CN101841303A - Predistortion estimation method based on polynomial - Google Patents

Abstract

The invention relates to a predistortion estimation method based on polynomial. The method comprises the following steps: receiving the polynomial of a predistorter, converting the polynomial to represent using a matrix, obtaining a predistortion matrix composed of K-numbered predistortion functions, wherein K represents the maximum order of the polynomial; performing singular value decomposition to the predistortion matrix to obtain a singular value matrix; obtaining the effective rank of the predistortion matrix according to the singular value matrix, and intending the effective rank to be the effective order of the polynomial, wherein the effective rank of the predistortion matrix is the minimal rank of an approaching predistortion matrix and reaching the predetermined approaching effect on the predistortion matrix, and the approaching predistortion matrix is the matrix approaching the predistortion matrix. The invention ensures the predistortion performance while reducing the complexity of the power amplifier predistorter based on polynomial so as to ensure that the power amplifier predistorter based on polynomial has better convergence and stability.

Description

Based on polynomial predistortion estimation method

Technical field

The present invention relates to based on polynomial predistortion estimation method, particularly, based on effective rank method of estimation of polynomial predistorter.

Background technology

Power amplifier (PA) is one of requisite parts in the wireless communication system, and the raising of its effect is for Wireless Telecom Equipment, and especially the energy-saving consumption-reducing to portable mobile terminal is significant.Portable terminals such as the base station that radio communication operator adopted, repeater, radio and television launching tower and the employed mobile phone of user, mobile TV, Miniaturized Communications equipment all relate to the efficacy problems of power amplifier.

The distortion of the transmission signals that nonlinear characteristic caused of power amplifier is to cause one of major reason that the power amplifier effect reduces, and therefore, it is most important for the development of radio communication to improve its effect that power amplifier is carried out linearization process.

The linearization process of power amplifier generally comprises back-off technology, envelope elimination and recovery technology, cartesian loop rear feed technology, feedforward linearized technology, nonlinear device linearization technique and digital pre-distortion technology etc., wherein, digital pre-distortion technology has advantages such as high, the suitable bandwidth of stability is wide, precision is high, the realization cost is lower, is widely used at present.

The power amplifier digital pre-distortion technology has based on look-up table with based on multinomial two class methods.Method based on look-up table needs a large amount of memory cell, and estimation and convergence rate are slow; , for look-up table method, can save a large amount of RAM (Random Access Memory) memory cell, and very fast convergence rate is arranged based on polynomial pre-distortion method.Basic thought based on polynomial pre-distortion method is: by regulating the contrary complex gain curve of multinomial coefficient match PA.

In based on polynomial power amplifier pre-distortion technology, when the used multinomial exponent number of predistorter is higher, system complexity is index to be increased, and correspondingly makes the hardware complexity of power amplifier pre-distortion device and realizes that difficulty increases greatly, and then cause the bad stability of predistorter.In addition, when the design predistorter, order of a polynomial has determined the free transmission range design and the linearisation effect of power amplifier median filter.

Current based on polynomial power amplifier pre-distortion technology in, make the predistortion effect obvious, can cause the computation complexity of higher predistorter usually; And to reduce complexity, must reduce the multinomial exponent number again, thereby reduce the predistortion effect.

Summary of the invention

The purpose of this invention is to provide a kind of based on polynomial predistortion estimation method, it reduces the complexity based on polynomial power amplifier pre-distortion device when guaranteeing the predistortion performance, and then guarantee better convergence and stability are arranged based on polynomial power amplifier pre-distortion utensil

For this reason, the invention provides a kind of based on polynomial predistortion estimation method, it is characterized in that, comprise: the multinomial that receives described predistorter, described multinomial is converted to matrix notation, the predistortion matrix that acquisition is made of K predistortion function, K represents described polynomial maximum order; Described predistortion matrix is carried out singular value decomposition, obtain singular value matrix; According to described singular value matrix, obtain effective order of described predistortion matrix, effective order of described predistortion matrix is defined as described polynomial effective rank, wherein, effective order of described predistortion matrix is can reach the predetermined minimum order of approaching predistortion matrix of approaching effect to approaching of described predistortion matrix, and the described predistortion matrix of approaching is the matrix that described predistortion matrix is approached.

