively. FIGS. 6 and 7 are signal error rates confirmed by the user 130 .

6 and 7 are BER plots. The x-axis is SNR and the y-axis is BER. BER indicates the probability that an error occurs according to the size of transmission power, and the ZF technique has the lowest error rate because the graph of the ZF technique is located at the bottom, and it can be confirmed that the JC technique proposed in the present invention has the next lowest error rate.

The conventional techniques are the ZF precoding technique using the existing JC technique, the ZF technique requiring an accurate inverse matrix, and the RI iteration technique and the damped JC technique, respectively. The number of base station antennas (N _T ) is fixed to 100. In all performance comparisons, the performance of the ZF technique is presented as an indicator.

In [Fig. 6a], when the number of users (K) is 30, [Fig. 6b] is the number of users (K) of 40, and [Fig. 6c] is the number of repetitions (n) when the number of users is 50. JC (Jacobi) technique in modulation order QPSK, It shows the performance between the RI (Richardson) iteration method and the proposed JC method (indicated as PJC in Fig. 6). It can be seen that JC does not perform well, and as the number of users (K) of the JC (Jacobi) technique presented in the present invention increases, the performance difference with RI increases. Also, it can be seen that even if the number of repetitions of RI increases, the performance improvement decreases as the number of users (K) increases in an environment with high SNR.

In [Fig. 7a], the number of repetitions (n) when the number of users (K) is 30, [Fig. 7b] is the number of users (K) is 40, and [Fig. 7b] is the number of users (K) of 50, modulation order 16QAM It shows the performance between JC (Jacobi) technique, damped JC (DJ), and the proposed JC technique (indicated as PJC in Fig. 7). It can be seen that the JC does not perform well, and it can be seen that the performance difference with the damped JC (DJ) increases as the number of users (K) of the technique presented in the present invention increases. In addition, even if the number of repetitions of the damped JC (DJ) increases, it can be seen that the performance improvement decreases as the number of users (K) increases in an environment with high SNR. The simulation results show that the preconditioning technique has a very large effect on the convergence speed. The method presented in the present invention can obtain higher error performance at the same complexity than the existing IM techniques required for the required performance. In addition, it can be seen that the BER performance of the ZF precoding technique, which requires an accurate inverse matrix, converges faster than that of the existing IM techniques, as the number of iterations increases.

On the other hand, although the technical idea of the present invention has been described in detail according to the above embodiment, it should be noted that the above embodiment is for the description and not the limitation. In addition, those skilled in the art will understand that various embodiments are possible within the scope of the technical spirit of the present invention.

110 data stream
Base station with 120 digital precoder
130 users
210 Channel Acquisition Steps
220 Precondition Matrix Generation Steps
230 Steps to create a modified JC scheme
240 Steps to perform precoding

Claims (10)

In a digital precoder of a base station of a large-capacity downlink MIMO system,
an estimation matrix generator for generating an estimation matrix W based on a channel matrix;
a decomposition matrix generator that decomposes the estimation matrix W into an M,N decomposition matrix and sets a decomposition matrix;
a precondition matrix generator for generating a _{precondition matrix P c} to which the decomposition matrix obtained through the decomposition matrix generator is applied;
By transforming the estimation matrix W by multiplying the estimation matrix W by the precondition P _{c , the transformed estimation matrix}

an estimation matrix transformation unit to generate;
Estimation matrix transformed through the estimation matrix transformation unit

is applied to the modified Jacobian technique (JC technique) to pre-code the transmission signal to generate a precoded signal;
The modified Jacobian technique applied in the modified JC technique application unit is,
It is characterized in that it is a JC technique modified so that the calculation complexity of the precoded transmission signal becomes a function of the square of the number of users K,
The decomposition matrix generator,
A digital precoder, characterized in that the decomposition matrix M, N is set according to Equation 1 below based on the RI (Richardson) iteration method.
(Formula 1)

(

, N _t is the number of base station antennas, K is the number of users, and W is the estimation matrix.

ego

is the Hermitian transform, M, N are nonsingular matrices)
The method of claim 1,
The precondition matrix generator,
The precondition matrix P _c expands polynomial expansion (PE) up to the first term, expands only the diagonal component up to the second term, and applies the decomposition matrix M,N to Equation 2 below.
(Equation 2)

