Simulation model creation device

Abstract

To easily perform circuit simulation of a power device.SOLUTION: A model creation system 10 changes default values of each of a plurality of parameters of a power module model 30 so as to reduce difference between electrical characteristics indicated by received electrical characteristics data and electrical characteristics output from a model of simulation of a power device. The power module model 30 includes a JFET-based operation unit and a MOS-based operation unit that perform operation of the electrical characteristics based on an input value to a plurality of terminals. When the electrical characteristics and temperature characteristics among a first electrode, a second electrode, and a third electrode are simulated, changed default values are used as input values input to a first terminal corresponding to the first electrode, the second electrode, and the third electrode and a second terminal used for operation of the temperature characteristics.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present disclosure relates to a technique such as a simulation model creation device that creates a simulation model of a circuit of a power device that uses a semiconductor substrate. [Background technology]

[0002] When simulating a power device circuit, the accuracy of the circuit model is important. For this reason, various models have been developed, as well as systems that perform the modeling itself.

[0003] For example, Patent Document 1 describes a system that models the characteristics of a physical device and obtains an optimal design.

[0004] Furthermore, Patent Document 2 describes an apparatus for creating a circuit simulation model of a power device. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 2023-82386 [Patent Document 2] Japanese Patent Application Laid-Open No. 2013-89696 Summary of the Invention [Problem to be solved by the invention]

[0006] Creating a circuit simulation model requires a large amount of advanced know-how, as well as a high level of knowledge and experience.

[0007] An object of the present disclosure is to provide a technology that can easily create a model used in circuit simulation of a power device. [Means for solving the problem]

[0008] A representative embodiment of the present disclosure has the following configuration. A simulation model creation device according to one embodiment includes: an input unit that inputs electrical characteristic data related to a power device; a model creation unit that changes default values ​​of a plurality of parameters of the model so as to reduce differences between electrical characteristics indicated by the electrical characteristic data received from the input unit and electrical characteristics output by a simulation model of the power device; and an output unit that outputs the default values ​​changed by the model creation unit. The plurality of parameters include parameters that indicate a physical state or a geometric state of the power device. The model includes a model calculation unit that calculates electrical characteristics based on input values ​​to a plurality of terminals. The plurality of terminals include a first terminal corresponding to a first electrode, a second electrode, and a third electrode of the power device, and a second terminal used to calculate temperature characteristics. When simulating the electrical characteristics between the first electrode, the second electrode, and the third electrode, as well as the temperature characteristics, the input values ​​input to the first terminal corresponding to the first electrode, the second electrode, and the third electrode, and the second terminal used to calculate the temperature characteristics use the default values ​​of the parameters after the change. [Effects of the Invention]

[0009] According to a representative embodiment of the present disclosure, a technique for easily creating a circuit simulation model of a power device can be provided. Problems, configurations, effects, etc. other than those described above will be described in the description of the present disclosure. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a diagram illustrating an example of functional blocks of a simulation model creating device according to a first embodiment. [Figure 2] FIG. 2 illustrates an example of a simulation engine according to the first embodiment. [Figure 3]FIG. 3 is a diagram illustrating an example of a simulation engine corresponding to a physics-based model in which four power module models according to the first embodiment are connected to a circuit. [Figure 4] FIG. 4 illustrates an example of the operation of the simulation engine according to the first embodiment. [Figure 5] FIG. 5 illustrates an example of the operation of the simulation engine according to the first embodiment. [Figure 6] FIG. 6 is a diagram illustrating an example of the configuration of the Si power MOSFET according to the first embodiment. [Figure 7] FIG. 7 is a diagram illustrating an example of the configuration of the Si power MOSFET according to the first embodiment. [Figure 8] FIG. 8 is a diagram illustrating an example of the configuration of the SiC power MOSFET according to the first embodiment. [Figure 9] FIG. 9 is a diagram illustrating an example of the configuration of the SiC power MOSFET according to the first embodiment. [Figure 10] FIG. 10 is a diagram illustrating an example of functional blocks of a simulator unit included in the simulation model creating device according to the second embodiment. [Figure 11] FIG. 11 illustrates an example of functional blocks of the model computation unit selection system according to the third embodiment. [Figure 12] FIG. 12 is a diagram illustrating another example of the display of the model selection UI according to the third embodiment. [Figure 13] FIG. 13 is a diagram illustrating an example of functional blocks of the model simulation condition selection system according to the fourth embodiment. [Figure 14] FIG. 14 is a diagram showing an example of a power module model in a modified example. DETAILED DESCRIPTION OF THE INVENTION

[0011] <Related technologies> In the field of power devices known as MOSFETs (Metal Oxide Semiconductor Field Effect Transistors), power MOSFETs using silicon (Si) substrates (hereinafter referred to as "Si power MOSFETs") have traditionally been mainstream. However, the electric field strength required for dielectric breakdown in silicon carbide (SiC) is approximately one order of magnitude greater than that in silicon.

[0012] For this reason, in a power MOSFET using a SiC substrate (hereinafter referred to as "SiC power MOSFET"), the thickness of the drift layer required to maintain a breakdown voltage can be reduced to about one-tenth of that of a Si power MOSFET, and the impurity concentration of the drift layer can be increased by about 100 times. As a result, in a SiC power MOSFET, the device resistance can theoretically be reduced by three orders of magnitude or more. Furthermore, since SiC has a band gap about three times larger than that of Si, a SiC power MOSFET can have a lower on-resistance than a Si power MOSFET at the same breakdown voltage, and can also operate in high-temperature environments. Therefore, SiC semiconductor devices such as SiC power MOSFETs are expected to have performance that exceeds that of Si semiconductor devices such as Si power MOSFETs.

[0013] One example of the use of SiC power MOSFETs is a circuit in which two SiC power MOSFETs connected in series are connected to a load, and the potential associated with the load is adjusted by alternately switching the two SiC power MOSFETs on and off. A circuit configured in this way is called a half-bridge circuit. In a half-bridge circuit, in one SiC power MOSFET, current flows but no voltage is applied when it is in the on state, and no current flows but a voltage is applied when it is in the off state. In this case, the operation of changing the gate voltage of one of the MOSFETs to switch the MOSFET's on / off state is called switching.

[0014] In a half-bridge circuit, when a SiC power MOSFET is on, the voltage is small, and when it is off, the current is small. Therefore, the power generated in a steady state is small in a SiC power MOSFET. However, when switching, a SiC power MOSFET consumes a large amount of power because high voltage and current are generated simultaneously. This power consumption is called switching loss. In order to increase the switching frequency, it is necessary to shorten the switching period and reduce power loss by increasing the rate of change of voltage and current (dv / dt and di / dt).

[0015] The performance index of a SiC power MOSFET is determined by the sum of the switching loss and the conduction loss consumed by the resistance while the switch is on. In other words, the smaller these losses are, the less heat is generated and the higher performance the power device can handle, allowing for large currents to flow.

[0016] On the other hand, being able to pass large currents through SiC power MOSFETs also means that the current value in the event of a fault is large, making circuit protection more difficult. In a half-bridge circuit, if one SiC power MOSFET is on and the other SiC power MOSFET falls into a short-circuit state, the maximum time it takes for the SiC power MOSFET to be returned to the off state is called short-circuit withstand capability. The higher the performance, the shorter the short-circuit withstand capability, so there is a trade-off between short-circuit withstand capability and loss. Because short-circuit protection involves detection and control processes, if the short-circuit withstand capability is less than 3us (microseconds), it is difficult to protect the circuit, and further improvements in SiC power MOSFET performance are difficult.

[0017] Therefore, if the control technology for SiC power MOSFETs, including short-circuit protection, can be improved, lower-loss SiC power MOSFETs will become available. Various systems have been developed to improve such control technology. A typical example is a circuit simulator that performs circuit simulation. By modeling and incorporating a SiC power MOSFET, the switching operation can be verified in a circuit simulator.

[0018] When performing a circuit simulation, the accuracy of the circuit model is important, and for this reason, the techniques described in the above-mentioned Patent Documents 1 and 2 have been developed.

[0019] The system in Patent Document 1 is aimed at device developers. Therefore, it is not a system that is useful for circuit developers, who are users. Specifically, the characteristics of the circuit to be used as input to the evaluation function differ depending on how the circuit is assembled by each user, making it difficult to define the characteristics to be modeled in Patent Document 1.

[0020] Furthermore, the system in Patent Document 2 describes that a model is created from electrical characteristic data. However, no matter how accurate the model, if the actual SiC power MOSFETs themselves vary widely, the accuracy of the model is meaningless. For example, among commercially available SiC power MOSFETs, there are some products with a variation of approximately 1 V in threshold voltage Vth and approximately 30% in on-voltage.

[0021] More specifically, even if a model can achieve an accuracy within 5%, if a SiC power MOSFET with a 30% variation in performance is delivered, the error between the test results of the actual device and the calculation results of the model will expand to 35%. Even if a design is made at the prototype stage using a median model and a SiC power MOSFET with performance close to the median, and actual device measurements are performed to satisfy the required specifications, there is a possibility that a SiC power MOSFET with characteristics that differ by up to 30% will be used in mass production. Such variation in characteristics could lead to failures when the SiC power MOSFET is used in the market.

