CN101527544B - Device and method for identifying inverse characteristic of nonlinear system, power amplifier and predistorter thereof - Google Patents

Abstract

The invention discloses a nonlinear system, especially an inverse characteristic identification device and method of a nonlinear power amplifier, a power amplifier predistorter, and power amplifying equipment using the nonlinear system inverse characteristic identification device and the predistorter. The method for identifying the inverse characteristics of the nonlinear system according to the original input signal x(n) of the nonlinear system and the corresponding feedback output signal _yG (n) of the present invention includes: according to the inverse filter signal _xF (n) and the feedback The output signal y _G (n) calculates the lookup table function Q{ }; generates the intermediate predistortion signal z(n) according to the feedback output signal y _G (n) and the lookup table function Q{ }; according to the intermediate predistortion signal z( n) and the original input signal x(n) construct a filter function F{ }; use the parameters of the filter function F{ } to inverse filter the original input signal x(n) to generate an inverse filter signal x _F (n); and iterate Repeat the above steps until the set conditions are met; among them, the look-up table function Q{·} and the filter function F{·} represent the inverse characteristics of the nonlinear system.

Description

Nonlinear system inverse characteristic identification device and method, power amplifier and predistorter thereof

technical field

The present invention relates generally to nonlinear systems, and in particular to nonlinear power amplifiers, especially for use in transmitters of wireless communication systems such as base stations and mobile stations. More specifically, the present invention relates to a nonlinear system, especially a nonlinear power amplifier inverse characteristic identification device and method, a power amplifier predistorter, and a power amplification device using a nonlinear system inverse characteristic identification device and a predistorter, which mainly utilizes Digital technology implements pre-distortion in the baseband to compensate for the non-memory non-linear and memory characteristics of the non-linear complementary power amplifier.

Background technique

As a typical representative of nonlinear systems, power amplifiers are an important part of many electronic devices, which can amplify weak electrical signals to meet the needs of long-distance transmission. Among them, the amplified energy comes from a DC power supply, that is, the power amplifier can convert DC energy into an AC signal.

According to the physical characteristics of the power amplifier, as the power of the input signal increases from small to large, the characteristic curve reflecting the power relationship between the input signal and the output signal of the power amplifier can be divided into a linear region, a nonlinear region and a saturation region. Fig. 1 shows the nonlinear input and output signal characteristics of the power amplifier PA. In the region where the amplitude of the input signal V _IN is small, the output V _OUT of the power amplifier PA is almost a linear amplification of the input signal V _IN , but as the input signal V _IN amplitude increases, the nonlinear characteristics of the power amplifier PA gradually become obvious until Finally saturated. The nonlinear performance is in the frequency domain. Due to the intermodulation effect, the amplified signal has spectral broadening outside the band and distortion in the band. As shown in Figure 2, it shows the spectral broadening caused by the nonlinear power amplifier. .

Ideally, it is hoped that the power amplifier can only play the role of linear amplification, that is, the output signal is only a simple linear amplification of the input signal, so let the power amplifier work in the linear region. However, at this time, the efficiency of the power amplifier in converting the DC signal into the AC signal is very low, resulting in a large amount of waste of energy, and additional heat dissipation equipment needs to be added.

Therefore, on the one hand, in order to improve the efficiency of the power amplifier, on the other hand, because the signals in many modern communication systems have a large dynamic range (peak-to-average power ratio), the power amplifier often needs to work in the nonlinear region, resulting in output Distortion of the signal (in the frequency domain, distortion occurs in the band, and spectrum broadening occurs outside the band). In the field of power amplifiers, the distortion of the nonlinear power amplifier whose current output is only affected by the current input is usually called the memoryless nonlinear characteristic of the power amplifier.

Among many methods to overcome the nonlinear characteristics of power amplifiers, baseband predistortion is a method that has attracted much attention. As shown in FIG. 3 , a schematic diagram of input and output signal characteristics of a power amplifier predistorter for canceling the nonlinear characteristic shown in FIG. 1 is shown. The baseband predistortion technology uses the predistorter PD to simulate the inverse characteristics of the power amplifier PA, so that the original input signal V _IN is predistorted before being input to the power amplifier, thereby compensating for the nonlinearity of the power amplifier PA, and at the output of the power amplifier An amplified output signal V _OUT without distortion is obtained. The input and output signal characteristics of a power amplifier with a predistortion processing function show very good linearity until it approaches the saturation region, as shown in Figure 4, which shows a schematic diagram of the input and output signal characteristics of a power amplifier equipped with a predistorter .

However, most predistorters that work well are for single-tone signals or narrowband frequency signals. That is, mostly used to compensate the memoryless nonlinearity of the power amplifier mentioned above. However, as the bandwidth of the signal becomes wider and wider, the power amplifier shows a certain memory characteristic (frequency selectivity), that is, the current output signal is not only related to the current input signal, but also related to the previous input signal, as shown in the figure 5 shows a schematic diagram of the input and output signal characteristics of a nonlinear power amplifier with memory characteristics. The memory characteristic of the power amplifier manifests as an asymmetric frequency spectrum near the carrier on its output, as shown in Figure 6, which shows a schematic diagram of the frequency spectrum of a signal amplified by a nonlinear power amplifier with memory characteristics. That is, although the spectrum of the carrier (desired signal) is completely symmetrical, the spurious spectrum caused by the distortion is asymmetrical with respect to the center carrier.

Therefore, in order to accurately simulate the inverse characteristics of the power amplifier with the above-mentioned memory characteristics, it is necessary to use more complex structures and methods to model and estimate the parameters of the inverse characteristics of the power amplifier in order to achieve pre-distortion. Current techniques for simultaneously compensating both memoryless nonlinearity and memory characteristics of power amplifiers mainly include memory polynomial models and two-dimensional look-up table methods.

The memory polynomial model is a simplified model of the Volterra model. The Volterra model uses the Volterra series to accurately model the nonlinear system. By simulating the inverse Volterra series model of the power amplifier, it can theoretically offset the non-memory nonlinear characteristics and memory characteristics of the power amplifier. While the Volterra model is very effective for counteracting the memoryless nonlinearity and memory characteristics of power amplifiers, its physical implementation is nearly impossible due to the use of complex formulas and the enormous amount of mathematical operations required. To this end, a simplified Volterra model, ie, a memory polynomial model, is proposed. As shown in Figure 7, it shows a schematic diagram of predistorting the input signal of a nonlinear power amplifier with memory characteristics using a memory polynomial model according to the prior art, wherein the memory polynomial model uses a polynomial function related to the current input and the previous input Indicates the output. The ability to cancel the non-linearity of the power amplifier depends on the number of previous inputs used and the order of the polynomial. To describe the nonlinear characteristics of power amplifiers finely requires the use of polynomials with high-order terms, which often leads to the generation of matrices with large condition numbers, making numerical calculations very difficult.

The two-dimensional look-up table method is based on the relationship between the current output of the power amplifier and its current input and previous input, and the inverse characteristics of the power amplifier are made into a look-up table, so that the input signal is used before power amplification from the look-up table. The resulting weighting coefficients predistort the input signal, thereby canceling out the non-linear characteristics of the power amplifier. As shown in FIG. 8 , it shows a schematic diagram of performing predistortion processing on an input signal of a nonlinear power amplifier with memory characteristics using a two-dimensional look-up table according to the prior art. However, the two-dimensional look-up table not only requires a large memory to store a large amount of pre-distortion data, but also has a limited memory depth, and often only uses the relationship between the current input and the previous input and output to pre-distort the input signal of the power amplifier. Counteracts the memory characteristics of power amplifiers. If memory depth needs to be increased, multidimensional lookup tables will become very complex and difficult to implement.

Similar technical problems as above also exist for other nonlinear systems similar to nonlinear power amplifiers.

Contents of the invention

In view of this, an object of the present invention is to provide a nonlinear system inverse characteristic identification device and method, which can effectively reduce the requirements for storage devices and numerical calculations.

Another object of the present invention is to provide a power amplifier inverse characteristic identification device and method, a power amplifier predistorter, and a power amplification device using a power amplifier inverse characteristic identification device and a predistorter, which are mainly realized in baseband using digital technology Pre-distortion to compensate for the non-memory nonlinearity and memory characteristics of the power amplifier, while effectively reducing the requirements for storage devices and numerical calculations.

