A stacked chip vertical interconnect coupling electromagnetic field extraction method and computer device

Abstract

The application discloses a stacked chip vertical interconnection coupling electromagnetic field extraction method and computer equipment, and the method comprises the following steps: determining a target position of an electromagnetic field generated by TSV vertical interconnection coupling in a stacked chip, running simulation to obtain electromagnetic field intensity of the target position, and storing TSV coupling structure parameters and the extracted electromagnetic field intensity as a data set; loading data from the data set and performing feature extraction, constructing a feature matrix by using the extracted features, taking the electromagnetic field intensity as a target variable, inputting the data set into each regression model for training to obtain an optimal model of the electromagnetic field intensity; comparing the model prediction result with the simulation result to calculate whether the error meets the requirement, and if the requirement cannot be met, repeating the feature extraction and model training; inputting TSV coupling structure parameters of a stacked chip to be measured into the prediction model to obtain electromagnetic field intensity of the target position. The application can quickly predict electromagnetic field interference of a chip internal circuit caused by TSV vertical interconnection electromagnetic coupling.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of chip electromagnetic interference evaluation, and particularly relates to a stacked chip vertical interconnection coupling electromagnetic field extraction method and a computer device. BACKGROUND

[0002] Chiplet stacked chips have a wide range of electromagnetic noise sources, which are greatly affected by stacked process size, working frequency, electromagnetic source spacing, shielding protection measures, etc. The stacked chips are vertically interconnected through TSV (Through-Silicon Via), but the size span of TSV and CMOS advanced process is large. For the needs of heterogeneous integration of radio frequency devices at different processes, the TSV and RDL (Redistribution Layer) paths in the silicon substrate will form external electromagnetic interference to the circuit units and elements. For example, the electromagnetic field coupled between the TSV high-speed interconnections of the transmission signal lines interferes with the input and output and internal signal timing of other chiplets.

[0003] For the internal electromagnetic interference evaluation problem in Chiplet stacked chips, the existing technology focuses on analytical algorithms and experimental tests. Among them, the analytical method extracts TSV high-frequency parasitic parameters to calculate the conducted electromagnetic field and the radiated electromagnetic field, but due to the complexity of variables in chiplet stacked chips and the large amount of analytical work, it is difficult to build an accurate analytical model. Experimental testing is a standard method for solving electromagnetic fields between board-level chips, but EMC (Electro Magnetic Compatibility) testing requires testing probes and other devices. Due to the size limitation of chiplet stacked chips, there is currently no corresponding solution to carry out testing between DIEs (chiplets or dies), resulting in EMC testing difficulties. SUMMARY

[0004] The purpose of the present application is to provide a stacked chip vertical interconnection coupling electromagnetic field extraction method, computer device, computer readable storage medium and computer program product, which can quickly predict the electromagnetic field interference of the internal circuit of the chiplet caused by the electromagnetic coupling of the TSV vertical interconnection.

[0005] In order to achieve the above-mentioned purpose, one aspect of the present application provides a stacked chip vertical interconnection coupling electromagnetic field extraction method, comprising:

[0006] Step S1, a three-dimensional model of the stacked chip is established by finite element simulation, the target position of the electromagnetic field generated by the TSV vertical interconnection coupling in the stacked chip is determined, the electric field intensity and the magnetic field intensity of the target position are obtained by running the simulation, and the TSV coupling structure parameters and the extracted electric field intensity and magnetic field intensity are stored as a data set;

[0007] Step S2, loading data from the data set and performing feature extraction, constructing a feature matrix using the extracted features, selecting multiple different types of regression models, using a grid search method to optimize the hyperparameters of the selected regression models, determining the optimal hyperparameters of each regression model, taking the electric field intensity and the magnetic field intensity as the target variables, inputting the data set into each regression model for training, and determining the best model for the electric field intensity and the best model for the magnetic field intensity according to the R 2 score and the MAPE comprehensive score.

[0008] Step S3, using the best model for the electric field intensity and the best model for the magnetic field intensity to predict the electric field intensity and the magnetic field intensity, respectively, comparing the prediction results with the electric field intensity and the magnetic field intensity at the target position obtained by simulation, calculating whether the error meets the requirements, and if not, returning to step S2 until the error requirements are met.

