A normal distribution weight-based wideband power amplifier circuit real frequency matching method

Abstract

The application discloses a normal distribution weight-based wideband power amplifier circuit real frequency matching method, which is applied to the communication field and aims at the frequency point performance compromise in the existing wideband matching method and the problem that the existing wideband matching method cannot adjust a cost function according to actual application requirements; the application adopts a normal distribution to give different weights to different frequency points in a bandwidth, wherein according to the properties of the normal distribution, proper mean value of the normal distribution can make the central frequency in the bandwidth have the highest weight, and further proper standard deviation can obtain the weights of other frequencies and the ratio of the weight of the central frequency, the method can design a better matching network by using the characteristics that most actual systems conform to the normal distribution, so as to improve the matching network performance of the power amplifier, and further design a radio frequency power amplifier with the characteristics of high bandwidth and high conversion efficiency.

Description

Technical Field

[0001] This invention belongs to the field of communications, and specifically relates to a broadband real-frequency matching technology. Background Technology

[0002] In modern communication systems, the most common method to increase data throughput is to increase the signal transmission rate. However, the higher the signal transmission rate, the greater the bandwidth required. How to efficiently transmit broadband signals is a pressing problem for communication systems. In communication systems, the performance of the power amplifier directly affects the signal transmission quality. Therefore, to improve the data throughput of a communication system, its power amplifier must possess broadband characteristics.

[0003] In radio frequency (RF) communication systems, parasitic parameters of components can easily lead to impedance mismatches between the source and load ends. To improve the bandwidth and conversion efficiency of RF communication systems, it is often necessary to design a matching network with high-performance broadband characteristics to convert the impedance of the load port to the optimal impedance required by each component. The commonly used broadband circuit matching technique is the real-frequency technique. However, because the traditional real-frequency technique requires the use of Hilbert transform, it increases design complexity. Therefore, a simplified real-frequency technique has been developed that does not require Hilbert transform and can simultaneously take into account performance parameters such as gain, noise figure, and return loss.

[0004] Existing real-frequency and simplified real-frequency techniques both consider impedance matching within the bandwidth equally. Although they can achieve good matching results, according to practical circuit knowledge, fairly considering the matching problem of the entire bandwidth not only increases the complexity of system optimization, but also means that the design of matching circuits does not focus on key frequency points, and a lot of optimization work is done on optimizing secondary frequency points. Furthermore, designing matching networks based on the principle of fairness and considering the overall system performance can only result in matching circuits that meet the compromise performance of each frequency point, and cannot achieve better matching results.

[0005] Existing broadband matching methods often employ real-frequency technology and simplified real-frequency technology, which equally consider the performance of each frequency point within the bandwidth to optimize the matching network. In other words, existing real-frequency technology assigns the same weight to each frequency point within the bandwidth.

[0006] The cost function (also called the error function) of traditional real-frequency technology is generally defined as:

[0007]

[0008] Where, ω i (i = 1, 2...) represent any discrete frequency point selected within the load's operating frequency band, N is the total number of discrete frequency points, and Z... opt (ω i ) represents the frequency ω iThe optimal impedance at Z (i.e., the actual measured impedance value) in (ω i Let Z be the impedance function of the port drive point of the load matching network. From equation (1), it can be seen that by minimizing ΔZ using an optimization algorithm, Z can be calculated. in (ω i The expression for ) can be further used to derive the matching network through network synthesis, thereby enabling the design of a matching circuit with broadband and high efficiency characteristics.

[0009] The drawbacks of existing technologies are: the cost function of optimizing the matching network assigns the same weight to each frequency point within the bandwidth, which can only achieve a compromise in the working performance of each frequency point and cannot further improve the overall system performance; moreover, it can only design matching circuits for general application scenarios and cannot adjust the cost function according to actual application requirements to meet various application needs. Summary of the Invention

[0010] To address the aforementioned technical problems, this invention proposes a broadband real-frequency matching method based on normal distribution weights suitable for RF power amplifiers. This method uses a normal distribution to assign different weights to optimize the performance of each frequency point within the bandwidth. Furthermore, different normal distribution functions can be designed according to actual application requirements to set different weights for optimizing the performance of each frequency point, which can effectively improve the performance of the broadband matching network.

[0011] The technical solution adopted in this invention is: a broadband real-frequency matching method based on normally distributed weights suitable for RF power amplifiers, comprising:

[0012] S1. Establish a new cost function based on the cutoff frequency and the required gain value;

[0013] S2. Based on the multiple sets of discrete frequency points and corresponding impedance values ​​obtained from actual testing, initialize the coefficients h using polynomial curve fitting.

