A predistortion device and method adaptable to the change of input signal power

Abstract

The present invention discloses a predistortion device and method adaptable to input signal power variation. The device includes a DPD actuator, a DAC module, a power amplifier PA, an ADC module, a power variation detection module, a switching switch, a coefficient pre-extraction module, and a parameter estimation module. The DPD actuator includes a DPD model, a first multiplier, a second multiplier, and an adder, where DPD refers to digital predistortion. When the signal power changes, the present invention only needs to adjust a parameter called a compensation factor, without the need to recalculate the DPD model coefficients, and can quickly adapt to the variation of the input signal power.

Description

Technical Field

[0001] The present invention relates to digital predistortion, and particularly to a predistortion device and method adaptable to changes in the power of an input signal. Background Art

[0002] With the development of 5G communication, the linearity requirements of handheld devices continue to grow. For handheld devices with a large power change range, the DPD technology design of transmitters is not yet mature, and existing methods mainly focus on the LUT model to reduce hardware complexity.

[0003] However, since the PA nonlinearity changes in real time due to the influence of signal power and load adaptation, static DPD cannot be used for compensation. The LUT model is generally developed based on the trimmed Volterra series. To perform dynamic compensation, updating and tabulating the model coefficients each time requires a large amount of time and computing resources, resulting in its inability to track the rapid changes in signal power and its inadaptability to scenarios with high environmental change complexity. Summary of the Invention

[0004] The purpose of the present invention is to overcome the deficiencies of the prior art and provide a predistortion device and method adaptable to changes in the power of an input signal. When the signal power changes, only a parameter called a compensation factor needs to be adjusted, and it is not necessary to recalculate the DPD model coefficients, enabling rapid adaptation to changes in the power of the input signal.

[0005] The purpose of the present invention is achieved through the following technical solutions: A power amplifier predistortion device adaptable to changes in the power of an input signal includes a DPD actuator, a DAC module, a power amplifier PA, an ADC module, a power change detection module, a switching switch, a coefficient pre-extraction module, and a parameter estimation module; the DPD actuator includes a DPD model, a first multiplier, a second multiplier, and an adder, and the DPD refers to digital predistortion;

[0006] The input signal is respectively transmitted to the first input ports of the DPD model and the first multiplier, and the output end of the DPD model is connected to the first input port of the second multiplier; the second input port of the first multiplier and the second output port of the second multiplier are both connected to the output end of the parameter estimation module; the output ports of the first multiplier and the second multiplier are both connected to the adder, and the signals are added by the adder and then transmitted to the power amplifier PA through the DAC;

[0007] The input end of the ADC module is connected to the output end of the power amplifier PA, the output end of the ADC module is connected to the power change detection module, and the output end of the power change detection module is connected to the switching switch, which is used to select to connect the coefficient pre-extraction module or the parameter estimation module; the coefficient pre-extraction module is used to extract the pre-distortion coefficient and assign the extracted coefficient to the DPD model, and the parameter estimation module is used to calculate the adjustment parameter when the power changes and transmit the received signal to the first multiplier and the second multiplier.

[0008] Among them, the power amplifier PA is used to amplify the power of the received signal and transmit the amplified signal to the transmitting antenna.

[0009] A power amplifier pre-distortion method adapted to the power change of the input signal includes the following steps:

[0010] S1. In the initial state, when the power of the input signal remains unchanged, use the coefficient pre-extraction module to extract the pre-distortion parameters of the power amplifier and assign the extracted coefficient to the DPD model;

[0011] The step S1 includes:

[0012] S101. The parameter estimation module initializes the adjustment parameter α = 1 and switches the switching switch to connect to the coefficient pre-extraction module;

[0013] S102. The parameter estimation module transmits the parameter 1-α to the second input end of the first multiplier and transmits the parameter α to the second input end of the second multiplier;

[0014] S103. When the signal is input, first keep the power of the signal unchanged. The signal after passing through the DPD actuator is transmitted to the power amplifier through the DAC, and then the signal output by the power amplifier is sampled by the ADC. After that, the obtained signal is transmitted to the coefficient pre-extraction module through the power change detection module and the switching switch;

