CN102142814A - Power amplifier related device, power amplifier predistortion system and modeling method - Google Patents

Abstract

The invention relates to a power amplifier related device, a power amplifier predistortion system and a modeling method. The power amplifier related device comprises a modulus calculation unit, a selection unit and a plurality of output calculation units, wherein the modulus calculation unit is used for calculating the modules of an input signal; the selection unit is used for determining one of the plurality of the output calculation units to serve as the output calculation unit into which an input signal needs to be input according to the modulus calculated by the modulus calculation unit; and each output calculation unit calculates a signal to be output according to the input signal input into the output calculation unit.

Description

Power amplifier relevant apparatus, power amplifier pre-distortion system and modeling method

Technical field

Embodiments of the present invention are relevant with power amplifier, and embodiments of the present invention are with that the memory effect nonlinear power amplifier is arranged is relevant.

Background technology

Power amplifier (abbreviation power amplifier) is a necessary part in the transmitting set.Power amplifier is carried out modeling, can be used for the performance simulation of communication system on the one hand, be convenient to the design of predistorter on the other hand.The behavior model of power amplifier is to utilize certain mathematic(al) representation to describe the input/output relation of power amplifier, efficiently becomes a kind of important power amplifier modeling method owing to it is simple.Fig. 1 shows the theory diagram of power amplifier behavior modeling method.As shown in Figure 1, the part of the radio-frequency input signals of power amplifier 102 obtains the digital equivalent baseband signal of radio-frequency input signals through behind coupler 101, quadrature down converter 104 and the A/D converter 107.Simultaneously, the part of the radio frequency output signal of power amplifier 102 is through obtaining the digital equivalent baseband signal of radio frequency output signal behind coupler 103, quadrature down converter 106 and the A/D converter 110.Local oscillator 105 is used to provide local oscillating frequency.Power amplifier Model Distinguish module 108 is with the input as the power amplifier model of the digital equivalent baseband signal of A/D converter 107 output, with the digital equivalent baseband signal of A/D converter 110 output desired output, utilize adaptive algorithm to carry out the identification of power amplifier model parameter as the power amplifier model.Adder 109 is used for calculating poor between the output of the actual output of power amplifier Model Distinguish module 108 and the power amplifier 102 that A/D converter 110 obtains, and this difference is used for the parameter of Adaptive Identification power amplifier Model Distinguish module 108.

In this type of modeling method, key issue is how to design the power amplifier structure of models, promptly adopts what kind of mathematic(al) representation.Non-linear is the inherent characteristic of power amplifier, and along with the continuous increase of communication system bandwidth, the power amplifier in the transmitter also shows tangible memory effect sometimes in addition.Therefore need to adopt the nonlinear power amplifier model structure of memory is arranged.

Volterra progression is to describe the universal method that the memory non linear system is arranged.But the shortcoming of traditional Volterra progression is that the numerous computation complexities of its parameter are very high, and the convergence rate when using identification algorithm simultaneously is slow.In the prior art, adopt the various simplified structure modelings of Volterra progression more.As document [1] (Clark C J, Chrisikos G, Muha M S, et al.Time-domain envelopemeasurement technique with application to wide-band power amplifiermodeling.IEEE Trans.Microwave Theory Tech., 1998,46 (12): use the Wiener model description that the memory power amplifier is arranged 2531-2540.), document [2] (Kim J, Konstantinou K.Digital predistortion of wideband signals based on power amplifier modelwith memory.Electron.Lett., 2001, the memory multinomial model that proposes in 37:1417-1418.) etc.The parameter of these models is less, simple in structurely is easy to identification.But when power amplifier shows different nonlinear characteristics at different input power places, this single Volterra progression or its simplified model will be difficult to describe the interior power amplifier characteristic of whole input signal dynamic range.On the other hand, when using the high-order nonlinear model, in the identification of Model Parameters process, may produce numerical value instability problem.

Adopt the method for segmentation modeling can effectively solve above problem.But the input sample of power amplifier model needs sequential processes in time, and the front and back sample value may fall into different piecewise intervals like this, is handled by different nonlinear functions.Therefore, be difficult to memory effect is introduced segmented model.

Summary of the invention

These problems that the present invention is directed to prior art are made, and in order to solve the shortcoming that exists in the prior art, provide a kind of useful selection at least.

To achieve these goals, the application provides following aspect.

Aspect 1, a kind of power amplifier relevant apparatus is characterized in that, this power amplifier relevant apparatus comprises mould value computing unit, selected cell and a plurality of output computing unit: wherein

Described mould value computing unit is used to calculate the mould value of input signal;

Described selected cell is used for the mould value that calculates according to described mould value computing unit, determines an output computing unit that should be input to as described input signal in described a plurality of output computing unit; And

Each described output computing unit calculates the signal that should export according to the described input signal of this output computing unit of input.

When each described output computing unit all adopts the power amplifier predistortion pattern function, described power amplifier relevant apparatus is as the power amplifier predistortion device, when each described output computing unit all adopted the power amplifier pattern function, described power amplifier relevant apparatus was as the power amplifier model equipment.

Aspect 2, according to aspect 1 described power amplifier relevant apparatus, it is characterized in that each described output computing unit calculates the described signal that should export according to the different same functions of parameter.

Aspect 3, according to aspect 2 described power amplifier relevant apparatus, it is characterized in that described function parameters is separate between each described output computing unit.

Aspect 4, according to aspect 1 described power amplifier relevant apparatus, it is characterized in that the employed function of each described output computing unit is separate between described a plurality of described output computing units.