According to one aspect of the present invention, a nonlinear system inverse characteristic identification method is provided, which is used to analyze the inverse characteristic of the nonlinear system according to the original input signal x(n) of the nonlinear system and the corresponding feedback output signal y _G (n) Carry out identification, including: look-up table function Q{ } generation step, calculate look-up table function Q{ } according to inverse filter signal x _F (n) and feedback output signal y _G (n); intermediate predistortion signal z(n) The generation step is to generate the intermediate predistortion signal z(n) according to the feedback output signal y _G (n) and the lookup table function Q{ }; the filter function F{ } construction step is to generate the intermediate predistortion signal z(n) and the original The input signal x(n) constructs the filter function F{ }; in the inverse filter step, the parameters of the filter function F{ } are used according to the formula x _F (n)=F ^-1 {x(n)} to the original input signal x( n) Perform inverse filtering to generate an inverse filtering signal x _F (n); and iteratively repeat the above steps until the set conditions are met; wherein, the lookup table function Q{ } and filter function F{ } represent the nonlinear system inverse characteristic.

According to another aspect of the present invention, there is provided a nonlinear system inverse characteristic identification device, which is used for the inverse of the nonlinear system according to the original input signal x(n) of the nonlinear system and the corresponding feedback output signal y _G (n) Identify the characteristics of the nonlinear system, including: parameter calculator, used to identify the inverse characteristics of the nonlinear system, generate a lookup table function Q{ } and a filter function F{ } representing the inverse characteristics of the nonlinear system; and a convergence decision device , used to determine whether the lookup table function Q{ } and filter function F{ } generated by the parameter calculator meet the set conditions, and control the parameter calculator to iteratively calculate the lookup table function Q{ } and filter function F{ }; wherein the parameter calculator includes: a look-up table generator, for calculating the look-up table function Q{ } according to the inverse filter signal x _F (n) and the feedback output signal y _G (n); intermediate predistorter, It is used to generate the intermediate predistortion signal z(n) according to the feedback output signal y _G (n) and the lookup table function Q{ } calculated by the lookup table generator; the filter parameter calculator is used to generate the intermediate predistorter according to The intermediate predistortion signal z(n) and the original input signal x(n) construct the filter function F{ }; and the inverse filter is used to use the filter function F{ } constructed by the filter parameter calculator according to the formula x _F (n)=F ⁻¹ {x(n)} performs inverse filtering on the original input signal x(n) to generate an inverse filtered signal x _F (n).

According to another aspect of the present invention, a power amplifier predistorter is provided, including: an address generator, which is used to convert the original input signal into a look-up table address after modulo and quantization; a one-dimensional look-up table, according to the look-up table function Q{ } configures and outputs the correction factor according to the address of the lookup table output by the address generator; the multiplier is used to multiply the original input signal with the correction factor output by the one-dimensional lookup table to obtain an intermediate predistortion signal; and filter The multiplier is configured according to the filter function F{ }, and the intermediate pre-distortion signal output by the multiplier is filtered to generate a pre-distortion signal; where the look-up table function Q{ } and filter function F{ } are based on the above generated by the above-mentioned nonlinear system inverse characteristic identification method or generated according to the above-mentioned nonlinear system inverse characteristic identification device.

According to another aspect of the present invention, a power amplifying device is provided, including: a pre-distortion module, used to pre-distort the original input signal, and output the pre-distorted input signal; a digital-to-analog converter, used to convert the The pre-distorted input signal is converted into an analog signal; the up-converter is used to up-convert the analog signal output from the digital-to-analog converter into a radio frequency signal; the power amplifier is used to amplify the power of the radio frequency signal output by the up-converter and output The output signal after power amplification; wherein, the structure of the pre-distortion module is the same as that of the power amplifier pre-distorter defined above.

According to still another aspect of the present invention, a power amplification device is provided, including: a pre-distortion module, the structure of which is the same as that of the power amplifier pre-distorter defined above, for performing pre-distortion on the original input signal and outputting the pre-distortion module. Distorted input signal; digital-to-analog converter, used to convert the pre-distorted input signal output by the pre-distortion module into an analog signal; up-converter, used to up-convert the analog signal output from the digital-to-analog converter into a radio frequency signal; power The amplifier is used to amplify the power of the RF signal output by the up-converter and output the amplified output signal; the directional coupler is used to branch the output of the power amplifier into two signals, one signal is output as an output signal, and the other One signal is fed back to the attenuator; the attenuator is used to receive the signal fed back by the directional coupler and attenuate the feedback signal; the down-converter is used to down-convert the attenuated signal output by the attenuator to a baseband signal; analog-to-digital conversion The device is used to convert the baseband signal output by the down-converter into a digital signal as a feedback output signal; the delayer is used to delay the original input signal to generate a delayed input signal; and the pre-distortion data updater is used to The method is to receive the feedback output signal output by the analog-to-digital converter and the delayed input signal output by the delayer, and update the look-up table function and the filter coefficient online according to the received feedback output signal and delayed input signal.

According to another aspect of the present invention, there is provided a method for online updating the parameters of the pre-distortion module of the power amplification device, comprising steps:

Comparing the original input signal x(n) of the power amplification device with the feedback output signal, the error signal is calculated according to the following formula:

$\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; [</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span><span\; class="notranslate">\; \}</span>$

$<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; \{</span><span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; [</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; [</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span><span\; class="notranslate">\; \}</span>$

$<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span>$

Among them, F{·}, Q{·}, P{·} respectively represent the filtering function of the filter, the lookup table function of the lookup table, and the amplification function of the power amplifier, and

z(n)=x(n) Q([|x(n)|]),

The query table function Q{ } is

Q([|x(n)|])=[q ⁰ ,q ¹ ,...,q ^K-1 ] ^T

The filter coefficients are

W=[w ₁ , w ₂ , . . . , w _L-1 ] ^T

L represents the memory depth of the filter, [|·|] represents modulo and quantization; and

Update the lookup table function and filter coefficients online according to the following formula

$\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\end{array}\right]<span\; class="notranslate">\; =</span>\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\end{array}\right]<span\; class="notranslate">\; +</span>\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; q</span>}}\\ \mathrm{<span\; class="notranslate">\; u</span>}\end{array}\right]<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>$

in

U=[μ ₁ , μ ₁ ,..., μ _L-1 ] ^T

Both U and _μq are convergence step sizes.

According to the above-mentioned nonlinear system inverse characteristic identification device and method, power amplifier predistorter, and power amplification equipment using the power amplifier inverse characteristic identification device and predistorter of the present invention, the numerical calculation difficulties caused by the high condition number of the matrix are avoided .

Further aspects of the invention are given in the descriptive section below, wherein the detailed description serves to fully disclose preferred embodiments of the invention without imposing limitations thereon.

Description of drawings

The above and other objects and advantages of the present invention will be further described below in conjunction with specific embodiments and with reference to the accompanying drawings. In the drawings, the same or corresponding technical features or components will be indicated by the same or corresponding reference numerals.

FIG. 1 is a schematic diagram showing the nonlinear characteristics of input and output signals of a power amplifier;

Figure 2 is a schematic diagram showing the frequency spectrum of a signal amplified by a nonlinear power amplifier;

3 is a schematic diagram showing input and output signal characteristics of a power amplifier predistorter for canceling the non-linear characteristic shown in FIG. 1;

4 is a schematic diagram showing input and output signal characteristics of a power amplifier provided with a predistorter;

5 is a schematic diagram showing the input and output signal characteristics of a nonlinear power amplifier with memory characteristics;

Fig. 6 is a schematic diagram showing the frequency spectrum of a signal amplified by a nonlinear power amplifier with memory characteristics;

Fig. 7 is a schematic diagram showing that a memory polynomial model is used to perform predistortion processing on an input signal of a nonlinear power amplifier with memory characteristics according to the prior art;

Fig. 8 is a schematic diagram showing that a two-dimensional look-up table is used to perform pre-distortion processing on an input signal of a nonlinear power amplifier with memory characteristics according to the prior art;

Fig. 9 is a structural schematic diagram showing the inverse characteristics of the nonlinear system constructed using the Hammerstein model;

Fig. 10 is a schematic diagram showing the principle of the indirect learning method used in the present invention to obtain the inverse characteristics of the nonlinear system;

11 is a schematic diagram showing the principle of an identification device for obtaining the inverse characteristics of a nonlinear system off-line according to the indirect learning method of the present invention;

Fig. 12 is a flow chart showing the optimization process of the identification method for obtaining the inverse characteristics of the nonlinear system off-line according to the indirect learning method of the present invention;

Fig. 13 is a block diagram showing a nonlinear system inverse characteristic identification device according to a preferred embodiment of the present invention;

FIG. 14 is a block diagram showing a device structure for collecting input and output data of a power amplifier;

Fig. 15 is a structural block diagram showing a nonlinear power amplifying device capable of realizing predistortion according to the present invention;

Fig. 16 is a structural principle diagram showing a power amplifying device capable of online updating and upgrading the parameters of the pre-distortion module according to the present invention; and

Fig. 17 is a block diagram showing an exemplary structure of a personal computer.

