Electromagnetic compatibility test method and storage system for PCB (Printed Circuit Board) design

Abstract

The invention provides an electromagnetic compatibility test method and system for PCB design, a PCB comprises a schematic circuit diagram and a structural model, and the method comprises the following steps: extracting high-frequency signal characteristics in the schematic circuit diagram to obtain an electromagnetic simulation parameter set; performing multi-physical field coupling modeling according to the electromagnetic simulation parameter set and the structure model to obtain an electromagnetic field-circuit model; performing frequency-time domain hybrid excitation loading on the electromagnetic field-circuit model to obtain an electromagnetic noise distribution diagram; performing gradient descent optimization on layout parameters of the structure model according to the electromagnetic noise distribution diagram to obtain an anti-interference layout scheme; and performing multi-working-condition Monte Carlo simulation and residual error calibration on the anti-interference layout scheme to generate a correction guidance library.

Description

Technical Field

[0001] This application relates to the technical field of electronic circuits, and particularly to an electromagnetic compatibility testing method and a memory system for PCB design. Background Art

[0002] Currently, in the traditional PCB design process, electromagnetic compatibility (EMC) verification usually relies on physical sample testing after design finalization, and interference sources are located through near-field probe scanning or anechoic chamber radiation testing. However, this method has significant drawbacks: when it is found that the EMC exceeds the standard, the PCB layout has been solidified, and readjusting the line width, stack-up structure, or grounding strategy requires re-sampling verification, resulting in an extended R & D cycle and a sharp increase in costs. Although existing simulation technologies can perform pre-analysis in the design stage, they are mostly limited to a single physical field (such as pure electromagnetic fields or circuit simulation), and it is difficult to accurately model the coupling effects of high-frequency signals, power supply noise, and three-dimensional structures. In addition, traditional optimization methods rely on empirical rules (such as the 3W principle) for manual adjustment, lack quantitative trade-offs for multi-objective conflicts (radiation suppression and signal integrity), and are even less able to predict the potential impact of manufacturing tolerances and environmental variations on EMC. These problems often lead to the "test - modification" loop iteration in PCB design, seriously affecting the product development efficiency. Summary of the Invention

[0003] This application provides an electromagnetic compatibility testing method and a memory system for PCB design, which are used to discover and solve electromagnetic compatibility problems in the design stage of the PCB and improve the development efficiency of PCB products.

[0004] In a first aspect, an embodiment of this application provides an electromagnetic compatibility testing method for PCB design. The PCB includes a circuit schematic diagram and a structure model. The method includes: Extracting high-frequency signal characteristics from the circuit schematic diagram to obtain an electromagnetic simulation parameter set; Performing multi-physical field coupling modeling based on the electromagnetic simulation parameter set and the structure model to obtain an electromagnetic field - circuit model; Performing frequency - time domain hybrid excitation loading on the electromagnetic field - circuit model to obtain an electromagnetic noise distribution map; Performing gradient descent optimization on the layout parameters of the structure model according to the electromagnetic noise distribution map to obtain an anti-interference layout scheme; Performing multi-condition Monte Carlo simulation and residual calibration on the anti-interference layout scheme to generate a correction guidance library.

[0005] Second aspect, an electromagnetic compatibility test system for PCB design provided by an embodiment of the present application is used to execute the electromagnetic compatibility test method for PCB design as described in any one of the embodiments of the present application. The PCB includes a circuit schematic diagram and a structural model. The system includes: A high-frequency analysis module for extracting high-frequency signal characteristics in the circuit schematic diagram to obtain an electromagnetic simulation parameter set; A parameter modeling module for performing multi-physical field coupling modeling based on the electromagnetic simulation parameter set and the structural model to obtain an electromagnetic field-circuit model; An excitation loading module for performing frequency-time domain hybrid excitation loading on the electromagnetic field-circuit model to obtain an electromagnetic noise distribution map; A layout optimization module for performing gradient descent optimization on the layout parameters of the structural model according to the electromagnetic noise distribution map to obtain an anti-interference layout scheme; A correction generation module for performing multi-condition Monte Carlo simulation and residual calibration on the anti-interference layout scheme to generate a correction guidance library.

[0006] Third aspect, an electronic device provided by an embodiment of the present application includes a memory and a processor; The memory is used to store a computer program; The processor is used to execute the computer program and implement the electromagnetic compatibility test method for PCB design as described in any one of the embodiments of the present application when executing the computer program.

[0007] Fourth aspect, a computer-readable storage medium provided by an embodiment of the present application stores a computer program. When the computer program is executed by a processor, the processor is enabled to implement the electromagnetic compatibility test method for PCB design as described in any one of the embodiments of the present application.

[0008] An embodiment of the present application provides an electromagnetic compatibility testing method for PCB design. The PCB includes a circuit schematic diagram and a structural model. The method includes: extracting high-frequency signal characteristics from the circuit schematic diagram to obtain an electromagnetic simulation parameter set; performing multi-physical field coupling modeling based on the electromagnetic simulation parameter set and the structural model to obtain an electromagnetic field-circuit model; performing frequency-time domain hybrid excitation loading on the electromagnetic field-circuit model to obtain an electromagnetic noise distribution map; performing gradient descent optimization on the layout parameters of the structural model according to the electromagnetic noise distribution map to obtain an anti-interference layout scheme; performing multi-condition Monte Carlo simulation and residual calibration on the anti-interference layout scheme to generate a correction guidance library. In the above method, by extracting the high-frequency characteristics of high-speed signal networks and power networks in the circuit schematic diagram, a multi-dimensional electromagnetic simulation parameter set is generated to accurately locate the radiation noise source. Based on the three-dimensional electromagnetic topology reconstruction and multi-physical field coupling modeling of the structural model, the circuit ports and the electromagnetic field distribution are jointly analyzed. Through the frequency-time domain hybrid excitation loading technology, electromagnetic noise is synchronously captured. Combining with the gradient descent algorithm, the layout parameters (line width, layer spacing) are constrained and optimized. On the premise of ensuring signal integrity, the radiation peak is reduced. Then, through multi-condition Monte Carlo simulation, the influence of manufacturing tolerances and environmental variations is quantitatively analyzed, and a residual-driven correction rule library is generated to improve the correction efficiency and the development efficiency of PCB products. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0010] Figure 1 FIG. is a schematic flowchart of an electromagnetic compatibility testing method for PCB design provided by an embodiment of the present application; Figure 2 FIG. is a schematic block diagram of an electromagnetic compatibility testing system for PCB design provided by an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0011] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the protection scope of the present invention.

[0012] The flowcharts shown in the accompanying drawings are merely illustrative examples, and do not necessarily include all contents and operations / steps, nor are they necessarily executed in the order described. For example, some operations / steps can be decomposed, combined or partially merged, so the actual execution order may change according to the actual situation.

[0013] It should also be understood that the terms used in the specification of this application are only for the purpose of describing specific embodiments and are not intended to limit this application. As used in the specification of this application and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0014] It should be further understood that the term "and / or" used in the specification of this application and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0015] Please refer to Figure 1 , Figure 1 which is a schematic flowchart of an electromagnetic compatibility test method for PCB design provided by an embodiment of this application. As Figure 1 shown, the specific steps of the electromagnetic compatibility test method for PCB design include: S101 - S105.

[0016] S101. Extract the high - frequency signal characteristics in the circuit schematic diagram to obtain an electromagnetic simulation parameter set.