According to the present invention,, under the prerequisite of guaranteeing certain good predistortion effect, the predistorted complex degree is reduced greatly, and then guaranteed the stability of predistortion by estimating polynomial effective order.

Description of drawings

Fig. 1 is the schematic flow sheet of the effective rank of multinomial of the present invention method of estimation embodiment.

Fig. 2 describes the part of effective order estimator.

Fig. 3 is first kind of implementation of approximation unit among the present invention.

Fig. 4 is second kind of implementation of approximation unit among the present invention.

Fig. 5 is the schematic flow sheet of radio transmission method embodiment of the present invention.

Fig. 6 has the composition schematic diagram of the wireless launcher embodiment of predistortion function for the present invention.

Embodiment

For making technical scheme of the present invention and beneficial effect clearer, the present invention is described in further detail below in conjunction with drawings and Examples, so that implementation procedure of the present invention can be fully understood and enforcement according to this.

Basic ideas of the present invention are: utilize the singular value decomposition technology that polynomial effective rank are estimated, with the effective order that estimates multinomial exponent number as predistorter, make when the computation complexity that makes predistorter reduces, guarantee the superior function of predistorter.Polynomial predistortion distorter under this effective order can obtain enough good predistortion effect.If do not carry out the estimation on effective rank, then can bring the problem that makes higher order polynomial predistorter implementation complexity height and poor stability.

The multinomial z (n) of predistorter can be expressed as:

$z\left(n\right)=\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{a}_k{\mathrm{\φ}}_k\left(x\left(n\right)\right)---\left(1\right)$

Above-mentioned expression formula (1) is an expression formula based on polynomial predistorter;

Wherein:

K represents the sequence number of multinomial exponent number;

a _kThe expression multinomial coefficient;

φ _k(x (n)) expression is corresponding to the polynomial predistortion function in k rank;

The input data sequence of x (n) expression predistorter;

N represents the sequence number of input data sequence;

K represents polynomial maximum order.

Expression formula (1) can utilize matrix-style to be expressed as:

z＝Φ _xa (2)

In the above-mentioned expression formula (2):

Z=[z (0) ..., z (N-1)] ^T, the polynomial matrix of expression predistorter, by N element z (0) ..., z (N-1) constitutes; Wherein, the transposition of subscript T representing matrix.

A=[a ₁..., a _K] ^T, represent K multinomial coefficient a ₁..., a _KThe coefficient matrix that constitutes;

Φ _x=[φ ₁(x) ..., φ _K(x)], the predistortion matrix that constitutes by K predistortion function of expression;

X=[x (0) ..., x (N-1)] ^T, represent N input data sequence x (0) by predistorter ..., the input matrix that x (N-1) constitutes;

φ _k(x)=[φ _k(x (0)) ..., φ _k(x (N-1))] ^T, the predistortion Jacobian matrix that expression is made of N predistorter input data sequence.Need to prove φ _k(x) matrix that constitutes corresponding to the polynomial predistortion function in k rank of expression, and top Φ _xThen represent the predistortion matrix that constitutes by a plurality of predistortion functions.

Like this, seek the problem on polynomial effective rank of predistorter, can be converted to and seek predistortion matrix Φ _xThe problem of effective order.The present invention provides two kinds of execution modes and seeks predistortion matrix Φ _xEffective order.Wherein first kind of execution mode is the predistortion matrix Φ that seeks K * N rank _xThe best under Fu Luobeiniusi (Frobenius) norm is approached Approach according to this best

Obtain predistortion matrix Φ _xEffective order and final obtain polynomial effective rank; Second kind of execution mode is to utilize the normalization singular value to obtain predistortion matrix Φ _xEffective order and final obtain polynomial effective rank.

