Electromagnetic co-simulation control system and method for multi-chip and package system

Abstract

The application provides an electromagnetic co-simulation control system and method of a multi-chip and package system, which comprises the following steps: obtaining structural parameters, electrical parameters and operating parameters, establishing and dynamically correcting an electromagnetic co-simulation model, and constructing an effective synthetic port group according to the spatial position and electromagnetic coupling relationship of the standby port; adjusting the excitation amplitude and excitation phase of the standby port in the group to form a synthetic virtual port with variable position, collecting bidirectional transfer response and identifying a dark mode field kernel, a dominant frequency band and an observable standby port set; determining an interference source chip and establishing an interference propagation model to obtain a risk level, a health margin and a control safety boundary; performing strategy optimization on a target standby port combination and a pre-echo control parameter, generating a pre-echo injection plan according to a source end pilot signal; performing pre-echo injection, collecting residual electromagnetic response and updating the model, and generating a control configuration.

Description

Technical Field

[0001] This invention relates to the field of electromagnetic simulation control technology for multi-chip packages, and in particular to an electromagnetic collaborative simulation control system and method for multi-chip and packaged systems. Background Technology

[0002] With the development of multi-chip integration and 3D packaging technologies, the electromagnetic coupling between chips, microbumps, through-silicon vias (TSVs), package substrate traces, and grounding return structures is becoming increasingly complex. Transient currents generated by chips during clock switching, data burst transmission, and task startup propagate along package interconnects and return paths, forming electromagnetic energy accumulation in local areas, which in turn leads to increased noise at functional ports, clock jitter, or higher bit error rates.

[0003] Existing electromagnetic simulations typically rely on static models built based on the design structure and nominal material parameters, making it difficult to accurately reflect the impact of manufacturing tolerances, temperature variations, and changes in operating conditions on propagation amplitude, phase, and time delay. During detection, reliance on functional ports or externally packaged ports means that when internal electromagnetic disturbances propagate along closed return paths, internal energy may have already accumulated significantly while the external response remains weak, making it difficult to determine the dark mode field core, its interference source chip, and the propagation path in a timely manner. Furthermore, existing active suppression methods usually implement compensation only after an anomaly is detected, meaning the first interference wavefront often has already reached a local high-energy region, and there is a lack of mechanisms to pre-generate inverse compensation waveforms using backup ports, source-end leader signals, and actual propagation time delays.

[0004] Therefore, this invention proposes an electromagnetic co-simulation control system and method for multi-chip and packaged systems. The information disclosed in the background section is only for enhancing understanding of the background of this disclosure and may therefore contain prior art information that is not common knowledge to those skilled in the art. Summary of the Invention

[0005] The purpose of this invention is to address the shortcomings of existing technologies by providing an electromagnetic collaborative simulation control system and method for multi-chip and packaged systems, thereby solving the technical problems mentioned in the background section.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] The electromagnetic co-simulation control method for multi-chip and packaged systems includes the following steps:

[0008] S1. Obtain the structural, electrical, and operational parameters of the multi-chip and packaging system, and establish an initial electromagnetic co-simulation model; construct an effective synthetic port group based on the spatial location and electromagnetic coupling relationship of the spare ports, and correct the model according to the operating status to obtain the current electromagnetic co-simulation model;

[0009] S2. Adjust the excitation amplitude and excitation phase of the backup ports within the effective synthetic port group to generate synthetic virtual ports with variable positions; collect bidirectional transfer responses and correct the current electromagnetic co-simulation model to determine the dark mode field core, dominant frequency band, and observable backup port set;

[0010] S3. Determine the actual interference source chip based on the dark mode field core, dominant frequency band, and observable backup port set; collect calibration responses and establish an interference propagation model; determine the dark mode runaway risk level, health margin, and control safety boundary based on the electromagnetic health state sequence and interference propagation model.

[0011] S4. Based on the interference propagation model and control safety boundary, optimize the target backup port combination, excitation amplitude, excitation phase and injection timing, establish the advance triggering mapping between the source leader signal and the pre-echo control parameters, and generate the pre-echo injection plan.

[0012] S5. Inject the pre-echo waveform into the target backup port according to the pre-echo injection plan and collect the remaining electromagnetic response. Update the electromagnetic co-simulation model and interference propagation model according to the remaining electromagnetic response, perform secondary strategy optimization, and generate an effective control configuration when the preset suppression conditions are met.

[0013] The electromagnetic collaborative simulation control system for multi-chip and packaged systems includes: a parameter acquisition unit, a multi-channel synchronous excitation acquisition unit, a control processor, a memory, a multi-channel digital-to-analog converter and a port driver unit. The memory stores the program executed by the control processor.

[0014] The beneficial effects of this invention are as follows:

[0015] This invention dynamically corrects the electromagnetic co-simulation model by combining structural parameters, electrical parameters, and real-time operating parameters, improving the model's adaptability to manufacturing tolerances, temperature variations, and changes in operating conditions, resulting in more accurate calculations of propagation amplitude, phase, and propagation delay. By synthesizing a variable-position synthetic virtual port using multiple backup ports and combining this with bidirectional transfer response scanning of the encapsulation's internal region, it can identify dark mode field cores with weak functional port responses but significant internal electromagnetic energy accumulation, reducing the possibility of missing hidden internal interference.

[0016] This invention calculates the electromagnetic contribution of each chip to the dark mode field core and establishes an interference propagation model based on the calibration response. This accurately identifies the actual interference source chip, main propagation path, propagation delay, and impulse response, providing a reliable basis for subsequent active control. By using the source-end leader signal to predict interference events and determining the advance injection time based on the interference propagation delay and the additional propagation delay of the target backup port, the pre-echo waveform can be injected before the first interference wave reaches the dark mode field core, improving the timeliness and suppression effect of active cancellation.

[0017] This invention optimizes the target backup port combination, excitation amplitude, excitation phase, and injection timing to reduce the residual electromagnetic response of the dark mode field core while limiting the injected energy of backup ports and the additional response of functional ports, thus preventing the generation of new local electromagnetic energy accumulation. By acquiring actual output waveforms and residual electromagnetic responses, it distinguishes between hardware execution errors and model errors, and updates the electromagnetic co-simulation model, interference propagation model, and effective control configuration to achieve closed-loop correction, risk prediction, and health management of the electromagnetic interference suppression process, thereby improving long-term operational stability. Attached Figure Description

[0018] Figure 1 This is a flowchart of the electromagnetic collaborative simulation control method for the multi-chip and packaging system of the present invention;

[0019] Figure 2 This is a framework diagram of the electromagnetic collaborative simulation control system for the multi-chip and packaging system of the present invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] Example 1: As Figure 1 As shown, this embodiment provides an electromagnetic co-simulation control method for a multi-chip and packaged system, including the following steps:

[0022] S1. Obtain the structural, electrical, and operational parameters of the multi-chip and packaging system, and establish an initial electromagnetic co-simulation model; construct an effective synthetic port group based on the spatial location and electromagnetic coupling relationship of the spare ports, and correct the model according to the operating status to obtain the current electromagnetic co-simulation model;

[0023] S2. Adjust the excitation amplitude and excitation phase of the backup ports within the effective synthetic port group to generate synthetic virtual ports with variable positions; collect bidirectional transfer responses and correct the current electromagnetic co-simulation model to determine the dark mode field core, dominant frequency band, and observable backup port set;

[0024] S3. Determine the actual interference source chip based on the dark mode field core, dominant frequency band, and observable backup port set; collect calibration responses and establish an interference propagation model; determine the dark mode runaway risk level, health margin, and control safety boundary based on the electromagnetic health state sequence and interference propagation model.

