Program, method, and calculation device

Abstract

To provide a program for calculating a circuit parameter with high accuracy, a method for generating a learning model, and a calculation device.SOLUTION: An acquisition unit configured to acquire first information on one or more bonding wires of a semiconductor device connected in parallel between a first port and a second port, the first information excluding the number of the one or more bonding wires, and second information on the number of the one or more bonding wires; Estimating at least one parameter of a first parameter that is a circuit parameter representing a connection between the first port and the second port when the number is one, a second parameter that is a circuit parameter when the number is two or more (N1), and a third parameter that is a circuit parameter when the number is more than N1 (N2), and calculating, from the second information, a calculation parameter that is a circuit parameter for the number indicated by the second information, based on the at least one parameter; SELECTED DRAWING: Figure 12

Description

[Technical Field]

[0001] The present disclosure relates to a program, a method, and a computing device. [Background technology]

[0002] In high-frequency circuits, it is known to use bonding wires as inductors (for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2022-138983 Summary of the Invention [Problem to be solved by the invention]

[0004] When designing high-frequency circuits, bonding wires are modeled. If bonding wires are modeled using an equivalent circuit model represented by a lumped constant circuit, it is not possible to accurately represent high-frequency characteristics over a wide frequency band. If bonding wires are modeled using circuit parameters such as S parameters, Y parameters, or Z parameters, it is possible to accurately represent high-frequency characteristics over a wide frequency band. However, when multiple bonding wires are connected in parallel, attempting to calculate highly accurate circuit parameters requires collecting a huge amount of data. As such, it is difficult to calculate highly accurate circuit parameters.

[0005] An object of the present disclosure is to provide a program, a method, and a calculation device that are capable of calculating circuit parameters of bonding wires with high accuracy. [Means for solving the problem]

[0006] An embodiment of the present disclosure is a program that causes a computer to function as an acquisition unit that acquires first information, excluding the number of one or more bonding wires, regarding one or more bonding wires connected in parallel between a first port and a second port, and second information regarding the number; and a calculation unit that estimates at least one parameter from the first information based on a trained model: a first parameter that is a circuit parameter representing the area between the first port and the second port when the number is one, a second parameter that is the circuit parameter when the number is N1 (two or more), and a third parameter that is the circuit parameter when the number is N2 (N2), which is greater than N1; and calculates from the second information, based on the at least one parameter, a calculated parameter that is the circuit parameter for the number indicated by the second information; and the trained model is generated by machine learning training data that defines the relationship between the first information and the number of wires and the circuit parameters obtained for the first information and the number of wires when the number is one, when the number is N1, and when the number is N2.

[0007] The present disclosure can be realized not only as such a characteristic program, but also as a calculation device and a method for processing such characteristic steps, as a semiconductor integrated circuit that realizes part or all of the calculation device, or as an estimation system including the estimation device. [Effects of the Invention]

[0008] According to the present disclosure, circuit parameters can be calculated with high accuracy. [Brief explanation of the drawings]

[0009] [Figure 1A] FIG. 1A is a schematic diagram of a bonding wire for estimating circuit parameters in the first embodiment. [Figure 1B] FIG. 1B is a cross-sectional view taken along line AA in FIG. 1A. [Figure 2]FIG. 2 is a circuit diagram of the semiconductor device. [Figure 3] FIG. 3 is a plan view of the semiconductor device. [Figure 4] FIG. 4 is a cross-sectional view taken along line AA in FIG. [Figure 5] FIG. 5 is a diagram showing Re(Y12) versus the number N in the simulation. [Figure 6] FIG. 6 is a diagram showing Im(Y12) versus the number N in the simulation. [Figure 7] FIG. 7 is a block diagram of a computer according to the first embodiment. [Figure 8] FIG. 8 is a flowchart showing a method for generating teacher data in the first embodiment. [Figure 9] FIG. 9 is a diagram illustrating an example of training data in the first embodiment. [Figure 10] FIG. 10 is a flowchart showing a method for generating a trained model in the first embodiment. [Figure 11] FIG. 11 is a functional block diagram of the calculation device according to the first embodiment. [Figure 12] FIG. 12 is a flowchart showing a method for calculating the Y parameter in the first embodiment. [Figure 13] FIG. 13 is a flowchart showing a method for generating a trained model in a first modified example of the first embodiment. [Figure 14] FIG. 14 is a flowchart showing a method for calculating the Y parameter in the first modified example of the first embodiment. [Figure 15] FIG. 15 is a schematic diagram showing Im(Y12) versus the number N in the second embodiment. [Figure 16] FIG. 16 is a diagram illustrating an example of training data in the second embodiment. [Figure 17] FIG. 17 is a flowchart showing a method for calculating the Y parameter in the second embodiment. [Figure 18] FIG. 18 is a schematic diagram of a bonding wire for which circuit parameters are estimated in the third embodiment. [Figure 19] FIG. 19 is a schematic diagram of a bonding wire. [Figure 20] FIG. 20 is a diagram illustrating an example of training data in the third embodiment. [Figure 21] FIG. 21 is a flowchart showing a method for generating a trained model in the third embodiment. [Figure 22] FIG. 22 is a flowchart showing a method for calculating the Y parameter in the third embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0010] [Description of the embodiments of the present disclosure] First, embodiments of the present disclosure will be listed and described.

