CN103036514B - Utilize the method for Volterra correction model rated output amplifier output variable - Google Patents

Abstract

The invention discloses the method utilizing Volterra correction model rated output amplifier output variable, belong to the technical field of Digital Signal Processing.The present invention revises constraints by setting; Measure power amplifier performance parameter to determine to revise constraints expression formula; Utilize correction constraint expression formula corrected output amplifier to connect and input data, obtain revised Volterra model; According to the output variable that power amplifier exports discrete data, revised Volterra model calculates power amplifier.Utilize rated output amplifier output variable of the present invention, both simplified the number of coefficients of Volterra model, turn improve the precision utilizing Volterra model rated output amplifier output variable.

Description

Utilize the method for Volterra correction model rated output amplifier output variable

Technical field

The invention discloses the method utilizing Volterra correction model rated output amplifier output variable, belong to the technical field of Digital Signal Processing.

Background technology

It is common method that rated output amplifier output variable evaluates its performance.Input according to power amplifier calculates output variable one by one, and workload is large, and the output data obtained are imperfect.Discrete input variable, the discrete output variable of scholars' method processing power amplifier of research modeling gradually, thus calculate all output variables of power method device.Evaluate the performance of power amplifier by analyzing output variable, and analyze its predistortion in wide-band communication system and apply.

The method of existing following several Modeling Calculation power amplifier: existing modeling method and respective Problems existing please be sketch.Set up the performance of accurate power amplifier model to research and predicted power amplifier and have great meaning.According to the difference of modeling method, power amplifier model can be divided into two classes: for the physical model (device model) of circuit-level simulation analysis and the behavior model (black-box model) for system integration project analysis.Consider that physical model is expressed as based on the physical circuit of power amplifier and the non-linear mechanism of production of component models the output that equivalent electric circuit carrys out analog amplifier, cannot express accurately with mathematical way, and behavior model only considers the mathematical relationship between system input/output signal, can well be applied in theory analysis.

The non-linear behavior model of power amplifier can be divided into two kinds: memoryless nonlinear model and memory nonlinear model.Memoryless behavior model is applied to narrow band power amplifier, because giving in fixed temperature and direct current biasing situation, model is the static function of input signal, AM/AM and AM/PM characteristic is indeclinable; And memory nonlinear models applying is in wideband power amplifer, because the heating of the single-frequency of Match circuits, DC bias circuit and transistor all can produce memory effect.

The power amplifier behavior model of current proposition mainly contains: without note multinomial model, Saleh model, Wiener model, Hammerstein model, Volterra model, memory polynomial model and neural network model etc.Except neural network model, other behavior models are all based on the improvement on Volterra model basis and correction model.Although neural net is the effective ways of nonlinear dynamic system modeling, but most of neural network model is all the structure adopting multilayer perceptron, difficulty extracts model parameter, and the sandwich construction of its complexity constrains realization and the application of predistortion to a certain extent.

The memory nonlinear model that Volterra progression Chang Zuowei is general, and used by many researchers with the relation describing power amplifier input and output.But the Volterra model of classics is owing to comprising all non-linear and Memorability items, and its core coefficient exponentially increases, therefore more difficult in practice.In recent years, a lot of scholar proposes a lot based on the simplified model of traditional Volterra model, but the number of coefficients of model still sharply can increase along with the increase of the non-linear of system and memory depth, make the system that these model applicable band are narrower, the requirement of modern broadband communication system cannot be met.

Summary of the invention

Technical problem to be solved by this invention is the deficiency for above-mentioned background technology, provides the method utilizing Volterra model rated output amplifier output variable.