[0025] S4. Based on the interference propagation model and control safety boundary, optimize the target backup port combination, excitation amplitude, excitation phase and injection timing, establish the advance triggering mapping between the source leader signal and the pre-echo control parameters, and generate the pre-echo injection plan.

[0026] S5. Inject the pre-echo waveform into the target backup port according to the pre-echo injection plan and collect the remaining electromagnetic response. Update the electromagnetic co-simulation model and interference propagation model according to the remaining electromagnetic response, perform secondary strategy optimization, and generate an effective control configuration when the preset suppression conditions are met.

[0027] S1 specifically includes the following sub-steps:

[0028] S110. Obtain the structural, electrical, and operational parameters of the multi-chip and packaging system, and establish an initial electromagnetic co-simulation model.

[0029] Structural parameters are derived from chip layout files, package design files, and package substrate wiring files, including the position, size, and connection relationship of the chip, microbumps, through-silicon vias, package traces, ground plane, and spare ports in a unified package coordinate system; material parameters are preferably derived from the actual test reports of the same batch of materials, including conductor conductivity, dielectric relative permittivity, dielectric loss tangent, and permeability. When actual test values ​​are unavailable, the nominal values ​​and tolerance ranges given in the material specifications are used.

[0030] Port parameters are derived from chip input / output models, scattering parameter files, or vector network analyzer test results, including reference impedance, reference ground, operating bandwidth, and allowable voltage and current ranges for functional and spare ports. Operating parameters are derived from the power management unit, clock control unit, performance counter, on-chip current monitoring unit, and temperature sensor, including supply voltage, operating frequency, output switching rate, port current, and chip temperature. If the packaged sample has not yet been manufactured, the port excitation waveform corresponding to the preset operating state is generated using the port model provided by the chip manufacturer.

[0031] After unifying the port direction, reference impedance, and reference ground, chip port models, package interconnect models, ground return current models, and spare port models are established respectively. The chip port model represents the excitation waveform and equivalent output impedance of the functional port; the package interconnect model represents the propagation of electromagnetic disturbances through microbumps, through-silicon vias, and package traces; the ground return current model represents the return current in the ground plane and ground vias; and the spare port model records the spare port location, operating bandwidth, allowable injection current, and equivalent impedance.

[0032] The finite element method is used for three-dimensional electromagnetic solution, and the mesh is continuously refined. The mesh is considered convergent when the rate of change of the amplitude and propagation delay at the target port after two consecutive mesh refinements does not exceed 2%. In this embodiment, 2% is determined by the solver numerical error and test repeatability, and can be adjusted according to actual accuracy requirements.

[0033] Each preset operating state is mapped to port excitation amplitude, excitation period, flip event density and initial phase. Port voltage, port current, response phase, propagation delay, spatial field distribution and return current distribution are calculated to form the basic data of the model. The initial electromagnetic co-simulation model is output for S120 and S130 to call.

[0034] S120. Construct an effective synthetic port group based on the spatial location and electromagnetic coupling relationship of the spare ports.

[0035] A power-limited linear sweep calibration signal is sequentially applied to each backup port. The peak value of the calibration signal does not exceed 10% of the rated voltage of the corresponding backup port, and the frequency range covers the chip's fundamental frequency and major harmonics. The remaining backup ports are connected to a reference impedance, and the induced voltage, current, phase, and response arrival time of each backup port and functional port are collected synchronously. The 10% value is taken in this embodiment, based on the rated capacity of the backup ports and the upper limit of the test that will not cause the logic to flip in the functional ports.

[0036] Based on the actual spatial distance, bidirectional response amplitude correlation, response phase correlation, and common return path overlap, the overall coupling degree between backup port i and backup port j is calculated:

[0037]

[0038] in, For overall coupling degree; Spatial proximity is obtained by normalizing the actual distance; The correlation degree of the bidirectional response amplitude; For response phase correlation degree; It is the ratio of the overlap length of the common return path to the total length of the corresponding return path; , , , The weights are denoted by , and the sum of the weights is 1. The weights are determined by the calibration data of healthy packaged samples. A healthy packaged sample refers to a sample that uses the same chip model, package structure, and manufacturing process as the system under test, and has passed functional, electrical, and package reliability tests.

[0039] A backup port coupling relationship matrix is ​​established using backup ports as nodes and overall coupling degree as edge weights. At least two backup ports are considered as candidate synthetic port groups when their overall coupling degree is not lower than the grouping threshold, they can form a phase-adjustable superposition within the same internal region, and they have a continuous common return path. The grouping threshold is determined based on the coupling degree distribution and measurement error of healthy packaged samples.

[0040] The net coupling of candidate synthetic port groups to each functional port is further calculated using the initial electromagnetic co-simulation model. The upper limit of functional port coupling is set as follows: the additional response of all functional ports does not exceed 2% of the swing of the corresponding functional signal. Only candidate synthetic port groups that meet the upper limit of functional port coupling are retained to form effective synthetic port groups.

[0041] Record the identifier, location, allowable amplitude range, allowable phase range, working bandwidth, and upper limit of functional port coupling in each group to form port constraint parameters, and output the effective synthetic port group, port constraint parameters, and backup port coupling relationship matrix for use by S130, S210, S410, and S510.

[0042] S130. Dynamically model the initial electromagnetic co-simulation model based on the real-time operating status to obtain the current electromagnetic co-simulation model. Dynamic modeling refers to online correction of chip port excitation, conductor loss, dielectric parameters, return impedance, and port transmission parameters based on the actual operating parameters of the current packaged sample and the background induction signal of the spare port, so that the model corresponds to the current operating status.

[0043] A unified time reference is used to collect power supply voltage, operating frequency, output toggle rate, port current, chip temperature, and background sensing signals from spare ports. The sampling period is less than 1 / 10 of the shortest stable operating state duration, and each update uses one complete operating window. The absolute deviation of each sampled value within the window relative to the median value within the window is calculated, and sampled values ​​with an absolute deviation exceeding 3 times the median absolute deviation are identified as outliers and removed.

[0044] The measured parameters are compared with the baseline parameters in the model's basic data under the same operating conditions. Dynamic modeling is triggered when the normalized deviation of any parameter exceeds the corresponding update threshold for five consecutive sampling points; the update threshold is determined based on the measurement equipment error, packaging manufacturing tolerance, and the upper limit of normal fluctuation of healthy packaged samples.