[0011] (1) An embodiment of the present disclosure is a program that causes a computer to function as an acquisition unit that acquires first information, excluding the number of one or more bonding wires, regarding one or more bonding wires connected in parallel between a first port and a second port, and second information regarding the number; and a calculation unit that estimates at least one parameter from the first information based on a trained model: a first parameter that is a circuit parameter representing the area between the first port and the second port when the number is one, a second parameter that is the circuit parameter when the number is N1 (two or more), and a third parameter that is the circuit parameter when the number is N2 (greater than N1); and calculates from the second information, based on the at least one parameter, a calculated parameter that is the circuit parameter for the number indicated by the second information; and the trained model is generated by machine learning training data that defines the relationship between the first information and the number of wires and the circuit parameters obtained for the first information and the number of wires when the number is one, when the number is N1, and when the number is N2. This allows for highly accurate calculation of the circuit parameters of the bonding wire. (2) In (1) above, the calculation unit may calculate the calculated parameters based on the second parameters and the third parameters when the number of lines indicated by the second information is other than 1, N1, and N2. This allows for highly accurate calculation of circuit parameters when the number N is other than 1, N1, and N2. (3) In the above (2), the calculation unit may calculate the first parameter as the calculated parameter when the number of lines indicated by the second information is 1, calculate the second parameter as the calculated parameter when the number of lines indicated by the second information is N1, and calculate the third parameter as the calculated parameter when the number of lines indicated by the second information is N2. This makes it possible to calculate highly accurate circuit parameters when the number N is other than 1, N1, or N2. (4) In any one of (1) to (3) above, the trained models may include a first trained model generated by machine learning first teacher data defining the relationship between the first information and the circuit parameters obtained for the first information when the number of wires is 1, a second trained model generated by machine learning second teacher data defining the relationship between the first information and the circuit parameters obtained for the first information when the number of wires is N1, and a third trained model generated by machine learning third teacher data defining the relationship between the first information and the circuit parameters obtained for the first information when the number of wires is N2, and the calculation unit may estimate the second parameters from the first information based on the second trained model and estimate the third parameters from the first information based on the third trained model when the number of wires indicated by the second information is other than 1, N1, or N2. This reduces the amount of data per trained model and reduces the load on the computer. (5) In any of (1) to (3) above, the trained model is a single trained model generated by machine learning a plurality of training data defining the relationship between the first information and the number of wires and the circuit parameters obtained for the first information and the number of wires, and the calculation unit may, when the number of wires indicated by the second information is other than 1, N1, or N2, estimate the second parameter based on the single trained model from the first information and the number of wires being N1, and estimate the third parameter based on the single trained model from the first information and the number of wires being N2. This allows the circuit parameters to be calculated using a single trained model. (6) In any of (1) to (5) above, the trained model may be generated by machine learning training data that defines the relationship between the first information, the number of wires, and the circuit parameters obtained for the first information and the number of wires when the number is 1, when the number is N1, when the number is N2, and when the number is N3, which is greater than N1 and less than N2. When the number indicated by the second information is less than N3, the calculation unit may estimate a fourth parameter, which is the circuit parameter when the number is N3, from the first information based on the trained model and calculate the calculated parameter based on the second parameter and the fourth parameter. When the number indicated by the second information is greater than N3, the calculation unit may calculate the calculated parameter based on the fourth parameter and the third parameter. This allows for accurate calculation of the circuit parameters of the bonding wires. (7) In any of (1) to (6) above, the one or more bonding wires are at least three bonding wires connected in parallel between the first port and the second port, a spacing between adjacent bonding wires of the at least three bonding wires has a first spacing and a second spacing different from the first spacing, the second information includes a first number of the first spacings and a second number of the second spacings, the N1 pieces correspond to when the first number is N1-1 pieces, and the N2 pieces correspond to when the first number is N2-1 pieces, and the calculation unit calculates, based on another trained model from the first information, a fourth parameter which is the circuit parameter when the second number is M1, and a second parameter which is the circuit parameter when the second number is M2 and estimating at least one parameter of a fifth parameter that is the circuit parameter in the second information, and calculating, from the second information, calculated parameters that are the circuit parameters for the first number and the second number indicated by the second information based on at least one parameter of the first parameter, the second parameter, and the third parameter and at least one parameter of the fourth parameter and the fifth parameter, and the other trained model may be generated by machine learning other training data that defines the relationship between the first information and the second number and the circuit parameters obtained for the first information and the second number when the second number is M1 and when the second number is M2. This allows the circuit parameters of bonding wires to be calculated with high accuracy even when the spacing between bonding wires varies. (8) In any one of (1) to (6) above, the one or more bonding wires may be at least three bonding wires connected in parallel between the first port and the second port, and the spacing between adjacent bonding wires of the at least three bonding wires may be constant. This allows for accurate calculation of the circuit parameters of the bonding wires. (9) In any of (1) to (8) above, the acquisition unit may acquire third information related to the frequency of a high-frequency signal transmitted between the first port and the second port, the calculation unit may estimate the at least one parameter from the first information and the third information based on a trained model, and calculate the calculated parameter from the second information based on the at least one parameter, and the trained model may be generated by machine learning a plurality of training data defining the relationship between the first information, the frequency, and the number of wires and the plurality of circuit parameters obtained for the first information, the frequency, and the number of wires when the number of wires is 1, when the number of wires is N1, and when the number of wires is N2. This allows the circuit parameters at any frequency to be calculated. (10) In any of the above (1) to (9), the circuit parameters may be S parameters, Y parameters, or Z parameters, thereby enabling bonding wires to be modeled. (11) An embodiment of the present disclosure is a method including: an acquisition step of acquiring first information, excluding the number of one or more bonding wires, for one or more bonding wires connected in parallel between a first port and a second port, and second information regarding the number; and a calculation step of estimating at least one parameter from the first information based on a trained model: a first parameter that is a circuit parameter representing the area between the first port and the second port when the number is one, a second parameter that is the circuit parameter when the number is N1 (two or more), and a third parameter that is the circuit parameter when the number is N2 (greater than N1); and calculating from the second information, based on the at least one parameter, a calculated parameter that is the circuit parameter for the number indicated by the second information; wherein the trained model is generated by machine learning training data that defines the relationship between the first information and the number of wires and the circuit parameters obtained for the first information and the number of wires when the number is one, when the number is N1, and when the number is N2. This allows for highly accurate calculation of the circuit parameters of the bonding wire. (12) An embodiment of the present disclosure is a calculation device including: an acquisition unit that acquires first information, excluding the number of one or more bonding wires, for one or more bonding wires connected in parallel between a first port and a second port, and second information regarding the number; and a calculation unit that estimates at least one parameter from the first information based on a trained model: a first parameter that is a circuit parameter representing the area between the first port and the second port when the number is one, a second parameter that is the circuit parameter when the number is N1 (two or more), and a third parameter that is the circuit parameter when the number is N2 (N2), which is greater than N1; and calculates from the second information, based on the at least one parameter, a calculated parameter that is the circuit parameter for the number indicated by the second information; and the trained model is generated by machine learning training data that defines the relationship between the first information and the number of wires and the circuit parameters obtained for the first information and the number of wires when the number is one, when the number is N1, and when the number is N2. This allows for highly accurate calculation of the circuit parameters of the bonding wire. (13) An embodiment of the present disclosure is a calculation device that includes: a memory; and a processor that acquires first information regarding one or more bonding wires connected in parallel between a first port and a second port, excluding the number of the one or more bonding wires, and second information regarding the number; and estimates at least one parameter from the first information based on a trained model: a first parameter that is a circuit parameter representing the area between the first port and the second port when the number is one, a second parameter that is the circuit parameter when the number is N1 (two or more), and a third parameter that is the circuit parameter when the number is N2 (N2), which is greater than N1; and calculates from the second information, based on the at least one parameter, a calculated parameter that is the circuit parameter for the number indicated by the second information; and the trained model is generated by machine learning training data that defines the relationship between the first information and the number of wires and the circuit parameters obtained for the first information and the number of wires when the number is one, when the number is N1, and when the number is N2. This allows for highly accurate calculation of the circuit parameters of the bonding wire.

[0012] [Details of the embodiments of the present disclosure] Specific examples of a program, a method, and a computing device according to embodiments of the present disclosure will be described below with reference to the drawings. Note that the present disclosure is not limited to these examples, but is defined by the claims, and is intended to include all modifications within the meaning and scope of the claims.

[0013] At least some of the embodiments described below may be combined in any manner. The calculation device is configured with a computer, and the functions of the estimation device are realized by a computer program stored in a storage device of the computer being executed by the computer's CPU (Central Processing Unit). The computer program can be stored in a storage medium such as a CD-ROM (Compact Disc Read Only Memory) or a DVD (Digital Versatile Disc).

[0014] (First embodiment) FIG. 1A is a schematic diagram of a bonding wire for estimating circuit parameters in the first embodiment. FIG. 1B is a cross-sectional view taken along line AA in FIG. 1A. As shown in FIGS. 1A and 1B, N bonding wires 10 are connected in parallel between a pad 11A corresponding to a first port P1 and a pad 11B corresponding to a second port P2. A high-frequency signal is transmitted from the pad 11A to the pad 11B via the bonding wires 10. The high-frequency signal is, for example, a microwave (300 MHz to 30 GHz) or a millimeter wave (30 GHz to 300 GHz). The bonding wires 10 have a diameter φ1, a length W1, a height H1, and an interval D1.

[0015] (Examples of bonding wires) A specific example in which bonding wires are used will be described using a semiconductor device having an amplifier circuit as an example. FIG. 2 is a circuit diagram of the semiconductor device. As shown in FIG. 2, a semiconductor device 18 includes an input terminal Tin, an output terminal Tout, a transistor Q1, and matching circuits 16 and 17. The transistor Q1 has a source S, a gate G, and a drain D. The source S is grounded. The gate G is electrically connected to the input terminal Tin via the matching circuit 16. The drain D is electrically connected to the output terminal Tout via the matching circuit 17. The matching circuit 16 includes inductors L1 and L2 and a capacitor C1. The inductors L1 and L2 are connected in series between the input terminal Tin and the gate G. The capacitor C1 is shunt-connected to a node between the inductors L1 and L2. The matching circuit 17 includes an inductor L3. A first end of the inductor L3 is electrically connected to the drain D, and a second end of the inductor L3 is electrically connected to the output terminal Tout.

[0016] The semiconductor device 18 is, for example, an amplifier circuit. A high-frequency signal is input to the input terminal Tin. When the semiconductor device 18 is used in a mobile communication base station, the frequency of the high-frequency signal is, for example, 0.5 GHz to 20 GHz. The high-frequency signal passes through a matching circuit 16 and is input to a gate G. The matching circuit 16 matches the impedance seen from the input terminal Tin to the impedance seen from the matching circuit 16 to the gate G. The transistor Q1 amplifies the high-frequency signal input to the gate G. The amplified high-frequency signal is output from an output terminal Tout via a matching circuit 17.