The present invention adopts following technical scheme for achieving the above object:

Utilize the method for Volterra correction model rated output amplifier output variable, comprise the steps:

Step 1, sets the correction constraints of each input item signal: α ∑ m+ β * p≤-ln ε,

Wherein: α, β and ε are correction factor, p is the non linear coefficient of each input item signal, ∑ m be each memory depth in every input item signal and;

Step 2, measures the input power P of power amplifier _in, gain G, q rank intermodulation component I _q, calculate correction factor α, β and ε, q be greater than 1 odd number, the correction constraints expression formula described in determining step 1:

Wherein: $\mathrm{\ϵ}=\frac{\sqrt{P_{\mathrm{in}}\×{10}^{-I_q/20}}}{G},$

$\mathrm{\β}=-\frac{1}{5}\ln \frac{\sqrt{P_{\mathrm{in}}\×{10}^{-I_q/20}}}{G},$

α＝2；

Step 3, the input discrete data measuring power amplifier is x (n), and correction conditions expression formula correction Volterra mode input discrete data x (n) utilizing step 2 to determine obtains new input item vector x _p,m' (n):

Step a, according to the correction conditions that step 2 is determined, determines the valued combinations of non-linearity of power amplifier coefficient p, memory depth and ∑ m;

Step b is choose the input item signal of the valued combinations meeting non-linearity of power amplifier coefficient p, memory depth and ∑ m x (n) from the input discrete data of power amplifier;

Step c, with the input item signal meeting non-linearity of power amplifier coefficient p, memory depth and ∑ m valued combinations chosen in step b for vector element, builds new input item vector x _p,m' (n);

Step 4, the output discrete data measuring power amplifier is y (n), according to the new input item obtained in step 3, determines the output y'(n of the power amplifier that Volterra correction model calculates) be:

y'(n)＝x _p,m'(n){([x _p,m'(n)] ^Tx _p,m'(n)) ^-1[x _p,m'(n)] ^Ty(n)}。

The present invention adopts technique scheme, has following beneficial effect: the number of coefficients of having simplified Volterra model, improves the precision utilizing Volterra model rated output amplifier output variable.

Accompanying drawing explanation

Fig. 1 is the simulate effect power spectrum function figure of Doherty power amplifier.

Fig. 2 is flow chart of the present invention.

Embodiment

Be described in detail below in conjunction with the technical scheme of accompanying drawing to invention:

Utilize Volterra correction model to calculate the method for Doherty power amplifier output variable, as shown in Fig. 2 flow chart, comprise the steps:

Step 1, sets the correction constraints of each input item signal: α ∑ m+ β * p≤-ln ε,

Wherein: α, β and ε are correction factor, p is the non linear coefficient of each input item signal, ∑ m be all input item signals memory depth and.

Step 2, measures the input power P of power amplifier _in=10W, gain G=20dB, 5 (q=5) rank intermodulation component I ₅=-40dBc, calculates correction factor α, β and ε, the correction constraints expression formula described in determining step 1:

$\mathrm{\ϵ}=\frac{\sqrt{P_{\mathrm{in}}\×{10}^{-I_5/20}}}{G}=0.0158,$

$\mathrm{\β}=-\frac{1}{5}\ln \frac{\sqrt{P_{\mathrm{in}}\×{10}^{-I_5/20}}}{G}=0.8,$

α＝2；

Revising constraints expression formula is: 2 ∑ m+0.8p≤4.2.

The intermodulation component of odd-order is comparatively large to the interference of PA signal, in general the having the greatest impact of 3 rank power amplifier precision, and for bandwidth power amplifier, the intermodulation component that exponent number is high is large to Accuracy.In the present embodiment, compromise gets 5 rank intermodulation components after considering.

Step 3, measuring the input discrete data of power amplifier is x (n) ofdm signal of 20MHz (x (n) to be bandwidth be), the correction conditions expression formula correction Volterra mode input item x utilizing step 2 to determine _p,mn () obtains new input item vector x _p,m' (n):

$x_{p,m}\left(n\right)={\mathrm{\Σ}}_{p=1}^P{\mathrm{\Σ}}_{m_1=0}^M...{\mathrm{\Σ}}_{m_p=0}^M{\mathrm{\Π}}_{i=1}^{p}x(n-m_i)$

P is the nonlinearity of power amplifier, arranges different model orders can impact the precision of model amplifier is analog, and value needs to determine according to the performance of concrete amplifier.In the present embodiment, P=5; M is the memory depth of power amplifier, M=5, new input item x _p,m' (n) as shown in table 1.