[0017] Exemplarily, a high-speed signal network is defined as a transmission path with a signal rise time less than 1 nanosecond. Through the frequency-domain energy distribution analysis technique, a vector network analyzer with a sweep range of 1 MHz to 20 GHz is used to collect the fundamental harmonic characteristics point by point at a step of 10 MHz, including harmonic amplitude, phase shift, and spectral density parameters, to ensure the accurate capture of key resonant frequencies (such as the 3rd and 5th harmonics of the clock fundamental frequency). A multi-port excitation source is loaded in the power plane segmentation area for modal impedance analysis of the power network. A three-dimensional full-wave electromagnetic field solver is used to calculate the impedance matrix in the frequency band of 0.1 GHz to 5 GHz, and the high-risk frequency band (such as the resonant valley at 1.2 GHz) where the impedance drop in the characteristic impedance curve exceeds 20% is extracted, and the corresponding planar resonant mode is marked. The method of moments (MoM) and partial element equivalent circuit (PEEC) hybrid algorithm are used to perform near-field coupling modeling of the interconnect structure, calculate the electromagnetic field distribution in the interconnect area where the adjacent trace spacing is less than 3 times the line width, quantify the distributed capacitance, mutual inductance, and crosstalk coefficient, and generate a parasitic parameter feature set (such as the distributed capacitance between adjacent traces is 0.05 pF / mm) through a parameter extraction engine. The above analysis results are integrated into an electromagnetic simulation parameter set by a hierarchical feature fusion module according to the frequency-band - space mapping relationship. Its data structure includes a frequency-domain response matrix (dimension: frequency × spatial coordinates), a spatial coupling coefficient tensor (dimension: trace spacing × frequency), and a network topology relationship map, providing accurate input for subsequent multi-physics field coupling modeling.

[0018] S102. Perform multi-physics field coupling modeling based on the electromagnetic simulation parameter set and the structural model to obtain an electromagnetic field - circuit model.

[0019] Exemplarily, the structural model contains information such as the stack-up thickness (such as the thickness of a 6-layer board is 1.6 mm), the dielectric constant distribution (such as the dielectric constant of FR4 material is 4.3 at 1 GHz), and the conductor surface roughness (such as the RMS roughness of copper foil is 0.8 μm). The three-dimensional geometric model is reconstructed through non-uniform rational B-spline (NURBS) surface fitting technology, and the surface fitting accuracy is controlled within 0.01 mm. The loading of frequency-varying material parameters needs to be based on the test data of the complex dielectric constant of the stack-up material (the real and imaginary parts are interpolated in the frequency band of 1 MHz to 10 GHz), establish an interpolation function of the dielectric constant varying with frequency, the frequency sampling interval is set to 50 MHz, and the material parameters are associated with the corresponding area of the geometric model through a field mapping algorithm (such as the conductivity of 5.8×10 7S / m). Adaptive mesh refinement uses a dual criterion of curvature-driven and field strength gradient. Local mesh refinement is implemented in regions with sudden field strength changes such as wire corners, via edges, and power split slots. The mesh size is dynamically adjusted from a reference of 0.5 mm to a minimum of 0.05 mm to ensure the convergence of the field distribution calculation (e.g., the mesh near the via edge is refined to 0.1 mm). For the circuit-electromagnetic port coupling process, the excitation source (such as a clock driver with a 50 Ω impedance) and the load impedance (such as a 100 Ω differential termination) in the schematic diagram need to be mapped to the port boundary of the 3D model, and the bidirectional coupling relationship between voltage - electric field and current - magnetic field is defined through the hybrid-domain interface matrix. The full-wave electromagnetic-circuit co-solver uses the joint iteration of the finite-difference time-domain (FDTD) and SPICE circuit simulation. The electromagnetic field distribution and circuit node voltages are updated at each time step, and the output includes the electromagnetic field-circuit model with spatial field strength distribution (such as the maximum electric field strength of 120 V / m), current density (such as the power plane current density of 5 A / mm 2 ), and S-parameters (such as insertion loss of -3 dB @ 5 GHz).

[0020] S103. Apply frequency-time domain hybrid excitation loading to the electromagnetic field-circuit model to obtain the electromagnetic noise distribution map.

[0021] Exemplarily, for the operation of applying frequency-time domain hybrid excitation loading to the electromagnetic field-circuit model, two types of excitation sources, namely a frequency-domain multi-tone signal and a time-domain transient pulse, are injected synchronously. The frequency-domain excitation source is generated by a pseudo-random multi-frequency synthesis algorithm, with frequency components covering 1 MHz to 10 GHz (at 100 MHz intervals), the amplitude decaying at -20 dB / decade to simulate the actual signal spectrum characteristics (such as the amplitude of 1 V at 1 GHz and 0.1 V at 10 GHz), and the phase adopting a uniform random distribution to avoid spectrum leakage. The time-domain excitation source uses a step pulse with a 50 ps rising edge combined with a PRBS (pseudo-random binary sequence) pattern, and the code length is set to 2 15-1 (32,767 bits), with a symbol rate of 1 Gbps to cover broadband excitation requirements. During field-circuit co-simulation, the excitation signal is synchronously applied to the radiation boundary (such as the PCB edge) of the 3D model and the circuit ports of the schematic (such as the clock signal input port) through the electromagnetic-circuit interface matrix. The time-domain simulation step size is set to 0.1 ps to meet the time resolution of the highest frequency band (10 GHz). The acquisition of the original noise dataset requires recording the time-domain waveform of the electric field intensity at the spatial grid nodes (sampling rate 10 THz), and converting it into a frequency-domain spectrum line through the short-time Fourier transform (STFT). The window function is selected as the Hamming window, and the window length is adaptively adjusted (such as 1 ns for the 1 GHz frequency band) to balance the frequency resolution and time locality. The noise characteristic parameter set is generated by quantifying the peak field strength in the near-field region (such as the maximum electric field of 150 V / m near the surface trace), the main lobe width of the far-field radiation pattern (such as the 3 dB beam width of 30 degrees), and the cross-polarization level (such as -20 dBc). The space-frequency mapping process uses the spherical wave expansion algorithm to project the 3D field data onto a virtual test sphere with a radius of 1 m (the azimuth and elevation angle resolutions are set to 5 degrees), generating a 3D electromagnetic noise distribution map including frequency-azimuth-polarization (the data dimension is frequency × azimuth angle × elevation angle × polarization direction).

[0022] S104. Gradient descent optimization is performed on the layout parameters of the structural model according to the electromagnetic noise distribution map to obtain an anti-interference layout scheme.

[0023] Exemplarily, based on the gradient descent optimization of the electromagnetic noise distribution map, the line width (such as a minimum of 0.1 mm), layer spacing (such as a minimum of 0.15 mm), and via position parameters (such as via spacing ≥ 0.3 mm) are initialized according to the preset layout constraint boundaries. The construction of the objective function gradient matrix requires calculating the partial derivatives of the radiation field strength with respect to each layout parameter, and solving the sensitivity coefficients (such as the influence coefficient of -0.5 dB / mm on the 3 GHz radiation field strength when the line width increases by 0.01 mm) through the adjoint field method during the backpropagation process. The momentum accelerated descent iteration introduces a historical gradient weighted accumulation mechanism, with the initial value of the momentum factor set to 0.9, and it is dynamically adjusted according to the persistence of the parameter update direction (such as increasing to 0.95 when the iteration directions are the same for three consecutive times). The constraint boundary projection process uses the Lagrange multiplier method to impose hard corrections on the parameters that exceed the process constraints. For example, the illegal line width is forced to return to the minimum allowable value (0.1 mm), and the adjacent trace spacing is synchronously adjusted (increased by 0.02 mm) to maintain impedance continuity (such as single-ended impedance of 50 Ω ± 5%). DRC (Design Rule Check) and LVS (Layout-Schematic) are performed after each iteration Figure 1Consistency verification) is performed to generate a sequence of physically feasible solutions, ensuring that parameter changes comply with manufacturing specifications (such as a minimum pad size of 0.2 mm × 0.2 mm). The convergence criterion is set such that the relative change rate of the objective function value is less than 0.1% and the gradient norm is below 1e-4. An optimized parameter update sequence is output through parameter sensitivity ranking, and key parameters with a contribution to radiation suppression greater than 5% are marked (such as the power layer segmentation distance being adjusted from 0.5 mm to 0.7 mm) for subsequent rule base construction.