[0017] FIG. 3 is a plan view of the semiconductor device. FIG. 4 is a cross-sectional view taken along line AA of FIG. 3. As shown in FIGS. 3 and 4, the semiconductor device 18 includes a base 12, leads 14 and 15, a semiconductor chip 20, a passive chip 25, and bonding wires 10A, 10B, and 10C. The base 12 and leads 14 and 15 are conductive. The semiconductor chip 20 includes a substrate 21 and electrodes 22 to 24. The electrodes 22 and 23 are provided on the upper surface of the substrate 21, and the electrode 24 is provided on the lower surface of the substrate 21. The passive chip 25 includes a substrate 26 and electrodes 27 and 28. The electrode 27 is provided on the upper surface of the substrate 26, and the electrode 28 is provided on the lower surface of the substrate 26. The electrodes 24 and 28 are bonded to the base 12 by a conductive bonding layer. The bonding wire 10A electrically connects the lead 14 to the electrode 27. The bonding wire 10B electrically connects the electrodes 27 and 22. The bonding wire 10C electrically connects the electrode 23 and the lead 15 together.

[0018] Leads 14 and 15 correspond to the input terminal Tin and the output terminal Tout, respectively. Bonding wires 10A, 10B, and 10C correspond to inductors L1, L2, and L3, respectively. In the passive chip 25, the substrate 26 and the electrodes 27 and 28 sandwiching the substrate 26 correspond to the capacitor C1. A reference potential such as ground potential is supplied to the base 12, which corresponds to ground. Electrodes 22, 23, and 24 of the semiconductor chip 20 correspond to the gate G, drain D, and source S of the transistor Q1, respectively.

[0019] The base 12 and leads 14 and 15 are metal plates such as copper plates or laminates of copper, molybdenum, and copper plates. The transistor Q1 is, for example, a FET (Field Effect Transistor). The substrate 21 is a semiconductor substrate such as a silicon carbide substrate or a silicon substrate. The substrate 26 is a dielectric substrate such as an alumina substrate or a barium titanate substrate. The electrodes 22 to 24, 27, and 28 are metal layers such as gold layers. The bonding wires 10A, 10B, and 10C are thin metal wires such as thin gold wires or thin aluminum wires.

[0020] When inductors L1 to L3 are used in, for example, matching circuits 16 and 17, it is difficult to design an amplifier circuit unless the high-frequency characteristics of inductors L1 to L3 are modeled with high accuracy. However, in semiconductor device 18, bonding wires 10A (or 10B, 10C) connected in parallel are used as inductor L1 (or L2, L3). In the case of a high-power amplifier circuit, the number of bonding wires 10A (or 10B, 10C) may be as many as 100 to 150. In this way, when a large number of bonding wires 10A are connected in parallel, it is difficult to model the bonding wires 10A (or 10B, 10C).

[0021] Consider the case of modeling multiple bonding wires 10 in parallel between a first port P1 and a second port P2, as shown in FIGS. 1A and 1B. First, it is possible to use an equivalent circuit model in which the multiple bonding wires 10 are represented by a lumped constant circuit. However, it is difficult to model with high accuracy over a wide frequency range using an equivalent circuit model. Therefore, it is possible to use circuit parameters such as S parameters, Y parameters, or Z parameters. The circuit parameters between the first port P1 and the second port P2 are measured or simulated for each frequency. This allows for high-accuracy modeling over a wide frequency range.

[0022] The high-frequency characteristics of the bonding wire 10 depend on the shape (e.g., diameter φ1, length W1, height H1, and spacing D1) of the bonding wire 10. Therefore, it is conceivable to create a model using a neural network with the shape, number N, and frequency f of the bonding wire 10 as explanatory variables and S parameters, Y parameters, Z parameters, or the like as objective variables.

[0023] The neural network model is generated based on training data. The training data is generated by actual measurements or electromagnetic field analysis while changing the shape, number N, and frequency f of the bonding wires 10. However, when many bonding wires 10 are connected in parallel, the number N of bonding wires 10 increases. In this case, when attempting to generate a highly accurate bonding wire model, the amount of training data becomes enormous. Furthermore, as the trained model becomes larger, the load on the computer increases. The first embodiment solves these problems.

[0024] (simulation) 1A and 1B, the number N of bonding wires 10 was changed, and the Y parameter Y12 between the first port P1 and the second port P2 was simulated. The N bonding wires 10 had the same shape. The frequency was 3 GHz.

[0025] FIG. 5 is a diagram showing Re(Y12) versus the number N in the simulation. FIG. 6 is a diagram showing Im(Y12) versus the number N in the simulation. Re(Y12) indicates the real part of Y12, and Im(Y12) indicates the imaginary part of Y12. White circles indicate simulated points, and solid lines are straight lines connecting the white circles. Dashed lines are straight lines linearly approximating the white circles when N is excluding 1. As shown in FIGS. 5 and 6, Re(Y12) and Im(Y12) can be linearly approximated when the number of lines is excluding 1. The coefficient of determination R in FIG. 5 2 The coefficient of determination R in Figure 6 is 0.98. 2 is 0.99. If the linear approximation line is N×yc-yc+ye, yc corresponds to the slope of the dashed line, and -yc+ye corresponds to the intercept. If the Y parameter (for example, Re(Y12) and Im(Y12)) for N lines is YN, YN is expressed by Equation 1.

[0026] YN = (N-1) × yc + ye (Equation 1)

[0027] (Calculation principle of Y parameter in the first embodiment) In FIG. 1A, the Y parameter Yc inside the bonding wire 10 at both ends corresponds to (N-1) x yc, and the Y parameter outside the bonding wire 10 at both ends corresponds to ye / 2.

[0028] When the number of bonding wires 10 is one, the Y parameter YN is expressed by Equation 2.

[0029] YN=Y1 (Equation 2) When the number of bonding wires 10 is N, which is other than 1, the Y parameter YN is expressed by Equation 3.

[0030] YN = (N-1) × yc + ye (Equation 3) Here, YN is an element of the admittance matrix of the Y parameter, and is the real and imaginary parts of Y11, Y12, Y21, and Y22, respectively.

[0031] The Y parameter YN for any number N of lines can be calculated from equations 2 and 3 if the Y parameter Y1 when N is 1 line, and the Y parameters YN1 and YN2 when N is 2 or more lines (N1 lines), and N2 when N is greater than N1 lines, respectively, are known.

[0032] When N is 1, YN can be calculated using Equation 1.

[0033] When N is other than 1, YN can be calculated using the following formula 4.

[0034] YN=(N-1)×(YN2-YN1) / (N2-N1)-((N1-1)×YN2-(N2-1)×YN1) / (N2-N1) (Formula 4)

[0035] Therefore, we propose a neural network model in which the explanatory variables are the shape of the bonding wire 10 (e.g., diameter φ1, length W1, height H1, and spacing D1) and frequency f, and the objective variables are Y1, YN1, and YN2. Note that once the length W1 and height H1 are determined, the three-dimensional shape of the bonding wire 10 is roughly determined. Therefore, the three-dimensional shape of the bonding wire 10 can be expressed using the length W1 and height H1. A more detailed three-dimensional shape of the bonding wire 10 may also be used as an explanatory variable.

[0036] A method for calculating the Y parameter of the bonding wire 10 in the first embodiment will be described below.

[0037] (computer block diagram) 7 is a block diagram of a computer in Example 1. The computer 30 works in cooperation with software and functions as a calculation device that calculates the circuit parameters of the bonding wire 10. The computer 30 executes a calculation program and performs a calculation method.

[0038] The computer 30 includes a processor 32, a memory 34, an input / output device 36, and an internal bus 38. The processor 32 is, for example, a CPU (Central Processing Unit) and executes programs and methods. The memory 34 is, for example, a volatile or non-volatile memory and stores data and the like used by the processor 32 when executing the programs and methods. The memory 34 may also store programs executed by the processor 32. The input / output device 36 inputs data acquired by the processor 32 from an external device and outputs data output by the processor 32 to an external device. The external device may be another computer or another program within the same computer. The internal bus 38 connects the processor 32, the memory 34, and the input / output device 36 and transmits data and the like. The programs are stored in a storage medium 35. The storage medium 35 is, for example, a non-transitory tangible medium such as a CD-ROM or DVD.

[0039] (How to generate training data) Fig. 8 is a flowchart showing a method for generating training data in the first embodiment. Each step in Fig. 8 may be executed by the computer shown in Fig. 7 or may be executed by a human. Fig. 9 is a diagram showing an example of training data in the first embodiment.

