JP2009200694A - Distortion compensation circuit and predistortion type distortion compensating amplifier - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a distortion compensation circuit which imparts predistortion to an input signal using a polynomial with input signals at a plurality of times taken into consideration, the distortion compensation circuit accurately calculating each coefficient of the polynomial. <P>SOLUTION: A predistortion compensation section 16 generates a predistortion signal y(n) obtained by imparting predistortion to an input signal x(n) based on a first polynomial being a polynomial of input signals x(n-k) at a plurality of times. A polynomial coefficient calculation section 70 calculates each coefficient of a second polynomial being a polynomial of the input signal x(n) as to a plurality of frequency domains based on the input signal x(n) and an output signal from an amplifier 12 when nonlinear characteristics of the amplifier 12 are represented by approximation by the second polynomial. A distortion coefficient calculation section 46 calculates each coefficient of the first polynomial based on the each coefficient of the second polynomial calculated for each of the frequency domains by the polynomial coefficient calculation section 70. <P>COPYRIGHT: (C)2009,JPO&INPIT

Description

  The present invention relates to a distortion compensation circuit and a predistortion type distortion compensation amplifier that perform distortion compensation by applying predistortion to an input signal to an amplifier having nonlinear characteristics.

  In a predistortion type distortion compensation amplifier, distortion compensation is performed by giving predistortion (predistortion) to the input signal to the amplifier in advance to cancel distortion components generated in the output signal from the amplifier due to the nonlinear characteristics of the amplifier. . In order to reduce the distortion component remaining in the output signal from the amplifier, it is necessary to bring the predistortion characteristic applied to the input signal to the amplifier close to the inverse characteristic of the input / output characteristic of the amplifier. In order to bring the predistortion characteristics closer to the inverse characteristics of the input / output characteristics of the amplifier, the inverse characteristics of the input / output characteristics of the amplifier are approximated by a polynomial of the input signal so as to reduce the distortion component remaining in the output signal from the amplifier The polynomial coefficients are determined.

  However, the distortion characteristics of the amplifier may be affected by the memory effect, and the distortion compensation amount may be limited due to asymmetric frequency characteristics of distortion components generated in the output signal from the amplifier. In polynomial approximation considering only the instantaneous value of the input signal, it is difficult to compensate for a distortion component whose frequency characteristics are asymmetric.

In order to compensate for distortion components whose frequency characteristics are asymmetric, a related technique for giving predistortion to an input signal using a polynomial that takes into account the input signal at a plurality of times has been proposed. It is disclosed. In this related technique, an input signal at the current sample time n (n is an integer) is x (n), and an input signal at time (n−k) before k samples (k is an integer of 0 or more) from the time n is x. Assuming that (n−k), a predistortion signal y (n) obtained by applying predistortion to the input signal x (n) is calculated by a polynomial expressed by the following equation (1). In Equation (1), the input signal x (n−k), the predistortion signal y (n), and the distortion compensation coefficients A _{1 and k} respectively set corresponding to the values of l and k are complex numbers. It is. Each distortion compensation coefficient A _{l, k} is determined using, for example, a perturbation method in which each distortion compensation coefficient A _{l, k} is updated in a direction in which distortion component power remaining in the output signal from the amplifier is reduced.

  By calculating the predistortion signal y (n) using the polynomial expressed by the equation (1), it becomes possible to give the predistortion signal y (n) a frequency characteristic, and a distortion component whose frequency characteristic is asymmetric is obtained. It becomes possible to compensate.

JP 2006-279780 A JP 2004-320598 A JP 2002-57533 A

  To improve predistortion distortion compensation performance when predistortion is applied to an input signal using a polynomial that takes into account input signals at multiple times, the distortion components remaining in the output signal from the amplifier are reduced. Therefore, it is necessary to calculate each coefficient of the polynomial for this purpose with high accuracy. As a calculation method for each coefficient of the polynomial, a perturbation method is generally used in which the value of each coefficient is updated in a direction in which distortion component power remaining in the output signal from the amplifier is reduced. However, in the perturbation method, it is necessary to update the value of each coefficient while determining whether or not the distortion component power, which is a scalar amount, is reduced before and after changing the value of each coefficient. It takes a lot of time to converge. Also, if noise is mixed in the detected distortion component power, the value of each coefficient is not necessarily updated in the direction in which the actual distortion component power is reduced, and each coefficient does not converge to the optimum value and is likely to diverge. . Therefore, in the perturbation method, it is difficult to accurately calculate each coefficient of a polynomial for reducing distortion components remaining in the output signal from the amplifier.

  The present invention reduces distortion components remaining in an output signal from an amplifier in a distortion compensation circuit and a predistortion type distortion compensation amplifier that give a predistortion to an input signal using a polynomial in consideration of an input signal at a plurality of times. The purpose is to calculate each coefficient of a polynomial for the purpose with high accuracy.

  The distortion compensation circuit and the predistortion type distortion compensation amplifier according to the present invention employ the following means in order to achieve the above-described object.

  The distortion compensation circuit according to the present invention generates a predistortion signal in which a predistortion is applied to an input signal to an amplifier so that a distortion component remaining in an output signal from an amplifier having nonlinear characteristics is reduced. A distortion compensation circuit for amplifying a distortion signal by an amplifier, wherein an input signal at a current sample time is x (n), and an input signal at a time before k samples (k is an integer of 0 or more) is x (n−k). ), A predistortion compensation unit that generates the predistortion signal based on a first polynomial that is a polynomial of the input signal x (n−k) at a plurality of times, and the nonlinear characteristic of the amplifier is represented by the input signal x ( n), when approximated by a second polynomial, which is a polynomial of n), each coefficient of the second polynomial is respectively expressed for a plurality of frequency domains based on the input signal x (n) and the output signal from the amplifier. The polynomial coefficient calculator to be operated and the polynomial coefficient calculator The gist of the present invention is to include a distortion compensation coefficient calculation unit that calculates each coefficient of the first polynomial based on each coefficient of the second polynomial calculated for each wave number region.

  In one aspect of the present invention, the distortion compensation coefficient calculation unit calculates the frequency characteristic of each coefficient of the second polynomial based on each coefficient of the second polynomial calculated for each frequency domain by the polynomial coefficient calculation unit. Then, it is preferable to convert the frequency characteristic of each coefficient of the second polynomial into an impulse response, and calculate each coefficient of the first polynomial based on the impulse response of each coefficient of the second polynomial.