[0024] S105. Conduct multi-condition Monte Carlo simulations and residual calibration on the anti-interference layout scheme to generate a correction guidance library.

[0025] Exemplarily, manufacturing tolerance perturbations include random deviations of line width ±10% (such as a nominal line width of 0.15 mm allowing a deviation of ±0.015 mm), dielectric constant ±5% (such as the dielectric constant of FR4 material being 4.3 ± 0.215), and copper thickness ±8% (such as a 1 oz copper thickness of 34 μm ± 2.72 μm). 200 sets of Monte Carlo sample sets are generated through Latin hypercube sampling. Environmental perturbation simulates the effects of temperature varying from -40°C to 125°C and humidity fluctuating from 10% to 90%. The drift relationship of material parameters with temperature and humidity is modeled based on the Arrhenius equation (such as a dielectric constant temperature coefficient of -50 ppm / °C) and the hygroscopic expansion coefficient (such as an increase in dielectric constant of 0.1 when the water absorption rate of FR4 is 0.2%). The multi-physics field joint simulation uses a parallel computing architecture. Each group of samples performs electromagnetic-thermal-stress coupling simulations, and a dataset containing radiation field strength (such as a 2 dB increase in 3 GHz radiation at a high temperature of 125°C), temperature rise distribution (such as a temperature rise ΔT of 25°C in the power supply area), and mechanical deformation (such as a warpage amount of 0.05 mm) is output. Principal component analysis (PCA) reduces the dimension of the EMC performance dataset and extracts the top 3 principal components with a cumulative contribution rate exceeding 85% (such as line width deviation, dielectric constant offset, temperature drift). Then, the samples are divided into three categories: compliant, critical, and non-compliant through K-means clustering (such as the proportion of critical class samples being 15%). The residual feature vector set is generated by calculating the mean square error between the measured field strength and the simulation prediction value (such as a root mean square error of the residual of 1.2 dB). Residual correction analysis uses a random forest regression model to establish the mapping relationship between process parameter deviations and radiation residuals, and outputs a key parameter correction coefficient matrix (such as a line width negative deviation of 8% corresponding to a layer spacing correction amount of +0.02 mm). The construction of the correction rule base converts the correction coefficients into IF-THEN type rules (such as "when the humidity > 70%, the power supply segmentation distance needs to be increased by 0.1 mm"), and redundant rules with a coverage rate lower than 5% are removed through cross-condition verification to generate an electromagnetic compliance correction guidance library (containing 200 priority ranking rules) that supports the invocation of automated design tools.

[0026] The embodiment of the present application provides an electromagnetic compatibility testing method for PCB design. The PCB includes a circuit schematic diagram and a structure model. The method includes: extracting the high-frequency signal characteristics in the circuit schematic diagram to obtain an electromagnetic simulation parameter set; performing multi-physical field coupling modeling based on the electromagnetic simulation parameter set and the structure model to obtain an electromagnetic field-circuit model; performing frequency-time domain hybrid excitation loading on the electromagnetic field-circuit model to obtain an electromagnetic noise distribution map; performing gradient descent optimization on the layout parameters of the structure model according to the electromagnetic noise distribution map to obtain an anti-interference layout scheme; performing multi-condition Monte Carlo simulation and residual calibration on the anti-interference layout scheme to generate a correction guidance library. In the above method, by extracting the high-frequency characteristics of the high-speed signal network and power network in the circuit schematic diagram, a multi-dimensional electromagnetic simulation parameter set is generated to accurately locate the radiation noise source. Based on the three-dimensional electromagnetic topology reconstruction and multi-physical field coupling modeling of the structure model, the circuit ports and the electromagnetic field distribution are co-analyzed. Through the frequency-time domain hybrid excitation loading technology, electromagnetic noise is synchronously captured. Combining with the gradient descent algorithm, the layout parameters (line width, layer spacing) are constrained and optimized. On the premise of ensuring signal integrity, the radiation peak is reduced. Then, through multi-condition Monte Carlo simulation, the influence of manufacturing tolerances and environmental variations is quantitatively analyzed, and a residual-driven correction rule library is generated to improve the correction efficiency and the development efficiency of PCB products.

[0027] To more clearly introduce the technical solution of the present application, the technical solution of the present application will also be introduced through specific embodiments below. It should be noted that the specific embodiment is used to expand the description of the technical solution of the present application, rather than limiting the present application.

[0028] In some embodiments, extracting the high-frequency signal characteristics in the circuit schematic diagram to obtain an electromagnetic simulation parameter set includes: S1011-S1015.

[0029] S1011. Determine the high-speed signal network, power network and interconnection structure according to the circuit schematic diagram.

[0030] Exemplarily, a high-speed signal network is defined as a transmission path with a signal rise time less than 1 nanosecond. It is necessary to perform time-domain reflectometry (TDR) simulation on all transmission lines in the schematic diagram based on a signal integrity analysis tool (such as ANSYS SIwave), filter out sensitive paths with a delay jitter exceeding 10%, and mark their termination impedance, trace length, and turning angle information. The power network includes a power plane, a decoupling capacitor network, and power supply pins. The voltage fluctuation range (such as 3.3V ± 5%), current-carrying capacity (such as a maximum of 10A), and plane segmentation topology of all power domains are extracted through a power integrity analysis module to identify high-risk coupling regions with a distance between adjacent power domains less than 0.2mm. The interconnect structure includes signal vias, connector pins, and traces on adjacent layers. Three-dimensional electromagnetic field analysis technology is used to perform topology marking on parallel trace regions with a distance less than 3 times the line width, and their geometric parameters are extracted (such as via diameter 0.2mm, pad size 0.4mm × 0.4mm). The above operations output a high-speed signal network topology diagram, a power network impedance distribution diagram, and a set of interconnect coupling region coordinates through a circuit network feature classification engine, providing a structured input for subsequent feature parameter extraction.

[0031] S1012. Analyze the frequency-domain energy distribution of the high-speed signal network to generate a set of fundamental frequency harmonic characteristic parameters.

[0032] Exemplarily, for the operation of analyzing the frequency-domain energy distribution of the high-speed signal network, the sweep frequency technology of a vector network analyzer (VNA) is adopted to measure the S-parameters (such as S21 insertion loss, S11 return loss) of the transmission line point by point at a step of 10MHz in the frequency band from 1MHz to 20GHz. The generation of the set of fundamental frequency harmonic characteristic parameters is to convert the time-domain signal (such as a 1GHz square wave) into a frequency-domain energy spectrum through a fast Fourier transform (FFT), and extract the amplitude (such as the 3rd harmonic -25dBc), phase shift (such as the 5th harmonic 30 degrees), and spectral density (such as the energy density at the 1GHz frequency point 0.8mW / GHz) of the fundamental frequency (1GHz) and its 3rd, 5th, and 7th harmonics. The energy density threshold is set to -30dBc, and harmonic components below this threshold are regarded as background noise and filtered out. The parameter set is stored through a frequency band - energy correlation matrix. The rows of the matrix correspond to frequency points (1MHz to 20GHz), the columns correspond to amplitude, phase, and energy density values, and key resonance frequency points are marked (such as a 50% sudden increase in energy density at 2.4GHz), providing frequency-domain excitation characteristics for electromagnetic coupling modeling.

[0033] S1013. Perform modal impedance analysis on the power network to generate a set of frequency-varying impedance parameters.