[0045] Supply voltage and output toggle rate are used to correct excitation amplitude and toggle event density, port current is used to correct equivalent output impedance, and chip temperature is used in conjunction with material temperature coefficient to correct conductor conductivity and dielectric loss.

[0046] To correct the equivalent dielectric parameter and port propagation delay, a pre-established model parameter sensitivity matrix is ​​used. The deviation between the measured propagation delay and the model's basic data is taken as input, and the correction amount of the equivalent dielectric parameter is obtained by back-calculating through the first-order partial derivative relationship of the sensitivity matrix. The background sensing signal of the backup port, combined with the backup port coupling relationship matrix, is used to correct the propagation parameters between backup ports and from the backup port to the functional port.

[0047] Reserve one uncorrected backup port response as verification data; when the amplitude error, phase error and propagation delay error of the corrected model to the verification data do not exceed the verification threshold determined by the measurement error and mesh convergence error, output the current electromagnetic co-simulation model and dynamic modeling trigger record; otherwise, cancel this correction and collect the data for the next running window.

[0048] Dynamic modeling trigger records are written into the S530's fault prediction and health management records and serve as the basis for subsequent model updates and status tracing.

[0049] S2 specifically includes the following sub-steps:

[0050] S210. Generate a composite virtual port with a variable position in the effective composite port group. A composite virtual port refers to an equivalent port characterized by applying a combination of microcurrents with a defined amplitude and phase relationship to multiple fixed spare ports within the same effective composite port group, causing a local electromagnetic response peak to form in a specified area inside the package, and the spatial position corresponding to the peak.

[0051] The system calls upon the effective synthetic port group, port constraint parameters, and spare port coupling relationship matrix output from S120, and the current electromagnetic co-simulation model output from S130. Candidate scanning regions are defined based on chip gaps, micro-bump array edges, through-silicon via array centers, ground plane openings, and adjacent areas of sensitive circuits. Candidate positions are generated according to the mesh nodes of the current electromagnetic co-simulation model.

[0052] The initial excitation weights of the backup ports within the group are determined based on the backup port coupling relationship matrix. Then, the complex field transfer coefficients from each backup port to the candidate position and the complex transfer coefficients from each backup port to the functional port are calculated using the current electromagnetic co-simulation model.

[0053] For each candidate location, a complex excitation weight is calculated, where the modulus of the complex excitation weight represents the excitation amplitude and the argument represents the excitation phase. The objective is to maximize the local electromagnetic response at the candidate location, while ensuring that the excitation amplitude, excitation phase, and operating frequency of each backup port are within the port constraint parameters, the additional response of each functional port does not exceed the coupling upper limit determined by S120, and the residual injected charge within one calibration cycle satisfies the following:

[0054]

[0055] in, This is the start time of the calibration cycle; For calibration cycle; The number of spare ports used to generate the current synthesized virtual port; The combined microcurrent output by the backup port i at time t; The residual injected charge threshold is determined based on the spare port input capacitance, rated voltage, and charge measurement error. It is a time integral infinitesimal element.

[0056] The candidate location with the largest local electromagnetic response after the combined microcurrent is determined as the synthetic virtual port location. A moving trajectory is generated according to the spatial distance between adjacent candidate locations, with a moving step size not exceeding 1 / 2 of the minimum conductor spacing in the candidate region. The complex excitation weights, synthetic virtual port locations, functional port additional responses, and residual injected charges corresponding to each location are recorded to form a virtual port scanning parameter set.

[0057] For example, four spare ports are arranged around a through-silicon via (TSV) array. By changing the amplitude ratio and phase difference of the four spare ports, the local response peaks can be located sequentially on the left, center, and right sides of the array, thereby achieving equivalent port movement. The virtual port scan parameter set is input into S220.

[0058] S220: Acquire the bidirectional transmission response during the movement of the synthesized virtual port and correct the current electromagnetic co-simulation model. The bidirectional transmission response refers to the two sets of transmission responses obtained by exchanging the excitation port and response port respectively, while maintaining consistency in chip operating state, port terminal state, and environmental conditions. It is not assumed that the responses in the two directions are equal.

[0059] According to the virtual port scanning parameter set, the multi-channel synchronous signal generation unit outputs a combined micro-current, and the multi-channel synchronous acquisition unit using the same trigger clock collects the induced voltage, current and phase of the spare port, functional port and on-chip noise monitoring point that are not involved in the current excitation.

[0060] During data acquisition, each functional port maintains a fixed test pattern or a preset idle pattern, and spare ports not involved in excitation are connected to a reference impedance. Attenuation and additional delay introduced by cables, probes, fixtures, and port pads are de-embedded using a standard through-structure or a known reference structure. When normal services cannot be interrupted, a calibration code distinct from the service signal is added to the combined microcurrent, and the corresponding response is extracted through relevant calculations.

[0061] The complex transfer coefficient is obtained by calculating the complex ratio of the induced voltage to the excitation current at each frequency point. The complex transfer coefficient is then subjected to phase expansion and inverse Fourier transform to obtain the time-domain impulse response. The moment when the time-domain impulse response first exceeds four times the root mean square value of the background noise under the same operating condition is defined as the response arrival time. A bidirectional transfer response matrix is ​​established using the synthetic virtual port location, response port, and frequency point as indices. Each matrix element records the complex transfer coefficient, response arrival time, excitation direction, acquisition signal-to-noise ratio, and complex excitation weight.

[0062] A portion of the bidirectional transfer response is selected as correction data. The equivalent conductivity of the conductor, the equivalent dielectric constant of the dielectric, the parasitic capacitance of the spare port, the equivalent impedance of the port, and the local propagation delay in the current scanning area are corrected by constrained least squares. This minimizes the weighted square difference between the measured complex transfer coefficient and the simulated complex transfer coefficient. The correction value must not exceed the measured range of the material obtained in S110, the packaging manufacturing tolerance, and the range of port constraint parameters determined in S120.

[0063] The remaining bidirectional transfer responses are used as independent verification data. When the amplitude error, phase error, and response arrival time error do not exceed the verification threshold determined by the acquisition error and grid convergence error, the bidirectional transfer response matrix and the electromagnetic co-simulation model corrected by the bidirectional transfer response are output. Otherwise, the correction is revoked and the signal-to-noise ratio, port status, and de-embedding parameters are rechecked.

[0064] S230. Determine the dark mode field core based on the bidirectional transfer response matrix and the electromagnetic co-simulation model corrected for the bidirectional transfer response. Dark mode refers to the electromagnetic propagation state in which, within a preset operating state and preset frequency band, the additional response of the functional port is lower than the external response threshold, while the response of the internal backup port, the return current intensity, or the local electromagnetic energy is higher than the internal response threshold. Dark mode field core refers to the connected region with the highest local electromagnetic energy density, composed of continuous grid cells, within the closed current path surrounding the dark mode.