Detailed ways

Embodiments of the present invention will be described below with reference to the drawings.

The identification of the inverse characteristics of the nonlinear system in the present invention is based on the commonly used Hammerstein model. The Hammerstein model is a nonlinear structural model that considers the memory characteristics of nonlinear systems. It is composed of two parts in series: a memoryless nonlinear model and a linear time-varying system. FIG. 9 shows a schematic structural diagram of constructing the inverse characteristics of a nonlinear system using a Hammerstein model 90 . As shown in FIG. 9 , the Hammerstein model 90 includes an address generator 91 , a look-up table (Look-Up Table, LUT) 92 , a FIR (finite impulse response) filter 93 and a multiplier 95 . The address generator 91 generates an address for inquiring the LUT 92 based on the input signal x(n), and then selects a coefficient represented by a complex number from the LUT 92 based on the address. After the complex coefficients are multiplied by the input signal x(n) in the multiplier 95, the resulting product is digitally filtered by the FIR filter 93 to form the final predistorted signal u(n). Here, the address of the LUT 92 is represented by k.

The one-dimensional LUT can be used to compensate the memoryless nonlinear characteristics of the nonlinear system, and the filter is used to compensate the memory characteristics of the nonlinear system, thereby avoiding the use of a large amount of memory and high-order polynomials, and reducing the amount of calculation. Therefore, the problem is transformed into how to construct the parameters of the one-dimensional LUT and FIR filter, so that the combination of the two can better fit the comprehensive nonlinear characteristics of the nonlinear system. For this reason, the present invention adopts " indirect learning method ". Fig. 10 shows the principle of the indirect learning method used in the present invention to obtain the inverse characteristics of the nonlinear system.

Usually, the characteristics of the nonlinear system do not change frequently over time, so the learning algorithm and the module of identifying the predistortion model A can collect some input and output data samples (for example, n input data samples and n Output data samples) and perform offline data processing. After identification, the new parameters are sent to the predistortion model A module of the nonlinear system to compensate for the integrated nonlinearity of the nonlinear system. As shown in Fig. 10, the process is to obtain the inverse characteristics of the nonlinear system (estimation model A parameters). In Figure 10, x(n) is the input of the nonlinear system, y(n) is the output of the nonlinear system, G is the expected gain coefficient of the nonlinear system, and y _G (n) is called the feedback of the nonlinear system output. The predistortion model A module can be implemented with various programmable devices such as field programmable gate array (FPGA), so that it is easy to update the time-varying parameters of the nonlinear system on the spot.

FIG. 11 shows an identification device for obtaining the inverse characteristics of the nonlinear system off-line according to the indirect learning method, that is, a schematic diagram of the learning algorithm shown in FIG. 10 and a module for identifying the predistortion model A. The principle of the device is based on the indirect learning method shown in FIG. 10 . The data collection and storage module 301 in the figure sends the collected and stored input and output data of the nonlinear system to the identification module 300, and the identification module 300 maps and outputs these data with the input data of the nonlinear system stored in the data collection and storage module 301 through The adder 306 compares and uses the error signal e(n) as a basis for modifying the parameters in the identification module 300 .

The identification module 300 includes an address generator 302 , a lookup table 303 , a multiplier 304 and a dual filter 305 . Here address generator 302, look-up table 303, multiplier 304 and double filter 305 correspond to address generator 91 shown in Figure 9, look-up table 92, multiplier 95 and FIR filter 93 respectively, and its specific working principle will Combined with the specific description below.

According to Figure 11, the error signal e(n) can be expressed as

$\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span>$

$<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span>$

$<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; \{</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; [</span><span\; class="notranslate">\; |</span>{\mathrm{<span\; class="notranslate">\; the\; y</span>}}_{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>$

where x(n) represents the input signal shown in Figure 10. In the case where the expected gain of the nonlinear system shown in Figure 10 is G, and the output of the nonlinear system is y(n), y _G (n) can be expressed by the following formula

_yG (n)=y(n)/G(2)

表示辨识信号(输入信号x(n)的估计值)，记为

Represents the identification signal (estimated value of the input signal x(n)), denoted as

$\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>$

The intermediate predistortion signal z(n) is

z(n)=y _G (n) Q([|y _G (n)|]) (4)

Here, F{·} represents the filter function of the filter 305 , and Q{·} represents the lookup table function of the lookup table 303 .

[|·|] represents the modulo and quantization performed by the address generator 302, namely

k=[|y _G (n)|] k=0,...,K-1 (5)

k corresponds to an address (for example, 0-255) in the lookup table.

Therefore, the problem of identifying the inverse characteristic of the nonlinear system is transformed into minimizing the error signal e(n) by designing the parameters of the filter 305 and the look-up table 303 . Considering that e(n) is a time series, the identification problem can be written as:

$\underset{\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; Q</span>}}{\mathrm{<span\; class="notranslate">\; min</span>}\mathrm{<span\; class="notranslate">\; E.</span>}}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; Q</span>}}{\mathrm{<span\; class="notranslate">\; min</span>}}\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 6</span>}<span\; class="notranslate">\; )</span>$

here

$\mathrm{<span\; class="notranslate">\; E.</span>}<span\; class="notranslate">\; =</span>\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 7</span>}<span\; class="notranslate">\; )</span>$

In order to effectively minimize the function E, the best way is to jointly optimize F{·} and Q{·}. This embodiment proposes a calculation method to minimize E by alternately optimizing F{ } and Q{ } to optimize the obtained F{ } and Q{ } (filter 305 and lookup table 303) to approximate the inverse properties of nonlinear systems.

First, let's analyze the calculation function Q{·}. The content of the lookup table can be obtained by analyzing the gain of each sample point, where the gain of each sample point can be expressed as

gp(n)=x(n)/y _G (n) (8)

That is to say, each gp(n) corresponds to a y _G (n)y corresponds to an address k in a lookup table, and the gp(n) with the same address k value is classified into one category (into a group) , write it as

GP _k = [gp(i), . . . gp(j)] 0<k<K-1,

                                                                                                                                                                                                                  ,

By averaging the modulus and phase of gp(n) in each group, a compromised gain characteristic can be obtained, namely

g _k = mean(|GP _k |) (10)

p _k ＝mean(∠GP _k ) (11)

Therefore, the query table function Q{ } can be obtained

Q{ }＝Gain⊙exp(j Phase) (12)

in

Gain=[g ₀ , . . . , g _i , . . . ] ^T (13)

Phase＝[p ₀ ，...p _i ，...] ^T (14)

And, ⊙ represents the Hadamard product, and the superscript T represents transposing the matrix.

After obtaining the lookup table function Q{·} according to formula (12), the intermediate predistortion signal z(n) can be calculated according to formula (4).

It should be pointed out that when calculating the compromised gain characteristics above, the average value of the modulus and phase of gp(n) in each group is selected. Of course, other algorithms known to those skilled in the art can also be used to obtain the compromised gain characteristics, such as the median method, the least square method, etc., and specific related descriptions are omitted here.

The following analyzes and calculates the filter function F{·}, which can be obtained by the least square method. After obtaining the intermediate predistortion signal z(n), in order to obtain the optimal filter function F{ } to approximate the input signal x(n), the filter coefficients can be obtained according to the least square method:

W＝(Z _M ^H Z _M ) ^-1 Z _M ^H X (15)

in

X=[x(1), x(2), . . . , x(N)] ^T (16)

${\mathrm{<span\; class="notranslate">\; Z</span>}}_{\mathrm{<span\; class="notranslate">\; m</span>}}<span\; class="notranslate">\; =</span>\left(\begin{array}{cccc}\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; 0</span>}& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& \mathrm{<span\; class="notranslate">\; 0</span>}\\ \mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 3</span>}<span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& \mathrm{<span\; class="notranslate">\; 0</span>}\\ <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>\\ & & <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& \mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 2</span>}<span\; class="notranslate">\; )</span>\\ \mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; )</span>& \mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>& <span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span><span\; class="notranslate">\; .</span>& \mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; N</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}<span\; class="notranslate">\; )</span>\end{array}\right)<span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 17</span>}<span\; class="notranslate">\; )</span>$

Here, N is the length of collected data, and L is the order (memory depth) of the FIR filter.