[0022] In this sense, a worst-case model of the SiC power MOSFET is necessary for model-based product design. Design using a worst-case model is safe and may even eliminate the need for prototypes in the development process.

[0023] The biggest challenge is how to obtain a worst-case model. Modeling involves measuring the electrical characteristics of an actual product and adjusting parameters to reproduce similar characteristics. For commercially available SiC power MOSFETs, experimental behavioral models are mainly used.

[0024] An experimental behavior model is a model that reproduces the characteristics of SiC power MOSFETs using functions such as tanh (a hyperbolic function) and polynomials. An experimental behavior model attempts to reproduce circuits without being bound by physical meaning. Because the function form can be freely changed, the experimental behavior model can easily match the target characteristics, and is fast because complex phenomena can be replaced with a single simple equation.

[0025] On the other hand, experimental behavioral models have the disadvantage that they are difficult to analyze because they have no physical meaning. For example, in a vertical SiC power MOSFET, the overall resistance is made up of multiple resistance components such as channel resistance, JFET resistance, and drift resistance. Because experimental behavioral models treat these multiple resistances as a single resistance, it is not possible to adjust individual components.

[0026] For these reasons, when creating a worst-case model using an experimental behavior model, it is necessary to actually prepare and measure such a worst-case product. However, the occurrence rate of worst-case products is low, and preparing them is not easy. Furthermore, there is not just one worst-case scenario, but multiple scenarios. This creates the need to cover all worst-case scenarios.

[0027] In the present disclosure described below, we propose a perturbation simulation system that uses a physics-based behavioral model that allows parameters to be input. A physics-based behavioral model is a model constructed using an equation that is well supported by physics or an approximation thereof. Examples of such models include the Berkeley Short-channel IGFET Model (BSIM) and the Hiroshima-University STARC IGFET Model (HiSIM).

[0028] In physics-based behavioral models, functions that do not have sufficient physical explanations are avoided as much as possible. Therefore, even if the current model cannot explain a phenomenon, we do not perform simple tasks such as applying polynomials and fine-tuning, as is done with experimental models.

[0029] Therefore, the accuracy of fitting to the individual SiC power MOSFET is inferior to that of an experimental model. Also, because the physical equations are generally complex, the calculation speed is also slow. Therefore, as modeling of a single case is poor in both accuracy and speed, it has not been used in general circuit simulations.

[0030] However, when considering worst-case estimation for SiC power MOSFETs, the constraints of physics-based behavioral models come into play. The parameters of experimental behavioral models have no physical meaning and cannot be interpreted by users. This means that users cannot specify the range within which the parameter values ​​are distributed. On the other hand, the parameters of physics-based behavioral models have physical meaning and are composed of physical constants and parameters. This means that users can specify the range within which the parameter values ​​are distributed.

[0031] Specifically, suppose an experimental behavioral model contains parameters A, B, and C whose meanings are unknown. There is no information on whether these parameters vary by 10% or 100%. Even if parameters A, B, and C exhibit interlocking characteristics, they do not interact in the experimental behavioral model. On the other hand, a physics-based behavioral model consists of, for example, gate dielectric thickness tox, JFET width Wjft, and epitaxial concentration Nepi. Therefore, with some knowledge of manufacturing, it is possible to infer the domain of the parameters, for example, that they may vary by up to 20%. Furthermore, physics-based behavioral models can be verified using actual data from the manufacturing line.

[0032] Therefore, in the case of a physics-based behavioral model, it is possible to combine parameters with known domains within the possible range, and create a worst-case model with a solid basis. Conversely, by creating a range of parameters that are practically acceptable and delivering power devices within that range, safety can be ensured without incurring extreme increases in cost.

[0033] <Technical Concept of the Present Embodiment> According to the technical concept of this embodiment, it is possible to provide a simulation model creation device that creates a physics-based model based on electrical characteristic data of a power device, in other words, a semiconductor device, and performs circuit simulation using the created model. The simulation model creation device includes a model creation system and a simulator unit. The model creation system inputs electrical characteristic data of the semiconductor device, calculates characteristics of the physics-based model, and creates parameters of the physics-based model based on the calculated characteristics. The simulator unit inputs parameters of the created physics-based model and parameter perturbation (variation) ranges, and changes the model parameters based on the input to perform a simulation.

[0034] The simulation model creation device also assumes a distribution for each parameter, performs random sampling, and repeats simulations to estimate the statistical distribution of the characteristic variations in the final product.

[0035] Furthermore, the simulation model creation device is equipped with a model selection system for effectively managing and operating multiple physics-based models, allowing users to identify different models by hash values ​​of source code, etc.

[0036] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. In the drawings, the same parts are generally designated by the same reference numerals, and repeated explanations will be omitted. In the drawings, the representation of components may not represent their actual positions, sizes, shapes, ranges, etc., in order to facilitate understanding of the invention.

[0037] For the sake of explanation, when describing processing by a program, the program, function, processing unit, etc. may be described as the main body, but the main hardware body for these is a processor, or a controller, device, computer system, system, etc. that is configured with the processor, etc. A computer system executes processing according to a program read into memory using resources such as memory and communication interfaces as appropriate through the processor. This realizes predetermined functions, processing units, etc. The processor is configured, for example, with semiconductor devices such as a CPU / MPU or GPU. Processing is not limited to software program processing, but can also be implemented using dedicated circuits. Dedicated circuits such as FPGAs, ASICs, and CPLDs can be used.

[0038] The program may be pre-installed as data on the target computer system, or may be distributed as data to the target computer system from a program source. The program source may be a program distribution server on a communication network or a non-transitory computer-readable storage medium, such as a memory card or disk. The program may be composed of multiple modules. The computer system may be composed of multiple devices. The computer system may be composed of a client-server system, a cloud computing system, an IoT system, etc. Various data and information may be composed of structures such as, but not limited to, tables and lists. Expressions such as identification information, identifiers, IDs, names, and numbers are interchangeable.

[0039] (First embodiment) 1 to 9, a simulation model creation device 100 according to a first embodiment of the present disclosure will be described. The simulation model creation device 100 is applied to computer systems such as personal computers, servers, tablet terminals, etc. Functions of the simulation model creation device 100 described below are realized, for example, by a control unit such as a CPU executing a program.

[0040] <Functional blocks of the simulation model creation device> The functions of the simulation model creation device 100 will be described with reference to Figures 1 and 2. Figure 1 is a diagram showing an example of functional blocks of the simulation model creation device 100. Figure 2 is a diagram showing an example of a simulation engine 25.

[0041] The simulation model creation device 100 is a device that creates a model by adjusting the parameters of a power module model 30 (described later), which is a physics-based model, in a circuit including the power module model 30, and uses the created model to simulate electrical characteristics, etc., of a plurality of terminals (which may also be referred to as ports) included in the power module model 30.

[0042] 1, the simulation model creation device 100 includes a model creation system 10 and a simulator unit 20. Here, the physics-based model is a model of a semiconductor device such as a power device. In this embodiment, a case where a SiC (silicon carbide) power MOSFET (metal-oxide-semiconductor field-effect transistor) is used as a model will be described, but the present invention is not limited to this.

[0043] The model creation system 10 changes the default values ​​of the parameters to be set in the power module model 30 and outputs the changed default values ​​to the simulator unit 20. The model creation system 10 has an electrical characteristic input unit 11, a model creation unit 12, and a model output unit 13.

[0044] The simulator unit 20 executes a simulation of the simulation engine 25. The simulator unit 20 has a default value input unit 21, a parameter perturbation input unit 22, a parameter adjustment unit 23, a simulator control unit 24, and a simulation engine 25. A power module model 30 is incorporated into the simulator unit 20. In this embodiment, the power module model 30 is incorporated into the simulation engine 25.

[0045] 2, the simulation engine 25 includes a power module model 30 and a circuit 40. The power module model 30 includes a JFET-based calculation unit 31 and a MOS-based calculation unit 32.

[0046] Each of the above-mentioned functions of the simulation model creation device 100 will be explained below.

[0047] The electrical characteristic input unit 11 inputs electrical characteristic data relating to a power device, i.e., a SiC power MOSFET. Here, the electrical characteristic data includes, for example, IV characteristics, capacitance characteristics, and switching waveforms. The input electrical characteristic data may be, for example, data obtained by evaluating an actual SiC power MOSFET procured as a sample and serving as the base of the power module model 30, values ​​on a specification sheet for the SiC power MOSFET, or data on the SiC power MOSFET provided by a vendor.

[0048] The model creation unit 12 calculates the characteristics of the power module model 30. More specifically, the model creation unit 12 changes the default values ​​of a plurality of parameters of the power module model 30 so as to reduce the difference between the electrical characteristics indicated by the electrical characteristic data received from the electrical characteristic input unit 11 and the electrical characteristics output by the power module model 30, which is a model for circuit simulation of a power device. The details of the plurality of parameters will be described later.

[0049] In this embodiment, the model creation unit 12 includes a parameter median calculation unit 12a. The parameter median calculation unit 12a calculates the median value of each of a plurality of parameters used in a simulation of the power module model 30 executed by the simulator unit 20. The parameter median calculation unit 12a, for example, compares the electrical characteristic data received from the electrical characteristic input unit 11 with the electrical characteristics calculated by the power module model 30, and changes the parameter values ​​of the power module model 30 based on a certain policy. Here, the certain policy is, for example, evaluating and minimizing the MAE (mean absolute error) between the electrical characteristic data and the electrical characteristics calculated by the power module model 30. The certain policy may also be a policy of evaluating the MSE (mean square error), or a policy of calculating and evaluating the error after taking a logarithm.

