Electromagnetic co-simulation control system and method of multi-chip and packaging system

Abstract

The invention discloses an electromagnetic collaborative simulation control system and method for a multi-chip and packaging system, and relates to the technical field of electromagnetic collaboration, and the method comprises the following steps: setting an initial electromagnetic simulation frequency domain range according to the working frequency range of a target system; according to the invention, the electromagnetic co-simulation control system extracts key characteristics such as multi-signal channel nonlinear crosstalk intensity and impedance resonance offset, and the harmonic change coefficient generated by machine learning is combined, so that the nonlinear interference and resonance mismatch influence of higher harmonics on the system can be accurately quantified. The system dynamically adjusts a simulation frequency domain range, classifies and identifies normal and abnormal processes, effectively captures higher harmonic interference of key frequency bands, avoids resource waste, improves simulation efficiency and system reliability, and realizes efficient and accurate electromagnetic analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of electromagnetic collaboration, and particularly to an electromagnetic collaborative simulation control system and method for multi-chip and packaging systems. Background Art

[0002] The electromagnetic collaborative simulation control of multi-chip and packaging systems refers to the use of electromagnetic simulation technology to collaboratively analyze the complex electromagnetic coupling and interference problems between multiple chips and their packaging structures during the design and optimization of integrated circuits (ICs) and their packages, and to optimize the design parameters through accurate modeling and simulation to improve system performance. The core is to analyze key issues such as signal integrity (SI), power integrity (PI), and electromagnetic compatibility (EMC) through the joint simulation of multiple physical fields (such as electromagnetic fields, thermal fields) and multiple scales (from chip level to package level), and at the same time, combine intelligent algorithms for dynamic optimization control to ensure the stability and efficiency of multi-chip systems in high-frequency and high-density integration environments.

[0003] Using electromagnetic simulation technology to collaboratively analyze the complex electromagnetic coupling and interference problems between multiple chips and their packaging structures is to establish an accurate electromagnetic simulation model, comprehensively consider the electromagnetic interactions between multiple chips, packaging structures, and their interconnections, and analyze key issues such as signal integrity (SI), power integrity (PI), and electromagnetic compatibility (EMC) caused by high-density integration and high-frequency operation. Through this collaborative simulation, potential electromagnetic interference sources and coupling paths in the system can be identified, and chip layout, package design, and interconnection structure can be optimized to ensure the reliability and performance stability of multi-chip systems in complex electromagnetic environments.

[0004] The prior art has the following deficiencies:

[0005] The prior art usually sets the simulation frequency domain range according to the operating frequency range of the target system to capture the coupling characteristics between the chip and the package. This usually meets the requirements of most scenarios, but some chips or circuits (such as power amplifiers or mixers) generate high-order harmonics during operation, and their frequencies are higher than the main frequency range. If the simulation does not cover these high-order harmonics, serious consequences may occur. High-order harmonics may cause crosstalk, reflection, and impedance mismatch in the signal path, leading to a decline in signal quality, such as increased jitter, increased bit error rate, and even loss of digital communication signals. For example, in a high-speed digital system, high-order harmonics may cause transmission line resonance, resulting in a complete breakdown of data transmission. In addition, high-order harmonics propagate through the power network and generate noise spikes at the resonance frequency, causing the chip to operate unstably or the function to fail. For example, when a power amplifier operates, it interferes with other chips, causing bias drift in analog circuits or false triggering in digital circuits, seriously affecting the reliability of the system.

[0006] The above information disclosed in the background section is only used to enhance the understanding of the background of the present disclosure. Therefore, it may include information that does not constitute the prior art known to those of ordinary skill in the art. Summary of the Invention

[0007] The object of the present invention is to provide an electromagnetic co - simulation control system and method for a multi - chip and packaging system. Through the electromagnetic co - simulation control system, by using key characteristics such as the non - linear crosstalk intensity of multi - signal channels and impedance resonance offset, combined with the harmonic change coefficient generated by a machine learning model, the non - linear interference of high - order harmonics on the system and the impact of resonance mismatch are accurately quantified. The system dynamically adjusts the simulation frequency domain range according to the harmonic change coefficient, classifies and identifies normal and abnormal processes, avoids resource waste caused by frequency domain expansion, and at the same time captures high - order harmonic interference in key frequency bands, improves simulation efficiency and system reliability, and realizes efficient and accurate electromagnetic analysis to solve the problems in the above - mentioned background technology.

[0008] To achieve the above object, the present invention provides the following technical solutions: An electromagnetic co - simulation control method for a multi - chip and packaging system, comprising the following steps:

[0009] Set an initial electromagnetic simulation frequency domain range according to the operating frequency range of the target system;

[0010] Conduct a comprehensive electromagnetic analysis to identify potential electromagnetic interference sources and coupling paths. At the same time, collect multi - dimensional frequency domain data information to establish a structured analysis set, and extract key characteristics reflecting high - order harmonics from the analysis set;

[0011] Analyze and process the extracted key features under a detection window, and input the key feature data after analysis and processing into a pre - trained machine learning model to conduct an intelligent evaluation of the recognition process through the machine learning model;

[0012] According to the output result of the machine learning model, divide the recognition process into a normal recognition process and an abnormal recognition process;

[0013] For the recognition process classified as normal, continue the simulation analysis within the originally set operating frequency range;

[0014] For the case classified as an abnormal recognition process, based on the operating frequency range of the target system, expand the simulation frequency domain range to cover the high - order harmonic frequencies and capture the electromagnetic phenomena caused by high - order harmonics.

[0015] Preferably, key characteristics reflecting high-order harmonics are extracted from the analysis set, including the change rate of the non-linear crosstalk intensity between multiple signal channels and the ratio of the impedance change amount at the resonance frequency point to the impedance characteristic drift amount under high-frequency excitation. After obtaining these, under the detection window, the change rate of the non-linear crosstalk intensity between multiple signal channels and the ratio of the impedance change amount at the resonance frequency point to the impedance characteristic drift amount under high-frequency excitation are analyzed and processed to generate a crosstalk intensity non-linear index and an impedance resonance offset index respectively. The crosstalk intensity non-linear index quantifies the strength of the non-linear coupling effect caused by high-order harmonics between multiple signal channels and the rate of change of the coupling with frequency, reflecting the degree of interference of high-order harmonics on signal transmission integrity; the impedance resonance offset index quantifies the offset degree of the impedance change amplitude caused by high-frequency excitation at the system resonance frequency point from the impedance drift, reflecting the resonance mismatch caused by high-order harmonics in the system and its impact on the overall stability.

[0016] Preferably, after obtaining the crosstalk intensity non-linear index and the impedance resonance offset index generated by analyzing the key features, the crosstalk intensity non-linear index and the impedance resonance offset index are input into a pre-trained machine learning model, and a harmonic change coefficient is generated by the machine learning model to intelligently evaluate the recognition process through the harmonic change coefficient.

[0017] Preferably, the harmonic change coefficient generated by analyzing the extracted key characteristics is compared with a pre-set harmonic change coefficient reference threshold to divide the recognition process. The division steps are as follows:

[0018] If the harmonic change coefficient is greater than or equal to the pre-set harmonic change coefficient reference threshold, the recognition process is divided into an abnormal recognition process;

[0019] If the harmonic change coefficient is less than the pre-set harmonic change coefficient reference threshold, the recognition process is divided into a normal recognition process.

