Circuit board printing control method and system based on signal transmission loss

Abstract

The invention discloses a circuit board printing control method and system based on signal transmission loss. And obtaining a target differential transmission line parameter and a circuit board lamination parameter. And establishing an electromagnetic simulation model, and simulating insertion loss, return loss and impedance discontinuous point distribution. The line width, the line spacing and the interlayer hollowed-out area of the differential transmission line are dynamically adjusted through the reinforcement learning model, the proportion of the line width to the shielding layer spacing is limited by a preset critical value, and the line spacing is configured based on the multiple relation of the shielding layer spacing. And a sectional optimization strategy is adopted for the corners of the transmission line. The ratio of the aperture to the diameter of the anti-pad is optimized, and impedance mutation is reduced. According to the method, the key parameters of the target differential transmission line and the circuit board lamination parameters are acquired, and the electromagnetic simulation model is combined with the multi-physical field coupling algorithm, so that the insertion loss, the return loss and the impedance discontinuous point distribution can be accurately simulated. The method improves the adjustment efficiency, guarantees the accuracy of an adjustment result, and effectively reduces the signal transmission loss.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of circuit board design, and particularly relates to a circuit board printing control method and system based on signal transmission loss. BACKGROUND

[0002] In the field of modern electronics manufacturing, circuit board printing technology, also known as PCB printing technology, is a core link in the electronic manufacturing process. It involves printing conductive materials such as copper onto an insulating substrate to form a circuit pattern, thereby achieving precise connections between electronic components. With the continuous miniaturization and high performance of electronic devices, circuit board printing technology must provide higher precision and more reliable connections to meet the growing performance demands.

[0003] With the advancement of technology, circuit board printing technology has evolved from traditional silk screen printing to more advanced photolithography technology, and recently laser direct imaging (LDI) technology. The advancement of these technologies allows for finer line widths and spacing on circuit boards, enabling the integration of more electronic components in a limited space. In addition, high-precision printing technology has greatly improved the performance and reliability of electronic devices, making it possible to achieve more complex functions.

[0004] Patent application CN116056341A discloses a wiring method and system for signal lines of a printed circuit board, which divides the signal lines into dynamic lines and static lines. According to the basic properties of the signal lines, such as characteristic impedance, signal frequency, and expected crosstalk rate, an equivalent transmission line model is constructed. Then, by simulating the crosstalk of the model, the actual crosstalk situation is obtained. Finally, according to the comparison between the simulation results and the target crosstalk value, the line spacing between dynamic lines and static lines and the thickness of the dielectric layer between them are scientifically designed. This scheme aims to solve the problems of excessive crosstalk rate and low space utilization in traditional wiring methods. In reality, circuit board printing still faces various problems, such as signal transmission loss. Signal transmission loss is one of the important factors affecting the performance of circuit boards, especially in high-frequency applications, where signal attenuation, reflection, and crosstalk can severely affect the performance of circuit boards. To reduce these losses, fine control of signal paths is required during circuit board design and manufacturing, which includes optimizing wire width, spacing, and layer structure. SUMMARY

[0005] The present application aims to overcome the shortcomings of the prior art and provide a circuit board printing control method and system based on signal transmission loss. By obtaining the key parameters of the target differential transmission line and the circuit board layer parameters, accurate data basis is provided for subsequent operations. Using electromagnetic simulation models combined with multi-physical field coupling algorithms, the insertion loss, return loss, and impedance discontinuity point distribution are accurately simulated, providing a scientific basis for the optimization of transmission lines.

[0006] The object of the present application is achieved by the following technical solutions:

[0007] A circuit board printing control method based on signal transmission loss, comprising the following steps: obtaining a target single-ended impedance Z0, a differential impedance Z diff and a circuit board stack parameter, the stack parameter comprising a shielding layer spacing h, a dielectric constant ε r and a copper thickness t;

[0008] Establishing an electromagnetic simulation model based on the shielding layer spacing, the dielectric constant and the copper thickness, simulating the insertion loss IL, the return loss RL and the impedance discontinuity point distribution by combining the multi-physical field coupling algorithm;

[0009] Dynamically adjusting the line width w, the line spacing s and the interlayer hollow area of the differential transmission line through a reinforcement learning model, wherein the ratio of the line width to the shielding layer spacing is limited by a preset critical value, and the line spacing is configured based on a multiple relationship with the shielding layer spacing;

[0010] Using a segmented optimization strategy for the transmission line corner, and the corner angle range is 90°-180°;

[0011] Optimizing the ratio of the via parasitic capacitance model aperture D drill to the anti-pad diameter D anti , and reducing the impedance mutation by hollowing out the adjacent reference layer;

[0012] Outputting the optimized layout rule, and generating a PCB printing control instruction in combination with real-time process data.

[0013] As a preferred mode, the line width w satisfies the formula:

[0014]

[0015] Wherein, α∈[5.0,6.5], β∈[5.5,6.5], γ∈[0.7,0.9], and α, β, γ are updated through feedback IL and RL.

[0016] As a preferred mode, the line spacing s satisfies the formula:

[0017]

[0018] Wherein, k∈[1.2,1.8], δ∈[0.8,1.2], and δ is calibrated through an online impedance monitoring module.

[0019] As a preferred mode, the via optimization comprises:

[0020] When D drill ≤D th , D anti ≥λDdrill And the ratio of via spacing to w is μ:1;

[0021] When D drill > D th , D anti ≥ ηD drill , and the impedance continuity is verified by the neural network;

[0022] Wherein, D th ∈ [6, 10] mil, λ ∈ [2.0, 2.5], μ ∈ [4, 6], η ∈ [2.0, 2.5].

[0023] As a preferred mode, the corner optimization includes:

[0024] For a transmission line with a length L th ≥ 500 mil, a segmented corner routing is adopted and the angle is dynamically adjusted;

[0025] A compensation capacitor array is introduced at the corner, and the capacitance C satisfies:

[0026]

[0027] Where ΔZ is the impedance deviation value at the corner, K ∈ [0.8, 1.2], and f is the highest frequency of the signal.

[0028] As a preferred mode, it further includes: controlling the copper foil roughness Ra ≤ 0.5 μm by laser micro-etching technology, and using electroplating process to make the loss increment not more than θ%, wherein θ ∈ [3, 5].

[0029] As a preferred mode, the dynamic adjustment includes: when IL exceeds the threshold value, w is preferentially increased and the cooperative optimization of D anti is triggered; when RL exceeds the threshold value, the hollowed-out area is preferentially optimized and C is adjusted in parallel.