[0034] Exemplarily, when performing modal impedance analysis on the power network, a multi-port excitation source (such as a 1A current source) is loaded on the power plane, and the input impedance matrix of each port within the frequency band from 0.1 GHz to 5 GHz is calculated through a three-dimensional full-wave electromagnetic field solver (such as CST Studio Suite). The impedance matrix is subjected to singular value decomposition (SVD) to extract the main modal impedance curve (such as the impedance of the first mode drops to 0.5 Ω at 1.2 GHz), and the resonant frequency points are marked (such as the impedance peak of 15 Ω at 3.5 GHz), generating a frequency-varying impedance parameter set. The planar resonance mode is visualized through the electric field distribution contour map to identify the strong resonance region at the edge of the power plane (such as 2 mm away from the boundary). The parameter set is stored as a frequency band-impedance association table, containing the real and imaginary impedance values at each frequency point (such as 0.8 + j0.2 Ω at 1 GHz), and the high-risk frequency bands (such as 0.8 GHz to 1.5 GHz) are marked through an impedance tolerance threshold (such as ±20%), providing frequency-varying impedance boundary conditions for multi-physics coupling modeling.

[0035] S1014. Perform near-field coupling modeling on the interconnect structure to generate a set of parasitic parameter characteristics.

[0036] Exemplarily, the method of moments (MoM) and partial element equivalent circuit (PEEC) hybrid algorithm are used to calculate the electromagnetic field distribution of adjacent traces with a spacing less than 0.3 mm. The mutual capacitance (such as 0.05 pF / mm between parallel traces), mutual inductance (such as 1.2 nH / mm for traces on adjacent layers), and crosstalk coefficient (such as -30 dB for near-end crosstalk) are quantified. The via parameters need to be extracted based on a three-dimensional field solver to calculate its equivalent inductance (such as 0.1 nH for a through hole) and capacitance (such as 0.02 pF for the pad to ground), and a coupling coefficient matrix of the via array is established (such as the coupling degree between adjacent vias is -25 dB), generating a set of parasitic parameter characteristics. The parameter set is stored through a spatial coordinate-parasitic parameter mapping table, marking the high-coupling regions (such as the crosstalk coefficient suddenly increases to -20 dB at the corner), and associating them with the network nodes in the schematic diagram (such as the coupling path of signal line A to signal line B), providing near-field coupling characteristics for electromagnetic noise source tracing.

[0037] S1015. Integrate the fundamental harmonic characteristic parameter set, the frequency-varying impedance parameter set, and the parasitic parameter characteristic set into an electromagnetic simulation parameter set.

[0038] Exemplarily, data alignment and association are achieved through a hierarchical feature fusion module. The frequency domain matrix (1 MHz - 20 GHz) of the fundamental frequency harmonic feature parameter set and the frequency band (0.1 GHz - 5 GHz) of the frequency-variable impedance parameter set are frequency-domain matched through an interpolation algorithm, and the interpolation step is set to 1 MHz. The spatial coordinate data of the parasitic parameter feature set is associated with the geometric position of the structural model through grid mapping (such as the via coordinates matching the drilling positions in the 3D model). The fused electromagnetic simulation parameter set is stored in a multi-dimensional tensor structure, including a frequency domain response matrix (frequency × amplitude / phase), a spatial coupling tensor (coordinate × mutual capacitance / mutual inductance), and a network topology relationship graph (node × connection relationship), and the parameter consistency is checked through a data verification engine (such as the logical correlation between impedance parameters and resonance frequencies), and the output is a standardized HDF5 format file, which supports direct import into electromagnetic simulation tools for multi-physics field modeling.

[0039] In some embodiments, a multi-physics field coupling model is established based on the electromagnetic simulation parameter set and the structural model to obtain an electromagnetic field - circuit model, including: S1021 - S1026.

[0040] S1021. Perform three-dimensional electromagnetic topology reconstruction on the structural model to generate a first geometric model.

[0041] Exemplarily, geometric model reconstruction is performed based on the PCB stack-up structure (such as a 6-layer board with a thickness of 1.6 mm) and material properties (such as the FR4 dielectric constant of 4.3). The three-dimensional electromagnetic topology reconstruction uses non-uniform rational B-spline (NURBS) surface fitting technology to convert the planar contours of each layer of conductors (such as the top signal layer and the middle power layer) and via arrays (such as through holes with a diameter of 0.2 mm) in the structural model into a smooth surface model, and the surface fitting accuracy is controlled within 0.01 mm. The surface roughness of the conductor (such as the copper foil RMS roughness of 0.8 μm) is modeled through a fractal geometry algorithm to generate microscopic concave and convex textures to accurately simulate the skin effect. The thickness tolerance of the dielectric layer (such as ±5%) generates a random deviation data set through Monte Carlo sampling and is mapped to the corresponding area of the geometric model. The first geometric model is output as a three-dimensional solid model containing the stack-up topology relationship (such as the layer-to-layer spacing of 0.1 mm), material property labels (such as the copper layer conductivity of 5.8×10 7 S / m), and surface roughness parameters, providing a geometric basis for subsequent material parameter loading.

[0042] S1022. Load the preset frequency-variable material parameters into the first geometric model to generate a second geometric model.

[0043] Exemplarily, a frequency-varying interpolation function is established based on the measured data of the material (such as the real and imaginary parts of the FR4 dielectric constant in the frequency band from 1 GHz to 10 GHz), so as to load the preset frequency-varying material parameters into the first geometric model. The loading of the frequency-varying material parameters adopts the field mapping algorithm, and the real and imaginary parts of the dielectric constant are respectively associated with the dielectric region of the geometric model (such as the real part of the dielectric constant 4.3@1 GHz is mapped to dielectric layer 1), and the conductivity parameter is associated with the conductor region (such as the conductivity of the copper layer 5.8×10 7 S / m). The frequency band sampling interval is set to 50 MHz, and the cubic spline algorithm is used for the interpolation function to ensure parameter continuity. The principal axis direction parameters of the anisotropic material (such as glass fiber cloth-reinforced epoxy resin) are loaded in tensor form, and the orthogonal axes of the material are marked (such as the dielectric constant difference in the X / Y / Z directions is ±0.2). The output of the second geometric model is an enhanced three-dimensional model containing frequency-varying material tags (such as the dielectric constant of dielectric layer 1 at 5 GHz is 4.1-j0.05) and anisotropic properties, which supports the frequency-domain adaptive calculation of the electromagnetic field solver.

[0044] S1023. Perform adaptive mesh refinement on the second geometric model to generate a discretized mesh model.

[0045] Exemplarily, the curvature-driven and field strength gradient dual criteria are used to dynamically adjust the mesh density to complete the adaptive mesh refinement. The curvature-driven refinement is applied to high-curvature regions such as trace corners (such as 90-degree turns), via hole edges (such as a drilling diameter of 0.2 mm), and pad gaps (such as a 0.15 mm pitch), and the reference mesh size of 0.5 mm is encrypted to 0.05 mm. The field strength gradient criterion is based on the initial electric field simulation results, and local encryption is implemented in the region where the rate of change of the electric field strength exceeds 20 V / (m·mm) (such as the edge of the power plane). The aspect ratio of the mesh elements is limited to 5:1 to avoid distortion, and a mixed tetrahedron and hexahedron mesh strategy is used to balance the calculation accuracy and efficiency. The output of the discretized mesh model is a refined data set containing 120 million mesh elements, and the conformal mesh nodes at the conductor-dielectric interface are marked (such as the mesh alignment accuracy at the copper foil and FR4 interface is ±0.005 mm) to ensure the numerical stability of the electromagnetic field calculation.

[0046] S1024. Perform circuit-electromagnetic port coupling processing on the discretized mesh model to generate a hybrid-domain interface matrix.