[0020] Preferably, based on the operating frequency range of the target system, the frequency domain range of the simulation is expanded to cover the high-order harmonic frequencies. The steps to capture the electromagnetic phenomena caused by high-order harmonics are as follows:

[0021] In the abnormal recognition process, according to the operating frequency range of the target system, combined with the harmonic change coefficient HVARC and the harmonic change coefficient reference threshold HVARC ref , the frequency domain range to be expanded is dynamically determined, and the calculation expression is as follows:

[0022]

[0023] , where f work is the operating frequency range of the target system, that is, the highest frequency of the normal operating frequency range of the target system, and Δfexpand is to expand the frequency domain range, that is, from the highest frequency f of the original operating frequency range of the target system work to the width of the frequency domain range expanded in the direction of higher harmonics, and ω is the expansion coefficient;

[0024] After expanding the frequency domain range, the simulation frequency domain is divided into different frequency bands to improve the simulation accuracy and efficiency. The calculation expression is as follows:

[0025]

[0026] , where f z is the center frequency of the z-th partition, used for refined simulation, and Q is the total number of partitions in the expanded frequency band, controlling the simulation granularity;

[0027] Perform weighted analysis on the higher harmonic characteristics of each partition to capture the key electromagnetic phenomena within the frequency band, calculate the harmonic weight of each partition, and the calculation expression is as follows:

[0028]

[0029] , where W z is the contribution weight of the frequency partition z to the overall simulation result;

[0030] After completing the simulation of all partitions, integrate the data of each frequency band, analyze the higher harmonic behavior within the expanded frequency domain range and its impact on the system, verify the effectiveness of the expanded range, and calculate the frequency coverage rate index. The calculation expression is as follows:

[0031]

[0032] , where C f is the frequency coverage rate, indicating the capture ratio of the impact of the expanded frequency domain range on higher harmonics, and P(f) is the power spectral density at the target frequency f, reflecting the energy distribution of higher harmonics;

[0033] By calculating the frequency coverage rate C f , clearly evaluate whether the currently expanded frequency domain range completely covers the frequency band where higher harmonics are located and the electromagnetic phenomena caused by them; when C f = 1, the expanded frequency domain range has fully covered the energy distribution of all higher harmonics, and the simulation results are complete without further adjustment; when C f < 1, the energy distribution of some higher harmonics falls outside the current simulation range, resulting in the failure to capture key electromagnetic phenomena; subsequent mechanisms include increasing the expanded frequency domain range Δf expand , and refining the simulation frequency band f in the partition manner zand focus on simulating and analyzing the harmonic power spectral density P(f) in the high-frequency region to ensure coverage of all frequency bands related to high-order harmonics and comprehensively evaluate its impact on system performance.

[0034] Preferably, under the detection window, the specific steps for analyzing the change rate of the nonlinear crosstalk intensity between multiple signal channels and generating the nonlinear index of crosstalk intensity are as follows:

[0035] Under the detection window, collect the coupled electromagnetic signals between multiple signal channels through high-precision electromagnetic simulation, and calculate the coupling strength between different channels. The calculation expression is as follows:

[0036]

[0037] , where C i,j (f) is the coupling strength, that is, the electromagnetic coupling characteristic between channels i and j at frequency f, and V j (f) and V i (f) are the signal voltage amplitudes of channels j and i, reflecting the strength of the channel signals, and φ i,j (f) is the phase difference between channels j and i, used to characterize the phase characteristics of signal coupling, and e is the natural base;

[0038] After obtaining the coupling strength C i,j (f), use a nonlinear enhancement operator to calculate the nonlinear change rate at each frequency point, that is, the derivative form of the coupling strength changing with frequency. The calculation expression is as follows:

[0039]

[0040] , where R i,j (f) is the nonlinear change rate, the rate of change of the nonlinear crosstalk intensity between channels i and j at frequency f, reflecting the dynamic characteristics of signal coupling changing with frequency, is the first derivative of the coupling strength C i,j (f) with respect to frequency f, representing the trend of the coupling strength between channels changing with frequency, is the second derivative of the coupling strength C i,j (f) with respect to frequency f, representing the acceleration of the change rate of the coupling strength;

[0041] After completing the calculation of the nonlinear change rate R i,j (f), project it into the target frequency range to generate the nonlinear index of crosstalk intensity. The calculation expression is as follows:

[0042]

[0043] , where XTNLI is the nonlinear index of crosstalk intensity, and |R i,j (f)|p It is the power amplification of the non - linear change rate. p is the weight parameter used to adjust the sensitivity to the region with drastic changes, and f min is the lower limit of the detection window frequency, and f max is the upper limit of the detection window frequency. π is the frequency - weighted periodic change, and sin(·) is the frequency - weighted function.

[0044] Preferably, under the detection window, the specific steps for analyzing the ratio of the impedance change amount at the resonance frequency point to the impedance characteristic drift amount under the high - frequency excitation to generate the impedance resonance offset index are as follows:

[0045] Under the detection window, first calculate the impedance change amount at the resonance frequency point. The calculation expression is as follows: ΔZ(f) = |Z(f) - Z ref (f)|. In the formula, ΔZ(f) is the impedance change amount, indicating the difference between the actual impedance value and the reference impedance value of the target system at the frequency f. Z ref (f) is the reference impedance value at the resonance frequency point, and Z(f) is the real - time impedance value within the detection window, which depends on the frequency f;

[0046] Calculate the impedance characteristic drift amount under the high - frequency excitation. The calculation expression is as follows:

[0047]

[0048] , where D Z (f) is the impedance characteristic drift amount, is the first - order derivative of the impedance Z(f) with respect to the frequency f, representing the instantaneous rate of change of the impedance with respect to the frequency. α is the weight factor, and f min is the lower limit of the detection window frequency, and f max is the upper limit of the detection window frequency, is the absolute - value integral of the second - order derivative of the impedance with respect to the frequency within the detection window [f min 、f max , which is used to capture the overall non - linear trend of the frequency change;

[0049] According to the impedance change amount ΔZ(f) and the impedance drift amount D Z (f), generate the impedance resonance offset index. The calculation expression is as follows:

[0050]

[0051] , where IRSPI is the impedance resonance offset index, is the sum of the impedance change amounts, which is used to quantify the degree of impedance deviation of the system, is the sum of the impedance drift amounts, which quantifies the response of the system impedance to the frequency change. max(ΔZ(f)) is the maximum value of the impedance change amount, and max(DZ (f)) is the maximum value of the impedance drift amount.

[0052] An electromagnetic co - simulation control system for a multi - chip and packaging system, including an initial frequency - domain setting module, a data analysis and characteristic extraction module, an intelligent evaluation module, a classification and recognition module, a normal analysis module, and an abnormal adjustment module:

[0053] The initial frequency - domain setting module sets an initial electromagnetic simulation frequency - domain range according to the operating frequency range of the target system;

[0054] The data analysis and characteristic extraction module conducts a comprehensive electromagnetic analysis, identifies potential electromagnetic interference sources and coupling paths, and at the same time collects multi - dimensional frequency - domain data information to establish a structured analysis set, and extracts key characteristics reflecting high - order harmonics from the analysis set;

[0055] The intelligent evaluation module analyzes and processes the extracted key features under a detection window, and inputs the analyzed key feature data into a pre - trained machine - learning model, and conducts an intelligent evaluation of the recognition process through the machine - learning model;

[0056] The classification and recognition module divides the recognition process into a normal recognition process and an abnormal recognition process according to the output result of the machine - learning model;

[0057] The normal analysis module continues to conduct simulation analysis within the originally set operating frequency range for the recognition process classified as normal;

[0058] The abnormal adjustment module, for the situation classified as an abnormal recognition process, based on the operating frequency range of the target system, expands the simulation frequency - domain range to cover high - order harmonic frequencies and captures electromagnetic phenomena caused by high - order harmonics.

