JP5290698B2 - Predistorter - Google Patents

Description

  The present invention relates to a predistorter that generates a predistortion signal applied in advance to an input signal in order to compensate for distortion of an output signal from a compensated circuit such as a signal amplifier.

Conventionally, the number of sampling points of an input / output signal used when updating a coefficient (or look-up table) of a polynomial that generates a distortion compensation value used in a predistorter has been fixed (see, for example, Patent Document 1). .
Japanese Patent No. 3560398

  For example, when the number of sampling points is increased, the distortion compensation accuracy is increased, but the update speed and the tracking speed corresponding to the update speed and the convergence speed are reduced. On the other hand, when the number of sampling points is reduced, the update speed and the follow-up speed are increased, but the distortion compensation accuracy is lowered. For this reason, when the distortion characteristics of the amplifier change due to changes in the number of transmission carriers or changes in temperature, the coefficient (or lookup table) of the polynomial that generates the distortion compensation value multiplied by the predistorter is changed immediately. However, there is a problem that maintaining the distortion compensation accuracy is not compatible with increasing the followability and the update speed.

  Accordingly, the present invention has been made to solve the above-described problems, and provides a predistorter that can reduce the amount of memory for distortion compensation and can achieve both high distortion compensation accuracy and distortion compensation follow-up performance. With the goal.

  In order to achieve the above object, the predistorter according to the present invention can change the number of sampling points of an input / output signal used when updating a coefficient (or lookup table) of a distortion compensation polynomial for generating a distortion compensation value. It was decided to.

  Specifically, the predistorter according to the present invention includes a distortion compensator that distorts an input signal with a distortion compensation value based on a predetermined criterion and outputs the input signal to a compensated circuit, and the input signal at different sampling points. A control unit that calculates an update amount of the reference from a plurality of input sampling signals acquired and a plurality of output sampling signals acquired from the compensated circuit at the sampling points, and causes the distortion compensation unit to update the reference The control unit can vary the number of sampling points.

  Here, the predetermined reference is a coefficient of a distortion compensation polynomial used when obtaining a distortion compensation value or a table value of a lookup table. Since the control unit can change the number of sampling points of the input / output signal, reduce the number of sampling points when emphasizing distortion compensation followability, and increase the number of sampling points when emphasizing distortion compensation accuracy. Can do.

  Therefore, the predistorter according to the present invention can achieve both high distortion compensation accuracy and distortion compensation followability.

  The distortion compensation unit of the predistorter according to the present invention has a distortion compensation polynomial, the reference is a coefficient of the distortion compensation polynomial, and the control unit is Nth (N is an integer equal to or greater than 1) The matrix element of a row and b column calculated by the value (a, b are integers of 1 or more) obtained by multiplying the output sampling signal taken in by (2a + 2b-2) up to the (N-1) th taken output sampling signal To calculate the first matrix by multiplying the complex conjugate transpose matrix of the output signal matrix from the left side and calculating the first matrix. The value obtained by multiplying the input sampling signal to the power of (2a + 2b-2) is added to the matrix element of a row and b column calculated up to the N-1th input sampling signal, and the Nth matrix of a row and b column is added. Calculate the input signal matrix as elements, The difference between the input signal matrix calculated from the input sampling signals and the output signal matrix calculated from the N output sampling signals is multiplied from the right side by a matrix of coefficients of the distortion compensation polynomial to obtain an error matrix, and the error A matrix is multiplied by a complex conjugate transpose matrix of N output signal matrices from the left side to calculate a second matrix, and an inverse matrix of the first matrix and the second matrix are multiplied to obtain a coefficient of the distortion compensation polynomial. The update amount of is calculated.

  The distortion compensator of the predistorter according to the present invention may have a lookup table. In this case, the reference is a table value of the look-up table, and the control unit obtains a value (a, 2b−2) raised to the Nth (N is an integer of 1 or more) output sampling signal (2a + 2b−2). b is an integer greater than or equal to 1) and is added to the matrix element of a row and b column calculated up to the (N-1) th output sampling signal, and the output signal matrix as the Nth a row and b column matrix element The first matrix is calculated by multiplying the complex conjugate transpose matrix of the output signal matrix from the left side, and the value obtained by multiplying the Nth input sampling signal to the (2a + 2b-2) th power is the (N-1) th. Is added to the matrix element of a row and b column calculated up to the input sampling signal taken in to calculate the input signal matrix as the matrix element of the Nth a row and b column, and calculated from the N input sampling signals The input signal The difference between the column and the output signal matrix calculated from the N output sampling signals is multiplied by a matrix of coefficients of the distortion compensation polynomial from the right side to obtain an error matrix, and the error matrix includes the complex of the N output signal matrices. A conjugate transpose matrix is multiplied from the left side to calculate a second matrix, and an inverse matrix of the first matrix and the second matrix are multiplied to calculate an update amount of the table value of the lookup table.

  By sequentially expressing each element of the matrix obtained in the middle of obtaining the distortion compensation polynomial coefficient for generating the distortion compensation value by a recurrence formula, it is possible to sequentially update. For this reason, the number of data to be stored is the number of matrix elements, not the number of sampling points (capture points) of the input / output signal, so the number of sampling points used for calculating the coefficient (or lookup table) of the distortion compensation polynomial is The greater the number, the higher the effect of reducing the amount of memory compared to the conventional method.