[0014] S3. With the goal of minimizing the new cost function established in step S1, the polynomial h obtained in step S2 is optimized using nonlinear least squares.

[0015] S4. Calculate the strict Herwitz polynomial g based on the polynomial h optimized in step S3.

[0016] S5. Based on the polynomial h optimized in step S3 and the strict Herwitz polynomial g obtained in step S4, the component parameters and corresponding component types of the matching network are obtained through network synthesis.

[0017] S6. Based on the component parameters and corresponding component types obtained in step S5, establish the input-output matching circuit structure of the power amplifier in ADS, and optimize the input-output matching circuit structure of the power amplifier.

[0018] S7. Use the parameters of the power amplifier's input-output matching circuit optimized in step S6 for power amplifier circuit design.

[0019] The beneficial effects of this invention are as follows: The technical solution of this invention obtains weight values ​​through a normal distribution function to optimize the cost function, thereby realizing a broadband matching network. Simultaneously, this invention also proposes a method for calculating the parameters of the normal distribution function based on the operating bandwidth, unifying the calculation method of weight values ​​and expanding the application scope of real-frequency technology. The real-frequency matching method based on normal distribution weights proposed in this invention can be used to solve the broadband matching problem of RF components such as transceiver antennas, power amplifiers, and low-noise amplifiers, thereby improving the bandwidth of the communication system. This invention is not only compatible with traditional real-frequency technology but also applicable to various application scenarios. Attached Figure Description

[0020] Figure 1 This invention presents the implementation process for broadbanding of radio frequency power amplifiers.

[0021] Figure 2 for Figure 1 A power amplifier input / output matching circuit design platform based on ADS simulation software was established. Detailed Implementation

[0022] To facilitate understanding of the technical content of this invention by those skilled in the art, the following description, in conjunction with the accompanying drawings, further illustrates the invention.

[0023] This invention proposes a broadband real-frequency matching method based on normal distribution weights for RF power amplifiers. It employs a normal distribution to assign different weights to different frequency points within the bandwidth. According to the properties of the normal distribution, appropriately designing the mean of the normal distribution ensures that the center frequency within the bandwidth has the highest weight. Further designing an appropriate standard deviation yields the weights of other frequencies, and the ratio of these weights to the center frequency weights is obtained. This method leverages the fact that most practical systems conform to a normal distribution to design better matching networks, thereby improving the matching network performance of power amplifiers and further enabling the design of RF power amplifiers with high bandwidth and high conversion efficiency.

[0024] This invention proposes a real-frequency matching method for broadband power amplifier circuits based on normal distribution weights. It constructs different weight values ​​for different frequencies using a normal distribution to change the weight of impedance error at each frequency point in the cost function, thereby focusing on optimizing frequencies of interest within the bandwidth as needed (generally, the center frequency is set to have the highest weight). The calculation formula for the proposed real-frequency matching method is:

[0025]

[0026] Where, k i(i = 1, 2...) represents the weight values ​​of discrete frequency points in the cost function. Based on this formula, a general matching network can be designed for any frequency point of interest within the bandwidth.

[0027] As can be seen from equation (1), the simplest method is to give k i (i = 1, 2...) assign different values, and directly give a constant value to the frequency point of interest. For example, when each k i When all (i = 1, 2...) are equal, the proposed cost function is consistent with the traditional cost function and is fully compatible with traditional methods. However, if k i If (i = 1, 2, ...) are constant values, then different values ​​of k need to be selected during the design process based on different application scenarios or measured data. i The algorithm iterates through (i = 1, 2, ...) values ​​until the selected value is the optimal value. However, this optimization is inefficient and lacks general applicability.

[0028] This invention draws upon the normal distribution's conformity to general developmental patterns. It designs a specific normal distribution function based on the actual designed bandwidth, then substitutes different frequency values ​​into the designed normal distribution function to calculate the normal distribution output value. This output value represents the weight of that frequency point within the entire bandwidth. While possessing versatility, it also considers the general principles of conventional circuit design. The proposed method for calculating the mean of the real-frequency normal distribution function is as follows:

[0029]

[0030] The proposed method for calculating the standard deviation of the real-frequency normal distribution function is as follows:

[0031]

[0032] Where p is an adjustable factor, a constant that can be set according to actual needs. The normal distribution function can be uniquely determined based on μ and σ:

[0033]

[0034] Furthermore, based on the actual frequency point, the weight value of that frequency point can be calculated by substituting it into formula (5), i.e.;

[0035]

[0036] Among them, Frq i For discrete frequency points, different frequency points can be selected according to the actual situation. The process for broadband design of RF power amplifiers proposed in this invention is as follows: Figure 1 As shown, it includes the following steps:

[0037] 1. Measure the input and output impedance of the power amplifier using load pulling;

[0038] For example, the input impedance R1…Rn for 91 frequency points within the operating frequency band from 800MHz to 1700MHz is given as 27.926-j*10.786,…,25.577-j*5.723.