[0059] In the above technical solution, the technical effects and advantages provided by the present invention:

[0060] Through the electromagnetic co - simulation control system, the present invention can comprehensively capture and analyze the electromagnetic coupling characteristics in the chip and packaging system, especially the influence of high - order harmonics on the system. By using key characteristics such as the non - linear crosstalk intensity of multi - signal channels and the impedance resonance offset amount, combined with the harmonic change coefficient generated by the detection window and the machine - learning model, the non - linear interference degree of high - order harmonics in signal transmission and the influence degree on resonance mismatch can be accurately quantified. When the harmonic change coefficient exceeds the preset threshold, the system automatically identifies the abnormality and dynamically adjusts the simulation frequency - domain range to ensure capturing the electromagnetic interference phenomenon caused by high - order harmonics. This accuracy solves the performance problems caused by insufficient frequency - domain range setting in traditional methods, such as failure to detect signal reflections, increased jitter, or data loss caused by high - order harmonics.

[0061] After the present invention eliminates these problems through optimized design, the overall reliability and stability of the system are significantly improved. By introducing a machine learning model, intelligent evaluation is carried out on the extracted key features (such as the crosstalk intensity nonlinear index and the impedance resonance offset index), and a harmonic change coefficient is generated, realizing the automatic classification of the recognition process. The distinction between the normal recognition process and the abnormal recognition process enables the system to maintain the simulation analysis within the original frequency domain range in normal scenarios, avoiding unnecessary expansion of the frequency domain range, thereby saving computing resources. For the abnormal recognition process, the system intelligently expands the simulation frequency domain range and focuses on capturing the key frequency bands affected by high-order harmonics, thus ensuring the pertinence and accuracy of the analysis. This classification and dynamic adjustment strategy not only improves the simulation efficiency but also avoids the resource waste caused by full-band simulation, realizing an efficient and economical electromagnetic simulation analysis process. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings described below are only some embodiments recorded in the present invention. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings.

[0063] Figure 1 It is a method flow chart of the electromagnetic co-simulation control method for the multi-chip and packaging system of the present invention.

[0064] Figure 2 It is a schematic diagram of the modules of the electromagnetic co-simulation control system for the multi-chip and packaging system of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0065] Example embodiments will now be described more fully with reference to the accompanying drawings. However, the example embodiments can be implemented in various forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that this disclosure will be more comprehensive and complete and will fully convey the concept of the example embodiments to those skilled in the art.

[0066] The present invention provides an electromagnetic co-simulation control method for a multi-chip and packaging system as Figure 1 shown, including the following steps:

[0067] Set an initial electromagnetic simulation frequency domain range according to the operating frequency range of the target system;

[0068] The initial electromagnetic simulation frequency domain range refers to the frequency interval initially set according to the operating frequency range of the target system when conducting electromagnetic simulation analysis. This frequency domain range defines the starting and ending frequencies of the simulation, that is, within this range, the simulation tool will calculate the electromagnetic response and characteristics of the system. The setting of the initial frequency domain range directly affects the accuracy and effectiveness of the simulation and is the basis for ensuring reliable simulation results.

[0069] Specifically, this frequency domain range should include the basic operating frequency of the system, possible harmonic frequencies (including sub-harmonics and higher harmonics), and other key frequency points that may affect the system performance. By reasonably setting the initial frequency domain range, the electromagnetic behavior of the system at different frequencies can be captured, such as signal transmission characteristics, interference coupling effects, and resonance phenomena, etc., providing comprehensive data support for subsequent analysis and optimization.

[0070] The initial electromagnetic simulation frequency domain range should meet the following conditions: First of all, it must cover all the operating frequencies of the target system and the possible harmonic frequencies generated, ensuring that the simulation can capture all electromagnetic phenomena related to the system performance. This means that the frequency domain range should be wide enough to include the main frequency, harmonics, and potential resonance frequencies, avoiding missing key electromagnetic effects.

[0071] Secondly, this frequency domain range needs to strike a balance between simulation accuracy and computational efficiency. An overly broad frequency domain range will increase the computational amount and time of the simulation, which may lead to waste of resources; while an overly narrow frequency domain range may not be able to capture important electromagnetic characteristics, affecting the accuracy of the simulation results. Therefore, the initial frequency domain range should be reasonably set according to the system characteristics and analysis requirements to ensure accurate and comprehensive simulation results under limited computational resources.

[0072] Conduct a comprehensive electromagnetic analysis to identify potential electromagnetic interference sources and coupling paths, and at the same time collect multi-dimensional frequency domain data information to establish a structured analysis set, and extract the key characteristics reflecting higher harmonics from within the analysis set;

[0073] Collecting multi-dimensional frequency domain data information means comprehensively collecting multiple physical characteristics of the system at different frequencies during the electromagnetic simulation process. These physical characteristics include but are not limited to electric field strength, magnetic field strength, current distribution, impedance parameters (such as input impedance and output impedance), scattering parameters (such as S parameters), and power loss, etc. These data reflect the electromagnetic behavior of the system within the frequency domain range and provide a comprehensive basis for analyzing electromagnetic interference sources, coupling paths, and the influence of higher harmonics.

[0074] The key to multi-dimensional data acquisition lies in capturing the relationships between different physical quantities. For example, by the variation relationship between frequency and electric field strength, the resonance frequency can be identified; by the distribution of frequency and scattering parameters, the signal transmission and reflection can be analyzed; by the relationship between frequency and power loss, the transmission and leakage paths of electromagnetic energy in the system can be evaluated. Multi-dimensional data acquisition can provide a panoramic view of the electromagnetic characteristics of the system, providing necessary information for subsequent feature extraction and problem location.

[0075] Establishing a structured analysis set means organizing, storing the collected multi-dimensional frequency domain data information so that it is presented in a logically clear, easy-to-retrieve and analyze form. This analysis set usually contains the correlation relationships of multi-dimensional data, such as the numerical correlations and variation laws between parameters at different frequencies. These data can be indexed and classified and stored by means of a database or a matrix table according to dimensions (such as frequency, position, parameter type).

[0076] The core purpose of the structured set is to transform complex raw data into a usable information set. For example, in the set, the electric and magnetic field distributions at a specific frequency can be quickly queried, the characteristic parameters with abnormal changes can be identified, or the trends of electromagnetic behavior in different frequency bands can be evaluated. This structured storage and representation method greatly improves the utilization efficiency of data, provides a standardized data interface for subsequent feature extraction, pattern recognition and input of machine learning models, thus making the whole analysis process more efficient and systematic.

