Adaptive pre-distortion method based on two-dimensional polling list in OFDM system

Technical Field

The invention relates to wireless communication Technical Field, in particular to an OFDM system based on two-dimensional lookup table in the adaptive predistortion method.

Background Art

Because the OFDM system memory of the plurality of sub-channel signal is superimposed, in the transmitter of the system therefore, synthetic signal will be large instantaneous peak, thus generating excessive peak-to-average power ratio. These instantaneous peak through the transmission system of the limited linear dynamic range of the amplifier, will produce distortion.

Amplifier distortion is divided into two kinds: linear distortion and non-linear distortion. Usually the input signal is of the amplifying circuit of the multi-frequency signal, if the amplifying circuit to the signal of different frequency components have different relative phase shift amplitude or gain change, the output waveform distortion takes place, the former is referred to as amplitude distortion, the latter referred to as a phase distortion, both collectively referred to as the frequency distortion. By the circuit due to frequency distortion is caused by the linear reactance element, and said linear distortion, is characterized in that the output signal does not produce the input signal in is not in the new frequency component. The non-linear distortion is composed of an amplifying device or load the non-linear current-voltage characteristic of the waveform distortion caused by the non-linearity, which is characterized in that the frequency components of the generated new, in order to input the signal is generated by a single-frequency components into a high-order harmonic components of the fundamental wave component. This invention is directed primarily to the amplifier nonlinear distortion correction for in-depth study.

At present there are many kinds of amplifier linearization method, often used in the method is that the power is withdrawn, negative feedback law , forward feed law , such as and LINC predistortion law. However, no matter what kind of method, the purpose of the utility model are in order to maintain a relatively high emission efficiency at the same time, obtain better linear input-output characteristic. Linearization technique in the development of a very important step is the emergence of predistortion technology. Predistortion technology to make the power amplifier linearity is improved greatly, increases the output power, suitable for broadband application, the stability of the system is good, so a more extensive application at present. The most important thing is that the pre-distortion method to realize the upper is relatively simple, is also relatively inexpensive price, following the increase of the economic benefit. Therefore, the pre-distortion technique is a very promising power amplifier linearization technique.

However, most of the pre-distortion method for the current are directed to non-memory power amplifier and proposed, to the memory the influence caused by the power amplifier have not been fully taken into account. In fact, the OFDM system on the existing increasingly popular many of the memoryless pre-distortion method provides new problems and challenges. For many power amplifier, OFDM is a broadband signal, that is to say, the amplitude-frequency characteristic of the power amplifier in the bandwidth of the OFDM signal in the range of is not completely flat, not only the magnification of the current input value, the history of the input values before a certain extent, this is the memory characteristics of the power amplifier. When the OFDM signal frequency band is of a width sufficient, with the historical value of the current value of the degree of coupling is large enough, this memory characteristic of the impact of the output signal cannot be easily overlooked. Therefore, adequate analysis of power amplifier non-linearity and Memonic the OFDM dual impact the broadband signal, digital predistortion has already become a new one in the field of research direction.

Although the current International the power amplifier predistortion technology has launched a certain research, but internally in this regard the relatively initial stage of the study, the current of the power amplifier with the memory of most predistortion scheme is based on a polynomial law is, the is very difficult to employ these programmes and to the actual project. Consider now the power amplifier memory effect of a lookup table pre-distortion technique is mainly a two categories: one is still using unidimensional table , the index way of a table address, however, include the memory effect of the power amplifier, such as PMC and Altera table address generation scheme. There is another type of multi-dimensional table is a, by increasing the dimensionality of the lookup table used to characterize the memory effect of the amplifier. Zhi-yong He (Terry   Hoh),   Jian-hua Ge, Geng   shu-jian,   Gang andWang, "An   Improved   Look-Up   Table   Predistortion   Technique   forHPA   with     Systems   Effects   Memory   in   OFDM", IEEE   NG   BROADCASTI   TRANSACTIONSON, VOL. 52, NO.1 proposed for the first time in the method of using the two-dimensional lookup table to the memory effect of the power amplifier can be corrected, because of increasing the two-dimensional index factor Y, so can effectively distinguish the different historical state information, so as to overcome the memory effect of the power amplifier. The proposal is a simple efficient pre-distortion method, but the performance is yet to be improved.