[0050] The model output unit 13 outputs the default values ​​changed by the model creation unit 12. In this embodiment, the model output unit 13 outputs the default values ​​of the power module model 30 calculated by the model creation unit 12 to the default value input unit 21. The model output unit 13 may output code used in a simulation or a program used in a simulation instead of the default values. Here, the code or program is, for example, information for operating the power module model 30 with the default values ​​changed by the model creation unit 12.

[0051] The default value input unit 21 inputs a parameter value that becomes the perturbation center of the power module model 30. In this embodiment, the default value output by the model output unit 13 is input to the default value input unit 21 as the parameter value that becomes the perturbation center. Alternatively, instead of the parameter value output by the model output unit 13, the user may manually set the parameter value that becomes the perturbation center, and input the parameter value to the default value input unit 21. In this case, for example, a predetermined interface may be displayed on a display unit (not shown) of a computer system, and the user may input the parameter value from the displayed interface.

[0052] The parameter perturbation input unit 22 receives the default value input from the model output unit 13 to the default value input unit 21, i.e., the perturbation range of the parameter value serving as the perturbation center. For example, using an input unit (not shown) of a computer system, a user inputs, as parameters to be used for perturbation, ranges (perturbation ranges) in which the parameter values ​​vary in relative or absolute values ​​from the parameter values ​​input to the default value input unit 21 for multiple parameters of the power module model 30.

[0053] Here, the multiple parameters are parameters that indicate the physical state or geometric state of the SiC power MOSFET. The parameters that indicate the physical state include, for example, a parameter that indicates the impurity concentration or defect density of a semiconductor region included in the SiC power MOSFET. Furthermore, the parameters that indicate the geometric state include at least one of the thickness of an insulating film included in the SiC power MOSFET, the dimensions of the semiconductor region, and the position of the semiconductor region.

[0054] More specifically, the parameters indicating the physical state include, for example, the MOS interface trap density (carrier amount) Ntrap, the MOS interface trap energy level Etrap, the MOS channel base mobility km, the MOS channel phonon scattering mobility km_sr, the cell pitch CP, the channel density Dch, the short channel coefficient lambda, the chip active area Aactive, and the drift layer minority carrier time constant tau_rec. More specifically, the parameters indicating other geometric states include, for example, the semiconductor substrate thickness Dsub, the semiconductor substrate concentration Nsub, the drift layer thickness Depi, the drift layer concentration Nepi, the JFET width Wjft, the JFET concentration Njft, the JFET length Ljft, and the gate insulating film thickness tox. The parameters indicating the physical state and the parameters indicating the geometric state are not limited to these.

[0055] On the other hand, SiC power MOSFET parameters that are not used for perturbation are parameters that are guaranteed to be constant in principle. If perturbations are applied to the values ​​of these parameters and a simulation is performed, there is a possibility that unnecessary margins will be calculated. Examples of parameters that are not used for perturbation include the bulk reference mobility muE and the phonon scattering order zeta. It is desirable to allow these values ​​to be changed dependently from other parameters using, for example, an equation for doping dependency. Note that, in the case of SiC power MOSFETs, channel mobility is highly process-dependent, so it is desirable to treat it as a parameter to be used for perturbation.

[0056] The perturbation range can be set to, for example, 10% or 0.3 μm for the dimension values ​​of the SiC power MOSFET, 20% for the concentration of the drift layer, and 10% for other concentrations. The SiC power MOSFET vendor itself can calculate the perturbation range based on its own production capacity. On the other hand, the SiC power MOSFET user can calculate using values ​​that are conventional perturbation ranges. If possible, it is desirable for the user to coordinate with the SiC power MOSFET vendor to set the perturbation range values.

[0057] The parameter adjusting unit 23 adjusts the values ​​of the parameters used in the simulation by the simulator unit 20 within the perturbation range input to the parameter perturbation input unit 22 from the default values, that is, the parameter values ​​that are the center of perturbation.

[0058] More specifically, the parameter adjustment unit 23 sets the default value received from the default value input unit 21 as the center of perturbation. Furthermore, the parameter adjustment unit 23 sets the perturbation range received from the parameter perturbation input unit 22 from the default value set as the center as the range in which the parameter value varies. Furthermore, the parameter adjustment unit 23 changes the parameter value to any value within the range in which the parameter value varies, and sets the changed parameter value in the power module model 30 of the simulation engine 25. The parameter adjustment unit 23 executes processing to set parameter values ​​for all parameters to be simulated by the simulation engine 25.

[0059] The simulation engine 25 is any software. The simulation engine 25 is executed upon receiving a command from the simulator control unit 24. The simulation engine 25 is a simulator that can execute a simulation of the power module model 30. The simulation engine 25 is, for example, SPICE, which is an electronic circuit simulator. The simulation engine 25 is not limited to SPICE.

[0060] <Configuration example of power module model> The power module model 30 includes a model calculation unit that calculates electrical characteristics based on input values ​​to multiple terminals. The model calculation unit may calculate the electrical characteristics based on defined parameters. As shown in FIG. 2, the power module model 30 includes a JFET-based calculation unit 31 and a MOS-based calculation unit 32 as model calculation units. The JFET-based calculation unit 31 corresponds to the JFET region of the SiC power MOSFET. The MOS-based calculation unit 32 corresponds to the MOSFET region of the SiC power MOSFET. The JFET region and the MOSFET region will be described in detail later.

[0061] The power module model 30 includes a plurality of terminals. The plurality of terminals includes a first terminal for calculating the electrical characteristics between the first electrode, the second electrode, and the third electrode, and a second terminal used for calculating the temperature characteristics. When simulating the electrical characteristics and the temperature characteristics between the first electrode, the second electrode, and the third electrode, the input values ​​to the first terminal for calculating the electrical characteristics between the first electrode, the second electrode, and the third electrode and the second terminal used for calculating the temperature characteristics are default parameter values. The first electrode, the second electrode, and the third electrode are, for example, a gate electrode, a source electrode, and a drain electrode.

[0062] As shown in Fig. 2, in this embodiment, the power module model 30 is configured by an equivalent JFET-equivalent MOSFET series circuit. The power module model 30 has a JFET-based calculation unit (JFET-based model) 31 and a MOS-based calculation unit (MOS-based model) 32. The JFET-based calculation unit 31 and the MOS-based calculation unit 32 are each set with parameter values ​​adjusted by the parameter adjustment unit 23. The JFET-based calculation unit 31 and the MOS-based calculation unit 32 will now be described.

[0063] The JFET-based arithmetic unit 31 has three first terminals. The three first terminals are a gate JFET-G (first electrode), a source JFET-S (second electrode), and a drain JFET-D (third electrode). The MOS-based arithmetic unit 32 has three first terminals. The three first terminals are a gate MOS-G (first electrode), a source MOS-S (second electrode), and a drain MOS-D (third electrode).

[0064] The source JFET-S and the drain MOS-D are connected. The drain JFET-D and the drain PM-D of the power module model 30 have equivalent electrical characteristics. The gate JFET-G and the source MOS-S are connected at a connection point P1. The electrical characteristics calculated at this connection point P1 are equivalent to the electrical characteristics of the source PM-S of the power module model 30. The gate MOS-G and the gate PM-G of the power module model 30 have equivalent electrical characteristics. In addition, the voltage between the drain MOS-D and the source MOS-S of the MOS-based calculation unit 32 is input as the voltage of the source terminal JFET-S and the gate JFET-G of the JFET-based calculation unit 31.

[0065] The JFET-based calculation unit 31 further has a temperature JFET-T, which is a second terminal for calculating the temperature. The MOS-based calculation unit 32 further has a temperature MOS-T, which is a second terminal for calculating the temperature. The temperature JFET-T and the temperature MOS-T are connected at a connection point P2. The temperature calculated at the connection point P2 is equivalent to the temperature PM-T of the power module model 30.

[0066] The first terminals of the power module model 30, namely, the gate PM-G (first electrode), the source PM-S (second electrode), and the drain PM-D (third electrode), and the second terminal, namely, the temperature PM-T, are each connected to the circuit 40. The circuit 40 may be any circuit that can be connected to the power module model 30.

[0067] <Example of simulation engine configuration> 3 is a diagram showing an example of a simulation engine 25a corresponding to a physics-based model in which four power module models 30a to 30d are connected to a circuit 40. Here, the physics-based model is a full-bridge inverter circuit. Each of the power module models 30a to 30d has the same configuration as the power module model 30.

[0068] 3, power module models 30a and 30b are connected in series to a power source V1. Power module models 30c and 30d are connected in series to the power source V1 so as to be in parallel with the power module models 30a and 30b. A smoothing capacitor C1 is connected to the power source V1.

[0069] The drains PM-D of the power module models 30a and 30c are each connected to the positive side of the power supply V1. The sources PM-S of the power module models 30a and 30c are each connected to the drains PM-D of the power module models 30b and 30d. The sources PM-S of the power module models 30b and 30d are each connected to the negative side of the power supply V1.

