A method for pre-processing amplitude weights of a large-scale beamforming matrix

Abstract

The application provides a large-scale beam forming matrix amplitude weight preprocessing method, which utilizes the different synthesis losses of each feed source caused by the different actual synthesis paths of the feed source signals, adds a certain amplitude weight margin for each original amplitude weight, that is, the synthesis loss margin of the feed source with less synthesis times is compensated to the power distribution circuit at the beam end, the amplitude weight of the feed source is reduced, and the reduced energy is used to increase the amplitude weight of other feed sources in the beam; then the amplitude weight after the addition of the margin is compensated from large to small, and in order to meet the energy conservation, the increased part of the energy also needs to be realized by reducing the energy of the large weight, and finally the energy balance at the system level between the amplitude accuracy adjustment and the insertion loss of the large-scale beam forming matrix is realized.

Description

Technical Field

[0001] This invention belongs to the fields of communication and radar technology, and in particular relates to a method for preprocessing amplitude weights of a large-scale beamforming matrix. Background Technology

[0002] Beamforming matrices are key components in multi-beam communication, radar, and other systems, and are widely used in satellite communication, radar systems, and other fields. The amplitude and phase weight distribution from the beam to the feed in a beamforming matrix is ​​a crucial parameter affecting beamforming. It is generally obtained by optimizing the coverage area of ​​each beam. The beamforming matrix implements this amplitude and phase weight distribution through circuit design, with amplitude and phase accuracy being key performance indicators. Furthermore, beam insertion loss must be considered when implementing the amplitude weight distribution, which is also a key performance indicator for beamforming matrices. When the beamforming matrix is ​​large, it requires complex system design. The system consists of multiple functional circuit modules and connecting cables between modules, all of which affect amplitude accuracy and insertion loss. Moreover, errors in circuit fabrication and the accumulation of errors at multiple stages can impact insertion loss during amplitude accuracy adjustments. Therefore, error analysis must be performed before circuit design to address the balance between amplitude and phase loss in advance.

[0003] Assume the beamforming matrix includes 109 beam input signal power dividers, 109 sets (each containing several feeds for the corresponding beam) of fixed phase shifters and fixed attenuators, and 64 power combiners outputting to the feeds. The input signal for each beam is first split, with the number of splits equal to the number of feeds for that beam. Unequal power dividers are used to ensure the amplitude of each signal meets the beam amplitude weighting requirements. Then, the signal passes through fixed phase shifters and fixed attenuators to ensure the signal meets the beam phase weighting requirements, and amplitude accuracy is fine-tuned within a small range. Finally, the signals from the feeds shared by all beams are combined by an equal-power, equal-phase combiner and transmitted to the feeds. The number of feeds, feed numbers, and amplitude and phase weights for forming the beam are all predetermined.

[0004] The principle block diagram of the beamforming matrix is ​​as follows: Figure 1 As shown, for convenience, the beamhead is uniformly defined as the power distribution end, and the feedhead is uniformly defined as the power combining end. The connection relationship between the output of the beamhead and the input of the feedhead is given by the amplitude and phase weighting table. From Figure 1 As shown in the block diagram of the beamforming matrix, it comprises three circuit modules: a power distribution circuit at the beam end, a phase-shifting attenuator circuit, and a power combining circuit at the feed end. Signal transmission between these modules is achieved via radio frequency (RF) transmission lines, such as RF cables (including RF connectors), microstrip lines, and striplines.

[0005] Through miniaturization design, it is possible to Figure 1 The various circuits in the design utilize LTCC (Low Temperature Cofired Ceramic) three-dimensional stacking technology to transform the planar microwave circuit into a three-dimensional stacked circuit, embedded within multiple layers of LTCC ceramic substrates. This reduces the circuit area and the number of connecting cables. Furthermore, by optimizing the beamforming matrix topology, the number of substrates used is reduced, further miniaturizing the circuit. Figure 2 The diagram shows a miniaturized beamforming matrix LTCC substrate. Signal transmission between different substrates and between the input and input terminals is achieved through radio frequency cables.

[0006] Miniaturized beamforming matrices based on LTCC multilayer ceramics include three types of LTCC substrates.

[0007] The first type of beam power distribution board divides 109 beam signals into two signals each. This type of board has 109 independent circuits distributed on 14 LTCC substrates with an area of ​​84mm×23mm, and the substrate numbers are LT2076-1 to LT2076-14.