After the parameter W of the filter is calculated according to formula (15), the filter function F{ } can be expressed in the form of matrix and vector as:

F{z(n)}＝Z _M ·W (18)

满足要求。这意味着当前所使用的查询表函数Q{·}和滤波函数F{·}很好地拟和了非线性系统的特性，可以用作图10所示的预失真模型A的参数。Then, the estimated value of the input signal x(n) can be obtained according to formula (3) and formula (18) The error signal e(n) is thus calculated according to formula (1). If the error signal e(n) satisfies the set conditions, the estimated value of the input signal x(n) is considered

fulfil requirements. This means that the currently used lookup table function Q{·} and filter function F{·} fit the characteristics of the nonlinear system well, and can be used as parameters of the predistortion model A shown in FIG. 10 .

不能满足要求，需要对查询表函数Q{·}和滤波函数F{·}进行进一步的修正。此时，利用滤波器305对输入信号x(n)按照下式进行逆滤波得到逆滤波信号x_F(n)：On the other hand, if the error signal e(n) does not meet the set conditions, the estimated value of the input signal x(n)

Can not meet the requirements, need to further modify the query table function Q{·} and filter function F{·}. At this time, the input signal x(n) is inversely filtered by the filter 305 according to the following formula to obtain the inverse filtered signal _xF (n):

x _F (n) = F ^-1 {x(n)} (19)

Then, use the inverse filtered signal x _F (n) to replace the original input signal x (n) in formula (8) together with the feedback output signal y _G (n) of the nonlinear system, using formulas (4), (5) And formula (12) to formula (18) recalculate the lookup table function Q{·} and filter function F{·}. Then, calculate the error signal e(n) according to formula (1) until the set conditions are met.

Here, the set condition can be to check whether the standard mean square error NMSE shown in the following formula (20) converges or to determine whether the number of iterations is greater than a certain threshold after obtaining the error signal e(n).

$\mathrm{<span\; class="notranslate">\; NMSE</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; dB</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 10</span>}{\mathrm{<span\; class="notranslate">\; log</span>}}_{\mathrm{<span\; class="notranslate">\; 10</span>}}<span\; class="notranslate">\; (</span>\frac{\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}{\underset{\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; N</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{<span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span>}^{\mathrm{<span\; class="notranslate">\; 2</span>}}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 20</span>}<span\; class="notranslate">\; )</span>$

Of course, those skilled in the art should understand that the set conditions are not limited to the standard mean square error NMSE mentioned here, and may also be other criteria, such as mean square error MSE, peak mean square error PMSE and so on.

Fig. 12 shows a flow chart of the optimization process performed by the identification module 300 shown in Fig. 11, that is, how to obtain The look-up table 303 function Q{·} of the inverse characteristic of the system and the function F{·} of the filter 305.

As shown in FIG. 12 , firstly, in step S1201, the original input signal x(n) of the nonlinear system is obtained, and the output signal y _G (n ).

Next, in step S1202, the output signal y _G (n) is subjected to modulo sampling and quantization operations sample by sample, and the address number k corresponding to the output signal y _G (n) of each sample is obtained according to formula (5). Afterwards, in step S1203, an initial standard mean square error NMSE ₀ is set, and the iteration variable i=0.

Then, in step S1204, grouping and averaging are performed according to formulas (8)-(14), and the lookup table function Q{·} is obtained. First, the point-by-point gain gp(n) is obtained according to formula (8), and the sample-by-sample gain is divided into several groups according to the corresponding output signal addresses according to formula (9). Next, obtain the query table function Q{·} according to (10)-(14).

After the look-up table function Q{·} is calculated, in step S1205, the intermediate predistortion signal z(n) is calculated according to formula (4). Afterwards, in step 1206, the iteration variable i is incremented by 1, that is, i=i+1.

并根据公式(20)计算标准均方误差NMSE_i。Then, perform filter design in step S1206 to obtain the filter function F{·}, and calculate the standard mean square error NMSE _i . First, construct the filter matrix Z _M according to formula (17) and obtain the filter function F{·} according to formula (15). The estimated value of the input signal x(n) is then calculated according to Equation (3) and Equation (18)

And calculate the standard mean square error NMSE _i according to formula (20).

After the standard mean square error NMSE _i is obtained, it is judged in step S1208 whether the standard mean square error NMSE _i converges or whether the number of iterations is greater than a certain threshold. If yes, the processing flow advances to step S1209, and the current look-up table function Q{·} and filter function F{·} are output to the field programmable gate array FPGA, so as to realize the predistorter in hardware.

If the judgment result in step S1208 shows that the standard mean square error NMSE _i does not converge and the number of iterations is not greater than a certain threshold value, then the processing flow advances to step S1210, and uses a filter to perform inverse filtering on the input signal x(n) to obtain x _F (n)=F ^-1 {x(n)}.

Then, the processing flow returns to step S1204, and steps S1204-S1208 are iteratively executed. Use this inverse filtered signal x _F (n) to replace the original input signal x (n) in formula (8) together with the feedback output signal y _G (n) of the nonlinear system, using formulas (4), (5) and formula (12) to formula (18) to recalculate the lookup table function Q{·} and filter function F{·}. Next, calculate the standard mean square error NMSE _i according to formula (20), until the standard mean square error NMSE _i converges or the number of iterations is greater than a certain threshold.

The structure of a preferred nonlinear system inverse characteristic identification device for implementing the above identification method will be described in detail below. As shown in FIG. 13 , it shows a block diagram of a nonlinear system inverse characteristic identification device according to a preferred embodiment of the present invention.

The original input signal x(n) and the attenuated feedback output signal y _G (n) are stored in the data storage device 810 shown in FIG. 13 , and these signals are sent to the parameter calculator 800 . The parameter calculator 800 includes an address generator 801 , a lookup table generator 802 , a filter parameter calculator 803 , an intermediate predistorter 804 , a filter 805 , an inverse filter 806 , and a selector 807 .

After the feedback output signal y _G (n) is input into the parameter calculator 800, it is converted into an address k by the address generator 801 according to formula (5). The converted address k is input to the look-up table generator 802 together with the feedback output signal y _G (n). In addition, the look-up table generator 802 also inputs the signal x(n) or the inverse filter signal x _F (n) from the selector 807, and generates a look-up table function Q{ } according to the formulas (8)-(14), the obtained The generated lookup table function Q{·} is output to the intermediate predistorter 804 .

The intermediate predistorter 804 receives the lookup table function Q{ } and the feedback output signal y _G (n) from the lookup table generator 802, thereby generating an intermediate predistortion signal z(n) according to the formula (4), and the generated The intermediate predistortion signal z(n) of is output to the filter parameter calculator 803 and the filter 805 respectively.

The filter parameter calculator 803 receives the intermediate predistortion signal z(n) and the original input signal x(n) from the intermediate predistorter 804, and generates a filter function F{·} according to formulas (15)-(18). The generated filter function F{·} is output to the filter 805 and the inverse filter 806, respectively.

并将所生成的原始输入信号的估计值

输出到NMSE计算器808。The filter 805 processes the intermediate predistortion signal z(n) generated by the intermediate predistorter 804 according to the filter function F{ } generated by the filter parameter calculator 803 according to formula (3) to generate an estimated value of the original input signal

and generate an estimate of the original input signal

Output to NMSE calculator 808.

和原始输入信号x(n)，并按照公式(20)计算输出标准均方误差NMSE到收敛判决器811。NMSE calculator 808 receives an estimate of the original input signal generated by filter 805

and the original input signal x(n), and calculate and output the standard mean square error NMSE to the convergence decision unit 811 according to formula (20).

The inverse filter 806 arranged symmetrically with the filter 805 also receives the filter function F{ } from the filter parameter calculator 803, and uses the filter function F{ } to inverse filter the original input signal x(n), and outputs the inverse Filtered signal x _F (n) to selector 807 .

The selector 807 is used to select the original input signal x(n) and the inverse filtered signal x _F (n), and output one of them to the lookup table generator 802 as required. Specifically, in general, when calculating the lookup table function Q{·} and filter function F{·} in the first iteration, the selector 807 outputs the original input signal x(n) to the lookup table generator 802, but in In the subsequent iterative calculation, the inverse filter signal x _F (n) generated by selecting the inverse filter 806 is output to the lookup table generator 802 .