Step 3 specifically comprises the steps:

Step a, according to the correction conditions that step 2 is determined, determines the valued combinations of non-linearity of power amplifier coefficient p, memory depth and m: during p=1, m=0,1,2; During p=2, m=0,1; During p=3, m=0,1; During p=4, m=0; During p=5, m=0;

Step b is choose the input item signal of the valued combinations meeting non-linearity of power amplifier coefficient p, memory depth and m x (n) from the input discrete data of power amplifier;

Step c, with the input item signal meeting non-linearity of power amplifier coefficient p, memory depth and m valued combinations chosen in step b for vector element, builds new new input item vector x _p,m' (n):

x _p,m'(n)＝[x(n),x(n-1),x(n-2)，x(n) ²,x(n)x(n-1)，x(n) ³,x(n) ²x(n-1)，x(n) ⁴，x(n) ⁵]。

The input item x that table 1 is new _p,m' (n)

Step 4, the output discrete data measuring power amplifier is y (n), according to the new input item obtained in step 3, determines the output y'(n of the power amplifier that Volterra correction model calculates) be:

y'(n)＝x _p,m'(n){([x _p,m'(n)] ^Tx _p,m'(n)) ^-1[x _p,m'(n)] ^Ty(n)}。

Wherein: [x _p,m' (n)] ^trepresent new input item vector x _p,m' (n) do matrix transpose computing, ([x _p,m' (n)] ^tx _p,m' (n)) ^-1represent ([x _p,m' (n)] ^tx _p,m' (n)) do inverse matrix operation.

Can find through correction input item number be far smaller than without revise input item, Fig. 1 gives the simulate effect power spectrum function figure of Doherty power amplifier and gives model error, can find the number of coefficients greatly reducing model while this analogy method can keep the precision of model.

The present invention has simplified the number of coefficients of Volterra model in sum, improves the precision utilizing Volterra model rated output amplifier output variable.Above-described embodiment is an embody rule of the present invention, and for the power amplifier of other type, the method for the invention is applicable equally.Every meet present inventive concept any equivalent or equivalents all within protection scope of the present invention.

Claims (1)

1. utilize the method for Volterra correction model rated output amplifier output variable, it is characterized in that comprising the steps:

Step 1, sets the correction constraints of each input item signal: α ∑ m+ β * p≤-ln ε,

Wherein: α, β and ε are correction factor, p is the non linear coefficient of each input item signal, ∑ m be each memory depth in every input item signal and;

Step 2, measures the input power P of power amplifier _in, gain G, q rank intermodulation component I _q, calculate correction factor α, β and ε, q be greater than 1 odd number, the correction constraints expression formula described in determining step 1:

Wherein: $\mathrm{\ϵ}=\frac{\sqrt{P_{\mathrm{in}}\×{10}^{{-I}_q/20}}}{G},$

$\mathrm{\β}=-\frac{1}{5}\ln \frac{\sqrt{P_{\mathrm{in}}\×{10}^{{-I}_q/20}}}{G},$

α＝2；

Step 3, the input discrete data measuring power amplifier is x (n), and correction conditions expression formula correction Volterra mode input discrete data x (n) utilizing step 2 to determine obtains new input item vector x _p,m' (n):

Step a, according to the correction conditions that step 2 is determined, determines the valued combinations of non-linearity of power amplifier coefficient p, memory depth and ∑ m;

Step b is choose the input item signal of the valued combinations meeting non-linearity of power amplifier coefficient p, memory depth and ∑ m x (n) from the input discrete data of power amplifier;

Step c, with the input item signal meeting non-linearity of power amplifier coefficient p, memory depth and ∑ m valued combinations chosen in step b for vector element, builds new input item vector x _p,m' (n);

Step 4, the output discrete data measuring power amplifier is y (n), according to the new input item obtained in step 3, determines that output y ' (n) of the power amplifier that Volterra correction model calculates is:

$y^{\′}\left(n\right)={x_{p,m}}^{\′}\left(n\right)\left\{{\left({\left[{x_{p,m}}^{\′}\left(n\right)\right]}^T{x_{p,m}}^{\′}\left(n\right)\right)}^{-1}{\left[{x_{p,m}}^{\′}\left(n\right)\right]}^{T}y\left(n\right)\right\}.$