  In one aspect of the present invention, the polynomial coefficient arithmetic unit extracts a specific frequency component in the output signal from the distortion generator that outputs the distortion of the input signal x (n) based on the second polynomial and outputs the distortion. A first filter that can change a frequency component to be extracted, and a second filter that extracts a specific frequency component in an output signal from the distortion generator, A second filter capable of changing the frequency component to be performed, a filter control unit for changing the frequency component extracted by the first and second filters in a plurality of ways, and the first and second filters by the filter control unit When the frequency component extracted by the filter is changed in a plurality of ways, the correlation value between the output signal from the amplifier that has passed through the first filter and the output signal from the distortion generator that has passed through the second filter A correlation calculation unit that calculates for each wave number component; a polynomial coefficient control unit that controls each coefficient of the second polynomial for each frequency region based on the correlation value calculated for each frequency component by the correlation calculation unit; It is preferable to have

  In one aspect of the present invention, the polynomial coefficient calculation unit calculates the input / output relationship of the amplifier with respect to a plurality of frequency domains based on the input signal x (n) and the output signal from the amplifier. And the characteristics of the input / output relationship with respect to the power value of the input signal x (n) based on the input signal x (n) and the input / output relationship calculated for each frequency region by the input / output relationship calculation unit. A non-linear characteristic calculating unit that calculates for each region; a polynomial coefficient setting unit that sets each coefficient of the second polynomial for each frequency region based on the characteristics calculated for each frequency region by the non-linear characteristic calculating unit; It is preferable to have

  The predistortion type distortion compensation amplifier according to the present invention includes an amplifier having nonlinear characteristics and a predistorter that applies predistortion to an input signal to the amplifier so that a distortion component remaining in the output signal from the amplifier is reduced. A predistortion type distortion compensation amplifier that amplifies the predistortion signal generated by the distortion compensation circuit with an amplifier, wherein the distortion compensation circuit is a distortion compensation circuit according to the present invention. The gist is that it is a compensation circuit.

  According to the present invention, in order to reduce distortion components remaining in the output signal from the amplifier, the input signal x (n) is based on the polynomial (first polynomial) of the input signal x (nk) at a plurality of times. ), The coefficients of the polynomial (second polynomial) of the input signal x (n) approximating the nonlinear characteristics of the amplifier are calculated for a plurality of frequency regions, respectively. Then, by calculating each coefficient of the first polynomial based on each coefficient of the second polynomial calculated for each frequency region, a first distortion component for reducing distortion components remaining in the output signal from the amplifier is obtained. Each coefficient of the polynomial of 1 can be calculated with high accuracy.

  DESCRIPTION OF EMBODIMENTS Hereinafter, embodiments for carrying out the present invention (hereinafter referred to as embodiments) will be described with reference to the drawings.

  1 to 3 are diagrams schematically illustrating a configuration of a predistortion type distortion compensation amplifier including a distortion compensation circuit according to an embodiment of the present invention. The predistortion type distortion compensation amplifier according to the present embodiment applies predistortion (predistortion) to the amplifier 12 for reducing distortion components remaining in the output signal from the amplifier (power amplifier) 12 having nonlinear characteristics. Distortion compensation is performed by giving the input signal in advance.

  In the control unit 14, an input signal x (n) that is a digital signal (data series) generated every predetermined sample period is input to the predistortion unit 16. Here, n represents the sampling time of the digital signal. The input signal x (n) is a complex signal (modulation signal) having a real part (in-phase component I (n)) and an imaginary part (quadrature component Q (n)), and has a frequency bandwidth corresponding to the modulation. . The input signal x (n) can be expressed by the following equation (2) using the in-phase component I (n) and the quadrature component Q (n). The input signal x (n) may be a baseband signal or an IF signal.

  x (n) = I (n) + j · Q (n) (2)

  The predistortion compensation unit 16 generates and outputs a predistortion signal (predistortion signal) y (n) obtained by adding predistortion (predistortion) to the input signal x (n) to the amplifier 12. The predistortion signal (digital signal) y (n) to which the predistortion is given by the predistortion unit 16 is output from the control unit 14 via the D / A converter 18, whereby the analog signal y (t ) And output. Here, t represents the time of the analog signal. The details of the predistortion given by the predistortion unit 16 will be described later.

  The mixer 20 mixes the predistortion signal y (t) supplied from the control unit 14 (D / A converter 18) with the oscillation signal output from the oscillator 22 to thereby convert the predistortion signal y (t) to RF. Upconvert to signal and output. The amplifier 12 amplifies the RF signal supplied from the mixer 20, that is, the predistortion signal y (t) to which predistortion is given, and outputs the amplified signal to the output terminal OUT. The predistortion signal y (t) amplified by the amplifier 12 becomes an RF signal having a frequency bandwidth corresponding to the modulation. Since the input / output characteristics of the amplifier 12 are non-linear characteristics, distortion occurs when the signal is amplified by the amplifier 12, and the output signal z (t) from the amplifier 12 includes an amplified input signal component (hereinafter referred to as a main signal). In addition to a component, a distortion component is also included. Therefore, in order to cancel the distortion component caused by the nonlinear characteristic of the amplifier 12, the predistortion unit 16 applies the predistortion to the input signal x (n) to the amplifier 12 in advance. At that time, the distortion component generated in the output signal z (t) from the amplifier 12 is canceled by matching the predistortion characteristic given by the predistortion unit 16 with the inverse characteristic of the input / output characteristic of the amplifier 12. be able to. Hereinafter, a configuration for applying the predistortion to the input signal x (n) to the amplifier 12 by the predistortion unit 16 will be described.

Assuming that the input signal at the current sample time n is x (n) and the input signal at time (n−k) k samples before the time n (k is an integer of 0 or more) is x (n−k), the predistortion The compensator 16 calculates the predistortion signal y (n) using the polynomial expressed by the above-described equation (1). As described above, when the distortion characteristics of the amplifier 12 are affected by the memory effect, the frequency characteristics of distortion components generated in the output signal z (t) from the amplifier become asymmetrical. By calculating the predistortion signal y (n) using the polynomial of the input signal x (nk) at the sample time, it is possible to give the predistortion signal y (n) a frequency characteristic, and the frequency characteristic is asymmetric. It becomes possible to compensate for the distortion component. However, in the formula (1), A _{l, k} · | x (n−k) | ^l · x (n) is not necessarily associated with all l values from 0 to (L−1) (L is an integer of 2 or more). there is no need to integrate the value of -k), for example the even only l values with respect to _{a l, k · | x (} n-k) | by integrating the values of ^{l · x (n-k)} y (n ) Value may be calculated. Further, in the expression (1), A _{l, k} · | x (n−k) | ^l · x (n) is not necessarily associated with all values of k from 0 to (K−1) (K is an integer of 2 or more). The value of -k) may not be integrated.