Figure 1 provides a two-dimensional lookup table the universal method for adaptive predistortion, input digital predistortion module in the OFDM digital baseband complex signal x_in, obtained after the transformation R/P ρexp (jθ). Ρ as a 1st after quantizing uygur rope pin value X; ρ times the value of the input signal history ρ_his 2nd joint calculated uygur rope pin value Y (Y have different definition mode). The index value of about (X, Y) at this moment can be obtained two-dimensional lookup table x_in of the multiplication correction factor corresponding to the β, after the output of the multiplier to obtain the pre-distorter x_pd = x_in *β.

Content of the invention

The aim of the invention is to overcome the deficiency of the prior art, provides a OFDM system based on two-dimensional lookup table in the adaptive predistortion method, so that the complexity of the algorithm is not increased effectively at the same time of a lookup table to accelerate the speed of convergence, the smaller the mean square error, at the same time reduces the complexity of hardware implementation of algorithm.

The purpose of this invention is realized through the following technical scheme:

A OFDM system based on two-dimensional lookup table in the adaptive predistortion method, which is characterized by comprising the following steps:

(1) 1st lookup table according to OFDM source signal to the predistortion coefficient v_i (n) predistortion by the predistortion signal v_d (n), the process of pre-distortion processing among them are as follows:

A) input OFDM source signal v_i (n) = ρ_in (n) *exp (jθ_in (n)), unidimensional 1st a lookup table to obtain the index address of the X (n) and two-dimensional index address Y (n), wherein ρ_in (n) the source signal for the amplitude of the input OFDM, θ_in (n) the input OFDM phase of source signal;

B) according to X (n) and Y (n) joint for indexing the lookup table 1st 1st web-hour meter to obtain amplitude predistortion coefficient α (n), searching the lookup table 1st 1st phase table to obtain phase predistortion coefficients σ (n), the calculated pre-distortion coefficient β (n) = α (n) exp (jρ (n));

C) under v_i (n) and β (n) calculated by the signal after the pre-distortion processing v_d (n):

v_d (n) = v_i (n)* β (n) = ρ_in (n)* α (n) *exp (j (θ_in (n) + σ (n)));

(2) pre-distortion signal v_d (n) through the digital-to-analog conversion processing sequentially, up-conversion processing, power amplifying processing, down-conversion processing and analog-to-digital conversion processing to obtain the signal   The lookup table according to the 2nd to the signal predistortion coefficients   Predistortion processing, to obtain the pre-distortion signal   The process of pre-distortion processing among them are as follows:

(D) input signal

v_i '(n) = ρ_in (n)* α (n)* G_ρ (ρ_in (n)* α (n)) *exp (j (θ_in (n) + σ (n) + F_θ (ρ_in (n)* α (n)))), unidimensional index a lookup table to obtain the address of the X' (n) and two-dimensional index address Y '(n), wherein G_ρ (ρ_in (n)* α (n)) AM/AM the distortion of the power amplifier, F_θ (ρ_in (n)* α (n)) AM/PM the distortion of the power amplifier, ρ_in (n)* α (n)* G_ρ (ρ_in (n)* α (n)) the input signal v_i' (n) amplitude, θ_in (n) + σ (n) + F_θ (ρ_in (n)* α (n)) the input signal v_i ' (n) phase;

(E) according to X '(n) and Y' (n) joint index search 2nd 2nd site in the look-up table to obtain amplitude predistortion coefficients α '(n), searching phase 2nd 2nd in the look-up table to obtain a σ phase predistortion coefficients' (n), the calculated pre-distortion coefficient β '(n) =α' (n) exp (jσ ' (n));

(F) under v_i '(n) and β' (n) is the signal after the pre-distortion processing  

v^d(n)=vi′(n)*β′(n)=ρin(n)*α(n)*α′(n)*Gρ(ρin(n)*α(n))*;

exp (j (θ_in (n) + σ (n) +σ ' (n) + F_θ (ρ_in (n)* α (n))))

(3) the step (1) of the predistortion signal in v_d (n) and step (2) of the predistortion signal in   The error calculation, to obtain an error signal e (n);

(4) according to the error signal e (n) to a self-adaptive iterative operation, updating step (2) of the lookup table in the predistortion coefficient 2nd, so that the error signal e (n) to approach to 0;

Wherein e (n) = e_ρ (n) *exp (je_θ (n)), e_ρ (n) to v_d (n) and   The amplitude error, e_θ (n) to v_d (n) and   The phase error, said n sampling points, is a positive integer.