[0065] The external and internal response thresholds are primarily determined by test data of healthy packaged samples under the same operating conditions; when healthy packaged samples are unavailable, the mean of the response under the current package's low load condition is used plus three times the standard deviation.

[0066] Based on the bidirectional transfer response matrix, the synthetic virtual port positions are selected where the additional responses of all functional ports are lower than the external response threshold, the responses of at least two unexcited backup ports are higher than the internal response threshold, and the acquisition signal-to-noise ratio meets the requirements; the corresponding current density distribution and electromagnetic energy density distribution are reconstructed in the electromagnetic co-simulation model corrected by the bidirectional transfer response.

[0067] Starting from a high current density grid node, current streamlines are traced along the current density vector direction. When the distance between the endpoint and the starting point of the streamline does not exceed the maximum side length of two grid cells, the total length of the streamline is greater than the perimeter of a single local microbump or a single through-silicon via, and the angle between the current direction of the endpoint and the starting point does not exceed a preset direction threshold, it is determined as a closed current path.

[0068] Starting with the grid cell with the highest electromagnetic energy density within the region enclosed by the closed current path, adjacent grid cells with electromagnetic energy density not less than 60% of the local maximum and spatially connected are included in the dark mode field core; the dark mode field core is confirmed only if the region exists at at least 3 adjacent frequency points or 3 consecutive sampling times.

[0069] The continuous frequency range in which the electromagnetic energy of the dark mode field core is not less than 50% of the peak energy is determined as the dominant frequency band, and the spare ports whose transmission amplitude and acquisition signal-to-noise ratio with the dark mode field core both reach the preset threshold are determined as the set of observable spare ports.

[0070] The combined microcurrent amplitude is changed, the local mesh is refined, and the backup ports that did not participate in the original scan are called for verification; only when the dark mode field core position and the closed current path remain stable, the dark mode field core, closed current path, dominant frequency band and observable backup port set are output for S310 to call.

[0071] S3 specifically includes the following sub-steps:

[0072] S310. Identify the actual interference source chip and generate a calibration response dataset.

[0073] The system calls upon the dark mode field core, closed current path, dominant frequency band, and observable spare port set output from S230, and the electromagnetic co-simulation model corrected for bidirectional transfer response output from S220. It then obtains the output toggle rate, operating frequency, port current, and task startup status for each chip. The output toggle rate is derived from the chip performance counter, the operating frequency from the clock control unit, the port current from the on-chip current monitoring unit or packaged power supply monitoring unit, and the task startup status from the task scheduling unit.

[0074] By sequentially retaining the current port excitation of only one chip, while keeping the current termination impedance of the remaining chips and ceasing active excitation, calculate the electromagnetic energy generated by this chip on the dark mode field core in the dominant frequency band, and determine the electromagnetic contribution of chip k:

[0075]

[0076] in, The electromagnetic contribution of chip k; To retain only the electromagnetic energy of the dark mode field core at frequency f when chip k is actively excited; n is the chip index involved in the calculation. To retain only the electromagnetic energy of the dark mode field core at frequency f when chip n is actively excited; and These are the lower and upper limits of the dominant frequency band, respectively. The number of chips involved in the calculation; The frequency integral element is used. Chips are sorted from largest to smallest based on their electromagnetic contribution and accumulated sequentially. Chips whose cumulative electromagnetic contribution first reaches 80% are identified as candidate interference source chips. In this embodiment, 80% is used to retain the main energy source.

[0077] Multi-tone calibration excitation covering the dominant frequency band is applied sequentially to each candidate interference source chip. The peak excitation current does not exceed 5% of the peak normal operating current of the corresponding port, and this is continued for one calibration cycle. The 5% is determined based on the port's rated capacity and the upper limit of the test that does not cause changes in logic state.

[0078] If the applied voltage at the functional port exceeds 2% of the normal signal swing, the calibration excitation should be stopped immediately. If at least two observable backup ports acquire responses consistent with the calibration code, the response arrival order matches the model prediction, and the increase in the dark mode field nuclear electromagnetic energy after the calibration excitation is not less than 10% relative to the reference energy before excitation, the corresponding chip is identified as the actual interference source chip; multiple chips meeting these conditions form an actual interference source chip set.

[0079] Record the chip identifier, excitation port, excitation waveform (i.e., source-end excitation template), excitation time, induced voltage, induced current, response phase, response arrival time, acquisition signal-to-noise ratio, and operating status of each observable backup port to form a calibration response dataset. For example, if the electromagnetic contributions of the three chips are 0.62, 0.28, and 0.10 respectively, the first two chips enter the candidate set; if only the calibration excitation of the first chip simultaneously satisfies the multi-port response and energy increment conditions, then the first chip is identified as the actual interference source chip. Input the actual interference source chip set and the calibration response dataset into S320.

[0080] S320. Establish and verify the interference propagation model based on the calibration response dataset. The interference propagation model refers to the model that characterizes the propagation segment connection relationship, propagation time delay of each propagation segment, overall impulse response, and spare port transmission relationship of electromagnetic disturbances from the actual interference source chip through chip ports, microbumps, through-silicon vias, package traces, and ground return structures to the dark mode field core.

[0081] The calibration response dataset is input into the electromagnetic co-simulation model corrected by bidirectional transfer response. Starting from the excitation port of the actual interference source chip, the connection is established from early to late according to the arrival time of the first effective response of the adjacent grid cells. Only grid cells with continuous energy transfer direction, response amplitude exceeding 4 times the root mean square value of background noise and finally reaching the dark mode field core are retained. The continuous grid cells are mapped to the corresponding micro bumps, through silicon vias, package traces and ground return structures to form the interference propagation path.

[0082] When multiple paths exist, calculate the proportion of energy transmitted by each path to the total transmitted energy, and delete paths with an energy proportion of less than 5%; the propagation delay of adjacent propagation segments is the difference between the arrival times of the first valid response at both ends, and the overall propagation delay of each path is the sum of the propagation segment delays.

[0083] The calibration excitation waveform and the observable standby port response waveform are subjected to Fourier transform, and the frequency domain transfer function is obtained through regularized deconvolution.

[0084]

[0085] in, The frequency domain transfer function from the actual interference source chip k to the observable backup port q; The calibration excitation spectrum of chip k; The response spectrum of the observable standby port q; for The complex conjugate; This is a regularization parameter, determined based on the ratio of background noise power to calibration excitation power.

[0086] An inverse Fourier transform is performed on the frequency domain transfer function to obtain the time domain impulse response. Then, combining the bidirectional transfer response matrix from S220, the dark mode core location determined by S230, and the closed current path, the transfer amplitude, transfer phase, and additional propagation delay from each spare port to the dark mode core are calculated. One set of calibration events not involved in model building is selected for verification. If the amplitude error, phase error, and arrival time error between the predicted and measured responses do not exceed the thresholds determined by the acquisition error and model verification error, the interference propagation model is output; otherwise, the propagation path, time synchronization, and de-embedding parameters are rechecked. The calibration response dataset is retained in S520 as the model update baseline.

