Method of influencing an input signal and predistorter

Abstract

The present invention relates to a method of influencing an input signal (x), in particular an input signal of a power amplifier of a wireless communication system, wherein said input signal (x) has a range of input amplitudes (r0, . . . , rM−1), wherein a correction signal (Γ(x)) depending on said input signal (x) is generated, and wherein said input signal (x) is influenced depending on said correction signal (Γ(x)).

Description

TECHNICAL FIELD

[0001] The present invention relates to a method of influencing an input signal and a predistorter.

[0002] The invention is based on a priority application, EP 03291281.8, which is hereby incorporated by reference.

BACKGROUND OF THE INVENTION

[0003] Such methods and corresponding predistorters are per se known from prior art and they are used to shape an input signal before said input signal is fed into a power amplifier. The power amplifier itself usually has a nonlinear characteristic which is undesirable since it causes the input signal to be distorted, i.e. a gain imparted on said input signal by said power amplifier is not constant but depends on an amplitude of said input signal.

[0004] To avoid this unwanted distortion, the input signal is predistorted by a predistorter prior to feeding it to said power amplifier. The predistorter distorts or shapes, respectively, the characteristic of the input signal in a manner opposite to the distortion caused by said power amplifier, and in a well-tuned system an overall distortion of said input signal may thus be avoided. Said predistortion is usually achieved by multiplying said input signal by a correction signal or by adding such a correction signal which depends on said input signal.

[0005] Contemporary methods use so-called polynomial predistorters wherein the correction signal is based on a polynomial. Since each function value of a polynomial depends on all coefficients of said polynomial, it is not possible to tune the correction signal or certain intervals of it, respectively, via changing one or few coefficients without changing the complete correction signal. As a further disadvantage, said polynomial predistorter only works well for a limited range of input amplitudes and is thus not flexible with respect to a varying mean power of said input signal. Furthermore, the polynomial predistorter is not flexible enough with respect to special required shapes of a characteristic of the non-linear power amplifier.

[0006] Another solution known from prior art employs a lookup table to determine function values of said correction signal. Consequently, it is possible to configure a predistorter wherein each function value depends on one parameter that can be read from said lookup table. This solution also lacks flexibility, because especially in adaptive systems, where a dynamic modification of said parameters is required, the number of parameters that must be updated is too high.

[0007] The present invention relates to a method of influencing an input signal, in particular an input signal of a power amplifier of a wireless communication system, wherein said input signal has a range of input amplitudes, wherein a correction signal depending on said input signal is generated, and wherein said input signal is influenced depending on said correction signal.

[0008] The present invention also relates to a predistorter for influencing an input signal, in particular an input signal of a power amplifier of a wireless communication system, wherein said input signal has a range of input amplitudes, wherein a correction signal depending on said input signal is generated, and wherein said input signal is influenced depending on said correction signal.

[0009] Therefore, it is an object of the present invention to provide an improved method of influencing an input signal and a corresponding predistorter which provide increased flexibility.

SUMMARY OF THE INVENTION

[0010] This object is achieved by a method of influencing an input signal, in particular an input signal of a power amplifier of a wireless communication system, wherein said input signal has a range of input amplitudes, wherein a correction signal depending on said input signal is generated, and wherein said input signal is influenced depending on said correction signal, wherein a generation of said correction signal is based on a plurality of base functions, and in that said range of input amplitudes is divided into intervals, wherein each of said base functions contributes to said correction signal in a limited number of intervals.

[0011] The object of the invention is further achieved by a predistorter for incluencing an input signal, in particular an input signal of a power amplifier of a wireless communication system, wherein said input signal has a range of input amplitudes, wherein a correction signal depending on said input signal is generated, and wherein said input signal is influenced depending on said correction signal, wherein a generation of said correction signal is based on a plurality of base functions, and wherein said range of input amplitudes is divided into intervals, wherein each of said base functions contributes to said correction signal in a limited number of intervals.

[0012] According to the present invention, the generation of said correction signal is based on a plurality of base functions, and said range of input amplitudes is divided into intervals, wherein each of said base functions contributes to said correction signal in a limited number of intervals.

[0013] Hence it is possible to shape a desired characteristic of said correction signal more accurately as compared to the approach using one polynomial for the definition of the whole correction signal, because changing one base function or its coefficients or the like does not affect the whole correction signal but only some portions of the correction signal which correspond to the intervals in which said changed base function contributes to the correction signal.