In the above-mentioned OFDM system based on two-dimensional lookup table in the adaptive predistortion method, step (1) 1st uygur index in address X (n) by ρ_in (n) quantization obtained through the address.

In the above-mentioned OFDM system based on two-dimensional lookup table in the adaptive predistortion method, steps (1) to calculate two-dimensional index address Y (n) of the 1st kind of method is as follows:

Y (n) =Int {[ ρ_in (n-Q) + ρ_in (n-Q + 1) +... + ρ_in (n-1)]/ A}

Wherein the function of the representative Int (·) is not greater than the maximum integer value; Q is a memory depth, the value is a positive integer; ρ_in (n) the input signal v_i (n) amplitude; A is a constant, the value of the section number directs the valueuygur rope 2nd; said n sampling points, is a positive integer.

In the above-mentioned OFDM system based on two-dimensional lookup table in the adaptive predistortion method, steps (1) to calculate two-dimensional index address Y (n) of the 2nd kind of method is as follows:

The input weighted average amplitude is:

ρ_wei (n) = λ_Q ρ_in (n-Q) + λ_Q-1 ρ_in (n-Q + 1) +L+λ₂ ρ_in (n-2) + λ₁ ρ_in (n-1)

Wherein λ_i = D^1-i; D exponential forgetting factor of the networks, D> 1, i=1, 2, L, Q -1, Q is a memory depth, the value is a positive integer,

if ρwei(n)ρin(n)<Umax,Y(n)=Int(ρwei(n)ρin(n)Umax*LY);

If   Y (n) = L_Y -1

wherein, U_max to   The upper limit value of the ideal, L_Yuygur ropedirects the value interval of the number of 2nd.

In the above-mentioned OFDM system based on two-dimensional lookup table in the adaptive predistortion method, steps (2) 1st uygur index in address X ' (n) by ρ_in (n)* α (n)* G_ρ (ρ_in (n)* α (n)) quantization obtained through the address.

In the above-mentioned OFDM system based on two-dimensional lookup table in the adaptive predistortion method, steps (2) to calculate two-dimensional index address Y ' (n) of the 1st kind of method is as follows:

Y '(n) =Int {[ ρ'_in (n-Q)+ρ '_in (n-Q + 1) +…+ρ'_in (n-1)]/ A}

Wherein the function is not greater than value Int (·) is the maximum integer; Q is a memory depth, the value is a positive integer; ρ '_in (n) = ρ_in (n)* α (n)* G_ρ (ρ_in (n)* α (n)) the input signal   The amplitude; A is a constant, the value of the section number directs the valueuygur rope 2nd; said n sampling points, is a positive integer.

In the above-mentioned OFDM system based on two-dimensional lookup table in the adaptive predistortion method, steps (2) to calculate two-dimensional index address Y ' (n) of the 2nd kind of method is as follows:

The input weighted average amplitude is:

[0043]   ρ '_wei (n) = λ_Q ρ'_in (n-Q) + λ_Q-1 ρ '_in (n-Q + 1) +L+λ₂ ρ'_in (n-2) + λ₁ ρ '_in (n-1)

[0044]   wherein λ_i = D^1-i, D exponential forgetting factor of the networks, D> 1, i=1, 2, L, Q -1, Q is a memory depth, the value is a positive integer,

if ρ′wei(n)ρ′in(n)<Umax,Y′(n)=Int(ρ′wei(n)ρ′in(n)Umax*LY);

If   Y ' (n) = L_Y -1

wherein, U_max to   The upper limit value of the ideal, L_Yuygur ropedirects the value interval of the number of 2nd, ρ '_in (n) = ρ_in (n)* α (n)* G_ρ (ρ_in (n)* α (n)) the input signal   The amplitude.

In the above-mentioned OFDM system based on two-dimensional lookup table in the adaptive predistortion method, step (4) self-adaptive iterative operation in the iterative equation as:

αi+1′(n)=αi′(n)+eρ(n)*μρ

σi+1′(n)=σi′(n)+e[!theta!](n)*μ[!theta!]

Wherein   v_d (n) states the step to the (1) in the pre-distortion processing of the signal,   states the step to the (2) in the signal after the pre-distortion processing, function to arg(.)the phase of the transducer, μ_ρ is amplitude of the step of iteration, μ_θ is the phase of the step of iteration.