Aspect 5, according to aspect 1 described power amplifier relevant apparatus, it is characterized in that each described output computing unit is all according to the non-linear power amplifier pattern function of memory being arranged or having the non-linear power amplifier predistortion pattern function of memory to calculate the signal that should export.

Aspect 6, according to aspect 5 described power amplifier relevant apparatus, it is characterized in that described have the non-linear power amplifier pattern function of memory or have the non-linear power amplifier predistortion pattern function of memory to comprise wiener power amplifier pattern function, memory multinomial power amplifier pattern function, Hammerstein pattern function or Hammerstein-Wiener power amplifier pattern function etc.

Aspect 7, according to aspect 1 described power amplifier relevant apparatus, it is characterized in that each described output computing unit all calculates the signal that should export according to memoryless power amplifier pattern function or memoryless power amplifier predistortion pattern function.

Aspect 8, according to aspect 7 described power amplifier relevant apparatus, it is characterized in that described memoryless power amplifier pattern function or memoryless power amplifier predistortion pattern function comprise memoryless multinomial power amplifier pattern function etc.

Aspect 9, a kind of power amplifier pre-distortion system, it is characterized in that, this power amplifier pre-distortion system comprises each described power amplifier relevant apparatus of aspect 1-8 as the power pre-distortion device, and the described output computing unit of each of wherein said power amplifier relevant apparatus calculates the signal that should export according to the power amplifier predistortion pattern function.

Aspect 10, a kind of power amplifier model or power amplifier pre-distortion method for establishing model, described method comprises:

The input and output obtaining step obtains the input of power amplifier model or power amplifier pre-distortion model and distinguishes corresponding output with described input;

Mould value calculation procedure is calculated each mould value of importing constantly in the described input;

The grouping step, according to each the residing interval of mould value imported constantly in the described input, be divided into a plurality of groups with described input with the corresponding respectively output of described input, and structure and described respectively import corresponding input vector and with the corresponding output of described input vector, thereby obtain a plurality of by described input vector and the input and output Vector Groups formed with the corresponding output of described input vector; And

The calculation of parameter step according to the input and output Vector Groups that comprises in each group in described a plurality of groups, at each group in described a plurality of groups, is calculated respectively each parameter of should group corresponding power amplifier pattern function, thereby is obtained a plurality of amplifier submodel functions.

Aspect 11, according to aspect 10 described methods, it is characterized in that described calculation of parameter step adopts least square method to come at each group in described a plurality of groups, calculate respectively each parameter of should group corresponding power amplifier pattern function.

Aspect 12, according to aspect 10 described methods, it is characterized in that, described calculation of parameter step adopts weighted least-squares method to come at each group in described a plurality of groups, calculate respectively each parameter of should group corresponding power amplifier pattern function, in each described group, with described input in followingly import corresponding input vector and give bigger weight:

The summit in the mould value of this input and the residing interval of described mould value differs less than predetermined value.

Aspect 13, according to aspect 10 described methods, it is characterized in that, corresponding power amplifier pattern function or the power amplifier pre-distortion model of each group in described a plurality of groups comprises amplitude model function and phase model function, described calculation of parameter step is at each group in described a plurality of groups, calculate each parameter of amplitude model function and phase model function respectively, and adopt the constraint least-squares algorithm that the average of the signal in the preset range of both sides, summit in described interval is equated.

Aspect 14, according to aspect 10 described power amplifier method for establishing model, it is characterized in that described input and output obtaining step comprises:

The first quadrature frequency conversion step is used for the part of the signal that is input to power amplifier is carried out down-conversion;

The first analog-to-digital conversion step is used for the signal that the described first quadrature frequency conversion step is obtained is carried out analog-to-digital conversion, thereby obtains described input;

The second quadrature frequency conversion step is used for the part of the signal of the output of power amplifier is carried out down-conversion;

The second analog-to-digital conversion step is used for the signal that the described second quadrature frequency conversion step is obtained is carried out analog-to-digital conversion, thereby obtains described output.

Aspect 15, a kind of power amplifier model or power amplifier pre-distortion modelling device, described amplifier model apparatus for establishing comprises:

The input and output acquiring unit obtains the input of amplifier model or power amplifier pre-distortion model and distinguishes corresponding output with described input;

Mould value computing unit calculates each mould value of importing constantly in the described input;

Grouped element, according to each the residing interval of mould value imported constantly in the described input, be divided into a plurality of groups with described input with the corresponding respectively output of described input, and structure and described respectively import corresponding input vector and with the corresponding output of described input vector, thereby obtain a plurality of by described input vector and the input and output Vector Groups formed with the corresponding output of described input vector; And

Parameter calculation unit, according to the input and output Vector Groups that comprises in each group in described a plurality of groups, at each group in described a plurality of groups, calculate respectively and should organize corresponding power amplifier pattern function or each parameter of power amplifier pre-distortion model, thereby obtain a plurality of amplifier submodel functions or power amplifier pre-distortion submodel function.

Aspect 16, according to aspect 15 described amplifier model apparatus for establishing, it is characterized in that, described parameter calculation unit adopts least square method to come at each group in described a plurality of groups, calculates respectively should group corresponding power amplifier pattern function or each parameter of power amplifier pre-distortion model.

Aspect 17, according to aspect 15 described amplifier model apparatus for establishing, it is characterized in that, described parameter calculation unit adopts weighted least-squares method to come at each group in described a plurality of groups, calculate respectively should group corresponding power amplifier pattern function or each parameter of power amplifier pre-distortion model, in each described group, with described input in followingly import corresponding input vector and give bigger weight:

The summit in the mould value of this input and the residing interval of described mould value differs less than predetermined value.