[0008] The second type of beam power distribution and feed power combining assembly board (hereinafter referred to as the assembly board) includes three functions. First, the power distribution at the beam end is performed in rows on the front side of the assembly board. Then, the first equal power and equal phase combining at the feed end is performed in columns on the back side of the assembly board. The second power distribution at the beam end and the first equal power and equal phase combining at the feed end are connected by phase shifters, attenuators, and embedded microwave transmission lines. The phase shifters and attenuators extend to the surface of the substrate, enabling fine-tuning of phase and amplitude accuracy within a small range. These 11 substrates are key components for miniaturizing the beamforming matrix, and each substrate measures 163mm × 163mm. The substrate numbers are LT2078-1 to LT2078-11.

[0009] The function of the third type of feed-side power combining board is to perform a second combining of feeds with the same number on the second type of substrate. This type of board has a total of 64 independent circuits, distributed on 5 LTCC substrates with an area of ​​100mm×100mm. The substrate numbers are LT2076-1 to LT2076-5;

[0010] according to Figure 2The beam-end power distribution circuit is implemented in two stages. The first stage is a two-way power divider implemented on substrates LT2076-1 to LT2076-14. The second stage consists of two multi-way power dividers implemented on substrates LT2078-1 to LT2078-11. Each substrate is divided into two circuit sections along its thickness. The upper section implements the second-stage power distribution circuit at the beam end, while the lower section implements the first-stage power combining circuit at the feed end. The connection between the two (including phase shifting) is achieved through microwave transmission lines embedded within the substrates. The amplitude weight distribution of each beam in the beamforming matrix is ​​achieved through these two stages of circuitry. The design method for these two stages of circuitry is described below.

[0011] For ease of explanation, this invention will be explained using one of the beams (denoted as beam 1) as an example. Beam 1 involves a total of 13 feed sources, and the amplitude weights of each feed source are shown in Table 1.

[0012] Table 1. Amplitude weight distribution of beam 1

[0013]

[0014]

[0015] The amplitude accuracy of each feed in the beam should meet the requirements of Table 2.

[0016] Table 2. Beamforming Matrix Amplitude Accuracy Requirements

[0017] Amplitude weight Precision control requirements 0dB to -6dB ±0.2dB -6dB to -13dB ±0.25dB -13dB to -20dB ±0.35dB -20dB to -30dB ±0.5dB

[0018] Assume that output port 1 of the first-stage power divider of beam 1 is connected to substrate LT2078-1, and output port 2 is connected to substrate LT2078-2; there are m feed sources related to beam 1 on LT2078-1, and n feed sources related to beam 1 on LT2078-2. Thus, the power distribution ratio of the two outputs of the first-stage power divider of beam 1 is the ratio of the sum of the output energy of the m output ports on LT2078-1 to the sum of the output energy of the n output ports on LT2078-2.

[0019] The second-stage power divider on the LT2078 substrate employs a combination of parallel and series connections, such as... Figure 3 As shown. First, a 2-way power divider is designed at the input. The power divider's distribution ratio is the ratio of the sum of the energy of all output terminals of one channel to the sum of the energy of all output terminals of the other channel. Then, two branch power dividers are designed in series. The power dividers for these two branches are implemented using series units mainly because series power divider circuits are simpler, have no intersecting connection lines, and are easier to miniaturize and simulate.

[0020] The above analysis shows that the power distribution circuit at the beam end can ultimately be decomposed into designing multiple 2-way unequal power dividers and connecting them in a certain order. To meet different power distribution ratio requirements, three different types of unequal power dividers were designed. The first type is the Wilkinson power divider. This type of power divider can use lumped-parameter capacitors and inductors to be equivalent to a quarter-wavelength impedance transformer. The lumped-parameter capacitors and inductors can be miniaturized through LTCC three-dimensional stacking. However, the power distribution ratio range of this type of unequal power divider is generally between 1:1 and 1:2 (corresponding to coefficients of -3.0dB:-3.0dB to -4.0dB:-2.0dB at the two output terminals). This is mainly limited by the manufacturing process, because a larger power ratio requires a finer metal line width (corresponding to a high-impedance impedance transformer). Whether it's high-frequency PCB technology or LTCC technology, the line width of their metal conductors generally needs to be greater than 0.1mm to ensure line accuracy. Therefore, for unequal power dividers with larger power ratios, couplers can be used. A coupler consists of a through end and a coupled end, and can achieve a relatively high power ratio. Although there is a 90° phase difference between the through end and the coupled end, this can be compensated for by a phase shifter. There are two types of couplers: wide-side couplers and narrow-side couplers. In a wide-side coupler, the through transmission line and the coupled transmission line are on two layers of dielectric material. Its coupling output (also called the coupling coefficient) is related to the thickness of the dielectric layer between the two lines and their relative displacement on the two layers. When the dielectric material and the dielectric layer thickness are fixed, the relative displacement between the two lines at the coupled end determines the coupling coefficient. The coupling is maximized when the two layers completely overlap; the greater the relative displacement, the smaller the coupling. Wide-side couplers can achieve a coupling coefficient between -4.0dB and -13.0dB, corresponding to a through coefficient of -2.0dB to -0.22dB at the through end. In a narrow-edge coupler, the through transmission line and the coupled transmission line are on the same dielectric layer. Its coupling coefficient is related to the spacing between the two lines. The larger the spacing, the smaller the coupling coefficient. In order to meet the requirements of LTCC process, the spacing between the two lines is required to be greater than 0.1mm. For couplers with a coupling coefficient of less than -13.0dB, a narrow-edge coupler can be used.