With the prior art, the invention has the following advantages:

(1) the iterative algorithm RASCAL, compared with the traditional LMS, RLS adaptive algorithm such as, parameter less, and has a good convergence characteristic.

(2) by adopting a two-dimensional lookup table approach to characterization of the memory effect of the power amplifier, wherein the input OFDM uygur rope 1st directs the value strayly of source signal, 2nd uygur ropedirects the value in order to utilize the index forgetting factor for different historical input amplitude of the value of the weighted sum, thereby characterize the current input signal of the input signal with the amplitude value of the history of the coupling situation strayly, effectively overcome the memory effect of the power amplifier.

(3) through the use of the amplitude of the input signal by way of the index, has the advantages of simple realization, less dependent on factors, with better performance characteristics.

The technical scheme of the invention greatly reduces the complexity of hardware implementation of algorithm, to achieve suitable for the project.

Figure illustrates

Figure 1 is a two-dimensional lookup table universal method for adaptive predistortion block diagram structure;

Figure 2 in the OFDM system of the present invention non-direct learning structure chart adaptive pre-distorter;

Figure 3 as in the present invention 1st the predistortion unit predistortion work flow chart;

Figure 4 is the invention in pre-distortion unit 2nd predistortion work flow chart;

Figure 5 and in the pre-distortion device to work the power amplifier output a demodulated constellation;

Figure 6 for various circumstances power amplifier output power spectral density graph;

Figure 7 in this invention two-dimensional index value for Y distribution histogram;

Figure 8 as when the 2nd out uygur size 32 and 512 of the power amplifier when the output power spectral density diagram.

Mode of execution

Mode of execution under the combination of the present invention and for further detailed description.

Figure 2 is the invention of the non-direct study structure self-adaptive predistortion system chart, the system will be a polynomial predistortion method in the structure of the lookup table of the non-direct learning adaptive pre-distortion method for combination, the invention pre-distorter including the predistortion unit 1st, 2nd pre-distortion unit two predistortion lookup table unit, error signal calculation unit and a self-adapting algorithm unit.

Wherein the predistortion unit includes 1st the 1st memory predistortion coefficient lookup table, a lookup table according to the 1st to the input of the source signal OFDM v_i (n) predistortion processing, and the signal of the predistortion processing after v_d (n) are respectively delivered to the error signal calculation unit and an external of the digital-to-analog converter DAC unit; 1st 2nd pre-distortion unit receives the predistortion unit the signal after the pre-distortion processing v_d (n) sequentially through an external digital-to-analog converter DAC unit, frequency conversion unit, power amplifier, lower frequency conversion unit and an analog-to-digital converter ADC unit processed signal v_i ' (n), of predistortion coefficients based on the stored lookup table 2nd predistortion processing, after processing and the signal of the predistortion   Conveying to the error signal calculating unit; error signal calculation unit receives the predistortion unit 1st the signal after the pre-distortion processing v_d (n) and 2nd after pre-distortion unit predistortion processing of the signal   And error calculation, to obtain an error signal e (n); and receiving the error adaptive algorithm unit complex signal e (n) to a self-adaptive iterative operation, update 2nd 2nd pre-distortion unit predistortion coefficient of the lookup table, so that the error signal e (n) to approach to 0, wherein e (n) = e_ρ (n) *exp (je_θ (n)), e_ρ (n) representative v_d (n) and   The amplitude error, e_θ (n) representative v_d (n) and   The phase error, said sampling point n, is a positive integer.

Can be obtained by a simulation comparing the following conclusions: for a typical AB type power amplifier, power index in any case are the worst of a method; in the dynamic power control system, μ-law amplitude dependence of the fallbacks level change the index of the minimum; relative to the two former method, amplitude index generating intermodulation power is relatively small; and the most optimum distribution scheme is superior to the amplitude of the index table entry. However, although the optimum distribution of the best performance of the programme, however, from the standpoint of engineering implementation, more inclined to amplitude index method, because it is simple to realize, less dependent on factors, the performance is the most close to the optimum distribution, so the technical proposal of the present invention using magnitude index method.