Aspect 18, according to aspect 15 described amplifier model apparatus for establishing, it is characterized in that, corresponding power amplifier pattern function or the power amplifier pre-distortion model of each group in described a plurality of groups comprises amplitude model function and phase model function, described parameter calculation unit is at each group in described a plurality of groups, calculate each parameter of amplitude model function and phase model function respectively, and adopt the constraint least-squares algorithm that the average of the signal in the preset range of both sides, summit in described interval is equated.

Aspect 19, a kind of software, can make during by computer or the execution of other logic machine when being carried out or through compiling or after explaining by computer or other logic machine described computer or other logic machine realize above-mentioned aspect described device of 1-18 or method.

Aspect 20, a kind of storage medium are preserved above-mentioned aspect 18 described softwares on it.

The modelling apparatus and method that embodiment of the present invention proposes are suitable for segmentation memory nonlinear model structure and identification algorithm thereof, and can be used for modeling has the memory effect power amplifier, also can be used for the design of predistorter simultaneously.When utilizing embodiments of the present invention to carry out the power amplifier modeling, according to the power amplifier characteristic dynamic range of input signal is divided into several sections, each section uses one to have memory nonlinear function (being called a submodel) to be described respectively, according to the mould value of current input signal, select the output of corresponding submodel output as final power amplifier model.In the Model Distinguish process, each submodel independently carries out identification.After using the method for this segmentation modeling, for the different qualities problem of power amplifier at the different input power place, can accurate more modeling.In addition, each segmentation can be adopted the low order multinomial, reduces numerical value instability and computation complexity.When carrying out the predistorter design, this model also has higher flexibility ratio.

Description of drawings

In order to be illustrated more clearly in the embodiment of the invention or technical scheme of the prior art, to do to introduce simply to the accompanying drawing of required use in embodiment or the description of the Prior Art below, apparently, accompanying drawing in describing below only is some embodiments of the present invention, for those of ordinary skills, under the prerequisite of not paying creative work, can also obtain other accompanying drawing according to these accompanying drawings.

Fig. 1 shows the theory diagram of power amplifier behavior modeling method;

Fig. 2 shows the functional block diagram according to the power amplifier Model Distinguish device of one embodiment of the present invention;

Fig. 3 shows the functional block diagram according to the power amplifier Model Distinguish device of another execution mode of the present invention;

Fig. 4 shows the flow chart according to power amplifier identification Method of the present invention;

Fig. 5 shows the functional block diagram according to a kind of power amplifier relevant apparatus of embodiment of the present invention;

Fig. 6 shows a kind of power amplification system of employing according to the power amplifier predistortion device of embodiment of the present invention; And

Fig. 7 shows the schematic block diagram that can be used for implementing according to the computer of the method and apparatus of embodiment of the present invention.

Embodiment

Below in conjunction with accompanying drawing embodiments of the present invention are described in detail.In order to make the present invention clear succinct, this paper omitted may cause the present invention unclear, to the description of the parts of prior art.In addition, identical in this article or similar parts describe with identical Reference numeral, and have omitted the repeat specification to it.

Embodiments of the present invention provide power amplifier Model Distinguish device, and it can be used for power amplifier Model Distinguish module 108 shown in Figure 1.Should note, though in Fig. 1, power amplifier Model Distinguish module 108 links together online with power amplifier 102 grades, but power amplifier Model Distinguish module 108 can the work of off-line ground, and Model Distinguish device promptly of the present invention can be operated on off-line ground after the input and output of having collected power amplifier.

It may be noted that because there is time-delay in analog link, need at first the inputoutput data of the power amplifier collected is carried out synchronously, correctly to find and the corresponding output signal of the input of current time.Synchronous method can adopt the whole bag of tricks (for example correlation method) known to those skilled in the art to realize.Input and output signal after input described in the embodiment of the present invention and corresponding output all refer to synchronously.

Fig. 2 shows the functional block diagram according to the power amplifier Model Distinguish device of one embodiment of the present invention.As shown in Figure 2, comprise power amplifier input and output acquiring unit 21, mould value computing unit 22, grouped element 23, parameter calculation unit 24 according to power amplifier Model Distinguish device 20 of the present invention.

In one embodiment, power amplifier input and output acquiring unit 21 can comprise coupler shown in Figure 1 101, quadrature down converter 104, A/D converter 107, local oscillator 105, coupler 103, quadrature down converter 106, A/D converter 110, thereby directly obtains the input and the corresponding output of power amplifier 102 from power amplifier 102.In this embodiment, power amplifier input and output acquiring unit 21 can comprise that memory cell stores the data that obtained.

In another embodiment, power amplifier input and output acquiring unit 21 can comprise quadrature down converter shown in Figure 1 104, A/D converter 107, local oscillator 105, quadrature down converter 106, A/D converter 110, thereby by obtaining the input and the corresponding outputs of power amplifiers 102 with the coupler 101 of the discrete setting of power amplifier Model Distinguish device of embodiment of the present invention and 103 from power amplifier 102.In this embodiment, power amplifier input and output acquiring unit 21 can comprise that memory cell stores the data that obtained.

In another execution mode, power amplifier input and output acquiring unit 21 can comprise memory cell and receiving element.Receiving element obtains the input and the corresponding output of power amplifier 102 from the outside of the power amplifier Model Distinguish device 20 of embodiment of the present invention.This receiving element can be from the input and the corresponding output of A/D converter 107 and A/D converter 110 reception power amplifiers 102.This receiving element also can for example pass through network, radio wave or USB interface etc. receive power amplifier 102 from the device of the input of storage such as remote server, external memory storage power amplifier 102 and corresponding output input and corresponding output.Memory cell can be used to store the data that this receiving element receives.