[0030] A circuit board printing control system based on signal transmission loss, applicable to the circuit board printing control method based on signal transmission loss described above, comprising:

[0031] A parameter acquisition module for collecting Z0, Z diff , f, h, ε r , and t;

[0032] A multi-physical field coupling simulation engine for calculating IL, RL, and impedance discontinuity points;

[0033] A reinforcement learning optimization module for adjusting w, s, and the hollowed-out area;

[0034] A via analysis unit for generating a via scheme based on the ratio of D drill to D anti .

[0035] An instruction generation module converts the optimization parameters into PCB control instructions.

[0036] As a preferred mode, the reinforcement learning module comprises a critical value database storing the maximum allowable proportion of w / h at different f, and is updated in linkage with real-time process data.

[0037] As a preferred mode, a roughness detection unit is further included to feedback the Ra three-dimensional topography through terahertz wave scanning and adjust the electroplating parameters in linkage.

[0038] The present application has at least the following beneficial effects: by obtaining the key parameters of the target differential transmission line and the circuit board lamination parameters, accurate data basis is provided for subsequent operations. By using the electromagnetic simulation model combined with the multi-physical field coupling algorithm, the insertion loss, return loss and impedance discontinuity point distribution can be accurately simulated, providing a scientific basis for the optimization of the transmission line. By dynamically adjusting the line width, line spacing and interlayer hollow area of the differential transmission line through the reinforcement learning model, not only the adjustment efficiency is improved, but also the accuracy of the adjustment result is ensured, effectively reducing the signal transmission loss. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to reveal the technical details of the embodiments of the present application, the drawings involved in the embodiments will be briefly introduced next. It should be emphasized that these drawings only present several embodiments of the present application, and should not be regarded as a limitation on the scope of the application. For those skilled in the art, other related drawings can be derived based on these drawings without creative labor.

[0040] Figure 1 A flow chart of a circuit board printing control method in the embodiment;

[0041] Figure 2 A schematic structural diagram of a circuit board printing control system in the embodiment. DETAILED DESCRIPTION

[0042] The technical solutions of the present application will be described in further detail below with reference to the drawings, but the protection scope of the present application is not limited to the following description.

[0043] In the following content, the embodiments of the present disclosure will be described in detail with the help of the drawings. However, it should be clear that the present disclosure is not limited to the specific forms shown here. On the contrary, it should be understood to cover various variations, equivalent forms and / or alternatives of the embodiments of the present disclosure. In the process of describing the drawings, the same reference numerals will be used to indicate similar components.

[0044] It should be clarified that while the following description provides detailed specific information to aid in a comprehensive understanding of the exemplary embodiments, those skilled in the art will recognize that the exemplary embodiments can be implemented even without these specific details. For example, the system may be illustrated using block diagrams to avoid excessive detail that could obscure the clarity of the example. In other cases, to maintain the clarity of the example, unnecessary details of well-known processes, structures, and techniques may be omitted.

[0045] like Figure 1 As shown, a circuit board printing control method includes the following steps: S1: Obtain the target single-ended impedance Z0 and differential impedance Z of the target differential transmission line. diff and circuit board stack-up parameters, including shielding layer spacing h, dielectric constant ε r and copper thickness (t);

[0046] S2: An electromagnetic simulation model is established based on the shielding layer spacing, dielectric constant, and copper thickness. The insertion loss IL, return loss RL, and impedance discontinuity distribution are simulated by combining a multiphysics coupling algorithm.

[0047] S3: The linewidth w, line spacing s, and interlayer cutout area of ​​the differential transmission line are dynamically adjusted through a reinforcement learning model. The ratio of linewidth to shielding layer spacing is limited by a preset critical value, and the line spacing is configured based on a multiple relationship of shielding layer spacing.

[0048] S4: A segmented optimization strategy is adopted for transmission line corners, combined with an adaptive compensation mechanism to shorten the equivalent line length, with the corner angle range being 90°-180°.

[0049] S5: Optimize the aperture D based on the via parasitic capacitance model. drill With anti-pad diameter D anti The proportion of [the target] is reduced by hollowing out adjacent reference layers to decrease impedance abruptness;

[0050] S6: Outputs optimized layout rules and generates PCB printing control instructions by combining them with real-time process data.

[0051] This embodiment discloses a circuit board printing control method based on signal transmission loss. By modeling and optimizing the key parameters of differential transmission lines, the insertion loss and return loss of high-speed signals during transmission are minimized, and the impact of impedance discontinuity is reduced.

[0052] First, the system acquires the basic electrical parameters of the target differential transmission line, including the target single-ended impedance Z0 and differential impedance Z. diff Where Z0 refers to the characteristic impedance relative to the reference ground when one conductor in the differential pair operates independently; while Z... diffThis means that the total impedance exhibited when two conductors simultaneously transmit signals of opposite polarities is equal to twice the single-ended impedance minus the effect of the coupling portion (i.e., Z). diff =2Z0(1-k), where k represents the coupling coefficient). In addition, it is necessary to obtain the stack-up structure parameters of the circuit board, such as the spacing h between shielding layers and the relative permittivity ε of the dielectric material. r The parameters, including the copper foil thickness (t), collectively determine the propagation speed, attenuation, and distributed capacitance and inductance of the electromagnetic wave in the transmission line. The dielectric loss tangent (tanδ) directly affects the insertion loss IL, with the relationship: IL∝αL, where α is the dielectric loss coefficient and L is the transmission line length.

[0053] A three-dimensional electromagnetic simulation model was established based on the above parameters. This model utilizes numerical calculation methods such as the finite element method (FEM) or the finite-difference time-domain method (FDTD), combined with multiphysics coupling algorithms (such as the FEM-BEM hybrid method), to simulate the insertion loss IL and return loss RL of the differential transmission structure at the actual operating frequency. Insertion loss refers to the energy attenuation caused by material absorption, radiation, scattering, and other factors during signal transmission from the input port to the output port, and is usually defined as: Among them, P in P represents the input power. out This represents the output power. Return loss reflects the signal reflection intensity caused by impedance mismatch, and its expression is: RL = 20log 10 (|Γ|), where Γ is the voltage reflection coefficient. Z L The load impedance is represented by the simulation result. The simulation can also identify the location and magnitude of impedance abrupt changes, forming an "impedance discontinuity distribution map," which provides a basis for subsequent optimization.

[0054] A reinforcement learning model is introduced to dynamically optimize the geometric layout of differential transmission lines. This model takes the current layout parameters (such as linewidth w, line spacing s, and the shape of the cutout area) as input, and uses insertion loss, return loss, and impedance fluctuation as objective functions. Through a trial-and-error strategy, it finds a set of optimal design parameters. Specifically, the selection of the linewidth w is limited by a preset proportional relationship, such as w / h < α, where α is a safety threshold set based on the material properties to prevent increased loss or crosstalk due to excessive linewidth. The line spacing s is usually set to a multiple of the shielding layer spacing h to ensure sufficient isolation and control coupling effects. s = nh, where n is an integer multiple, can reduce edge field interference.