FIG. 2 shows a configuration example of a distortion compensation circuit for generating the predistortion signal y (n) expressed by the equation (1). In FIG. 2, the sum (A _{l, 0} · | x (n) | ^l · x (n) + A _{l, 1} · | (x (n−1) | ^l · x (n−1) +...)) is shown, but in reality, a power operation unit 152, a multiplier 154, a delay element, which will be described below, are shown. The configurations of 156-1 to 156-4 and multipliers 158-1 to 158-5 calculate the sum of terms of each order in the polynomial y (n) of x (n−k). A plurality are provided corresponding to

The power calculator 152 calculates and outputs the l-th power (power multiplier) | x (n) | ^l of the absolute value (amplitude level) of the input signal x (n) based on the input signal x (n). Multiplier 154, the input signal x (n) and the exponent | x (n) | product of ^l | x (n) | by calculating ^l · x (n) and outputs. The cascade-connected delay elements 156-1 to 156-4 are connected to a signal | x (n) || (x (n) | ^l · x (n) delayed by k samples for a plurality of one or more values of k. n−k) | ^l × x (n−k). In the configuration example shown in FIG. 2, the signal | x (n-1) | ^l · x (n−) obtained by delaying the signal | x (n) | ^l · x (n) by one sample from the delay element 156-1. 1) is output, and the signal | x (n-2) | ^l · x (n-2) obtained by delaying the signal | x (n) | ^l · x (n) by two samples from the delay element 156-2 Is output from the delay element 156-3, and the signal | x (n-3) | ^l · x (n-3) obtained by delaying the signal | x (n) | ^l · x (n) by three samples is output. The delay element 156-4 outputs a signal | x (n-4) | ^l · x (n-4) obtained by delaying the signal | x (n) | ^l · x (n) by four samples. . In each of the delay elements 156-1 to 156-4, processing for delaying the complex number | x (n) | ^l · x (n) is performed.

Multiplier 158-1～158-5 is distortion compensation coefficient A _{l, k} and the signal | x (n-k) | product A _l with ^{l · x (n-k)} , k · | x (n-k ) | ^l · x (n−k) is calculated for each of a plurality of 0 or more k values and output. In the configuration example illustrated in FIG. 2, the multiplier 158-1 includes a product A _{l, 0} · | of the distortion compensation coefficient A _{l, 0} and the signal | x (n) | ^l · x (n) from the multiplier 154. x (n) | ^l · x (n) is calculated, and the multiplier 158-2 calculates the distortion compensation coefficient A _{l, 1} and the signal | x (n-1) | ^l · x (from the delay element 156-1. n−1) and the product A _{l, 1} · | x (n−1) | ^l · x (n−1) are calculated, and the multiplier 158-3 calculates the distortion compensation coefficient A _{l, 2} and the delay element 156. -2 from the signal | x (n-2) | ^l · x (n-2) and the product A _{l, 2} · | x (n-2) | ^l · x (n-2) The unit 158-4 generates a product A _{l, 3} · | x (n) of the distortion compensation coefficient A _{l, 3} and the signal | x (n-3) | ^l · x (n-3) from the delay element 156-3. -3) | ^l · x (n-3) is calculated, and the multiplier 158-5 calculates the distortion compensation coefficient A _{l, 4} and the signal | x (n-4) | ^l · x from the delay element 156-4. the product of (n-4) and _{a l, 4 · | x (} n-4) | calculating the ^l · x (n-4) That. In each multiplier 158-1 to 158-5, multiplication of the complex number A _{l, k} and the complex number | x (n−k) | ^l · x (n−k) is performed.

The adder 160 is a product A _{l, k} · | x (n−k) | ^l · x (n−k) calculated by the multipliers 158-1 to 158-5 (k in the configuration example shown in FIG. 2). = 0 to 4) and calculating the sum thereof _, the convolution sum ΣA _{l, k} of the distortion compensation coefficient A _{l, k} and the signal | x (n−k) | ^l × x (n−k) Calculate | x (n−k) | ^l × x (n−k). In the configuration example shown in FIG. 2, the convolution sum ΣA _{l, k} · | x (n−k) | ^l · x (n−k) is expressed as A _{l, 0} · | x (n) | ^l · x (n) +... + A _{l, 4} · | x (n−4) | ^l · x (n−4). However, the number of delay elements 156-1 to 156-4 and multipliers 158-1 to 158-5 can be arbitrarily set, and the convolution sum ΣA _{l, k} · | x (n−k) | signal is integrated in calculating the ^{l · x (n-k)} a l, k · | x (n-k) | for the number of ^l · x (n-k) is possible to arbitrarily set It is. As described above, the arithmetic unit 152, the multiplier 154, the delay elements 156-1 to 156-4, the multipliers 158-1 to 158-5, and the adder 160 are included, and the convolution sum ΣA _{l, k} · | x It is possible to configure a convolution operation unit that calculates (n−k) | ^l · x (n−k).

The convolution operation units (power operation unit 152, multiplier 154, delay elements 156-1 to 156-4, and multipliers 158-1 to 158-5) are each order in the polynomial y (n) of x (nk). A plurality of products A _{l, k} · | x (nk) | ^l · x (nk) are respectively calculated with respect to a plurality of values of l and k. In the adder 160, the convolution sum ΣA _{l, k} · | x (n−k) | ^l · x (n−k) is calculated for each of a plurality of values of l. For example, the exponent operation unit 152, the multiplier 154, the delay elements 156-1 to 156-4, and the multipliers 158-1 to 158-5 are connected to odd-orders (first order, third order, fifth order, ...) And a convolution sum ΣA _{l, k} · | x (n−k) | ^l · x (n−k) for each of a plurality of even-numbered l values (0, 2, 4,...). ) Can also be calculated. The adder 160 accumulates the convolution sums ΣA _{l, k} · | x (n−k) | ^l · x (n−k) calculated for each value of ^l and calculates the sum. , (1) A predistortion signal y (n) represented by a polynomial is generated and output. As described above, in the configuration shown in FIG. 2, the predistortion signal y (n) is generated using the convolution operation (FIR filter), and each distortion compensation coefficient A _{l, k} corresponds to the tap coefficient of the FIR filter. To do. When calculating the predistortion signal y (n), the product A _{l, k} · | x (n−k) | ^l · x (n−k) calculated for each value of k is given first. The product A _{l, k} · | x (nk) | ^l · x (nk) calculated for each value of l can be accumulated first.

