Efficient Optimization Method for Designing Defected Ground Structure-Based Common-Mode Filters

Abstract

An efficient optimization method for designing defected ground structure (DGS)-based common-mode filters (CMFs) is proposed, utilizing equation-based transmission line models integrated with a genetic algorithm (GA). Designing an optimal DGS-based CMF using full-wave simulation tools is time-consuming due to its process-intensive nature. The proposed optimization method implements transmission line theory to allow for direct S-parameter calculation, enabling integration with an optimization algorithm to identify optimal parameters within a confined 5 mm × 10 mm design space. This work demonstrates a compact asymmetric DGS design to illustrate the method’s capability. The resulting compact asymmetric DGS-based CMF achieves wideband common-mode suppression with a –10 dB bandwidth from 3.18 GHz to 12.89 GHz. The optimization method significantly reduces design time by minimizing the need for lengthy and repetitive full-wave simulations. The measured S-parameters of the fabricated CMF closely match the simulated results, validating the model’s accuracy. Compared with traditional methods for designing DGS-based CMFs, the proposed method utilizes transmission line theory to optimize the design efficiently, providing a practical and efficient solution.

1. Introduction

In high-speed digital circuits, maintaining high-quality signals is crucial to signal integrity. Differential signals are often used to achieve good signal integrity due to their high tolerance for noise [1]. The characteristics of differential signals allow the signals to be resistant to external noise, crosstalk, and electromagnetic interference (EMI) [2]. As the demand for compact design structures increases, systems are becoming more prone to unbalanced routing and timing skew in differential line designs, leading to increased common-mode (CM) noise [3]. Consequently, designing compact common-mode filters (CMFs) has become essential to fitting within the constrained design space. Among compact CMFs, defected ground structure (DGS)-based CMFs are widely used for their compactness, ease of implementation, and effectiveness in reducing CM noise.

DGSs are defects made on the ground plane of microwave circuits to disrupt the return current distribution and control electrical properties. Their simple design and cost-effectiveness make them suitable for differential circuits, where they suppress CM noise without significantly affecting differential-mode signal quality. DGSs can be modeled as parallel LC circuits for CM signals, generating a stopband that suppresses CM noise while minimally impacting differential-mode signals and preserving signal integrity.

Diverse DGS-based CMFs have been proposed, focusing on wideband CM suppression in the lower frequency range. These CMFs utilize DGSs of various shapes, such as dumbbell [4], U-H [5], and C [6]. Although these designs achieve wideband CM suppression, they lack compactness. To address this, several compact DGS-based CMFs have been introduced [7,8,9,10,11,12,13,14,15,16,17,18,19,20,21], including recent advancements in absorptive CMFs [22,23,24,25,26,27,28]. Efficient optimization methods for designing an optimal filter remain underdeveloped. Previous methods, such as the proximal policy optimization (PPO) algorithm-assisted design approach [29], rely on multiple iterations of full-wave simulations, making them highly time-intensive. Additionally, the typical design process for DGS-based CMFs involves designing resonators with full-wave simulation tools, calculating mutual coupling using LC components, and determining the optimal coupling distance.

This paper proposes an efficient optimization method for designing DGS-based CMFs using equation-based transmission line modeling and introduces an asymmetric CMF design optimized using the proposed method. A previous approach to modeling DGS-based CMFs using transmission line models [30] calculated the equivalent circuit through a two-step process that relied on analyzing the resonant frequency along with the impedance and electrical length of the transmission lines. This two-step process restricts the ability to calculate the equivalent circuit parameters dynamically, limiting its utility in dynamic optimization processes. In contrast, our method overcomes this limitation by dynamically calculating the equivalent circuit directly, representing the DGS structure with multiple slotlines of varying characteristic impedance and effective dielectric constants. An asymmetric CMF design, optimized using the proposed method, is introduced to demonstrate its applicability and effectiveness. This application highlights the versatility of the method in achieving efficient optimization for a compact, asymmetric DGS design.