Figure 3 as in the present invention 1st the predistortion unit predistortion work flow chart, first of all input OFDM source signal v_i (n) = ρ_in (n) *exp (jθ_in (n)), wherein ρ_in (n) the source signal for the amplitude of the input OFDM, θ_in (n) the input OFDM phase of source signal, according to ρ_in (n) unidimensional quantization obtain the lookup table of the index address X (n), the historical input amplitude value is calculated to obtain 2nd-dimensional index address Y (n), the present invention provides a computing Y (n) two kinds of method:

Method 1:

A most simple method to improve the two-dimensional index value Y calculated. Because of introducing two-dimensional index factor Y is aimed at in order to identify different historical input signal, to eliminate the memory effect of the power amplifier, thus can be used for a simple method to obtain the two-dimensional index value Y:

Y (n) =Int {[ ρ_in (n-Q) + ρ_in (n-Q + 1) +... + ρ_in (n-1)]/ A}     (1)

Wherein the function of the representative Int (·) is not greater than the maximum integer value; Q is a memory depth, the value is a positive integer; ρ_in (n) the input signal v_i (n) amplitude; A is a constant, the value of the section number directs the valueuygur rope 2nd; said n sampling points, is a positive integer. .

Method 2:

The input weighted average amplitude is:

ρ_wei (n) = λ_Q ρ_in (n-Q) + λ_Q-1 ρ_in (n-Q + 1) +…+λ₂ ρ_in (n-2) + λ₁ ρ_in (n-1)     (2)

Wherein λ_i = D^1-i, D> 1 is index forgetting factor of the networks, i=1, 2, the [...] , Q-1, Q is a memory depth. In order to compress the ρ_wei (n) the range of values of the, while at the same time does not affect the description of the nature of the historical state, can be to the current amplitude ρ_in (n) normalized, is   According to   The distribution of the method of the use of the average quantization interval, can be two-dimensional index value Y (n).

if ρwei(n)ρin(n)<Umax,Y(n)=Int(ρwei(n)ρin(n)Umax*LY);---(3)

if ρwei(n)ρin(n)[!GreaterEqual!]Umax,Y (n) = L_Y -1 (4)

wherein, U_max representative   The upper limit value of the ideal, L_Y Y of two-dimensional index value representative of the number of the section. If the memory feature is not very strong power amplifier, can handle type simplified to   Consider only the largest weight of one of the last.

By adopting the above-mentioned calculation method is that the memory on the basis of a plurality of typical HPA coefficient distribution of the model, the more similar the time can be found on the historical value, of the current value of the greater the impact. Therefore, the index to the Y value of the forgetting factor λ be improved, relatively close to the time of use of the right input, the input value of the earlier of the time according to an exponential law to reduce power. This kind of method can be used as the unknown HPA to the actual memory characteristics in a more reasonable estimates.

According to {X (n), Y (n)} watt-hour meter site 1st joint index lookup to obtain amplitude predistortion coefficient α (n), searching 1st phase table to obtain phase predistortion coefficients σ (n), the calculated pre-distortion coefficient β (n) = α (n) exp (jσ (n)); β (n) and v_i (n) the operation by the multiplier by the signal after the pre-distortion processing   v_d (n):

v_d (n) = v_i (n)* β (n) = ρ_in (n)* α (n) *exp (j (θ_in (n) + σ (n))) (5)

Signal v_d (n) through an external digital-to-analog converter DAC unit, frequency conversion unit, power amplifier, lower frequency conversion unit and an analog-to-digital converter receives the signal processing after the ADC unit v_i '(n), the invention temporarily does not consider the signal through the digital-to-analog/analog-to-digital conversion and the impact of the frequency conversion process, considering only the power amplifier of the distortion to the signal, thereby the signal v_d (n) after processing by a power amplifier of the signal v_i' (n) to:

v_i ' (n) = ρ_in (n)* α (n)* G_ρ (ρ_in (n)* α (n)) *exp (j (θ_in (n) + σ (n) + F_θ (ρ_in (n)* α (n))))     (6)

Wherein G_ρ (ρ_in (n)* α (n)) representative AM/AM distortion of the power amplifier, F_θ (ρ_in (n)* α (n)) representative AM/PM distortion of the power amplifier, ρ_in (n) the source signal for the amplitude of the input OFDM, θ_in (n) the input OFDM source signal phase, signal v_i ' (n) input 2nd pre-distortion unit.