[0039] 2. Normalize the impedance;

[0040] For each impedance, calculate Ric = Ri / Rs, where Rs is the reference impedance value, typically taken as 50 ohms.

[0041] 3. Construct the mean and variance of the normal distribution function based on the cutoff frequency and the required gain value; the expressions are as follows:

[0042]

[0043] Where μ is the mean, σ is the variance, ω1 is the initial frequency, and ω N This is the cutoff frequency.

[0044] 4. Based on the multiple sets of discrete frequency points and corresponding impedance values ​​obtained from actual testing, initialize the coefficients h0 using polynomial curve fitting; the implementation process of step 4 is as follows: based on the multiple sets of discrete frequency points and corresponding impedance values ​​obtained from actual testing, use the polyfit function in MATLAB to perform polynomial fitting to obtain the initialized polynomial coefficients h0.

[0045] A41. Set x = ω 2 Generate row vector Xd to assist in linear regression of the polynomial; ω 2 Let ω be the angular frequency. For example, when the input angular frequency Wa = [0.2000 0.2790 0.3580 0.4370 0.5160 0.5950 0.6740 0.7530 0.8320 0.9110 0.9900], then x = ω 2 That is, Xd=Wa*Wa;

[0046] A42. Generate the initial denominator polynomial g0, and calculate g0(jω) = g R +jg X g R and g x Let gval be the real part and the imaginary part of the polynomial evaluation of g0 at each frequency point, respectively. That is, gval = polyval(g0, p); g R =real(gval); g X =imag(gval);

[0047] The process of generating the denominator polynomial g0 is as follows:

[0048] M1, by setting x = ω 2To define the auxiliary polynomial P a (x) The regression array Xd, for example, when the input angular frequency Wa = [0.2000 0.2790 0.3580 0.4370 0.5160 0.5950 0.6740 0.7530 0.8320 0.9110 0.9900], x = ω 2 That is, Xd=Wa*Wa; we get

[0049] M 2, Generate P a The data point Pd of (x) is represented as follows:

[0050]

[0051] Therefore, using the data provided in this embodiment,

[0052] Pd=[5.7690 2.2857 1.0886 1.1405 1.5536 1.9505 2.2921 2.5819 2.82483.0236 3.1794]

[0053] In the next step, set x = ω 2 Used for regression

[0054] M3. Perform polynomial fitting on the given data of Xd and Pd with a selected degree nr = n / 2; use MATLAB's polyfit function to generate the polynomial P. a (x), Pa = polyfit(Xd,Pd,2), to obtain P a (x)=P1x 2 Given the coefficients of P2x and P3, Pa = [7.1550 -6.5538 3.3516].

[0055] M 4, Calculation

[0056] MATLAB function G = conv(P a ,P a Generate strict positive polynomials We obtain G = [51.1936 -93.7841 90.9129 -43.9310 11.2331]

[0057] M5, Factorization

[0058] The result can be easily obtained using the MATLAB function Xr = root(G). The root; then, set The MATLAB function P, which generates roots in RHP, can be used. r =sqrt(-X) r To select the LHP root construction g(p), we then set prm = -Pr to find the root of the LHP.

[0059] Therefore, there is

[0060] >>Xr=root(G)

[0061] Xr =

[0062] 0.4584+0.5090i

[0063] 0.4584-0.5090i

[0064] 0.4575+0.5082i

[0065] 0.4575-0.5082i

[0066] >>P r =sqrt(-X) r )

[0067] Pr =

[0068] 0.3366-0.7561i

[0069] 0.3366+0.7561i

[0070] 0.3364-0.7554i

[0071] 0.3364+0.7554i

[0072] >>prm=-Pr

[0073] prm=

[0074] -0.3366+0.7561i

[0075] -0.3366-0.7561i

[0076] -0.3364+0.7554i

[0077] -0.3364-0.7554i

[0078] M 6. Construct a linear polynomial C(p) on the LHP roots of prm.