[0040] 8, i=1 and j=1 are set (step S10), where i is an integer from 1 to m1, and j is an integer from 1 to m2.

[0041] Next, information A is set to A(i), and frequency f is set to f(j) (step S11). A(i) is information about the shape of the bonding wire 10, such as diameter φ1(i), length W1(i), height H1(i), and interval D1(i). i corresponds to 1 to m1, and the value of each piece of information is made different. j corresponds to 1 to m2, and the frequency f(j) is made different.

[0042] Next, for the information A(i) and frequency f(j), a simulation (e.g., electromagnetic field analysis) is performed to calculate the Y parameters Y1(i,j), YN1(i,j), and YN2(i,j) when the number of bonding wires 10 is 1, N1, and N2 (step S12). The Y parameters Y1(i,j), YN1(i,j), and YN2(i,j) may be measured. Here, each Y parameter is a 2-row, 2-column admittance matrix, and the elements of the admittance matrix are Y11, Y12, Y21, and Y22. Each element is a complex number. The real parts of each element are Re(Y11), Re(Y12), Re(Y21), and Re(Y22). The imaginary parts of each element are Im(Y11), Im(Y12), Im(Y21), and Im(Y22). The Y parameters Y1(i,j), YN1(i,j), and YN2(i,j) are at least one parameter among Re(Y11), Re(Y12), Re(Y21), Re(Y22), Im(Y11), Im(Y12), Im(Y21), and Im(Y22).

[0043] Next, teacher data T1(i,j), TN1(i,j), and TN2(i,j) are generated (step S13). For example, when i = 1 and j = 1, teacher data T1(1,1), TN1(1,1), and TN2(1,1) are generated, which respectively define the relationship between information A(1) and frequency f(1) and Y1(1,1), YN1(1,1), and YN2(1,1), as shown in FIG.

[0044] Next, it is determined whether i = m1 (step S14). If no, in step S15, i = i + 1 is set and the process returns to step S11. By changing i from 1 to m1, m1 pieces of training data T1(i,1) that define the relationship between information A(i) and Y1(i,1) are generated. m1 pieces of training data TN1(i,1) and m1 pieces of training data TN2(i,1) that define the relationship between information A(i) and YN1(i,1) and YN2(i,1), respectively, are generated.

[0045] If the answer is Yes in step S14, it is determined whether j=m2 (step S16). If the answer is No, j=j+1 is set in step S17, and the process returns to step S11. By changing j from 1 to m2, m1×m2 pieces of training data T1(i,j) are generated, which define the relationship between information A(i) and Y1(i,j). m1×m2 pieces of training data TN1(i,j) and m1×m2 pieces of training data TN2(i,j) are generated, which define the relationship between information A(i) and YN1(i,j) and YN2(i,j), respectively.

[0046] If the answer is Yes in step S16, the training data T1, TN1, and TN2 are output to the memory 34 or an external device (step S18), and then the process ends.

[0047] (How to generate a trained model) FIG. 10 is a flowchart showing a method for generating a trained model in the first embodiment. As shown in FIG. 10, the processor 32 acquires m1×m2 pieces of training data T1, m1×m2 pieces of training data TN1, and m1×m2 pieces of training data T2 from the memory 34 or an external device (step S20). The processor 32 performs machine learning based on the acquired training data T1, TN1, and TN3 to generate a trained model LM (step S21). The processor 32 or a human verifies the trained model LM (step S22). For example, Y1, YN1, and YN2 are estimated using the trained model LM from information A, where the relationship between information A and Y1, YN1, and YN2 is known. If the estimated Y1, YN1, and YN2 approximately match the known Y1, YN1, and YN2, step S22 is judged as "Yes." If they do not match, step S22 is judged as "No." If the judgment is "No," the process returns to step S21. If the answer is Yes in step S22, the trained model LM is output to the memory 34 or an external device (step S23).

[0048] (Functional block diagram of the calculation device) Fig. 11 is a functional block diagram of a calculation device according to the first embodiment. As shown in Fig. 11, the calculation device 40 includes an acquisition unit 42, a calculation unit 44, and an output unit 46. The processor 32 functions as the acquisition unit 42, the calculation unit 44, and the output unit 46 in cooperation with a program. The acquisition unit 42 acquires information A, frequency f, and number N. The calculation unit 44 calculates a Y parameter YN based on the learned model LM from the acquired information A, frequency f, and number N. The output unit 46 outputs the calculated Y parameter YN.

[0049] (Y parameter calculation method) Fig. 12 is a flowchart showing a method for calculating the Y parameter in the first embodiment. Each step in Fig. 12 is executed by the computer 30. As shown in Fig. 12, the acquisition unit 42 acquires information A (first information) about the bonding wire 10, information (third information) about the frequency f of the high-frequency signal, and information (second information) about the number N of the bonding wires 10 (step S30).

[0050] Next, the calculation unit 44 acquires the learned model LM (step S31). The calculation unit 44 determines whether N=1 (step S32). If Yes, the calculation unit 44 estimates Y1 based on the learned model LM from the acquired information A and frequency f (step S33). Next, the calculation unit 44 sets the estimated Y1 to YN (step S34). Thereafter, the process proceeds to step S43.

[0051] If the result of step S32 is No, the calculation unit 44 determines whether N=N1 (step S35). If the result is Yes, the calculation unit 44 estimates YN1 based on the learned model LM from the acquired information A and frequency f (step S36). Next, the calculation unit 44 sets the estimated YN1 to YN (step S37). Then, the process proceeds to step S43.

[0052] If the result of step S35 is No, the calculation unit 44 determines whether N=N2 (step S38). If the result is Yes, the calculation unit 44 estimates YN2 based on the learned model LM from the acquired information A and frequency f (step S39). Next, the calculation unit 44 sets the estimated YN2 to YN (step S40). Then, the process proceeds to step S43.

[0053] If the result of step S38 is No, the calculation unit 44 estimates YN1 and YN2 based on the learned model LM from the acquired information A and frequency f (step S41). Next, the calculation unit 44 calculates YN from the estimated YN1 and YN2 (step S42). For example, the calculation unit 44 calculates YN by substituting the estimated YN1 and YN2 into Equation 4.

[0054] Thereafter, the output unit 46 outputs the Y parameter YN to the memory 34 or an external device (step S43), and then the process ends. Note that the order of steps S32 to S34, S35 to S37, and S38 to S40 can be set as appropriate.

[0055] (First comparative example) As a first comparative example, an example of modeling a bonding wire using a neural network will be described. The explanatory variables and objective variables of the neural network are as follows: Explanatory variables Diameter φ1, length W1, height H1, spacing D1, number N, frequency f Response variable Re(S11), Re(S12), Re(S21), Re(S22), Im(S11), Im(S12), Im(S21), Im(S22) S11, S12, S21, and S22 are S parameters between the first port P1 and the second port P2. In the first comparative example, training data is created according to the number N. This results in a huge amount of training data. In addition, the trained model of the neural network becomes large.

[0056] (Description of the First Embodiment) Examples of explanatory variables and response variables of the neural network in the first embodiment are as follows: Explanatory variables Diameter φ1, length W1, height H1, spacing D1, frequency f Response variable Re(Y1_11), Re(Y1_12), Re(Y1_Y21), Re(Y1_Y22), Im(Y1_11), Im(Y1_12), Im(Y1_21), Im(Y1_22) Re(YN1_11), Re(YN1_12), Re(YN1_Y21), Re(YN1_Y22), Im(YN1_11), Im(YN1_12), Im(YN1_21), Im(YN1_22) Re(YN2_11), Re(YN2_12), Re(YN2_Y21), Re(YN2_Y22), Im(YN2_11), Im(YN2_12), Im(YN2_21), Im(YN1_22) In the first embodiment, although the number of objective variables is three times that of the first comparative embodiment, the number N may be three types: 1, N1, and N2. Therefore, it is possible to reduce the amount of training data compared to the first comparative embodiment. Also, it is possible to reduce the size of the trained model of the neural network. Note that the objective variable may be at least one parameter out of the eight Y parameters in each of Y1, YN1, and YN2.

[0057] According to the program, method, and calculation device of the first embodiment, as shown in Figures 8 and 9, the training data T defines the relationship between the information A(i) (first information) and the number N when the number N is 1, when the number N is N1, and when the number N is N2, and the Y parameters Y1, YN1, and YN2 obtained for the information A(i) and the number N.