[0009] Step S4, inputting the TSV coupling structure parameters of the stacked chip to be measured, and outputting the electric field intensity and the magnetic field intensity at the target position using the best model for the electric field intensity and the best model for the magnetic field intensity.

[0010] Preferably, the TSV vertical interconnection coupling includes TSV coupling between same-layer core particles, TSV coupling between different-layer core particles, coupling between TSV and RDL, and coupling between TSV and metal connection of RDL.

[0011] Preferably, the features include original features, interaction features, and square features, wherein the interaction features are features generated by combining the original features two by two, and the square features are squares of the original features.

[0012] Preferably, the original features include the number of coupling paths, distance, frequency, TSV diameter, core particle thickness, RDL width, and length.

[0013] Preferably, the regression models include Ridge regression, Lasso regression, random forest regression, gradient boosting regression, and support vector machine.

[0014] Preferably, in step S2, the data set is divided into 5 test sets, and the model is repeatedly trained and verified under different test sets, the negative MAPE is used as the evaluation index for each training and verification, the average value of 5 times is used as the final evaluation index of the model, and the hyperparameters of the model with the minimum MAPE index are used as the optimal hyperparameters.

[0015] Preferably, in step S3, the key features affecting the change of the electromagnetic field among all features are used as input variables, and other non-key features are set to fixed values, and fixed data set setting is performed to reduce model overhead.

[0016] Another aspect of the present invention provides a computer device including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described above.

[0017] Another aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described above.

[0018] Another aspect of the present invention provides a computer program product including a computer program that, when executed by a processor, implements the steps of the method described above.

[0019] According to the stacked chip vertical interconnect coupling electromagnetic field extraction method, computer device, computer-readable storage medium and computer program product of the present invention, electromagnetic field interference of the chip internal circuit caused by TSV vertical interconnect electromagnetic coupling can be quickly predicted. Attached Figure Description

[0020] To more clearly illustrate the technical solutions of the present invention, the accompanying drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort:

[0021] Figure 1 This is a flowchart of a method for extracting coupled electromagnetic fields in vertical interconnects of stacked chips according to an embodiment of the present invention;

[0022] Figure 2 This is a schematic diagram of a three-dimensional model of a stacked chip according to an embodiment of the present invention;

[0023] Figure 3 This is a schematic diagram of the TSV coupling structure under test according to an embodiment of the present invention, wherein (a) is a cross-sectional view and (b) is a top view;

[0024] Figure 4 This is a structural diagram of a computer device according to an embodiment of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0026] One embodiment of the present invention provides a method for extracting the coupled electromagnetic field of a vertical interconnect of stacked chips, such as... Figure 1 As shown, the method for extracting the coupled electromagnetic field of stacked chip vertical interconnect in this embodiment of the invention includes steps S1 to S4.

[0027] In step S1, electromagnetic field data generated by the vertical interconnect coupling of TSVs in the stacked chips are extracted through finite element simulation. Specifically, in finite element simulation software such as HFSS or CST, a simulation model is established as follows: Figure 2 The three-dimensional model of the stacked chip shown determines the target location of the electromagnetic field generated by the vertical interconnect coupling of TSVs in the stacked chip. For example... Figure 2 As shown, the coupling objects include: TSV coupling between cores in the same layer, TSV coupling between cores in different layers, coupling between TSV and RDL, and coupling between the metal connections of TSV and RDL. The media between the coupling objects include intra-substrate coupling, intra-cavity coupling, and multi-dielectric intra-coupling penetrating the interlayer insulating material. The target location point from which the electromagnetic field strength needs to be extracted is determined, simulation is run, and the electric field strength (E) and magnetic field strength (H) at the target location are extracted. The physical parameters of the TSV coupling structure of the stacked chip and the simulation results of the extracted electromagnetic field strength are stored as a dataset file in CSV format.

[0028] In step S2, an electromagnetic field strength prediction model is trained based on machine learning methods, including steps S21 to S24.