[0014] The other portions of said correction signal remain unchanged, because of which the mechanism may also be said to have a “local effect” on said correction signal, and which enables an elaborated adaption of said correction signal.

[0015] To control the local effect more precisely, a variant of the inventive method is characterized in that a set of base functions contributes to said correction signal only in a corresponding interval and/or in a limited number of neighbour intervals. I.e. changing such a base function will affect the corresponding interval and a limited number of neighbour intervals thus affecting a portion of said correction signal that is larger than one single interval but yet significantly smaller than the whole correction signal. In this case, base functions of neighbour intervals may overlap each other.

[0016] Contributing to said correction signal in this sense of the present invention means to significantly change the correction signal, regardless of the mode by which said change is performed. I.e. if various base functions are added together, a contribution in the sense of the present invention means that said base functions have non-zero function values large enough to change the resulting sum in a noticeable way.

[0017] Consequently, a base function may also have non-zero function values and may not contribute to said correction signal in the above sense, as long as said non-zero function values are very small compared to any other function value of the above mentioned sum.

[0018] According to a further advantageous embodiment of the present invention, said base functions are chosen such that said correction signal and a first and/or higher derivatives of said correction signal are continuous. This enables the correction signal to be modified by changing said base functions or coefficients of said base functions or the like while maintaining a smooth correction signal. Algorithms for changing said base functions must hence not obey any restriction to changing said base functions in order to maintain a smooth correction signal.

[0019] A further variant of the invention is characterized in that said base functions have non-zero function values within a number of neighbour intervals and function values and/or derivatives of first or higher order of zero at all but one interval bounds of said neighbour intervals, in that either a function value or a derivative of said base functions is different from zero at one interval bound of said neighbour intervals, and in that said base functions are smooth. This guarantees the above explained local effect while at the same time ensuring a smooth correction signal.

[0020] According to a very advantageous variant of the present invention, a maximum of four different base functions has non-zero function values at a distinct input amplitude, i.e. at a distinct interval of said range of input amplitudes. Hence only said four base functions and their parameters contribute to a function value of the correction signal used for predistorting the input signal thus again achieving said “local effect”.

[0021] Instead of having non-zero function values, said four different base functions may have rather small function values, either, that do not contribute to said correction signal in said interval significantly, or in case said base functions are multiplied with each other, they may as well have function values of about one in order not to contribute to said correction signal significantly.

[0022] It is not necessary to restrict the number of base functions to four, every other number of base functions may also be used, as long as the overall number of base functions is e.g. smaller than the number of parameters to be changed with a prior art lookup table.

[0023] Another variant of the method according to the present invention is characterized in that said correction signal is a weighted sum of said base functions according to 1$\Gamma \left(x\right)=\sum _{i=0}^{N-1}{a}_i{\phi }_i\left(r\left(x\right)\right),$

[0024] wherein complex coefficients

αiεC,i=0, . . . ,N−1

[0025] are provided for each base function

φi,i=0, . . . ,N−1,

[0026] and wherein

r=r (x),rεR

[0027] is a function of said input signal (x).

[0028] Since in most cases, especially with digital baseband predistortion for power amplifiers of wireless communication systems, said input signal x is a complex variable, r(x) may e.g. be chosen as an absolute value of said input signal, if the correction signal is desired to only depend on the input signal's amplitude. The base functions φi,i=0, . . . ,N−1, in turn, depend on r(x).

[0029] The coefficients αiεC,i=0, . . . ,N−1 are complex numbers and are used as weights which determine to what extent each of the base functions contributes to said correction signal. The coefficients may also attain a value of zero, if a reduced number of base functions is sufficient for shaping the desired correction signal.

[0030] According to a further variant of the present invention, at least one of said base functions is a polynomial, in particular a polynomial of third order, which allows an easy evaluation of function values and a simple base function definition.