[0015] S104. The coefficient pre-extraction module performs coefficient pre-extraction of the DPD model according to the original input signal and the signal output by the switching switch:

[0016] Assume the maximum memory depth Q and the maximum order K of the pre-distortion model, and the pre-distortion model coefficient is denoted as:

[0017] ω = [ω 1,0 , ω 1,1 , …, ω 1,Q , ω 3,0 , ω 3,1 , …, ω 3,Q ,..., ω K,0 , ω K,1 , …, ω K,Q ​

[0018] In the initial state, the first term ω in ω 1,0 is 1, and the rest are all 0, that is, the signal remains unchanged after passing through the DPD model in the initial state;

[0019] Assume that the length of the original input signal is N, denote the original input signal as y, and denote the signal output by the switch as X. The estimated value of the model coefficient is calculated by the least squares method

[0020] ω LS =(X H X) -1 X H y

[0021] where X = [x 1,0 , x 1,1 , …, x 1,Q , x 3,0 , x 3,1 , …, x 3,Q ,..., x K,0 , x K,1 , …, x K,Q , X is an N×R - dimensional matrix,

[0022] R=(K + 1) / 2*(Q + 1); y is an N×1 - dimensional column vector containing the original input signal of length N;

[0023] where x i,j , i = 1, 3, 5,..., K; j = 0, 1, 2, 3,..., Q in the matrix X are N×1 - dimensional column vectors representing the switch output data corresponding to the order i and the memory depth j. Among them, K is an odd number and Q is a positive integer;

[0024] S105. Assign the pre - extracted model coefficients to the DPD model.

[0025] S2. The power change detection module detects the power of the input signal in real - time:

[0026] When the power does not change, perform pre - distortion processing of the power amplifier based on the DPD model;

[0027] When the power changes, calculate the adjustment parameters through the parameter estimation module, update the DPD actuator, and realize the pre - distortion processing after the power change.

[0028] In step S2, when the power does not change, the process of performing pre - distortion processing of the power amplifier based on the DPD model is as follows:

[0029] The input signal is multiplied by the parameter 1-α in the first multiplier and transmitted to the adder; at the same time, after the input signal is pre-distorted by the DPD model, it is multiplied by the parameter α in the second multiplier and transmitted to the adder; the adder adds the received signals, converts them through the DAC module, and then transmits the converted signal to the power amplifier PA for power amplification, and the power amplifier PA transmits the amplified signal to the transmitting antenna for transmission.

[0030] In step S2, when the power changes, the adjustment parameter is calculated by the parameter estimation module to update the DPD actuator, and the process of pre-distortion processing after the power change is as follows:

[0031] S201. When it is found that the power changes, the change-over switch is switched to be connected to the parameter estimation module, and the power change detection module transmits the signal powers before and after the power change to the parameter estimation module;

[0032] S202. The parameter estimation module calculates the adjustment parameter α according to the signal powers before and after the power change and updates the DPD actuator;

[0033] S203. Then, based on the updated DPD actuator, pre-distortion processing of the signal after the power change is realized:

[0034] The input signal is multiplied by the updated parameter 1-α in the first multiplier and transmitted to the adder; at the same time, after the input signal is pre-distorted by the DPD model, it is multiplied by the updated parameter α in the second multiplier and transmitted to the adder; the adder adds the received signals, converts them through the DAC module, and then transmits the converted signal to the power amplifier PA for power amplification, and the power amplifier PA transmits the amplified signal to the transmitting antenna for transmission.

[0035] The calculation method of the adjustment parameter α is as follows:

[0036]

[0037] Among them, represents the signal power detected by the power change detection module after the power change, and P out represents the signal power in the initial state.

[0038] The beneficial effects of the present invention are: 1. After the input signal power changes, there is no need to spend a lot of time recalculating the model coefficients using traditional algorithms (such as LS, LMS algorithms, etc.), which greatly improves the efficiency.

[0039] 2. By measuring the magnitude of the nonlinearity through the feedback path and adjusting the compensation factor, the consumed hardware resources are small, and the algorithm complexity is greatly reduced.