[0029] Step S21: Data Loading and Feature Engineering

[0030] Data is loaded from the dataset and features are extracted. These features are categorized into material characteristic variables, process characteristic variables, and working characteristic variables. A feature matrix is ​​constructed using the extracted features. The feature matrix can include N original features, N(N-1) / 2 interaction features generated by pairwise combinations of the original features, and the squared features of the N original features, for a total of (N... 2 +3N) / 2 features. The target variables are E (electric field strength) and H (magnetic field strength). Among them, the original features include the number of coupling paths (n), distance (d), frequency (f), TSV process diameter (D), core thickness (t), RDL width (w), length (1), etc., and the interactive features are such as the number of paths * frequency, TSV process diameter * RDL width, etc. The squared features are the squares of the original features, such as n 2 d 2 f 2 wait.

[0031] Step S22: Model selection and hyperparameter tuning

[0032] Model selection: Ridge regression, Lasso regression, Random Forest regression, Gradient Boosting regression, Support Vector Machine (SVR) and other different types of regression models are used for initial training.

[0033] Hyperparameter tuning: GridSearchCV is used to tune the hyperparameters of the selected regression models, such as regularization terms, iteration times, and decision tree depth, which directly affect model performance and cannot be learned from training.

[0034] To establish the optimal hyperparameters of each regression model, the following steps are taken: use 5-fold cross-validation to evaluate the performance of the model based on each hyperparameter on the dataset, divide the dataset into 5 test sets, and then repeatedly train and validate the model on different test sets. Each training and validation uses the Mean Absolute Percentage Error (MAPE) as the evaluation indicator, and the average of 5 times is used as the final evaluation indicator of the model. The hyperparameters of the model with the smallest MAPE are the optimal hyperparameters.

[0035] Step S23: Best model selection

[0036] After completing the hyperparameter tuning of each regression model, the dataset is input into each regression model for training. According to the R 2 score and MAPE comprehensive score (R 2 -MAPE) as the final score. The model with the highest score is the best model. Here, the models for target variables E and H are independent, so we get a best model for E and a best model for H.

[0037] Step S24: Model validation and visualization

[0038] By analyzing the residual plot to predict the error distribution and by plotting the predicted value vs actual value scatter plot to verify the model fitting degree, the performance of the best model on the test set is demonstrated.

[0039] In step S3, the key features that affect the electromagnetic field change, such as the number of coupling paths n and the TSV diameter D, are set as input variables, and other non-key features are set as fixed values to reduce the model overhead.

[0040] Load a new dataset (containing TSV structural parameters not used in training), obtain the E and H results using the best E and H models, compare the predicted results with the E and H values ​​of the target location obtained from simulations using software such as HFSS, and calculate whether the mean squared error or error rate meets the requirements. If the requirements are not met (e.g., the error rate exceeds the threshold of 5%), return to step S2, select more regression models for combined training, and construct a larger feature matrix for optimization.

[0041] For example, various types of regression models can be used, including Ridge regression, Lasso regression, random forest regression, gradient boosting regression, and support vector machines (SVR). During initial training, one regression model can be selected, while during optimization, other models are chosen to achieve the desired error. The feature matrix includes single features and multiple interactive features. During initial training, N original features can be selected as the matrix; during optimization, interactive features can be added to form a larger feature matrix.

[0042] Repeat steps S2 to S3 until the error requirement is met.

[0043] In step S4, according to the application requirements, the coupling structure parameters of the stacked chip under test are input, and the electromagnetic field strength at the target location is output through the E / H optimal model as the final structural data to be predicted.

[0044] The following specific example verifies the effectiveness of the method in this embodiment of the invention.

[0045] Using the number of pathways n, the horizontal distance between centers d, and the target operating frequency f as variable features, and other parameters such as the TSV diameter as constants, the following parameters are used: Figure 3 The coupling structure in this study involves n coupling paths on a two-layer core substrate for predicting the magnetic field strength (H). A random forest regression model is used for training during the prediction process. The original features are the number of paths n, distance d, and frequency f. Commonly used interaction features n*s, n*f, d*f, and squared features n are selected. 2 d 2 f 2 A feature matrix is ​​composed of nine features. The E and H values ​​on the central target core are obtained. Through residual analysis and comparison of predicted and actual values, the simulated H value is basically consistent with the model prediction value, with the error controlled within 0.01, verifying the effectiveness of the method in this embodiment.