[0058] As shown in FIG. 10, a trained model LN is generated by machine learning the training data T. As shown in step S30 of FIG. 12, the acquisition unit 42 acquires information A excluding the number of bonding wires 10 and second information regarding the number N. As shown in steps S32 to S42, the calculation unit 44 estimates a Y parameter Y1 (first parameter) when the number N is 1, a Y parameter YN1 (second parameter) when the number N is N1, and a Y parameter YN2 (third parameter) when the number N is N2 based on the trained model LM from the information A. From the second information indicating the number N, the calculation unit 44 calculates a Y parameter YN (calculated parameter) for the number N indicated by the second information based on at least one parameter of Y1, YN1, and YN1. As a result, compared to the first comparative example, it is possible to reduce the amount of training data and the trained model, and therefore it is possible to calculate the Y parameter of the bonding wire 10 with high accuracy even with a low load on the computer. Furthermore, the calculated Y parameters Y1, YN1, YN2, and YN may each be all eight parameters, or may be at least one parameter out of the eight parameters.

[0059] As in steps S41 and S42, the calculation unit 44 calculates YN based on YN1 and YN2 when the number N is other than 1, N1, and N2. As a result, when the number N is other than 1, N1, and N2, YN can be calculated with high accuracy using, for example, Equation 4.

[0060] As in steps S34, S37, and S40, the calculation unit 44 calculates Y1 as YN when the number N is 1, calculates YN1 as YN when the number N is N1, and calculates YN3 as YN when the number N is N2. This allows for accurate calculation of YN when the number N is 1, N1, or N2.

[0061] As shown in Figures 8 to 10, the learned model LM is a single learned model generated by machine learning multiple pieces of training data T that define the relationship between information A, number N, and multiple YNs obtained for the information A and number N. As shown in step S41 of Figure 12, when the number N is other than 1, N1, or N2, the calculation unit 44 estimates YN1 based on the learned model LM from the information A and number N as N1, and estimates YN2 based on the learned model LM from the information A and number N as N2. In step S42, YN is calculated based on YN1 and YN2. This allows YN to be calculated using a single learned model LN.

[0062] As in steps S33 and S34, when the number N is 1, the calculation unit 44 estimates Y1 based on the information A and the learned model LM from the number N of 1, and calculates Y1 as YN. As in steps S36 and S37, when the number N is N1, the calculation unit 44 estimates YN1 based on the information A and the learned model LM from the number N of N1, and calculates YN1 as YN. As in steps S39 and S40, when the number N is N2, the calculation unit 44 estimates YN2 based on the learned model LN from the information A and the number N of N2, and calculates YN2 as YN. This allows YN to be calculated using one learned model LN.

[0063] When there are at least three bonding wires 10, the spacing D1 between adjacent bonding wires of the at least three bonding wires 10 is constant. As a result, the Y parameter YN can be approximated by Equation 3, allowing the Y parameter of the bonding wires 10 to be calculated with high accuracy. Note that a constant spacing D1 does not necessarily mean that the spacings D1 are strictly equal. For example, when the maximum value of the multiple spacings D1 is MAX and the minimum value is MIN, 2×(MAX−MIN) / (MAX+MIN)≦0.1.

[0064] As shown in FIGS. 8 to 10, the learned model LN is generated by machine learning training data T that defines the relationship between information A, frequency f, number N, and multiple Y parameters obtained for the information A, frequency f, and number N. As shown in step S30 of FIG. 12, the acquisition unit 42 acquires third information regarding the frequency f of the high-frequency signal transmitted between the first port P1 and the second port P2. As shown in steps S33, S36, S39, and S41, the calculation unit 44 estimates Y1, YN1, or YN2 based on the learned model LM from the information f and the third information. As shown in steps S34, S37, S40, and S42, YN for number N is calculated based on at least one of Y1, YN1, and YN2 from the number N. This allows the Y parameter YN at any frequency to be calculated.

[0065] (First Modification of the First Embodiment) The first modified example of the first embodiment is an example in which a plurality of trained models are used.

[0066] (How to generate a trained model) Fig. 13 is a flowchart showing a method for generating a trained model in a first modified example of the first embodiment. As shown in Fig. 13, the processor 32 acquires m1 x m2 pieces of training data T1 from the memory 34 or an external device (step S20A). The processor 32 performs machine learning based on the acquired training data T1 to generate a trained model LM1 (step S21A). The processor 32 or a human verifies the trained model LM1 (step S22A). If the result is No, the process returns to step S21A.

[0067] If the answer is Yes in step S22A, the processor 32 acquires m1×m2 pieces of training data TN1 from the memory 34 or an external device (step S20B). The processor 32 performs machine learning based on the acquired training data TN1 to generate a trained model LMN1 (step S21B). The processor 32 or a human verifies the trained model LMN1 (step S22B). If the answer is No, the process returns to step S21B.

[0068] If the answer is Yes in step S22B, the processor 32 acquires m1 × m2 pieces of training data TN2 from the memory 34 or an external device (step S20C). The processor 32 performs machine learning based on the acquired training data TN2 to generate a trained model LMN2 (step S21C). The processor 32 or a human verifies the trained model LMN2 (step S22C). If the answer is No, the process returns to step S21C. If the answer is Yes in step S22C, the trained models LM1, LMN1, and LMN2 are output to the memory 34 or an external device (step S23). The process then ends. Note that the order of steps S20A to S22A, S20B to S22B, and S20C to S22C can be set as appropriate. Furthermore, the training data T1, TN1, and TN2 may be acquired in step S20A, and steps S20B and S20C may not be performed.

[0069] (Y parameter calculation method) FIG. 14 is a flowchart showing a method for calculating the Y parameter in the first modified example of the first embodiment. As shown in FIG. 14, the calculation unit 44 does not acquire a trained model between steps S30 and S32. If the answer is Yes in step S32, the calculation unit 44 acquires the trained model LM1 (step S31A). Next, the calculation unit 44 estimates Y1 based on the trained model LM1 from the acquired information A and frequency f (step S33). The calculation unit 44 sets the estimated Y1 to YN (step S34). Thereafter, the process proceeds to step S43.

[0070] If the answer is Yes in step S35, the calculation unit 44 acquires the learned model LMN1 (step S31B). Next, the calculation unit 44 estimates YN1 based on the learned model LMN1 from the acquired information A and frequency f (step S36). Next, the calculation unit 44 sets the estimated YN1 to YN (step S37). Thereafter, the process proceeds to step S43.

[0071] If the answer is Yes in step S37, the calculation unit 44 acquires the learned model LMN2 (step S30C). Next, the calculation unit 44 estimates YN2 based on the learned model LMN2 from the acquired information A and frequency f (step S39). Next, the calculation unit 44 sets the estimated YN2 to YN (step S40). Thereafter, the process proceeds to step S43.

[0072] If the answer is No in step S38, the calculation unit 44 acquires the learned models LMN1 and LMN2 (step S30D). Next, the calculation unit 44 estimates YN1 based on the learned model LM1 from the acquired information A and frequency f, and estimates YN2 based on the learned model LM2 from the acquired information A and frequency f (step S41). Next, the calculation unit 44 calculates YN from the estimated YN1 and YN2 (step S42). The rest of the flow is the same as in FIG. 12, and therefore a description thereof will be omitted. Note that the acquisition of the learned models LM1, LMN1, and LMN2 may be performed between steps S30 and S32, and steps S31A to S31D may not be performed.