[0077] Analyze and process the extracted key features under the detection window, and input the key feature data after analysis and processing into a pre-trained machine learning model to intelligently evaluate the recognition process through the machine learning model;

[0078] Extract the key characteristics reflecting high-order harmonics from the analysis set, including the change rate of the non-linear crosstalk intensity between multi-signal channels and the ratio of the impedance change amount at the resonance frequency point to the impedance characteristic drift amount under high-frequency excitation. After obtaining them, under the detection window, analyze and process the change rate of the non-linear crosstalk intensity between multi-signal channels and the ratio of the impedance change amount at the resonance frequency point to the impedance characteristic drift amount under high-frequency excitation, and generate the crosstalk intensity non-linear index and the impedance resonance offset index respectively. The crosstalk intensity non-linear index quantifies the strength of the non-linear coupling effect caused by high-order harmonics between multi-signal channels and the rate of change of the coupling with frequency, reflecting the interference degree of high-order harmonics on the signal transmission integrity; the impedance resonance offset index quantifies the offset degree of the impedance change amplitude caused by high-frequency excitation at the system resonance frequency point and the impedance drift, reflecting the resonance mismatch caused by high-order harmonics in the system and its impact on the overall stability;

[0079] After obtaining the crosstalk intensity non - linear index and impedance resonance offset index generated after analyzing the key features, input the crosstalk intensity non - linear index and impedance resonance offset index into a pre - learned machine learning model. Generate a harmonic change coefficient through the machine learning model, and use the harmonic change coefficient to intelligently evaluate the recognition process;

[0080] The pre - learned machine learning model refers to that before the actual detection and analysis of the target system, a large amount of historical data and simulation data are used to train the machine learning algorithm, and finally an intelligent model with specific task capabilities is obtained. This model has learned the laws and patterns of the influence of high - order harmonics in the target system from the training data and can quickly analyze and evaluate the input key characteristic parameters (such as crosstalk intensity non - linear index and impedance resonance offset index). The pre - training of the model is usually based on the supervised learning method, taking the key characteristic parameters as input variables and the frequency change coefficient or other evaluation indicators as output targets. Through repeated training and verification, the model can extract the deep associations between parameters and master the complex characteristics of non - linear electromagnetic effects.

[0081] The pre - learned model can not only reduce the time overhead of online training, but also ensure the real - time and accurate evaluation of the influence of high - order harmonics. This model is optimized to be efficient and stable when processing complex input data, and is especially suitable for scenarios with a wide frequency change range and multi - dimensional parameters. Compared with traditional static evaluation methods, the pre - learned model has dynamic adaptability, can quickly capture the change trends of non - linear phenomena such as crosstalk and resonance, and provides strong support for subsequent intelligent recognition and simulation optimization.

[0082] The implementation of this model requires the following stages: First, collect a large amount of relevant data, including electromagnetic simulation results and test data of the target system at different frequencies, ensuring that the data covers a wide frequency range and various possible high - order harmonic scenarios. Second, perform feature engineering on the data, extract key parameters, such as crosstalk intensity non - linear index and impedance resonance offset index, as model inputs; at the same time, set the frequency change coefficient or other evaluation indicators as the model output target. Subsequently, use machine learning algorithms (such as support vector machines, decision trees, deep learning networks, etc.) to train and verify the model, and optimize its fitting ability for non - linear relationships.

[0083] When the model is applied to an actual system, it can quickly perform intelligent analysis on the input data and generate evaluation results such as the frequency change coefficient. This process combines analysis and prediction, helping the system dynamically adjust the simulation frequency domain range, identify the specific frequency bands affected by high-order harmonics, and provide accurate references for further interference suppression or design optimization. At the same time, the model can continuously learn new detection data, perform adaptive updates, enhance its perception of system changes, and thus continuously improve the accuracy and efficiency of the recognition process.

[0084] When the change rate of the non-linear crosstalk intensity between multiple signal channels rapidly increases, it indicates the existence of a significant high-order harmonic coupling effect in the system. This phenomenon is usually caused by non-linear electromagnetic interference induced by high-order harmonics. However, if the set simulation frequency domain range does not include the frequency range of high-order harmonics, the simulation tool will not be able to capture these high-frequency phenomena, resulting in the system's inability to identify the existence of high-order harmonics. This is because the initial simulation frequency domain only focuses on the main frequency and its limited sub-harmonic frequency bands, lacking the basis for calculating and analyzing high-order harmonics outside the main frequency. The non-linear crosstalk effect induced by high-order harmonics is manifested as a sharp change in the signal coupling strength in the target system and has a significant impact on the signal integrity and stability of specific frequency bands. Simulation settings that do not cover these frequency ranges cannot provide a quantitative description of these phenomena or further abnormal location. This situation may lead to the neglect of potential system problems, especially in applications involving high-speed digital signal transmission or complex radio frequency systems, which may cause serious performance degradation or even functional failure.

[0085] The specific steps for generating the non-linear index of crosstalk intensity by analyzing the change rate of the non-linear crosstalk intensity between multiple signal channels under the detection window are as follows:

[0086] Under the detection window, through high-precision electromagnetic simulation, collect the coupled electromagnetic signals between multiple signal channels, and calculate the coupling strength between different channels. The calculation expression is as follows:

[0087]

[0088] , where C i,j (f) is the coupling strength, that is, the electromagnetic coupling characteristics between channels i and j at frequency f, including the signal amplitude relationship and phase relationship, V j (f) and V i (f) are the signal voltage amplitudes of channels j and i, reflecting the strength of the channel signals, φ i,j (f) is the phase difference between channels j and i, used to characterize the phase characteristics of signal coupling, and e is the natural base;

[0089] Obtain the coupling strength C i,j(f), the non - linear change rate at each frequency point is calculated using a non - linear enhancement operator, that is, the derivative form of the coupling strength with respect to frequency. The calculation expression is as follows:

[0090]

[0091] , where R i,j (f) is the non - linear change rate, the rate of change of the non - linear crosstalk strength between channels i and j at frequency f, reflecting the dynamic characteristics of signal coupling with frequency change. is the coupling strength C i,j (f) with respect to the first - order derivative of frequency f, representing the trend of the coupling strength between channels changing with frequency. is the coupling strength C i,j (f) with respect to the second - order derivative of frequency f, representing the acceleration of the rate of change of the coupling strength.

[0092] Calculating the non - linear change rate at each frequency point using a non - linear enhancement operator means amplifying the non - linear characteristics caused by frequency changes through mathematical operations to more sensitively capture the drastic changes caused by high - order harmonics. Specifically, the non - linear enhancement operator combines the first - order derivative of the coupling strength (representing the basic change trend) and the second - order derivative (representing the acceleration of the rate of change) for weighted processing. For example, by taking the absolute value of the second - order derivative and performing non - linear transformations (such as logarithmic or exponential operations) on it to enhance the detection ability for high - frequency small - amplitude mutations. The role of this method is to highlight the fast - changing characteristics caused by high - order harmonics without being overly sensitive to slow - changing background interference, thereby more accurately quantifying the non - linear behavior in frequency changes, especially when there are complex non - linear coupling effects in the target system, providing accurate indicators for subsequent analysis and optimization.

[0093] After completing the calculation of the non - linear change rate R i,j (f), project it into the target frequency range to generate the non - linear index of crosstalk strength. The calculation expression is as follows:

[0094]

[0095] , where XTNLI is the non - linear index of crosstalk strength, |R i,j (f)| p is the power amplification of the non - linear change rate, p is the weight parameter used to adjust the sensitivity to the drastic change region, f min is the lower limit of the detection window frequency, f max is the upper limit of the detection window frequency, π is the frequency - weighted periodic change, and sin(·) is the frequency - weighted function.

[0096] The weight parameter is a regulating factor used to adjust the influence degree of certain characteristics or indicators in a formula on the final result. In the formula, the weight parameter (such as the power amplification parameter p determines the contribution size of certain characteristics (such as the non-linear change rate R i,j (f) to the comprehensive result (the non-linear index of crosstalk intensity XTNLI). For example, the power parameter p assigns higher weights to the frequency bands with intense non-linear changes, thereby amplifying the influence of these frequency bands. The role of the weight parameter is to prioritize the characteristics that have a greater impact on the system performance by adjusting the relative importance of key variables, help accurately identify the non-linear problems caused by high-order harmonics, and improve the sensitivity and accuracy of the analysis.

