Experimental Characterization of Millimeter-Wave Substrate-Integrated Waveguide Interconnect with Slot Transition in Flexible Printed Circuit Boards

: For high-speed communication services such as 5G technology, the use of millimeter-wave (mmWave) components substantially increases in mobile applications. The interconnect based on a substrate-integrated waveguide (SIW) is an efﬁcient solution for connecting these devices. However, the SIW characteristics in the mmWave frequency range are not sufﬁciently presented from the practical viewpoint. In this paper, the experimental characterization of mmWave SIWs in ﬂexible printed circuit boards (FPCBs) and their simulation results are presented. A practical method using balanced/single slot transition is proposed for microstrip-to-SIW transition. Using a full-wave simulation and genetic algorithm, the proposed slot technique is optimized. It is experimentally demonstrated that the cutoff frequency affects the operating band of the SIW differently. The per-unit-length losses of the full-mode and half-mode SIWs are obtained as 0.0375 dB/mm and 0.0609 dB/mm, respectively. Using the measurements, the SIW type effect on the transmission loss is quantitatively analyzed, and the loss is increased up to 62.4% at 39 GHz.


Introduction
High-speed communication technologies for transferring multi-media data are of importance in modern electronics.Especially, fifth generation (5G) wireless technology has received significant attention in recent years due to its extremely high data rate which is enabled using millimeter wave (mmWave) bands.Numerous industrial applications using 5G technology such as mobile cell phones and tablet devices are rapidly emerging in the market.To achieve reliable and successful mmWave communication in mobile devices, the design of interconnects between 5G devices is one of the important factors.The critical considerations of the interconnect design for practical mobile devices are a low loss for mmWave signal transmission, low electromagnetic interference (EMI), and flexible form factors.
An interconnect based on a substrate-integrated waveguide (SIW) can provide an efficient solution for these requirements.The extensive study on SIWs has been initialized mainly in microwave applications since it shows high power capability compared with conventional microstrip lines [1][2][3][4][5][6][7][8][9][10].Research on improving SIW characteristics has been continuously proposed [11][12][13][14][15][16][17][18][19].To miniaturize the SIW size, a half mode SIW (HMSIW) was proposed [11,12].The width of the HMSIW is half of that of the full-mode SIW (FMSIW).The size reduction is achieved; however, the signal transmission loss and leakage field are the drawbacks compared with the FMSIW.To overcome the problem, a folded SIW and a ridged SIW using more than three conductor layers were introduced [13][14][15][16][17][18][19].These SIW show good performance, however, the use of multiple conductor layers severely limits the flexibility of the SIW-based interconnect.For the mmWave interconnect in a flexible printed circuit board (FPCB) substrate, the minimum use of the conductor layer is preferred in practical mobile applications.Consequently, a two-layer SIW is considered a solution for FPCB interconnect operating in mmWave bands.
In addition to the layer constraint, the input and output ports of a SIW interconnect should be connected to a transmission line based on TEM mode propagation such as a microstrip line.Since the input/output interconnections of most integrated circuit (IC) devices in industrial mobile applications commonly employ a microstrip line, a mode transition structure from a quasi-TEM mode of a microstrip line to a TE mode of a SIW is necessary.Moreover, the legacy design of interconnects in mobile devices needs to be reused, and thus the compatibility between them is of importance.Many kinds of research on the transition structure are found [20][21][22][23][24].A tapered microstrip line in [20,21] and an additional via at the input port in [22][23][24] are employed for the microstrip-to-SIW transition.These previous techniques show good performance, however, they require a longer length of the microstrip line region and more conductor layers, and induce interference with the structure previously designed.Considering the small form factor of mobile applications and the need for reusing a previous design without any change, a transition structure implemented only in a SIW region is necessary.In addition, the transition structure needs to allow an engineer-friendly approach to its design.
The other crucial factor for a SIW-based interconnect for an industrial application is estimating the per-unit-length loss (i.e., dB/mm) of the interconnect through measurements.To achieve a reliable design for a system-level channel, an accurate prediction of signal transmission loss is essential.A method using an electromagnetic (EM) simulation based on the finite element method, finite difference time domain method, or method of moments is generally employed for estimating an interconnect loss.However, it is difficult to achieve an accurate result using EM simulations due to the limitation of the information on surface roughness and complex permittivity over the mmWave frequency range which significantly affect the results of the signal transmission characteristics.Surface roughness and complex permittivity of a target dielectric material can be extracted by measuring a specific FPCB pattern.However, directly measuring a transmission loss from the interconnect pattern is more convenient in the practical viewpoint if intending to fabricate a pattern anyway.
In this paper, the experimental characterization of the mmWave FMSIW and HMSIW in an FPCB substrate is presented for 5G mobile applications.The optimization method for the slot transition structure is presented based on a full-wave simulation and genetic algorithm.The engineer-friendly method for the proposed transition structure of the mmWave FMSIW and HMSIW in an FPCB substrate is developed.From the experimental results, the accurate loss characterization is performed and the useful design guide of the SIW interconnects is presented.