  The control unit of the predistorter according to the present invention monitors an integrated value of error power between the input signal and the output signal, and when the integrated value exceeds a predetermined threshold, the control unit is connected to the distortion compensating unit. May be updated. The control unit can automatically determine whether followability or accuracy is required, and can update the coefficient (or look-up table) of the distortion compensation polynomial.

  The control unit of the predistorter according to the present invention monitors the error power between the input signal and the output signal, and when the error power exceeds a predetermined threshold, the sampling point number (capture point number) is The reference may be updated by reducing the error power as compared with a case where the error power is within a predetermined threshold. The control unit can automatically determine whether followability or accuracy is required, and can update the coefficient (or look-up table) of the distortion compensation polynomial. In addition, at the initial stage of creating the coefficient (or lookup table) of the polynomial that generates the distortion compensation value, the number of sampling points is reduced to speed up the drawing of distortion compensation. It can also be increased.

  According to the present invention, since the amount of memory is not the number of sample points (number of acquisition points) but the number of matrix elements, the amount of memory for distortion compensation can be reduced. In addition, since the number of sampling points (capture points) of the input / output signal used when updating the coefficient (or lookup table) of the polynomial that generates the distortion compensation value can be arbitrarily set, any accuracy, any update speed, Distortion compensation at an arbitrary tracking speed and an arbitrary convergence speed is possible. For this reason, this invention can provide the predistorter which can make high distortion compensation precision and distortion compensation followability compatible.

  Embodiments of the present invention will be described with reference to the accompanying drawings. The embodiment described below is an example of the configuration of the present invention, and the present invention is not limited to the following embodiment. In the present specification and drawings, the same reference numerals denote the same components.

(Embodiment 1)
FIG. 1 is a block diagram illustrating the configuration of the predistorter 301 of the present embodiment. The predistorter 301 outputs the predistortion compensation signal A obtained by compensating the distortion of the input signal X with a distortion compensation value based on a predetermined reference to the compensated circuit 401, and the input signal X at different sampling points. Control for calculating the reference update amount from a plurality of input sampling signals acquired and a plurality of output sampling signals acquiring the output signal Y from the compensated circuit 401 at the sampling points, and causing the distortion compensator 11 to update the reference Unit 13. For example, the compensated circuit 401 is a signal amplifier. Further, the standard possessed by the distortion compensation unit 11 is a coefficient of a distortion compensation polynomial or a table value of a lookup table, but will be described below as a coefficient of a distortion compensation polynomial.

  The control unit 13 can vary the coefficient of the distortion compensation polynomial or the number of sampling points (number of capture points) used for calculating the lookup table value. This is because the control unit 13 calculates the coefficient update amount of the distortion compensation polynomial by multiplying the inverse matrix of the first matrix and the second matrix. Here, the first matrix is a matrix of a rows and b columns calculated from the value obtained by raising the Nth output sampling signal to the power of (2a + 2b-2) up to the N−1th output sampling signal. It is a matrix obtained by adding an element and calculating an output signal matrix as a matrix element of the Nth a row and b column and multiplying the complex conjugate transpose matrix of the output signal matrix from the left side. The second matrix is an error matrix obtained by multiplying the difference between an input signal matrix made up of N input sampling signals and an output signal matrix made up of N output sampling signals by a matrix of distortion compensation polynomial coefficients from the right side. And a matrix obtained by multiplying the error column from the left by a complex conjugate transpose matrix of an output signal matrix composed of N output sampling signals. Details of the calculation of the update amount will be described later.

  For example, the control unit 13 may monitor an integrated value of error power between the input signal X and the output signal Y, and cause the distortion compensation unit 11 to update the reference when the integrated value exceeds a predetermined threshold value. . The control unit 13 calculates the difference in power between the input signal X and the output signal Y to obtain error power. For example, the error power for each sampling point is integrated, and this is used as the accumulated error power. The control unit 13 is set with a threshold of accumulated error power. When the accumulated error power exceeds the threshold, the control unit 13 calculates the update amount of the distortion compensation polynomial coefficient or the lookup table value, and the distortion compensation polynomial coefficient or lookup table table of the distortion compensation unit 11. Update the value.

  Further, the control unit 13 monitors the accumulated error power between the input signal X and the output signal Y, and when the accumulated error power exceeds a predetermined threshold, the number of sampling points is determined when the accumulated error power is within the predetermined threshold. The reference may be updated with fewer. For example, a threshold value of accumulated error power is set in the control unit 13. When the accumulated error power is within the threshold value, the control unit 13 calculates the update amount of the coefficient of the distortion compensation polynomial or the table value of the lookup table by the predetermined number of sampling points, and the distortion compensation polynomial of the distortion compensation unit 11 is calculated. Update the coefficient or the table value in the lookup table. When the accumulated error power exceeds the threshold value, the control unit 13 updates the coefficient of the distortion compensation polynomial or the table value of the lookup table with a smaller number of sampling points than when the accumulated error power is within the threshold value. And the coefficient of the distortion compensation polynomial of the distortion compensation unit 11 or the table value of the lookup table is updated. As a result, it is possible to achieve both tracking ability and distortion compensation accuracy.