[0047] Exemplarily, for the operation of circuit-electromagnetic port coupling processing on the discretized grid model, the excitation source (such as a 50 Ω impedance clock driver) and the load (such as a 100 Ω differential terminal) in the circuit schematic are mapped to the port boundaries of the three-dimensional grid model. The circuit-electromagnetic port coupling is achieved through a mixed-domain interface matrix, where the matrix rows correspond to circuit node voltages / currents, and the columns correspond to the tangential electric / magnetic field components of the electromagnetic field. The voltage-electric field coupling relationship is defined by the potential continuity condition at the port boundary (e.g., a port voltage of 1 V corresponds to a tangential electric field strength of 50 V / m), and the current-magnetic field coupling is established through Ampere's circuital law (e.g., a current of 1 A corresponds to a magnetic field strength of 0.2 A / m). For multi-port coupling (such as 8 differential pairs), impedance matrix normalization is performed to eliminate the phase error introduced by the mutual coupling between ports. The output of the mixed-domain interface matrix is a sparse matrix in complex form (dimension 2000×2000), which is stored in CSR (compressed sparse row) format to support the call of an efficient solver.

[0048] S1025. Calculate the full-wave electromagnetic-circuit co-relationship of the mixed-domain interface matrix to generate the field-circuit co-simulation result.

[0049] Exemplarily, at each time step (such as 0.1 ps), the field-circuit co-solver synchronously updates the electromagnetic field distribution (such as electric field strength E and magnetic field strength H) and the circuit node voltages / currents (such as the driver output waveform). The electromagnetic field solution is based on the discretization of Maxwell's equations on a Yee grid, and the circuit solution uses the modified nodal analysis (MNA). The field-circuit data interaction is achieved through the mixed-domain interface matrix. The tangential components of the electromagnetic field are converted into equivalent voltage sources / current sources at the circuit ports, and the circuit response is inversely updated to the field domain boundary conditions. The field-circuit co-simulation results are output as a time-domain field distribution sequence (such as recording the full-field electric field data every 10 ps) and a frequency-domain S-parameter matrix (such as S11 at 5 GHz is -15 dB), covering the electromagnetic characteristics in the frequency band from 0.1 GHz to 10 GHz. S1026. Perform multi-dimensional verification processing on the field-circuit co-simulation results to generate the electromagnetic field-circuit model.

[0050] Exemplarily, the operations for multi-dimensional verification processing of the field-circuit co-simulation results include grid convergence verification, S-parameter error checking, and energy conservation analysis. Grid convergence verification is performed by comparing the field distribution differences under three different grid densities (such as reference sizes of 0.5 mm, 0.2 mm, and 0.05 mm), and it is confirmed that the maximum electric field strength deviation is less than 5%. S-parameter error checking is carried out by comparing the measured data of a vector network analyzer (such as the S21 insertion loss in the frequency band from 1 GHz to 10 GHz) with the simulation results, and the root mean square error is controlled within 1 dB. Energy conservation analysis is conducted by calculating the sum of the input power (such as injecting 1 W at the circuit port) and the radiation loss, dielectric loss, and conductor loss to ensure that the energy deviation is less than 2%. The output of the electromagnetic field-circuit model is a three-dimensional field distribution database (including electric field, magnetic field, and current density), frequency-domain network parameters, and circuit response waveforms that pass all verification items, supporting accurate modeling for subsequent electromagnetic noise analysis.

[0051] In some embodiments, a frequency-time domain hybrid excitation is loaded onto the electromagnetic field-circuit model to obtain an electromagnetic noise distribution map, including: S1031 - S1034.

[0052] S1031. Add a frequency-time domain excitation source to the electromagnetic field-circuit model to generate a set of hybrid excitation signals.

[0053] Exemplarily, by analyzing the fundamental frequency harmonic distribution (such as the 1 GHz fundamental wave and its 3rd and 5th harmonics) and time-domain waveform parameters (such as a rise time of 50 ps) in the electromagnetic simulation parameter set, the spectrum and transient characteristics of the hybrid excitation source are defined. The frequency-domain excitation is configured as a linear swept-frequency signal from 1 GHz to 10 GHz, with a step size set to 100 MHz, and a continuous wave signal with an injected power of 0 dBm at each frequency point. The time-domain excitation generates a differential signal with a rate of 5 Gbps based on a pseudo-random binary sequence (PRBS31), and at the same time, a transient step current (amplitude 2 A, rise time 10 ns) is configured for the power supply network. The spectrum of the time-domain PRBS signal is decomposed into discrete frequency point components through Fourier series and superimposed with the frequency-domain swept-frequency signal to form a set of hybrid excitation signals covering the entire frequency band. This library contains frequency-domain complex voltage vectors (such as V f = [1 GHz: 1∠0°, 2 GHz: 0.8∠45°, …]) and time-domain transient current waveform sampling points (such as I t = [0 ns: 0 A, 0.2 ns: 1.8 A, …]), providing input conditions for subsequent field-circuit co-simulation.

[0054] S1032. Perform field-circuit co-loading on the set of hybrid excitation signals to generate an original noise data set.

[0055] Exemplarily, a set of mixed excitation signals is loaded into the corresponding ports of the electromagnetic field - circuit model. The frequency - domain excitation is injected into the high - speed signal line (such as the TX end of the PCIe channel) through the S - parameter port, and the port impedance matching is set to 100Ω differential. The time - domain transient excitation is loaded into the VRM output of the power network through a current source, and the initial condition is set to a steady - state DC voltage (such as 12 V). A FDTD - FEM hybrid solver is used to perform the field - circuit co - simulation: in the frequency domain, the finite - element method is used to calculate the electric - field distribution (such as the maximum field strength of 120 V / m at 10 GHz), and in the time domain, the FDTD is used to iteratively solve the current propagation of the transmission line (such as the peak - to - peak reflection noise of 60 mA at the via). The simulation outputs the original noise data set, which includes: the frequency - domain S - parameter matrix (dimension N×N, where N is the number of ports), the time - domain radiation - field time series (such as the E - field values sampled every 0.1 ns), and the conduction - noise current spectrum of the power network (0 - 100 MHz). The data is stored as a three - dimensional matrix containing frequency, time, and spatial coordinates, with the dimension of (number of frequency points×number of time steps×number of grid cells).

[0056] S1033. Quantify the near - field and far - field noise of the original noise data set to generate a set of noise characteristic parameters.

[0057] Exemplarily, near - field and far - field quantization analysis is performed on the original noise data output from the simulation. For near - field analysis, the magnetic - probe equivalent algorithm is used. A virtual probe array (grid density 1 mm×1 mm) is set at a height of 1 cm above the PCB surface, and the H - field distribution at each probe point is calculated through the Biot - Savart law (such as the peak H z = 5 A / m@3 GHz in the BGA area). For far - field analysis, based on the equivalent radiation model, the Stratton - Chu formula is used to extrapolate the near - field data to a distance of 1 meter to calculate the electric - field intensity (such as E theta = 58 dBμV / m at 2.4 GHz). Time - domain noise analysis converts the conduction current into a spectral - density curve through FFT (such as the noise - current density of 12 mA / √Hz at 50 MHz), and the peak - to - peak value of the common - mode voltage is extracted (such as V cm = 200 mVpp at the VRM output). The generated set of noise characteristic parameters includes a list of field strengths in the frequency domain, a time - domain spectral envelope, and spatial - coordinate - related data, providing structured input for noise - distribution visualization.

[0058] S1034. Perform spatial - frequency - domain mapping processing on the set of noise characteristic parameters to generate an electromagnetic - noise distribution map.

[0059] Exemplarily, the noise characteristic parameters are mapped to the PCB layout space to generate a multi-dimensional electromagnetic noise distribution map. In the layout coordinate system, with the X-Y position as the horizontal and vertical axes and the frequency as the third dimension, a field strength data cube is constructed (such as a size of 200 mm × 150 mm × 100 frequency points). The field strength data of discrete probe points is spatially interpolated through the Kriging interpolation algorithm to generate a continuous radiation field equipotential surface (such as the area covered by the 55 dBμV / m equipotential surface is 120 mm²). A heat map is used to render the over-standard frequency bands (such as the red area above the FCC Class B limit of 54 dBμV / m), and the high-radiation positions are marked on the map (such as the coordinates (45 mm, 80 mm) under the clock driver). For conducted noise, noise current density vector arrows are superimposed on the power plane (such as the current density in the PDN edge area > 15 mA / mm²). The output electromagnetic noise distribution map contains the radiation and conducted noise characteristics associated with the frequency-spatial domain, providing a visual criterion for layout optimization.