The convergence determiner 811 records the standard mean square error NMSE and the number of iteration steps, and uses the standard mean square error NMSE or the number of iteration steps as a judgment condition. If the determination condition is not satisfied, that is, if the standard mean square error NMSE does not converge or the number of iteration steps does not exceed a predetermined threshold, the convergence determiner 811 controls the parameter calculator 800 to iteratively execute the above process. If the judgment condition is met, that is, if the standard mean square error NMSE converges or the number of iteration steps exceeds a predetermined threshold, then the convergence decision unit 811 controls the parameter calculator 800 to output the latest calculated filter function F{ } and lookup table function Q{ ·}.

The following will take a nonlinear power amplifier as an example to illustrate how to use the above-mentioned nonlinear system inverse characteristic identification method and identification device to compensate the memoryless nonlinear characteristics and memory characteristics of the power amplifier, thereby expanding its linear amplification region. That is to say, how to use the above-mentioned nonlinear system inverse characteristic identification method and identification device to predistort the nonlinear power amplifier.

Firstly, it is necessary to collect the input and output data of the nonlinear power amplifier in order to identify its inverse characteristics. Fig. 14 is a block diagram showing the structure of a device for collecting input and output data of a power amplifier. It should be pointed out that the power amplifier here is described by taking the power amplifier commonly used by the transmitters of the base station and the mobile station in the communication system as an example. Of course, the power amplifier can also be applied to other occasions.

As shown in Figure 14, the device for collecting the input and output data of the power amplifier includes: a digital-to-analog converter 101, which is used to convert the digital input signal x(n) into an analog signal (baseband signal); an up-converter 102 , for upconverting the baseband signal output from the digital-to-analog converter 101 into a radio frequency signal; the power amplifier 103 is used to amplify the power of the radio frequency signal output by the up converter 102; the directional coupler 104 is used to convert the power amplifier 103 The output branch of is two-way signal, and one-way signal is output as output signal y (n), and another way signal is fed back to attenuator 105; Attenuator 105 is used for receiving the output of power amplifier 103 fed back by directional coupler 104, and to The feedback signal is attenuated; the down converter 106 is used to down-convert the attenuated signal output by the attenuator 105 into a baseband signal; the analog-to-digital converter 107 is used to convert the baseband signal obtained by down-converting the down converter 106 into The digital signal, namely, the feedback output signal y _G (n); and the data collection and storage module 100, used to collect and store the original input signal x(n) and the feedback output signal y _G (n). The original input signal x(n) and the feedback output signal _yG (n) stored in the data collection storage module 100 are expressed in vector form as follows:

X = [x(1), x(2), ..., x(n)] (21)

Y _G = [y _G (1), y _G (2), ..., y _G (n)] (22)

After using the device shown in Figure 14 to collect the original input signal x(n) and feedback output signal _yG (n) of the power amplifier, you can use the nonlinear system inverse characteristic identification method shown in Figure 12 or the The nonlinear system inverse characteristic identification device identifies the inverse characteristic of the power amplifier, thereby obtaining the filter function F{·} and the look-up table function Q{·}.

After obtaining the filter function F{ } and the lookup table function Q{ }, the pre-distortion module of the power amplifier can be constructed and used to pre-distort the input signal x(n) of the power amplifier, so that the power amplifier can obtain a good linear output. Fig. 15 shows a structural block diagram of a nonlinear power amplifying device capable of realizing predistortion according to an embodiment of the present invention.

As shown in FIG. 15 , the non-linear power amplification device capable of pre-distortion includes: a pre-distortion module 400, configured to pre-distort an input signal x(n), and output a pre-distorted signal; a digital-to-analog converter 405 configured to The predistortion signal output by the predistortion module 400 is converted into an analog signal (baseband signal); the upconverter 406 is used to upconvert the baseband signal output from the digital-to-analog converter 405 into a radio frequency signal; the power amplifier 407 is used to It is used to amplify the power of the radio frequency signal output by the up-converter 406 and output the amplified output signal y(n). At this time, the signal y(n) output by the power amplifying device has eliminated the non-memory nonlinear characteristic and the memory characteristic of the power amplifier, and has good linearity.

The pre-distortion module 400 shown in Figure 15 includes: an address generator 401, which is used to convert the input signal x(n) into a look-up table address after modulo and quantization; a one-dimensional look-up table 402, according to the above-mentioned identification The lookup table function Q{ } obtained by the method or identification device is configured, and the correction factor is output according to the lookup table address output by the address generator 401; the multiplier 403 is used to combine the original input signal x(n) with the one-dimensional query The correction factors output by table 402 are multiplied to obtain the intermediate predistortion signal; and the filter 404 is configured according to the filter function F{ } obtained by using the identification method or identification device described above, and the output of the multiplier 403 The intermediate pre-distortion signal is filtered to complete the pre-distortion, and the pre-distortion signal is output to the digital-to-analog converter 405 .

In addition, the directional coupler 104, the attenuator 105, the down converter 106, the analog-to-digital converter 107, and the data collection and storage module 100 in the device for collecting the input and output data of the power amplifier shown in FIG. into the power amplifying device shown in FIG. 15 to constitute a power amplifying device having both a data collection and storage function and a pre-distortion function.

According to the non-linear power amplification device with pre-distortion module configured as above, the pre-distortion is realized in the baseband by using digital technology, which can not only compensate for the non-memory nonlinear characteristics and memory characteristics of the power amplifier, but also effectively reduce the capacity requirements for storage devices And greatly reduce the requirements for numerical calculation.

What is described above is to use the previously collected and stored input signal x(n) and feedback output signal y _G (n) of the power amplifier to conduct offline identification of the inverse characteristics of the power amplifier, and then configure the pre-distortion module of the power amplifier according to the identification results.

However, as mentioned above, as a nonlinear system, the output characteristics of the nonlinear power amplifier will change with time. Although this change is not significant in a short period of time, its characteristics will inevitably change as time goes on. Therefore, it is best to regularly identify the inverse characteristics of the power amplifier so that its output always shows a good linear form.

In order to avoid the cumbersome work of regularly identifying the inverse characteristics of the power amplifier and periodically updating the parameters of the pre-distortion module, this embodiment also proposes a power amplification method and device for online updating and upgrading the parameters of the pre-distortion module. As shown in FIG. 16 , it shows a schematic structural diagram of a power amplification device capable of updating parameters of an upgraded predistortion module online according to the present invention.

The power amplifying device shown in FIG. 16 includes a predistortion module 400, a digital-to-analog converter 405, an up-converter 406, a power amplifier 407, a directional coupler 104, an attenuator 105, a down-converter 106, an analog-to-digital converter 107, a pre- Distortion data updater 600, and delayer 601. Except for the predistortion data updater 600 and the delayer 601, the other units have the same structure as the corresponding units shown in Fig. 14 and Fig. 15, and are denoted by the same reference numerals, and their detailed descriptions are omitted here.

In the power amplifying device with an online update predistortion module shown in FIG. 16 , first the original digital input signal x(n) is input to the predistortion module 400, and converted into a lookup table address after modulo and quantization by the address generator 401. The correction factor is read from the one-dimensional lookup table 402 according to the lookup table address, and multiplied by the original digital input signal x(n) in the multiplier 403 . The multiplied result is then filtered by the filter 404 to complete pre-distortion processing and generate a pre-distortion signal. Next, the predistortion signal is converted into a radio frequency analog signal by the digital-to-analog converter 405 and the up-converter 406, and sent to the power amplifier 407 for power amplification and outputting an amplified signal y(n).

The amplified output signal y(n) is fed back through the directional coupler 104, then attenuated by the attenuator 105, and then the down-converter 106 down-converts the signal to the baseband, and then the baseband signal is converted into a digital signal by the analog-to-digital converter 107 , and input to the predistortion data updater 600.

One of the original digital input signals x(n) is delayed by the delayer 601 and then input to the predistortion data updater 600 . The predistortion data updater 600 is equivalent to the identification module 300 shown in FIG. 11 , and updates the filter function F{·} and the lookup table function Q{·} in the predistortion module 400 according to the following algorithm.