Theoretical Background: Design of Millimeter-Wave SIW with Slot Transition
In this section, the test vehicles (TVs) of a mmWave SIW, including a slot transition technique, are designed to experimentally characterize their transmission and reflection characteristics.TVs are fabricated using a commercial FPCB process enabling mass production.Since the SIW-based interconnect in this study focuses on industrial mobile applications, highly challenging constraints are encountered in the design process.One of the design constraints is the need for a microstrip-to-SIW transition structure without using a tapered microstrip line.Due to the extremely high density of interconnects, the area for a transition structure at the input of the SIW is very limited.Hence, the tapered type of a microstrip-to-SIW transition is difficult to be adopted.In this study, the balanced slots and single slot techniques for FMSIW and HMSIW, respectively, are presented.
Figure 1 depicts the back-to-back FMSIW and HMSIW to which the slot transition technique is applied.The FMSIW consists of two conductors between which a flexible dielectric material is placed.Two conductors are connected with each other at their edges using the via array which is equivalent to the perfect electric wall (PEC).This configuration enables forming a quasi-rectangular waveguide where a TE 10 mode is the dominant mode of wave propagation.Different from a rectangular waveguide, an FMSIW has the advantages of a low profile and ease of integration into an FPCB while determining a cutoff frequency of a dominant mode is similar to that of a rectangular waveguide.The cutoff frequency of the FMSIW is given as follows [1]. where, w eff is the effective length of the SIW for cutoff frequency estimation.Since the boundary condition implemented by the via an array of the FM-SIW is the approximation of a perfect electric wall, the width needs to be modified as seen in ( 2).D and s are a via diameter and spacing between the vias, respectively.c and ε r are the speed of light in a vacuum and the relative dielectric constant of the flexible dielectric material, respectively.dielectric material is placed.Two conductors are connected with each other at their edges using the via array which is equivalent to the perfect electric wall (PEC).This configuration enables forming a quasi-rectangular waveguide where a TE10 mode is the dominant mode of wave propagation.Different from a rectangular waveguide, an FMSIW has the advantages of a low profile and ease of integration into an FPCB while determining a cutoff frequency of a dominant mode is similar to that of a rectangular waveguide.The cutoff frequency of the FMSIW is given as follows [1]. where, weff is the effective length of the SIW for cutoff frequency estimation.Since the boundary condition implemented by the via an array of the FM-SIW is the approximation of a perfect electric wall, the width needs to be modified as seen in ( 2).D and s are a via diameter and spacing between the vias, respectively.c and εr are the speed of light in a vacuum and the relative dielectric constant of the flexible dielectric material, respectively.The HMSIW consists of a similar configuration to the FMSIW.The differences between them are the horizontal width and the boundary condition.The HMSIW width is half of that of the FMSIW.Due to the reduced width, the HMSIW has the advantage of higher interconnect density than the FMSIW, whereas the insertion loss of the HMSIW is higher because of the conductor loss and radiation loss [11].The radiation loss is induced by the leakage field of the different boundary conditions.One boundary of the HMSIW is the same as the FMSIW.In other words, the FMSIW includes the quasi-PEC implemented by the via array.The other side comprises the open boundary as shown in Figure 1b.The open boundary is supposed to be a perfect magnetic wall (PMC) ideally.However, it behaves as an aperture exhibiting a frequency-dependent radiation efficiency, which results in increasing the insertion loss of the HMSIW.Moreover, the surface roughness of the conductor and dielectric loss should be considered for justifying the insertion loss of the SIWs.These factors in the mmWave band are difficult to be obtained before the fabrication The HMSIW consists of a similar configuration to the FMSIW.The differences between them are the horizontal width and the boundary condition.The HMSIW width is half of that of the FMSIW.Due to the reduced width, the HMSIW has the advantage of higher interconnect density than the FMSIW, whereas the