[0097] Frequency-weighted periodic change refers to dynamically weighting the indicators within a specific frequency range through the frequency-weighting function sin(·), so that the contribution values at different frequency points have certain periodic characteristics. In the formula, the frequency-weighting function changes with the frequency from f min to f max to form a smooth sine curve. Its role is to prioritize the characteristics of the middle frequency bands within the detection frequency range and suppress the noise or interference in the boundary frequency bands. This periodic change ensures the smoothness and stability of the weight distribution, enabling the frequency weighting to more accurately reflect the actual characteristics within the target frequency band.

[0098] Under the detection window, the larger the value of the non-linear index of crosstalk intensity generated after analyzing the change rate of the non-linear crosstalk intensity between multiple signal channels, the more significant the non-linear coupling effect between multiple signal channels and the more obvious the change rate. This is usually due to the presence of high-order harmonics in the target system that are not included in the initial simulation frequency domain range. Therefore, the set simulation frequency domain range in this case cannot capture these high-frequency phenomena and cannot identify the high-order harmonics. On the contrary, when the value of the non-linear index of crosstalk intensity is small or remains stable, it indicates that there are no significant high-order harmonics in the current target system, and the initially set simulation frequency domain range can cover the main frequency range of the system and correctly identify the electromagnetic behavior in the target system.

[0099] The ratio of the impedance change at the resonance frequency point to the impedance characteristic drift under high-frequency excitation shows an asymmetric change, indicating the presence of high-order harmonics in the target system. This is because this phenomenon is usually triggered by the excitation of the system's nonlinear characteristics by high-frequency signals. If the set simulation frequency domain does not cover the frequencies corresponding to the high-order harmonics, the simulation tool cannot calculate the impedance characteristics at these frequencies, resulting in the failure to monitor the changes in key parameters. In this case, the simulation results are limited to the impedance characteristics within the main frequency range, and the high-frequency excitations beyond the range and the resonance and impedance drift phenomena they trigger cannot be captured. This will lead to the inability to quantify and analyze the effects of high-order harmonics, and thus unable to effectively identify and address the nonlinear interference problems in the system, directly affecting the comprehensive assessment and handling of the potential risks of the system.

[0100] Under the detection window, the specific steps for analyzing the ratio of the impedance change at the resonance frequency point to the impedance characteristic drift under high-frequency excitation and generating the impedance resonance offset index are as follows:

[0101] Under the detection window, first calculate the impedance change at the resonance frequency point. The calculation expression is as follows: ΔZ(f) = |Z(f) - Z ref (f)|, where ΔZ(f) is the impedance change, indicating the difference between the actual impedance value and the reference impedance value of the target system at frequency f. Z ref (f) is the reference impedance value at the resonance frequency point, and Z(f) is the real-time impedance value within the detection window, which depends on frequency f;

[0102] The reference impedance value at the resonance frequency point can be obtained through theoretical calculation, simulation analysis, or experimental measurement. Theoretical calculation is based on the design parameters of the system (such as circuit topology, material properties, and geometric dimensions), and uses the equivalent circuit model or transmission line theory to deduce the ideal impedance value at the resonance frequency point; simulation analysis uses high-precision electromagnetic simulation tools (such as HFSS, CST, etc.) to simulate the impedance characteristics of the system under ideal conditions without the influence of high-order harmonics; experimental measurement is to directly measure the impedance value at the resonance frequency point through equipment such as a network analyzer under the actual working conditions of the system, excluding the interference of high-order harmonics. The combination of these methods can ensure the accuracy of the reference impedance value and provide a reliable reference benchmark for subsequent offset analysis.

[0103] In this step, the key is to determine the impedance offset behavior at each frequency point within the window, and capture the impedance change signal caused by high-frequency harmonics in the system by comparing the actual measurement with the theoretical reference value.

[0104] Calculate the impedance characteristic drift under high-frequency excitation. The calculation expression is as follows:

[0105]

[0106] , where DZ (f) is the impedance characteristic drift amount, is the first derivative of the impedance Z(f) with respect to the frequency f, representing the instantaneous rate of change of the impedance with frequency. α is a weighting factor, and f min is the lower limit of the detection window frequency, and f max is the upper limit of the detection window frequency, is the absolute value integral of the second derivative of the impedance with respect to frequency within the detection window [f min , f max , which is used to capture the overall non-linear trend of the frequency change;

[0107] The role of the weighting factor α is to balance the contributions of the instantaneous rate of change of the impedance with frequency (first derivative) and the change in the overall trend (second derivative integral) to the impedance drift amount. High-order harmonics may cause drastic local changes in the impedance, and may also have a slowly accumulating global non-linear effect on the impedance characteristics, and the weights of the two may be different in the actual system. By adjusting α, according to the characteristics of the target system, the attention to a certain part of the effect can be enhanced. For example, when the system is more sensitive to the global non-linear trend caused by high-order harmonics, the value of α can be increased; while when the local transient changes are more significant, the value of α can be decreased. Therefore, as an adjustment parameter, α can improve the adaptability of the formula, enabling it to more accurately capture the main characteristics of the impedance drift in the target system.

[0108] According to the impedance change amount ΔZ(f) and the impedance drift amount D Z (f), an impedance resonance offset index is generated, and the calculation expression is as follows:

[0109]

[0110] , where IRSPI is the impedance resonance offset index, is the sum of the impedance change amounts, used to quantify the degree of deviation of the system's impedance, is the sum of the impedance drift amounts, quantifying the response of the system's impedance to frequency changes. max(ΔZ(f)) is the maximum value of the impedance change amount, representing the maximum impedance offset value at the frequency points within the detection window, reflecting the possible resonance points or non-linear peaks in the system. max(D Z (f)) is the maximum value of the impedance drift amount, representing the maximum impedance drift intensity at the frequency points within the detection window, usually related to high-order harmonics or non-linear behavior.

[0111] Under the detection window, the larger the impedance resonance offset index value generated after analyzing the ratio of the impedance change amount at the resonance frequency point to the impedance characteristic drift amount under the high-frequency excitation, the more significant the nonlinear coupling effect exists in the current target system, which is a direct manifestation of the influence of high-order harmonics on the system. In this case, the set simulation frequency domain range does not include these high-order harmonic frequencies, so the relevant electromagnetic characteristics and interference phenomena cannot be captured, resulting in the failure of identification. On the contrary, if the value of the crosstalk intensity nonlinear index is small, it indicates that the influence of high-order harmonics can be ignored, the system works normally within the initial simulation frequency domain range, and the currently set frequency domain range is sufficient to capture the electromagnetic characteristics of the system and complete the identification.

[0112] The machine learning model is not limited here. Any machine learning model that can comprehensively analyze the crosstalk intensity nonlinear index XTNLI and the impedance resonance offset index IRSPI to generate the harmonic variation coefficient HVARC is acceptable. To implement the technical solution of the present invention, the present invention provides a specific implementation method;

[0113] The generation formula of the harmonic variation coefficient HVARC is as follows:

[0114]

[0115] , where k 1 , k 2 are the preset proportionality coefficients of the crosstalk intensity nonlinear index XTNLI and the impedance resonance offset index IRSPI respectively, and k 1 , k 2 are both greater than 0.