Mould value computing unit 22 is used to calculate the mould value of described input.Grouped element 23 is according to the residing interval of mould value (mould value interval) of each described input, each input signal and corresponding output signal branch thereof is constantly gone into a plurality of intervals, make up each input vector constantly simultaneously and also determine to export accordingly, thereby obtain a plurality of input and output Vector Groups (hereinafter will describe in detail) with this input vector.Parameter calculation unit 24 is according to the input vector and the corresponding output that comprise in each group in described a plurality of groups, at each group in described a plurality of groups, calculate each parameter of the power amplifier submodel function corresponding respectively with each group, thereby obtain a plurality of power amplifier submodel functions, these power amplifier submodel functions can be the power amplifier pattern functions of same type, for example all be the polynomial module type function or all be the wiener pattern function, it also can be dissimilar power amplifier pattern functions, be the wiener pattern function for example in first interval, and be the polynomial module type function in second interval, in the 3rd interval is the polynomial module type function, is Hammerstein-wiener pattern function etc. in the 4th interval.The Hammerstein-wiener pattern function for example can be referring to " vector machine identification of Hammerstein-wiener model least square and the application thereof " delivered on the 25th volume the 3rd phase " control theory and application " in June, 2008 such as Gui Weihua, by reference this document is incorporated herein, as describing in detail in this article.The interval corresponding power amplifier submodel function of these isotype respectively values has constituted the power amplifier model.Thereby parameter calculation unit 24 comprises a plurality of independently power amplifier submodel function calculation unit.

Fig. 3 shows the functional block diagram according to the power amplifier Model Distinguish device of another execution mode of the present invention.As shown in Figure 3,, except comprising power amplifier input and output acquiring unit 21 shown in Figure 2, mould value computing unit 22, grouped element 23, parameter calculation unit 24, also comprise unit 25 is set according to power amplifier Model Distinguish device of the present invention.

In one embodiment, unit 25 is set is used for mould value interval is provided with and changes, for example can be provided with or change the starting point in number, each mould value length of an interval degree, each mould value interval in mould value interval and end point etc.

In another execution mode, unit 25 is set is used for being provided with and changing each interval pairing power amplifier submodel function (also can abbreviate submodel as).For example the first interval pairing multinomial power amplifier pattern function can be changed over wiener power amplifier pattern function etc.

In another execution mode, unit 25 is set both had been used for being provided with and changing each interval pairing power amplifier submodel function, also be used for being provided with and changing each mould value interval.

Below each power amplifier submodel function calculation unit is described.

For the convenience that illustrates, in the following description, each power amplifier submodel function all adopts memory polynomial module type function.But should note; this is not a limiting the scope of the invention; each power amplifier submodel function of the present invention can adopt identical power amplifier pattern function also can adopt different power amplifier pattern functions, and in one embodiment, each power amplifier pattern function is for there being the non-linear power amplifier pattern function of memory.

When submodel adopts the memory multinomial, the expression formula of power amplifier model as the formula (1),

$y\left(n\right)=\left\{\begin{array}{c}\underset{i=1}{\overset{{\mathrm{<figure-callout\; id="1"\; label="K"\; filenames="HSA00000010485200052.png"\; state="\{\{state\}\}">K</figure-callout>}}_1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{Q_1-1}{\mathrm{\&Sigma;}}}{a}_{\mathrm{1,2}i-1,j}x(n-j){\left|x(n-j)\right|}^{2(i-1)},0\&le;\left|x\left(n\right)\right|<{\mathrm{<figure-callout\; id="1"\; label="S"\; filenames="HSA00000010485200052.png"\; state="\{\{state\}\}">S</figure-callout>}}_1\\ \underset{i=1}{\overset{K_2}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{Q_2-1}{\mathrm{\&Sigma;}}}{a}_{\mathrm{2,2}i-1,j}x(n-j){\left|x(n-j)\right|}^{2(i-1)},{\mathrm{<figure-callout\; id="1"\; label="S"\; filenames="HSA00000010485200052.png"\; state="\{\{state\}\}">S</figure-callout>}}_1\&le;\left|x\left(n\right)\right|<S_2\\ .\\ .\\ .\\ \underset{i=1}{\overset{K_N}{\mathrm{\&Sigma;}}}\stackrel{Q_N-1}{\underset{j=0}{\mathrm{\&Sigma;}}}{a}_{N,2i-1,j}x(n-j){\left|x(n-j)\right|}^{2(i-1)},\left|x\left(n\right)\right|\&GreaterEqual;S_{N-1}\end{array}\right.---\left(1\right)$

Here, model be input as x (n), model is output as y (n), input signal amplitude (mould value) dynamic range is by waypoint (segmentation mould value) S ₁, S ₂..., S _N-1Be divided into N interval.For submodel l, its memory depth is Q _l, non-linear exponent number is 2K _l-1, submodel parameter a _{L, 2i-1, j}Be complex coefficient.

When this model is used for system shown in Figure 1 the modeling of memory effect power amplifier is arranged, need the by stages to the independent identification of each submodel.In one embodiment, as previously mentioned, power amplifier input and output acquiring unit 21 collect one group of A/D converter 107 continuously constantly output and one group of corresponding output of corresponding A/D converter 110, respectively as the input signal and the desired output signal of power amplifier model.Mould value computing unit 22 calculates each input signal mould value constantly.Grouped element 23 is according to the result of calculation of mould value computing unit 22, each input signal and corresponding desired output signal branch thereof is constantly gone into a plurality of intervals, make up each input vector constantly simultaneously and also determine to export accordingly, thereby obtain a plurality of input and output Vector Groups with this input vector.At each interval, by each each power amplifier submodel function of power amplifier submodel function calculation unit identification.The example that is recognized as with submodel l describes below.