insertion loss of the HMSIW is higher because of the conductor loss and radiation loss [11].The radiation loss is induced by the leakage field of the different boundary conditions.One boundary of the HMSIW is the same as the FMSIW.In other words, the FMSIW includes the quasi-PEC implemented by the via array.The other side comprises the open boundary as shown in Figure 1b.The open boundary is supposed to be a perfect magnetic wall (PMC) ideally.However, it behaves as an aperture exhibiting a frequency-dependent radiation efficiency, which results in increasing the insertion loss of the HMSIW.Moreover, the surface roughness of the conductor and dielectric loss should be considered for justifying the insertion loss of the SIWs.These factors in the mmWave band are difficult to be obtained before the fabrication of the products; the experimental characterization and the analysis based on it are of importance for the SIW interconnect in 5G mobile applications.
For impedance matching between the microstrip line and the SIW, the transition structure using balanced slots is presented in Figure 1a.For the microstrip-to-SIW transition region, two symmetric slots are formed at the input of the SIW.The rectangular slots are arranged in the longitudinal axis of the SIW.The transition structure is completely implemented inside the FMSIW region.Thus, the region adjacent to the microstrip line can be used for other interconnects and power/ground planes.It enables the compact design of mobile applications.The design parameters for the balanced slots are depicted in Figure 2a.A slot width, a slot length, and a distance between the slot and the microstrip line are represented as w s , L s , and d s , respectively.The proposed transition structure is simple to be designed and fabricated.Thus, it tolerates manufacturing process variations which are highly demanded in industrial applications.The HMSIW includes a single slot for the microstrip-to-SIW transition as shown in Figure 1b.Though the SIW region of the HMSIW can be simply designed by splitting the SIW region of the FMSIW in half along the longitudinal axis, it notes that the same results of the slot design parameters of the FMSIW cannot be used for the HMSIW.The microstrip region of the HMSIW is not half of the microstrip region of the FMSIW.The slot transition for the FMSIW and HMSIW should be designed differently for acquiring an optimized result.
Electronics 2022, 11, x FOR PEER REVIEW 5 the compact design of mobile applications.The design parameters for the balanced are depicted in Figure 2a.A slot width, a slot length, and a distance between the slot the microstrip line are represented as ws, Ls, and ds, respectively.The proposed trans structure is simple to be designed and fabricated.Thus, it tolerates manufacturing pro variations which are highly demanded in industrial applications.The HMSIW inclu single slot for the microstrip-to-SIW transition as shown in Figure 1b.Though the region of the HMSIW can be simply designed by splitting the SIW region of the FM in half along the longitudinal axis, it notes that the same results of the slot de parameters of the FMSIW cannot be used for the HMSIW.The microstrip region o HMSIW is not half of the microstrip region of the FMSIW.The slot transition fo FMSIW and HMSIW should be designed differently for acquiring an optimized resu For the experimental characterization of the FPCB SIW over mmWave bands, are fabricated.The design parameters of the FMSIW and HMSIW are shown in Fi 2a,b, respectively.The width and length of the SIW region for the FMSIW are denot wi and Li, respectively.According to the objective value of the cutoff frequency, determined as described in (2).The design variables for the balanced slot transition the slot width, length, and distance of ws, Ls, and ds.For the microstrip line, wm an represent the width and length of the line.The via diameter and spacing are represe as D and sv.
One of the critical variables is the design parameters of signal-launching pad measurements.Though the measurement pads are not shown in Figure 1 for the sa simplicity, they should be included in the TV design.To minimize the unwanted e from the signal-launching pads during the measurements, a measurement techn using a GSG-type/150 μm-pitch probe is employed as seen in Figure 2. The width o signal line is denoted as wp.The size of a ground pad is dp × dp.The spacing betwee signal line and the ground pad is sp.The distance from the microstrip line to the gro pad is gp.For the HMSIW, the design parameters of TVs are determined in a sim manner.The design parameters of the HMSIW are denoted using the additional subs 'h'.
