Rapid selection method for magnet material of LTCF magnetic bead with simple structure

Abstract

The invention relates to the technical field of LTCF laminated devices, in particular to a rapid selection method for a simple-structure LTCF magnetic bead magnet material. On the basis of an equivalent circuit, equivalent circuit simulation is adopted, a prediction calculation process integrated with machine learning and a genetic algorithm is added before actual preparation, and therefore higher and faster magnet material selection is achieved. Due to the fact that machine learning and a genetic algorithm are integrated in the prediction calculation process, the calculation efficiency and precision are greatly improved, the finally selected magnet material is higher and faster, and compared with the prior art, on the basis of meeting performance indexes, the time cost of LTCF magnetic bead design is reduced; and the working frequency range of the magnetic beads can be widened better, so that the use quantity of the LTCF magnetic beads is reduced, and the loss of effective signals caused by the use of the magnetic beads is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of LTCF laminated chip devices, in particular to a method for quickly selecting magnetic materials for LTCF magnetic beads with a simple structure. Background Art

[0002] As modern electronic devices continue to increase their demand for high-frequency signal processing, ferrite beads, as important electromagnetic interference (EMI) suppression components, play an important role in high-frequency noise filtering and signal integrity protection.

[0003] The traditional magnetic bead design process is to select an LTCF (low-temperature co-fired ferrite) magnetic bead with similar performance indicators based on the performance index requirements of the LTCF magnetic bead, and then use a certain brand of ferrite powder to prepare the magnetic beads of this structure based on experience; if the prepared magnetic beads basically meet the performance indicators, then slightly modify the magnetic bead design structure until the performance indicators are fully met; if the prepared magnetic beads deviate significantly from the performance indicators, reselect the powder preparation until the performance indicators are basically met. Although this method of designing magnetic beads is effective, it has the problem of a long magnetic bead design cycle, often going through more than 5 magnetic bead preparation cycles.

[0004] For wide-band signal transmission, a single ferrite bead usually cannot meet the needs of suppressing noise signals, and multiple ferrite beads are often required to be used in series and parallel to ensure the signal transmission effect. However, this also increases the number of components in the circuit, which will lead to excessive signal attenuation, damage circuit impedance matching, aggravate self-resonance problems, and increase the difficulty of design and manufacturing, which is not conducive to the stability of electronic circuits and the development trend of miniaturization. Summary of the Invention

[0005] In response to the above-mentioned problems or shortcomings, and in order to solve the problem of long design cycle of existing LTCF magnetic beads, the present invention proposes a method for quickly selecting magnetic materials for simple-structure LTCF magnetic beads, which is based on equivalent circuits and combined with machine learning or genetic algorithms to achieve higher and faster magnetic material selection, thereby reducing the time cost of LTCF magnetic bead design while meeting performance indicators; it can also better consider broadening the operating frequency band of the magnetic beads, thereby reducing the number of LTCF magnetic beads used and reducing the loss of effective signals caused by the use of magnetic beads.

[0006] The present invention adopts the following technical solutions:

[0007] A quick selection method for simple structure LTCF magnetic bead magnet materials, the specific steps are as follows:

[0008] Step 1: Generate equivalent circuit structure parameters for simple structure LTCF magnetic beads; the simple structure refers to a structure in which the magnetic beads are prepared with a single component and do not have heterogeneous stacking.

[0009] According to the working principle of LTCF magnetic beads and their working status at different frequencies, their impedance sources are divided into equivalent resistance, DC resistance, parasitic capacitance, and equivalent inductance; the three equivalent components of equivalent resistance, equivalent inductance, and parasitic capacitance are connected in parallel and then in series with the DC resistance to form the equivalent circuit of the LTCF magnetic bead.

[0010] Step 2: Calculate the impedance characteristic curve of the magnetic bead using the circuit element impedance series-parallel calculation method, specifically:

[0011]

[0012] Where Z is the impedance of the bead; R DC It is the resistance of the electrode inside the magnetic bead, and its value is only related to the material used for the electrode inside the magnetic bead; the equivalent resistance of the magnetic bead R = μ″ωL0, L0 is the ring coil constant, μ" is the imaginary part of the magnetic permeability, ω is the frequency; j is the imaginary unit; L is the equivalent inductance of the magnetic bead, L = μ′L0, μ′ is the real part of the magnetic permeability; C is the parasitic capacitance.