[0079] The polynomial C(p) is constructed on the LHP roots of g(p) using the MATLAB function C = poly(prm).

[0080] Therefore, there is

[0081] >>C = poly(prm)

[0082] C=1.000 1.3459 1.8217 0.9212 0.4684

[0083] Finally, we obtain g(p), that is...

[0084] >>g0=aqrt(abs(G(1)))*C

[0085] g0=[7.1550 9.6299 13.0342 6.5908 3.3516]

[0086] A43. Generate array vectors Ad and Bd to produce polynomials A(x) and B(x);

[0087] Let SRLa and SXLa be the real and imaginary parts of the load impedance obtained after normalization in step 2. For each sampling frequency j, let SRL = SRLa(j) and SXL = SXLa(j); then A(j) = g R SRL+g X *SXL; and B(j) = -(g X *SRL-g R *SXL) / W; where W is the value of Wa at point j, i.e. W = Wa(j);

[0088] A44. Find polynomials A(x) and B(x) by polynomial curve fitting;

[0089] A45. The initial coefficients h0 are obtained as 0.0606, -0.2120, and -0.8452.

[0090] Based on the polynomials A and B obtained in step A44, h can be calculated; that is, h(p) = A(-p) 2 )+pB(p 2 )

[0091] 5. To minimize the cost function, the nonlinear least squares optimization polynomial h is used;

[0092] Where, the objective function to be minimized is ε is the cost function shown in equation (2), and nopt is the number of impedances at each frequency point. The implementation process of step A5 is as follows:

[0093] A51. Set the unknown vector x = [.] and set the coefficient vector h = x; assign the h0 obtained in the previous step to the vector x;

[0094] A52. Generate the strict Hurwitz polynomial g; calculate the value of g using the values ​​of h0 and k, where k is the order of the DC transmission zero; for example, when h = [-3.2944 -3.1010 -4.1546 -1.8843 -0.5035]; k = 0, we get g = [3.2944 4.4538 5.7057 3.4847 1.1196];

[0095] A53. Generate the complex form of the load reflection coefficient L11. L11 = hval / gval; gval = polyval(g,x); hval = polyval(h,x); those skilled in the art should know that gval = polyval(g,x) is a computer function that uses the polynomial coefficients g and the x vector to calculate the gval vector. The method of use is to return the value of the nth degree polynomial g at x.

[0096] A54. Calculate the normalized trigger point impedance z in Based on the relationship between reflection coefficient and impedance, we obtain:

[0097] z in = (gval + hval) / (gval - hval);

[0098] The specific derivation process is as follows:

[0099]

[0100] In the above formula, λ is the ω in formula (2). i S 11 This is the reflection coefficient L11 in step A53; R represents the load resistance.

[0101] A55. Using the cost function of equation (2) at the selected sampling frequency w, the optimization vector x, i.e., h, is obtained by minimizing the objective function value at the sampling frequency; h = [h n h n-1 …h0]

[0102] Therefore, by setting h = x, the optimized values ​​of h are h = -0.1360, 0.0825, and -0.5637.

[0103] 6. Calculate the strict Herwitz polynomial g using polynomial h;

[0104] The implementation process of step 6 is as follows:

[0105] A61. Set U = [h] n hn-1 …h0] and V=[(-1] n h n (-1) n-1 h n-1 …h0]. H(-p) is calculated through the convolution of U and V. 2 )=h(p)h(-p)=H0+H1p+…+H n p n .

[0106] A62. Generating Even Polynomials

[0107] G(-p 2 )=H(-p 2 )+(-1) k p 2k =G0+G1p 2 +…+G n p 2n

[0108] A63, By setting X = -p 2 This yields a coefficient vector X = [(-1)] n G n (-1) n-1 G n-1 …G0], calculate the root X of X. k .

[0109] A64, According to X k Calculate G(-p) 2 Find the roots pk and pkm in the right and left half-planes, respectively. Their calculation is as follows:

[0110] pk = sqrt(-Xk)

[0111] pkm = -pk

[0112] A65. Generating a single polynomial It is C = poly(pkm). Those skilled in the art should know that poly() is a C language function that generates a polynomial based on parameters.

[0113] A66, through calculation Obtain the strict Herwitz polynomial g(p). Here, gc is an (n+1)th order MATLAB row vector containing g(p) = g0 + g1p + ... + g n p n The coefficients {g0, g1, ..., g n}, that is, gc = [g n g n-1 …g0].