[0021] According to the principle of energy conservation, when the energy at the input of the two unequal power divider is 1, the sum of the power at the two outputs, without considering physical losses (such as dielectric loss, conductor loss, etc.), is also 1. That is, if the output energy of one channel is greater, the energy of the other channel will definitely be less. Therefore, the power difference between the two outputs determines the distribution of the amplitude weights of the beamforming matrix. Figures 4(a) to 4(c)The diagram shows models of three unequal power dividers. The Wilkinson unequal power divider exhibits a relatively small power difference between its two outputs, with a maximum difference not exceeding 2dB. The power difference error primarily stems from the line width accuracy of the two circuits. The line width accuracy of LTCC technology can be controlled within 0.005mm, resulting in a relatively small power difference error. Simulation and actual test results show that its power difference error is less than 0.2dB. Given a fixed dielectric material and thickness, the power difference error of the wide-edge coupler mainly originates from the alignment accuracy between the through transmission lines and coupled transmission lines located on the two dielectric layers. For a 163mm × 163mm LTCC substrate, even with a high-precision automated stacking process, the alignment accuracy can theoretically only be controlled within ±0.02mm. In some areas of the substrate, the error may be larger. Based on simulation and test results, the maximum power difference error can reach 0.7dB, and this error is random across the entire substrate. The power difference error of a narrow-edge coupler originates from the spacing between the through transmission line and the coupled transmission line located on the same layer. Due to the relatively small coupling, the coupling coefficient is highly sensitive to the spacing between them. According to simulation and test results, the power difference error between the two output ports of this coupler can reach up to 1 dB, mainly due to the large variation in the coupling coefficient, while the variation in the through coefficient is very small. This error is random across the entire substrate.

[0022] As can be seen from the above analysis, due to the processing error of the LTCC substrate, the smaller the coupling amount of the coupler used, the greater the power difference error between the two output ports. When the entire circuit is connected in series with multiple couplers, the error will accumulate, leading to a larger error.

[0023] Because substrate processing errors are inevitable and random, they particularly affect small weights in the beamforming matrix, impacting the insertion loss of the entire beam when adjusting its amplitude accuracy. Let's take the amplitude weight of beam 1 as an example. Assume that due to processing errors, and ignoring other losses, the actual amplitude of feed 5 becomes -26.55dB, 2dB smaller than the target value. Since the feed's original weight was already small, its reduced amplitude has negligible impact on other channels, assuming the amplitude and amplitude weights of other channels are required to be consistent. In this case, the amplitude accuracy of feed 5 does not meet the requirements of Table 2. To make the amplitude accuracy of the feed meet the requirements, there are two methods. One is to increase the amplitude of this channel, but it is impossible to increase the amplitude when the substrate has already been processed. The other method is to reduce the amplitude of all other channels. This can be achieved by combining the attenuation network reserved on the surface of the substrate. Theoretically, increasing the attenuation of each other channel by 1.5dB can ensure that the amplitude accuracy of all channels meets the requirements (at this time, the amplitude accuracy of feed 10 is -0.5dB, and the amplitude accuracy of other channels is 0dB). However, if this is adjusted, the insertion loss of the entire beam will increase by 1.5dB due to the addition of the attenuation network. This increase in insertion loss is undesirable in the design.