[0087] S330 implements fault prediction and health management and generates control safety boundaries. The electromagnetic health status sequence refers to a multi-parameter sequence formed by recording the dark mode field core electromagnetic energy, the center frequency of the dominant frequency band, the location of the closed current path, the transmission amplitude of the spare port, the propagation delay, and the chip temperature in the order of the operating window; the first 5 items come from S220, S230, and S320, and the chip temperature comes from the temperature sensor used in S130.

[0088] Fault prediction and health management refers to the process of predicting the response changes of dark mode field cores based on electromagnetic health state sequences and interference propagation models, and determining the risk level, health margin, and control safety boundary of dark mode runaway. Dark mode runaway refers to a state in which the peak value of the dark mode field core is expected to exceed the upper limit of permissible electromagnetic energy, or the additional voltage, power supply noise, bit error rate, or clock jitter of the functional port is expected to exceed the corresponding permissible upper limit.

[0089] The additional voltage and power supply noise of the functional port are obtained by the port monitoring unit; the bit error rate and clock jitter are obtained by the chip error counter and clock monitoring unit, respectively, and the correspondence between dark mode response parameters and functional indicators is established by synchronous test data of healthy packaged samples under different dark mode energy conditions.

[0090] Using 1ms as one operating window, data from no less than 100 operating windows are continuously acquired. Sliding linear fitting is performed on the nuclear electromagnetic energy of the dark mode field, the center frequency of the dominant frequency band, and the propagation delay. The parameter values ​​for the next 20 operating windows are predicted based on the fitting slope, and the predicted peak value is calculated by inputting the interference propagation model.

[0091] The predicted peak value, the degree to which the dominant frequency band approaches the chip's operating harmonics, the changes in closed current paths and propagation delay, and the changes in chip temperature are normalized relative to the healthy baseline and the runaway boundary using the Min-Max Normalization method. Then, the weights of each are determined based on the historical data of healthy packaged samples and abnormal samples, and the weighted sum is obtained to obtain the dark mode runaway risk index.

[0092] Based on the upper limit of normal fluctuations for healthy packaged samples and the allowable upper limit specified in the chip datasheet, two risk thresholds are set to classify risks into low, medium, and high risks. The normalized margins corresponding to the runaway boundaries for electromagnetic energy, bit error rate, and clock jitter distance are calculated respectively, and the minimum value is taken as the health margin.

[0093] Based on the risk level, health margin, backup port rated capacity, and permissible impact of the functional port, the upper limit of the target remaining response and the maximum injected energy of the backup port are determined. In the interference propagation model, the pre-echo phase and injection time are perturbed respectively. The maximum deviation that still allows the remaining response to be below the target upper limit is determined as the phase tolerance and the early injection time tolerance, forming the control safety boundary. The dark mode runaway risk level, health margin, and control safety boundary are output for S410 to use.

[0094] S4 specifically includes the following sub-steps:

[0095] S410. Perform pre-echo strategy optimization based on the interference propagation model and control safety boundary. Strategy optimization refers to jointly solving for the target backup port combination, the excitation amplitude, excitation phase, and injection timing of each target backup port under the constraints of the interference propagation model and control safety boundary, so as to minimize the residual electromagnetic response after the first interference wavefront cancellation, while limiting the energy injected into the backup ports and the additional response of the functional ports.

[0096] The system utilizes the dominant frequency band and observable backup port set output from S230, the port constraint parameters output from S120, the bidirectional transfer response matrix output from S220, the calibration response dataset output from S310, the interference propagation model output from S320, and the dark mode runaway risk level, health margin, and control safety boundary output from S330. Based on the source-end excitation template in the calibration response dataset and the interference source impulse response in the interference propagation model, the predicted first-wave spectrum at the dark mode field core is determined.

[0097] The candidate target backup port is determined as follows: the transmission amplitude of the dark mode field core is not lower than the median transmission amplitude of all observable backup ports under the current operating state; the excitation can form an electromagnetic response at the dark mode field core with the phase opposite to the expected first interference wavefront after transmission; the additional propagation delay can meet the early injection requirements; and the rated current, operating bandwidth and functional port additional response all meet the port constraint parameters.

[0098] Establish strategy evaluation values ​​at discrete frequency points in the dominant frequency band:

[0099]

[0100] Where J is the strategy evaluation value; The first in the dominant frequency band L is the number of discrete frequency points; The expected first wave spectrum at the core of the dark mode field; This represents the number of backup ports for the current candidate target. The complex transfer function from candidate target backup port i to the dark mode field kernel; Pre-echo the frequency domain of the spare port; This is the energy penalty coefficient; the higher the risk level, the higher the penalty coefficient. The smaller the value, the less likely the total injected energy must exceed the control safety boundary.

[0101] During the solution process, the Sequential Quadratic Programming (SQP) algorithm is used as the underlying solver for joint optimization. The excitation amplitude, excitation phase and injection timing of each backup port are used as continuous design variables. The objective function is to minimize the strategy evaluation value J. The peak current, peak voltage and operating frequency of each backup port are constrained to be within the range of port constraint parameters. The residual injected charge in one control cycle does not exceed the residual injected charge threshold determined by S210. The additional response of all functional ports does not exceed the coupling upper limit. Furthermore, no new dark mode field cores are allowed to form in other packaging areas.

[0102] For discrete port combination variables, following a greedy strategy, target backup ports are gradually added in descending order of transmission amplitude. For each additional port, the above-mentioned sequential quadratic programming algorithm is called to solve for the optimal excitation parameters under the current combination and a full-encapsulation simulation is performed. The expansion stops when the expected reduction ratio of the remaining response is less than 5%, or when the maximum number of ports allowed by the control safety boundary is reached.

[0103] The feasible strategy with the smallest strategy evaluation value is determined as the target control strategy, and the remaining feasible strategies are determined as backup control strategies, forming a candidate set of pre-echo control parameters, which is then input into S420 and S430.

[0104] S420 establishes an advance trigger mapping between the source-end leader signal and the pre-echo control parameters. The source-end leader signal refers to the hardware status signal output by the port control unit, clock control unit, memory controller, on-chip interconnect controller, or task scheduling unit before the actual interference source chip generates port interference current, and which can determine the type of subsequent interference event and the expected time of occurrence. The output enable signal, clock domain switching signal, data burst request signal, and task start signal originate from the corresponding control unit on the chip. These signals are led out to the parameter acquisition unit outside the package through the chip's reserved test pins (such as the JTAG interface) or a dedicated debug bus, and timestamps are recorded using a unified time base consistent with S220.

[0105] For each type of source leader signal, record the signal occurrence time, port interference current generation time, operating frequency, port load, data burst length, and task type in no less than 20 calibration events. Only when the signal stably corresponds to the first wave waveform of the same type, and the time difference fluctuation between the source leader signal and the port interference current generation time does not exceed the advance injection time tolerance determined by S330, is it determined to be a valid source leader signal.