[0040] 3. Replace the traditional cumbersome dynamic predistortion model with a lightweight model coefficient update method, which can quickly adapt to scenarios with frequent power changes and effectively improve the flexibility of the predistortion model. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 Schematic diagram of the principle of the present invention;

[0042] Figure 2 Schematic diagram of the principle of the predistortion link;

[0043] Figure 3 Simplified schematic diagram of the predistortion principle;

[0044] Figure 4 Schematic diagram of the principle of the DPD actuator based on power change;

[0045] Figure 5 Schematic diagram of the PA output power spectrum before and after power change. DETAILED DESCRIPTION OF THE INVENTION

[0046] The technical solution of the present invention will be further described in detail below with reference to the drawings, but the protection scope of the present invention is not limited to the following.

[0047] The predistortion model coefficient update architecture of the present invention is as Figure 1 shown.

[0048] The update process includes three steps: pre-extraction of basic model coefficients, power change detection, and update of model coefficients.

[0049] The process of pre-extracting basic model coefficients is as follows:

[0050] Under the current input signal power input condition, the sliding switch in the figure is turned to S1, and at this time, α = 1; the memory depth Q and order K of the basic predistortion model are determined by experimental tests (generally ensuring that the adjacent channel leakage ratio suppression ability reaches 20 dBc or above);

[0051] The predistortion model coefficients are denoted as:

[0052] ω = [ω 1,0 , ω 1,1 , …, ω 1,Q , ω 3,0 , ω 3,1 , …, ω 3,Q ,..., ω K,0 , ω K,1 , …, ω K,Q

[0053] In the initial state, the first term ω 1,0 ​is 1 and the rest are all 0, that is, the signal remains unchanged after passing through the DPD model in the initial state;

[0054] Assume that the length of the original input signal is N. Denote the original input signal as y and the signal output by the switching switch as X. The estimated value of the model coefficient is calculated by the least squares method

[0055] Formula (1) ω LS =(X H X) -1 X H y

[0056] where X = [x 1,0 , x 1,1 , …, x 1,Q , x 3,0 , x 3,1 , …, x 3,Q ,..., x K,0 , x K,1 , …, x K,Q , X is an N×R-dimensional matrix,

[0057] R = (K + 1) / 2 * (Q + 1); y is an N×1-dimensional column vector containing the original input signal of length N;

[0058] where the x i,j inside the matrix X, i = 1, 3, 5,..., K; j = 0, 1, 2, 3,..., Q are N×1-dimensional column vectors representing the switching switch output data corresponding to the order i and the memory depth j. Among them, K is an odd number and Q is a positive integer;

[0059] In some other embodiments of the present application, the predistortion model can also be pre-trained offline in advance, so that the process of extracting the predistortion coefficient can be omitted, and the model coefficient can be directly updated based on power detection. In this case, there is no need to configure a high-precision ADC because only the power change needs to be detected later, and the ADC accuracy requirement is relatively low, so as to achieve the purpose of cost savings.

[0060] The process of power change detection and updating the model coefficient is as follows:

[0061] After detecting the power change, the sliding switch is turned to S2, and the model adjustment factor α is started to be calculated;

[0062] Update α to the proposed DPD actuator to achieve model adaptation in the power change scenario.

[0063] Among them, the compensation factor α and the update of the proposed DPD actuator include the following multiple sub-steps:

[0064] A general predistortion link can be represented by Figure 2 where

[0065] Formula (2)

[0066] Formula (3)

[0067] Formulas (2) and (3) can be expressed as:

[0068] Formula (4)

[0069] Formula (5)

[0070] Where λ(x(n)) represents the compensation of the DPD model, which includes the non-linear out-of-band extension, and represents the peak compression effect introduced by the PA. Combining Formulas (4) and (5), the PA output signal can be expressed as

[0071] Formula (6)

[0072] Since the power of λ(x) is almost negligible compared to the input signal x(n), Formula (6) can be simplified to Formula (7) y 2 (n) ≈ x(n) + λ(x) + δ(x)

[0073] Because the function of DPD is to cancel the non-linearity generated by the PA, the purpose is to make

