RU2786412C1 - Method for compensating nonlinear distortions of high-frequency power amplifiers and a device for its implementation - Google Patents

Abstract

FIELD: radio engineering.

SUBSTANCE: invention relates to the field of radio engineering. By subtracting the input signal from the output, a distortion signal is isolated and the functional dependence of these distortions on the input signal is determined. Using this functional dependence in the form of a separate parameterized function of pre-orders, a compensating signal is synthesized, which is mixed into the input of a power amplifier. At the same time, in order to calculate the vector of coefficients of the pre-distortion function, the input signal amplitude is quantized by M levels and the coefficients for each quantization level are determined as a sample covariance coefficient normalized relative to the input signal power.

EFFECT: reducing the level of residual distortion at the output of the power amplifier.

2 cl, 1 dwg

Description

Technical field

The invention relates to the field of radio engineering, in particular to radio transmitters for creating power amplifiers with a low level of out-of-band radiation used in radio communications and television and radio broadcasting.

State of the art

Intensive use of the radio spectrum exacerbates the EMC problem of radio equipment operating in adjacent frequency bands. The main source of interference in such conditions are out-of-band emissions, which occur mainly in the output high-frequency power amplifiers (PA). The reduced level of out-of-band radiations has PA operating in class A mode, however, they have low efficiency and are not economically efficient. Energy-efficient PAs with high efficiency operate in non-linear modes and create an unacceptable level of out-of-band emissions.

Methods and devices are known for linearizing such amplifiers or for compensating out-of-band radiation. Patent RU 2731135 "Radio transmitting device with digital nonlinearity correction" describes a device for minimizing the amplitude and phase distortions of the transmitted signal under conditions of rapidly changing and a priori uncertain non-linearity of the radio path transfer characteristic. The device contains a digital non-linearity corrector using a weighted sum of E-polynomials, and a combinational components meter that calculates the amplitude and phase of distortion using the Fourier transform. The resulting distortion amplitudes are normalized to the input signal, and the phases are rotated by 180º, after which the distortions are fed to a digital non-linearity corrector, in which, using E-polynomials, a transfer characteristic is formed that compensates for the distortions introduced by the PA. The formation of the transfer characteristic of the digital corrector is carried out on a test (two-tone) signal in the absence of an information signal. The disadvantages of the presented device are the need to measure and preliminary describe the transfer characteristic of the digital nonlinear corrector using E-polynomials, as well as the need for periodic calibration of the device with a test signal with interruption of the transmission of the information signal.

In the proceedings of the 2009 IEEE International Symposium on Radio-Frequency Integration Technology, a Power Amplifier Linearization Using Feedforward Technique for Wide Band Communication System is described. In this method, the PA distortion compensation is carried out in two stages. At the first stage, after equalizing the amplitudes, phases and delays of the input and output signals of the PA, the input signal is subtracted from the output signal, as a result, signal distortions are highlighted at the output of the subtractor. At the second stage, the selected distortions are subtracted from the PA output signal, thereby excluding them from the PA output signal. For the successful implementation of this method at each stage, it is necessary to accurately align the amplitudes, phases and delays of the interacting signals. Devices that implement this method contain at least two servo control loops that provide equalization of the signal parameters. In this method, distortion compensation occurs at the PA output, therefore, before subtracting, the distortion must be amplified to the required level by a linear PA, which leads to a decrease in the efficiency of the device as a whole. In addition, the need to use at least two circuits reduces the stability of the device.

The prototype of the invention is the method and device described in patent application 2020143494/07(081150). In accordance with it, the compensation of non-linear distortions in the PA is carried out in two stages. At the first stage, after equalizing the amplitudes, phases and delays of the input and output signals of the PA, the input signal is subtracted from the output. As a result, distortions are highlighted. By observing the input signal and the distortion, the functional dependence of the distortion signal on the input signal is determined. After the functional dependence is determined, it is isolated as a separate parameterized function - the predistortion function. Then, using the amplitude of the current input signal as an argument of this function, a compensating signal is synthesized, which is a model of distortion in the power amplifier. In the second step, the compensating signal is subtracted from the input signal.

The functional dependence of the distortion un on the input signal xin at a discrete time n is approximated as the product of the input signal and the parameterized function of the input signal amplitude f(|xin(n)|,Λ). Mathematically, when representing signals in complex form, this can be represented as:

where u̇n is the synthesized distortion signal;
n - discrete moment of time;
ẋin - input signal;
f(|ẋin(n)|,Λ) – parameterized function (pre-distortion function), relates the distortion signal to the amplitude of the input signal;
Λ - row vector of coefficients of the predistortion function.

Here and below, it is assumed that the signal processing is performed in digital form, for example, in the form of quadrature components I(t) and Q(t).

