CN101764577B - Baseband pre-distortion power amplifier linearization method based on one-way feedback and non-iterative technique - Google Patents

Abstract

The invention discloses a baseband pre-distortion power amplifier linearization method based on one-way feedback and a non-iterative technique, belonging to the technical field of electronics and relating to a power amplifier linearization technique. Two orthogonal training sequences of linearly increased amplitudes are input into a power amplifier successively, and an improved look-up table is constructed by an I path feedback signal sequence output by power amplifier; the input signals of the whole power amplifier system calculates and looks for corresponding look-up table values by the amplitudes, then corresponding pre-distortion process is carried out on the I path and Q path baseband input signals, and the input signals after pre-distortion are output to the power amplifier after D/A conversion, filtration and up-conversion, thus obtaining the output signals which show linear amplification with the baseband input signals of the whole power amplifier system. The whole pre-distortion process can be realized in a digital processing chip (comprising DSP or FPGA). Compared with the prior art, the method does not need arctangent or inverse cosine calculation and does not need coordinate transformation, thus reducing calculation amount, increasing calculation efficiency and ensuring the system structure to be simpler.

Description

Baseband pre-distortion power amplifier linearization method based on single channel feedback and non-iterative technique

Technical field

The invention belongs to electronic technology field, relate to the power amplifier linearization technology, be specifically related to a kind of wide-band radio frequency power amplifier linearization technique based on the digital baseband pre-distortion technology.

Technical background

Utopian power amplifier should be a linear unit, promptly for an input signal x (t)=re ^{J θ}(r is the amplitude of input signal, and θ is a phase of input signals), output signal y (t)=Krcos (ω t+ θ) (K is the gain of ideal linearity power amplifier, the ω carrier angular frequencies).Yet actual power amplifier is a nonlinear device, for an input signal x (t)=re ^{J θ}Its output signal is y (t)=B (r) cos (ω t+ θ+Φ (r)), wherein B (r) is the amplitude function of nonlinear power amplifier output signal, Φ (r) is the additive phase function (studies show that all the amplitude r with input signal is relevant with additive phase in the gain of nonlinear power amplifier) of nonlinear power amplifier.Therefore, nonlinear characteristic is an inherent characteristic of power amplifier.

Along with mobile communication technology develops rapidly, it is more and more crowded that wireless communication frequency band becomes, band resource is more and more nervous, in order to hold more communication channel in limited spectral range, people have proposed the modulation system (M-QAM etc.) of some wideband digital transmission technologys (OFDM, WCDMA etc.) and high spectrum utilization.The signal that adopts these technology to transmit has characteristics such as non-constant envelope, broadband and high peak-to-average ratio, after modulation signal is by nonlinear power amplifier, will produce in the band and out-of-band distortion, adjacent channel is disturbed in the output signal spectrum expansion, increases the communication system error rate.Therefore, Modern Communication System has proposed very high requirement to the linearity of radio-frequency power amplifier, and the power amplifier linearization technology has become one of key technology of next generation wireless communication system, becomes a big focus of world today's research.

The power amplifier linearization technology is a lot, and commonly used have back-off, feedforward, negative feedback and a predistortion etc.The back-off method is traditional and effective a kind of method, yet the rollback of working point has reduced the power utilization efficient of power amplifier and caused very high heat dissipation; Feedforward can accomplish that amplify in the broadband and high linearity, is a kind of linearization technique of present comparative maturity, but that its shortcoming is an efficient is low and complicated; Negative feedback is the method that exchanges distortion suppression with the sacrifice Amplifier Gain for, and its narrow bandwidth and stability problem have limited its application; Predistortion is divided into analog predistortion and digital pre-distortion, and advantage such as that digital pre-distortion technology has is stable, efficient, wide bandwidth and self adaptation can reach moderate linearisation, is more promising a kind of linearization technique.