[0013] The above formula can be written as follows:

[0014] Z=R E +Xj

[0015] in:

[0016]

[0017] Step 3: Since L0 and C are parameters related to the bead structure, they can be regarded as constants when the bead structure remains unchanged. The impedance characteristic curve of the LTCF bead is obtained by measurement (such as using an impedance analyzer). The impedance characteristic curve is decomposed using the equivalent circuit described in step 2 to obtain the equivalent parameters: the toroidal coil constant L o and the value of the parasitic capacitance C.

[0018] Step 4: The working frequency band of the magnetic bead is the frequency range where the real part modulus of the magnetic bead impedance ≥ the imaginary part modulus of the impedance, that is, it satisfies |R E |≥|X|, we can make R E =X, R E =-X to solve the two frequencies ω1 and ω2, and ω1~ω2 is the working frequency band of the magnetic bead.

[0019] To expand the working frequency band of the existing magnetic beads, first clarify the working frequency band of the magnetic beads required by the design, ω1~ω2, and use the formula in step 2 as a function. When ω=ω1, R E =X, ω = ω2 when R E =-X, as two constraints, are solved using genetic algorithms or machine learning, and ultimately the material properties that meet the magnetic permeability requirements of the magnetic beads design are obtained.

[0020] Step 5: Search the material database for materials that meet the properties solved in step 4, or develop new materials with this goal in mind, and use materials that meet the requirements to prepare magnetic beads.

[0021] Furthermore, if the shapes of the electrode coils between the layers inside the magnetic bead structure are the same, the solution is more accurate and the present invention is more applicable.

[0022] Furthermore, if the inner electrodes of the magnetic bead structure are all in a series structure, the solution is more accurate and the present invention is more applicable.

[0023] Furthermore, the machine learning in step 4 is specifically as follows: the data set is constructed from the parameters of the magnetic beads and the ferrite used to prepare the magnetic beads: its feature input is the working parameters of the magnetic beads, including the working starting frequency f s , operating bandwidth Δf; the characteristic output is the performance parameters of the ferrite used to prepare the magnetic beads, including the initial magnetic permeability μ i , the maximum value of the real part of magnetic permeability μ′ max , the maximum imaginary part of magnetic permeability μ" max , cutoff frequency f r Divide the data set into a training set and a test set (e.g., 80% of the data is used as the training set and the remaining 20% of the data is used as the test set). Substitute the feature output into the formula in step 2 to obtain the difference between ω1~ω2 and the design requirements as the residual function, and use the residual function to evaluate its calculation effect.

[0024] Furthermore, the working parameters also include the working cutoff frequency f e 、Impedance Z at 100MHz 100 and the maximum impedance Z max , in order to improve the accuracy of the final result.

[0025] Furthermore, the model used in the machine learning is a random forest, naive Bayes or multi-layer perceptron MLP model.

[0026] Furthermore, the genetic algorithm in step 4 is specifically as follows: because the shape of the permeability spectrum of the ferrite beads is relatively fixed, such as Figure 5As shown. Fit the ferrite permeability spectrum into a polynomial or spline curve; then substitute the fitting result into the impedance calculation formula in step 2, and the problem of solving the ferrite material permeability is transformed into the problem of solving the coefficient of the fitting result. The inverse sum of the squares of the difference between the calculated theoretical magnetic bead operating bandwidth and the magnetic bead operating bandwidth design requirement is used as the fitness function, and then use selection operators (such as roulette method, random competitive selection method), crossover operators (such as single point crossover, uniform crossover), and mutation operators (such as Gaussian mutation, uniform mutation); the termination condition is 300-1000 iteration cycles or the goal is to reach the optimal solution. The solution result is the performance corresponding to the target ferrite material.