ただし、ｘ（ｎ）は入力信号、ａ（ｎ）は予歪補償信号、ｗ’_ｄ，ｋ、ｗ’_{ｒ，ｄ，ｌ}及びｗ’_{ｒ，ｄ，ｌ，ｋ}は係数、ｄ及びｒは正規化先行時間又は正規化遅延時間、ｋ及びｌは次数である。また、Ｄ_ａ＝ｍａｘ（Ｄ１，Ｒ１）、Ｄ_ｂ＝ｍａｘ（Ｄ２，Ｒ２）、およびＫ_ｂ＝Ｋ＋Ｌ−１である。ここで、ｍａｘ（ｃ_１，ｃ_２，・・・，ｃ_Ｊ）はｃ_１、ｃ_２、・・・、ｃ_Ｊの最大値を表す。

FIG. 2 is a block diagram illustrating the distortion compensator 11 of the predistorter 301. The distortion compensation unit 11 stores a distortion compensation polynomial that can be expressed by Equation 1 or Equation 2. The distortion compensation unit 11 generates a predistortion compensation signal A obtained by compensating the distortion of the input signal X using the distortion compensation polynomial, and outputs the predistortion compensation signal A to the compensated circuit 401.

Where x (n) is an input signal, a (n) is a predistortion compensation signal, w ′ _{d, k} , w ′ _{r, d, l} and w ′ _{r, d, l, k} are coefficients, d and r are Normalized lead time or normalized delay time, k and l are orders. Also, D _a = max (D1, R1), D _b = max (D2, R2), and K _b = K + L-1. _{Here, max (c 1, c 2} , ···, c J) represents the maximum value of _{c 1, c 2, ···,} c J.

  The control unit 13 calculates the coefficient of the distortion compensation polynomial by a coefficient calculation method described later. Further, the control unit 13 outputs the calculated coefficient to the distortion compensation unit 11. The distortion compensator 11 generates a predistortion compensation signal A using this coefficient. Further, the distortion compensation unit 11 periodically updates the coefficient of the distortion compensation polynomial.

とおく。ここで、ｊは要素増幅器５１３の番号であり、Ｊは要素増幅器５１３の全個数である。また、ｄは入力信号の正規化先行時間、または正規化遅延時間を表し、Ｄ１_ｊはｊ番目の要素増幅器の最大正規化先行時間、Ｄ２_ｊはｊ番目の要素増幅器の最大正規化遅延時間、ｋは次数であり、Ｋ_ｊはｊ番目の要素増幅器の歪成分の最大次数を表す。 (Modeling of amplifier circuit and estimation of distortion characteristics)
First, the distortion characteristics of the amplifier circuit are modeled. FIG. 3 is a diagram modeling an amplifier circuit composed of a plurality of amplifiers. Input signal X is input and output signal Y is output. Here, a discrete time signal obtained by sampling the input signal X with a period T _s is x (nT _s ), and a discrete time signal obtained by sampling the output signal Y with a period T _s is y (nT _s ) to simplify the notation. Therefore, each is represented by x (n) and y (n). Further, both x (n) and y (n) are complex signals having a real component and an imaginary component, and multiplication and addition for x (n) and y (n) are complex multiplication and complex addition, respectively. It shall be shown. Each of the distortion characteristics of the element amplifier 513 constituting the amplifier circuit

far. Here, j is the number of the element amplifier 513, and J is the total number of element amplifiers 513. Further, d represents a normalization leading time of the input signal, or a normalization delay time, D1 _j is a maximum normalization leading time of the _jth element amplifier, D2 _j is a maximum normalization delay time of the jth element amplifier, k is the order, and K _j represents the maximum order of the distortion component of the j-th element amplifier.

とおく。ここで、ｒは入力信号の正規化先行時間、または正規化遅延時間を表し、Ｒ１_ｊはｊ番目の要素増幅器に対する合成する比率において関係する入力信号の最大正規化先行時間、Ｒ２_ｊはｊ番目の要素増幅器に対する合成する比率において関係する入力信号の最大正規化遅延時間、ｌは次数であり、Ｌ_ｊはｊ番目の要素増幅器に対する合成する比率において関係する入力信号の最大次数から１次高い次数を表す。 The ratio of combining the element amplifiers 513 is a function (combining polynomial) 512 of the amplitude value of the input signal X.

far. Here, r represents the normalization leading time of the input signal, or the normalization delay time, R1 _j is the maximum normalization leading time of the input signal related in the combining ratio with respect to the jth element amplifier, and R2 _j is the jth The maximum normalized delay time of the input signal related in the combining ratio with respect to the element amplifier of L, l is the order, and L _j is the first order higher than the maximum order of the input signal related in the ratio of combining with the j-th element amplifier. Represents.

である。 At this time, the output of the amplifier circuit is

It is.

Here, Formula 5 is arranged. First, the same terms for different amplifiers (different j)

である。 Next, if the term of r = d of Formula 6 is put together,

It is.