[0031] According to a very advantageous variant of the present invention, the definition of said N base functions is: 2$\begin{array}{c}{\phi }_0\left(r\right)=\{\begin{array}{cc}1-s^2\left(3-2s\right),& r_0\le r<r_1\\ 0& \mathrm{else}\end{array},\\ {\phi }_1\left(r\right)=\{\begin{array}{cc}\left(r_1-r_0\right){s\left(1-s\right)}^2,& r_0\le r<r_1\\ 0& \mathrm{else}\end{array},\\ {\phi }_{2i}\left(r\right)=\left\{\begin{array}{cc}{s}^2\left(3-2s\right),& r_{i-1}\le r<r_i\\ 1-s^2\left(3-2s\right),& r_i\le r<r_{i+1}\\ 0& \mathrm{else}\end{array}\right\}i=1,\dots ,M-2,\\ {\phi }_{2i+1}\left(r\right)=\left\{\begin{array}{cc}\left(r_i-r_{i-1}\right)s^2\left(s-1\right),& r_{i-1}\le r<r_i\\ \left(r_{i+1}-r_i\right){s\left(1-s\right)}^2,& r_i\le r<r_{i+1}\\ 0& \mathrm{else}\end{array}\right\}i=1,\dots ,M-2,\\ {\phi }_{2M-2}\left(r\right)=\{\begin{array}{cc}{s}^2\left(3-2s\right),& r_{M-2}\le r<r_{M-1}\\ 0& \mathrm{else}\end{array},\\ {\phi }_{2M-1}\left(r\right)=\{\begin{array}{cc}\left(r_{M-1}-r_{M-2}\right)s^2\left(s-1\right),& r_{M-2}\le r<r_{M-1}\\ 0& \mathrm{else}\end{array},\mathrm{wherein}\end{array}$$S=\{\begin{array}{c}\frac{r-r_{i-1}}{r_i-r_{i-1}},r_{i-1}\le r<r_i\\ \frac{r-r_i}{r_{i+1}-r_i},r_i\le r<r_{i+1}\end{array}$

[0032] defines an offset from a lower limit of an interval ranging from a first input signal amplitude ri to a second input signal amplitude ri+1, and wherein 3$M=\frac{N}{2}.$

[0033] As can be seen from the above definition, the range (r0, . . . ,rM−1) of input amplitudes is subdivided into M−1 intervals (r0,r1),(r1,r2), . . . ,(rM−2,rM−1), wherein for each interval corresponding base functions are defined. Each of said base functions has only a local effect on said correction signal in that for each interval (ri,ri+1) not more than four base functions have non-zero function values.

[0034] The weighted sum representing the correction signal can hence be simplified to

Γ(x)=α2iφ2i(r(x))+α2i+1φ2i+1(r(x))+α2i+2φ2i+2(r(x))+α2i+3φ2i+3(r(x))

[0035] for ri≦r≦ri+1

[0036] A further embodiment of the inventive method proposes to use the auxiliary functions

Ψ0(s)=1−s2(3−2s)

Ψ1(s)=s(1−s)2

Ψ2 (s)=s2(3−2s)

Ψ3(s)=s2(s−1)

[0037] to calculate said base functions.

[0038] At least one of said auxiliary functions Ψ0(s),Ψ1(s),Ψ2(s),Ψ3(s) is stored to/read from a lookup table in a further advantageous embodiment of the present invention.

[0039] According to yet a further variant of the invention, an interval width ri+1−ri is independent of i which leads to a constant interval width ri+1−ri for i=0, . . . ,M−1, which enables a simplified evaluation of said base functions and said correction signal.

[0040] A reciprocal value 4$q_i=\frac{1}{r_{i+1}-r_i}$

[0041] of an interval width ri+1−ri is stored to/read from a lookup table in a further variant of the invention which eliminates the need of repeatedly performing the corresponding division operations.

[0042] A further advantageous embodiment of the method according to the present invention is characterized in that said input signal is delayed before being influenced by said correction signal, which may be necessary in most cases and depends on the computation time required for obtaining said correction signal from said input signal.

[0043] A further solution to the object of the present invention is given by a predistorter according to the invention.

BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Further advantages and details are presented in the following detailed description with reference to the drawings.

[0045] FIG. 1 shows a block diagram of an inventive predistorter,

[0046] FIG. 2 shows base functions according to the present invention, and

[0047] FIG. 3 shows auxiliary functions used to obtain said base functions of FIG. 2.

DETAILED DESCRIPTION OF THE DRAWINGS

[0048] As can be seen from the block diagram of FIG. 1, a predistorter 100 receives a complex input signal x which, after a certain delay realised by delay means 10, is multiplied by a complex correction signal Γ(x) that depends on said input signal x. Said delay caused by the delay means 10 accounts for a computation time required by the predistorter 100 to calculate the correction signal Γ(x) out of the input signal x.

[0049] The calculation of the correction signal Γ(x) is described in the following passages.