[0024] Based on the above analysis, during the fabrication of the substrate for a large-scale beamforming matrix, due to the existence of fabrication errors that cannot be precisely controlled, if the circuit is designed according to the actual amplitude weights, it may result in some channels with small weights having amplitudes smaller than the required specifications. In order to ensure that the amplitude accuracy of such channels meets the requirements, the amplitudes of other channels must be reduced through attenuation networks. Although this satisfies the amplitude accuracy requirements, the insertion loss of the entire beam will increase, and may even exceed the insertion loss specifications. Summary of the Invention

[0025] To address the balance between amplitude accuracy tuning and insertion loss in large-scale beamforming matrices, this invention provides a method for preprocessing amplitude weights in large-scale beamforming matrices, which can achieve system-level energy balance between amplitude accuracy tuning and insertion loss in large-scale beamforming matrices.

[0026] A method for preprocessing amplitude weights in a large-scale beamforming matrix involves treating each beam as the current beam and performing preprocessing operations to preprocess the amplitude weights of each beam. The preprocessing operations are as follows:

[0027] Based on the total number of synthesis times of each feed in the beamforming matrix, the initial amplitude weight is obtained by superimposing the original amplitude weights between the current beam and the feeds it needs to connect to with the set amplitude weight margin. The amplitude weight margins corresponding to different total synthesis times are not exactly the same.

[0028] Increase the minimum value in the initial amplitude weights by the maximum margin ΔM. max The maximum value increases by the minimum value ΔM min The remaining amplitude weights are in ΔM max ~ΔM min Add a margin within the range to obtain the updated amplitude weight;

[0029] Convert each update magnitude weight into an energy value, and obtain the sum of the energy values ​​of all update magnitude weights except the maximum update magnitude weight. Then the energy value The corresponding magnitude weight replaces the maximum value in the updated magnitude weight, while the other updated magnitude weights remain unchanged.

[0030] Furthermore, the remaining magnitude weights are proportional to the minimum and maximum values ​​in ΔM. max ~ΔM min Increase the margin within the specified range, specifically as follows:

[0031] M new =M+((ΔM) min -ΔM max )×M-ΔM min ×M min +ΔM max ×M max ) / (ΔM max -ΔM min )

[0032] Among them, M new M represents the amplitude weight after adding a margin, and M represents the initial amplitude weight. max M represents the maximum value among all initial magnitude weights. min This represents the minimum value among all initial magnitude weights.

[0033] Furthermore, the method for converting each update magnitude weight into an energy value is as follows:

[0034]

[0035] Among them, M p M represents the energy value. dB This indicates the update magnitude weight.

[0036] Furthermore, after obtaining the final amplitude weight, the final amplitude weight is used as the target amplitude weight for circuit simulation design to obtain the actual amplitude weight of the circuit simulation design.

[0037] The difference between each actual amplitude weight and its corresponding original amplitude weight is obtained. The larger the difference, the greater the channel loss of the feed channel corresponding to that difference.

[0038] The channel containing the feed source with the maximum difference is taken as the reference channel. The channel loss of the other feed sources is then adjusted through the attenuator network of the beamforming matrix itself, so that the channel loss of the other channels is consistent with the loss of the reference channel.

[0039] Furthermore, when the beamforming matrix has 109 beams and 64 feeds, the total number of feed combining times 16 and 20 both correspond to a -3.0dB amplitude weight margin, the total number of feed combining times 24 corresponds to a -2.0dB amplitude weight margin, the total number of feed combining times 28 corresponds to a -1.5dB amplitude weight margin, the total number of feed combining times 32 corresponds to a -1.0dB amplitude weight margin, the total number of feed combining times 36, 40, and 44 all correspond to a 0.0dB amplitude weight margin, the total number of feed combining times 48 corresponds to a +0.5dB amplitude weight margin, and the total number of feed combining times 52 corresponds to a +1.5dB amplitude weight margin.

[0040] Beneficial effects:

[0041] 1. This invention provides a method for preprocessing amplitude weights of a large-scale beamforming matrix. It utilizes the different synthesis losses of each feed source due to the varying actual number of synthesis paths of the feed signal to set an amplitude weight margin for the superposition of each original amplitude weight. This is equivalent to compensating the power distribution circuit at the beam end with the synthesis loss margin of feed sources with fewer synthesis paths, thus reducing the amplitude weights of these feed sources. The reduced energy is used to increase the amplitude weights of other feed sources in the beam. Then, compensation is performed according to the amplitude weights after superposition margin from largest to smallest. Simultaneously, to satisfy energy conservation, the increased energy needs to be achieved by reducing the energy of large weights, ultimately achieving system-level energy balance between amplitude accuracy adjustment and insertion loss in the large-scale beamforming matrix.