Equation 7 is

further,

That is, in order to estimate the distortion characteristics of the amplifier circuit that can be modeled as shown in FIG. 3, the coefficients h ′ _{d, k} , h ′ _{r, d, l} , and h ′ _{r, d, l, k} are calculated. good. The coefficients h ′ _{d, k} , h ′ _{r, d, l} , and h ′ _{r, d, l, k} are calculated by applying the input signal X and the output signal Y to the model of Equation 9 and using the least squares method. May be used.

となり、線形であるのが理想的である。但し、Ｇは増幅装置の利得を表す実数定数である。ここでは、以降の議論を簡単にする目的で、Ｇ＝１とおくこととする。 (Distortion compensation method)
Next, a distortion compensation method for the amplifier circuit based on the model of FIG. 3 will be described. The relationship between the input and output signals of the entire amplifier is

Ideally, it should be linear. Here, G is a real constant representing the gain of the amplifier. Here, G = 1 is set for the purpose of simplifying the following discussion.

However, in an actual amplifying device, when the amplitude (or power) of the input signal increases, the relationship between the input and output signals is not linear but nonlinear as expressed by Equation 5 or Equation 9. In order to compensate for distortion in the amplifier circuit based on the model of FIG. 3, a distortion compensation polynomial in which the input signal X and the output signal Y are interchanged in Equation 5 or Equation 9, which is a distortion characteristic polynomial expressing the model of FIG.

Similar to the arrangement from Equation 5 to Equation 9,

If the coefficients w ′ _{d, k} , w ′ _{r, d, l} , and w ′ _{r, d, l, k} in Equation 12 are estimated using the least square method, the input signal X is compensated according to the distortion compensation polynomial. Good. In calculating the coefficients w ′ _{d, k} , w ′ _{r, d, l} and w ′ _{r, d, l, k} by the least square method, the coefficients w ′ _{d, k} , w ′ _{r, d, l} , And a set of x (n) and y (n) obtained by sampling more input signals X and output signals Y than the total number of w ′ _{r, d, l, k} .

  That is, for the purpose of linearizing the input / output relationship of the amplifying apparatus using the coefficient of Equation 12 obtained using the least square method, the nonlinear distortion characteristic (distortion value) of the amplifying apparatus provided in the previous stage of the amplifying apparatus. In a predistorter 301 that generates a reverse distortion characteristic (distortion compensation characteristic) for the input signal X, a predistortion compensation signal A in which a distortion compensation value is given in advance to the input signal X is generated and input to an amplifying apparatus. At this time, in order to obtain a predistortion compensation signal A in which the output signal Y of the amplification device becomes y (n) = x (n), the predistorter 301 uses the distortion compensation polynomial expressed by Equation 1 as the input signal X A distortion compensation value is given in advance to generate a predistortion compensation signal A and input to the amplifying apparatus.

A predistortion compensation signal A that gives a reverse distortion characteristic (distortion compensation characteristic) that compensates for a non-linear distortion given by a mathematical expression 5 or 9, which is a distortion characteristic polynomial expressing the model of FIG. In order to obtain (n), the coefficient w ′ _{j, r, d, l, k} , which satisfies the relationship of Expression 11 or Expression 12 in which x (n) and y (n) are replaced in Expression 5 or 9, Alternatively, the coefficients w ′ _{d, k} , w ′ _{r, d, l} , and w ′ _{r, d, l, k} may be obtained.

ここで、Ｔは行列の転置を表す。 (Calculation method of distortion compensation polynomial coefficients)
Here, a method for obtaining the coefficient of the distortion compensation polynomial using Expression 12 will be described. However, here, for simplicity of expression, Da = 0, R1 = 0, and D1 = 0. A coefficient vector in which all the coefficients w ′ _{d, k} , w ′ _{r, d, l} , and w ′ _{r, d, l, k} are arranged is set as in Expression 13.

Here, T represents transposition of the matrix.

但し、

歪補償多項式の係数を推定する際に使用する異なるｎに対する数式１２の個数はＮ（Ｎ≧Ｑ＋１）であり、数式１２をＮ個まとめると数式１４は数式１６の連立方程式となる。

Here, in order to obtain the coefficients w ₀ , w ₁ ,..., W _Q of the distortion compensation polynomial in Expression 12, Q + 1 or more Expressions 12 in different n are required. Here, when Expression 12 is expressed as a matrix, Expression 14 is obtained.

However,

The number of Equations 12 for different n used for estimating the coefficient of the distortion compensation polynomial is N (N ≧ Q + 1), and when Equations 12 are combined, Equation 14 becomes the simultaneous equations of Equation 16.

但し、Ｈは表列の複素共役転置を表す。このようにして得られた係数ｗ_０、ｗ_１、・・・、ｗ_Ｑをａ（ｎ）に適用し、ｘ（ｎ）に応じた歪補償値を算出すれば歪補償ができる。 The coefficients w ₀ , w ₁ ,..., W _Q are obtained by solving the simultaneous equations of Expression 16 or Expression 18. In order to solve the simultaneous equations of Expression 16 or Expression 18, the sweeping method may be used, or Expression 19 may be used by using the least square method.

However, H represents the complex conjugate transpose of a table row. If the coefficients w ₀ , w ₁ ,..., W _Q obtained in this way are applied to a (n) and a distortion compensation value corresponding to x (n) is calculated, distortion compensation can be performed.