[0027] In summary, the present invention is based on an equivalent circuit and uses equivalent circuit simulation. Before actual preparation, a predictive calculation process that incorporates machine learning and genetic algorithms is added, thereby achieving higher and faster selection of magnetic materials. Since the integration of machine learning and genetic algorithms into the predictive calculation process greatly improves the calculation efficiency and accuracy, the magnetic material finally selected is higher and faster. Compared with the existing technology, on the basis of meeting performance indicators, the time cost of LTCF magnetic bead design is reduced; the present invention can effectively guide the expansion of the working frequency band of the magnetic beads, cover wider application scenarios, reduce the negative impact caused by the traditional method of using multiple magnetic beads in series and parallel, and reduce the loss of effective signals caused by the use of multiple magnetic beads. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 is an equivalent circuit diagram of the LTCF magnetic bead of the embodiment;

[0029] Figure 2 1 is a schematic diagram of the overall structure of the magnetic beads in an embodiment;

[0030] Figure 3 Schematic diagram of the internal structure of the magnetic beads in the embodiment;

[0031] Figure 4 is an impedance characteristic curve measured in the embodiment;

[0032] Figure 5 This is an example of the magnetic permeability of ferrite used in preparing magnetic beads in the embodiment. DETAILED DESCRIPTION

[0033] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0034] Although the traditional method of designing magnetic beads is effective, it has the problem of a long magnetic bead design cycle, which often requires more than 5 magnetic bead preparation cycles. Based on this, the present invention proposes a method for quickly selecting magnet materials for simple structure LTCF magnetic beads to reduce the magnetic bead design time.

[0035] Example:

[0036] Step 1: The equivalent circuit of the simple structure LTCF magnetic bead involved in this embodiment is as follows: Figure 1 As shown, including the equivalent inductance R, equivalent inductance jωL, parasitic capacitance connected in parallel Three equivalent components, and the DC resistance R in series DC .

[0037] Step 2: According to the equivalent circuit, the impedance characteristic calculation formula can be calculated as follows:

[0038]

[0039] The above formula can be expanded according to the phase circuit series and parallel calculation method:

[0040]

[0041] Simplifying the above formula, the impedance calculation formula is:

[0042]

[0043] Step 3: Select the LTCF magnetic beads with a certain structure and measure their impedance characteristic curve. For example, in this embodiment, select Figure 3 The impedance characteristic curve of the LTCF magnetic bead is as follows: Figure 4 shown.

[0044] Use an impedance analyzer (or mathematical tool) to decompose the impedance curve and fit the value of the equivalent parameter. For example, in this embodiment, the DC resistance R DC , equivalent inductance R, inductance L, capacitance C.

[0045] Step 4: The working frequency band of the magnetic bead is the frequency range where the real part modulus of the magnetic bead impedance ≥ the imaginary part modulus of the impedance, that is, R E ≥X frequency band, R E =X, R E =-X to solve the two frequencies of ω1 and ω2, ω1~ω2 is the working frequency band of the magnetic beads;

[0046] To expand the working frequency band of the existing magnetic beads, first clarify the working frequency band of the magnetic beads required by the design, ω1~ω2, and use the formula in step 2 as a function. When ω=ω1, R E =X, ω = v2 when R E =-X, as two constraints; using genetic algorithms or machine learning to solve, ultimately derive the material properties that meet the magnetic permeability requirements of the magnetic bead design;

[0047] Step 5: Search the material database for materials that meet the properties solved in step 4, or develop new materials with this goal in mind, and use materials that meet the requirements to prepare magnetic beads.

[0048] As verification example 1, the genetic algorithm used in step 4 is specifically as follows:

[0049] Substitute different brands of magnets and use the ferrite permeability to derive the impedance curve. For example, in this embodiment, the substitution method is:

[0050]

[0051] Among them L n is the inductance after substitution; L1 is the inductance L fitted in the previous step; μ′1 is the real part of the magnetic permeability of the ferrite material of the LTCF magnetic bead whose impedance is measured; μ′ n is the real part of the permeability of the new brand of ferrite; R n is the equivalent resistance after substitution; R1 is the equivalent resistance R fitted in the above step 2; μ"1 is the imaginary part of the magnetic permeability of the ferrite material of the LTCF magnetic bead whose impedance is measured; μ" n is the imaginary part of the magnetic permeability of the new brand of ferrite.