The stack-up of the FPCB manufacturing process is depicted in Figure 3.Both FMSIW and HMSIW have two layers for a conductor, namely layers 1 and 2. Copp For the experimental characterization of the FPCB SIW over mmWave bands, TVs are fabricated.The design parameters of the FMSIW and HMSIW are shown in Figure 2a,b, respectively.The width and length of the SIW region for the FMSIW are denoted as w i and L i , respectively.According to the objective value of the cutoff frequency, w i is determined as described in (2).The design variables for the balanced slot transition are the slot width, length, and distance of w s , L s , and d s .For the microstrip line, w m and L m represent the width and length of the line.The via diameter and spacing are represented as D and s v .
One of the critical variables is the design parameters of signal-launching pads for measurements.Though the measurement pads are not shown in Figure 1 for the sake of simplicity, they should be included in the TV design.To minimize the unwanted effect from the signal-launching pads during the measurements, a measurement technique using a GSG-type/150 µm-pitch probe is employed as seen in Figure 2. The width of the signal line is denoted as w p .The size of a ground pad is d p × d p .The spacing between the signal line and the ground pad is s p .The distance from the microstrip line to the ground pad is g p .For the HMSIW, the design parameters of TVs are determined in a similar manner.The design parameters of the HMSIW are denoted using the additional subscript 'h'.
The stack-up of the FPCB manufacturing process is depicted in Figure 3.Both the FMSIW and HMSIW have two layers for a conductor, namely layers 1 and 2. Copper is used for the conductor layer.Layer 1 further consists of Cu plating and copper layers of which thicknesses are 15 ± 5 µm and 12 µm.Layer 2 includes a copper layer (12 µm), Cu plating (15 ± 5 µm), and PSR (25 ± 10 µm).A dielectric layer comprises two polyimide layers and two prepreg layers as shown in Figure 3.Each thickness is 50 µm.The dielectric constant (DK) values of the polyimide and prepreg are 3.39 and 2.90, respectively.The dissipation factor (DF) of the polyimide and prepreg are 0.0025 and 0.0023, respectively.The stack-up employed in this study is widely used for practical applications.
signal line and the ground pad is sp.The distance from the microstrip line to the ground pad is gp.For the HMSIW, the design parameters of TVs are determined in a similar manner.The design parameters of the HMSIW are denoted using the additional subscript 'h'.
The stack-up of the FPCB manufacturing process is depicted in Figure 3.Both the FMSIW and HMSIW have two layers for a conductor, namely layers 1 and 2. Copper is used for the conductor layer.Layer 1 further consists of Cu plating and copper layers of which thicknesses are 15 ± 5 μm and 12 μm.Layer 2 includes a copper layer (12 μm), Cu plating (15 ± 5 μm), and PSR (25 ± 10 μm).A dielectric layer comprises two polyimide layers and two prepreg layers as shown in Figure 3.Each thickness is 50 μm.The dielectric constant (DK) values of the polyimide and prepreg are 3.39 and 2.90, respectively.The dissipation factor (DF) of the polyimide and prepreg are 0.0025 and 0.0023, respectively.The stack-up employed in this study is widely used for practical applications.Using the design parameters and the stack-up, four TVs are fabricated.TV 1 and TV 2 are the FMSIWs of which the SIW lengths are different from each other.Similarly, TV 3 and TV 4 are the HMSIWs with short and long SIW regions, respectively.The reason why the TVs with different lengths are fabricated is that the transmission loss is simply extracted using the different length method.In general, the characteristics of the back-toback configuration of the SIW shown in Figure 1 are obtained by de-embedding the microprobe pads using a de-embedding technique such as a thru-reflect-line (TRL) method.However, the SIW structures analyzed herein exhibit a large-scale difference between the SIW and the microprobe pad, which results in inaccurate results after de-embedding.Moreover, one of the main purposes of the SIW experimental characterization is to obtain the signal transmission loss in the mmWave bands rather than to obtain overall characteristics.Therefore, the difference length method requiring two different-length TVs is adopted in this study.Li and Lih are appropriately determined by this different length method.