[0060] In some embodiments, the layout parameters of the structural model are optimized by gradient descent according to the electromagnetic noise distribution map to obtain an anti-interference layout scheme, including: S1041 - S1044.

[0061] S1041. Determine an initial layout parameter set according to the preset layout constraint boundaries and the layout parameters of the structural model.

[0062] Exemplarily, based on the over-standard frequency points and high-noise regions in the electromagnetic noise distribution map, the initial optimization variables of the signal line width, interlayer spacing, via number, and power split spacing are defined. The layout constraint boundaries are set through topology constraint rules (such as the 3W principle and the 20H principle).

[0063] By analyzing the over-standard frequency bands (such as the 2.4 GHz radiation exceeding the standard) and spatial hotspots (such as the ±5 mm area around the power pin) in the electromagnetic noise distribution map, the key layout parameters affecting the EMC performance are extracted. The initial layout parameter set includes the signal line width (default 0.15 mm), interlayer spacing (default 0.2 mm), via number (default 3 vias per centimeter of trace), and power split spacing (default 1 mm). Based on the 3W principle (line spacing ≥ 3 times the line width), the lower limit of the spacing between adjacent signal lines is set, and based on the 20H principle (the power plane indents 20 times the dielectric thickness), the indentation boundaries between the power layer and the ground layer are defined. For high-noise regions (such as under the clock driver), an additional density constraint on shielded ground vias is added (such as 1 via per square millimeter). The feasible region boundary is jointly determined by the PCB manufacturing process limits (such as the minimum line width of 0.1 mm and the minimum hole diameter of 0.2 mm) and the signal integrity requirements (such as an impedance deviation of ±10%). The generated initial parameter set is stored in vector form, with a dimension of N×1 (N is the number of optimization variables, and the typical value N = 50), serving as the starting point for gradient descent optimization.

[0064] S1042. Calculate the electromagnetic performance indicators of the initial layout parameter set and generate the objective function gradient matrix.

[0065] Exemplarily, through the full-parameterization simulation of the electromagnetic field-circuit model, quantify the influence of layout parameters on the noise indicators. The objective function J = 0.5·I noise + 0.3·E rad + 0.2·X talk In this formula, the conduction noise current density I noise takes the RMS value of the PDN network (such as the integral value in the 0-100 MHz frequency band), the radiation field strength E rad takes the peak value at 1 meter (such as the 2.4 GHz frequency point), and the crosstalk amplitude X talk takes the near-end crosstalk coefficient between adjacent differential line pairs. The central difference method is used to calculate the partial derivative of the objective function with respect to the layout parameters: ∂J / ∂w i ≈ [J(w i +Δw) - J(w i -Δw)] / (2Δw), where Δw = 0.01 mm is the step size, and w i is the i-th line width parameter. For discrete variables such as the number of vias, they are mapped to continuous variables through relaxation processing (such as via density ρ ∈ [0,1]). The dimension of the Jacobian matrix is M×N (M = 3 objective functions, N = 50 parameters), which stores the sensitivity coefficients of each parameter to the three types of noise indicators. During the calculation process, the electromagnetic simulator (such as HFSS) is called in parallel to perform fast solution after parameter perturbation. The time-consuming for a single gradient calculation is about 20 minutes (assuming 100 simulations are required when N = 50), and the objective function gradient matrix is generated.

[0066] S1043. Perform momentum accelerated descent iteration on the objective function gradient matrix to generate the optimized parameter update sequence.

[0067] Exemplarily, the Adam optimization algorithm is used to implement the momentum accelerated gradient descent iteration. The parameter update formula is: m t =μ·m {t-1} +(1-μ)·∇J(w t ); v t =ν·v {t-1} +(1-ν)·(∇J(w t )); 2 ; w {t+1} =w t -η·m t / ( +ε); Among them, the hyperparameters are set as μ = 0.9 (momentum term), ν = 0.999 (adaptive learning rate term), ε = 1e - 8 (to prevent zero perturbation). The initial learning rate η = 0.1 is dynamically adjusted according to the parameter sensitivity: if the objective function decrease rate in three consecutive iterations < 1%, then η ← η × 0.5; if the decrease rate > 5%, then η ← η × 1.2. For the line width parameter, the update amplitude each time is limited within ±0.02 mm to prevent violation of manufacturing constraints; the via density parameter is mapped back to discrete values through the Sigmoid function (for example, when ρ > 0.7, one via is added). After each iteration, verify whether the layout parameters are within the feasible region (such as line spacing ≥ 3W), and if they exceed the boundary, perform projection correction. After about 50 iterations (taking about 17 hours), the objective function J usually converges to less than 30% of the initial value (such as from 120 to 35), and the radiation field strength and conduction noise are significantly reduced. The optimized parameter sequence is stored indexed by time steps for compliance verification.

[0068] S1044. Perform EMC compliance verification on the optimized parameter update sequence to generate an anti - interference layout scheme.

[0069] Exemplarily, compare the noise metrics before and after optimization: the radiation field strength needs to be lower than the FCC Class B limit (such as 54 dBμV / m@3 m), the integrated value of the conduction noise within 100 MHz < 60 mA·MHz, and the crosstalk amplitude ≤ - 30 dB. If the verification passes, output the anti - interference layout scheme, including the updated line width (such as the clock line is widened to 0.18 mm), the layer - to - layer spacing (such as the power - ground layer spacing is increased to 0.3 mm), the via layout (such as the shielding ground via density is increased to 5 per centimeter), and the power splitting rule (such as the sensitive analog area is indented by 0.5 mm). The scheme is stored in the form of a three - dimensional layout rule library, including constraint conditions (such as the differential pair length deviation < 5 mil) and priority strategies (such as preferentially optimizing the high - frequency signal layer). For the unqualified parameters, trigger the residual feedback mechanism to re - execute the S503 iteration until all metrics meet the EMC compliance requirements, that is, generate an anti - interference layout scheme that can be directly imported into the PCB design tool (such as Cadence Allegro).

[0070] In some embodiments, perform momentum - accelerated descent iteration on the objective function gradient matrix to generate an optimized parameter update sequence, including: S431 - S434.

[0071] S431. Construct a momentum correction term based on the objective function gradient matrix to generate an iterative direction correction vector.

[0072] S432. Update the layout parameters according to the iterative direction correction vector and the preset momentum factor to generate an intermediate solution set of parameter iteration.

[0073] S433. Perform constraint boundary projection on the intermediate solution set of parameter iteration to generate a sequence of physically feasible solutions.

[0074] S434. Generate an updated sequence of optimization parameters according to the convergence criterion of the sequence of physically feasible solutions.