[0073] In the first modification of the first embodiment, there are three trained neural network models LM1, LMN1, and LMN2. Examples of explanatory variables and response variables in each model are as follows: Trained model LM1 Explanatory variables Diameter φ1, length W1, height H1, spacing D1, frequency f Response variable Re(Y1_11), Re(Y1_12), Re(Y1_Y21), Re(Y1_Y22), Im(Y1_11), Im(Y1_12), Im(Y1_21), Im(Y1_22) Trained model LMN1 Explanatory variables Diameter φ1, length W1, height H1, spacing D1, frequency f Response variable Re(YN1_11), Re(YN1_12), Re(YN1_Y21), Re(YN1_Y22), Im(YN1_11), Im(YN1_12), Im(YN1_21), Im(YN1_22) Trained model LMN1 Explanatory variables Diameter φ1, length W1, height H1, spacing D1, frequency f Response variable Re(YN2_11), Re(YN2_12), Re(YN2_Y21), Re(YN2_Y22), Im(YN2_11), Im(YN2_12), Im(YN2_21), Im(YN1_22)

[0074] According to a first modification of the first embodiment, the trained models include LM1 (first trained model), LMN1 (second trained model), and LMN2 (third trained model). As shown in FIGS. 9 and 14, LM1 is generated by machine learning training of training data T1 (first training data) that defines the relationship between information A and Y1 obtained for the information A when the number N is 1. LMN1 is generated by machine learning training of training data TN1 (second training data) that defines the relationship between information A and YN1 obtained for the information A when the number N is N1. LMN2 is generated by machine learning training of training data TN2 (third training data) that defines the relationship between information A and YN2 obtained for the information A when the number N is N2. While this increases the number of trained models, it is possible to reduce the amount of data per trained model. This reduces the load on the computer 30.

[0075] As in steps S31D and S41, when the number N is other than 1, N1, and N2, the calculation unit 44 estimates YN1 based on the learned model LMN1 from information A, and estimates YN2 based on the learned model LMN2 from information A. As in step S42, the calculation unit 44 calculates YN based on the estimated YN1 and YN2. This allows YN to be calculated with high accuracy.

[0076] As in steps S31A, S33, and S34, when the number N is 1, the calculation unit 44 estimates Y1 from information A based on the learned model LM1 and calculates Y1 as YN. As in steps S31B, S36, and S37, when the number N is N1, the calculation unit 44 estimates YN1 from information A based on the learned model LMN1 and calculates YN1 as YN. As in steps S31C, S39, and S40, when the number N is N2, the calculation unit 44 estimates YN2 from information A based on the learned model LMN2 and calculates YN2 as YN. This allows for accurate calculation of YN.

[0077] (Second embodiment) The second embodiment is an example in which the number N used in the training data has four or more levels. Fig. 15 is a schematic diagram showing Im(Y12) versus the number N in the second embodiment. As shown in Fig. 15, when the number N is 2 or more, Im(Y12) deviates from a straight line. Therefore, in the second embodiment, the Y parameter YN3 is estimated when the number N is N3, which is between N1 and N2.

[0078] (Teacher data) FIG. 16 is a diagram showing an example of training data in the second embodiment. For simplicity, in FIG. 16, A(1) to A(m1) in FIG. 9 are represented as A(i), and f(1) to f(m2) in FIG. 9 are represented as f(j). Y1(1,1) to Y1(m1,m2) in FIG. 9 are represented as Y1(i,j), and T1(1,1) to T1(m1,m2) in FIG. 9 are represented as T1(i,j). The same applies to YN1(i,j), TN1(i,j), YN2(i,j), TN2(i,j), YN3(i,j), and T3(i,j). As shown in FIG. 16, in the second embodiment, TN3(i,j) is used as training data in addition to T1(i,j), TN1(i,j), and TN2(i,j).

[0079] (Y parameter calculation method) 17 is a flowchart showing a method for calculating the Y parameter in the second embodiment. As shown in FIG. 17, if the result in step S38 is No, the calculation unit 44 determines whether N=N3 (step S44). If the result is Yes, the calculation unit 44 estimates YN3 based on the learned model LM from the acquired information A and frequency f (step S45). Next, the calculation unit 44 sets the estimated YN3 to YN (step S46). Thereafter, the process proceeds to step S43.

[0080] When the answer is No in step S44, the calculation unit 44 determines whether 1 < N < N3 (step S47). When the answer is Yes, the calculation unit 44 estimates YN1 and YN3 based on the learned model LM from the acquired information A and the frequency f (step S41A). The calculation unit 44 calculates YN from the estimated YN1 and YN3 (step S42A). Then, the process proceeds to step S43.

[0081] When the answer is No in step S47, the calculation unit 44 estimates YN3 and YN2 based on the learned model LM from the acquired information A and the frequency f (step S41B). The calculation unit 44 calculates YN from the estimated YN3 and YN3 (step S42A). Then, the process proceeds to step S43. Note that the order of steps S32 to S34, S35 to S37, S38 to S40, and S44 to S46 can be set as appropriate. Other flows are the same as those in FIG. 12 and the description thereof is omitted.

[0082] The formula 5 for calculating YN in steps S42A and S42B is as follows.

[0083] YN = (N - 1)×(YB - YA) / (NB - NA) - ((NA - 1)×YB - (NB - 1)×YA) / (NB - NA) (Formula 5) Here, in step S42A, NA = N1, NB = N3, YA = YN1, and YB = YN3. In step S42B, NA = N3, NB = N2, YA = YN3, and YB = YN2.

[0084] According to the second embodiment, as shown in FIG. 16, in addition to the teacher data T1, TN1, and TN3 in FIG. 9 of the first embodiment, the learned model LM is based on the number N being N3, the information A and the number N3, and a plurality of YN3 obtained for the information A and the number N3. A plurality of teacher data TN3 that define the relationship are generated by machine learning. As in steps S41A and S42A of FIG. 17, when the number N is smaller than N3, the calculation unit 44 estimates YN1 and YN3 (the fourth parameter) based on the learned model LM, and calculates YN based on YN1 and YN3. As in steps S41B and S42B, when the number N is larger than N3, the calculation unit 44 estimates YN3 and YN2 based on the learned model LM, and calculates YN based on YN3 and YN2. Thus, even when the slope of YN with respect to the number N changes depending on N, the Y parameter of the bonding wire 10 can be accurately calculated.

[0085] The number of N for estimating the Y parameter may be 5 or more. For example, a learned model is generated using teacher data for the numbers 1, 2, N1, N3, and N2. When the number N for calculating the Y parameter is 2 < N < N1, Y2 and YN1 are estimated, and YN is calculated from Y2 and YN1 using Equation 5. When the number N is N1 < N < N3, YN1 and YN3 are estimated, and YN is calculated from YN1 and YN3 using Equation 5. When the number N is N3 < N, YN3 and YN2 are estimated, and YN is calculated from YN3 and YN2 using Equation 5.

[0086] The learned model may include four learned models when the number N is 1, N1, N3, and N2, as in the first modification of the first embodiment.

[0087] (Third Embodiment) The third embodiment is an example in which the spacing between bonding wires 10 is different. FIG. 18 is a schematic diagram of a bonding wire for estimating circuit parameters in the third embodiment. As shown in FIG. 18, the spacing between adjacent bonding wires 10 connected in parallel between pads 11A and 11B has two levels of spacing, D1 and D2. If the number of spacings D1 is n and the number of spacings D2 is m, then the number N of bonding wires 10 is N=n+m+1. In the example of FIG. 18, n=9, m=2, and N=12.

[0088] The Y parameter Ynm between the first port P1 and the second port P2 is Ynm=Ye+9×yc+2×yc2. More generally, Ynm is expressed by Equation 6.

[0089] Ynm=ye+n×yc+m×yc2 (Equation 6)

[0090] If Yn=ye+n×yc and Ym=m×yc2, Ynm is given by Equation 7.

[0091] Ynm=Yn+Ym (Equation 7) ye and yc are the same as in formula 1. Therefore, if the number N in FIG. 1A is set to n+1, Yn is given by formula 8.

[0092] Yn=n×(YN2-YN1) / (N2-N1)-((N1-1)×YN2-(N2-1)×YN1) / (N2-N1) (Formula 8)

[0093] FIG. 19 is a schematic diagram of a bonding wire. As shown in FIG. 19, m+1 bonding wires 10 are connected between pads 11A and 11B. The spacing D2 between adjacent bonding wires 10 is constant. The number of spacings D2 is m. Ym, excluding ye / 2 at both ends, is m×yc2. Therefore, as the number m, mutually different arbitrary M1 and M2 are set. The Y parameters between the first port P1 and the second port P2 when the numbers are M1 and M2 are YM1 and YM2. In this case, Ym can be calculated from Equation 9.

[0094] Ym=m×(YM2-YM2) / (M2-M1) (Formula 9)

[0095] (Teacher data) Fig. 20 is a diagram showing an example of training data in the third embodiment. As shown in Fig. 20, training data T1(i,j), TN1(i,j), and TN2(i,j) are generated in the same manner as in Fig. 9. Training data T1(i,j), TN1(i,j), and TN2(i,j) are generated using constant samples at an interval D1 between adjacent bonding wires 10, as in Fig. 1A. The number of wires corresponds to n+1.