Next, a configuration example for calculating each distortion compensation coefficient A _{l, k} will be described. As shown in FIG. 1, a directional coupler (coupler) 34 is provided between the amplifier 12 and the output terminal OUT, and a part of the output signal z (t) from the amplifier 12 is directional coupler. 34. The mixer 36 mixes the oscillation signal output from the oscillator 22 with the output signal (hereinafter referred to as a feedback signal) resp (t) extracted by the directional coupler 34, whereby the feedback signal (RF signal) resp. (t) is down-converted to an IF signal or a baseband signal and output. Only a low-frequency component (IF component or baseband component) is extracted from the signal output from the mixer 36 by a filter 38 provided on the output side of the mixer 36. The feedback signal resp (t) after passing through the filter 38 is converted from an analog signal to a digital signal (data series) resp (n) by the A / D converter 40 and then to the polynomial coefficient calculation unit 70 in the control unit 14. Supplied. The polynomial coefficient calculation unit 70 is also supplied with an input signal x (n) before being given predistortion. The polynomial coefficient calculation unit 70 approximates the input signal x (n) and the feedback signal resp (when the nonlinear characteristic of the amplifier 12 is approximated by a polynomial of the input signal x (n) expressed by the following equation (3). n) Based on (the output signal from the amplifier 12), each coefficient B _{l, f} of the polynomial represented by the equation (3) is calculated and output for a plurality of frequency bands. Then, the distortion compensation coefficient calculation unit 46 calculates each distortion compensation coefficient A _{l, k} based on each coefficient B _{l, f} calculated for each frequency band by the polynomial coefficient calculation unit 70.

  A configuration example of the polynomial coefficient calculation unit 70 is shown in FIG. The distortion generator 71 calculates a signal (hereinafter referred to as an approximate signal) fit (n) that distorts the input signal x (n) by a polynomial of the input signal x (n) expressed by the equation (3). And output. Since equation (3) is a polynomial that approximates the nonlinear characteristic of the amplifier 12, a distortion corresponding to the nonlinear distortion of the amplifier 12 is given to the approximate signal fit (n). The variable filter 72 can extract a specific frequency component in the feedback signal resp (n) (output signal from the amplifier 12), and can further change the frequency component to be extracted (pass frequency band) in a plurality of ways. . The variable filter 74 can extract a specific frequency component in the approximate signal fit (n) output from the distortion generator 71, and can further change the frequency component (pass frequency band) to be extracted in a plurality of ways. . The filter characteristic control unit 76 controls the pass frequency band of the variable filters 72 and 74 (frequency components extracted by the variable filters 72 and 74), and the frequency band range of the approximate signal fit (n) and the feedback signal resp (n). The pass frequency bands of the variable filters 72 and 74 are changed in a plurality of ways. The filter characteristic control unit 76 changes the frequency components extracted by the variable filters 72 and 74 in a plurality of ways while keeping the pass frequency characteristics of the variable filters 72 and 74 equal.

The correlation calculation unit 78 passes the feedback signal resp (n) that has passed through the variable filter 72 and the variable filter 74 when the filter characteristic control unit 76 changes the frequency components extracted by the variable filters 72 and 74 in a plurality of ways. A correlation value with the approximate signal fit (n) is calculated and output for each frequency component. The polynomial coefficient control unit 80 controls each coefficient B _{l, f} of the polynomial expressed by equation (3) for each frequency band based on the correlation value calculated for each frequency component by the correlation calculation unit 78. . Assuming that the feedback signal resp (n) and the approximate signal fit (n) that have passed through the variable filters 72 and 74 are e2 (n) and e1 (n), respectively, the complex signal e2 (n) and the complex signal e1 (n) The correlation value (complex cross-correlation coefficient) ρ _m can be expressed by the following equation (4). In Equation (4), R is an integer representing the correlation length, and m is a relative time difference between the two signals e2 (n) and e1 (n) (m = 0 is the same time).

For example, as shown in FIG. 4, the filter characteristic control unit 76 changes the pass frequency bands (center frequencies) of the variable filters 72 and 74 to f0, f1, f2, f3, and f4, respectively. The correlation calculation unit 78 calculates a complex cross-correlation coefficient ρ _m between the feedback signal resp (n) and the approximate signal fit (n) for each frequency band f0 to f4. The polynomial coefficient control unit 80 searches for each coefficient B _{l, f} for each frequency band f0 to f4 such that the real part of the complex cross-correlation coefficient ρ _m calculated for each frequency band f0 to f4 is maximized. To do. As the search algorithm here, for example, a known algorithm such as a full search method or RMS or LMS can be used. By calculating each coefficient B _{l, f} that maximizes the correlation value between the feedback signal resp (n) and the approximate signal fit (n) for each frequency band f0 to f4, the nonlinear characteristic of the amplifier 12 is approximated. Can be obtained for each frequency band f0 to f4. When the frequency characteristic of the distortion component is asymmetric due to the memory effect of the amplifier 12, the coefficients B _{l, f} have different values depending on the frequency bands f0 to f4. Further, when calculating each coefficient B _{l, f} , the influence of the pass frequency characteristics of the variable filters 72 and 74 is canceled by equalizing the pass frequency characteristics of the variable filters 72 and 74. Note that the number of frequency bands for calculating the coefficients B _{l, f} is not limited to the five described above, and can be arbitrarily set.

An example of a polynomial that approximates the nonlinear characteristic of the amplifier 12 calculated for each frequency band f0 to f4 is shown in the following equation (5). In equation (5), the coefficient of the polynomial corresponding to the frequency band f0 is B _{l, f0} , the coefficient of the polynomial corresponding to the frequency band f1 is B _{l, f1} , and the coefficient of the polynomial corresponding to the frequency band f2 is B _l, The coefficient of the polynomial corresponding to _f2 and the frequency band f3 is B _{l, f3} , and the coefficient of the polynomial corresponding to the frequency band f4 is B _{l, f4} . The coefficients B _{l, f0 to} B _{l, f4} indicate the coefficients corresponding to the (l + 1) th order term in the polynomial expressed by the expression (5), and the expression (5) represents the first order, the third order, and the fifth order. An example in which the nonlinear characteristic of the amplifier 12 is approximated by the next term (odd-order term) is shown. However, the order of each term of the polynomial that approximates the nonlinear characteristic of the amplifier 12 is not limited to the first order, the third order, and the fifth order (odd order) shown in the equation (5), and may be arbitrarily set. Is possible. In order to eliminate uncertainty, all the coefficients B _{0, f0 to} B _{0, f4} are set to 1.