  The distortion compensation unit 11 receives the coefficient calculated by the control unit 13 and applies it to the distortion compensation polynomial. The predistorter 301 generates a predistortion compensation signal A from the input signal X using a distortion compensation polynomial to which this coefficient is applied. Since the distortion compensation polynomial stored in the distortion compensation unit 11 is obtained by appropriately modeling the compensated circuit 401 and obtained from the distortion characteristics of the model, the predistorter 301 can calculate the distortion compensation value with a small amount of calculation. The accuracy of distortion compensation can be increased.

(Update algorithm)
Next, an update algorithm for updating the coefficient of the distortion compensation polynomial over time will be described.

  When updating the coefficient of the distortion compensation polynomial while compensating for distortion of the amplifier circuit, the signal input to the amplifier circuit is not x (n), but y (n), which is the output signal Y of the amplifier circuit, is changed to x (n). This is a predistortion compensation signal A that can be expressed by Formula 2 in which distortion compensation is performed so as to be close, that is, a ′ (n). Formula 2 plays the role of Formula 1 described so far when performing distortion compensation while updating the coefficient of the distortion compensation polynomial.

も成立する。 At this time, if distortion compensation of the amplifier circuit is accurately performed, y (n) = x (n) is established.

Also holds.

とおくと、誤差

が得られる。この誤差ｅ（ｎ）が零になるように歪補償多項式の係数を更新する。 If the distortion compensation of the amplifier circuit is not sufficient, y (n) = x (n) is not satisfied,

Error

Is obtained. The coefficient of the distortion compensation polynomial is updated so that this error e (n) becomes zero.

That is, the coefficients w ′ _{d, k} (i), w ′ _{r, d, l} (i), and w ′ _r satisfying the condition y (n) = x (n) that makes the input / output relationship of the entire amplifier circuit linear. _{, D, l, k} (i) satisfies Expression 12. Here, the coefficients w ′ _{d, k} , w ′ _{r, d, l} , and w ′ _{r, d, l, k} are all arranged as w ₀ , w _{1 ,.} because represented, i-th coefficient _w 0 obtained _{updates, w 1, ···,} w _Q each _{w 0 (i), w 1} (i), ···, w Q (i) and In other words, the coefficients w ₀ (i), w ₁ (i),..., W _Q (i) satisfying the condition y (n) = x (n) that makes the input / output relationship of the entire amplifier circuit linear 12 is satisfied. Accordingly, w ₀ (i), w ₁ (i),..., W _Q (i) may be obtained so that a ′ (n) and a ″ (n) match.

とおく。但し、ｉ≧０である。なお、数式２３の係数ｗ（ｉ）の初期値としては、ｗ（０）＝（１，０，…，０）^Ｔを用いるか、同じ被補償部に対して十分な取り込み点数Ｎのときに数式１６、数式１８、または数式１９を用いて予め求めておいた係数ｗをｗ（０）＝ｗとして用いるか、同じ被補償部に対して十分に更新回数を確保して予め求めておいた係数の行列ｗ（∞）をｗ（０）＝ｗ（∞）として用いればよい。 Here, the update of w ₀ (i), w ₁ (i),..., W _Q (i) will be described.

far. However, i ≧ 0. Note that as the initial value of the coefficient w (i) of Equation 23, w (0) = (1, 0,..., 0) ^T is used, or when a sufficient number of capture points N is obtained for the same compensated portion. The coefficient w obtained in advance using Expression 16, Expression 18 or Expression 19 is used as w (0) = w, or is obtained in advance by sufficiently ensuring the number of updates for the same compensated portion. The coefficient matrix w (∞) may be used as w (0) = w (∞).

である。
また、行ｘ_ｎ（ｉ）を構成するベクトルｘ１_ｎ（ｉ）、ｘ２_ｎ（ｉ）およびｘ３_ｎ（ｉ）はそれぞれ数式２５、数式２６及び数式２７で表せる。

である。
また、行ｙ_ｎ（ｉ）を構成するベクトルｙ１_ｎ（ｉ）、ｙ２_ｎ（ｉ）およびｙ３_ｎ（ｉ）はそれぞれ数式２９、数式３０及び数式３１で表せる。

但し、Ｎは歪補償値を推定する際に使用する入出力信号の取り込み点数、Ｍは次の多項式係数の更新時に用いる入出力信号の取り込み開始までのサンプリング間隔数である。 Here, if placed with Da = 0, R1 = 0, D1 = 0 for simplicity of _{representation, w 0 (i), w} 1 (i), the input signal X corresponding to the ··· _w Q (i) And the matrix of the output signal Y can be expressed as Equation 24 and Equation 28, respectively.

It is.
Further, the vectors x1 _n (i), x2 _n (i), and x3 _n (i) constituting the row x _n (i) can be expressed by Expression 25, Expression 26, and Expression 27, respectively.

It is.
Further, the vectors y1 _n (i), y2 _n (i), and y3 _n (i) constituting the row y _n (i) can be expressed by Expression 29, Expression 30, and Expression 31, respectively.

Here, N is the number of input / output signal capture points used when estimating the distortion compensation value, and M is the number of sampling intervals until the start of input / output signal capture used when updating the next polynomial coefficient.