[0055] For the corner part in the transmission line, the system adopts a segmented optimization strategy, i.e. dividing a corner into multiple small segments and performing local optimization respectively. On this basis, an adaptive compensation mechanism is introduced to dynamically adjust the line length according to the change of the corner angle, so as to reduce the additional phase delay and reflection caused by the broken line. The angle range of the corner is limited between 90° and 180°, which ensures the flexibility of the layout while avoiding the significant electromagnetic concentration effect of high-frequency signals at sharp corners. For the 90° corner, by changing the right angle to an arc or a cut corner, the edge radiation loss and impedance discontinuity can be reduced.

[0056] In addition, for the design of via structure, the system also optimizes the aperture D drill based on the parasitic capacitance model anti . The parasitic capacitance C via of the via can be approximately expressed as:

[0057]

[0058] where ε0 is the vacuum permittivity, A pad is the area of the anti-pad, and d is the via depth. By adjusting the ratio of D drill and D anti , the parasitic capacitance can be effectively reduced, thereby reducing signal delay and bandwidth limitation. At the same time, hollowing out the adjacent reference layer (such as the ground layer or power layer) can reduce the coupling capacitance between the via and the reference layer, alleviating the impedance discontinuity problem.

[0059] The system outputs the optimized layout rules and generates PCB printing control instructions in combination with real-time process data (such as etching precision, lamination error, etc.). These instructions directly drive the laser engraving or photolithography process through numerical control equipment, ensuring high-fidelity implementation of design parameters in the manufacturing stage and achieving signal integrity optimization of high-speed circuit boards.

[0060] It should be noted that when establishing a three-dimensional electromagnetic simulation model based on target differential transmission line parameters / circuit board stack parameters, the goal is to solve Maxwell's equations by numerical methods to simulate the distribution and behavior of electromagnetic fields in complex geometric structures. We divide it into the following steps: geometric modeling, material definition, mesh division, boundary condition setting, solver selection, and post-processing analysis, as follows:

[0061] Firstly, geometric modeling requires constructing a three-dimensional structure according to design requirements. For a differential transmission line, parameters such as line width w, line spacing s, and shielding layer spacing h need to be defined accurately, and the interlayer hollow area or via structure needs to be considered. Modeling tools can use built-in geometry generators (such as cubes, cylinders, etc.) or import external CAD files (such as STEP, STL format). In CST or ANSYS HFSS, users can combine basic geometric bodies through Boolean operations to form a complete PCB stack model. Key parameters need to meet physical constraint conditions, such as the ratio limit w / h < α of line width and shielding layer spacing, where α is a safety threshold set according to the performance of the board.

[0062] Material property definition needs to assign electromagnetic characteristics to different regions in the model. Conductors (such as copper wires) need to set electrical conductivity σ, and dielectric layers (such as FR-4) need to define relative permittivity ε r and loss tangent tan δ. These parameters directly affect the propagation speed v p and attenuation coefficient α' of electromagnetic waves, which can be related by the formula:

[0063]

[0064] where c is the speed of light, ω is the angular frequency, and ε0 is the vacuum permittivity.

[0065] Meshing is a key step in discretizing the geometric model. Tetrahedral or hexahedral meshes are usually used, and high gradient areas (such as conductor-dielectric interfaces) need to be meshed to improve accuracy. For example, in ANSYS HFSS, adaptive meshing technology can automatically adjust the mesh density to ensure that the local error is less than the preset threshold. Mesh quality directly affects the efficiency and accuracy of the calculation, and numerical instability caused by distorted meshes needs to be avoided.

[0066] Boundary conditions and excitation source settings need to be selected according to the simulation target. For a differential transmission line, common boundary conditions include: perfect matched layer (PML), which absorbs electromagnetic waves to simulate infinite space; periodic boundary condition (PBC), which is used for periodic structures (such as array antennas); ideal conductor boundary (PEC), which simulates the tangential electric field of a metal surface to be zero. The excitation source can be set as a wave port or a voltage source to drive the differential signal transmission. For example, a wave port can be defined as:

[0067]

[0068] where E0 is the input electric field amplitude, k0 = ω / c is the free-space wave number, ω is the angular frequency, c is the speed of light, E in (x, y, z) is the complex expression of the input electric field, representing the electric field intensity at a certain point (x, y, z) in three-dimensional space, z is the direction of electromagnetic wave propagation, and j here represents the imaginary unit.

[0069] Solver selection needs to match the algorithm according to the problem type. High frequency problems are usually solved by Finite Element Method (FEM) or Method of Moments (MoM), while wideband transient response is suitable for Finite Difference Time Domain (FDTD). In CST, FDTD is suitable for analyzing the influence of via parasitic capacitance on signal integrity, and its discrete form is:

[0070]

[0071] The time-domain electromagnetic field distribution can be obtained by iteratively solving the above equation. is the rate of change of magnetic field intensity H with respect to time t, is the rate of change of electric field intensity W with respect to time t. μ is the magnetic permeability, ε is the dielectric constant, is the curl operator.

[0072] Post-processing analysis needs to extract key performance indicators such as insertion loss IL, return loss RL, impedance discontinuity point distribution, etc. For example, the insertion loss calculation formula is: The impedance mutation position can be intuitively identified by the field distribution diagram (such as electric field intensity E or current density J).

[0073] The simulation results need to be compared with the theoretical values or measured data for verification. If the deviation is large, the parameters (such as line width w, line spacing s) or grid density need to be adjusted in the modeling stage, forming a closed-loop design process. Through this process, the layout design of PCB can be effectively guided, and the integrity and stability of high-speed signal transmission can be improved. In a preferred embodiment, a reinforcement learning model optimizes the geometry layout of differential transmission lines. Including problem modeling and reinforcement learning framework, the optimization of the geometry layout of differential transmission lines is modeled as a Markov decision process, including the following contents. State space, state S includes current design parameters and constraint conditions. Geometric parameters: line width w, line spacing s, hollow region shape (such as rectangle, trapezoid, etc.), dielectric layer thickness h, material properties of the stack (such as ε r ). Constraint conditions: w / h < α, where α is a safety threshold set according to the performance of the board (to prevent excessive line width from causing loss or crosstalk). s = nh (n is an integer multiple to ensure isolation). Other physical limitations (such as minimum line width w min , maximum line width w max ). Action space: action A is a fine-tuning strategy for state parameters. Continuous action: adjust the increment of line width w or line spacing s (such as Δw ∈ [-Δmax, Δmax]). Discrete action: select the shape of the hollow region (such as switching from rectangle to trapezoid). Combined action: adjust multiple parameters at the same time (such as joint optimization of w and s). For the reward function, the reward R is directly related to the objective function (insertion loss IL, return loss RL, impedance fluctuation Z fluct ): R = -(β1·IL + β2·RL + β3·Zfluct ), where β1, β2, β3 are weight coefficients to balance the importance of different objectives. Insertion loss IL: reflects the energy loss of the signal during transmission, the smaller the better. Return loss RL: reflects the degree of signal reflection, the larger the better (i.e. the smaller the reflected energy). Impedance fluctuation Z fluct : reflects the consistency of impedance, the smaller the fluctuation the better. Termination condition (any of the following conditions): ① reaches the maximum number of iterations. ② the reward value converges to a stable interval (such as the reward changes less than the threshold δ for consecutive N iterations). ③ the parameters violate the physical constraints (such as w / h≥α).