[0042] 2. This invention provides a method for preprocessing amplitude weights of a large-scale beamforming matrix. After obtaining the initial amplitude weights of the superposition margin, compensation is performed according to the amplitude weights after superposition margin in descending order, with the smallest amplitude weight receiving the largest compensation and the largest amplitude weight receiving the smallest compensation. The remaining amplitude weights are then proportional to the minimum and maximum values ​​within ΔM. max ~ΔM min By increasing the margin within the range, the small-weight channels are prevented from becoming the main factor affecting beam loss, thus achieving system-level energy balance, while minimizing the impact of large-weight energy on beam loss.

[0043] 3. This invention provides a method for preprocessing amplitude weights of a large-scale beamforming matrix. After obtaining the final amplitude weights, the final amplitude weights are used as the target amplitude weights for circuit simulation design. This method can adjust the amplitude error caused by processing errors through an attenuation network without significantly increasing the insertion loss of the entire beam, so that the amplitude accuracy of all channels can meet the requirements. Attached Figure Description

[0044] Figure 1 A block diagram illustrating the principle of a beamforming matrix;

[0045] Figure 2 Block diagram of LTCC substrate for beamforming matrix;

[0046] Figure 3 This is the second-stage power distribution circuit at the beam end;

[0047] Figure 4(a) shows the Wilkinson unequal power divider;

[0048] Figure 4(b) shows the wide-side coupler;

[0049] Figure 4(c) shows a narrow-side coupler;

[0050] Figure 5 This is a flowchart of a large-scale beamforming matrix amplitude weight preprocessing method according to the present invention. Detailed Implementation

[0051] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings.

[0052] Based on the previous analysis of the beamforming matrix circuit design process, the smaller the amplitude weight, the lower the energy, and the more sensitive it is to substrate processing errors. Therefore, before circuit design, in order to effectively resolve the contradiction between amplitude weight distribution accuracy and insertion loss, this invention proposes an amplitude weight preprocessing method for large-scale beamforming matrices. The basic idea of ​​preprocessing is to reflect potential errors in the processing in the amplitude weight distribution. By preprocessing the amplitude weight distribution and designing the circuit according to the preprocessed amplitude weight distribution, the insertion loss can be minimized while ensuring amplitude accuracy.

[0053] A method for preprocessing amplitude weights in a large-scale beamforming matrix involves treating each beam as the current beam and performing preprocessing operations to complete the amplitude weight preprocessing for each beam. For example... Figure 5 As shown, the preprocessing operation is as follows:

[0054] Based on the total number of synthesis times of each feed in the beamforming matrix, the initial amplitude weight is obtained by superimposing the original amplitude weights between the current beam and the feeds it needs to connect to with the set amplitude weight margin. The amplitude weight margins corresponding to different total synthesis times are not exactly the same.

[0055] Increase the minimum value in the initial amplitude weights by the maximum margin ΔM. max The maximum value increases by the minimum value ΔM min The remaining amplitude weights are in ΔM max ~ΔM min Add a margin within the range to obtain the updated amplitude weight;

[0056] Convert each update magnitude weight into an energy value, and obtain the sum of the energy values ​​of all update magnitude weights except the maximum update magnitude weight. Then the energy value The corresponding magnitude weight replaces the maximum value in the updated magnitude weight, while the other updated magnitude weights remain unchanged.

[0057] The present invention will now be described using a beamforming matrix with 64 feeds forming 109 beams. Beamforming matrices of other sizes can also achieve a balance between amplitude accuracy and insertion loss using the present invention.

[0058] First, we will introduce how to preprocess the amplitude weights based on the actual number of synthesis iterations for each feed source. Figure 2In this process, the power combining of the feed source is performed in two stages. The first stage occurs on the combined board, with the combining circuits arranged in columns on the back of the substrate. To achieve equal power and equal phase combining, the first stage of feed source power combining circuits uses standard 16-in-1, 8-in-1, and 4-in-1 circuits. Within the same type of circuit, the combined feed sources are all of equal power and equal phase. However, different types of combining circuits are not of equal power and equal phase. For example, without considering physical losses, the combining loss of the 16-in-1 circuit is 6dB greater than that of the 4-in-1 circuit. Therefore, to achieve equal power combining for the same feed source, a power divider with unequal power must be used in the second stage of feed source combining to compensate for the loss difference in the first stage of power combining circuit. Assuming that feed source 1 appears three times on the 11 combined boards, namely 16-in-1, 8-in-1, and 4-in-1, then a 4:2:1 unequal power combiner is needed to combine these three signals in the second stage of combining, where 4 corresponds to 16-in-1, 2 corresponds to 8-in-1, and 1 corresponds to 4-in-1. After the second synthesis, feed 1 can achieve equal-power synthesis, with 28 synthesis cycles and a corresponding synthesis loss of 14.47 dB. Since the beamforming matrix itself is a sparse matrix, the number of synthesis cycles required for each feed is not the same. In actual design, the minimum number of synthesis cycles for a single feed is 16 (synthesis loss 12 dB), and the maximum is 52 (synthesis loss 17.16 dB). Therefore, while each feed has equal-power synthesis within its own section, different feeds will have different synthesis losses due to the varying number of synthesis cycles. Without considering physical losses, the synthesis loss from 52 synthesis cycles is 5.16 dB greater than that from 16 synthesis cycles. Therefore, based on the loss from 52 synthesis cycles, the feed loss from 16 synthesis cycles has a margin of 5.12 dB, which can be used for preprocessing the amplitude weights.