とおくと、連立方程式

によりｗ（ｉ）を更新すればよい。但し、μは０＜μ≦１．０を満たす。 Also,

Then, simultaneous equations

The w (i) may be updated by However, μ satisfies 0 <μ ≦ 1.0.

  Specifically, the control unit 13 generates an input replica signal in which the predistortion compensation signal A is estimated by multiplying the input signal X by a value obtained by substituting the input signal X into the distortion compensation polynomial. The control unit 13 generates an output replica signal in which the predistortion compensation signal A is estimated by multiplying the output signal Y by a value obtained by substituting the output signal Y into the distortion compensation polynomial. The coefficient of the distortion compensation polynomial is updated so that the error between the input replica signal and the output replica signal is minimized.

  The predistorter 301 in FIG. 2 calculates a ′ (i (N + M) + n) of the predistortion signal A in the distortion compensation unit 11 using w (i) calculated in the distortion compensation coefficient calculation unit 13. To the compensated circuit 401. Since the distortion-compensated predistortion compensation signal A is input, the compensated circuit 401 obtains y (i (N + M) + n) of the output signal Y.

ここで、Ｙ（ｉ）は出力信号Ｙの出力サンプリング信号の行列、Ｙ^Ｈ（ｉ）はＹ（ｉ）の複素共役転置行列、ｅ（ｉ）は前記歪補償多項式に入力信号を代入して得た値と前記入力信号とを乗算して入力レプリカ信号と前記歪補償多項式に出力信号を代入して得た値と前記出力信号とを乗算して出力レプリカ信号との差分行列である。 (Update amount calculation algorithm)
A matrix of an amount Δw (i) for updating the coefficient w (i) of the distortion compensation polynomial is expressed by Expression 34.

Here, Y (i) is the matrix of the output sampling signal of the output signal Y, Y ^H (i) is the complex conjugate transpose matrix of Y (i), and e (i) is the input signal substituted into the distortion compensation polynomial. It is a difference matrix between the output replica signal obtained by multiplying the output signal by the value obtained by multiplying the obtained value by the input signal and substituting the output signal into the distortion compensation polynomial.

但し、μは０＜μ≦１．０を満たす。 For this reason, the distortion compensation unit 11 updates the coefficient w (i) as in Expression 35.

However, μ satisfies 0 <μ ≦ 1.0.

In the conventional method, the matrix Y (i) is calculated using the output signal Y, the complex conjugate transpose matrix Y ^H (i) is calculated, and Y ^H (i) Y (i) and Y ^H (i) e are calculated. Since (i) was calculated, it was necessary to hold the input / output signals of N samples as they were.

とおく。但し、

である。このとき、行列Ｙ^Ｈ（ｉ）Ｙ（ｉ）は

となるので、例えば行列Ｙ^Ｈ（ｉ）Ｙ（ｉ）の一部である行列Ｙ_１ ^Ｈ（ｉ）Ｙ_１（ｉ）のａ行ｂ列要素Ｚ_{１，１，ａ，ｂ}は

であるので、従来はＮ回の乗算とＮ−１回の加算が必要であり、かつ加算結果Ｚ_{１，１，ａ，ｂ}とＮ個の出力信号を保持しておくメモリが必要であった。ここで、ｌは０以上Ｎ−１以下の整数である。但し、ここでは記述の簡単のためにｙ（ｉ（Ｎ＋Ｍ）＋ｌ）をｙ（ｌ）としている。 here,

far. However,

It is. At this time, the matrix Y ^H (i) Y (i) is

Since the, for example, the matrix ^Y H (i) a line of Y is part matrix _Y ¹ H in _{(i) (i) Y 1} (i) b sequence component _{Z 1, 1, a, b} is

Therefore, conventionally, N multiplications and N-1 additions are necessary, and a memory for holding the addition results Z _{1, 1, a, b} and N output signals is required. . Here, l is an integer of 0 or more and N−1 or less. However, here, y (i (N + M) + l) is set to y (l) for simplicity of description.

と変形できる。nサンプル目までの出力信号を用いて求めたＺ_{１，１，ａ，ｂ}をＺ_{１，１，ａ，ｂ}（ｎ）とおくと、数式３９、および数式４０のＺ_{１，１，ａ，ｂ}はＺ_{１，１，ａ，ｂ}（Ｎ−１）となり、

で表される。但し、ここでも記述の簡単のためにｙ（ｉ（Ｎ＋Ｍ）＋ｎ）をｙ（ｎ）としている。 Here, Z _{1, 1, a, b} in Equation 39 is

And can be transformed. If Z _{1,1, a, b} obtained using the output signals up to the nth sample is Z _{1,1, a, b} (n), Z _{1,1, a, b} becomes Z _{1,1, a, b} (N−1),

It is represented by Here, however, y (i (N + M) + n) is set to y (n) for simplicity of description.