[0074] In addition, the selection and implementation of reinforcement learning algorithms are also included. For the continuity of the parameter space and complex constraints of the differential transmission line design, the following algorithms can be selected: ① Proximal Policy Optimization. Advantages: suitable for continuous action space, supports high-dimensional state representation, and strong stability. Implementation: State encoding: normalize parameters w, s, h, ε r , etc. to vector S ∈ R n . Policy network π θ : input S, output action A (such as Δw, Δs). Value function network V φ : evaluates the long-term reward expectation of state S. Training process: calculate IL, RL, Z fluct , generate reward signal R, update policy network and value function network. ② Deep Deterministic Policy Gradient. Advantages: suitable for continuous control problems, combining Q-learning and policy gradient. Implementation: Action space: directly output parameter adjustment amount (such as Δw, Δs). Accelerate convergence through experience replay mechanism.

[0075] For the handling of constraints. Reinforcement learning models need to dynamically satisfy design constraints (such as w / h < α, s = nh), which can be achieved through the following ways: ① Hard constraints. Action clipping: directly truncate values that exceed the constraint range when generating action A. For example: w / h≥α, then set Δw to 0. Virtual penalty: add a penalty term to the reward function: R = -(IL+RL+Z fluct )-γ·max(0,w / h-α), where γ is the Lagrange multiplier, adjusted iteratively to satisfy the constraints. ② Soft constraints. Probability distribution design: set constraint boundaries for the sampling distribution of action A (such as Gaussian distribution) in the policy network. For example: if s = nh, then the sampling range of Δs is limited to ±h. Lagrange multiplier method: convert the constraint into part of the objective function: where λ is the Lagrange multiplier, adjusted iteratively to satisfy the constraints.

[0076] Simulation and training process: ① Initialization: Randomly generate initial parameters w, s, h, satisfying w / h < a, s = nh. Calculate initial IL, RL, Z using electromagnetic simulation tools (e.g. HFSS) fluct . ② Interaction and learning: The model attempts random actions (e.g. adjusting w or s), generating new parameters. Simulation verification: Run electromagnetic simulation on new parameters, calculate IL, RL, Z fluct . Reward calculation: Generate reward R according to objective function and constraints. Strategy update: Update policy network and value function network through PPO / DDPG algorithm. ③ Convergence judgment: If the reward value converges (e.g. the change is less than d for consecutive N iterations), output the optimal parameters. If not converged, return to step ② for further training.

[0077] In a preferred embodiment, the line width w satisfies the formula:

[0078]

[0079] where a ∈ [5.0, 6.5], β ∈ [5.5, 6.5], γ ∈ [0.7, 0.9], and a, β, γ are updated by feedback IL and RL. And a, β, γ are iteratively optimized to make IL and RL lower than the preset threshold. The line width w satisfies the dynamic adjustment formula, by controlling the impedance Z0, the dielectric layer thickness h and the material properties (such as the dielectric constant ε r ), combined with the feedback insertion loss IL and return loss RL, the parameters are optimized to achieve the performance target. The a, β, γ in the formula are adjustable coefficients, the specific values are dynamically adjusted according to the real-time measured insertion loss IL and return loss RL. For example, when IL or RL exceeds the preset threshold, the system will adjust a, β, γ according to the current simulation results, so that the line width w automatically changes to a better direction. For example, if IL is too high, a can be reduced to increase the line width and reduce conductor loss, and IL will decrease; if RL is insufficient, the line spacing s can be increased or the line width w can be reduced, and β can be increased. γ is a secondary parameter for fine tuning. The ultimate goal is to make IL and RL meet the design requirements at the same time. This process is similar to "trial and error learning", after adjusting the parameters each time, the performance is verified by simulation, and the coefficients are updated according to the feedback results, and the optimal solution is gradually approached. This dynamic optimization method can adapt to complex design constraints, such as board performance limitations (e.g. w / h < a) and electromagnetic compatibility requirements, so as to achieve efficient and reliable layout scheme in high frequency circuit design.

[0080] In a preferred embodiment, the line spacing s satisfies the formula:

[0081]

[0082] where k ∈ [1.2, 1.8], δ ∈ [0.8, 1.2], and δ is calibrated by the online impedance monitoring module.

[0083] Line spacing s needs to meet the requirement of differential impedance Z through dynamic adjustment of parameters diff k in the formula is a geometric proportionality coefficient (usually between 1.2-1.8) for controlling the relationship between line spacing and dielectric thickness h; δ is a calibration factor (usually between 0.8-1.2) for correcting the influence of manufacturing errors or material property variations. When transmitting differential signals, electromagnetic coupling between the two wires directly affects the differential impedance. If the line spacing s is too small, the coupling is enhanced, resulting in a decrease in impedance; if s is too large, the coupling is weakened, and the impedance rises. By adjusting k and δ, the coupling effect can be dynamically balanced, so that s automatically adapts to the target impedance Z diff . The line impedance monitoring module collects the actual impedance value of the differential line and compares it with the target value. If there is a deviation, the module will automatically update the value of δ (for example, from 0.9 to 1.1) and recalculate the line spacing s. Similar to "adaptive adjustment", through closed-loop feedback, it ensures that the impedance always meets the design requirements. Finally, the calculation result of s is synchronized to the manufacturing process, realizing high-precision impedance matching.