[0106] Let the effective source leader signal occur at time . The time difference between the signal and the actual moment when the port interference current is generated is The actual propagation delay from the interference source chip to the dark mode field core is The additional propagation delay from the target backup port i to the dark mode field core is The total system latency is The corresponding control command is issued at the following time:

[0107]

[0108] in, The timing of issuing the control command for the target backup port i; This includes the latency of the pilot signal detection, control calculation, digital-to-analog conversion, and port drive, which are obtained through the controller timestamp and the actual output response of the backup port. When the control command is issued earlier than the pilot signal can be detected, or when the available execution time is less than the hardware execution margin, the backup control strategy is invoked or the source pilot signal that appears earlier is selected.

[0109] A mapping is established by mapping the actual interference source chip identifier, source-end precursor signal type, event characteristic range, applicable operating state, risk level, target backup port combination, control command issuance time, and pre-echo control parameter candidate set index. The event characteristic range includes the operating frequency range, port load range, data burst length range, and task type.

[0110] When multiple control tasks occupy the same backup port, the control task with the higher risk level and smaller health margin is executed first, and the remaining tasks call the backup control strategy. The pre-triggered mapping input S430 will be used.

[0111] S430. Generate and verify the pre-echo injection plan based on the advance trigger mapping. The pre-echo waveform refers to the waveform obtained by reverse calculation based on the expected first interference wavefront and the transmission relationship from the target backup port to the dark mode field core, and injected through the target backup port before the first interference wavefront reaches the dark mode field core, so that the two form an anti-phase superimposed current waveform at the dark mode field core.

[0112] After detecting a valid source leader signal, the target control strategy is obtained based on the advance trigger mapping. Within the event characteristic range, the normalized absolute deviation between the current source event and the historical event in terms of operating frequency, port load, and data burst length is calculated. The deviations are sorted from smallest to largest according to the weighted sum of these three deviations (the weights are determined by the sensitivity of each parameter to the first interference wave shape). The source excitation template with the smallest weighted sum is selected from the calibration response dataset of S310 and convolved with the interference source impulse response of S320 to obtain the expected first interference wavefront.

[0113] The expected first interference wavefront is converted to the frequency domain. The frequency domain pre-echoes of each target backup port obtained by S410 are called. The inverse Fourier transform is performed to obtain the port-side time domain pre-echo waveforms. Then, the waveforms are discretized and the amplitude is quantized according to the sampling rate of the digital-to-analog converter and the port drive accuracy.

[0114] The hardware transfer function between the digital control value and the actual output of the backup port is measured using a standard calibration waveform, and compensation is only provided for amplitude attenuation and phase delay caused by the digital-to-analog converter, port driver unit, and port switch. If the theoretical waveform contains frequency components that exceed the operating bandwidth of the target backup port, the overbandwidth components are removed, and the corresponding compensation is borne by the remaining target backup ports within their rated capabilities.

[0115] The processed pre-echo waveform is input into the electromagnetic co-simulation model after bidirectional transfer response correction for full encapsulation verification. An effective pre-echo injection plan is generated only when the expected residual response of the dark mode field core is lower than the upper limit of the target residual response, the additional response of all functional ports is lower than the upper limit of coupling, no new dark mode field cores are formed in other regions, and the peak current, total injected energy and output bandwidth of each target spare port are within the allowable range.

[0116] The plan records the actual interference source chip identifier, source-end leader signal type, target backup port identifier, discrete pre-echo waveform, control command issuance time, target output time, allowable amplitude adjustment range, allowable phase adjustment range, allowable time adjustment range, expected residual response, and planned failure conditions. Planned failure conditions include changes in the dark mode field core position, interference propagation model updates, event characteristics exceeding the applicable range, or abnormal backup port status. The effective pre-echo injection plan is input to S510.

[0117] S5 specifically includes the following sub-steps:

[0118] S510. Perform entity injection and acquire the remaining electromagnetic response according to the effective pre-echo injection plan.

[0119] The effective pre-echo injection plan output by S430 is invoked. The multi-channel synchronous controller triggers the multi-channel digital-to-analog converter according to the control command, converting the discrete pre-echo waveform into an analog current, which is then output to the target backup port via the port drive unit. The channel clock calibration unit corrects the output channel time difference, and the protection unit monitors whether the target backup port exceeds the port constraint parameters determined by S120.

[0120] Using the timestamp of the source-end pilot signal as the unified time starting point of the current control cycle, the time when the control command is issued, the target output time, the actual output time, and the arrival time of the first valid response are recorded, and the actual output voltage and actual output current of the target backup port are collected.

[0121] The injection is considered valid only if the amplitude error, phase error, and timing error of the actual output waveform relative to the planned waveform are all within the allowable adjustment range recorded in S430; otherwise, the injection is stopped, the reference impedance state of the target spare port is restored, the out-of-tolerance type and the reason for termination are written into the control execution record, and the propagation parameters are not updated using this control cycle.

[0122] The residual electromagnetic response refers to the control result consisting of the residual response of the dark mode field core acquired by the observable backup port after pre-echo injection and the influence of the operation of the functional port. It includes the residual response amplitude, phase, arrival time, electromagnetic energy reconstructed by the dark mode field core, bit error rate of the functional port, power supply noise and clock jitter. The relevant data are acquired by the observable backup port, chip error counter, power supply monitoring unit and clock monitoring unit.

[0123] To isolate the response directly coupled to the pre-echo waveform, only the pre-echo waveform is output during the calibration phase without triggering actual interference source events, and the pre-echo direct response is acquired. During actual control, this direct response is corrected according to the actual output amplitude, phase, and chip temperature, and subtracted from the measured response. The reference response preferentially uses uncontrolled historical responses of the same package with the same event type, operating frequency, port load, and temperature range; when no matching data exists, the uncontrolled response is reconstructed from the S320's interference propagation model.

[0124] Port weights are determined based on the propagation amplitude and acquisition signal-to-noise ratio of each observable backup port to the dark mode field core, and the residual response suppression rate is calculated:

[0125]

[0126] in, The residual response suppression rate; The number of observable backup ports; Let q be the weight of the backup port, and the sum of all weights is 1, where t is the time variable; To remove the residual response after the direct pre-echo response; As the baseline response; and These are the start and end times of the first wave analysis time window, respectively; It is a time-integral infinitesimal element. It outputs the actual output waveform, actual output time, remaining electromagnetic response, and control execution record for use by S520 and S530.

[0127] S520. Update the model based on the remaining electromagnetic response and perform secondary strategy optimization.

[0128] The system calls upon the expected residual response recorded by S430, the actual output waveform, actual output time, and residual electromagnetic response output by S510, the interference propagation model output by S320, the calibration response dataset retained by S310, and the control safety boundary output by S330. First, the actual output waveform is compared with the planned waveform. If the actual output amplitude, phase, or time exceeds the allowable adjustment range, only the transfer function of the execution hardware or the total system delay in S420 is corrected; the packaged electromagnetic parameters are not corrected.