[0114] Therefore, the expression for the polynomial g is: g = 0.1360, 0.4071, 1.1479.

[0115] 7. Obtain the component parameters and types (capacitors, inductors, resistors) of the matching network through network synthesis;

[0116] By using the inputs h and g, the component parameters and initial component types of the matching network are obtained through long division. The calculation formula is as follows:

[0117]

[0118] Based on this, the matching circuit parameters are: CV = 2.9296, 2.1479, 1.1934.

[0119] CV is the component value, which is obtained according to the long division formula and m is obtained through the following formula:

[0120] n1 = length(h);

[0121] k = 0;

[0122] for i=1:n1

[0123] h_poly(i) = h(n1-i+1);

[0124] g_poly(i) = g(n1-i+1);

[0125] end

[0126] m = g_poly(1) / h_poly(1);

[0127] If m = -1, the components appear sequentially as capacitors and inductors; if m = 1, they appear sequentially as inductors and capacitors. Furthermore, the last component is a resistor.

[0128] 8. Calculate the transmission gain at different frequency points based on polynomials h and g; the implementation process of step 8 is as follows:

[0129] A81. Define a complex variable p = jW; j = sqrt(-1); then p = sqrt(-1) * W

[0130] A82. Calculate the polynomial values ​​hval and gval using h and g. gval = polyval(g,p); hval = polyval(h,p) to obtain the following values: gval = g(0)*p^n + g(1)*p^n-1 + ... + g(n-1)*p + g(n), hval = h(0)*p^n + h(1)*p^n-1 + ... + h(n-1)*p + h(n);

[0131] A83. Obtain the desired transmission gain by inputting the reflection coefficient (gval+hval) / (gval-hval).

[0132] By using gval = polyval(g,x);

[0133] hval = polyval(h,x);

[0134] fval = x^k;

[0135] And the reflection coefficient L11 = hval / gval, calculate

[0136] Weight=fval*conj(fval)*(1-L11*conj(L11));

[0137] D=hval*conj(hval)*(1+L11*conj(L11))+fval*conj(fval)-2*real(L11*hval*conj(gval));

[0138] T = Weight / D;

[0139] Obtain the transmission gain T. The transmission gain is the decisive factor in determining the key performance indicator of the matching circuit—the insertion loss. The better the transmission gain, the less attenuation the RF signal will experience in the designed input / output matching circuit, and the less attenuation the matching circuit will transmit the broadband signal to subsequent circuits. If the transmission gain does not meet the requirements, return to step 1 and reconstruct the expected transmission gain and weight allocation by adjusting the adjustable factor p; specifically, this involves iterating through different values ​​within a small range of p values.

[0140] 9. Based on the above calculation results of h, g and CV values, establish the input-output matching circuit structure of the power amplifier in ADS;

[0141] 10. Simulate and optimize the output matching circuit;

[0142] 11. Maintain the output matching circuit connection, and simulate and optimize the input matching circuit;

[0143] 12. Determine if the transmission efficiency requirement is met. If it is met, proceed to step 13; otherwise, proceed to step 9.

[0144] The optimization process in steps 9-12 is based on the ADS platform. The ideally calculated broadband matching circuit structure will experience varying degrees of performance degradation in actual simulations due to the parasitic parameters of the devices. Therefore, optimization is required. Specifically, optimization is carried out based on the optimization controls of the ADS simulation platform until the PA performance that meets the expectations is achieved.

[0145] 13. The matching circuit parameters obtained from the simulation are used for power amplifier circuit design.

[0146] The power amplifier input / output matching circuit design platform established based on ADS simulation software in the process of this invention is as follows: Figure 2 As shown.

[0147] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Various modifications and variations can be made to the invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the scope of the claims of the invention.

Claims

1. A broadband real-frequency matching method based on normally distributed weights for RF power amplifiers, characterized in that, include: S1. Based on the cutoff frequency and the required gain value, and considering the weights of different frequencies, establish a new cost function. Where, ω i Z is a discrete frequency point arbitrarily selected within the load's operating frequency band. in (ω i Z represents the port drive point impedance function of the load matching network. opt (ω i ) represents the frequency point ω i The optimal impedance at the specified frequency point, where N represents the number of discrete frequency points; S2. Based on the multiple sets of discrete frequency points and corresponding impedance values ​​obtained from actual testing, the initial polynomial coefficients h0 are obtained by polynomial curve fitting. S3. With the goal of minimizing the new cost function established in step S1, the optimized polynomial h is obtained by using the initial coefficients h0 obtained in the nonlinear least squares optimization step S2. S4. Calculate the strict Herwitz polynomial g based on the polynomial h optimized in step S3. S5. Based on the polynomial h optimized in step S3 and the strict Herwitz polynomial g obtained in step S4, the component parameters and corresponding component types of the matching network are obtained by long division. S6. Based on the component parameters and corresponding component types obtained in step S5, establish the input-output matching circuit structure of the power amplifier in ADS, and optimize the input-output matching circuit structure of the power amplifier. S7. Use the parameters of the power amplifier's input-output matching circuit optimized in step S6 for power amplifier circuit design.