The output of the A/D converter 107 that power amplifier input and output acquiring unit 21 is collected is designated as x (1), x (2) ..., x (n) ..., be called power amplifier model input signal; The output of the A/D converter 110 of its correspondence of collecting is designated as y (1), y (2) ..., y (n) ..., be called power amplifier model desired output signal.The maximal memory degree of depth in N submodel of definition power amplifier is Q _Max=max{Q ₁, Q ₂..., Q _N.Like this from Q _MaxConstantly begin, calculate each input signal x (Q constantly _Max), x (Q _Max+ 1) ..., x (n) ... the mould value.With moment n is example, if input signal x (n) satisfies formula S _L-1≤ | x (n) |＜S _l, then with this signal and Q before thereof _l-1 signal record constantly, and calculate a line of input vector that belongs to interval l according to following form,

$[x\left(n\right),x(n-1),\&CenterDot;\&CenterDot;\&CenterDot;,x(n-Q_l+1),$

$x\left(n\right)\&CenterDot;{\left|x\left(n\right)\right|}^2,\&CenterDot;\&CenterDot;\&CenterDot;,x(n-Q_l+1)\&CenterDot;{\left|x(n-Q_l+1)\right|}^2,$

$...$

$x\left(n\right)\&CenterDot;{\left|x\left(n\right)\right|}^{2(K_l-1)},\&CenterDot;\&CenterDot;\&CenterDot;,x(n-Q_l+1)\&CenterDot;{\left|x(n-Q_l+1)\right|}^{2(K_l-1)}]$

Write down the corresponding desired output signal of the output signal y (n) of this moment A/D converter 110 simultaneously as submodel l.The rest may be inferred constantly for other.For example in one embodiment, at moment n+1,, can obtain next row vector if this input signal constantly falls into interval l:

$[x(n+1),x\left(n\right),\&CenterDot;\&CenterDot;\&CenterDot;,x(n-Q_l+2),$

$x(n+1)\&CenterDot;{\left|x(n+1)\right|}^2,\&CenterDot;\&CenterDot;\&CenterDot;,x(n-Q_l+2)\&CenterDot;{\left|x(n-Q_l+2)\right|}^2,$

$...$

$x(n+1)\&CenterDot;{\left|x(n+1)\right|}^{2(K_l-1)},\&CenterDot;\&CenterDot;\&CenterDot;,x(n-Q_l+2)\&CenterDot;{\left|x(n-Q_l+2)\right|}^{2(K_l-1)}]$

Each interval all obtains one group of line of input vector and corresponding desired output signal thereof like this.

One group of line of input vector that will belong to interval l is designated as X _l(1), X _l(2) ..., X _l(M), its corresponding desired output signal is designated as y _l(1), y _l(2) ..., y _l(M).These row vectors constitute the input matrix of submodel l

$X_l=\left[\begin{array}{c}{X}_l\left(1\right)\\ X_l\left(2\right)\\ .\\ .\\ .\\ X_l\left(M\right)\end{array}\right]$

Note submodel l coefficient vector to be asked is

$A_l={[a_{l,\mathrm{1,0}},a_{l,\mathrm{1,1}},\&CenterDot;\&CenterDot;\&CenterDot;,a_{l,1,Q_1-1},a_{l,\mathrm{3,0}},\&CenterDot;\&CenterDot;\&CenterDot;,a_{l,3,Q_l-1},\&CenterDot;\&CenterDot;\&CenterDot;,a_{l,2,K_l-\mathrm{1,0}},\&CenterDot;\&CenterDot;\&CenterDot;,a_{l,2K_l-1,Q_l-1}]}^T$

Output vector is

Y _l＝[y _l(1)，y _l(2)，…，y _l(M)] ^T

Then the coefficient of submodel l can obtain according to least-squares algorithm

$A_l={(X_l^H\&CenterDot;X_l)}^{-1}\&CenterDot;(X_l^H\&CenterDot;Y_l)---\left(2\right)$

It is noted that because different interval submodel is an independent process, guarantee that the continuity of submodel output valve at interval intersection point place between two adjacent regions is favourable.Can adopt following two kinds of methods to reduce discontinuous influence according to the embodiment of the present invention.

Method one, use weighted least square algorithm, increase the weight of near the weight of the data interval border point (waypoint).Just give bigger weight for input vector in each described group, corresponding with following input: the summit in the mould value of this input and the residing interval of described mould value differs less than predetermined value.Still with the example that is recognized as of submodel l.On the basis of formula (2), increase a weighting matrix C _l, in one embodiment, it is the diagonal matrix of a M * M, diagonal entry be (0, the M) real number between, satisfying the diagonal entry sum simultaneously is M.The Elements C of weighting matrix _l(i, i) i of expression belongs to the pairing weight of line of input vector of interval l.Do not add C temporary _lIt is a unit matrix.When adopting weighting algorithm, determine a bounds Δ, find all to satisfy S _l-Δ≤| x (n) |≤S _lAnd S _L-1≤ | x (n) |≤S _L-1Line of input vector in the+Δ scope, its corresponding weight is set to the value greater than 1, and all the other weights are that the condition of M calculates by the element sum.The coefficient of so model l is obtained by formula (3)

$A_l={(X_l^H\&CenterDot;C_l\&CenterDot;X_l)}^{-1}\&CenterDot;(X_l^H\&CenterDot;C_l\&CenterDot;Y_l)---\left(3\right)$

This weighted least-squares can be so that more accurate near the data modeling at interval border point place, thereby reduce the noncontinuity at separation place.