[0070] The gate PM-G of the power module model 30a is connected to a connection point P11a between the source PM-S of the power module model 30a and the drain PM-D of the power module model 30b via a load R1 and an inductance L1. The temperature PM-T of the power module model 30a is grounded via a load R2 and a power supply V2. The load R2 is, for example, a thermistor that detects temperature. The power module model 30c is configured similarly to the power module model 30a, except that the connection point P11a becomes a connection point P11c.

[0071] The gate PM-G of the power module model 30b is connected to a connection point P11b between the source PM-S of the power module model 30a and the negative side of the power supply V1 via a load R1 and an inductance L1. The power module model 30d is configured similarly to the power module model 30b, except that the connection point P11b becomes a connection point P11d.

[0072] A load R3 is also placed between the connection point P12 and the connection point P13. The connection point P12 is provided between the connection point P11a and the drain PM-D of the power module model 30b. The connection point P13 is provided between the connection point P11c and the drain PM-D of the power module model 30d.

[0073] <Simulation engine operation> Next, the operation of the simulation engine 25a shown in Fig. 3 will be described with reference to Fig. 4 and Fig. 5. Fig. 4 and Fig. 5 are diagrams each showing an example of the operation of the simulation engine 25a.

[0074] In Fig. 4, the flow of current I when the power module models 30a and 30d are ON and the power module models 30b and 30c are OFF is indicated by an arrow AW1 and a thick line. In Fig. 5, the flow of current I when the power module models 30a and 30d are OFF and the power module models 30b and 30c are ON is indicated by an arrow AW2 and a thick line. In this way, in the simulation engine 25a, the flow of current can be switched by switching the power module models 30a and 30d and the power module models 30c and 30b ON and OFF.

[0075] The simulator unit 20 can output the electrical characteristics between the gate PM-G, source PM-S, and drain PM-D of each of the power module models 30a to 30d, as well as the temperature of PM-T, in the simulation engine 25a. Furthermore, the simulator unit 20 may output the electrical characteristics between the gate JFET-G, source JFET-S, drain JFET-D, gate MOS-G, source MOS-S, and drain MOS-D of each of the power module models 30a to 30d, as well as the temperatures of JFET-T and MOS-T.

[0076] Next, the differences between the configuration of the SiC power MOSFET, which is the product based on the power module model 30, and the configuration of MOSFETs of other technologies will be described. Here, the MOSFETs of other technologies refer to Si power MOSFETs made of Si.

[0077] <Configuration of Si Power MOSFET> First, the configuration of the Si power MOSFET will be described. FIGS. 6 and 7 are diagrams showing an example of the configuration of the Si power MOSFET. In FIGS. 6 and 7, the Y direction indicates the vertical direction of the Si power MOSFETs 200A and 200B, and the X direction indicates the horizontal direction of the Si power MOSFETs 200A and 200B.

[0078] As shown in FIG. 6, in the Si power MOSFET 200A, a drift layer 220 is formed on the drain electrode 210. On the drift layer 220, body layers 230A and 230B, and a semiconductor layer 240 are formed. The semiconductor layer 240 is formed between the body layers 230A and 230B. A source layer 250 is formed in a part of one of the body layers 230A adjacent to the semiconductor layer 240. A source electrode 260 is formed so as to contact a part of the upper side of the source layer 250. A gate insulating film 270 is provided so as to cover the source electrode 260, another part of the source layer 250, and the semiconductor layer 240. Through the gate insulating film 270, a gate electrode 280 is formed on a part of the source layer 250, the body layers 230A and 230B, and the semiconductor layer 240.

[0079] The drift layer 220, the semiconductor layer 240, and the source layer 250 are, for example, semiconductor regions having N-type impurities. The impurity concentration of the N-type impurities is, for example, if the drift layer 220 has a concentration N, the semiconductor layer 240 has a concentration N + and the source layer 250 has a concentration N ++ The body layers 230A and 230B are, for example, semiconductor regions having P-type impurities. The impurity concentration of the P-type impurities in the body layers 230A and 230B is, for example, a concentration P.

[0080] Also, AR11 indicated by the dashed-dotted line frame in the figure represents the JFET region. The double-headed arrow shown in the figure indicates the JFET width W1, which is the width of the JFET region AR11. AR12 indicated by the dashed-dotted line frame in the figure represents the MOSFET region AR12. The dashed frame indicates the depletion layers DL11 and DL12 where no current flows when the drain electrode 210 and the source electrode 260 are energized. Current flows from the drain electrode 210 to the source electrode 260 through the semiconductor layer 240 where the depletion layers DL11 and DL12 are not formed.

[0081] As shown in FIG. 7, the Si power MOSFET 200B has the same configuration as the Si power MOSFET 200A, except that the widths of the depletion layers DL11 and DL12 in the left-right direction are different. That is, due to the solid difference of the Si power MOSFET, there is a difference in the width where the depletion layer is formed. If the widths of the depletion layers DL11 and DL12 shown in FIG. 6 are width W11 and width W12 respectively, the widths of the depletion layers DL21 and DL22 shown in FIG. 7 are width W21 (>W11) and width W22 (>W12) respectively.

[0082] <Configuration of SiC Power MOSFET> Next, the configuration of the SiC power MOSFET will be described. FIGS. 8 and 9 are diagrams showing an example of the configuration of the SiC power MOSFET. In FIGS. 8 and 9, the Y direction indicates the up-down direction of the SiC power MOSFET, and the X direction indicates the left-right direction of the SiC power MOSFET.

[0083] As shown in FIGS. 8 and 9, the configurations of the SiC power MOSFETs 300A and 300B are the same as those of the Si power MOSFET, except that the width of the semiconductor layer 240 in the left-right direction, that is, the JFET width W2, is narrower than the JFET width W1 of the Si power MOSFET.

[0084] The SiC power MOSFET 300A includes depletion layers DL31, DL32, a JFET region AR31, and a MOSFET region AR32. The SiC power MOSFET 300B includes depletion layers DL41, DL42, a JFET region AR41, and a MOSFET region AR42. The JFET regions AR31, AR41 respectively correspond to the JFET base arithmetic units 31 of the power module model 30. The MOSFET regions AR32, AR42 respectively correspond to the MOS base arithmetic units 32 of the power module model 30.

[0085] Comparing FIGS. 8 and 9, the widths in the X direction of the depletion layers DL31, DL3 of the SiC power MOSFET 300A and the depletion layers DL41, DL42 of the SiC power MOSFET 300B are different. That is, due to the solid difference of the SiC power MOSFET, a difference occurs in the width in which the depletion layer is formed. If the widths of the depletion layers DL31, DL32 shown in FIG. 8 are widths W31, W32 respectively, the widths of the depletion layers DL41, DL42 shown in FIG. 9 are widths W41 (>W31) and width W42 (>W32) respectively.

[0086] <Comparison between Si Power MOSFET and SiC Power MOSFET> Furthermore, comparing FIGS. 8 and 9 with FIGS. 6 and 7, the JFET widths W2 of the SiC power MOSFETs 300A, 300B shown in FIGS. 8 and 9 are narrower than the JFET widths W1 of the Si power MOSFETs 200A, 200B shown in FIGS. 6 and 7. Therefore, in the SiC power MOSFETs 300A, 300B, the influence of perturbation on the JFET width W2 becomes larger. On the other hand, in the case of the Si power MOSFETs 200A, 200B, the voltage applied to the source electrode 260 and the drain electrode 210 is small, that is, the driving electric field is small and the JFET width W1 is wide. Therefore, the inventor of the present application regarded the JFET regions AR11, AR21 of the Si power MOSFET as mere resistors and performed a simulation, but the accuracy of the simulation did not result in a bad result.

[0087] However, in the SiC power MOSFETs 300A and 300B, which exhibit higher withstand voltage performance than the Si power MOSFETs 200A and 200B, a large voltage is applied to the source electrode 360 ​​and the drain electrode 310, that is, they are driven in a high electric field, and therefore the electric field of the gate insulating film 370 is relatively large compared to the electric field of the gate insulating film 270 of the Si power MOSFETs 200A and 200B. For this reason, it becomes necessary to protect the SiC power MOSFETs 300A and 300B by narrowing the JFET width W2 compared to the JFET width W1, as shown in Figures 8 and 9.

[0088] Therefore, in the case of the SiC power MOSFETs 300A and 300B, the JFET regions AR31 and AR41 cannot be regarded as simple resistors, as in the case of the Si power MOSFETs 200A and 200B. In the case of the SiC power MOSFETs 300A and 300B, if a simulation is performed by regarding the JFET regions AR31 and AR41 as simple resistors, this leads to a decrease in the accuracy of the simulation.

[0089] The inventors of the present application have found, through their many years of experience in developing power devices using wide bandgap materials (hereinafter referred to as "WBG"), that when designing a power device with a narrow JFET width W2, such as the SiC power MOSFET 300A shown in Fig. 8, if the resistance value of the MOSFET region AR32 (MOS channel) increases, the resistance value of the JFET region AR31 also increases at the same time. The inventors of the present application also found that this phenomenon is caused by the application of an effective reverse gate bias to the JFET region AR31 when the voltage across the MOSFET region AR32 increases.