GA-Based Design Results
For the design of the balanced slot transition, the optimization method combining genetic algorithm (GA) with finite element method (FEM) is adopted in this study.Considering industrial needs, the FEM-based electromagnetic simulation is preferred instead Using the design parameters and the stack-up, four TVs are fabricated.TV 1 and TV 2 are the FMSIWs of which the SIW lengths are different from each other.Similarly, TV 3 and TV 4 are the HMSIWs with short and long SIW regions, respectively.The reason why the TVs with different lengths are fabricated is that the transmission loss is simply extracted using the different length method.In general, the characteristics of the back-to-back configuration of the SIW shown in Figure 1 are obtained by de-embedding the microprobe pads using a deembedding technique such as a thru-reflect-line (TRL) method.However, the SIW structures analyzed herein exhibit a large-scale difference between the SIW and the microprobe pad, which results in inaccurate results after de-embedding.Moreover, one of the main purposes of the SIW experimental characterization is to obtain the signal transmission loss in the mmWave bands rather than to obtain overall characteristics.Therefore, the difference length method requiring two different-length TVs is adopted in this study.L i and L ih are appropriately determined by this different length method.

GA-Based Design Results
For the design of the balanced slot transition, the optimization method combining genetic algorithm (GA) with finite element method (FEM) is adopted in this study.Considering industrial needs, the FEM-based electromagnetic simulation is preferred instead of extracting an analytical equation because the analytical equation is commonly limited to stack-up and dielectric materials.However, the optimization method of the FEM and GA enables an engineer to achieve a more flexible design for various configurations.The GA architecture in this study consists of the generation of the initial population, objective function evaluation, selection, crossover, and mutation.The population is the set of chromosomes comprising the single array of the values of w s , L s , and d s for FMSIW or w sh , L sh , and d sh for HMSIW.The cost function for the genetic algorithm is determined by the reflection coefficient of the SIW as follows:  To demonstrate the improvement of the reflection and transmission character the S11 and S21 parameters of the 4th, 126th, and 635th iterations are exhibited.In th iteration, the values of the slot design parameters are (ws, Ls, ds) of (0.3 mm, 1.32 mm mm).The reflection characteristics (S11) are more than −15 dB over the wideband quency range and thus the signal transmission loss is substantially high.In the 126th ation, partial improvement is observed.The values of the slot design parameters ar Ls, ds) of (0.09 mm, 0.76 mm, 0.17 mm).Most of the S11 values are less than −10 dB.Fi a substantial improvement of the S11 and S21 parameters is observed in the 635th iter The overall S11 values in the frequency of interest are less than −15 dB which results good characteristics of the signal transmission.The values of the slot design param are (ws, Ls, ds) of (0.87 mm, 0.98 mm, 0.19 mm).The design parameters of this iter To demonstrate the improvement of the reflection and transmission characteristics, the S 11 and S 21 parameters of the 4th, 126th, and 635th iterations are exhibited.In the 4th iteration, the values of the slot design parameters are (w s , L s , d s ) of (0.3 mm, 1.32 mm, 0.38 mm).The reflection characteristics (S 11 ) are more than −15 dB over the wideband frequency range and thus the signal transmission loss is substantially high.In the 126th iteration, partial improvement is observed.The values of the slot design parameters are (w s , L s , d s ) of (0.09 mm, 0.76 mm, 0.17 mm).Most of the S 11 values are less than −10 dB.Finally, a substantial improvement of the S 11 and S 21 parameters is observed in the 635th iteration.The overall S 11 values in the frequency of interest are less than −15 dB which results in the good characteristics of the signal transmission.The values of the slot design parameters are (w s , L s , d s ) of (0.87 mm, 0.98 mm, 0.19 mm).The design parameters of this iteration step are selected as an optimized solution since the value of the cost function is minimized in this step.
The complete dimensions including the optimized slot design parameters for FMSIW are depicted in Table 1.For HMSIW, a similar optimization method is applied, and the dimensions are obtained as shown in Table 1.The optimized slot parameters for HMSIW are (w s , L s , d s ) of (0.44 mm, 1.02 mm, 0.07 mm).As previously explained, the different optimized values are acquired for FMSIW and HMSIW.For the difference length method, the short (TV 1) and long (TV 2) lengths of the SIW region in the FMSIW are 10.84 mm and 23.64 mm, respectively.The SIW region lengths of TV 3 and TV 4 are 10.76 mm and 23.56 mm, respectively.The proposed transition technique is compared with the previous methods as shown in Table 2.The proposed method shows the advantage of occupation in the SIW region only and an engineer-friendly optimization method.