[0059] The actual number of times the feed source is synthesized includes 16, 20, 24, 28, 32, 36, 40, 44, 48, and 52. The amplitude weight can be preprocessed according to the number of times each feed source is synthesized. The relationship between the preprocessed amplitude weight and the number of times the feed source is synthesized is shown in Table 5.

[0060] Table 5. Relationship between amplitude weight preprocessing value and feed synthesis number.

[0061]

[0062] The second type of preprocessing for amplitude weights involves increasing the amplitude weights of the smaller weight channels in advance, according to their magnitude. To satisfy energy conservation, this increased energy needs to be achieved by reducing the energy of the larger weight channels. However, for the larger weight channels, this reduction has little impact on beam loss and is acceptable.

[0063] By combining the two preprocessing methods above, the amplitude weights of each beam can be preprocessed before circuit design. The preprocessing steps are described below using beam 1 as an example.

[0064] The amplitude weight distribution table for beam 1 is shown in columns 1 and 2 of Table 6. Column 1 is the feed number involved in this beam, and column 2 is the amplitude weight.

[0065] Table 6. Beam 1 Amplitude Weight Preprocessing Process

[0066]

[0067]

[0068] Step 1: Based on the actual total number of synthesizations for each feed in the beam (column 3 of Table 6), perform the first preprocessing on each amplitude weight according to the method in Table 5. The result after preprocessing is shown in column 4 of Table 6, and is called the initial amplitude weight.

[0069] Step 2: Based on Step 1, select the channel with the smallest initial amplitude weight, and increase the amplitude weight of this channel by ΔM. max =3dB. If the amplitude weight of beam 1 to feed 7 is the smallest (column 4 in Table 6), increase its amplitude weight by 3dB margin, from -26.05dB to -23.05dB; select the channel with the largest amplitude weight, and increase the amplitude weight by ΔM. min =0dB margin, other channels follow the linear proportional relationship of amplitude weights in column 4 of Table 6 in ΔM max ~ΔM min The following calculation method applies to the addition of a margin within the specified range:

[0070] M new =M+((ΔM) min -ΔM max )×M-ΔM min ×M min +ΔM max ×M max ) / (ΔM max -ΔM min M new M is the preprocessed amplitude weight, and M is the initial amplitude weight. max M min These are the maximum and minimum values ​​among all initial magnitude weights, ΔM. max ,ΔM min These are the maximum and minimum values ​​of the added margin, which can be preset according to actual conditions. This invention uses ΔM... max =3dB,ΔM min =0dB is used for explanation.

[0071] This updates the amplitude weights of all channels; the smaller the amplitude weight, the greater the increase in margin. To ensure energy conservation, the updated amplitude weights need to be processed. This processing involves calculating the energy of all channels except those with the largest amplitude weight. The calculation formula is as follows: Among them, M p M represents the energy value. dB This involves updating the amplitude weights and then obtaining the total energy of these channels. Since the total energy is 1, the energy of the channel with the largest amplitude weight is... Finally, according to M dB =10log(1-M) pt The updated maximum weights are obtained. Column 5 of Table 6 shows the distribution of amplitude weights after the second preprocessing. The sum of the energy of these amplitude weights is 1, satisfying energy conservation. Column 6 of Table 6 shows the difference between the amplitude weights after the second preprocessing and those after the first preprocessing.

[0072] Column 7 of Table 6 shows the difference between the amplitude weights after two preprocessing steps and the original amplitude weights. The table shows that the amplitude weights of some channels increased, especially those with smaller amplitude weights, while those of others decreased. This is mainly due to the fewer feed combining operations and lower combining losses in these channels, allowing excess feed energy to be transferred to the beam end. This energy can then be used to compensate for channels in the beam that require increased energy. This approach provides higher redundancy for substrate processing errors in the smaller weight channels while maintaining energy conservation.