At first kind of execution mode, promptly seek the predistortion matrix Φ on K * N rank _xThe best under Fu Luobeiniusi (Frobenius) norm is approached

Problem in, K represents polynomial maximum order, N represents the number of predistorter input data sequence.

Predistortion matrix is carried out singular value decomposition, Φ _x=U ∑ V ^H, obtaining the singular value matrix ∑, it comes down to a matrix that comprises singular value information.Wherein, U is the matrix of a K * K, and V is the matrix of a N * N, and ∑ then is the matrix of a K * N, diagonal entry diag (∑)=(σ of matrix ∑ ₁₁, σ ₂₂... .., σ _Kk), singular value σ _KkBe the diagonal entry that the capable k of k through the matrix ∑ that obtains after the singular value decomposition lists, the span of k be (1, min (K, N)), K, N represent predistortion matrix Φ respectively _xLine number and columns.Singular value σ _KkComprised predistortion matrix Φ _xThe useful information of characteristic of order.

Constant and by preceding k the singular value that keeps the matrix ∑ with other singular value zero setting, the matrix ∑ that obtains _k, ∑ _kThe order k that is called ∑ approaches matrix.Obtain expression formula as follows (3).

${\mathrm{\Φ}}_x^{\left(k\right)}={\mathrm{U\Σ}}_k{V}^H---\left(3\right)$

Wherein, Φ _x ^(k)Expression predistortion matrix Φ _xOrder k approach predistortion matrix; U is the K rank (unitary matrice of K * K); V is the N rank (unitary matrice of N * N).

Above-mentioned effect of approaching can be used the matrix difference Φ shown in following expression formula (4) _x-Φ _x ^(k)The Frobenius norm of (also can be referred to as the approximate error vector) || Ф _x-Φ _x ^(k)|| _FWeigh.|| Ф _x-Φ _x ^(k)|| _FMore little, the effect of then approaching is good more.

${|{|\mathrm{\Φ}}_x-{\mathrm{\Φ}}_x^{\left(k\right)}||}_F={\left|\right|{\mathrm{U\ΣV}}^H-{\mathrm{U\Σ}}_k{V}^H\left|\right|}_F---\left(4\right)$

For K rank unitary matrice U and N rank unitary matrice V, its norm is respectively:

${\left|\right|U\left|\right|}_F=\sqrt{K}---\left(5\right)$

${\left|\right|V\left|\right|}_F=\sqrt{N}---\left(6\right)$

Therefore, above-mentioned expression formula (4) can be reduced to:

${\left|\right|{\mathrm{\Φ}}_x-{\mathrm{\Φ}}_x^{\left(k\right)}\left|\right|}_F={\left|\right|U\left|\right|}_F\·{\left|\right|\mathrm{\Σ}-{\mathrm{\Σ}}_k\left|\right|}_F\·{\left|\right|V^H\left|\right|}_F=\sqrt{\mathrm{KN}}{\left|\right|\mathrm{\Σ}-{\mathrm{\Σ}}_k\left|\right|}_F=\sqrt{\mathrm{KN}}{\left[\begin{array}{cc}\underset{i=k+1}{\overset{\min (K,N)}{\mathrm{\Σ}}}& {\mathrm{\σ}}_{\mathrm{ii}}^2\end{array}\right]}^{1/2}---\left(7\right)$

By expression formula (7) as seen, order k approaches predistortion matrix Φ _x ^(k)For predistortion matrix Φ _xApproach accuracy, depend on by those square of singular values of zero setting and.

Obvious according to expression formula (7), if k is big more, then || Ф _x-Φ _x ^(k)|| _FMore little, and || Ф _x-Φ _x ^(k)|| _F(K trends towards zero in the time of N) at k=min.Can reach the predetermined integer p that approaches the minimum of effect and be called predistortion matrix Φ _xEffective order.For example, when k value p, approximate error vector Φ _x-Φ _x ^(k)The Frobenius norm less than predetermined threshold, show to reach the predetermined effect of approaching, and when k＞p, approach effect and there is no obvious raising, then p can be defined as predistortion matrix Φ _xEffective order.