[0050] At first, an absolute value r of said input signal x is obtained by successively calculating the squared absolute value within calculation means 11 and applying a square root function to said squared absolute value within calculation means 12:

r (x)={square root}{square root over (xI2+xQ2)},

[0051] wherein

[0052] X=xI+jXQ and xI,xQεR.

[0053] In a digital baseband system of a wireless communications system, r(x) corresponds to an amplitude of said input signal x.

[0054] The square root calculation operation performed within calculation means 12 may be approximated by a polynomial. This is particularly useful if the predistorter 100 is realized on e.g. a digital signal processor (DSP).

[0055] After obtaining said absolute value r(x), which in the following explanations will be referred to as amplitude r of said input signal, a corresponding interval number is evaluated by interval number evaluation means 14.

[0056] This is necessary to determine the correct set of base functions to be used to calculate the portion of the correction signal Γ(x) which is associated to the respective amplitude r.

[0057] As can be seen from FIG. 2, the range (r0, . . . ,rM−1) of input amplitudes is subdivided into intervals (r0,r1),(r1,r2), . . . ,(rM−2,rM−1), wherein for each interval corresponding base functions 5$\begin{array}{c}{\phi }_0\left(r\right)=\{\begin{array}{cc}1-s^2\left(3-2s\right),& r_0\le r<r_1\\ 0& \mathrm{else}\end{array},\\ {\phi }_1\left(r\right)=\{\begin{array}{cc}\left(r_1-r_0\right){s\left(1-s\right)}^2,& r_0\le r<r_1\\ 0& \mathrm{else}\end{array},\\ {\phi }_{2i}\left(r\right)=\left\{\begin{array}{cc}{s}^2\left(3-2s\right),& r_{i-1}\le r<r_i\\ 1-s^2\left(3-2s\right),& r_i\le r<r_{i+1}\\ 0& \mathrm{else}\end{array}\right\}i=1,\dots ,M-2,\\ {\phi }_{2i+1}\left(r\right)=\left\{\begin{array}{cc}\left(r_i-r_{i-1}\right)s^2\left(s-1\right),& r_{i-1}\le r<r_i\\ \left(r_{i+1}-r_i\right){s\left(1-s\right)}^2,& r_i\le r<r_{i+1}\\ 0& \mathrm{else}\end{array}\right\}i=1,\dots ,M-2,\\ {\phi }_{2M-2}\left(r\right)=\{\begin{array}{cc}{s}^2\left(3-2s\right),& r_{M-2}\le r<r_{M-1}\\ 0& \mathrm{else}\end{array},\\ {\phi }_{2M-1}\left(r\right)=\{\begin{array}{cc}\left(r_{M-1}-r_{M-2}\right)s^2\left(s-1\right),& r_{M-2}\le r<r_{M-1}\\ 0& \mathrm{else}\end{array},\end{array}$

[0058] are provided, and wherein 6$s=\{\begin{array}{c}\frac{r-r_{i-1}}{r_i-r_{i-1}},r_{i-1}\le r<r_i\\ \frac{r-r_i}{r_{i+1}-r_i},r_i\le r<r_{i+1}\end{array}$

[0059] defines an offset from a lower limit of an interval ranging from a first input signal amplitude ri to a second input signal amplitude ri+1, and wherein 7$M=\frac{N}{2},$

[0060] i.e. the overall number N of base functions is 2M.

[0061] The base functions of a distinct interval (ri,ri+1) have non-zero function values only within two neighbour intervals, which results in a simplified expression for the respective portion of the correction signal

Γ(x)=α2iφ2i(r(x))+α2i+1φ2i+1(r(x))+α2i+2φ2i+2(r(x))+α2i+3φ2i+3(r(x))

[0062] at that interval (ri,ri+1).

[0063] According to the above simplified expression for the correction signal, the correction signal for the interval (ri,ri+1) only depends on four of said N base functions:

φ2i(r(x)),φ2i+1(r(x)),φ2i+2(r(x)),φ2i+3(r(x)),

[0064] and on the respective coefficients.

[0065] Consequently, said portion of the correction signal may individually be shaped according to system requirements concerning an optimal predistortion of said input signal x without influencing other portions of said correction signal. This “local effect” is achieved by limiting the number of base functions that contribute to said correction signal Γ(x) and by providing extra sets of base functions φ2i(r(x)),φ2i+1(r(x)),φ2i+2(r(x)),φ2i+3(r(x)) for individual intervals.