により逐次的に算出できることとなる。ここまでの説明で、行列要素Ｚ_{１、１、ａ、ｂ}（ｎ）が逐次的に算出できることを明らかにした。しかし、｜ｙ（ｎ）｜^{２ａ＋２ｂ−２}を求める際、全ての行列要素Ｚ_{１、１、ａ、ｂ}（ｎ）においてｙ（ｎ）を（２ａ＋２ｂ−２）乗するのでは、従来の方法に比べてかえって計算量が多くなってしまう。そこで、本発明では出力信号のサンプルｙ（ｎ）毎に数式２８の出力信号ベクトルを予め算出しておき、数式４３によりＺ_{１、１、ａ、ｂ}（ｎ）を算出する。

ただし、上付きの＊は複素共役を表す。 That is, in general, the matrix element obtained using all the output signals up to an arbitrary n-th sample is a recurrence formula

Can be calculated sequentially. In the description so far, it has been clarified that the matrix elements _{Z1,1, a, b} (n) can be calculated sequentially. However, when obtaining | y (n) | ^{2a + 2b-2} , if y (n) is raised to the power of (2a + 2b-2) in all matrix elements Z _{1, 1, a, b} (n), the conventional method is used. In comparison, the amount of calculation increases. Therefore, in the present invention, the output signal vector of Equation 28 is calculated in advance for each sample y (n) of the output signal, and Z _{1, 1, a, b} (n) is calculated by Equation 43.

However, the superscript * represents a complex conjugate.

Therefore, in the calculation of the matrix Y ₁ ^H (i) Y ₁ (i), N multiplications and N−1 additions are necessary as in the conventional case, but it is necessary to store them in the memory. Only the addition result Z _{1,1, a, b} (n) and the latest output signal y (n) are obtained. Therefore, according to the method of the present invention, the memory of the output signal is reduced to 1 / N, and the larger the N, the smaller the amount of memory to be held.

より

となる。但し、記述の簡単のために
ｙ_ｌ，ａ（ｉ）＝ｙ（ｉ（ｎ＋Ｍ）＋ｎ）｜ｙ（ｉ（Ｎ＋Ｍ）＋ｌ）｜^２ａ−２
、および
ｅ_ｌ（ｉ）＝ｅ（ｉ（Ｎ＋Ｍ）＋ｌ）
をそれぞれ、ｙ（ｌ）、及びｅ（ｌ）としている。ここでも、Ｚ_{１，１，ａ，ｂ}のときと同様にＶ_１，ａは

と変形できる。ここで、nサンプル目までの出力信号を用いて求めたＶ_１，ａをＶ_１，ａ（ｎ）とおくと、数式４５、および数式４６のＶ_１，ａはＶ_１，ａ（Ｎ−１）となり、

で表される。 Similarly, a row element _{V 1, a} part of the matrix ^{Y H (i) e (i} ) Y 1 H (i) e (i) is

Than

It becomes. However, for simplicity of description, y _{l, a} (i) = y (i (n + M) + n) | y (i (N + M) + l) | ^2a−2
And e _l (i) = e (i (N + M) + l)
Are y (l) and e (l), respectively. Here again, V _{1, a} is the same as Z _{1,1, a, b.}

And can be transformed. Here, placing the _{V 1, a} obtained by using the output signal to the n-th sample and _{V 1, a (n),} V 1, a formula 45, and formula 46 is _{V 1,} a (N- 1)

It is represented by

により逐次的に算出できる。 That is, in general, the matrix element obtained using all the output signals up to an arbitrary n-th sample is a recurrence formula

Can be calculated sequentially.

により入出力信号のサンプル毎に算出できる。ただし、

である。 Regarding each element of the error signal e (n),

Can be calculated for each sample of input / output signals. However,

It is.

Thus, in the operation of the matrix Y ^{H e,} conventionally requires N multiplications and N-1 additions, and hold the addition result V _{1, a} and N output signals N input signals However, according to the present invention, only a memory for holding one output signal and one input signal is required according to the present invention, although N multiplications and N-1 additions are necessary. Necessary.

(Comparison of memory)
In this section, the amount of memory used in the present method and the present invention is compared. In the conventional method, an N-row (Q + 1) column matrix Y and a matrix X corresponding to N sampling points are calculated, and an error vector e of N rows and 1 column is calculated (Q + 1) rows (Q + 1). The matrix Y ^H Y of the column and the matrix Y ^H e of (Q + 1) rows and 1 column were calculated. ^{Thus, Y} H Y, and ^{Y H} e ^Y in operation H Y, and ^{Y H} Excluding memory to hold the e 2N (Q + 1) + N
An amount of memory proportional to was required.

になる。一般にＮは数千以上であるので、大幅にメモリ量が削減できる。従って、多項式係数（もしくはルックアップテーブル）の更新に用いる入出力信号のサンプリングポイント数が多いほど、本発明によるメモリ量の削減効果が高い。 On the other hand, in the present invention, in the calculation of Y ^H Y and Y ^H e, only the vectors y and x of 1 row (Q + 1) column corresponding to one sampling point, and the error vector e of N rows 1 column only. Except for the memory that holds the calculation results of Y ^H Y and Y ^H e that are also necessary in the conventional method,
2 (Q + 1) +1
An amount of memory proportional to is required. That is, the amount of memory required in the present invention compared to the conventional method is

become. Since N is generally several thousand or more, the amount of memory can be greatly reduced. Therefore, as the number of input / output signal sampling points used for updating the polynomial coefficient (or look-up table) increases, the memory amount reduction effect according to the present invention increases.