Method two, respectively during modeling, use the constraint least-squares algorithm, equate to reduce discontinuity by the signal average that guarantees interval intersection point both sides at amplitude and phase place.At amplitude and phase place modeling respectively, be meant that for each interval sub-power amplifier pattern function comprises sub-power amplifier pattern function of amplitude and phason power amplifier pattern function.For the sub-power amplifier model of amplitude, the signal average equates to be meant that power or amplitude average equate.For phason power amplifier model, the signal average equates to be meant that average phase equates.Below describe with the amplitude characteristic modeling.When submodel adopted the memory multinomial, the amplitude characteristic expression formula of power amplifier model as the formula (4)

$y\left(n\right)=\left\{\begin{array}{c}\underset{i=1}{\overset{{\mathrm{<figure-callout\; id="1"\; label="K"\; filenames="HSA00000010485200052.png"\; state="\{\{state\}\}">K</figure-callout>}}_1}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{Q_1-1}{\mathrm{\&Sigma;}}}{\mathrm{<figure-callout\; id="1"\; label="c"\; filenames="HSA00000010485200052.png"\; state="\{\{state\}\}">c</figure-callout>}}_{\mathrm{1,2}i-1,j}{\left|x(n-j)\right|}^{2i-1},0\&le;\left|x\left(n\right)\right|<{\mathrm{<figure-callout\; id="1"\; label="S"\; filenames="HSA00000010485200052.png"\; state="\{\{state\}\}">S</figure-callout>}}_1\\ \underset{i=1}{\overset{K_2}{\mathrm{\&Sigma;}}}\underset{j=0}{\overset{Q_2-1}{\mathrm{\&Sigma;}}}{c}_{\mathrm{2,2}i-1,j}{\left|x(n-j)\right|}^{2i-1},{\mathrm{<figure-callout\; id="1"\; label="S"\; filenames="HSA00000010485200052.png"\; state="\{\{state\}\}">S</figure-callout>}}_1\&le;\left|x\left(n\right)\right|<S_2\\ .\\ .\\ .\\ \underset{i=1}{\overset{K_N}{\mathrm{\&Sigma;}}}\stackrel{Q_N-1}{\underset{j=0}{\mathrm{\&Sigma;}}}{c}_{N,2i-1,j}{\left|x(n-j)\right|}^{2i-1},\left|x\left(n\right)\right|\&GreaterEqual;S_{N-1}\end{array}\right.---\left(4\right)$

The parameter meaning here is identical with formula (1), and difference is the submodel parameter c _{L, 2i-1, j}Be real coefficient.

Guarantee that the required constraints of continuity describes with the example that is recognized as of submodel l.Suppose that submodel l-1 identification finishes.Determine a bounds Δ, under submodel l-1, find all to satisfy S in the interval _L-1-Δ≤| x (n) |≤S _L-1The line of input vector, each line of input vector has following form

$[\left|x\left(n\right)\right|,\left|x(n-1)\right|,\&CenterDot;\&CenterDot;\&CenterDot;,\left|x(n-Q_{l-1}+1)\right|,{\left|x\left(n\right)\right|}^3,\&CenterDot;\&CenterDot;\&CenterDot;,{\left|x(n-Q_{l-1}+1)\right|}^3,$

$\&CenterDot;\&CenterDot;\&CenterDot;,{\left|x\left(n\right)\right|}^{2K_{l-1}-1},\&CenterDot;\&CenterDot;\&CenterDot;,{\left|x(n-Q_{l-1}+1)\right|}^{2K_{l-1}-1}]$

These input vectors are designated as

The coefficient of the submodel l-1 that note has picked out is

$C_{l-1}={[c_{l-\mathrm{1,1,0}},c_{l-\mathrm{1,1,1}},\&CenterDot;\&CenterDot;\&CenterDot;,c_{l-\mathrm{1,1},Q_{l-1}-1},c_{l-\mathrm{1,3,0}},\&CenterDot;\&CenterDot;\&CenterDot;,c_{l-\mathrm{1,3},Q_{l-1}-1},\&CenterDot;\&CenterDot;\&CenterDot;,c_{l-\mathrm{1,2}{K}_{l-1}-\mathrm{1,0}},\&CenterDot;\&CenterDot;\&CenterDot;,c_{l-\mathrm{1,2}{K}_{l-1}-1Q_{l-1}-1}]}^T$

Then in interval point of intersection S _L-1The signal average in left side is

$\mathrm{cons}=\left(\frac{1}{R}\underset{t=1}{\overset{R}{\mathrm{\&Sigma;}}}{x}_{l-1,S_{l-1}}\left(t\right)\right)\&CenterDot;C_{l-1}$

This value is then as the binding occurrence of identification submodel l.

Under submodel l, find all to satisfy S in the interval _L-1≤ | x (n) |≤S _L-1The line of input vector of+Δ is designated as

The coefficient of remembering submodel l to be identified is

$C_l={[c_{l,\mathrm{1,0}},c_{l,\mathrm{1,1}},\&CenterDot;\&CenterDot;\&CenterDot;,c_{l,1,Q_{l1}-1},c_{l,\mathrm{3,0}},\&CenterDot;\&CenterDot;\&CenterDot;,c_{l,3,Q_l-1},\&CenterDot;\&CenterDot;\&CenterDot;,c_{l,2K_l-\mathrm{1,0}},\&CenterDot;\&CenterDot;\&CenterDot;,c_{l,2K_l-1,Q_l-1}]}^T$

Then when finding the solution the least square coefficient, increase constraints (5)

$\left(\frac{1}{P}\underset{t=1}{\overset{P}{\mathrm{\&Sigma;}}}{X}_{l,S_{l-1}}\left(t\right)\right)\&CenterDot;C_l=\mathrm{con}---\left(5\right)$

The least square of this belt restraining condition is found the solution and can be used Lagrange (Lagrange) solving method to finish.