(W is the estimated inverse matrix, W ^-1 is the pseudo-inverse of the estimation matrix, called the estimated inverse matrix. M,N are the decomposition matrix, and i is the iteration order)
3. The method of claim 2,
The precondition matrix generator,
Digital precoder, characterized in that for generating a precondition matrix as shown in Equation 3 below.
(Equation 3)

(P _c is the precondition matrix, I is the identity matrix,

,N _t is the number of base station antennas, K is the number of users, D is the strict diagonal matrix of the estimation matrix W)
4. The method of claim 3,
A digital precoder, characterized in that the precondition matrix as shown in Equation 3 is transformed as shown in Equation 4 below.
(Equation 4)

(P _c is the precondition matrix, I is the identity matrix,

,N _t is the number of base station antennas, K is the number of users, W is the estimation matrix, D is the strict diagonal matrix of the estimation matrix W)
The method of claim 1,
The digital precoder, characterized in that the transmission signal vector of the precoded signal is determined by Equation 5 below by applying the modified Jacobian technique.
(Equation 5)

(

is the n+1-th repeated transmission signal vector,

is the n-th repeated transmission signal vector,

is the transformed estimated inverse matrix

strict diagonal of

Inverse matrix of, W estimation matrix, P _c is the precondition matrix, x is the symbol vector x (linear equation x=Ws converging in the IM technique))
A method for digitally precoding a transmission signal in a base station of a large-capacity downlink MIMO system, the method comprising:
an estimation matrix generation step of generating an estimation matrix W based on a channel matrix;
a decomposition matrix generation step of decomposing the estimation matrix W into an M,N decomposition matrix and setting a decomposition matrix;
a precondition matrix generation step of generating a _{precondition matrix P c} to which the decomposition matrix obtained through the decomposition matrix step is applied;
By transforming the estimation matrix W by multiplying the estimation matrix W by the precondition P _{c , the transformed estimation matrix}

generating an estimation matrix transformation step;
Estimation matrix transformed through the step of transforming the estimation matrix

A modified JC method applying step of precoding a transmission signal by applying the modified Jacobian method (JC method) to generate a precoded signal;
The modified Jacobian technique applied in the step of applying the modified JC technique is,
It is characterized in that it is a JC technique modified so that the calculation complexity of the precoded transmission signal becomes a function of the square of the number of users K,
The decomposition matrix generation step is,
A method for obtaining a digital precoding matrix, characterized in that the decomposition matrix M, N is set according to Equation 1 below based on the RI (Richardson) iteration method.
(Formula 1)

(

, N _t is the number of base station antennas, K is the number of users, and W is the estimation matrix.

ego

is the Hermitian transform, M, N are nonsingular matrices)
7. The method of claim 6,
The precondition matrix generation step is,
The precondition matrix P _c is a digital precoding matrix, characterized in that polynomial expansion (PE) is expanded up to the first term, only the diagonal component is expanded up to the second term, and the decomposition matrix M, N is applied to the following Equation 2 How to get it.
(Equation 2)

(W is the estimated inverse matrix, W ^-1 is the pseudo-inverse of the estimation matrix, called the estimated inverse matrix. M,N are the decomposition matrix, and i is the iteration order)
8. The method of claim 7,
The precondition matrix generation step includes:
A method of obtaining a digital precoding matrix, characterized in that it generates a precondition matrix as shown in Equation 3 below.
(Equation 3)

(P _c is the precondition matrix, I is the identity matrix, W is the estimation matrix,

,N _t is the number of base station antennas, K is the number of users, D is the strict diagonal matrix of the estimation matrix W)
9. The method of claim 8,
A method for obtaining a digital precoding matrix, characterized in that the precondition matrix as shown in Equation 3 is transformed as shown in Equation 4 below.

(Equation 4)

(P _c is the precondition matrix, I is the identity matrix, W is the estimation matrix,

,N _t is the number of base station antennas, K is the number of users, W is the estimation matrix, D is the strict diagonal matrix of the estimation matrix W)
7. The method of claim 6,
A method for obtaining a digital precoding matrix, characterized in that the precoded signal is transformed as shown in Equation 5 below.
(Equation 5)

(

is the n+1-th repeated transmission signal vector,

is the n-th repeated transmission signal vector,

is the transformed estimated inverse matrix

strict diagonal of

Inverse matrix of, W estimation matrix, P _c is the precondition matrix, x is the symbol vector x (linear equation x=Ws converging in the IM technique))

Images