[0090] Therefore, as shown in FIG. 2, the inventor of the present application connected the source JFET-S of the JFET-based arithmetic unit 31 to the drain MOS-D of the MOS-based arithmetic unit 32, and connected the gate JFET-G of the JFET-based arithmetic unit 31 to the source MOS-S of the MOS-based arithmetic unit 32, thereby forming an equivalent circuit configuration of the power module model 30. As a result, the inventor of the present application succeeded in reproducing, for example, in the SiC power MOSFET 300A shown in FIG. 8, a configuration in which an increase in the resistance of the MOSFET region AR32 is propagated to the JFET region AR31. This device is necessary to reproduce, by simulation, manufacturing variations caused by the JFET region AR31, which are specific to the case where SiC is newly used as a material.

[0091] For example, in the SiC power MOSFET 300A in which the JFET region AR31 shown in FIG. 8 is formed, the electrical characteristics around the JFET region AR31 can be incorporated into the simulation engine 25 as parameters. Here, the parameters are, for example, parameters that have a dominant influence on the feedback capacitance Cgd, the short-circuit saturation current Isat, and the breakdown voltage Vbd.

[0092] Also, generally, in a power device, the JFET region is vertical. This is because the JFET region is formed by implanting impurities into the body region through a mask by photolithography. Therefore, in the JFET region, variations due to the accuracy of the photolithography process, that is, variations in the final electrical characteristics of the actual product occur. Therefore, from the viewpoint of estimating variations in electrical characteristics, which is one of the problems that the technology of the present disclosure attempts to solve, incorporating terminals equivalent to the electrical characteristics of the JFET region into the power module model 30 is an important factor.

[0093] <Detailed description of SiC power MOSFET> The simulation model creation device 100 using the power module model 30 of the present disclosure requires a product whose parameters can be explicitly estimated, for example, SiC power MOSFETs 300A and 300B shown in FIGS. 8 and 9. The model creation system 10 estimates parameters from electrical characteristics. Therefore, an accurate physics-based model is required for the model creation system 10 to accurately estimate parameters. Below, details of the SiC power MOSFET 300A shown in FIG. 8, which is used as an example of a product in this embodiment, are described.

[0094] First, the background to the lack of a sufficient number of physics-based models using WBG, such as the SiC power MOSFET 300A, will be explained. This is because, for example, there is a difference in design concept between the Si power MOSFET 200A shown in Fig. 6 and the SiC power MOSFET 300A using WBG.

[0095] WBG generally has a breakdown field characteristic several to several tens of times higher than that of Si. This characteristic enables a low-resistance design that allows for a high doping concentration and thin thickness in the semiconductor impurity region. This is the most important motivation for adopting WBG as a power device material. However, the high electric field that accompanies such a low-resistance design poses extremely severe conditions for materials other than semiconductors, such as gate insulating films.

[0096] In the Si power MOSFET 200A shown in FIG. 6, since the electric field to be handled is small to begin with, electric field relaxation of the gate insulating film 270 is not necessary, and the JFET width W1 of the JFET region AR11 can be widened. On the other hand, in the SiC power MOSFET 300A shown in FIG. 8, the JFET width W2 needs to be narrowed to relax the electric field of the gate insulating film 370. In the Si power MOSFET 200A, the JFET width W1 is, for example, 10 μm. On the other hand, in the SiC power MOSFET 300A, the JFET width W2 is, for example, 1 μm.

[0097] 8, when a perturbation occurs in the parameters and the voltage across the MOSFET region AR32 increases, the voltage between the gate electrode 380 and the source electrode 360 ​​in the equivalent JFET region AR31 increases to the negative side. At this time, physically, the depletion layers DL31 and DL32 in the JFET region AR31 extend in the X direction. As a result, the width of the current path for the current flowing from the drain electrode 310 to the source electrode 360 ​​decreases.

[0098] In contrast, in the case of the Si power MOSFET 200A using Si, even if the depletion layers DL11 and DL12 extend in the X direction, as described above, the current path width of the JFET region AR11 is sufficiently wide due to the design features, so there is almost no effect on the resistance value. Therefore, in the Si power MOSFET 200A, there is a rational reason to regard the JFET region AR11 as a simple resistor, and there was no need to make the electrical characteristics of the MOSFET region AR12 and the JFET region AR11 equivalent. In other words, in the Si power MOSFET 200A, there was no element that could conceive the technical idea of ​​making the electrical characteristics of the MOSFET region AR12 and the electrical characteristics of the JFET region AR11 equivalent.

[0099] On the other hand, in the case of the SiC power MOSFET 300A, the original JFET width W2 is narrow. Therefore, even a slight perturbation can cause the on / off state of the equivalent JFET region AR31 to fluctuate. Therefore, for the SiC power MOSFET 300A, it becomes necessary to develop a new power module model incorporating an equivalent JFET.

[0100] Therefore, in this disclosure, a model has been developed that can fully incorporate the effect of this perturbation by arranging an equivalent JFET region AR31 and an equivalent MOSFET region AR32 in series, and connecting the drain MOS-D of the equivalent MOSFET to the source JFET-S of the equivalent JFET, and connecting the body of the equivalent MOSFET, i.e., the source MOS-S, to the gate JFET-G of the equivalent JFET. This model is the power module model 30 already described with reference to Figure 2.

[0101] Furthermore, in the power module model 30 of this embodiment, parameters of interface defects, which are particularly important when considering channel perturbations of the SiC power MOSFET 300A, can also be introduced. These parameters are, for example, energy Etrap and carrier amount Ntrap. When the temperature of the SiC power MOSFET 300A is increased from room temperature to a high temperature of about 150°C, the threshold value decreases by about 1 V. By introducing the energy Etrap and carrier amount Ntrap as parameters into the power module model 30, the simulation engine 25 can reproduce the effect of the threshold value decreasing by about 1 V.

[0102] Next, the specific operation of the simulation engine 25 will be described. The simulation engine 25 assumes that the amount of trapped carriers is Ntrap × (1-exp(-qEtrap / kT)) and performs a process of multiplying the channel mobility by the ratio of the amount of charge remaining after subtracting this amount of trapped carriers from the total inversion charge to the original total inversion charge. This process allows the simulation engine 25 to reproduce the phenomenon in which a large amount of charge is trapped at low temperatures, increasing the threshold, and almost no charge is trapped at high temperatures. In this way, by introducing the power module model 30 including carrier trapping parameters, the simulation model creation device 100 can accurately determine the threshold of the interface defects in the MOSFET region AR32. This allows the user to determine the appropriate perturbation range to be set.

[0103] Furthermore, in the power module model 30, a parameter is introduced that indicates the temperature dependence of bulk mobility so that it is proportional to T^(-ζ). This parameter indicates phonon scattering of semiconductor bulk mobility. This temperature dependence of bulk mobility can be implemented so that it affects all resistance components in the semiconductor. For example, in FIG. 2, the temperature PM-T of the power module model 30 is calculated by adding the temperature JFET-T of the JFET-based calculation unit 31 and the temperature MOS-T of the MOS-based calculation unit 32.

[0104] The simulation model creation device 100 can incorporate, for example, a JFET-based calculation unit 31 corresponding to an equivalent JFET region that has only a minor effect in a conventional material such as Si but has a significant effect in a SiC power MOSFET using WBG, into the power module model 30. This enables the simulation model creation device 100 to accurately calculate the effect that perturbations in the parameters of the MOS-based calculation unit 32 corresponding to the equivalent MOSFET have on the final performance of the SiC power MOSFET.

[0105] Furthermore, the simulation model creation device 100 can introduce parameters relating to interface defects and temperature dependency, which are particularly important for the SiC power MOSFET 300A. This enables the simulation model creation device 100 to estimate accurate values ​​of the parameters to be input to the default value input unit 21 from the measurement data input from the electrical characteristic input unit 11.

[0106] Furthermore, in the simulation model creation device 100, the parameter median calculation unit 12a first calculates a parameter value (default value) for each parameter based on the electrical characteristics input from the electrical characteristic input unit 11 in the model creation system 10, and inputs the calculated default values ​​to the default value input unit 21. The simulator unit 20 executes a simulation using the input default values ​​as the median value of perturbation. This makes it possible to obtain statistically valid results for the electrical characteristics of the terminals in the power module model 30, such as the gate PM-G, drain PM-D, and source PM-S, and the temperature PM-T.

[0107] According to this method, even in the case where a user who does not know the internal design of the SiC power MOSFET 300A performs a simulation, the simulation model creation device 100 can perform a simulation in which the JFET-based calculation unit 31 and the MOS-based calculation unit 32 have equivalent electrical characteristics, and obtain parameter values ​​of multiple parameters as the simulation results.

[0108] Furthermore, by adding known perturbations to the parameter values ​​thus obtained, the simulation model creation device 100 can, for example, reproduce in a simulation a SiC power MOSFET with worst-case electrical characteristics. Conventionally, users had to search for a worst-case SiC power MOSFET product, which appears with an extremely low probability, from among many other products. However, the simulation model creation device 100 can estimate the worst-case SiC power MOSFET based on the electrical characteristics of a single SiC power MOSFET product. This allows the simulation model creation device 100 to have a significant advantage in terms of cost.

[0109] (Second embodiment) The simulation model creation device 100 of the second embodiment simulates risks that may occur in the final product, a SiC power MOSFET, in addition to the simulation of the first embodiment. The simulation model creation device 100 of the present embodiment samples parameters using a specified distribution function from within the range of perturbation given to the parameter perturbation input unit 22, applies the parameters to the power module model 30, and repeats the operation of executing the simulation.