The functional dependence in the prototype is proposed to be presented in the form of a lookup table with interpolation of the degree p. In this case, it is proposed to search for the coefficients λ in the table by the least squares method (LSM), solving an overdetermined system of linear equations of the form

where A is the matrix of instantaneous values of the input signal;
AH is the Hermitian conjugate matrix;
Λ=[λ1, λ2, … λm, … λM]T - column vector of coefficients of the functional dependence of the distortion signal on the input signal;
b is a column vector of instantaneous values of the distortion signal.

The device that implements the method described in the prototype contains the first adder and PA connected in series. The input of the adder is the input of the device, and the output of the PA is the output of the device. The device also contains a control loop for mutual equalization of the amplitudes, phases and delays of the input and output signals. In this case, one input of the control loop is connected to the output of the device, and the other - to the input of the device. The device contains a second adder, the subtractive input of which is connected to the first output of the control loop, and the other input of the adder is connected to the second output of the control loop. In this case, respectively, at the first and second outputs of the control loop, the input and output signals of the PA are formed, balanced in amplitude, phase and delay. The device also contains a block for calculating the parameters of the pre-distortion function, a key, a block for the pre-distortion function and a phase equalization block connected in series. In this case, the output of the phase alignment unit is connected to the second input of the first adder. The second input of the predistortion function block is connected to the input of the device. The first input of the block for calculating the parameters of the predistortion function is connected to the first output of the control loop, and the second input is connected to the output of the second adder.

In the device, after equalization in the loop for regulating the amplitudes, phases and delays of the input and output signals, they go to the second adder, at the output of which a distortion signal is allocated. Using the distortion signal and the input signal, in the block for calculating the parameters of the predistortion function, a vector Λ is calculated containing the coefficients of the functional dependence of the distortion signal on the input signal, which is transmitted through a closed key to the block for calculating the parameterized function. In this block, the parameterized function f(|xin(n)|,Λ) is calculated for the current value of the input signal amplitude, and the value of the synthesized distortion signal is formed at the output of the block in accordance with expression (1). Then, in the phase equalization block, the phase of the synthesized distortion signal is changed so that in the first adder the compensating signal is subtracted from the input signal of the device. By introducing synthesized pre-distortions at the PA input in antiphase to the expected distortions in the PA, the distortions are compensated at its output, which corresponds to an increase in the linearity of the amplitude and phase-amplitude characteristics of the PA.

Thus, to perform signal conversion operations according to the method described in the prototype when determining the coefficients of the functional dependence of the distortion signal on the input signal, it is necessary to solve an overdetermined system of equations in matrix form, which requires large computational, time and energy costs.

Disclosure of the essence of the invention

In accordance with the proposed method, the compensation of non-linear distortions in the PA is performed in two stages. At the first stage, after equalizing the amplitudes, phases and delays of the input and output signals, the input signal is subtracted from the output signal. As a result, a distortion signal is emitted. By observing the input signal and the distortion signal, the functional dependence of the distortion on the input signal is determined. After the functional dependence is defined, it is isolated as a separate parameterized function. Then, using the amplitude of the current input signal as an argument of this function, a compensating signal is synthesized, which is a model of distortion in the power amplifier. In the second step, the compensating signal is subtracted from the input signal. In this case, the proposed method differs from the prototype in that to calculate the vector of coefficients of the parameterized function, quantize the amplitude of the input signal |xin| on the interval from zero to the maximum value |xin|max on M levels, and determine the coefficient of the parameterized function, for each quantization level from the expression:

where λ̇m is the parameterized function coefficient, for the m-th level;
km is the number of signal values at the m-th quantization level during the observation time n;
ẋmb is the input signal at the m-th quantization level, balanced in amplitude, phase and delay with the output signal of the power amplifier;
- - sign of complex conjugation of a number;
ẏmb is the distortion signal at the m-th quantization level.

Or

where

Rmn(ẋmb,ẏmb) is the sample covariance coefficient of the corresponding signals at the n-th step for the m-th quantization level;

which corresponds to the average power of the input signal at the m-th quantization level.

Thus, to form the vector Λ of coefficients of the parameterized function, on the observation interval n, for each quantization level m, the coefficient λm is determined by calculating the sample covariance coefficient between the input signal xmb balanced in amplitude, phase and delay and the distortion signal ymb, with its subsequent normalization relative to the power of the specified xmb input signal. The resulting coefficients are written as an element of the vector Λ in position m.

This method does not require the solution of an overdetermined system of equations in matrix form, and thus, in practical terms, reduces computational, time and energy costs. The device that implements this method can be built on the basis of typical functional units and blocks of electronic equipment.