The distortion compensation coefficient calculation unit 46 is expressed by equation (1) based on the coefficients B _{l, f0 to} B _{l, f4} corresponding to the (l + 1) th order term in the polynomial calculated for each frequency band _{f0 to f4.} The distortion compensation coefficient A _{l, k} corresponding to the (l + 1) th order term in the polynomial is calculated. Here, the discrete coefficients B _{l, f0 to} B _{l, f4} for each frequency band _{f0 to f4} are functions representing the frequency characteristics of the coefficient B _{l, f in} the equation (3) as shown by the solid line in FIG. Is considered part of FIG. 5A shows the frequency characteristic of the amplitude of the coefficient B _{l, f} , and FIG. 5B shows the frequency characteristic of the phase of the coefficient B _{l, f} . In FIG. 5, the frequency on the horizontal axis is a normalized frequency based on f2 (set to 0). The distortion compensation coefficient calculation unit 46 calculates a function representing the frequency characteristic of the coefficient B _{l, f} as shown by the solid line in FIG. 5 based on each coefficient B _{l, f0 to} B _{l, f4} . In that case, the types of functions representing the frequency characteristics of the coefficients B _{l, f} are prepared in advance, and each coefficient of the function is calculated using a known method such as a least square method or linear interpolation. Then, the distortion compensation coefficient calculation unit 46 converts the function representing the frequency characteristics of the coefficients B _{l, f} from the frequency domain to the time domain using IFFT (inverse frequency transform), so that the time axis as shown in FIG. To the impulse response function (B _{l, 0} , B _{l, 1} , B _{l, 2} ). The tap coefficients (B _{l, 0} , B _{l, 1} , B _{l, 2} ) of the impulse response function here are complex numbers. Further, the sign of tap coefficients (B _{l, 0} , B _{l, 2 in the} example shown in FIG. 6) other than the center tap (B _{l, 1 in the} example shown in FIG. 6) of the impulse response is inverted. Here, the tap coefficient is considered as a complex vector, and complex vectors other than the center tap are inverted (rotated 180 °). Each tap coefficient (complex vector) of the inverted impulse response can be considered as each distortion compensation coefficient A _{l, k} . In the example shown in FIG. 6, A _{l, 0} = −B _{l, 0} , A _{l, 1} = B _{l, 1} , A _{l, 2} = −B _{l, 2} . However, the number of taps of the impulse response (the number of distortion compensation coefficients A1 _{, k} to be calculated) is not limited to 3 described above, and can be arbitrarily set. As described above, the distortion compensation coefficient A _{l, k} corresponding to the (l + 1) th order term can be obtained based on the impulse response function (B _{l, 0} , B _{l, 1} , B _{l, 2} ) on the time axis. it can. Then, the distortion compensation coefficient A _{l, k} is calculated for each term (each order) in the polynomial expressed by equation (1).

In the present embodiment described above, each coefficient B _{l, f} of the polynomial of the input signal x (n) approximating the nonlinear characteristic of the amplifier 12 is calculated with respect to a plurality of frequency bands f0 to f4, and each frequency band f0 to f4 is calculated. Each distortion compensation coefficient A _{l, k} is calculated based on each B _{l, f0 to} B _{l, f4} calculated for each _f4 . Perturbation method whereby, for updating the value of each distortion compensation coefficient A _l, the distortion compensation coefficient while determining whether or not the distortion component power is a scalar quantity reduced before and after changing the value of _k A _{l, k} Since each distortion compensation coefficient A _{l, k} can be calculated without using, each distortion compensation coefficient A _{l, k} can be set to an optimum value without diverging. Therefore, each distortion compensation coefficient A _{l, k} for compensating for distortion components whose frequency characteristics are asymmetric due to the memory effect of the amplifier 12 can be calculated with high accuracy.

  FIG. 7 shows the result of confirming the distortion compensation effect by the configuration of the present embodiment by actual measurement. FIG. 7 shows the result of actual measurement of the frequency characteristics of the output signal power from the amplifier 12, and as a comparison target with the configuration of the present embodiment (corresponding to the memory effect), distortion compensation is not performed (no distortion compensation). And the actual measurement result in the case where distortion compensation is performed considering only the instantaneous value of the input signal (memory effect not supported) are also shown. However, in FIG. 7, the frequency on the horizontal axis is a normalized frequency with the center frequency being zero. As shown in FIG. 7, when distortion compensation is performed in consideration of only the instantaneous value of the input signal (when the memory effect is not supported), the frequency characteristic of the distortion component generated in the output signal from the amplifier becomes asymmetric and distortion occurs. Although the amount of compensation is limited, it can be seen that the configuration of the present embodiment can compensate for a distortion component having an asymmetric frequency characteristic and improve the distortion compensation performance.

  Hereinafter, another configuration example of the present embodiment will be described.

  Another configuration example of the polynomial coefficient calculation unit 70 is shown in FIG. The frequency conversion unit 81 converts the feedback signal resp (n) (output signal from the amplifier 12) from time domain to frequency domain data resp (f) (f is a frequency) using FFT (frequency conversion) and outputs the result. To do. The data selection unit 82 can extract a specific frequency component in the feedback signal resp (f) converted into the frequency domain by the frequency conversion unit 81, and can change the frequency component to be extracted in a plurality of ways. The inverse frequency transform unit 83 transforms the feedback signal resp (f) from which the specific frequency component is extracted by the data selection unit 82 from the frequency domain to the time domain data resp (n) using IFFT (inverse frequency transform). Output. The frequency conversion unit 84 converts the input signal x (n) from time domain to frequency domain data x (f) using FFT (frequency conversion) and outputs the result. The data selection unit 85 can extract a specific frequency component in the input signal x (f) converted into the frequency domain by the frequency conversion unit 84, and can change the extracted frequency component in a plurality of ways. The inverse frequency transform unit 86 transforms the input signal x (f) from which the specific frequency component has been extracted by the data selection unit 85 from the frequency domain to the time domain data x (n) using IFFT (Inverse Frequency Transform). Output. Instead of the frequency conversion units 81 and 84, the data selection units 82 and 85, and the inverse frequency conversion units 83 and 86, a variable filter that can change a plurality of frequency components to be extracted can be used.