Digital pre-distortion technology is to add a predistorter with power amplifier nonlinear characteristic contrary before power amplifier, predistorter carries out preliminary treatment to the IQ baseband signal in numeric field, after D/A conversion, filtering, up-conversion, send into power amplifier through pretreated baseband signal, power amplifier is launched after signal is amplified, owing in the predistorter of base band, signal has been carried out non-linear preliminary treatment, and the nonlinear characteristic contrary of its nonlinear characteristic and power amplifier then can obtain the radiofrequency signal of linear amplification at the power amplifier output.Usually, look-up table parameter or polynomial coefficient for adaptive acquisition predistorter, the part of power amplifier output signal need be fed back after decay, feedback signal is used for the adaptive learning algorithm after down-conversion, filtering, A/D conversion, obtain or adjust the parameter (look-up table or polynomial coefficient) of predistorter by adaptive algorithm, make the nonlinear characteristic of predistorter and the nonlinear characteristic contrary of power amplifier, thereby reach linearizing purpose.

In the digital baseband predistortion, the normal adaptive algorithm that adopts has LMS (lowest mean square), RLS iterative algorithms such as (recurrence least squares), these algorithms all need through study and iteration, the prerequisite that algorithm can obtain optimum predistorter parameter is to guarantee that algorithm can restrain, and all needs usually could obtain the result that is satisfied with through a large amount of calculating based on the adaptive algorithm of iteration.

At the shortcoming that iterative method exists, people have proposed based on non-iterative pre-distortion method, and existing non-iteration pre-distortion method principle is as follows:

At first adopt training sequence generator to produce one group of training sequence that amplitude linearity increases, the amplitude peak of establishing input signal is a normalization level 1, and the initial phase of training sequence is

Then i input signal constantly is (i=1 wherein, 2 ..., N), memoryless non-linear power amplification system shows as AM/AM conversion and AM/PM conversion, and this signal is (the IQ signal being write as the form of complex signal) through the signal that feeds back behind power amplifier:

$=y_I\left(i\right)+{\mathrm{jy}}_Q\left(i\right)$

Wherein, y _I(i) and y _Q(i) I road signal that feeds back constantly for i and Q road signal, With Expression amplitude respectively is

Signal through the range value of the signal that feeds back behind the power amplifier and the change amount of phase place, order $r_i=A\left(\frac{i}{N}\right),$ ${\mathrm{\θ}}_i=\mathrm{\Φ}\left(\frac{i}{N}\right)$ The i that then can obtain feeding back IQ two paths of signals constantly is respectively:

According to above two formulas, the IQ two paths of signals that feeds back to just can be obtained r through following computing _iAnd θ _iValue

$r_i=\sqrt{y_I{\left(i\right)}^2+y_Q{\left(i\right)}^2},(i=\mathrm{1,2},...,N)---\left(4\right)$

In order to simplify computing, can be with the initial phase of training sequence

Be set at 0.For i moment input range be

And initial phase is 0 signal, has become r through its amplitude of the non-linear amplification of power amplifier back _i, phase place has changed θ on the basis of original signal _iBased on the principle (accompanying drawing 1) of indirect learning structure, the power amplifier inversion model parameter that available back distortion methods obtains is as the parameter of predistorter, for make after distortion is later handled signal with import power amplifier before signal equate, be r then for input range _iSignal, its amplitude should become i/N after distortion is later handled, and the change amount of phase place is-θ _iIf tried to achieve about signal amplitude r _iWith phase change amount-θ _iSequence, then can obtain the look-up table parameter of predistorter according to back distortion principle, the look-up table of corresponding predistorter is:

Table 1

According to above method, the training sequence that increases with amplitude linearity passes through power amplification system earlier, and the IQ two paths of signals that feeds back to is gathered, according to (4) (5) two formulas, through calculating the parameter r that just can obtain look-up table in the predistorter _iWith-θ _iThereby the predistorter that can make up in the form 1 is searched table.

When being pre-distortion, the major defect of this non-alternative manner the IQ conversion of signals need be become polar form, amplitude according to input signal is revised amplitude and phase place respectively, and the signal of revised polar form need be converted again to the form of IQ signal, therefore, pre-distortion needs twice Coordinate Conversion to finish, and is obtaining look-up table parameter θ _iProcess in need arctangent cp cp operation, computing is complicated.

Summary of the invention

The invention provides a kind of baseband pre-distortion power amplifier linearization method based on single channel feedback and non-iterative technique, this method can be simplified calculating process greatly, raise the efficiency based on single channel feedback and non-iterative technique, and can simplify and realize circuit and reduce system cost.