[0073] The circuit simulation design was performed according to the preprocessed amplitude weights in column 5 of Table 6, and the design results are shown in Table 7. In Table 7, column 1 is the feed number involved in beam 1, column 2 is the original amplitude weight, column 3 is the preprocessed amplitude weight, which is also the target value for circuit design, column 4 is the circuit simulation design result, and column 5 is the difference between the circuit simulation design value and the original amplitude weight. This difference can also be regarded as the channel loss to each feed (including physical losses such as dielectric loss and metal loss, as well as power combining loss at the feed end). Channel 18, which has the highest loss among all channels, is selected as the reference channel (it can also be used as the loss of this beam). The difference between the loss of the other channels and that of this channel is the design margin of the channel, as shown in column 6 of Table 7. It can be seen from this column that the reference channel of this beam is the channel with the largest amplitude weight. This is because some of the energy borrowed during amplitude weight preprocessing comes from this channel, and the channel with a large amplitude weight is less affected by substrate processing errors. In addition, the smaller the amplitude weight, the larger the margin, which can compensate for the impact of processing errors. Even without processing errors, the attenuator network pre-designed on the surface of the composite substrate in the system can be used for debugging. Through the attenuation network, the loss of all channels is made consistent with that of channel 18. In this way, the loss of the entire beam is basically the same as that of channel 18, and the amplitude accuracy of the other channels can be guaranteed.

[0074] Table 7 Simulation Design Results of Beam 1

[0075]

[0076] Table 8 shows the overall test results for Beam 1, without amplitude adjustment. Column 3 of Table 8 corresponds to the simulation design values ​​in column 4 of Table 7, using feed channel 18 as a baseline. There is approximately a 2dB difference between the two. This difference arises because the overall test includes the loss of the RF cables connecting the substrates, while Table 7 only shows the cascaded results of the substrate circuits and does not consider the loss of the connecting cables. Due to manufacturing errors, the channel with the highest insertion loss in Table 8 is feed channel 10. Using this channel as a baseline (as the loss of this beam), the difference between the amplitude values ​​of other channels and this channel is shown in column 5 of Table 8. The data in column 5 shows that due to manufacturing errors, the margin for each channel has changed. For example, the margin for channel 5 decreased from the designed 4.26dB to 1.51dB. Although it decreased, it can still be adjusted to around 0dB using an attenuation network. In theory, the amplitude accuracy of all channels can be adjusted to around 0dB by using the attenuation network bits reserved on the surface of the composite substrate. The beam insertion loss is determined by the high-weight channel (feed 10), and the amplitude of the low-weight channel has been resolved by preprocessing due to the influence of processing errors.

[0077] Table 8. Test Results of Beam 1 System

[0078]

[0079] It should be noted that beam power distribution circuits in large-scale beamforming matrices mostly employ couplers. The coupling coefficient of the coupler is highly sensitive to the precision of the substrate fabrication, and the smaller the coupling coefficient, the more sensitive it is to fabrication errors. After the accumulation of errors across multiple circuit stages, the output amplitude of small-weight channels may decrease significantly. To ensure that the amplitude weight accuracy of a particular channel meets the requirements, attenuation networks designed on the substrate surface must be used to attenuate other channels, ensuring that all channels meet the amplitude accuracy requirements. However, this also increases the overall beam insertion loss, potentially causing the beam insertion loss to fail to meet the requirements. This invention proposes an amplitude weight preprocessing method. By preprocessing the amplitude weights, a design margin is increased for small-weight channels. Even if the amplitude of small-weight channels decreases due to fabrication errors, this pre-designed margin will compensate for it, thus preventing small-weight channels from becoming the main factor affecting beam insertion loss.

[0080] In general, the preprocessing of amplitude weights in this invention is divided into two steps. The first step utilizes the different synthesis losses of each feed source due to the different actual synthesis number of feed source signals. The synthesis loss margin of feed sources with fewer synthesis times can be compensated to the power distribution circuit at the beam end, thereby reducing the amplitude weights of these feed sources. The reduced energy is used to increase the amplitude weights of other feed sources in the beam.

[0081] After the first amplitude weight preprocessing, compensation is performed according to the updated amplitude weight from largest to smallest. The maximum and minimum compensation values ​​can be preset, with the smallest amplitude weight receiving the largest compensation and the smallest amplitude weight receiving the smallest compensation. The compensation value of each feed source is linearly related to the magnitude of the feed source amplitude weight.