The proposed compact wideband CMF design is optimally designed using the proposed optimization method. In contrast to previous approaches that utilize LC modeling and require a time-consuming two-step process of performing full-wave simulations for analysis and the construction of equivalent circuits, our equation-based method directly models the DGS with transmission line theory. Our approach allows S-parameters to be calculated in real time, enabling its integration with optimization algorithms. The model is integrated with a genetic algorithm (GA) [31], an optimization algorithm, to efficiently design an optimized DGS-based CMF. Unlike methods that apply GA directly to full-wave simulation tools, which require multiple simulations and result in long computation times, our method integrates GA with the transmission line model, resulting in reduced simulation time. This method requires only a single full-wave simulation for final validation. An optimal DGS-based CMF within a specified design space can be efficiently designed using the proposed optimization method.

2. Design Configuration

Figure 1 illustrates the proposed DGS-based CMF design with differential transmission lines (DTLs). Configured on a printed circuit board (PCB) made of FR4 epoxy (${\epsilon }_r=4.4$ and $\tan \delta =0.02$) with a height of 0.4 mm, the design consists of DTLs with width w of 0.68 mm and separation s of 1.0 mm. The DTLs are designed to establish a differential-mode characteristic impedance of 100 $\Omega$. To achieve compactness, the optimized DGS pattern etched into the ground layer of the PCB has dimensions of 5 mm × 10 mm ($l_{\mathrm{dgs}}\times h_{\mathrm{dgs}})$.

The asymmetric DGS pattern, designed with rectangular features, is modeled using transmission lines. While its inhomogeneous nature presents challenges for accurate modeling, it can still be modeled efficiently using the proposed method. The inhomogeneities at the junctions between the different transmission line sections of the DGS, caused by discontinuities, are compensated using correction factors derived by comparing the circuit model with ANSYS HFSS 2023 R1 simulation results. To account for the most significant coupling effect induced by the coupling between the left and right DGS sections, one coupling section is modeled as a coupled slotline. Other coupling areas, where the left and right DGS sections are closely positioned, are not considered, which reduces the model’s accuracy. However, even without including these additional coupling effects, the model accurately predicts the common-mode suppression bandwidth. The structure is intentionally designed with rectangular segments, enabling modeling through transmission lines and the application of optimization algorithms. This design scheme overcomes modeling challenges and fits the DGS design into a compact form factor, making it suitable for practical applications.

Even-mode half-circuit analysis is employed to model the proposed CMF design for the efficient evaluation of $S_{\mathrm{cc}21}$ (common-mode insertion loss). The half-circuit analysis simplifies the model to be calculated using the proposed modeling method. The CMF design consists of five slotline segments with widths $w_1$ to $w_5$ and lengths $l_1$ to $l_5$, as well as one symmetric coupled slotline with width $w_3$ = $w_4$ and length $l_{\mathrm{cs}}$, as shown in Figure 2. These dimensions serve as parameters for the optimization algorithm. Transmission line theory is applied to each segment of the DGS to incorporate the characteristic impedances ($Z_0$) and effective dielectric constants (${\epsilon }_e$) of the slotlines and coupled slotlines. With the overall size of the DGS fixed, the parameters of each segment are optimized within the design space to achieve an optimal design for CM suppression.