It can be deduced from Figure 16 that the error signal can be obtained by comparing the original digital input signal x(n) with the fed back digital output signal:

$\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; [</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span><span\; class="notranslate">\; \}</span>$

$<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; \{</span><span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; [</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; (</span><span\; class="notranslate">\; [</span><span\; class="notranslate">\; |</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; |</span><span\; class="notranslate">\; ]</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span><span\; class="notranslate">\; \}</span>$

$<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span>\frac{\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; G</span>}}<span\; class="notranslate">\; \&Center\; Dot;</span>\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; \{</span>\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; +</span>\underset{\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\overset{\mathrm{<span\; class="notranslate">\; L</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}{\mathrm{<span\; class="notranslate">\; \&Sigma;</span>}}}{\mathrm{<span\; class="notranslate">\; w</span>}}_{\mathrm{<span\; class="notranslate">\; l</span>}}\mathrm{<span\; class="notranslate">\; z</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; l</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \}</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; twenty\; three</span>}<span\; class="notranslate">\; )</span>$

where P{ } represents the power amplifier function, and

z(n)=x(n)·Q([|x(n)|]) (24)

The query table function Q{ } can be written

Q([|x(n)|])=[q ⁰ ,q ¹ ,...,q ^K-1 ] ^T (25)

The filter coefficients can be organized as

W=[w ₁ , w ₂ , . . . , w _L-1 ] ^T (26)

L still stands for memory depth.

The lookup table function and filter coefficients can then be updated as follows:

$\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\end{array}\right]<span\; class="notranslate">\; =</span>\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; q</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}^{\mathrm{<span\; class="notranslate">\; k</span>}}\\ {\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; i</span>}}\end{array}\right]<span\; class="notranslate">\; +</span>\left[\begin{array}{c}{\mathrm{<span\; class="notranslate">\; \&mu;</span>}}_{\mathrm{<span\; class="notranslate">\; q</span>}}\\ \mathrm{<span\; class="notranslate">\; u</span>}\end{array}\right]<span\; class="notranslate">\; \&CenterDot;</span>\mathrm{<span\; class="notranslate">\; e</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; no</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; 27</span>}<span\; class="notranslate">\; )</span>$

in

U=[μ ₁ , μ ₁ , . . . , μ _L-1 ] ^T (28)

Both U and _μq are convergence step sizes.

Of course, the predistortion data updater 600 can also use the newly acquired original input signal x(n) and feedback output signal _yG (n) to conduct online identification of the inverse characteristic of the power amplifier according to the above-mentioned nonlinear inverse characteristic identification method , to obtain the real-time query table function Q{·} and filter function F{·}.

After obtaining the new look-up table function and filtering function, the corresponding parameters in the pre-distortion module 400 can be modified, and the steps performed by the pre-distortion module 400 and the pre-distortion data updater 600 shown in FIG. 16 are omitted here. A detailed description of the process flow of the processing method.

According to the identification device and identification method disclosed in this embodiment, not only the requirements for storage devices and numerical calculations can be reduced simultaneously, but also the online update of the parameters of the predistortion module can be realized, thereby ensuring the linearity of the power amplifier in real time.

Taking the nonlinear power amplifier as an example above, how to identify the time-varying nonlinear system offline and online inverse characteristics is described in detail, and pre-distort the nonlinear system according to the identification result.

As mentioned above, the present invention uses a one-dimensional look-up table in series with a finite impulse response (FIR) filter to describe the nonlinear characteristics of the power amplifier, thereby avoiding the need to accurately describe many high-order terms used in the memory polynomial model Due to the nonlinear characteristics of the power amplifier, the use of high-order terms often leads to a high condition number of the corresponding matrix, which makes it difficult to accurately obtain memory polynomial parameters and pre-distortion. Moreover, the power amplifier predistorter and the power amplifying device using the power amplifier inverse characteristic identification device and the predistorter according to the present invention use a one-dimensional look-up table, which greatly saves memory requirements.

In addition, it should also be noted that the series of processes and devices described above may also be implemented by software and/or firmware. In the case of realization by software and/or firmware, the program constituting the software is installed from a storage medium or a network to a computer having a dedicated hardware configuration, such as a general-purpose personal computer 700 shown in FIG. , can perform various functions and so on.

In FIG. 17 , a central processing unit (CPU) 701 executes various processes according to programs stored in a read only memory (ROM) 702 or programs loaded from a storage section 708 to a random access memory (RAM) 703 . In the RAM 703, data required when the CPU 701 executes various processing and the like is also stored as necessary.

The CPU 701, ROM 702, and RAM 703 are connected to each other via a bus 704. The input/output interface 705 is also connected to the bus 704 .

The following components are connected to the input/output interface 705: an input section 706 including a keyboard, a mouse, etc.; an output section 707 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker; The storage section 708 includes a hard disk and the like; and the communication section 709 includes a network interface card such as a LAN card, a modem, and the like. The communication section 709 performs communication processing via a network such as the Internet.

A driver 710 is also connected to the input/output interface 705 as needed. A removable medium 711 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 710 as necessary, so that a computer program read therefrom is installed into the storage section 708 as necessary.

In the case of realizing the above-described series of processing by software, the programs constituting the software are installed from a network such as the Internet or a storage medium such as the removable medium 711 .

Those skilled in the art should understand that such a storage medium is not limited to the removable medium 711 shown in FIG. 17 in which the program is stored and distributed separately from the device to provide the program to the user. Examples of the removable media 711 include magnetic disks (including floppy disks (registered trademark)), optical disks (including compact disk read only memory (CD-ROM) and digital versatile disks (DVD)), magneto-optical disks (including )) and semiconductor memory. Alternatively, the storage medium may be a ROM 702, a hard disk contained in the storage section 708, or the like, in which the programs are stored and distributed to users together with devices containing them.

It should also be pointed out that the steps for executing the above series of processes can naturally be executed in chronological order according to the illustrated order, but it does not need to be executed in chronological order. Certain steps may be performed in parallel or independently of each other.

Although the present invention and its advantages have been described in detail, it should be understood that various changes, substitutions and alterations can be made hereto without departing from the spirit and scope of the invention as defined by the appended claims. Moreover, the terms "comprising", "comprising", or any other variation thereof in this application are intended to cover a non-exclusive inclusion such that a process, method, article, or apparatus that includes a set of elements includes not only those elements, but also includes none. other elements specifically listed, or also include elements inherent in such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus comprising said element.

Claims (29)

1. an identifying inverse characteristic of nonlinear system method is used for original input signal x (n) and corresponding feedback output signal y according to non linear system _G(n) the contrary characteristic of non linear system is carried out identification, comprising:

Question blank function Q { } generates step, according to liftering signal x _F(n) and feedback loop output signal y _G(n) calculate question blank function Q { };

Middle pre-distorted signals z (n) generates step, according to feedback loop output signal y _G(n) and question blank function Q { } generate in the middle of pre-distorted signals z (n);

Filter function F{} constitution step is according to middle pre-distorted signals z (n) and original input signal x (n) structure filter function F{};

The liftering step utilizes filter function F{} according to formula x _F(n)=F ^-1{ x (n) } carries out liftering to original input signal x (n) and generates liftering signal x _F(n); And

Iteration repeats above each step till satisfying the condition of setting;

Wherein, question blank function Q { } and filter function F{} represent the contrary characteristic of non linear system;

Wherein, the condition of said setting is to confirm standard mean square error NMSE convergence shown in the following formula or iterations greater than predetermined threshold,

$\mathrm{NMSE}\left(\mathrm{dB}\right)=10{\log }_{10}\left(\frac{\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{|x\left(n\right)-\hat{x}\left(n\right)|}^2}{\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\left|x\left(n\right)\right|}^2}\right)$

Formula

according to the following formula using the filter function F {·} and the intermediate pre-distortion signal z (n) to generate the original input signal x (n) the estimated value of

$\hat{x}\left(n\right)=F\left\{z\left(n\right)\right\}.$

2. identifying inverse characteristic of nonlinear system method according to claim 1, wherein in first time during generated query table function Q{}, with original input signal x (n) as liftering signal x _F(n) with feedback loop output signal y _G(n) calculate question blank function Q { } together.

3. identifying inverse characteristic of nonlinear system method according to claim 1 also comprises delivery and quantization step, according to following formula with feedback loop output signal y _G(n) be quantified as some address k in the question blank,

k＝[|y _G(n)|] k＝0，...，K-1

Wherein, [||] represented delivery and quantification.

4. identifying inverse characteristic of nonlinear system method according to claim 3, wherein question blank function Q { } generation step comprises:

According to formula gp (n)=x _F(n)/y _G(n) calculate the gain of each sampling point;

The gp (n) that will have identical address is divided into a group according to following formula,

GP _k＝[gp(i)，...gp(j)] 0＜k＜K-1，

1＜i，j＜N

So that it is corresponding with an address k in the question blank; And

Obtain the gain of each group, thus generated query table function Q{}.