Experimental Setup
Using the FPCB process, TVs are fabricated.The photograph of the TVs is shown in Figure 5.The measurement setup for S-parameters in mmWave bands is exhibited in Figure 6.The frequency range is from 15 GHz to 40 GHz with 534 points.SOLT calibration is performed.To minimize the effect of parasitics of measurement pads, GSG-type microprobes with 150 µm pitch are used.The procedure of the experimental setup is summarized as follows.1. Prepare test vehicles; 2. Prepare two microprobes and connect them to a vector network analyzer (VNA) through high-frequency cables; 3. Power a VNA on for over 30 min before starting a measurement; 4. Set start and stop frequencies, number of points, and IF bandwidth; 5. Prepare calsub corresponding to the microprobes for calibration; 6. Conduct calibration using SOLT method; 7. Confirm the calibration result using calsub; 8. Start measurements of TVs; 9. Save the data of the measurements as snp file.
The procedure of the experimental setup can be clarified using a standard method referred to ASTM/ISO standards for further mass production.

Experimental Results and SIW Characterization
Figure 7a,b depict the measured S-parameters of the FMSIW (TV 1, TV 2, TV2-bent) and the HMSIW (TV 3, TV 4), respectively.According to the measurements, the cutoff frequencies of all TVs are approximately 22.9 GHz.The cutoff frequency is defined as the frequency where the S11 value is equal to −15 dB.Using (1) and ( 2), weff is equal to 3.7 mm and the cutoff frequency is calculated as 22.6 GHz.It shows good agreement with the measurements.
Prepare two microprobes and connect them to a vector network analyzer (VNA) through high-frequency cables; 3.
Power a VNA on for over 30 min before starting a measurement; 4.
Set start and stop frequencies, number of points, and IF bandwidth; 5.
Prepare calsub corresponding to the microprobes for calibration; 6.
Confirm the calibration result using calsub; 8.
Start measurements of TVs; 9.
Save the data of the measurements as snp file.
The procedure of the experimental setup can be clarified using a standard method referred to ASTM/ISO standards for further mass production.However, it is not ensured in practical applications considering a cutoff frequency as the start frequency of the operating band.Characterizing the transmission characteristics at the cutoff frequency, the S 21 values of TVs 1, 2, 3, and 4 are −1.79 dB, −2.89 dB, −2.75 dB, and −5.14 dB, respectively.Especially, TV 4 (HMSIW with the long length) shows a relatively high loss of the S 21 parameters.Considering that a loss of 3 dB is commonly used as the critical value for signal transmission, the cutoff frequency is not ensured for estimating the low limit of the operating frequency for the SIW-based interconnect.The experimental results indicate that the operating frequency of the SIW-based interconnect should be compensated.The frequency of 3 dB loss in TV 4 is 24.7 GHz which is obtained from the measurements.The low limit of the operating band in TV 4 is increased up to 9.3% compared to (1).