[0066] The base functions φi(r) and their first derivatives are continuous, and the coefficients of the polynomials the base functions φi(r) are composed of are defined by function values of and derivatives of the respective base function at three subsequent input signal amplitudes. The base functions can be divided into two different classes, the even-numbered base functions φ2i(r) and the odd-numbered base functions φ2i+1(r).

[0067] The series of weighted base functions yielding said correction signal

Γ(x)=α2iφ2i(r(x))+α2i+1φ2i+1(r(x))+α2i+2φ2i+2(r(x))+α2i+3φ2i+3(r(x))

[0068] itself is continuous, as well as its first derivative with respect to r. Furthermore, the series is a linear function of the coefficients αi.

[0069] The values of even-numbered coefficients α2i,α2i+2, . . . define function values of said correction signal Γ(x) at said input amplitudes ri, and the values of odd-numbered coefficients α2i+1,α2i+3, . . . define derivatives of said correction signal Γ(x) at said input amplitudes ri.

[0070] Due to the definition of the correction signal Γ(x), its smoothness is guaranteed for arbitrary coefficient values, and the correction signal Γ(x) is suitable for approximating a required shape/characteristic in the complete range of input amplitudes ri.

[0071] After determining the correct set of base functions to be used to calculate the portion of the correction signal Γ(x) which is associated to the respective amplitude r, i.e. after finding the index variable i which relates a set of base functions and their coefficients to said input amplitude, said index variable i is fed to an address calculation system 16 which calculates memory addresses of the required complex coefficients of said base functions and outputs said memory addresses to a coefficient memory 18 which provides the selected coefficients to complex multipliers, cf. FIG. 2.

[0072] At the same time, said offset 8$s=\{\begin{array}{c}\frac{r-r_{i-1}}{r_i-r_{i-1}},r_{i-1}\le r<r_i\\ \frac{r-r_i}{r_{i+1}-r_i},r_i\le r<r_{i+1}\end{array}$

[0073] is evaluated by offset calculation means 13, which enables to calculate the function values of the base functions by means of auxiliary functions

Ψ0(s)=1−s2(3−2s)

Ψ1(s)=s(1−s)2

Ψ2(s)=s2(3−2s)

Ψ3(s)=s2(s−1)

[0074] which require said offset s as an input variable.

[0075] Said auxiliary functions can be pre-calculated and the resulting functions values can be stored to/read from a lookup table.

[0076] Due to the following symmetry

Ψ2(s)=Ψ0(1−s)

Ψ3(s)=−Ψ1(1−s)

[0077] in a further embodiment of the present invention it is sufficient to calculate and/or store the function values of said auxiliary functions only for a first, lower half 9$\left(r_i,\frac{r_i+r_{i+1}}{2}\right)$

[0078] of an interval (ri,ri+1) and to use a modified offset 10$s^{\prime }=1-s=\frac{r_{i+1}-r}{r_{i+1}-r_i}$

[0079] for attaining auxiliary function values of the second, upper half 11$\left(\frac{r_i+r_{i+1}}{2},r_{i+1}\right)$

[0080] of the interval (ri,ri+1).

[0081] To make use of said symmetry, the sign of the function values of said auxiliary functions as well as the calculation of the coefficient addresses within said calculation system 16 must be change appropriately. By using said symmetry, memory requirements can be decreased by a factor of two as far as a storage of said function values is concerned.

[0082] Based on said index variable i, block 15 calculates an interval width ri+1−ri which is required for obtaining said base functions 12$\begin{array}{c}{\phi }_{2i}\left(r\right)={\Psi }_0\left(\frac{r-r_i}{r_{i+1}-r_i}\right)\\ {\phi }_{2i+1}\left(r\right)=\left(r_{i+1}-r_i\right){\Psi }_1\left(\frac{r-r_i}{r_{i+1}-r_i}\right)\\ {\phi }_{2i+2}\left(r\right)={\Psi }_2\left(\frac{r-r_i}{r_{i+1}-r_i}\right)\\ {\phi }_{2i+3}\left(r\right)=\left(r_{i+1}-r_i\right){\Psi }_3\left(\frac{r-r_i}{r_{i+1}-r_i}\right)\end{array}$

[0083] out of said auxiliary functions Ψ0,Ψ1,Ψ2,Ψ3.

[0084] The interval width ri+1−ri, or its reciprocal value, can also be stored to a lookup table.