[0110] <Simulator functional block> Fig. 10 is a diagram showing an example of functional blocks of the simulator unit 20 included in the simulation model creation device 100 of this embodiment. As shown in Fig. 10, the simulator unit 20 further includes a simulation result analysis unit 26 and a result output unit 27 in addition to a default value input unit 21, a parameter perturbation input unit 22, a parameter adjustment unit 23, and a simulation engine 25. Note that the simulator control unit 24 and the model creation system 10 are omitted from Fig. 10.

[0111] In the simulator unit 20, the process up to the process in which a reference parameter value (default value) is introduced from the default value input unit 21 is the same as the process in the simulator unit 20 of the first embodiment. In this embodiment, instead of the range of parameter perturbation, information on a distribution function indicating the distribution of perturbation values ​​(hereinafter referred to as "perturbation information") is input from the parameter perturbation input unit 22 to the parameter adjustment unit 23. Here, the perturbation information is information indicating feature quantities of the distribution function, such as the distribution shape (Gaussian, polynomial, etc.), variance, and mean value, for example.

[0112] Parameter adjustment unit 23 samples parameter perturbation values ​​to be implemented in each trial (iter in FIG. 10 ) based on the probability indicated by the distribution function from the perturbation information input from parameter perturbation input unit 22. Here, parameter adjustment unit 23 may perform sampling by controlling pseudo-random numbers using a seed value. Parameter adjustment unit 23 adds the perturbation values ​​sampled for each parameter to the default values ​​for each parameter input from default value input unit 21, and inputs the results to power module model 30.

[0113] The simulation result analysis unit 26 analyzes the results of the simulation performed by the simulator unit 20 based on the parameter values ​​adjusted by the parameter adjustment unit 23.

[0114] More specifically, the simulation result analysis unit 26 assumes a distribution for each parameter, performs random sampling, and repeats the simulation to accumulate the simulation results in the simulation result analysis unit 26. In this way, the simulation result analysis unit 26 acquires a probability distribution of the final product performance of the SiC power MOSFET. The simulation result analysis unit 26 may acquire statistical results instead of the probability distribution.

[0115] The result output unit 27 outputs the probability distribution of the simulation results analyzed by the simulation result analysis unit 26 or characteristic information that characterizes the statistical results.

[0116] More specifically, the result output unit 27 outputs feature information characterizing the probability distribution of the product performance acquired by the simulation result analysis unit 26, such as statistical quantities characterizing the probability distribution, such as the maximum value, minimum value, standard deviation, and average. The probability distribution may be output together with the feature values. Similarly, when statistical results are obtained, the result output unit 27 outputs the feature values ​​obtained from the statistical results, or the feature values ​​and the statistical results. The output destination is, for example, a display unit of a computer system to which the simulation model creation device 100 is applied. Alternatively, the output destination may be a storage unit of the computer system.

[0117] The simulation model creation device 100 of this embodiment can simulate the electrical characteristics and temperature characteristics of a SiC power MOSFET, perform sampling by assuming a distribution for parameters that cause variations, and output a statistical distribution of the characteristic variations. By visually checking this output result, the user can estimate the cost of the SiC power MOSFET, which is the final product. In other words, by determining the limiting characteristics of the SiC power MOSFET, the user can determine what proportion of all the products to be manufactured can be manufactured as non-defective products. This allows the user to quantitatively grasp the cost of manufacturing a SiC power MOSFET.

[0118] <Comparison with other methods for determining costs> Next, we will explain the effect of this method compared with other methods for determining costs. Here, the other methods are methods that calculate yields by assuming that the characteristics of SiC power MOSFETs themselves follow a normal distribution or the like. This method results in large deviations from the distribution, particularly for devices such as SiC power MOSFETs, where the JFET region is important. Specifically, when the JFET width Wjft becomes narrower than a certain width, current no longer flows, resulting in a sudden increase in resistance. In this case, the distribution clearly does not follow a normal distribution and has a long distribution on the high resistance side. In contrast, the simulation model creation device 100 of this embodiment can reproduce a distribution that is long on the high resistance side via the power module model 30 by, for example, assuming a distribution on the JFET width Wjft side.

[0119] Furthermore, when the perturbation range of the parameters is clear, parameter-adjusting unit 23 may sample the endpoints of the perturbation range by linear programming. In this case, the number of calculations is small, so that simulation model creation device 100 can quickly obtain worst-case results for the SiC power MOSFET.

[0120] <Comparison with other methods for controlling variability> Next, we will explain a comparison with other methods for managing the variation in the perturbation values ​​of sampled parameters. Here, the other methods are methods that estimate the distribution and yield by measuring, for example, the on-state voltage Vdson of samples of products that have actually been mass-produced. This method has the disadvantages of being high cost because it is assumed that the products will be mass-produced as samples, and not being able to know the distribution and yield until actual mass production is carried out.

[0121] In contrast, the simulation model creation device 100 of this embodiment utilizes the distribution of each manufacturing process that causes variation, such as film thickness inspection data for an insulating film deposition process, to enable detailed distribution acquisition at low cost before mass production. Such a method using the distribution of manufacturing processes requires a physics-based model that estimates how the parameters of the manufacturing processes affect the characteristics of the final product. In this embodiment, the physics-based model is realized by a simulation engine 25 that includes a power module model 30.

[0122] (Third embodiment) The third embodiment adds a model calculation unit selection system for effectively utilizing the power module model 30 used in the first and second embodiments. The model calculation unit selection system is included in the simulation model creation device 100. By having the model calculation unit selection system, the simulation model creation device 100 can effectively manage and operate the power module model 30.

[0123] <Function block of model calculation unit selection system> 11 is a diagram showing an example of functional blocks of the model computation unit selection system 45. As shown in FIG. 11, the model computation unit selection system 45 has a model name DB (database) 50 and a model selection UI (user interface) 60.

[0124] The model name DB 50 is configured with a storage unit such as an HDD (hard disk drive). The model name DB 50 stores information on a plurality of model calculation units of different types. In this embodiment, the model name DB 50 stores the correspondence between the model calculation unit version name and the model calculation unit hash value.

[0125] The model selection UI 60 is configured with a display unit such as a display, for example. When some or all of multiple model calculation units are selected, the model selection UI 60 displays information that distinguishes the selected multiple model calculation units from one another. The model selection UI 60 displays a model name 61, a parameter set name 62, a model calculation unit hash value 63, a model calculation unit version name 64, and a waveform comparison sample 65. The displayed model calculation units are distinguished from one another by the model calculation unit hash value 63 and the model calculation unit version name 64. The information displayed on the model selection UI 60 may be stored in association with the model name in the model name DB 50, for example, or may be stored in association with the model name in another database.

[0126] The model calculation units M1 and M2 are incorporated into the power module model 30 for use. The model calculation units M1 and M2 identified by the model calculation unit version name 64 are, for example, the JFET-based calculation unit 31 and the MOS-based calculation unit 32 described above. For example, the model calculation unit M1 is provided with four terminals, namely, a gate MOS-G1, a source MOS-S1, a drain MOS-D1, and a temperature MOS-T1, and can be incorporated into the simulation engine 25 as an element. In this embodiment, a case is illustrated in which the MOS-based calculation unit 32 has two versions (MOS_V1, MOS_V2). The JFET-based calculation unit 31 is not shown in the figure.

[0127] In the model selection UI 60, for example, a model name 61, "Model A," a parameter set name 62, "Product A," a model calculation unit hash value 63, "d3f4ae42faffcb3," a model calculation unit version name 64, "MOS_V1," and a waveform comparison sample 65, "Available," are displayed in association with each other. By visually checking the display of the model selection UI 60, the user can recognize information specifying the model calculation unit, such as the model calculation unit hash value 63 and the model calculation unit version name 64 in the example of FIG. 11, and identify the model calculation unit. In addition to the display shown in FIG. 11, the model selection UI 60 may also display information such as the corresponding product name, a description field, and the provider of the parameters. The model selection UI 60 may also display data indicating the electrical characteristics used by the model creation unit 12 when calculating default values ​​for the parameters of the model calculation unit.

[0128] The model calculation unit hash value 63 is, for example, the hash value of the source code or the compiled program. The hash value will be different even if a small modification is made to the source code. Therefore, by visually checking the model calculation unit hash value 63, the user can clearly distinguish between model calculation units such as model calculation units M1 and M2. For example, when a small modification is made, even if the model calculation unit version name is not changed, the model calculation unit hash values ​​before and after the modification are different, so the user can visually recognize that they are different model calculation units.

[0129] The model computation unit version names 64 of the model computation units M1, M2, etc. may be assigned manually by the user, but the model computation unit version names 64 alone do not, in principle, guarantee the identity of the model computation units. For this reason, the model computation unit selection system 45 may search the model name DB 50 using the model computation unit hash value 63, and automatically obtain from the model name DB 50 the model computation unit version name that matches the model computation unit hash value 63, and display it on the model selection UI 60. In this way, the model computation unit selection system 45 can ensure the identity of the model computation units M1, M2 and also promote linguistic understanding by the user.

[0130] <Effects of the model calculation unit selection system> The simulation model creation device 100 including the model calculation unit selection system 45 can display information about the model calculation units M1 and M2 in the model selection UI 60 so that the model calculation units M1 and M2 used in the power module model 30 can be easily compared.