Experimental Results and SIW Characterization
Moreover, as previously explained, TV 1 and TV 3 are the FMSIW and HMSIW with the short length of the SIW region whereas TV 2 and TV 4 are the structure having the long length of the SIW region.It is experimentally observed that the signal loss increases as the SIW region becomes longer for both the FMSIW and the HMSIW which can be anticipated.The lowest S 21 values in the frequency range above the cutoff frequency are −0.95dB for TV 1, −1.28 dB for TV 2, −1.13 dB for TV 3, and −1.75 dB for TV 4. The length difference induces the loss increase in 0.33 dB of the FMSIW and 0.62 dB of the HMSIW for the lowest S 21 value, whereas the S 21 differences for the FMSIW and the HMSIW are 1.10 dB and 2.39 dB, respectively, for the S 21 value at the cutoff frequency.The loss increase at the cutoff frequency is more significant than that in the operating band.Additionally, the HMSIW shows more critical characteristics than the FMSIW.Considering these results of the experimental characterization, in practical applications, a design with a sufficient margin is necessary when a cutoff frequency is determined according to an operating band.In addition, the straight and bent TVs are compared in Figure 7a.As can be seen here, the overall characteristics of the measured S-parameters between the straight and bent TVs are similar.
One of the important characteristics of the mmWave SIW-based interconnect is the per-unit-length loss at specific frequency points.This study mainly focuses on the perunit-length losses at the frequencies of 28 GHz and 39 GHz that are used for mmWave devices in a mobile application.To clarify the losses, the measurements zoomed in on the S 21 value from −3.0 dB to 0 dB are shown in Figure 8.As can be seen in the Figure, the losses for the SIWs with the short and long lengths at the frequencies of 28 GHz and 39 GHz can be acquired.To minimize the ripple effect in the passband, the moving average is applied to the S 21 value that represents filtered measurements.The ripple is induced from the measurement pads, so it can be ignored by using the moving average value.For TV 1 and TV 2, the S 21 values at the frequency of 28 GHz are −1.13dB and −1.59 dB, whereas those at the frequency of 39 GHz are −1.24dB and −1.72 dB.From these results, the per-unit-length loss is extracted using the difference method as follows.
Loss (dB/mm) = |S21| 28, 39 GHz According to the results, the per-unit-length loss of the FMSIW at the frequencies of 28 and 39 GHz are 0.0359 dB/mm and 0.0375 dB/mm, respectively.For the HMSIW, 0.0484 dB/mm at 28 GHz and 0.0609 dB/mm at 39 GHz are obtained.The loss of the HMSIW is increased up to 34.8% and 62.4% at the frequencies of 28 and 39 GHz, respectively, compared to the FMSIW.The HMSIW has the advantage of the narrower width, however, it has the drawback of the loss increase in the passband.This result indicates that an engineer should cautiously select the type of the SIW depending on the application and the experimental characterization should be performed for achieving a successful design of a mmWave interconnect in mobile applications.

Conclusions
In this paper, the experimental characterization of the SIW-based interconnects for mmWave-band applications including 5G mobile devices was presented.The efficient microstrip-to-SIW transition using the balanced/single slot techniques was developed.Additionally, the practical design method based on the genetic algorithm for the transition structure was presented.Using the fabricated FPCB SIW interconnects, the effect of the cutoff frequency and the per-unit-length loss were thoroughly characterized.The experimental results indicate that the low limit of the passband should be compensated by the cutoff frequency.Moreover, the operating band design based on the cutoff frequency is more critical in the HMSIW design.The quantitative analysis of the per-unit-length loss for the FMSIW and HMSIW provides an insight into the practical design of the mmWave SIW interconnect in FPCBs.The experimental characterization of a crosstalk effect in the mmWave SIW interconnect will be conducted in further work.According to the results, the per-unit-length loss of the FMSIW at the frequencies of 28 and 39 GHz are 0.0359 dB/mm and 0.0375 dB/mm, respectively.For the HMSIW, 0.0484 dB/mm at 28 GHz and 0.0609 dB/mm at 39 GHz are obtained.The loss of the HMSIW is increased up to 34.8% and 62.4% at the frequencies of 28 and 39 GHz, respectively, compared to the FMSIW.The HMSIW has the advantage of the narrower width, however, it has the drawback of the loss increase in the passband.This result indicates that an engineer should cautiously select the type of the SIW depending on the application and the experimental characterization should be performed for achieving a successful design of a mmWave interconnect in mobile applications.