Therefore, seeking the effective rank of multinomial is converted into and seeks predistortion matrix Φ _xEffective order.Approach predistortion matrix Φ by order k _x ^(k)To predistortion matrix Φ _xApproach effect, determine predistortion matrix Φ _xEffective order.That is to say that effective order of predistortion matrix is herein represented can reach the predetermined minimum order of approaching predistortion matrix of approaching effect to approaching of described predistortion matrix.Describe the effective order that how to obtain predistortion matrix below in detail.

For example, first threshold ε is set ₁, make and work as || Ф _x-Φ _x ^(k)|| _FLess than first threshold ε ₁The time, can reach the predetermined effect of approaching.Calculate || Φ _x-Φ _x ^(k)|| _F, judge || Ф _x-Φ _x ^(k)|| _FWhether less than first threshold ε ₁, will make || Ф _x-Φ _x ^(k)|| _FLess than first threshold ε ₁Smallest positive integral k be defined as matrix Φ _xEffective order p (being polynomial effective rank of predistorter).

After obtaining the singular value matrix ∑, the present invention also can utilize the normalization singular value to judge and approach effect, therefore, also can utilize the method for normalization singular value to determine predistortion matrix Φ _xEffective order.Specifically, can adopt expression formula as follows (8) to obtain the normalization singular value

${\stackrel{\‾}{\mathrm{\σ}}}_{\mathrm{kk}}\stackrel{\mathrm{def}}{=}{\mathrm{\σ}}_{\mathrm{kk}}/{\mathrm{\σ}}_{11},1\≤k\≤\min (K,N)---\left(8\right)$

Obviously ${\stackrel{\‾}{\mathrm{\σ}}}_{11}=1.$

Be provided with one and approach zero positive number ε ₂As second threshold value, make and determine matrix Φ as follows _xEffective order the time, can reach the predetermined effect of approaching: calculate the normalization singular value

Judge the normalization singular value

Whether greater than second threshold epsilon ₂, will make the normalization singular value

Greater than second threshold epsilon ₂Maximum integer k be taken as matrix Φ _xEffective order p (being polynomial effective rank of predistorter).

Adopt approximate error vector Φ _x-Φ _x ^(k)The Frobenius norm to approach and adopt the normalization singular value to approach be the method on two kinds of parallel definite effective rank.

Fig. 1 is the schematic flow sheet of the effective rank of multinomial of the present invention method of estimation embodiment, and this method embodiment illustrates in above-mentioned first kind of mode of approaching.As shown in Figure 1, this method of estimation mainly comprises the steps:

Step S110, the multinomial of reception predistorter The multinomial z (n) of predistorter is converted to matrix z=Φ _xA obtains predistortion matrix Φ _x=[φ ₁(x) ..., φ _K(x)], predistortion matrix Φ _xConstitute by K predistortion function, wherein, φ _k(x) expression is corresponding to the matrix of the polynomial predistortion function formation in k rank;

Step S120 is to predistortion matrix Φ _xCarry out singular value decomposition Φ _x=U ∑ V ^H, obtain the singular value matrix ∑.Diagonal entry diag (∑)=(σ of singular value matrix ∑ ₁₁, σ ₂₂... .., σ _Kk), singular value σ _KkIt is the diagonal entry that the capable k of k through the singular value matrix ∑ that obtains after the singular value decomposition lists;

Step S130, the order k that obtains ∑ according to this ∑ approaches the matrix ∑ _k, and approach ∑ according to this order k _kObtain predistortion matrix Φ _xOrder k approach predistortion matrix Φ _x ^(k), wherein, preceding k the singular value that keeps ∑ is constant, and simultaneously with other singular value zero setting, the order k that obtains the singular value ∑ in view of the above approaches the matrix ∑ _k, this Φ _x ^(k)Obtain and see also above-mentioned expression formula (3);