5. identifying inverse characteristic of nonlinear system method according to claim 4, the step that wherein obtains the gain of each group comprises:

Gp (n) in each group is made even all to mould value and phase place respectively according to following formula,

g _k＝mean(|GP _k|)

p _k＝mean(∠GP _k)

Thereby obtain question blank function Q { }

Q{·}＝Gain?⊙exp(j·Phase)

Wherein

Gain＝[g ₀，...，g _i，...] ^T

Phase＝[p ₀，...p _i，...] ^T

⊙ representes the Hadamard product, and subscript T representes matrix transpose.

6. identifying inverse characteristic of nonlinear system method according to claim 5 wherein generates in the step at middle pre-distorted signals z (n), uses feedback loop output signal y according to following formula _G(n) and question blank function Q { } generate in the middle of pre-distorted signals z (n):

z(n)＝y _G(n)·Q([|y _G(n)|])。

7. identifying inverse characteristic of nonlinear system method according to claim 6 wherein in filter function F{} constitution step, is used middle pre-distorted signals z (n) and original input signal x (n) structure filter factor according to following formula

W＝(Z _M ^HZ _M) ^-1·Z _M ^H·X

Wherein

X＝[x(1)，x(2)，...，x(N)] ^T

$Z_M=\left(\begin{array}{cccc}z\left(1\right)& 0& \&CenterDot;\&CenterDot;\&CenterDot;& 0\\ z\left(2\right)& z\left(1\right)& \&CenterDot;\&CenterDot;\&CenterDot;& 0\\ z\left(3\right)& z\left(2\right)& \&CenterDot;\&CenterDot;\&CenterDot;& 0\\ \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;\\ & & \&CenterDot;\&CenterDot;\&CenterDot;& z(N-L+2)\\ z\left(N\right)& z(N-1)& \&CenterDot;\&CenterDot;\&CenterDot;& z(N-L+1)\end{array}\right)$

N is the length of the data of collection, and L is the exponent number of filtering, thereby filter function F{} is expressed as with matrix and vector form:

F{z(n)}＝Z _M·W。

8. identifying inverse characteristic of nonlinear system method according to claim 7 is wherein exported the contrary characteristic as non linear system with question blank function Q { } and filter function F{} when satisfying the condition of setting.

9. according to the arbitrary described identifying inverse characteristic of nonlinear system method of claim 1～8, wherein feedback loop output signal y _G(n) generate according to following formula

y _G(n)＝y(n)/G

G is the expected gain of non linear system, and y (n) is the output of non linear system.

10. identifying inverse characteristic of nonlinear system method according to claim 9 also comprises data collection step, is used for non linear system is applied specific original input signal x (n) and obtains corresponding output signal y (n).

11. an identifying inverse characteristic of nonlinear system device is used for original input signal x (n) and corresponding feedback output signal y according to non linear system _G(n) the contrary characteristic of non linear system is carried out identification, comprising:

Parameter calculator is used for the contrary characteristic of non linear system is carried out identification, generates the question blank function Q { } and the filter function F{} of the contrary characteristic of expression non linear system; And

The convergence decision device is used for the condition whether question blank function Q { } that the critical parameter calculator generated and filter function F{} satisfy setting, and Control Parameter calculator iterative computation question blank function Q { } and filter function F{};

Wherein, said parameter calculator comprises:

The question blank maker is used for according to liftering signal x _F(n) and feedback loop output signal y _G(n) calculate question blank function Q { };

Middle predistorter is used for according to feedback loop output signal y _G(n) and the question blank function Q { } calculated of question blank maker generate in the middle of pre-distorted signals z (n);

The filtering parameter calculator is used for the middle pre-distorted signals z (n) and original input signal x (n) the structure filter function F{} that generate according to middle predistorter; And

Inverse filter is used to utilize filter function F{} that the filtering parameter calculator constructed according to formula x _F(n)=F ^-1{ x (n) } carries out liftering to original input signal x (n) and generates liftering signal x _F(n).

12. identifying inverse characteristic of nonlinear system device according to claim 11 also comprises selector, is used for selecting original input signal x (n) as liftering signal x in first time during generated query table function Q{} _F(n) be input to the question blank maker, make the question blank maker according to original input signal x (n) and feedback loop output signal y _G(n) calculate question blank function Q { }, and after during generated query table function Q{}, select the liftering signal x that inverse filter generated _F(n) be input to the question blank maker, make the question blank maker according to liftering signal x _F(n) and feedback loop output signal y _G(n) calculate question blank function Q { }.

13. identifying inverse characteristic of nonlinear system device according to claim 12; Also comprise: filter is used for the estimated value

of using filter function F{} that the filtering parameter calculator constructed and middle pre-distorted signals z (n) to generate original input signal x (n) according to formula

Standard mean square error NMSE calculator is used for according to following formula basis of calculation mean square error NMSE

$\mathrm{NMSE}\left(\mathrm{dB}\right)=10{\log }_{10}\left(\frac{\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{|x\left(n\right)-\hat{x}\left(n\right)|}^2}{\underset{n=1}{\overset{N}{\mathrm{\&Sigma;}}}{\left|x\left(n\right)\right|}^2}\right)$

Wherein, restrain whether decision device restrains according to standard mean square error NMSE or whether iterations judges whether satisfy the condition of setting greater than predetermined threshold.

14. identifying inverse characteristic of nonlinear system device according to claim 13 also comprises address generator, is used for according to following formula feedback loop output signal y _G(n) be quantified as some address k in the question blank that the question blank maker generated,

k＝[|y _G(n)|] k＝0，...，K-1

Wherein, [||] represented delivery and quantification.

15. identifying inverse characteristic of nonlinear system device according to claim 14, wherein the question blank maker is configured to

According to formula gp (n)=x _F(n)/y _G(n) calculate the gain of each sampling point;

The gp (n) that will have identical address is divided into a group according to following formula,

GP _k＝[gp(i)，...gp(j)] 0＜k＜K-1，

1＜i，j＜N

So that it is corresponding with an address k in the question blank; And

Obtain the gain of each group, thus generated query table function Q{}.

16. identifying inverse characteristic of nonlinear system device according to claim 15, wherein the question blank maker is configured to

Gp (n) in each group is made even all to mould value and phase place respectively according to following formula,

g _k＝mean(|GP _k|)

p _k＝mean(∠GP _k)

Thereby obtain question blank function Q { }

Q{·}＝Gain?⊙exp(j.Phase)

Wherein

Gain＝[g? ₀，...，g _i，...] ^T

Phase＝[p ₀，...p _i，...] ^T

⊙ representes the Hadamard product, and subscript T representes matrix transpose.

17. identifying inverse characteristic of nonlinear system device according to claim 16, wherein middle predistorter uses feedback loop output signal y according to following formula _G(n) and the question blank function Q { } calculated of question blank maker generate in the middle of pre-distorted signals z (n):

z(n)＝y _G(n)·Q([|y _G(n)|])。

18. identifying inverse characteristic of nonlinear system device according to claim 17, predistorter was generated in the middle of wherein the filtering parameter calculator used according to following formula middle pre-distorted signals z (n) and original input signal x (n) structure filter coefficient

W＝(Z _M ^HZ _M) ^-1·Z _M ^H·X

Wherein

X＝[x(1)，x(2)，...，x(N)] ^T

$Z_M=\left(\begin{array}{cccc}z\left(1\right)& 0& \&CenterDot;\&CenterDot;\&CenterDot;& 0\\ z\left(2\right)& z\left(1\right)& \&CenterDot;\&CenterDot;\&CenterDot;& 0\\ z\left(3\right)& z\left(2\right)& \&CenterDot;\&CenterDot;\&CenterDot;& 0\\ \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;& \&CenterDot;\&CenterDot;\&CenterDot;\\ & & \&CenterDot;\&CenterDot;\&CenterDot;& z(N-L+2)\\ z\left(N\right)& z(N-1)& \&CenterDot;\&CenterDot;\&CenterDot;& z(N-L+1)\end{array}\right)$

N is the length of the data of collection, and L is the exponent number of filtering, thereby filter function F{} is expressed as with matrix and vector form:

F{z(n)}＝Z _M·W。

19. identifying inverse characteristic of nonlinear system device according to claim 18 is wherein exported the contrary characteristic as non linear system with question blank function Q { } and filter function F{} when satisfying the condition of setting.

20. according to the arbitrary described identifying inverse characteristic of nonlinear system device of claim 11～19, wherein feedback loop output signal y _G(n) generate according to following formula

y _G(n)＝y(n)/G

G is the expected gain of non linear system, and y (n) is the output of non linear system.