[0116] The preset proportionality coefficients here refer to the predefined coefficients used to adjust the weights of different parameters in the formula. These coefficients f 1 and f 2 are used to balance the influence of the crosstalk intensity nonlinear index XTNLI and the impedance resonance offset index IRSPI on the harmonic variation coefficient HVARC. In other words, they determine the relative importance of these two parameters in the final calculation result. By reasonably setting f 1 and f 2 , the actual physical phenomena or design requirements in the target system can be reflected. For example, if the influence of high-order harmonics is mainly reflected in crosstalk, f 1 can be set larger, and if the influence of resonance offset is more significant, the value of f 2 can be increased. The selection of the preset proportionality coefficients is usually based on the physical characteristics of the actual application scenario, simulation analysis or experimental data to ensure that the model accurately reflects the actual dynamic behavior of the system.

[0117] From the harmonic variation coefficient, it can be seen that under the detection window, the larger the crosstalk intensity non-linear index value generated after analyzing the change rate of the non-linear crosstalk intensity between multiple signal channels, and the larger the impedance resonance offset index value generated after analyzing the ratio of the impedance change amount at the resonance frequency point to the impedance characteristic drift amount under high-frequency excitation, the larger the harmonic variation coefficient value generated after analyzing the extracted key characteristics under the detection window, indicating that there are high-order harmonics in the target system that are not included in the initial simulation frequency domain range and cannot identify the high-order harmonics. On the contrary, it indicates that the initially set simulation frequency domain range can cover the main frequency range of the system and correctly identify the electromagnetic behavior in the target system.

[0118] According to the output result of the machine learning model, the recognition process is divided into a normal recognition process and an abnormal recognition process;

[0119] Compare and analyze the harmonic variation coefficient generated after analyzing the extracted key characteristics with the pre-set reference threshold of the harmonic variation coefficient, and divide the recognition process as follows:

[0120] If the harmonic variation coefficient is greater than or equal to the pre-set reference threshold of the harmonic variation coefficient, the recognition process is divided into an abnormal recognition process;

[0121] If the harmonic variation coefficient is less than the pre-set reference threshold of the harmonic variation coefficient, the recognition process is divided into a normal recognition process;

[0122] The normal recognition process refers to the situation where in electromagnetic simulation and analysis, the system operating parameters (such as frequency, impedance, signal strength, etc.) are within the expected working range, and there are no obvious abnormal characteristics or behaviors beyond the design range. The abnormal recognition process refers to the situation where in electromagnetic simulation and analysis, abnormal phenomena beyond expectations are detected in the electromagnetic behavior of the target system, such as enhanced signal crosstalk caused by high-order harmonics, impedance mismatch, resonance offset, or noise spikes.

[0123] For the recognition process classified as normal, continue the simulation analysis within the originally set working frequency range;

[0124] For the recognition process classified as normal, the purpose of continuing the simulation analysis within the originally set operating frequency range is to deeply explore potential anomalies that may exist between the chip and the package, such as subtle interferences caused by parasitic capacitance, inductance, or other minor effects. Although these effects do not manifest as significant anomalies in the preliminary analysis, they may pose potential risks to the system performance under specific conditions. Through refined simulation, these subtle characteristics can be accurately captured, and their potential impacts on signal integrity, power integrity, and electromagnetic compatibility can be evaluated, thereby ensuring that the electromagnetic performance of the system fully meets the design requirements within the operating frequency range. This process helps to verify the reliability and stability of the system and provides a reliable basis for subsequent product optimization and performance assurance.

[0125] For the case classified as an abnormal recognition process, based on the operating frequency range of the target system, expand the frequency domain range of the simulation to cover high-order harmonic frequencies and capture the electromagnetic phenomena caused by high-order harmonics;

[0126] Based on the operating frequency range of the target system, the steps to expand the frequency domain range of the simulation to cover high-order harmonic frequencies and capture the electromagnetic phenomena caused by high-order harmonics are as follows:

[0127] During the abnormal recognition process, according to the operating frequency range of the target system, combined with the harmonic variation coefficient HVARFC and the harmonic variation coefficient reference threshold HVARC ref , dynamically determine the frequency domain range that needs to be expanded, and the calculation expression is as follows:

[0128]

[0129] , where f work is the operating frequency range of the target system, that is, the highest frequency of the normal operating frequency range of the target system, and Δf expand is the expanded frequency domain range, that is, the width of the frequency domain range extended from the highest frequency f work of the original operating frequency range of the target system in the direction of high-order harmonics, and ω is the expansion coefficient;

[0130] After expanding the frequency domain range, divide the simulation frequency domain into different frequency bands to improve the simulation accuracy and efficiency, and the calculation expression is as follows:

[0131]

[0132] , where f z is the center frequency of the z-th partition for refined simulation, Q is the total number of partitions of the expanded frequency band, which controls the simulation granularity;

[0133] Perform weighted analysis on the high-order harmonic characteristics of each partition to capture the key electromagnetic phenomena within the frequency band, calculate the harmonic weight of each partition, and the calculation expression is as follows:

[0134]

[0135] , where W z is the contribution weight of frequency partition z to the overall simulation result;

[0136] The contribution weight of frequency partition z to the overall simulation result refers to the influence degree of the harmonic characteristics in a specific frequency range on the simulation result, which is represented by the weight value W z and is used to quantify the importance of this frequency band in capturing high-order harmonic related electromagnetic phenomena. The calculation of the weight comprehensively considers the distance between the center frequency of the frequency band and the target operating frequency, as well as the influence of the harmonic variation coefficient HVARC on the system dynamic characteristics. Its role is to highlight the key frequency bands, make the simulation analysis resources focus on the frequency regions with high influence, so as to optimize the simulation efficiency and accuracy, and at the same time avoid the interference of invalid calculations to the overall analysis. Through this weight allocation, the key electromagnetic effects caused by high-order harmonics can be captured more accurately, providing a reliable basis for the subsequent system design optimization.

[0137] After completing the simulations of all partitions, integrate the data of each frequency band, analyze the high-order harmonic behavior in the extended frequency domain and its influence on the system, verify the effectiveness of the extended range, and calculate the frequency coverage rate index. The calculation expression is as follows:

[0138]

[0139] , where C f is the frequency coverage rate, indicating the capture ratio of the influence of the extended frequency domain range on high-order harmonics. P(f) is the power spectral density at the target frequency f, reflecting the energy distribution of high-order harmonics;

[0140] By calculating the frequency coverage rate C f , clearly evaluate whether the currently extended frequency domain range completely covers the frequency bands where high-order harmonics are located and the electromagnetic phenomena they cause; when C f = 1, the extended frequency domain range has fully covered the energy distribution of all high-order harmonics, and the simulation result has integrity and no further adjustment is required; when C f < 1, the energy distribution of some high-order harmonics falls outside the current simulation range, resulting in the failure to capture key electromagnetic phenomena; subsequent mechanisms include increasing the extended frequency domain range Δf expand , refining the simulation frequency band f according to the partition method z , and conducting key simulation analysis on the harmonic power spectral density P(f) in the high-frequency region to ensure that all frequency bands related to high-order harmonics are covered and comprehensively evaluate their influence on the system performance.

[0141] Through the electromagnetic co - simulation control system, the present invention can comprehensively capture and analyze the electromagnetic coupling characteristics in the chip and packaging system, especially the impact of high - order harmonics on the system. By using key characteristics such as the multi - signal - channel non - linear crosstalk intensity and impedance resonance offset, combined with the harmonic change coefficient generated by the detection window and the machine - learning model, the non - linear interference degree of high - order harmonics in signal transmission and the impact degree on resonance mismatch can be accurately quantified. When the harmonic change coefficient exceeds the preset threshold, the system automatically identifies the anomaly and dynamically adjusts the simulation frequency domain range to ensure the capture of electromagnetic interference phenomena caused by high - order harmonics. This accuracy solves the performance problems caused by insufficient frequency domain range setting in traditional methods, such as the failure to detect signal reflections, increased jitter, or data loss caused by high - order harmonics.