As shown in Figure 4 to the invention pre-distortion unit 2nd predistortion work flow chart, first of all input signal v_i '(n) = ρ_in (n)* α (n) *G_ρ (ρ_in (n)* α (n)) *exp (j (θ_in (n) + σ (n) + F_θ (ρ_in (n)* α (n)))), wherein ρ_in (n)* α (n)* G_ρ (ρ_in (n)* α (n)) the input signal v_i' (n) amplitude, θ_in (n) + σ (n) + F_θ (ρ_in (n)* α (n)) the input signal v_i '(n) phase, amplitude value ρ_in (n)* α (n)* G_ρ (ρ_in (n)* α (n)) obtained through the address lookup table of the amount of 2nd unidimensional index address X' (n), the historical input amplitude value is calculated to obtain 2nd-dimensional index address Y '(n), the present invention provides a computing Y' (n) two kinds of method, the calculating process Y (n) calculation:

Method 1:

Y '(n) =Int {[ ρ'_in (n-Q)+ρ '_in (n-Q + 1) +...+ρ'_in (n-1)]/ A}     (7)

Wherein the function is not greater than value representative Int (·) of the large integer, on behalf of the memory depth Q, the value is a positive integer, ρ '_in (n) = ρ_in (n)* α (n)* G_ρ (ρ_in (n)* α (n)) the input signal v_i' (n) amplitude, A is a constant, the value of the number of 2nd directs the valueuygur rope interval, said n sampling points, is a positive integer.

Method 2:

The input weighted average amplitude is:

ρ '_wei (n) = λ_Q ρ'_in (n-Q) + λ_Q-1 ρ '_in (n-Q + 1) +…+λ₂ ρ'_in (n-2) + λ₁ ρ '_in (n-1)     (8)

Wherein λ_i = D^1-i, D> 1 is index forgetting factor of the networks, i=1, 2, the [...] , Q-1, Q is a memory depth. In order to compress the ρ '_wei the range of values of the, while at the same time does not affect the description of the nature of the historical state, can be to the current amplitude ρ_in' (n) normalized, is   According to   The distribution of the method of the use of the average quantization interval, can be two-dimensional index value Y ' (n).

if ρ′wei(n)ρ′in(n)<Umax,Y′(n)=Int(ρ′wei(n)ρ′in(n)Umax*LY);---(9)

if ρ′wei(n)ρ′in(n)[!GreaterEqual!]Umax,Y ' (n) = L_Y -1     (10)

Wherein U_max representative   The upper limit value of the ideal, L_Yuygur ropedirects the value interval of the representative number of 2nd, ρ '_in (n) = ρ_in (n)* α (n) *G_ρ (ρ_in (n)* α (n)) the input signal v_i' (n) amplitude. If the memory feature is not very strong power amplifier, can handle type simplified to   Consider only the largest weight of one of the last.

Predistortion correction coefficient β is calculated '(n) =α' (n) exp (jσ '(n)); β' (n) and v_i ' (n) the operation by the multiplier by the signal after the pre-distortion processing  

v^d(n)=vi′(n)*β′(n)=ρin(n)*α(n)*α′(n)*Gρ(ρin(n)*α(n))*---(11)

exp (j (θ_in (n) + σ (n) +σ ' (n) + F_θ (ρ_in (n)* α (n))))

When the system is converged, should satisfy v_i (n) =v '_i (n), at this time vd(n)=v^d(n).RASCAL algorithm using the adaptive process, then the iterative equation as:

eρ(n)=|vd(n)|-|v^d(n)|---(12)

e[!theta!](n)=arg(vd(n))-arg(v^d(n))---(13)

α '_i+1 (n) =α'_i (n) + e_ρ * μ_ρ     (14)

σ '_i+1 (n) =σ'_i (n) +e_θ (n)* μ_θ     (15)

wherein, e_ρ (n), e_θ (n) are respectively representative v_d (n) and   The error of the amplitude and phase of, arg (v_d (n)) and   The phase of the transducer and said, μ_ρ and μ_θ is amplitude and phase of the step of iteration. According to the type (14)   and (15) gives a self-adaptive iterative equation to update to 2nd 2nd time lookup table of the pre-distortion processing of the pre-distortion coefficient α '(n) and σ' (n), so that the error signal e (n) to approach to 0, wherein e (n) = e_ρ (n) *exp (je_θ (n)).