Based on the feedback signal resp (n) (output signal from the amplifier 12) from the inverse frequency converter 83 and the input signal x (n) from the inverse frequency converter 86, the input / output relation calculation unit 87 Are calculated for a plurality of frequency bands. Here, as the input / output relationship of the amplifier 12, a ratio resp / x (hereinafter referred to as a transfer function) between the feedback signal resp and the input signal x is calculated for each frequency band. The histogram calculation unit 88 calculates a histogram representing the characteristics of the transfer function resp / x with respect to the power value | x (n) | ² of the input signal x (n) for each frequency band. Here, the histogram class is the power value | x (n) | ² of the input signal x (n), and the value of the complex transfer function resp / x corresponding to the class (power value | x (n) | ² ) is set. Cumulative addition. When a certain time has elapsed, the cumulative addition value of the complex transfer function resp / x is divided by the frequency (the cumulative addition number of each class) and averaged. By this calculation, the characteristic of the complex transfer function resp / x (average value) with respect to the input signal power value | x (n) | ² is obtained. The gain characteristic of the complex transfer function resp / x with respect to the input signal power value (instantaneous input power) | x (n) | ² represents the AM-AM characteristic of the amplifier 12 as shown in FIG. The phase characteristic of the complex transfer function resp / x with respect to the value (instantaneous input power) | x (n) | ² represents the AM-PM characteristic of the amplifier 12 as shown in FIG. The characteristic of this complex transfer function resp / x can be approximated by a polynomial of the input signal power value | x (n) | ² .

The polynomial coefficient setting unit 90 is based on the characteristic of the complex transfer function resp / x with respect to the input signal power value | x (n) | ² calculated for each frequency band by the histogram calculation unit 88 (6) Each coefficient B _{l, f} of the polynomial expressed by the equation is set for each frequency band. Here, a known method such as a least square method is used so that the polynomial represented by the equation (6) matches (or substantially matches) the characteristics of the complex transfer function resp / x calculated by the histogram calculation unit 88. It is possible to set each coefficient B _{l, f} using. When the frequency characteristic of the distortion component is asymmetric due to the memory effect of the amplifier 12, the characteristic of the complex transfer function resp / x with respect to the input signal power value | x (n) | ² is different depending on the frequency band. The coefficients B _{l, f} have different values depending on the frequency band. In addition, the polynomial represented by the equation (6) represents a complex transfer function, and is a first order lower expression than the polynomial represented by the equation (5).

For example, the data selecting unit 82 extracts a frequency component extracted from the feedback signal resp (f) as a high frequency component higher than the center frequency of the feedback signal resp (f) and a low frequency lower than the center frequency of the feedback signal resp (f). The component can be changed in a plurality of ways including the total frequency component of the feedback signal resp (f). Similarly, the data selecting unit 85 extracts a frequency component extracted from the input signal x (f) as a high frequency component higher than the center frequency of the input signal x (f) and a low frequency lower than the center frequency of the input signal x (f). It can be changed in a plurality of ways, including the band component and all frequency components of the input signal x (f). In that case, the input / output relationship calculation unit 87 calculates the complex transfer function resp / x for the high frequency component, the low frequency component, and the total frequency component, respectively, and the histogram calculation unit 88 calculates the high frequency component and the low frequency component. The characteristic of the complex transfer function resp / x with respect to the input signal power value | x (n) | ² is calculated for all frequency components. Then, the polynomial coefficient setting unit 90 calculates each coefficient B _{l, f} of the polynomial expressed by the equation (6) for the high frequency component, the low frequency component, and the total frequency component. However, the number of frequency bands for calculating the coefficients B _{l, f} is not limited to the three described above, and can be set arbitrarily. Further, the order of each term of the polynomial that approximates the complex transfer function resp / x is not limited to the 0th order, the second order, and the fourth order shown in the equation (6), and can be arbitrarily set. is there.

  In this embodiment, the input signal at the current sample time n is x (n), and the input signal at time (n−k) k samples before time n (k is an integer of 0 or more) is x (n−k). ), Assuming that an input signal at time (n−m) m samples before time n (m is an integer of 0 or more) is x (n−m), the predistortion compensation unit 16 has the following equation (7): The predistortion signal y (nm) can also be calculated by a polynomial represented by

From the equation (7), y (nm) / x (nm) (hereinafter referred to as a transfer function) corresponding to the inverse characteristic of the input / output characteristics of the amplifier 12 is expressed by the following equation (8). The The above equation (1) gives the frequency characteristic to the predistortion signal y (n) itself, whereas the equation (8) shows the transfer function y (nm) / x (nm). Is different from the equation (1) in that frequency characteristics are given to. However, in the expressions (7) and (8), A _{l, k} · | x (n−k) | is not necessarily associated with all l values from 0 to (L−1) (L is an integer of 2 or more). ^There is no need to integrate the values of ^l . For example, the values of A _{l, k} · x (n−k) | ^l are integrated with respect to even values of _l, and y (n−m) and y (n−m). ) / X (nm) may be calculated. In the expressions (7) and (8), A _{l, k} · | x (n−k) | is not necessarily associated with all k values from 0 to (K−1) (K is an integer of 2 or more). The value of ^l does not have to be integrated.

Predistortion signal for generating a predistortion signal y (nm) expressed by equation (7) (transfer function y (nm) / x (nm) expressed by equation (8)) A configuration example of the compensation unit 16 is shown in FIG. In FIG. 10, the sum (A _{l, 0} · | x (n) | ^l + A _{l, 1} ) of l-order terms in the polynomial y (nm) / x (nm) represented by the equation (8). Although a configuration for calculating | x (n−1) | ^l +... Is shown, actually, a power calculation unit 52, delay elements 56-1 to 56-4, and In order that the configuration of the multipliers 58-1 to 58-5 calculates the sum of the terms of each order in the polynomial y (nm) / x (nm) of | x (nk) | A plurality is provided corresponding to each order (excluding the 0th order).

The power calculator 52 calculates and outputs the l-th power (power multiplier) | x (n) | ^l of the absolute value (amplitude level) of the input signal x (n) based on the input signal x (n). The cascaded delay elements 56-1 to 56-4 are connected to a power multiplier | x (n−k) of a power multiplier | x (n) | ^l delayed by k samples with respect to a plurality of one or more k values. ) | Output as ^l . In the configuration example shown in FIG. 10, a signal | x (n-1) | ^{l obtained} by delaying the signal | x (n) | ^l by one sample is output from the delay element 56-1. Outputs a signal | x (n-2) | ^{l obtained} by delaying the signal | x (n) | ^l by two samples, and the delay element 56-3 delays the signal | x (n) | ^l by three samples. Signal | x (n-3) | ^l is output, and the signal | x (n-4) | ^{l obtained} by delaying the signal | x (n) | ^l by four samples is output from the delay element 56-4. Is done. In each of the delay elements 56-1 to 56-4, processing for delaying the scalar quantity (real number) | x (n) | ^l is performed.