[0046] In summary, the stacked chip vertical interconnection coupling electromagnetic field extraction method of the embodiment of the application generates a high-precision electromagnetic field database through finite element simulation, adopts a machine learning method to perform feature extraction on influencing factors of multiple electromagnetic noise sources, predicts electromagnetic field interference of internal circuits of stacked chips caused by TSV vertical interconnection electromagnetic coupling, and solves the problem of difficult evaluation of interference caused by coupled electromagnetic fields between stacked chip dies.

[0047] 1) The electromagnetic field inside the die that cannot be measured can be extracted, and the speed is faster.

[0048] 2) The model can be expanded and is applicable to interference caused by different TSV high-frequency coupling, and has strong applicability.

[0049] The embodiment of the application further provides a computer device, which can be a server, and an internal structure diagram of the computer device can be as shown in Figure 4 The computer device includes a processor, a memory and a network interface connected through a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for running of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store running parameter data of each framework. The network interface of the computer device is used to communicate with an external terminal through network connection. The computer program is executed by the processor to implement the steps of the method of the embodiment of the application.

[0050] Those skilled in the art can understand that Figure 4 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0051] The embodiment of the application further provides a computer readable storage medium, which stores a computer program, and the computer program is executed by the processor to implement the steps of the method of the embodiment of the application.

[0052] The embodiment of the application further provides a computer program product, which includes a computer program, and the computer program is executed by the processor to implement the steps of the method of the embodiment of the application.

[0053] The foregoing merely illustrates some exemplary embodiments of the application, and it will be appreciated that those skilled in the art will be able to devise various modifications without departing from the spirit and scope of the application. The appended drawings and description are illustrative only, and are not intended to be limiting.

Claims

1. A method for extracting the coupled electromagnetic field of a vertical interconnect of stacked chips, characterized in that, include: Step S1: Establish a three-dimensional model of the stacked chip through finite element simulation, determine the target location of the electromagnetic field generated by the vertical interconnect coupling of TSV in the stacked chip, run the simulation to obtain the electric field strength and magnetic field strength at the target location, and store the TSV coupling structure parameters and the extracted electric field strength and magnetic field strength as a dataset. Step S2: Load data from the dataset and extract features. Construct a feature matrix using the extracted features. Select various types of regression models and use a grid search method to fine-tune the hyperparameters of the selected regression models. Determine the optimal hyperparameters for each regression model. Using electric field strength and magnetic field strength as target variables, input the dataset into each regression model for training. Based on R... 2 The optimal models for electric field strength and magnetic field strength are obtained by combining the scores and MAPE scores. Step S3: Predict the electric field strength and magnetic field strength using the optimal models for electric field strength and magnetic field strength, respectively. Compare the prediction results with the electric field strength and magnetic field strength at the target location obtained from the simulation to calculate whether the error meets the requirements. If the error does not meet the requirements, return to step S2 until the error requirement is met. Step S4: Input the TSV coupling structure parameters of the stacked chip under test, and output the electric field strength and magnetic field strength at the target location using the optimal model of electric field strength and the optimal model of magnetic field strength.

2. The method as described in claim 1, characterized in that, TSV vertical interconnect coupling includes TSV coupling between cores in the same layer, TSV coupling between cores in different layers, coupling between TSV and RDL, and coupling between the metal connections of TSV and RDL.

3. The method as described in claim 1 or 2, characterized in that, The features include original features, interactive features, and squared features, where interactive features are generated by combining two original features in pairs, and squared features are the squares of the original features.

4. The method as described in claim 3, characterized in that, The original characteristics include the number of coupling paths, distance, frequency, TSV diameter, core thickness, RDL width, and length.

5. The method as described in claim 1 or 2, characterized in that, The regression models include Ridge regression, Lasso regression, random forest regression, gradient boosting regression, and support vector machine.

6. The method as described in claim 1 or 2, characterized in that, In step S2, the dataset is divided into 5 test sets. The model is trained and validated repeatedly on different test sets. Negative MAPE is used as the evaluation metric for each training and validation. The average of the 5 tests is used as the final evaluation metric for the model. The hyperparameters of the model with the smallest MAPE metric are taken as the optimal hyperparameters.

7. The method as described in claim 1 or 2, characterized in that, In step S3, the key features that affect the change of electromagnetic field among all features are used as input variables, and other non-key features are set to constant values. Fixed dataset settings are used to reduce model overhead.

8. A computer device, comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-7.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-7.

Images