3. Modeling Methodology

3.1. Setting Up the Equivalent Circuit Model

The proposed design is modeled as an equation-based even-mode equivalent circuit model and implemented in MATLAB R2024a to enable efficient optimization. Figure 3 illustrates a circuit using three transmission line types: microstrip lines derived from the DTLs, slotlines originating in the DGS, and coupled slotlines also part of the DGS. The characteristics of each transmission line, including their $Z_0$ and ${\epsilon }_e$, are inputs required for modeling. The $Z_0$ and ${\epsilon }_e$ of the microstrip line ($Z_{0,\mathrm{ms}}$ and ${\epsilon }_{e,\mathrm{ms}}$) are calculated using closed-form expressions from [32]. For the slotlines, the frequency-dependent $Z_0$ and ${\epsilon }_e$ ($Z_{0,1}$ to $Z_{0,5}$ and ${\epsilon }_{e,1}$ to ${\epsilon }_{e,5}$) are extracted using ANSYS Q2D Extractor 2023 R1. The same tool is also used to obtain the frequency-dependent even- and odd-mode $Z_0$ and ${\epsilon }_e$ ($Z_{0,e}$, $Z_{0,o}$, ${\epsilon }_{e,e}$, and ${\epsilon }_{e,o}$) of the coupled slotline.

The extracted $Z_0$ and ${\epsilon }_e$ of the slotlines and coupled slotline vary with frequency, making them well-suited for an equation-based transmission line model. Using frequency-dependent $Z_0$ and ${\epsilon }_e$ in the model improves the accuracy of S-parameter calculations over the entire frequency range. For instance, a slotline with width and height of 0.4 mm yields $Z_0$ increasing from 101.7 $\Omega$ to 102.8 $\Omega$ and ${\epsilon }_e$ decreasing from 1.79 to 1.69 as the frequency increases from 50 MHz to 15 GHz, providing an example of how dimensions affect frequency-dependent behavior.

3.2. Transmission Line Modeling

When modeling the CMF design, the DGS section is calculated first and then combined with the microstrip line section. The DGS section consists of slotline and coupled slotline models. The accuracy of using transmission line modeling for the DGS is validated by dividing the design into the left and right sections, excluding the coupling effect in the coupled slotline in the middle. Each section is modeled separately in MATLAB R2024a, and the calculated $S_{\mathrm{cc}21}$ results are compared with full-wave simulation results using ANSYS HFSS 2023 R1.

To accurately model each section of the DGS, the resonant frequency of the left and right sections of the DGS in MATLAB R2024a should align with the ANSYS HFSS 2023 R1 results. As shown in Figure 2, uncovered gaps are present at the junctions in the DGS that require consideration during modeling. The lengths of the slotlines are modified to account for the effects of junctions among several slotlines. Two junctions with uncovered gaps are identified: one at the intersection of $l_1$, $l_2$, and $l_3$ and the other at the intersection of $l_4$ and $l_5$. Correction factors are applied to the slotline lengths to compensate for the uncovered gaps, resulting in adjusted lengths $l_1^{\prime }$, $l_2^{\prime }$, and $l_5^{\prime }$, defined as

$$\begin{array}{cc}\hfill l_1^{\prime }& =l_1+k(w_3/2)\hfill \\ \hfill l_2^{\prime }& =l_2+k(w_3/2)\hfill \\ \hfill l_5^{\prime }& =l_5+w_4.\hfill \end{array}$$

(1)

The adjusted lengths are reflected in the equivalent circuit presented in Figure 3.

A correction factor of k = 1.2 is derived by fine-tuning to match the resonant frequency of the left DGS section. In the right DGS section, the width $w_4$ is added to $l_5$, covering the gap at the intersection where $l_4$ and $l_5$ meet. Figure 4a,b compare the $S_{\mathrm{cc}21}$ results from ANSYS HFSS 2023 R1 and MATLAB R2024a for the left and right sections of the DGS, with the correction factor applied to the model. The results displaying accurate resonant frequencies for both sections confirm the validity of using slotlines to model DGSs. With the left and right DGS sections having geometric asymmetry, the two shapes create high and low resonance frequencies at 8.01 and 6.89 GHz, respectively, leading to electrical asymmetry. With the asymmetric structure coupled, the asymmetric DGS produces wideband common-mode suppression bandwidth.