[0074] Formula (8) λ(x) + δ(x) = 0

[0075] Formula (7) can be simply expressed as Figure 3 :

[0076] When the power of the input signal x(n) decreases from to px(n) (0 < p < 1), the non-linearity generated by the PA also weakens. Because power normalization is performed when calculating the DPD model coefficients, we can assume that the response of the PA changes to:

[0077] Formula (9)

[0078] Where is the power attenuation factor and is the power compensation factor. This assumption is reasonable. Consider the following simple PA model:

[0079] Formula (10) δ(x(n)) = x(n)|x(n)| 2

[0080] Then the response of the PA to px(n) can be expressed as

[0081] Formula (11)

[0082] In this case, α = p 2 . When δ(x(n)) becomes more complex, the compensation factor α can be obtained by calculating the power spectral density (PSD) of the PA output signal. According to Equation (9), the nonlinearity of the PA becomes αδ(x(n)). To maintain Equation (8), the pre-distortion nonlinear compensation also needs to be multiplied by α. Thus, the pre-distorted signal is corrected to:

[0083] Equation (12)

[0084] According to Equation (12), the entire power-variation-based DPD actuator can be represented by Figure 4 as follows:

[0085] The specific calculation steps for estimating the compensation factor α and updating the model coefficients are as follows:

[0086] The ADC samples the data output by the power amplifier. When the power changes, first, the power of the output signal is normalized to the power at the base point, and then the degree of power change in the nonlinear region of the power spectrum is calculated, as shown in Figure 5 as follows:

[0087] The general power calculation range for the nonlinear region is where B represents the signal bandwidth. According to the rate of change of the power in the upper sideband or lower sideband nonlinear region, the formula for calculating α is as follows:

[0088] Equation (13)

[0089] After obtaining α, the transmitted pre-distorted signal changes from the original to:

[0090] (1 - α)x(n) + αg(x(n)), thereby achieving dynamic model correction based on power variation.

[0091] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. For example, the basic pre-distortion model can be an MP, ML, GMP model, etc.; the LS algorithm for calculating the initial model coefficients can be replaced with an LMS algorithm, neural network training, etc.; the detection of the power change in the nonlinear region of the output signal obtained from the power spectrum calculated by FFT can be replaced with filtering out the in-band signal by a filter and then calculating the power change rate of the out-of-band signal. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A power amplifier predistortion method adapted to input signal power variation, characterized in that: The following steps are involved: S1. In the initial state, when the power of the input signal remains unchanged, the pre-distortion parameters of the power amplifier are extracted using the coefficient pre-extraction module, and the extracted coefficients are assigned to the DPD model; S2. The power change detection module detects the power of the input signal in real time: When the power does not change, the pre-distortion processing of the power amplifier is performed based on the DPD model; When the power changes, the adjustment parameters are calculated through the parameter estimation module, and the DPD actuator is updated to achieve pre-distortion processing after the power changes; In step S2, when the power does not change, the pre-distortion processing process of the power amplifier based on the DPD model is as follows: The input signal is multiplied by the parameter 1-α in the first multiplier and transmitted to the adder; at the same time, the input signal is pre-distorted by the DPD model, multiplied by the parameter α in the second multiplier, and transmitted to the adder; the adder performs addition processing on the received signal, converts it through the DAC module, and then transmits the converted signal to the power amplifier PA for power amplification, and the power amplifier PA transmits the amplified signal to the transmitting antenna for transmission; In step S2, when the power changes, the adjustment parameters are calculated by the parameter estimation module, and the DPD actuator is updated to implement the pre-distortion processing after the power changes as follows: S201. When a power change is found, the switch is switched to connect to the parameter estimation module, and the power change detection module transmits the signal power before and after the power change to the parameter estimation module; S202. The parameter estimation module calculates the adjustment parameter α to update the DPD actuator according to the signal power before and after the power change; S203. Then, based on the updated DPD actuator, the pre-distortion processing of the signal after the power change is implemented: The input signal is multiplied by the updated parameter 1-α in the first multiplier and transmitted to the adder; at the same time, the input signal is pre-distorted by the DPD model, multiplied by the updated parameter α in the second multiplier, and transmitted to the adder; the adder performs addition processing on the received signal, converts it through the DAC module, and then transmits the converted signal to the power amplifier PA for power amplification, and the power amplifier PA transmits the amplified signal to the transmitting antenna for transmission; The adjustment parameter α is calculated as follows: in, It indicates the signal power detected by the power change detection module after the power change, P out Represents the signal power in the initial state.