[0084] In a preferred embodiment, the ratio of the area of the hollowed-out region between the layers to the projected area of the transmission line is 1:1.5 to 1:2, the hollowed-out shape is optimized by a genetic algorithm to be a polygon or an arc to reduce near-end crosstalk, and the hollowed-out depth covers the adjacent two layers of dielectric. A smaller hollowed-out area cannot fully block the field coupling of adjacent signal layers, while a too large hollowed-out area will reduce the mechanical strength of the circuit board. This ratio has been verified by a large number of simulations to balance between signal integrity and structural reliability. The genetic algorithm encodes the hollowed-out contour as a gene sequence, with each gene segment corresponding to a vertex coordinate or a curvature parameter. The initial population contains randomly generated polygon and arc schemes, and the near-end crosstalk value of each scheme is calculated by full-wave electromagnetic simulation. The fitness function is set as the weighted sum of crosstalk suppression ratio and structural strain to ensure the balance between electromagnetic performance and mechanical strength. In the evolution process, the crossover operation exchanges the corner features of different shapes, and the mutation operation randomly adjusts the vertex position or the curvature radius. After multiple iterations, the algorithm converges to a set of Pareto optimal solutions.

[0085] In a preferred embodiment, the via optimization includes:

[0086] When D drill ≤D th , D anti ≥λD drill , and the ratio of the via spacing to w is μ:1;

[0087] When D drill >D th , D anti ≥ηD drill , and the impedance continuity is verified by a neural network;

[0088] wherein D th ∈[6,10] mil, λ∈[2.0,2.5], μ∈[4,6], η∈[2.0,2.5].

[0089] Via optimization strategy implements differentiated control mechanism according to aperture size. When the hole diameter D drill does not exceed the threshold value D th (6-10 mil), the reverse pad diameter D anti is required to be at least 2-2.5 times the aperture size, which can effectively expand the electric field diffusion space and reduce the parasitic capacitance. The center distance of the via is simultaneously constrained to be 4-6 times the line width w, and this layout can block the electromagnetic coupling path between adjacent vias, suppressing the crosstalk amplitude below -35 dB. For large aperture vias (D drill > D th ), the reverse pad ratio is adjusted to 2.0-2.5 times, at which time the skin effect of high-frequency signals needs to be considered. The trained convolutional neural network predicts the impedance change curve in real time, which learns from more than ten thousand sets of three-dimensional electromagnetic field simulation data to identify the influence mode of the aperture and reverse pad combination on the characteristic impedance. When the impedance fluctuation exceeds 5%, the system automatically triggers the compensation algorithm to adjust the window shape of the adjacent reference layer to ensure the continuity of the signal return path.

[0090] In a preferred embodiment, the corner optimization includes:

[0091] For transmission lines with a length L th ≥ 500 mil, segmented corner routing is used and the angle is dynamically adjusted;

[0092] A compensation capacitor array is introduced at the corner, with a capacitance C satisfying:

[0093]

[0094] where ΔZ is the impedance deviation value at the corner, K ∈ [0.8, 1.2], f is the highest frequency of the signal, and K is adjusted by a dynamic impedance matching algorithm.

[0095] The transmission line corner optimization uses a hierarchical compensation strategy to solve the high-frequency signal distortion problem. For long-distance routing with a length exceeding 500 mil, the system will automatically decompose the right-angle turn into multiple continuous small-angle bends. This segmented structure can make the electric field distribution more gentle. For example, converting a 90-degree right angle into three small-angle bevels (such as 36°+36°+18°), the signal reflection loss can be improved from -15 dB to -25 dB. The angle adjustment algorithm tracks the impedance changes in real time, and when it detects that the local ΔZ exceeds 3Ω, it dynamically increases the number of segments to maintain a constant phase velocity.

[0096] A programmable capacitor array is implanted at the critical inflection point to form a distributed compensation network through micro-MEMS capacitors (0.1-1.2pF per cell). The configuration values of these capacitors follow the relationship where the highest frequency f of the signal determines the base compensation, the impedance deviation ΔZ comes from the real-time probe measurement, and the adjustment coefficient K is automatically adjusted in the range of 0.8-1.2 according to the coupling strength of adjacent segments.

[0097] In a preferred embodiment, it further comprises: controlling the roughness R a of the copper foil through laser micro-etching technology ≤0.5μm, and using an electroplating process to make the loss increment not more than θ%, where θ ∈ [3, 5], and the electroplating parameters are inversely optimized by real-time IL data. The conductor surface optimization is realized by combining precision machining with closed-loop control. The laser micro-etching technology uses femtosecond pulse laser to etch nanoscale groove structures on the surface of the copper foil, and by adjusting the laser power and scanning speed, the roughness R a is controlled in the range of 0.3-0.5μm. This micro-texture design can reduce the skin effect loss of high-frequency signals. The electroplating process uses an adaptive parameter adjustment system to collect the transmission line insertion loss IL value in real time and inversely calculate the optimal process parameters. When IL exceeds the threshold value of θ% (θ takes 3-5), the system dynamically adjusts the electroplating solution temperature (±2℃), current density (±0.5A / dm 2 ), and additive concentration (±5%), such as when the 12GHz frequency point loss is detected to be excessive, the current density is increased to 2.8A / dm 2 and the temperature is reduced to 48℃, making the copper layer grain size refined from 35nm to 22nm. Dynamic adjustment improves the uniformity of the skin depth of the conductor surface.

[0098] In a preferred embodiment, the dynamic adjustment includes: when IL exceeds the threshold value, preferentially increasing w and triggering the coordinated optimization of D anti ; when RL exceeds the threshold value, preferentially optimizing the hollowed-out area and adjusting C in coordination. In the dynamic adjustment mechanism, C represents the equivalent capacitance value parameter of the compensation capacitor array.

[0099] In a high-speed signal transmission system, the dynamic adjustment mechanism maintains signal integrity through parameter coordinated optimization. When the insertion loss IL exceeds the preset threshold value, the control system first increases the line width w, which can directly reduce the unit length resistance of the conductor and preliminarily improve the signal attenuation characteristics. The line width increment Δw follows the empirical algorithm: for every 1mil increase in line width, about 0.25dB of loss is compensated, but at the same time the characteristic impedance is reduced by about 3Ω. To avoid exacerbating the return loss RL due to impedance mismatch, the adjustment of the reverse pad diameter D anti is triggered simultaneously. After increasing the line width, the reverse pad is enlarged in proportion to maintain the balance of the capacitance, and the typical adjustment rule is ΔD anti= 1.5Δw. For example, when w increases from 5 mil to 6 mil, D anti Synchronously adjust from 12 mil to 13.5 mil.

[0100] When the return loss RL is out of tolerance, first expand the hollow area to reduce the impedance mutation, but this will change the local electric field distribution. At this time, the value of C is essentially the total capacitance of the micro-capacitance network distributed at the key nodes of the transmission line (such as corners, vias), and the capacitance neutralizes the inductive mutation caused by the change of structure by providing a controllable capacitive load. The specific operation process is as follows: when RL = -18 dB is detected, the hollow area is expanded by 15% to weaken the reflection, and the value of C is increased from the initial 0.8 pF to 1.2 pF. The adjustment of the capacitance is equivalent to implanting an "electronic buffer" on the signal path to compensate for the phase delay caused by the hollowing. During the adjustment process, the increase and decrease of the value of C follows the rule of where f" is the main frequency of the signal, and Δφ is the amount of phase deviation caused by the hollowing.