[0052] Because the shape of the permeability spectrum of ferrite beads is relatively fixed, such as Figure 5 As shown in the figure, the parasitic capacitance is related to the structure of the LTCF bead, so the parasitic capacitance can be calculated using the previously fitted results without changing the structure. The DC resistance is only related to the printed electrode silver paste used in the preparation of the LTCF bead, so the previously fitted results can be used without changing the type of electrode silver paste.

[0053] After substitution, a new impedance characteristic curve is obtained. For example, in this embodiment, the formula after substitution is:

[0054]

[0055] Determine whether the newly obtained impedance characteristic curve meets the design requirements.

[0056] As a verification example 2, the machine learning in step 4 is specifically as follows:

[0057] The parameters of magnetic beads and ferrite used to prepare magnetic beads are selected to construct a data set, and its feature input is the working parameters of the magnetic beads (working starting frequency f s , working cut-off frequency f e , operating bandwidth Δf, impedance Z at 100MHz 100 , impedance maximum Z max ); the characteristic output is the performance parameters of the ferrite for preparing the magnetic beads (initial magnetic permeability μ i , the maximum value of the real part of magnetic permeability μ′max , the maximum imaginary part of magnetic permeability μ" max , cutoff frequency f r ).

[0058] 80% of the data in the dataset is used as the training set, and the remaining 20% of the data is used as the test set. The output results (ferrite material properties) are substituted into the impedance calculation formula. The substitution method is:

[0059]

[0060] The relationship between permeability and impedance can be obtained. Substituting the output data into the theoretical operating bandwidth of the bead can be obtained. The difference between the theoretical operating bandwidth of the bead and the operating bandwidth of the bead that meets the design requirements is used as the residual function to evaluate the solution.

[0061] A multi-layer perceptron (MLP) model is used as the model used in machine learning to solve the performance of the required ferrite.

[0062] Compare the calculated ferrite material properties with the existing material library and select a ferrite material that meets the requirements for manufacturing the magnetic beads. If no fully matching material exists, the solution results can be used as a design requirement to guide material preparation. Finally, the selected ferrite material can be used to manufacture the magnetic beads.

[0063] It can be seen from the above embodiments that the present invention is based on an equivalent circuit, and adds a predictive calculation process that incorporates machine learning and genetic algorithms before actual preparation, utilizing the inherent advantages of machine learning and genetic algorithms in computing efficiency and accuracy. Compared with the existing technology, on the basis of meeting performance indicators, it achieves higher and faster magnetic material selection and reduces the time cost of LTCF magnetic bead design; it can also better consider broadening the working frequency band of the magnetic beads, thereby reducing the number of LTCF magnetic beads used and reducing the loss of effective signals caused by the use of magnetic beads.

Claims

1. A method for quickly selecting magnet materials for simple structure LTCF magnetic beads, characterized in that: The specific steps are as follows: Step 1: Generate equivalent circuit structure parameters for simple structure LTCF magnetic beads; the simple structure refers to a structure in which the magnetic beads are prepared with a single component and do not have heterogeneous stacking; According to the working principle of LTCF magnetic beads and the working state at different frequencies, its impedance sources are divided into equivalent resistance, DC resistance, parasitic capacitance, and equivalent inductance. The three equivalent components of equivalent resistance, equivalent inductance, and parasitic capacitance are connected in parallel and then connected in series with the DC resistance to form the equivalent circuit of the LTCF magnetic bead. Step 2: Calculate the impedance characteristic curve of the magnetic bead using the circuit element impedance series-parallel calculation method, specifically: Where Z is the impedance of the bead; R DC is the resistance of the electrode inside the magnetic bead; the equivalent resistance of the magnetic bead R = μ″ωL0, L0 is the ring coil constant, μ″ is the imaginary part of the magnetic permeability, ω is the frequency; j is the imaginary unit; L is the equivalent inductance of the magnetic bead, L = μ′L0, μ′ is the real part of the magnetic permeability; C is the parasitic capacitance; The above formula can be written as follows: Z=R E +Xj in: Step 3. Since L0 and C are parameters related to the bead structure, they can be regarded as constants when the bead structure remains unchanged. The impedance characteristic curve of the LTCF bead is obtained by measurement. The impedance characteristic curve is decomposed using the equivalent circuit described in step 2 to obtain the equivalent parameters: the toroidal coil constant L0 and the parasitic capacitance C. Step 4: The working frequency band of the magnetic bead is the frequency range where the real part modulus of the magnetic bead impedance ≥ the imaginary part modulus of the impedance, that is, it satisfies |R E |≥|X|, we can make R E =X, R E =-X to solve the two frequencies of ω1 and ω2, ω1~ω2 is the working frequency band of the magnetic beads; To expand the working frequency band of the existing magnetic beads, first clarify the working frequency band of the magnetic beads required by the design, ω1~ω2, and use the formula in step 2 as a function. When ω=ω1, R E =ω, ω = ω2 when R E =-X, as two constraints; using genetic algorithms or machine learning to solve, ultimately derive the material properties that meet the magnetic permeability requirements of the magnetic bead design; Step 5: Search the material database for materials that meet the properties solved in step 4, or develop new materials with this goal in mind, and use materials that meet the requirements to prepare magnetic beads.