The objective of the invention is to be achieved through the following technical solutions:

A kind of baseband pre-distortion power amplifier linearization method based on single channel feedback and non-iterative technique as shown in Figure 2, may further comprise the steps:

Step 1: one group of amplitude linearity increased and phase place is 0 training sequence

I=1 wherein, 2 ..., i ... N is input to power amplifier as input signal after D/A conversion, filtering and up-conversion; Gather I road feedback signal then, be about to the power amplifier output signal through coupling, down-conversion, filtering and A/D conversion; If the amplitude of the maximum input signal that power amplifier allows is a normalization level 1, then the input signal of i moment power amplifier is

I moment I road feedback signal is I ₁(i); Then corresponding to training sequence

Feedback sequence be { I ₁(i) }.Training sequence

The available sequences generator produces.

Step 2: with another group amplitude linearity growth and phase place is the training sequences of 90 degree

I=1 wherein, 2 ..., i ... N is input to power amplifier as input signal after D/A conversion, filtering and up-conversion; Gather I road feedback signal then, be about to the power amplifier output signal through coupling, down-conversion, filtering and A/D conversion; If the amplitude of the maximum input signal that power amplifier allows is a normalization level 1, then the input signal of i moment power amplifier is

I moment I road feedback signal is I ₂(i); Then corresponding to training sequence

Feedback sequence be { I ₂(i) }.Training sequence

The available sequences generator produces.

Step 3: make up initial pre-distortion look-up table.

The pre-distortion look-up table comprises index entry and searches list item and form that wherein index entry is input signal amplitude r _i, search list item and comprise look-up table LUT1 and look-up table LUT2.

Wherein, input signal amplitude $r_i=\sqrt{I_1{\left(i\right)}^2+I_2{\left(i\right)}^2},$ Corresponding input signal amplitude r _iThe value of searching of following look-up table LUT1 is iI ₁(i)/r _i ²N, corresponding input signal amplitude r _iThe value of searching of following look-up table LUT2 is iI ₂(i)/r _i ²N.

Concrete initial pre-distortion look-up table is as shown in table 2:

Table 2

Step 4: the initial pre-distortion look-up table to step 3 gained carries out interpolation processing, obtains final pre-distortion look-up table.

The index entry of the initial pre-distortion look-up table of step 3 gained is input signal amplitude sequence r _iThis sequence is not equally spaced, search list item in order to obtain equally spaced other, that can adopt here that two kinds of methods obtain other range value correspondence searches the list item parameter: the one, obtain the value of pairing look-up table LUT1 of other input signal amplitude and look-up table LUT2 with linear interpolation method; Another kind is to utilize the method for curve fit to come match amplitude and phase characteristic curve with multinomial, and then fills the value of corresponding look-up table LUT1 and LUT2 by the calculating back according to the multinomial of match.

Step 5: for input signal x (the n)=re of whole power amplification system ^{J θ}=x _I(n)+jx _Q(n), n representation signal input time wherein, r is an input signal amplitude, θ is the input signal phase place; Adopt predistorter as shown in Figure 3 to handle, detailed process is: at first according to the amplitude of input signal x (n), find look-up table LUT1 under the corresponding input signal amplitude and value LTU1 and the LTU2 of look-up table LUT2 in the final pre-distortion look-up table of step 4 gained; Calculate x then respectively _I(n) LTU1, x _I(n) LTU2, x _Q(n) LTU1 and x _Q(n) LTU2; Calculate (x again _I(n) LTU1-x _QLTU2) and (x (n) _Q(n) LTU1+x _I(n) LTU2), with (x _I(n) LTU1-x _Q(n) LTU2) as the I road output signal pr of predistorter _I(n), with (x _Q(n) LTU1+x _I(n) LTU2) as the Q road output signal pr of predistorter _Q(n).

Step 6: with the I road output signal pr of the predistorter of step 5 gained _I(n) and Q road output signal pr _Q(n) after D/A conversion, filtering and up-conversion, be input to power amplifier, can obtain linearizing output signal from the power amplifier output.