[0129] The electromagnetic co-simulation model and interference propagation model corrected by bidirectional transfer response are only updated when the pre-echo waveform is effectively executed, but the amplitude difference, phase difference, or arrival time difference between the actual residual electromagnetic response and the expected residual response still exceeds the model verification threshold.

[0130] When the residual response amplitude is generally large while the arrival time is consistent, correct the transfer amplitude from the target backup port to the dark mode field core; when there is a consistent phase deviation in the residual response, correct the transfer phase; when there is a consistent deviation in the arrival time, correct the additional propagation delay; when the amplitude changes in each backup port are inconsistent and the high-energy region moves, redetermine the dark mode field core and the closed current path; when the additional response of the functional port increases, correct the complex transfer coefficient from the target backup port to the functional port; when a new dark mode field core is formed in other regions, mark the current strategy as unavailable.

[0131] Parameter adjustments are limited to the measured material range obtained in S110, the packaging manufacturing tolerances, the port constraint parameters determined in S120, and the allowable range of local parameters determined in S220. After obtaining the updated electromagnetic co-simulation model and the updated interference propagation model, the advance injection time, excitation phase, and excitation amplitude are adjusted sequentially. If the existing port combination still cannot meet the upper limit of the target residual response, the backup control strategy output in S410 is invoked. Each adjustment must not exceed the allowable adjustment range recorded in S430; if it does, the process returns to S410 to re-execute the strategy optimization.

[0132] Verification is performed using an independent control cycle. If the updated model's response amplitude, phase, arrival time, functional port influence, and dark mode core position error all do not exceed the verification threshold, the update is considered valid; otherwise, the update is revoked. Secondary strategy optimization is stopped when the suppression rate increment after two consecutive adjustments is less than 5%, the maximum number of optimizations is reached, the next adjustment will exceed the control safety boundary, or a new dark mode core appears.

[0133] The updated electromagnetic co-simulation model, the updated interference propagation model, and the updated pre-echo injection plan or strategy failure flag are output for the S530 to use.

[0134] S530, Determine the suppression result and update the fault prediction and health management record.

[0135] The preset suppression condition refers to the simultaneous fulfillment of the following limiting conditions: the residual response of the dark mode field core, the impact of the operation of the functional port, the detection result of the new dark mode, and the stability of the control execution. These conditions include: the residual response suppression rate reaching the minimum suppression rate determined by S330; the electromagnetic energy reconstructed by the dark mode field core being lower than the upper limit of the target residual response; the bit error rate, power supply noise, and clock jitter of the functional port not exceeding the allowable upper limit; no new dark mode field cores forming in other regions; no parameter exceeding the limit occurring at the target spare port; and five consecutive source events of the same type satisfying the above conditions.

[0136] When the conditions are met, the target backup port combination, discrete pre-echo waveform, control command issuance time, event characteristic range, operating state range, temperature range, and model version are determined as the effective control configuration. An effective control configuration refers to a set of control parameters with a clearly defined applicable range and verified through continuous control cycles. When a source-end leader signal within the applicable range is detected, the effective control configuration is invoked as the initial configuration for generating the pre-echo injection plan in S430. If the conditions are not met, the updated pre-echo injection plan is invoked to continue executing S510. If a strategy failure flag exists, a request for frequency reduction, task staggering, task migration, or functional port isolation is output.

[0137] Fault prediction and health management records refer to a data set that stores dark mode identification, risk prediction, control execution, suppression results, and model updates in chronological order. It includes dark mode field core, dominant frequency band, closed current path, actual interference source chip, risk level, health margin, dynamic modeling trigger record, pre-echo injection plan, actual output waveform, residual electromagnetic response, control termination reason, model parameter update value, effective control configuration, and cumulative injected energy of spare port.

[0138] The recorded data is divided into uncontrolled state data, controlled state data, and reconstructed data after removing the control contribution. When S130 performs dynamic modeling, it only calls the uncontrolled state data or the reconstructed data and does not directly use the original low response in the controlled state, so as to avoid misjudging the suppression effect brought by active electromagnetic control as an improvement in the physical characteristics of the encapsulated body.

[0139] The updated electromagnetic co-simulation model and the updated interference propagation model will serve as the model basis for subsequent S130 dynamic modeling and S330 health prediction, and the strategy failure record and the cumulative usage status of the backup port will be fed back to S410. If the dark mode field core location, closed current path, dominant frequency band, propagation delay, target backup port transmission parameters, or operating temperature exceed the applicable range of the effective control configuration, the configuration will be marked as failed, and S130 to S430 will be re-executed.

[0140] Example 2: Figure 2 As shown, this embodiment provides an electromagnetic collaborative simulation control system for a multi-chip and packaged system, including:

[0141] The system includes a parameter acquisition unit, a multi-channel synchronous excitation acquisition unit, a control processor, a memory, a multi-channel digital-to-analog converter, and a port driver unit. The memory stores a program executed by the control processor, which enables the system to function as follows:

[0142] The model building and dynamic correction module is used to obtain the structural, electrical and operational parameters of the multi-chip and packaging system, establish an initial electromagnetic co-simulation model, construct an effective synthetic port group based on the spatial location and electromagnetic coupling relationship of the spare ports, and obtain the current electromagnetic co-simulation model based on the real-time operating status.

[0143] The dark mode identification module is used to adjust the excitation amplitude and excitation phase of the backup ports within the effective synthetic port group, generate synthetic virtual ports with variable positions, acquire bidirectional transmission responses through a multi-channel synchronous excitation acquisition unit, correct the current electromagnetic co-simulation model, and determine the dark mode field core, dominant frequency band, and observable backup port set.

[0144] The propagation modeling and health management module is used to determine the actual interference source chip based on the dark mode field core, dominant frequency band and observable backup port set, collect calibration response and establish interference propagation model, and determine the dark mode runaway risk level, health margin and control safety boundary based on electromagnetic health state sequence.

[0145] The pre-echo plan generation module is used to optimize the target backup port combination, excitation amplitude, excitation phase and injection timing according to the interference propagation model and control safety boundary, establish the advance trigger mapping between the source leader signal and the pre-echo control parameters, and generate the pre-echo injection plan.

[0146] The injection and feedback update module is used to control the multi-channel digital-to-analog converter and port driver unit to inject pre-echo waveforms into the target backup port according to the pre-echo injection plan, collect the remaining electromagnetic response, update the electromagnetic co-simulation model and interference propagation model according to the remaining electromagnetic response, and generate an effective control configuration when the preset suppression conditions are met.

[0147] All the above formulas are performed using dimensionless numerical calculations; the relevant formulas are based on empirical models that approximate the real situation, obtained through extensive data collection and software simulation fitting. The preset parameters and thresholds involved in the formulas can be conventionally set and adjusted by those skilled in the art according to the physical constraints of the actual application scenario.