Step S140 judges and approaches effect, can adopt matrix difference Φ _x-Φ _x ^(k)The Frobenius norm || Ф _x-Φ _x ^(k)|| _FSize weigh the above-mentioned effect of approaching, wherein || Ф _x-Φ _x ^(k)|| _FSee also above-mentioned expression formula (4) to expression formula (7);

Step S150 approaches effect according to this and obtains predistortion matrix Φ _xEffective order p, effectively order p is polynomial effective rank.

Fig. 2 is the composition schematic diagram of multinomial effective order estimator embodiment of the present invention.In conjunction with method embodiment shown in Figure 1, this estimation module shown in Figure 2 mainly comprises converting unit 210, resolving cell 220 and approximation unit 230, wherein:

Converting unit 210 is used to receive the multinomial z (n) of predistorter, wherein

And according to this multinomial z (n) acquisition one predistortion matrix Φ _xAbove-mentioned multinomial z (n) is converted to matrix z=Φ _xA obtains above-mentioned matrix Φ _x

Resolving cell 220 is used for predistortion matrix is carried out singular value decomposition, Φ _x=U ∑ V ^H, obtain the singular value matrix ∑; Wherein, U is the matrix of a K * K, and V is the matrix of a N * N, and ∑ then is the matrix of a K * N, and K, N represent predistortion matrix Φ respectively _xLine number and columns;

Approximation unit 230 is used for obtaining predistortion matrix Φ according to singular value matrix ∑ and the default threshold value of approaching _xEffective order, predistortion matrix Φ _xEffective order be effective rank of multinomial z (n).

Fig. 3 is first kind of implementation of approximation unit 230 among the present invention, and as shown in Figure 3, the approximation unit 230 in this first kind of implementation mainly comprises the first operator unit 310, the second operator unit 320 and the 3rd operator unit 330, wherein:

The first operator unit 310 be used to keep preceding k singular value of singular value matrix, and with all the other singular value zero setting, the order k that obtains singular value matrix approaches matrix;

The second operator unit 320 is used for approaching the order k that matrix obtains predistortion matrix according to order k and approaches predistortion matrix;

The 3rd operator unit 330 calculates the size of Fu Luobeiniusi norm that predistortion matrix and order k approach the difference of predistortion matrix;

Determine subelement 340, be used for approaching approach the effective order that effect determine predistortion matrix of predistortion matrix predistortion matrix according to order k; Wherein first threshold is for approaching threshold value, the size of Fu Luobeiniusi norm that adopts predistortion matrix and order k to approach the difference of predistortion matrix is weighed and is approached effect, suppose that the Fu Luobeiniusi norm of difference of approaching predistortion matrix as predistortion matrix and order k is during less than first threshold, reach and approach effect, then approach the Fu Luobeiniusi norm of difference of predistortion matrix smaller or equal to first threshold as predistortion matrix and order p, and (the Fu Luobeiniusi norm of difference that k＜p) approaches predistortion matrix is defined as predistortion matrix Φ with p during all greater than first threshold for predistortion matrix and order k _xEffective order, promptly polynomial effective rank.Predistortion matrix and order k are approached the Fu Luobeiniusi norm calculation of the difference of predistortion matrix, see also above-mentioned expression formula (4) to expression formula (7).

Fig. 4 is second kind of implementation of approximation unit 230 among the present invention, and as shown in Figure 4, the approximation unit 230 in this second kind of implementation mainly comprises normalization subelement 410 and definite subelement 420, wherein:

Normalized value computation subunit 410, the element that is used for being listed as according to singular value matrix first row first is to the singular value σ in the singular value matrix _KkCarry out normalization, obtain singular value σ _KkNormalized value

Determine subelement 420, adopt normalized value

Judge and approach effect, suppose when the normalization singular value

Greater than second threshold epsilon ₂The time, can not reach and approach effect, then according to second threshold value and normalized value, obtain effective order of predistortion matrix, particularly, if the normalization singular value

More than or equal to second threshold epsilon ₂, and

All less than second threshold epsilon ₂, then integer p can be taken as matrix Φ _xEffective order, wherein second threshold value is for approaching threshold value.