[0085] The auxiliary functions Ψ1,Ψ3 need not be multiplied with said interval width ri+1−ri, if said interval width ri+1−ri is constant for each value of said index variable i. The corresponding multipliers of the predistorter 100 of FIG. 1 may be omitted in this case, as well as block 15.

[0086] Finally, the evaluated auxiliary function values are added in an adder 17, which yields said correction signal Γ(x) that is multiplied with the delayed input signal x.

[0087] The presented method and predistorter advantageously provide a correction signal Γ(x) and an output signal y which is a linear function of a real part of the coefficients αi, of an imaginary part of the coefficients αi and of the coefficients αi themselves, too, which is very useful for some adaptation algorithms.

[0088] Furthermore, in comparison to a lookup-table approach of prior art, the number of coefficients for the base functions is very small which enables the method to be used in systems with limited feedback information, where only a small number of coefficients can be adapted to optimally shape the characteristic of the correction signal.

[0089] Since smoothness of the correction signal is guaranteed by the presented solution, adaptive algorithms can alter coefficients without obeying restrictions for assuring a continuous correction signal. The predistorter can also be easily implemented.

[0090] If at a certain point only feedback information for adapting coefficients corresponding to small signal amplitudes are available, an existing approximation of the correction signal via the base functions for large signal amplitudes is not influenced and remains constant until feedback information for improving coefficients corresponding to large signal amplitudes will be available again. The predistorter according to the present invention is therefore not as sensitive to power variations of the input signal as prior art polynomial predistorters.

[0091] Furthermore, with the various sets of base functions, any desired characteristic of the correction signal can be implemented more easily as compared to a prior art lookup table approach. The proposed method does also enable a faster adaptation as is given with contemporary lookup table solutions.

[0092] The method according to the present invention and/or the predistorter are not limited to be used with a power amplifier, they may be used with any non-linear device or channel which requires a process of predistortion.

Claims

1. Method of influencing an input signal, in particular an input signal of a power amplifier of a wireless communication system, wherein said input signal has a range of input amplitudes, wherein a correction signal depending on said input signal is generated, and wherein said input signal is influenced depending on said correction signal, wherein a generation of said correction signal is based on a plurality of base functions, and in that said range of input amplitudes is divided into intervals, wherein each of said base functions contributes to said correction signal in a limited number of intervals.

2. Method according to claim 1, wherein a set of base functions contributes to said correction signal only in a corresponding interval and/or in a limited number of neighbour intervals.

3. Method according to claim 1, wherein said base functions are chosen such that said correction signal and a first and/or higher derivatives of said correction signal are continuous.

4. Method according to claim 1, wherein said base functions have non-zero function values within a number of neighbour intervals and function values and/or derivatives of first or higher order of zero at all but one interval bounds of said neighbour intervals, wherein either a function value or a derivative of said base functions is different from zero at one interval bound of said neighbour intervals, and wherein said base functions are smooth.

5. Method according to claim 1, wherein said correction signal is a weighted sum of said base functions according to

6. Method according to claim 1, wherein at least one of said base functions is a polynomial, in particular a polynomial of third order.

7. Method according to claim 1, wherein a maximum of four different base functions has non-zero function values at a distinct input amplitude.

8. Method according to claim 1, wherein the definition of said N base functions is:

9. Method according to claim 8, wherein an interval width ri+1−ri is independent of i.

10. Method according to claim 7, wherein a reciprocal value

11. Method according to claim 8, wherein auxiliary functions

12. Method according to claim 8, wherein at least one of said auxiliary functions is stored to/read from a lookup table.

13. Method according to claim 1, wherein said input signal is delayed before being influenced by said correction signal.

14. Predistorter for influencing an input signal, in particular an input signal of a power amplifier of a wireless communication system, wherein said input signal has a range of input amplitudes, wherein a correction signal depending on said input signal is generated, and wherein said input signal is influenced depending on said correction signal, wherein a generation of said correction signal is based on a plurality of base functions, and wherein said range of input amplitudes is divided into intervals, wherein each of said base functions contributes to said correction signal in a limited number of intervals.

15. Predistorter for influencing an input signal, in particular an input signal of a power amplifier of a wireless communication system, wherein said input signal has a range of input amplitudes, wherein a correction signal depending on said input signal is generated, and wherein said input signal is influenced depending on said correction signal, wherein a generation of said correction signal is based on a plurality of base functions, and wherein said range of input amplitudes is divided into intervals, wherein each of said base functions contributes to said correction signal in a limited number of intervals, said predistorter being capable of performing the method according to claim 2.