2. The broadband real-frequency matching method based on normally distributed weights for RF power amplifiers according to claim 1, characterized in that, The calculation process for the weights of different frequencies in step S1 is as follows: S11. Construct the mean and variance of the normal distribution function based on the cutoff frequency and the required gain value; the expressions are as follows: and Where μ is the mean, σ is the variance, ω1 is the initial frequency, and ω N Where p is the cutoff frequency, and p is an adjustable factor; S12. The normal distribution function uniquely determined by μ and σ is: S13. Substitute the actual frequency point into the normal distribution function formula to calculate the weight value of that frequency point: Among them, Frq i These are discrete frequency points.

3. The broadband real-frequency matching method based on normally distributed weights for RF power amplifiers according to claim 1, characterized in that, Step S2 specifically involves using the polyfit function in MATLAB to perform polynomial fitting.

4. A broadband real-frequency matching method based on normally distributed weights for RF power amplifiers according to claim 1, characterized in that, Step S3 further includes: constructing a minimized objective function based on the cost function of step S1. ε is the cost function of step S1, and nopt is the number of impedances at each frequency point.

5. A broadband real-frequency matching method based on normally distributed weights for RF power amplifiers according to claim 4, characterized in that, The implementation process of step S3 is as follows: S31. Set the unknown vector x = [.] and set the coefficient vector h = x; assign the initial coefficient h0 obtained in step S2 to the vector x; S32. Generate the strict Hurwitz polynomial g using h0 and k; where k is the order of the DC transmission zero. S33. Generate the complex form of the load reflection coefficient L11; L11 = hval / gval; gval = polyval(g,p); hval = polyval(h,p); S34. Calculate the normalized driving point impedance; based on the relationship between the transmission coefficient and the impedance, we obtain Zin=(gval+hval) / (gval-hval); S35. Using the new cost function constructed in step S1 at the selected sampling frequency w, the optimization vector x, i.e., h, is obtained by minimizing the objective function value at the sampling frequency; h = [h n h n-1 …h0].

6. A broadband real-frequency matching method based on normally distributed weights for RF power amplifiers according to claim 1, characterized in that, Step S4 specifically includes the following sub-steps: S41. Set U = [h] n h n-1 …h0] and V=[(-1] n h n (-1) n-1 h n-1 …h0], by convolution of U and V, H(-p) is calculated. 2 )=h(p)h(-p)=H0+H1p+…+H n p n ; S42, Generating Even Polynomials G(-p 2 )=H(-p 2 )+(-1) k p 2k =G0+G1p 2 +…+G n p 2n S43. Set X = -p 2 This yields the coefficient vector X = [(-1)]. n G n (-1) n-1 G n-1 …G0], calculate the root X of X. k ; S44. Calculate G(-p) 2 The roots pk and pkm in the right and left half-planes are calculated as follows: pk = sqrt(-Xk) pkm = -pk S45. Generating a single polynomial Its value is C = poly(pkm); S46, through calculation Obtain the strict Herwitz polynomial g(p); where gc is an (n+1)th order MATLAB row vector containing g(p) = g0 + g1p + ... + g n p n The coefficients {g0, g1, ..., g n }, that is, gc = [g n g n-1 …g0].

7. A broadband real-frequency matching method based on normally distributed weights for RF power amplifiers according to claim 1, characterized in that, The implementation process of step S5 is as follows: Based on the polynomial h optimized in step S3 and the strict Herwitz polynomial g obtained in step S4, the component parameters and initial component types of the matching network are obtained using long division. Specifically, m is obtained through the following formula: n1 = length(h); k＝0； for i=1:n1 h_poly(i) = h(n1-i+1); g_poly(i) = g(n1-i+1); end m = g_poly(1) / h_poly(1); If m = -1, the components appear in sequence as capacitors and inductors; if m = 1, they appear in sequence as inductors and capacitors. In addition, the last component is a resistor.

Images