Adopted aforesaid effective rank to estimate the power amplifier technology of handling, can guarantee the superior function of predistorter under low complex degree, thereby can effectively improve the power efficiency of power amplifier.

Fig. 5 is the schematic flow sheet of radio transmission method embodiment of the present invention.In conjunction with effective rank method of estimation embodiment and effective order estimator embodiment shown in Figure 2 shown in Figure 1, power-magnifying method embodiment shown in Figure 5 mainly comprises the steps:

Step S510 receives coupling data based on polynomial predistorter, obtains the multinomial of predistorter;

Step S520 carries out effective rank to this multinomial and estimates, obtains this polynomial effective rank;

Step S530, filter receives the polynomial predistorting data based on effective rank, and this predistorting data is carried out certain conversion, changes signal spectrum, make it more appropriate to Channel Transmission;

Step 540, up-conversion and D/A module are carried out upconversion process with baseband signal, make it to become radiofrequency signal;

Step S550 is that radiofrequency signal is amplified to the frequency-conversion processing result, obtains to amplify result and output.

Among the above-mentioned steps S520 this multinomial is carried out effective rank and estimate, and obtain the process on these effective rank, the effective rank method of estimation embodiment that please refer to as shown in Figure 1 understands, and repeats no more herein.

Fig. 6 is the composition schematic diagram with wireless launcher embodiment of predistortion function.In conjunction with effective rank method of estimation embodiment shown in Figure 1, effective order estimator embodiment shown in Figure 2 and the radio communication launching technique embodiment with predistortion function shown in Figure 5, the wireless launcher with predistortion function shown in Figure 6 mainly comprises:

Predistorter 610, be used to receive the data of coupling and process base band demodulating, obtain multinomial and send to effective order estimator 620, receive polynomial effective rank of effective order estimator 620 feedbacks, and effective rank are sent to filter 630 as the multinomial exponent number;

Effective order estimator 620 links to each other with predistorter 610, is used for that multinomial is carried out effective rank and estimates, obtains polynomial effective rank;

Filter 630 links to each other with predistorter 610, is used for receiving based on the polynomial predistorting data in effective rank, and predistorting data is carried out filtering, and filtered result is imported D/A and up-conversion module 650;

D/A and up-conversion module 650 are carried out digital-to-analogue conversion and upconversion process to the output result of filter 630, obtain the frequency-conversion processing result, i.e. rf data;

Power amplifier 640 links to each other with D/A and up-conversion module 650, and this rf data is amplified, and obtains to amplify result and output;

Down-conversion and A/D module 660 are carried out down-conversion and analog-to-digital conversion to the data sequence that is coupled back, and export transformation result to delay estimation module 670;

Postpone estimation module 670, to through the former input data sequence that postpones be coupled back and compare through the data sequence of the base band of demodulation, analyze and to be coupled back and through the data sequence of the base band of demodulation and the difference of former data (being ideal data), so that predistorter 610 is adjusted.

620 pairs of these multinomials of above-mentioned effective order estimator carry out effective rank to be estimated, and obtains the process on these effective rank, please refer to as shown in Figure 1 effective rank method of estimation embodiment and effective rank estimation module shown in Figure 2 understand, repeat no more herein.

Among Fig. 6, " postponing to estimate " module is positioned at " left side " end, carries out the data contrast.

Effective rank estimation technique among the present invention makes predistorter under the prerequisite of guaranteeing the predistortion effect, reduces overhead and implementation complexity greatly, and owing to reduced the multinomial exponent number, makes the convergence of predistorter and stability improve.