21. identifying inverse characteristic of nonlinear system device according to claim 20 also comprises the data collection memory device, is used for non linear system is applied specific input signal x (n) and obtains corresponding output signal y (n).

22. a power amplifier pre-distortion device comprises:

Address generator is used for original input signal is carried out being converted into the question blank address after delivery, the quantification;

The one dimension question blank is configured according to question blank function Q { }, and according to the question blank address output calibration factor of address generator output;

Multiplier is used for the correction factor of original input signal and the output of one dimension question blank being multiplied each other pre-distorted signals in the middle of obtaining; And

Filter, F{} is configured according to filter function, and the middle pre-distorted signals that multiplier is exported is carried out filtering, generates pre-distorted signals;

Wherein, question blank function Q { } and filter function F{} generate according to the arbitrary described identifying inverse characteristic of nonlinear system method of claim 1～8 or generate according to the arbitrary described identifying inverse characteristic of nonlinear system device of claim 11～19.

23. power amplifier pre-distortion device according to claim 22 also comprises the data collection storage device, is used to collect the original input signal and the feedback loop output signal of power amplifier, this data collection storage device comprises:

Digital to analog converter is used for converting original input signal into analog signal;

Upconverter is used for the analog signal from digital to analog converter output is up-converted to radiofrequency signal;

Power amplifier is used for the radiofrequency signal of upconverter output is carried out power amplification;

Directional coupler is used for the output of power amplifier is branched into two paths of signals, and one road signal is exported as the output signal, and another road signal feedback is given attenuator;

Attenuator is used to receive the signal that directional coupler feeds back, and this feedback signal is decayed;

Low-converter is used for the deamplification of attenuator output is down-converted to baseband signal;

Analog to digital converter is used for converting the baseband signal of low-converter output into digital signal, as feedback loop output signal; And

The data collection memory module is used for collecting storage original input signal and feedback loop output signal.

24. a power amplification device comprises:

Predistortion module is used for original input signal is carried out predistortion, the output predistorted input signal;

Digital to analog converter is used for converting the predistorted input signal that predistortion module is exported into analog signal;

Upconverter is used for the analog signal from digital to analog converter output is up-converted to radiofrequency signal;

Power amplifier is used for the radiofrequency signal of upconverter output is carried out signal after power amplification and power output are amplified as the output signal;

Wherein, the structure of said predistortion module is identical with the structure of power amplifier pre-distortion device according to claim 22.

25. power amplification device according to claim 24 also comprises:

Directional coupler is used for the output of power amplifier is branched into two paths of signals, and one road signal is exported as the output signal, and another road signal feedback is given attenuator;

Attenuator is used to receive the signal that directional coupler feeds back, and this feedback signal is decayed;

Low-converter is used for the deamplification of attenuator output is down-converted to baseband signal;

Analog to digital converter is used for converting the baseband signal of low-converter output into digital signal, as feedback loop output signal; And

The data collection memory module is used for collecting storage original input signal and feedback loop output signal.

26. a power amplification device comprises:

Predistortion module, its structure is identical with the structure of power amplifier pre-distortion device according to claim 22, is used for original input signal is carried out predistortion the output predistorted input signal;

Digital to analog converter is used for converting the predistorted input signal that predistortion module is exported into analog signal;

Upconverter is used for the analog signal from digital to analog converter output is up-converted to radiofrequency signal;

Power amplifier is used for the radiofrequency signal of upconverter output is carried out signal after power amplification and power output are amplified as the output signal;

Directional coupler is used for the output of power amplifier is branched into two paths of signals, and one road signal is exported as the output signal, and another road signal feedback is given attenuator;

Attenuator is used to receive the signal that directional coupler feeds back, and this feedback signal is decayed;

Low-converter is used for the deamplification of attenuator output is down-converted to baseband signal;

Analog to digital converter is used for converting the baseband signal of low-converter output into digital signal, as feedback loop output signal;

Delayer is used for original input signal is delayed time, and generates delay input signal; And

The predistorting data renovator; Be used to receive the delay input signal that feedback loop output signal that analog to digital converter exports and delayer are exported, and question blank function and filter coefficient carried out online updating according to feedback loop output signal that is received and delay input signal.

27. power amplification device according to claim 26, wherein the predistorting data renovator is configured to question blank function and the filter function according to the arbitrary described identifying inverse characteristic of nonlinear system method online updating predistortion module of claim 1～8.

28. power amplification device according to claim 26, wherein the predistorting data renovator is configured to the question blank function and the filter function of online updating predistortion module in the following manner:

Relatively original input signal x (n) and feedback loop output signal, according to following formula calculation error signal:

$e\left(n\right)=x\left(n\right)-\frac{1}{G}\&CenterDot;P\left\{F\right\{x\left(n\right)\&CenterDot;Q\left(\right[\left|x\left(n\right)\right|\left]\right)\left\}\right\}$

$=x\left(n\right)-\frac{1}{G}\&CenterDot;P\left\{\right\{x\left(n\right)\&CenterDot;Q\left(\right[\left|x\left(n\right)\right|\left]\right)+\underset{l=1}{\overset{L-1}{\mathrm{\&Sigma;}}}{w}_{l}x(n-l)\&CenterDot;Q\left(\right[\left|x(n-l)\right|\left]\right)\left\}\right\}$

$=x\left(n\right)-\frac{1}{G}\&CenterDot;P\{z\left(n\right)+\underset{l=1}{\overset{L-1}{\mathrm{\&Sigma;}}}{w}_{l}z(n-l)\}$

Wherein, F{}, Q{}, P{} represent filter function, question blank function and power amplifier function respectively, and

z(n)＝x(n)·Q([|x(n)|])，

Question blank function Q { } does

Q([|x(n)|])＝[q ⁰，q ¹，...，q ^K-1] ^T

Filter coefficient does

W＝[w ₁，w ₂，...，w _L-1] ^T

L represents the memory depth of filter, and [||] represented delivery and quantification; And

According to following formula online updating question blank function and filter coefficient

$\left[\begin{array}{c}{q}_{i+1}^k\\ W_{i+1}\end{array}\right]=\left[\begin{array}{c}{q}_i^k\\ W_i\end{array}\right]+\left[\begin{array}{c}{\mathrm{\&mu;}}_q\\ U\end{array}\right]\&CenterDot;e\left(n\right)$

Wherein

U＝[μ ₁，μ ₁，...，μ _L-1] ^T

U and μ _qBe the convergence step-length.

29. the method for the parameter of the predistortion module of an online updating power amplification device comprises step:

Relatively the original input signal x (n) and the feedback loop output signal of power amplification device, according to following formula calculation error signal:

$e\left(n\right)=x\left(n\right)-\frac{1}{G}\&CenterDot;P\left\{F\right\{x\left(n\right)\&CenterDot;Q\left(\right[\left|x\left(n\right)\right|\left]\right)\left\}\right\}$

$=x\left(n\right)-\frac{1}{G}\&CenterDot;P\left\{\right\{x\left(n\right)\&CenterDot;Q\left(\right[\left|x\left(n\right)\right|\left]\right)+\underset{l=1}{\overset{L-1}{\mathrm{\&Sigma;}}}{w}_{l}x(n-l)\&CenterDot;Q\left(\right[\left|x(n-l)\right|\left]\right)\left\}\right\}$

$=x\left(n\right)-\frac{1}{G}\&CenterDot;P\{z\left(n\right)+\underset{l=1}{\overset{L-1}{\mathrm{\&Sigma;}}}{w}_{l}z(n-l)\}$

Wherein, F{}, Q{}, P{} represent the filter function of filter, the question blank function of question blank and the magnification function of power amplifier respectively, and

z(n)＝x(n)·Q([|x(n)|])，

Question blank function Q { } does

Q([|x(n)|])＝[q ⁰，q ¹，...，q ^K-1] ^T

Filter coefficient does

W＝[w ₁，w ₂，...，w _L-1] ^T

L represents the memory depth of filter, and [||] represented delivery and quantification; And

According to following formula online updating question blank function and filter coefficient

$\left[\begin{array}{c}{q}_{i+1}^k\\ W_{i+1}\end{array}\right]=\left[\begin{array}{c}{q}_i^k\\ W_i\end{array}\right]+\left[\begin{array}{c}{\mathrm{\&mu;}}_q\\ U\end{array}\right]\&CenterDot;e\left(n\right)$

Wherein

U＝[μ ₁，μ ₁，...，μ _L-1] ^T

U and μ _qBe the convergence step-length.

Images