[0082] After compensating all channels, the channel with the largest amplitude weight is corrected so that the sum of the energy of all channels equals 1. This satisfies energy conservation and allows the amplitude weight distribution to be implemented using physical circuitry. The physical meaning of this correction step is to sacrifice the energy of the channel with the largest amplitude weight to compensate for the other channels. As a result, the loss of the channel with the largest amplitude weight becomes the loss benchmark for the entire beam. Although this increases the overall beam loss, the increase is relatively small and controllable.

[0083] Therefore, the amplitude weight preprocessing proposed in this invention actually compensates for manufacturing errors through energy transfer. This is primarily because channels sensitive to manufacturing errors inherently transmit relatively little energy. Therefore, transferring some energy from high-weight channels to low-weight channels has little impact on high-weight channels or the entire beam, but it increases the redundancy of low-weight channels, preventing them from becoming the main factor affecting beam loss. The amplitude weight preprocessing also fully utilizes the fact that different feed sources in the beamforming matrix system have different actual synthesis times. This difference results in some feed sources having relatively low synthesis losses. The excess energy from these feed sources can be transferred to the beam power distribution end. By reducing the amplitude weights of these feed sources, the excess energy is used to compensate other channels, achieving system-level energy balance.

[0084] Of course, the present invention may have other various embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding changes and modifications according to the present invention, but these corresponding changes and modifications should all fall within the protection scope of the appended claims.

Claims

1. A method for pre-processing of amplitude weights of a large-scale beamforming matrix, characterized in that, The pre-processing operation is performed on each beam as a current beam to complete amplitude weight pre-processing of each beam, and the pre-processing operation is: According to the total synthesis times of each feed in the beam forming matrix, an amplitude weight margin is set for the original amplitude weight superposition between the current beam and the feed to which the current beam needs to be connected, to obtain an initial amplitude weight, wherein the amplitude weight margins corresponding to different total synthesis times are not completely the same; increasing the minimum value in the initial amplitude weight by a margin maximum value , increasing the maximum value by a margin minimum value , and increasing the remaining amplitude weights in the initial amplitude weight by a margin within a range of to obtain updated amplitude weights; convert the update amplitude weight into an energy value, and obtain the energy sum of the energy values of all update amplitude weights except the maximum update amplitude weight , and then convert the energy value The corresponding amplitude weight replaces the maximum value in the update amplitude weight, and the remaining update amplitude weights remain unchanged.

2. A method for pre-processing of amplitude weights of a large-scale beamforming matrix as claimed in claim 1, characterized by, The remaining amplitude weights are proportional to the minimum and maximum values. Increase the margin within the specified range, specifically as follows: wherein, represents the amplitude weight after the increase of the margin, represents the initial amplitude weight, represents the maximum value among all the initial amplitude weights, represents the minimum value among all the initial amplitude weights.

3. A method for pre-processing of amplitude weights of a large-scale beamforming matrix as claimed in claim 1, wherein, The method for converting each updated amplitude weight into an energy value is: wherein, represents an energy value, represents an update amplitude weight.

4. A method for pre-processing of amplitude weights of a large-scale beamforming matrix as claimed in claim 1, wherein, After obtaining the final amplitude weight, the final amplitude weight is taken as a target amplitude weight for circuit simulation design to obtain actual amplitude weights of the circuit simulation design; The difference between each actual amplitude weight and the corresponding original amplitude weight is obtained, wherein the greater the difference, the greater the channel loss of the channel in which the feed corresponding to the difference is located; The channel in which the feed corresponding to the maximum difference is located is taken as a reference channel, and the channel losses of the channels in which the remaining feeds are located are adjusted through the attenuator network of the beam forming matrix itself, so that the channel losses of the remaining channels are consistent with the loss of the reference channel.

5. A method for pre-processing amplitude weights of a large-scale beamforming matrix as claimed in any one of claims 1 to 4, characterized in that, When the beam forming matrix has 109 beams and 64 feeds, the total synthesis times 16 and the total synthesis times 20 of the feeds both correspond to-3.0 dB amplitude weight margin, the total synthesis times 24 of the feeds correspond to-2.0 dB amplitude weight margin, the total synthesis times 28 of the feeds correspond to-1.5 dB amplitude weight margin, the total synthesis times 32 of the feeds correspond to-1.0 dB amplitude weight margin, the total synthesis times 36, the total synthesis times 40 and the total synthesis times 44 of the feeds all correspond to 0.0 dB amplitude weight margin, the total synthesis times 48 of the feeds correspond to +0.5 dB amplitude weight margin, and the total synthesis times 52 of the feeds correspond to +1.5 dB amplitude weight margin.

Images