The predistorter that the present invention realizes has the advantage of low, stable height of complexity and fast convergence rate, goes for portable mobile termianls such as the communication network device such as base station, repeater, broadcasting television tower of large-scale communication equipment and mobile phone, mobile TV.

Though embodiments of the present invention as above, described content is not in order to limit the present invention just for the ease of understanding the present invention.Technical staff in the technical field of the invention; under the prerequisite that does not break away from the disclosed spirit and scope of the present invention; can on concrete form of implementing and details, do any modification and variation, but these are revised and variation all should be included within the scope of patent protection of the present invention.

Claims (5)

1. one kind based on polynomial predistortion estimation method, it is characterized in that, comprising:

Receive the multinomial of described predistorter, described multinomial is converted to matrix notation, the predistortion matrix that acquisition is made of K predistortion function, K represents described polynomial maximum order;

Described predistortion matrix is carried out singular value decomposition, obtain singular value matrix;

According to described singular value matrix, obtain effective order of described predistortion matrix, effective order of described predistortion matrix is defined as described polynomial effective rank, wherein,

Effective order of described predistortion matrix is can reach the predetermined minimum order of approaching predistortion matrix of approaching effect to approaching of described predistortion matrix, and the described predistortion matrix of approaching is the matrix that described predistortion matrix is approached.

2. according to claim 1ly it is characterized in that based on polynomial predistortion estimation method described multinomial is converted to uses matrix notation, to obtain the step of described predistortion matrix, this step comprises:

The multinomial of described predistorter It is z=Φ that described multinomial z (n) is converted to matrix notation _xA is to obtain described predistortion matrix Φ _xWherein:

K represents the sequence number of described multinomial exponent number;

a _kRepresent described polynomial coefficient;

φ _k(x (n)) expression is corresponding to the described polynomial predistortion function in k rank;

The input data sequence of the described predistorter of x (n) expression;

N represents the sequence number of described input data sequence;

Z=[z (0) ..., z (N-1)] ^TThe polynomial matrix of representing described predistorter, by N element z (0) ..., z (N-1) constitutes, the transposition of T representing matrix;

A=[a ₁..., a _K] ^TRepresent K multinomial coefficient a ₁..., a _KThe coefficient matrix that constitutes;

Φ _x=[φ ₁(x) ..., φ _K(x)] the described predistortion matrix that constitutes by K predistortion function of expression;

X=[x (0) ..., x (N-1)] ^TExpression is by N the input data sequence x (0) of predistorter ..., the input matrix that x (N-1) constitutes;

φ _k(x)=[φ _k(x (0)) ..., φ _k(x (N-1))] ^TThe described predistortion Jacobian matrix that expression is made of N input data sequence.

3. according to claim 1ly it is characterized in that based on polynomial predistortion estimation method,

Preceding k the singular value that keeps described singular value matrix, and with all the other singular value zero setting, the order k that obtains described singular value matrix approaches singular value matrix, approaches the described predistortion matrix of approaching that singular value matrix obtains described predistortion matrix according to described order k; And

When the Fu Luobeiniusi norm of described predistortion matrix and described difference of approaching predistortion matrix during, be judged as the described predistortion matrix of approaching and reach the predetermined effect of approaching smaller or equal to first threshold.

4. according to claim 1ly it is characterized in that based on polynomial predistortion estimation method,

Based on the element of first row first row in the described singular value matrix to singular value σ in the described singular value matrix _KkCarry out normalized, obtain singular value σ _PpNormalized value

And

If described normalized value More than or equal to second threshold value, the described predistortion matrix of approaching that then is judged as order and is p can reach predetermined and approaches effect, wherein, and 1＜p＜K.

5. according to claim 4ly it is characterized in that based on polynomial predistortion estimation method,

If for arbitrary k value that satisfies p＜k≤K,

All less than described second threshold value, then p can reach the predetermined minimum order of approaching predistortion matrix of approaching effect to approaching of described predistortion matrix.

Images