While the slotlines are modeled using transmission line theory, the coupled slotline is defined using the ABCD parameters as given by the following equation [33]:

$$\begin{array}{ll}\hfill A& =D={\displaystyle \frac{Y_{0,e}\cot {\theta }_e+Y_{0,o}\cot {\theta }_o}{\Delta }}\hfill \\ \hfill B& ={\displaystyle \frac{2j}{\Delta }}\hfill \\ \hfill C& ={\displaystyle \frac{j}{2\Delta }}\left[Y_{0,e}^2+Y_{0,o}^2-2Y_{0,e}{Y}_{0,o}\left(\cot {\theta }_e\cot {\theta }_o+\csc {\theta }_e\csc {\theta }_o\right)\right]\hfill \\ \hfill \Delta & =Y_{0,e}\csc {\theta }_e-Y_{0,o}\csc {\theta }_o,\hfill \end{array}$$

(2)

where the even- and odd-mode electrical lengths (${\theta }_e$, ${\theta }_o$) and admittances ($Y_{0,e}$, $Y_{0,o}$) are determined from even- and odd-mode $Z_0$ and ${\epsilon }_e$ ($Z_{0,e}$, $Z_{0,o}$, ${\epsilon }_{e,e}$, and ${\epsilon }_{e,o}$). Equation (2) calculates the frequency-dependent ABCD parameters for a short-circuited interdigital coupled line section.

The ABCD parameters of the slotlines and coupled slotline are cascaded by multiplying their respective matrices. The cascaded ABCD matrix is added through impedance parameters in series with the microstrip line, with length $l_{\mathrm{dgs}}=5\mathrm{mm}$, above the DGS section. Finally, the resulting series ABCD parameters are cascaded with the remaining microstrip line sections, with length $l_{\mathrm{ms}}=7.5\mathrm{mm}$. The system’s $S_{\mathrm{cc}21}$ is then calculated by transforming the total ABCD parameters.

3.3. DGS Modeling Validation

Figure 5 shows the S-parameter magnitudes obtained from the equation-based transmission line model and full-wave simulation of the optimized DGS design. The dimensions of the optimized DGS design labeled in

Table 1. Dimensions of the optimized DGS design.

w1 (mm)	w2 (mm)	w3 (mm)	w4 (mm)	w5 (mm)	l1 (mm)	l2 (mm)	l3 (mm)	l4 (mm)	l5 (mm)	lcs (mm)
0.4	0.4	1.4	1.4	0.4	1.8	1.8	0.5	0.5	5	3.2

were found using the optimization algorithm described in Section 4. Discrepancies in S-parameter magnitude arise when comparing the equation-based transmission line model with full-wave simulations. These stem from treating DGS sections as isolated slotlines, except for one coupled slotline section with the most coupling effect. Modeling the DGS sections as slotlines limits the model’s applicability to rectangular shapes that can be represented as transmission lines. As coupling and parasitic effects become more significant, the model’s accuracy would be reduced. The model prioritizes computational efficiency by excluding parasitic effects inherent to full-wave simulations. Nevertheless, frequency-dependent $Z_0$ and ${\epsilon }_e$ effectively capture slotline behavior over a wide range, enabling the model to handle larger variations. Although these simplifications cause slight deviations in $S_{\mathrm{cc}21}$ suppression levels, they do not compromise the model’s ability to predict the suppression bandwidth.

The accuracy of the proposed modeling method is validated by comparing the $S_{\mathrm{cc}21}$ results from MATLAB R2024a and ANSYS HFSS 2023 R1. Both results exhibit high accuracy in calculating the CM suppression bandwidth. The CM suppression bandwidth under −10 dB ranges from 3.27 GHz to 12.81 GHz for the full-wave simulation and from 3.12 GHz to 12.70 GHz for the proposed modeling method. These results show close agreement in the CM suppression bandwidth.