(Embodiment 2)
FIG. 4 is a block diagram illustrating the configuration of the predistorter 302 according to this embodiment. The predistorter 302 outputs a predistortion compensation signal A obtained by distortion-compensating the input signal X with a distortion compensation value based on a predetermined reference to the compensated circuit 401, and a predistortion compensation signal at different sampling points. An update amount of the reference is calculated from a plurality of predistortion sampling signals that include A and a plurality of output sampling signals that include the output signal Y from the compensated circuit 401 at the same sampling point. A control unit 13 to be updated.

  The predistorter 302 operates in the same manner as the predistorter 301 described in the first embodiment. Specifically, the control unit 13 captures the predistortion sampling signal and the output sampling signal from the predistortion compensation signal A and the output signal Y, respectively, at predetermined sampling points. Further, the control unit 13 calculates the update amount of the distortion compensation polynomial coefficient or the table value of the lookup table, and updates the distortion compensation polynomial coefficient or lookup table value held by the compensation unit 11.

  The predistorter according to the present invention can be applied to a power amplifier of a radio transmitter used in a mobile communication base station or the like.

It is a block diagram explaining the structure of the predistorter which concerns on this invention. It is a block diagram explaining the distortion compensation part of the predistorter which concerns on this invention. It is a block diagram explaining the structure of the predistorter which concerns on this invention. It is a block diagram explaining the structure of the predistorter which concerns on this invention.

Explanation of symbols

301, 302: Predistorter 11: Distortion compensation unit 13: Control unit 401: Compensated circuit 511: Delay element 512-j: Amplitude value function (j is a natural number)
513: Element amplifier 514: Multiplier 515: Adder X: Input signal Y: Output signal A: Predistortion compensation signal

Claims (5)

A distortion compensation unit that compensates the distortion of the input signal with a distortion compensation value based on a predetermined criterion and outputs the distortion to the compensated circuit;
A plurality of input sampling signals that capture the input signal at different sampling points, and a plurality of output sampling signals that capture the output signal from the compensated circuit at the sampling point, and the orthogonal coordinate that handles the complex update number of the reference update amount . calculated by the matrix equation, a control unit for updating the reference to the distortion compensating unit,
A predistorter comprising:
The predistorter, wherein the control unit can vary the number of sampling points.
The distortion compensation unit
Possesses a distortion compensation polynomial, and the criterion is a coefficient of the distortion compensation polynomial;
The controller is
A value obtained by multiplying the Nth (N is an integer of 1 or more) output sampling signal to the power of (2a + 2b−2) (a and b are integers of 1 or more) up to the N−1th output sampling signal. Is added to the matrix element of a row and b column calculated in step (b) to calculate an output signal matrix as a matrix element of the Nth a row and b column, and the complex conjugate transpose matrix of the output signal matrix is multiplied from the left side to obtain the first Calculate the matrix
A value obtained by raising the Nth input sampling signal to the power of (2a + 2b−2) is added to the matrix element of a row and b column calculated up to the N−1th input sampling signal, and the Nth a The distortion compensation polynomial is calculated by calculating an input signal matrix having matrix elements of row b columns and calculating the difference between the input signal matrix calculated from N input sampling signals and the output signal matrix calculated from N output sampling signals. Is multiplied by a matrix of coefficients from the right side to obtain an error matrix, and the error matrix is multiplied by a complex conjugate transpose matrix of N output signal matrices from the left side to calculate a second matrix,
The predistorter according to claim 1, wherein an update amount of a coefficient of the distortion compensation polynomial is calculated by multiplying an inverse matrix of the first matrix and the second matrix. The distortion compensation unit
Has a lookup table, and the criterion is a table value of the lookup table;
The controller is
A value obtained by multiplying the Nth (N is an integer of 1 or more) output sampling signal to the power of (2a + 2b−2) (a and b are integers of 1 or more) up to the N−1th output sampling signal. Is added to the matrix element of a row and b column calculated in step (b) to calculate an output signal matrix as a matrix element of the Nth a row and b column, and the complex conjugate transpose matrix of the output signal matrix is multiplied from the left side to obtain the first Calculate the matrix
A value obtained by raising the Nth input sampling signal to the power of (2a + 2b−2) is added to the matrix element of a row and b column calculated up to the N−1th input sampling signal, and the Nth a The distortion compensation polynomial is calculated by calculating an input signal matrix having matrix elements of row b columns and calculating the difference between the input signal matrix calculated from N input sampling signals and the output signal matrix calculated from N output sampling signals. Is multiplied by a matrix of coefficients from the right side to obtain an error matrix, and the error matrix is multiplied by a complex conjugate transpose matrix of N output signal matrices from the left side to calculate a second matrix,
The predistorter according to claim 1, wherein an update amount of a table value of the lookup table is calculated by multiplying an inverse matrix of the first matrix and the second matrix.   The control unit monitors an integrated value of error power between the input signal and the output signal, and causes the distortion compensation unit to update the reference when the integrated value exceeds a predetermined threshold value. The predistorter according to claim 1.   The control unit monitors error power between the input signal and the output signal, and reduces the number of sampling points when the error power exceeds a predetermined threshold than when the error power is within a predetermined threshold. The predistorter according to claim 1, wherein the reference is updated.

Images