Should be noted that though above when power amplifier Model Distinguish device of the present invention is described, each submodel has all adopted the power amplifier of memory model, thereby can carry out modeling to the power amplifier that memory effect is arranged.But each submodel also can all adopt memoryless power amplifier model, thereby can carry out modeling to the power amplifier of memory-less effect.The for example memoryless multinomial power amplifier of memoryless model pattern function etc.

The functional block diagram of power amplifier Model Distinguish device shown in Figure 3 also can be used for the power amplifier predistortion Model Distinguish.In power amplifier input and output acquiring unit 21,, the input of the power amplifier desired output as the predistortion model is got final product the output of power amplifier input as the predistortion model.

Fig. 4 shows the flow chart according to power amplifier identification Method of the present invention, and it also can be considered as is the operational flowchart of power amplifier Model Distinguish device 20.

As shown in Figure 4, at first in step 401, by power amplifier input and output acquiring unit 21 obtain the input of power amplifier and respectively with the input corresponding output.In step 402, mould value computing unit 22 carries out the mould value and calculates then.In step 403, grouped element 23 divides into groups by mould value interval to input signal according to mould value result calculated, and obtains the output of each interval input vector and each input vector correspondence.In step 404, parameter calculation unit 24 is calculated the pattern function parameter value of each group, thereby determines the pattern function of each group.

Fig. 5 shows the functional block diagram according to a kind of power amplifier relevant apparatus of embodiment of the present invention.As shown in Figure 5, the power amplifier relevant apparatus according to one embodiment of the present invention comprises: mould value computing unit 51, a plurality of output computing unit 52 (52-1 is to 52-N) and selector 53.

Mould value calculating part 51 calculates the mould value of current input signal, the interval that falls into according to this mould value, selector 53 selects the output of corresponding output computing unit 52 as final output, promptly determines an output computing unit that should be input to as described input signal in described a plurality of output computing units.

In one embodiment, each is exported computing unit 52 and carries out work according to certain power amplifier submodel.In one embodiment, these submodels are for there being the memory nonlinear model, for example multiple typical model such as Wiener model, memory multinomial model.In another embodiment, these submodels are memoryless model, for example memoryless multinomial power amplifier pattern function etc.In this case, this power amplifier relevant apparatus is as the power amplifier simulator.Each exports computing unit 52 is separate, can adopt separate dissimilar pattern function, also can adopt the separate same model function of parameter between the different output computing units.

In another embodiment, each is exported computing unit 52 and carries out work according to certain power amplifier predistortion submodel.At this moment this power amplifier relevant apparatus is as the power amplifier predistortion device.

Fig. 6 shows a kind of power amplification system of employing according to the power amplifier predistortion device of embodiment of the present invention.The operation principle of this power amplifier predistortion device is referring to Lei Ding, G.Tong Zhou, DennisR.Morgan, et al.A robust digital baseband predistorter constructed usingmemory polynomials.IEEE Trans.Communications, 2004,52 (1): 159-165 is incorporated herein this document, by reference as describing in detail in this article.

As shown in Figure 6, behind the base band information source information process predistorter 602 (power amplifier pre-distortion device) from information source 601, obtain signals after pre-distortion, this signal is converted to analog signal by digital to analog converter 604.After this analog signal is radiofrequency signal through upconverter 605 frequency up-converted, input power amplifier 606.Signal after power amplifier 606 amplifies is through antenna transmission.Feed back to analog to digital converter 608 after a part of signal process low-converter 607 down-conversions of the output of power amplifier 606 simultaneously, through the power amplifier output digital baseband signal that obtains feeding back after analog to digital converter 608 samplings.Gain normalization module 609 is exported digital baseband signal with the power amplifier of feedback and is carried out normalization according to the linear gain value G of power amplifier, with this signal value input predistortion training aids 610.Here predistortion training aids 610 adopts identical structure with predistorter 602, it all can be realized by model structure shown in Figure 5, respectively export computing unit 52 carry out work according to certain power amplifier predistortion submodel this moment, and each power amplifier predistortion submodel can be that the power amplifier predistortion submodel of memory is arranged also can be memoryless power amplifier predistortion submodel.Adder 603 is used for calculating poor between the output of the actual output of predistortion training aids 610 and predistorter 602, and this difference is used for the parameter identification of predistortion training aids 610.Predistortion training aids 610 with the output of gain normalization module 609 as input signal, with the output of the predistorter 602 desired output signal as training aids 610, the predistortion function of each section that is adopted can get as the power amplifier predistortion Model Distinguish apparatus and method of front in identification.Simultaneously, the parameter of predistortion training aids 610 is copied to predistorter 602, obtain the parameter of predistorter.Under the parameter of this new predistorter 602, the parameter of the predistorter 602 of the parameter of calculation training device 610, and renewal once more.This process circulation is carried out, and restrains until predistorter.

Power amplification system shown in Figure 6 is exemplary, is not limitation of the present invention, and pre-distortion device of the present invention can be applied in other the power amplification system.

Each composition module, unit, subelement can be configured by the mode of software, firmware, hardware or its combination in the said apparatus.Dispose spendable concrete means or mode and be well known to those skilled in the art, do not repeat them here.Under situation about realizing by software or firmware, to the computer with specialized hardware structure (being incorporated in computer or all-purpose computer for example shown in Figure 7 the transmitter) program that constitutes this software is installed from storage medium or network, this computer can be carried out various functions etc. when various program is installed.

Fig. 7 shows the schematic block diagram that can be used for implementing according to the computer of the method and apparatus of the embodiment of the invention.