Previous equivalent circuits modeled using LC modeling in [7,8,9,10,11,12,13] relied on LC modeling, which requires a two-step process of calculating the equivalent circuit. The two-step process involves identifying resonant frequencies through full-wave simulation and then using those frequencies to calculate the S-parameters. However, this approach cannot dynamically update the equivalent circuit when DGS dimensions vary, making it unsuitable for optimization algorithms. As a result, LC modeling is unsuitable for iterative calculations of CM suppression bandwidth across multiple DGS dimension settings. The proposed method overcomes this limitation by dynamically calculating the equivalent circuit using equation-based transmission line modeling, representing the DGS with slotlines and coupled slotlines. Based on the ABCD parameters of the transmission line segments, the equivalent circuit model accurately calculates the CM bandwidth. This approach enables the efficient optimization of the CMF design within a limited design space for various dimensional configurations.

Summary of the Equation-Based Modeling Method

• Each segment of the DGS is represented as slotlines of lengths $l_1$ to $l_5$.
• The coupled slotline is modeled as a short-circuited interdigital coupled line.
• The ABCD parameters for each transmission line are calculated and cascaded.
• The DGS and the microstrip line above it, of length $l_{\mathrm{dgs}}$, are combined in series.
• The remaining microstrip lines are cascaded with the combined structure.

4. Optimization Algorithm

The performance of the design is evaluated based on how low the starting frequency of the CM suppression bandwidth is, provided that the bandwidth extends to 12 GHz. Combining DGS modeling with GA, both implemented using MATLAB R2024a, the optimization algorithm sets the fitness function as the performance metric to design a DGS-based CMF optimally. The fitness function is defined as

$$\begin{array}{c}{f}_{\mathrm{fitness}}=\left\{\begin{array}{cc}{f}_s,\hfill & \mathrm{if}\mathrm{dB}\left(S_{\mathrm{cc}21}\left(f\right)\right)<-10\mathrm{dB}\mathrm{for}\mathrm{all}\hfill \\ \hfill & f\in [f_s,12\mathrm{GHz}]\hfill \\ \infty ,\hfill & \mathrm{otherwise}\hfill \end{array}\right.\hfill \\ \mathrm{Minimize}{f}_{\mathrm{fitness}}.\hfill \end{array}$$

(3)

The goal is to minimize $f_{\mathrm{fitness}}$, where ∞ corresponds to a maximum frequency for computational purposes. The fitness function ensures that suppression remains below −10 dB for all frequencies between $f_s$ and 12 GHz, with $f_s$ representing the starting frequency of CM suppression bandwidth. The upper frequency limit of 12 GHz is selected to ensure that the $S_{\mathrm{cc}21}$ bandwidth extends to this frequency, enabling CM suppression across the target range.

The optimization algorithm consists of the proposed modeling method combined with a GA. The modeling method extracts $Z_0$ and ${\epsilon }_e$ values for slotlines with widths from 0.4 mm to 3 mm in 0.1 mm increments (with 0.4 mm as the minimum value for fabrication). For coupled slotlines, widths range from 0.4 mm to 2 mm, with spacings from 0.4 mm to 3 mm. The extracted values are saved in a lookup table, which is imported by the GA. When the GA selects a DGS design with specified dimensions, the algorithm retrieves the corresponding $Z_0$ and ${\epsilon }_e$ values from the lookup table to calculate the ABCD parameters of each transmission line segment in the DGS. These segments are then cascaded, and the S-parameters are calculated from the combined ABCD parameters. The resulting S-parameters are used to evaluate the fitness of the DGS design.