[0075] Exemplarily, the parameter update path is dynamically adjusted by combining historical gradient information with the current gradient direction, thereby overcoming the problem that the traditional gradient descent method is prone to falling into local minima or oscillatory convergence. The construction of the gradient matrix of the objective function is based on the sensitivity relationship between the electromagnetic noise distribution map and the layout parameters. The partial derivatives of the radiation field strength with respect to parameters such as line width, layer spacing, and via position are calculated by the adjoint field method, forming a gradient matrix with the dimension of the number of parameters × the number of frequency points (such as 100 parameters × 100 frequency points). The exponential weighted average method is used to accumulate historical gradient information to generate the momentum correction term. The initial value of the momentum factor is set to 0.9 and is dynamically adjusted according to the consistency of the iteration direction (such as increasing to 0.95 when the iteration directions are consistent for three consecutive times). In the calculation of the iteration direction correction vector, the current gradient matrix and the momentum correction term are added according to weights (such as the current gradient weight is 0.1 and the historical momentum weight is 0.9), and the dimension difference is eliminated through normalization to generate a unit direction vector (such as the magnitude of the line width gradient direction vector is 1). Combining the preset learning rate (such as the initial learning rate of 0.01) with the correction direction vector, the layout parameters are adjusted along the negative gradient direction (such as the line width is increased by 0.02 mm and the layer spacing is decreased by 0.01 mm). At the same time, an adaptive learning rate decay mechanism is introduced (such as the learning rate decays by 5% every 10 iterations) to avoid excessive parameter update step sizes causing oscillations and to achieve the update of the intermediate solution set of the parameter iteration. During the constraint boundary projection, the intermediate solution set is hard-corrected according to the manufacturing process limitations (such as the minimum line width of 0.1 mm and the minimum line spacing of 0.15 mm). The Lagrange multiplier method is used to project the violated parameters into the feasible region (such as forcing the line width of 0.09 mm to 0.1 mm and compensating the adjacent trace spacing to 0.16 mm), and quadratic programming optimization is used to ensure that the parameters after correction satisfy impedance continuity (such as the single-ended trace impedance is 50 Ω ± 5%). After each iteration, design rule checking (DRC) and electromagnetic compatibility (EMC) pre-verification are performed to generate a sequence of physically feasible solutions, eliminating parameter combinations that violate process specifications or radiation exceedance (such as the via density exceeds 5 per cm² or the radiation field strength at 3 GHz exceeds 30 V / m). The setting of the convergence criterion includes dual conditions: the relative change rate of the objective function value is less than 0.1% (such as the difference in the total radiation energy between adjacent iterations is lower than 0.5 mW) and the magnitude of the gradient matrix is lower than the threshold (such as 1e-4). At the same time, an early stopping mechanism is introduced (such as terminating if the convergence condition is not met for 5 consecutive iterations). Through parameter sensitivity ranking, key parameters with a contribution to radiation suppression greater than 5% are screened out (such as adjusting the power supply layer spacing reduces the radiation at 3 GHz by 8 dB), and the mapping relationship between their iteration trajectories (such as the line width is gradually optimized from 0.15 mm to 0.18 mm) and the radiation suppression effect is recorded, forming an optimized sequence of layout parameters that supports the import of automated design tools, covering parameter names, adjustment ranges, and quantitative indicators of EMC performance improvement (such as the radiation in the 5 GHz band is reduced by 12 dB after optimization), obtaining an optimized parameter update sequence.

[0076] In some embodiments, a multi - condition Monte Carlo simulation and residual calibration are performed on the anti - interference layout scheme to generate a correction guidance library, including: S1051 - S1056.

[0077] S1051. Perturb the multi - physical - field parameters of the anti - interference layout scheme to generate a Monte Carlo sample set.

[0078] S1052. Perform a multi - physical - field joint simulation according to the Monte Carlo sample set to generate an EMC performance data set.

[0079] S1053. Perform principal component analysis and clustering processing on the EMC performance data set to generate a residual feature vector set.

[0080] S1054. Perform residual correction analysis on the residual feature vector set to generate a key parameter correction coefficient matrix.

[0081] S1055. Refine rules and prioritize the correction coefficient matrix to generate a correction rule library.

[0082] S1056. Perform cross - condition generalization verification on the correction rule library to generate an electromagnetic compliance correction guidance library.

[0083] Exemplarily, the process of performing multi-condition Monte Carlo simulation and residual calibration on the anti-interference layout scheme to generate a correction guidance library is achieved through a closed-loop process of parameter perturbation, simulation verification, data analysis, and rule refinement. The multi-physical field parameter perturbation mainly focuses on manufacturing tolerances and environmental variables: The manufacturing tolerance perturbation covers the random deviations of line width (±10%), dielectric constant (±5%), and copper thickness (±8%). The Latin Hypercube Sampling is used to construct 200 groups of Monte Carlo sample sets to ensure uniform coverage of the parameter space. The environmental perturbation simulates the changes in temperature (-40°C to 125°C) and humidity (10% to 90%). The temperature drift characteristics of the material are modeled based on the Arrhenius equation (such as the temperature coefficient of the FR4 dielectric constant is -50 ppm / °C), and the humidity effect is quantified through the hygroscopic expansion model (such as the dielectric constant increases by 0.1 when the water absorption rate is 0.2%). The multi-physical field co-simulation uses the co-simulation framework of ANSYS HFSS and Mechanical. Each group of samples performs electromagnetic-thermal-stress coupling calculations: The electromagnetic field solution frequency band covers 1 MHz to 10 GHz (step size 100 MHz), the thermal simulation calculates the steady-state temperature rise distribution (such as ΔT 25°C in the power supply area), and the mechanical simulation evaluates the warpage deformation amount (such as the maximum deformation is 0.05 mm). The EMC performance dataset includes the radiation field strength (such as the average field strength at 3 GHz is 35 V / m), the conducted noise (such as the power pin noise ripple is ±5%), the temperature rise sensitivity (such as the radiation increases by 1.2 dB when the temperature rises by 10°C), and the impact of deformation on impedance (such as a warpage of 0.1 mm causes an impedance shift of 3%). The data is stored in a multi-dimensional tensor structure (sample × frequency point × physical quantity). Principal Component Analysis (PCA) reduces the dimension of the dataset, extracts the first 3 principal components with a cumulative contribution rate exceeding 85% (such as principal component 1 is related to the line width and the radiation field strength, with a contribution rate of 52%), and divides the samples into three categories: compliant (radiation field strength < 30 V / m), critical (30 - 50 V / m), and non-compliant (> 50 V / m) through K-means clustering, and labels the high-risk parameter combinations of the critical class samples (such as a negative line width deviation of 8% superimposed with a high temperature of 125°C). The residual feature vector set is generated by calculating the residual between the measured field strength and the simulation prediction value (such as the root mean square error of the residual is 1.5 dB). The residual correction analysis uses a random forest regression model to establish the non-linear mapping relationship between the process parameter deviation (such as the dielectric constant + 3%) and the residual, and outputs the key parameter correction coefficient matrix (such as a negative line width deviation of 5% corresponding to a layer spacing correction of +0.03 mm can reduce the radiation by 2 dB). The refinement of the correction rule library converts the correction coefficient into an IF-THEN type conditional rule (such as "IF temperature > 85°C AND line width < 90% of the nominal value THEN increase the power supply layer spacing by 0.1 mm"), and the rule priorities are sorted based on parameter sensitivity (such as the contribution degree of the power supply layer spacing adjustment to the 5 GHz radiation suppression is 12%) and the working condition coverage rate (such as the rule applies to 80% of the high temperature and high humidity samples), and redundant rules with a coverage rate lower than 5% are removed.Cross - operating - condition generalization verification resamples 100 groups of test samples under an extended temperature range (-55°C to 150°C), humidity range (5% to 95%), and manufacturing tolerance (line width ±15%) to verify the effectiveness of the correction rules (e.g., the residual correction error of 90% of the rules in the rule base is <1 dB under the extended operating conditions), and shrinks the parameter range or replaces the failure rules (e.g., when the humidity > 95%, the dielectric constant drift exceeds the model prediction range) with alternative rules (e.g., increasing the decoupling capacitance density). The electromagnetic compliance correction guidance library outputs a structured database containing 150 prioritized rules, with each rule marked with the applicable operating condition range, parameter adjustment amount, and expected EMC performance improvement value (e.g., the 3 - GHz radiation is reduced by 6 dB ±0.5 dB), supports integration into EDA tools to achieve automatic layout parameter optimization, and tracks the applicability iteration update of the rules under different process nodes (such as 10 - layer boards and HDI boards) through the version management module.