[0101] In a preferred embodiment, when f ≥ 10 GHz, a gradient permittivity design is adopted, and the permittivity ε r Optimize the matching Z0 by the particle swarm algorithm to reduce the high-frequency signal attenuation (IL) and the dielectric material meets the requirements of D k ≤ 3.5, D f ≤ 0.002. In the field of high-frequency signal transmission, when the signal frequency exceeds 10 GHz, the traditional homogeneous dielectric material will cause significant signal attenuation problems. The gradient permittivity design configures different dielectric parameters in different depth directions of the circuit board stack to build an electric field control structure similar to an optical lens. For example, in a 12-layer board, the top layer uses low-dielectric material with ε r = 3.0 to reduce edge field coupling, the middle signal layer uses conventional material with ε r = 3.3 to maintain impedance stability, and the bottom layer is configured with high-dielectric material with ε r = 3.6 to suppress electromagnetic leakage. The particle swarm optimization algorithm takes each layer permittivity as an independent particle, and takes the global impedance deviation and loss value as the fitness function, and finds the optimal combination through iterative calculation. Strict material performance indicators (D k ≤ 3.5, D f ≤ 0.002) are the basis for implementing this technology.

[0102] In a preferred embodiment, the generated printing instructions control the laser engraving device to perform manufacturing with a W tolerance of ±ω1%, and an s tolerance of ±ω1%, where ω1 ∈ [5, 8], ω2 ∈ [8, 10], and the tolerance parameters are dynamically tightened through yield feedback of IL and RL.

[0103] The generation and execution of printing instructions adopt an adaptive control strategy. The laser engraving equipment is initially set with a line width tolerance of ±8% and a pitch tolerance of ±10%, which is a balanced value considering production efficiency and accuracy. After producing several sample boards in the first round, the insertion loss IL and return loss RL data of each board are collected. When the proportion of abnormal boards with IL>0.9dB / inch or RL>-18dB exceeds 5%, the tolerance tightening mechanism is triggered. The dynamic adjustment algorithm gradually compresses the tolerance range according to the yield distribution curve: if the yield of the last three batches is ≥95%, the line width tolerance ω1 is tightened from 8% to 6% and the pitch tolerance ω2 is reduced from 10% to 9%, eliminating hidden dangers; otherwise, if the yield drops below 90%, the tolerance is appropriately relaxed to improve production capacity and avoid collapse.

[0104] As shown in Figure 2 , a circuit board printing control system suitable for the above-mentioned circuit board printing control method based on signal transmission loss, comprising:

[0105] A parameter acquisition module is used to collect Z0, Z diff , f and h, ε r , t;

[0106] A multi-physical field coupling simulation engine is used to calculate IL, RL and impedance discontinuity points;

[0107] A reinforcement learning optimization module is used to adjust w, s and hollowed-out areas;

[0108] A via analysis unit generates a via scheme based on the ratio of D drill and D anti ;

[0109] An instruction generation module converts the optimized parameters into PCB control instructions.

[0110] The circuit board printing control system realizes signal integrity optimization through multi-module cooperation. The parameter acquisition module collects key parameters such as target impedance Z0, differential impedance Z diff , maximum signal frequency f, and medium thickness h, dielectric constant ε r , copper thickness t, etc. in real time, forming an initial design constraint library. The multi-physical field coupling simulation engine then builds a three-dimensional electromagnetic model, simulates the electric field distribution and energy loss during signal transmission, accurately predicts the distribution trend of insertion loss IL and return loss RL, and marks the coordinates of regions with impedance mutations exceeding 5%.

[0111] The reinforcement learning optimization module initiates adaptive adjustment based on simulation results, taking line width w and line spacing s as core adjustment variables, and combining shape iteration of the medium layer hollow region. The module gradually converges to the optimal solution in several iterations, such as optimizing the initial line width of 5 mil to 5.8 mil, while generating a quincunx-shaped hollow structure to reduce edge field coupling. The via analysis unit operates synchronously, automatically matching the via diameter D drill with the optimal ratio of anti-pad size D anti , recommending a 2.2 times anti-pad scheme for a 0.5 mm aperture to effectively suppress via resonance. The instruction generation module finally converts the optimized parameters into a digital instruction set for a laser engraving machine, controlling the device to perform line etching with a precision of ±0.3 mil. The system is embedded with a process compensation algorithm, such as automatically increasing the line width by 0.1 mil to compensate for copper foil roughness.

[0112] The instruction generation module finally converts the optimized parameters into a digital instruction set for a laser engraving machine, which is the solution of the present application and the inventive concept to be protected. As for what kind of parameters generate what kind of instructions, it belongs to the embodiments of the present application. For example, in terms of geometric parameter conversion, the adjustment of line width w is directly related to the regulation of laser power P. When the line width changes by Δw from the reference value, the laser power is adjusted by ΔP = 2000Δw, and this coefficient is calibrated through copper foil etching rate experiments. The expansion of line spacing s is adjusted by moving speed through an exponential function, F new =F base e -0.12Δs , where F base is the reference moving speed, i.e. the initial speed when the line spacing is not adjusted, Δs is the change in line spacing, and F new is the adjusted moving speed, ensuring the uniformity of the heat affected zone when the spacing is expanded. The thickness deviation h err of the medium is compensated by focal length compensation ΔZ = 0.03h err , where h err = h actual -h design , h actual is the thickness of the medium, and h design is the designed thickness of the medium layer, to avoid abnormal etching depth caused by lamination fluctuations. For electrical performance optimization, for every 0.5 dB increase in insertion loss IL, the system will reduce the laser power by 5% in the signal line edge area to weaken the edge field coupling effect. When the return loss RL exceeds the standard, the path overlap rate is adjusted by η = 1.2 - 0.03|RL|, where η is the path overlap rate, RL is the return loss, and 0.03 is 0.03 dB -1 , with the unit dB -1 , which is well known in the art and will not be repeated. The impedance mutation zone ΔZ adopts a curvature radius κ ∝ 1 / ΔZ 2Adaptive spiral scanning, enhanced impedance continuity. Process constraint, line width tolerance ω1 is converted to pulse interval t pulse = 50(1 + ω1 / 10)μs, where t pulse is the laser pulse interval time, ω1 is the line width tolerance. Copper foil roughness R a determines the path repetition number of engraving N = [2.5R a ], where [·] is the upward rounding function, R a is brought into numerical calculation, which is an empirical formula, and the coefficient 2.5 can also be understood as a unit conversion, which is actually 2.5 times / μm. It should be noted that the present application involves multiple formulas, in order to ensure the unit consistency on both sides of the formula, it can be understood that the numerical coefficient implies the unit, or it can be understood as an empirical formula, which can be calculated by bringing in numerical values (standard units). Dynamic feedback link uses PID control algorithm to correct power in real time: ΔR = K p e + K i ∫edt + K d (de / dt), e = w target -w meas . Where, ΔP is the laser power adjustment amount, K p is the proportional gain coefficient, K i is the integral gain coefficient, K d is the differential gain coefficient, e is the line width error, w target is the design target line width, w meas is the real-time measured line width. Temperature fluctuation compensation formula F adj = F[1-0.015(T-T0)] ensures deceleration and heat dissipation in high temperature environment, T0 = 25°C is the reference temperature, F adj is the actual moving speed after temperature compensation, F is the reference moving speed, T is the real-time monitored temperature, and 0.015 is the temperature compensation coefficient (unit: 1 / ℃). The abnormal processing mechanism includes three levels of response: when the burr of 0.5μm or more is detected, the reverse engraving is performed (power-20%, speed+30%); 40kHz ultrasonic cleaning is triggered when the medium carbonizes; when the impedance mutation exceeds 8%, a compensation via hole (diameter 0.5w) is implanted and the reverse pad is expanded by 20%.