The GA is selected as the optimization algorithm due to its strong performance in finding an optimal solution in an ample search space. The algorithm optimizes six parameters ($w_1$, $w_3$, $w_5$, $l_1$, $l_3$, and $l_{\mathrm{cs}}$) to design the CMF while restricting the design space to a compact 5 mm × 5 mm area for the half-circuit. Once these six parameters are optimized, the remaining dimensions of the DGS are automatically determined based on the design space, assigning fixed values to all other dimensions. The dimensions of the DGS are calculated using the following equations:

$$\begin{array}{ll}\hfill & h_{\mathrm{dgs}}=10\mathrm{mm}\hfill \\ \hfill & l_{\mathrm{dgs}}=5\mathrm{mm}\hfill \\ \hfill & l_5+w_3=h_{\mathrm{dgs}}/2\hfill \\ \hfill & l_1+w_3+l_2=h_{\mathrm{dgs}}/2\hfill \\ \hfill & w_1=w_2\hfill \\ \hfill & w_1+l_3+l_{\mathrm{cs}}+l_4+w_5=l_{\mathrm{dgs}}/2.\hfill \end{array}$$

(4)

The parameters of the GA are configured to optimize the proposed DGS-based CMF design. The algorithm is set up with a population size of 50, a crossover probability of 80%, a mutation probability of 10%, and the number of generations set to 300. The genes are configured such that each optimization parameter corresponds to a specific gene, resulting in six genes per individual ($w_1$, $w_3$, $w_5$, $l_1$, $l_3$, and $l_{\mathrm{cs}}$). As illustrated in Figure 6, the algorithm [34] begins with the initialization of a population, where random values within a specified range are assigned to the genes. When assigning values to the genes, the minimum value is set to 0.4 mm, corresponding to the minimum etching size for the DGS in PCB fabrication.

At each generation, 50 individuals are evaluated to identify the individual with the lowest (best) fitness value. This value is compared with the lowest fitness value from the previous generation, and the individual with the smallest value is saved as the overall best solution. The next generation’s population is generated by copying the parent population and through tournament selection/crossover, mutation, and elitism (the insertion of the best individual). After copying the parent population, two independent nonlinear rank-based tournament selections [35] are executed per iteration using the method from [36] with a tournament size of 6, resulting in two parent chromosomes for crossover.

With an 80% crossover probability, the two selected parent chromosomes undergo single-point crossover at a randomly chosen crossover point, generating two new offspring chromosomes. This tournament selection and crossover process repeats for 9 iterations (population/tournament size: 50/6 ≈ 9), producing 18 offspring chromosomes that replace an equivalent number of parents in the current population. Then, the mutation process replaces each gene with a new value with 10% probability when a uniformly random number in $(0,1)$ falls below 0.1 (10% mutation probability), affecting approximately 30 genes (population × number of genes × mutation probability: 50 × 6 × 0.1). Finally, elitism is applied by inserting the best individual from the prior generation into the new population, replacing a randomly selected chromosome. The process continues until 300 generations are completed. The algorithm converges well within 300 generations, as demonstrated in Figure 7, with the fitness value showing no further reduction beyond this point. Through this process, the optimization algorithm efficiently optimized the DGS-based CMF design, achieving an optimal solution within the specified design space and finding the design with the optimal CM suppression bandwidth.

5. Experimental Validation

With the optimization algorithm, it was possible to efficiently design a DGS-based CMF design with limited design space by simulating various design combinations. The algorithm evaluated 15,000 design combinations (50 individuals × 300 generations) and identified the optimal combination. The simulation for the proposed method performed on an Intel Core i9-14900 processor with 64 GB of RAM, took 580 s (∼9.6 min). In comparison, the full-wave simulation method (ANSYS HFSS 2023 R1) would take roughly 21 days to complete (121 s × 15,000 iterations). Optimally designing a DGS-based CMF with such an extensive search space is very difficult to simulate using full-wave simulation tools.