Conclusions
In this paper, the experimental characterization of the SIW-based interconnects for mmWave-band applications including 5G mobile devices was presented.The efficient microstrip-to-SIW transition using the balanced/single slot techniques was developed.Additionally, the practical design method based on the genetic algorithm for the transition structure was presented.Using the fabricated FPCB SIW interconnects, the effect of the cutoff frequency and the per-unit-length loss were thoroughly characterized.The experimental results indicate that the low limit of the passband should be compensated by the cutoff frequency.Moreover, the operating band design based on the cutoff frequency is more critical in the HMSIW design.The quantitative analysis of the per-unit-length loss for the FMSIW and HMSIW provides an insight into the practical design of the mmWave SIW interconnect in FPCBs.The experimental characterization of a crosstalk effect in the mmWave SIW interconnect will be conducted in further work.

Figure 2 .
Figure 2. Fabricated TVs and their design parameters for experimental characterization.(a) FM for TV 1 and TV 2; (b) HMSIW for TV 3 and TV 4.

Figure 2 .
Figure 2. Fabricated TVs and their design parameters for experimental characterization.(a) FMSIW for TV 1 and TV 2; (b) HMSIW for TV 3 and TV 4.

Figure 3 .
Figure 3. Stack-up of FPCB process for the experimental characterization of FMSIW and HMSIW.

Figure 3 .
Figure 3. Stack-up of FPCB process for the experimental characterization of FMSIW and HMSIW.

Electronics 2022, 11 , 2995 6 of 11 where S 11
is the reflection coefficient obtained from the full-wave simulation based on FEM.In this study, Ansys HFSS software is used for a FEM solver.w is the weight factor for calculation convenience.The value of 15 dB is employed as the reference value of the wellmatched case.The purpose of the GA is to obtain the slot design parameters minimizing the cost function.The number of individuals is 30.Roulette selection is adopted.Simulated binary crossover and polynomial mutation are employed in GA.The maximum number of generations is 1000.The GA results during iterations for FMSIW are shown in Figure4.The overall value of the cost function gradually decreases as the iteration number increases.mosomescomprising the single array of the values of ws, Ls, and ds for FMSIW or ws and dsh for HMSIW.The cost function for the genetic algorithm is determined by th flection coefficient of the SIW as follows:  = ∑ (| 11 |dB + 15 dB) 40 GHz = where S11 is the reflection coefficient obtained from the full-wave simulation base FEM.In this study, Ansys HFSS software is used for a FEM solver.w is the weight f for calculation convenience.The value of 15 dB is employed as the reference value o well-matched case.The purpose of the GA is to obtain the slot design parameters mizing the cost function.The number of individuals is 30.Roulette selection is ado Simulated binary crossover and polynomial mutation are employed in GA.The maxi number of generations is 1000.The GA results during iterations for FMSIW are show Figure4.The overall value of the cost function gradually decreases as the iteration nu increases.

Figure 4 .
Figure 4.The GA results for the FMSIW with a balanced slot transition show the improvem the reflection and transmission characteristics.

Figure 4 .
Figure 4.The GA results for the FMSIW with a balanced slot transition show the improvement of the reflection and transmission characteristics.

Figure 6 .
Figure 6.Measurement setup for S-parameters with a vector network analyzer and microprobes.

Figure 6 .
Figure 6.Measurement setup for S-parameters with a vector network analyzer and microprobes.

Figure
Figure7a,b depict the measured S-parameters of the FMSIW (TV 1, TV 2, TV2-bent) and the HMSIW (TV 3, TV 4), respectively.According to the measurements, the cutoff frequencies of all TVs are approximately 22.9 GHz.The cutoff frequency is defined as the frequency where the S 11 value is equal to −15 dB.Using (1) and (2), w eff is equal to 3.7 mm and the cutoff frequency is calculated as 22.6 GHz.It shows good agreement with the measurements.

Figure 8 .
Figure 8. S 21 measurements of (a) TV 1 and TV 2 and (b) TV 3 and TV 4 for per-unit-length characterization using difference length method.

Table 1 .
Design parameters and dimensions of TVs for FMSIW and HMSIW (units: mm).

Table 2 .
Comparison of transition technique for SIWs.