The device for implementing this method (see figure) contains at least the first adder (1) and PA (2) connected in series. The input of the adder (1) is the input of the device, and the output of the PA (2) is the output of the device. The device also contains a control loop (3) for mutual equalization of amplitudes, phases and delays of the input and output signals. In this case, one input of the control loop (3) is connected to the output of the device, and the other - to the input of the device. The device contains a second adder (4), the subtractive input of which is connected to the first output of the control loop (3), and the other input of the adder (4) is connected to the second output of the control loop (3). In this case, respectively, at the first and second outputs of the control loop (3), the input and output signals of the PA are formed, balanced in amplitude, phase and delay. The device also contains a block for calculating the parameters of the predistortion function (5), a key (6) (an optional element), a block for the predistortion function (7) and a phase equalization block (8) connected in series. In this case, the input of the phase equalization block (8) is connected to the first output of the predistortion function block (7), and its output is connected to the second input of the first adder (1). The second input of the predistortion function block (7) is connected to the input of the device. The first input of the block for calculating the parameters of the predistortion function (5) is connected to the first output of the control loop (3), and the second input is connected to the output of the second adder (4).

The device differs in that a functional divider (9) is introduced into the block for calculating the parameters of the predistortion function (5), the output of which is the output of the block (5), as well as two identical circuits (10) and (11), each of which contains a multiplier connected in series conjugate quadrature signal components (12, 13), adder (14, 15), random access memory (RAM) (16, 17) for M memory cells. The RAM output (16, 17) is connected to the second input of the adder (14, 15) in the corresponding circuits (10, 11) and is the output of these circuits. The first inputs from the multipliers of the conjugate quadrature components of the signals (12, 13) are complex conjugate (marked with a dot in the figure) and correspond to the first inputs, and their second inputs correspond to the second inputs of the circuits (10, 11). The inputs of the first circuit (10) are combined with each other and with the first input of the second circuit (11) and are the first input of the block for calculating the parameters of the predistortion function (5). The second input of the second circuit (11) is the second input of the block for calculating the parameters of the predistortion function (5). The input of the denominator (divider) of the functional divider (9) is connected to the output of the first circuit (10), and the input of the dividend is connected to the output of the second circuit (11).

To control the writing-reading processes in the RAM (16, 17), the block for calculating the parameters of the predistortion function (5) includes a signal amplitude calculator (18), an amplitude quantizer (19), and an address generator (20) connected in series. In this case, the input of the signal amplitude calculator (18) is connected to the first input of the first circuit (10), and the output of the address generator (20) is connected to the control inputs (address inputs) of the RAM (16, 17). The quantizer (19) quantizes the range of the input signal amplitude into M levels, the address shaper (20) recodes the quantized amplitude levels into the addresses of the RAM memory cells (16, 17). Thus, each RAM contains M memory cells.

The device works as follows. After equalizing the amplitudes, phases and delays of the input and output signals in the control loop (3), they go to the second adder (4), at the output of which a distortion signal yb is formed. The input signal after equalization in the control loop (3) xb is fed to the inputs of the first circuit (10) and to the first input of the second circuit (11), the second input of which receives the distortion signal yb. At the output of the multiplier of the conjugate quadrature signal components (12), the square of the amplitude (i.e., power) of the instantaneous value of the input signal at a discrete time n is calculated, and at the output of the multiplier of the quadrature signal components (13), the value of the current term in the sample covariance coefficient at the time time n (see expression 5). At the same time, from the amplitude of the input signal obtained at the output of the signal amplitude calculator (18), the address of the memory cells m in the RAM (16, 17) is formed in the circuit from the quantizer (19) and the address generator (20), of which at time n the values of the input signal power accumulated in the RAM (16) and the values of the products of the input signal and distortions accumulated in the RAM (17) are read (in accordance with the denominator in expression 3). At the inputs of the adders (14, 15) of the first and second circuits (10, 11), new values of these quantities are obtained and they update the values in the m-th RAM cells (16, 17). The order of managing the process of reading-writing data in RAM is determined by the features of its implementation and is not the subject of protection of this invention. At the output of the functional divider (9), the m-th coefficient of the parameterized function is formed at the n-th discrete time moment, which is transmitted through the closed key (6) to the predistortion function block (7). The process of transferring the coefficients of the parameterized function λ̇m to the predistortion function block (7) can be carried out in various ways, for example, continuously, sequentially or periodically by polling RAM cells (16, 17) or in other ways. The organization of this process is not the subject of protection of this invention. The key (6) is used in accordance with the selected option for updating the parameter vector in the predistortion function block.

At the output of the predistortion function block (7), a distortion signal is synthesized in accordance with expression (1). Then, in the phase equalization block (8), the phase of the synthesized distortion signal is changed so that in the first adder (1) the compensating signal is subtracted from the input signal of the device. By introducing synthesized pre-distortions at the PA input in antiphase to the expected distortions in the PA, the distortions are compensated at its output, which corresponds to an increase in the linearity of the amplitude and phase-amplitude characteristics of the PA.