[0113] In a preferred embodiment, the multi-physics coupled simulation engine integrates S-parameter extraction algorithm and predicts metamaterial structure loss characteristics of f > 20GHz signals, including artificial electromagnetic surfaces or photonic crystals, through convolutional neural network. S-parameters are a key indicator matrix for describing the reflection and transmission characteristics of electromagnetic waves in high-frequency circuits, which quantifies the energy distribution relationship of signals at different ports to evaluate the signal integrity of the circuit under high-frequency working conditions. Reflection characteristics: S11 / S22 represents the reflection coefficient of port 1 / 2, transmission characteristics: S21 / S12 represents the transmission coefficient of port 2 to 1 / 1 to 2. S-parameters are a bridge connecting electromagnetic simulation and actual measurement, and their matrix data provide quantitative basis for optimizing PCB layout and suppressing signal degradation. These are prior art and will not be described here. The multi-physics coupled simulation engine builds a three-dimensional behavior model of metamaterial by integrating the interaction of electromagnetic field, thermal field and structural stress field. For artificial electromagnetic surfaces, the engine simulates the electromagnetic response of the unit structure in the millimeter wave band, calculates the surface current distribution and near-field radiation pattern; for photonic crystals, it analyzes the photonic bandgap characteristics caused by periodic permittivity changes. Above 20GHz, the engine uses the finite difference time domain (FDTD) algorithm to capture field distribution details with a spatial resolution of λ / 20. The S-parameter extraction algorithm is embedded in the post-processing link of the simulation, and the scattering matrix containing amplitude and phase information is generated by calculating the port excitation and response signals. Wideband adaptive sampling technology can extract 5000 S21 parameters at 0.1GHz steps in the range of 26.5-110GHz, while recording the local temperature rise ΔT (accuracy ±0.5°C) and mechanical deformation δ (accuracy ±0.1μm) of the material. These data form a four-dimensional feature space: frequency, transmission loss IL, thermal stress TS, and structural distortion SD. By correlating the data of the four dimensions, the non-linear interaction between each factor can be analyzed. The convolutional neural network uses a five-layer deep architecture, and the input layer receives 256x256 pixel electromagnetic field distribution snapshots and corresponding S-parameter curves. The first convolution kernel focuses on extracting surface plasmon resonance features and identifying local field enhancement hotspots; the second convolution kernel correlates the nonlinear relationship between the thermal expansion coefficient and the permittivity drift. After several iterations of training, the network can predict the loss characteristics of the metamaterial at specific frequency points (such as 28 / 39 / 60GHz) based on its geometric parameters only.

[0114] In a preferred embodiment, the reinforcement learning module contains a threshold database storing the maximum allowed ratio of w / h at different f, and is updated in real-time with process data. The threshold database in the reinforcement learning module is essentially a dynamic design rule library, used to constrain the key parameter combination of high-frequency transmission lines. Taking the ratio of line width to dielectric thickness h as an example, when the signal frequency increases, excessive w / h will cause the edge field coupling to intensify, resulting in excessive insertion loss. The initial database is based on electromagnetic field simulation and process history data, and pre-stores safety thresholds at different frequency bands, for example, w / h ≤ 5.2 is allowed at f = 10 GHz, and w / h ≤ 3.8 is tightened at f = 28 GHz.

[0115] The real-time linkage update mechanism continuously collects two types of data through line sensors: process parameters (actual etching line width w real , dielectric thickness h real after pressing, and performance indicators (measured insertion loss IL, return loss RL). When it is detected that w / h = 4.0 at f = 24 GHz but IL is still better than the standard, the algorithm determines that the current threshold is conservative, automatically relaxes the upper limit of w / h from 3.8 to 4.1, and marks the adjustment for engineers to review. Conversely, if RL deteriorates beyond the limit when w / h = 3.7, the threshold is immediately tightened to 3.5 and the process is triggered for troubleshooting.

[0116] In a preferred embodiment, a roughness detection unit is also included, which feeds back the three-dimensional topography of Ra (resolution ≤ 1 μm) through terahertz wave scanning and adjusts the electroplating parameters in linkage. The roughness detection unit uses terahertz time-domain spectroscopy technology to scan the copper foil surface by emitting a pulse wave in the 0.1-3 THz frequency band. When electromagnetic waves encounter microscopic pits or protrusions, the phase and amplitude of the reflected signal produce characteristic distortions. A 128-element detector scans at a step of 50 μm, constructing a three-dimensional surface topography map (lateral resolution 0.8 μm, vertical resolution 0.3 μm). For the detected rough peak-valley structure, the system calculates the arithmetic average roughness Ra and the ten-point height Rz, and determines that it is a high-risk area when Ra ≥ 0.4 μm. The detection data are transmitted in real time to the electroplating control center, triggering a dynamic parameter adjustment mechanism: if Ra ∈ [0.4, 0.6] μm, increase the electroplating current density to 2.8 A / dm 2 and increase the concentration of organic additives by 12% to fill in the depressions by accelerating deposition; when Ra ≤ 0.2 μm, reduce the current density to 1.5 A / dm 2 to avoid crystallization of the plated layer. For the detected local abnormal protrusions (Rz ≥ 2 μm), the system locates the coordinates and then enables micro-area reverse electrolysis in the area, applying a voltage of -0.5 V for 3 seconds to achieve selective flattening.