[0096] Furthermore, when the number m of intervals D2 is M1, training data TM1(i,j) is generated in which information A(j) and frequency f(j) are associated with YM1(i,j) simulated from the information A(j) and frequency f(j). When the number m of intervals D2 is M2, training data TM2(i,j) is generated in which information A(j) and frequency f(j) are associated with YM2(i,j) simulated from the information A(j) and frequency f(j).

[0097] (How to generate a trained model) Fig. 21 is a flowchart showing a method for generating a trained model in the third embodiment. As shown in Fig. 21, the processor 32 acquires training data T1, TN1, and TN2 from the memory 34 or an external device (step S20D). The processor 32 performs machine learning based on the acquired training data T1, TN1, and TN2 to generate a trained model LMn (step S21D). The processor 32 or a human verifies the trained model LMn (step S22D). If the result is No, the process returns to step S21D.

[0098] If the answer is Yes in step S22D, the processor 32 acquires the training data TM1 and TM2 from the memory 34 or an external device (step S20E). The processor 32 performs machine learning based on the acquired training data TM1 and TM2 to generate a trained model LMm (step S21E). The processor 32 or a human verifies the trained model LMm (step S22E). If the answer is No, the process returns to step S21E. If the answer is Yes in step S22E, the trained models LMn and LMm are output to the memory 34 or an external device (step S23). The process then ends. Note that the order of steps S20D to S22D and S20E to S22E can be set as appropriate. Furthermore, the training data TM1 and TM2 may be acquired in step S20D, and step S20E may not be performed.

[0099] (Y parameter calculation method) Fig. 22 is a flowchart showing a method for calculating the Y parameter in the third embodiment. As shown in Fig. 22, the acquiring unit 42 acquires information A about the bonding wire 10, information about the frequency f, the number n of the intervals D1, and the number m of the intervals D2 (step S30C).

[0100] Next, the calculation unit 44 acquires a learned model LMn (step S31C). The calculation unit 44 determines whether n+1=1 (step S32C). If yes, the calculation unit 44 estimates Y1 based on the learned model LMn from the acquired information A and frequency f (step S33C). Next, the calculation unit 44 sets the estimated Y1 to Yn (step S34C). Thereafter, the process proceeds to step S31E.

[0101] If the result of step S32C is No, the calculation unit 44 determines whether n+1=N1 (step S35C). If the result is Yes, the calculation unit 44 estimates YN1 based on the learned model LMn from the acquired information A and frequency f (step S36S). Next, the calculation unit 44 sets the estimated YN1 to Yn (step S37C). Then, the process proceeds to step S31E.

[0102] If the result of step S35C is No, the calculation unit 44 determines whether n+1=N2 (step S38C). If the result is Yes, the calculation unit 44 estimates YN2 based on the learned model LMn from the acquired information A and frequency f (step S39D). Next, the calculation unit 44 sets the estimated YN2 to Yn (step S40C). Then, the process proceeds to step S31E.

[0103] If the result of step S38C is No, the calculation unit 44 estimates YN1 and YN2 based on the learned model LMn from the acquired information A and frequency f (step S41C). Next, the calculation unit 44 calculates Yn from the estimated YN1 and YN2 (step S42C). For example, the calculation unit 44 calculates Yn from YN1 and YN2 using Equation 8.

[0104] Thereafter, the calculation unit 44 acquires the learned model LMm (step S31E). The calculation unit 44 determines whether m=M1 (step S35D). If yes, the calculation unit 44 estimates YM1 based on the learned model LMm from the acquired information A and frequency f (step S36D). Next, the calculation unit 44 sets the estimated YM1 as Ym (step S37D). Thereafter, the process proceeds to step S48.

[0105] If the result of step S35D is No, the calculation unit 44 determines whether m=M2 (step S38D). If the result is Yes, the calculation unit 44 estimates YM2 based on the learned model LMm from the acquired information A and frequency f (step S39D). Next, the calculation unit 44 sets the estimated YM2 as Ym (step S40D). Thereafter, the process proceeds to step S48.

[0106] If the result of step S38D is No, the calculation unit 44 estimates YM1 and YM2 based on the learned model LMm from the acquired information A and frequency f (step S41D). Next, the calculation unit 44 calculates Ym from the estimated YM1 and YM2 (step S42D). For example, the calculation unit 44 calculates Ym from YM1 and YM2 using Equation 9.

[0107] Next, the calculation unit 44 calculates Ynm from the calculated Yn and Ym (step S48). For example, the calculation unit 44 calculates Ynm from Yn and Ym using Equation 7. Thereafter, the output unit 46 outputs the Y parameter Ynm to the memory 34 or an external device (step S43). Then, the process ends. Note that the order of steps S32C to S34C, S35C to S37C, and S38C to S40C can be set as appropriate. The order of steps S35D to S37D, and S38D to S40D can be set as appropriate. The learned model LMm is acquired in step S31C, and step S31E may not be performed.

[0108] According to the third embodiment, for at least three bonding wires 10 connected in parallel between the first port P1 and the second port P2, the intervals between adjacent bonding wires 10 are D1 (first interval) and D2 (second interval different from the first interval). The number of intervals D1 is n (first number), and the number of intervals D2 is m (second number). The number n of intervals D1 is N-1, where N is the number in the first embodiment. That is, in the first embodiment, N1 corresponds to the case where the number n is N1-1, and N2 corresponds to the case where the number n is N2-1.

[0109] 19 and 20, trained models LMn and LMm are generated. Trained model LMn is generated by machine learning multiple pieces of training data T1, TN1, and TN2 that define the relationship between information A and number n+1 and Y parameters Y1, YN1, and YN2 obtained for information A and number n+1 when the number n+1 is 1, N1, and N2.

[0110] The trained model LMm is generated by machine learning multiple training data TM1 and TM2 that define the relationship between the information A and the number m, and the Y parameters YM1 and YM2 obtained for the information A and the number m, when the number m is M1 and N2.

[0111] 22, the calculation unit 44 estimates at least one of Y1, YN1, and YN2 based on the trained model LMn. As in steps S36D, S39D, and S41D, the calculation unit 44 estimates at least one of YM1 (fourth parameter) and YM2 (fifth parameter) based on another trained model LMm. As in steps S34C, S37C, S40C, S42C, S37D, S40D, S42D, and S48, the calculation unit 44 calculates Ynm (calculation parameter) for the number n and number m indicated by the second information based on at least one of Y1, YN1, YN2, YM1, and YM2. This allows for accurate calculation of Ynm when the spacing between the bonding wires 10 is two levels, D1 and D2.

[0112] The trained model LMn may include three trained models when the number n is 0, N1-1, and N2-1. The trained model LMm may include two trained models when the number m is M1 and M2.

[0113] In the first to third embodiments, the Y parameters have been described as an example of the circuit parameters. The circuit parameters are parameters expressed by matrices of the first port P1 and the second port P2. The circuit parameters may be, for example, S parameters (scattering matrices) or Z parameters (impedance matrices). The circuit parameters may also be parameters that can be uniquely calculated from the Y parameters.

[0114] The Y parameters provided in parallel can be calculated by addition. Therefore, the processes of step S42 in Figures 12 and 14, steps S42A and S42B in Figure 17, and steps S42C and S42D in Figure 22 are performed using the Y parameters. At least some of the other processes may use parameters that can be calculated uniquely from the Y parameters (for example, S parameters or Z parameters).

[0115] It becomes difficult to calculate the circuit parameters with high accuracy when there are a large number of bonding wires 10. From this viewpoint, when the number of bonding wires 10 is 10 or more or 50 or more, the first to third embodiments can be used.

[0116] Each process (each function) in the above-described embodiments is realized by a processing circuit including one or more processors. The processing circuit may be configured as an integrated circuit or the like that combines one or more memories, various analog circuits, and various digital circuits in addition to the one or more processors. The one or more memories store programs (instructions) that cause the one or more processors to execute each of the above processes. The one or more processors may execute each of the above processes according to the program read from the one or more memories, or may execute each of the above processes according to a logic circuit designed in advance to execute each of the above processes.