Brief description of the drawings

The figure shows a functional diagram of a device that implements a method for compensating for non-linear distortions of high-frequency power amplifiers.

Implementation of the invention

The invention can be carried out in accordance with the diagram shown in the figure. The device for forming the coefficients of the parameterized function uses typical functional elements of digital signal processing, such as adders, multipliers, RAM, etc. Functional signal conversion operations can be performed digitally in the baseband (baseband) with the representation of signals in the form of quadrature components (Richard Lyons, Digital Signal Processing: Second Edition, Translated from English - M.: OOO "Binom-Press", 2006-656, pp. 351 - 356).

Examples of the implementation of the procedure for equalizing the amplitudes, phases and delays of the input and output signals in the control loop (3) can be devices for automatic gain control, phase locked loop and adjustable delay lines.

Blocks of the pre-distortion function (7) and calculation of the parameters of the pre-distortion function (5) can be performed, for example (but not limited to), based on a programmable logic integrated circuit (FPGA), a digital signal processor (DSP processor), a computer and other computing devices.

The block for calculating the pre-emphasis function in expression (1) can be implemented in various ways, for example, as a lookup table with power p interpolation. Then the parameter vector Λ is a vector of length M whose elements are the values of the function f at M points of the argument of the function f, usually (but not always) uniformly distributed over the domain of the function f, i.e. on the interval from zero to the maximum value of the input signal amplitude |xin|max.

Claims (2)

1. The method for compensating for non-linear distortions of a high-frequency power amplifier includes the process of extracting distortions of the power amplifier by subtracting the input signal from the output signal after their alignment in amplitudes, phases and delays, contains the process of generating a compensating signal corresponding to the selected distortions, and mixing it in antiphase to the input signal of the amplifier power, while observing the input signal and distortions, approximate the functional dependence of distortions on the input signal in the form of the product of the input signal by the parameterized function on the amplitude of the input signal, select this functional dependence in the form of a separate predistortion function and, using the input signal as an argument of this function , a compensating signal is synthesized, the method differs in that to calculate the coefficient vector of the parameterized function, the input signal amplitude is quantized into M levels and the parameterized func coefficient is determined. fractions for each quantization level by calculating the power-normalized input signal, the sample covariance of the input signal, and the distortion signal. 2. The device for implementing the method according to claim 1 contains a first adder and a high-frequency power amplifier connected in series, while the first input of the adder is the input of the device, and the output of the power amplifier is the output of the device, the device contains a control loop for mutual alignment of the amplitudes, phases and delay of the input and output signals, while one input of the control loop is connected to the output of the device, and the other to the input of the device, the device also contains a second adder, the subtracting input of which is connected to the first output of the control loop, and the other input of the adder is connected to the second output of the control loop, with at the same time, respectively, at the first and second outputs of the control loop, the input and output signals of the power amplifier are formed, balanced in amplitude, phase and delay, and the device also contains a block for calculating the parameters of the pre-distortion function, a block of the pre-distortion function and a phase equalization block connected in series, In this case, the block for calculating the parameters of the predistortion function and the block for the predistortion function are connected directly or through a switch, the output of the phase equalization block is connected to the second input of the first adder, the second input of the block for the predistortion function is connected to the input of the device, the first input of the block for calculating the parameters of the predistortion function is connected to the first output of the circuit regulation, and the second input - to the output of the second adder, while the output of the block for calculating the parameters of the predistortion function is directly connected to the first input of the block of the predistortion function, thereby the predistortion function is updated continuously, or they are connected through a key, thereby the predistortion function is updated when the key is closed, at the same time, the device differs in that a functional divider is introduced into the block for calculating the parameters of the predistortion function, the output of which is the output of this block, as well as two identical circuits, each of which contains a multiplier of conjugate quadrature components connected in series signals, an adder and a random access memory (RAM) for M memory cells, while the output of each RAM is connected to the second input of the adder of the corresponding circuit and is the output of this circuit, the first inputs of the multipliers of the conjugate quadrature signal components are complex conjugate and correspond to the first inputs, and their the second inputs - to the second inputs of the circuits, the inputs of the first circuit are combined with each other and with the first input of the second circuit and are the first input of the block for calculating the parameters of the predistortion function, the second input of the second circuit is the second input of the block for calculating the parameters of the predistortion function, while the input of the denominator (divider) of the functional the divider is connected to the output of the first circuit, and the input of the dividend is connected to the output of the second circuit, to control the process of writing reading to the RAM, the signal amplitude calculator, the quantizer and the address generator connected in series are introduced into the block for calculating the parameters of the predistortion function, while the input of the signal amplitude calculator is connected with the first input of the first circuit, and the output of the address shaper with the address inputs of the RAM, the quantizer quantizes the range of the input signal amplitude into M levels, and the address shaper recodes the quantized amplitude levels into addresses of RAM memory cells, thus each RAM contains M memory cells.

Images