In Fig. 7, CPU (CPU) 701 carries out various processing according to program stored among read-only memory (ROM) 702 or from the program that storage area 708 is loaded into random-access memory (ram) 703.In RAM 703, also store data required when CPU 701 carries out various processing or the like as required.CPU 701, ROM 702 and RAM 703 are connected to each other via bus 704.Input/output interface 705 also is connected to bus 704.

Following parts are connected to input/output interface 705: importation 706 (comprising keyboard, mouse or the like), output 707 (comprise display, such as cathode ray tube (CRT), LCD (LCD) etc. and loud speaker etc.), storage area 708 (comprising hard disk etc.), communications portion 709 (comprising that network interface unit is such as LAN card, modulator-demodulator etc.).Communications portion 709 is handled such as the internet executive communication via network.As required, driver 710 also can be connected to input/output interface 1705.Detachable media 711 can be installed on the driver 710 as required such as disk, CD, magneto optical disk, semiconductor memory or the like, makes the computer program of therefrom reading be installed to as required in the storage area 708.

Realizing by software under the situation of above-mentioned series of processes, such as detachable media 711 program that constitutes software is being installed such as internet or storage medium from network.

It will be understood by those of skill in the art that this storage medium is not limited to shown in Figure 7 wherein having program stored therein, distribute separately so that the detachable media 711 of program to be provided to the user with equipment.The example of detachable media 711 comprises disk (comprising floppy disk (registered trade mark)), CD (comprising compact disc read-only memory (CD-ROM) and digital universal disc (DVD)), magneto optical disk (comprising mini-disk (MD) (registered trade mark)) and semiconductor memory.Perhaps, storage medium can be hard disk that comprises in ROM 702, the storage area 708 or the like, computer program stored wherein, and be distributed to the user with the equipment that comprises them.

The present invention also proposes a kind of program product that stores the instruction code that machine readable gets.When described instruction code is read and carried out by machine, can carry out above-mentioned method according to the embodiment of the invention.

Correspondingly, being used for carrying the above-mentioned storage medium that stores the program product of the instruction code that machine readable gets is also included within of the present invention open.Described storage medium includes but not limited to floppy disk, CD, magneto optical disk, storage card, memory stick or the like.

Claims (10)

1. a power amplifier relevant apparatus is characterized in that, this power amplifier relevant apparatus comprises mould value computing unit, selected cell and a plurality of output computing unit: wherein

Described mould value computing unit is used to calculate the mould value of input signal;

Described selected cell is used for the mould value that calculates according to described mould value computing unit, determines an output computing unit that should be input to as described input signal in described a plurality of output computing unit; And

Each described output computing unit calculates the signal that should export according to the described input signal of this output computing unit of input.

2. power amplifier relevant apparatus according to claim 1 is characterized in that, each described output computing unit calculates the described signal that should export according to the different same functions of parameter.

3. power amplifier relevant apparatus according to claim 1 is characterized in that, the employed function of each described output computing unit is separate between described a plurality of described output computing units.

4. power amplifier relevant apparatus according to claim 1 is characterized in that, each described output computing unit is all according to the non-linear power amplifier pattern function of memory being arranged or having the non-linear power amplifier predistortion pattern function of memory to calculate the signal that should export.

5. power amplifier relevant apparatus according to claim 1 is characterized in that, each described output computing unit all calculates the signal that should export according to memoryless power amplifier pattern function or memoryless power amplifier predistortion pattern function.

6. power amplifier pre-distortion system, it is characterized in that, this power amplifier pre-distortion system comprises each described power amplifier relevant apparatus of claim 1-5 as the power amplifier predistortion device, and the described output computing unit of each of wherein said power amplifier relevant apparatus calculates the signal that should export according to the power amplifier predistortion pattern function.

7. power amplifier model or power amplifier pre-distortion method for establishing model, described method comprises:

The input and output obtaining step obtains the input of power amplifier model or power amplifier pre-distortion model and distinguishes corresponding output with described input;

Mould value calculation procedure is calculated each mould value of importing constantly in the described input;

The grouping step, according to each the residing interval of mould value imported constantly in the described input, be divided into a plurality of groups with described input with the corresponding respectively output of described input, and structure and described respectively import corresponding input vector and with the corresponding output of described input vector, thereby obtain a plurality of by described input vector and the input and output Vector Groups formed with the corresponding output of described input vector; And

The calculation of parameter step, according to the input and output Vector Groups that comprises in each group in described a plurality of groups, at each group in described a plurality of groups, calculate respectively should group corresponding power amplifier pattern function or each parameter of power amplifier pre-distortion model, thereby obtain a plurality of amplifier submodel functions.

8. according to the described method of claim 7, it is characterized in that described calculation of parameter step adopts least square method to come at each group in described a plurality of groups, calculate respectively each parameter of should group corresponding power amplifier pattern function.

9. according to the described method of claim 8, it is characterized in that, described calculation of parameter step adopts weighted least-squares method to come at each group in described a plurality of groups, calculate respectively should group corresponding power amplifier pattern function or each parameter of power amplifier pre-distortion model, in each described group, with described input in followingly import corresponding input vector and give bigger weight:

The summit in the mould value of this input and the residing interval of described mould value differs less than predetermined value.

10. according to the described method of claim 7, it is characterized in that, corresponding power amplifier pattern function or the power amplifier pre-distortion pattern function of each group in described a plurality of groups comprises amplitude model function and phase model function, described calculation of parameter step is at each group in described a plurality of groups, calculate each parameter of amplitude model function and phase model function respectively, and adopt the constraint least-squares algorithm that the average of the signal in the preset range of both sides, summit in described interval is equated.

Images