[0142] After the present invention eliminates these problems through optimized design, the overall reliability and stability of the system are significantly improved. By introducing a machine - learning model to intelligently evaluate the extracted key features (such as the non - linear index of crosstalk intensity and the impedance resonance offset index) and generate a harmonic change coefficient, an automated classification of the identification process is achieved. The distinction between the normal identification process and the abnormal identification process enables the system to maintain the simulation analysis within the original frequency domain range in normal scenarios, avoiding unnecessary expansion of the frequency domain range and thus saving computing resources. For the abnormal identification process, the system intelligently expands the simulation frequency domain range to focus on capturing the key frequency bands affected by high - order harmonics, thereby ensuring the pertinence and accuracy of the analysis. This classification and dynamic adjustment strategy not only improve the efficiency of the simulation but also avoid resource waste caused by full - band simulation, realizing an efficient and economical electromagnetic simulation analysis process.

[0143] For cases classified as the abnormal identification process, the frequency domain range of the simulation is expanded based on the operating frequency range of the target system to cover high - order harmonic frequencies. The purpose is to more comprehensively capture the electromagnetic phenomena caused by high - order harmonics, thereby analyzing and locating the source of the anomaly. High - order harmonics may cause complex electromagnetic interference phenomena, such as non - linear crosstalk in the signal path, impedance mismatch, parasitic resonance, increased power loss, or enhanced electromagnetic radiation. These phenomena usually exceed the initial simulation frequency domain range and may therefore be ignored in the initial simulation. By expanding the frequency domain range, the simulation tool can calculate the electromagnetic response in higher frequency bands and capture the dynamic behavior caused by high - order harmonics, such as unexpected parasitic coupling effects or high - frequency noise spikes.

[0144] In addition, this step can also reveal potential threats of high-order harmonics to system performance, such as issues like signal integrity degradation, instability of the power network, and electromagnetic compatibility failure. By comprehensively capturing these high-order harmonic-related phenomena, designers can accurately evaluate the impact of high-order harmonics on the system and optimize the design accordingly, such as adjusting the chip layout, optimizing the package structure, adding filtering components, or improving shielding performance, so as to ensure that the electromagnetic performance and stability of the target system meet the design requirements and avoid potential performance bottlenecks or failure risks. Ultimately, this step is a crucial link in ensuring the reliability and overall performance of the system.

[0145] The present invention provides an electromagnetic co-simulation control system for a multi-chip and package system as Figure 2 shown, including an initial frequency domain setting module, a data analysis and characteristic extraction module, an intelligent evaluation module, a classification and recognition module, a normal analysis module, and an abnormal adjustment module:

[0146] The initial frequency domain setting module sets an initial electromagnetic simulation frequency domain range according to the operating frequency range of the target system;

[0147] The data analysis and characteristic extraction module conducts a comprehensive electromagnetic analysis to identify potential electromagnetic interference sources and coupling paths. At the same time, it collects multi-dimensional frequency domain data information to establish a structured analysis set and extracts key characteristics reflecting high-order harmonics from the analysis set;

[0148] The intelligent evaluation module analyzes and processes the extracted key features under a detection window and inputs the analyzed key feature data into a pre-trained machine learning model to conduct an intelligent evaluation of the recognition process through the machine learning model;

[0149] The classification and recognition module divides the recognition process into a normal recognition process and an abnormal recognition process according to the output result of the machine learning model;

[0150] The normal analysis module continues to conduct simulation analysis within the originally set operating frequency range for the recognition process classified as normal;

[0151] The abnormal adjustment module, for the situation classified as an abnormal recognition process, expands the simulation frequency domain range based on the operating frequency range of the target system to cover the high-order harmonic frequencies and capture the electromagnetic phenomena caused by high-order harmonics.

[0152] The electromagnetic co-simulation control method for a multi-chip and package system provided by an embodiment of the present invention is implemented through the above-mentioned electromagnetic co-simulation control system for a multi-chip and package system. The specific methods and processes of the electromagnetic co-simulation control system for a multi-chip and package system are detailed in the embodiments of the above-mentioned electromagnetic co-simulation control method for a multi-chip and package system, and will not be elaborated here.

[0153] Only certain exemplary embodiments of the present invention have been described above by way of illustration. Without doubt, for those of ordinary skill in the art, the described embodiments can be modified in various different ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and description are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. An electromagnetic collaborative simulation control method for a multi-chip and packaging system, characterized in that: The following steps are involved: According to the operating frequency range of the target system, set the initial electromagnetic simulation frequency domain range; Conduct comprehensive electromagnetic analysis to identify potential electromagnetic interference sources and coupling paths, collect multi-dimensional frequency domain data information to establish a structured analysis set, and extract key characteristics reflecting high-order harmonics from the analysis set; The extracted key features are analyzed and processed under the detection window, and the analyzed key feature data are input into the pre-trained machine learning model, and the recognition process is intelligently evaluated through the machine learning model; According to the output results of the machine learning model, the recognition process is divided into normal recognition process and abnormal recognition process; For the identification process classified as normal, continue to perform simulation analysis within the originally set working frequency range; For the cases classified as abnormal identification processes, the frequency domain range of the simulation is expanded to cover high-order harmonic frequencies based on the operating frequency range of the target system to capture the electromagnetic phenomena caused by high-order harmonics.

2. The electromagnetic collaborative simulation control method of a multi-chip and packaging system according to claim 1, characterized in that: Extract key characteristics reflecting high-order harmonics from the analysis set, including the rate of change of nonlinear crosstalk intensity between multiple signal channels and the ratio of the impedance change at the resonance frequency point to the drift of the impedance characteristic under the high-frequency excitation. After obtaining, analyze and process the rate of change of nonlinear crosstalk intensity between multiple signal channels and the ratio of the impedance change at the resonance frequency point to the drift of the impedance characteristic under the high-frequency excitation under the detection window, and generate a crosstalk intensity nonlinear index and an impedance resonance shift index, respectively. The crosstalk intensity nonlinear index quantifies the strength of the nonlinear coupling effect caused by high-order harmonics between multiple signal channels, as well as the rate at which the coupling changes with frequency, reflecting the degree of interference of high-order harmonics on the integrity of signal transmission; The impedance resonance shift index quantifies the impedance change amplitude and impedance drift deviation caused by high-frequency excitation at the system resonant frequency point, reflecting the resonance mismatch caused by high-order harmonics in the system and its impact on the overall stability.

3. The electromagnetic collaborative simulation control method of a multi-chip and packaging system according to claim 2, characterized in that: After obtaining the crosstalk intensity nonlinear index and impedance resonance shift index generated after analyzing the key features, the crosstalk intensity nonlinear index and impedance resonance shift index are input into a pre-learned machine learning model, and the harmonic variation coefficient is generated by the machine learning model. The recognition process is intelligently evaluated by the harmonic variation coefficient.

4. The electromagnetic collaborative simulation control method of a multi-chip and packaging system according to claim 3, characterized in that: The harmonic variation coefficient generated after analyzing the extracted key characteristics is compared with the preset harmonic variation coefficient reference threshold, and the identification process is divided. The division steps are as follows: If the harmonic variation coefficient is greater than or equal to a preset harmonic variation coefficient reference threshold, the identification process is classified as an abnormal identification process; If the harmonic variation coefficient is less than a preset harmonic variation coefficient reference threshold, the identification process is classified as a normal identification process.