The optimized DGS-based CMF was fabricated and measured to validate the simulated results. An image of the optimized CMF after fabrication is shown in Figure 8. The S-parameters of the fabricated CMF were measured using Keysight’s E5071C ENA Vector Network Analyzer. The CMF was connected in a four-port configuration to measure differential and common-mode S-parameters. The measured results displayed in Figure 9 agree with the simulated results from ANSYS HFSS 2023 R1. The measured results indicate that the optimized CMF has a −10 dB CM suppression bandwidth ranging from 3.18 GHz to 12.89 GHz. These results are in close agreement with the ANSYS HFSS 2023 R1 results, with a CM suppression bandwidth of 3.27 GHz to 12.81 GHz. The proposed method used transmission line theory to design the DGS-based CMF optimally within the constrained design space, and the measured results validate the agreement with the simulation results.

As shown in

Table 2. Performance comparison between DGS-based CMF designs.

Reference	${\mathit{S}}_{\mathbf{cc}\mathbf{21}}\mathbf{<}\mathbf{-}\mathbf{10}$ dB (GHz)	DGS Length × Width (mm)	Equivalent Circuit Modeling Type	Algorithm-Assisted Design	Iteration Count	Optimization Time
[5]	3.6–9.1	10 × 10	LC	No	N/A	N/A
[8]	3–13	10 × 10	LC	No	N/A	N/A
[6]	2.4–6.35	12.5 × 10	Transmission line	No	N/A	N/A
[11]	9–19	19 × 9.2	LC	No	N/A	N/A
This work	3.18–12.89	5 × 10	Transmission line	Yes	15,000	9.6 min

, traditional methods do not incorporate optimization algorithms to identify optimal designs. In contrast, the proposed approach applies transmission line modeling with GA, enabling the rapid identification of optimal solutions. Specifically, the algorithm evaluated the CM suppression bandwidth for 15,000 design combinations within 9.6 min, a process that would have required approximately 21 days using ANSYS HFSS 2023 R1 simulations alone. This demonstrates the superior efficiency of the proposed method in designing compact asymmetric DGS-based CMF designs.

The optimization method presented in this work is applicable to any DGS-based CMF designs with rectangular segments, as these can be effectively modeled as transmission lines. The DGS-based CMF design developed in this work is a compact DGS-based CMF design that also achieves wideband common-mode suppression. By simulating a wide range of dimensional combinations, the algorithm can efficiently identify an optimal DGS design.

The fitness function developed in this work was specifically designed to maximize the suppression bandwidth while ensuring a low starting frequency, which is essential to wideband CMF designs. This function is flexible and can be reconfigured to target specified bandwidth requirements as needed. Overall, the proposed optimization method offers both flexibility and computational efficiency, enabling the rapid identification of optimal designs within a constrained design space.

6. Conclusions

This paper proposes an efficient optimization method for designing a DGS-based CMF that integrates equation-based transmission line modeling with GA, overcoming the inapplicability of traditional LC modeling to optimization algorithms. The proposed method provides a more practical approach to designing an optimal CMF by reducing simulation time and overcoming the limitations of full-wave simulation-based optimization. By efficiently evaluating a large number of design combinations (15,000 in just 9.6 min), the method significantly accelerates the design process compared with conventional approaches, which require substantially more computation time. The method is broadly applicable to DGS-based CMF designs with rectangular segments, which consists of most practical implementations, and enables the design of compact structures with wideband common-mode suppression. Additionally, the approach is scalable and can be extended to other DGS-based CMF topologies that can be modeled as transmission lines. The fitness function used in this approach is flexible and can be customized to target specific bandwidth requirements, further enhancing the method’s versatility. To demonstrate the method’s capabilities, this paper presents a compact asymmetric DGS-based CMF (5 mm × 10 mm). Both simulation and measured results validated the design method, with the simulated and measured bandwidths ranging from 3.27 to 12.81 GHz and from 3.18 to 12.89 GHz, respectively, demonstrating its practicality. Compared with traditional methods, the proposed method provides a more efficient approach for optimally designing DGS-based CMFs to enhance signal integrity in high-speed circuits. Overall, the proposed method offers a flexible, computationally efficient, and widely applicable solution for designing DGS-based CMFs.