[0084] Please refer to Figure 2 , Figure 2 FIG. Figure 2 is a schematic block diagram of an electromagnetic compatibility test system for PCB design provided by an embodiment of the present application. The electromagnetic compatibility test system 200 for PCB design is used to execute the aforementioned electromagnetic compatibility test method for PCB design. Among them, the electromagnetic compatibility test system 200 for PCB design can be configured in a server.

[0085] Among them, the server can be an independent server, a server cluster, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, Content Delivery Network (CDN), and big data and artificial intelligence platforms.

[0086] As Figure 2 shown, the electromagnetic compatibility test system 200 for PCB design includes: a high - frequency analysis module 201, a parameter modeling module 202, an excitation loading module 203, a layout optimization module 204, and a correction generation module 205.

[0087] The high - frequency analysis module 201 is used to extract the high - frequency signal characteristics in the circuit schematic diagram to obtain an electromagnetic simulation parameter set.

[0088] The parameter modeling module 202 is used to perform multi - physical - field coupling modeling based on the electromagnetic simulation parameter set and the structure model to obtain an electromagnetic field - circuit model.

[0089] The excitation loading module 203 is used to perform frequency - time - domain hybrid excitation loading on the electromagnetic field - circuit model to obtain an electromagnetic noise distribution map.

[0090] The layout optimization module 204 is configured to perform gradient descent optimization on the layout parameters of the structural model according to the electromagnetic noise distribution map to obtain an anti-interference layout scheme.

[0091] The correction generation module 205 is configured to perform multi-condition Monte Carlo simulation and residual calibration on the anti-interference layout scheme to generate a correction guidance library.

[0092] An embodiment of the present application provides an electronic device, which includes a memory and a processor; the memory is used to store a computer program; the processor is configured to execute the computer program and implement the electromagnetic compatibility test method for PCB design according to any one of the embodiments of the present application when executing the computer program.

[0093] An embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor is caused to implement the electromagnetic compatibility test method for PCB design according to any one of the embodiments of the present application.

[0094] The above is only the specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present application can easily think of various equivalent modifications or substitutions, and these modifications or substitutions should all be covered within the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. An electromagnetic compatibility testing method for PCB design, characterized in that, The PCB includes: A circuit schematic diagram and a structure model, and the method includes: Extracting high-frequency signal features in the circuit schematic diagram to obtain an electromagnetic simulation parameter set; Performing multi-physics field coupling modeling according to the electromagnetic simulation parameter set and the structure model to obtain an electromagnetic field-circuit model; Performing frequency-time domain hybrid excitation loading on the electromagnetic field-circuit model to obtain an electromagnetic noise distribution map; Performing gradient descent optimization on the layout parameters of the structure model according to the electromagnetic noise distribution map to obtain an anti-interference layout scheme; Performing multi-condition Monte Carlo simulation and residual calibration on the anti-interference layout scheme to generate a correction guidance library.

2. The electromagnetic compatibility test method for PCB design according to claim 1, wherein, The extracting high-frequency signal features in the circuit schematic diagram to obtain an electromagnetic simulation parameter set includes: Determining high-speed signal networks, power supply networks, and interconnection structures according to the circuit schematic diagram; Performing frequency-domain energy distribution analysis on the high-speed signal networks to generate a fundamental frequency harmonic feature parameter set; Performing modal impedance analysis on the power supply networks to generate a frequency-variable impedance parameter set; Performing near-field coupling modeling on the interconnection structures to generate a parasitic parameter feature set; Integrating the fundamental frequency harmonic feature parameter set, the frequency-variable impedance parameter set, and the parasitic parameter feature set into an electromagnetic simulation parameter set.

3. The electromagnetic compatibility test method for PCB design according to claim 1, characterized in that, The performing multi-physics field coupling modeling according to the electromagnetic simulation parameter set and the structure model to obtain an electromagnetic field-circuit model includes: Performing three-dimensional electromagnetic topology reconstruction on the structure model to generate a first geometric model; Loading preset frequency-variable material parameters to the first geometric model to generate a second geometric model; Performing adaptive mesh division on the second geometric model to generate a discretized mesh model; Performing circuit-electromagnetic port coupling processing on the discretized mesh model to generate a mixed-domain interface matrix; Calculating the full-wave electromagnetic-circuit collaborative relationship of the mixed-domain interface matrix to generate a field-circuit joint simulation result; Performing multi-dimensional verification processing on the field-circuit joint simulation result to generate an electromagnetic field-circuit model.

4. The electromagnetic compatibility testing method for PCB design according to claim 1, characterized in that The performing frequency-time domain hybrid excitation loading on the electromagnetic field-circuit model to obtain an electromagnetic noise distribution map includes: Adding a frequency-time domain excitation source to the electromagnetic field-circuit model to generate a mixed excitation signal set; Performing field-circuit collaborative loading on the mixed excitation signal set to generate an original noise data set; Quantifying the near / far field noise of the original noise data set to generate a noise feature parameter set; Performing space-frequency domain mapping processing on the noise feature parameter set to generate an electromagnetic noise distribution map.

5. The electromagnetic compatibility test method for PCB design according to claim 1, characterized in that The performing gradient descent optimization on the layout parameters of the structure model according to the electromagnetic noise distribution map to obtain an anti-interference layout scheme includes: Determining an initial layout parameter set according to a preset layout constraint boundary and the layout parameters of the structure model; Calculating the electromagnetic performance index of the initial layout parameter set to generate a target function gradient matrix; Performing momentum accelerated descent iteration on the target function gradient matrix to generate an optimized parameter update sequence; Performing EMC compliance verification on the optimized parameter update sequence to generate an anti-interference layout scheme.

6. The electromagnetic compatibility testing method for PCB design according to claim 5, wherein, The performing momentum accelerated descent iteration on the target function gradient matrix to generate an optimized parameter update sequence includes: Construct a momentum correction term based on the gradient matrix of the objective function to generate an iterative direction correction vector; Update the layout parameters according to the iterative direction correction vector and a preset momentum factor to generate an intermediate solution set of parameter iteration; Perform constraint boundary projection on the intermediate solution set of parameter iteration to generate a sequence of physically feasible solutions; Generate an updated sequence of optimization parameters according to the convergence criterion of the sequence of physically feasible solutions.

7. The electromagnetic compatibility test method for PCB design according to claim 1, wherein The multi-condition Monte Carlo simulation and residual calibration of the anti-interference layout scheme generate a correction guidance library, including: Perform multi-physical-field parameter perturbation on the anti-interference layout scheme to generate a Monte Carlo sample set; Perform multi-physical-field joint simulation according to the Monte Carlo sample set to generate an EMC performance data set; Perform principal component analysis and clustering processing on the EMC performance data set to generate a set of residual feature vectors; Perform residual correction analysis on the set of residual feature vectors to generate a key parameter correction coefficient matrix; Refine rules and prioritize the correction coefficient matrix to generate a correction rule library; Perform cross-condition generalization verification on the correction rule library to generate an electromagnetic compliance correction guidance library.

8. An electromagnetic compatibility test system for PCB design, characterized in that The electromagnetic compatibility test system for PCB design is used to execute the electromagnetic compatibility test method for PCB design according to any one of claims 1 to 7. The PCB includes a circuit schematic diagram and a structure model. The electromagnetic compatibility test system for PCB design includes: A high-frequency analysis module for extracting high-frequency signal characteristics in the circuit schematic diagram to obtain an electromagnetic simulation parameter set; A parameter modeling module for performing multi-physical-field coupling modeling according to the electromagnetic simulation parameter set and the structure model to obtain an electromagnetic field-circuit model; An excitation loading module for performing frequency-time domain hybrid excitation loading on the electromagnetic field-circuit model to obtain an electromagnetic noise distribution map; A layout optimization module for performing gradient descent optimization on the layout parameters of the structure model according to the electromagnetic noise distribution map to obtain an anti-interference layout scheme; A correction generation module for performing multi-condition Monte Carlo simulation and residual calibration on the anti-interference layout scheme to generate a correction guidance library.

Images