The present invention adopts single channel feedback and non-iterative technique that the power amplification system base-band input signal is carried out digital pre-distortion and handles, the training sequence that the amplitude linearity of two-way quadrature is increased successively is input to power amplifier, then the follow-on look-up table of I road feedback signal sequence construct by power amplifier output; The input signal of whole power amplification system calculates, searches corresponding look-up table value by amplitude earlier, respectively I, Q two-way base-band input signal are done the true processing of corresponding indication again, input signal after the pre-distortion is passed to power amplifier after D/A conversion, filtering and up-conversion, can obtain being with whole power amplification system base-band input signal the output signal of linear amplification.Whole pre-distortion process can realize that in digital processing chip (comprising DSP or FPGA) compared with prior art, the present invention need not arc tangent or anticosine computing, need not to carry out coordinate transform, thereby reduced amount of calculation, improved computational efficiency, it realizes that system configuration is more simple.

Description of drawings

Fig. 1 indirect learning structural principle block diagram.

Fig. 2 is a distortion linearizing method theory diagram of the present invention.

Fig. 3 is the structural principle block diagram of look-up table predistorter of the present invention

Embodiment

Describe the baseband pre-distortion power amplifier linearization method based on single channel feedback and non-iterative technique of the present invention in detail below in conjunction with accompanying drawing.

A kind of baseband pre-distortion power amplifier linearization method based on single channel feedback and non-iterative technique as shown in Figure 2, may further comprise the steps:

Step 1: one group of amplitude linearity increased and phase place is 0 training sequence

I=1 wherein, 2 ..., i ... N is input to power amplifier as input signal after D/A conversion, filtering and up-conversion; Gather I road feedback signal then, be about to the power amplifier output signal through coupling, down-conversion, filtering and A/D conversion; If the amplitude of the maximum input signal that power amplifier allows is a normalization level 1, then the input signal of i moment power amplifier is

I moment I road feedback signal is $I_1\left(i\right)=A\left(\frac{i}{N}\right)\cos \left(\mathrm{\Φ}\left(\frac{i}{N}\right)\right),$ Wherein

Amplitude function,

Be the power amplifier additive phase; Then corresponding to training sequence

Feedback sequence be { I ₁(i) }.Training sequence

The available sequences generator produces.

Step 2: with another group amplitude linearity growth and phase place is the training sequences of 90 degree I=1 wherein, 2 ..., i ... N is input to power amplifier as input signal after D/A conversion, filtering and up-conversion; Gather I road feedback signal then, be about to the power amplifier output signal through coupling, down-conversion, filtering and A/D conversion; If the amplitude of the maximum input signal that power amplifier allows is a normalization level 1, then the input signal of i moment power amplifier is I moment I road feedback signal is $I_2\left(i\right)=A\left(\frac{i}{N}\right)\cos (\frac{\mathrm{\π}}{2}+\mathrm{\Φ}\left(\frac{i}{N}\right))=-A\left(\frac{i}{N}\right)\sin \left(\mathrm{\Φ}\left(\frac{i}{N}\right)\right);$ Then corresponding to training sequence

Feedback sequence be { I ₂(i) }.Training sequence

The available sequences generator produces.

Step 3: make up initial pre-distortion look-up table.

The pre-distortion look-up table comprises index entry and searches list item and form that wherein index entry is input signal amplitude r _i, search list item and comprise look-up table LUT1 and look-up table LUT2.

Wherein, input signal amplitude $r_i=\sqrt{I_1{\left(i\right)}^2+I_2{\left(i\right)}^2},$ Corresponding input signal amplitude r _iThe value of searching of following look-up table LUT1 is iI ₁(i)/r _i ²N, corresponding input signal amplitude r _iThe value of searching of following look-up table LUT2 is iI ₂(i)/r _i ²N.

Concrete initial pre-distortion look-up table is as shown in the table:

Step 4: the initial pre-distortion look-up table to step 3 gained carries out interpolation processing, obtains final pre-distortion look-up table.

The index entry of the initial pre-distortion look-up table of step 3 gained is input signal amplitude sequence r _iThis sequence is not equally spaced, search list item in order to obtain equally spaced other, that can adopt here that two kinds of methods obtain other range value correspondence searches the list item parameter: the one, obtain the value of pairing look-up table LUT1 of other input signal amplitude and look-up table LUT2 with linear interpolation method; Another kind is to utilize the method for curve fit to come match amplitude and phase characteristic curve with multinomial, and then fills the value of corresponding look-up table LUT1 and LUT2 by the calculating back according to the multinomial of match.