It should be explained that the predistortion unit 1st 1st 2nd in the lookup table is a lookup table pre-distortion unit duplication of 2nd, when the adaptive algorithm in the look-up unit to the 2nd when updating the predistortion coefficient, 1st lookup table corresponding to the pre-distortion coefficient is updated, convenient for describing the present invention, a lookup table 2nd 1st lookup table and the variable in the said given to the different symbols.

Because the power amplifier to the input phase of signal in the original phase distortion is then overlaid a phase offset, the phase offset and the amplitude value of the input signal. Therefore, when the amplitude distortion sufficient hours, the following formula can be utilized to eliminate all of the residual phase error.

σ '_i+1 (n) =σ'_i (n) + μ_θ     (16)

Figure 5, 6 provides the predistortion scheme of the present invention is applied to a digital television terrestrial multimedia broadcast (DTMB) the results of the performance of the system, power amplifier model the polynomial model.

Figure 5 and in the pre-distortion device to work the power amplifier output a demodulated constellation, Figure (a) representative 64QAM constellation of signal source, Figure (b) representative of the demodulated constellation without predistortion device, it can be seen the presence of the memory effect, has been seriously distorted constellation, Figure (c) is the one-dimensional lookup table pre-distorter of the constellation diagram, and Figure (d) is provided by the invention of a two-dimensional lookup table pre-distorter the constellation diagram. As can be seen, the one-dimensional lookup table pre-distorter can also be better restore the constellation point of distortion, but the inhibition of the proliferation of the constellation points is still relatively limited capacity, and two-dimensional lookup table pre-distorter can be clearly more perfect restoration of the original constellation diagram.

Figure 6 is in all cases, the power amplifier output power spectral density graph, wherein (a) the signal of the predistortion correction-free ; (b) is a one-dimensional lookup table pre-distortion device corrected signal ; (c) provided by the invention is the method for pre-distorter 1 corrected signal ; (d) provided by the invention is the method for pre-distorter 2 correction signal ; (e) the original signal. As can be seen by the diagram, the method of using one-dimensional look-up-of-band spectral leakage can be suppressed, the performance of the one-dimensional lookup table pre-distorter, however, than the two-dimensional lookup table pre-distorter performance is bad, the two can be achieved the most performance difference 20dB, and can be found through careful observation of the PSD (Power   Density   Spectrum: power spectral density) curve will present the double-shoulder of uneven phenomenon, this is because of the memory effect of the power amplifier, can see one-dimensional lookup table to correct or not very good memory effect of the power amplifier. And (c), (d) of the two two-dimensional lookup table can be a very good method to overcome the memory effect of the power amplifier. Because the (d) to take in the method for weighted summation of forgetting factor, compared with (c) in a direct summation, this kind of method for estimation of the power amplifier memory effect more accurate and more reasonable. Therefore (d) than the performance of (e) with the property of almost 5dB improved, already close to the PSD curve of the original signal.

In the two-dimensional lookup table in the design of the pre-distorter, it is worth noting that is one of the problems with the two-dimensional table on the size of the pre-distorter performance compromise. Usually the greater the size of the lookup table, the convergence slower, but the smaller the error. Figure 7 of the present invention shown in two-dimensional index value Y distribution of the columnar pattern. As can be seen from the chart, only in a small part of the probability of Y in the relatively large, while the present invention adopts a uniform quantization scheme, therefore, if the 2nd-dimensional size and fetch of the excessive, will lead to many lookup table unit is not used, thus causing waste of storage space. Figure 8 as when the 2nd out uygur size 32 and 512 of the power amplifier output power spectral density graph, wherein (a) the signal of the predistortion correction-free ; (b) is a one-dimensional lookup table pre-distortion device corrected signal ; (c) provided by the invention for the programme 1 after the pre-distortion corrector corrects the signal ; (d) is the scheme provided by the invention 2 after the pre-distortion corrector corrects the signal ; (e) the original signal. As can be seen by the map when the 2nd-dimensional Y size of the respectively 32 and 512 time, the impact on the system performance is not particularly large, so in an actual system design in the pre-distorter, the performance and complexity to be considered a compromise.

Of the above, only the best of the Mode of execution of this invention, the scope of protection of the present invention, however, is not limited to this, any familiar with the technical staff Technical Field in this invention discloses the technical range, can be easily imagined that the change or replacement, should be covered in the scope of protection of the invention.

The invention has been described in detail in the specification of the content belonging to the field of known technology professional technical personnel.