[0148] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0149] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0150] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An electromagnetic co-simulation control method for multi-chip and packaged systems, characterized in that, Includes the following steps: S1. Obtain the structural, electrical, and operational parameters of the multi-chip and packaging system, and establish an initial electromagnetic co-simulation model; construct an effective synthetic port group based on the spatial location and electromagnetic coupling relationship of the spare ports, and correct the model according to the operating status to obtain the current electromagnetic co-simulation model; S2. Adjust the excitation amplitude and excitation phase of the backup ports within the effective synthetic port group to generate synthetic virtual ports with variable positions; collect bidirectional transfer responses and correct the current electromagnetic co-simulation model to determine the dark mode field core, dominant frequency band, and observable backup port set; S3. Determine the actual interference source chip based on the dark mode field core, dominant frequency band, and observable backup port set; collect calibration responses and establish an interference propagation model; determine the dark mode runaway risk level, health margin, and control safety boundary based on the electromagnetic health state sequence and interference propagation model. S4. Based on the interference propagation model and control safety boundary, optimize the target backup port combination, excitation amplitude, excitation phase and injection timing, establish the advance trigger mapping between the source leader signal and the pre-echo control parameters, and generate the pre-echo injection plan.

2. The electromagnetic collaborative simulation control method for multi-chip and packaging systems according to claim 1, characterized in that, Also includes: S5. Inject the pre-echo waveform into the target backup port according to the pre-echo injection plan and collect the remaining electromagnetic response. Update the electromagnetic co-simulation model and interference propagation model according to the remaining electromagnetic response, perform secondary strategy optimization, and generate an effective control configuration when the preset suppression conditions are met.

3. The electromagnetic co-simulation control method for multi-chip and packaging systems according to claim 1, characterized in that, S1 specifically includes: The system structure, electrical and operating parameters are obtained, the port direction, reference impedance and reference ground are unified, and chip port model, package interconnection model, ground return current model and spare port model are established. After grid convergence verification, the initial electromagnetic co-simulation model and model basic data are formed. A frequency sweep calibration signal with limited energy is applied to the backup port. The comprehensive coupling degree is calculated based on spatial proximity, bidirectional response amplitude correlation, response phase correlation, and common return path overlap. A backup port coupling relationship matrix is ​​constructed, and effective synthetic port groups and port constraint parameters that meet the upper limit of functional port coupling are selected.

4. The electromagnetic co-simulation control method for multi-chip and packaging systems according to claim 3, characterized in that, Also includes: The system collects operating parameters and background induction signals from the backup port. Based on the normalized deviation relative to the model's base data, it triggers dynamic modeling, corrects chip port excitation, conductor loss, dielectric parameters, return impedance, and port transmission parameters, and obtains the current electromagnetic co-simulation model after verification by the backup port response.

5. The electromagnetic co-simulation control method for multi-chip and packaging systems according to claim 1, characterized in that, S2 specifically includes: By calling the effective synthetic port group, port constraint parameters, spare port coupling relationship matrix and the current electromagnetic co-simulation model, the complex excitation weights of the spare ports in the group are obtained. Under the conditions of satisfying the upper limit of functional port coupling and the residual injected charge threshold, a synthetic virtual port with variable position and a set of virtual port scanning parameters are formed. The bidirectional transmission response obtained by collecting the excitation port and response port of the virtual port scanning parameter set is used to establish the bidirectional transmission response matrix. The transmission parameters of the local conductor, medium and port are constrained and corrected according to the measured response to obtain the electromagnetic co-simulation model corrected by the bidirectional transmission response. Based on the bidirectional transfer response matrix and the corrected model, the current density and electromagnetic energy density are reconstructed, closed current paths and dark mode field cores are identified, and the dominant frequency band and observable backup port set are determined after stability verification.

6. The electromagnetic collaborative simulation control method for a multi-chip and packaging system according to claim 1, characterized in that, S3 specifically includes: The electromagnetic co-simulation model, which includes the dark mode field core, closed current path, dominant frequency band, set of observable spare ports, and corrected bidirectional transfer response, is used to calculate the electromagnetic contribution of each chip to the dark mode field core, screen candidate interference source chips, and determine the actual interference source chip through energy-limited calibration excitation, thus forming a calibration response dataset. Based on the calibration response dataset and the corrected model, the interference propagation path, propagation delay and impulse response are determined, and the transmission amplitude, transmission phase and additional propagation delay from the backup port to the dark mode field core are calculated. After verification by independent calibration events, the interference propagation model is established.

7. The electromagnetic co-simulation control method for multi-chip and packaging systems according to claim 6, characterized in that, Also includes: An electromagnetic health state sequence is constructed based on the dark mode field core electromagnetic energy, dominant frequency band, closed current path, port transmission parameters, and chip temperature. The dark mode response change is predicted by combining the interference propagation model, and the dark mode runaway risk level, health margin, and control safety boundary are determined.

8. The electromagnetic co-simulation control method for multi-chip and packaging systems according to claim 1, characterized in that, S4 specifically includes: The dominant frequency band, observable backup port set, port constraint parameters, calibration response dataset, interference propagation model, and control safety boundary are used to screen candidate target backup ports. The combination of target backup ports, excitation amplitude, excitation phase, and injection timing are jointly optimized to form a candidate set of pre-echo control parameters. The source-end leader signal of the actual interference source chip is collected. Based on the time difference between the source-end leader signal and the port interference current, the interference propagation delay, the additional propagation delay of the target backup port, and the total system delay, the control command issuance time is determined and an advance trigger mapping is established. Based on the target control strategy selected by the early trigger mapping, a discrete pre-echo waveform with hardware characteristic compensation is generated, and an effective pre-echo injection plan is formed after simulation verification.

9. The electromagnetic co-simulation control method for a multi-chip and packaged system according to claim 2, characterized in that, S5 specifically includes: According to the effective pre-echo injection plan, output discrete pre-echo waveforms to the target backup port, collect the actual output waveforms, times and port responses, obtain the residual electromagnetic response after stripping the direct pre-echo response, and calculate the residual response suppression rate. The actual output waveform and residual electromagnetic response are compared with the planned waveform and expected residual response to distinguish between hardware execution error and model error. The electromagnetic co-simulation model and interference propagation model are updated, and the advance injection time, excitation phase, excitation amplitude and target spare port combination are adjusted to form the updated pre-echo injection plan. The preset suppression conditions are determined based on the residual response suppression rate, the electromagnetic energy reconstructed from the dark mode field core, the influence of the functional port, and the detection results of the new dark mode. If the conditions are met, an effective control configuration is generated and the fault prediction and health management records are updated. If the conditions are not met, the updated pre-echo injection plan is executed or a safety control request is output.

10. An electromagnetic co-simulation control system for multi-chip and packaged systems, employing the electromagnetic co-simulation control method for multi-chip and packaged systems as described in any one of claims 1 to 9, characterized in that, include: The system includes a parameter acquisition unit, a multi-channel synchronous excitation acquisition unit, a control processor, a memory, a multi-channel digital-to-analog converter, and a port driver unit. The memory stores the program executed by the control processor.

Images