[0117] The processor may be any of various processors suitable for computer control, such as a CPU, a GPU (Graphics Processing Unit), a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), or an ASIC (Application Specification Integrated Circuit). The physically separated processors may cooperate with each other to execute the processes. For example, the processors installed in the physically separated computers may cooperate with each other via a network such as a LAN (Local Area Network), a WAN (Wide Area Network), or the Internet to execute the processes.

[0118] The program may be installed into the memory from an external server device or the like via the network, or may be distributed stored on a recording medium such as a CD-ROM, DVD-ROM, or semiconductor memory, and then installed into the memory from the recording medium.

[0119] The embodiments disclosed herein should be considered to be illustrative and not restrictive in all respects. The scope of the present disclosure is defined by the claims, not by the above meaning, and is intended to include all modifications within the meaning and scope equivalent to the claims. [Explanation of symbols]

[0120] 10, 10A, 10B, 10C Bonding Wire 11A, 11B pads 12 base 14, 15 leads 16, 17 Matching circuit 18 Semiconductor Devices 20 Semiconductor chips 21, 26 PCB 22, 23, 24, 27, 28 electrodes 25 Passive Chips 30 Computer 32 processors 34 memory 36 Input / Output Devices 38 Internal Bus 40 Calculation Device 42 Acquisition Department 44 Calculation section 46 Output section A (First Information) Information N (second information) Number f (third information) frequency D1 (first interval), D2 (second interval) interval Y1 (first parameter), YN1 (second parameter), YN2 (third parameter), YM1 (fourth parameter), YM2 (fifth parameter), YN (calculated parameter) Y parameter T1 (first teacher data), TN1 (second teacher data), TN2 (third teacher data) Teacher data LM1 (first trained model), LMN1 (second trained model), LMN2 (third trained model), LMn, LMm (another trained model) Trained model P1 First port P2 Second port

Claims

1. Computer, an acquiring unit that acquires first information regarding one or more bonding wires connected in parallel between a first port and a second port, excluding the number of the one or more bonding wires, and second information regarding the number of the bonding wires; a calculation unit that estimates at least one parameter from the first information based on a trained model, the first parameter being a circuit parameter representing the relationship between the first port and the second port when the number of ports is one, the second parameter being the circuit parameter when the number of ports is N1, which is two or more, and the third parameter being the circuit parameter when the number of ports is N2, which is greater than N1, and calculates from the second information, based on the at least one parameter, the calculation unit calculating a calculation parameter being the circuit parameter for the number of ports indicated by the second information; It functions as The trained model is generated by machine learning training data that defines the relationship between the first information and the number of lines and the circuit parameters obtained for the first information and the number of lines when the number of lines is 1, when the number of lines is N1, and when the number of lines is N2. program.

2. The program according to claim 1 , wherein the calculation unit calculates the calculation parameter based on the second parameter and the third parameter when the number indicated by the second information is other than 1, N1, or N2.

3. The calculation unit When the number indicated by the second information is one, the first parameter is calculated as the calculated parameter; When the number indicated by the second information is N1, the second parameter is calculated as the calculated parameter; When the number indicated by the second information is N2, the third parameter is calculated as the calculation parameter. The program according to claim 2.

4. The trained model is a first trained model generated by machine learning first teacher data that defines the relationship between the first information and the circuit parameters obtained for the first information when the number of the first circuit is one; and a second trained model generated by machine learning second teacher data that defines the relationship between the first information and the circuit parameters obtained for the first information when the number is N1; and a third trained model generated by machine learning third teacher data that defines the relationship between the first information and the circuit parameters obtained for the first information when the number is N2; and Including, The calculation unit When the number indicated by the second information is other than 1, N1, and N2, the second parameter is estimated from the first information based on the second trained model, and the third parameter is estimated from the first information based on the third trained model. The program according to any one of claims 1 to 3.

5. the trained model is a single trained model generated by machine learning a plurality of training data defining a relationship between the first information and the number of lines, and the circuit parameters obtained for the first information and the number of lines; When the number indicated by the second information is other than 1, N1, or N2, the calculation unit estimates the second parameter based on the one trained model from the first information and the number of N1, and estimates the third parameter based on the one trained model from the first information and the number of N2. The program according to any one of claims 1 to 3.

6. The trained model is generated by machine learning training data that defines a relationship between the first information and the number of lines, and the circuit parameters obtained for the first information and the number of lines, when the number is 1, when the number is N1, when the number is N2, and when the number is N3, which is greater than N1 and less than N2; when the number indicated by the second information is smaller than N3, the calculation unit estimates a fourth parameter, which is the circuit parameter when the number is N3, from the first information based on the trained model, and calculates the calculation parameter based on the second parameter and the fourth parameter; When the number indicated by the second information is greater than N3, the calculation parameter is calculated based on the fourth parameter and the third parameter. The program according to any one of claims 1 to 3.

7. the one or more bonding wires are at least three bonding wires connected in parallel between the first port and the second port; the spacing between adjacent bonding wires of the at least three bonding wires has a first spacing and a second spacing different from the first spacing; the second information includes a first number of the first intervals and a second number of the second intervals; The N1 number corresponds to when the first number is N1-1, and the N2 number corresponds to when the first number is N2-1, the calculation unit estimates at least one of a fourth parameter that is the circuit parameter when the second number is M1 and a fifth parameter that is the circuit parameter when the second number is M2 from the first information based on another learned model, and calculates, from the second information, a calculation parameter that is the circuit parameter for the first number and the second number indicated by the second information based on at least one parameter of the first parameter, the second parameter, and the third parameter and at least one parameter of the fourth parameter and the fifth parameter; The other trained model is generated by machine learning another training data that defines a relationship between the first information and the second number, and the circuit parameters obtained for the first information and the second number when the second number is M1 and when the second number is M2. The program according to any one of claims 1 to 3.

8. the one or more bonding wires are at least three bonding wires connected in parallel between the first port and the second port; the spacing between adjacent bonding wires of the at least three bonding wires is constant; The program according to any one of claims 1 to 3.

9. the acquisition unit acquires third information related to a frequency of a high-frequency signal transmitted between the first port and the second port; the calculation unit estimates the at least one parameter based on a trained model from the first information and the third information, and calculates the calculated parameter from the second information based on the at least one parameter; The trained model is generated by machine learning a plurality of training data that defines the relationship between the first information, the frequency, and the number of lines, and a plurality of the circuit parameters obtained for the first information, the frequency, and the number of lines, when the number of lines is 1, when the number of lines is N1, and when the number of lines is N2. The program according to any one of claims 1 to 3.

10. 4. The program according to claim 1, wherein the circuit parameters are S parameters, Y parameters, or Z parameters.

11. an acquiring step of acquiring first information regarding one or more bonding wires connected in parallel between a first port and a second port, excluding the number of the one or more bonding wires, and second information regarding the number of the bonding wires; a calculation step of estimating, based on a learned model from the first information, at least one parameter selected from the group consisting of a first parameter that is a circuit parameter representing the relationship between the first port and the second port when the number of ports is one, a second parameter that is the circuit parameter when the number of ports is N1, which is two or more, and a third parameter that is the circuit parameter when the number of ports is N2, which is greater than N1, and calculating, based on the at least one parameter from the second information, a calculation parameter that is the circuit parameter for the number of ports indicated by the second information; Including, The trained model is generated by machine learning training data that defines the relationship between the first information and the number of lines and the circuit parameters obtained for the first information and the number of lines when the number of lines is 1, when the number of lines is N1, and when the number of lines is N2. method.

12. an acquiring unit that acquires first information regarding one or more bonding wires connected in parallel between a first port and a second port, excluding the number of the one or more bonding wires, and second information regarding the number of the bonding wires; a calculation unit that estimates at least one parameter from the first information based on a trained model, the first parameter being a circuit parameter representing the relationship between the first port and the second port when the number of ports is one, the second parameter being the circuit parameter when the number of ports is N1, which is two or more, and the third parameter being the circuit parameter when the number of ports is N2, which is greater than N1, and calculates from the second information, based on the at least one parameter, the calculation unit calculating a calculation parameter being the circuit parameter for the number of ports indicated by the second information; Equipped with The trained model is generated by machine learning training data that defines the relationship between the first information and the number of lines and the circuit parameters obtained for the first information and the number of lines when the number of lines is 1, when the number of lines is N1, and when the number of lines is N2. Calculation device.

Images