[0117] In a preferred embodiment, the instruction generation module synchronously controls the processing error of the laser engraving, drilling and electroplating equipment through a digital twin model, and the calibration period of the model and the physical equipment is ≤1 hour. The digital twin system of the instruction generation module realizes multi-device collaborative control through a three-stage process. The real-time state data of the physical equipment is collected by the sensor once every second to construct a holographic mirror image of the virtual equipment. In the second stage, the twin model runs a Monte Carlo simulation to predict the future 1-hour processing error trend. When the predicted laser engraving line width error is about to exceed the ±0.25 mil threshold, a pre-compensation algorithm is triggered: the laser path is adjusted in advance, and 0.1 mil of line width redundancy is added in the key area. When the spindle temperature rise ΔT=8°C is monitored by the drilling synchronization module, the hole diameter deviation ΔD=0.002ΔT caused by thermal expansion is immediately calculated in the virtual model, and reverse compensation drilling coordinates are generated. In the third stage, dynamic calibration is implemented, and the actual processing data of the physical equipment is compared with the predicted value of the twin model every 60 minutes, and the model parameters are updated through the gradient descent algorithm.

[0118] The circuit board printing control method and system disclosed in the application are based on signal transmission loss. An electrical and lamination parameter acquisition module is used to collect electrical and lamination parameters, construct a multi-physical field coupling simulation model, accurately predict loss characteristics and impedance mutation distribution, and strengthen a learning optimization module to iteratively adjust line width, line spacing and hollow area layout based on dynamic constraint rules, balance electromagnetic performance and structural reliability by combining corner segmentation compensation and via parasitic capacitance optimization strategies, and use an instruction generation module to synchronize process parameters of laser engraving, drilling and electroplating equipment through a digital twin model, and realize process error closed-loop control by using real-time roughness detection and dynamic tolerance adjustment. The three modules are combined through electromagnetic characteristic simulation, intelligent algorithm optimization and physical equipment linkage to realize a three-way collaborative mechanism, and signal integrity design is applied throughout the modeling, optimization and manufacturing chain to ensure high-frequency transmission performance and significantly improve processing precision and production efficiency.

[0119] Although the preferred embodiments of the application have been described in detail, other changes and modifications can be made by those skilled in the art once they understand the basic inventive concepts. Therefore, the appended claims are intended to cover these preferred embodiments and all changes and modifications that fall within the scope of the application. It should be emphasized that the above description is only representative of the preferred embodiments of the application and does not constitute a limitation. Any modification, equivalent replacement or improvement made under the guidance of the spirit and principles of the application should be considered within the scope of the application.

Claims

1. A circuit board printing control method based on signal transmission loss, characterized in that, Includes the following steps: Obtain the target single-ended impedance Z0 and differential impedance Z of the target differential transmission line. diff and circuit board stack-up parameters, including shielding layer spacing h, dielectric constant ε r and copper thickness (t); An electromagnetic simulation model is established based on the shielding layer spacing, dielectric constant, and copper thickness. The insertion loss IL, return loss RL, and impedance discontinuity distribution are simulated by combining a multiphysics coupling algorithm. The linewidth w, line spacing s, and interlayer cutout area of ​​the differential transmission line are dynamically adjusted by a reinforcement learning model. The ratio of linewidth to shielding layer spacing is limited by a preset critical value, and the line spacing is configured based on a multiple relationship of shielding layer spacing. A segmented optimization strategy is adopted for transmission line corners, with the corner angle range being 90°-180°; Optimize aperture D based on via parasitic capacitance model drill With anti-solder pad diameter D anti The proportion of [the target] is reduced by hollowing out adjacent reference layers to decrease impedance abruptness; The optimized layout rules are output and combined with real-time process data to generate PCB printing control instructions.

2. The circuit board printing control method based on signal transmission loss according to claim 1, characterized in that, The line width w satisfies the formula: Where α∈[5.0,6.5], β∈[5.5,6.5], and γ∈[0.7,0.9], α, β, and γ are updated through the feedback IL and RL.

3. The circuit board printing control method based on signal transmission loss according to claim 1, characterized in that, The line spacing s satisfies the formula: Where k∈[1.2,1.8] and δ∈[0.8,1.2], and δ is calibrated by the online impedance monitoring module.

4. The circuit board printing control method based on signal transmission loss according to claim 1, characterized in that, Via optimization includes: When D drill ≤D th At that time, D anti ≥λD drill Furthermore, the ratio of the via spacing to w is μ:1; When D drill >D th At that time, D anti ≥ηD drill The impedance continuity was verified using a neural network. Among them, D th ∈[6,10]mil,λ∈[2.0,2.5],μ∈[4,6],η∈[2.0,2.5].

5. The circuit board printing control method based on signal transmission loss according to claim 1, characterized in that, The corner optimization includes: For length L th For transmission lines ≥500mil, segmented corner wiring with dynamic angle adjustment is used; A compensation capacitor array is introduced at the corner, and its capacitance C satisfies: Where ΔZ is the impedance deviation at the corner, K∈[0.8,1.2], and f is the highest frequency of the signal.

6. The circuit board printing control method based on signal transmission loss according to claim 1, characterized in that, Also includes: Controlling the roughness R of copper foil using laser micro-engraving technology a ≤0.5μm, and use electroplating process to ensure that the loss increment does not exceed θ%, where θ∈[3,5].

7. The circuit board printing control method based on signal transmission loss according to claim 1, characterized in that, Dynamic adjustment includes: when IL exceeds the threshold, preferentially increasing w and triggering D. anti Collaborative optimization; when RL exceeds the threshold, prioritize optimization of the hollowed-out area and adjust C accordingly.

8. A circuit board printing control system based on signal transmission loss, characterized in that, A circuit board printing control method based on signal transmission loss according to any one of claims 1-7, comprising: The parameter acquisition module is used to collect Z0 and Z. diff f and h, ε r , t; A multiphysics coupling simulation engine for calculating IL, RL, and impedance discontinuities; The reinforcement learning optimization module is used to adjust w, s, and the cut-out region; Via analysis unit, based on D drill With D anti Proportional via generation scheme; The instruction generation module converts optimized parameters into PCB control instructions.

9. A circuit board printing control system based on signal transmission loss according to claim 8, characterized in that, The reinforcement learning module includes a critical value database that stores the maximum allowable ratio of w / h under different f values ​​and updates it in conjunction with real-time process data.

10. A circuit board printing control system based on signal transmission loss according to claim 8 or 9, characterized in that, It also includes a roughness detection unit, which uses terahertz wave scanning to provide feedback on the three-dimensional morphology of Ra and adjusts the electroplating parameters accordingly.

Images