Multiplier 58-1～58-5 is distortion compensation coefficient A _{l, k} and exponent | x (n-k) | product A _l and _{^{l, k · | x (n}} -k) | multiple ^l Street K values of 0 or more are calculated and output. In the configuration example shown in FIG. 10, the multiplier 58-1, the distortion compensation coefficients A _{l, 0} and the signal from the power arithmetic unit 52 | x (n) | ^l and the product _{A l, 0 · | x (} n) ^l is calculated, and the multiplier 58-2 calculates the product A _{l, 1} · | x (n) of the distortion compensation coefficient A _{l, 1} and the signal | x (n−1) | ^l from the delay element 56-1. -1) | ^l is calculated, and the multiplier 58-3 calculates the product A _{l, 2} · | of the distortion compensation coefficient A _{l, 2} and the signal | x (n-2) | ^l from the delay element 56-2. x (n-2) | computes ^l, multiplier 58-4, the signal from the delay element 56-3 and the distortion compensation coefficient _{a l, 3 | x (n} -3) | product of ^l a _{l, 3} · | x (n−3) | ^l is calculated, and the multiplier 58-5 calculates the product of the distortion compensation coefficient A _{l, 4} and the signal | x (n−4) | ^l from the delay element 56-4. A _{l, 4} · | x (n−4) | ^l is calculated. In each of the multipliers 58-1 to 58-5, multiplication of the complex number A _{l, k} and the scalar quantity (real number) | x (n−k) | ^l is performed.

The adder 60 integrates the products A _{l, k} · | x (n−k) | ^l (k = 0 to 4 in the configuration example shown in FIG. 10) calculated by the multipliers 58-1 to 58-5. the sum by calculating the distortion compensation coefficient a _{l, k} and exponent by | computes the ^{l | x (n-k)} | convolution sum of _{^{l ΣA l, k · | x}} (n-k) . In the configuration example shown in FIG. 10, the convolution sum ΣA _{l, k} · | x (n−k) | ^l is A _{l, 0} · | x (n) | ^l + A _{l, 1} · | x (n−1) ^l + ... + _{Al, 4} · | x (n−4) | ^l . However, the number of delay elements 56-1 to 56-4 and multipliers 58-1 to 58-5 can be set arbitrarily, and the convolution sum ΣA _{l, k} · | x (n−k) The number of signals A _{l, k} · | x (n−k) | ^l to be integrated when calculating | ^l can be arbitrarily set. In this way, the power calculation unit 52, the delay elements 56-1 to 56-4, the multipliers 58-1 to 58-5, and the adder 60 are included, and the convolution sum ΣA _{l, k} · | x (n−k ) | It is possible to configure a convolution operation unit for calculating ^l .

The convolution operation unit (power operation unit 52, delay elements 56-1 to 56-4, and multipliers 58-1 to 58-5) is a polynomial y (nm) / x (| x (n−k) | n−m) are provided in correspondence with each order (excluding the 0th order), and the product A _{l, k of} the distortion compensation coefficient A _{l, k} and the power multiplier | x (n−k) | ^l _k · | x (n−k) | ^l is calculated for each of a plurality of values of l and k. In the adder 60, the convolution sum ΣA _{l, k} · | x (n−k) | ^l is calculated for each of a plurality of values of l. For example, the power calculation unit 52, the delay elements 56-1 to 56-4, and the multipliers 58-1 to 58-5 are connected to the even order (second order, fourth, fourth) in the polynomial y (nm) / x (nm). Next,... Are provided in correspondence with each other, and a convolution sum ΣA _{l, k} · | x (n−k) | ^l is calculated for each of a plurality of even-numbered l values (0, 2, 4,...). Is also possible. However, the convolution sum ΣA _{l, k} · | x (n−k) | ^l corresponding to l = 0 is simply the distortion compensation coefficient A _{0, k} (A _{0,0 to} A _{0 in the} configuration example shown in FIG. 10). _{, 4} ). Then, the adder 60 adds the convolution sum ΣA _{l, k} · | x (n−k) | ^l calculated for each value of l and calculates the sum, thereby calculating the polynomial in the equation (8). A transfer function y (nm) / x (nm) represented by As described above, in the configuration shown in FIG. 10, the transfer function y (nm) / x (nm) is generated using the convolution operation (FIR filter), and each distortion compensation coefficient A _{l, k} Corresponds to the tap coefficient of the FIR filter. When the transfer function y (n−m) / x (n−m) is calculated, the product A _{l, k} · | x (n−k) | ^l calculated for each value of k is given first. Or the product A _{l, k} · | x (n−k) | ^l calculated for each value of l can be accumulated first.

The delay element 61 outputs the input signal x (n) delayed by m samples as the input signal x (nm). The multiplier 62 calculates the sum of the convolution sums ΣA _{l, k} · | x (n−k) | ^l (transfer function y (n−m) / x (n−n−)) calculated for each l value by the adder 60. m)) and the input signal x (n−m) and the product y (n−m) / x (n−m) · x (n−m) are calculated by the equation (7). A predistortion signal y (nm) is generated and output. FIG. 10 shows an example in which the delay element 61 delays the input signal x (n) by two samples (the value of m is 2), but the delay element 61 delays the input signal x (n). About the time (value of m) to carry out, it can set arbitrarily (for example, in the range of m> = 1). Alternatively, the delay element 61 can be omitted (the value of m is set to 0).