2. A predistortion method adapted to input signal power variation according to claim 1, characterized in that: The step S1 comprises: S101. The parameter estimation module initializes the adjustment parameter α=1, and switches the switch to connect with the coefficient pre-extraction module; S102. The parameter estimation module transmits the parameter 1-α to the second input terminal of the first multiplier, and transmits the parameter α to the second input terminal of the second multiplier; S103. When the signal is input, the power of the signal is kept constant. The signal after the DPD actuator is transmitted to the power amplifier through the DAC. Then, the ADC samples the signal output by the power amplifier, and then the obtained signal is transmitted to the coefficient pre-extraction module through the power change detection module and the switch. S104. The coefficient pre-extraction module performs coefficient pre-extraction of the DPD model according to the original input signal and the signal output by the switching switch: Assuming the maximum memory depth Q and the maximum order K of the predistortion model, the predistortion model coefficients are recorded as: ω=[ω 1,0 ,oh 1,1 ,…,oh 1,Q ,oh 3,0 ,oh 3,1 ,…,oh 3,Q ,...,oh K,0 ,oh K,1 ,…,oh K,Q ] In the initial state, the first term ω in ω 1,0 is 1, and the rest are all 0, that is, the signal in the initial state remains unchanged after passing through the DPD model; Assume that the length of the original input signal is N, the original input signal is recorded as y, the signal output by the switching switch is recorded as X, and the estimated value of the model coefficient is calculated by the least squares method ω LS =(X H X) -1 X H y Where X = [x 1,0 ,x 1,1 ,…,x 1,Q ,x 3,0 ,x 3,1 ,…,x 3,Q ,...,x K,0 ,x K,1 ,…,x K,Q ], X is an N×R dimensional matrix, R = (K + 1) / 2 * (Q + 1); y is an N × 1-dimensional column vector containing the original input signal of length N; where x inside the matrix X i,j ,i=1,3,5,...,K;j=0,1,2,3,...,Q is an N×1-dimensional column vector, representing the output data of the switch corresponding to the order i and the memory depth j, where K is an odd number and Q is a positive integer; S105. Assign the pre-extracted model coefficients to the DPD model.

3. A power amplifier predistortion device adapted to input signal power variation, using the method according to any one of claims 1 to 2, characterized in that: It includes a DPD actuator, a DAC module, a power amplifier PA, an ADC module, a power change detection module, a switch, a coefficient pre-extraction module and a parameter estimation module; the DPD actuator includes a DPD model, a first multiplier, a second multiplier and an adder, and the DPD refers to digital pre-distortion; The input signal is transmitted to the DPD model and the first input port of the first multiplier respectively, and the output end of the DPD model is connected to the first input port of the second multiplier; the second input port of the first multiplier and the second output port of the second multiplier are both connected to the output end of the parameter estimation module; the output ports of the first multiplier and the second multiplier are both connected to the adder, and the adder adds the signal and transmits it to the power amplifier PA through the DAC; The input end of the ADC module is connected to the output end of the power amplifier PA, the output end of the ADC module is connected to the power change detection module, the output end of the power change detection module is connected to the switching switch, and the switching switch is used to select the connection coefficient pre-extraction module or the parameter estimation module; the coefficient pre-extraction module is used to extract the pre-distortion coefficient and assign the extracted coefficient to the DPD model, and the parameter estimation module is used to receive the signal when the power changes, calculate the adjustment parameter and transmit it to the first multiplier and the second multiplier.

4. The predistortion device adapted to input signal power variation according to claim 3, characterized in that: The power amplifier PA is used to amplify the power of the received signal and transmit the amplified signal to the transmitting antenna.

Images