Step 5: for input signal x (the n)=re of whole power amplification system ^{J θ}=x _I(n)+jx _Q(n), n representation signal input time wherein, r is an input signal amplitude, θ is the input signal phase place; Adopt predistorter as shown in Figure 3 to handle, detailed process is: at first according to the amplitude of input signal x (n), find look-up table LUT1 under the corresponding input signal amplitude and value LTU1 and the LTU2 of look-up table LUT2 in the final pre-distortion look-up table of step 4 gained; Calculate x then respectively _I(n) LTU1, x _I(n) LTU2, x _Q(n) LTU1 and x _Q(n) LTU2; Calculate (x again _I(n) LTU1-x _QLTU2) and (x (n) _Q(n) LTU1+x _I(n) LTU2), with (x _I(n) LTU1-x _Q(n) LTU2) as the I road output signal pr of predistorter _I(n), with (x _Q(n) LTU1+x _I(n) LTU2) as the Q road output signal pr of predistorter _Q(n).

Step 6: with the I road output signal pr of the predistorter of step 5 gained _I(n) and Q road output signal pr _Q(n) after D/A conversion, filtering and up-conversion, be input to power amplifier, can obtain linearizing output signal from the power amplifier output.

Below sketch pre-distortion principle of the present invention:

If whole power amplification system input signal is x (n)=re ^{J θ}=x _I(n)+jx _Q(n), then through becoming after the predistorter processing:

$\mathrm{pr}\left(n\right)=A\left(r\right)e^{j(\mathrm{\θ}+\mathrm{\Φ}\left(r\right))}$

$=x\left(n\right)\frac{A\left(r\right)}{r}{e}^{\mathrm{j\Φ}\left(r\right)}$

$=[x_I\left(n\right)+{\mathrm{jx}}_Q\left(n\right)][\frac{A\left(r\right)}{r}\cos \left(\mathrm{\Φ}\left(r\right)\right)+j\frac{A\left(r\right)}{r}\sin \left(\mathrm{\Φ}\left(r\right)\right)]---\left(6\right)$

$=[x_I\left(n\right)\frac{A\left(r\right)}{r}\cos \left(\mathrm{\Φ}\left(r\right)\right)-x_Q\left(n\right)\frac{A\left(r\right)}{r}\sin \left(\mathrm{\Φ}\left(r\right)\right)]$

$+j[x_Q\left(n\right)\frac{A\left(r\right)}{r}\cos \left(\mathrm{\Φ}\left(r\right)\right)+x_I\left(n\right)\frac{A\left(r\right)}{r}\sin \left(\mathrm{\Φ}\left(r\right)\right)]$

Store respectively with two look-up table LUT1 and LUT2 according to following formula With

Value, then according to the amplitude of input signal, in look-up table LUT1 and LUT2, index corresponding value, carry out multiplication and add operation then and just can directly obtain IQ two-way output signal.

By step 1,2, have:

$r_i=\frac{i}{N}---\left(7\right)$

$I_1\left(i\right)=A\left(\frac{i}{N}\right)\cos \left(\mathrm{\Φ}\left(\frac{i}{N}\right)\right)---\left(8\right)$

$I_2\left(i\right)=A\left(\frac{i}{N}\right)\cos (\frac{\mathrm{\π}}{2}+\mathrm{\Φ}\left(\frac{i}{N}\right))=-A\left(\frac{i}{N}\right)\sin \left(\mathrm{\Φ}\left(\frac{i}{N}\right)\right)---\left(9\right)$

In conjunction with (6), (7), (8) and (9) formula, can get:

Corresponding to amplitude is r _iInput signal, the value among look-up table LUT1 and the LUT2 is

${\mathrm{LUT}}_1\left(r_i\right)=\frac{A\left(r_i\right)}{r_i}\cos \left(\mathrm{\Φ}\left(r_i\right)\right)=\frac{{\mathrm{iI}}_1\left(i\right)}{{r_i}^{2}N},(i=\mathrm{1,2},...,N)$

${\mathrm{LUT}}_2\left(r_i\right)=\frac{A\left(r_i\right)}{r_i}\sin \left(\mathrm{\Φ}\left(r_i\right)\right)=\frac{{\mathrm{iI}}_2\left(i\right)}{{r_i}^{2}N},(i=\mathrm{1,2},...,N)$