2. The method for quickly selecting a magnet material for a simple structure LTCF magnetic bead according to claim 1, wherein: The shapes of the electrode coils between the layers inside the magnetic bead structure are the same.

3. The method for quickly selecting a magnet material for a simple structure LTCF magnetic bead according to claim 1, wherein: The inner electrodes of the magnetic bead structure are all in series structure.

4. The method for quickly selecting a simple structure LTCF magnetic bead magnet material according to claim 1, wherein: The machine learning in step 4 is specifically as follows: The dataset is constructed from the parameters of magnetic beads and ferrite used to prepare the beads: its feature input is the working parameter of the magnetic bead, including the working starting frequency f s and operating bandwidth Δf; the characteristic output is the performance parameters of the ferrite used to prepare the magnetic beads, including the initial magnetic permeability μ i , the maximum value of the real part of magnetic permeability μ′ max , the maximum value of the imaginary part of magnetic permeability μ″ max , cutoff frequency f r ; The data set is divided into a training set and a test set. The difference between ω1~ω2 obtained by solving the formula in step 2 using the feature output and the design requirements is used as the residual function, and the residual function is used to evaluate its calculation effect.

5. The method for quickly selecting a magnet material for a simple structure LTCF magnetic bead according to claim 4, wherein: The working parameters also include the working cut-off frequency f e 、Impedance Z at 100MHz 100 and the maximum impedance Z max .

6. The method for quickly selecting a magnet material for a simple structure LTCF magnetic bead according to claim 4, wherein: The model used in the machine learning is a random forest, naive Bayes or multi-layer perceptron MLP model.

7. The method for quickly selecting a simple structure LTCF magnetic bead magnet material according to claim 1, wherein: The genetic algorithm in step 4 is specifically as follows: Fit the ferrite permeability spectrum into a polynomial or spline curve; then substitute the fitting result into the impedance calculation formula in step 2, converting the problem of solving the ferrite material permeability into the problem of solving the coefficient of the fitting result; The inverse of the sum of the squares of the differences between the calculated theoretical operating bandwidth of the ferrite bead and the design requirements of the operating bandwidth of the ferrite bead is used as the fitness function, and then combined with the selection operator, crossover operator, and mutation operator; the termination condition is 300-1000 iterations or the target reaches the optimal solution; the solution result is the performance corresponding to the target ferrite material.

8. The method for quickly selecting a magnetic material for a simple-structure LTCF magnetic bead according to claim 7, wherein: The selection operator is a roulette method or a random competition selection method.

9. The method for quickly selecting a magnetic material for a simple structure LTCF magnetic bead according to claim 7, wherein: Said: The crossover operator is single-point crossover or uniform crossover.

10. The method for quickly selecting magnetic materials for simple-structure LTCF magnetic beads according to claim 7, characterized in that: Said: The mutation operator is Gaussian mutation or uniform mutation.

Images