  In contrast to the configuration example shown in FIG. 2 in which the predistortion signal y (n) itself is given a frequency characteristic, the transfer function y (nm) / x (nm) is given a frequency characteristic. In the configuration example shown in FIG. 10, the order of the polynomial by the transfer function y (nm) / x (nm) is lower than the order of the polynomial by the predistortion signal y (n). Therefore, the multiplier 154 having the configuration of FIG. 2 can be omitted. In the configuration example shown in FIG. 2, the multiplications by the multipliers 158-1 to 158-5 are multiplications of complex numbers, whereas in the configuration example shown in FIG. 10, the multipliers 58-1 to 58-58. Multiplication by -5 is multiplication of complex numbers and real numbers. Therefore, the circuit scale of each multiplier 58-1 to 58-5 can be reduced. Further, in the configuration example shown in FIG. 2, the delay processing in each delay element 156-1 to 156-4 is processing for delaying the complex number, whereas in the configuration example shown in FIG. 10, each delay element 56- The delay processing at 1 to 56-4 is processing for delaying the real number. Therefore, the circuit scale of each delay element 56-1 to 56-4 can be reduced. Therefore, according to the configuration example shown in FIG. Note that the memory effect of the amplifier 12 is considered to be one of the causes that the bias line fluctuates due to the influence of signal envelope fluctuation and the signal is modulated. Based on this idea, the configuration of FIG. 10 in which the frequency characteristic is given to the transfer function y (nm) / x (nm) is considered to be extremely appropriate.

  As mentioned above, although the form for implementing this invention was demonstrated, this invention is not limited to such embodiment at all, and it can implement with a various form in the range which does not deviate from the summary of this invention. Of course.

It is a figure which shows the outline of a structure of a predistortion type distortion compensation amplifier provided with the distortion compensation circuit which concerns on embodiment of this invention. It is a figure which shows the outline of a structure of the distortion compensation circuit which concerns on embodiment of this invention. It is a figure which shows the outline of a structure of the distortion compensation circuit which concerns on embodiment of this invention. It is a figure explaining an example of the frequency band in which the coefficient _{Bl, f of a} polynomial is calculated. It is a figure explaining an example of the frequency characteristic of coefficient _{B1 and f} of a polynomial. It is a figure explaining an example of the impulse response of coefficient B _{l, f} of a polynomial. It is a figure which shows the result of having confirmed the distortion compensation effect by the distortion compensation circuit which concerns on embodiment of this invention by measurement. It is a figure which shows the outline of the other structure of the distortion compensation circuit which concerns on embodiment of this invention. It is a figure explaining an example of the characteristic of complex transfer function resp / x with respect to input signal power value | x (n) | ² . It is a figure which shows the outline of the other structure of the distortion compensation circuit which concerns on embodiment of this invention.

Explanation of symbols

  12 amplifier, 14 control unit, 16 predistortion compensation unit, 18 D / A converter, 20, 36 mixer, 22 oscillator, 34 directional coupler, 38 filter, 40 A / D converter, 46 distortion compensation coefficient calculation unit, 52,152 exponentiation operation unit, 56-1 to 56-4, 156-1 to 156-4 delay element, 58-1 to 58-5, 62,154, 158-1 to 158-5 multiplier, 60,160 Adder, 70 Polynomial coefficient calculation unit, 71 Distortion generator, 72, 74 Variable filter, 76 Filter characteristic control unit, 78 Correlation calculation unit, 80 Polynomial coefficient control unit, 81, 84 Frequency conversion unit, 82, 85 Data selection unit 83, 86 Inverse frequency conversion unit, 87 Input / output relation calculation unit, 88 Histogram calculation unit, 90 Polynomial coefficient setting unit.

Claims (5)

A predistortion signal in which predistortion is applied to the input signal to the amplifier is generated so as to reduce distortion components remaining in the output signal from the amplifier having nonlinear characteristics, and the predistortion signal is amplified by the amplifier. A distortion compensation circuit,
If the input signal at the current sample time is x (n), and the input signal at the time before k samples (k is an integer of 0 or more) is x (n−k), the input signal x (n−k) at a plurality of times A predistorter that generates the predistortion signal based on a first polynomial that is a polynomial;
When the nonlinear characteristic of the amplifier is approximated by a second polynomial that is a polynomial of the input signal x (n), each of the second polynomials is determined based on the input signal x (n) and the output signal from the amplifier. A polynomial coefficient calculation unit for calculating coefficients for a plurality of frequency domains,
A distortion compensation coefficient calculation unit that calculates each coefficient of the first polynomial based on each coefficient of the second polynomial calculated for each frequency region by the polynomial coefficient calculation unit;
A distortion compensation circuit comprising: The distortion compensation circuit according to claim 1,
The distortion compensation coefficient calculator is
Based on each coefficient of the second polynomial calculated for each frequency region in the polynomial coefficient calculation unit, the frequency characteristic of each coefficient of the second polynomial is calculated,
Convert the frequency characteristics of each coefficient of the second polynomial into an impulse response,
A distortion compensation circuit that calculates each coefficient of the first polynomial based on an impulse response of each coefficient of the second polynomial. The distortion compensation circuit according to claim 1 or 2,
The polynomial coefficient calculator is
A distortion generator that distorts and outputs the input signal x (n) based on the second polynomial;
A first filter for extracting a specific frequency component in an output signal from the amplifier, the first filter capable of changing the extracted frequency component;
A second filter for extracting a specific frequency component in the output signal from the distortion generator, the second filter capable of changing the extracted frequency component;
A filter control unit that changes a plurality of frequency components extracted by the first and second filters;
When the filter controller changes the frequency components extracted by the first and second filters in a plurality of ways, the output signal from the amplifier that has passed through the first filter and the distortion generator that has passed through the second filter A correlation calculation unit for calculating a correlation value with the output signal for each frequency component;
A polynomial coefficient control unit that controls each coefficient of the second polynomial for each frequency region based on the correlation value calculated for each frequency component by the correlation calculation unit;
A distortion compensation circuit. The distortion compensation circuit according to claim 1 or 2,
The polynomial coefficient calculator is
An input / output relationship calculation unit for calculating the input / output relationship of the amplifier with respect to a plurality of frequency domains based on the input signal x (n) and the output signal from the amplifier;
Based on the input signal x (n) and the input / output relationship calculated for each frequency region by the input / output relationship calculation unit, the characteristics of the input / output relationship with respect to the power value of the input signal x (n) are A non-linear characteristic calculation unit for calculating
A polynomial coefficient setting unit that sets each coefficient of the second polynomial for each frequency domain based on the characteristics calculated for each frequency domain by the nonlinear characteristic calculation unit;
A distortion compensation circuit. An amplifier having nonlinear characteristics, and a distortion compensation circuit that generates a predistortion signal in which predistortion is applied to an input signal to the amplifier so that a distortion component remaining in the output signal from the amplifier is reduced, A predistortion type distortion compensation amplifier that amplifies a predistortion signal generated by a compensation circuit with an amplifier,
A predistortion type distortion compensation amplifier, wherein the distortion compensation circuit is the distortion compensation circuit according to claim 1.

Images