[0131] Currently, each vendor creates its own physics-based model for SiC power MOSFETs and provides it to users. Therefore, users of SiC power MOSFETs may incorporate the physics-based model provided by each vendor into the simulation engine 25 as a power module model 30 and compare the products of each vendor in a simulation.

[0132] However, there is no guarantee that the physics-based models provided by each vendor have been created using the same methodology. As a result, the accuracy and operating conditions of the physics-based models provided by each vendor are not uniform. Therefore, for example, even if a product from vendor A is superior to a product from vendor B in a simulation, it is entirely possible that the relationship between the two products is reversed when the actual measured values ​​of the products are compared.

[0133] Furthermore, even for the same product, there may be multiple physics-based models. If the physics-based models incorporated into the simulation engine 25 as the power module model 30 are not the same, the simulation cannot be reproduced. For example, in FIG. 11, two model calculation units, Model C1 and Model C2, are provided for "Product C." Therefore, if a simulation is performed using these two model calculation units, the results of the two simulations will not be the same. In this way, if different model calculation units are used, it becomes meaningless to compare the results of the simulations.

[0134] Therefore, the simulation model creation device 100 displays information indicating the identity of the model calculation units in the model selection UI 60. In this embodiment, for example, as shown in FIG. 11, Model A, Model B, and Model C1 have the same model calculation unit hash value 63 of "d3f4ae42faffcb3." Furthermore, Model C2 and Model D have the same model calculation unit hash value 63 of "75653e0977b82." In this way, the display of the model calculation unit hash value 63 indicates the identity when the same model calculation unit is used.

[0135] Furthermore, referring to the parameter set name 62, Model A, Model B, and Model C are applied to Product A, Product B, and Product C, respectively. Therefore, the simulation model creation device 100 makes it possible to standardize and compare the accuracy and operating conditions of products from different vendors using the same model calculation unit (model calculation unit version name MOS_V1). In this case, it is necessary to be able to detect even if the implementation of the model calculation unit is falsified. In this embodiment, the identity of the model calculation unit can be guaranteed by visually checking whether the model calculation unit hash value 63 is the same. Therefore, in principle, it is possible to prevent the model calculation unit from being falsified. A user can select models having the same model calculation unit, run a simulation, and compare the electrical characteristics and temperature characteristics that are the execution results, thereby making it possible to perform a highly reliable comparison.

[0136] <Other display examples> FIG. 12 is a diagram showing another example of the display of the model selection UI 60. The model selection UI 60 displays default values ​​66 of multiple parameters of the model calculation units such as the model calculation units M1 and M2. Here, the default values ​​are the default values ​​of the multiple parameters set by the model creation system 10.

[0137] 12, in addition to the display shown in FIG. 11, the model selection UI 60 displays default values ​​for multiple parameters X11 and X12 in association with a model name 61, etc. Furthermore, the user can visually check default values ​​66 for all parameters by sliding a slide bar 67. This allows the user to check the default values ​​of all parameters set in the model calculation units M1, M2, etc. The default values ​​66 for all parameters may be stored in association with the model name in the model name DB 50, for example, or may be stored in association with the model name in another database.

[0138] (Fourth embodiment) The fourth embodiment adds a model / simulation condition selection system that allows a user to select a model and simulation conditions using natural language. The model / simulation condition selection system is included in the simulation model creation device 100.

[0139] <Functional blocks of the model simulation condition selection system> 13 is a diagram showing an example of functional blocks of a model / simulation condition selection system 70. The model / simulation condition selection system 70 has a platform language input unit 71, a platform unit 72, a platform execution unit 75, a platform calculation result output unit 76, and a platform display unit 90. The model / simulation condition selection system 70 is included in a simulation model creation device 100. In FIG. 13, the model / simulation condition selection system 70 and the simulator unit 20 are illustrated, but the model creation system 10 is not illustrated.

[0140] The platform language input unit 71 accepts input in a natural language and inputs, for example, text input by a user in a predetermined input format to the platform unit 72 as text data.

[0141] The platform unit 72 determines simulation conditions from the natural language input to the platform language input unit 71 .

[0142] The platform unit 72 has LLMs (Large Language Models) 73 and a platform condition determination unit 74. The LLMs 73 are configured using text data and deep learning. The LLMs 73 create queries based on the natural language input from the platform language input unit 71, i.e., the text data. The platform condition determination unit 74 is configured by a program or database that determines the simulation conditions described above from the query. The platform condition determination unit 74 determines the simulation conditions based on the query created by the LLMs 73. The LLMs 73 and the platform condition determination unit 74 are described in detail below.

[0143] The LLM 73 accepts input in natural language from the platform language input unit 71 or responds in natural language to the platform language input unit 71. For example, the LLM 73 creates a query from text data input from the platform language input unit 71 in accordance with a platform specified in the platform condition determination unit 74, and inputs the created query to the platform condition determination unit 74. If the text data cannot be converted into a query to be input to the platform condition determination unit 74, the LLM 73 may, for example, display content on the platform language input unit 71 instructing the user to make an appropriate input. As a result, a query for determining simulation conditions is input from the LLM 73 to the platform condition determination unit 74.

[0144] The platform condition determination unit 74 is realized, for example, by the control unit executing a condition inference program. The platform condition determination unit 74 determines the simulation conditions when a simulation is executed by the simulator unit 20 based on a query input from the LLM 73. Here, the simulation conditions include, for example, a model calculation unit, parameter values ​​(default values) of multiple parameters, a repetition method, and a sampling method. The platform condition determination unit 74 inputs the determined simulation conditions to the platform execution unit 75.

[0145] The platform execution unit 75 outputs an instruction to the simulator unit 20 to execute a simulation based on the simulation conditions determined by the platform condition determination unit 74 and the parameter values ​​(default values) of multiple parameters set in the determined model calculation unit.

[0146] The simulator unit 20 executes the simulation described in the above embodiment in accordance with the input simulation conditions and parameter values ​​(default values) of a plurality of parameters. The simulator unit 20 outputs the execution result of the simulation to the platform execution unit 75.

[0147] The platform calculation result output unit 76 has a query creation unit 77, an LLM 78, a model database (MDB) 81, and a default value database (DDB) 82. The model database 81 stores information on multiple model calculation units of different types. The default value database 82 stores default values ​​for multiple parameters of the model calculation units. The model database 81 and the default value database 82 may be common to, for example, the model name database 50 described above.

[0148] The platform calculation result output unit 76 outputs the simulation execution results of the simulator unit 20 in numerical values ​​or natural language. The platform calculation result output unit 76 further searches for a model calculation unit to execute the simulation or default values ​​of parameters based on the execution results of the simulator unit 20, the model database 81, and the default value database 82, and outputs one or more model calculation units or default values. For example, the platform calculation result output unit 76 may output a model calculation unit whose performance is closest to the model calculation unit used in the power module model 30, or default values ​​of multiple other parameters that are close to the parameter values ​​of multiple parameters used as default values.

[0149] The query creation unit 77 creates a query based on the input simulation execution results. Here, the query is, for example, an answer template. The query creation unit 77 inputs the created query to the LLM 78.

[0150] When a query is input, LLM78 outputs an explanation of the simulation results and instructions for the user's actions based on the query. This allows the user to check the simulation results. Also, if the model calculation section or default values ​​are output, the user can use the output contents as a reference for the next simulation.

[0151] Here, we will explain the challenges of simulating a physics-based power module model. Simulation requires a large amount of advanced know-how, such as the meaning and influence of parameters, guidelines for setting parameters, and interpretation of errors resulting from the method of acquiring the actual measured data to be compared (measurement method). Therefore, while anyone can run a simulation, feeding the results back into actual design and maintenance requires advanced knowledge and experience.

[0152] However, such extensive and sophisticated know-how is only available to simulation vendors and engineers who create power module models, and users who actually use the power module models often lack such sophisticated know-how. As a result, there are limited cases where highly accurate simulations can be effectively implemented.

[0153] Therefore, in the simulation model creation device 100 of this embodiment, a system is constructed that quantifies the user's objectives and determines the optimal conditions using the LLM 73 and LLM 78. As a result, the simulation model creation device 100 allows even a user without advanced know-how to effectively operate the simulation of the power module model 30.

[0154] Furthermore, in this embodiment, the simulation model creation device 100 does not directly determine simulation conditions using the LLM 73. The LLM 73 is responsible only for generating queries, and the platform condition determination unit 74 determines simulation conditions. This configuration, which separates functions, plays an important role in the model / simulation condition selection system 70. Generally, an LLM only performs statistical inference on text data (sentences). Therefore, mathematical accuracy cannot be expected from the output results of the LLM. Therefore, the model / simulation condition selection system 70 has the LLM 73 create a query in a specific format, i.e., a query fixed to the platform of the platform condition determination unit 74. Based on the created query, the platform condition determination unit 74 determines accurate simulation conditions using a condition inference program, thereby enabling the model / simulation condition selection system 70 to achieve both accuracy and usability.

[0155] Therefore, the simulation model creation device 100 including the model / simulation condition selection system 70 of this embodiment allows even users with little know-how to operate a simulation effectively through a natural language interface. Furthermore, the model / simulation condition selection system 70 is internally configured to use only the LLM 73 for query generation, thereby improving the accuracy of the simulation conditions to be set.