5. The electromagnetic collaborative simulation control method of a multi-chip and packaging system according to claim 4, characterized in that: Based on the operating frequency range of the target system, the steps to expand the frequency domain range of the simulation to cover higher harmonic frequencies and capture the electromagnetic phenomena caused by higher harmonics are as follows: In the abnormality identification process, according to the operating frequency range of the target system, combined with the harmonic variation coefficient HVARFC and the harmonic variation coefficient reference threshold HVARC ref , dynamically determine the frequency domain range that needs to be expanded, and the calculation expression is as follows: In the formula, f work is the operating frequency range of the target system, that is, the highest frequency of the normal operating frequency range of the target system, Δf expand is the extended frequency domain range, that is, the highest frequency f from the original operating frequency range of the target system work The width of the frequency domain range extending toward higher harmonics, ω is the expansion coefficient; After expanding the frequency domain range, the simulation frequency domain is divided into different frequency bands to improve the simulation accuracy and efficiency. The calculation expression is as follows: In the formula, f z is the center frequency of the zth partition, which is used for refined simulation, Q is the total number of partitions in the extended frequency band, which controls the simulation granularity; The high-order harmonic characteristics of each partition are weighted analyzed to capture the key electromagnetic phenomena within the frequency band, and the harmonic weight of each partition is calculated. The calculation expression is as follows: Where W z is the contribution weight of frequency partition z to the overall simulation results; After completing the simulation of all partitions, the data of each frequency band is integrated to analyze the high-order harmonic behavior within the extended frequency domain and its impact on the system, verify the effectiveness of the extended range, and calculate the frequency coverage index. The calculation expression is as follows: In the formula, C f is the frequency coverage, which indicates the capture ratio of the extended frequency domain range on the impact of higher harmonics, P(f) is the power spectrum density at the target frequency f, which reflects the energy distribution of higher harmonics; By calculating the frequency coverage C f , clearly evaluate whether the currently expanded frequency domain range completely covers the frequency band where the high-order harmonics are located and the electromagnetic phenomena caused by them; when C f = 1, the extended frequency domain has fully covered the energy distribution of all high-order harmonics, and the simulation results are complete and no further adjustment is required; when C f <1, some high-order harmonic energy distribution falls outside the current simulation range, resulting in the failure to capture key electromagnetic phenomena; subsequent mechanisms include increasing the extended frequency domain range Δf expand , refine the simulation frequency band f according to the partition method z , and focus on simulating and analyzing the harmonic power spectrum density P(f) in the high-frequency area to ensure that all high-order harmonic-related frequency bands are covered and their impact on system performance is fully evaluated.

6. The electromagnetic collaborative simulation control method of a multi-chip and packaging system according to claim 2, characterized in that: Under the detection window, the change rate of the nonlinear crosstalk intensity between multiple signal channels is analyzed, and the specific steps of generating the crosstalk intensity nonlinear index are as follows: In the detection window, high-precision electromagnetic simulation is used to collect the coupled electromagnetic signals between multiple signal channels and calculate the coupling strength between different channels. The calculation expression is as follows: In the formula, C i,j (f) is the coupling strength, i.e., the electromagnetic coupling characteristics of channels i and j at frequency f, V j (f) and V i (f) is the signal voltage amplitude of channel j and channel i, reflecting the strength of the channel signal, φ i,j (f) is the phase difference between channel j and channel i, which is used to characterize the phase characteristics of signal coupling, and e is the natural base; Get the coupling strength C i,j (f), the nonlinear enhancement operator is used to calculate the nonlinear change rate at each frequency point, that is, the derivative form of the coupling intensity changing with frequency. The calculation expression is as follows: In the formula, R i,j (f) is the nonlinear change rate, the rate at which the nonlinear crosstalk intensity of channels i and j changes at frequency f, reflecting the dynamic characteristics of signal coupling changing with frequency. is the coupling strength C i,j (f) The first-order derivative of frequency f, which indicates the trend of inter-channel coupling strength changing with frequency, is the coupling strength C i,j (f) The second derivative with respect to frequency f, which represents the acceleration of the rate of change of coupling strength; After completing the nonlinear change rate R i,j After calculating (f), it is projected into the target frequency range to generate the crosstalk intensity nonlinear index. The calculation expression is as follows: Where XTNLI is the crosstalk intensity nonlinear index, |R i,j (f)| p is the power amplification of the nonlinear change rate, p is the weight parameter used to adjust the sensitivity to the area of ​​​​drastic changes, and f min is the lower limit of the detection window frequency, f max is the upper frequency limit of the detection window, π is the frequency-weighted periodic variation, and sin(·) is the frequency-weighted function.

7. The electromagnetic collaborative simulation control method of a multi-chip and packaging system according to claim 2, characterized in that: In the detection window, the ratio of the impedance change at the resonance frequency point to the impedance characteristic drift under high-frequency excitation is analyzed, and the specific steps for generating the impedance resonance shift index are as follows: In the detection window, first calculate the impedance change at the resonant frequency point. The calculation expression is as follows: ΔZ(f)=|Z(f)-Z ref (f)|, where ΔZ(f) is the impedance change, which represents the difference between the actual impedance value of the target system and the reference impedance value at frequency f, and Z ref (f) is the reference impedance value at the vibration frequency point, and Z(f) is the real-time impedance value within the detection window, which depends on the frequency f; Calculate the impedance characteristic drift under high-frequency excitation. The calculation expression is as follows: Where D Z (f) is the impedance characteristic drift, is the first-order derivative of impedance Z(f) with respect to frequency f, indicating the instantaneous rate at which impedance changes with frequency. α is the weighting factor, f min is the lower limit of the detection window frequency, f max is the upper limit of the detection window frequency, is the second derivative of impedance with respect to frequency in the detection window [f min 、f max ] is used to capture the overall nonlinear trend of frequency variation; According to the impedance change ΔZ(f) and impedance drift D Z (f), generate the impedance resonance shift index, the calculation expression is as follows: Where IRSPI is the impedance resonance shift index, It is the sum of the impedance changes and is used to quantify the impedance deviation of the system. is the sum of the impedance drift, quantifying the response of the system impedance to frequency changes, max(ΔZ(f)) is the maximum impedance change, max(D Z (f)) is the maximum value of the impedance drift.

8. An electromagnetic collaborative simulation control system for a multi-chip and packaging system, used to implement the electromagnetic collaborative simulation control method for a multi-chip and packaging system as described in any one of claims 1 to 7, characterized in that: It includes the initial frequency domain setting module, data analysis and feature extraction module, intelligent evaluation module, classification and recognition module, normal analysis module and abnormal adjustment module: An initial frequency domain setting module sets the initial electromagnetic simulation frequency domain range according to the operating frequency range of the target system; The data analysis and feature extraction module conducts comprehensive electromagnetic analysis, identifies potential electromagnetic interference sources and coupling paths, collects multi-dimensional frequency domain data information to establish a structured analysis set, and extracts key features reflecting high-order harmonics from the analysis set; The intelligent evaluation module analyzes and processes the extracted key features under the detection window, and inputs the analyzed and processed key feature data into the pre-trained machine learning model to perform intelligent evaluation of the recognition process through the machine learning model; The classification and recognition module divides the recognition process into normal recognition process and abnormal recognition process according to the output results of the machine learning model; The normal analysis module continues to perform simulation analysis within the originally set working frequency range for the identification process classified as normal; The abnormal adjustment module, for situations classified as abnormal identification processes, expands the frequency domain range of the simulation based on the operating frequency range of the target system to cover high-order harmonic frequencies and capture electromagnetic phenomena caused by high-order harmonics.

Images