Claims (4)

1. baseband pre-distortion power amplifier linearization method based on single channel feedback and non-iterative technique may further comprise the steps:

Step 1: one group of amplitude linearity increased and phase place is 0 training sequence

I=1 wherein, 2 ..., i ... N is input to power amplifier as input signal after D/A conversion, filtering and up-conversion; Gather I road feedback signal then, be about to the power amplifier output signal through coupling, down-conversion, filtering and A/D conversion; If the amplitude of the maximum input signal that power amplifier allows is a normalization level 1, then the input signal of i moment power amplifier is

I moment I road feedback signal is I ₁(i); Then corresponding to training sequence Feedback sequence be { I ₁(i) };

Step 2: with another group amplitude linearity growth and phase place is the training sequences of 90 degree

I=1 wherein, 2 ..., i ... N is input to power amplifier as input signal after D/A conversion, filtering and up-conversion; Gather I road feedback signal then, be about to the power amplifier output signal through coupling, down-conversion, filtering and A/D conversion; If the amplitude of the maximum input signal that power amplifier allows is a normalization level 1, then the input signal of i moment power amplifier is

I moment I road feedback signal is I ₂(i); Then corresponding to training sequence

Feedback sequence be { I ₂(i) };

Step 3: make up initial pre-distortion look-up table;

The pre-distortion look-up table comprises index entry and searches list item and form that wherein index entry is input signal amplitude r _i, search list item and comprise look-up table LUT1 and look-up table LUT2;

Wherein, input signal amplitude

Corresponding input signal amplitude r _iThe value of searching of following look-up table LUT1 is Corresponding input signal amplitude r _iThe value of searching of following look-up table LUT2 is

Concrete initial pre-distortion look-up table is:

Step 4: the initial pre-distortion look-up table to step 3 gained carries out interpolation processing, obtains final pre-distortion look-up table;

Step 5: for input signal x (the n)=re of whole power amplification system ^{J θ}=x _I(n)+jx _Q(n), n representation signal input time wherein, r is an input signal amplitude, θ is the input signal phase place, x _I(n) be the in-phase component of x (n), x _Q(n) be the quadrature component of x (n); Adopt predistorter to handle, detailed process is: at first according to the amplitude of input signal x (n), find look-up table LUT1 under the corresponding input signal amplitude and value LTU1 and the LTU2 of look-up table LUT2 in the final pre-distortion look-up table of step 4 gained; Calculate x then respectively _I(n) LTU1, x _I(n) LTU2, x _Q(n) LTU1 and x _Q(n) LTU2; Calculate (x again _I(n) LTU1-x _QLTU2) and (x (n) _Q(n) LTU1+x _I(n) LTU2), with (x _I(n) LTU1-x _Q(n) LTU2) as the I road output signal pr of predistorter _I(n), with (x _Q(n) LTU1+x _I(n) LTU2) as the Q road output signal pr of predistorter _Q(n);

Step 6: with the I road output signal pr of the predistorter of step 5 gained _I(n) and Q road output signal pr _Q(n) after D/A conversion, filtering and up-conversion, be input to power amplifier, can obtain linearizing output signal from the power amplifier output;

Step 1 is to step 6, and described N is a natural number.

2. the baseband pre-distortion power amplifier linearization method based on single channel feedback and non-iterative technique according to claim 1 is characterized in that the training sequence in step 1 and the step 2 is produced by sequence generator.

3. the baseband pre-distortion power amplifier linearization method based on single channel feedback and non-iterative technique according to claim 1, it is characterized in that, when the initial pre-distortion look-up table of step 4 pair step 3 gained carries out interpolation processing, adopt linear interpolation method to obtain the value of pairing look-up table LUT1 of other input signal amplitude and look-up table LUT2.

4. the baseband pre-distortion power amplifier linearization method based on single channel feedback and non-iterative technique according to claim 1, it is characterized in that, when the initial pre-distortion look-up table of step 4 pair step 3 gained carries out interpolation processing, adopt the method for curve fit to come match amplitude and phase characteristic curve, and then fill the value of corresponding look-up table LUT1 and LUT2 according to the multinomial of match by the calculating back with multinomial.

Images