[0156] Furthermore, the model / simulation condition selection system 70 can create model calculation units for SiC power MOSFETs manufactured by multiple different vendors based on the simulation conditions. The multiple model calculation units created in this way are stored in, for example, a model database 81 and a default value database 82.

[0157] The platform display unit 90 can provide selectable model calculation units and default values ​​for SiC power MOSFETs manufactured by multiple different vendors on a common platform. That is, the user can use the platform display unit 90 to display and freely select the model calculation units and default values.

[0158] <Modifications of the power module model> Next, a description will be given of a modified example of the power module model 30. The power module model 30A is a modified example of the power module model 30.

[0159] Fig. 14 is a diagram showing an example of a power module model 30A. As shown in Fig. 14, a plurality of terminals included in the power module model 30A are provided separately as terminals for calculating voltage characteristics and terminals for calculating current characteristics.

[0160] In the power module model 30A, the gate PM-GI, source PM-SI, and drain PM-DI are provided as terminals for calculating current, and the gate PM-GV, source PM-SV, and drain PM-DV are provided as terminals for calculating voltage.

[0161] In the JFET-based calculation unit 31, the gate JFET-GI, source JFET-SI, and drain JFET-DI are provided as terminals for calculating current, and the gate JFET-GV, source JFET-SV, and drain JFET-DV are provided as terminals for calculating voltage.

[0162] In the MOS-based calculation unit 32, a gate MOS-GI, a source MOS-SI, and a drain MOS-DI are provided as terminals for calculating current, and a gate MOS-GV, a source MOS-SV, and a drain MOS-DV are provided as terminals for calculating voltage.

[0163] In this way, by providing separate terminals for calculating the voltage characteristics and the current characteristics in the power module model 30A, the simulator unit 20 can perform simulations with higher accuracy.

[0164] In the above embodiments, the terminals provided on the power module model 30 are described as being gate, source, and drain terminals. However, depending on the physics-based model that serves as the basis for the power module model 30, the terminals provided on the power module model 30 may be terminals of a base electrode, an emitter electrode, and a collector electrode.

[0165] <Other variations> Although the embodiments of the present disclosure have been specifically described above, they are not limited to the above-described embodiments and various modifications are possible without departing from the spirit of the present disclosure. In each embodiment, components can be added, deleted, or replaced, except for essential components. Unless otherwise specified, each component may be singular or plural. Combinations of the embodiments and their modifications are also possible. Some or all of the above-described configurations, functions, processing units, etc. may be realized by hardware, such as an integrated circuit design, or by software in which a processor interprets and executes a program. Data and information such as programs, tables, and files that realize each function can be stored in a storage device such as a memory, hard disk, or SSD, or on a storage medium such as an IC card, SD card, or DVD. [Explanation of symbols]

[0166] 10...Model creation system, 11...Electrical characteristic input section, 12...Model creation section, 12a...Parameter median calculation section, 13...Model output section, 20...Simulator section, 21...Default value input section, 22...Parameter perturbation input section, 23...Parameter adjustment section, 24...Simulator control section, 25...Simulation engine, 26...Simulation result analysis section, 27...Result output section, 30...Power module model, 31...JFET-based calculation section, 32...MOS-based calculation section, 40...Circuit, 45...Model calculation section selection system, 50...Model name database, 60...Model selection user interface, 61...Model name, 62...Parameter dataset name, 63...model calculation unit hash value, 64...model calculation unit version name, 65...waveform comparison sample, 66...default value, 70...model simulation condition selection system, 71...platform language input unit, 72...platform unit, 73,78...LLM, 74...platform condition determination unit, 75...platform execution unit, 76...platform calculation result output unit, 77...query creation unit, 81...model database, 82...default value database, 100...simulation model creation device, 200A,200B...Si power MOSFET, 300A,300B...SiC power MOSFET

Claims

1. an input unit for inputting electrical characteristic data relating to a power device; a model creation unit that changes default values ​​of a plurality of parameters of the model so as to reduce a difference between the electrical characteristics indicated by the electrical characteristic data received from the input unit and the electrical characteristics output by the model of the simulation of the power device; an output unit that outputs the default values ​​changed by the model creation unit; Equipped with the plurality of parameters include a parameter indicating a physical state or a geometric state of the power device; The model is a model calculation unit that calculates electrical characteristics based on input values ​​to a plurality of terminals; the plurality of terminals include a plurality of first terminals corresponding to first electrodes, second electrodes, and third electrodes of the power device, and a second terminal used for calculating temperature characteristics; When simulating the electrical characteristics between the first electrode, the second electrode, and the third electrode, and the temperature characteristics, the input values ​​to be input to first terminals corresponding to the first electrode, the second electrode, and the third electrode, and to second terminals used in calculating the temperature characteristics, use default values ​​after the parameters are changed. Simulation model creation device.

2. 2. The simulation model creation device according to claim 1, the parameter indicating the physical state includes a parameter indicating an impurity concentration or a defect density of a semiconductor region included in the power device, the parameters indicating the geometric state include at least one of a thickness of an insulating film included in the power device, a size of a semiconductor region, and a position of the semiconductor region; Simulation model creation device.

3. 2. The simulation model creation device according to claim 1, The power device is a MOSFET region and a JFET region, the model includes a MOS-based model corresponding to the MOSFET region and a JFET-based model corresponding to the JFET region; the plurality of first terminals corresponding to the first electrode, the second electrode, and the third electrode of the MOS base model are a gate terminal, a source terminal, and a drain terminal; the plurality of first terminals corresponding to the first electrode, the second electrode, and the third electrode of the JFET-based model are a gate terminal, a source terminal, and a drain terminal; a voltage between the drain terminal and the source terminal of the MOS-based model is input as a voltage between the source terminal and the gate terminal of the JFET-based model; Simulation model creation device.

4. 2. The simulation model creation device according to claim 1, Further comprising a simulator unit in which the model is incorporated, The simulator unit a default value input section into which the changed default value is input; Simulation model creation device.

5. 5. The simulation model creation device according to claim 4, a perturbation input unit to which a perturbation range of the default value is input; a parameter adjusting unit that adjusts parameter values ​​for the simulator unit to execute a simulation from the default values ​​within a perturbation range input to the perturbation input unit; a simulation result analysis unit that analyzes a simulation result obtained by the simulator unit executing a simulation based on the parameter values ​​adjusted by the parameter adjustment unit; a result output unit that outputs the probability distribution of the simulation result analyzed by the simulation result analysis unit or characteristic information that characterizes the statistical result, Simulation model creation device.

6. 2. The simulation model creation device according to claim 1, a model database that stores information on a plurality of models of different types; and an interface that, when some or all of the plurality of models are selected, displays information that distinguishes the selected plurality of models from one another. Simulation model creation device.

7. 7. The simulation model creation device according to claim 6, the interface further displays default values ​​for the plurality of parameters of the model. Simulation model creation device.

8. 7. The simulation model creation device according to claim 6, The information that distinguishes the plurality of models from one another is a source code of the model or a hash value of a compiled file of the model. Simulation model creation device.

9. 2. The simulation model creation device according to claim 1, a platform language input unit that accepts input in natural language; a platform unit that determines simulation conditions from a natural language input to the platform language input unit; a platform execution unit that outputs an instruction to a simulator unit to execute a simulation based on the simulation conditions determined by the platform unit and the default values ​​of the plurality of parameters set in the model; and a platform calculation result output unit that outputs the simulation execution result of the simulator unit in numerical values ​​or natural language. Simulation model creation device.

10. 10. The simulation model creation device according to claim 9, The platform portion is a large-scale language model constructed using text data and deep learning; a platform condition determination unit configured by a program or database that determines the simulation conditions from a query, the large-scale language model creates the query based on the natural language input from the platform language input unit; the platform condition determination unit determines the simulation conditions based on the query created by the large-scale language model. Simulation model creation device.

11. 10. The simulation model creation device according to claim 9, a model database that stores information on a plurality of models of different types; a default value database storing default values ​​for the plurality of parameters of the model; Furthermore, The platform calculation result output unit further searches for the model or the default value for which a simulation should be performed based on the calculation result of the simulator unit, the model database, and the default value database, and outputs one or more of the model or the default value. Simulation model creation device.

12. 2. The simulation model creation device according to claim 1, a model simulation condition selection system that creates the models of the power devices manufactured by a plurality of different vendors; a platform display unit that displays the models of the power devices manufactured by the plurality of different vendors and the default values ​​so that the models and default values ​​can be selected on a common platform. Simulation model creation device.

13. 2. The simulation model creation device according to claim 1, the first electrode, the second electrode, and the third electrode are a gate electrode, a source electrode, and a drain electrode, or a base electrode, an emitter electrode, and a collector electrode, respectively; The electrical characteristics are voltage characteristics or current characteristics. Simulation model creation device.

14. 2. The simulation model creation device according to claim 1, the electrical characteristics are voltage characteristics or current characteristics, the plurality of first terminals include a terminal for calculating a current between each of the first electrode, the second electrode, and the third electrode, and a terminal for calculating a voltage between each of the first electrode, the second electrode, and the third electrode; Simulation model creation device.

15. 2. The simulation model creation device according to claim 1, the output unit outputs a code to be used in the simulation or a program to be used in the simulation instead of the default value. Simulation model creation device.

Images